From: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Date: Tue, 10 Oct 2017 16:50:16 -0400
Subject: drm/amdgpu: Refactor amdgpu_cs_ib_vm_chunk and amdgpu_cs_ib_fill.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Git-commit: ad864d243826cedc53404a1c0db7d1e38ddceb84
Patch-mainline: v4.15-rc1
References: FATE#326289 FATE#326079 FATE#326049 FATE#322398 FATE#326166

This enables old fence waiting before reservation lock is aquired
which in turn is part of a bigger solution to deadlock happening
when gpu reset with VRAM recovery accures during intensive rendering.

Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Petr Tesarik <ptesarik@suse.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c |  112 +++++++++++++++++----------------
 1 file changed, 61 insertions(+), 51 deletions(-)

--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -843,15 +843,60 @@ static int amdgpu_cs_ib_vm_chunk(struct
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct amdgpu_vm *vm = &fpriv->vm;
 	struct amdgpu_ring *ring = p->job->ring;
-	int i, r;
+	int i, j, r;
+
+	for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) {
+
+		struct amdgpu_cs_chunk *chunk;
+		struct amdgpu_ib *ib;
+		struct drm_amdgpu_cs_chunk_ib *chunk_ib;
+
+		chunk = &p->chunks[i];
+		ib = &p->job->ibs[j];
+		chunk_ib = (struct drm_amdgpu_cs_chunk_ib *)chunk->kdata;
+
+		if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB)
+			continue;
+
+		if (p->job->ring->funcs->parse_cs) {
+			struct amdgpu_bo_va_mapping *m;
+			struct amdgpu_bo *aobj = NULL;
+			uint64_t offset;
+			uint8_t *kptr;
+
+			r = amdgpu_cs_find_mapping(p, chunk_ib->va_start,
+					&aobj, &m);
+			if (r) {
+				DRM_ERROR("IB va_start is invalid\n");
+				return r;
+			}
 
-	/* Only for UVD/VCE VM emulation */
-	if (ring->funcs->parse_cs) {
-		for (i = 0; i < p->job->num_ibs; i++) {
-			r = amdgpu_ring_parse_cs(ring, p, i);
+			if ((chunk_ib->va_start + chunk_ib->ib_bytes) >
+				(m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
+				DRM_ERROR("IB va_start+ib_bytes is invalid\n");
+				return -EINVAL;
+			}
+
+			/* the IB should be reserved at this point */
+			r = amdgpu_bo_kmap(aobj, (void **)&kptr);
+			if (r) {
+				return r;
+			}
+
+			offset = m->start * AMDGPU_GPU_PAGE_SIZE;
+			kptr += chunk_ib->va_start - offset;
+
+			memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
+			amdgpu_bo_kunmap(aobj);
+
+			/* Only for UVD/VCE VM emulation */
+			r = amdgpu_ring_parse_cs(ring, p, j);
 			if (r)
 				return r;
+
 		}
+
+		j++;
 	}
 
 	if (p->job->vm) {
@@ -917,54 +962,18 @@ static int amdgpu_cs_ib_fill(struct amdg
 
 		parser->job->ring = ring;
 
-		if (ring->funcs->parse_cs) {
-			struct amdgpu_bo_va_mapping *m;
-			struct amdgpu_bo *aobj = NULL;
-			uint64_t offset;
-			uint8_t *kptr;
-
-			r = amdgpu_cs_find_mapping(parser, chunk_ib->va_start,
-						   &aobj, &m);
-			if (r) {
-				DRM_ERROR("IB va_start is invalid\n");
-				return r;
-			}
-
-			if ((chunk_ib->va_start + chunk_ib->ib_bytes) >
-			    (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) {
-				DRM_ERROR("IB va_start+ib_bytes is invalid\n");
-				return -EINVAL;
-			}
-
-			/* the IB should be reserved at this point */
-			r = amdgpu_bo_kmap(aobj, (void **)&kptr);
-			if (r) {
-				return r;
-			}
-
-			offset = m->start * AMDGPU_GPU_PAGE_SIZE;
-			kptr += chunk_ib->va_start - offset;
-
-			r =  amdgpu_ib_get(adev, vm, chunk_ib->ib_bytes, ib);
-			if (r) {
-				DRM_ERROR("Failed to get ib !\n");
-				return r;
-			}
-
-			memcpy(ib->ptr, kptr, chunk_ib->ib_bytes);
-			amdgpu_bo_kunmap(aobj);
-		} else {
-			r =  amdgpu_ib_get(adev, vm, 0, ib);
-			if (r) {
-				DRM_ERROR("Failed to get ib !\n");
-				return r;
-			}
-
+		r =  amdgpu_ib_get(adev, vm,
+					ring->funcs->parse_cs ? chunk_ib->ib_bytes : 0,
+					ib);
+		if (r) {
+			DRM_ERROR("Failed to get ib !\n");
+			return r;
 		}
 
 		ib->gpu_addr = chunk_ib->va_start;
 		ib->length_dw = chunk_ib->ib_bytes / 4;
 		ib->flags = chunk_ib->flags;
+
 		j++;
 	}
 
@@ -1210,6 +1219,10 @@ int amdgpu_cs_ioctl(struct drm_device *d
 		goto out;
 	}
 
+	r = amdgpu_cs_ib_fill(adev, &parser);
+	if (r)
+		goto out;
+
 	r = amdgpu_cs_parser_bos(&parser, data);
 	if (r) {
 		if (r == -ENOMEM)
@@ -1220,9 +1233,6 @@ int amdgpu_cs_ioctl(struct drm_device *d
 	}
 
 	reserved_buffers = true;
-	r = amdgpu_cs_ib_fill(adev, &parser);
-	if (r)
-		goto out;
 
 	r = amdgpu_cs_dependencies(adev, &parser);
 	if (r) {