Skip to contentSkip to conversation

drm/amdgpu: SRIOV flr_work should take write_lock

PATCH In review Jingwen Chen 2 participants 2 messages Jul 1, 2021 – Jul 6, 2021 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu: SRIOV flr_work should take write_lock

[Why]
If flr_work takes read_lock, then other threads who takes read_lock can access hardware when host is doing vf flr.

[How]
flr_work should take write_lock to avoid this case.

---
 drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c | 4 ++--
 drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +2 -2
index 3ee481557fc9..ff2307d7ee0f 100644
@@ -252,7 +252,7 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work)
 	 * otherwise the mailbox msg will be ruined/reseted by
 	 * the VF FLR.
 	 */
-	if (!down_read_trylock(&adev->reset_sem))
+	if (!down_write_trylock(&adev->reset_sem))
 		return;
 
 	amdgpu_virt_fini_data_exchange(adev);
@@ -268,7 +268,7 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work)
 
 flr_done:
 	atomic_set(&adev->in_gpu_reset, 0);
-	up_read(&adev->reset_sem);
+	up_write(&adev->reset_sem);
 
 	/* Trigger recovery for world switch failure if no TDR */
 	if (amdgpu_device_should_recover_gpu(adev)
drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c +2 -2
index 48e588d3c409..9f7aac435d69 100644
@@ -273,7 +273,7 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work)
 	 * otherwise the mailbox msg will be ruined/reseted by
 	 * the VF FLR.
 	 */
-	if (!down_read_trylock(&adev->reset_sem))
+	if (!down_write_trylock(&adev->reset_sem))
 		return;
 
 	amdgpu_virt_fini_data_exchange(adev);
@@ -289,7 +289,7 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work)
 
 flr_done:
 	atomic_set(&adev->in_gpu_reset, 0);
-	up_read(&adev->reset_sem);
+	up_write(&adev->reset_sem);
 
 	/* Trigger recovery for world switch failure if no TDR */
 	if (amdgpu_device_should_recover_gpu(adev)
Signature · 6 lines
2.25.1

_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

A message that is not in the archive (sent off-list or not yet received): 20210701101233.wlskqc2kjzwhiudv@wayne-build

Reviewed-by
#2 lore raw

[AMD Official Use Only]

Thanks

------------------------------------------
Monk Liu | Cloud-GPU Core team
------------------------------------------

-----Original Message-----
From: Jingwen Chen <[email protected]>
Sent: Thursday, July 1, 2021 6:13 PM
To: [email protected]
Cc: Liu, Monk <[email protected]>; Chen, Horace <[email protected]>
Subject: Re: [PATCH] drm/amdgpu: SRIOV flr_work should take write_lock

ping..

Quoted 60 lines · Jingwen Chen

[Why]
If flr_work takes read_lock, then other threads who takes read_lock can access hardware when host is doing vf flr.

[How]
flr_work should take write_lock to avoid this case.

---
 drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c | 4 ++--  
drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +2 -2
index 3ee481557fc9..ff2307d7ee0f 100644
@@ -252,7 +252,7 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work)
 	 * otherwise the mailbox msg will be ruined/reseted by
 	 * the VF FLR.
 	 */
-	if (!down_read_trylock(&adev->reset_sem))
+	if (!down_write_trylock(&adev->reset_sem))
 		return;
 
 	amdgpu_virt_fini_data_exchange(adev);
@@ -268,7 +268,7 @@ static void xgpu_ai_mailbox_flr_work(struct 
 work_struct *work)
 
 flr_done:
 	atomic_set(&adev->in_gpu_reset, 0);
-	up_read(&adev->reset_sem);
+	up_write(&adev->reset_sem);
 
 	/* Trigger recovery for world switch failure if no TDR */
drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +0 -0
 	if (amdgpu_device_should_recover_gpu(adev)
drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c +2 -2
index 48e588d3c409..9f7aac435d69 100644
@@ -273,7 +273,7 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work)
 	 * otherwise the mailbox msg will be ruined/reseted by
 	 * the VF FLR.
 	 */
-	if (!down_read_trylock(&adev->reset_sem))
+	if (!down_write_trylock(&adev->reset_sem))
 		return;
 
 	amdgpu_virt_fini_data_exchange(adev);
@@ -289,7 +289,7 @@ static void xgpu_nv_mailbox_flr_work(struct 
 work_struct *work)
 
 flr_done:
 	atomic_set(&adev->in_gpu_reset, 0);
-	up_read(&adev->reset_sem);
+	up_write(&adev->reset_sem);
 
 	/* Trigger recovery for world switch failure if no TDR */
Quoted 4 lines

if (amdgpu_device_should_recover_gpu(adev)
--
2.25.1

Best Regards,
JingWen Chen
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx