[PATCH Reivew 1/1] drm/amdgpu: fix use-after-free during gpu recovery

[PATCH Reivew 1/1] drm/amdgpu: fix use-after-free during gpu recovery

[Stanley.Yang]

[Why]
    [  754.862560] refcount_t: underflow; use-after-free.
    [  754.862898] Call Trace:
    [  754.862903]  <TASK>
    [  754.862913]  amdgpu_job_free_cb+0xc2/0xe1 [amdgpu]
    [  754.863543]  drm_sched_main.cold+0x34/0x39 [amd_sched]

[How]
    The fw_fence may be not init, check whether dma_fence_init
    is performed before job free

Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 8771df97d590..ddee6a6b133d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -169,7 +169,11 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
 	amdgpu_sync_free(&job->sync);
 	amdgpu_sync_free(&job->sched_sync);
 
-	dma_fence_put(&job->hw_fence);
+	/* only put the hw fence if has embedded fence */
+	if (!job->hw_fence.ops)
+		kfree(job);
+	else
+		dma_fence_put(&job->hw_fence);
 }
 
 void amdgpu_job_set_gang_leader(struct amdgpu_job *job,
-- 
2.17.1

RE: [PATCH Reivew 1/1] drm/amdgpu: fix use-after-free during gpu recovery

[Zhou1, Tao]

[AMD Official Use Only - General]

Reviewed-by: Tao Zhou <tao.zhou1@amd.com>

> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of
> Stanley.Yang
> Sent: Thursday, November 17, 2022 11:01 AM
> To: amd-gfx@lists.freedesktop.org
> Cc: Wang, YuBiao <YuBiao.Wang@amd.com>; andrey.grodzovsky@amd.com;
> Yang, Stanley <Stanley.Yang@amd.com>
> Subject: [PATCH Reivew 1/1] drm/amdgpu: fix use-after-free during gpu
> recovery
> 
> [Why]
>     [  754.862560] refcount_t: underflow; use-after-free.
>     [  754.862898] Call Trace:
>     [  754.862903]  <TASK>
>     [  754.862913]  amdgpu_job_free_cb+0xc2/0xe1 [amdgpu]
>     [  754.863543]  drm_sched_main.cold+0x34/0x39 [amd_sched]
> 
> [How]
>     The fw_fence may be not init, check whether dma_fence_init
>     is performed before job free
> 
> Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> index 8771df97d590..ddee6a6b133d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> @@ -169,7 +169,11 @@ static void amdgpu_job_free_cb(struct drm_sched_job
> *s_job)
>  	amdgpu_sync_free(&job->sync);
>  	amdgpu_sync_free(&job->sched_sync);
> 
> -	dma_fence_put(&job->hw_fence);
> +	/* only put the hw fence if has embedded fence */
> +	if (!job->hw_fence.ops)
> +		kfree(job);
> +	else
> +		dma_fence_put(&job->hw_fence);
>  }
> 
>  void amdgpu_job_set_gang_leader(struct amdgpu_job *job,
> --
> 2.17.1