drm/virtio: add dma sync for dma mapped virtio gpu framebuffer pages

With virtio gpu ttm-pages being dma mapped, dma sync is needed when
swiotlb is used as bounce buffers, before TRANSFER_TO_HOST_2D/3D
commands are sent.

Signed-off-by: Jiandi An <jiandi.an@amd.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20180919070931.91168-1-jiandi.an@amd.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>

drivers/gpu/drm/virtio/virtgpu_drv.h

@@ -36,6 +36,7 @@
 #include <drm/drm_atomic.h>
 #include <drm/drm_crtc_helper.h>
 #include <drm/drm_encoder.h>
+#include <drm/drm_fb_helper.h>
 #include <drm/ttm/ttm_bo_api.h>
 #include <drm/ttm/ttm_bo_driver.h>
 #include <drm/ttm/ttm_placement.h>
@@ -133,6 +134,13 @@ struct virtio_gpu_framebuffer {
 #define to_virtio_gpu_framebuffer(x) \
 	container_of(x, struct virtio_gpu_framebuffer, base)
 
+struct virtio_gpu_fbdev {
+	struct drm_fb_helper           helper;
+	struct virtio_gpu_framebuffer  vgfb;
+	struct virtio_gpu_device       *vgdev;
+	struct delayed_work            work;
+};
+
 struct virtio_gpu_mman {
 	struct ttm_bo_global_ref        bo_global_ref;
 	struct drm_global_reference	mem_global_ref;

drivers/gpu/drm/virtio/virtgpu_fb.c

@@ -29,13 +29,6 @@
 
 #define VIRTIO_GPU_FBCON_POLL_PERIOD (HZ / 60)
 
-struct virtio_gpu_fbdev {
-	struct drm_fb_helper           helper;
-	struct virtio_gpu_framebuffer  vgfb;
-	struct virtio_gpu_device       *vgdev;
-	struct delayed_work            work;
-};
-
 static int virtio_gpu_dirty_update(struct virtio_gpu_framebuffer *fb,
 				   bool store, int x, int y,
 				   int width, int height)

drivers/gpu/drm/virtio/virtgpu_vq.c

@@ -490,6 +490,15 @@ void virtio_gpu_cmd_transfer_to_host_2d(struct virtio_gpu_device *vgdev,
 {
 	struct virtio_gpu_transfer_to_host_2d *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
+	struct virtio_gpu_fbdev *vgfbdev = vgdev->vgfbdev;
+	struct virtio_gpu_framebuffer *fb = &vgfbdev->vgfb;
+	struct virtio_gpu_object *obj = gem_to_virtio_gpu_obj(fb->base.obj[0]);
+	bool use_dma_api = !virtio_has_iommu_quirk(vgdev->vdev);
+
+	if (use_dma_api)
+		dma_sync_sg_for_device(vgdev->vdev->dev.parent,
+				       obj->pages->sgl, obj->pages->nents,
+				       DMA_TO_DEVICE);
 
 	cmd_p = virtio_gpu_alloc_cmd(vgdev, &vbuf, sizeof(*cmd_p));
 	memset(cmd_p, 0, sizeof(*cmd_p));
@@ -789,6 +798,15 @@ void virtio_gpu_cmd_transfer_to_host_3d(struct virtio_gpu_device *vgdev,
 {
 	struct virtio_gpu_transfer_host_3d *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
+	struct virtio_gpu_fbdev *vgfbdev = vgdev->vgfbdev;
+	struct virtio_gpu_framebuffer *fb = &vgfbdev->vgfb;
+	struct virtio_gpu_object *obj = gem_to_virtio_gpu_obj(fb->base.obj[0]);
+	bool use_dma_api = !virtio_has_iommu_quirk(vgdev->vdev);
+
+	if (use_dma_api)
+		dma_sync_sg_for_device(vgdev->vdev->dev.parent,
+				       obj->pages->sgl, obj->pages->nents,
+				       DMA_TO_DEVICE);
 
 	cmd_p = virtio_gpu_alloc_cmd(vgdev, &vbuf, sizeof(*cmd_p));
 	memset(cmd_p, 0, sizeof(*cmd_p));