rpi-4.1.y pull: vc4: kernel-side fix for X rendering lag, plus 3D testing fix

drivers/gpu/drm/drm_fb_cma_helper.c

@@ -279,7 +279,7 @@ static int drm_fbdev_cma_create(struct drm_fb_helper *helper,
 	if (!fbi) {
 		dev_err(dev->dev, "Failed to allocate framebuffer info.\n");
 		ret = -ENOMEM;
-		goto err_drm_gem_cma_free_object;
+		goto err_gem_free_object;
 	}
 
 	fbdev_cma->fb = drm_fb_cma_alloc(dev, &mode_cmd, &obj, 1);
@@ -322,8 +322,8 @@ static int drm_fbdev_cma_create(struct drm_fb_helper *helper,
 	drm_fb_cma_destroy(fb);
 err_framebuffer_release:
 	framebuffer_release(fbi);
-err_drm_gem_cma_free_object:
-	drm_gem_cma_free_object(&obj->base);
+err_gem_free_object:
+	dev->driver->gem_free_object(&obj->base);
 	return ret;
 }
 

drivers/gpu/drm/drm_gem_cma_helper.c

@@ -121,7 +121,7 @@ struct drm_gem_cma_object *drm_gem_cma_create(struct drm_device *drm,
 	return cma_obj;
 
 error:
-	drm_gem_cma_free_object(&cma_obj->base);
+	drm->driver->gem_free_object(&cma_obj->base);
 	return ERR_PTR(ret);
 }
 EXPORT_SYMBOL_GPL(drm_gem_cma_create);
@@ -171,7 +171,7 @@ drm_gem_cma_create_with_handle(struct drm_file *file_priv,
 	return cma_obj;
 
 err_handle_create:
-	drm_gem_cma_free_object(gem_obj);
+	drm->driver->gem_free_object(gem_obj);
 
 	return ERR_PTR(ret);
 }

drivers/gpu/drm/vc4/vc4_bo.c

@@ -213,10 +213,10 @@ struct vc4_bo *vc4_bo_create(struct drm_device *dev, size_t unaligned_size,
 	size_t size = roundup(unaligned_size, PAGE_SIZE);
 	struct vc4_dev *vc4 = to_vc4_dev(dev);
 	struct drm_gem_cma_object *cma_obj;
-	int pass;
+	int pass, ret;
 
 	if (size == 0)
-		return NULL;
+		return ERR_PTR(-EINVAL);
 
 	/* First, try to get a vc4_bo from the kernel BO cache. */
 	if (from_cache) {
@@ -247,14 +247,17 @@ struct vc4_bo *vc4_bo_create(struct drm_device *dev, size_t unaligned_size,
 			 * unreferenced BOs to the cache, and then
 			 * free the cache.
 			 */
-			vc4_wait_for_seqno(dev, vc4->emit_seqno, ~0ull, true);
+			ret = vc4_wait_for_seqno(dev, vc4->emit_seqno, ~0ull,
+						 true);
+			if (ret)
+				return ERR_PTR(ret);
 			vc4_job_handle_completed(vc4);
 			vc4_bo_cache_purge(dev);
 			break;
 		case 3:
 			DRM_ERROR("Failed to allocate from CMA:\n");
 			vc4_bo_stats_dump(vc4);
-			return NULL;
+			return ERR_PTR(-ENOMEM);
 		}
 	}
 
@@ -276,8 +279,8 @@ int vc4_dumb_create(struct drm_file *file_priv,
 		args->size = args->pitch * args->height;
 
 	bo = vc4_bo_create(dev, args->size, false);
-	if (!bo)
-		return -ENOMEM;
+	if (IS_ERR(bo))
+		return PTR_ERR(bo);
 
 	ret = drm_gem_handle_create(file_priv, &bo->base.base, &args->handle);
 	drm_gem_object_unreference_unlocked(&bo->base.base);
@@ -460,8 +463,8 @@ int vc4_create_bo_ioctl(struct drm_device *dev, void *data,
 	 * get zeroed, and that might leak data between users.
 	 */
 	bo = vc4_bo_create(dev, args->size, false);
-	if (!bo)
-		return -ENOMEM;
+	if (IS_ERR(bo))
+		return PTR_ERR(bo);
 
 	ret = drm_gem_handle_create(file_priv, &bo->base.base, &args->handle);
 	drm_gem_object_unreference_unlocked(&bo->base.base);
@@ -513,8 +516,8 @@ vc4_create_shader_bo_ioctl(struct drm_device *dev, void *data,
 	}
 
 	bo = vc4_bo_create(dev, args->size, true);
-	if (!bo)
-		return -ENOMEM;
+	if (IS_ERR(bo))
+		return PTR_ERR(bo);
 
 	ret = copy_from_user(bo->base.vaddr,
 			     (void __user *)(uintptr_t)args->data,

drivers/gpu/drm/vc4/vc4_debugfs.c

@@ -17,6 +17,7 @@
 
 static const struct drm_info_list vc4_debugfs_list[] = {
 	{"bo_stats", vc4_bo_stats_debugfs, 0},
+	{"gem_exec", vc4_gem_exec_debugfs, 0},
 	{"hdmi_regs", vc4_hdmi_debugfs_regs, 0},
 	{"hvs_regs", vc4_hvs_debugfs_regs, 0},
 	{"crtc0_regs", vc4_crtc_debugfs_regs, 0, (void *)(uintptr_t)0},

drivers/gpu/drm/vc4/vc4_drv.h

@@ -403,6 +403,7 @@ void vc4_job_handle_completed(struct vc4_dev *vc4);
 int vc4_queue_seqno_cb(struct drm_device *dev,
 		       struct vc4_seqno_cb *cb, uint64_t seqno,
 		       void (*func)(struct vc4_seqno_cb *cb));
+int vc4_gem_exec_debugfs(struct seq_file *m, void *arg);
 
 /* vc4_hdmi.c */
 extern struct platform_driver vc4_hdmi_driver;

drivers/gpu/drm/vc4/vc4_gem.c

@@ -31,6 +31,20 @@
 #include "vc4_regs.h"
 #include "vc4_trace.h"
 
+#ifdef CONFIG_DEBUG_FS
+int vc4_gem_exec_debugfs(struct seq_file *m, void *unused)
+{
+	struct drm_info_node *node = (struct drm_info_node *)m->private;
+	struct drm_device *dev = node->minor->dev;
+	struct vc4_dev *vc4 = to_vc4_dev(dev);
+
+	seq_printf(m, "Emitted  seqno:   0x%016llx\n", vc4->emit_seqno);
+	seq_printf(m, "Finished seqno:   0x%016llx\n", vc4->finished_seqno);
+
+	return 0;
+}
+#endif /* CONFIG_DEBUG_FS */
+
 static void
 vc4_queue_hangcheck(struct drm_device *dev)
 {
@@ -338,12 +352,7 @@ vc4_wait_for_seqno(struct drm_device *dev, uint64_t seqno, uint64_t timeout_ns,
 	finish_wait(&vc4->job_wait_queue, &wait);
 	trace_vc4_wait_for_seqno_end(dev, seqno);
 
-	if (ret && ret != -ERESTARTSYS) {
-		DRM_ERROR("timeout waiting for render thread idle\n");
-		return ret;
-	}
-
-	return 0;
+	return ret;
 }
 
 static void
@@ -579,9 +588,9 @@ vc4_get_bcl(struct drm_device *dev, struct vc4_exec_info *exec)
 	}
 
 	bo = vc4_bo_create(dev, exec_size, true);
-	if (!bo) {
+	if (IS_ERR(bo)) {
 		DRM_ERROR("Couldn't allocate BO for binning\n");
-		ret = PTR_ERR(exec->exec_bo);
+		ret = PTR_ERR(bo);
 		goto fail;
 	}
 	exec->exec_bo = &bo->base;
@@ -747,6 +756,9 @@ vc4_wait_bo_ioctl(struct drm_device *dev, void *data,
 	struct drm_gem_object *gem_obj;
 	struct vc4_bo *bo;
 
+	if (args->pad != 0)
+		return -EINVAL;
+
 	gem_obj = drm_gem_object_lookup(dev, file_priv, args->handle);
 	if (!gem_obj) {
 		DRM_ERROR("Failed to look up GEM BO %d\n", args->handle);

drivers/gpu/drm/vc4/vc4_irq.c

@@ -57,7 +57,7 @@ vc4_overflow_mem_work(struct work_struct *work)
 	struct vc4_bo *bo;
 
 	bo = vc4_bo_create(dev, 256 * 1024, true);
-	if (!bo) {
+	if (IS_ERR(bo)) {
 		DRM_ERROR("Couldn't allocate binner overflow mem\n");
 		return;
 	}

drivers/gpu/drm/vc4/vc4_render_cl.c

@@ -316,20 +316,11 @@ static int vc4_create_rcl_bo(struct drm_device *dev, struct vc4_exec_info *exec,
 	size += xtiles * ytiles * loop_body_size;
 
 	setup->rcl = &vc4_bo_create(dev, size, true)->base;
-	if (!setup->rcl)
-		return -ENOMEM;
+	if (IS_ERR(setup->rcl))
+		return PTR_ERR(setup->rcl);
 	list_add_tail(&to_vc4_bo(&setup->rcl->base)->unref_head,
 		      &exec->unref_list);
 
-	rcl_u8(setup, VC4_PACKET_TILE_RENDERING_MODE_CONFIG);
-	rcl_u32(setup,
-		(setup->color_write ? (setup->color_write->paddr +
-				       args->color_write.offset) :
-		 0));
-	rcl_u16(setup, args->width);
-	rcl_u16(setup, args->height);
-	rcl_u16(setup, args->color_write.bits);
-
 	/* The tile buffer gets cleared when the previous tile is stored.  If
 	 * the clear values changed between frames, then the tile buffer has
 	 * stale clear values in it, so we have to do a store in None mode (no
@@ -349,6 +340,15 @@ static int vc4_create_rcl_bo(struct drm_device *dev, struct vc4_exec_info *exec,
 		rcl_u32(setup, 0); /* no address, since we're in None mode */
 	}
 
+	rcl_u8(setup, VC4_PACKET_TILE_RENDERING_MODE_CONFIG);
+	rcl_u32(setup,
+		(setup->color_write ? (setup->color_write->paddr +
+				       args->color_write.offset) :
+		 0));
+	rcl_u16(setup, args->width);
+	rcl_u16(setup, args->height);
+	rcl_u16(setup, args->color_write.bits);
+
 	for (y = min_y_tile; y <= max_y_tile; y++) {
 		for (x = min_x_tile; x <= max_x_tile; x++) {
 			bool first = (x == min_x_tile && y == min_y_tile);

drivers/gpu/drm/vc4/vc4_validate.c

@@ -401,8 +401,8 @@ validate_tile_binning_config(VALIDATE_ARGS)
 	tile_bo = vc4_bo_create(dev, exec->tile_alloc_offset + tile_alloc_size,
 				true);
 	exec->tile_bo = &tile_bo->base;
-	if (!exec->tile_bo)
-		return -ENOMEM;
+	if (IS_ERR(exec->tile_bo))
+		return PTR_ERR(exec->tile_bo);
 	list_add_tail(&tile_bo->unref_head, &exec->unref_list);
 
 	/* tile alloc address. */

include/uapi/drm/Kbuild

@@ -14,6 +14,7 @@ header-y += radeon_drm.h
 header-y += savage_drm.h
 header-y += sis_drm.h
 header-y += tegra_drm.h
+header-y += vc4_drm.h
 header-y += via_drm.h
 header-y += vmwgfx_drm.h
 header-y += msm_drm.h