[PATCH 10/14] drm/amdgpu: Fix interrupt handling in GFX v9.4.3

[Alex Deucher]

From: Lijo Lazar <lijo.lazar at amd.com>

IH follows a different identification scheme for its clients. Get the
right mapping of xcc instance from IH node id.

Signed-off-by: Lijo Lazar <lijo.lazar at amd.com>
Reviewed-by: Le Ma <le.ma at amd.com>
Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 35 +++++++++++++++++++++----
 1 file changed, 30 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index caa0b3e1c877..ca55e522a168 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -2736,11 +2736,24 @@ static int gfx_v9_4_3_set_eop_interrupt_state(struct amdgpu_device *adev,
 	return 0;
 }
 
+static int gfx_v9_4_3_ih_to_xcc_inst(struct amdgpu_device *adev, int ih_node)
+{
+	int xcc;
+
+	xcc = hweight8(adev->gfx.xcc_mask & GENMASK(ih_node / 2, 0));
+	if (!xcc) {
+		dev_err(adev->dev, "Couldn't find xcc mapping from IH node");
+		return -EINVAL;
+	}
+
+	return xcc - 1;
+}
+
 static int gfx_v9_4_3_eop_irq(struct amdgpu_device *adev,
 			    struct amdgpu_irq_src *source,
 			    struct amdgpu_iv_entry *entry)
 {
-	int i, phys_id;
+	int i, xcc_id;
 	u8 me_id, pipe_id, queue_id;
 	struct amdgpu_ring *ring;
 
@@ -2749,14 +2762,19 @@ static int gfx_v9_4_3_eop_irq(struct amdgpu_device *adev,
 	pipe_id = (entry->ring_id & 0x03) >> 0;
 	queue_id = (entry->ring_id & 0x70) >> 4;
 
-	phys_id = node_id_to_phys_map[entry->node_id];
+	xcc_id = gfx_v9_4_3_ih_to_xcc_inst(adev, entry->node_id);
+
+	if (xcc_id == -EINVAL)
+		return -EINVAL;
 
 	switch (me_id) {
 	case 0:
 	case 1:
 	case 2:
 		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-			ring = &adev->gfx.compute_ring[i + phys_id * adev->gfx.num_compute_rings];
+			ring = &adev->gfx.compute_ring
+					[i +
+					 xcc_id * adev->gfx.num_compute_rings];
 			/* Per-queue interrupt is supported for MEC starting from VI.
 			  * The interrupt can only be enabled/disabled per pipe instead of per queue.
 			  */
@@ -2774,18 +2792,25 @@ static void gfx_v9_4_3_fault(struct amdgpu_device *adev,
 {
 	u8 me_id, pipe_id, queue_id;
 	struct amdgpu_ring *ring;
-	int i;
+	int i, xcc_id;
 
 	me_id = (entry->ring_id & 0x0c) >> 2;
 	pipe_id = (entry->ring_id & 0x03) >> 0;
 	queue_id = (entry->ring_id & 0x70) >> 4;
 
+	xcc_id = gfx_v9_4_3_ih_to_xcc_inst(adev, entry->node_id);
+
+	if (xcc_id == -EINVAL)
+		return;
+
 	switch (me_id) {
 	case 0:
 	case 1:
 	case 2:
 		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
-			ring = &adev->gfx.compute_ring[i];
+			ring = &adev->gfx.compute_ring
+					[i +
+					 xcc_id * adev->gfx.num_compute_rings];
 			if (ring->me == me_id && ring->pipe == pipe_id &&
 			    ring->queue == queue_id)
 				drm_sched_fault(&ring->sched);
-- 
2.39.2