+ Casey
On Wednesday, March 03/07/18, 2018 at 13:17:23 +1100, Alexey Kardashevskiy wrote:
> On 06/03/18 17:05, Arjun Vynipadath wrote:
> > Hi Alexey,
> > Commit 2170dd0("vfio-pci: Mask INTx if a device is not capabable of enabling it") is causing 'Masking broken INTx support' messages everytime we powerup a VM with a SRIOV Virtual Function device attached.
> >
> > PCI SRIOV Virtual Functions aren't allowed to support traditional INTx Pin Interrupts as per Section 5 of the Single Root I/O Virtualization and Sharing Specification, Revision 1.0 (September 11, 2007).
> > Can you make us sure if this is what you intended?
>
> The PCI device does not support INTx, VFIO detects it and masks the
> support, all correct. VFIO does not make distinction between PF and VF,
> hence no special handling for SRIOV. If the message is the only problem, we
> could do something like this, I am not sure there is much value in it though:
>
>
> diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c
> index b0f7594..b181ca6 100644
> --- a/drivers/vfio/pci/vfio_pci.c
> +++ b/drivers/vfio/pci/vfio_pci.c
> @@ -245,7 +245,13 @@ static int vfio_pci_enable(struct vfio_pci_device *vdev)
>
> if (likely(!nointxmask)) {
> if (vfio_pci_nointx(pdev)) {
> - dev_info(&pdev->dev, "Masking broken INTx support\n");
> + u8 pin = 0;
> +
> + pci_read_config_byte(vdev->pdev, PCI_INTERRUPT_PIN,
> + &pin);
> + if (pin)
> + dev_info(&pdev->dev,
> + "Masking broken INTx support\n");
> vdev->nointx = true;
> pci_intx(pdev, 0);
>
>
>
> --
> Alexey
