1
0
mirror of https://passt.top/passt synced 2024-06-30 23:12:39 +00:00
passt/tcp.c

3496 lines
93 KiB
C
Raw Normal View History

passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
// SPDX-License-Identifier: AGPL-3.0-or-later
/* PASST - Plug A Simple Socket Transport
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* for qemu/UNIX domain socket mode
*
* PASTA - Pack A Subtle Tap Abstraction
* for network namespace/tap device mode
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* tcp.c - TCP L2-L4 translation state machine
*
* Copyright (c) 2020-2022 Red Hat GmbH
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* Author: Stefano Brivio <sbrivio@redhat.com>
*/
/**
* DOC: Theory of Operation
*
*
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* PASST mode
* ==========
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* This implementation maps TCP traffic between a single L2 interface (tap) and
* native TCP (L4) sockets, mimicking and reproducing as closely as possible the
* inferred behaviour of applications running on a guest, connected via said L2
* interface. Four connection flows are supported:
* - from the local host to the guest behind the tap interface:
* - this is the main use case for proxies in service meshes
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* - we bind to configured local ports, and relay traffic between L4 sockets
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* with local endpoints and the L2 interface
* - from remote hosts to the guest behind the tap interface:
* - this might be needed for services that need to be addressed directly,
* and typically configured with special port forwarding rules (which are
* not needed here)
* - we also relay traffic between L4 sockets with remote endpoints and the L2
* interface
* - from the guest to the local host:
* - this is not observed in practice, but implemented for completeness and
* transparency
* - from the guest to external hosts:
* - this might be needed for applications running on the guest that need to
* directly access internet services (e.g. NTP)
*
* Relevant goals are:
* - transparency: sockets need to behave as if guest applications were running
* directly on the host. This is achieved by:
* - avoiding port and address translations whenever possible
* - mirroring TCP dynamics by observation of socket parameters (TCP_INFO
* socket option) and TCP headers of packets coming from the tap interface,
* reapplying those parameters in both flow directions (including TCP_MSS,
* TCP_WINDOW_CLAMP socket options)
* - simplicity: only a small subset of TCP logic is implemented here and
* delegated as much as possible to the TCP implementations of guest and host
* kernel. This is achieved by:
* - avoiding a complete TCP stack reimplementation, with a modified TCP state
* machine focused on the translation of observed events instead
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* - mirroring TCP dynamics as described above and hence avoiding the need for
* segmentation, explicit queueing, and reassembly of segments
* - security:
* - no dynamic memory allocation is performed
* - TODO: synflood protection
*
* Portability is limited by usage of Linux-specific socket options.
*
*
* Limits
* ------
*
* To avoid the need for dynamic memory allocation, a maximum, reasonable amount
* of connections is defined by TCP_MAX_CONNS (currently 128k).
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
*
* Data needs to linger on sockets as long as it's not acknowledged by the
* guest, and is read using MSG_PEEK into preallocated static buffers sized
* to the maximum supported window, 16 MiB ("discard" buffer, for already-sent
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
* data) plus a number of maximum-MSS-sized buffers. This imposes a practical
* limitation on window scaling, that is, the maximum factor is 256. Larger
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
* factors will be accepted, but resulting, larger values are never advertised
* to the other side, and not used while queueing data.
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
*
* Ports
* -----
*
* To avoid the need for ad-hoc configuration of port forwarding or allowed
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* ports, listening sockets can be opened and bound to all unbound ports on the
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* host, as far as process capabilities allow. This service needs to be started
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* after any application proxy that needs to bind to local ports. Mapped ports
* can also be configured explicitly.
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* No port translation is needed for connections initiated remotely or by the
* local host: source port from socket is reused while establishing connections
* to the guest.
*
* For connections initiated by the guest, it's not possible to force the same
* source port as connections are established by the host kernel: that's the
* only port translation needed.
*
*
* Connection tracking and storage
* -------------------------------
*
* Connections are tracked by the @tc array of struct tcp_conn, containing
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* addresses, ports, TCP states and parameters. This is statically allocated and
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* indexed by an arbitrary connection number. The array is compacted whenever a
* connection is closed, by remapping the highest connection index in use to the
* one freed up.
*
* References used for the epoll interface report the connection index used for
* the @tc array.
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* IPv4 addresses are stored as IPv4-mapped IPv6 addresses to avoid the need for
* separate data structures depending on the protocol version.
*
* - Inbound connection requests (to the guest) are mapped using the triple
* < source IP address, source port, destination port >
* - Outbound connection requests (from the guest) are mapped using the triple
* < destination IP address, destination port, source port >
* where the source port is the one used by the guest, not the one used by the
* corresponding host socket
*
*
* Initialisation
* --------------
*
* Up to 2^15 + 2^14 listening sockets (excluding ephemeral ports, repeated for
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* IPv4 and IPv6) can be opened and bound to wildcard addresses. Some will fail
* to bind (for low ports, or ports already bound, e.g. by a proxy). These are
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* added to the epoll list, with no separate storage.
*
*
* Events and states
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* -----------------
*
* Instead of tracking connection states using a state machine, connection
* events are used to determine state and actions for a given connection. This
* makes the implementation simpler as most of the relevant tasks deal with
* reactions to events, rather than state-associated actions. For user
* convenience, approximate states are mapped in logs from events by
* @tcp_state_str.
*
* The events are:
*
* - SOCK_ACCEPTED connection accepted from socket, SYN sent to tap/guest
*
* - TAP_SYN_RCVD tap/guest initiated connection, SYN received
*
* - TAP_SYN_ACK_SENT SYN, ACK sent to tap/guest, valid for TAP_SYN_RCVD only
*
* - ESTABLISHED connection established, the following events are valid:
*
* - SOCK_FIN_RCVD FIN (EPOLLRDHUP) received from socket
*
* - SOCK_FIN_SENT FIN (write shutdown) sent to socket
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* - TAP_FIN_RCVD FIN received from tap/guest
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* - TAP_FIN_SENT FIN sent to tap/guest
*
* - TAP_FIN_ACKED ACK to FIN seen from tap/guest
*
* Setting any event in CONN_STATE_BITS (SOCK_ACCEPTED, TAP_SYN_RCVD,
* ESTABLISHED) clears all the other events, as those represent the fundamental
* connection states. No events (events == CLOSED) means the connection is
* closed.
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* Connection setup
* ----------------
*
* - inbound connection (from socket to guest): on accept() from listening
* socket, the new socket is mapped in connection tracking table, and
* three-way handshake initiated towards the guest, advertising MSS and window
* size and scaling from socket parameters
* - outbound connection (from guest to socket): on SYN segment from guest, a
* new socket is created and mapped in connection tracking table, setting
* MSS and window clamping from header and option of the observed SYN segment
*
*
* Aging and timeout
* -----------------
*
* Timeouts are implemented by means of timerfd timers, set based on flags:
*
* - SYN_TIMEOUT: if no ACK is received from tap/guest during handshake (flag
* ACK_FROM_TAP_DUE without ESTABLISHED event) within this time, reset the
* connection
*
* - ACK_TIMEOUT: if no ACK segment was received from tap/guest, after sending
* data (flag ACK_FROM_TAP_DUE with ESTABLISHED event), re-send data from the
* socket and reset sequence to what was acknowledged. If this persists for
* more than TCP_MAX_RETRANS times in a row, reset the connection
*
* - FIN_TIMEOUT: if a FIN segment was sent to tap/guest (flag ACK_FROM_TAP_DUE
* with TAP_FIN_SENT event), and no ACK is received within this time, reset
* the connection
*
* - FIN_TIMEOUT: if a FIN segment was acknowledged by tap/guest and a FIN
* segment (write shutdown) was sent via socket (events SOCK_FIN_SENT and
* TAP_FIN_ACKED), but no socket activity is detected from the socket within
* this time, reset the connection
*
* - ACT_TIMEOUT, in the presence of any event: if no activity is detected on
* either side, the connection is reset
*
* - ACK_INTERVAL elapsed after data segment received from tap without having
* sent an ACK segment, or zero-sized window advertised to tap/guest (flag
* ACK_TO_TAP_DUE): forcibly check if an ACK segment can be sent
*
*
* Summary of data flows (with ESTABLISHED event)
* ----------------------------------------------
*
* @seq_to_tap: next sequence for packets to tap/guest
* @seq_ack_from_tap: last ACK number received from tap/guest
* @seq_from_tap: next sequence for packets from tap/guest (expected)
* @seq_ack_to_tap: last ACK number sent to tap/guest
*
* @seq_init_from_tap: initial sequence number from tap/guest
* @seq_init_to_tap: initial sequence number from tap/guest
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* @wnd_from_tap: last window size received from tap, never scaled
* @wnd_from_tap: last window size advertised from tap, never scaled
*
* - from socket to tap/guest:
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* - on new data from socket:
* - peek into buffer
* - send data to tap/guest:
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* - starting at offset (@seq_to_tap - @seq_ack_from_tap)
* - in MSS-sized segments
* - increasing @seq_to_tap at each segment
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
* - up to window (until @seq_to_tap - @seq_ack_from_tap <= @wnd_from_tap)
* - on read error, send RST to tap/guest, close socket
* - on zero read, send FIN to tap/guest, set TAP_FIN_SENT
* - on ACK from tap/guest:
* - set @ts_ack_from_tap
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* - check if it's the second duplicated ACK
* - consume buffer by difference between new ack_seq and @seq_ack_from_tap
* - update @seq_ack_from_tap from ack_seq in header
* - on two duplicated ACKs, reset @seq_to_tap to @seq_ack_from_tap, and
* resend with steps listed above
* - set TCP_WINDOW_CLAMP from TCP header from tap
*
* - from tap/guest to socket:
* - on packet from tap/guest:
* - set @ts_tap_act
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* - set TCP_WINDOW_CLAMP from TCP header from tap
* - check seq from header against @seq_from_tap, if data is missing, send
* two ACKs with number @seq_ack_to_tap, discard packet
* - otherwise queue data to socket, set @seq_from_tap to seq from header
* plus payload length
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
* - in ESTABLISHED state, send ACK to tap as soon as we queue to the
* socket. In other states, query socket for TCP_INFO, set
* @seq_ack_to_tap to (tcpi_bytes_acked + @seq_init_from_tap) % 2^32 and
* send ACK to tap/guest
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
*
*
* PASTA mode
* ==========
*
* For traffic directed to TCP ports configured for mapping to the tuntap device
* in the namespace, and for non-local traffic coming from the tuntap device,
* the implementation is identical as the PASST mode described in the previous
* section.
*
* For local traffic directed to TCP ports configured for direct mapping between
* namespaces, see the implementation in tcp_splice.c.
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
#include <sched.h>
#include <fcntl.h>
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#include <stdio.h>
#include <stdlib.h>
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#include <errno.h>
#include <limits.h>
#include <net/ethernet.h>
#include <net/if.h>
#include <netinet/in.h>
#include <netinet/ip.h>
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#include <stdint.h>
#include <stddef.h>
#include <string.h>
#include <sys/epoll.h>
#ifdef HAS_GETRANDOM
#include <sys/random.h>
#endif
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#include <sys/socket.h>
#include <sys/timerfd.h>
#include <sys/types.h>
#include <sys/uio.h>
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#include <unistd.h>
#include <time.h>
#include <linux/tcp.h> /* For struct tcp_info */
#include "checksum.h"
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
#include "util.h"
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#include "passt.h"
#include "tap.h"
#include "siphash.h"
#include "pcap.h"
#include "conf.h"
#include "tcp_splice.h"
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define TCP_FRAMES_MEM 128
#define TCP_FRAMES \
(c->mode == MODE_PASST ? TCP_FRAMES_MEM : 1)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define TCP_FILE_PRESSURE 30 /* % of c->nofile */
#define TCP_CONN_PRESSURE 30 /* % of c->tcp.conn_count */
#define TCP_HASH_BUCKET_BITS (TCP_CONN_INDEX_BITS + 1)
#define TCP_HASH_TABLE_LOAD 70 /* % */
#define TCP_HASH_TABLE_SIZE (TCP_MAX_CONNS * 100 / \
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
TCP_HASH_TABLE_LOAD)
#define MAX_WS 8
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define MAX_WINDOW (1 << (16 + (MAX_WS)))
/* MSS rounding: see SET_MSS() */
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define MSS_DEFAULT 536
struct tcp4_l2_head { /* For MSS4 macro: keep in sync with tcp4_l2_buf_t */
uint32_t psum;
uint32_t tsum;
#ifdef __AVX2__
uint8_t pad[18];
#else
uint8_t pad[2];
#endif
uint32_t vnet_len;
struct ethhdr eh;
struct iphdr iph;
struct tcphdr th;
#ifdef __AVX2__
} __attribute__ ((packed, aligned(32)));
#else
} __attribute__ ((packed, aligned(__alignof__(unsigned int))));
#endif
struct tcp6_l2_head { /* For MSS6 macro: keep in sync with tcp6_l2_buf_t */
#ifdef __AVX2__
uint8_t pad[14];
#else
uint8_t pad[2];
#endif
uint32_t vnet_len;
struct ethhdr eh;
struct ipv6hdr ip6h;
struct tcphdr th;
#ifdef __AVX2__
} __attribute__ ((packed, aligned(32)));
#else
} __attribute__ ((packed, aligned(__alignof__(unsigned int))));
#endif
#define MSS4 ROUND_DOWN(USHRT_MAX - sizeof(struct tcp4_l2_head), 4)
#define MSS6 ROUND_DOWN(USHRT_MAX - sizeof(struct tcp6_l2_head), 4)
#define WINDOW_DEFAULT 14600 /* RFC 6928 */
#ifdef HAS_SND_WND
# define KERNEL_REPORTS_SND_WND(c) (c->tcp.kernel_snd_wnd)
#else
# define KERNEL_REPORTS_SND_WND(c) (0 && (c))
#endif
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define ACK_INTERVAL 50 /* ms */
#define SYN_TIMEOUT 10 /* s */
#define ACK_TIMEOUT 2
#define FIN_TIMEOUT 60
#define ACT_TIMEOUT 7200
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define TCP_SOCK_POOL_TSH 16 /* Refill in ns if > x used */
#define LOW_RTT_TABLE_SIZE 8
#define LOW_RTT_THRESHOLD 10 /* us */
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/* We need to include <linux/tcp.h> for tcpi_bytes_acked, instead of
* <netinet/tcp.h>, but that doesn't include a definition for SOL_TCP
*/
#define SOL_TCP IPPROTO_TCP
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
#define SEQ_LE(a, b) ((b) - (a) < MAX_WINDOW)
#define SEQ_LT(a, b) ((b) - (a) - 1 < MAX_WINDOW)
#define SEQ_GE(a, b) ((a) - (b) < MAX_WINDOW)
#define SEQ_GT(a, b) ((a) - (b) - 1 < MAX_WINDOW)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define FIN (1 << 0)
#define SYN (1 << 1)
#define RST (1 << 2)
#define ACK (1 << 4)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/* Flags for internal usage */
#define DUP_ACK (1 << 5)
#define ACK_IF_NEEDED 0 /* See tcp_send_flag() */
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define OPT_EOL 0
#define OPT_NOP 1
#define OPT_MSS 2
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
#define OPT_MSS_LEN 4
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define OPT_WS 3
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
#define OPT_WS_LEN 3
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define OPT_SACKP 4
#define OPT_SACK 5
#define OPT_TS 8
/**
* struct tcp_conn - Descriptor for a TCP connection (not spliced)
* @next_index: Connection index of next item in hash chain, -1 for none
* @tap_mss: MSS advertised by tap/guest, rounded to 2 ^ TCP_MSS_BITS
* @sock: Socket descriptor number
* @events: Connection events, implying connection states
* @timer: timerfd descriptor for timeout events
* @flags: Connection flags representing internal attributes
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @hash_bucket: Bucket index in connection lookup hash table
* @retrans: Number of retransmissions occurred due to ACK_TIMEOUT
* @ws_from_tap: Window scaling factor advertised from tap/guest
* @ws_to_tap: Window scaling factor advertised to tap/guest
* @sndbuf: Sending buffer in kernel, rounded to 2 ^ SNDBUF_BITS
* @seq_dup_ack_approx: Last duplicate ACK number sent to tap
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @a.a6: IPv6 remote address, can be IPv4-mapped
* @a.a4.zero: Zero prefix for IPv4-mapped, see RFC 6890, Table 20
* @a.a4.one: Ones prefix for IPv4-mapped
* @a.a4.a: IPv4 address
* @tap_port: Guest-facing tap port
* @sock_port: Remote, socket-facing port
* @wnd_from_tap: Last window size from tap, unscaled (as received)
* @wnd_to_tap: Sending window advertised to tap, unscaled (as sent)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @seq_to_tap: Next sequence for packets to tap
* @seq_ack_from_tap: Last ACK number received from tap
* @seq_from_tap: Next sequence for packets from tap (not actually sent)
* @seq_ack_to_tap: Last ACK number sent to tap
* @seq_init_from_tap: Initial sequence number from tap
*/
struct tcp_conn {
int next_index :TCP_CONN_INDEX_BITS + 2;
#define TCP_RETRANS_BITS 3
unsigned int retrans :TCP_RETRANS_BITS;
#define TCP_MAX_RETRANS ((1U << TCP_RETRANS_BITS) - 1)
#define TCP_WS_BITS 4 /* RFC 7323 */
#define TCP_WS_MAX 14
unsigned int ws_from_tap :TCP_WS_BITS;
unsigned int ws_to_tap :TCP_WS_BITS;
int sock :SOCKET_REF_BITS;
uint8_t events;
#define CLOSED 0
#define SOCK_ACCEPTED BIT(0) /* implies SYN sent to tap */
#define TAP_SYN_RCVD BIT(1) /* implies socket connecting */
#define TAP_SYN_ACK_SENT BIT( 3) /* implies socket connected */
#define ESTABLISHED BIT(2)
#define SOCK_FIN_RCVD BIT( 3)
#define SOCK_FIN_SENT BIT( 4)
#define TAP_FIN_RCVD BIT( 5)
#define TAP_FIN_SENT BIT( 6)
#define TAP_FIN_ACKED BIT( 7)
#define CONN_STATE_BITS /* Setting these clears other flags */ \
(SOCK_ACCEPTED | TAP_SYN_RCVD | ESTABLISHED)
int timer :SOCKET_REF_BITS;
uint8_t flags;
#define STALLED BIT(0)
#define LOCAL BIT(1)
#define WND_CLAMPED BIT(2)
#define IN_EPOLL BIT(3)
#define ACTIVE_CLOSE BIT(4)
#define ACK_TO_TAP_DUE BIT(5)
#define ACK_FROM_TAP_DUE BIT(6)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
unsigned int hash_bucket :TCP_HASH_BUCKET_BITS;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define TCP_MSS_BITS 14
unsigned int tap_mss :TCP_MSS_BITS;
#define MSS_SET(conn, mss) (conn->tap_mss = (mss >> (16 - TCP_MSS_BITS)))
#define MSS_GET(conn) (conn->tap_mss << (16 - TCP_MSS_BITS))
#define SNDBUF_BITS 24
unsigned int sndbuf :SNDBUF_BITS;
#define SNDBUF_SET(conn, bytes) (conn->sndbuf = ((bytes) >> (32 - SNDBUF_BITS)))
#define SNDBUF_GET(conn) (conn->sndbuf << (32 - SNDBUF_BITS))
uint8_t seq_dup_ack_approx;
union {
struct in6_addr a6;
struct {
uint8_t zero[10];
uint8_t one[2];
struct in_addr a;
} a4;
} a;
#define CONN_V4(conn) IN6_IS_ADDR_V4MAPPED(&conn->a.a6)
#define CONN_V6(conn) (!CONN_V4(conn))
in_port_t tap_port;
in_port_t sock_port;
uint16_t wnd_from_tap;
uint16_t wnd_to_tap;
uint32_t seq_to_tap;
uint32_t seq_ack_from_tap;
uint32_t seq_from_tap;
uint32_t seq_ack_to_tap;
uint32_t seq_init_from_tap;
};
#define CONN_IS_CLOSING(conn) \
((conn->events & ESTABLISHED) && \
(conn->events & (SOCK_FIN_RCVD | TAP_FIN_RCVD)))
#define CONN_HAS(conn, set) ((conn->events & (set)) == (set))
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
#define CONN(index) (tc + (index))
/* We probably don't want to use gcc statement expressions (for portability), so
* use this only after well-defined sequence points (no pre-/post-increments).
*/
#define CONN_OR_NULL(index) \
(((int)(index) >= 0 && (index) < TCP_MAX_CONNS) ? (tc + (index)) : NULL)
static const char *tcp_event_str[] __attribute((__unused__)) = {
"SOCK_ACCEPTED", "TAP_SYN_RCVD", "ESTABLISHED", "TAP_SYN_ACK_SENT",
"SOCK_FIN_RCVD", "SOCK_FIN_SENT", "TAP_FIN_RCVD", "TAP_FIN_SENT",
"TAP_FIN_ACKED",
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
};
static const char *tcp_state_str[] __attribute((__unused__)) = {
"SYN_RCVD", "SYN_SENT", "ESTABLISHED",
"SYN_RCVD", /* approximately maps to TAP_SYN_ACK_SENT */
/* Passive close: */
"CLOSE_WAIT", "CLOSE_WAIT", "LAST_ACK", "LAST_ACK", "LAST_ACK",
/* Active close (+5): */
"CLOSING", "FIN_WAIT_1", "FIN_WAIT_1", "FIN_WAIT_2", "TIME_WAIT",
};
static const char *tcp_flag_str[] __attribute((__unused__)) = {
"STALLED", "LOCAL", "WND_CLAMPED", "IN_EPOLL", "ACTIVE_CLOSE",
"ACK_TO_TAP_DUE", "ACK_FROM_TAP_DUE",
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
};
/* Listening sockets, used for automatic port forwarding in pasta mode only */
static int tcp_sock_init_lo [NUM_PORTS][IP_VERSIONS];
static int tcp_sock_init_ext [NUM_PORTS][IP_VERSIONS];
static int tcp_sock_ns [NUM_PORTS][IP_VERSIONS];
/* Table of destinations with very low RTT (assumed to be local), LRU */
static struct in6_addr low_rtt_dst[LOW_RTT_TABLE_SIZE];
/* Static buffers */
/**
* tcp4_l2_buf_t - Pre-cooked IPv4 packet buffers for tap connections
* @psum: Partial IP header checksum (excluding tot_len and saddr)
* @tsum: Partial TCP header checksum (excluding length and saddr)
* @pad: Align TCP header to 32 bytes, for AVX2 checksum calculation only
* @vnet_len: 4-byte qemu vnet buffer length descriptor, only for passt mode
* @eh: Pre-filled Ethernet header
* @iph: Pre-filled IP header (except for tot_len and saddr)
* @uh: Headroom for TCP header
* @data: Storage for TCP payload
*/
static struct tcp4_l2_buf_t {
uint32_t psum; /* 0 */
uint32_t tsum; /* 4 */
#ifdef __AVX2__
uint8_t pad[18]; /* 8, align th to 32 bytes */
#else
uint8_t pad[2]; /* align iph to 4 bytes 8 */
#endif
uint32_t vnet_len; /* 26 10 */
struct ethhdr eh; /* 30 14 */
struct iphdr iph; /* 44 28 */
struct tcphdr th; /* 64 48 */
uint8_t data[MSS4]; /* 84 68 */
/* 65541 65525 */
#ifdef __AVX2__
} __attribute__ ((packed, aligned(32)))
#else
} __attribute__ ((packed, aligned(__alignof__(unsigned int))))
#endif
tcp4_l2_buf[TCP_FRAMES_MEM];
static unsigned int tcp4_l2_buf_used;
static size_t tcp4_l2_buf_bytes;
/**
* tcp6_l2_buf_t - Pre-cooked IPv6 packet buffers for tap connections
* @pad: Align IPv6 header for checksum calculation to 32B (AVX2) or 4B
* @vnet_len: 4-byte qemu vnet buffer length descriptor, only for passt mode
* @eh: Pre-filled Ethernet header
* @ip6h: Pre-filled IP header (except for payload_len and addresses)
* @th: Headroom for TCP header
* @data: Storage for TCP payload
*/
struct tcp6_l2_buf_t {
#ifdef __AVX2__
uint8_t pad[14]; /* 0 align ip6h to 32 bytes */
#else
uint8_t pad[2]; /* align ip6h to 4 bytes 0 */
#endif
uint32_t vnet_len; /* 14 2 */
struct ethhdr eh; /* 18 6 */
struct ipv6hdr ip6h; /* 32 20 */
struct tcphdr th; /* 72 60 */
uint8_t data[MSS6]; /* 92 80 */
/* 65639 65627 */
#ifdef __AVX2__
} __attribute__ ((packed, aligned(32)))
#else
} __attribute__ ((packed, aligned(__alignof__(unsigned int))))
#endif
tcp6_l2_buf[TCP_FRAMES_MEM];
static unsigned int tcp6_l2_buf_used;
static size_t tcp6_l2_buf_bytes;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
/* recvmsg()/sendmsg() data for tap */
static char tcp_buf_discard [MAX_WINDOW];
static struct iovec iov_sock [TCP_FRAMES_MEM + 1];
static struct iovec tcp4_l2_iov [TCP_FRAMES_MEM];
static struct iovec tcp6_l2_iov [TCP_FRAMES_MEM];
static struct iovec tcp4_l2_flags_iov [TCP_FRAMES_MEM];
static struct iovec tcp6_l2_flags_iov [TCP_FRAMES_MEM];
static struct mmsghdr tcp_l2_mh [TCP_FRAMES_MEM];
/* sendmsg() to socket */
static struct iovec tcp_iov [UIO_MAXIOV];
/**
* tcp4_l2_flags_buf_t - IPv4 packet buffers for segments without data (flags)
* @psum: Partial IP header checksum (excluding tot_len and saddr)
* @tsum: Partial TCP header checksum (excluding length and saddr)
* @pad: Align TCP header to 32 bytes, for AVX2 checksum calculation only
* @vnet_len: 4-byte qemu vnet buffer length descriptor, only for passt mode
* @eh: Pre-filled Ethernet header
* @iph: Pre-filled IP header (except for tot_len and saddr)
* @th: Headroom for TCP header
* @opts: Headroom for TCP options
*/
static struct tcp4_l2_flags_buf_t {
uint32_t psum; /* 0 */
uint32_t tsum; /* 4 */
#ifdef __AVX2__
uint8_t pad[18]; /* 8, align th to 32 bytes */
#else
uint8_t pad[2]; /* align iph to 4 bytes 8 */
#endif
uint32_t vnet_len; /* 26 10 */
struct ethhdr eh; /* 30 14 */
struct iphdr iph; /* 44 28 */
struct tcphdr th; /* 64 48 */
char opts[OPT_MSS_LEN + OPT_WS_LEN + 1];
#ifdef __AVX2__
} __attribute__ ((packed, aligned(32)))
#else
} __attribute__ ((packed, aligned(__alignof__(unsigned int))))
#endif
tcp4_l2_flags_buf[TCP_FRAMES_MEM];
static unsigned int tcp4_l2_flags_buf_used;
static size_t tcp4_l2_flags_buf_bytes;
/**
* tcp6_l2_flags_buf_t - IPv6 packet buffers for segments without data (flags)
* @pad: Align IPv6 header for checksum calculation to 32B (AVX2) or 4B
* @vnet_len: 4-byte qemu vnet buffer length descriptor, only for passt mode
* @eh: Pre-filled Ethernet header
* @ip6h: Pre-filled IP header (except for payload_len and addresses)
* @th: Headroom for TCP header
* @opts: Headroom for TCP options
*/
static struct tcp6_l2_flags_buf_t {
#ifdef __AVX2__
uint8_t pad[14]; /* 0 align ip6h to 32 bytes */
#else
uint8_t pad[2]; /* align ip6h to 4 bytes 0 */
#endif
uint32_t vnet_len; /* 14 2 */
struct ethhdr eh; /* 18 6 */
struct ipv6hdr ip6h; /* 32 20 */
struct tcphdr th /* 72 */ __attribute__ ((aligned(4))); /* 60 */
char opts[OPT_MSS_LEN + OPT_WS_LEN + 1];
#ifdef __AVX2__
} __attribute__ ((packed, aligned(32)))
#else
} __attribute__ ((packed, aligned(__alignof__(unsigned int))))
#endif
tcp6_l2_flags_buf[TCP_FRAMES_MEM];
static unsigned int tcp6_l2_flags_buf_used;
static size_t tcp6_l2_flags_buf_bytes;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/* TCP connections */
static struct tcp_conn tc[TCP_MAX_CONNS];
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/* Table for lookup from remote address, local port, remote port */
static struct tcp_conn *tc_hash[TCP_HASH_TABLE_SIZE];
/* Pools for pre-opened sockets */
int init_sock_pool4 [TCP_SOCK_POOL_SIZE];
int init_sock_pool6 [TCP_SOCK_POOL_SIZE];
int ns_sock_pool4 [TCP_SOCK_POOL_SIZE];
int ns_sock_pool6 [TCP_SOCK_POOL_SIZE];
/**
* tcp_conn_epoll_events() - epoll events mask for given connection state
* @events: Current connection events
* @conn_flags Connection flags
*
* Return: epoll events mask corresponding to implied connection state
*/
static uint32_t tcp_conn_epoll_events(uint8_t events, uint8_t conn_flags)
{
if (!events)
return 0;
if (events & ESTABLISHED) {
if (events & TAP_FIN_SENT)
return EPOLLET;
if (conn_flags & STALLED)
return EPOLLIN | EPOLLOUT | EPOLLRDHUP | EPOLLET;
return EPOLLIN | EPOLLRDHUP;
}
if (events == TAP_SYN_RCVD)
return EPOLLOUT | EPOLLET | EPOLLRDHUP;
return EPOLLRDHUP;
}
static void conn_flag_do(const struct ctx *c, struct tcp_conn *conn,
unsigned long flag);
#define conn_flag(c, conn, flag) \
do { \
trace("TCP: flag at %s:%i", __func__, __LINE__); \
conn_flag_do(c, conn, flag); \
} while (0)
/**
* tcp_epoll_ctl() - Add/modify/delete epoll state from connection events
* @c: Execution context
* @conn: Connection pointer
*
* Return: 0 on success, negative error code on failure (not on deletion)
*/
static int tcp_epoll_ctl(const struct ctx *c, struct tcp_conn *conn)
{
int m = (conn->flags & IN_EPOLL) ? EPOLL_CTL_MOD : EPOLL_CTL_ADD;
union epoll_ref ref = { .r.proto = IPPROTO_TCP, .r.s = conn->sock,
.r.p.tcp.tcp.index = conn - tc,
.r.p.tcp.tcp.v6 = CONN_V6(conn) };
struct epoll_event ev = { .data.u64 = ref.u64 };
if (conn->events == CLOSED) {
if (conn->flags & IN_EPOLL)
epoll_ctl(c->epollfd, EPOLL_CTL_DEL, conn->sock, &ev);
if (conn->timer != -1)
epoll_ctl(c->epollfd, EPOLL_CTL_DEL, conn->timer, &ev);
return 0;
}
ev.events = tcp_conn_epoll_events(conn->events, conn->flags);
if (epoll_ctl(c->epollfd, m, conn->sock, &ev))
return -errno;
conn->flags |= IN_EPOLL; /* No need to log this */
if (conn->timer != -1) {
union epoll_ref ref_t = { .r.proto = IPPROTO_TCP,
.r.s = conn->sock,
.r.p.tcp.tcp.timer = 1,
.r.p.tcp.tcp.index = conn - tc };
struct epoll_event ev_t = { .data.u64 = ref_t.u64,
.events = EPOLLIN | EPOLLET };
if (epoll_ctl(c->epollfd, EPOLL_CTL_MOD, conn->timer, &ev_t))
return -errno;
}
return 0;
}
/**
* tcp_timer_ctl() - Set timerfd based on flags/events, create timerfd if needed
* @c: Execution context
* @conn: Connection pointer
*
* #syscalls timerfd_create timerfd_settime
*/
static void tcp_timer_ctl(const struct ctx *c, struct tcp_conn *conn)
{
struct itimerspec it = { { 0 }, { 0 } };
if (conn->events == CLOSED)
return;
if (conn->timer == -1) {
union epoll_ref ref = { .r.proto = IPPROTO_TCP,
.r.s = conn->sock,
.r.p.tcp.tcp.timer = 1,
.r.p.tcp.tcp.index = conn - tc };
struct epoll_event ev = { .data.u64 = ref.u64,
.events = EPOLLIN | EPOLLET };
int fd;
fd = timerfd_create(CLOCK_MONOTONIC, 0);
if (fd == -1 || fd > SOCKET_MAX) {
debug("TCP: failed to get timer: %s", strerror(errno));
return;
}
conn->timer = fd;
if (epoll_ctl(c->epollfd, EPOLL_CTL_ADD, conn->timer, &ev)) {
debug("TCP: failed to add timer: %s", strerror(errno));
close(conn->timer);
conn->timer = -1;
return;
}
}
if (conn->flags & ACK_TO_TAP_DUE) {
it.it_value.tv_nsec = (long)ACK_INTERVAL * 1000 * 1000;
} else if (conn->flags & ACK_FROM_TAP_DUE) {
if (!(conn->events & ESTABLISHED))
it.it_value.tv_sec = SYN_TIMEOUT;
else
it.it_value.tv_sec = ACK_TIMEOUT;
} else if (CONN_HAS(conn, SOCK_FIN_SENT | TAP_FIN_ACKED)) {
it.it_value.tv_sec = FIN_TIMEOUT;
} else {
it.it_value.tv_sec = ACT_TIMEOUT;
}
debug("TCP: index %li, timer expires in %lu.%03lus", conn - tc,
it.it_value.tv_sec, it.it_value.tv_nsec / 1000 / 1000);
timerfd_settime(conn->timer, 0, &it, NULL);
}
/**
* conn_flag_do() - Set/unset given flag, log, update epoll on STALLED flag
* @c: Execution context
* @conn: Connection pointer
* @flag: Flag to set, or ~flag to unset
*/
static void conn_flag_do(const struct ctx *c, struct tcp_conn *conn,
unsigned long flag)
{
if (flag & (flag - 1)) {
if (!(conn->flags & ~flag))
return;
conn->flags &= flag;
if (fls(~flag) >= 0) {
debug("TCP: index %li: %s dropped", conn - tc,
tcp_flag_str[fls(~flag)]);
}
} else {
if (conn->flags & flag)
return;
conn->flags |= flag;
if (fls(flag) >= 0) {
debug("TCP: index %li: %s", conn - tc,
tcp_flag_str[fls(flag)]);
}
}
if (flag == STALLED || flag == ~STALLED)
tcp_epoll_ctl(c, conn);
if (flag == ACK_FROM_TAP_DUE || flag == ACK_TO_TAP_DUE ||
(flag == ~ACK_FROM_TAP_DUE && (conn->flags & ACK_TO_TAP_DUE)) ||
(flag == ~ACK_TO_TAP_DUE && (conn->flags & ACK_FROM_TAP_DUE)))
tcp_timer_ctl(c, conn);
}
/**
* conn_event_do() - Set and log connection events, update epoll state
* @c: Execution context
* @conn: Connection pointer
* @event: Connection event
*/
static void conn_event_do(const struct ctx *c, struct tcp_conn *conn,
unsigned long event)
{
int prev, new, num = fls(event);
if (conn->events & event)
return;
prev = fls(conn->events);
if (conn->flags & ACTIVE_CLOSE)
prev += 5;
if ((conn->events & ESTABLISHED) && (conn->events != ESTABLISHED))
prev++; /* i.e. SOCK_FIN_RCVD, not TAP_SYN_ACK_SENT */
if (event == CLOSED || (event & CONN_STATE_BITS))
conn->events = event;
else
conn->events |= event;
new = fls(conn->events);
if ((conn->events & ESTABLISHED) && (conn->events != ESTABLISHED)) {
num++;
new++;
}
if (conn->flags & ACTIVE_CLOSE)
new += 5;
if (prev != new) {
debug("TCP: index %li, %s: %s -> %s", conn - tc,
num == -1 ? "CLOSED" : tcp_event_str[num],
prev == -1 ? "CLOSED" : tcp_state_str[prev],
(new == -1 || num == -1) ? "CLOSED" : tcp_state_str[new]);
} else {
debug("TCP: index %li, %s", conn - tc,
num == -1 ? "CLOSED" : tcp_event_str[num]);
}
if ((event == TAP_FIN_RCVD) && !(conn->events & SOCK_FIN_RCVD))
conn_flag(c, conn, ACTIVE_CLOSE);
else
tcp_epoll_ctl(c, conn);
if (CONN_HAS(conn, SOCK_FIN_SENT | TAP_FIN_ACKED))
tcp_timer_ctl(c, conn);
}
#define conn_event(c, conn, event) \
do { \
trace("TCP: event at %s:%i", __func__, __LINE__); \
conn_event_do(c, conn, event); \
} while (0)
/**
* tcp_rtt_dst_low() - Check if low RTT was seen for connection endpoint
* @conn: Connection pointer
*
* Return: 1 if destination is in low RTT table, 0 otherwise
*/
static int tcp_rtt_dst_low(const struct tcp_conn *conn)
{
int i;
for (i = 0; i < LOW_RTT_TABLE_SIZE; i++)
if (IN6_ARE_ADDR_EQUAL(&conn->a.a6, low_rtt_dst + i))
return 1;
return 0;
}
/**
* tcp_rtt_dst_check() - Check tcpi_min_rtt, insert endpoint in table if low
* @conn: Connection pointer
* @tinfo: Pointer to struct tcp_info for socket
*/
static void tcp_rtt_dst_check(const struct tcp_conn *conn,
const struct tcp_info *tinfo)
{
#ifdef HAS_MIN_RTT
int i, hole = -1;
if (!tinfo->tcpi_min_rtt ||
(int)tinfo->tcpi_min_rtt > LOW_RTT_THRESHOLD)
return;
for (i = 0; i < LOW_RTT_TABLE_SIZE; i++) {
if (IN6_ARE_ADDR_EQUAL(&conn->a.a6, low_rtt_dst + i))
return;
if (hole == -1 && IN6_IS_ADDR_UNSPECIFIED(low_rtt_dst + i))
hole = i;
}
/* Keep gcc 12 happy: this won't actually happen because the table is
* guaranteed to have a hole, see the second memcpy() below.
*/
if (hole == -1)
return;
memcpy(low_rtt_dst + hole++, &conn->a.a6, sizeof(conn->a.a6));
if (hole == LOW_RTT_TABLE_SIZE)
hole = 0;
memcpy(low_rtt_dst + hole, &in6addr_any, sizeof(conn->a.a6));
#else
(void)conn;
(void)tinfo;
#endif /* HAS_MIN_RTT */
}
/**
* tcp_get_sndbuf() - Get, scale SO_SNDBUF between thresholds (1 to 0.5 usage)
* @conn: Connection pointer
*/
static void tcp_get_sndbuf(struct tcp_conn *conn)
{
int s = conn->sock, sndbuf;
socklen_t sl;
uint64_t v;
sl = sizeof(sndbuf);
if (getsockopt(s, SOL_SOCKET, SO_SNDBUF, &sndbuf, &sl)) {
SNDBUF_SET(conn, WINDOW_DEFAULT);
return;
}
v = sndbuf;
if (v >= SNDBUF_BIG)
v /= 2;
else if (v > SNDBUF_SMALL)
v -= v * (v - SNDBUF_SMALL) / (SNDBUF_BIG - SNDBUF_SMALL) / 2;
SNDBUF_SET(conn, MIN(INT_MAX, v));
}
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
/**
* tcp_sock_set_bufsize() - Set SO_RCVBUF and SO_SNDBUF to maximum values
* @s: Socket, can be -1 to avoid check in the caller
*/
void tcp_sock_set_bufsize(const struct ctx *c, int s)
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
{
int v = INT_MAX / 2; /* Kernel clamps and rounds, no need to check */
if (s == -1)
return;
if (!c->low_rmem && setsockopt(s, SOL_SOCKET, SO_RCVBUF, &v, sizeof(v)))
trace("TCP: failed to set SO_RCVBUF to %i", v);
if (!c->low_wmem && setsockopt(s, SOL_SOCKET, SO_SNDBUF, &v, sizeof(v)))
trace("TCP: failed to set SO_SNDBUF to %i", v);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
/**
* tcp_update_check_ip4() - Update IPv4 with variable parts from stored one
* @buf: L2 packet buffer with final IPv4 header
*/
static void tcp_update_check_ip4(struct tcp4_l2_buf_t *buf)
{
uint32_t sum = buf->psum;
sum += buf->iph.tot_len;
sum += (buf->iph.saddr >> 16) & 0xffff;
sum += buf->iph.saddr & 0xffff;
buf->iph.check = (uint16_t)~csum_fold(sum);
}
/**
* tcp_update_check_tcp4() - Update TCP checksum from stored one
* @buf: L2 packet buffer with final IPv4 header
*/
static void tcp_update_check_tcp4(struct tcp4_l2_buf_t *buf)
{
uint16_t tlen = ntohs(buf->iph.tot_len) - 20;
uint32_t sum = buf->tsum;
sum += (buf->iph.saddr >> 16) & 0xffff;
sum += buf->iph.saddr & 0xffff;
sum += htons(ntohs(buf->iph.tot_len) - 20);
buf->th.check = 0;
buf->th.check = csum(&buf->th, tlen, sum);
}
/**
* tcp_update_check_tcp6() - Calculate TCP checksum for IPv6
* @buf: L2 packet buffer with final IPv6 header
*/
static void tcp_update_check_tcp6(struct tcp6_l2_buf_t *buf)
{
int len = ntohs(buf->ip6h.payload_len) + sizeof(struct ipv6hdr);
buf->ip6h.hop_limit = IPPROTO_TCP;
buf->ip6h.version = 0;
buf->ip6h.nexthdr = 0;
buf->th.check = 0;
buf->th.check = csum(&buf->ip6h, len, 0);
buf->ip6h.hop_limit = 255;
buf->ip6h.version = 6;
buf->ip6h.nexthdr = IPPROTO_TCP;
}
/**
* tcp_update_l2_buf() - Update L2 buffers with Ethernet and IPv4 addresses
* @eth_d: Ethernet destination address, NULL if unchanged
* @eth_s: Ethernet source address, NULL if unchanged
* @ip_da: Pointer to IPv4 destination address, NULL if unchanged
*/
void tcp_update_l2_buf(const unsigned char *eth_d, const unsigned char *eth_s,
const uint32_t *ip_da)
{
int i;
for (i = 0; i < TCP_FRAMES_MEM; i++) {
struct tcp4_l2_flags_buf_t *b4f = &tcp4_l2_flags_buf[i];
struct tcp6_l2_flags_buf_t *b6f = &tcp6_l2_flags_buf[i];
struct tcp4_l2_buf_t *b4 = &tcp4_l2_buf[i];
struct tcp6_l2_buf_t *b6 = &tcp6_l2_buf[i];
if (eth_d) {
memcpy(b4->eh.h_dest, eth_d, ETH_ALEN);
memcpy(b6->eh.h_dest, eth_d, ETH_ALEN);
memcpy(b4f->eh.h_dest, eth_d, ETH_ALEN);
memcpy(b6f->eh.h_dest, eth_d, ETH_ALEN);
}
if (eth_s) {
memcpy(b4->eh.h_source, eth_s, ETH_ALEN);
memcpy(b6->eh.h_source, eth_s, ETH_ALEN);
memcpy(b4f->eh.h_source, eth_s, ETH_ALEN);
memcpy(b6f->eh.h_source, eth_s, ETH_ALEN);
}
if (ip_da) {
b4f->iph.daddr = b4->iph.daddr = *ip_da;
if (!i) {
b4f->iph.saddr = b4->iph.saddr = 0;
b4f->iph.tot_len = b4->iph.tot_len = 0;
b4f->iph.check = b4->iph.check = 0;
b4f->psum = b4->psum = sum_16b(&b4->iph, 20);
b4->tsum = ((*ip_da >> 16) & 0xffff) +
(*ip_da & 0xffff) +
htons(IPPROTO_TCP);
b4f->tsum = b4->tsum;
} else {
b4f->psum = b4->psum = tcp4_l2_buf[0].psum;
b4f->tsum = b4->tsum = tcp4_l2_buf[0].tsum;
}
}
}
}
/**
* tcp_sock4_iov_init() - Initialise scatter-gather L2 buffers for IPv4 sockets
*/
static void tcp_sock4_iov_init(void)
{
struct iovec *iov;
int i;
for (i = 0; i < ARRAY_SIZE(tcp4_l2_buf); i++) {
tcp4_l2_buf[i] = (struct tcp4_l2_buf_t) { 0, 0,
{ 0 },
0, L2_BUF_ETH_IP4_INIT, L2_BUF_IP4_INIT(IPPROTO_TCP),
{ .doff = sizeof(struct tcphdr) / 4, .ack = 1 }, { 0 },
};
}
for (i = 0; i < ARRAY_SIZE(tcp4_l2_flags_buf); i++) {
tcp4_l2_flags_buf[i] = (struct tcp4_l2_flags_buf_t) { 0, 0,
{ 0 },
0, L2_BUF_ETH_IP4_INIT, L2_BUF_IP4_INIT(IPPROTO_TCP),
{ 0 }, { 0 },
};
}
for (i = 0, iov = tcp4_l2_iov; i < TCP_FRAMES_MEM; i++, iov++) {
iov->iov_base = &tcp4_l2_buf[i].vnet_len;
iov->iov_len = MSS_DEFAULT;
}
for (i = 0, iov = tcp4_l2_flags_iov; i < TCP_FRAMES_MEM; i++, iov++)
iov->iov_base = &tcp4_l2_flags_buf[i].vnet_len;
}
/**
* tcp_sock6_iov_init() - Initialise scatter-gather L2 buffers for IPv6 sockets
*/
static void tcp_sock6_iov_init(void)
{
struct iovec *iov;
int i;
for (i = 0; i < ARRAY_SIZE(tcp6_l2_buf); i++) {
tcp6_l2_buf[i] = (struct tcp6_l2_buf_t) {
{ 0 },
0, L2_BUF_ETH_IP6_INIT, L2_BUF_IP6_INIT(IPPROTO_TCP),
{ .doff = sizeof(struct tcphdr) / 4, .ack = 1 }, { 0 },
};
}
for (i = 0; i < ARRAY_SIZE(tcp6_l2_flags_buf); i++) {
tcp6_l2_flags_buf[i] = (struct tcp6_l2_flags_buf_t) {
{ 0 },
0, L2_BUF_ETH_IP6_INIT, L2_BUF_IP6_INIT(IPPROTO_TCP),
{ 0 }, { 0 },
};
}
for (i = 0, iov = tcp6_l2_iov; i < TCP_FRAMES_MEM; i++, iov++) {
iov->iov_base = &tcp6_l2_buf[i].vnet_len;
iov->iov_len = MSS_DEFAULT;
}
for (i = 0, iov = tcp6_l2_flags_iov; i < TCP_FRAMES_MEM; i++, iov++)
iov->iov_base = &tcp6_l2_flags_buf[i].vnet_len;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_opt_get() - Get option, and value if any, from TCP header
* @opts: Pointer to start of TCP options in header
* @len: Length of buffer, excluding TCP header -- NOT checked here!
* @type_find: Option type to look for
* @optlen_set: Optional, filled with option length if passed
* @value_set: Optional, set to start of option value if passed
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* Return: option value, meaningful for up to 4 bytes, -1 if not found
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static int tcp_opt_get(const char *opts, size_t len, uint8_t type_find,
uint8_t *optlen_set, const char **value_set)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
uint8_t type, optlen;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (!len)
return -1;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
for (; len >= 2; opts += optlen, len -= optlen) {
switch (*opts) {
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
case OPT_EOL:
return -1;
case OPT_NOP:
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
optlen = 1;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
break;
default:
type = *(opts++);
optlen = *(opts++) - 2;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
len -= 2;
if (type != type_find)
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
break;
if (optlen_set)
*optlen_set = optlen;
if (value_set)
*value_set = opts;
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
switch (optlen) {
case 0:
return 0;
case 1:
return *opts;
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
case 2:
return ntohs(*(uint16_t *)opts);
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
default:
return ntohl(*(uint32_t *)opts);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
}
}
return -1;
}
/**
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* tcp_hash_match() - Check if a connection entry matches address and ports
* @conn: Connection entry to match against
* @af: Address family, AF_INET or AF_INET6
* @addr: Remote address, pointer to sin_addr or sin6_addr
* @tap_port: tap-facing port
* @sock_port: Socket-facing port
*
* Return: 1 on match, 0 otherwise
*/
static int tcp_hash_match(const struct tcp_conn *conn, int af, const void *addr,
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
in_port_t tap_port, in_port_t sock_port)
{
if (af == AF_INET && CONN_V4(conn) &&
!memcmp(&conn->a.a4.a, addr, sizeof(conn->a.a4.a)) &&
conn->tap_port == tap_port && conn->sock_port == sock_port)
return 1;
if (af == AF_INET6 &&
IN6_ARE_ADDR_EQUAL(&conn->a.a6, addr) &&
conn->tap_port == tap_port && conn->sock_port == sock_port)
return 1;
return 0;
}
/**
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* tcp_hash() - Calculate hash value for connection given address and ports
* @c: Execution context
* @af: Address family, AF_INET or AF_INET6
* @addr: Remote address, pointer to sin_addr or sin6_addr
* @tap_port: tap-facing port
* @sock_port: Socket-facing port
*
* Return: hash value, already modulo size of the hash table
*/
#if TCP_HASH_NOINLINE
__attribute__((__noinline__)) /* See comment in Makefile */
#endif
static unsigned int tcp_hash(const struct ctx *c, int af, const void *addr,
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
in_port_t tap_port, in_port_t sock_port)
{
uint64_t b = 0;
if (af == AF_INET) {
struct {
struct in_addr addr;
in_port_t tap_port;
in_port_t sock_port;
} __attribute__((__packed__)) in = {
*(struct in_addr *)addr, tap_port, sock_port,
};
b = siphash_8b((uint8_t *)&in, c->tcp.hash_secret);
} else if (af == AF_INET6) {
struct {
struct in6_addr addr;
in_port_t tap_port;
in_port_t sock_port;
} __attribute__((__packed__)) in = {
*(struct in6_addr *)addr, tap_port, sock_port,
};
b = siphash_20b((uint8_t *)&in, c->tcp.hash_secret);
}
return (unsigned int)(b % TCP_HASH_TABLE_SIZE);
}
/**
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* tcp_hash_insert() - Insert connection into hash table, chain link
* @c: Execution context
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @conn: Connection pointer
* @af: Address family, AF_INET or AF_INET6
* @addr: Remote address, pointer to sin_addr or sin6_addr
*/
static void tcp_hash_insert(const struct ctx *c, struct tcp_conn *conn,
int af, const void *addr)
{
int b;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
b = tcp_hash(c, af, addr, conn->tap_port, conn->sock_port);
conn->next_index = tc_hash[b] ? tc_hash[b] - tc : -1;
tc_hash[b] = conn;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->hash_bucket = b;
debug("TCP: hash table insert: index %li, sock %i, bucket: %i, next: "
"%p", conn - tc, conn->sock, b, CONN_OR_NULL(conn->next_index));
}
/**
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* tcp_hash_remove() - Drop connection from hash table, chain unlink
* @conn: Connection pointer
*/
static void tcp_hash_remove(const struct tcp_conn *conn)
{
struct tcp_conn *entry, *prev = NULL;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
int b = conn->hash_bucket;
for (entry = tc_hash[b]; entry;
prev = entry, entry = CONN_OR_NULL(entry->next_index)) {
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (entry == conn) {
if (prev)
prev->next_index = conn->next_index;
else
tc_hash[b] = CONN_OR_NULL(conn->next_index);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
break;
}
}
debug("TCP: hash table remove: index %li, sock %i, bucket: %i, new: %p",
conn - tc, conn->sock, b,
prev ? CONN_OR_NULL(prev->next_index) : tc_hash[b]);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
}
/**
* tcp_hash_update() - Update pointer for given connection
* @old: Old connection pointer
* @new: New connection pointer
*/
static void tcp_hash_update(struct tcp_conn *old, struct tcp_conn *new)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
{
struct tcp_conn *entry, *prev = NULL;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
int b = old->hash_bucket;
for (entry = tc_hash[b]; entry;
prev = entry, entry = CONN_OR_NULL(entry->next_index)) {
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (entry == old) {
if (prev)
prev->next_index = new - tc;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
else
tc_hash[b] = new;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
break;
}
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
debug("TCP: hash table update: old index %li, new index %li, sock %i, "
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
"bucket: %i, old: %p, new: %p",
old - tc, new - tc, new->sock, b, old, new);
}
/**
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* tcp_hash_lookup() - Look up connection given remote address and ports
* @c: Execution context
* @af: Address family, AF_INET or AF_INET6
* @addr: Remote address, pointer to sin_addr or sin6_addr
* @tap_port: tap-facing port
* @sock_port: Socket-facing port
*
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* Return: connection pointer, if found, -ENOENT otherwise
*/
static struct tcp_conn *tcp_hash_lookup(const struct ctx *c, int af,
const void *addr,
in_port_t tap_port, in_port_t sock_port)
{
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
int b = tcp_hash(c, af, addr, tap_port, sock_port);
struct tcp_conn *conn;
for (conn = tc_hash[b]; conn; conn = CONN_OR_NULL(conn->next_index)) {
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (tcp_hash_match(conn, af, addr, tap_port, sock_port))
return conn;
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
return NULL;
}
/**
* tcp_table_compact() - Perform compaction on connection table
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @c: Execution context
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @hole: Pointer to recently closed connection
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static void tcp_table_compact(struct ctx *c, struct tcp_conn *hole)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
struct tcp_conn *from, *to;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if ((hole - tc) == --c->tcp.conn_count) {
debug("TCP: hash table compaction: maximum index was %li (%p)",
hole - tc, hole);
memset(hole, 0, sizeof(*hole));
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
return;
}
from = CONN(c->tcp.conn_count);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
memcpy(hole, from, sizeof(*hole));
to = hole;
tcp_hash_update(from, to);
tcp_epoll_ctl(c, to);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
debug("TCP: hash table compaction: old index %li, new index %li, "
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
"sock %i, from: %p, to: %p",
from - tc, to - tc, from->sock, from, to);
memset(from, 0, sizeof(*from));
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
/**
* tcp_conn_destroy() - Close sockets, trigger hash table removal and compaction
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @c: Execution context
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @conn: Connection pointer
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static void tcp_conn_destroy(struct ctx *c, struct tcp_conn *conn)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
close(conn->sock);
if (conn->timer != -1)
close(conn->timer);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
tcp_hash_remove(conn);
tcp_table_compact(c, conn);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
static void tcp_rst_do(struct ctx *c, struct tcp_conn *conn);
#define tcp_rst(c, conn) \
do { \
debug("TCP: index %li, reset at %s:%i", conn - tc, \
__func__, __LINE__); \
tcp_rst_do(c, conn); \
} while (0)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/**
* tcp_l2_buf_write_one() - Write a single buffer to tap file descriptor
* @c: Execution context
* @iov: struct iovec item pointing to buffer
* @ts: Current timestamp
*
* Return: 0 on success, negative error code on failure (tap reset possible)
*/
static int tcp_l2_buf_write_one(struct ctx *c, const struct iovec *iov)
{
if (write(c->fd_tap, (char *)iov->iov_base + 4, iov->iov_len - 4) < 0) {
debug("tap write: %s", strerror(errno));
if (errno != EAGAIN && errno != EWOULDBLOCK)
tap_handler(c, c->fd_tap, EPOLLERR, NULL);
return -errno;
}
return 0;
}
/**
* tcp_l2_buf_flush_part() - Ensure a complete last message on partial sendmsg()
* @c: Execution context
* @mh: Message header that was partially sent by sendmsg()
* @sent: Bytes already sent
*/
static void tcp_l2_buf_flush_part(const struct ctx *c,
const struct msghdr *mh, size_t sent)
{
size_t end = 0, missing;
struct iovec *iov;
unsigned int i;
char *p;
for (i = 0, iov = mh->msg_iov; i < mh->msg_iovlen; i++, iov++) {
end += iov->iov_len;
if (end >= sent)
break;
}
missing = end - sent;
p = (char *)iov->iov_base + iov->iov_len - missing;
if (send(c->fd_tap, p, missing, MSG_NOSIGNAL))
debug("TCP: failed to flush %lu missing bytes to tap", missing);
}
/**
* tcp_l2_flags_buf_flush() - Send out buffers for segments with or without data
* @c: Execution context
* @mh: Message header pointing to buffers, msg_iovlen not set
* @buf_used: Pointer to count of used buffers, set to 0 on return
* @buf_bytes: Pointer to count of buffer bytes, set to 0 on return
*/
static void tcp_l2_buf_flush(struct ctx *c, struct msghdr *mh,
unsigned int *buf_used, size_t *buf_bytes)
{
if (!(mh->msg_iovlen = *buf_used))
return;
if (c->mode == MODE_PASST) {
size_t n = sendmsg(c->fd_tap, mh, MSG_NOSIGNAL | MSG_DONTWAIT);
if (n > 0 && n < *buf_bytes)
tcp_l2_buf_flush_part(c, mh, n);
} else {
size_t i;
for (i = 0; i < mh->msg_iovlen; i++) {
struct iovec *iov = &mh->msg_iov[i];
if (tcp_l2_buf_write_one(c, iov))
i--;
}
}
*buf_used = *buf_bytes = 0;
pcapm(mh);
}
/**
* tcp_l2_flags_buf_flush() - Send out buffers for segments with no data (flags)
* @c: Execution context
*/
static void tcp_l2_flags_buf_flush(struct ctx *c)
{
struct msghdr mh = { 0 };
unsigned int *buf_used;
size_t *buf_bytes;
mh.msg_iov = tcp6_l2_flags_iov;
buf_used = &tcp6_l2_flags_buf_used;
buf_bytes = &tcp6_l2_flags_buf_bytes;
tcp_l2_buf_flush(c, &mh, buf_used, buf_bytes);
mh.msg_iov = tcp4_l2_flags_iov;
buf_used = &tcp4_l2_flags_buf_used;
buf_bytes = &tcp4_l2_flags_buf_bytes;
tcp_l2_buf_flush(c, &mh, buf_used, buf_bytes);
}
/**
* tcp_l2_data_buf_flush() - Send out buffers for segments with data
* @c: Execution context
*/
static void tcp_l2_data_buf_flush(struct ctx *c)
{
struct msghdr mh = { 0 };
unsigned int *buf_used;
size_t *buf_bytes;
mh.msg_iov = tcp6_l2_iov;
buf_used = &tcp6_l2_buf_used;
buf_bytes = &tcp6_l2_buf_bytes;
tcp_l2_buf_flush(c, &mh, buf_used, buf_bytes);
mh.msg_iov = tcp4_l2_iov;
buf_used = &tcp4_l2_buf_used;
buf_bytes = &tcp4_l2_buf_bytes;
tcp_l2_buf_flush(c, &mh, buf_used, buf_bytes);
}
/**
* tcp_defer_handler() - Handler for TCP deferred tasks
* @c: Execution context
*/
void tcp_defer_handler(struct ctx *c)
{
int max_conns = c->tcp.conn_count / 100 * TCP_CONN_PRESSURE;
int max_files = c->nofile / 100 * TCP_FILE_PRESSURE;
struct tcp_conn *conn;
tcp_l2_flags_buf_flush(c);
tcp_l2_data_buf_flush(c);
tcp_splice_defer_handler(c);
if (c->tcp.conn_count < MIN(max_files, max_conns))
return;
for (conn = CONN(c->tcp.conn_count - 1); conn >= tc; conn--) {
if (conn->events == CLOSED)
tcp_conn_destroy(c, conn);
}
}
/**
* tcp_l2_buf_fill_headers() - Fill 802.3, IP, TCP headers in pre-cooked buffers
* @c: Execution context
* @conn: Connection pointer
* @p: Pointer to any type of TCP pre-cooked buffer
* @plen: Payload length (including TCP header options)
* @check: Checksum, if already known
* @seq: Sequence number for this segment
*
* Return: 802.3 length, host order
*/
static size_t tcp_l2_buf_fill_headers(const struct ctx *c,
const struct tcp_conn *conn,
void *p, size_t plen,
const uint16_t *check, uint32_t seq)
{
size_t ip_len, eth_len;
#define SET_TCP_HEADER_COMMON_V4_V6(b, conn, seq) \
do { \
b->th.source = htons(conn->sock_port); \
b->th.dest = htons(conn->tap_port); \
b->th.seq = htonl(seq); \
b->th.ack_seq = htonl(conn->seq_ack_to_tap); \
if (conn->events & ESTABLISHED) { \
b->th.window = htons(conn->wnd_to_tap); \
} else { \
unsigned wnd = conn->wnd_to_tap << conn->ws_to_tap; \
\
b->th.window = htons(MIN(wnd, USHRT_MAX)); \
} \
} while (0)
if (CONN_V6(conn)) {
struct tcp6_l2_buf_t *b = (struct tcp6_l2_buf_t *)p;
ip_len = plen + sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
b->ip6h.payload_len = htons(plen + sizeof(struct tcphdr));
b->ip6h.saddr = conn->a.a6;
if (IN6_IS_ADDR_LINKLOCAL(&b->ip6h.saddr))
b->ip6h.daddr = c->ip6.addr_ll_seen;
else
b->ip6h.daddr = c->ip6.addr_seen;
memset(b->ip6h.flow_lbl, 0, 3);
SET_TCP_HEADER_COMMON_V4_V6(b, conn, seq);
tcp_update_check_tcp6(b);
b->ip6h.flow_lbl[0] = (conn->sock >> 16) & 0xf;
b->ip6h.flow_lbl[1] = (conn->sock >> 8) & 0xff;
b->ip6h.flow_lbl[2] = (conn->sock >> 0) & 0xff;
eth_len = ip_len + sizeof(struct ethhdr);
if (c->mode == MODE_PASST)
b->vnet_len = htonl(eth_len);
} else {
struct tcp4_l2_buf_t *b = (struct tcp4_l2_buf_t *)p;
ip_len = plen + sizeof(struct iphdr) + sizeof(struct tcphdr);
b->iph.tot_len = htons(ip_len);
b->iph.saddr = conn->a.a4.a.s_addr;
b->iph.daddr = c->ip4.addr_seen;
if (check)
b->iph.check = *check;
else
tcp_update_check_ip4(b);
SET_TCP_HEADER_COMMON_V4_V6(b, conn, seq);
tcp_update_check_tcp4(b);
eth_len = ip_len + sizeof(struct ethhdr);
if (c->mode == MODE_PASST)
b->vnet_len = htonl(eth_len);
}
#undef SET_TCP_HEADER_COMMON_V4_V6
return eth_len;
}
/**
* tcp_update_seqack_wnd() - Update ACK sequence and window to guest/tap
* @c: Execution context
* @conn: Connection pointer
* @force_seq: Force ACK sequence to latest segment, instead of checking socket
* @tinfo: tcp_info from kernel, can be NULL if not pre-fetched
*
* Return: 1 if sequence or window were updated, 0 otherwise
*/
static int tcp_update_seqack_wnd(const struct ctx *c, struct tcp_conn *conn,
int force_seq, struct tcp_info *tinfo)
{
uint32_t prev_wnd_to_tap = conn->wnd_to_tap << conn->ws_to_tap;
uint32_t prev_ack_to_tap = conn->seq_ack_to_tap;
socklen_t sl = sizeof(*tinfo);
struct tcp_info tinfo_new;
uint32_t new_wnd_to_tap = prev_wnd_to_tap;
int s = conn->sock;
#ifndef HAS_BYTES_ACKED
(void)force_seq;
conn->seq_ack_to_tap = conn->seq_from_tap;
if (SEQ_LT(conn->seq_ack_to_tap, prev_ack_to_tap))
conn->seq_ack_to_tap = prev_ack_to_tap;
#else
if ((unsigned)SNDBUF_GET(conn) < SNDBUF_SMALL || tcp_rtt_dst_low(conn)
|| CONN_IS_CLOSING(conn) || (conn->flags & LOCAL) || force_seq) {
conn->seq_ack_to_tap = conn->seq_from_tap;
} else if (conn->seq_ack_to_tap != conn->seq_from_tap) {
if (!tinfo) {
tinfo = &tinfo_new;
if (getsockopt(s, SOL_TCP, TCP_INFO, tinfo, &sl))
return 0;
}
conn->seq_ack_to_tap = tinfo->tcpi_bytes_acked +
conn->seq_init_from_tap;
if (SEQ_LT(conn->seq_ack_to_tap, prev_ack_to_tap))
conn->seq_ack_to_tap = prev_ack_to_tap;
}
#endif /* !HAS_BYTES_ACKED */
if (!KERNEL_REPORTS_SND_WND(c)) {
tcp_get_sndbuf(conn);
new_wnd_to_tap = MIN(SNDBUF_GET(conn), MAX_WINDOW);
conn->wnd_to_tap = MIN(new_wnd_to_tap >> conn->ws_to_tap,
USHRT_MAX);
goto out;
}
if (!tinfo) {
if (prev_wnd_to_tap > WINDOW_DEFAULT) {
goto out;
}
tinfo = &tinfo_new;
if (getsockopt(s, SOL_TCP, TCP_INFO, tinfo, &sl)) {
goto out;
}
}
#ifdef HAS_SND_WND
if ((conn->flags & LOCAL) || tcp_rtt_dst_low(conn)) {
new_wnd_to_tap = tinfo->tcpi_snd_wnd;
} else {
tcp_get_sndbuf(conn);
new_wnd_to_tap = MIN((int)tinfo->tcpi_snd_wnd,
SNDBUF_GET(conn));
}
#endif
new_wnd_to_tap = MIN(new_wnd_to_tap, MAX_WINDOW);
if (!(conn->events & ESTABLISHED))
new_wnd_to_tap = MAX(new_wnd_to_tap, WINDOW_DEFAULT);
conn->wnd_to_tap = MIN(new_wnd_to_tap >> conn->ws_to_tap, USHRT_MAX);
if (!conn->wnd_to_tap)
conn_flag(c, conn, ACK_TO_TAP_DUE);
out:
return new_wnd_to_tap != prev_wnd_to_tap ||
conn->seq_ack_to_tap != prev_ack_to_tap;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_send_flag() - Send segment with flags to tap (no payload)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @c: Execution context
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @conn: Connection pointer
* @flags: TCP flags: if not set, send segment only if ACK is due
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
* Return: negative error code on connection reset, 0 otherwise
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static int tcp_send_flag(struct ctx *c, struct tcp_conn *conn, int flags)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
uint32_t prev_ack_to_tap = conn->seq_ack_to_tap;
uint32_t prev_wnd_to_tap = conn->wnd_to_tap;
struct tcp4_l2_flags_buf_t *b4 = NULL;
struct tcp6_l2_flags_buf_t *b6 = NULL;
struct tcp_info tinfo = { 0 };
socklen_t sl = sizeof(tinfo);
size_t optlen = 0, eth_len;
int s = conn->sock;
struct iovec *iov;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
struct tcphdr *th;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
char *data;
void *p;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (SEQ_GE(conn->seq_ack_to_tap, conn->seq_from_tap) &&
!flags && conn->wnd_to_tap)
return 0;
if (getsockopt(s, SOL_TCP, TCP_INFO, &tinfo, &sl)) {
conn_event(c, conn, CLOSED);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
return -ECONNRESET;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
#ifdef HAS_SND_WND
if (!c->tcp.kernel_snd_wnd && tinfo.tcpi_snd_wnd)
c->tcp.kernel_snd_wnd = 1;
#endif
if (!(conn->flags & LOCAL))
tcp_rtt_dst_check(conn, &tinfo);
if (!tcp_update_seqack_wnd(c, conn, flags, &tinfo) && !flags)
return 0;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (CONN_V4(conn)) {
iov = tcp4_l2_flags_iov + tcp4_l2_flags_buf_used;
p = b4 = tcp4_l2_flags_buf + tcp4_l2_flags_buf_used++;
th = &b4->th;
/* gcc 11.2 would complain on data = (char *)(th + 1); */
data = b4->opts;
} else {
iov = tcp6_l2_flags_iov + tcp6_l2_flags_buf_used;
p = b6 = tcp6_l2_flags_buf + tcp6_l2_flags_buf_used++;
th = &b6->th;
data = b6->opts;
}
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (flags & SYN) {
int mss;
/* Options: MSS, NOP and window scale (8 bytes) */
optlen = OPT_MSS_LEN + 1 + OPT_WS_LEN;
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
*data++ = OPT_MSS;
*data++ = OPT_MSS_LEN;
if (c->mtu == -1) {
mss = tinfo.tcpi_snd_mss;
} else {
mss = c->mtu - sizeof(struct tcphdr);
if (CONN_V4(conn))
mss -= sizeof(struct iphdr);
else
mss -= sizeof(struct ipv6hdr);
if (c->low_wmem &&
!(conn->flags & LOCAL) && !tcp_rtt_dst_low(conn))
mss = MIN(mss, PAGE_SIZE);
else if (mss > PAGE_SIZE)
mss = ROUND_DOWN(mss, PAGE_SIZE);
}
*(uint16_t *)data = htons(MIN(USHRT_MAX, mss));
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
data += OPT_MSS_LEN - 2;
th->doff += OPT_MSS_LEN / 4;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
conn->ws_to_tap = MIN(MAX_WS, tinfo.tcpi_snd_wscale);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
*data++ = OPT_NOP;
*data++ = OPT_WS;
*data++ = OPT_WS_LEN;
*data++ = conn->ws_to_tap;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
th->ack = !!(flags & ACK);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
} else {
th->ack = !!(flags & (ACK | DUP_ACK)) ||
conn->seq_ack_to_tap != prev_ack_to_tap ||
!prev_wnd_to_tap;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
th->doff = (sizeof(*th) + optlen) / 4;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
th->rst = !!(flags & RST);
th->syn = !!(flags & SYN);
th->fin = !!(flags & FIN);
eth_len = tcp_l2_buf_fill_headers(c, conn, p, optlen,
NULL, conn->seq_to_tap);
iov->iov_len = eth_len + sizeof(uint32_t);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (CONN_V4(conn))
tcp4_l2_flags_buf_bytes += iov->iov_len;
else
tcp6_l2_flags_buf_bytes += iov->iov_len;
if (th->ack)
conn_flag(c, conn, ~ACK_TO_TAP_DUE);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (th->fin)
conn_flag(c, conn, ACK_FROM_TAP_DUE);
/* RFC 793, 3.1: "[...] and the first data octet is ISN+1." */
if (th->fin || th->syn)
conn->seq_to_tap++;
if (CONN_V4(conn)) {
if (flags & DUP_ACK) {
memcpy(b4 + 1, b4, sizeof(*b4));
(iov + 1)->iov_len = iov->iov_len;
tcp4_l2_flags_buf_used++;
tcp4_l2_flags_buf_bytes += iov->iov_len;
}
if (tcp4_l2_flags_buf_used > ARRAY_SIZE(tcp4_l2_flags_buf) - 2)
tcp_l2_flags_buf_flush(c);
} else {
if (flags & DUP_ACK) {
memcpy(b6 + 1, b6, sizeof(*b6));
(iov + 1)->iov_len = iov->iov_len;
tcp6_l2_flags_buf_used++;
tcp6_l2_flags_buf_bytes += iov->iov_len;
}
if (tcp6_l2_flags_buf_used > ARRAY_SIZE(tcp6_l2_flags_buf) - 2)
tcp_l2_flags_buf_flush(c);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return 0;
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/**
* tcp_rst_do() - Reset a tap connection: send RST segment to tap, close socket
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @c: Execution context
* @conn: Connection pointer
*/
static void tcp_rst_do(struct ctx *c, struct tcp_conn *conn)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
{
if (conn->events == CLOSED)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
return;
if (!tcp_send_flag(c, conn, RST))
conn_event(c, conn, CLOSED);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_get_tap_ws() - Get Window Scaling option for connection from tap/guest
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @conn: Connection pointer
* @opts: Pointer to start of TCP options
* @optlen: Bytes in options: caller MUST ensure available length
*/
static void tcp_get_tap_ws(struct tcp_conn *conn,
const char *opts, size_t optlen)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
int ws = tcp_opt_get(opts, optlen, OPT_WS, NULL, NULL);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (ws >= 0 && ws <= TCP_WS_MAX)
conn->ws_from_tap = ws;
else
conn->ws_from_tap = 0;
}
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
/**
* tcp_clamp_window() - Set new window for connection, clamp on socket
* @c: Execution context
* @conn: Connection pointer
* @window: Window value, host order, unscaled
*/
static void tcp_clamp_window(const struct ctx *c, struct tcp_conn *conn,
unsigned wnd)
{
uint32_t prev_scaled = conn->wnd_from_tap << conn->ws_from_tap;
int s = conn->sock;
wnd <<= conn->ws_from_tap;
wnd = MIN(MAX_WINDOW, wnd);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (conn->flags & WND_CLAMPED) {
if (prev_scaled == wnd)
return;
/* Discard +/- 1% updates to spare some syscalls. */
if ((wnd > prev_scaled && wnd * 99 / 100 < prev_scaled) ||
(wnd < prev_scaled && wnd * 101 / 100 > prev_scaled))
return;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
conn->wnd_from_tap = MIN(wnd >> conn->ws_from_tap, USHRT_MAX);
if (setsockopt(s, SOL_TCP, TCP_WINDOW_CLAMP, &wnd, sizeof(wnd)))
trace("TCP: failed to set TCP_WINDOW_CLAMP on socket %i", s);
conn_flag(c, conn, WND_CLAMPED);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
/**
* tcp_seq_init() - Calculate initial sequence number according to RFC 6528
* @c: Execution context
* @af: Address family, AF_INET or AF_INET6
* @addr: Remote address, pointer to sin_addr or sin6_addr
* @dstport: Destination port, connection-wise, network order
* @srcport: Source port, connection-wise, network order
udp: Connection tracking for ephemeral, local ports, and related fixes As we support UDP forwarding for packets that are sent to local ports, we actually need some kind of connection tracking for UDP. While at it, this commit introduces a number of vaguely related fixes for issues observed while trying this out. In detail: - implement an explicit, albeit minimalistic, connection tracking for UDP, to allow usage of ephemeral ports by the guest and by the host at the same time, by binding them dynamically as needed, and to allow mapping address changes for packets with a loopback address as destination - set the guest MAC address whenever we receive a packet from tap instead of waiting for an ARP request, and set it to broadcast on start, otherwise DHCPv6 might not work if all DHCPv6 requests time out before the guest starts talking IPv4 - split context IPv6 address into address we assign, global or site address seen on tap, and link-local address seen on tap, and make sure we use the addresses we've seen as destination (link-local choice depends on source address). Similarly, for IPv4, split into address we assign and address we observe, and use the address we observe as destination - introduce a clock_gettime() syscall right after epoll_wait() wakes up, so that we can remove all the other ones and pass the current timestamp to tap and socket handlers -- this is additionally needed by UDP to time out bindings to ephemeral ports and mappings between loopback address and a local address - rename sock_l4_add() to sock_l4(), no semantic changes intended - include <arpa/inet.h> in passt.c before kernel headers so that we can use <netinet/in.h> macros to check IPv6 address types, and remove a duplicate <linux/ip.h> inclusion Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-04-29 14:59:20 +00:00
* @now: Current timestamp
*
* Return: initial TCP sequence
*/
static uint32_t tcp_seq_init(const struct ctx *c, int af, const void *addr,
udp: Connection tracking for ephemeral, local ports, and related fixes As we support UDP forwarding for packets that are sent to local ports, we actually need some kind of connection tracking for UDP. While at it, this commit introduces a number of vaguely related fixes for issues observed while trying this out. In detail: - implement an explicit, albeit minimalistic, connection tracking for UDP, to allow usage of ephemeral ports by the guest and by the host at the same time, by binding them dynamically as needed, and to allow mapping address changes for packets with a loopback address as destination - set the guest MAC address whenever we receive a packet from tap instead of waiting for an ARP request, and set it to broadcast on start, otherwise DHCPv6 might not work if all DHCPv6 requests time out before the guest starts talking IPv4 - split context IPv6 address into address we assign, global or site address seen on tap, and link-local address seen on tap, and make sure we use the addresses we've seen as destination (link-local choice depends on source address). Similarly, for IPv4, split into address we assign and address we observe, and use the address we observe as destination - introduce a clock_gettime() syscall right after epoll_wait() wakes up, so that we can remove all the other ones and pass the current timestamp to tap and socket handlers -- this is additionally needed by UDP to time out bindings to ephemeral ports and mappings between loopback address and a local address - rename sock_l4_add() to sock_l4(), no semantic changes intended - include <arpa/inet.h> in passt.c before kernel headers so that we can use <netinet/in.h> macros to check IPv6 address types, and remove a duplicate <linux/ip.h> inclusion Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-04-29 14:59:20 +00:00
in_port_t dstport, in_port_t srcport,
const struct timespec *now)
{
uint32_t ns, seq = 0;
if (af == AF_INET) {
struct {
struct in_addr src;
in_port_t srcport;
struct in_addr dst;
in_port_t dstport;
} __attribute__((__packed__)) in = {
.src = *(struct in_addr *)addr,
.srcport = srcport,
.dst = { c->ip4.addr },
.dstport = dstport,
};
seq = siphash_12b((uint8_t *)&in, c->tcp.hash_secret);
} else if (af == AF_INET6) {
struct {
struct in6_addr src;
in_port_t srcport;
struct in6_addr dst;
in_port_t dstport;
} __attribute__((__packed__)) in = {
.src = *(struct in6_addr *)addr,
.srcport = srcport,
.dst = c->ip6.addr,
.dstport = dstport,
};
seq = siphash_36b((uint8_t *)&in, c->tcp.hash_secret);
}
udp: Connection tracking for ephemeral, local ports, and related fixes As we support UDP forwarding for packets that are sent to local ports, we actually need some kind of connection tracking for UDP. While at it, this commit introduces a number of vaguely related fixes for issues observed while trying this out. In detail: - implement an explicit, albeit minimalistic, connection tracking for UDP, to allow usage of ephemeral ports by the guest and by the host at the same time, by binding them dynamically as needed, and to allow mapping address changes for packets with a loopback address as destination - set the guest MAC address whenever we receive a packet from tap instead of waiting for an ARP request, and set it to broadcast on start, otherwise DHCPv6 might not work if all DHCPv6 requests time out before the guest starts talking IPv4 - split context IPv6 address into address we assign, global or site address seen on tap, and link-local address seen on tap, and make sure we use the addresses we've seen as destination (link-local choice depends on source address). Similarly, for IPv4, split into address we assign and address we observe, and use the address we observe as destination - introduce a clock_gettime() syscall right after epoll_wait() wakes up, so that we can remove all the other ones and pass the current timestamp to tap and socket handlers -- this is additionally needed by UDP to time out bindings to ephemeral ports and mappings between loopback address and a local address - rename sock_l4_add() to sock_l4(), no semantic changes intended - include <arpa/inet.h> in passt.c before kernel headers so that we can use <netinet/in.h> macros to check IPv6 address types, and remove a duplicate <linux/ip.h> inclusion Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-04-29 14:59:20 +00:00
ns = now->tv_sec * 1E9;
ns += now->tv_nsec >> 5; /* 32ns ticks, overflows 32 bits every 137s */
return seq + ns;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_conn_new_sock() - Get socket for new connection from pool or make new one
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @c: Execution context
* @af: Address family
*
* Return: socket number if available, negative code if socket creation failed
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static int tcp_conn_new_sock(const struct ctx *c, sa_family_t af)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
int *p = af == AF_INET6 ? init_sock_pool6 : init_sock_pool4, i, s = -1;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
for (i = 0; i < TCP_SOCK_POOL_SIZE; i++, p++) {
SWAP(s, *p);
if (s >= 0)
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
break;
}
if (s < 0)
s = socket(af, SOCK_STREAM | SOCK_NONBLOCK, IPPROTO_TCP);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (s > SOCKET_MAX) {
close(s);
return -EIO;
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (s < 0)
return -errno;
tcp_sock_set_bufsize(c, s);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
return s;
}
/**
* tcp_conn_tap_mss() - Get MSS value advertised by tap/guest
* @c: Execution context
* @conn: Connection pointer
* @opts: Pointer to start of TCP options
* @optlen: Bytes in options: caller MUST ensure available length
*
* Return: clamped MSS value
*/
static uint16_t tcp_conn_tap_mss(const struct ctx *c,
const struct tcp_conn *conn,
const char *opts, size_t optlen)
{
unsigned int mss;
int ret;
if ((ret = tcp_opt_get(opts, optlen, OPT_MSS, NULL, NULL)) < 0)
mss = MSS_DEFAULT;
else
mss = ret;
/* Don't upset qemu */
if (c->mode == MODE_PASST) {
if (CONN_V4(conn))
mss = MIN(MSS4, mss);
else
mss = MIN(MSS6, mss);
}
return MIN(mss, USHRT_MAX);
}
/**
* tcp_conn_from_tap() - Handle connection request (SYN segment) from tap
* @c: Execution context
* @af: Address family, AF_INET or AF_INET6
* @addr: Remote address, pointer to sin_addr or sin6_addr
* @th: TCP header from tap: caller MUST ensure it's there
* @opts: Pointer to start of options
* @optlen: Bytes in options: caller MUST ensure available length
* @now: Current timestamp
*/
static void tcp_conn_from_tap(struct ctx *c, int af, const void *addr,
const struct tcphdr *th, const char *opts,
size_t optlen, const struct timespec *now)
{
struct sockaddr_in addr4 = {
.sin_family = AF_INET,
.sin_port = th->dest,
.sin_addr = *(struct in_addr *)addr,
};
struct sockaddr_in6 addr6 = {
.sin6_family = AF_INET6,
.sin6_port = th->dest,
.sin6_addr = *(struct in6_addr *)addr,
};
const struct sockaddr *sa;
struct tcp_conn *conn;
socklen_t sl;
int s, mss;
if (c->tcp.conn_count >= TCP_MAX_CONNS)
return;
if ((s = tcp_conn_new_sock(c, af)) < 0)
return;
if (!c->no_map_gw) {
if (af == AF_INET && addr4.sin_addr.s_addr == c->ip4.gw)
addr4.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
if (af == AF_INET6 && IN6_ARE_ADDR_EQUAL(addr, &c->ip6.gw))
addr6.sin6_addr = in6addr_loopback;
}
if (af == AF_INET6 && IN6_IS_ADDR_LINKLOCAL(&addr6.sin6_addr)) {
struct sockaddr_in6 addr6_ll = {
.sin6_family = AF_INET6,
.sin6_addr = c->ip6.addr_ll,
.sin6_scope_id = c->ifi6,
};
if (bind(s, (struct sockaddr *)&addr6_ll, sizeof(addr6_ll))) {
close(s);
return;
}
}
conn = CONN(c->tcp.conn_count++);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->sock = s;
conn->timer = -1;
conn_event(c, conn, TAP_SYN_RCVD);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
conn->wnd_to_tap = WINDOW_DEFAULT;
mss = tcp_conn_tap_mss(c, conn, opts, optlen);
if (setsockopt(s, SOL_TCP, TCP_MAXSEG, &mss, sizeof(mss)))
trace("TCP: failed to set TCP_MAXSEG on socket %i", s);
MSS_SET(conn, mss);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
tcp_get_tap_ws(conn, opts, optlen);
/* RFC 7323, 2.2: first value is not scaled. Also, don't clamp yet, to
* avoid getting a zero scale just because we set a small window now.
*/
if (!(conn->wnd_from_tap = (htons(th->window) >> conn->ws_from_tap)))
conn->wnd_from_tap = 1;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (af == AF_INET) {
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
sa = (struct sockaddr *)&addr4;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
sl = sizeof(addr4);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
memset(&conn->a.a4.zero, 0, sizeof(conn->a.a4.zero));
memset(&conn->a.a4.one, 0xff, sizeof(conn->a.a4.one));
memcpy(&conn->a.a4.a, addr, sizeof(conn->a.a4.a));
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
} else {
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
sa = (struct sockaddr *)&addr6;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
sl = sizeof(addr6);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
memcpy(&conn->a.a6, addr, sizeof(conn->a.a6));
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->sock_port = ntohs(th->dest);
conn->tap_port = ntohs(th->source);
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->seq_init_from_tap = ntohl(th->seq);
conn->seq_from_tap = conn->seq_init_from_tap + 1;
conn->seq_ack_to_tap = conn->seq_from_tap;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->seq_to_tap = tcp_seq_init(c, af, addr, th->dest, th->source, now);
conn->seq_ack_from_tap = conn->seq_to_tap + 1;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_hash_insert(c, conn, af, addr);
if (!bind(s, sa, sl)) {
tcp_rst(c, conn); /* Nobody is listening then */
return;
}
if (errno != EADDRNOTAVAIL && errno != EACCES)
conn_flag(c, conn, LOCAL);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (connect(s, sa, sl)) {
if (errno != EINPROGRESS) {
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_rst(c, conn);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return;
}
tcp_get_sndbuf(conn);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
} else {
tcp_get_sndbuf(conn);
if (tcp_send_flag(c, conn, SYN | ACK))
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
return;
conn_event(c, conn, TAP_SYN_ACK_SENT);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
tcp_epoll_ctl(c, conn);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
/**
* tcp_sock_consume() - Consume (discard) data from buffer, update ACK sequence
* @conn: Connection pointer
* @ack_seq: ACK sequence, host order
*
* Return: 0 on success, negative error code from recv() on failure
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static int tcp_sock_consume(struct tcp_conn *conn, uint32_t ack_seq)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
/* Simply ignore out-of-order ACKs: we already consumed the data we
* needed from the buffer, and we won't rewind back to a lower ACK
* sequence.
*/
if (SEQ_LE(ack_seq, conn->seq_ack_from_tap))
return 0;
if (recv(conn->sock, NULL, ack_seq - conn->seq_ack_from_tap,
MSG_DONTWAIT | MSG_TRUNC) < 0)
return -errno;
conn->seq_ack_from_tap = ack_seq;
return 0;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/**
* tcp_data_to_tap() - Finalise (queue) highest-numbered scatter-gather buffer
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @c: Execution context
* @conn: Connection pointer
* @plen: Payload length at L4
* @no_csum: Don't compute IPv4 checksum, use the one from previous buffer
* @seq: Sequence number to be sent
* @now: Current timestamp
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
*/
static void tcp_data_to_tap(struct ctx *c, struct tcp_conn *conn,
ssize_t plen, int no_csum, uint32_t seq)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
{
struct iovec *iov;
size_t len;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (CONN_V4(conn)) {
struct tcp4_l2_buf_t *b = &tcp4_l2_buf[tcp4_l2_buf_used];
uint16_t *check = no_csum ? &(b - 1)->iph.check : NULL;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
len = tcp_l2_buf_fill_headers(c, conn, b, plen, check, seq);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
iov = tcp4_l2_iov + tcp4_l2_buf_used++;
tcp4_l2_buf_bytes += iov->iov_len = len + sizeof(b->vnet_len);
if (tcp4_l2_buf_used > ARRAY_SIZE(tcp4_l2_buf) - 1)
tcp_l2_data_buf_flush(c);
} else if (CONN_V6(conn)) {
struct tcp6_l2_buf_t *b = &tcp6_l2_buf[tcp6_l2_buf_used];
len = tcp_l2_buf_fill_headers(c, conn, b, plen, NULL, seq);
iov = tcp6_l2_iov + tcp6_l2_buf_used++;
tcp6_l2_buf_bytes += iov->iov_len = len + sizeof(b->vnet_len);
if (tcp6_l2_buf_used > ARRAY_SIZE(tcp6_l2_buf) - 1)
tcp_l2_data_buf_flush(c);
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
/**
* tcp_data_from_sock() - Handle new data from socket, queue to tap, in window
* @c: Execution context
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @conn: Connection pointer
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*
* Return: negative on connection reset, 0 otherwise
*
* #syscalls recvmsg
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static int tcp_data_from_sock(struct ctx *c, struct tcp_conn *conn)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
uint32_t wnd_scaled = conn->wnd_from_tap << conn->ws_from_tap;
int fill_bufs, send_bufs = 0, last_len, iov_rem = 0;
int sendlen, len, plen, v4 = CONN_V4(conn);
int s = conn->sock, i, ret = 0;
struct msghdr mh_sock = { 0 };
uint16_t mss = MSS_GET(conn);
uint32_t already_sent;
struct iovec *iov;
already_sent = conn->seq_to_tap - conn->seq_ack_from_tap;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (SEQ_LT(already_sent, 0)) {
/* RFC 761, section 2.1. */
trace("TCP: ACK sequence gap: ACK for %u, sent: %u",
conn->seq_ack_from_tap, conn->seq_to_tap);
conn->seq_to_tap = conn->seq_ack_from_tap;
already_sent = 0;
}
if (!wnd_scaled || already_sent >= wnd_scaled) {
conn_flag(c, conn, STALLED);
conn_flag(c, conn, ACK_FROM_TAP_DUE);
return 0;
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/* Set up buffer descriptors we'll fill completely and partially. */
fill_bufs = DIV_ROUND_UP(wnd_scaled - already_sent, mss);
if (fill_bufs > TCP_FRAMES) {
fill_bufs = TCP_FRAMES;
iov_rem = 0;
} else {
iov_rem = (wnd_scaled - already_sent) % mss;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
mh_sock.msg_iov = iov_sock;
mh_sock.msg_iovlen = fill_bufs + 1;
iov_sock[0].iov_base = tcp_buf_discard;
iov_sock[0].iov_len = already_sent;
if (( v4 && tcp4_l2_buf_used + fill_bufs > ARRAY_SIZE(tcp4_l2_buf)) ||
(!v4 && tcp6_l2_buf_used + fill_bufs > ARRAY_SIZE(tcp6_l2_buf))) {
tcp_l2_data_buf_flush(c);
/* Silence Coverity CWE-125 false positive */
tcp4_l2_buf_used = tcp6_l2_buf_used = 0;
}
for (i = 0, iov = iov_sock + 1; i < fill_bufs; i++, iov++) {
if (v4)
iov->iov_base = &tcp4_l2_buf[tcp4_l2_buf_used + i].data;
else
iov->iov_base = &tcp6_l2_buf[tcp6_l2_buf_used + i].data;
iov->iov_len = mss;
}
if (iov_rem)
iov_sock[fill_bufs].iov_len = iov_rem;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/* Receive into buffers, don't dequeue until acknowledged by guest. */
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
recvmsg:
len = recvmsg(s, &mh_sock, MSG_PEEK);
if (len < 0) {
if (errno == EINTR)
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
goto recvmsg;
goto err;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (!len)
goto zero_len;
sendlen = len - already_sent;
if (sendlen <= 0) {
conn_flag(c, conn, STALLED);
return 0;
}
conn_flag(c, conn, ~STALLED);
send_bufs = DIV_ROUND_UP(sendlen, mss);
last_len = sendlen - (send_bufs - 1) * mss;
/* Likely, some new data was acked too. */
tcp_update_seqack_wnd(c, conn, 0, NULL);
/* Finally, queue to tap */
plen = mss;
for (i = 0; i < send_bufs; i++) {
int no_csum = i && i != send_bufs - 1 && tcp4_l2_buf_used;
if (i == send_bufs - 1)
plen = last_len;
tcp_data_to_tap(c, conn, plen, no_csum, conn->seq_to_tap);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
conn->seq_to_tap += plen;
}
conn_flag(c, conn, ACK_FROM_TAP_DUE);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
return 0;
err:
if (errno != EAGAIN && errno != EWOULDBLOCK) {
ret = -errno;
tcp_rst(c, conn);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
return ret;
zero_len:
if ((conn->events & (SOCK_FIN_RCVD | TAP_FIN_SENT)) == SOCK_FIN_RCVD) {
if ((ret = tcp_send_flag(c, conn, FIN | ACK))) {
tcp_rst(c, conn);
return ret;
}
conn_event(c, conn, TAP_FIN_SENT);
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return 0;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
/**
* tcp_data_from_tap() - tap/guest data for established connection
* @c: Execution context
* @conn: Connection pointer
* @p: Pool of TCP packets, with TCP headers
*
* #syscalls sendmsg
*/
static void tcp_data_from_tap(struct ctx *c, struct tcp_conn *conn,
const struct pool *p)
{
int i, iov_i, ack = 0, fin = 0, retr = 0, keep = -1, partial_send = 0;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
uint16_t max_ack_seq_wnd = conn->wnd_from_tap;
uint32_t max_ack_seq = conn->seq_ack_from_tap;
uint32_t seq_from_tap = conn->seq_from_tap;
struct msghdr mh = { .msg_iov = tcp_iov };
size_t len;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
ssize_t n;
for (i = 0, iov_i = 0; i < (int)p->count; i++) {
uint32_t seq, seq_offset, ack_seq;
struct tcphdr *th;
char *data;
size_t off;
if (!packet_get(p, i, 0, 0, &len)) {
tcp_rst(c, conn);
return;
}
th = packet_get(p, i, 0, sizeof(*th), NULL);
if (!th) {
tcp_rst(c, conn);
return;
}
off = th->doff * 4UL;
if (off < sizeof(*th) || off > len) {
tcp_rst(c, conn);
return;
}
if (th->rst) {
conn_event(c, conn, CLOSED);
return;
}
len -= off;
data = packet_get(p, i, off, len, NULL);
if (!data)
continue;
seq = ntohl(th->seq);
ack_seq = ntohl(th->ack_seq);
if (th->ack) {
ack = 1;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (SEQ_GE(ack_seq, conn->seq_ack_from_tap) &&
SEQ_GE(ack_seq, max_ack_seq)) {
/* Fast re-transmit */
retr = !len && !th->fin &&
ack_seq == max_ack_seq &&
ntohs(th->window) == max_ack_seq_wnd;
max_ack_seq_wnd = ntohs(th->window);
max_ack_seq = ack_seq;
}
}
if (th->fin)
fin = 1;
if (!len)
continue;
seq_offset = seq_from_tap - seq;
/* Use data from this buffer only in these two cases:
*
* , seq_from_tap , seq_from_tap
* |--------| <-- len |--------| <-- len
* '----' <-- offset ' <-- offset
* ^ seq ^ seq
* (offset >= 0, seq + len > seq_from_tap)
*
* discard in these two cases:
* , seq_from_tap , seq_from_tap
* |--------| <-- len |--------| <-- len
* '--------' <-- offset '-----| <- offset
* ^ seq ^ seq
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
* (offset >= 0, seq + len <= seq_from_tap)
*
* keep, look for another buffer, then go back, in this case:
* , seq_from_tap
* |--------| <-- len
* '===' <-- offset
* ^ seq
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
* (offset < 0)
*/
if (SEQ_GE(seq_offset, 0) && SEQ_LE(seq + len, seq_from_tap))
continue;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (SEQ_LT(seq_offset, 0)) {
if (keep == -1)
keep = i;
continue;
}
tcp_iov[iov_i].iov_base = data + seq_offset;
tcp_iov[iov_i].iov_len = len - seq_offset;
seq_from_tap += tcp_iov[iov_i].iov_len;
iov_i++;
if (keep == i)
keep = -1;
if (keep != -1)
i = keep - 1;
}
tcp_clamp_window(c, conn, max_ack_seq_wnd);
if (ack) {
if (max_ack_seq == conn->seq_to_tap) {
conn_flag(c, conn, ~ACK_FROM_TAP_DUE);
conn->retrans = 0;
}
tcp_sock_consume(conn, max_ack_seq);
}
if (retr) {
trace("TCP: fast re-transmit, ACK: %u, previous sequence: %u",
max_ack_seq, conn->seq_to_tap);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
conn->seq_ack_from_tap = max_ack_seq;
conn->seq_to_tap = max_ack_seq;
tcp_data_from_sock(c, conn);
}
if (!iov_i)
goto out;
mh.msg_iovlen = iov_i;
eintr:
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
n = sendmsg(conn->sock, &mh, MSG_DONTWAIT | MSG_NOSIGNAL);
if (n < 0) {
if (errno == EPIPE) {
/* Here's the wrap, said the tap.
* In my pocket, said the socket.
* Then swiftly looked away and left.
*/
conn->seq_from_tap = seq_from_tap;
tcp_send_flag(c, conn, ACK);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
if (errno == EINTR)
goto eintr;
if (errno == EAGAIN || errno == EWOULDBLOCK) {
tcp_send_flag(c, conn, ACK_IF_NEEDED);
return;
}
tcp_rst(c, conn);
return;
}
if (n < (int)(seq_from_tap - conn->seq_from_tap)) {
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
partial_send = 1;
conn->seq_from_tap += n;
tcp_send_flag(c, conn, ACK_IF_NEEDED);
} else {
conn->seq_from_tap += n;
}
out:
if (keep != -1) {
/* We use an 8-bit approximation here: the associated risk is
* that we skip a duplicate ACK on 8-bit sequence number
* collision. Fast retransmit is a SHOULD in RFC 5681, 3.2.
*/
if (conn->seq_dup_ack_approx != (conn->seq_from_tap & 0xff)) {
conn->seq_dup_ack_approx = conn->seq_from_tap & 0xff;
tcp_send_flag(c, conn, DUP_ACK);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
return;
}
if (ack && conn->events & TAP_FIN_SENT &&
conn->seq_ack_from_tap == conn->seq_to_tap)
conn_event(c, conn, TAP_FIN_ACKED);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (fin && !partial_send) {
conn->seq_from_tap++;
conn_event(c, conn, TAP_FIN_RCVD);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
} else {
tcp_send_flag(c, conn, ACK_IF_NEEDED);
}
}
/**
* tcp_conn_from_sock_finish() - Complete connection setup after connect()
* @c: Execution context
* @conn: Connection pointer
* @th: TCP header of SYN, ACK segment: caller MUST ensure it's there
* @opts: Pointer to start of options
* @optlen: Bytes in options: caller MUST ensure available length
*/
static void tcp_conn_from_sock_finish(struct ctx *c, struct tcp_conn *conn,
const struct tcphdr *th,
const char *opts, size_t optlen)
{
tcp_clamp_window(c, conn, ntohs(th->window));
tcp_get_tap_ws(conn, opts, optlen);
/* First value is not scaled */
if (!(conn->wnd_from_tap >>= conn->ws_from_tap))
conn->wnd_from_tap = 1;
MSS_SET(conn, tcp_conn_tap_mss(c, conn, opts, optlen));
conn->seq_init_from_tap = ntohl(th->seq) + 1;
conn->seq_from_tap = conn->seq_init_from_tap;
conn->seq_ack_to_tap = conn->seq_from_tap;
conn_event(c, conn, ESTABLISHED);
/* The client might have sent data already, which we didn't
* dequeue waiting for SYN,ACK from tap -- check now.
*/
tcp_data_from_sock(c, conn);
tcp_send_flag(c, conn, ACK_IF_NEEDED);
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_tap_handler() - Handle packets from tap and state transitions
* @c: Execution context
* @af: Address family, AF_INET or AF_INET6
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @addr: Destination address
* @p: Pool of TCP packets, with TCP headers
udp: Connection tracking for ephemeral, local ports, and related fixes As we support UDP forwarding for packets that are sent to local ports, we actually need some kind of connection tracking for UDP. While at it, this commit introduces a number of vaguely related fixes for issues observed while trying this out. In detail: - implement an explicit, albeit minimalistic, connection tracking for UDP, to allow usage of ephemeral ports by the guest and by the host at the same time, by binding them dynamically as needed, and to allow mapping address changes for packets with a loopback address as destination - set the guest MAC address whenever we receive a packet from tap instead of waiting for an ARP request, and set it to broadcast on start, otherwise DHCPv6 might not work if all DHCPv6 requests time out before the guest starts talking IPv4 - split context IPv6 address into address we assign, global or site address seen on tap, and link-local address seen on tap, and make sure we use the addresses we've seen as destination (link-local choice depends on source address). Similarly, for IPv4, split into address we assign and address we observe, and use the address we observe as destination - introduce a clock_gettime() syscall right after epoll_wait() wakes up, so that we can remove all the other ones and pass the current timestamp to tap and socket handlers -- this is additionally needed by UDP to time out bindings to ephemeral ports and mappings between loopback address and a local address - rename sock_l4_add() to sock_l4(), no semantic changes intended - include <arpa/inet.h> in passt.c before kernel headers so that we can use <netinet/in.h> macros to check IPv6 address types, and remove a duplicate <linux/ip.h> inclusion Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-04-29 14:59:20 +00:00
* @now: Current timestamp
*
* Return: count of consumed packets
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
int tcp_tap_handler(struct ctx *c, int af, const void *addr,
const struct pool *p, const struct timespec *now)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
struct tcp_conn *conn;
size_t optlen, len;
struct tcphdr *th;
int ack_due = 0;
char *opts;
if (!packet_get(p, 0, 0, 0, &len))
return 1;
th = packet_get(p, 0, 0, sizeof(*th), NULL);
if (!th)
return 1;
optlen = th->doff * 4UL - sizeof(*th);
/* Static checkers might fail to see this: */
optlen = MIN(optlen, ((1UL << 4) /* from doff width */ - 6) * 4UL);
opts = packet_get(p, 0, sizeof(*th), optlen, NULL);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn = tcp_hash_lookup(c, af, addr, htons(th->source), htons(th->dest));
/* New connection from tap */
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (!conn) {
if (opts && th->syn && !th->ack)
tcp_conn_from_tap(c, af, addr, th, opts, optlen, now);
return 1;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
trace("TCP: packet length %lu from tap for index %lu", len, conn - tc);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (th->rst) {
conn_event(c, conn, CLOSED);
return p->count;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
if (th->ack) {
conn_flag(c, conn, ~ACK_FROM_TAP_DUE);
conn->retrans = 0;
}
conn_flag(c, conn, ~STALLED);
passt: Assorted fixes from "fresh eyes" review A bunch of fixes not worth single commits at this stage, notably: - make buffer, length parameter ordering consistent in ARP, DHCP, NDP handlers - strict checking of buffer, message and option length in DHCP handler (a malicious client could have easily crashed it) - set up forwarding for IPv4 and IPv6, and masquerading with nft for IPv4, from demo script - get rid of separate slow and fast timers, we don't save any overhead that way - stricter checking of buffer lengths as passed to tap handlers - proper dequeuing from qemu socket back-end: I accidentally trashed messages that were bundled up together in a single tap read operation -- the length header tells us what's the size of the next frame, but there's no apparent limit to the number of messages we get with one single receive - rework some bits of the TCP state machine, now passive and active connection closes appear to be robust -- introduce a new FIN_WAIT_1_SOCK_FIN state indicating a FIN_WAIT_1 with a FIN flag from socket - streamline TCP option parsing routine - track TCP state changes to stderr (this is temporary, proper debugging and syslogging support pending) - observe that multiplying a number by four might very well change its value, and this happens to be the case for the data offset from the TCP header as we check if it's the same as the total length to find out if it's a duplicated ACK segment - recent estimates suggest that the duration of a millisecond is closer to a million nanoseconds than a thousand of them, this trend is now reflected into the timespec_diff_ms() convenience routine Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-21 10:33:38 +00:00
/* Establishing connection from socket */
if (conn->events & SOCK_ACCEPTED) {
if (th->syn && th->ack && !th->fin)
tcp_conn_from_sock_finish(c, conn, th, opts, optlen);
else
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_rst(c, conn);
return 1;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/* Establishing connection from tap */
if (conn->events & TAP_SYN_RCVD) {
if (!(conn->events & TAP_SYN_ACK_SENT)) {
tcp_rst(c, conn);
return p->count;
tcp: Proper error handling for sendmmsg() to UNIX domain socket As data from socket is forwarded to the guest, sendmmsg() might send fewer bytes than requested in three different ways: - failing altogether with a negative error code -- ignore that, we'll get an error on the UNIX domain socket later if there's really an issue with it and reset the connection to the guest - sending less than 'vlen' messages -- instead of assuming success in that case and waiting for the guest to send a duplicate ACK indicating missing data, update the sequence number according to what was actually sent and spare some retransmissions - somewhat unexpectedly to me, sending 'vlen' or less than 'vlen' messages, returning up to 'vlen', with the last message being partially sent, and no further indication of errors other than the returned msg_len for the last partially sent message being less than iov_len. In this case, we would assume success and proceed as nothing happened. However, qemu would fail to parse any further message, having received a partial descriptor, and eventually close the connection, logging: serious error: oversized packet received,connection terminated. as the length descriptor for the next message would be sourced from the middle of the next successfully sent message, not from its header. Handle this by checking the msg_len returned for the last (even partially) sent message, and force re-sending the missing bytes, if any, with a blocking sendmsg() -- qemu must not receive anything else than that anyway. While at it, allow to send up to 64KiB for each message, the previous 32KiB limit isn't actually required, and just switch to a new message at each iteration on sending buffers, they are already MSS-sized anyway, so the check in the loop isn't really needed. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-08-26 12:37:48 +00:00
}
conn_event(c, conn, ESTABLISHED);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (th->fin) {
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
conn->seq_from_tap++;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
shutdown(conn->sock, SHUT_WR);
tcp_send_flag(c, conn, ACK);
conn_event(c, conn, SOCK_FIN_SENT);
return p->count;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
if (!th->ack) {
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_rst(c, conn);
return p->count;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
tcp_clamp_window(c, conn, ntohs(th->window));
tcp_data_from_sock(c, conn);
if (p->count == 1)
return 1;
}
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
/* Established connections not accepting data from tap */
if (conn->events & TAP_FIN_RCVD) {
if (conn->events & SOCK_FIN_RCVD &&
conn->seq_ack_from_tap == conn->seq_to_tap)
conn_event(c, conn, CLOSED);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
return 1;
}
/* Established connections accepting data from tap */
tcp_data_from_tap(c, conn, p);
if (conn->seq_ack_to_tap != conn->seq_from_tap)
ack_due = 1;
if ((conn->events & TAP_FIN_RCVD) && !(conn->events & SOCK_FIN_SENT)) {
shutdown(conn->sock, SHUT_WR);
conn_event(c, conn, SOCK_FIN_SENT);
tcp_send_flag(c, conn, ACK);
ack_due = 0;
}
if (ack_due)
conn_flag(c, conn, ACK_TO_TAP_DUE);
return p->count;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
/**
* tcp_connect_finish() - Handle completion of connect() from EPOLLOUT event
* @c: Execution context
* @conn: Connection pointer
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
static void tcp_connect_finish(struct ctx *c, struct tcp_conn *conn)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
socklen_t sl;
int so;
sl = sizeof(so);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (getsockopt(conn->sock, SOL_SOCKET, SO_ERROR, &so, &sl) || so) {
tcp_rst(c, conn);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return;
}
if (tcp_send_flag(c, conn, SYN | ACK))
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return;
conn_event(c, conn, TAP_SYN_ACK_SENT);
conn_flag(c, conn, ACK_FROM_TAP_DUE);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
}
/**
* tcp_conn_from_sock() - Handle new connection request from listening socket
* @c: Execution context
* @ref: epoll reference of listening socket
* @now: Current timestamp
*/
static void tcp_conn_from_sock(struct ctx *c, union epoll_ref ref,
const struct timespec *now)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
{
struct sockaddr_storage sa;
struct tcp_conn *conn;
socklen_t sl;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
int s;
if (c->tcp.conn_count >= TCP_MAX_CONNS)
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
return;
sl = sizeof(sa);
s = accept4(ref.r.s, (struct sockaddr *)&sa, &sl, SOCK_NONBLOCK);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (s < 0)
return;
conn = CONN(c->tcp.conn_count++);
conn->sock = s;
conn->timer = -1;
conn->ws_to_tap = conn->ws_from_tap = 0;
conn_event(c, conn, SOCK_ACCEPTED);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (ref.r.p.tcp.tcp.v6) {
struct sockaddr_in6 sa6;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
memcpy(&sa6, &sa, sizeof(sa6));
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (IN6_IS_ADDR_LOOPBACK(&sa6.sin6_addr) ||
IN6_ARE_ADDR_EQUAL(&sa6.sin6_addr, &c->ip6.addr_seen) ||
IN6_ARE_ADDR_EQUAL(&sa6.sin6_addr, &c->ip6.addr)) {
struct in6_addr *src;
if (IN6_IS_ADDR_LINKLOCAL(&c->ip6.gw))
src = &c->ip6.gw;
else
src = &c->ip6.addr_ll;
memcpy(&sa6.sin6_addr, src, sizeof(*src));
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
memcpy(&conn->a.a6, &sa6.sin6_addr, sizeof(conn->a.a6));
conn->sock_port = ntohs(sa6.sin6_port);
conn->tap_port = ref.r.p.tcp.tcp.index;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->seq_to_tap = tcp_seq_init(c, AF_INET6, &sa6.sin6_addr,
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->sock_port,
conn->tap_port,
now);
tcp_hash_insert(c, conn, AF_INET6, &sa6.sin6_addr);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
} else {
struct sockaddr_in sa4;
in_addr_t s_addr;
memcpy(&sa4, &sa, sizeof(sa4));
s_addr = ntohl(sa4.sin_addr.s_addr);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
memset(&conn->a.a4.zero, 0, sizeof(conn->a.a4.zero));
memset(&conn->a.a4.one, 0xff, sizeof(conn->a.a4.one));
if (s_addr >> IN_CLASSA_NSHIFT == IN_LOOPBACKNET ||
s_addr == INADDR_ANY || htonl(s_addr) == c->ip4.addr_seen)
s_addr = ntohl(c->ip4.gw);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
s_addr = htonl(s_addr);
memcpy(&conn->a.a4.a, &s_addr, sizeof(conn->a.a4.a));
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->sock_port = ntohs(sa4.sin_port);
conn->tap_port = ref.r.p.tcp.tcp.index;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->seq_to_tap = tcp_seq_init(c, AF_INET, &s_addr,
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
conn->sock_port,
conn->tap_port,
now);
tcp_hash_insert(c, conn, AF_INET, &s_addr);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
}
conn->seq_ack_from_tap = conn->seq_to_tap + 1;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
conn->wnd_from_tap = WINDOW_DEFAULT;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_send_flag(c, conn, SYN);
conn_flag(c, conn, ACK_FROM_TAP_DUE);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
tcp_get_sndbuf(conn);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_timer_handler() - timerfd events: close, send ACK, retransmit, or reset
* @c: Execution context
* @ref: epoll reference of timer (not connection)
*
* #syscalls timerfd_gettime
*/
static void tcp_timer_handler(struct ctx *c, union epoll_ref ref)
{
struct tcp_conn *conn = CONN_OR_NULL(ref.r.p.tcp.tcp.index);
struct itimerspec check_armed = { { 0 }, { 0 } };
if (!conn)
return;
/* We don't reset timers on ~ACK_FROM_TAP_DUE, ~ACK_TO_TAP_DUE. If the
* timer is currently armed, this event came from a previous setting,
* and we just set the timer to a new point in the future: discard it.
*/
timerfd_gettime(conn->timer, &check_armed);
if (check_armed.it_value.tv_sec || check_armed.it_value.tv_nsec)
return;
if (conn->flags & ACK_TO_TAP_DUE) {
tcp_send_flag(c, conn, ACK_IF_NEEDED);
conn_flag(c, conn, ~ACK_TO_TAP_DUE);
} else if (conn->flags & ACK_FROM_TAP_DUE) {
if (!(conn->events & ESTABLISHED)) {
debug("TCP: index %li, handshake timeout", conn - tc);
tcp_rst(c, conn);
} else if (CONN_HAS(conn, SOCK_FIN_SENT | TAP_FIN_ACKED)) {
debug("TCP: index %li, FIN timeout", conn - tc);
tcp_rst(c, conn);
} else if (conn->retrans == TCP_MAX_RETRANS) {
debug("TCP: index %li, retransmissions count exceeded",
conn - tc);
tcp_rst(c, conn);
} else {
debug("TCP: index %li, ACK timeout, retry", conn - tc);
conn->retrans++;
conn->seq_to_tap = conn->seq_ack_from_tap;
tcp_data_from_sock(c, conn);
tcp_timer_ctl(c, conn);
}
} else {
struct itimerspec new = { { 0 }, { ACT_TIMEOUT, 0 } };
struct itimerspec old = { { 0 }, { 0 } };
/* Activity timeout: if it was already set, reset the
* connection, otherwise, it was a left-over from ACK_TO_TAP_DUE
* or ACK_FROM_TAP_DUE, so just set the long timeout in that
* case. This avoids having to preemptively reset the timer on
* ~ACK_TO_TAP_DUE or ~ACK_FROM_TAP_DUE.
*/
timerfd_settime(conn->timer, 0, &new, &old);
if (old.it_value.tv_sec == ACT_TIMEOUT) {
debug("TCP: index %li, activity timeout", conn - tc);
tcp_rst(c, conn);
}
}
}
/**
* tcp_sock_handler() - Handle new data from socket, or timerfd event
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @c: Execution context
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
* @ref: epoll reference
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @events: epoll events bitmap
udp: Connection tracking for ephemeral, local ports, and related fixes As we support UDP forwarding for packets that are sent to local ports, we actually need some kind of connection tracking for UDP. While at it, this commit introduces a number of vaguely related fixes for issues observed while trying this out. In detail: - implement an explicit, albeit minimalistic, connection tracking for UDP, to allow usage of ephemeral ports by the guest and by the host at the same time, by binding them dynamically as needed, and to allow mapping address changes for packets with a loopback address as destination - set the guest MAC address whenever we receive a packet from tap instead of waiting for an ARP request, and set it to broadcast on start, otherwise DHCPv6 might not work if all DHCPv6 requests time out before the guest starts talking IPv4 - split context IPv6 address into address we assign, global or site address seen on tap, and link-local address seen on tap, and make sure we use the addresses we've seen as destination (link-local choice depends on source address). Similarly, for IPv4, split into address we assign and address we observe, and use the address we observe as destination - introduce a clock_gettime() syscall right after epoll_wait() wakes up, so that we can remove all the other ones and pass the current timestamp to tap and socket handlers -- this is additionally needed by UDP to time out bindings to ephemeral ports and mappings between loopback address and a local address - rename sock_l4_add() to sock_l4(), no semantic changes intended - include <arpa/inet.h> in passt.c before kernel headers so that we can use <netinet/in.h> macros to check IPv6 address types, and remove a duplicate <linux/ip.h> inclusion Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-04-29 14:59:20 +00:00
* @now: Current timestamp
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
void tcp_sock_handler(struct ctx *c, union epoll_ref ref, uint32_t events,
const struct timespec *now)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
struct tcp_conn *conn;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (ref.r.p.tcp.tcp.timer) {
tcp_timer_handler(c, ref);
return;
}
if (ref.r.p.tcp.tcp.splice) {
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_sock_handler_splice(c, ref, events);
return;
}
if (ref.r.p.tcp.tcp.listen) {
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_conn_from_sock(c, ref, now);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return;
}
if (!(conn = CONN_OR_NULL(ref.r.p.tcp.tcp.index)))
return;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (conn->events == CLOSED)
return;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
if (events & EPOLLERR) {
tcp_rst(c, conn);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return;
}
if ((conn->events & TAP_FIN_SENT) && (events & EPOLLHUP)) {
conn_event(c, conn, CLOSED);
return;
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}
if (conn->events & ESTABLISHED) {
if (CONN_HAS(conn, SOCK_FIN_SENT | TAP_FIN_ACKED))
conn_event(c, conn, CLOSED);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (events & (EPOLLRDHUP | EPOLLHUP))
conn_event(c, conn, SOCK_FIN_RCVD);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (events & EPOLLIN)
tcp_data_from_sock(c, conn);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (events & EPOLLOUT)
tcp_update_seqack_wnd(c, conn, 0, NULL);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
return;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
/* EPOLLHUP during handshake: reset */
if (events & EPOLLHUP) {
tcp_rst(c, conn);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
return;
}
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
/* Data during handshake tap-side: check later */
if (conn->events & SOCK_ACCEPTED)
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
return;
if (conn->events == TAP_SYN_RCVD) {
if (events & EPOLLOUT)
tcp_connect_finish(c, conn);
/* Data? Check later */
}
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
/**
* tcp_sock_init() - Initialise listening sockets for a given port
* @c: Execution context
* @ns: In pasta mode, if set, bind with loopback address in namespace
* @af: Address family to select a specific IP version, or AF_UNSPEC
* @addr: Pointer to address for binding, NULL if not configured
* @port: Port, host order
*/
void tcp_sock_init(const struct ctx *c, int ns, sa_family_t af,
const void *addr, in_port_t port)
{
union tcp_epoll_ref tref = { .tcp.listen = 1 };
const void *bind_addr;
int s;
if (ns) {
tref.tcp.index = (in_port_t)(port + c->tcp.fwd_out.delta[port]);
} else {
tref.tcp.index = (in_port_t)(port + c->tcp.fwd_in.delta[port]);
}
if (af == AF_INET || af == AF_UNSPEC) {
if (!addr && c->mode == MODE_PASTA)
bind_addr = &c->ip4.addr;
else
bind_addr = addr;
tref.tcp.v6 = 0;
tref.tcp.splice = 0;
if (!ns) {
s = sock_l4(c, AF_INET, IPPROTO_TCP, bind_addr, port,
tref.u32);
if (s >= 0)
tcp_sock_set_bufsize(c, s);
else
s = -1;
if (c->tcp.fwd_in.mode == FWD_AUTO)
tcp_sock_init_ext[port][V4] = s;
}
if (c->mode == MODE_PASTA) {
bind_addr = &(uint32_t){ htonl(INADDR_LOOPBACK) };
tref.tcp.splice = 1;
s = sock_l4(c, AF_INET, IPPROTO_TCP, bind_addr, port,
tref.u32);
if (s >= 0)
tcp_sock_set_bufsize(c, s);
else
s = -1;
if (c->tcp.fwd_out.mode == FWD_AUTO) {
if (ns)
tcp_sock_ns[port][V4] = s;
else
tcp_sock_init_lo[port][V4] = s;
}
}
}
if (af == AF_INET6 || af == AF_UNSPEC) {
if (!addr && c->mode == MODE_PASTA)
bind_addr = &c->ip6.addr;
else
bind_addr = addr;
tref.tcp.v6 = 1;
tref.tcp.splice = 0;
if (!ns) {
s = sock_l4(c, AF_INET6, IPPROTO_TCP, bind_addr, port,
tref.u32);
if (s >= 0)
tcp_sock_set_bufsize(c, s);
else
s = -1;
if (c->tcp.fwd_in.mode == FWD_AUTO)
tcp_sock_init_ext[port][V6] = s;
}
if (c->mode == MODE_PASTA) {
bind_addr = &in6addr_loopback;
tref.tcp.splice = 1;
s = sock_l4(c, AF_INET6, IPPROTO_TCP, bind_addr, port,
tref.u32);
if (s >= 0)
tcp_sock_set_bufsize(c, s);
else
s = -1;
if (c->tcp.fwd_out.mode == FWD_AUTO) {
if (ns)
tcp_sock_ns[port][V6] = s;
else
tcp_sock_init_lo[port][V6] = s;
}
}
}
}
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
/**
* tcp_sock_init_ns() - Bind sockets in namespace for inbound connections
* @arg: Execution context
*
passt, pasta: Namespace-based sandboxing, defer seccomp policy application To reach (at least) a conceptually equivalent security level as implemented by --enable-sandbox in slirp4netns, we need to create a new mount namespace and pivot_root() into a new (empty) mountpoint, so that passt and pasta can't access any filesystem resource after initialisation. While at it, also detach IPC, PID (only for passt, to prevent vulnerabilities based on the knowledge of a target PID), and UTS namespaces. With this approach, if we apply the seccomp filters right after the configuration step, the number of allowed syscalls grows further. To prevent this, defer the application of seccomp policies after the initialisation phase, before the main loop, that's where we expect bad things to happen, potentially. This way, we get back to 22 allowed syscalls for passt and 34 for pasta, on x86_64. While at it, move #syscalls notes to specific code paths wherever it conceptually makes sense. We have to open all the file handles we'll ever need before sandboxing: - the packet capture file can only be opened once, drop instance numbers from the default path and use the (pre-sandbox) PID instead - /proc/net/tcp{,v6} and /proc/net/udp{,v6}, for automatic detection of bound ports in pasta mode, are now opened only once, before sandboxing, and their handles are stored in the execution context - the UNIX domain socket for passt is also bound only once, before sandboxing: to reject clients after the first one, instead of closing the listening socket, keep it open, accept and immediately discard new connection if we already have a valid one Clarify the (unchanged) behaviour for --netns-only in the man page. To actually make passt and pasta processes run in a separate PID namespace, we need to unshare(CLONE_NEWPID) before forking to background (if configured to do so). Introduce a small daemon() implementation, __daemon(), that additionally saves the PID file before forking. While running in foreground, the process itself can't move to a new PID namespace (a process can't change the notion of its own PID): mention that in the man page. For some reason, fork() in a detached PID namespace causes SIGTERM and SIGQUIT to be ignored, even if the handler is still reported as SIG_DFL: add a signal handler that just exits. We can now drop most of the pasta_child_handler() implementation, that took care of terminating all processes running in the same namespace, if pasta started a shell: the shell itself is now the init process in that namespace, and all children will terminate once the init process exits. Issuing 'echo $$' in a detached PID namespace won't return the actual namespace PID as seen from the init namespace: adapt demo and test setup scripts to reflect that. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2022-02-07 20:11:37 +00:00
* Return: 0
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
*/
static int tcp_sock_init_ns(void *arg)
{
struct ctx *c = (struct ctx *)arg;
unsigned port;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
ns_enter(c);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
for (port = 0; port < NUM_PORTS; port++) {
if (!bitmap_isset(c->tcp.fwd_out.map, port))
continue;
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
tcp_sock_init(c, 1, AF_UNSPEC, NULL, port);
passt: Add PASTA mode, major rework PASTA (Pack A Subtle Tap Abstraction) provides quasi-native host connectivity to an otherwise disconnected, unprivileged network and user namespace, similarly to slirp4netns. Given that the implementation is largely overlapping with PASST, no separate binary is built: 'pasta' (and 'passt4netns' for clarity) both link to 'passt', and the mode of operation is selected depending on how the binary is invoked. Usage example: $ unshare -rUn # echo $$ 1871759 $ ./pasta 1871759 # From another terminal # udhcpc -i pasta0 2>/dev/null # ping -c1 pasta.pizza PING pasta.pizza (64.190.62.111) 56(84) bytes of data. 64 bytes from 64.190.62.111 (64.190.62.111): icmp_seq=1 ttl=255 time=34.6 ms --- pasta.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 34.575/34.575/34.575/0.000 ms # ping -c1 spaghetti.pizza PING spaghetti.pizza(2606:4700:3034::6815:147a (2606:4700:3034::6815:147a)) 56 data bytes 64 bytes from 2606:4700:3034::6815:147a (2606:4700:3034::6815:147a): icmp_seq=1 ttl=255 time=29.0 ms --- spaghetti.pizza ping statistics --- 1 packets transmitted, 1 received, 0% packet loss, time 0ms rtt min/avg/max/mdev = 28.967/28.967/28.967/0.000 ms This entails a major rework, especially with regard to the storage of tracked connections and to the semantics of epoll(7) references. Indexing TCP and UDP bindings merely by socket proved to be inflexible and unsuitable to handle different connection flows: pasta also provides Layer-2 to Layer-2 socket mapping between init and a separate namespace for local connections, using a pair of splice() system calls for TCP, and a recvmmsg()/sendmmsg() pair for UDP local bindings. For instance, building on the previous example: # ip link set dev lo up # iperf3 -s $ iperf3 -c ::1 -Z -w 32M -l 1024k -P2 | tail -n4 [SUM] 0.00-10.00 sec 52.3 GBytes 44.9 Gbits/sec 283 sender [SUM] 0.00-10.43 sec 52.3 GBytes 43.1 Gbits/sec receiver iperf Done. epoll(7) references now include a generic part in order to demultiplex data to the relevant protocol handler, using 24 bits for the socket number, and an opaque portion reserved for usage by the single protocol handlers, in order to track sockets back to corresponding connections and bindings. A number of fixes pertaining to TCP state machine and congestion window handling are also included here. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-07-17 06:34:53 +00:00
}
return 0;
}
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
/**
* struct tcp_sock_refill_arg - Arguments for tcp_sock_refill()
* @c: Execution context
* @ns: Set to refill pool of sockets created in namespace
*/
struct tcp_sock_refill_arg {
struct ctx *c;
int ns;
};
/**
* tcp_sock_refill() - Refill pool of pre-opened sockets
* @arg: See @tcp_sock_refill_arg
*
* Return: 0
*/
static int tcp_sock_refill(void *arg)
{
struct tcp_sock_refill_arg *a = (struct tcp_sock_refill_arg *)arg;
int i, *p4, *p6;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
if (a->ns) {
passt, pasta: Namespace-based sandboxing, defer seccomp policy application To reach (at least) a conceptually equivalent security level as implemented by --enable-sandbox in slirp4netns, we need to create a new mount namespace and pivot_root() into a new (empty) mountpoint, so that passt and pasta can't access any filesystem resource after initialisation. While at it, also detach IPC, PID (only for passt, to prevent vulnerabilities based on the knowledge of a target PID), and UTS namespaces. With this approach, if we apply the seccomp filters right after the configuration step, the number of allowed syscalls grows further. To prevent this, defer the application of seccomp policies after the initialisation phase, before the main loop, that's where we expect bad things to happen, potentially. This way, we get back to 22 allowed syscalls for passt and 34 for pasta, on x86_64. While at it, move #syscalls notes to specific code paths wherever it conceptually makes sense. We have to open all the file handles we'll ever need before sandboxing: - the packet capture file can only be opened once, drop instance numbers from the default path and use the (pre-sandbox) PID instead - /proc/net/tcp{,v6} and /proc/net/udp{,v6}, for automatic detection of bound ports in pasta mode, are now opened only once, before sandboxing, and their handles are stored in the execution context - the UNIX domain socket for passt is also bound only once, before sandboxing: to reject clients after the first one, instead of closing the listening socket, keep it open, accept and immediately discard new connection if we already have a valid one Clarify the (unchanged) behaviour for --netns-only in the man page. To actually make passt and pasta processes run in a separate PID namespace, we need to unshare(CLONE_NEWPID) before forking to background (if configured to do so). Introduce a small daemon() implementation, __daemon(), that additionally saves the PID file before forking. While running in foreground, the process itself can't move to a new PID namespace (a process can't change the notion of its own PID): mention that in the man page. For some reason, fork() in a detached PID namespace causes SIGTERM and SIGQUIT to be ignored, even if the handler is still reported as SIG_DFL: add a signal handler that just exits. We can now drop most of the pasta_child_handler() implementation, that took care of terminating all processes running in the same namespace, if pasta started a shell: the shell itself is now the init process in that namespace, and all children will terminate once the init process exits. Issuing 'echo $$' in a detached PID namespace won't return the actual namespace PID as seen from the init namespace: adapt demo and test setup scripts to reflect that. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2022-02-07 20:11:37 +00:00
ns_enter(a->c);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
p4 = ns_sock_pool4;
p6 = ns_sock_pool6;
} else {
p4 = init_sock_pool4;
p6 = init_sock_pool6;
}
for (i = 0; a->c->ifi4 && i < TCP_SOCK_POOL_SIZE; i++, p4++) {
if (*p4 >= 0)
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
break;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
*p4 = socket(AF_INET, SOCK_STREAM | SOCK_NONBLOCK, IPPROTO_TCP);
if (*p4 > SOCKET_MAX) {
close(*p4);
*p4 = -1;
return -EIO;
}
if (*p4 >= 0)
tcp_sock_set_bufsize(a->c, *p4);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
for (i = 0; a->c->ifi6 && i < TCP_SOCK_POOL_SIZE; i++, p6++) {
if (*p6 >= 0)
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
break;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
*p6 = socket(AF_INET6, SOCK_STREAM | SOCK_NONBLOCK,
IPPROTO_TCP);
if (*p6 > SOCKET_MAX) {
close(*p6);
*p6 = -1;
return -EIO;
}
if (*p6 >= 0)
tcp_sock_set_bufsize(a->c, *p6);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
return 0;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_init() - Get initial sequence, hash secret, initialise per-socket data
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @c: Execution context
*
* Return: 0, doesn't return on failure
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
int tcp_init(struct ctx *c)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
struct tcp_sock_refill_arg refill_arg = { c, 0 };
int i;
#ifndef HAS_GETRANDOM
int dev_random = open("/dev/random", O_RDONLY);
unsigned int random_read = 0;
while (dev_random && random_read < sizeof(c->tcp.hash_secret)) {
int ret = read(dev_random,
(uint8_t *)&c->tcp.hash_secret + random_read,
sizeof(c->tcp.hash_secret) - random_read);
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
if (ret == -1 && errno == EINTR)
continue;
if (ret <= 0)
break;
random_read += ret;
}
if (dev_random >= 0)
close(dev_random);
if (random_read < sizeof(c->tcp.hash_secret)) {
#else
if (getrandom(&c->tcp.hash_secret, sizeof(c->tcp.hash_secret),
GRND_RANDOM) < 0) {
#endif /* !HAS_GETRANDOM */
perror("TCP initial sequence getrandom");
exit(EXIT_FAILURE);
}
for (i = 0; i < ARRAY_SIZE(tcp_l2_mh); i++)
tcp_l2_mh[i] = (struct mmsghdr) { .msg_hdr.msg_iovlen = 1 };
if (c->ifi4)
tcp_sock4_iov_init();
if (c->ifi6)
tcp_sock6_iov_init();
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
memset(init_sock_pool4, 0xff, sizeof(init_sock_pool4));
memset(init_sock_pool6, 0xff, sizeof(init_sock_pool6));
memset(ns_sock_pool4, 0xff, sizeof(ns_sock_pool4));
memset(ns_sock_pool6, 0xff, sizeof(ns_sock_pool6));
memset(tcp_sock_init_lo, 0xff, sizeof(tcp_sock_init_lo));
memset(tcp_sock_init_ext, 0xff, sizeof(tcp_sock_init_ext));
memset(tcp_sock_ns, 0xff, sizeof(tcp_sock_ns));
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
tcp_sock_refill(&refill_arg);
if (c->mode == MODE_PASTA) {
tcp_splice_init(c);
NS_CALL(tcp_sock_init_ns, c);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
refill_arg.ns = 1;
NS_CALL(tcp_sock_refill, &refill_arg);
tcp_splice_timer(c);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
return 0;
}
/**
* struct tcp_port_detect_arg - Arguments for tcp_port_detect()
* @c: Execution context
* @detect_in_ns: Detect ports bound in namespace, not in init
*/
struct tcp_port_detect_arg {
struct ctx *c;
int detect_in_ns;
};
/**
* tcp_port_detect() - Detect ports bound in namespace or init
* @arg: See struct tcp_port_detect_arg
*
* Return: 0
*/
static int tcp_port_detect(void *arg)
{
struct tcp_port_detect_arg *a = (struct tcp_port_detect_arg *)arg;
if (a->detect_in_ns) {
ns_enter(a->c);
get_bound_ports(a->c, 1, IPPROTO_TCP);
} else {
get_bound_ports(a->c, 0, IPPROTO_TCP);
}
return 0;
}
/**
* struct tcp_port_rebind_arg - Arguments for tcp_port_rebind()
* @c: Execution context
* @bind_in_ns: Rebind ports in namespace, not in init
*/
struct tcp_port_rebind_arg {
struct ctx *c;
int bind_in_ns;
};
/**
* tcp_port_rebind() - Rebind ports in namespace or init
* @arg: See struct tcp_port_rebind_arg
*
* Return: 0
*/
static int tcp_port_rebind(void *arg)
{
struct tcp_port_rebind_arg *a = (struct tcp_port_rebind_arg *)arg;
unsigned port;
if (a->bind_in_ns) {
ns_enter(a->c);
for (port = 0; port < NUM_PORTS; port++) {
if (!bitmap_isset(a->c->tcp.fwd_out.map, port)) {
if (tcp_sock_ns[port][V4] >= 0) {
close(tcp_sock_ns[port][V4]);
tcp_sock_ns[port][V4] = -1;
}
if (tcp_sock_ns[port][V6] >= 0) {
close(tcp_sock_ns[port][V6]);
tcp_sock_ns[port][V6] = -1;
}
continue;
}
/* Don't loop back our own ports */
if (bitmap_isset(a->c->tcp.fwd_in.map, port))
continue;
if ((a->c->ifi4 && tcp_sock_ns[port][V4] == -1) ||
(a->c->ifi6 && tcp_sock_ns[port][V6] == -1))
tcp_sock_init(a->c, 1, AF_UNSPEC, NULL, port);
}
} else {
for (port = 0; port < NUM_PORTS; port++) {
if (!bitmap_isset(a->c->tcp.fwd_in.map, port)) {
if (tcp_sock_init_ext[port][V4] >= 0) {
close(tcp_sock_init_ext[port][V4]);
tcp_sock_init_ext[port][V4] = -1;
}
if (tcp_sock_init_ext[port][V6] >= 0) {
close(tcp_sock_init_ext[port][V6]);
tcp_sock_init_ext[port][V6] = -1;
}
if (tcp_sock_init_lo[port][V4] >= 0) {
close(tcp_sock_init_lo[port][V4]);
tcp_sock_init_lo[port][V4] = -1;
}
if (tcp_sock_init_lo[port][V6] >= 0) {
close(tcp_sock_init_lo[port][V6]);
tcp_sock_init_lo[port][V6] = -1;
}
continue;
}
/* Don't loop back our own ports */
if (bitmap_isset(a->c->tcp.fwd_out.map, port))
continue;
if ((a->c->ifi4 && tcp_sock_init_ext[port][V4] == -1) ||
(a->c->ifi6 && tcp_sock_init_ext[port][V6] == -1))
tcp_sock_init(a->c, 0, AF_UNSPEC, NULL, port);
}
}
return 0;
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
/**
* tcp_timer() - Periodic tasks: port detection, closed connections, pool refill
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
* @c: Execution context
* @ts: Unused
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
*/
void tcp_timer(struct ctx *c, const struct timespec *ts)
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
{
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
struct tcp_sock_refill_arg refill_arg = { c, 0 };
struct tcp_conn *conn;
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
(void)ts;
if (c->mode == MODE_PASTA) {
struct tcp_port_detect_arg detect_arg = { c, 0 };
struct tcp_port_rebind_arg rebind_arg = { c, 0 };
if (c->tcp.fwd_in.mode == FWD_AUTO) {
detect_arg.detect_in_ns = 0;
tcp_port_detect(&detect_arg);
rebind_arg.bind_in_ns = 1;
NS_CALL(tcp_port_rebind, &rebind_arg);
}
if (c->tcp.fwd_out.mode == FWD_AUTO) {
detect_arg.detect_in_ns = 1;
NS_CALL(tcp_port_detect, &detect_arg);
rebind_arg.bind_in_ns = 0;
tcp_port_rebind(&rebind_arg);
}
}
for (conn = CONN(c->tcp.conn_count - 1); conn >= tc; conn--) {
if (conn->events == CLOSED)
tcp_conn_destroy(c, conn);
}
tcp_sock_refill(&refill_arg);
if (c->mode == MODE_PASTA) {
refill_arg.ns = 1;
if ((c->ifi4 && ns_sock_pool4[TCP_SOCK_POOL_TSH] < 0) ||
(c->ifi6 && ns_sock_pool6[TCP_SOCK_POOL_TSH] < 0))
NS_CALL(tcp_sock_refill, &refill_arg);
tcp_splice_timer(c);
tcp: Rework window handling, timers, add SO_RCVLOWAT and pools for sockets/pipes This introduces a number of fundamental changes that would be quite messy to split. Summary: - advertised window scaling can be as big as we want, we just need to clamp window sizes to avoid exceeding the size of our "discard" buffer for unacknowledged data from socket - add macros to compare sequence numbers - force sending ACK to guest/tap on PSH segments, always in pasta mode, whenever we see an overlapping segment, or when we reach a given threshold compared to our window - we don't actually use recvmmsg() here, fix comments and label - introduce pools for pre-opened sockets and pipes, to decrease latency on new connections - set receiving and sending buffer sizes to the maximum allowed, kernel will clamp and round appropriately - defer clean-up of spliced and non-spliced connection to timer - in tcp_send_to_tap(), there's no need anymore to keep a large buffer, shrink it down to what we actually need - introduce SO_RCVLOWAT setting and activity tracking for spliced connections, to coalesce data moved by splice() calls as much as possible - as we now have a compacted connection table, there's no need to keep sparse bitmaps tracking connection activity -- simply go through active connections with a loop in the timer handler - always clamp the advertised window to half our sending buffer, too, to minimise retransmissions from the guest/tap - set TCP_QUICKACK for originating socket in spliced connections, there's no need to delay them - fix up timeout for unacknowledged data from socket Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-09-19 00:29:05 +00:00
}
passt: New design and implementation with native Layer 4 sockets This is a reimplementation, partially building on the earlier draft, that uses L4 sockets (SOCK_DGRAM, SOCK_STREAM) instead of SOCK_RAW, providing L4-L2 translation functionality without requiring any security capability. Conceptually, this follows the design presented at: https://gitlab.com/abologna/kubevirt-and-kvm/-/blob/master/Networking.md The most significant novelty here comes from TCP and UDP translation layers. In particular, the TCP state and translation logic follows the intent of being minimalistic, without reimplementing a full TCP stack in either direction, and synchronising as much as possible the TCP dynamic and flows between guest and host kernel. Another important introduction concerns addressing, port translation and forwarding. The Layer 4 implementations now attempt to bind on all unbound ports, in order to forward connections in a transparent way. While at it: - the qemu 'tap' back-end can't be used as-is by qrap anymore, because of explicit checks now introduced in qemu to ensure that the corresponding file descriptor is actually a tap device. For this reason, qrap now operates on a 'socket' back-end type, accounting for and building the additional header reporting frame length - provide a demo script that sets up namespaces, addresses and routes, and starts the daemon. A virtual machine started in the network namespace, wrapped by qrap, will now directly interface with passt and communicate using Layer 4 sockets provided by the host kernel. Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
2021-02-16 06:25:09 +00:00
}