[PATCH AUTOSEL 6.6 09/40] drm/amdgpu: not to save bo in the case of RAS err_event_athub

Sasha Levin sashal at kernel.org
Tue Nov 7 12:16:11 UTC 2023


From: "David (Ming Qiang) Wu" <David.Wu3 at amd.com>

[ Upstream commit fa1f1cc09d588a90c8ce3f507c47df257461d148 ]

err_event_athub will corrupt VCPU buffer and not good to
be restored in amdgpu_vcn_resume() and in this case
the VCPU buffer needs to be cleared for VCN firmware to
work properly.

Acked-by: Leo Liu <leo.liu at amd.com>
Signed-off-by: David (Ming Qiang) Wu <David.Wu3 at amd.com>
Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
Signed-off-by: Sasha Levin <sashal at kernel.org>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
index 36b55d2bd51a9..03b4bcfca1963 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c
@@ -292,8 +292,15 @@ int amdgpu_vcn_suspend(struct amdgpu_device *adev)
 	void *ptr;
 	int i, idx;
 
+	bool in_ras_intr = amdgpu_ras_intr_triggered();
+
 	cancel_delayed_work_sync(&adev->vcn.idle_work);
 
+	/* err_event_athub will corrupt VCPU buffer, so we need to
+	 * restore fw data and clear buffer in amdgpu_vcn_resume() */
+	if (in_ras_intr)
+		return 0;
+
 	for (i = 0; i < adev->vcn.num_vcn_inst; ++i) {
 		if (adev->vcn.harvest_config & (1 << i))
 			continue;
-- 
2.42.0



More information about the amd-gfx mailing list