[Intel-xe] [PATCH 3/3] drm/xe: Use proper vram offset

[Niranjana Vishwanathapura]

In xe_migrate functions, use proper vram io offset of the
tiles while calculating addresses.

Reviewed-by: Matthew Brost <matthew.brost at intel.com>
Signed-off-by: Niranjana Vishwanathapura <niranjana.vishwanathapura at intel.com>
---
 drivers/gpu/drm/xe/xe_bo.c      | 12 ++++++------
 drivers/gpu/drm/xe/xe_bo.h      |  1 +
 drivers/gpu/drm/xe/xe_migrate.c | 13 ++++++++-----
 3 files changed, 15 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_bo.c b/drivers/gpu/drm/xe/xe_bo.c
index e4d079b61d52..5460e6fe3c1f 100644
--- a/drivers/gpu/drm/xe/xe_bo.c
+++ b/drivers/gpu/drm/xe/xe_bo.c
@@ -1207,12 +1207,12 @@ struct xe_bo *xe_bo_create_from_data(struct xe_device *xe, struct xe_gt *gt,
  * XXX: This is in the VM bind data path, likely should calculate this once and
  * store, with a recalculation if the BO is moved.
  */
-static uint64_t vram_region_io_offset(struct xe_bo *bo)
+uint64_t vram_region_io_offset(struct ttm_resource *res)
 {
-	struct xe_device *xe = xe_bo_device(bo);
-	struct xe_gt *gt = mem_type_to_gt(xe, bo->ttm.resource->mem_type);
+	struct xe_device *xe = ttm_to_xe_device(res->bo->bdev);
+	struct xe_gt *gt = mem_type_to_gt(xe, res->mem_type);
 
-	if (bo->ttm.resource->mem_type == XE_PL_STOLEN)
+	if (res->mem_type == XE_PL_STOLEN)
 		return xe_ttm_stolen_gpu_offset(xe);
 
 	return gt->mem.vram.io_start - xe->mem.vram.io_start;
@@ -1299,7 +1299,7 @@ int xe_bo_pin(struct xe_bo *bo)
 			XE_BUG_ON(!(place->flags & TTM_PL_FLAG_CONTIGUOUS));
 
 			place->fpfn = (xe_bo_addr(bo, 0, PAGE_SIZE, &vram) -
-				       vram_region_io_offset(bo)) >> PAGE_SHIFT;
+				       vram_region_io_offset(bo->ttm.resource)) >> PAGE_SHIFT;
 			place->lpfn = place->fpfn + (bo->size >> PAGE_SHIFT);
 
 			spin_lock(&xe->pinned.lock);
@@ -1443,7 +1443,7 @@ dma_addr_t xe_bo_addr(struct xe_bo *bo, u64 offset,
 
 		xe_res_first(bo->ttm.resource, page << PAGE_SHIFT,
 			     page_size, &cur);
-		return cur.start + offset + vram_region_io_offset(bo);
+		return cur.start + offset + vram_region_io_offset(bo->ttm.resource);
 	}
 }
 
diff --git a/drivers/gpu/drm/xe/xe_bo.h b/drivers/gpu/drm/xe/xe_bo.h
index 9b26049521de..8f5a7ad10d09 100644
--- a/drivers/gpu/drm/xe/xe_bo.h
+++ b/drivers/gpu/drm/xe/xe_bo.h
@@ -223,6 +223,7 @@ void xe_bo_vunmap(struct xe_bo *bo);
 bool mem_type_is_vram(u32 mem_type);
 bool xe_bo_is_vram(struct xe_bo *bo);
 bool xe_bo_is_stolen(struct xe_bo *bo);
+uint64_t vram_region_io_offset(struct ttm_resource *res);
 
 bool xe_bo_can_migrate(struct xe_bo *bo, u32 mem_type);
 
diff --git a/drivers/gpu/drm/xe/xe_migrate.c b/drivers/gpu/drm/xe/xe_migrate.c
index deb4233b94fa..e8978440c725 100644
--- a/drivers/gpu/drm/xe/xe_migrate.c
+++ b/drivers/gpu/drm/xe/xe_migrate.c
@@ -392,6 +392,7 @@ static u64 xe_migrate_res_sizes(struct xe_res_cursor *cur)
 
 static u32 pte_update_size(struct xe_migrate *m,
 			   bool is_vram,
+			   struct ttm_resource *res,
 			   struct xe_res_cursor *cur,
 			   u64 *L0, u64 *L0_ofs, u32 *L0_pt,
 			   u32 cmd_size, u32 pt_ofs, u32 avail_pts)
@@ -417,7 +418,8 @@ static u32 pte_update_size(struct xe_migrate *m,
 		cmds += cmd_size;
 	} else {
 		/* Offset into identity map. */
-		*L0_ofs = xe_migrate_vram_ofs(cur->start);
+		*L0_ofs = xe_migrate_vram_ofs(cur->start +
+					      vram_region_io_offset(res));
 		cmds += cmd_size;
 	}
 
@@ -467,6 +469,7 @@ static void emit_pte(struct xe_migrate *m,
 					addr |= GEN12_PTE_PS64;
 				}
 
+				addr += vram_region_io_offset(bo->ttm.resource);
 				addr |= GEN12_PPGTT_PTE_LM;
 			}
 			addr |= PPAT_CACHED | GEN8_PAGE_PRESENT | GEN8_PAGE_RW;
@@ -646,17 +649,17 @@ struct dma_fence *xe_migrate_copy(struct xe_migrate *m,
 
 		src_L0 = min(src_L0, dst_L0);
 
-		batch_size += pte_update_size(m, src_is_vram, &src_it, &src_L0,
+		batch_size += pte_update_size(m, src_is_vram, src, &src_it, &src_L0,
 					      &src_L0_ofs, &src_L0_pt, 0, 0,
 					      NUM_PT_PER_BLIT);
 
-		batch_size += pte_update_size(m, dst_is_vram, &dst_it, &src_L0,
+		batch_size += pte_update_size(m, dst_is_vram, dst, &dst_it, &src_L0,
 					      &dst_L0_ofs, &dst_L0_pt, 0,
 					      NUM_PT_PER_BLIT, NUM_PT_PER_BLIT);
 
 		if (copy_system_ccs) {
 			ccs_size = xe_device_ccs_bytes(xe, src_L0);
-			batch_size += pte_update_size(m, false, &ccs_it, &ccs_size,
+			batch_size += pte_update_size(m, false, NULL, &ccs_it, &ccs_size,
 						      &ccs_ofs, &ccs_pt, 0,
 						      2 * NUM_PT_PER_BLIT,
 						      NUM_PT_PER_BLIT);
@@ -879,7 +882,7 @@ struct dma_fence *xe_migrate_clear(struct xe_migrate *m,
 
 		/* Calculate final sizes and batch size.. */
 		batch_size = 2 +
-			pte_update_size(m, clear_vram, &src_it,
+			pte_update_size(m, clear_vram, src, &src_it,
 					&clear_L0, &clear_L0_ofs, &clear_L0_pt,
 					emit_clear_cmd_len(xe), 0,
 					NUM_PT_PER_BLIT);
-- 
2.21.0.rc0.32.g243a4c7e27