On Mon, May 06, 2019 at 10:20:05AM -0700, sathyanarayanan.kuppuswamy@xxxxxxxxxxxxxxx wrote:
> From: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@xxxxxxxxxxxxxxx>
>
> If PF does not implement ATS and VF implements/uses it, it might lead to
> runtime issues. Also, as per spec r4.0, sec 9.3.7.8, PF should implement
> ATS if VF implements it. So add additional check to confirm given device
> aligns with the spec.
"might lead to runtime issues" is pretty wishy-washy. I really don't
want to prevent some device from working merely because it has
something in config space that doesn't follow the spec exactly but we
never touch.
> Cc: Ashok Raj <ashok.raj@xxxxxxxxx>
> Cc: Keith Busch <keith.busch@xxxxxxxxx>
> Suggested-by: Ashok Raj <ashok.raj@xxxxxxxxx>
> Reviewed-by: Keith Busch <keith.busch@xxxxxxxxx>
> Signed-off-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@xxxxxxxxxxxxxxx>
> ---
> drivers/pci/ats.c | 12 ++++++++++++
> 1 file changed, 12 insertions(+)
>
> diff --git a/drivers/pci/ats.c b/drivers/pci/ats.c
> index e7a904e347c3..718e6f414680 100644
> --- a/drivers/pci/ats.c
> +++ b/drivers/pci/ats.c
> @@ -19,6 +19,7 @@
> void pci_ats_init(struct pci_dev *dev)
> {
> int pos;
> + struct pci_dev *pdev;
>
> if (pci_ats_disabled())
> return;
> @@ -27,6 +28,17 @@ void pci_ats_init(struct pci_dev *dev)
> if (!pos)
> return;
>
> + /*
> + * Per PCIe r4.0, sec 9.3.7.8, if VF implements Address Translation
> + * Services (ATS) Extended Capability then corresponding PF should
> + * also implement it.
> + */
> + if (dev->is_virtfn) {
> + pdev = pci_physfn(dev);
> + if (!pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_ATS))
> + return;
> + }
> +
> dev->ats_cap = pos;
> }
>
> --
> 2.20.1
>
