2020-04-24 13:00:33

by Colin King

Subject: [PATCH][next] drm/amdgpu: fix unlocks on error return path

From: Colin Ian King <[email protected]>

Currently the error returns paths are unlocking lock kiq->ring_lock
however it seems this should be dev->gfx.kiq.ring_lock as this
is the lock that is being locked and unlocked around the ring
operations. This looks like a bug, fix it by unlocking the
correct lock.

[ Note: untested ]

Addresses-Coverity: ("Missing unlock")
Fixes: 82478876eaac ("drm/amdgpu: protect ring overrun")
Signed-off-by: Colin Ian King <[email protected]>
---
drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 2 +-
drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
index b120f9160f13..edaa50d850a6 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
@@ -430,7 +430,7 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
if (r) {
amdgpu_ring_undo(ring);
- spin_unlock(&kiq->ring_lock);
+ spin_unlock(&adev->gfx.kiq.ring_lock);
return -ETIME;
}

diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
index 0a6026308343..055ecba754ff 100644
--- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
@@ -624,7 +624,7 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
if (r) {
amdgpu_ring_undo(ring);
- spin_unlock(&kiq->ring_lock);
+ spin_unlock(&adev->gfx.kiq.ring_lock);
return -ETIME;
}

--
2.25.1

2020-04-24 14:19:02

by Alex Deucher

Subject: Re: [PATCH][next] drm/amdgpu: fix unlocks on error return path

On Fri, Apr 24, 2020 at 8:56 AM Colin King <[email protected]> wrote:
>
> From: Colin Ian King <[email protected]>
>
> Currently the error returns paths are unlocking lock kiq->ring_lock
> however it seems this should be dev->gfx.kiq.ring_lock as this
> is the lock that is being locked and unlocked around the ring
> operations. This looks like a bug, fix it by unlocking the
> correct lock.
>
> [ Note: untested ]
>
> Addresses-Coverity: ("Missing unlock")
> Fixes: 82478876eaac ("drm/amdgpu: protect ring overrun")
> Signed-off-by: Colin Ian King <[email protected]>

It's the same lock, just accessed via a local pointer. I'll take the
patch and update the commit message when I apply it to avoid confusion
in the future.

Alex

> ---
> drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c | 2 +-
> drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
> index b120f9160f13..edaa50d850a6 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v10_0.c
> @@ -430,7 +430,7 @@ static int gmc_v10_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
> r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
> if (r) {
> amdgpu_ring_undo(ring);
> - spin_unlock(&kiq->ring_lock);
> + spin_unlock(&adev->gfx.kiq.ring_lock);
> return -ETIME;
> }
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> index 0a6026308343..055ecba754ff 100644
> --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c
> @@ -624,7 +624,7 @@ static int gmc_v9_0_flush_gpu_tlb_pasid(struct amdgpu_device *adev,
> r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
> if (r) {
> amdgpu_ring_undo(ring);
> - spin_unlock(&kiq->ring_lock);
> + spin_unlock(&adev->gfx.kiq.ring_lock);
> return -ETIME;
> }
>
> --
> 2.25.1
>
> _______________________________________________
> dri-devel mailing list
> [email protected]
> https://lists.freedesktop.org/mailman/listinfo/dri-devel