Commit Graph

7443 Commits

Author SHA1 Message Date
Samuel Ortiz
db5b4763c2 vfio: Initial PCI support
This brings the initial PCI support to the VFIO crate.
The VfioPciDevice is the main structure and holds an inner VfioDevice.

VfioPciDevice implements the PCI trait, leaving the IRQ assignments
empty as this will be driven by both the guest and the VFIO PCI device,
not by the VMM.

As we must trap BAR programming from the guest (We don't want to program
the actual device with guest addresses), we use our local PCI
configuration cache to read and write BARs.

Signed-off-by: Zhang, Xiong Y <xiong.y.zhang@intel.com>
Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-24 11:55:08 +02:00
Samuel Ortiz
2cec3aad7f vfio: VFIO API wrappers and helpers
The Virtual Function I/O (VFIO) kernel subsystem exposes a vast and
relatively complex userspace API. This commit abstracts and simplifies
this API into both an internal and external API.

The external API is to be consumed by VFIO device implementation through
the VfioDevice structure. A VfioDevice instance can:

- Enable and disable all interrupts (INTX, MSI and MSI-X) on the
  underlying VFIO device.
- Read and write all of the VFIO device memory regions.
- Set the system's IOMMU tables for the underlying device.

Signed-off-by: Zhang, Xiong Y <xiong.y.zhang@intel.com>
Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-24 11:55:08 +02:00
Samuel Ortiz
5372554ed4 vfio-bindings: Initial commit
The default bindings are generated from the 5.0.0 Linux userspace API.

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-24 11:55:08 +02:00
Samuel Ortiz
4e48309660 vm: Factorize all virtio devices creation routines
Our DeviceManager::new() routine is reaching north of 250 lines.
For simplicity and readbility sake, extract all virtio devices creation
code into their own routines.

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-23 08:41:37 +01:00
fazlamehrab
8ba54af71d vm-virtio: Add integration test for virtio console device
Two integration tests are added for testing the implemented virtio
console device for single port operation. One checks the presence
and the simple stdout operation. The other test checks the stdout
on file (option: file) using virtio console.

Signed-off-by: A K M Fazla Mehrab <fazla.mehrab.akm@intel.com>
2019-07-22 23:08:56 +01:00
fazlamehrab
24438e0390 vm-virtio: Enable the vmm support for virtio-console
To use the implemented virtio console device, the users can select one
of the three options ("off", "tty" or "file=/path/to/the/file") with
the command line argument "--console". By default, the console is
enabled as a device named "hvc0" (option: tty). When "off" option is
used, the console device is not added to the VM configuration at all.

Signed-off-by: A K M Fazla Mehrab <fazla.mehrab.akm@intel.com>
2019-07-22 23:08:56 +01:00
fazlamehrab
577d44c8eb vm-virtio: Add virtio console device for single port operation
The virtio console device is a console for the communication between
the host and guest userspace. It has two parts: the device and the
driver. The console device is implemented here as a virtio-pci device
to the guest. On the other side, the guest OS expected to have a
character device driver which provides an interface to the userspace
applications.

The console device can have multiple ports where each port has one
transmit queue and one receive queue. The current implementation only
supports one port. For data IO communication, one or more empty
buffers are placed in the receive queue for incoming data, and
outgoing characters are placed in the transmit queue. Details spec
can be found from the following link.

https://docs.oasis-open.org/virtio/virtio/v1.1/csprd01/virtio-v1.1-csprd01.pdf#e7

Apart from the console, for the communication between guest and host,
the Cloud Hypervisor has a legacy serial device implemented. However,
the implementation of a console device lets us be independent of legacy
pin-based interrupts without losing the logs and access to the VM.

Signed-off-by: A K M Fazla Mehrab <fazla.mehrab.akm@intel.com>
2019-07-22 23:08:56 +01:00
Sebastien Boeuf
f98a69f42e vm-allocator: Introduce an MMIO hole address allocator
With this new AddressAllocator as part of the SystemAllocator, the
VMM can now decide with finer granularity where to place memory.

By allocating the RAM and the hole into the MMIO address space, we
ensure that no memory will be allocated by accident where the RAM or
where the hole is.
And by creating the new MMIO hole address space, we create a subset
of the entire MMIO address space where we can place 32 bits BARs for
example.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:51:16 -07:00
Sebastien Boeuf
a761b820c7 vm-allocator: Fix the aligned address check
The requested address for a range can be the base of the entire
address space, this is a valid use case.

In particular, when creating an MMIO address space of 0-64GiB, we
might want to create a range of 0-1GiB if the RAM of our VM is 1G.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:51:16 -07:00
Sebastien Boeuf
709148803e vm-allocator: Fix free range allocation
This patch fixes the function first_available_range() responsible
for finding the first range that could fit the requested size.

The algorithm was working, that is allocating ranges from the end
of the address space because we created an empty region right at the
end. But the problem is, the VMM might request for some specific
allocations at fixed address to allocate the RAM for example. In this
case, the RAM range could be 0-1GiB, which means with the previous
algorithm, the new available range would have been found right after
1GiB.

This is not the intended behavior, and that's why the algorithm has
been fixed by this patch, making sure to walk down existing ranges
starting from the end.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:51:16 -07:00
Samuel Ortiz
0a04a950a1 vm-allocator: Expand the IRQ allocation API to support GSI
GSI (Global System Interrupt) is an extension of just a linear array of
IRQs. It takes IOAPICs into account for example.

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-22 09:51:16 -07:00
Chao Peng
96fb38a5aa vm-allocator: Align address at allocation time
There is alignment support for AddressAllocator but there are occations
that the alignment is known only when we call allocate(). One example
is PCI BAR which is natually aligned, means for which we have to align
the base address to its size.

Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
2019-07-22 09:51:16 -07:00
Chao Peng
af7cd74e04 vm-allocator: Make port IO non optional
This is only for allocating the port IO address range.
If a platform does not have PIO devices at all, the address
range will simply be unused.
So, simplify the vm-allocator data structure by making both
MMIO and PIO mandatory.

Signed-off-by: Chao Peng <chao.p.peng@linux.intel.com>
2019-07-22 09:51:16 -07:00
Sebastien Boeuf
1268165040 pci: Allow for registering IO and Memory BAR
This patch adds the support for both IO and Memory BARs by expecting
the function allocate_bars() to identify the type of each BAR.
Based on the type, register_mapping() insert the address range on the
appropriate bus (PIO or MMIO).

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:50:10 -07:00
Sebastien Boeuf
b157181656 pci: Fix the way PCI configuration registers are being written
The way the function write_reg() was implemented, it was not keeping
the bits supposed to be read-only whenever the guest was writing to one
of those. That's why this commit takes care of protecting those bits,
preventing them from being updated.

The tricky part is about the BARs since we also need to handle the very
specific case where the BAR is being written with all 1's. In that case
we want to return the size of the BAR instead of its address.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:50:10 -07:00
Sebastien Boeuf
185b1082fb pci: Add a helper to set the BAR type
A BAR can be three different types: IO, 32 bits Memory, or 64 bits
Memory. The VMM needs a way to set the right type depending on its
needs.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:50:10 -07:00
Sebastien Boeuf
ee39e46568 pci: Add MSI capability structure
In order to support use cases that require MSI, the pci crate is
being expanded with the description of an MSI PCI capability
structure through the new MsiCap Rust structure.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:50:10 -07:00
Sebastien Boeuf
72007f016a pci: Improve MSI-X code to let VFIO rely on it
This commit enhances the current msi-x code hosted in the pci crate
in order to be reused by the vfio crate. Specifically, it creates
several useful methods for the MsixCap structure that can simplify
the caller's code.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-22 09:50:10 -07:00
Samuel Ortiz
29878956bd pci: Implement the From trait for the PciCapabilityID structure
This will be needed by the VFIO crate for managing MSI capabilities.

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-22 09:50:10 -07:00
Rob Bradford
3f02ccaa8c qcow: Add support for QCOW v2 header
The QCOW2 format is documented here:
https://git.qemu.org/?p=qemu.git;a=blob;f=docs/interop/qcow2.txt;hb=HEAD

The only difference between v2 and v3 is the addition of some extra
fields into the header in v3 for which there are default values in v2.

This introduces a new unit test for the behaviour but it has been
manually verified by the converting the image from v3 to v2
with a command like:

qemu-img convert -O qcow2 -o compat=0.10 clear-29620-cloud.img clear-29620-cloud.img.v2

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-19 17:21:54 +02:00
Rob Bradford
6f65f3406e build: Ensure caps needed for unit test are set
In some situations it is possible for the setting of the capabilities to
fail due to the variable naming of the build artifacts resulting in the
first parameter to setcap being rejected and thus the whole command
failing.

Use xargs -n 1 to ensure that every potential target independently has
its caps set.

Further it was observed that in some situations the binary produced by
cargo test --all --no-run would not be used and instead a new binary
would be produced when the test was run using the second method. This
again would result in test failures as that binary did not have the
desired capabilities set. Therefore build the test binaries with the
same methodology used to run them.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-19 12:00:10 +02:00
Rob Bradford
998140f1b0 tests: Remove single test limit
Run the tests with default parallelisation.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
492ab7a1a8 build: Use tmpfs for /tmp
On the Jenkins build slaves disk I/O is a bottlneck so make /tmp a tmpfs
which removes I/O issues when running lots of VMs at the same time.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
80f33113cb tests: Use incrementing IP and mac address for VMs
This allows us to test multiple VMs at once.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
93c2099ab6 tests: Abstract guest management under a struct
Create a struct to handle all the details for the guest under test
including details of network and disks.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
eab639efe3 tests: Support customising the cloud-init network details
Allow replacement of the network details used for the VM. By replacing
those from the file checked into the source tree we can continue to use
the file in the tree for manual testing but adjust the network per-VM to
allow parallel testing.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
e9f01740e8 tests: Create cloud-init image from source files in tests
In the future this will provide the basis for the ability to customise
the cloud-init file per VM.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
0776d9d7ae tests: Sleep more in order to speed up tests
By sleeping more earlier this will speed up the tests as the SSH
connection will complete on the first attempt and thus alleviate timeout
and backoff delays.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
7ebfe90985 tests: Use a temporary directory for the temporary test files
Use the tempdir crate to create a temporary directory that is deleted
when the structure goes out of scope.

Use this temporary directory for all temporary test files created by the
tests. The cloud init file is still in /tmp as that is created by the
test wrapper code.

This is the first stage towards being able to run the integration tests
in parallel.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-18 18:01:18 +02:00
Rob Bradford
78fe807284 build: Run unit tests on the Jenkins server
The addition of [workspace] to the top level Cargo.toml is necessary to
have the binaries colocated together.

The Cargo.lock files have also been refreshed by the change to the
Cargo.toml.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
1dfe16c382 vhost_rs: vhost_user: Update unit tests
The Master::connect() function now takes a queue count.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
7499210d0c vm-virtio: net: Remove attributes for test exclusions
Now that the tests are in use this import and function is used.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
af15ce9dc3 vm-virtio: Update test activate() function
The type of interrupt_evt has changed along with the addition of an
msix_config member for the virtio device.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
a50c54671c qcow: Make unit tests pass
Rather than relying on shared memory for a temporary file for QCOW
testing instead use tempfile crate to get a temporary file. The vector
cache tests also need a trivial update after the refactor.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
9a17871630 pci: Make unit tests compile
Another member was added to the configuration struct.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
74d079f7da pci: Mark add_capability test as #[ignore] as it is currently failing
See #105

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
18d52869c5 arch: x86_64: Make unit tests pass
With the adoption for rust-vmm linux-loader crate some small changes
were needed to update the unit tests to reflect this change:

* configure_system now takes an extra parameter
* the e820 entry structure comes from the linux-loader crate

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
224f77500c devices: serial: Make the serial unit tests pass
Some refactoring has taken place since the unit tests were written:

The read/write in BusDevice now take a base address and the interrupt
handling code has changed necessitating the need for a new TestInterrupt
struct.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
9e372a8e48 net_util: Bump pnet dependency
The older version of pnet had a bug which broke some of the behaviour
that the unit tests relied upon.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-16 17:09:05 +02:00
Rob Bradford
cb81f8be5b vmm: Make serial port controllable via command line
Add a "--serial" command line that takes as input either "off", "tty"
(default and current behaviour) and "file=/path/to/file".

When "--serial off" is used the serial device is not added to the VM
configuration at all.

Integration tests added that check for interrupts present (or not) and
that when sending to a file the file contains the expected serial
output.

Signed-off-by: Rob Bradford <robert.bradford@intel.com>
2019-07-11 12:17:58 +01:00
Samuel Ortiz
00df79a530 README: Really fix the memory parameters examples
For real now...

Fixes: #96

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-11 08:49:38 +01:00
Samuel Ortiz
bc742d6a5f main: Add a PCI bridge class integration test
Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-10 09:53:02 +02:00
Samuel Ortiz
2b2c31d206 pci: Use device PCI header type for our root bridge
The Bridge header type is really meant to be for PCI-to-PCI bridges.

Fixes: #85

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-10 09:53:02 +02:00
Samuel Ortiz
d6d9b76c7e README: Fix the memory parameter examples
We need to give it a unit.

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-10 08:35:10 +02:00
Samuel Ortiz
7ed073805d config: Fix default memory size parameter
We need to give it a suffix.

Fixes: #96

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-10 08:35:10 +02:00
Sebastien Boeuf
d9ce29117e vmm: Flag --disk should be optional
Now that cloud-hypervisor VMM supports virtio-pmem, it can directly
boot a VM from an image exposed as a persistent memory block device.

That's why there is no need to force the --disk option as being
mandatory.

Fixes #90

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-09 21:58:02 +02:00
Sebastien Boeuf
f0a76ad424 vmm: Add support for multiple virtio-net devices
Until now, the VMM was only accepting a single instance of virtio-net
device. This commit extends the virtio-net support by allowing several
devices to be created for a single VM.

Fixes #71

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-09 18:55:30 +01:00
Sebastien Boeuf
a2947f9a9f cli: Accept K,M,G suffixes for size parameters
For every parameter dealing with a size as option, such as memory or
virtio-pmem, the CLI can now parse sizes with the suffixes K, M or G.

Fixes #70

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
2019-07-09 15:22:26 +01:00
Jing Liu
2bb0b22cc1 pci: Refine pci topology
PciConfigIo is a legacy pci bus dispatcher, which manages all pci
devices including a pci root bridge. However, it is unnecessary to
design a complex hierarchy which redirects every access by PciRoot.

Since pci root bridge is also a pci device instance, and only contains
easy config space read/write, and PciConfigIo actually acts as a pci bus
to dispatch resource based resolving when VMExit, we re-arrange to make
the pci hierarchy clean.

Signed-off-by: Jing Liu <jing2.liu@linux.intel.com>
2019-07-09 10:01:18 +02:00
Samuel Ortiz
4605ecf1a8 pci: Extend the Device trait to carry the device BARs
When reading from or writing to a PCI BAR to handle a VM exit, we need
to have the BAR address itself to be able to support multiple BARs PCI
devices.

Fixes: #87

Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
2019-07-08 07:39:21 +02:00