[PATCH resend] pci: ARI is a PCIe v2 feature

[PATCH resend] pci: ARI is a PCIe v2 feature

[Chris Wright]

The function pci_enable_ari() may mistakenly set the downstream port
of a v1 PCIe switch in ARI Forwarding mode.  This is a PCIe v2 feature,
and with an SR-IOV device on that switch port believing the switch above
is ARI capable it may attempt to use functions 8-255, translating into
invalid (non-zero) device numbers for that bus.  This has been seen
to cause Completion Timeouts and general misbehaviour including hangs
and panics.

Cc: stable@kernel.org
Acked-by: Don Dutile <ddutile@redhat.com>
Tested-by: Don Dutile <ddutile@redhat.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
---

 drivers/pci/pci.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 2c5b9b9..7f83236 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -1905,7 +1905,7 @@ void pci_enable_ari(struct pci_dev *dev)
 {
 	int pos;
 	u32 cap;
-	u16 ctrl;
+	u16 flags, ctrl;
 	struct pci_dev *bridge;
 
 	if (!pci_is_pcie(dev) || dev->devfn)
@@ -1923,6 +1923,11 @@ void pci_enable_ari(struct pci_dev *dev)
 	if (!pos)
 		return;
 
+	/* ARI is a PCIe v2 feature */
+	pci_read_config_word(bridge, pos + PCI_EXP_FLAGS, &flags);
+	if ((flags & PCI_EXP_FLAGS_VERS) < 2)
+		return;
+
 	pci_read_config_dword(bridge, pos + PCI_EXP_DEVCAP2, &cap);
 	if (!(cap & PCI_EXP_DEVCAP2_ARI))
 		return;

Re: [PATCH resend] pci: ARI is a PCIe v2 feature

[Jesse Barnes]

On Wed, 13 Jul 2011 10:14:33 -0700
Chris Wright <chrisw@sous-sol.org> wrote:

> The function pci_enable_ari() may mistakenly set the downstream port
> of a v1 PCIe switch in ARI Forwarding mode.  This is a PCIe v2
> feature, and with an SR-IOV device on that switch port believing the
> switch above is ARI capable it may attempt to use functions 8-255,
> translating into invalid (non-zero) device numbers for that bus.
> This has been seen to cause Completion Timeouts and general
> misbehaviour including hangs and panics.


Thanks Chris, your earlier one is still sitting in my queue, but I
wasn't going to send it for 3.0 as it didn't seem that critical.  Do
you think it is?  If so you can send it directly to Linus with my ack
as I don't have anything else queued, or just wait until the merge
window when it will be put into the stable tree as well.

Thanks,
Jesse

Re: [PATCH resend] pci: ARI is a PCIe v2 feature

[Chris Wright]

* Jesse Barnes (jbarnes@virtuousgeek.org) wrote:
> On Wed, 13 Jul 2011 10:14:33 -0700
> Chris Wright <chrisw@sous-sol.org> wrote:
> 
> > The function pci_enable_ari() may mistakenly set the downstream port
> > of a v1 PCIe switch in ARI Forwarding mode.  This is a PCIe v2
> > feature, and with an SR-IOV device on that switch port believing the
> > switch above is ARI capable it may attempt to use functions 8-255,
> > translating into invalid (non-zero) device numbers for that bus.
> > This has been seen to cause Completion Timeouts and general
> > misbehaviour including hangs and panics.
> 
> Thanks Chris, your earlier one is still sitting in my queue, but I
> wasn't going to send it for 3.0 as it didn't seem that critical.  Do
> you think it is?  If so you can send it directly to Linus with my ack
> as I don't have anything else queued, or just wait until the merge
> window when it will be put into the stable tree as well.

Cool, just wanted to make sure it wasn't lost.  I think it's fine to
merge in next window and let it filter back to -stable.  While I've seen
it cause panics, it's not a regression and the conditions required are
very specific (possibly including buggy BIOS).

thanks,
-chris

Re: [PATCH resend] pci: ARI is a PCIe v2 feature

[Jesse Barnes]

On Wed, 13 Jul 2011 10:14:33 -0700
Chris Wright <chrisw@sous-sol.org> wrote:

> The function pci_enable_ari() may mistakenly set the downstream port
> of a v1 PCIe switch in ARI Forwarding mode.  This is a PCIe v2 feature,
> and with an SR-IOV device on that switch port believing the switch above
> is ARI capable it may attempt to use functions 8-255, translating into
> invalid (non-zero) device numbers for that bus.  This has been seen
> to cause Completion Timeouts and general misbehaviour including hangs
> and panics.

Applied, thanks.

-- 
Jesse Barnes, Intel Open Source Technology Center