2015-04-16 14:42:25

by Oded Gabbay

Subject: [PATCH] iommu/amd: Fix bug in put_pasid_state_wait

This patch fixes a bug in put_pasid_state_wait that appeared in kernel 4.0
The bug is that pasid_state->count wasn't decremented before entering the
wait_event. Thus, the condition in wait_event will never be true.

The fix is to decrement (atomically) the pasid_state->count before the
wait_event.

Signed-off-by: Oded Gabbay <[email protected]>
Cc: [email protected]
---
drivers/iommu/amd_iommu_v2.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/drivers/iommu/amd_iommu_v2.c b/drivers/iommu/amd_iommu_v2.c
index 6d5a5c4..173e70d 100644
--- a/drivers/iommu/amd_iommu_v2.c
+++ b/drivers/iommu/amd_iommu_v2.c
@@ -266,6 +266,7 @@ static void put_pasid_state(struct pasid_state *pasid_state)

static void put_pasid_state_wait(struct pasid_state *pasid_state)
{
+ atomic_dec(&pasid_state->count);
wait_event(pasid_state->wq, !atomic_read(&pasid_state->count));
free_pasid_state(pasid_state);
}
--
1.9.1

2015-04-26 09:15:09

by Oded Gabbay

Subject: Re: [PATCH] iommu/amd: Fix bug in put_pasid_state_wait

Hi Joerg,
Could you please take a look at this patch ?
I want it to be included in stable asap so Fedora 22 will pick it up.

Thanks,
Oded

On 04/16/2015 05:08 PM, Oded Gabbay wrote:
> This patch fixes a bug in put_pasid_state_wait that appeared in kernel 4.0
> The bug is that pasid_state->count wasn't decremented before entering the
> wait_event. Thus, the condition in wait_event will never be true.
>
> The fix is to decrement (atomically) the pasid_state->count before the
> wait_event.
>
> Signed-off-by: Oded Gabbay <[email protected]>
> Cc: [email protected]
> ---
> drivers/iommu/amd_iommu_v2.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/iommu/amd_iommu_v2.c b/drivers/iommu/amd_iommu_v2.c
> index 6d5a5c4..173e70d 100644
> --- a/drivers/iommu/amd_iommu_v2.c
> +++ b/drivers/iommu/amd_iommu_v2.c
> @@ -266,6 +266,7 @@ static void put_pasid_state(struct pasid_state *pasid_state)
>
> static void put_pasid_state_wait(struct pasid_state *pasid_state)
> {
> + atomic_dec(&pasid_state->count);
> wait_event(pasid_state->wq, !atomic_read(&pasid_state->count));
> free_pasid_state(pasid_state);
> }
>

2015-05-04 11:22:05

by Joerg Roedel

Subject: Re: [PATCH] iommu/amd: Fix bug in put_pasid_state_wait

On Thu, Apr 16, 2015 at 05:08:44PM +0300, Oded Gabbay wrote:
> This patch fixes a bug in put_pasid_state_wait that appeared in kernel 4.0
> The bug is that pasid_state->count wasn't decremented before entering the
> wait_event. Thus, the condition in wait_event will never be true.
>
> The fix is to decrement (atomically) the pasid_state->count before the
> wait_event.

Gah! This issue was already pointed out in the original review of the
patch, looks like I applied the wrong version. Thanks for the fix, I
applied it to iommu/fixes.


Joerg