kata-containers/cloud-hypervisor
1
0
Fork 0
mirror of https://github.com/cloud-hypervisor/cloud-hypervisor.git synced 2026-08-05 02:19:16 +00:00
Code Issues Packages Projects Releases Wiki Activity

docs: vfio: Remove documentation regarding VFIO limitations

Browse Source 
Since INTx is now supported for VFIO PCI devices, there is no more
limitations regarding VFIO, which means the documentation must be
updated accordingly.

Signed-off-by: Sebastien Boeuf <sebastien.boeuf@intel.com>
This commit is contained in:
Sebastien Boeuf
2021-02-09 15:27:01 +01:00
committed by Rob Bradford
parent 19167e7647
commit 169a6bb83b
1 changed files with 0 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files

66
docs/vfio.md
View File

@@ -78,69 +78,3 @@ takes the device's sysfs path as an argument. In our example it is
The guest kernel will then detect the card reader on its PCI bus and provided The guest kernel will then detect the card reader on its PCI bus and provided
that support for this device is enabled, it will probe and enable it for the that support for this device is enabled, it will probe and enable it for the
guest to use. guest to use.
## Limitations
Cloud-Hypervisor does not implement legacy IRQ for VFIO devices. The choice is
intentional, based on the fact that recent PCI cards should either support MSI
or MSI-X. This prevents from adding extra complexity to the project.
A PCI card works in combination with a driver, meaning the combination of
hardware and software must support either MSI or MSI-X to be compatible with
Cloud-Hypervisor.
### NVIDIA cards
Some NVIDIA graphic cards may support only MSI, therefore one could think they
would work with Cloud-Hypervisor. Unfortunately, because of the implementation
of the NVIDIA proprietary driver (observed on version `460.39`), the driver
will fail to be probed. As shown below, if there is no legacy IRQ support, the
driver will search for MSI-X capability, ignoring a potential MSI support.
```
static int
nv_pci_probe
(
struct pci_dev *pci_dev,
const struct pci_device_id *id_table
)
{
...
if ((pci_dev->irq == 0 && !pci_find_capability(pci_dev, PCI_CAP_ID_MSIX))
&& nv_treat_missing_irq_as_error())
{
nv_printf(NV_DBG_ERRORS, "NVRM: Can't find an IRQ for your NVIDIA card!\n");
nv_printf(NV_DBG_ERRORS, "NVRM: Please check your BIOS settings.\n");
nv_printf(NV_DBG_ERRORS, "NVRM: [Plug & Play OS] should be set to NO\n");
nv_printf(NV_DBG_ERRORS, "NVRM: [Assign IRQ to VGA] should be set to YES \n");
goto failed;
}
```
This means if one tries to use NVIDIA proprietary driver with Cloud-Hypervisor,
the card __MUST__ support MSI-X.
The alternatives to be able to use NVIDIA cards with MSI only support along with
Cloud-Hypervisor are:
- Use the Open Source driver `nouveau` provided by the Linux kernel
- Modify the NVIDIA proprietary driver to allow for MSI support
### Identify PCI capabilities
A quick way to identify if a PCI card supports MSI and/or MSI-X capabilities is
by running `lspci` and by parsing its output. Assuming the card is located at
`01:00.0` in the PCI tree, here is the command one could run:
```
sudo lspci -vvv -s 01:00.0 | grep MSI
```
Generating the following possible output:
```
Capabilities: [68] MSI: Enable- Count=1/1 Maskable- 64bit+
Capabilities: [78] Express (v2) Legacy Endpoint, MSI 00
```