By merging receive buffers through the VIRTIO_NET_F_MRG_RXBUF feature,
as well as enabling the use of indirect descriptors through
VIRTIO_RING_F_INDIRECT_DESC feature, we achieve better throughput for
the virtio-net device without hurting its latency.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Since each segment must have a non-overlapping memory range associated
with it the device memory must be equally divided amongst all segments.
A new allocator is used for each segment to ensure that BARs are
allocated from the correct address ranges. This requires changes to
PciDevice::allocate/free_bars to take that allocator and when
reallocating BARs the correct allocator must be identified from the
ranges.
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
Move the decision on whether to use a 64-bit bar up to the DeviceManager
so that it can use both the device type (e.g. block) and the PCI segment
ID to decide what size bar should be used.
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
Relying on the vm-virtio/virtio-queue crate from rust-vmm which has been
copied inside the Cloud Hypervisor tree, the entire codebase is moved to
the new definition of a Queue and other related structures.
The reason for this move is to follow the upstream until we get some
agreement for the patches that we need on top of that to make it
properly work with Cloud Hypervisor.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This was causing some issues because of the use of 2 different versions
for the vm-memmory crate. We'll wait for all dependencies to be properly
resolved before we move to 0.7.0.
This reverts commit 76b6c62d07.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
These structs are not read on the VMM side but are used in communication
with the guest.
As identified by the new beta clippy.
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
There's no need to patch the vhost crate anymore since the fixes we were
looking for have been released as part of 0.2.0 on crates.io.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Setting the reply_ack should depend on the set of acknowledged features
containing the REPLY_ACK flag.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
In order to support correctly the snapshot/restore and migration use
cases, we must be careful with the ranges that we discard by punching
holes. On restore, there might be some ranges already plugged in,
meaning they should not be discarded. That's why we loop over the list
of blocks to discard only the ranges that are marked as unplugged.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
By creating the BlocksState object in the MemoryManager, we can directly
provide it to the virtio-mem device when being created. This will allow
the MemoryManager through each VirtioMemZone to have a handle onto the
blocks that are plugged at any point in time.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This is going to be useful to let virtio-mem report the list of ranges
that are currently plugged, so that both snapshot/restore and migration
will copy only what is needed.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This will be helpful to support the creation of a MemoryRangeTable from
virtio-mem, as it uses 2M pages.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Adding the snapshot/restore support along with migration as well,
allowing a VM with virtio-mem devices attached to be properly
migrated.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Implement the infrastructure that lets a virtio-mem device map the guest
memory into the device. This is necessary since with virtio-mem zones
memory can be added or removed and the vfio-user device must be
informed.
Fixes: #3025
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
For vfio-user the mapping handler is per device and needs to be removed
when the device in unplugged.
For VFIO the mapping handler is for the default VFIO container (used
when no vIOMMU is used - using a vIOMMU does not require mappings with
virtio-mem)
To represent these two use cases use an enum for the handlers that are
stored.
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
Adding the snapshot/restore support along with migration as well,
allowing a VM with a virtio-balloon device attached to be properly
migrated.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
The MSI IOVA address on X86 and AArch64 is different.
This commit refactored the code to receive the MSI IOVA address and size
from device_manager, which provides the actual IOVA space data for both
architectures.
Signed-off-by: Michael Zhao <michael.zhao@arm.com>
For most use cases, there is no need to create multiple VFIO containers
as it causes unwanted behaviors. Especially when passing multiple
devices from the same IOMMU group, we need to use the same container so
that it can properly list the groups that have been already opened. The
correct logic was already there in vfio-ioctls, but it was incorrectly
used from our VMM implementation.
For the special case where we put a VFIO device behind a vIOMMU, we must
create one container per device, as we need to control the DMA mappings
per device, which is performed at the container level. Because we must
keep one container per device, the vIOMMU use case prevents multiple
devices attached to the same IOMMU group to be passed through the VM.
But this is a limitation that we are fine with, especially since the
vIOMMU doesn't let us group multiple devices in the same group from a
guest perspective.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
When a pty is resized (using the TIOCSWINSZ ioctl -- see ioctl_tty(2)),
the kernel will send a SIGWINCH signal to the pty's foreground process
group to notify it of the resize. This is the only way to be notified
by the kernel of a pty resize.
We can't just make the cloud-hypervisor process's process group the
foreground process group though, because a process can only set the
foreground process group of its controlling terminal, and
cloud-hypervisor's controlling terminal will often be the terminal the
user is running it in. To work around this, we fork a subprocess in a
new process group, and set its process group to be the foreground
process group of the pty. The subprocess additionally must be running
in a new session so that it can have a different controlling
terminal. This subprocess writes a byte to a pipe every time the pty
is resized, and the virtio-console device can listen for this in its
epoll loop.
Alternatives I considered were to have the subprocess just send
SIGWINCH to its parent, and to use an eventfd instead of a pipe.
I decided against the signal approach because re-purposing a signal
that has a very specific meaning (even if this use was only slightly
different to its normal meaning) felt unclean, and because it would
have required using pidfds to avoid race conditions if
cloud-hypervisor had terminated, which added complexity. I decided
against using an eventfd because using a pipe instead allows the child
to be notified (via poll(2)) when nothing is reading from the pipe any
more, meaning it can be reliably notified of parent death and
terminate itself immediately.
I used clone3(2) instead of fork(2) because without
CLONE_CLEAR_SIGHAND the subprocess would inherit signal-hook's signal
handlers, and there's no other straightforward way to restore all signal
handlers to their defaults in the child process. The only way to do
it would be to iterate through all possible signals, or maintain a
global list of monitored signals ourselves (vmm:vm::HANDLED_SIGNALS is
insufficient because it doesn't take into account e.g. the SIGSYS
signal handler that catches seccomp violations).
Signed-off-by: Alyssa Ross <hi@alyssa.is>
This prepares us to be able to handle console resizes in the console
device's epoll loop, which we'll have to do if the output is a pty,
since we won't get SIGWINCH from it.
Signed-off-by: Alyssa Ross <hi@alyssa.is>
Musl often uses mmap to allocate memory where Glibc would use brk.
This has caused seccomp violations for me on the API and signal
handling threads.
Signed-off-by: Alyssa Ross <hi@alyssa.is>
As well as reducing the amount of code this also improves the binary
size slightly:
cargo bloat --release -n 2000 --bin cloud-hypervisor | grep virtio_devices::seccomp_filters::get_seccomp_rules
Before:
0.1% 0.2% 7.8KiB virtio_devices virtio_devices::seccomp_filters::get_seccomp_rules
After:
0.0% 0.1% 3.0KiB virtio_devices virtio_devices::seccomp_filters::get_seccomp_rules
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
Introduce a common solution for spawning the virtio threads which will
make it easier to add the panic handling.
During this effort I discovered that there were no seccomp filters
registered for the vhost-user-net thread nor the vhost-user-block
thread. This change also incorporates basic seccomp filters for those as
part of the refactoring.
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
Move the processing of the input from stdin, PTY or file from the VMM
thread to the existing virtio-console thread. The handling of the resize
of a virtio-console has not changed but the name of the struct used to
support that has been renamed to reflect its usage.
Fixes: #3060
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
vhdx_sync.rs in block_util implements traits to represent the vhdx
crate as a supported block device in the cloud hypervisor. The vhdx
is added to the block device list in device_manager.rs at the vmm
crate so that it can automatically detect a vhdx disk and invoke the
corresponding crate.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Signed-off-by: Fazla Mehrab <akm.fazla.mehrab@intel.com>
We are relying on applying empty 'seccomp' filters to support the
'--seccomp false' option, which will be treated as an error with the
updated 'seccompiler' crate. This patch fixes this issue by explicitly
checking whether the 'seccomp' filter is empty before applying the
filter.
Signed-off-by: Bo Chen <chen.bo@intel.com>
Introducing a new structure VhostUserCommon allowing to factorize a lot
of the code shared between the vhost-user devices (block, fs and net).
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
We cannot let vhost-user devices connect to the backend when the Block,
Fs or Net object is being created during a restore/migration. The reason
is we can't have two VMs (source and destination) connected to the same
backend at the same time. That's why we must delay the connection with
the vhost-user backend until the restoration is performed.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Introducing a new function to factorize a small part of the
initialization that is shared between a full reinitialization and a
restoration.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
In order to prevent the vhost-user devices from reconnecting to the
backend after the migration has been successfully performed, we make
sure to kill the thread in charge of handling the reconnection
mechanism.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
During a migration, the vhost-user device talks to the backend to
retrieve the dirty pages. Once done with this, a snapshot will be taken,
meaning there's no need to communicate with the backend anymore. Closing
the communication is needed to let the destination VM being able to
connect to the same backend.
That's why we shutdown the communication with the backend in case a
migration has been started and we're asked for a snapshot.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This anticipates the need for creating a new Blk, Fs or Net object
without having performed the connection with the vhost-user backend yet.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
It was incorrect to call Vec::from_raw_parts() on the address pointing
to the shared memory log region since Vec is a Rust specific structure
that doesn't directly translate into bytes. That's why we use the same
function from std::slice in order to create a proper slice out of the
memory region, which is then copied into a Vec.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Now that the common vhost-user code can handle logging dirty pages
through shared memory, we need to advertise it to the vhost-user
backends with the protocol feature VHOST_USER_PROTOCOL_F_LOG_SHMFD.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Allow vsocks to connect to Unix sockets on the host running
cloud-hypervisor with enabled seccomp.
Reported-by: Philippe Schaaf <philippe.schaaf@secunet.com>
Tested-by: Franz Girlich <franz.girlich@tu-ilmenau.de>
Signed-off-by: Markus Theil <markus.theil@tu-ilmenau.de>
This doesn't really affect the build as we ship a Cargo.lock with fixed
versions in. However for clarity it makes sense to use fixed versions
throughout and let dependabot update them.
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
Adding the common vhost-user code for starting logging dirty pages when
the migration is started, and its counterpart for stopping, as well as
the code in charge of retrieving the bitmap of the dirty pages that have
been logged.
All these functions are meant to be leveraged from vhost-user devices.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Adding a simple field `migration_support` to VhostUserHandle in order to
store the information about the device supporting migration or not. The
value of this flag depends on the feature set negotiated with the
backend. It's considered as supporting migration if VHOST_F_LOG_ALL is
present in the virtio features and if VHOST_USER_PROTOCOL_F_LOG_SHMFD is
present in the vhost-user protocol features.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
backend like SPDK required to know how many virt queues to be handled
before gets VHOST_USER_SET_INFLIGHT_FD message.
fix dpdk core dump while processing vhost_user_set_inflight_fd:
#0 0x00007fffef47c347 in vhost_user_set_inflight_fd (pdev=0x7fffe2895998, msg=0x7fffe28956f0, main_fd=545) at ../lib/librte_vhost/vhost_user.c:1570
#1 0x00007fffef47e7b9 in vhost_user_msg_handler (vid=0, fd=545) at ../lib/librte_vhost/vhost_user.c:2735
#2 0x00007fffef46bac0 in vhost_user_read_cb (connfd=545, dat=0x7fffdc0008c0, remove=0x7fffe2895a64) at ../lib/librte_vhost/socket.c:309
#3 0x00007fffef45b3f6 in fdset_event_dispatch (arg=0x7fffef6dc2e0 <vhost_user+8192>) at ../lib/librte_vhost/fd_man.c:286
#4 0x00007ffff09926f3 in rte_thread_init (arg=0x15ee180) at ../lib/librte_eal/common/eal_common_thread.c:175
Signed-off-by: Arafatms <arafatms@outlook.com>
This patch adds all the seccomp rules missing for MSHV.
With this patch MSFT internal CI runs with seccomp enabled.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
The vhost-user-net backend needs to prepare all queues before enabling vring.
For example, DPVGW will report 'the RX queue can't find' error, if we enable
vring immediately after kicking it out.
Signed-off-by: Arafatms <arafatms@outlook.com>
Adding the support for snapshot/restore feature for all supported
vhost-user devices.
The complexity of vhost-user-fs device makes it only partially
compatible with the feature. When using the DAX feature, there's no way
to store and remap what was previously mapped in the DAX region. And
when not using the cache region, if the filesystem is mounted, it fails
to be properly restored as this would require a special command to let
the backend know that it must remount what was already mounted before.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This patch moves all vhost-user common functions behind a new structure
VhostUserHandle. There is no functional changes intended, the only goal
being to prepare for storing information through this new structure,
limiting the amount of parameters that are needed for each function.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
With the new beta version, clippy complains about redundant allocation
when using Arc<Box<dyn T>>, and suggests replacing it simply with
Arc<dyn T>.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
There are some seccomp rules needed for MSHV
in virtio-devices but not for KVM. We only want to
add those rules based on MSHV feature guard.
Signed-off-by: Muminul Islam <muislam@microsoft.com>
The cloud hypervisor tells the VM and the backend to support the PACKED_RING feature,
but it actually processes various variables according to the split ring logic, such
as last_avail_index. Eventually it will cause the following error (SPDK as an example):
vhost.c: 516:vhost_vq_packed_ring_enqueue: *ERROR*: descriptor has been used before
vhost_blk.c: 596:process_blk_task: *ERROR*: ====== Task 0x200113784640 req_idx 0 failed ======
vhost.c: 629:vhost_vring_desc_payload_to_iov: *ERROR*: gpa_to_vva((nil)) == NULL
Signed-off-by: Arafatms <arafatms@outlook.com>
If writing to the TAP returns EAGAIN then listen for the TAP to be
writable. When the TAP becomes writable attempt to process the TX queue
again.
Fixes: #2807
Signed-off-by: Rob Bradford <robert.bradford@intel.com>
This dependency bump needed some manual handling since the API changed
quite a lot regarding some RawFd being changed into either File or
AsRawFd traits.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Issue from beta verion of clippy:
Error: --> vm-virtio/src/queue.rs:700:59
|
700 | if let Some(used_event) = self.get_used_event(&mem) {
| ^^^^ help: change this to: `mem`
|
= note: `-D clippy::needless-borrow` implied by `-D warnings`
= help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#needless_borrow
Signed-off-by: Bo Chen <chen.bo@intel.com>
Now that vhost crate allows the caller to set the header flags, we can
set NEED_REPLY whenever the REPLY_ACK protocol feature is supported from
both ends.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
For vhost-user devices, memory should be shared between CLH and
vhost-user backend. However, madvise DONTNEED doesn't working in
this case. So, let's use fallocate PUNCH_HOLE to discard those
memory regions instead.
Signed-off-by: Li Hangjing <lihangjing@bytedance.com>
Sometimes we need balloon deflate automatically to give memory
back to guest, especially for some low priority guest processes
under memory pressure. Enable deflate_on_oom to support this.
Usage: --balloon "size=0,deflate_on_oom=on" \
Signed-off-by: Fei Li <lifei.shirley@bytedance.com>
Since using the VIRTIO configuration to expose the virtual IOMMU
topology has been deprecated, the virtio-iommu implementation must be
updated.
In order to follow the latest patchset that is about to be merged in the
upstream Linux kernel, it must rely on ACPI, and in particular the newly
introduced VIOT table to expose the information about the list of PCI
devices attached to the virtual IOMMU.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
Since the reconnection thread took on the responsibility to handle
backend initiated requests as well, the variable naming should reflect
this by avoiding the "reconnect" prefix.
Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>