[PATCH v3 3/8] drm/amdgpu/gfx11: Fall back to pipe reset if per-queue reset ring test fails

[Jesse Zhang <Jesse.Zhang@amd.com>]

After a per-queue reset via MES, verify that the queue is functional by
performing a ring test. If the test fails, fall back to a full pipe reset
to ensure proper recovery.

This adds a fallback path similar to the one already present when the
initial per-queue reset attempt fails, improving the robustness of KCQ
reset handling on GFX11 hardware.

Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
index e29e8e620699..fbef19ed46f9 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c
@@ -7075,14 +7075,18 @@ static int gfx_v11_0_reset_kcq(struct amdgpu_ring *ring,
 			       struct amdgpu_fence *timedout_fence)
 {
 	struct amdgpu_device *adev = ring->adev;
+	int reset_mode = AMDGPU_RESET_TYPE_PER_QUEUE;
 	int r = 0;
 
 	amdgpu_ring_reset_helper_begin(ring, timedout_fence);
 
 	r = amdgpu_mes_reset_legacy_queue(ring->adev, ring, vmid, true, 0);
+
+pipe_reset:
 	if (r) {
 		dev_warn(adev->dev, "fail(%d) to reset kcq and try pipe reset\n", r);
 
+		reset_mode = AMDGPU_RESET_TYPE_PER_PIPE;
 		amdgpu_amdkfd_suspend(adev, true);
 		r = amdgpu_gfx_mec_pipe_reset_run(adev,
 						     ring->xcc_id, ring->me, ring->pipe,
@@ -7105,6 +7109,13 @@ static int gfx_v11_0_reset_kcq(struct amdgpu_ring *ring,
 		return r;
 	}
 
+	if (reset_mode == AMDGPU_RESET_TYPE_PER_QUEUE) {
+		if (amdgpu_ring_reset_helper_end(ring, timedout_fence))
+			goto pipe_reset;
+		else
+			return 0;
+	}
+
 	return amdgpu_ring_reset_helper_end(ring, timedout_fence);
 }
 
-- 
2.49.0