OSDN Git Service

drm/amdgpu: add cache flush workaround to gfx8 emit_fence
authorPierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Thu, 28 Nov 2019 11:08:58 +0000 (12:08 +0100)
committerAlex Deucher <alexander.deucher@amd.com>
Thu, 5 Dec 2019 21:24:58 +0000 (16:24 -0500)
The same workaround is used for gfx7.
Both PAL and Mesa use it for gfx8 too, so port this commit to
gfx_v8_0_ring_emit_fence_gfx.

Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

index 9b0b225..cfc1403 100644 (file)
@@ -6145,7 +6145,23 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
        bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT;
        bool int_sel = flags & AMDGPU_FENCE_FLAG_INT;
 
-       /* EVENT_WRITE_EOP - flush caches, send int */
+       /* Workaround for cache flush problems. First send a dummy EOP
+        * event down the pipe with seq one below.
+        */
+       amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
+       amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
+                                EOP_TC_ACTION_EN |
+                                EOP_TC_WB_ACTION_EN |
+                                EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+                                EVENT_INDEX(5)));
+       amdgpu_ring_write(ring, addr & 0xfffffffc);
+       amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
+                               DATA_SEL(1) | INT_SEL(0));
+       amdgpu_ring_write(ring, lower_32_bits(seq - 1));
+       amdgpu_ring_write(ring, upper_32_bits(seq - 1));
+
+       /* Then send the real EOP event down the pipe:
+        * EVENT_WRITE_EOP - flush caches, send int */
        amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
        amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
                                 EOP_TC_ACTION_EN |
@@ -6887,7 +6903,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
                5 +  /* COND_EXEC */
                7 +  /* PIPELINE_SYNC */
                VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 9 + /* VM_FLUSH */
-               8 +  /* FENCE for VM_FLUSH */
+               12 +  /* FENCE for VM_FLUSH */
                20 + /* GDS switch */
                4 + /* double SWITCH_BUFFER,
                       the first COND_EXEC jump to the place just
@@ -6899,7 +6915,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
                31 + /* DE_META */
                3 + /* CNTX_CTRL */
                5 + /* HDP_INVL */
-               8 + 8 + /* FENCE x2 */
+               12 + 12 + /* FENCE x2 */
                2, /* SWITCH_BUFFER */
        .emit_ib_size = 4, /* gfx_v8_0_ring_emit_ib_gfx */
        .emit_ib = gfx_v8_0_ring_emit_ib_gfx,