From a3b3bef33519fa8edba60849dea29d93a503ebc4 Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Mon, 21 Sep 2020 12:56:48 +0200
Subject: drm/vmwgfx: remove unused placement combination
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Just some dead code cleanup.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Link: https://patchwork.freedesktop.org/patch/391599/?series=81973&rev=1
---
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 30 ------------------------------
 1 file changed, 30 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 7454f797d37b..d7ea658e9910 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -65,13 +65,6 @@ static const struct ttm_place gmr_placement_flags = {
 	.flags = TTM_PL_FLAG_CACHED
 };
 
-static const struct ttm_place gmr_ne_placement_flags = {
-	.fpfn = 0,
-	.lpfn = 0,
-	.mem_type = VMW_PL_GMR,
-	.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_NO_EVICT
-};
-
 static const struct ttm_place mob_placement_flags = {
 	.fpfn = 0,
 	.lpfn = 0,
@@ -128,29 +121,6 @@ struct ttm_placement vmw_vram_gmr_placement = {
 	.busy_placement = &gmr_placement_flags
 };
 
-static const struct ttm_place vram_gmr_ne_placement_flags[] = {
-	{
-		.fpfn = 0,
-		.lpfn = 0,
-		.mem_type = TTM_PL_VRAM,
-		.flags = TTM_PL_FLAG_CACHED |
-			 TTM_PL_FLAG_NO_EVICT
-	}, {
-		.fpfn = 0,
-		.lpfn = 0,
-		.mem_type = VMW_PL_GMR,
-		.flags = TTM_PL_FLAG_CACHED |
-			 TTM_PL_FLAG_NO_EVICT
-	}
-};
-
-struct ttm_placement vmw_vram_gmr_ne_placement = {
-	.num_placement = 2,
-	.placement = vram_gmr_ne_placement_flags,
-	.num_busy_placement = 1,
-	.busy_placement = &gmr_ne_placement_flags
-};
-
 struct ttm_placement vmw_vram_sys_placement = {
 	.num_placement = 1,
 	.placement = &vram_placement_flags,
-- 
cgit 


From b254557cb244e2c18e59ee1cc2293128c52d2473 Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Mon, 21 Sep 2020 14:14:32 +0200
Subject: drm/vmwgfx: stop using ttm_bo_create v2
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Implement in the driver instead since it is the only user of that function.

v2: fix usage of ttm_bo_init_reserved

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Link: https://patchwork.freedesktop.org/patch/391614/?series=81973&rev=1
---
 drivers/gpu/drm/vmwgfx/vmwgfx_bo.c         | 43 ++++++++++++++++++++++++++++++
 drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c     |  6 ++---
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.h        |  4 +++
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  8 +++---
 4 files changed, 53 insertions(+), 8 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
index 813f1b148094..c8ca09f0e627 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
@@ -487,6 +487,49 @@ static void vmw_user_bo_destroy(struct ttm_buffer_object *bo)
 	ttm_prime_object_kfree(vmw_user_bo, prime);
 }
 
+/**
+ * vmw_bo_create_kernel - Create a pinned BO for internal kernel use.
+ *
+ * @dev_priv: Pointer to the device private struct
+ * @size: size of the BO we need
+ * @placement: where to put it
+ * @p_bo: resulting BO
+ *
+ * Creates and pin a simple BO for in kernel use.
+ */
+int vmw_bo_create_kernel(struct vmw_private *dev_priv, unsigned long size,
+			 struct ttm_placement *placement,
+			 struct ttm_buffer_object **p_bo)
+{
+	unsigned npages = PAGE_ALIGN(size) >> PAGE_SHIFT;
+	struct ttm_operation_ctx ctx = { false, false };
+	struct ttm_buffer_object *bo;
+	size_t acc_size;
+	int ret;
+
+	bo = kzalloc(sizeof(*bo), GFP_KERNEL);
+	if (unlikely(!bo))
+		return -ENOMEM;
+
+	acc_size = ttm_round_pot(sizeof(*bo));
+	acc_size += ttm_round_pot(npages * sizeof(void *));
+	acc_size += ttm_round_pot(sizeof(struct ttm_tt));
+	ret = ttm_bo_init_reserved(&dev_priv->bdev, bo, size,
+				   ttm_bo_type_device, placement, 0,
+				   &ctx, acc_size, NULL, NULL, NULL);
+	if (unlikely(ret))
+		goto error_free;
+
+	ttm_bo_pin(bo);
+	ttm_bo_unreserve(bo);
+	*p_bo = bo;
+
+	return 0;
+
+error_free:
+	kfree(bo);
+	return ret;
+}
 
 /**
  * vmw_bo_init - Initialize a vmw buffer object
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c
index 3b41cf63110a..9a9fe10d829b 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cmdbuf.c
@@ -1245,9 +1245,9 @@ int vmw_cmdbuf_set_pool_size(struct vmw_cmdbuf_man *man,
 		    !dev_priv->has_mob)
 			return -ENOMEM;
 
-		ret = ttm_bo_create(&dev_priv->bdev, size, ttm_bo_type_device,
-				    &vmw_mob_ne_placement, 0, false,
-				    &man->cmd_space);
+		ret = vmw_bo_create_kernel(dev_priv, size,
+					   &vmw_mob_placement,
+					   &man->cmd_space);
 		if (ret)
 			return ret;
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index 9ceee4eb0b13..5d07de5183e1 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -845,6 +845,10 @@ extern void vmw_bo_get_guest_ptr(const struct ttm_buffer_object *buf,
 				 SVGAGuestPtr *ptr);
 extern void vmw_bo_pin_reserved(struct vmw_buffer_object *bo, bool pin);
 extern void vmw_bo_bo_free(struct ttm_buffer_object *bo);
+extern int vmw_bo_create_kernel(struct vmw_private *dev_priv,
+				unsigned long size,
+				struct ttm_placement *placement,
+				struct ttm_buffer_object **p_bo);
 extern int vmw_bo_init(struct vmw_private *dev_priv,
 		       struct vmw_buffer_object *vmw_bo,
 		       size_t size, struct ttm_placement *placement,
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index d7ea658e9910..39a2f720f4ed 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -793,11 +793,9 @@ int vmw_bo_create_and_populate(struct vmw_private *dev_priv,
 	struct ttm_buffer_object *bo;
 	int ret;
 
-	ret = ttm_bo_create(&dev_priv->bdev, bo_size,
-			    ttm_bo_type_device,
-			    &vmw_sys_ne_placement,
-			    0, false, &bo);
-
+	ret = vmw_bo_create_kernel(dev_priv, bo_size,
+				   &vmw_sys_placement,
+				   &bo);
 	if (unlikely(ret != 0))
 		return ret;
 
-- 
cgit 


From fbe86ca567919b22bbba1220ce55020b1868879f Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Mon, 21 Sep 2020 14:37:25 +0200
Subject: drm/vmwgfx: switch over to the new pin interface v2
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Stop using TTM_PL_FLAG_NO_EVICT.

v2: fix unconditional pinning

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Link: https://patchwork.freedesktop.org/patch/391601/?series=81973&rev=1
---
 drivers/gpu/drm/vmwgfx/vmwgfx_blit.c       |  4 +--
 drivers/gpu/drm/vmwgfx/vmwgfx_bo.c         | 49 +++++++++++++++---------------
 drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c    |  4 +--
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.c        |  2 +-
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.h        |  7 +----
 drivers/gpu/drm/vmwgfx/vmwgfx_fb.c         |  2 +-
 drivers/gpu/drm/vmwgfx/vmwgfx_resource.c   |  4 +--
 drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c       |  4 +--
 drivers/gpu/drm/vmwgfx/vmwgfx_shader.c     |  4 +--
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 42 -------------------------
 drivers/gpu/drm/vmwgfx/vmwgfx_validation.c |  2 +-
 11 files changed, 39 insertions(+), 85 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c b/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c
index e8d66182cd7b..ea2f2f937eb3 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_blit.c
@@ -459,9 +459,9 @@ int vmw_bo_cpu_blit(struct ttm_buffer_object *dst,
 	int ret = 0;
 
 	/* Buffer objects need to be either pinned or reserved: */
-	if (!(dst->mem.placement & TTM_PL_FLAG_NO_EVICT))
+	if (!(dst->pin_count))
 		dma_resv_assert_held(dst->base.resv);
-	if (!(src->mem.placement & TTM_PL_FLAG_NO_EVICT))
+	if (!(src->pin_count))
 		dma_resv_assert_held(src->base.resv);
 
 	if (!ttm_tt_is_populated(dst->ttm)) {
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
index c8ca09f0e627..9a66ba254326 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
@@ -106,7 +106,7 @@ int vmw_bo_pin_in_placement(struct vmw_private *dev_priv,
 	if (unlikely(ret != 0))
 		goto err;
 
-	if (buf->pin_count > 0)
+	if (buf->base.pin_count > 0)
 		ret = ttm_bo_mem_compat(placement, &bo->mem,
 					&new_flags) == true ? 0 : -EINVAL;
 	else
@@ -155,7 +155,7 @@ int vmw_bo_pin_in_vram_or_gmr(struct vmw_private *dev_priv,
 	if (unlikely(ret != 0))
 		goto err;
 
-	if (buf->pin_count > 0) {
+	if (buf->base.pin_count > 0) {
 		ret = ttm_bo_mem_compat(&vmw_vram_gmr_placement, &bo->mem,
 					&new_flags) == true ? 0 : -EINVAL;
 		goto out_unreserve;
@@ -246,12 +246,12 @@ int vmw_bo_pin_in_start_of_vram(struct vmw_private *dev_priv,
 	if (bo->mem.mem_type == TTM_PL_VRAM &&
 	    bo->mem.start < bo->num_pages &&
 	    bo->mem.start > 0 &&
-	    buf->pin_count == 0) {
+	    buf->base.pin_count == 0) {
 		ctx.interruptible = false;
 		(void) ttm_bo_validate(bo, &vmw_sys_placement, &ctx);
 	}
 
-	if (buf->pin_count > 0)
+	if (buf->base.pin_count > 0)
 		ret = ttm_bo_mem_compat(&placement, &bo->mem,
 					&new_flags) == true ? 0 : -EINVAL;
 	else
@@ -343,23 +343,13 @@ void vmw_bo_pin_reserved(struct vmw_buffer_object *vbo, bool pin)
 
 	dma_resv_assert_held(bo->base.resv);
 
-	if (pin) {
-		if (vbo->pin_count++ > 0)
-			return;
-	} else {
-		WARN_ON(vbo->pin_count <= 0);
-		if (--vbo->pin_count > 0)
-			return;
-	}
+	if (pin == !!bo->pin_count)
+		return;
 
 	pl.fpfn = 0;
 	pl.lpfn = 0;
 	pl.mem_type = bo->mem.mem_type;
 	pl.flags = bo->mem.placement;
-	if (pin)
-		pl.flags |= TTM_PL_FLAG_NO_EVICT;
-	else
-		pl.flags &= ~TTM_PL_FLAG_NO_EVICT;
 
 	memset(&placement, 0, sizeof(placement));
 	placement.num_placement = 1;
@@ -368,8 +358,12 @@ void vmw_bo_pin_reserved(struct vmw_buffer_object *vbo, bool pin)
 	ret = ttm_bo_validate(bo, &placement, &ctx);
 
 	BUG_ON(ret != 0 || bo->mem.mem_type != old_mem_type);
-}
 
+	if (pin)
+		ttm_bo_pin(bo);
+	else
+		ttm_bo_unpin(bo);
+}
 
 /**
  * vmw_bo_map_and_cache - Map a buffer object and cache the map
@@ -539,6 +533,7 @@ error_free:
  * @size: Buffer object size in bytes.
  * @placement: Initial placement.
  * @interruptible: Whether waits should be performed interruptible.
+ * @pin: If the BO should be created pinned at a fixed location.
  * @bo_free: The buffer object destructor.
  * Returns: Zero on success, negative error code on error.
  *
@@ -547,9 +542,10 @@ error_free:
 int vmw_bo_init(struct vmw_private *dev_priv,
 		struct vmw_buffer_object *vmw_bo,
 		size_t size, struct ttm_placement *placement,
-		bool interruptible,
+		bool interruptible, bool pin,
 		void (*bo_free)(struct ttm_buffer_object *bo))
 {
+	struct ttm_operation_ctx ctx = { interruptible, false };
 	struct ttm_bo_device *bdev = &dev_priv->bdev;
 	size_t acc_size;
 	int ret;
@@ -563,11 +559,16 @@ int vmw_bo_init(struct vmw_private *dev_priv,
 	vmw_bo->base.priority = 3;
 	vmw_bo->res_tree = RB_ROOT;
 
-	ret = ttm_bo_init(bdev, &vmw_bo->base, size,
-			  ttm_bo_type_device, placement,
-			  0, interruptible, acc_size,
-			  NULL, NULL, bo_free);
-	return ret;
+	ret = ttm_bo_init_reserved(bdev, &vmw_bo->base, size,
+				   ttm_bo_type_device, placement,
+				   0, &ctx, acc_size, NULL, NULL, bo_free);
+	if (unlikely(ret))
+		return ret;
+
+	if (pin)
+		ttm_bo_pin(&vmw_bo->base);
+	ttm_bo_unreserve(&vmw_bo->base);
+	return 0;
 }
 
 
@@ -656,7 +657,7 @@ int vmw_user_bo_alloc(struct vmw_private *dev_priv,
 	ret = vmw_bo_init(dev_priv, &user_bo->vbo, size,
 			  (dev_priv->has_mob) ?
 			  &vmw_sys_placement :
-			  &vmw_vram_sys_placement, true,
+			  &vmw_vram_sys_placement, true, false,
 			  &vmw_user_bo_destroy);
 	if (unlikely(ret != 0))
 		return ret;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c
index 65e8e7a97724..984d8884357d 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c
@@ -410,8 +410,8 @@ static int vmw_cotable_resize(struct vmw_resource *res, size_t new_size)
 	if (!buf)
 		return -ENOMEM;
 
-	ret = vmw_bo_init(dev_priv, buf, new_size, &vmw_mob_ne_placement,
-			  true, vmw_bo_bo_free);
+	ret = vmw_bo_init(dev_priv, buf, new_size, &vmw_mob_placement,
+			  true, true, vmw_bo_bo_free);
 	if (ret) {
 		DRM_ERROR("Failed initializing new cotable MOB.\n");
 		return ret;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
index 31e3e5c9f362..bdb7a5e96560 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -372,7 +372,7 @@ static int vmw_dummy_query_bo_create(struct vmw_private *dev_priv)
 		return -ENOMEM;
 
 	ret = vmw_bo_init(dev_priv, vbo, PAGE_SIZE,
-			  &vmw_sys_ne_placement, false,
+			  &vmw_sys_placement, false, true,
 			  &vmw_bo_bo_free);
 	if (unlikely(ret != 0))
 		return ret;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index 5d07de5183e1..b45becbb00f8 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -99,7 +99,6 @@ struct vmw_fpriv {
  * struct vmw_buffer_object - TTM buffer object with vmwgfx additions
  * @base: The TTM buffer object
  * @res_tree: RB tree of resources using this buffer object as a backing MOB
- * @pin_count: pin depth
  * @cpu_writers: Number of synccpu write grabs. Protected by reservation when
  * increased. May be decreased without reservation.
  * @dx_query_ctx: DX context if this buffer object is used as a DX query MOB
@@ -110,7 +109,6 @@ struct vmw_fpriv {
 struct vmw_buffer_object {
 	struct ttm_buffer_object base;
 	struct rb_root res_tree;
-	s32 pin_count;
 	atomic_t cpu_writers;
 	/* Not ref-counted.  Protected by binding_mutex */
 	struct vmw_resource *dx_query_ctx;
@@ -852,7 +850,7 @@ extern int vmw_bo_create_kernel(struct vmw_private *dev_priv,
 extern int vmw_bo_init(struct vmw_private *dev_priv,
 		       struct vmw_buffer_object *vmw_bo,
 		       size_t size, struct ttm_placement *placement,
-		       bool interruptible,
+		       bool interruptible, bool pin,
 		       void (*bo_free)(struct ttm_buffer_object *bo));
 extern int vmw_user_bo_verify_access(struct ttm_buffer_object *bo,
 				     struct ttm_object_file *tfile);
@@ -1009,15 +1007,12 @@ extern void vmw_validation_mem_init_ttm(struct vmw_private *dev_priv,
 
 extern const size_t vmw_tt_size;
 extern struct ttm_placement vmw_vram_placement;
-extern struct ttm_placement vmw_vram_ne_placement;
 extern struct ttm_placement vmw_vram_sys_placement;
 extern struct ttm_placement vmw_vram_gmr_placement;
 extern struct ttm_placement vmw_sys_placement;
-extern struct ttm_placement vmw_sys_ne_placement;
 extern struct ttm_placement vmw_evictable_placement;
 extern struct ttm_placement vmw_srf_placement;
 extern struct ttm_placement vmw_mob_placement;
-extern struct ttm_placement vmw_mob_ne_placement;
 extern struct ttm_placement vmw_nonfixed_placement;
 extern struct ttm_bo_driver vmw_bo_driver;
 extern const struct vmw_sg_table *
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c b/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
index c59806d40e15..4d60201037d1 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_fb.c
@@ -406,7 +406,7 @@ static int vmw_fb_create_bo(struct vmw_private *vmw_priv,
 
 	ret = vmw_bo_init(vmw_priv, vmw_bo, size,
 			      &vmw_sys_placement,
-			      false,
+			      false, false,
 			      &vmw_bo_bo_free);
 	if (unlikely(ret != 0))
 		goto err_unlock; /* init frees the buffer on failure */
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
index c0f156078dda..5e922d9d5f2c 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -370,7 +370,7 @@ static int vmw_resource_buf_alloc(struct vmw_resource *res,
 
 	ret = vmw_bo_init(res->dev_priv, backup, res->backup_size,
 			      res->func->backup_placement,
-			      interruptible,
+			      interruptible, false,
 			      &vmw_bo_bo_free);
 	if (unlikely(ret != 0))
 		goto out_no_bo;
@@ -1002,7 +1002,7 @@ int vmw_resource_pin(struct vmw_resource *res, bool interruptible)
 			vbo = res->backup;
 
 			ttm_bo_reserve(&vbo->base, interruptible, false, NULL);
-			if (!vbo->pin_count) {
+			if (!vbo->base.pin_count) {
 				ret = ttm_bo_validate
 					(&vbo->base,
 					 res->func->backup_placement,
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c b/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c
index 4bf0f5ec4fc2..0a53d189fceb 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_scrn.c
@@ -451,8 +451,8 @@ vmw_sou_primary_plane_prepare_fb(struct drm_plane *plane,
 	 */
 	vmw_overlay_pause_all(dev_priv);
 	ret = vmw_bo_init(dev_priv, vps->bo, size,
-			      &vmw_vram_ne_placement,
-			      false, &vmw_bo_bo_free);
+			      &vmw_vram_placement,
+			      false, true, &vmw_bo_bo_free);
 	vmw_overlay_resume_all(dev_priv);
 	if (ret) {
 		vps->bo = NULL; /* vmw_bo_init frees on error */
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
index e139fdfd1635..f328aa5839a2 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
@@ -978,8 +978,8 @@ int vmw_compat_shader_add(struct vmw_private *dev_priv,
 	if (unlikely(!buf))
 		return -ENOMEM;
 
-	ret = vmw_bo_init(dev_priv, buf, size, &vmw_sys_ne_placement,
-			      true, vmw_bo_bo_free);
+	ret = vmw_bo_init(dev_priv, buf, size, &vmw_sys_placement,
+			      true, true, vmw_bo_bo_free);
 	if (unlikely(ret != 0))
 		goto out;
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 39a2f720f4ed..fc68f54df46a 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -37,13 +37,6 @@ static const struct ttm_place vram_placement_flags = {
 	.flags = TTM_PL_FLAG_CACHED
 };
 
-static const struct ttm_place vram_ne_placement_flags = {
-	.fpfn = 0,
-	.lpfn = 0,
-	.mem_type = TTM_PL_VRAM,
-	.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_NO_EVICT
-};
-
 static const struct ttm_place sys_placement_flags = {
 	.fpfn = 0,
 	.lpfn = 0,
@@ -51,13 +44,6 @@ static const struct ttm_place sys_placement_flags = {
 	.flags = TTM_PL_FLAG_CACHED
 };
 
-static const struct ttm_place sys_ne_placement_flags = {
-	.fpfn = 0,
-	.lpfn = 0,
-	.mem_type = TTM_PL_SYSTEM,
-	.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_NO_EVICT
-};
-
 static const struct ttm_place gmr_placement_flags = {
 	.fpfn = 0,
 	.lpfn = 0,
@@ -72,13 +58,6 @@ static const struct ttm_place mob_placement_flags = {
 	.flags = TTM_PL_FLAG_CACHED
 };
 
-static const struct ttm_place mob_ne_placement_flags = {
-	.fpfn = 0,
-	.lpfn = 0,
-	.mem_type = VMW_PL_MOB,
-	.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_NO_EVICT
-};
-
 struct ttm_placement vmw_vram_placement = {
 	.num_placement = 1,
 	.placement = &vram_placement_flags,
@@ -128,13 +107,6 @@ struct ttm_placement vmw_vram_sys_placement = {
 	.busy_placement = &sys_placement_flags
 };
 
-struct ttm_placement vmw_vram_ne_placement = {
-	.num_placement = 1,
-	.placement = &vram_ne_placement_flags,
-	.num_busy_placement = 1,
-	.busy_placement = &vram_ne_placement_flags
-};
-
 struct ttm_placement vmw_sys_placement = {
 	.num_placement = 1,
 	.placement = &sys_placement_flags,
@@ -142,13 +114,6 @@ struct ttm_placement vmw_sys_placement = {
 	.busy_placement = &sys_placement_flags
 };
 
-struct ttm_placement vmw_sys_ne_placement = {
-	.num_placement = 1,
-	.placement = &sys_ne_placement_flags,
-	.num_busy_placement = 1,
-	.busy_placement = &sys_ne_placement_flags
-};
-
 static const struct ttm_place evictable_placement_flags[] = {
 	{
 		.fpfn = 0,
@@ -213,13 +178,6 @@ struct ttm_placement vmw_mob_placement = {
 	.busy_placement = &mob_placement_flags
 };
 
-struct ttm_placement vmw_mob_ne_placement = {
-	.num_placement = 1,
-	.num_busy_placement = 1,
-	.placement = &mob_ne_placement_flags,
-	.busy_placement = &mob_ne_placement_flags
-};
-
 struct ttm_placement vmw_nonfixed_placement = {
 	.num_placement = 3,
 	.placement = nonfixed_placement_flags,
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c b/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c
index e69bc373ae2e..f2e2bf6d1421 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_validation.c
@@ -540,7 +540,7 @@ int vmw_validation_bo_validate_single(struct ttm_buffer_object *bo,
 	if (atomic_read(&vbo->cpu_writers))
 		return -EBUSY;
 
-	if (vbo->pin_count > 0)
+	if (vbo->base.pin_count > 0)
 		return 0;
 
 	if (validate_as_mob)
-- 
cgit 


From 279a3010218d517284641b7dac2d65c0ecef78ab Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Tue, 6 Oct 2020 10:06:41 +1000
Subject: drm/vmwgfx: move null mem checks outside move notifies
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Both fns checked mem == NULL, just move the check outside.

Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201006000644.1005758-3-airlied@gmail.com
---
 drivers/gpu/drm/vmwgfx/vmwgfx_bo.c         | 3 ---
 drivers/gpu/drm/vmwgfx/vmwgfx_resource.c   | 2 +-
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 2 ++
 3 files changed, 3 insertions(+), 4 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
index 9a66ba254326..263d76ae43f0 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
@@ -1192,9 +1192,6 @@ void vmw_bo_move_notify(struct ttm_buffer_object *bo,
 {
 	struct vmw_buffer_object *vbo;
 
-	if (mem == NULL)
-		return;
-
 	/* Make sure @bo is embedded in a struct vmw_buffer_object? */
 	if (bo->destroy != vmw_bo_bo_free &&
 	    bo->destroy != vmw_user_bo_destroy)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
index 5e922d9d5f2c..00b535831a7a 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -867,7 +867,7 @@ void vmw_query_move_notify(struct ttm_buffer_object *bo,
 	mutex_lock(&dev_priv->binding_mutex);
 
 	dx_query_mob = container_of(bo, struct vmw_buffer_object, base);
-	if (mem == NULL || !dx_query_mob || !dx_query_mob->dx_query_ctx) {
+	if (!dx_query_mob || !dx_query_mob->dx_query_ctx) {
 		mutex_unlock(&dev_priv->binding_mutex);
 		return;
 	}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index fc68f54df46a..2f88d2d79f9a 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -707,6 +707,8 @@ static void vmw_move_notify(struct ttm_buffer_object *bo,
 			    bool evict,
 			    struct ttm_resource *mem)
 {
+	if (!mem)
+		return;
 	vmw_bo_move_notify(bo, mem);
 	vmw_query_move_notify(bo, mem);
 }
-- 
cgit 


From bcff5d3e3bab979d820bbabd3f0be13ae6ce6cf5 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Tue, 6 Oct 2020 10:06:42 +1000
Subject: drm/vmwgfx: add a move callback.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This just copies the fallback to vmwgfx, I'm going to iterate on this
a bit until it's not the same as the fallback path.

Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201006000644.1005758-4-airlied@gmail.com
---
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 2f88d2d79f9a..7b5fd5288870 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -725,6 +725,24 @@ static void vmw_swap_notify(struct ttm_buffer_object *bo)
 	(void) ttm_bo_wait(bo, false, false);
 }
 
+static int vmw_move(struct ttm_buffer_object *bo,
+		    bool evict,
+		    struct ttm_operation_ctx *ctx,
+		    struct ttm_resource *new_mem)
+{
+	struct ttm_resource_manager *old_man = ttm_manager_type(bo->bdev, bo->mem.mem_type);
+	struct ttm_resource_manager *new_man = ttm_manager_type(bo->bdev, new_mem->mem_type);
+
+	if (old_man->use_tt && new_man->use_tt) {
+		if (bo->mem.mem_type == TTM_PL_SYSTEM) {
+			ttm_bo_assign_mem(bo, new_mem);
+			return 0;
+		}
+		return ttm_bo_move_ttm(bo, ctx, new_mem);
+	} else {
+		return ttm_bo_move_memcpy(bo, ctx, new_mem);
+	}
+}
 
 struct ttm_bo_driver vmw_bo_driver = {
 	.ttm_tt_create = &vmw_ttm_tt_create,
@@ -735,7 +753,7 @@ struct ttm_bo_driver vmw_bo_driver = {
 	.ttm_tt_destroy = &vmw_ttm_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = vmw_evict_flags,
-	.move = NULL,
+	.move = vmw_move,
 	.verify_access = vmw_verify_access,
 	.move_notify = vmw_move_notify,
 	.swap_notify = vmw_swap_notify,
-- 
cgit 


From 1b4ea4c5980ff3a64607166298269c30a9671d33 Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Wed, 30 Sep 2020 10:38:48 +0200
Subject: drm/ttm: set the tt caching state at creation time
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

All drivers can determine the tt caching state at creation time,
no need to do this on the fly during every validation.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Link: https://patchwork.freedesktop.org/patch/394253/
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c    |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    | 11 ++++++++--
 drivers/gpu/drm/drm_gem_vram_helper.c      |  2 +-
 drivers/gpu/drm/nouveau/nouveau_sgdma.c    | 13 +++++++++++-
 drivers/gpu/drm/qxl/qxl_ttm.c              |  2 +-
 drivers/gpu/drm/radeon/radeon_ttm.c        | 16 ++++++++++++--
 drivers/gpu/drm/ttm/ttm_agp_backend.c      |  2 +-
 drivers/gpu/drm/ttm/ttm_page_alloc.c       | 26 +++++++++++------------
 drivers/gpu/drm/ttm/ttm_page_alloc_dma.c   | 20 +++++++++---------
 drivers/gpu/drm/ttm/ttm_tt.c               | 33 +++++++++++++++--------------
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  6 ++++--
 include/drm/ttm/ttm_caching.h              | 34 ++++++++++++++++++++++++++++++
 include/drm/ttm/ttm_tt.h                   | 16 ++++++--------
 13 files changed, 123 insertions(+), 60 deletions(-)
 create mode 100644 include/drm/ttm/ttm_caching.h

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 213ef090bb0e..3c5ad69eff19 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -124,7 +124,7 @@ uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo)
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
 	struct ttm_dma_tt *ttm;
 
-	if (bo->num_pages != 1 || bo->ttm->caching_state == tt_cached)
+	if (bo->num_pages != 1 || bo->ttm->caching == ttm_cached)
 		return AMDGPU_BO_INVALID_OFFSET;
 
 	ttm = container_of(bo->ttm, struct ttm_dma_tt, ttm);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 399961035ae6..7f41a47e7353 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -1292,7 +1292,9 @@ static void amdgpu_ttm_backend_destroy(struct ttm_bo_device *bdev,
 static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo,
 					   uint32_t page_flags)
 {
+	struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo);
 	struct amdgpu_ttm_tt *gtt;
+	enum ttm_caching caching;
 
 	gtt = kzalloc(sizeof(struct amdgpu_ttm_tt), GFP_KERNEL);
 	if (gtt == NULL) {
@@ -1300,8 +1302,13 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo,
 	}
 	gtt->gobj = &bo->base;
 
+	if (abo->flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC)
+		caching = ttm_write_combined;
+	else
+		caching = ttm_cached;
+
 	/* allocate space for the uninitialized page entries */
-	if (ttm_sg_tt_init(&gtt->ttm, bo, page_flags)) {
+	if (ttm_sg_tt_init(&gtt->ttm, bo, page_flags, caching)) {
 		kfree(gtt);
 		return NULL;
 	}
@@ -1525,7 +1532,7 @@ uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_resource *mem)
 	if (mem && mem->mem_type == TTM_PL_TT) {
 		flags |= AMDGPU_PTE_SYSTEM;
 
-		if (ttm->caching_state == tt_cached)
+		if (ttm->caching == ttm_cached)
 			flags |= AMDGPU_PTE_SNOOPED;
 	}
 
diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index 3213429f8444..ad58d0af5141 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -918,7 +918,7 @@ static struct ttm_tt *bo_driver_ttm_tt_create(struct ttm_buffer_object *bo,
 	if (!tt)
 		return NULL;
 
-	ret = ttm_tt_init(tt, bo, page_flags);
+	ret = ttm_tt_init(tt, bo, page_flags, ttm_cached);
 	if (ret < 0)
 		goto err_ttm_tt_init;
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
index 806d9ec310f5..cd6fdebae795 100644
--- a/drivers/gpu/drm/nouveau/nouveau_sgdma.c
+++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
@@ -5,6 +5,7 @@
 #include "nouveau_drv.h"
 #include "nouveau_mem.h"
 #include "nouveau_ttm.h"
+#include "nouveau_bo.h"
 
 struct nouveau_sgdma_be {
 	/* this has to be the first field so populate/unpopulated in
@@ -67,13 +68,23 @@ nouveau_sgdma_unbind(struct ttm_bo_device *bdev, struct ttm_tt *ttm)
 struct ttm_tt *
 nouveau_sgdma_create_ttm(struct ttm_buffer_object *bo, uint32_t page_flags)
 {
+	struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
+	struct nouveau_bo *nvbo = nouveau_bo(bo);
 	struct nouveau_sgdma_be *nvbe;
+	enum ttm_caching caching;
+
+	if (nvbo->force_coherent)
+		caching = ttm_uncached;
+	else if (drm->agp.bridge)
+		caching = ttm_write_combined;
+	else
+		caching = ttm_cached;
 
 	nvbe = kzalloc(sizeof(*nvbe), GFP_KERNEL);
 	if (!nvbe)
 		return NULL;
 
-	if (ttm_dma_tt_init(&nvbe->ttm, bo, page_flags)) {
+	if (ttm_dma_tt_init(&nvbe->ttm, bo, page_flags, caching)) {
 		kfree(nvbe);
 		return NULL;
 	}
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index 669bceb58205..f50863493f64 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -133,7 +133,7 @@ static struct ttm_tt *qxl_ttm_tt_create(struct ttm_buffer_object *bo,
 	ttm = kzalloc(sizeof(struct ttm_tt), GFP_KERNEL);
 	if (ttm == NULL)
 		return NULL;
-	if (ttm_tt_init(ttm, bo, page_flags)) {
+	if (ttm_tt_init(ttm, bo, page_flags, ttm_cached)) {
 		kfree(ttm);
 		return NULL;
 	}
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 63e38b05a5bc..130a7cea35c3 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -546,7 +546,7 @@ static int radeon_ttm_backend_bind(struct ttm_bo_device *bdev,
 		WARN(1, "nothing to bind %lu pages for mreg %p back %p!\n",
 		     ttm->num_pages, bo_mem, ttm);
 	}
-	if (ttm->caching_state == tt_cached)
+	if (ttm->caching == ttm_cached)
 		flags |= RADEON_GART_PAGE_SNOOP;
 	r = radeon_gart_bind(rdev, gtt->offset, ttm->num_pages,
 			     ttm->pages, gtt->ttm.dma_address, flags);
@@ -590,6 +590,10 @@ static struct ttm_tt *radeon_ttm_tt_create(struct ttm_buffer_object *bo,
 {
 	struct radeon_device *rdev;
 	struct radeon_ttm_tt *gtt;
+	enum ttm_caching caching;
+	struct radeon_bo *rbo;
+
+	rbo = container_of(bo, struct radeon_bo, tbo);
 
 	rdev = radeon_get_rdev(bo->bdev);
 #if IS_ENABLED(CONFIG_AGP)
@@ -603,7 +607,15 @@ static struct ttm_tt *radeon_ttm_tt_create(struct ttm_buffer_object *bo,
 	if (gtt == NULL) {
 		return NULL;
 	}
-	if (ttm_dma_tt_init(&gtt->ttm, bo, page_flags)) {
+
+	if (rbo->flags & RADEON_GEM_GTT_UC)
+		caching = ttm_uncached;
+	else if (rbo->flags & RADEON_GEM_GTT_WC)
+		caching = ttm_write_combined;
+	else
+		caching = ttm_cached;
+
+	if (ttm_dma_tt_init(&gtt->ttm, bo, page_flags, caching)) {
 		kfree(gtt);
 		return NULL;
 	}
diff --git a/drivers/gpu/drm/ttm/ttm_agp_backend.c b/drivers/gpu/drm/ttm/ttm_agp_backend.c
index a98fd795b752..a723062d37e7 100644
--- a/drivers/gpu/drm/ttm/ttm_agp_backend.c
+++ b/drivers/gpu/drm/ttm/ttm_agp_backend.c
@@ -136,7 +136,7 @@ struct ttm_tt *ttm_agp_tt_create(struct ttm_buffer_object *bo,
 	agp_be->mem = NULL;
 	agp_be->bridge = bridge;
 
-	if (ttm_tt_init(&agp_be->ttm, bo, page_flags)) {
+	if (ttm_tt_init(&agp_be->ttm, bo, page_flags, ttm_write_combined)) {
 		kfree(agp_be);
 		return NULL;
 	}
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index 111031cbb6df..c8f6790962b9 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -220,14 +220,14 @@ static struct ttm_pool_manager *_manager;
 /**
  * Select the right pool or requested caching state and ttm flags. */
 static struct ttm_page_pool *ttm_get_pool(int flags, bool huge,
-					  enum ttm_caching_state cstate)
+					  enum ttm_caching cstate)
 {
 	int pool_index;
 
-	if (cstate == tt_cached)
+	if (cstate == ttm_cached)
 		return NULL;
 
-	if (cstate == tt_wc)
+	if (cstate == ttm_write_combined)
 		pool_index = 0x0;
 	else
 		pool_index = 0x1;
@@ -441,17 +441,17 @@ static void ttm_pool_mm_shrink_fini(struct ttm_pool_manager *manager)
 }
 
 static int ttm_set_pages_caching(struct page **pages,
-		enum ttm_caching_state cstate, unsigned cpages)
+		enum ttm_caching cstate, unsigned cpages)
 {
 	int r = 0;
 	/* Set page caching */
 	switch (cstate) {
-	case tt_uncached:
+	case ttm_uncached:
 		r = ttm_set_pages_array_uc(pages, cpages);
 		if (r)
 			pr_err("Failed to set %d pages to uc!\n", cpages);
 		break;
-	case tt_wc:
+	case ttm_write_combined:
 		r = ttm_set_pages_array_wc(pages, cpages);
 		if (r)
 			pr_err("Failed to set %d pages to wc!\n", cpages);
@@ -486,7 +486,7 @@ static void ttm_handle_caching_failure(struct page **failed_pages,
  * pages returned in pages array.
  */
 static int ttm_alloc_new_pages(struct list_head *pages, gfp_t gfp_flags,
-			       int ttm_flags, enum ttm_caching_state cstate,
+			       int ttm_flags, enum ttm_caching cstate,
 			       unsigned count, unsigned order)
 {
 	struct page **caching_array;
@@ -566,7 +566,7 @@ out:
  * pages is small.
  */
 static void ttm_page_pool_fill_locked(struct ttm_page_pool *pool, int ttm_flags,
-				      enum ttm_caching_state cstate,
+				      enum ttm_caching cstate,
 				      unsigned count, unsigned long *irq_flags)
 {
 	struct page *p;
@@ -626,7 +626,7 @@ static void ttm_page_pool_fill_locked(struct ttm_page_pool *pool, int ttm_flags,
 static int ttm_page_pool_get_pages(struct ttm_page_pool *pool,
 				   struct list_head *pages,
 				   int ttm_flags,
-				   enum ttm_caching_state cstate,
+				   enum ttm_caching cstate,
 				   unsigned count, unsigned order)
 {
 	unsigned long irq_flags;
@@ -703,7 +703,7 @@ out:
 
 /* Put all pages in pages list to correct pool to wait for reuse */
 static void ttm_put_pages(struct page **pages, unsigned npages, int flags,
-			  enum ttm_caching_state cstate)
+			  enum ttm_caching cstate)
 {
 	struct ttm_page_pool *pool = ttm_get_pool(flags, false, cstate);
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
@@ -821,7 +821,7 @@ static void ttm_put_pages(struct page **pages, unsigned npages, int flags,
  * cached pages.
  */
 static int ttm_get_pages(struct page **pages, unsigned npages, int flags,
-			 enum ttm_caching_state cstate)
+			 enum ttm_caching cstate)
 {
 	struct ttm_page_pool *pool = ttm_get_pool(flags, false, cstate);
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
@@ -1040,7 +1040,7 @@ ttm_pool_unpopulate_helper(struct ttm_tt *ttm, unsigned mem_count_update)
 
 put_pages:
 	ttm_put_pages(ttm->pages, ttm->num_pages, ttm->page_flags,
-		      ttm->caching_state);
+		      ttm->caching);
 	ttm_tt_set_unpopulated(ttm);
 }
 
@@ -1057,7 +1057,7 @@ int ttm_pool_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 		return -ENOMEM;
 
 	ret = ttm_get_pages(ttm->pages, ttm->num_pages, ttm->page_flags,
-			    ttm->caching_state);
+			    ttm->caching);
 	if (unlikely(ret != 0)) {
 		ttm_pool_unpopulate_helper(ttm, 0);
 		return ret;
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
index 1045a5c26ee3..6625b43f6256 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
@@ -325,15 +325,15 @@ static struct dma_page *__ttm_dma_alloc_page(struct dma_pool *pool)
 	}
 	return d_page;
 }
-static enum pool_type ttm_to_type(int flags, enum ttm_caching_state cstate)
+static enum pool_type ttm_to_type(int flags, enum ttm_caching cstate)
 {
 	enum pool_type type = IS_UNDEFINED;
 
 	if (flags & TTM_PAGE_FLAG_DMA32)
 		type |= IS_DMA32;
-	if (cstate == tt_cached)
+	if (cstate == ttm_cached)
 		type |= IS_CACHED;
-	else if (cstate == tt_uncached)
+	else if (cstate == ttm_uncached)
 		type |= IS_UC;
 	else
 		type |= IS_WC;
@@ -663,7 +663,7 @@ static struct dma_pool *ttm_dma_find_pool(struct device *dev,
  * are pages that have changed their caching state already put them to the
  * pool.
  */
-static void ttm_dma_handle_caching_state_failure(struct dma_pool *pool,
+static void ttm_dma_handle_caching_failure(struct dma_pool *pool,
 						 struct list_head *d_pages,
 						 struct page **failed_pages,
 						 unsigned cpages)
@@ -734,7 +734,7 @@ static int ttm_dma_pool_alloc_new_pages(struct dma_pool *pool,
 				r = ttm_set_pages_caching(pool, caching_array,
 							  cpages);
 				if (r)
-					ttm_dma_handle_caching_state_failure(
+					ttm_dma_handle_caching_failure(
 						pool, d_pages, caching_array,
 						cpages);
 			}
@@ -760,7 +760,7 @@ static int ttm_dma_pool_alloc_new_pages(struct dma_pool *pool,
 				r = ttm_set_pages_caching(pool, caching_array,
 							  cpages);
 				if (r) {
-					ttm_dma_handle_caching_state_failure(
+					ttm_dma_handle_caching_failure(
 					     pool, d_pages, caching_array,
 					     cpages);
 					goto out;
@@ -773,7 +773,7 @@ static int ttm_dma_pool_alloc_new_pages(struct dma_pool *pool,
 	if (cpages) {
 		r = ttm_set_pages_caching(pool, caching_array, cpages);
 		if (r)
-			ttm_dma_handle_caching_state_failure(pool, d_pages,
+			ttm_dma_handle_caching_failure(pool, d_pages,
 					caching_array, cpages);
 	}
 out:
@@ -904,7 +904,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
 	INIT_LIST_HEAD(&ttm_dma->pages_list);
 	i = 0;
 
-	type = ttm_to_type(ttm->page_flags, ttm->caching_state);
+	type = ttm_to_type(ttm->page_flags, ttm->caching);
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 	if (ttm->page_flags & TTM_PAGE_FLAG_DMA32)
@@ -1000,7 +1000,7 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 	unsigned count, i, npages = 0;
 	unsigned long irq_flags;
 
-	type = ttm_to_type(ttm->page_flags, ttm->caching_state);
+	type = ttm_to_type(ttm->page_flags, ttm->caching);
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 	pool = ttm_dma_find_pool(dev, type | IS_HUGE);
@@ -1032,7 +1032,7 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 		return;
 
 	is_cached = (ttm_dma_find_pool(pool->dev,
-		     ttm_to_type(ttm->page_flags, tt_cached)) == pool);
+		     ttm_to_type(ttm->page_flags, ttm_cached)) == pool);
 
 	/* make sure pages array match list and count number of pages */
 	count = 0;
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index 23e9604bc924..a465f51df027 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -114,31 +114,30 @@ static int ttm_sg_tt_alloc_page_directory(struct ttm_dma_tt *ttm)
 	return 0;
 }
 
-static int ttm_tt_set_caching(struct ttm_tt *ttm,
-			      enum ttm_caching_state c_state)
+static int ttm_tt_set_caching(struct ttm_tt *ttm, enum ttm_caching caching)
 {
-	if (ttm->caching_state == c_state)
+	if (ttm->caching == caching)
 		return 0;
 
 	/* Can't change the caching state after TT is populated */
 	if (WARN_ON_ONCE(ttm_tt_is_populated(ttm)))
 		return -EINVAL;
 
-	ttm->caching_state = c_state;
+	ttm->caching = caching;
 
 	return 0;
 }
 
 int ttm_tt_set_placement_caching(struct ttm_tt *ttm, uint32_t placement)
 {
-	enum ttm_caching_state state;
+	enum ttm_caching state;
 
 	if (placement & TTM_PL_FLAG_WC)
-		state = tt_wc;
+		state = ttm_write_combined;
 	else if (placement & TTM_PL_FLAG_UNCACHED)
-		state = tt_uncached;
+		state = ttm_uncached;
 	else
-		state = tt_cached;
+		state = ttm_cached;
 
 	return ttm_tt_set_caching(ttm, state);
 }
@@ -162,20 +161,22 @@ void ttm_tt_destroy(struct ttm_bo_device *bdev, struct ttm_tt *ttm)
 
 static void ttm_tt_init_fields(struct ttm_tt *ttm,
 			       struct ttm_buffer_object *bo,
-			       uint32_t page_flags)
+			       uint32_t page_flags,
+			       enum ttm_caching caching)
 {
 	ttm->num_pages = bo->num_pages;
-	ttm->caching_state = tt_cached;
+	ttm->caching = ttm_cached;
 	ttm->page_flags = page_flags;
 	ttm_tt_set_unpopulated(ttm);
 	ttm->swap_storage = NULL;
 	ttm->sg = bo->sg;
+	ttm->caching = caching;
 }
 
 int ttm_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo,
-		uint32_t page_flags)
+		uint32_t page_flags, enum ttm_caching caching)
 {
-	ttm_tt_init_fields(ttm, bo, page_flags);
+	ttm_tt_init_fields(ttm, bo, page_flags, caching);
 
 	if (ttm_tt_alloc_page_directory(ttm)) {
 		pr_err("Failed allocating page table\n");
@@ -193,11 +194,11 @@ void ttm_tt_fini(struct ttm_tt *ttm)
 EXPORT_SYMBOL(ttm_tt_fini);
 
 int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
-		    uint32_t page_flags)
+		    uint32_t page_flags, enum ttm_caching caching)
 {
 	struct ttm_tt *ttm = &ttm_dma->ttm;
 
-	ttm_tt_init_fields(ttm, bo, page_flags);
+	ttm_tt_init_fields(ttm, bo, page_flags, caching);
 
 	INIT_LIST_HEAD(&ttm_dma->pages_list);
 	if (ttm_dma_tt_alloc_page_directory(ttm_dma)) {
@@ -209,12 +210,12 @@ int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
 EXPORT_SYMBOL(ttm_dma_tt_init);
 
 int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
-		   uint32_t page_flags)
+		   uint32_t page_flags, enum ttm_caching caching)
 {
 	struct ttm_tt *ttm = &ttm_dma->ttm;
 	int ret;
 
-	ttm_tt_init_fields(ttm, bo, page_flags);
+	ttm_tt_init_fields(ttm, bo, page_flags, caching);
 
 	INIT_LIST_HEAD(&ttm_dma->pages_list);
 	if (page_flags & TTM_PAGE_FLAG_SG)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 7b5fd5288870..1fa7f9438ec4 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -647,9 +647,11 @@ static struct ttm_tt *vmw_ttm_tt_create(struct ttm_buffer_object *bo,
 	vmw_be->mob = NULL;
 
 	if (vmw_be->dev_priv->map_mode == vmw_dma_alloc_coherent)
-		ret = ttm_dma_tt_init(&vmw_be->dma_ttm, bo, page_flags);
+		ret = ttm_dma_tt_init(&vmw_be->dma_ttm, bo, page_flags,
+				      ttm_cached);
 	else
-		ret = ttm_tt_init(&vmw_be->dma_ttm.ttm, bo, page_flags);
+		ret = ttm_tt_init(&vmw_be->dma_ttm.ttm, bo, page_flags,
+				  ttm_cached);
 	if (unlikely(ret != 0))
 		goto out_no_init;
 
diff --git a/include/drm/ttm/ttm_caching.h b/include/drm/ttm/ttm_caching.h
new file mode 100644
index 000000000000..161624dcf6be
--- /dev/null
+++ b/include/drm/ttm/ttm_caching.h
@@ -0,0 +1,34 @@
+/*
+ * Copyright 2020 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Christian König
+ */
+
+#ifndef _TTM_CACHING_H_
+#define _TTM_CACHING_H_
+
+enum ttm_caching {
+	ttm_uncached,
+	ttm_write_combined,
+	ttm_cached
+};
+
+#endif
diff --git a/include/drm/ttm/ttm_tt.h b/include/drm/ttm/ttm_tt.h
index 5d1835d44084..c39c722d5184 100644
--- a/include/drm/ttm/ttm_tt.h
+++ b/include/drm/ttm/ttm_tt.h
@@ -28,6 +28,7 @@
 #define _TTM_TT_H_
 
 #include <linux/types.h>
+#include <drm/ttm/ttm_caching.h>
 
 struct ttm_tt;
 struct ttm_resource;
@@ -42,12 +43,6 @@ struct ttm_operation_ctx;
 
 #define TTM_PAGE_FLAG_PRIV_POPULATED  (1 << 31)
 
-enum ttm_caching_state {
-	tt_uncached,
-	tt_wc,
-	tt_cached
-};
-
 /**
  * struct ttm_tt
  *
@@ -69,7 +64,7 @@ struct ttm_tt {
 	unsigned long num_pages;
 	struct sg_table *sg; /* for SG objects via dma-buf */
 	struct file *swap_storage;
-	enum ttm_caching_state caching_state;
+	enum ttm_caching caching;
 };
 
 static inline bool ttm_tt_is_populated(struct ttm_tt *tt)
@@ -121,6 +116,7 @@ int ttm_tt_create(struct ttm_buffer_object *bo, bool zero_alloc);
  * @ttm: The struct ttm_tt.
  * @bo: The buffer object we create the ttm for.
  * @page_flags: Page flags as identified by TTM_PAGE_FLAG_XX flags.
+ * @caching: the desired caching state of the pages
  *
  * Create a struct ttm_tt to back data with system memory pages.
  * No pages are actually allocated.
@@ -128,11 +124,11 @@ int ttm_tt_create(struct ttm_buffer_object *bo, bool zero_alloc);
  * NULL: Out of memory.
  */
 int ttm_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo,
-		uint32_t page_flags);
+		uint32_t page_flags, enum ttm_caching caching);
 int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
-		    uint32_t page_flags);
+		    uint32_t page_flags, enum ttm_caching caching);
 int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
-		   uint32_t page_flags);
+		   uint32_t page_flags, enum ttm_caching caching);
 
 /**
  * ttm_tt_fini
-- 
cgit 


From 1cf65c45183a6c8b4703675d40e709f7ffed935c Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Wed, 30 Sep 2020 11:17:44 +0200
Subject: drm/ttm: add caching state to ttm_bus_placement
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

And implement setting it up correctly in the drivers.

This allows getting rid of the placement flags for this.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Link: https://patchwork.freedesktop.org/patch/394254/
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    |  1 +
 drivers/gpu/drm/drm_gem_vram_helper.c      |  1 +
 drivers/gpu/drm/nouveau/nouveau_bo.c       | 11 +++++++++++
 drivers/gpu/drm/qxl/qxl_ttm.c              |  2 ++
 drivers/gpu/drm/radeon/radeon_ttm.c        |  2 ++
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  1 +
 include/drm/ttm/ttm_resource.h             |  8 +++++---
 7 files changed, 23 insertions(+), 3 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 7f41a47e7353..5b56a66063fd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -769,6 +769,7 @@ static int amdgpu_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_reso
 
 		mem->bus.offset += adev->gmc.aper_base;
 		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_write_combined;
 		break;
 	default:
 		return -EINVAL;
diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index ad58d0af5141..b9e7ce1adf25 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -981,6 +981,7 @@ static int bo_driver_io_mem_reserve(struct ttm_bo_device *bdev,
 	case TTM_PL_VRAM:
 		mem->bus.offset = (mem->start << PAGE_SHIFT) + vmm->vram_base;
 		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_write_combined;
 		break;
 	default:
 		return -EINVAL;
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 0c0ca44a6802..cb878c0e8276 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1134,6 +1134,8 @@ nouveau_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_resource *reg)
 	struct nouveau_drm *drm = nouveau_bdev(bdev);
 	struct nvkm_device *device = nvxx_device(&drm->client.device);
 	struct nouveau_mem *mem = nouveau_mem(reg);
+	struct nvif_mmu *mmu = &drm->client.mmu;
+	const u8 type = mmu->type[drm->ttm.type_vram].type;
 	int ret;
 
 	mutex_lock(&drm->ttm.io_reserve_mutex);
@@ -1149,6 +1151,7 @@ retry:
 			reg->bus.offset = (reg->start << PAGE_SHIFT) +
 				drm->agp.base;
 			reg->bus.is_iomem = !drm->agp.cma;
+			reg->bus.caching = ttm_write_combined;
 		}
 #endif
 		if (drm->client.mem->oclass < NVIF_CLASS_MEM_NV50 ||
@@ -1162,6 +1165,14 @@ retry:
 		reg->bus.offset = (reg->start << PAGE_SHIFT) +
 			device->func->resource_addr(device, 1);
 		reg->bus.is_iomem = true;
+
+		/* Some BARs do not support being ioremapped WC */
+		if (drm->client.device.info.family >= NV_DEVICE_INFO_V0_TESLA &&
+		    type & NVIF_MEM_UNCACHED)
+			reg->bus.caching = ttm_uncached;
+		else
+			reg->bus.caching = ttm_write_combined;
+
 		if (drm->client.mem->oclass >= NVIF_CLASS_MEM_NV50) {
 			union {
 				struct nv50_mem_map_v0 nv50;
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index f50863493f64..61eb06dbbce8 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -83,11 +83,13 @@ int qxl_ttm_io_mem_reserve(struct ttm_bo_device *bdev,
 	case TTM_PL_VRAM:
 		mem->bus.is_iomem = true;
 		mem->bus.offset = (mem->start << PAGE_SHIFT) + qdev->vram_base;
+		mem->bus.caching = ttm_cached;
 		break;
 	case TTM_PL_PRIV:
 		mem->bus.is_iomem = true;
 		mem->bus.offset = (mem->start << PAGE_SHIFT) +
 			qdev->surfaceram_base;
+		mem->bus.caching = ttm_cached;
 		break;
 	default:
 		return -EINVAL;
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 130a7cea35c3..9b53a1d80632 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -372,6 +372,7 @@ static int radeon_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_reso
 			mem->bus.offset = (mem->start << PAGE_SHIFT) +
 				rdev->mc.agp_base;
 			mem->bus.is_iomem = !rdev->ddev->agp->cant_use_aperture;
+			mem->bus.caching = ttm_write_combined;
 		}
 #endif
 		break;
@@ -382,6 +383,7 @@ static int radeon_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_reso
 			return -EINVAL;
 		mem->bus.offset += rdev->mc.aper_base;
 		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_write_combined;
 #ifdef __alpha__
 		/*
 		 * Alpha: use bus.addr to hold the ioremap() return,
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 1fa7f9438ec4..fae88969a15a 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -688,6 +688,7 @@ static int vmw_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_resourc
 		mem->bus.offset = (mem->start << PAGE_SHIFT) +
 			dev_priv->vram_start;
 		mem->bus.is_iomem = true;
+		mem->bus.caching = ttm_cached;
 		break;
 	default:
 		return -EINVAL;
diff --git a/include/drm/ttm/ttm_resource.h b/include/drm/ttm/ttm_resource.h
index 1b2f56163c6c..f48a70d39ac5 100644
--- a/include/drm/ttm/ttm_resource.h
+++ b/include/drm/ttm/ttm_resource.h
@@ -29,6 +29,7 @@
 #include <linux/mutex.h>
 #include <linux/dma-fence.h>
 #include <drm/drm_print.h>
+#include <drm/ttm/ttm_caching.h>
 
 #define TTM_MAX_BO_PRIORITY	4U
 
@@ -148,9 +149,10 @@ struct ttm_resource_manager {
  * Structure indicating the bus placement of an object.
  */
 struct ttm_bus_placement {
-	void		*addr;
-	phys_addr_t	offset;
-	bool		is_iomem;
+	void			*addr;
+	phys_addr_t		offset;
+	bool			is_iomem;
+	enum ttm_caching	caching;
 };
 
 /**
-- 
cgit 


From ce65b874001d756294e0b7cf06c51137af964f38 Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Wed, 30 Sep 2020 16:44:16 +0200
Subject: drm/ttm: nuke caching placement flags
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Changing the caching on the fly never really worked
flawlessly.

So stop this completely and just let drivers specific the
desired caching in the tt or bus object.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michael J. Ruhl <michael.j.ruhl@intel.com>
Link: https://patchwork.freedesktop.org/patch/394256/
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 20 ++++---------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    | 12 ++------
 drivers/gpu/drm/drm_gem_vram_helper.c      |  7 ++---
 drivers/gpu/drm/nouveau/nouveau_bo.c       | 36 ++++++-----------------
 drivers/gpu/drm/qxl/qxl_object.c           | 10 +++----
 drivers/gpu/drm/qxl/qxl_ttm.c              |  2 +-
 drivers/gpu/drm/radeon/radeon_object.c     | 46 ++++++------------------------
 drivers/gpu/drm/radeon/radeon_ttm.c        | 18 +++---------
 drivers/gpu/drm/ttm/ttm_agp_backend.c      |  2 +-
 drivers/gpu/drm/ttm/ttm_bo.c               | 44 +++-------------------------
 drivers/gpu/drm/ttm/ttm_bo_util.c          | 10 ++-----
 drivers/gpu/drm/ttm/ttm_tt.c               | 29 -------------------
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 30 +++++++++----------
 include/drm/ttm/ttm_placement.h            | 14 ---------
 include/drm/ttm/ttm_tt.h                   | 15 ----------
 15 files changed, 61 insertions(+), 234 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 8b96e7aaeff1..1aa516429c80 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -137,7 +137,7 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		places[c].fpfn = 0;
 		places[c].lpfn = 0;
 		places[c].mem_type = TTM_PL_VRAM;
-		places[c].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED;
+		places[c].flags = 0;
 
 		if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)
 			places[c].lpfn = visible_pfn;
@@ -154,11 +154,6 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		places[c].lpfn = 0;
 		places[c].mem_type = TTM_PL_TT;
 		places[c].flags = 0;
-		if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC)
-			places[c].flags |= TTM_PL_FLAG_WC |
-				TTM_PL_FLAG_UNCACHED;
-		else
-			places[c].flags |= TTM_PL_FLAG_CACHED;
 		c++;
 	}
 
@@ -167,11 +162,6 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		places[c].lpfn = 0;
 		places[c].mem_type = TTM_PL_SYSTEM;
 		places[c].flags = 0;
-		if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC)
-			places[c].flags |= TTM_PL_FLAG_WC |
-				TTM_PL_FLAG_UNCACHED;
-		else
-			places[c].flags |= TTM_PL_FLAG_CACHED;
 		c++;
 	}
 
@@ -179,7 +169,7 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		places[c].fpfn = 0;
 		places[c].lpfn = 0;
 		places[c].mem_type = AMDGPU_PL_GDS;
-		places[c].flags = TTM_PL_FLAG_UNCACHED;
+		places[c].flags = 0;
 		c++;
 	}
 
@@ -187,7 +177,7 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		places[c].fpfn = 0;
 		places[c].lpfn = 0;
 		places[c].mem_type = AMDGPU_PL_GWS;
-		places[c].flags = TTM_PL_FLAG_UNCACHED;
+		places[c].flags = 0;
 		c++;
 	}
 
@@ -195,7 +185,7 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		places[c].fpfn = 0;
 		places[c].lpfn = 0;
 		places[c].mem_type = AMDGPU_PL_OA;
-		places[c].flags = TTM_PL_FLAG_UNCACHED;
+		places[c].flags = 0;
 		c++;
 	}
 
@@ -203,7 +193,7 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain)
 		places[c].fpfn = 0;
 		places[c].lpfn = 0;
 		places[c].mem_type = TTM_PL_SYSTEM;
-		places[c].flags = TTM_PL_MASK_CACHING;
+		places[c].flags = 0;
 		c++;
 	}
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 5b56a66063fd..8cdec58b9106 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -92,7 +92,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_SYSTEM,
-		.flags = TTM_PL_MASK_CACHING
+		.flags = 0
 	};
 
 	/* Don't handle scatter gather BOs */
@@ -538,19 +538,13 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict,
 	placements.fpfn = 0;
 	placements.lpfn = 0;
 	placements.mem_type = TTM_PL_TT;
-	placements.flags = TTM_PL_MASK_CACHING;
+	placements.flags = 0;
 	r = ttm_bo_mem_space(bo, &placement, &tmp_mem, ctx);
 	if (unlikely(r)) {
 		pr_err("Failed to find GTT space for blit from VRAM\n");
 		return r;
 	}
 
-	/* set caching flags */
-	r = ttm_tt_set_placement_caching(bo->ttm, tmp_mem.placement);
-	if (unlikely(r)) {
-		goto out_cleanup;
-	}
-
 	r = ttm_tt_populate(bo->bdev, bo->ttm, ctx);
 	if (unlikely(r))
 		goto out_cleanup;
@@ -599,7 +593,7 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, bool evict,
 	placements.fpfn = 0;
 	placements.lpfn = 0;
 	placements.mem_type = TTM_PL_TT;
-	placements.flags = TTM_PL_MASK_CACHING;
+	placements.flags = 0;
 	r = ttm_bo_mem_space(bo, &placement, &tmp_mem, ctx);
 	if (unlikely(r)) {
 		pr_err("Failed to find GTT space for blit to VRAM\n");
diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index b9e7ce1adf25..7aeb5daf2805 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -147,15 +147,12 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo,
 
 	if (pl_flag & DRM_GEM_VRAM_PL_FLAG_VRAM) {
 		gbo->placements[c].mem_type = TTM_PL_VRAM;
-		gbo->placements[c++].flags = TTM_PL_FLAG_WC |
-					     TTM_PL_FLAG_UNCACHED |
-					     invariant_flags;
+		gbo->placements[c++].flags = invariant_flags;
 	}
 
 	if (pl_flag & DRM_GEM_VRAM_PL_FLAG_SYSTEM || !c) {
 		gbo->placements[c].mem_type = TTM_PL_SYSTEM;
-		gbo->placements[c++].flags = TTM_PL_MASK_CACHING |
-					     invariant_flags;
+		gbo->placements[c++].flags = invariant_flags;
 	}
 
 	gbo->placement.num_placement = c;
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index cb878c0e8276..4ccb3329014b 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -343,37 +343,23 @@ nouveau_bo_new(struct nouveau_cli *cli, u64 size, int align,
 }
 
 static void
-set_placement_list(struct nouveau_drm *drm, struct ttm_place *pl, unsigned *n,
-		   uint32_t domain, uint32_t flags)
+set_placement_list(struct ttm_place *pl, unsigned *n, uint32_t domain)
 {
 	*n = 0;
 
 	if (domain & NOUVEAU_GEM_DOMAIN_VRAM) {
-		struct nvif_mmu *mmu = &drm->client.mmu;
-		const u8 type = mmu->type[drm->ttm.type_vram].type;
-
 		pl[*n].mem_type = TTM_PL_VRAM;
-		pl[*n].flags = flags & ~TTM_PL_FLAG_CACHED;
-
-		/* Some BARs do not support being ioremapped WC */
-		if (drm->client.device.info.family >= NV_DEVICE_INFO_V0_TESLA &&
-		    type & NVIF_MEM_UNCACHED)
-			pl[*n].flags &= ~TTM_PL_FLAG_WC;
-
+		pl[*n].flags = 0;
 		(*n)++;
 	}
 	if (domain & NOUVEAU_GEM_DOMAIN_GART) {
 		pl[*n].mem_type = TTM_PL_TT;
-		pl[*n].flags = flags;
-
-		if (drm->agp.bridge)
-			pl[*n].flags &= ~TTM_PL_FLAG_CACHED;
-
+		pl[*n].flags = 0;
 		(*n)++;
 	}
 	if (domain & NOUVEAU_GEM_DOMAIN_CPU) {
 		pl[*n].mem_type = TTM_PL_SYSTEM;
-		pl[(*n)++].flags = flags;
+		pl[(*n)++].flags = 0;
 	}
 }
 
@@ -415,18 +401,14 @@ void
 nouveau_bo_placement_set(struct nouveau_bo *nvbo, uint32_t domain,
 			 uint32_t busy)
 {
-	struct nouveau_drm *drm = nouveau_bdev(nvbo->bo.bdev);
 	struct ttm_placement *pl = &nvbo->placement;
-	uint32_t flags = nvbo->force_coherent ? TTM_PL_FLAG_UNCACHED :
-						TTM_PL_MASK_CACHING;
 
 	pl->placement = nvbo->placements;
-	set_placement_list(drm, nvbo->placements, &pl->num_placement,
-			   domain, flags);
+	set_placement_list(nvbo->placements, &pl->num_placement, domain);
 
 	pl->busy_placement = nvbo->busy_placements;
-	set_placement_list(drm, nvbo->busy_placements, &pl->num_busy_placement,
-			   domain | busy, flags);
+	set_placement_list(nvbo->busy_placements, &pl->num_busy_placement,
+			   domain | busy);
 
 	set_placement_range(nvbo, domain);
 }
@@ -888,7 +870,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict,
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_TT,
-		.flags = TTM_PL_MASK_CACHING
+		.flags = 0
 	};
 	struct ttm_placement placement;
 	struct ttm_resource tmp_reg;
@@ -930,7 +912,7 @@ nouveau_bo_move_flips(struct ttm_buffer_object *bo, bool evict,
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_TT,
-		.flags = TTM_PL_MASK_CACHING
+		.flags = 0
 	};
 	struct ttm_placement placement;
 	struct ttm_resource tmp_reg;
diff --git a/drivers/gpu/drm/qxl/qxl_object.c b/drivers/gpu/drm/qxl/qxl_object.c
index 940e99354f49..547d46c14d56 100644
--- a/drivers/gpu/drm/qxl/qxl_object.c
+++ b/drivers/gpu/drm/qxl/qxl_object.c
@@ -64,21 +64,21 @@ void qxl_ttm_placement_from_domain(struct qxl_bo *qbo, u32 domain)
 	qbo->placement.busy_placement = qbo->placements;
 	if (domain == QXL_GEM_DOMAIN_VRAM) {
 		qbo->placements[c].mem_type = TTM_PL_VRAM;
-		qbo->placements[c++].flags = TTM_PL_FLAG_CACHED | pflag;
+		qbo->placements[c++].flags = pflag;
 	}
 	if (domain == QXL_GEM_DOMAIN_SURFACE) {
 		qbo->placements[c].mem_type = TTM_PL_PRIV;
-		qbo->placements[c++].flags = TTM_PL_FLAG_CACHED | pflag;
+		qbo->placements[c++].flags = pflag;
 		qbo->placements[c].mem_type = TTM_PL_VRAM;
-		qbo->placements[c++].flags = TTM_PL_FLAG_CACHED | pflag;
+		qbo->placements[c++].flags = pflag;
 	}
 	if (domain == QXL_GEM_DOMAIN_CPU) {
 		qbo->placements[c].mem_type = TTM_PL_SYSTEM;
-		qbo->placements[c++].flags = TTM_PL_MASK_CACHING | pflag;
+		qbo->placements[c++].flags = pflag;
 	}
 	if (!c) {
 		qbo->placements[c].mem_type = TTM_PL_SYSTEM;
-		qbo->placements[c++].flags = TTM_PL_MASK_CACHING;
+		qbo->placements[c++].flags = 0;
 	}
 	qbo->placement.num_placement = c;
 	qbo->placement.num_busy_placement = c;
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index 61eb06dbbce8..e3ed20215f18 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -56,7 +56,7 @@ static void qxl_evict_flags(struct ttm_buffer_object *bo,
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_SYSTEM,
-		.flags = TTM_PL_MASK_CACHING
+		.flags = 0
 	};
 
 	if (!qxl_ttm_bo_is_qxl_bo(bo)) {
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index ad0e6e9ef922..ab81e35cb060 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -113,57 +113,29 @@ void radeon_ttm_placement_from_domain(struct radeon_bo *rbo, u32 domain)
 			rbo->placements[c].fpfn =
 				rbo->rdev->mc.visible_vram_size >> PAGE_SHIFT;
 			rbo->placements[c].mem_type = TTM_PL_VRAM;
-			rbo->placements[c++].flags = TTM_PL_FLAG_WC |
-						     TTM_PL_FLAG_UNCACHED;
+			rbo->placements[c++].flags = 0;
 		}
 
 		rbo->placements[c].fpfn = 0;
 		rbo->placements[c].mem_type = TTM_PL_VRAM;
-		rbo->placements[c++].flags = TTM_PL_FLAG_WC |
-					     TTM_PL_FLAG_UNCACHED;
+		rbo->placements[c++].flags = 0;
 	}
 
 	if (domain & RADEON_GEM_DOMAIN_GTT) {
-		if (rbo->flags & RADEON_GEM_GTT_UC) {
-			rbo->placements[c].fpfn = 0;
-			rbo->placements[c].mem_type = TTM_PL_TT;
-			rbo->placements[c++].flags = TTM_PL_FLAG_UNCACHED;
-
-		} else if ((rbo->flags & RADEON_GEM_GTT_WC) ||
-			   (rbo->rdev->flags & RADEON_IS_AGP)) {
-			rbo->placements[c].fpfn = 0;
-			rbo->placements[c].mem_type = TTM_PL_TT;
-			rbo->placements[c++].flags = TTM_PL_FLAG_WC |
-				TTM_PL_FLAG_UNCACHED;
-		} else {
-			rbo->placements[c].fpfn = 0;
-			rbo->placements[c].mem_type = TTM_PL_TT;
-			rbo->placements[c++].flags = TTM_PL_FLAG_CACHED;
-		}
+		rbo->placements[c].fpfn = 0;
+		rbo->placements[c].mem_type = TTM_PL_TT;
+		rbo->placements[c++].flags = 0;
 	}
 
 	if (domain & RADEON_GEM_DOMAIN_CPU) {
-		if (rbo->flags & RADEON_GEM_GTT_UC) {
-			rbo->placements[c].fpfn = 0;
-			rbo->placements[c].mem_type = TTM_PL_SYSTEM;
-			rbo->placements[c++].flags = TTM_PL_FLAG_UNCACHED;
-
-		} else if ((rbo->flags & RADEON_GEM_GTT_WC) ||
-		    rbo->rdev->flags & RADEON_IS_AGP) {
-			rbo->placements[c].fpfn = 0;
-			rbo->placements[c].mem_type = TTM_PL_SYSTEM;
-			rbo->placements[c++].flags = TTM_PL_FLAG_WC |
-				TTM_PL_FLAG_UNCACHED;
-		} else {
-			rbo->placements[c].fpfn = 0;
-			rbo->placements[c].mem_type = TTM_PL_SYSTEM;
-			rbo->placements[c++].flags = TTM_PL_FLAG_CACHED;
-		}
+		rbo->placements[c].fpfn = 0;
+		rbo->placements[c].mem_type = TTM_PL_SYSTEM;
+		rbo->placements[c++].flags = 0;
 	}
 	if (!c) {
 		rbo->placements[c].fpfn = 0;
 		rbo->placements[c].mem_type = TTM_PL_SYSTEM;
-		rbo->placements[c++].flags = TTM_PL_MASK_CACHING;
+		rbo->placements[c++].flags = 0;
 	}
 
 	rbo->placement.num_placement = c;
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 9b53a1d80632..d6f42fbc81f4 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -89,7 +89,7 @@ static void radeon_evict_flags(struct ttm_buffer_object *bo,
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_SYSTEM,
-		.flags = TTM_PL_MASK_CACHING
+		.flags = 0
 	};
 
 	struct radeon_bo *rbo;
@@ -225,17 +225,12 @@ static int radeon_move_vram_ram(struct ttm_buffer_object *bo,
 	placements.fpfn = 0;
 	placements.lpfn = 0;
 	placements.mem_type = TTM_PL_TT;
-	placements.flags = TTM_PL_MASK_CACHING;
+	placements.flags = 0;
 	r = ttm_bo_mem_space(bo, &placement, &tmp_mem, ctx);
 	if (unlikely(r)) {
 		return r;
 	}
 
-	r = ttm_tt_set_placement_caching(bo->ttm, tmp_mem.placement);
-	if (unlikely(r)) {
-		goto out_cleanup;
-	}
-
 	r = ttm_tt_populate(bo->bdev, bo->ttm, ctx);
 	if (unlikely(r)) {
 		goto out_cleanup;
@@ -275,7 +270,7 @@ static int radeon_move_ram_vram(struct ttm_buffer_object *bo,
 	placements.fpfn = 0;
 	placements.lpfn = 0;
 	placements.mem_type = TTM_PL_TT;
-	placements.flags = TTM_PL_MASK_CACHING;
+	placements.flags = 0;
 	r = ttm_bo_mem_space(bo, &placement, &tmp_mem, ctx);
 	if (unlikely(r)) {
 		return r;
@@ -389,12 +384,7 @@ static int radeon_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_reso
 		 * Alpha: use bus.addr to hold the ioremap() return,
 		 * so we can modify bus.base below.
 		 */
-		if (mem->placement & TTM_PL_FLAG_WC)
-			mem->bus.addr =
-				ioremap_wc(mem->bus.offset, bus_size);
-		else
-			mem->bus.addr =
-				ioremap(mem->bus.offset, bus_size);
+		mem->bus.addr = ioremap_wc(mem->bus.offset, bus_size);
 		if (!mem->bus.addr)
 			return -ENOMEM;
 
diff --git a/drivers/gpu/drm/ttm/ttm_agp_backend.c b/drivers/gpu/drm/ttm/ttm_agp_backend.c
index a723062d37e7..4f76c9287159 100644
--- a/drivers/gpu/drm/ttm/ttm_agp_backend.c
+++ b/drivers/gpu/drm/ttm/ttm_agp_backend.c
@@ -54,7 +54,7 @@ int ttm_agp_bind(struct ttm_tt *ttm, struct ttm_resource *bo_mem)
 	struct page *dummy_read_page = ttm_bo_glob.dummy_read_page;
 	struct drm_mm_node *node = bo_mem->mm_node;
 	struct agp_memory *mem;
-	int ret, cached = (bo_mem->placement & TTM_PL_FLAG_CACHED);
+	int ret, cached = ttm->caching == ttm_cached;
 	unsigned i;
 
 	if (agp_be->mem)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 3732dcb58aad..b97ed6ca8765 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -252,10 +252,6 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 		if (ret)
 			goto out_err;
 
-		ret = ttm_tt_set_placement_caching(bo->ttm, mem->placement);
-		if (ret)
-			goto out_err;
-
 		if (mem->mem_type != TTM_PL_SYSTEM) {
 			ret = ttm_tt_populate(bdev, bo->ttm, ctx);
 			if (ret)
@@ -843,29 +839,6 @@ static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
 	return ttm_bo_add_move_fence(bo, man, mem, ctx->no_wait_gpu);
 }
 
-static uint32_t ttm_bo_select_caching(struct ttm_resource_manager *man,
-				      uint32_t cur_placement,
-				      uint32_t proposed_placement)
-{
-	uint32_t caching = proposed_placement & TTM_PL_MASK_CACHING;
-	uint32_t result = proposed_placement & ~TTM_PL_MASK_CACHING;
-
-	/**
-	 * Keep current caching if possible.
-	 */
-
-	if ((cur_placement & caching) != 0)
-		result |= (cur_placement & caching);
-	else if ((TTM_PL_FLAG_CACHED & caching) != 0)
-		result |= TTM_PL_FLAG_CACHED;
-	else if ((TTM_PL_FLAG_WC & caching) != 0)
-		result |= TTM_PL_FLAG_WC;
-	else if ((TTM_PL_FLAG_UNCACHED & caching) != 0)
-		result |= TTM_PL_FLAG_UNCACHED;
-
-	return result;
-}
-
 /**
  * ttm_bo_mem_placement - check if placement is compatible
  * @bo: BO to find memory for
@@ -884,18 +857,13 @@ static int ttm_bo_mem_placement(struct ttm_buffer_object *bo,
 {
 	struct ttm_bo_device *bdev = bo->bdev;
 	struct ttm_resource_manager *man;
-	uint32_t cur_flags = 0;
 
 	man = ttm_manager_type(bdev, place->mem_type);
 	if (!man || !ttm_resource_manager_used(man))
 		return -EBUSY;
 
-	cur_flags = ttm_bo_select_caching(man, bo->mem.placement,
-					  place->flags);
-	cur_flags |= place->flags & ~TTM_PL_MASK_CACHING;
-
 	mem->mem_type = place->mem_type;
-	mem->placement = cur_flags;
+	mem->placement = place->flags;
 
 	spin_lock(&ttm_bo_glob.lru_lock);
 	ttm_bo_del_from_lru(bo);
@@ -1028,8 +996,7 @@ static bool ttm_bo_places_compat(const struct ttm_place *places,
 			continue;
 
 		*new_flags = heap->flags;
-		if ((*new_flags & mem->placement & TTM_PL_MASK_CACHING) &&
-		    (mem->mem_type == heap->mem_type) &&
+		if ((mem->mem_type == heap->mem_type) &&
 		    (!(*new_flags & TTM_PL_FLAG_CONTIGUOUS) ||
 		     (mem->placement & TTM_PL_FLAG_CONTIGUOUS)))
 			return true;
@@ -1083,9 +1050,6 @@ int ttm_bo_validate(struct ttm_buffer_object *bo,
 		ret = ttm_bo_move_buffer(bo, placement, ctx);
 		if (ret)
 			return ret;
-	} else {
-		bo->mem.placement &= TTM_PL_MASK_CACHING;
-		bo->mem.placement |= new_flags & ~TTM_PL_MASK_CACHING;
 	}
 	/*
 	 * We might need to add a TTM.
@@ -1153,7 +1117,7 @@ int ttm_bo_init_reserved(struct ttm_bo_device *bdev,
 	bo->mem.bus.offset = 0;
 	bo->mem.bus.addr = NULL;
 	bo->moving = NULL;
-	bo->mem.placement = TTM_PL_FLAG_CACHED;
+	bo->mem.placement = 0;
 	bo->acc_size = acc_size;
 	bo->pin_count = 0;
 	bo->sg = sg;
@@ -1484,7 +1448,7 @@ int ttm_bo_swapout(struct ttm_operation_ctx *ctx)
 
 		evict_mem = bo->mem;
 		evict_mem.mm_node = NULL;
-		evict_mem.placement = TTM_PL_MASK_CACHING;
+		evict_mem.placement = 0;
 		evict_mem.mem_type = TTM_PL_SYSTEM;
 
 		ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, &ctx);
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 0542097dc419..ba7ab5ed85d0 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -72,10 +72,6 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 		old_mem->mem_type = TTM_PL_SYSTEM;
 	}
 
-	ret = ttm_tt_set_placement_caching(ttm, new_mem->placement);
-	if (unlikely(ret != 0))
-		return ret;
-
 	if (new_mem->mem_type != TTM_PL_SYSTEM) {
 
 		ret = ttm_tt_populate(bo->bdev, ttm, ctx);
@@ -135,7 +131,7 @@ static int ttm_resource_ioremap(struct ttm_bo_device *bdev,
 	} else {
 		size_t bus_size = (size_t)mem->num_pages << PAGE_SHIFT;
 
-		if (mem->placement & TTM_PL_FLAG_WC)
+		if (mem->bus.caching == ttm_write_combined)
 			addr = ioremap_wc(mem->bus.offset, bus_size);
 		else
 			addr = ioremap(mem->bus.offset, bus_size);
@@ -427,7 +423,7 @@ static int ttm_bo_ioremap(struct ttm_buffer_object *bo,
 		map->virtual = (void *)(((u8 *)bo->mem.bus.addr) + offset);
 	} else {
 		map->bo_kmap_type = ttm_bo_map_iomap;
-		if (mem->placement & TTM_PL_FLAG_WC)
+		if (mem->bus.caching == ttm_write_combined)
 			map->virtual = ioremap_wc(bo->mem.bus.offset + offset,
 						  size);
 		else
@@ -457,7 +453,7 @@ static int ttm_bo_kmap_ttm(struct ttm_buffer_object *bo,
 	if (ret)
 		return ret;
 
-	if (num_pages == 1 && (mem->placement & TTM_PL_FLAG_CACHED)) {
+	if (num_pages == 1 && ttm->caching == ttm_cached) {
 		/*
 		 * We're mapping a single page, and the desired
 		 * page protection is consistent with the bo.
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index a465f51df027..3e5dd6271d4c 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -114,35 +114,6 @@ static int ttm_sg_tt_alloc_page_directory(struct ttm_dma_tt *ttm)
 	return 0;
 }
 
-static int ttm_tt_set_caching(struct ttm_tt *ttm, enum ttm_caching caching)
-{
-	if (ttm->caching == caching)
-		return 0;
-
-	/* Can't change the caching state after TT is populated */
-	if (WARN_ON_ONCE(ttm_tt_is_populated(ttm)))
-		return -EINVAL;
-
-	ttm->caching = caching;
-
-	return 0;
-}
-
-int ttm_tt_set_placement_caching(struct ttm_tt *ttm, uint32_t placement)
-{
-	enum ttm_caching state;
-
-	if (placement & TTM_PL_FLAG_WC)
-		state = ttm_write_combined;
-	else if (placement & TTM_PL_FLAG_UNCACHED)
-		state = ttm_uncached;
-	else
-		state = ttm_cached;
-
-	return ttm_tt_set_caching(ttm, state);
-}
-EXPORT_SYMBOL(ttm_tt_set_placement_caching);
-
 void ttm_tt_destroy_common(struct ttm_bo_device *bdev, struct ttm_tt *ttm)
 {
 	ttm_tt_unpopulate(bdev, ttm);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index fae88969a15a..112253246f08 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -34,28 +34,28 @@ static const struct ttm_place vram_placement_flags = {
 	.fpfn = 0,
 	.lpfn = 0,
 	.mem_type = TTM_PL_VRAM,
-	.flags = TTM_PL_FLAG_CACHED
+	.flags = 0
 };
 
 static const struct ttm_place sys_placement_flags = {
 	.fpfn = 0,
 	.lpfn = 0,
 	.mem_type = TTM_PL_SYSTEM,
-	.flags = TTM_PL_FLAG_CACHED
+	.flags = 0
 };
 
 static const struct ttm_place gmr_placement_flags = {
 	.fpfn = 0,
 	.lpfn = 0,
 	.mem_type = VMW_PL_GMR,
-	.flags = TTM_PL_FLAG_CACHED
+	.flags = 0
 };
 
 static const struct ttm_place mob_placement_flags = {
 	.fpfn = 0,
 	.lpfn = 0,
 	.mem_type = VMW_PL_MOB,
-	.flags = TTM_PL_FLAG_CACHED
+	.flags = 0
 };
 
 struct ttm_placement vmw_vram_placement = {
@@ -70,12 +70,12 @@ static const struct ttm_place vram_gmr_placement_flags[] = {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_VRAM,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}, {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = VMW_PL_GMR,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}
 };
 
@@ -84,12 +84,12 @@ static const struct ttm_place gmr_vram_placement_flags[] = {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = VMW_PL_GMR,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}, {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_VRAM,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}
 };
 
@@ -119,22 +119,22 @@ static const struct ttm_place evictable_placement_flags[] = {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_SYSTEM,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}, {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_VRAM,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}, {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = VMW_PL_GMR,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}, {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = VMW_PL_MOB,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}
 };
 
@@ -143,17 +143,17 @@ static const struct ttm_place nonfixed_placement_flags[] = {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = TTM_PL_SYSTEM,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}, {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = VMW_PL_GMR,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}, {
 		.fpfn = 0,
 		.lpfn = 0,
 		.mem_type = VMW_PL_MOB,
-		.flags = TTM_PL_FLAG_CACHED
+		.flags = 0
 	}
 };
 
diff --git a/include/drm/ttm/ttm_placement.h b/include/drm/ttm/ttm_placement.h
index 50e72df48b8d..aa6ba4d0cf78 100644
--- a/include/drm/ttm/ttm_placement.h
+++ b/include/drm/ttm/ttm_placement.h
@@ -43,27 +43,13 @@
 #define TTM_PL_PRIV             3
 
 /*
- * Other flags that affects data placement.
- * TTM_PL_FLAG_CACHED indicates cache-coherent mappings
- * if available.
- * TTM_PL_FLAG_SHARED means that another application may
- * reference the buffer.
- * TTM_PL_FLAG_NO_EVICT means that the buffer may never
- * be evicted to make room for other buffers.
  * TTM_PL_FLAG_TOPDOWN requests to be placed from the
  * top of the memory area, instead of the bottom.
  */
 
-#define TTM_PL_FLAG_CACHED      (1 << 16)
-#define TTM_PL_FLAG_UNCACHED    (1 << 17)
-#define TTM_PL_FLAG_WC          (1 << 18)
 #define TTM_PL_FLAG_CONTIGUOUS  (1 << 19)
 #define TTM_PL_FLAG_TOPDOWN     (1 << 22)
 
-#define TTM_PL_MASK_CACHING     (TTM_PL_FLAG_CACHED | \
-				 TTM_PL_FLAG_UNCACHED | \
-				 TTM_PL_FLAG_WC)
-
 /**
  * struct ttm_place
  *
diff --git a/include/drm/ttm/ttm_tt.h b/include/drm/ttm/ttm_tt.h
index c39c722d5184..e042dec5e6c1 100644
--- a/include/drm/ttm/ttm_tt.h
+++ b/include/drm/ttm/ttm_tt.h
@@ -164,21 +164,6 @@ void ttm_tt_destroy_common(struct ttm_bo_device *bdev, struct ttm_tt *ttm);
  * Swap in a previously swap out ttm_tt.
  */
 int ttm_tt_swapin(struct ttm_tt *ttm);
-
-/**
- * ttm_tt_set_placement_caching:
- *
- * @ttm A struct ttm_tt the backing pages of which will change caching policy.
- * @placement: Flag indicating the desired caching policy.
- *
- * This function will change caching policy of any default kernel mappings of
- * the pages backing @ttm. If changing from cached to uncached or
- * write-combined,
- * all CPU caches will first be flushed to make sure the data of the pages
- * hit RAM. This function may be very costly as it involves global TLB
- * and cache flushes and potential page splitting / combining.
- */
-int ttm_tt_set_placement_caching(struct ttm_tt *ttm, uint32_t placement);
 int ttm_tt_swapout(struct ttm_bo_device *bdev, struct ttm_tt *ttm);
 
 /**
-- 
cgit 


From c37d951cb42aa340513c0bc2df10b7324fa0d856 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Mon, 19 Oct 2020 17:13:13 +1000
Subject: drm/ttm: add move old to system to drivers.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Uninline ttm_bo_move_ttm. Eventually want to unhook the unbind out.

Reviewed-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201019071314.1671485-5-airlied@gmail.com
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    | 15 ++++++++++++---
 drivers/gpu/drm/nouveau/nouveau_bo.c       | 11 +++++++++--
 drivers/gpu/drm/radeon/radeon_ttm.c        | 15 ++++++++++++---
 drivers/gpu/drm/ttm/ttm_bo_util.c          |  5 +++--
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  7 ++++++-
 include/drm/ttm/ttm_bo_driver.h            |  2 ++
 6 files changed, 44 insertions(+), 11 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index f8c9d66d3ef7..00bec9924b5b 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -562,7 +562,11 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict,
 	}
 
 	/* move BO (in tmp_mem) to new_mem */
-	r = ttm_bo_move_ttm(bo, ctx, new_mem);
+	r = ttm_bo_move_to_system(bo, ctx);
+	if (unlikely(r))
+		goto out_cleanup;
+
+	ttm_bo_assign_mem(bo, new_mem);
 out_cleanup:
 	ttm_resource_free(bo, &tmp_mem);
 	return r;
@@ -672,8 +676,13 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 	}
 
 	if (old_mem->mem_type == TTM_PL_TT &&
-	    new_mem->mem_type == TTM_PL_SYSTEM)
-		return ttm_bo_move_ttm(bo, ctx, new_mem);
+	    new_mem->mem_type == TTM_PL_SYSTEM) {
+		r = ttm_bo_move_to_system(bo, ctx);
+		if (r)
+			return r;
+		ttm_bo_assign_mem(bo, new_mem);
+		return 0;
+	}
 
 	if (old_mem->mem_type == AMDGPU_PL_GDS ||
 	    old_mem->mem_type == AMDGPU_PL_GWS ||
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 63baa9367851..ec79c3b251e8 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -897,7 +897,11 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict,
 	if (ret)
 		goto out;
 
-	ret = ttm_bo_move_ttm(bo, ctx, new_reg);
+	ret = ttm_bo_move_to_system(bo, ctx);
+	if (ret)
+		goto out;
+
+	ttm_bo_assign_mem(bo, &tmp_reg);
 out:
 	ttm_resource_free(bo, &tmp_reg);
 	return ret;
@@ -1048,7 +1052,10 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict,
 
 	if (old_reg->mem_type == TTM_PL_TT &&
 	    new_reg->mem_type == TTM_PL_SYSTEM) {
-		ret = ttm_bo_move_ttm(bo, ctx, new_reg);
+		ret = ttm_bo_move_to_system(bo, ctx);
+		if (ret)
+			return ret;
+		ttm_bo_assign_mem(bo, new_reg);
 		goto out;
 	}
 
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 9437d29248fd..426643120e3a 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -244,7 +244,11 @@ static int radeon_move_vram_ram(struct ttm_buffer_object *bo,
 	if (unlikely(r)) {
 		goto out_cleanup;
 	}
-	r = ttm_bo_move_ttm(bo, ctx, new_mem);
+	r = ttm_bo_move_to_system(bo, ctx);
+	if (unlikely(r))
+		goto out_cleanup;
+
+	ttm_bo_assign_mem(bo, new_mem);
 out_cleanup:
 	ttm_resource_free(bo, &tmp_mem);
 	return r;
@@ -319,9 +323,14 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, bool evict,
 	}
 
 	if (old_mem->mem_type == TTM_PL_TT &&
-	    new_mem->mem_type == TTM_PL_SYSTEM)
-		return ttm_bo_move_ttm(bo, ctx, new_mem);
+	    new_mem->mem_type == TTM_PL_SYSTEM) {
+		r = ttm_bo_move_to_system(bo, ctx);
+		if (r)
+			return r;
 
+		ttm_bo_assign_mem(bo, new_mem);
+		return 0;
+	}
 	if (!rdev->ring[radeon_copy_ring_index(rdev)].ready ||
 	    rdev->asic->copy.copy == NULL) {
 		/* use memcpy */
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 7ea0482c1aee..4a74a297e529 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -66,8 +66,8 @@ int ttm_bo_move_to_new_tt_mem(struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_bo_move_to_new_tt_mem);
 
-static int ttm_bo_move_to_system(struct ttm_buffer_object *bo,
-				 struct ttm_operation_ctx *ctx)
+int ttm_bo_move_to_system(struct ttm_buffer_object *bo,
+			  struct ttm_operation_ctx *ctx)
 {
 	struct ttm_resource *old_mem = &bo->mem;
 	int ret;
@@ -87,6 +87,7 @@ static int ttm_bo_move_to_system(struct ttm_buffer_object *bo,
 	old_mem->mem_type = TTM_PL_SYSTEM;
 	return 0;
 }
+EXPORT_SYMBOL(ttm_bo_move_to_system);
 
 int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 		   struct ttm_operation_ctx *ctx,
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 112253246f08..c5cf81c09971 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -735,13 +735,18 @@ static int vmw_move(struct ttm_buffer_object *bo,
 {
 	struct ttm_resource_manager *old_man = ttm_manager_type(bo->bdev, bo->mem.mem_type);
 	struct ttm_resource_manager *new_man = ttm_manager_type(bo->bdev, new_mem->mem_type);
+	int ret;
 
 	if (old_man->use_tt && new_man->use_tt) {
 		if (bo->mem.mem_type == TTM_PL_SYSTEM) {
 			ttm_bo_assign_mem(bo, new_mem);
 			return 0;
 		}
-		return ttm_bo_move_ttm(bo, ctx, new_mem);
+		ret = ttm_bo_move_to_system(bo, ctx);
+		if (ret)
+			return ret;
+		ttm_bo_assign_mem(bo, new_mem);
+		return 0;
 	} else {
 		return ttm_bo_move_memcpy(bo, ctx, new_mem);
 	}
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 1f4d2b1febd0..9e60e6814ba1 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -596,6 +596,8 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 int ttm_bo_move_to_new_tt_mem(struct ttm_buffer_object *bo,
 			      struct ttm_operation_ctx *ctx,
 			      struct ttm_resource *new_mem);
+int ttm_bo_move_to_system(struct ttm_buffer_object *bo,
+			  struct ttm_operation_ctx *ctx);
 
 /**
  * ttm_bo_move_memcpy
-- 
cgit 


From 29a1d482e4044ab76d0c0f6341212f1a51f48236 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Tue, 20 Oct 2020 11:03:15 +1000
Subject: drm/ttm: add move to system into drivers
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This moves the to system move into the drivers, and moves all
the unbinds in the move path under driver control

Note: radeon/nouveau already wait so don't duplicate it.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201020010319.1692445-4-airlied@gmail.com
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    | 12 +++++++++---
 drivers/gpu/drm/nouveau/nouveau_bo.c       | 10 ++++++----
 drivers/gpu/drm/radeon/radeon_ttm.c        | 12 +++++++-----
 drivers/gpu/drm/ttm/ttm_bo_util.c          | 21 ---------------------
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  5 ++++-
 include/drm/ttm/ttm_bo_driver.h            |  2 --
 6 files changed, 26 insertions(+), 36 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 11f464e0951d..56f16bdba936 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -66,6 +66,8 @@
 static int amdgpu_ttm_backend_bind(struct ttm_bo_device *bdev,
 				   struct ttm_tt *ttm,
 				   struct ttm_resource *bo_mem);
+static void amdgpu_ttm_backend_unbind(struct ttm_bo_device *bdev,
+				      struct ttm_tt *ttm);
 
 static int amdgpu_ttm_init_on_chip(struct amdgpu_device *adev,
 				    unsigned int type,
@@ -561,11 +563,12 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict,
 		goto out_cleanup;
 	}
 
-	/* move BO (in tmp_mem) to new_mem */
-	r = ttm_bo_move_to_system(bo, ctx);
+	r = ttm_bo_wait_ctx(bo, ctx);
 	if (unlikely(r))
 		goto out_cleanup;
 
+	amdgpu_ttm_backend_unbind(bo->bdev, bo->ttm);
+	ttm_resource_free(bo, &bo->mem);
 	ttm_bo_assign_mem(bo, new_mem);
 out_cleanup:
 	ttm_resource_free(bo, &tmp_mem);
@@ -682,9 +685,12 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 
 	if (old_mem->mem_type == TTM_PL_TT &&
 	    new_mem->mem_type == TTM_PL_SYSTEM) {
-		r = ttm_bo_move_to_system(bo, ctx);
+		r = ttm_bo_wait_ctx(bo, ctx);
 		if (r)
 			return r;
+
+		amdgpu_ttm_backend_unbind(bo->bdev, bo->ttm);
+		ttm_resource_free(bo, &bo->mem);
 		ttm_bo_assign_mem(bo, new_mem);
 		return 0;
 	}
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 526bbf4c399a..1ccd9ea6da95 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -46,6 +46,7 @@
 
 static int nouveau_ttm_tt_bind(struct ttm_bo_device *bdev, struct ttm_tt *ttm,
 			       struct ttm_resource *reg);
+static void nouveau_ttm_tt_unbind(struct ttm_bo_device *bdev, struct ttm_tt *ttm);
 
 /*
  * NV10-NV40 tiling helpers
@@ -897,10 +898,12 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict,
 	if (ret)
 		goto out;
 
-	ret = ttm_bo_move_to_system(bo, ctx);
+	ret = ttm_bo_wait_ctx(bo, ctx);
 	if (ret)
 		goto out;
 
+	nouveau_ttm_tt_unbind(bo->bdev, bo->ttm);
+	ttm_resource_free(bo, &bo->mem);
 	ttm_bo_assign_mem(bo, &tmp_reg);
 out:
 	ttm_resource_free(bo, &tmp_reg);
@@ -1056,9 +1059,8 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict,
 
 	if (old_reg->mem_type == TTM_PL_TT &&
 	    new_reg->mem_type == TTM_PL_SYSTEM) {
-		ret = ttm_bo_move_to_system(bo, ctx);
-		if (ret)
-			return ret;
+		nouveau_ttm_tt_unbind(bo->bdev, bo->ttm);
+		ttm_resource_free(bo, &bo->mem);
 		ttm_bo_assign_mem(bo, new_reg);
 		goto out;
 	}
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index c1cdee33be11..42c85afe7955 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -59,6 +59,8 @@ static void radeon_ttm_debugfs_fini(struct radeon_device *rdev);
 static int radeon_ttm_tt_bind(struct ttm_bo_device *bdev,
 			      struct ttm_tt *ttm,
 			      struct ttm_resource *bo_mem);
+static void radeon_ttm_tt_unbind(struct ttm_bo_device *bdev,
+				 struct ttm_tt *ttm);
 
 struct radeon_device *radeon_get_rdev(struct ttm_bo_device *bdev)
 {
@@ -244,10 +246,12 @@ static int radeon_move_vram_ram(struct ttm_buffer_object *bo,
 	if (unlikely(r)) {
 		goto out_cleanup;
 	}
-	r = ttm_bo_move_to_system(bo, ctx);
+	r = ttm_bo_wait_ctx(bo, ctx);
 	if (unlikely(r))
 		goto out_cleanup;
 
+	radeon_ttm_tt_unbind(bo->bdev, bo->ttm);
+	ttm_resource_free(bo, &bo->mem);
 	ttm_bo_assign_mem(bo, new_mem);
 out_cleanup:
 	ttm_resource_free(bo, &tmp_mem);
@@ -329,10 +333,8 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, bool evict,
 
 	if (old_mem->mem_type == TTM_PL_TT &&
 	    new_mem->mem_type == TTM_PL_SYSTEM) {
-		r = ttm_bo_move_to_system(bo, ctx);
-		if (r)
-			return r;
-
+		radeon_ttm_tt_unbind(bo->bdev, bo->ttm);
+		ttm_resource_free(bo, &bo->mem);
 		ttm_bo_assign_mem(bo, new_mem);
 		return 0;
 	}
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 5cfcb92a6d30..65f042d6da7c 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -65,27 +65,6 @@ int ttm_bo_move_to_new_tt_mem(struct ttm_buffer_object *bo,
 	return 0;
 }
 
-int ttm_bo_move_to_system(struct ttm_buffer_object *bo,
-			  struct ttm_operation_ctx *ctx)
-{
-	int ret;
-
-	if (bo->mem.mem_type == TTM_PL_SYSTEM)
-		return 0;
-
-	ret = ttm_bo_wait_ctx(bo, ctx);
-	if (unlikely(ret != 0)) {
-		if (ret != -ERESTARTSYS)
-			pr_err("Failed to expire sync object before unbinding TTM\n");
-		return ret;
-	}
-
-	ttm_bo_tt_unbind(bo);
-	ttm_resource_free(bo, &bo->mem);
-	return 0;
-}
-EXPORT_SYMBOL(ttm_bo_move_to_system);
-
 int ttm_mem_io_reserve(struct ttm_bo_device *bdev,
 		       struct ttm_resource *mem)
 {
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index c5cf81c09971..1d220a9794e6 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -742,9 +742,12 @@ static int vmw_move(struct ttm_buffer_object *bo,
 			ttm_bo_assign_mem(bo, new_mem);
 			return 0;
 		}
-		ret = ttm_bo_move_to_system(bo, ctx);
+		ret = ttm_bo_wait_ctx(bo, ctx);
 		if (ret)
 			return ret;
+
+		vmw_ttm_unbind(bo->bdev, bo->ttm);
+		ttm_resource_free(bo, &bo->mem);
 		ttm_bo_assign_mem(bo, new_mem);
 		return 0;
 	} else {
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index a89728cb9a23..81a1618b9535 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -574,8 +574,6 @@ void ttm_mem_io_free(struct ttm_bo_device *bdev,
 int ttm_bo_move_to_new_tt_mem(struct ttm_buffer_object *bo,
 			      struct ttm_operation_ctx *ctx,
 			      struct ttm_resource *new_mem);
-int ttm_bo_move_to_system(struct ttm_buffer_object *bo,
-			  struct ttm_operation_ctx *ctx);
 
 /**
  * ttm_bo_move_memcpy
-- 
cgit 


From f227ccc9612f5d2f6315874c93acf3945fd51dfb Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Tue, 20 Oct 2020 11:03:16 +1000
Subject: drm/ttm: drop unbind callback.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The drivers now control this, so drop unbinding.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201020010319.1692445-5-airlied@gmail.com
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    |  1 -
 drivers/gpu/drm/nouveau/nouveau_bo.c       |  1 -
 drivers/gpu/drm/qxl/qxl_ttm.c              |  7 -------
 drivers/gpu/drm/radeon/radeon_ttm.c        |  1 -
 drivers/gpu/drm/ttm/ttm_bo.c               |  5 -----
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  1 -
 include/drm/ttm/ttm_bo_driver.h            | 18 ------------------
 7 files changed, 34 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 56f16bdba936..d7a2f912055d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -1722,7 +1722,6 @@ static struct ttm_bo_driver amdgpu_bo_driver = {
 	.ttm_tt_populate = &amdgpu_ttm_tt_populate,
 	.ttm_tt_unpopulate = &amdgpu_ttm_tt_unpopulate,
 	.ttm_tt_bind = &amdgpu_ttm_backend_bind,
-	.ttm_tt_unbind = &amdgpu_ttm_backend_unbind,
 	.ttm_tt_destroy = &amdgpu_ttm_backend_destroy,
 	.eviction_valuable = amdgpu_ttm_bo_eviction_valuable,
 	.evict_flags = &amdgpu_evict_flags,
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 1ccd9ea6da95..43e6e089319e 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1394,7 +1394,6 @@ struct ttm_bo_driver nouveau_bo_driver = {
 	.ttm_tt_populate = &nouveau_ttm_tt_populate,
 	.ttm_tt_unpopulate = &nouveau_ttm_tt_unpopulate,
 	.ttm_tt_bind = &nouveau_ttm_tt_bind,
-	.ttm_tt_unbind = &nouveau_ttm_tt_unbind,
 	.ttm_tt_destroy = &nouveau_ttm_tt_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = nouveau_bo_evict_flags,
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index e3ed20215f18..95c4f2c7ab79 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -113,12 +113,6 @@ static int qxl_ttm_backend_bind(struct ttm_bo_device *bdev,
 	return -1;
 }
 
-static void qxl_ttm_backend_unbind(struct ttm_bo_device *bdev,
-				   struct ttm_tt *ttm)
-{
-	/* Not implemented */
-}
-
 static void qxl_ttm_backend_destroy(struct ttm_bo_device *bdev,
 				    struct ttm_tt *ttm)
 {
@@ -180,7 +174,6 @@ static struct ttm_bo_driver qxl_bo_driver = {
 	.ttm_tt_create = &qxl_ttm_tt_create,
 	.ttm_tt_bind = &qxl_ttm_backend_bind,
 	.ttm_tt_destroy = &qxl_ttm_backend_destroy,
-	.ttm_tt_unbind = &qxl_ttm_backend_unbind,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = &qxl_evict_flags,
 	.move = &qxl_bo_move,
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 42c85afe7955..914b2ca86936 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -817,7 +817,6 @@ static struct ttm_bo_driver radeon_bo_driver = {
 	.ttm_tt_populate = &radeon_ttm_tt_populate,
 	.ttm_tt_unpopulate = &radeon_ttm_tt_unpopulate,
 	.ttm_tt_bind = &radeon_ttm_tt_bind,
-	.ttm_tt_unbind = &radeon_ttm_tt_unbind,
 	.ttm_tt_destroy = &radeon_ttm_tt_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = &radeon_evict_flags,
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index cbc74a320db2..593e55879019 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -1496,8 +1496,3 @@ int ttm_bo_tt_bind(struct ttm_buffer_object *bo, struct ttm_resource *mem)
 {
 	return bo->bdev->driver->ttm_tt_bind(bo->bdev, bo->ttm, mem);
 }
-
-void ttm_bo_tt_unbind(struct ttm_buffer_object *bo)
-{
-	bo->bdev->driver->ttm_tt_unbind(bo->bdev, bo->ttm);
-}
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 1d220a9794e6..6e07ea982961 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -760,7 +760,6 @@ struct ttm_bo_driver vmw_bo_driver = {
 	.ttm_tt_populate = &vmw_ttm_populate,
 	.ttm_tt_unpopulate = &vmw_ttm_unpopulate,
 	.ttm_tt_bind = &vmw_ttm_bind,
-	.ttm_tt_unbind = &vmw_ttm_unbind,
 	.ttm_tt_destroy = &vmw_ttm_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = vmw_evict_flags,
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 81a1618b9535..fbbcf10670c1 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -104,17 +104,6 @@ struct ttm_bo_driver {
 	 */
 	int (*ttm_tt_bind)(struct ttm_bo_device *bdev, struct ttm_tt *ttm, struct ttm_resource *bo_mem);
 
-	/**
-	 * ttm_tt_unbind
-	 *
-	 * @bdev: Pointer to a ttm device
-	 * @ttm: Pointer to a struct ttm_tt.
-	 *
-	 * Unbind previously bound backend pages. This function should be
-	 * able to handle differences between aperture and system page sizes.
-	 */
-	void (*ttm_tt_unbind)(struct ttm_bo_device *bdev, struct ttm_tt *ttm);
-
 	/**
 	 * ttm_tt_destroy
 	 *
@@ -647,13 +636,6 @@ pgprot_t ttm_io_prot(struct ttm_buffer_object *bo, struct ttm_resource *res,
  */
 int ttm_bo_tt_bind(struct ttm_buffer_object *bo, struct ttm_resource *mem);
 
-/**
- * ttm_bo_tt_bind
- *
- * Unbind the object tt from a memory resource.
- */
-void ttm_bo_tt_unbind(struct ttm_buffer_object *bo);
-
 /**
  * ttm_bo_tt_destroy.
  */
-- 
cgit 


From 6d820003295977f865257f1845bcdebc5dab4fb5 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Tue, 20 Oct 2020 11:03:18 +1000
Subject: drm/ttm: drop move notify around move.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The drivers now do this in the move callback.

move_notify is still needed in the destroy path.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201020010319.1692445-7-airlied@gmail.com
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    | 13 +++++++--
 drivers/gpu/drm/drm_gem_vram_helper.c      | 11 +++++++-
 drivers/gpu/drm/nouveau/nouveau_bo.c       | 12 ++++++--
 drivers/gpu/drm/qxl/qxl_ttm.c              | 45 ++++++++++++++++++------------
 drivers/gpu/drm/radeon/radeon_ttm.c        | 11 ++++++--
 drivers/gpu/drm/ttm/ttm_bo.c               | 12 +-------
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 14 ++++++++--
 7 files changed, 78 insertions(+), 40 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index d7a2f912055d..fba8ada99b0e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -666,6 +666,8 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 	struct ttm_resource *old_mem = &bo->mem;
 	int r;
 
+	amdgpu_bo_move_notify(bo, evict, new_mem);
+
 	/* Can't move a pinned BO */
 	abo = ttm_to_amdgpu_bo(bo);
 	if (WARN_ON_ONCE(abo->tbo.pin_count > 0))
@@ -687,7 +689,7 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 	    new_mem->mem_type == TTM_PL_SYSTEM) {
 		r = ttm_bo_wait_ctx(bo, ctx);
 		if (r)
-			return r;
+			goto fail;
 
 		amdgpu_ttm_backend_unbind(bo->bdev, bo->ttm);
 		ttm_resource_free(bo, &bo->mem);
@@ -728,12 +730,12 @@ memcpy:
 		if (!amdgpu_mem_visible(adev, old_mem) ||
 		    !amdgpu_mem_visible(adev, new_mem)) {
 			pr_err("Move buffer fallback to memcpy unavailable\n");
-			return r;
+			goto fail;
 		}
 
 		r = ttm_bo_move_memcpy(bo, ctx, new_mem);
 		if (r)
-			return r;
+			goto fail;
 	}
 
 	if (bo->type == ttm_bo_type_device &&
@@ -748,6 +750,11 @@ memcpy:
 	/* update statistics */
 	atomic64_add((u64)bo->num_pages << PAGE_SHIFT, &adev->num_bytes_moved);
 	return 0;
+fail:
+	swap(*new_mem, bo->mem);
+	amdgpu_bo_move_notify(bo, false, new_mem);
+	swap(*new_mem, bo->mem);
+	return r;
 }
 
 /**
diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index 7aeb5daf2805..19087b22bdbb 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -590,7 +590,16 @@ static int drm_gem_vram_bo_driver_move(struct drm_gem_vram_object *gbo,
 				       struct ttm_operation_ctx *ctx,
 				       struct ttm_resource *new_mem)
 {
-	return ttm_bo_move_memcpy(&gbo->bo, ctx, new_mem);
+	int ret;
+
+	drm_gem_vram_bo_driver_move_notify(gbo, evict, new_mem);
+	ret = ttm_bo_move_memcpy(&gbo->bo, ctx, new_mem);
+	if (ret) {
+		swap(*new_mem, gbo->bo.mem);
+		drm_gem_vram_bo_driver_move_notify(gbo, false, new_mem);
+		swap(*new_mem, gbo->bo.mem);
+	}
+	return ret;
 }
 
 /*
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 43e6e089319e..ecd16847fbd8 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1032,9 +1032,10 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict,
 	struct nouveau_drm_tile *new_tile = NULL;
 	int ret = 0;
 
+	nouveau_bo_move_ntfy(bo, evict, new_reg);
 	ret = ttm_bo_wait_ctx(bo, ctx);
 	if (ret)
-		return ret;
+		goto out_ntfy;
 
 	if (nvbo->bo.pin_count)
 		NV_WARN(drm, "Moving pinned object %p!\n", nvbo);
@@ -1042,7 +1043,7 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict,
 	if (drm->client.device.info.family < NV_DEVICE_INFO_V0_TESLA) {
 		ret = nouveau_bo_vm_bind(bo, new_reg, &new_tile);
 		if (ret)
-			return ret;
+			goto out_ntfy;
 	}
 
 	/* Fake bo copy. */
@@ -1090,7 +1091,12 @@ out:
 		else
 			nouveau_bo_vm_cleanup(bo, new_tile, &nvbo->tile);
 	}
-
+out_ntfy:
+	if (ret) {
+		swap(*new_reg, bo->mem);
+		nouveau_bo_move_ntfy(bo, false, new_reg);
+		swap(*new_reg, bo->mem);
+	}
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index 95c4f2c7ab79..a6149e3cc3d2 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -136,24 +136,6 @@ static struct ttm_tt *qxl_ttm_tt_create(struct ttm_buffer_object *bo,
 	return ttm;
 }
 
-static int qxl_bo_move(struct ttm_buffer_object *bo, bool evict,
-		       struct ttm_operation_ctx *ctx,
-		       struct ttm_resource *new_mem)
-{
-	struct ttm_resource *old_mem = &bo->mem;
-	int ret;
-
-	ret = ttm_bo_wait_ctx(bo, ctx);
-	if (ret)
-		return ret;
-
-	if (old_mem->mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) {
-		ttm_bo_move_null(bo, new_mem);
-		return 0;
-	}
-	return ttm_bo_move_memcpy(bo, ctx, new_mem);
-}
-
 static void qxl_bo_move_notify(struct ttm_buffer_object *bo,
 			       bool evict,
 			       struct ttm_resource *new_mem)
@@ -170,6 +152,33 @@ static void qxl_bo_move_notify(struct ttm_buffer_object *bo,
 		qxl_surface_evict(qdev, qbo, new_mem ? true : false);
 }
 
+static int qxl_bo_move(struct ttm_buffer_object *bo, bool evict,
+		       struct ttm_operation_ctx *ctx,
+		       struct ttm_resource *new_mem)
+{
+	struct ttm_resource *old_mem = &bo->mem;
+	int ret;
+
+	qxl_bo_move_notify(bo, evict, new_mem);
+
+	ret = ttm_bo_wait_ctx(bo, ctx);
+	if (ret)
+		goto out;
+
+	if (old_mem->mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) {
+		ttm_bo_move_null(bo, new_mem);
+		return 0;
+	}
+	ret = ttm_bo_move_memcpy(bo, ctx, new_mem);
+out:
+	if (ret) {
+		swap(*new_mem, bo->mem);
+		qxl_bo_move_notify(bo, false, new_mem);
+		swap(*new_mem, bo->mem);
+	}
+	return ret;
+}
+
 static struct ttm_bo_driver qxl_bo_driver = {
 	.ttm_tt_create = &qxl_ttm_tt_create,
 	.ttm_tt_bind = &qxl_ttm_backend_bind,
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 914b2ca86936..e9c95ef0859b 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -311,9 +311,11 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, bool evict,
 	struct ttm_resource *old_mem = &bo->mem;
 	int r;
 
+	radeon_bo_move_notify(bo, evict, new_mem);
+
 	r = ttm_bo_wait_ctx(bo, ctx);
 	if (r)
-		return r;
+		goto fail;
 
 	/* Can't move a pinned BO */
 	rbo = container_of(bo, struct radeon_bo, tbo);
@@ -359,13 +361,18 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, bool evict,
 memcpy:
 		r = ttm_bo_move_memcpy(bo, ctx, new_mem);
 		if (r) {
-			return r;
+			goto fail;
 		}
 	}
 
 	/* update statistics */
 	atomic64_add((u64)bo->num_pages << PAGE_SHIFT, &rdev->num_bytes_moved);
 	return 0;
+fail:
+	swap(*new_mem, bo->mem);
+	radeon_bo_move_notify(bo, false, new_mem);
+	swap(*new_mem, bo->mem);
+	return r;
 }
 
 static int radeon_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_resource *mem)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index a41be3dba23d..cd9c2bb183f4 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -263,19 +263,9 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 		}
 	}
 
-	if (bdev->driver->move_notify)
-		bdev->driver->move_notify(bo, evict, mem);
-
 	ret = bdev->driver->move(bo, evict, ctx, mem);
-	if (ret) {
-		if (bdev->driver->move_notify) {
-			swap(*mem, bo->mem);
-			bdev->driver->move_notify(bo, false, mem);
-			swap(*mem, bo->mem);
-		}
-
+	if (ret)
 		goto out_err;
-	}
 
 	ctx->bytes_moved += bo->num_pages << PAGE_SHIFT;
 	return 0;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 6e07ea982961..fd82c9ba2d77 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -737,6 +737,8 @@ static int vmw_move(struct ttm_buffer_object *bo,
 	struct ttm_resource_manager *new_man = ttm_manager_type(bo->bdev, new_mem->mem_type);
 	int ret;
 
+	vmw_move_notify(bo, evict, new_mem);
+
 	if (old_man->use_tt && new_man->use_tt) {
 		if (bo->mem.mem_type == TTM_PL_SYSTEM) {
 			ttm_bo_assign_mem(bo, new_mem);
@@ -744,15 +746,23 @@ static int vmw_move(struct ttm_buffer_object *bo,
 		}
 		ret = ttm_bo_wait_ctx(bo, ctx);
 		if (ret)
-			return ret;
+			goto fail;
 
 		vmw_ttm_unbind(bo->bdev, bo->ttm);
 		ttm_resource_free(bo, &bo->mem);
 		ttm_bo_assign_mem(bo, new_mem);
 		return 0;
 	} else {
-		return ttm_bo_move_memcpy(bo, ctx, new_mem);
+		ret = ttm_bo_move_memcpy(bo, ctx, new_mem);
+		if (ret)
+			goto fail;
 	}
+	return 0;
+fail:
+	swap(*new_mem, bo->mem);
+	vmw_move_notify(bo, false, new_mem);
+	swap(*new_mem, bo->mem);
+	return ret;
 }
 
 struct ttm_bo_driver vmw_bo_driver = {
-- 
cgit 


From bfe5e585b44fb810e33fe54d4d822c6b2b40488c Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Tue, 20 Oct 2020 11:03:19 +1000
Subject: drm/ttm: move last binding into the drivers.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

This moves the call to tt binding into the driver move,
and drops the driver callback.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201020010319.1692445-8-airlied@gmail.com
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    |  7 ++++++-
 drivers/gpu/drm/nouveau/nouveau_bo.c       |  7 ++++++-
 drivers/gpu/drm/qxl/qxl_ttm.c              | 14 --------------
 drivers/gpu/drm/radeon/radeon_ttm.c        |  6 +++++-
 drivers/gpu/drm/ttm/ttm_bo.c               |  4 ----
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  7 ++++++-
 include/drm/ttm/ttm_bo_driver.h            | 14 --------------
 7 files changed, 23 insertions(+), 36 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index fba8ada99b0e..87e10a212b8a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -666,6 +666,12 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 	struct ttm_resource *old_mem = &bo->mem;
 	int r;
 
+	if (new_mem->mem_type == TTM_PL_TT) {
+		r = amdgpu_ttm_backend_bind(bo->bdev, bo->ttm, new_mem);
+		if (r)
+			return r;
+	}
+
 	amdgpu_bo_move_notify(bo, evict, new_mem);
 
 	/* Can't move a pinned BO */
@@ -1728,7 +1734,6 @@ static struct ttm_bo_driver amdgpu_bo_driver = {
 	.ttm_tt_create = &amdgpu_ttm_tt_create,
 	.ttm_tt_populate = &amdgpu_ttm_tt_populate,
 	.ttm_tt_unpopulate = &amdgpu_ttm_tt_unpopulate,
-	.ttm_tt_bind = &amdgpu_ttm_backend_bind,
 	.ttm_tt_destroy = &amdgpu_ttm_backend_destroy,
 	.eviction_valuable = amdgpu_ttm_bo_eviction_valuable,
 	.evict_flags = &amdgpu_evict_flags,
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index ecd16847fbd8..70b6f3b1ae85 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1032,6 +1032,12 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict,
 	struct nouveau_drm_tile *new_tile = NULL;
 	int ret = 0;
 
+	if (new_reg->mem_type == TTM_PL_TT) {
+		ret = nouveau_ttm_tt_bind(bo->bdev, bo->ttm, new_reg);
+		if (ret)
+			return ret;
+	}
+
 	nouveau_bo_move_ntfy(bo, evict, new_reg);
 	ret = ttm_bo_wait_ctx(bo, ctx);
 	if (ret)
@@ -1399,7 +1405,6 @@ struct ttm_bo_driver nouveau_bo_driver = {
 	.ttm_tt_create = &nouveau_ttm_tt_create,
 	.ttm_tt_populate = &nouveau_ttm_tt_populate,
 	.ttm_tt_unpopulate = &nouveau_ttm_tt_unpopulate,
-	.ttm_tt_bind = &nouveau_ttm_tt_bind,
 	.ttm_tt_destroy = &nouveau_ttm_tt_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = nouveau_bo_evict_flags,
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index a6149e3cc3d2..1cc3c14bc684 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -100,19 +100,6 @@ int qxl_ttm_io_mem_reserve(struct ttm_bo_device *bdev,
 /*
  * TTM backend functions.
  */
-
-static int qxl_ttm_backend_bind(struct ttm_bo_device *bdev,
-				struct ttm_tt *ttm,
-				struct ttm_resource *bo_mem)
-{
-	if (!ttm->num_pages) {
-		WARN(1, "nothing to bind %lu pages for mreg %p back %p!\n",
-		     ttm->num_pages, bo_mem, ttm);
-	}
-	/* Not implemented */
-	return -1;
-}
-
 static void qxl_ttm_backend_destroy(struct ttm_bo_device *bdev,
 				    struct ttm_tt *ttm)
 {
@@ -181,7 +168,6 @@ out:
 
 static struct ttm_bo_driver qxl_bo_driver = {
 	.ttm_tt_create = &qxl_ttm_tt_create,
-	.ttm_tt_bind = &qxl_ttm_backend_bind,
 	.ttm_tt_destroy = &qxl_ttm_backend_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = &qxl_evict_flags,
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index e9c95ef0859b..cd454e5c802f 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -311,6 +311,11 @@ static int radeon_bo_move(struct ttm_buffer_object *bo, bool evict,
 	struct ttm_resource *old_mem = &bo->mem;
 	int r;
 
+	if (new_mem->mem_type == TTM_PL_TT) {
+		r = radeon_ttm_tt_bind(bo->bdev, bo->ttm, new_mem);
+		if (r)
+			return r;
+	}
 	radeon_bo_move_notify(bo, evict, new_mem);
 
 	r = ttm_bo_wait_ctx(bo, ctx);
@@ -823,7 +828,6 @@ static struct ttm_bo_driver radeon_bo_driver = {
 	.ttm_tt_create = &radeon_ttm_tt_create,
 	.ttm_tt_populate = &radeon_ttm_tt_populate,
 	.ttm_tt_unpopulate = &radeon_ttm_tt_unpopulate,
-	.ttm_tt_bind = &radeon_ttm_tt_bind,
 	.ttm_tt_destroy = &radeon_ttm_tt_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = &radeon_evict_flags,
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index cd9c2bb183f4..2b578012cdef 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -256,10 +256,6 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 			ret = ttm_tt_populate(bo->bdev, bo->ttm, ctx);
 			if (ret)
 				goto out_err;
-
-			ret = bdev->driver->ttm_tt_bind(bo->bdev, bo->ttm, mem);
-			if (ret)
-				goto out_err;
 		}
 	}
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index fd82c9ba2d77..de25cf016be2 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -737,6 +737,12 @@ static int vmw_move(struct ttm_buffer_object *bo,
 	struct ttm_resource_manager *new_man = ttm_manager_type(bo->bdev, new_mem->mem_type);
 	int ret;
 
+	if (new_man->use_tt && new_mem->mem_type != TTM_PL_SYSTEM) {
+		ret = vmw_ttm_bind(bo->bdev, bo->ttm, new_mem);
+		if (ret)
+			return ret;
+	}
+
 	vmw_move_notify(bo, evict, new_mem);
 
 	if (old_man->use_tt && new_man->use_tt) {
@@ -769,7 +775,6 @@ struct ttm_bo_driver vmw_bo_driver = {
 	.ttm_tt_create = &vmw_ttm_tt_create,
 	.ttm_tt_populate = &vmw_ttm_populate,
 	.ttm_tt_unpopulate = &vmw_ttm_unpopulate,
-	.ttm_tt_bind = &vmw_ttm_bind,
 	.ttm_tt_destroy = &vmw_ttm_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = vmw_evict_flags,
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 0c4efc169f46..72f106b335e9 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -90,20 +90,6 @@ struct ttm_bo_driver {
 	 */
 	void (*ttm_tt_unpopulate)(struct ttm_bo_device *bdev, struct ttm_tt *ttm);
 
-	/**
-	 * ttm_tt_bind
-	 *
-	 * @bdev: Pointer to a ttm device
-	 * @ttm: Pointer to a struct ttm_tt.
-	 * @bo_mem: Pointer to a struct ttm_resource describing the
-	 * memory type and location for binding.
-	 *
-	 * Bind the backend pages into the aperture in the location
-	 * indicated by @bo_mem. This function should be able to handle
-	 * differences between aperture and system page sizes.
-	 */
-	int (*ttm_tt_bind)(struct ttm_bo_device *bdev, struct ttm_tt *ttm, struct ttm_resource *bo_mem);
-
 	/**
 	 * ttm_tt_destroy
 	 *
-- 
cgit 


From 6a6e5988a2657cd0c91f6f1a3e7d194599248b6d Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Wed, 21 Oct 2020 14:40:29 +1000
Subject: drm/ttm: replace last move_notify with delete_mem_notify
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The move notify callback is only used in one place, this should
be removed in the future, but for now just rename it to the use
case which is to notify the driver that the GPU memory is to be
deleted.

Drivers can be cleaned up after this separately.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201021044031.1752624-2-airlied@gmail.com
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    |  8 +++++++-
 drivers/gpu/drm/drm_gem_vram_helper.c      |  8 +++-----
 drivers/gpu/drm/nouveau/nouveau_bo.c       |  8 +++++++-
 drivers/gpu/drm/qxl/qxl_ttm.c              |  7 ++++++-
 drivers/gpu/drm/radeon/radeon_ttm.c        |  8 +++++++-
 drivers/gpu/drm/ttm/ttm_bo.c               |  4 ++--
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  8 +++++++-
 include/drm/ttm/ttm_bo_driver.h            | 10 ++--------
 8 files changed, 41 insertions(+), 20 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 87e10a212b8a..62f9194b1dd1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -1730,6 +1730,12 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo,
 	return ret;
 }
 
+static void
+amdgpu_bo_delete_mem_notify(struct ttm_buffer_object *bo)
+{
+	amdgpu_bo_move_notify(bo, false, NULL);
+}
+
 static struct ttm_bo_driver amdgpu_bo_driver = {
 	.ttm_tt_create = &amdgpu_ttm_tt_create,
 	.ttm_tt_populate = &amdgpu_ttm_tt_populate,
@@ -1739,7 +1745,7 @@ static struct ttm_bo_driver amdgpu_bo_driver = {
 	.evict_flags = &amdgpu_evict_flags,
 	.move = &amdgpu_bo_move,
 	.verify_access = &amdgpu_verify_access,
-	.move_notify = &amdgpu_bo_move_notify,
+	.delete_mem_notify = &amdgpu_bo_delete_mem_notify,
 	.release_notify = &amdgpu_bo_release_notify,
 	.io_mem_reserve = &amdgpu_ttm_io_mem_reserve,
 	.io_mem_pfn = amdgpu_ttm_io_mem_pfn,
diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index 19087b22bdbb..9da823eb0edd 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -949,9 +949,7 @@ static void bo_driver_evict_flags(struct ttm_buffer_object *bo,
 	drm_gem_vram_bo_driver_evict_flags(gbo, placement);
 }
 
-static void bo_driver_move_notify(struct ttm_buffer_object *bo,
-				  bool evict,
-				  struct ttm_resource *new_mem)
+static void bo_driver_delete_mem_notify(struct ttm_buffer_object *bo)
 {
 	struct drm_gem_vram_object *gbo;
 
@@ -961,7 +959,7 @@ static void bo_driver_move_notify(struct ttm_buffer_object *bo,
 
 	gbo = drm_gem_vram_of_bo(bo);
 
-	drm_gem_vram_bo_driver_move_notify(gbo, evict, new_mem);
+	drm_gem_vram_bo_driver_move_notify(gbo, false, NULL);
 }
 
 static int bo_driver_move(struct ttm_buffer_object *bo,
@@ -1002,7 +1000,7 @@ static struct ttm_bo_driver bo_driver = {
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = bo_driver_evict_flags,
 	.move = bo_driver_move,
-	.move_notify = bo_driver_move_notify,
+	.delete_mem_notify = bo_driver_delete_mem_notify,
 	.io_mem_reserve = bo_driver_io_mem_reserve,
 };
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 70b6f3b1ae85..acff82afe260 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1401,6 +1401,12 @@ nouveau_bo_fence(struct nouveau_bo *nvbo, struct nouveau_fence *fence, bool excl
 		dma_resv_add_shared_fence(resv, &fence->base);
 }
 
+static void
+nouveau_bo_delete_mem_notify(struct ttm_buffer_object *bo)
+{
+	nouveau_bo_move_ntfy(bo, false, NULL);
+}
+
 struct ttm_bo_driver nouveau_bo_driver = {
 	.ttm_tt_create = &nouveau_ttm_tt_create,
 	.ttm_tt_populate = &nouveau_ttm_tt_populate,
@@ -1408,7 +1414,7 @@ struct ttm_bo_driver nouveau_bo_driver = {
 	.ttm_tt_destroy = &nouveau_ttm_tt_destroy,
 	.eviction_valuable = ttm_bo_eviction_valuable,
 	.evict_flags = nouveau_bo_evict_flags,
-	.move_notify = nouveau_bo_move_ntfy,
+	.delete_mem_notify = nouveau_bo_delete_mem_notify,
 	.move = nouveau_bo_move,
 	.verify_access = nouveau_bo_verify_access,
 	.io_mem_reserve = &nouveau_ttm_io_mem_reserve,
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index 1cc3c14bc684..b52a4563b47b 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -166,6 +166,11 @@ out:
 	return ret;
 }
 
+static void qxl_bo_delete_mem_notify(struct ttm_buffer_object *bo)
+{
+	qxl_bo_move_notify(bo, false, NULL);
+}
+
 static struct ttm_bo_driver qxl_bo_driver = {
 	.ttm_tt_create = &qxl_ttm_tt_create,
 	.ttm_tt_destroy = &qxl_ttm_backend_destroy,
@@ -173,7 +178,7 @@ static struct ttm_bo_driver qxl_bo_driver = {
 	.evict_flags = &qxl_evict_flags,
 	.move = &qxl_bo_move,
 	.io_mem_reserve = &qxl_ttm_io_mem_reserve,
-	.move_notify = &qxl_bo_move_notify,
+	.delete_mem_notify = &qxl_bo_delete_mem_notify,
 };
 
 static int qxl_ttm_init_mem_type(struct qxl_device *qdev,
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index cd454e5c802f..321c09d20c6c 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -824,6 +824,12 @@ bool radeon_ttm_tt_is_readonly(struct radeon_device *rdev,
 	return !!(gtt->userflags & RADEON_GEM_USERPTR_READONLY);
 }
 
+static void
+radeon_bo_delete_mem_notify(struct ttm_buffer_object *bo)
+{
+	radeon_bo_move_notify(bo, false, NULL);
+}
+
 static struct ttm_bo_driver radeon_bo_driver = {
 	.ttm_tt_create = &radeon_ttm_tt_create,
 	.ttm_tt_populate = &radeon_ttm_tt_populate,
@@ -833,7 +839,7 @@ static struct ttm_bo_driver radeon_bo_driver = {
 	.evict_flags = &radeon_evict_flags,
 	.move = &radeon_bo_move,
 	.verify_access = &radeon_verify_access,
-	.move_notify = &radeon_bo_move_notify,
+	.delete_mem_notify = &radeon_bo_delete_mem_notify,
 	.io_mem_reserve = &radeon_ttm_io_mem_reserve,
 };
 
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 03b40ce7d2dc..5b411252a857 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -284,8 +284,8 @@ out_err:
 
 static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo)
 {
-	if (bo->bdev->driver->move_notify)
-		bo->bdev->driver->move_notify(bo, false, NULL);
+	if (bo->bdev->driver->delete_mem_notify)
+		bo->bdev->driver->delete_mem_notify(bo);
 
 	ttm_bo_tt_destroy(bo);
 	ttm_resource_free(bo, &bo->mem);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index de25cf016be2..88be48ad0344 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -771,6 +771,12 @@ fail:
 	return ret;
 }
 
+static void
+vmw_delete_mem_notify(struct ttm_buffer_object *bo)
+{
+	vmw_move_notify(bo, false, NULL);
+}
+
 struct ttm_bo_driver vmw_bo_driver = {
 	.ttm_tt_create = &vmw_ttm_tt_create,
 	.ttm_tt_populate = &vmw_ttm_populate,
@@ -780,7 +786,7 @@ struct ttm_bo_driver vmw_bo_driver = {
 	.evict_flags = vmw_evict_flags,
 	.move = vmw_move,
 	.verify_access = vmw_verify_access,
-	.move_notify = vmw_move_notify,
+	.delete_mem_notify = vmw_delete_mem_notify,
 	.swap_notify = vmw_swap_notify,
 	.io_mem_reserve = &vmw_ttm_io_mem_reserve,
 };
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 72f106b335e9..29f6a1d1c853 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -156,15 +156,9 @@ struct ttm_bo_driver {
 			     struct file *filp);
 
 	/**
-	 * Hook to notify driver about a driver move so it
-	 * can do tiling things and book-keeping.
-	 *
-	 * @evict: whether this move is evicting the buffer from the graphics
-	 * address space
+	 * Hook to notify driver about a resource delete.
 	 */
-	void (*move_notify)(struct ttm_buffer_object *bo,
-			    bool evict,
-			    struct ttm_resource *new_mem);
+	void (*delete_mem_notify)(struct ttm_buffer_object *bo);
 
 	/**
 	 * notify the driver that we're about to swap out this bo
-- 
cgit 


From e34b8feeaa4b65725b25f49c9b08a0f8707e8e86 Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Wed, 21 Oct 2020 14:06:49 +0200
Subject: drm/ttm: merge ttm_dma_tt back into ttm_tt
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

It makes no difference to kmalloc if the structure
is 48 or 64 bytes in size.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/396950/
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c    | 10 ++----
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    | 14 ++++----
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c     |  7 ++--
 drivers/gpu/drm/nouveau/nouveau_bo.c       | 12 +++----
 drivers/gpu/drm/nouveau/nouveau_mem.c      |  8 +++--
 drivers/gpu/drm/nouveau/nouveau_mem.h      |  4 +--
 drivers/gpu/drm/nouveau/nouveau_sgdma.c    |  6 ++--
 drivers/gpu/drm/qxl/qxl_ttm.c              |  2 +-
 drivers/gpu/drm/radeon/radeon_ttm.c        |  8 ++---
 drivers/gpu/drm/ttm/ttm_bo.c               |  2 +-
 drivers/gpu/drm/ttm/ttm_page_alloc.c       | 30 ++++++++--------
 drivers/gpu/drm/ttm/ttm_page_alloc_dma.c   | 44 +++++++++++-------------
 drivers/gpu/drm/ttm/ttm_tt.c               | 55 ++++++++++++------------------
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 26 +++++++-------
 include/drm/ttm/ttm_page_alloc.h           | 12 +++----
 include/drm/ttm/ttm_tt.h                   | 35 ++++++-------------
 16 files changed, 119 insertions(+), 156 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
index 3c5ad69eff19..0e35023b5703 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
@@ -45,12 +45,10 @@ void amdgpu_gmc_get_pde_for_bo(struct amdgpu_bo *bo, int level,
 			       uint64_t *addr, uint64_t *flags)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
-	struct ttm_dma_tt *ttm;
 
 	switch (bo->tbo.mem.mem_type) {
 	case TTM_PL_TT:
-		ttm = container_of(bo->tbo.ttm, struct ttm_dma_tt, ttm);
-		*addr = ttm->dma_address[0];
+		*addr = bo->tbo.ttm->dma_address[0];
 		break;
 	case TTM_PL_VRAM:
 		*addr = amdgpu_bo_gpu_offset(bo);
@@ -122,16 +120,14 @@ int amdgpu_gmc_set_pte_pde(struct amdgpu_device *adev, void *cpu_pt_addr,
 uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
-	struct ttm_dma_tt *ttm;
 
 	if (bo->num_pages != 1 || bo->ttm->caching == ttm_cached)
 		return AMDGPU_BO_INVALID_OFFSET;
 
-	ttm = container_of(bo->ttm, struct ttm_dma_tt, ttm);
-	if (ttm->dma_address[0] + PAGE_SIZE >= adev->gmc.agp_size)
+	if (bo->ttm->dma_address[0] + PAGE_SIZE >= adev->gmc.agp_size)
 		return AMDGPU_BO_INVALID_OFFSET;
 
-	return adev->gmc.agp_start + ttm->dma_address[0];
+	return adev->gmc.agp_start + bo->ttm->dma_address[0];
 }
 
 /**
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 0a3270f25e40..beacd00221d8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -294,11 +294,9 @@ static int amdgpu_ttm_map_buffer(struct ttm_buffer_object *bo,
 	cpu_addr = &job->ibs[0].ptr[num_dw];
 
 	if (mem->mem_type == TTM_PL_TT) {
-		struct ttm_dma_tt *dma;
 		dma_addr_t *dma_address;
 
-		dma = container_of(bo->ttm, struct ttm_dma_tt, ttm);
-		dma_address = &dma->dma_address[offset >> PAGE_SHIFT];
+		dma_address = &bo->ttm->dma_address[offset >> PAGE_SHIFT];
 		r = amdgpu_gart_map(adev, 0, num_pages, dma_address, flags,
 				    cpu_addr);
 		if (r)
@@ -841,7 +839,7 @@ uint64_t amdgpu_ttm_domain_start(struct amdgpu_device *adev, uint32_t type)
  * TTM backend functions.
  */
 struct amdgpu_ttm_tt {
-	struct ttm_dma_tt	ttm;
+	struct ttm_tt	ttm;
 	struct drm_gem_object	*gobj;
 	u64			offset;
 	uint64_t		userptr;
@@ -1292,7 +1290,7 @@ static void amdgpu_ttm_backend_unbind(struct ttm_bo_device *bdev,
 	r = amdgpu_gart_unbind(adev, gtt->offset, ttm->num_pages);
 	if (r)
 		DRM_ERROR("failed to unbind %u pages at 0x%08llX\n",
-			  gtt->ttm.ttm.num_pages, gtt->offset);
+			  gtt->ttm.num_pages, gtt->offset);
 	gtt->bound = false;
 }
 
@@ -1306,7 +1304,7 @@ static void amdgpu_ttm_backend_destroy(struct ttm_bo_device *bdev,
 	if (gtt->usertask)
 		put_task_struct(gtt->usertask);
 
-	ttm_dma_tt_fini(&gtt->ttm);
+	ttm_tt_fini(&gtt->ttm);
 	kfree(gtt);
 }
 
@@ -1340,7 +1338,7 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo,
 		kfree(gtt);
 		return NULL;
 	}
-	return &gtt->ttm.ttm;
+	return &gtt->ttm;
 }
 
 /**
@@ -1507,7 +1505,7 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start,
 	/* Return false if no part of the ttm_tt object lies within
 	 * the range
 	 */
-	size = (unsigned long)gtt->ttm.ttm.num_pages * PAGE_SIZE;
+	size = (unsigned long)gtt->ttm.num_pages * PAGE_SIZE;
 	if (gtt->userptr > end || gtt->userptr + size <= start)
 		return false;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 3e6243623082..f0e6fafd0938 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -1781,7 +1781,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 		resv = vm->root.base.bo->tbo.base.resv;
 	} else {
 		struct drm_gem_object *obj = &bo->tbo.base;
-		struct ttm_dma_tt *ttm;
 
 		resv = bo->tbo.base.resv;
 		if (obj->import_attach && bo_va->is_xgmi) {
@@ -1794,10 +1793,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va,
 		}
 		mem = &bo->tbo.mem;
 		nodes = mem->mm_node;
-		if (mem->mem_type == TTM_PL_TT) {
-			ttm = container_of(bo->tbo.ttm, struct ttm_dma_tt, ttm);
-			pages_addr = ttm->dma_address;
-		}
+		if (mem->mem_type == TTM_PL_TT)
+			pages_addr = bo->tbo.ttm->dma_address;
 	}
 
 	if (bo) {
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 06a1f4c4e96e..75fddbcd7832 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -547,7 +547,7 @@ void
 nouveau_bo_sync_for_device(struct nouveau_bo *nvbo)
 {
 	struct nouveau_drm *drm = nouveau_bdev(nvbo->bo.bdev);
-	struct ttm_dma_tt *ttm_dma = (struct ttm_dma_tt *)nvbo->bo.ttm;
+	struct ttm_tt *ttm_dma = (struct ttm_tt *)nvbo->bo.ttm;
 	int i;
 
 	if (!ttm_dma)
@@ -557,7 +557,7 @@ nouveau_bo_sync_for_device(struct nouveau_bo *nvbo)
 	if (nvbo->force_coherent)
 		return;
 
-	for (i = 0; i < ttm_dma->ttm.num_pages; i++)
+	for (i = 0; i < ttm_dma->num_pages; i++)
 		dma_sync_single_for_device(drm->dev->dev,
 					   ttm_dma->dma_address[i],
 					   PAGE_SIZE, DMA_TO_DEVICE);
@@ -567,7 +567,7 @@ void
 nouveau_bo_sync_for_cpu(struct nouveau_bo *nvbo)
 {
 	struct nouveau_drm *drm = nouveau_bdev(nvbo->bo.bdev);
-	struct ttm_dma_tt *ttm_dma = (struct ttm_dma_tt *)nvbo->bo.ttm;
+	struct ttm_tt *ttm_dma = (struct ttm_tt *)nvbo->bo.ttm;
 	int i;
 
 	if (!ttm_dma)
@@ -577,7 +577,7 @@ nouveau_bo_sync_for_cpu(struct nouveau_bo *nvbo)
 	if (nvbo->force_coherent)
 		return;
 
-	for (i = 0; i < ttm_dma->ttm.num_pages; i++)
+	for (i = 0; i < ttm_dma->num_pages; i++)
 		dma_sync_single_for_cpu(drm->dev->dev, ttm_dma->dma_address[i],
 					PAGE_SIZE, DMA_FROM_DEVICE);
 }
@@ -1309,7 +1309,7 @@ static int
 nouveau_ttm_tt_populate(struct ttm_bo_device *bdev,
 			struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 {
-	struct ttm_dma_tt *ttm_dma = (void *)ttm;
+	struct ttm_tt *ttm_dma = (void *)ttm;
 	struct nouveau_drm *drm;
 	struct device *dev;
 	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
@@ -1345,7 +1345,7 @@ static void
 nouveau_ttm_tt_unpopulate(struct ttm_bo_device *bdev,
 			  struct ttm_tt *ttm)
 {
-	struct ttm_dma_tt *ttm_dma = (void *)ttm;
+	struct ttm_tt *ttm_dma = (void *)ttm;
 	struct nouveau_drm *drm;
 	struct device *dev;
 	bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG);
diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.c b/drivers/gpu/drm/nouveau/nouveau_mem.c
index 269d8707acc3..0a4c04aaad30 100644
--- a/drivers/gpu/drm/nouveau/nouveau_mem.c
+++ b/drivers/gpu/drm/nouveau/nouveau_mem.c
@@ -92,7 +92,7 @@ nouveau_mem_fini(struct nouveau_mem *mem)
 }
 
 int
-nouveau_mem_host(struct ttm_resource *reg, struct ttm_dma_tt *tt)
+nouveau_mem_host(struct ttm_resource *reg, struct ttm_tt *tt)
 {
 	struct nouveau_mem *mem = nouveau_mem(reg);
 	struct nouveau_cli *cli = mem->cli;
@@ -116,8 +116,10 @@ nouveau_mem_host(struct ttm_resource *reg, struct ttm_dma_tt *tt)
 		mem->comp = 0;
 	}
 
-	if (tt->ttm.sg) args.sgl = tt->ttm.sg->sgl;
-	else            args.dma = tt->dma_address;
+	if (tt->sg)
+		args.sgl = tt->sg->sgl;
+	else
+		args.dma = tt->dma_address;
 
 	mutex_lock(&drm->master.lock);
 	cli->base.super = true;
diff --git a/drivers/gpu/drm/nouveau/nouveau_mem.h b/drivers/gpu/drm/nouveau/nouveau_mem.h
index 3fe1cfed57a1..7df3848e85aa 100644
--- a/drivers/gpu/drm/nouveau/nouveau_mem.h
+++ b/drivers/gpu/drm/nouveau/nouveau_mem.h
@@ -1,7 +1,7 @@
 #ifndef __NOUVEAU_MEM_H__
 #define __NOUVEAU_MEM_H__
 #include <drm/ttm/ttm_bo_api.h>
-struct ttm_dma_tt;
+struct ttm_tt;
 
 #include <nvif/mem.h>
 #include <nvif/vmm.h>
@@ -24,7 +24,7 @@ int nouveau_mem_new(struct nouveau_cli *, u8 kind, u8 comp,
 		    struct ttm_resource *);
 void nouveau_mem_del(struct ttm_resource *);
 int nouveau_mem_vram(struct ttm_resource *, bool contig, u8 page);
-int nouveau_mem_host(struct ttm_resource *, struct ttm_dma_tt *);
+int nouveau_mem_host(struct ttm_resource *, struct ttm_tt *);
 void nouveau_mem_fini(struct nouveau_mem *);
 int nouveau_mem_map(struct nouveau_mem *, struct nvif_vmm *, struct nvif_vma *);
 #endif
diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
index cd6fdebae795..a2e23fd4906c 100644
--- a/drivers/gpu/drm/nouveau/nouveau_sgdma.c
+++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c
@@ -11,7 +11,7 @@ struct nouveau_sgdma_be {
 	/* this has to be the first field so populate/unpopulated in
 	 * nouve_bo.c works properly, otherwise have to move them here
 	 */
-	struct ttm_dma_tt ttm;
+	struct ttm_tt ttm;
 	struct nouveau_mem *mem;
 };
 
@@ -23,7 +23,7 @@ nouveau_sgdma_destroy(struct ttm_bo_device *bdev, struct ttm_tt *ttm)
 	if (ttm) {
 		nouveau_sgdma_unbind(bdev, ttm);
 		ttm_tt_destroy_common(bdev, ttm);
-		ttm_dma_tt_fini(&nvbe->ttm);
+		ttm_tt_fini(&nvbe->ttm);
 		kfree(nvbe);
 	}
 }
@@ -88,5 +88,5 @@ nouveau_sgdma_create_ttm(struct ttm_buffer_object *bo, uint32_t page_flags)
 		kfree(nvbe);
 		return NULL;
 	}
-	return &nvbe->ttm.ttm;
+	return &nvbe->ttm;
 }
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index b52a4563b47b..9609eeb52821 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -116,7 +116,7 @@ static struct ttm_tt *qxl_ttm_tt_create(struct ttm_buffer_object *bo,
 	ttm = kzalloc(sizeof(struct ttm_tt), GFP_KERNEL);
 	if (ttm == NULL)
 		return NULL;
-	if (ttm_tt_init(ttm, bo, page_flags, ttm_cached)) {
+	if (ttm_dma_tt_init(ttm, bo, page_flags, ttm_cached)) {
 		kfree(ttm);
 		return NULL;
 	}
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index c51dcbc818ef..0a6d7ea847db 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -437,7 +437,7 @@ static int radeon_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_reso
  * TTM backend functions.
  */
 struct radeon_ttm_tt {
-	struct ttm_dma_tt		ttm;
+	struct ttm_tt		ttm;
 	u64				offset;
 
 	uint64_t			userptr;
@@ -602,7 +602,7 @@ static void radeon_ttm_backend_destroy(struct ttm_bo_device *bdev, struct ttm_tt
 	radeon_ttm_backend_unbind(bdev, ttm);
 	ttm_tt_destroy_common(bdev, ttm);
 
-	ttm_dma_tt_fini(&gtt->ttm);
+	ttm_tt_fini(&gtt->ttm);
 	kfree(gtt);
 }
 
@@ -640,7 +640,7 @@ static struct ttm_tt *radeon_ttm_tt_create(struct ttm_buffer_object *bo,
 		kfree(gtt);
 		return NULL;
 	}
-	return &gtt->ttm.ttm;
+	return &gtt->ttm;
 }
 
 static struct radeon_ttm_tt *radeon_ttm_tt_to_gtt(struct radeon_device *rdev,
@@ -653,7 +653,7 @@ static struct radeon_ttm_tt *radeon_ttm_tt_to_gtt(struct radeon_device *rdev,
 
 	if (!ttm)
 		return NULL;
-	return container_of(ttm, struct radeon_ttm_tt, ttm.ttm);
+	return container_of(ttm, struct radeon_ttm_tt, ttm);
 }
 
 static int radeon_ttm_tt_populate(struct ttm_bo_device *bdev,
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 5b411252a857..40c72a0f9325 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -1192,7 +1192,7 @@ size_t ttm_bo_dma_acc_size(struct ttm_bo_device *bdev,
 
 	size += ttm_round_pot(struct_size);
 	size += ttm_round_pot(npages * (2*sizeof(void *) + sizeof(dma_addr_t)));
-	size += ttm_round_pot(sizeof(struct ttm_dma_tt));
+	size += ttm_round_pot(sizeof(struct ttm_tt));
 	return size;
 }
 EXPORT_SYMBOL(ttm_bo_dma_acc_size);
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index 661b75d19cad..29e6c29ad60e 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -1081,28 +1081,28 @@ void ttm_pool_unpopulate(struct ttm_tt *ttm)
 }
 EXPORT_SYMBOL(ttm_pool_unpopulate);
 
-int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt,
+int ttm_populate_and_map_pages(struct device *dev, struct ttm_tt *tt,
 					struct ttm_operation_ctx *ctx)
 {
 	unsigned i, j;
 	int r;
 
-	r = ttm_pool_populate(&tt->ttm, ctx);
+	r = ttm_pool_populate(tt, ctx);
 	if (r)
 		return r;
 
-	for (i = 0; i < tt->ttm.num_pages; ++i) {
-		struct page *p = tt->ttm.pages[i];
+	for (i = 0; i < tt->num_pages; ++i) {
+		struct page *p = tt->pages[i];
 		size_t num_pages = 1;
 
-		for (j = i + 1; j < tt->ttm.num_pages; ++j) {
-			if (++p != tt->ttm.pages[j])
+		for (j = i + 1; j < tt->num_pages; ++j) {
+			if (++p != tt->pages[j])
 				break;
 
 			++num_pages;
 		}
 
-		tt->dma_address[i] = dma_map_page(dev, tt->ttm.pages[i],
+		tt->dma_address[i] = dma_map_page(dev, tt->pages[i],
 						  0, num_pages * PAGE_SIZE,
 						  DMA_BIDIRECTIONAL);
 		if (dma_mapping_error(dev, tt->dma_address[i])) {
@@ -1111,7 +1111,7 @@ int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt,
 					       PAGE_SIZE, DMA_BIDIRECTIONAL);
 				tt->dma_address[i] = 0;
 			}
-			ttm_pool_unpopulate(&tt->ttm);
+			ttm_pool_unpopulate(tt);
 			return -EFAULT;
 		}
 
@@ -1124,21 +1124,21 @@ int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt,
 }
 EXPORT_SYMBOL(ttm_populate_and_map_pages);
 
-void ttm_unmap_and_unpopulate_pages(struct device *dev, struct ttm_dma_tt *tt)
+void ttm_unmap_and_unpopulate_pages(struct device *dev, struct ttm_tt *tt)
 {
 	unsigned i, j;
 
-	for (i = 0; i < tt->ttm.num_pages;) {
-		struct page *p = tt->ttm.pages[i];
+	for (i = 0; i < tt->num_pages;) {
+		struct page *p = tt->pages[i];
 		size_t num_pages = 1;
 
-		if (!tt->dma_address[i] || !tt->ttm.pages[i]) {
+		if (!tt->dma_address[i] || !tt->pages[i]) {
 			++i;
 			continue;
 		}
 
-		for (j = i + 1; j < tt->ttm.num_pages; ++j) {
-			if (++p != tt->ttm.pages[j])
+		for (j = i + 1; j < tt->num_pages; ++j) {
+			if (++p != tt->pages[j])
 				break;
 
 			++num_pages;
@@ -1149,7 +1149,7 @@ void ttm_unmap_and_unpopulate_pages(struct device *dev, struct ttm_dma_tt *tt)
 
 		i += num_pages;
 	}
-	ttm_pool_unpopulate(&tt->ttm);
+	ttm_pool_unpopulate(tt);
 }
 EXPORT_SYMBOL(ttm_unmap_and_unpopulate_pages);
 
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
index a9aaed7e618a..c0353c25efd6 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
@@ -832,11 +832,10 @@ static int ttm_dma_page_pool_fill_locked(struct dma_pool *pool,
  * return dma_page pointer if success, otherwise NULL.
  */
 static struct dma_page *ttm_dma_pool_get_pages(struct dma_pool *pool,
-				  struct ttm_dma_tt *ttm_dma,
+				  struct ttm_tt *ttm,
 				  unsigned index)
 {
 	struct dma_page *d_page = NULL;
-	struct ttm_tt *ttm = &ttm_dma->ttm;
 	unsigned long irq_flags;
 	int count;
 
@@ -845,8 +844,8 @@ static struct dma_page *ttm_dma_pool_get_pages(struct dma_pool *pool,
 	if (count) {
 		d_page = list_first_entry(&pool->free_list, struct dma_page, page_list);
 		ttm->pages[index] = d_page->p;
-		ttm_dma->dma_address[index] = d_page->dma;
-		list_move_tail(&d_page->page_list, &ttm_dma->pages_list);
+		ttm->dma_address[index] = d_page->dma;
+		list_move_tail(&d_page->page_list, &ttm->pages_list);
 		pool->npages_in_use += 1;
 		pool->npages_free -= 1;
 	}
@@ -854,9 +853,8 @@ static struct dma_page *ttm_dma_pool_get_pages(struct dma_pool *pool,
 	return d_page;
 }
 
-static gfp_t ttm_dma_pool_gfp_flags(struct ttm_dma_tt *ttm_dma, bool huge)
+static gfp_t ttm_dma_pool_gfp_flags(struct ttm_tt *ttm, bool huge)
 {
-	struct ttm_tt *ttm = &ttm_dma->ttm;
 	gfp_t gfp_flags;
 
 	if (ttm->page_flags & TTM_PAGE_FLAG_DMA32)
@@ -883,11 +881,10 @@ static gfp_t ttm_dma_pool_gfp_flags(struct ttm_dma_tt *ttm_dma, bool huge)
  * On success pages list will hold count number of correctly
  * cached pages. On failure will hold the negative return value (-ENOMEM, etc).
  */
-int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
+int ttm_dma_populate(struct ttm_tt *ttm, struct device *dev,
 			struct ttm_operation_ctx *ctx)
 {
 	struct ttm_mem_global *mem_glob = &ttm_mem_glob;
-	struct ttm_tt *ttm = &ttm_dma->ttm;
 	unsigned long num_pages = ttm->num_pages;
 	struct dma_pool *pool;
 	struct dma_page *d_page;
@@ -901,7 +898,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
 	if (ttm_check_under_lowerlimit(mem_glob, num_pages, ctx))
 		return -ENOMEM;
 
-	INIT_LIST_HEAD(&ttm_dma->pages_list);
+	INIT_LIST_HEAD(&ttm->pages_list);
 	i = 0;
 
 	type = ttm_to_type(ttm->page_flags, ttm->caching);
@@ -912,7 +909,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
 
 	pool = ttm_dma_find_pool(dev, type | IS_HUGE);
 	if (!pool) {
-		gfp_t gfp_flags = ttm_dma_pool_gfp_flags(ttm_dma, true);
+		gfp_t gfp_flags = ttm_dma_pool_gfp_flags(ttm, true);
 
 		pool = ttm_dma_pool_init(dev, gfp_flags, type | IS_HUGE);
 		if (IS_ERR_OR_NULL(pool))
@@ -922,21 +919,21 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
 	while (num_pages >= HPAGE_PMD_NR) {
 		unsigned j;
 
-		d_page = ttm_dma_pool_get_pages(pool, ttm_dma, i);
+		d_page = ttm_dma_pool_get_pages(pool, ttm, i);
 		if (!d_page)
 			break;
 
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
 						pool->size, ctx);
 		if (unlikely(ret != 0)) {
-			ttm_dma_unpopulate(ttm_dma, dev);
+			ttm_dma_unpopulate(ttm, dev);
 			return -ENOMEM;
 		}
 
 		d_page->vaddr |= VADDR_FLAG_UPDATED_COUNT;
 		for (j = i + 1; j < (i + HPAGE_PMD_NR); ++j) {
 			ttm->pages[j] = ttm->pages[j - 1] + 1;
-			ttm_dma->dma_address[j] = ttm_dma->dma_address[j - 1] +
+			ttm->dma_address[j] = ttm->dma_address[j - 1] +
 				PAGE_SIZE;
 		}
 
@@ -949,7 +946,7 @@ skip_huge:
 
 	pool = ttm_dma_find_pool(dev, type);
 	if (!pool) {
-		gfp_t gfp_flags = ttm_dma_pool_gfp_flags(ttm_dma, false);
+		gfp_t gfp_flags = ttm_dma_pool_gfp_flags(ttm, false);
 
 		pool = ttm_dma_pool_init(dev, gfp_flags, type);
 		if (IS_ERR_OR_NULL(pool))
@@ -957,16 +954,16 @@ skip_huge:
 	}
 
 	while (num_pages) {
-		d_page = ttm_dma_pool_get_pages(pool, ttm_dma, i);
+		d_page = ttm_dma_pool_get_pages(pool, ttm, i);
 		if (!d_page) {
-			ttm_dma_unpopulate(ttm_dma, dev);
+			ttm_dma_unpopulate(ttm, dev);
 			return -ENOMEM;
 		}
 
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
 						pool->size, ctx);
 		if (unlikely(ret != 0)) {
-			ttm_dma_unpopulate(ttm_dma, dev);
+			ttm_dma_unpopulate(ttm, dev);
 			return -ENOMEM;
 		}
 
@@ -980,10 +977,9 @@ skip_huge:
 EXPORT_SYMBOL_GPL(ttm_dma_populate);
 
 /* Put all pages in pages list to correct pool to wait for reuse */
-void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev)
+void ttm_dma_unpopulate(struct ttm_tt *ttm, struct device *dev)
 {
 	struct ttm_mem_global *mem_glob = &ttm_mem_glob;
-	struct ttm_tt *ttm = &ttm_dma->ttm;
 	struct dma_pool *pool;
 	struct dma_page *d_page, *next;
 	enum pool_type type;
@@ -997,7 +993,7 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 	pool = ttm_dma_find_pool(dev, type | IS_HUGE);
 	if (pool) {
 		count = 0;
-		list_for_each_entry_safe(d_page, next, &ttm_dma->pages_list,
+		list_for_each_entry_safe(d_page, next, &ttm->pages_list,
 					 page_list) {
 			if (!(d_page->vaddr & VADDR_FLAG_HUGE_POOL))
 				continue;
@@ -1027,7 +1023,7 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 
 	/* make sure pages array match list and count number of pages */
 	count = 0;
-	list_for_each_entry_safe(d_page, next, &ttm_dma->pages_list,
+	list_for_each_entry_safe(d_page, next, &ttm->pages_list,
 				 page_list) {
 		ttm->pages[count] = d_page->p;
 		count++;
@@ -1048,7 +1044,7 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 		pool->nfrees += count;
 	} else {
 		pool->npages_free += count;
-		list_splice(&ttm_dma->pages_list, &pool->free_list);
+		list_splice(&ttm->pages_list, &pool->free_list);
 		/*
 		 * Wait to have at at least NUM_PAGES_TO_ALLOC number of pages
 		 * to free in order to minimize calls to set_memory_wb().
@@ -1059,10 +1055,10 @@ void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 	}
 	spin_unlock_irqrestore(&pool->lock, irq_flags);
 
-	INIT_LIST_HEAD(&ttm_dma->pages_list);
+	INIT_LIST_HEAD(&ttm->pages_list);
 	for (i = 0; i < ttm->num_pages; i++) {
 		ttm->pages[i] = NULL;
-		ttm_dma->dma_address[i] = 0;
+		ttm->dma_address[i] = 0;
 	}
 
 	/* shrink pool if necessary (only on !is_cached pools)*/
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index dc1dad982f28..65c4254eea5c 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -92,21 +92,22 @@ static int ttm_tt_alloc_page_directory(struct ttm_tt *ttm)
 	return 0;
 }
 
-static int ttm_dma_tt_alloc_page_directory(struct ttm_dma_tt *ttm)
+static int ttm_dma_tt_alloc_page_directory(struct ttm_tt *ttm)
 {
-	ttm->ttm.pages = kvmalloc_array(ttm->ttm.num_pages,
-					  sizeof(*ttm->ttm.pages) +
-					  sizeof(*ttm->dma_address),
-					  GFP_KERNEL | __GFP_ZERO);
-	if (!ttm->ttm.pages)
+	ttm->pages = kvmalloc_array(ttm->num_pages,
+				    sizeof(*ttm->pages) +
+				    sizeof(*ttm->dma_address),
+				    GFP_KERNEL | __GFP_ZERO);
+	if (!ttm->pages)
 		return -ENOMEM;
-	ttm->dma_address = (void *) (ttm->ttm.pages + ttm->ttm.num_pages);
+
+	ttm->dma_address = (void *)(ttm->pages + ttm->num_pages);
 	return 0;
 }
 
-static int ttm_sg_tt_alloc_page_directory(struct ttm_dma_tt *ttm)
+static int ttm_sg_tt_alloc_page_directory(struct ttm_tt *ttm)
 {
-	ttm->dma_address = kvmalloc_array(ttm->ttm.num_pages,
+	ttm->dma_address = kvmalloc_array(ttm->num_pages,
 					  sizeof(*ttm->dma_address),
 					  GFP_KERNEL | __GFP_ZERO);
 	if (!ttm->dma_address)
@@ -138,8 +139,10 @@ static void ttm_tt_init_fields(struct ttm_tt *ttm,
 	ttm->num_pages = bo->num_pages;
 	ttm->caching = ttm_cached;
 	ttm->page_flags = page_flags;
+	ttm->dma_address = NULL;
 	ttm->swap_storage = NULL;
 	ttm->sg = bo->sg;
+	INIT_LIST_HEAD(&ttm->pages_list);
 	ttm->caching = caching;
 }
 
@@ -158,20 +161,21 @@ EXPORT_SYMBOL(ttm_tt_init);
 
 void ttm_tt_fini(struct ttm_tt *ttm)
 {
-	kvfree(ttm->pages);
+	if (ttm->pages)
+		kvfree(ttm->pages);
+	else
+		kvfree(ttm->dma_address);
 	ttm->pages = NULL;
+	ttm->dma_address = NULL;
 }
 EXPORT_SYMBOL(ttm_tt_fini);
 
-int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
+int ttm_dma_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo,
 		    uint32_t page_flags, enum ttm_caching caching)
 {
-	struct ttm_tt *ttm = &ttm_dma->ttm;
-
 	ttm_tt_init_fields(ttm, bo, page_flags, caching);
 
-	INIT_LIST_HEAD(&ttm_dma->pages_list);
-	if (ttm_dma_tt_alloc_page_directory(ttm_dma)) {
+	if (ttm_dma_tt_alloc_page_directory(ttm)) {
 		pr_err("Failed allocating page table\n");
 		return -ENOMEM;
 	}
@@ -179,19 +183,17 @@ int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_dma_tt_init);
 
-int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
+int ttm_sg_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo,
 		   uint32_t page_flags, enum ttm_caching caching)
 {
-	struct ttm_tt *ttm = &ttm_dma->ttm;
 	int ret;
 
 	ttm_tt_init_fields(ttm, bo, page_flags, caching);
 
-	INIT_LIST_HEAD(&ttm_dma->pages_list);
 	if (page_flags & TTM_PAGE_FLAG_SG)
-		ret = ttm_sg_tt_alloc_page_directory(ttm_dma);
+		ret = ttm_sg_tt_alloc_page_directory(ttm);
 	else
-		ret = ttm_dma_tt_alloc_page_directory(ttm_dma);
+		ret = ttm_dma_tt_alloc_page_directory(ttm);
 	if (ret) {
 		pr_err("Failed allocating page table\n");
 		return -ENOMEM;
@@ -200,19 +202,6 @@ int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_sg_tt_init);
 
-void ttm_dma_tt_fini(struct ttm_dma_tt *ttm_dma)
-{
-	struct ttm_tt *ttm = &ttm_dma->ttm;
-
-	if (ttm->pages)
-		kvfree(ttm->pages);
-	else
-		kvfree(ttm_dma->dma_address);
-	ttm->pages = NULL;
-	ttm_dma->dma_address = NULL;
-}
-EXPORT_SYMBOL(ttm_dma_tt_fini);
-
 int ttm_tt_swapin(struct ttm_tt *ttm)
 {
 	struct address_space *swap_space;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 88be48ad0344..92a5d245ff4d 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -186,7 +186,7 @@ struct ttm_placement vmw_nonfixed_placement = {
 };
 
 struct vmw_ttm_tt {
-	struct ttm_dma_tt dma_ttm;
+	struct ttm_tt dma_ttm;
 	struct vmw_private *dev_priv;
 	int gmr_id;
 	struct vmw_mob *mob;
@@ -374,8 +374,8 @@ static int vmw_ttm_map_dma(struct vmw_ttm_tt *vmw_tt)
 		return 0;
 
 	vsgt->mode = dev_priv->map_mode;
-	vsgt->pages = vmw_tt->dma_ttm.ttm.pages;
-	vsgt->num_pages = vmw_tt->dma_ttm.ttm.num_pages;
+	vsgt->pages = vmw_tt->dma_ttm.pages;
+	vsgt->num_pages = vmw_tt->dma_ttm.num_pages;
 	vsgt->addrs = vmw_tt->dma_ttm.dma_address;
 	vsgt->sgt = &vmw_tt->sgt;
 
@@ -483,7 +483,7 @@ static void vmw_ttm_unmap_dma(struct vmw_ttm_tt *vmw_tt)
 const struct vmw_sg_table *vmw_bo_sg_table(struct ttm_buffer_object *bo)
 {
 	struct vmw_ttm_tt *vmw_tt =
-		container_of(bo->ttm, struct vmw_ttm_tt, dma_ttm.ttm);
+		container_of(bo->ttm, struct vmw_ttm_tt, dma_ttm);
 
 	return &vmw_tt->vsgt;
 }
@@ -493,7 +493,7 @@ static int vmw_ttm_bind(struct ttm_bo_device *bdev,
 			struct ttm_tt *ttm, struct ttm_resource *bo_mem)
 {
 	struct vmw_ttm_tt *vmw_be =
-		container_of(ttm, struct vmw_ttm_tt, dma_ttm.ttm);
+		container_of(ttm, struct vmw_ttm_tt, dma_ttm);
 	int ret = 0;
 
 	if (!bo_mem)
@@ -537,7 +537,7 @@ static void vmw_ttm_unbind(struct ttm_bo_device *bdev,
 			   struct ttm_tt *ttm)
 {
 	struct vmw_ttm_tt *vmw_be =
-		container_of(ttm, struct vmw_ttm_tt, dma_ttm.ttm);
+		container_of(ttm, struct vmw_ttm_tt, dma_ttm);
 
 	if (!vmw_be->bound)
 		return;
@@ -562,13 +562,13 @@ static void vmw_ttm_unbind(struct ttm_bo_device *bdev,
 static void vmw_ttm_destroy(struct ttm_bo_device *bdev, struct ttm_tt *ttm)
 {
 	struct vmw_ttm_tt *vmw_be =
-		container_of(ttm, struct vmw_ttm_tt, dma_ttm.ttm);
+		container_of(ttm, struct vmw_ttm_tt, dma_ttm);
 
 	vmw_ttm_unbind(bdev, ttm);
 	ttm_tt_destroy_common(bdev, ttm);
 	vmw_ttm_unmap_dma(vmw_be);
 	if (vmw_be->dev_priv->map_mode == vmw_dma_alloc_coherent)
-		ttm_dma_tt_fini(&vmw_be->dma_ttm);
+		ttm_tt_fini(&vmw_be->dma_ttm);
 	else
 		ttm_tt_fini(ttm);
 
@@ -583,7 +583,7 @@ static int vmw_ttm_populate(struct ttm_bo_device *bdev,
 			    struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 {
 	struct vmw_ttm_tt *vmw_tt =
-		container_of(ttm, struct vmw_ttm_tt, dma_ttm.ttm);
+		container_of(ttm, struct vmw_ttm_tt, dma_ttm);
 	struct vmw_private *dev_priv = vmw_tt->dev_priv;
 	struct ttm_mem_global *glob = vmw_mem_glob(dev_priv);
 	int ret;
@@ -612,7 +612,7 @@ static void vmw_ttm_unpopulate(struct ttm_bo_device *bdev,
 			       struct ttm_tt *ttm)
 {
 	struct vmw_ttm_tt *vmw_tt = container_of(ttm, struct vmw_ttm_tt,
-						 dma_ttm.ttm);
+						 dma_ttm);
 	struct vmw_private *dev_priv = vmw_tt->dev_priv;
 	struct ttm_mem_global *glob = vmw_mem_glob(dev_priv);
 
@@ -650,12 +650,12 @@ static struct ttm_tt *vmw_ttm_tt_create(struct ttm_buffer_object *bo,
 		ret = ttm_dma_tt_init(&vmw_be->dma_ttm, bo, page_flags,
 				      ttm_cached);
 	else
-		ret = ttm_tt_init(&vmw_be->dma_ttm.ttm, bo, page_flags,
+		ret = ttm_tt_init(&vmw_be->dma_ttm, bo, page_flags,
 				  ttm_cached);
 	if (unlikely(ret != 0))
 		goto out_no_init;
 
-	return &vmw_be->dma_ttm.ttm;
+	return &vmw_be->dma_ttm;
 out_no_init:
 	kfree(vmw_be);
 	return NULL;
@@ -813,7 +813,7 @@ int vmw_bo_create_and_populate(struct vmw_private *dev_priv,
 	ret = vmw_ttm_populate(bo->bdev, bo->ttm, &ctx);
 	if (likely(ret == 0)) {
 		struct vmw_ttm_tt *vmw_tt =
-			container_of(bo->ttm, struct vmw_ttm_tt, dma_ttm.ttm);
+			container_of(bo->ttm, struct vmw_ttm_tt, dma_ttm);
 		ret = vmw_ttm_map_dma(vmw_tt);
 	}
 
diff --git a/include/drm/ttm/ttm_page_alloc.h b/include/drm/ttm/ttm_page_alloc.h
index a6b6ef5f9bf4..8fa1e7df6213 100644
--- a/include/drm/ttm/ttm_page_alloc.h
+++ b/include/drm/ttm/ttm_page_alloc.h
@@ -61,13 +61,13 @@ void ttm_pool_unpopulate(struct ttm_tt *ttm);
 /**
  * Populates and DMA maps pages to fullfil a ttm_dma_populate() request
  */
-int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt,
+int ttm_populate_and_map_pages(struct device *dev, struct ttm_tt *tt,
 				struct ttm_operation_ctx *ctx);
 
 /**
  * Unpopulates and DMA unmaps pages as part of a
  * ttm_dma_unpopulate() request */
-void ttm_unmap_and_unpopulate_pages(struct device *dev, struct ttm_dma_tt *tt);
+void ttm_unmap_and_unpopulate_pages(struct device *dev, struct ttm_tt *tt);
 
 /**
  * Output the state of pools to debugfs file
@@ -90,9 +90,9 @@ void ttm_dma_page_alloc_fini(void);
  */
 int ttm_dma_page_alloc_debugfs(struct seq_file *m, void *data);
 
-int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
+int ttm_dma_populate(struct ttm_tt *ttm_dma, struct device *dev,
 			struct ttm_operation_ctx *ctx);
-void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev);
+void ttm_dma_unpopulate(struct ttm_tt *ttm_dma, struct device *dev);
 
 #else
 static inline int ttm_dma_page_alloc_init(struct ttm_mem_global *glob,
@@ -107,13 +107,13 @@ static inline int ttm_dma_page_alloc_debugfs(struct seq_file *m, void *data)
 {
 	return 0;
 }
-static inline int ttm_dma_populate(struct ttm_dma_tt *ttm_dma,
+static inline int ttm_dma_populate(struct ttm_tt *ttm_dma,
 				struct device *dev,
 				struct ttm_operation_ctx *ctx)
 {
 	return -ENOMEM;
 }
-static inline void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma,
+static inline void ttm_dma_unpopulate(struct ttm_tt *ttm_dma,
 				      struct device *dev)
 {
 }
diff --git a/include/drm/ttm/ttm_tt.h b/include/drm/ttm/ttm_tt.h
index 931a31355870..df9a80650feb 100644
--- a/include/drm/ttm/ttm_tt.h
+++ b/include/drm/ttm/ttm_tt.h
@@ -47,12 +47,13 @@ struct ttm_operation_ctx;
  * struct ttm_tt
  *
  * @pages: Array of pages backing the data.
+ * @page_flags: see TTM_PAGE_FLAG_*
  * @num_pages: Number of pages in the page array.
- * @bdev: Pointer to the current struct ttm_bo_device.
- * @be: Pointer to the ttm backend.
+ * @sg: for SG objects via dma-buf
+ * @dma_address: The DMA (bus) addresses of the pages
  * @swap_storage: Pointer to shmem struct file for swap storage.
- * @caching_state: The current caching state of the pages.
- * @state: The current binding state of the pages.
+ * @pages_list: used by some page allocation backend
+ * @caching: The current caching state of the pages.
  *
  * This is a structure holding the pages, caching- and aperture binding
  * status for a buffer object that isn't backed by fixed (VRAM / AGP)
@@ -62,8 +63,10 @@ struct ttm_tt {
 	struct page **pages;
 	uint32_t page_flags;
 	uint32_t num_pages;
-	struct sg_table *sg; /* for SG objects via dma-buf */
+	struct sg_table *sg;
+	dma_addr_t *dma_address;
 	struct file *swap_storage;
+	struct list_head pages_list;
 	enum ttm_caching caching;
 };
 
@@ -72,23 +75,6 @@ static inline bool ttm_tt_is_populated(struct ttm_tt *tt)
 	return tt->page_flags & TTM_PAGE_FLAG_PRIV_POPULATED;
 }
 
-/**
- * struct ttm_dma_tt
- *
- * @ttm: Base ttm_tt struct.
- * @dma_address: The DMA (bus) addresses of the pages
- * @pages_list: used by some page allocation backend
- *
- * This is a structure holding the pages, caching- and aperture binding
- * status for a buffer object that isn't backed by fixed (VRAM / AGP)
- * memory.
- */
-struct ttm_dma_tt {
-	struct ttm_tt ttm;
-	dma_addr_t *dma_address;
-	struct list_head pages_list;
-};
-
 /**
  * ttm_tt_create
  *
@@ -115,9 +101,9 @@ int ttm_tt_create(struct ttm_buffer_object *bo, bool zero_alloc);
  */
 int ttm_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo,
 		uint32_t page_flags, enum ttm_caching caching);
-int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
+int ttm_dma_tt_init(struct ttm_tt *ttm_dma, struct ttm_buffer_object *bo,
 		    uint32_t page_flags, enum ttm_caching caching);
-int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
+int ttm_sg_tt_init(struct ttm_tt *ttm_dma, struct ttm_buffer_object *bo,
 		   uint32_t page_flags, enum ttm_caching caching);
 
 /**
@@ -128,7 +114,6 @@ int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo,
  * Free memory of ttm_tt structure
  */
 void ttm_tt_fini(struct ttm_tt *ttm);
-void ttm_dma_tt_fini(struct ttm_dma_tt *ttm_dma);
 
 /**
  * ttm_ttm_destroy:
-- 
cgit 


From 8567d51555c12d169c4e0f796030051fff1c318d Mon Sep 17 00:00:00 2001
From: Christian König <christian.koenig@amd.com>
Date: Sat, 24 Oct 2020 13:16:05 +0200
Subject: drm/vmwgfx: switch to new allocator
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

It should be able to handle all cases now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Madhav Chauhan <madhav.chauhan@amd.com>
Tested-by: Huang Rui <ray.huang@amd.com>
Link: https://patchwork.freedesktop.org/patch/397083/?series=83051&rev=1
---
 drivers/gpu/drm/vmwgfx/vmwgfx_drv.c        |  4 ----
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c | 36 +++---------------------------
 2 files changed, 3 insertions(+), 37 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
index 7bd1be26afe4..b3a60959b5d5 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c
@@ -595,10 +595,6 @@ static int vmw_dma_select_mode(struct vmw_private *dev_priv)
 	else
 		dev_priv->map_mode = vmw_dma_map_populate;
 
-        if (!IS_ENABLED(CONFIG_DRM_TTM_DMA_PAGE_POOL) &&
-	    (dev_priv->map_mode == vmw_dma_alloc_coherent))
-		return -EINVAL;
-
 	DRM_INFO("DMA map mode: %s\n", names[dev_priv->map_mode]);
 	return 0;
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 92a5d245ff4d..aa0cdf476a53 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -28,7 +28,6 @@
 #include "vmwgfx_drv.h"
 #include <drm/ttm/ttm_bo_driver.h>
 #include <drm/ttm/ttm_placement.h>
-#include <drm/ttm/ttm_page_alloc.h>
 
 static const struct ttm_place vram_placement_flags = {
 	.fpfn = 0,
@@ -582,30 +581,11 @@ static void vmw_ttm_destroy(struct ttm_bo_device *bdev, struct ttm_tt *ttm)
 static int vmw_ttm_populate(struct ttm_bo_device *bdev,
 			    struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 {
-	struct vmw_ttm_tt *vmw_tt =
-		container_of(ttm, struct vmw_ttm_tt, dma_ttm);
-	struct vmw_private *dev_priv = vmw_tt->dev_priv;
-	struct ttm_mem_global *glob = vmw_mem_glob(dev_priv);
-	int ret;
-
+	/* TODO: maybe completely drop this ? */
 	if (ttm_tt_is_populated(ttm))
 		return 0;
 
-	if (dev_priv->map_mode == vmw_dma_alloc_coherent) {
-		size_t size =
-			ttm_round_pot(ttm->num_pages * sizeof(dma_addr_t));
-		ret = ttm_mem_global_alloc(glob, size, ctx);
-		if (unlikely(ret != 0))
-			return ret;
-
-		ret = ttm_dma_populate(&vmw_tt->dma_ttm, dev_priv->dev->dev,
-					ctx);
-		if (unlikely(ret != 0))
-			ttm_mem_global_free(glob, size);
-	} else
-		ret = ttm_pool_populate(ttm, ctx);
-
-	return ret;
+	return ttm_pool_alloc(&bdev->pool, ttm, ctx);
 }
 
 static void vmw_ttm_unpopulate(struct ttm_bo_device *bdev,
@@ -613,9 +593,6 @@ static void vmw_ttm_unpopulate(struct ttm_bo_device *bdev,
 {
 	struct vmw_ttm_tt *vmw_tt = container_of(ttm, struct vmw_ttm_tt,
 						 dma_ttm);
-	struct vmw_private *dev_priv = vmw_tt->dev_priv;
-	struct ttm_mem_global *glob = vmw_mem_glob(dev_priv);
-
 
 	if (vmw_tt->mob) {
 		vmw_mob_destroy(vmw_tt->mob);
@@ -623,14 +600,7 @@ static void vmw_ttm_unpopulate(struct ttm_bo_device *bdev,
 	}
 
 	vmw_ttm_unmap_dma(vmw_tt);
-	if (dev_priv->map_mode == vmw_dma_alloc_coherent) {
-		size_t size =
-			ttm_round_pot(ttm->num_pages * sizeof(dma_addr_t));
-
-		ttm_dma_unpopulate(&vmw_tt->dma_ttm, dev_priv->dev->dev);
-		ttm_mem_global_free(glob, size);
-	} else
-		ttm_pool_unpopulate(ttm);
+	ttm_pool_free(&bdev->pool, ttm);
 }
 
 static struct ttm_tt *vmw_ttm_tt_create(struct ttm_buffer_object *bo,
-- 
cgit 


From ebdf565169af006ee3be8c40eecbfc77d28a3b84 Mon Sep 17 00:00:00 2001
From: Dave Airlie <airlied@redhat.com>
Date: Thu, 29 Oct 2020 13:58:52 +1000
Subject: drm/ttm: add multihop infrastrucutre (v3)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Currently drivers get called to move a buffer, but if they have to
move it temporarily through another space (SYSTEM->VRAM via TT)
then they can end up with a lot of ttm->driver->ttm call stacks,
if the temprorary space moves requires eviction.

Instead of letting the driver do all the placement/space for the
temporary, allow it to report back (-EMULTIHOP) and a placement (hop)
to the move code, which will then do the temporary move, and the
correct placement move afterwards.

This removes a lot of code from drivers, at the expense of
adding some midlayering. I've some further ideas on how to turn
it inside out, but I think this is a good solution to the call
stack problems.

v2: separate out the driver patches, add WARN for getting
MULTHOP in paths we shouldn't (Daniel)
v3: use memset (Christian)

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: hristian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20201109005432.861936-2-airlied@gmail.com
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c    |  3 +-
 drivers/gpu/drm/drm_gem_vram_helper.c      |  3 +-
 drivers/gpu/drm/nouveau/nouveau_bo.c       |  3 +-
 drivers/gpu/drm/qxl/qxl_ttm.c              |  3 +-
 drivers/gpu/drm/radeon/radeon_ttm.c        |  3 +-
 drivers/gpu/drm/ttm/ttm_bo.c               | 74 ++++++++++++++++++++++++++----
 drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c |  3 +-
 include/drm/ttm/ttm_bo_driver.h            |  7 ++-
 8 files changed, 83 insertions(+), 16 deletions(-)

(limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 12aa35ac0eb4..fad7ca959316 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -656,7 +656,8 @@ static bool amdgpu_mem_visible(struct amdgpu_device *adev,
  */
 static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
 			  struct ttm_operation_ctx *ctx,
-			  struct ttm_resource *new_mem)
+			  struct ttm_resource *new_mem,
+			  struct ttm_place *hop)
 {
 	struct amdgpu_device *adev;
 	struct amdgpu_bo *abo;
diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c
index 5e1b09ba5f2b..889a06696f7e 100644
--- a/drivers/gpu/drm/drm_gem_vram_helper.c
+++ b/drivers/gpu/drm/drm_gem_vram_helper.c
@@ -955,7 +955,8 @@ static void bo_driver_delete_mem_notify(struct ttm_buffer_object *bo)
 static int bo_driver_move(struct ttm_buffer_object *bo,
 			  bool evict,
 			  struct ttm_operation_ctx *ctx,
-			  struct ttm_resource *new_mem)
+			  struct ttm_resource *new_mem,
+			  struct ttm_place *hop)
 {
 	struct drm_gem_vram_object *gbo;
 
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 8133377d865d..fee07b9d19ed 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1023,7 +1023,8 @@ nouveau_bo_vm_cleanup(struct ttm_buffer_object *bo,
 static int
 nouveau_bo_move(struct ttm_buffer_object *bo, bool evict,
 		struct ttm_operation_ctx *ctx,
-		struct ttm_resource *new_reg)
+		struct ttm_resource *new_reg,
+		struct ttm_place *hop)
 {
 	struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
 	struct nouveau_bo *nvbo = nouveau_bo(bo);
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index a80d59634143..128c38c8a837 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -140,7 +140,8 @@ static void qxl_bo_move_notify(struct ttm_buffer_object *bo,
 
 static int qxl_bo_move(struct ttm_buffer_object *bo, bool evict,
 		       struct ttm_operation_ctx *ctx,
-		       struct ttm_resource *new_mem)
+		       struct ttm_resource *new_mem,
+		       struct ttm_place *hop)
 {
 	struct ttm_resource *old_mem = &bo->mem;
 	int ret;
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 95038ac3382e..29062dbea299 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -303,7 +303,8 @@ out_cleanup:
 
 static int radeon_bo_move(struct ttm_buffer_object *bo, bool evict,
 			  struct ttm_operation_ctx *ctx,
-			  struct ttm_resource *new_mem)
+			  struct ttm_resource *new_mem,
+			  struct ttm_place *hop)
 {
 	struct radeon_device *rdev;
 	struct radeon_bo *rbo;
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index e2a124b3affb..e6bcbfe530ec 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -231,7 +231,8 @@ EXPORT_SYMBOL(ttm_bo_bulk_move_lru_tail);
 
 static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 				  struct ttm_resource *mem, bool evict,
-				  struct ttm_operation_ctx *ctx)
+				  struct ttm_operation_ctx *ctx,
+				  struct ttm_place *hop)
 {
 	struct ttm_bo_device *bdev = bo->bdev;
 	struct ttm_resource_manager *old_man = ttm_manager_type(bdev, bo->mem.mem_type);
@@ -259,9 +260,12 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 		}
 	}
 
-	ret = bdev->driver->move(bo, evict, ctx, mem);
-	if (ret)
+	ret = bdev->driver->move(bo, evict, ctx, mem, hop);
+	if (ret) {
+		if (ret == -EMULTIHOP)
+			return ret;
 		goto out_err;
+	}
 
 	ctx->bytes_moved += bo->num_pages << PAGE_SHIFT;
 	return 0;
@@ -566,8 +570,11 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
 	struct ttm_bo_device *bdev = bo->bdev;
 	struct ttm_resource evict_mem;
 	struct ttm_placement placement;
+	struct ttm_place hop;
 	int ret = 0;
 
+	memset(&hop, 0, sizeof(hop));
+
 	dma_resv_assert_held(bo->base.resv);
 
 	placement.num_placement = 0;
@@ -596,8 +603,9 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
 		goto out;
 	}
 
-	ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, ctx);
+	ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, ctx, &hop);
 	if (unlikely(ret)) {
+		WARN(ret == -EMULTIHOP, "Unexpected multihop in eviction - likely driver bug\n");
 		if (ret != -ERESTARTSYS)
 			pr_err("Buffer eviction failed\n");
 		ttm_resource_free(bo, &evict_mem);
@@ -936,15 +944,45 @@ error:
 }
 EXPORT_SYMBOL(ttm_bo_mem_space);
 
+static int ttm_bo_bounce_temp_buffer(struct ttm_buffer_object *bo,
+				     struct ttm_resource *mem,
+				     struct ttm_operation_ctx *ctx,
+				     struct ttm_place *hop)
+{
+	struct ttm_placement hop_placement;
+	int ret;
+	struct ttm_resource hop_mem = *mem;
+
+	hop_mem.mm_node = NULL;
+	hop_mem.mem_type = TTM_PL_SYSTEM;
+	hop_mem.placement = 0;
+
+	hop_placement.num_placement = hop_placement.num_busy_placement = 1;
+	hop_placement.placement = hop_placement.busy_placement = hop;
+
+	/* find space in the bounce domain */
+	ret = ttm_bo_mem_space(bo, &hop_placement, &hop_mem, ctx);
+	if (ret)
+		return ret;
+	/* move to the bounce domain */
+	ret = ttm_bo_handle_move_mem(bo, &hop_mem, false, ctx, NULL);
+	if (ret)
+		return ret;
+	return 0;
+}
+
 static int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
 			      struct ttm_placement *placement,
 			      struct ttm_operation_ctx *ctx)
 {
 	int ret = 0;
+	struct ttm_place hop;
 	struct ttm_resource mem;
 
 	dma_resv_assert_held(bo->base.resv);
 
+	memset(&hop, 0, sizeof(hop));
+
 	mem.num_pages = bo->num_pages;
 	mem.size = mem.num_pages << PAGE_SHIFT;
 	mem.page_alignment = bo->mem.page_alignment;
@@ -954,12 +992,25 @@ static int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
 
 	/*
 	 * Determine where to move the buffer.
+	 *
+	 * If driver determines move is going to need
+	 * an extra step then it will return -EMULTIHOP
+	 * and the buffer will be moved to the temporary
+	 * stop and the driver will be called to make
+	 * the second hop.
 	 */
+bounce:
 	ret = ttm_bo_mem_space(bo, placement, &mem, ctx);
 	if (ret)
-		goto out_unlock;
-	ret = ttm_bo_handle_move_mem(bo, &mem, false, ctx);
-out_unlock:
+		return ret;
+	ret = ttm_bo_handle_move_mem(bo, &mem, false, ctx, &hop);
+	if (ret == -EMULTIHOP) {
+		ret = ttm_bo_bounce_temp_buffer(bo, &mem, ctx, &hop);
+		if (ret)
+			return ret;
+		/* try and move to final place now. */
+		goto bounce;
+	}
 	if (ret)
 		ttm_resource_free(bo, &mem);
 	return ret;
@@ -1432,15 +1483,20 @@ int ttm_bo_swapout(struct ttm_operation_ctx *ctx)
 	if (bo->mem.mem_type != TTM_PL_SYSTEM) {
 		struct ttm_operation_ctx ctx = { false, false };
 		struct ttm_resource evict_mem;
+		struct ttm_place hop;
+
+		memset(&hop, 0, sizeof(hop));
 
 		evict_mem = bo->mem;
 		evict_mem.mm_node = NULL;
 		evict_mem.placement = 0;
 		evict_mem.mem_type = TTM_PL_SYSTEM;
 
-		ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, &ctx);
-		if (unlikely(ret != 0))
+		ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, &ctx, &hop);
+		if (unlikely(ret != 0)) {
+			WARN(ret == -EMULTIHOP, "Unexpected multihop in swaput - likely driver bug.\n");
 			goto out;
+		}
 	}
 
 	/**
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
index 51f70bea41cc..6a04261ce760 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c
@@ -695,7 +695,8 @@ static void vmw_swap_notify(struct ttm_buffer_object *bo)
 static int vmw_move(struct ttm_buffer_object *bo,
 		    bool evict,
 		    struct ttm_operation_ctx *ctx,
-		    struct ttm_resource *new_mem)
+		    struct ttm_resource *new_mem,
+		    struct ttm_place *hop)
 {
 	struct ttm_resource_manager *old_man = ttm_manager_type(bo->bdev, bo->mem.mem_type);
 	struct ttm_resource_manager *new_man = ttm_manager_type(bo->bdev, new_mem->mem_type);
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index da8208f43378..f02f7cf9ae90 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -121,6 +121,8 @@ struct ttm_bo_driver {
 	 * Return the bo flags for a buffer which is not mapped to the hardware.
 	 * These will be placed in proposed_flags so that when the move is
 	 * finished, they'll end up in bo->mem.flags
+	 * This should not cause multihop evictions, and the core will warn
+	 * if one is proposed.
 	 */
 
 	void (*evict_flags)(struct ttm_buffer_object *bo,
@@ -134,12 +136,15 @@ struct ttm_bo_driver {
 	 * the graphics address space
 	 * @ctx: context for this move with parameters
 	 * @new_mem: the new memory region receiving the buffer
+	 @ @hop: placement for driver directed intermediate hop
 	 *
 	 * Move a buffer between two memory regions.
+	 * Returns errno -EMULTIHOP if driver requests a hop
 	 */
 	int (*move)(struct ttm_buffer_object *bo, bool evict,
 		    struct ttm_operation_ctx *ctx,
-		    struct ttm_resource *new_mem);
+		    struct ttm_resource *new_mem,
+		    struct ttm_place *hop);
 
 	/**
 	 * struct ttm_bo_driver_member verify_access
-- 
cgit