drm/amd/amdgpu: fix corner case in SRIOV tdr

[Why] In SRIOV multi-vf, after using ordered workqueue for tdr, there could be a chance that a ring timeout continuously makes an innocent ring timeout. [How] 1. Use advance tdr mode in SRIOV as default 2. Use mdelay in flr work to make sure the waiting won't exceeds ring timeout. Signed-off-by: Jingwen Chen <Jingwen.Chen2@amd.com> Acked-by: Alex Deucher <alexander.deucher@amd.com>
SamuelMarks · Sep 29, 2021 · 4ff45ec · 4ff45ec
1 parent 87e71bc
commit 4ff45ec
Showing 3 changed files with 6 additions and 2 deletions.
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c
@@ -63,6 +63,10 @@ void amdgpu_virt_init_setting(struct amdgpu_device *adev)
 #endif
 	adev->cg_flags = 0;
 	adev->pg_flags = 0;
+
+	/*use advance recovery mode for SRIOV*/
+	if (amdgpu_gpu_recovery)
+		amdgpu_gpu_recovery = 2;
 }
 
 void amdgpu_virt_kiq_reg_write_reg_wait(struct amdgpu_device *adev,

diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c
@@ -265,7 +265,7 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work)
 		if (xgpu_ai_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL)
 			goto flr_done;
 
-		msleep(10);
+		mdelay(10);
 		timeout -= 10;
 	} while (timeout > 1);
 

diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_nv.c
@@ -294,7 +294,7 @@ static void xgpu_nv_mailbox_flr_work(struct work_struct *work)
 		if (xgpu_nv_mailbox_peek_msg(adev) == IDH_FLR_NOTIFICATION_CMPL)
 			goto flr_done;
 
-		msleep(10);
+		mdelay(10);
 		timeout -= 10;
 	} while (timeout > 1);