2014-06-12 15:28:19

by Alex Williamson

Subject: [PATCH] pci: Fix unaligned access testing AF transaction pending

pci_wait_for_pending() uses word access, so we shouldn't be passing
an offset that is only byte aligned. Use the control register offset
instead, shifting the mask to match.

Signed-off-by: Alex Williamson <[email protected]>
Reported-by: Ben Hutchings <[email protected]>
Cc: [email protected]
---
drivers/pci/pci.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 759475e..81012ec 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -3109,8 +3109,13 @@ static int pci_af_flr(struct pci_dev *dev, int probe)
if (probe)
return 0;

- /* Wait for Transaction Pending bit clean */
- if (pci_wait_for_pending(dev, pos + PCI_AF_STATUS, PCI_AF_STATUS_TP))
+ /*
+ * Wait for Transaction Pending bit to clear. A word-aligned test
+ * is used, so we use the conrol offset rather than status and shift
+ * the test bit to match.
+ */
+ if (pci_wait_for_pending(dev, pos + PCI_AF_CTRL,
+ PCI_AF_STATUS_TP << 8))
goto clear;

dev_err(&dev->dev, "transaction is not cleared; "

2014-06-12 23:33:47

by Gavin Shan

Subject: Re: [PATCH] pci: Fix unaligned access testing AF transaction pending

On Thu, Jun 12, 2014 at 09:28:07AM -0600, Alex Williamson wrote:
>pci_wait_for_pending() uses word access, so we shouldn't be passing
>an offset that is only byte aligned. Use the control register offset
>instead, shifting the mask to match.
>
>Signed-off-by: Alex Williamson <[email protected]>
>Reported-by: Ben Hutchings <[email protected]>
>Cc: [email protected]

Acked-by: Gavin Shan <[email protected]>

>---
> drivers/pci/pci.c | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
>diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
>index 759475e..81012ec 100644
>--- a/drivers/pci/pci.c
>+++ b/drivers/pci/pci.c
>@@ -3109,8 +3109,13 @@ static int pci_af_flr(struct pci_dev *dev, int probe)
> if (probe)
> return 0;
>
>- /* Wait for Transaction Pending bit clean */
>- if (pci_wait_for_pending(dev, pos + PCI_AF_STATUS, PCI_AF_STATUS_TP))
>+ /*
>+ * Wait for Transaction Pending bit to clear. A word-aligned test
>+ * is used, so we use the conrol offset rather than status and shift
>+ * the test bit to match.
>+ */
>+ if (pci_wait_for_pending(dev, pos + PCI_AF_CTRL,
>+ PCI_AF_STATUS_TP << 8))
> goto clear;
>
> dev_err(&dev->dev, "transaction is not cleared; "
>

Thanks,
Gavin

2014-06-17 21:47:16

by Bjorn Helgaas

Subject: Re: [PATCH] pci: Fix unaligned access testing AF transaction pending

On Thu, Jun 12, 2014 at 09:28:07AM -0600, Alex Williamson wrote:
> pci_wait_for_pending() uses word access, so we shouldn't be passing
> an offset that is only byte aligned. Use the control register offset
> instead, shifting the mask to match.
>
> Signed-off-by: Alex Williamson <[email protected]>
> Reported-by: Ben Hutchings <[email protected]>
> Cc: [email protected]

Applied with Gavin's ack to for-linus for v3.16, thanks!

> ---
> drivers/pci/pci.c | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 759475e..81012ec 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -3109,8 +3109,13 @@ static int pci_af_flr(struct pci_dev *dev, int probe)
> if (probe)
> return 0;
>
> - /* Wait for Transaction Pending bit clean */
> - if (pci_wait_for_pending(dev, pos + PCI_AF_STATUS, PCI_AF_STATUS_TP))
> + /*
> + * Wait for Transaction Pending bit to clear. A word-aligned test
> + * is used, so we use the conrol offset rather than status and shift
> + * the test bit to match.
> + */
> + if (pci_wait_for_pending(dev, pos + PCI_AF_CTRL,
> + PCI_AF_STATUS_TP << 8))
> goto clear;
>
> dev_err(&dev->dev, "transaction is not cleared; "
>