[PATCH 1/1] drm/radeon: Initialize fences array entries in radeon_sa_bo_next_hole

[PATCH 1/1] drm/radeon: Initialize fences array entries in radeon_sa_bo_next_hole

[Xiaohui Zhang]

Similar to the handling of amdgpu_sa_bo_next_hole in commit 6a15f3ff19a8
("drm/amdgpu: Initialize fences array entries in amdgpu_sa_bo_next_hole"),
we thought a patch might be needed here as well.

The entries were only initialized once in radeon_sa_bo_new. If a fence
wasn't signalled yet in the first radeon_sa_bo_next_hole call, but then
got signalled before a later radeon_sa_bo_next_hole call, it could
destroy the fence but leave its pointer in the array, resulting in
use-after-free in radeon_sa_bo_new.

Signed-off-by: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>
---
 drivers/gpu/drm/radeon/radeon_sa.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c
index 310c322c7112..0981948bd9ed 100644
--- a/drivers/gpu/drm/radeon/radeon_sa.c
+++ b/drivers/gpu/drm/radeon/radeon_sa.c
@@ -267,6 +267,8 @@ static bool radeon_sa_bo_next_hole(struct radeon_sa_manager *sa_manager,
 	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
 		struct radeon_sa_bo *sa_bo;
 
+		fences[i] = NULL;
+
 		if (list_empty(&sa_manager->flist[i])) {
 			continue;
 		}
@@ -332,10 +334,8 @@ int radeon_sa_bo_new(struct radeon_device *rdev,
 
 	spin_lock(&sa_manager->wq.lock);
 	do {
-		for (i = 0; i < RADEON_NUM_RINGS; ++i) {
-			fences[i] = NULL;
+		for (i = 0; i < RADEON_NUM_RINGS; ++i)
 			tries[i] = 0;
-		}
 
 		do {
 			radeon_sa_bo_try_free(sa_manager);
-- 
2.17.1

Re: [PATCH 1/1] drm/radeon: Initialize fences array entries in radeon_sa_bo_next_hole

[Alex Deucher]

Applied.  Thanks!

Alex

On Tue, Jun 7, 2022 at 12:04 PM Xiaohui Zhang <xiaohuizhang@ruc.edu.cn> wrote:
>
> Similar to the handling of amdgpu_sa_bo_next_hole in commit 6a15f3ff19a8
> ("drm/amdgpu: Initialize fences array entries in amdgpu_sa_bo_next_hole"),
> we thought a patch might be needed here as well.
>
> The entries were only initialized once in radeon_sa_bo_new. If a fence
> wasn't signalled yet in the first radeon_sa_bo_next_hole call, but then
> got signalled before a later radeon_sa_bo_next_hole call, it could
> destroy the fence but leave its pointer in the array, resulting in
> use-after-free in radeon_sa_bo_new.
>
> Signed-off-by: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>
> ---
>  drivers/gpu/drm/radeon/radeon_sa.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c
> index 310c322c7112..0981948bd9ed 100644
> --- a/drivers/gpu/drm/radeon/radeon_sa.c
> +++ b/drivers/gpu/drm/radeon/radeon_sa.c
> @@ -267,6 +267,8 @@ static bool radeon_sa_bo_next_hole(struct radeon_sa_manager *sa_manager,
>         for (i = 0; i < RADEON_NUM_RINGS; ++i) {
>                 struct radeon_sa_bo *sa_bo;
>
> +               fences[i] = NULL;
> +
>                 if (list_empty(&sa_manager->flist[i])) {
>                         continue;
>                 }
> @@ -332,10 +334,8 @@ int radeon_sa_bo_new(struct radeon_device *rdev,
>
>         spin_lock(&sa_manager->wq.lock);
>         do {
> -               for (i = 0; i < RADEON_NUM_RINGS; ++i) {
> -                       fences[i] = NULL;
> +               for (i = 0; i < RADEON_NUM_RINGS; ++i)
>                         tries[i] = 0;
> -               }
>
>                 do {
>                         radeon_sa_bo_try_free(sa_manager);
> --
> 2.17.1
>

Re: [PATCH 1/1] drm/radeon: Initialize fences array entries in radeon_sa_bo_next_hole

[Christian König]

Am 07.06.22 um 17:19 schrieb Xiaohui Zhang:
> Similar to the handling of amdgpu_sa_bo_next_hole in commit 6a15f3ff19a8
> ("drm/amdgpu: Initialize fences array entries in amdgpu_sa_bo_next_hole"),
> we thought a patch might be needed here as well.
>
> The entries were only initialized once in radeon_sa_bo_new. If a fence
> wasn't signalled yet in the first radeon_sa_bo_next_hole call, but then
> got signalled before a later radeon_sa_bo_next_hole call, it could
> destroy the fence but leave its pointer in the array, resulting in
> use-after-free in radeon_sa_bo_new.

I would rather like to see the sub allocator moved into a common drm helper.

Regards,
Christian.

>
> Signed-off-by: Xiaohui Zhang <xiaohuizhang@ruc.edu.cn>
> ---
>   drivers/gpu/drm/radeon/radeon_sa.c | 6 +++---
>   1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/radeon/radeon_sa.c b/drivers/gpu/drm/radeon/radeon_sa.c
> index 310c322c7112..0981948bd9ed 100644
> --- a/drivers/gpu/drm/radeon/radeon_sa.c
> +++ b/drivers/gpu/drm/radeon/radeon_sa.c
> @@ -267,6 +267,8 @@ static bool radeon_sa_bo_next_hole(struct radeon_sa_manager *sa_manager,
>   	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
>   		struct radeon_sa_bo *sa_bo;
>   
> +		fences[i] = NULL;
> +
>   		if (list_empty(&sa_manager->flist[i])) {
>   			continue;
>   		}
> @@ -332,10 +334,8 @@ int radeon_sa_bo_new(struct radeon_device *rdev,
>   
>   	spin_lock(&sa_manager->wq.lock);
>   	do {
> -		for (i = 0; i < RADEON_NUM_RINGS; ++i) {
> -			fences[i] = NULL;
> +		for (i = 0; i < RADEON_NUM_RINGS; ++i)
>   			tries[i] = 0;
> -		}
>   
>   		do {
>   			radeon_sa_bo_try_free(sa_manager);