Merge tag 'drm-fixes-2023-10-13' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Weekly fixes, the core is msm and amdgpu with some scattered fixes
  across vmwgfx, panel and the core stuff.

  atomic-helper:
   - Relax checks for unregistered connectors

  dma-buf:
   - Work around race condition when retrieving fence timestamp

  gem:
   - Avoid OOB access in BO memory range

  panel:
   - boe-tv101wun-ml6: Fix flickering

  simpledrm:
   - Fix error output

  vwmgfx:
   - Fix size calculation in texture-state code
   - Ref GEM BOs in surfaces

  msm:
   - PHY/link training reset fix
   - msm8998 - correct highest bank bit
   - skip video mode if timing engine disabled
   - check irq_of_parse_and_map return code
   - add new lines to some prints
   - fail atomic check for max mdp clk test

  amdgpu:
   - Seamless boot fix
   - Fix TTM BO resource check
   - SI fix for doorbell handling"

* tag 'drm-fixes-2023-10-13' of git://anongit.freedesktop.org/drm/drm:
  drm/tiny: correctly print `struct resource *` on error
  drm: Do not overrun array in drm_gem_get_pages()
  drm/atomic-helper: relax unregistered connector check
  drm/panel: boe-tv101wum-nl6: Completely pull GPW to VGL before TP term
  drm/amdgpu: fix SI failure due to doorbells allocation
  drm/amdgpu: add missing NULL check
  drm/amd/display: Don't set dpms_off for seamless boot
  drm/vmwgfx: Keep a gem reference to user bos in surfaces
  drm/vmwgfx: fix typo of sizeof argument
  drm/msm/dpu: fail dpu_plane_atomic_check() based on mdp clk limits
  dma-buf: add dma_fence_timestamp helper
  drm/msm/dp: Add newlines to debug printks
  drm/msm/dpu: change _dpu_plane_calc_bw() to use u64 to avoid overflow
  drm/msm/dsi: fix irq_of_parse_and_map() error checking
  drm/msm/dsi: skip the wait for video mode done if not applicable
  drm/msm/mdss: fix highest-bank-bit for msm8998
  drm/msm/dp: do not reinitialize phy unless retry during link training

27 files changed, 165 insertions, 100 deletions

diff --git a/drivers/dma-buf/dma-fence-unwrap.c b/drivers/dma-buf/dma-fence-unwrap.c
index c625bb2b5d56..628af51c81af 100644
--- a/drivers/dma-buf/dma-fence-unwrap.c
+++ b/drivers/dma-buf/dma-fence-unwrap.c
@@ -76,16 +76,11 @@ struct dma_fence *__dma_fence_unwrap_merge(unsigned int num_fences,
 		dma_fence_unwrap_for_each(tmp, &iter[i], fences[i]) {
 			if (!dma_fence_is_signaled(tmp)) {
 				++count;
-			} else if (test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT,
-					    &tmp->flags)) {
-				if (ktime_after(tmp->timestamp, timestamp))
-					timestamp = tmp->timestamp;
 			} else {
-				/*
-				 * Use the current time if the fence is
-				 * currently signaling.
-				 */
-				timestamp = ktime_get();
+				ktime_t t = dma_fence_timestamp(tmp);
+
+				if (ktime_after(t, timestamp))
+					timestamp = t;
 			}
 		}
 	}
diff --git a/drivers/dma-buf/sync_file.c b/drivers/dma-buf/sync_file.c
index af57799c86ce..2e9a316c596a 100644
--- a/drivers/dma-buf/sync_file.c
+++ b/drivers/dma-buf/sync_file.c
@@ -268,13 +268,10 @@ static int sync_fill_fence_info(struct dma_fence *fence,
 		sizeof(info->driver_name));
 
 	info->status = dma_fence_get_status(fence);
-	while (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags) &&
-	       !test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags))
-		cpu_relax();
 	info->timestamp_ns =
-		test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags) ?
-		ktime_to_ns(fence->timestamp) :
-		ktime_set(0, 0);
+		dma_fence_is_signaled(fence) ?
+			ktime_to_ns(dma_fence_timestamp(fence)) :
+			ktime_set(0, 0);
 
 	return info->status;
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
index da4be0bbb446..8eee5d783a92 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell_mgr.c
@@ -142,6 +142,10 @@ int amdgpu_doorbell_create_kernel_doorbells(struct amdgpu_device *adev)
 	int r;
 	int size;
 
+	/* SI HW does not have doorbells, skip allocation */
+	if (adev->doorbell.num_kernel_doorbells == 0)
+		return 0;
+
 	/* Reserve first num_kernel_doorbells (page-aligned) for kernel ops */
 	size = ALIGN(adev->doorbell.num_kernel_doorbells * sizeof(u32), PAGE_SIZE);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
index f3ee83cdf97e..d28e21baef16 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
@@ -252,7 +252,7 @@ static inline bool amdgpu_bo_in_cpu_visible_vram(struct amdgpu_bo *bo)
 	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
 	struct amdgpu_res_cursor cursor;
 
-	if (bo->tbo.resource->mem_type != TTM_PL_VRAM)
+	if (!bo->tbo.resource || bo->tbo.resource->mem_type != TTM_PL_VRAM)
 		return false;
 
 	amdgpu_res_first(bo->tbo.resource, 0, amdgpu_bo_size(bo), &cursor);
diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c
index 3a9077b60029..d08e60dff46d 100644
--- a/drivers/gpu/drm/amd/display/dc/core/dc.c
+++ b/drivers/gpu/drm/amd/display/dc/core/dc.c
@@ -1262,6 +1262,9 @@ static void disable_vbios_mode_if_required(
 		if (stream == NULL)
 			continue;
 
+		if (stream->apply_seamless_boot_optimization)
+			continue;
+
 		// only looking for first odm pipe
 		if (pipe->prev_odm_pipe)
 			continue;
diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
index 292e38eb6218..60794fcde1d5 100644
--- a/drivers/gpu/drm/drm_atomic_helper.c
+++ b/drivers/gpu/drm/drm_atomic_helper.c
@@ -290,7 +290,8 @@ static int
 update_connector_routing(struct drm_atomic_state *state,
 			 struct drm_connector *connector,
 			 struct drm_connector_state *old_connector_state,
-			 struct drm_connector_state *new_connector_state)
+			 struct drm_connector_state *new_connector_state,
+			 bool added_by_user)
 {
 	const struct drm_connector_helper_funcs *funcs;
 	struct drm_encoder *new_encoder;
@@ -339,9 +340,13 @@ update_connector_routing(struct drm_atomic_state *state,
 	 * there's a chance the connector may have been destroyed during the
 	 * process, but it's better to ignore that then cause
 	 * drm_atomic_helper_resume() to fail.
+	 *
+	 * Last, we want to ignore connector registration when the connector
+	 * was not pulled in the atomic state by user-space (ie, was pulled
+	 * in by the driver, e.g. when updating a DP-MST stream).
 	 */
 	if (!state->duplicated && drm_connector_is_unregistered(connector) &&
-	    crtc_state->active) {
+	    added_by_user && crtc_state->active) {
 		drm_dbg_atomic(connector->dev,
 			       "[CONNECTOR:%d:%s] is not registered\n",
 			       connector->base.id, connector->name);
@@ -620,7 +625,10 @@ drm_atomic_helper_check_modeset(struct drm_device *dev,
 	struct drm_connector *connector;
 	struct drm_connector_state *old_connector_state, *new_connector_state;
 	int i, ret;
-	unsigned int connectors_mask = 0;
+	unsigned int connectors_mask = 0, user_connectors_mask = 0;
+
+	for_each_oldnew_connector_in_state(state, connector, old_connector_state, new_connector_state, i)
+		user_connectors_mask |= BIT(i);
 
 	for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) {
 		bool has_connectors =
@@ -685,7 +693,8 @@ drm_atomic_helper_check_modeset(struct drm_device *dev,
 		 */
 		ret = update_connector_routing(state, connector,
 					       old_connector_state,
-					       new_connector_state);
+					       new_connector_state,
+					       BIT(i) & user_connectors_mask);
 		if (ret)
 			return ret;
 		if (old_connector_state->crtc) {
diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
index 6129b89bb366..44a948b80ee1 100644
--- a/drivers/gpu/drm/drm_gem.c
+++ b/drivers/gpu/drm/drm_gem.c
@@ -540,7 +540,7 @@ struct page **drm_gem_get_pages(struct drm_gem_object *obj)
 	struct page **pages;
 	struct folio *folio;
 	struct folio_batch fbatch;
-	int i, j, npages;
+	long i, j, npages;
 
 	if (WARN_ON(!obj->filp))
 		return ERR_PTR(-EINVAL);
@@ -564,11 +564,13 @@ struct page **drm_gem_get_pages(struct drm_gem_object *obj)
 
 	i = 0;
 	while (i < npages) {
+		long nr;
 		folio = shmem_read_folio_gfp(mapping, i,
 				mapping_gfp_mask(mapping));
 		if (IS_ERR(folio))
 			goto fail;
-		for (j = 0; j < folio_nr_pages(folio); j++, i++)
+		nr = min(npages - i, folio_nr_pages(folio));
+		for (j = 0; j < nr; j++, i++)
 			pages[i] = folio_file_page(folio, i);
 
 		/* Make sure shmem keeps __GFP_DMA32 allocated pages in the
diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
index c2aaaded07ed..0be195f9149c 100644
--- a/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
+++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_plane.c
@@ -119,6 +119,7 @@ static u64 _dpu_plane_calc_bw(const struct dpu_mdss_cfg *catalog,
 	struct dpu_sw_pipe_cfg *pipe_cfg)
 {
 	int src_width, src_height, dst_height, fps;
+	u64 plane_pixel_rate, plane_bit_rate;
 	u64 plane_prefill_bw;
 	u64 plane_bw;
 	u32 hw_latency_lines;
@@ -136,13 +137,12 @@ static u64 _dpu_plane_calc_bw(const struct dpu_mdss_cfg *catalog,
 	scale_factor = src_height > dst_height ?
 		mult_frac(src_height, 1, dst_height) : 1;
 
-	plane_bw =
-		src_width * mode->vtotal * fps * fmt->bpp *
-		scale_factor;
+	plane_pixel_rate = src_width * mode->vtotal * fps;
+	plane_bit_rate = plane_pixel_rate * fmt->bpp;
 
-	plane_prefill_bw =
-		src_width * hw_latency_lines * fps * fmt->bpp *
-		scale_factor * mode->vtotal;
+	plane_bw = plane_bit_rate * scale_factor;
+
+	plane_prefill_bw = plane_bw * hw_latency_lines;
 
 	if ((vbp+vpw) > hw_latency_lines)
 		do_div(plane_prefill_bw, (vbp+vpw));
@@ -733,9 +733,11 @@ static int dpu_plane_check_inline_rotation(struct dpu_plane *pdpu,
 static int dpu_plane_atomic_check_pipe(struct dpu_plane *pdpu,
 		struct dpu_sw_pipe *pipe,
 		struct dpu_sw_pipe_cfg *pipe_cfg,
-		const struct dpu_format *fmt)
+		const struct dpu_format *fmt,
+		const struct drm_display_mode *mode)
 {
 	uint32_t min_src_size;
+	struct dpu_kms *kms = _dpu_plane_get_kms(&pdpu->base);
 
 	min_src_size = DPU_FORMAT_IS_YUV(fmt) ? 2 : 1;
 
@@ -774,6 +776,12 @@ static int dpu_plane_atomic_check_pipe(struct dpu_plane *pdpu,
 		return -EINVAL;
 	}
 
+	/* max clk check */
+	if (_dpu_plane_calc_clk(mode, pipe_cfg) > kms->perf.max_core_clk_rate) {
+		DPU_DEBUG_PLANE(pdpu, "plane exceeds max mdp core clk limits\n");
+		return -E2BIG;
+	}
+
 	return 0;
 }
 
@@ -899,12 +907,13 @@ static int dpu_plane_atomic_check(struct drm_plane *plane,
 		r_pipe_cfg->dst_rect.x1 = pipe_cfg->dst_rect.x2;
 	}
 
-	ret = dpu_plane_atomic_check_pipe(pdpu, pipe, pipe_cfg, fmt);
+	ret = dpu_plane_atomic_check_pipe(pdpu, pipe, pipe_cfg, fmt, &crtc_state->adjusted_mode);
 	if (ret)
 		return ret;
 
 	if (r_pipe->sspp) {
-		ret = dpu_plane_atomic_check_pipe(pdpu, r_pipe, r_pipe_cfg, fmt);
+		ret = dpu_plane_atomic_check_pipe(pdpu, r_pipe, r_pipe_cfg, fmt,
+						  &crtc_state->adjusted_mode);
 		if (ret)
 			return ret;
 	}
diff --git a/drivers/gpu/drm/msm/dp/dp_ctrl.c b/drivers/gpu/drm/msm/dp/dp_ctrl.c
index a7a5c7e0ab92..77a8d9366ed7 100644
--- a/drivers/gpu/drm/msm/dp/dp_ctrl.c
+++ b/drivers/gpu/drm/msm/dp/dp_ctrl.c
@@ -1774,13 +1774,6 @@ int dp_ctrl_on_link(struct dp_ctrl *dp_ctrl)
 		return rc;
 
 	while (--link_train_max_retries) {
-		rc = dp_ctrl_reinitialize_mainlink(ctrl);
-		if (rc) {
-			DRM_ERROR("Failed to reinitialize mainlink. rc=%d\n",
-					rc);
-			break;
-		}
-
 		training_step = DP_TRAINING_NONE;
 		rc = dp_ctrl_setup_main_link(ctrl, &training_step);
 		if (rc == 0) {
@@ -1832,6 +1825,12 @@ int dp_ctrl_on_link(struct dp_ctrl *dp_ctrl)
 			/* stop link training before start re training  */
 			dp_ctrl_clear_training_pattern(ctrl);
 		}
+
+		rc = dp_ctrl_reinitialize_mainlink(ctrl);
+		if (rc) {
+			DRM_ERROR("Failed to reinitialize mainlink. rc=%d\n", rc);
+			break;
+		}
 	}
 
 	if (ctrl->link->sink_request & DP_TEST_LINK_PHY_TEST_PATTERN)
diff --git a/drivers/gpu/drm/msm/dp/dp_link.c b/drivers/gpu/drm/msm/dp/dp_link.c
index 42427129acea..6375daaeb98e 100644
--- a/drivers/gpu/drm/msm/dp/dp_link.c
+++ b/drivers/gpu/drm/msm/dp/dp_link.c
@@ -1090,7 +1090,7 @@ int dp_link_process_request(struct dp_link *dp_link)
 	} else if (dp_link_read_psr_error_status(link)) {
 		DRM_ERROR("PSR IRQ_HPD received\n");
 	} else if (dp_link_psr_capability_changed(link)) {
-		drm_dbg_dp(link->drm_dev, "PSR Capability changed");
+		drm_dbg_dp(link->drm_dev, "PSR Capability changed\n");
 	} else {
 		ret = dp_link_process_link_status_update(link);
 		if (!ret) {
@@ -1107,7 +1107,7 @@ int dp_link_process_request(struct dp_link *dp_link)
 		}
 	}
 
-	drm_dbg_dp(link->drm_dev, "sink request=%#x",
+	drm_dbg_dp(link->drm_dev, "sink request=%#x\n",
 				dp_link->sink_request);
 	return ret;
 }
diff --git a/drivers/gpu/drm/msm/dsi/dsi_host.c b/drivers/gpu/drm/msm/dsi/dsi_host.c
index 5d9ec27c89d3..3d6fb708dc22 100644
--- a/drivers/gpu/drm/msm/dsi/dsi_host.c
+++ b/drivers/gpu/drm/msm/dsi/dsi_host.c
@@ -1082,9 +1082,21 @@ static void dsi_wait4video_done(struct msm_dsi_host *msm_host)
 
 static void dsi_wait4video_eng_busy(struct msm_dsi_host *msm_host)
 {
+	u32 data;
+
 	if (!(msm_host->mode_flags & MIPI_DSI_MODE_VIDEO))
 		return;
 
+	data = dsi_read(msm_host, REG_DSI_STATUS0);
+
+	/* if video mode engine is not busy, its because
+	 * either timing engine was not turned on or the
+	 * DSI controller has finished transmitting the video
+	 * data already, so no need to wait in those cases
+	 */
+	if (!(data & DSI_STATUS0_VIDEO_MODE_ENGINE_BUSY))
+		return;
+
 	if (msm_host->power_on && msm_host->enabled) {
 		dsi_wait4video_done(msm_host);
 		/* delay 4 ms to skip BLLP */
@@ -1894,10 +1906,9 @@ int msm_dsi_host_init(struct msm_dsi *msm_dsi)
 	}
 
 	msm_host->irq = irq_of_parse_and_map(pdev->dev.of_node, 0);
-	if (msm_host->irq < 0) {
-		ret = msm_host->irq;
-		dev_err(&pdev->dev, "failed to get irq: %d\n", ret);
-		return ret;
+	if (!msm_host->irq) {
+		dev_err(&pdev->dev, "failed to get irq\n");
+		return -EINVAL;
 	}
 
 	/* do not autoenable, will be enabled later */
diff --git a/drivers/gpu/drm/msm/msm_mdss.c b/drivers/gpu/drm/msm/msm_mdss.c
index 2e87dd6cb17b..348c66b14683 100644
--- a/drivers/gpu/drm/msm/msm_mdss.c
+++ b/drivers/gpu/drm/msm/msm_mdss.c
@@ -511,7 +511,7 @@ static int mdss_remove(struct platform_device *pdev)
 static const struct msm_mdss_data msm8998_data = {
 	.ubwc_enc_version = UBWC_1_0,
 	.ubwc_dec_version = UBWC_1_0,
-	.highest_bank_bit = 1,
+	.highest_bank_bit = 2,
 };
 
 static const struct msm_mdss_data qcm2290_data = {
diff --git a/drivers/gpu/drm/panel/panel-boe-tv101wum-nl6.c b/drivers/gpu/drm/panel/panel-boe-tv101wum-nl6.c
index 5ac926281d2c..c9087f474cbc 100644
--- a/drivers/gpu/drm/panel/panel-boe-tv101wum-nl6.c
+++ b/drivers/gpu/drm/panel/panel-boe-tv101wum-nl6.c
@@ -1342,9 +1342,7 @@ static const struct panel_init_cmd starry_himax83102_j02_init_cmd[] = {
 	_INIT_DCS_CMD(0xB1, 0x01, 0xBF, 0x11),
 	_INIT_DCS_CMD(0xCB, 0x86),
 	_INIT_DCS_CMD(0xD2, 0x3C, 0xFA),
-	_INIT_DCS_CMD(0xE9, 0xC5),
-	_INIT_DCS_CMD(0xD3, 0x00, 0x00, 0x00, 0x00, 0x80, 0x0C, 0x01),
-	_INIT_DCS_CMD(0xE9, 0x3F),
+	_INIT_DCS_CMD(0xD3, 0x00, 0x00, 0x44, 0x00, 0x00, 0x00, 0x00, 0x00, 0x80, 0x0C, 0x01),
 	_INIT_DCS_CMD(0xE7, 0x02, 0x00, 0x28, 0x01, 0x7E, 0x0F, 0x7E, 0x10, 0xA0, 0x00, 0x00, 0x20, 0x40, 0x50, 0x40),
 	_INIT_DCS_CMD(0xBD, 0x02),
 	_INIT_DCS_CMD(0xD8, 0xFF, 0xFF, 0xBF, 0xFE, 0xAA, 0xA0, 0xFF, 0xFF, 0xBF, 0xFE, 0xAA, 0xA0),
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 506371c42745..5a3a622fc672 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -929,7 +929,7 @@ drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched)
 
 		if (next) {
 			next->s_fence->scheduled.timestamp =
-				job->s_fence->finished.timestamp;
+				dma_fence_timestamp(&job->s_fence->finished);
 			/* start TO timer for next job */
 			drm_sched_start_timeout(sched);
 		}
diff --git a/drivers/gpu/drm/tiny/simpledrm.c b/drivers/gpu/drm/tiny/simpledrm.c
index ff86ba1ae1b8..8ea120eb8674 100644
--- a/drivers/gpu/drm/tiny/simpledrm.c
+++ b/drivers/gpu/drm/tiny/simpledrm.c
@@ -745,7 +745,7 @@ static struct simpledrm_device *simpledrm_device_create(struct drm_driver *drv,
 
 		ret = devm_aperture_acquire_from_firmware(dev, res->start, resource_size(res));
 		if (ret) {
-			drm_err(dev, "could not acquire memory range %pr: %d\n", &res, ret);
+			drm_err(dev, "could not acquire memory range %pr: %d\n", res, ret);
 			return ERR_PTR(ret);
 		}
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
index c43853597776..2bfac3aad7b7 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.c
@@ -34,6 +34,8 @@
 
 static void vmw_bo_release(struct vmw_bo *vbo)
 {
+	WARN_ON(vbo->tbo.base.funcs &&
+		kref_read(&vbo->tbo.base.refcount) != 0);
 	vmw_bo_unmap(vbo);
 	drm_gem_object_release(&vbo->tbo.base);
 }
@@ -497,7 +499,7 @@ static int vmw_user_bo_synccpu_release(struct drm_file *filp,
 		if (!(flags & drm_vmw_synccpu_allow_cs)) {
 			atomic_dec(&vmw_bo->cpu_writers);
 		}
-		vmw_user_bo_unref(vmw_bo);
+		vmw_user_bo_unref(&vmw_bo);
 	}
 
 	return ret;
@@ -539,7 +541,7 @@ int vmw_user_bo_synccpu_ioctl(struct drm_device *dev, void *data,
 			return ret;
 
 		ret = vmw_user_bo_synccpu_grab(vbo, arg->flags);
-		vmw_user_bo_unref(vbo);
+		vmw_user_bo_unref(&vbo);
 		if (unlikely(ret != 0)) {
 			if (ret == -ERESTARTSYS || ret == -EBUSY)
 				return -EBUSY;
@@ -612,7 +614,6 @@ int vmw_user_bo_lookup(struct drm_file *filp,
 	}
 
 	*out = to_vmw_bo(gobj);
-	ttm_bo_get(&(*out)->tbo);
 
 	return 0;
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.h b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.h
index 1d433fceed3d..0d496dc9c6af 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_bo.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_bo.h
@@ -195,12 +195,19 @@ static inline struct vmw_bo *vmw_bo_reference(struct vmw_bo *buf)
 	return buf;
 }
 
-static inline void vmw_user_bo_unref(struct vmw_bo *vbo)
+static inline struct vmw_bo *vmw_user_bo_ref(struct vmw_bo *vbo)
 {
-	if (vbo) {
-		ttm_bo_put(&vbo->tbo);
-		drm_gem_object_put(&vbo->tbo.base);
-	}
+	drm_gem_object_get(&vbo->tbo.base);
+	return vbo;
+}
+
+static inline void vmw_user_bo_unref(struct vmw_bo **buf)
+{
+	struct vmw_bo *tmp_buf = *buf;
+
+	*buf = NULL;
+	if (tmp_buf)
+		drm_gem_object_put(&tmp_buf->tbo.base);
 }
 
 static inline struct vmw_bo *to_vmw_bo(struct drm_gem_object *gobj)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c b/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c
index c0b24d1cacbf..a7c07692262b 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_cotable.c
@@ -432,7 +432,7 @@ static int vmw_cotable_resize(struct vmw_resource *res, size_t new_size)
 	 * for the new COTable. Initially pin the buffer object to make sure
 	 * we can use tryreserve without failure.
 	 */
-	ret = vmw_bo_create(dev_priv, &bo_params, &buf);
+	ret = vmw_gem_object_create(dev_priv, &bo_params, &buf);
 	if (ret) {
 		DRM_ERROR("Failed initializing new cotable MOB.\n");
 		goto out_done;
@@ -502,7 +502,7 @@ static int vmw_cotable_resize(struct vmw_resource *res, size_t new_size)
 
 	vmw_resource_mob_attach(res);
 	/* Let go of the old mob. */
-	vmw_bo_unreference(&old_buf);
+	vmw_user_bo_unref(&old_buf);
 	res->id = vcotbl->type;
 
 	ret = dma_resv_reserve_fences(bo->base.resv, 1);
@@ -521,7 +521,7 @@ out_map_new:
 out_wait:
 	ttm_bo_unpin(bo);
 	ttm_bo_unreserve(bo);
-	vmw_bo_unreference(&buf);
+	vmw_user_bo_unref(&buf);
 
 out_done:
 	MKS_STAT_TIME_POP(MKSSTAT_KERN_COTABLE_RESIZE);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
index 58bfdf203eca..3cd5090dedfc 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.h
@@ -853,6 +853,10 @@ static inline bool vmw_resource_mob_attached(const struct vmw_resource *res)
 /**
  * GEM related functionality - vmwgfx_gem.c
  */
+struct vmw_bo_params;
+int vmw_gem_object_create(struct vmw_private *vmw,
+			  struct vmw_bo_params *params,
+			  struct vmw_bo **p_vbo);
 extern int vmw_gem_object_create_with_handle(struct vmw_private *dev_priv,
 					     struct drm_file *filp,
 					     uint32_t size,
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
index 98e0723ca6f5..36987ef3fc30 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c
@@ -1151,7 +1151,7 @@ static int vmw_translate_mob_ptr(struct vmw_private *dev_priv,
 				 SVGAMobId *id,
 				 struct vmw_bo **vmw_bo_p)
 {
-	struct vmw_bo *vmw_bo;
+	struct vmw_bo *vmw_bo, *tmp_bo;
 	uint32_t handle = *id;
 	struct vmw_relocation *reloc;
 	int ret;
@@ -1164,7 +1164,8 @@ static int vmw_translate_mob_ptr(struct vmw_private *dev_priv,
 	}
 	vmw_bo_placement_set(vmw_bo, VMW_BO_DOMAIN_MOB, VMW_BO_DOMAIN_MOB);
 	ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo);
-	vmw_user_bo_unref(vmw_bo);
+	tmp_bo = vmw_bo;
+	vmw_user_bo_unref(&tmp_bo);
 	if (unlikely(ret != 0))
 		return ret;
 
@@ -1206,7 +1207,7 @@ static int vmw_translate_guest_ptr(struct vmw_private *dev_priv,
 				   SVGAGuestPtr *ptr,
 				   struct vmw_bo **vmw_bo_p)
 {
-	struct vmw_bo *vmw_bo;
+	struct vmw_bo *vmw_bo, *tmp_bo;
 	uint32_t handle = ptr->gmrId;
 	struct vmw_relocation *reloc;
 	int ret;
@@ -1220,7 +1221,8 @@ static int vmw_translate_guest_ptr(struct vmw_private *dev_priv,
 	vmw_bo_placement_set(vmw_bo, VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM,
 			     VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM);
 	ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo);
-	vmw_user_bo_unref(vmw_bo);
+	tmp_bo = vmw_bo;
+	vmw_user_bo_unref(&tmp_bo);
 	if (unlikely(ret != 0))
 		return ret;
 
@@ -1619,7 +1621,7 @@ static int vmw_cmd_tex_state(struct vmw_private *dev_priv,
 {
 	VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdSetTextureState);
 	SVGA3dTextureState *last_state = (SVGA3dTextureState *)
-	  ((unsigned long) header + header->size + sizeof(header));
+	  ((unsigned long) header + header->size + sizeof(*header));
 	SVGA3dTextureState *cur_state = (SVGA3dTextureState *)
 		((unsigned long) header + sizeof(*cmd));
 	struct vmw_resource *ctx;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_gem.c b/drivers/gpu/drm/vmwgfx/vmwgfx_gem.c
index c0da89e16e6f..8b1eb0061610 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_gem.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_gem.c
@@ -111,6 +111,20 @@ static const struct drm_gem_object_funcs vmw_gem_object_funcs = {
 	.vm_ops = &vmw_vm_ops,
 };
 
+int vmw_gem_object_create(struct vmw_private *vmw,
+			  struct vmw_bo_params *params,
+			  struct vmw_bo **p_vbo)
+{
+	int ret = vmw_bo_create(vmw, params, p_vbo);
+
+	if (ret != 0)
+		goto out_no_bo;
+
+	(*p_vbo)->tbo.base.funcs = &vmw_gem_object_funcs;
+out_no_bo:
+	return ret;
+}
+
 int vmw_gem_object_create_with_handle(struct vmw_private *dev_priv,
 				      struct drm_file *filp,
 				      uint32_t size,
@@ -126,12 +140,10 @@ int vmw_gem_object_create_with_handle(struct vmw_private *dev_priv,
 		.pin = false
 	};
 
-	ret = vmw_bo_create(dev_priv, &params, p_vbo);
+	ret = vmw_gem_object_create(dev_priv, &params, p_vbo);
 	if (ret != 0)
 		goto out_no_bo;
 
-	(*p_vbo)->tbo.base.funcs = &vmw_gem_object_funcs;
-
 	ret = drm_gem_handle_create(filp, &(*p_vbo)->tbo.base, handle);
 out_no_bo:
 	return ret;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
index 1489ad73c103..818b7f109f53 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c
@@ -1471,8 +1471,8 @@ static int vmw_create_bo_proxy(struct drm_device *dev,
 	/* Reserve and switch the backing mob. */
 	mutex_lock(&res->dev_priv->cmdbuf_mutex);
 	(void) vmw_resource_reserve(res, false, true);
-	vmw_bo_unreference(&res->guest_memory_bo);
-	res->guest_memory_bo = vmw_bo_reference(bo_mob);
+	vmw_user_bo_unref(&res->guest_memory_bo);
+	res->guest_memory_bo = vmw_user_bo_ref(bo_mob);
 	res->guest_memory_offset = 0;
 	vmw_resource_unreserve(res, false, false, false, NULL, 0);
 	mutex_unlock(&res->dev_priv->cmdbuf_mutex);
@@ -1666,7 +1666,7 @@ static struct drm_framebuffer *vmw_kms_fb_create(struct drm_device *dev,
 err_out:
 	/* vmw_user_lookup_handle takes one ref so does new_fb */
 	if (bo)
-		vmw_user_bo_unref(bo);
+		vmw_user_bo_unref(&bo);
 	if (surface)
 		vmw_surface_unreference(&surface);
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
index fb85f244c3d0..c45b4724e414 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_overlay.c
@@ -451,7 +451,7 @@ int vmw_overlay_ioctl(struct drm_device *dev, void *data,
 
 	ret = vmw_overlay_update_stream(dev_priv, buf, arg, true);
 
-	vmw_user_bo_unref(buf);
+	vmw_user_bo_unref(&buf);
 
 out_unlock:
 	mutex_unlock(&overlay->mutex);
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
index 71eeabf001c8..ca300c7427d2 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_resource.c
@@ -141,7 +141,7 @@ static void vmw_resource_release(struct kref *kref)
 		if (res->coherent)
 			vmw_bo_dirty_release(res->guest_memory_bo);
 		ttm_bo_unreserve(bo);
-		vmw_bo_unreference(&res->guest_memory_bo);
+		vmw_user_bo_unref(&res->guest_memory_bo);
 	}
 
 	if (likely(res->hw_destroy != NULL)) {
@@ -338,7 +338,7 @@ static int vmw_resource_buf_alloc(struct vmw_resource *res,
 		return 0;
 	}
 
-	ret = vmw_bo_create(res->dev_priv, &bo_params, &gbo);
+	ret = vmw_gem_object_create(res->dev_priv, &bo_params, &gbo);
 	if (unlikely(ret != 0))
 		goto out_no_bo;
 
@@ -457,11 +457,11 @@ void vmw_resource_unreserve(struct vmw_resource *res,
 			vmw_resource_mob_detach(res);
 			if (res->coherent)
 				vmw_bo_dirty_release(res->guest_memory_bo);
-			vmw_bo_unreference(&res->guest_memory_bo);
+			vmw_user_bo_unref(&res->guest_memory_bo);
 		}
 
 		if (new_guest_memory_bo) {
-			res->guest_memory_bo = vmw_bo_reference(new_guest_memory_bo);
+			res->guest_memory_bo = vmw_user_bo_ref(new_guest_memory_bo);
 
 			/*
 			 * The validation code should already have added a
@@ -551,7 +551,7 @@ out_no_reserve:
 	ttm_bo_put(val_buf->bo);
 	val_buf->bo = NULL;
 	if (guest_memory_dirty)
-		vmw_bo_unreference(&res->guest_memory_bo);
+		vmw_user_bo_unref(&res->guest_memory_bo);
 
 	return ret;
 }
@@ -727,7 +727,7 @@ int vmw_resource_validate(struct vmw_resource *res, bool intr,
 		goto out_no_validate;
 	else if (!res->func->needs_guest_memory && res->guest_memory_bo) {
 		WARN_ON_ONCE(vmw_resource_mob_attached(res));
-		vmw_bo_unreference(&res->guest_memory_bo);
+		vmw_user_bo_unref(&res->guest_memory_bo);
 	}
 
 	return 0;
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
index 1e81ff2422cf..a01ca3226d0a 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_shader.c
@@ -180,7 +180,7 @@ static int vmw_gb_shader_init(struct vmw_private *dev_priv,
 
 	res->guest_memory_size = size;
 	if (byte_code) {
-		res->guest_memory_bo = vmw_bo_reference(byte_code);
+		res->guest_memory_bo = vmw_user_bo_ref(byte_code);
 		res->guest_memory_offset = offset;
 	}
 	shader->size = size;
@@ -809,7 +809,7 @@ static int vmw_shader_define(struct drm_device *dev, struct drm_file *file_priv,
 				    shader_type, num_input_sig,
 				    num_output_sig, tfile, shader_handle);
 out_bad_arg:
-	vmw_user_bo_unref(buffer);
+	vmw_user_bo_unref(&buffer);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
index 5db403ee8261..3829be282ff0 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_surface.c
@@ -686,9 +686,6 @@ static void vmw_user_surface_base_release(struct ttm_base_object **p_base)
 	    container_of(base, struct vmw_user_surface, prime.base);
 	struct vmw_resource *res = &user_srf->srf.res;
 
-	if (res->guest_memory_bo)
-		drm_gem_object_put(&res->guest_memory_bo->tbo.base);
-
 	*p_base = NULL;
 	vmw_resource_unreference(&res);
 }
@@ -855,23 +852,21 @@ int vmw_surface_define_ioctl(struct drm_device *dev, void *data,
 	 * expect a backup buffer to be present.
 	 */
 	if (dev_priv->has_mob && req->shareable) {
-		uint32_t backup_handle;
-
-		ret = vmw_gem_object_create_with_handle(dev_priv,
-							file_priv,
-							res->guest_memory_size,
-							&backup_handle,
-							&res->guest_memory_bo);
+		struct vmw_bo_params params = {
+			.domain = VMW_BO_DOMAIN_SYS,
+			.busy_domain = VMW_BO_DOMAIN_SYS,
+			.bo_type = ttm_bo_type_device,
+			.size = res->guest_memory_size,
+			.pin = false
+		};
+
+		ret = vmw_gem_object_create(dev_priv,
+					    &params,
+					    &res->guest_memory_bo);
 		if (unlikely(ret != 0)) {
 			vmw_resource_unreference(&res);
 			goto out_unlock;
 		}
-		vmw_bo_reference(res->guest_memory_bo);
-		/*
-		 * We don't expose the handle to the userspace and surface
-		 * already holds a gem reference
-		 */
-		drm_gem_handle_delete(file_priv, backup_handle);
 	}
 
 	tmp = vmw_resource_reference(&srf->res);
@@ -1512,7 +1507,7 @@ vmw_gb_surface_define_internal(struct drm_device *dev,
 		if (ret == 0) {
 			if (res->guest_memory_bo->tbo.base.size < res->guest_memory_size) {
 				VMW_DEBUG_USER("Surface backup buffer too small.\n");
-				vmw_bo_unreference(&res->guest_memory_bo);
+				vmw_user_bo_unref(&res->guest_memory_bo);
 				ret = -EINVAL;
 				goto out_unlock;
 			} else {
@@ -1526,8 +1521,6 @@ vmw_gb_surface_define_internal(struct drm_device *dev,
 							res->guest_memory_size,
 							&backup_handle,
 							&res->guest_memory_bo);
-		if (ret == 0)
-			vmw_bo_reference(res->guest_memory_bo);
 	}
 
 	if (unlikely(ret != 0)) {
diff --git a/include/linux/dma-fence.h b/include/linux/dma-fence.h
index 0d678e9a7b24..ebe78bd3d121 100644
--- a/include/linux/dma-fence.h
+++ b/include/linux/dma-fence.h
@@ -568,6 +568,25 @@ static inline void dma_fence_set_error(struct dma_fence *fence,
 	fence->error = error;
 }
 
+/**
+ * dma_fence_timestamp - helper to get the completion timestamp of a fence
+ * @fence: fence to get the timestamp from.
+ *
+ * After a fence is signaled the timestamp is updated with the signaling time,
+ * but setting the timestamp can race with tasks waiting for the signaling. This
+ * helper busy waits for the correct timestamp to appear.
+ */
+static inline ktime_t dma_fence_timestamp(struct dma_fence *fence)
+{
+	if (WARN_ON(!test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)))
+		return ktime_get();
+
+	while (!test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, &fence->flags))
+		cpu_relax();
+
+	return fence->timestamp;
+}
+
 signed long dma_fence_wait_timeout(struct dma_fence *,
 				   bool intr, signed long timeout);
 signed long dma_fence_wait_any_timeout(struct dma_fence **fences,