Virtualization - Dec 2016 - [RFC 1/5] drm/virtio: add virtio_gpu_alloc_fence()

From: Gustavo Padovan <gustavo.padovan at collabora.co.uk>

Refactor fence creation to remove the potential allocation failure from
the cmd_submit and atomic_commit paths. Now the fence should be allocated
first and just after we should proceed with the rest of the execution.

Signed-off-by: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
---
 drivers/gpu/drm/virtio/virtgpu_drv.h   | 17 +++++++------
 drivers/gpu/drm/virtio/virtgpu_fence.c | 36 +++++++++++++++++---------
 drivers/gpu/drm/virtio/virtgpu_ioctl.c | 32 ++++++++++++++++++++---
 drivers/gpu/drm/virtio/virtgpu_plane.c | 46 ++++++++++++++++++++++++++++++----
 drivers/gpu/drm/virtio/virtgpu_vq.c    | 16 ++++++------
 5 files changed, 111 insertions(+), 36 deletions(-)

diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.h
b/drivers/gpu/drm/virtio/virtgpu_drv.h
index 08906c8..806c98b 100644
--- a/drivers/gpu/drm/virtio/virtgpu_drv.h
+++ b/drivers/gpu/drm/virtio/virtgpu_drv.h
@@ -127,6 +127,7 @@ struct virtio_gpu_framebuffer {
 	int x1, y1, x2, y2; /* dirty rect */
 	spinlock_t dirty_lock;
 	uint32_t hw_res_handle;
+	struct virtio_gpu_fence *fence;
 };
 #define to_virtio_gpu_framebuffer(x) \
 	container_of(x, struct virtio_gpu_framebuffer, base)
@@ -268,7 +269,7 @@ void virtio_gpu_cmd_transfer_to_host_2d(struct
virtio_gpu_device *vgdev,
 					uint32_t resource_id, uint64_t offset,
 					__le32 width, __le32 height,
 					__le32 x, __le32 y,
-					struct virtio_gpu_fence **fence);
+					struct virtio_gpu_fence *fence);
 void virtio_gpu_cmd_resource_flush(struct virtio_gpu_device *vgdev,
 				   uint32_t resource_id,
 				   uint32_t x, uint32_t y,
@@ -280,7 +281,7 @@ void virtio_gpu_cmd_set_scanout(struct virtio_gpu_device
*vgdev,
 int virtio_gpu_object_attach(struct virtio_gpu_device *vgdev,
 			     struct virtio_gpu_object *obj,
 			     uint32_t resource_id,
-			     struct virtio_gpu_fence **fence);
+			     struct virtio_gpu_fence *fence);
 int virtio_gpu_attach_status_page(struct virtio_gpu_device *vgdev);
 int virtio_gpu_detach_status_page(struct virtio_gpu_device *vgdev);
 void virtio_gpu_cursor_ping(struct virtio_gpu_device *vgdev,
@@ -304,21 +305,21 @@ void virtio_gpu_cmd_context_detach_resource(struct
virtio_gpu_device *vgdev,
 					    uint32_t resource_id);
 void virtio_gpu_cmd_submit(struct virtio_gpu_device *vgdev,
 			   void *data, uint32_t data_size,
-			   uint32_t ctx_id, struct virtio_gpu_fence **fence);
+			   uint32_t ctx_id, struct virtio_gpu_fence *fence);
 void virtio_gpu_cmd_transfer_from_host_3d(struct virtio_gpu_device *vgdev,
 					  uint32_t resource_id, uint32_t ctx_id,
 					  uint64_t offset, uint32_t level,
 					  struct virtio_gpu_box *box,
-					  struct virtio_gpu_fence **fence);
+					  struct virtio_gpu_fence *fence);
 void virtio_gpu_cmd_transfer_to_host_3d(struct virtio_gpu_device *vgdev,
 					uint32_t resource_id, uint32_t ctx_id,
 					uint64_t offset, uint32_t level,
 					struct virtio_gpu_box *box,
-					struct virtio_gpu_fence **fence);
+					struct virtio_gpu_fence *fence);
 void
 virtio_gpu_cmd_resource_create_3d(struct virtio_gpu_device *vgdev,
 				  struct virtio_gpu_resource_create_3d *rc_3d,
-				  struct virtio_gpu_fence **fence);
+				  struct virtio_gpu_fence *fence);
 void virtio_gpu_ctrl_ack(struct virtqueue *vq);
 void virtio_gpu_cursor_ack(struct virtqueue *vq);
 void virtio_gpu_fence_ack(struct virtqueue *vq);
@@ -345,9 +346,11 @@ void virtio_gpu_ttm_fini(struct virtio_gpu_device *vgdev);
 int virtio_gpu_mmap(struct file *filp, struct vm_area_struct *vma);
 
 /* virtio_gpu_fence.c */
+struct virtio_gpu_fence *virtio_gpu_fence_alloc(
+					struct virtio_gpu_device *vgdev);
 int virtio_gpu_fence_emit(struct virtio_gpu_device *vgdev,
 			  struct virtio_gpu_ctrl_hdr *cmd_hdr,
-			  struct virtio_gpu_fence **fence);
+			  struct virtio_gpu_fence *fence);
 void virtio_gpu_fence_event_process(struct virtio_gpu_device *vdev,
 				    u64 last_seq);
 
diff --git a/drivers/gpu/drm/virtio/virtgpu_fence.c
b/drivers/gpu/drm/virtio/virtgpu_fence.c
index 2335352..4dbfe44 100644
--- a/drivers/gpu/drm/virtio/virtgpu_fence.c
+++ b/drivers/gpu/drm/virtio/virtgpu_fence.c
@@ -74,28 +74,40 @@ static const struct dma_fence_ops virtio_fence_ops = {
 	.timeline_value_str  = virtio_timeline_value_str,
 };
 
+struct virtio_gpu_fence *virtio_gpu_fence_alloc(struct virtio_gpu_device
*vgdev)
+{
+	struct virtio_gpu_fence_driver *drv = &vgdev->fence_drv;
+	struct virtio_gpu_fence *fence;
+	unsigned long irq_flags;
+
+	fence = kmalloc(sizeof(struct virtio_gpu_fence), GFP_ATOMIC);
+	if (!fence)
+		return NULL;
+
+	spin_lock_irqsave(&drv->lock, irq_flags);
+	fence->drv = drv;
+	fence->seq = ++drv->sync_seq;
+	dma_fence_init(&fence->f, &virtio_fence_ops, &drv->lock,
+		       drv->context, fence->seq);
+	spin_unlock_irqrestore(&drv->lock, irq_flags);
+
+	return fence;
+}
+
 int virtio_gpu_fence_emit(struct virtio_gpu_device *vgdev,
 			  struct virtio_gpu_ctrl_hdr *cmd_hdr,
-			  struct virtio_gpu_fence **fence)
+			  struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_fence_driver *drv = &vgdev->fence_drv;
 	unsigned long irq_flags;
 
-	*fence = kmalloc(sizeof(struct virtio_gpu_fence), GFP_ATOMIC);
-	if ((*fence) == NULL)
-		return -ENOMEM;
-
 	spin_lock_irqsave(&drv->lock, irq_flags);
-	(*fence)->drv = drv;
-	(*fence)->seq = ++drv->sync_seq;
-	dma_fence_init(&(*fence)->f, &virtio_fence_ops, &drv->lock,
-		       drv->context, (*fence)->seq);
-	dma_fence_get(&(*fence)->f);
-	list_add_tail(&(*fence)->node, &drv->fences);
+	dma_fence_get(&fence->f);
+	list_add_tail(&fence->node, &drv->fences);
 	spin_unlock_irqrestore(&drv->lock, irq_flags);
 
 	cmd_hdr->flags |= cpu_to_le32(VIRTIO_GPU_FLAG_FENCE);
-	cmd_hdr->fence_id = cpu_to_le64((*fence)->seq);
+	cmd_hdr->fence_id = cpu_to_le64(fence->seq);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
index 61f3a96..da281103 100644
--- a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
+++ b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
@@ -164,8 +164,15 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 		ret = PTR_ERR(buf);
 		goto out_unresv;
 	}
+
+	fence = virtio_gpu_fence_alloc(vgdev);
+	if (!fence) {
+		kfree(buf);
+		ret = -ENOMEM;
+		goto out_unresv;
+	}
 	virtio_gpu_cmd_submit(vgdev, buf, exbuf->size,
-			      vfpriv->ctx_id, &fence);
+			      vfpriv->ctx_id, fence);
 
 	ttm_eu_fence_buffer_objects(&ticket, &validate_list,
&fence->f);
 
@@ -281,8 +288,14 @@ static int virtio_gpu_resource_create_ioctl(struct
drm_device *dev, void *data,
 		rc_3d.nr_samples = cpu_to_le32(rc->nr_samples);
 		rc_3d.flags = cpu_to_le32(rc->flags);
 
+		fence = virtio_gpu_fence_alloc(vgdev);
+		if (!fence) {
+			ret = -ENOMEM;
+			goto fail_unref;
+		}
+
 		virtio_gpu_cmd_resource_create_3d(vgdev, &rc_3d, NULL);
-		ret = virtio_gpu_object_attach(vgdev, qobj, res_id, &fence);
+		ret = virtio_gpu_object_attach(vgdev, qobj, res_id, fence);
 		if (ret) {
 			ttm_eu_backoff_reservation(&ticket, &validate_list);
 			goto fail_unref;
@@ -376,10 +389,16 @@ static int virtio_gpu_transfer_from_host_ioctl(struct
drm_device *dev,
 		goto out_unres;
 
 	convert_to_hw_box(&box, &args->box);
+
+	fence = virtio_gpu_fence_alloc(vgdev);
+	if (!fence) {
+		ret = -ENOMEM;
+		goto out_unres;
+	}
 	virtio_gpu_cmd_transfer_from_host_3d
 		(vgdev, qobj->hw_res_handle,
 		 vfpriv->ctx_id, offset, args->level,
-		 &box, &fence);
+		 &box, fence);
 	reservation_object_add_excl_fence(qobj->tbo.resv,
 					  &fence->f);
 
@@ -425,10 +444,15 @@ static int virtio_gpu_transfer_to_host_ioctl(struct
drm_device *dev, void *data,
 			(vgdev, qobj->hw_res_handle, offset,
 			 box.w, box.h, box.x, box.y, NULL);
 	} else {
+		fence = virtio_gpu_fence_alloc(vgdev);
+		if (!fence) {
+			ret = -ENOMEM;
+			goto out_unres;
+		}
 		virtio_gpu_cmd_transfer_to_host_3d
 			(vgdev, qobj->hw_res_handle,
 			 vfpriv ? vfpriv->ctx_id : 0, offset,
-			 args->level, &box, &fence);
+			 args->level, &box, fence);
 		reservation_object_add_excl_fence(qobj->tbo.resv,
 						  &fence->f);
 		dma_fence_put(&fence->f);
diff --git a/drivers/gpu/drm/virtio/virtgpu_plane.c
b/drivers/gpu/drm/virtio/virtgpu_plane.c
index 05022ef..77415e5 100644
--- a/drivers/gpu/drm/virtio/virtgpu_plane.c
+++ b/drivers/gpu/drm/virtio/virtgpu_plane.c
@@ -115,6 +115,41 @@ static void virtio_gpu_primary_plane_update(struct
drm_plane *plane,
 				      plane->state->src_h >> 16);
 }
 
+static int virtio_gpu_cursor_prepare_fb(struct drm_plane *plane,
+					struct drm_plane_state *new_state)
+{
+	struct drm_device *dev = plane->dev;
+	struct virtio_gpu_device *vgdev = dev->dev_private;
+	struct virtio_gpu_framebuffer *vgfb;
+	struct virtio_gpu_object *bo;
+
+	if (!new_state->fb)
+		return 0;
+
+	vgfb = to_virtio_gpu_framebuffer(new_state->fb);
+	bo = gem_to_virtio_gpu_obj(vgfb->obj);
+	if (bo && bo->dumb && (plane->state->fb !=
new_state->fb)) {
+		vgfb->fence = virtio_gpu_fence_alloc(vgdev);
+		if (!vgfb->fence)
+			return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void virtio_gpu_cursor_cleanup_fb(struct drm_plane *plane,
+					 struct drm_plane_state *old_state)
+{
+	struct virtio_gpu_framebuffer *vgfb;
+
+	if (!plane->state->fb)
+		return;
+
+	vgfb = to_virtio_gpu_framebuffer(plane->state->fb);
+	if (vgfb->fence)
+		dma_fence_put(&vgfb->fence->f);
+}
+
 static void virtio_gpu_cursor_plane_update(struct drm_plane *plane,
 					   struct drm_plane_state *old_state)
 {
@@ -122,7 +157,6 @@ static void virtio_gpu_cursor_plane_update(struct drm_plane
*plane,
 	struct virtio_gpu_device *vgdev = dev->dev_private;
 	struct virtio_gpu_output *output = NULL;
 	struct virtio_gpu_framebuffer *vgfb;
-	struct virtio_gpu_fence *fence = NULL;
 	struct virtio_gpu_object *bo = NULL;
 	uint32_t handle;
 	int ret = 0;
@@ -148,13 +182,13 @@ static void virtio_gpu_cursor_plane_update(struct
drm_plane *plane,
 			(vgdev, handle, 0,
 			 cpu_to_le32(plane->state->crtc_w),
 			 cpu_to_le32(plane->state->crtc_h),
-			 0, 0, &fence);
+			 0, 0, vgfb->fence);
 		ret = virtio_gpu_object_reserve(bo, false);
 		if (!ret) {
 			reservation_object_add_excl_fence(bo->tbo.resv,
-							  &fence->f);
-			dma_fence_put(&fence->f);
-			fence = NULL;
+							  &vgfb->fence->f);
+			dma_fence_put(&vgfb->fence->f);
+			vgfb->fence = NULL;
 			virtio_gpu_object_unreserve(bo);
 			virtio_gpu_object_wait(bo, false);
 		}
@@ -196,6 +230,8 @@ static const struct drm_plane_helper_funcs
virtio_gpu_primary_helper_funcs = {
 };
 
 static const struct drm_plane_helper_funcs virtio_gpu_cursor_helper_funcs = {
+	.prepare_fb		= virtio_gpu_cursor_prepare_fb,
+	.cleanup_fb		= virtio_gpu_cursor_cleanup_fb,
 	.atomic_check		= virtio_gpu_plane_atomic_check,
 	.atomic_update		= virtio_gpu_cursor_plane_update,
 };
diff --git a/drivers/gpu/drm/virtio/virtgpu_vq.c
b/drivers/gpu/drm/virtio/virtgpu_vq.c
index 974f941..e7c3e8d 100644
--- a/drivers/gpu/drm/virtio/virtgpu_vq.c
+++ b/drivers/gpu/drm/virtio/virtgpu_vq.c
@@ -350,7 +350,7 @@ static int virtio_gpu_queue_ctrl_buffer(struct
virtio_gpu_device *vgdev,
 static int virtio_gpu_queue_fenced_ctrl_buffer(struct virtio_gpu_device *vgdev,
 					       struct virtio_gpu_vbuffer *vbuf,
 					       struct virtio_gpu_ctrl_hdr *hdr,
-					       struct virtio_gpu_fence **fence)
+					       struct virtio_gpu_fence *fence)
 {
 	struct virtqueue *vq = vgdev->ctrlq.vq;
 	int rc;
@@ -515,7 +515,7 @@ void virtio_gpu_cmd_transfer_to_host_2d(struct
virtio_gpu_device *vgdev,
 					uint32_t resource_id, uint64_t offset,
 					__le32 width, __le32 height,
 					__le32 x, __le32 y,
-					struct virtio_gpu_fence **fence)
+					struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_transfer_to_host_2d *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
@@ -539,7 +539,7 @@ virtio_gpu_cmd_resource_attach_backing(struct
virtio_gpu_device *vgdev,
 				       uint32_t resource_id,
 				       struct virtio_gpu_mem_entry *ents,
 				       uint32_t nents,
-				       struct virtio_gpu_fence **fence)
+				       struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_resource_attach_backing *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
@@ -795,7 +795,7 @@ void virtio_gpu_cmd_context_detach_resource(struct
virtio_gpu_device *vgdev,
 void
 virtio_gpu_cmd_resource_create_3d(struct virtio_gpu_device *vgdev,
 				  struct virtio_gpu_resource_create_3d *rc_3d,
-				  struct virtio_gpu_fence **fence)
+				  struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_resource_create_3d *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
@@ -814,7 +814,7 @@ void virtio_gpu_cmd_transfer_to_host_3d(struct
virtio_gpu_device *vgdev,
 					uint32_t resource_id, uint32_t ctx_id,
 					uint64_t offset, uint32_t level,
 					struct virtio_gpu_box *box,
-					struct virtio_gpu_fence **fence)
+					struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_transfer_host_3d *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
@@ -836,7 +836,7 @@ void virtio_gpu_cmd_transfer_from_host_3d(struct
virtio_gpu_device *vgdev,
 					  uint32_t resource_id, uint32_t ctx_id,
 					  uint64_t offset, uint32_t level,
 					  struct virtio_gpu_box *box,
-					  struct virtio_gpu_fence **fence)
+					  struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_transfer_host_3d *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
@@ -856,7 +856,7 @@ void virtio_gpu_cmd_transfer_from_host_3d(struct
virtio_gpu_device *vgdev,
 
 void virtio_gpu_cmd_submit(struct virtio_gpu_device *vgdev,
 			   void *data, uint32_t data_size,
-			   uint32_t ctx_id, struct virtio_gpu_fence **fence)
+			   uint32_t ctx_id, struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_cmd_submit *cmd_p;
 	struct virtio_gpu_vbuffer *vbuf;
@@ -877,7 +877,7 @@ void virtio_gpu_cmd_submit(struct virtio_gpu_device *vgdev,
 int virtio_gpu_object_attach(struct virtio_gpu_device *vgdev,
 			     struct virtio_gpu_object *obj,
 			     uint32_t resource_id,
-			     struct virtio_gpu_fence **fence)
+			     struct virtio_gpu_fence *fence)
 {
 	struct virtio_gpu_mem_entry *ents;
 	struct scatterlist *sg;
-- 
2.5.5

From: Gustavo Padovan <gustavo.padovan at collabora.co.uk>

Add a new field called fence_fd that will be used by userspace to send
in-fences to the kernel and receive out-fences created by the kernel.

This uapi enables virtio to take advantage of explicit synchronization of
dma-bufs.

There are two new flags:

* VIRTGPU_EXECBUF_FENCE_FD_IN to be used when passing an in-fence fd.
* VIRTGPU_EXECBUF_FENCE_FD_OUT to be used when requesting an out-fence fd

The execbuffer IOCTL is now read-write to allow the userspace to read the
out-fence.

On error -1 should be returned in the fence_fd field.

Signed-off-by: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
---
 drivers/gpu/drm/virtio/virtgpu_ioctl.c |  2 ++
 include/uapi/drm/virtgpu_drm.h         | 13 ++++++++++---
 2 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
index da281103..d164b54 100644
--- a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
+++ b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
@@ -113,6 +113,8 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 	struct ww_acquire_ctx ticket;
 	void *buf;
 
+	exbuf->fence_fd = -1;
+
 	if (vgdev->has_virgl_3d == false)
 		return -ENOSYS;
 
diff --git a/include/uapi/drm/virtgpu_drm.h b/include/uapi/drm/virtgpu_drm.h
index 91a31ff..d1d69be 100644
--- a/include/uapi/drm/virtgpu_drm.h
+++ b/include/uapi/drm/virtgpu_drm.h
@@ -47,6 +47,13 @@ extern "C" {
 #define DRM_VIRTGPU_WAIT     0x08
 #define DRM_VIRTGPU_GET_CAPS  0x09
 
+#define VIRTGPU_EXECBUF_FENCE_FD_IN	0x01
+#define VIRTGPU_EXECBUF_FENCE_FD_OUT	0x02
+#define VIRTGPU_EXECBUF_FLAGS  (\
+		VIRTGPU_EXECBUF_FENCE_FD_IN |\
+		VIRTGPU_EXECBUF_FENCE_FD_OUT |\
+		0)
+
 struct drm_virtgpu_map {
 	__u64 offset; /* use for mmap system call */
 	__u32 handle;
@@ -54,12 +61,12 @@ struct drm_virtgpu_map {
 };
 
 struct drm_virtgpu_execbuffer {
-	__u32		flags;		/* for future use */
+	__u32 flags;
 	__u32 size;
 	__u64 command; /* void* */
 	__u64 bo_handles;
 	__u32 num_bo_handles;
-	__u32 pad;
+	__s32 fence_fd;
 };
 
 #define VIRTGPU_PARAM_3D_FEATURES 1 /* do we have 3D features in the hw */
@@ -136,7 +143,7 @@ struct drm_virtgpu_get_caps {
 	DRM_IOWR(DRM_COMMAND_BASE + DRM_VIRTGPU_MAP, struct drm_virtgpu_map)
 
 #define DRM_IOCTL_VIRTGPU_EXECBUFFER \
-	DRM_IOW(DRM_COMMAND_BASE + DRM_VIRTGPU_EXECBUFFER,\
+	DRM_IOWR(DRM_COMMAND_BASE + DRM_VIRTGPU_EXECBUFFER,\
 		struct drm_virtgpu_execbuffer)
 
 #define DRM_IOCTL_VIRTGPU_GETPARAM \
-- 
2.5.5

From: Gustavo Padovan <gustavo.padovan at collabora.co.uk>

When the execbuf call receives an in-fence it will get the dma_fence
related to that fence fd and wait on it before submitting the draw call.

Signed-off-by: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
---
 drivers/gpu/drm/virtio/virtgpu_ioctl.c | 41 ++++++++++++++++++++++++++--------
 1 file changed, 32 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
index d164b54..ac0b4b0 100644
--- a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
+++ b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
@@ -29,6 +29,7 @@
 #include "virtgpu_drv.h"
 #include <drm/virtgpu_drm.h>
 #include "ttm/ttm_execbuf_util.h"
+#include <linux/sync_file.h>
 
 static void convert_to_hw_box(struct virtio_gpu_box *dst,
 			      const struct drm_virtgpu_3d_box *src)
@@ -111,6 +112,8 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 	struct ttm_validate_buffer *buflist = NULL;
 	int i;
 	struct ww_acquire_ctx ticket;
+	struct dma_fence *in_fence = NULL;
+	int in_fence_fd = exbuf->fence_fd;
 	void *buf;
 
 	exbuf->fence_fd = -1;
@@ -118,6 +121,19 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 	if (vgdev->has_virgl_3d == false)
 		return -ENOSYS;
 
+	/* TODO: if the fence is a fence array we need to check
+	 * the context of every single fence */
+	if (exbuf->flags & VIRTGPU_EXECBUF_FENCE_FD_IN) {
+		in_fence = sync_file_get_fence(in_fence_fd);
+		if (!in_fence)
+			return -EINVAL;
+
+		if (in_fence->context == vgdev->fence_drv.context) {
+			dma_fence_put(in_fence);
+			return -EINVAL;
+		}
+	}
+
 	INIT_LIST_HEAD(&validate_list);
 	if (exbuf->num_bo_handles) {
 
@@ -126,26 +142,22 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 		buflist = drm_calloc_large(exbuf->num_bo_handles,
 					   sizeof(struct ttm_validate_buffer));
 		if (!bo_handles || !buflist) {
-			drm_free_large(bo_handles);
-			drm_free_large(buflist);
-			return -ENOMEM;
+			ret = -ENOMEM;
+			goto out_in_fence;
 		}
 
 		user_bo_handles = (void __user *)(uintptr_t)exbuf->bo_handles;
 		if (copy_from_user(bo_handles, user_bo_handles,
 				   exbuf->num_bo_handles * sizeof(uint32_t))) {
 			ret = -EFAULT;
-			drm_free_large(bo_handles);
-			drm_free_large(buflist);
-			return ret;
+			goto out_in_fence;
 		}
 
 		for (i = 0; i < exbuf->num_bo_handles; i++) {
 			gobj = drm_gem_object_lookup(drm_file, bo_handles[i]);
 			if (!gobj) {
-				drm_free_large(bo_handles);
-				drm_free_large(buflist);
-				return -ENOENT;
+				ret = -ENOENT;
+				goto out_in_fence;
 			}
 
 			qobj = gem_to_virtio_gpu_obj(gobj);
@@ -154,6 +166,7 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 			list_add(&buflist[i].head, &validate_list);
 		}
 		drm_free_large(bo_handles);
+		bo_handles = NULL;
 	}
 
 	ret = virtio_gpu_object_list_validate(&ticket, &validate_list);
@@ -173,6 +186,13 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 		ret = -ENOMEM;
 		goto out_unresv;
 	}
+
+	if (in_fence) {
+		dma_fence_wait(in_fence, true);
+		dma_fence_put(in_fence);
+		in_fence = NULL;
+	}
+
 	virtio_gpu_cmd_submit(vgdev, buf, exbuf->size,
 			      vfpriv->ctx_id, fence);
 
@@ -188,7 +208,10 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 	ttm_eu_backoff_reservation(&ticket, &validate_list);
 out_free:
 	virtio_gpu_unref_list(&validate_list);
+out_in_fence:
 	drm_free_large(buflist);
+	drm_free_large(bo_handles);
+	dma_fence_put(in_fence);
 	return ret;
 }
 
-- 
2.5.5

From: Gustavo Padovan <gustavo.padovan at collabora.co.uk>

On the out-fence side we get fence returned by the submitted draw call
and attach it to a sync_file and send the sync_file fd to userspace. On
error -1 is returned to userspace.

Signed-off-by: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
---
 drivers/gpu/drm/virtio/virtgpu_ioctl.c | 51 ++++++++++++++++++++++++++--------
 1 file changed, 39 insertions(+), 12 deletions(-)

diff --git a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
index ac0b4b0..f441928 100644
--- a/drivers/gpu/drm/virtio/virtgpu_ioctl.c
+++ b/drivers/gpu/drm/virtio/virtgpu_ioctl.c
@@ -103,7 +103,7 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 	struct virtio_gpu_device *vgdev = dev->dev_private;
 	struct virtio_gpu_fpriv *vfpriv = drm_file->driver_priv;
 	struct drm_gem_object *gobj;
-	struct virtio_gpu_fence *fence;
+	struct virtio_gpu_fence *out_fence;
 	struct virtio_gpu_object *qobj;
 	int ret;
 	uint32_t *bo_handles = NULL;
@@ -113,7 +113,9 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 	int i;
 	struct ww_acquire_ctx ticket;
 	struct dma_fence *in_fence = NULL;
+	struct sync_file *sync_file;
 	int in_fence_fd = exbuf->fence_fd;
+	int out_fence_fd = -1;
 	void *buf;
 
 	exbuf->fence_fd = -1;
@@ -134,6 +136,14 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 		}
 	}
 
+	if (exbuf->flags & VIRTGPU_EXECBUF_FENCE_FD_OUT) {
+		out_fence_fd = get_unused_fd_flags(O_CLOEXEC);
+		if (out_fence_fd < 0) {
+			ret = out_fence_fd;
+			goto out_in_fence;
+		}
+	}
+
 	INIT_LIST_HEAD(&validate_list);
 	if (exbuf->num_bo_handles) {
 
@@ -143,21 +153,21 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 					   sizeof(struct ttm_validate_buffer));
 		if (!bo_handles || !buflist) {
 			ret = -ENOMEM;
-			goto out_in_fence;
+			goto out_unused_fd;
 		}
 
 		user_bo_handles = (void __user *)(uintptr_t)exbuf->bo_handles;
 		if (copy_from_user(bo_handles, user_bo_handles,
 				   exbuf->num_bo_handles * sizeof(uint32_t))) {
 			ret = -EFAULT;
-			goto out_in_fence;
+			goto out_unused_fd;
 		}
 
 		for (i = 0; i < exbuf->num_bo_handles; i++) {
 			gobj = drm_gem_object_lookup(drm_file, bo_handles[i]);
 			if (!gobj) {
 				ret = -ENOENT;
-				goto out_in_fence;
+				goto out_unused_fd;
 			}
 
 			qobj = gem_to_virtio_gpu_obj(gobj);
@@ -180,11 +190,22 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 		goto out_unresv;
 	}
 
-	fence = virtio_gpu_fence_alloc(vgdev);
-	if (!fence) {
-		kfree(buf);
+	out_fence = virtio_gpu_fence_alloc(vgdev);
+	if(!out_fence) {
 		ret = -ENOMEM;
-		goto out_unresv;
+		goto out_memdup;
+	}
+
+	if (out_fence_fd >= 0) {
+		sync_file = sync_file_create(dma_fence_get(&out_fence->f));
+		if (!sync_file) {
+			dma_fence_put(&out_fence->f);
+			ret = -ENOMEM;
+			goto out_memdup;
+		}
+
+		exbuf->fence_fd = out_fence_fd;
+		fd_install(out_fence_fd, sync_file->file);
 	}
 
 	if (in_fence) {
@@ -194,23 +215,29 @@ static int virtio_gpu_execbuffer_ioctl(struct drm_device
*dev, void *data,
 	}
 
 	virtio_gpu_cmd_submit(vgdev, buf, exbuf->size,
-			      vfpriv->ctx_id, fence);
+			      vfpriv->ctx_id, out_fence);
 
-	ttm_eu_fence_buffer_objects(&ticket, &validate_list,
&fence->f);
+	ttm_eu_fence_buffer_objects(&ticket, &validate_list,
&out_fence->f);
 
 	/* fence the command bo */
 	virtio_gpu_unref_list(&validate_list);
 	drm_free_large(buflist);
-	dma_fence_put(&fence->f);
+	dma_fence_put(&out_fence->f);
 	return 0;
 
+out_memdup:
+	kfree(buf);
 out_unresv:
 	ttm_eu_backoff_reservation(&ticket, &validate_list);
 out_free:
 	virtio_gpu_unref_list(&validate_list);
-out_in_fence:
+out_unused_fd:
 	drm_free_large(buflist);
 	drm_free_large(bo_handles);
+
+	if (out_fence_fd >= 0)
+		put_unused_fd(out_fence_fd);
+out_in_fence:
 	dma_fence_put(in_fence);
 	return ret;
 }
-- 
2.5.5

From: Gustavo Padovan <gustavo.padovan at collabora.co.uk>

To reflect the (backward compatible) changes in the uabi we are bumping
the driver's version.

Signed-off-by: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
---
 drivers/gpu/drm/virtio/virtgpu_drv.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.h
b/drivers/gpu/drm/virtio/virtgpu_drv.h
index 806c98b..b9ab010 100644
--- a/drivers/gpu/drm/virtio/virtgpu_drv.h
+++ b/drivers/gpu/drm/virtio/virtgpu_drv.h
@@ -45,8 +45,8 @@
 #define DRIVER_DATE "0"
 
 #define DRIVER_MAJOR 0
-#define DRIVER_MINOR 0
-#define DRIVER_PATCHLEVEL 1
+#define DRIVER_MINOR 1
+#define DRIVER_PATCHLEVEL 0
 
 /* virtgpu_drm_bus.c */
 int drm_virtio_init(struct drm_driver *driver, struct virtio_device *vdev);
-- 
2.5.5

Hi,
> +struct virtio_gpu_fence *virtio_gpu_fence_alloc(struct virtio_gpu_device
*vgdev)
> +{
> +	struct virtio_gpu_fence_driver *drv = &vgdev->fence_drv;
> +	struct virtio_gpu_fence *fence;
> +	unsigned long irq_flags;
> +
> +	fence = kmalloc(sizeof(struct virtio_gpu_fence), GFP_ATOMIC);
> +	if (!fence)
> +		return NULL;
> +
> +	spin_lock_irqsave(&drv->lock, irq_flags);
> +	fence->drv = drv;
> +	fence->seq = ++drv->sync_seq;
> +	dma_fence_init(&fence->f, &virtio_fence_ops,
&drv->lock,
> +		       drv->context, fence->seq);
> +	spin_unlock_irqrestore(&drv->lock, irq_flags);
seq assignment ...
> +
> +	return fence;
> +}
> +
>  int virtio_gpu_fence_emit(struct virtio_gpu_device *vgdev,
>  			  struct virtio_gpu_ctrl_hdr *cmd_hdr,
> -			  struct virtio_gpu_fence **fence)
> +			  struct virtio_gpu_fence *fence)
>  {
>  	struct virtio_gpu_fence_driver *drv = &vgdev->fence_drv;
>  	unsigned long irq_flags;
>  
> -	*fence = kmalloc(sizeof(struct virtio_gpu_fence), GFP_ATOMIC);
> -	if ((*fence) == NULL)
> -		return -ENOMEM;
> -
>  	spin_lock_irqsave(&drv->lock, irq_flags);
> -	(*fence)->drv = drv;
> -	(*fence)->seq = ++drv->sync_seq;
> -	dma_fence_init(&(*fence)->f, &virtio_fence_ops,
&drv->lock,
> -		       drv->context, (*fence)->seq);
... must stay here.  Otherwise requests can be submitted to the virt
queue with fence sequence numbers out of order.

cheers,
  Gerd

2016-12-13 Gerd Hoffmann <kraxel at redhat.com>:
>   Hi,
> 
> > +struct virtio_gpu_fence *virtio_gpu_fence_alloc(struct
virtio_gpu_device *vgdev)
> > +{
> > +	struct virtio_gpu_fence_driver *drv = &vgdev->fence_drv;
> > +	struct virtio_gpu_fence *fence;
> > +	unsigned long irq_flags;
> > +
> > +	fence = kmalloc(sizeof(struct virtio_gpu_fence), GFP_ATOMIC);
> > +	if (!fence)
> > +		return NULL;
> > +
> 
> > +	spin_lock_irqsave(&drv->lock, irq_flags);
> > +	fence->drv = drv;
> > +	fence->seq = ++drv->sync_seq;
> > +	dma_fence_init(&fence->f, &virtio_fence_ops,
&drv->lock,
> > +		       drv->context, fence->seq);
> > +	spin_unlock_irqrestore(&drv->lock, irq_flags);
> 
> seq assignment ...
> 
> > +
> > +	return fence;
> > +}
> > +
> >  int virtio_gpu_fence_emit(struct virtio_gpu_device *vgdev,
> >  			  struct virtio_gpu_ctrl_hdr *cmd_hdr,
> > -			  struct virtio_gpu_fence **fence)
> > +			  struct virtio_gpu_fence *fence)
> >  {
> >  	struct virtio_gpu_fence_driver *drv = &vgdev->fence_drv;
> >  	unsigned long irq_flags;
> >  
> > -	*fence = kmalloc(sizeof(struct virtio_gpu_fence), GFP_ATOMIC);
> > -	if ((*fence) == NULL)
> > -		return -ENOMEM;
> > -
> >  	spin_lock_irqsave(&drv->lock, irq_flags);
> > -	(*fence)->drv = drv;
> > -	(*fence)->seq = ++drv->sync_seq;
> > -	dma_fence_init(&(*fence)->f, &virtio_fence_ops,
&drv->lock,
> > -		       drv->context, (*fence)->seq);
> 
> ... must stay here.  Otherwise requests can be submitted to the virt
> queue with fence sequence numbers out of order.
Yes, makes sense. So I'll just leave the kmalloc in there.

Gustavo