Re: [PATCH] PCI: Relax function 0 VPD test and relocate

[Bjorn Helgaas]

On Tue, Sep 15, 2015 at 10:24:46PM -0600, Alex Williamson wrote:
> When we quirk a device with PCI_DEV_FLAGS_VPD_REF_F0 we're expecting
> to find a device where all the functions are identical.  If we don't
> find that, we don't make VPD accessible through pci_vpd_ops.  That
> means that if we quirk devices we shouldn't, we filter them out by
> hiding VPD entirely rather than allowing default access.  Instead, we
> can flip this around to only quirk devices that match a slightly more
> rigorous test in the quirk, allowing regular access for anything else.
> 
> Tests for the multifunction flag are removed since a) function 0 and
> the function under test are clearly a multifunction device if we're
> scanning a non-zero function in the same slot and b) at this point the
> flag is only set in the device under test if the multifunction bit is
> set in the PCI HEADER, which is a point of interpretation for the PCI
> spec.
> 
> Signed-off-by: Alex Williamson <alex.williamson@xxxxxxxxxx>
> ---
> 
> This is potentially another stable candiate since we're continuing to
> iterate on 932c435caba8, but since we don't actually know of a device
> where VPD is blocked (we don't think my Skylake example actually
> supports VPD), I'm not including it.  I would support it if requested
> though.

Applied with acks from Myron & Mark to for-linus for v4.3 with the
changelog below.  I added a stable tag on the theory that there's not
really any benefit in having different flavors of this brand-new code.

    PCI: Use function 0 VPD for identical functions, regular VPD for others
    
    932c435caba8 ("PCI: Add dev_flags bit to access VPD through function 0")
    added PCI_DEV_FLAGS_VPD_REF_F0.  Previously, we set the flag on every
    non-zero function of quirked devices.  If a function turned out to be
    different from function 0, i.e., it had a different class, vendor ID, or
    device ID, the flag remained set but we didn't make VPD accessible at all.
    
    Flip this around so we only set PCI_DEV_FLAGS_VPD_REF_F0 for functions that
    are identical to function 0, and allow regular VPD access for any other
    functions.
    
    [bhelgaas: changelog, stable tag]
    Fixes: 932c435caba8 ("PCI: Add dev_flags bit to access VPD through function 0")
    Signed-off-by: Alex Williamson <alex.williamson@xxxxxxxxxx>
    Signed-off-by: Bjorn Helgaas <helgaas@xxxxxxxxxx>
    Acked-by: Myron Stowe <myron.stowe@xxxxxxxxxx>
    Acked-by: Mark Rustad <mark.d.rustad@xxxxxxxxx>
    CC: stable@xxxxxxxxxxxxxxx

> 
>  drivers/pci/access.c |   22 ----------------------
>  drivers/pci/quirks.c |   20 ++++++++++++++++++--
>  2 files changed, 18 insertions(+), 24 deletions(-)
> 
> diff --git a/drivers/pci/access.c b/drivers/pci/access.c
> index 5a5f0a7..59ac36f 100644
> --- a/drivers/pci/access.c
> +++ b/drivers/pci/access.c
> @@ -475,23 +475,6 @@ static const struct pci_vpd_ops pci_vpd_f0_ops = {
>  	.release = pci_vpd_pci22_release,
>  };
>  
> -static int pci_vpd_f0_dev_check(struct pci_dev *dev)
> -{
> -	struct pci_dev *tdev = pci_get_slot(dev->bus,
> -					    PCI_DEVFN(PCI_SLOT(dev->devfn), 0));
> -	int ret = 0;
> -
> -	if (!tdev)
> -		return -ENODEV;
> -	if (!tdev->vpd || !tdev->multifunction ||
> -	    dev->class != tdev->class || dev->vendor != tdev->vendor ||
> -	    dev->device != tdev->device)
> -		ret = -ENODEV;
> -
> -	pci_dev_put(tdev);
> -	return ret;
> -}
> -
>  int pci_vpd_pci22_init(struct pci_dev *dev)
>  {
>  	struct pci_vpd_pci22 *vpd;
> @@ -500,12 +483,7 @@ int pci_vpd_pci22_init(struct pci_dev *dev)
>  	cap = pci_find_capability(dev, PCI_CAP_ID_VPD);
>  	if (!cap)
>  		return -ENODEV;
> -	if (dev->dev_flags & PCI_DEV_FLAGS_VPD_REF_F0) {
> -		int ret = pci_vpd_f0_dev_check(dev);
>  
> -		if (ret)
> -			return ret;
> -	}
>  	vpd = kzalloc(sizeof(*vpd), GFP_ATOMIC);
>  	if (!vpd)
>  		return -ENOMEM;
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
> index 6a30252..b03373f 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -1907,11 +1907,27 @@ static void quirk_netmos(struct pci_dev *dev)
>  DECLARE_PCI_FIXUP_CLASS_HEADER(PCI_VENDOR_ID_NETMOS, PCI_ANY_ID,
>  			 PCI_CLASS_COMMUNICATION_SERIAL, 8, quirk_netmos);
>  
> +/*
> + * Quirk non-zero PCI functions to route VPD access through function 0 for
> + * devices that share VPD resources between functions.  The functions are
> + * expected to be identical devices.
> + */
>  static void quirk_f0_vpd_link(struct pci_dev *dev)
>  {
> -	if (!dev->multifunction || !PCI_FUNC(dev->devfn))
> +	struct pci_dev *f0;
> +
> +	if (!PCI_FUNC(dev->devfn))
>  		return;
> -	dev->dev_flags |= PCI_DEV_FLAGS_VPD_REF_F0;
> +
> +	f0 = pci_get_slot(dev->bus, PCI_DEVFN(PCI_SLOT(dev->devfn), 0));
> +	if (!f0)
> +		return;
> +
> +	if (f0->vpd && dev->class == f0->class &&
> +	    dev->vendor == f0->vendor && dev->device == f0->device)
> +		dev->dev_flags |= PCI_DEV_FLAGS_VPD_REF_F0;
> +
> +	pci_dev_put(f0);
>  }
>  DECLARE_PCI_FIXUP_CLASS_EARLY(PCI_VENDOR_ID_INTEL, PCI_ANY_ID,
>  			      PCI_CLASS_NETWORK_ETHERNET, 8, quirk_f0_vpd_link);
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-pci" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/