Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1354558 > unrolled thread

[PATCH 3.13.y-ckt 078/138] drm/radeon: hold reference to fences in radeon_sa_bo_new

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH 3.13.y-ckt 078/138] drm/radeon: hold reference to fences in radeon_sa_bo_new Kamal Mostafa <kamal@canonical.com> - 2016-03-10 00:20 +0100
    Re: [PATCH 3.13.y-ckt 078/138] drm/radeon: hold reference to fences  in radeon_sa_bo_new Nicolai Hähnle <nicolai.haehnle@amd.com> - 2016-03-10 00:50 +0100
      Re: [PATCH 3.13.y-ckt 078/138] drm/radeon: hold reference to fences  in radeon_sa_bo_new Kamal Mostafa <kamal@canonical.com> - 2016-03-10 17:20 +0100

#1354558 — [PATCH 3.13.y-ckt 078/138] drm/radeon: hold reference to fences in radeon_sa_bo_new

3.13.11-ckt36 -stable review patch.  If anyone has any objections, please let me know.

---8<------------------------------------------------------------

From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle@amd.com>

commit f6ff4f67cdf8455d0a4226eeeaf5af17c37d05eb upstream.

An arbitrary amount of time can pass between spin_unlock and
radeon_fence_wait_any, so we need to ensure that nobody frees the
fences from under us.

Based on the analogous fix for amdgpu.

Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
---
 drivers/gpu/drm/radeon/radeon_sa.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c
index f0bac68..bb16684 100644
--- a/drivers/gpu/drm/radeon/radeon_sa.c
+++ b/drivers/gpu/drm/radeon/radeon_sa.c
@@ -349,8 +349,13 @@ int radeon_sa_bo_new(struct radeon_device *rdev,
 			/* see if we can skip over some allocations */
 		} while (radeon_sa_bo_next_hole(sa_manager, fences, tries));
 
+		for (i = 0; i < RADEON_NUM_RINGS; ++i)
+			radeon_fence_ref(fences[i]);
+
 		spin_unlock(&sa_manager->wq.lock);
 		r = radeon_fence_wait_any(rdev, fences, false);
+		for (i = 0; i < RADEON_NUM_RINGS; ++i)
+			radeon_fence_unref(&fences[i]);
 		spin_lock(&sa_manager->wq.lock);
 		/* if we have nothing to wait for block */
 		if (r == -ENOENT && block) {
-- 
2.7.0

[toc] | [next] | [standalone]

#1354684 — Re: [PATCH 3.13.y-ckt 078/138] drm/radeon: hold reference to fences in radeon_sa_bo_new

On 09.03.2016 18:13, Kamal Mostafa wrote:
> 3.13.11-ckt36 -stable review patch.  If anyone has any objections, please let me know.

Please drop the patch for now, it causes a NULL pointer dereference on 
kernels <= 3.17. We will follow up with a correctly backported patch.

Thanks,
Nicolai

>
> ---8<------------------------------------------------------------
>
> From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle@amd.com>
>
> commit f6ff4f67cdf8455d0a4226eeeaf5af17c37d05eb upstream.
>
> An arbitrary amount of time can pass between spin_unlock and
> radeon_fence_wait_any, so we need to ensure that nobody frees the
> fences from under us.
>
> Based on the analogous fix for amdgpu.
>
> Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
> Reviewed-by: Christian König <christian.koenig@amd.com>
> Signed-off-by: Kamal Mostafa <kamal@canonical.com>
> ---
>   drivers/gpu/drm/radeon/radeon_sa.c | 5 +++++
>   1 file changed, 5 insertions(+)
>
> diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c
> index f0bac68..bb16684 100644
> --- a/drivers/gpu/drm/radeon/radeon_sa.c
> +++ b/drivers/gpu/drm/radeon/radeon_sa.c
> @@ -349,8 +349,13 @@ int radeon_sa_bo_new(struct radeon_device *rdev,
>   			/* see if we can skip over some allocations */
>   		} while (radeon_sa_bo_next_hole(sa_manager, fences, tries));
>
> +		for (i = 0; i < RADEON_NUM_RINGS; ++i)
> +			radeon_fence_ref(fences[i]);
> +
>   		spin_unlock(&sa_manager->wq.lock);
>   		r = radeon_fence_wait_any(rdev, fences, false);
> +		for (i = 0; i < RADEON_NUM_RINGS; ++i)
> +			radeon_fence_unref(&fences[i]);
>   		spin_lock(&sa_manager->wq.lock);
>   		/* if we have nothing to wait for block */
>   		if (r == -ENOENT && block) {
>

[toc] | [prev] | [next] | [standalone]

#1355249 — Re: [PATCH 3.13.y-ckt 078/138] drm/radeon: hold reference to fences in radeon_sa_bo_new

On Wed, 2016-03-09 at 18:42 -0500, Nicolai Hähnle wrote:
> On 09.03.2016 18:13, Kamal Mostafa wrote:
> > 
> > 3.13.11-ckt36 -stable review patch.  If anyone has any objections,
> > please let me know.
> Please drop the patch for now, it causes a NULL pointer dereference
> on 
> kernels <= 3.17. We will follow up with a correctly backported patch.
> 
> Thanks,
> Nicolai


Dropped from 3.13-stable.  Thanks for the heads-up, Nicolai!

 -Kamal


> > 
> > 
> > ---8<------------------------------------------------------------
> > 
> > From: =?UTF-8?q?Nicolai=20H=C3=A4hnle?= <nicolai.haehnle@amd.com>
> > 
> > commit f6ff4f67cdf8455d0a4226eeeaf5af17c37d05eb upstream.
> > 
> > An arbitrary amount of time can pass between spin_unlock and
> > radeon_fence_wait_any, so we need to ensure that nobody frees the
> > fences from under us.
> > 
> > Based on the analogous fix for amdgpu.
> > 
> > Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
> > Reviewed-by: Christian König <christian.koenig@amd.com>
> > Signed-off-by: Kamal Mostafa <kamal@canonical.com>
> > ---
> >   drivers/gpu/drm/radeon/radeon_sa.c | 5 +++++
> >   1 file changed, 5 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/radeon/radeon_sa.c
> > b/drivers/gpu/drm/radeon/radeon_sa.c
> > index f0bac68..bb16684 100644
> > --- a/drivers/gpu/drm/radeon/radeon_sa.c
> > +++ b/drivers/gpu/drm/radeon/radeon_sa.c
> > @@ -349,8 +349,13 @@ int radeon_sa_bo_new(struct radeon_device
> > *rdev,
> >   			/* see if we can skip over some
> > allocations */
> >   		} while (radeon_sa_bo_next_hole(sa_manager,
> > fences, tries));
> > 
> > +		for (i = 0; i < RADEON_NUM_RINGS; ++i)
> > +			radeon_fence_ref(fences[i]);
> > +
> >   		spin_unlock(&sa_manager->wq.lock);
> >   		r = radeon_fence_wait_any(rdev, fences, false);
> > +		for (i = 0; i < RADEON_NUM_RINGS; ++i)
> > +			radeon_fence_unref(&fences[i]);
> >   		spin_lock(&sa_manager->wq.lock);
> >   		/* if we have nothing to wait for block */
> >   		if (r == -ENOENT && block) {
> > 

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web