1 files changed, 496 insertions, 162 deletions

diff --git a/drivers/media/v4l2-core/videobuf2-core.c b/drivers/media/v4l2-core/videobuf2-core.c
index a127925..f9059bb 100644
--- a/drivers/media/v4l2-core/videobuf2-core.c
+++ b/drivers/media/v4l2-core/videobuf2-core.c
@@ -33,17 +33,74 @@ module_param(debug, int, 0644);
 			printk(KERN_DEBUG "vb2: " fmt, ## arg);		\
 	} while (0)
 
-#define call_memop(q, op, args...)					\
-	(((q)->mem_ops->op) ?						\
-		((q)->mem_ops->op(args)) : 0)
+#ifdef CONFIG_VIDEO_ADV_DEBUG
+
+/*
+ * If advanced debugging is on, then count how often each op is called,
+ * which can either be per-buffer or per-queue.
+ *
+ * If the op failed then the 'fail_' variant is called to decrease the
+ * counter. That makes it easy to check that the 'init' and 'cleanup'
+ * (and variations thereof) stay balanced.
+ */
+
+#define call_memop(vb, op, args...)					\
+({									\
+	struct vb2_queue *_q = (vb)->vb2_queue;				\
+	dprintk(2, "call_memop(%p, %d, %s)%s\n",			\
+		_q, (vb)->v4l2_buf.index, #op,				\
+		_q->mem_ops->op ? "" : " (nop)");			\
+	(vb)->cnt_mem_ ## op++;						\
+	_q->mem_ops->op ? _q->mem_ops->op(args) : 0;			\
+})
+#define fail_memop(vb, op) ((vb)->cnt_mem_ ## op--)
+
+#define call_qop(q, op, args...)					\
+({									\
+	dprintk(2, "call_qop(%p, %s)%s\n", q, #op,			\
+		(q)->ops->op ? "" : " (nop)");				\
+	(q)->cnt_ ## op++;						\
+	(q)->ops->op ? (q)->ops->op(args) : 0;				\
+})
+#define fail_qop(q, op) ((q)->cnt_ ## op--)
+
+#define call_vb_qop(vb, op, args...)					\
+({									\
+	struct vb2_queue *_q = (vb)->vb2_queue;				\
+	dprintk(2, "call_vb_qop(%p, %d, %s)%s\n",			\
+		_q, (vb)->v4l2_buf.index, #op,				\
+		_q->ops->op ? "" : " (nop)");				\
+	(vb)->cnt_ ## op++;						\
+	_q->ops->op ? _q->ops->op(args) : 0;				\
+})
+#define fail_vb_qop(vb, op) ((vb)->cnt_ ## op--)
+
+#else
+
+#define call_memop(vb, op, args...)					\
+	((vb)->vb2_queue->mem_ops->op ? (vb)->vb2_queue->mem_ops->op(args) : 0)
+#define fail_memop(vb, op)
 
 #define call_qop(q, op, args...)					\
-	(((q)->ops->op) ? ((q)->ops->op(args)) : 0)
+	((q)->ops->op ? (q)->ops->op(args) : 0)
+#define fail_qop(q, op)
 
+#define call_vb_qop(vb, op, args...)					\
+	((vb)->vb2_queue->ops->op ? (vb)->vb2_queue->ops->op(args) : 0)
+#define fail_vb_qop(vb, op)
+
+#endif
+
+/* Flags that are set by the vb2 core */
 #define V4L2_BUFFER_MASK_FLAGS	(V4L2_BUF_FLAG_MAPPED | V4L2_BUF_FLAG_QUEUED | \
 				 V4L2_BUF_FLAG_DONE | V4L2_BUF_FLAG_ERROR | \
 				 V4L2_BUF_FLAG_PREPARED | \
 				 V4L2_BUF_FLAG_TIMESTAMP_MASK)
+/* Output buffer flags that should be passed on to the driver */
+#define V4L2_BUFFER_OUT_FLAGS	(V4L2_BUF_FLAG_PFRAME | V4L2_BUF_FLAG_BFRAME | \
+				 V4L2_BUF_FLAG_KEYFRAME | V4L2_BUF_FLAG_TIMECODE)
+
+static void __vb2_queue_cancel(struct vb2_queue *q);
 
 /**
  * __vb2_buf_mem_alloc() - allocate video memory for the given buffer
@@ -61,7 +118,7 @@ static int __vb2_buf_mem_alloc(struct vb2_buffer *vb)
 	for (plane = 0; plane < vb->num_planes; ++plane) {
 		unsigned long size = PAGE_ALIGN(q->plane_sizes[plane]);
 
-		mem_priv = call_memop(q, alloc, q->alloc_ctx[plane],
+		mem_priv = call_memop(vb, alloc, q->alloc_ctx[plane],
 				      size, q->gfp_flags);
 		if (IS_ERR_OR_NULL(mem_priv))
 			goto free;
@@ -73,9 +130,10 @@ static int __vb2_buf_mem_alloc(struct vb2_buffer *vb)
 
 	return 0;
 free:
+	fail_memop(vb, alloc);
 	/* Free already allocated memory if one of the allocations failed */
 	for (; plane > 0; --plane) {
-		call_memop(q, put, vb->planes[plane - 1].mem_priv);
+		call_memop(vb, put, vb->planes[plane - 1].mem_priv);
 		vb->planes[plane - 1].mem_priv = NULL;
 	}
 
@@ -87,11 +145,10 @@ free:
  */
 static void __vb2_buf_mem_free(struct vb2_buffer *vb)
 {
-	struct vb2_queue *q = vb->vb2_queue;
 	unsigned int plane;
 
 	for (plane = 0; plane < vb->num_planes; ++plane) {
-		call_memop(q, put, vb->planes[plane].mem_priv);
+		call_memop(vb, put, vb->planes[plane].mem_priv);
 		vb->planes[plane].mem_priv = NULL;
 		dprintk(3, "Freed plane %d of buffer %d\n", plane,
 			vb->v4l2_buf.index);
@@ -104,12 +161,11 @@ static void __vb2_buf_mem_free(struct vb2_buffer *vb)
  */
 static void __vb2_buf_userptr_put(struct vb2_buffer *vb)
 {
-	struct vb2_queue *q = vb->vb2_queue;
 	unsigned int plane;
 
 	for (plane = 0; plane < vb->num_planes; ++plane) {
 		if (vb->planes[plane].mem_priv)
-			call_memop(q, put_userptr, vb->planes[plane].mem_priv);
+			call_memop(vb, put_userptr, vb->planes[plane].mem_priv);
 		vb->planes[plane].mem_priv = NULL;
 	}
 }
@@ -118,15 +174,15 @@ static void __vb2_buf_userptr_put(struct vb2_buffer *vb)
  * __vb2_plane_dmabuf_put() - release memory associated with
  * a DMABUF shared plane
  */
-static void __vb2_plane_dmabuf_put(struct vb2_queue *q, struct vb2_plane *p)
+static void __vb2_plane_dmabuf_put(struct vb2_buffer *vb, struct vb2_plane *p)
 {
 	if (!p->mem_priv)
 		return;
 
 	if (p->dbuf_mapped)
-		call_memop(q, unmap_dmabuf, p->mem_priv);
+		call_memop(vb, unmap_dmabuf, p->mem_priv);
 
-	call_memop(q, detach_dmabuf, p->mem_priv);
+	call_memop(vb, detach_dmabuf, p->mem_priv);
 	dma_buf_put(p->dbuf);
 	memset(p, 0, sizeof(*p));
 }
@@ -137,11 +193,10 @@ static void __vb2_plane_dmabuf_put(struct vb2_queue *q, struct vb2_plane *p)
  */
 static void __vb2_buf_dmabuf_put(struct vb2_buffer *vb)
 {
-	struct vb2_queue *q = vb->vb2_queue;
 	unsigned int plane;
 
 	for (plane = 0; plane < vb->num_planes; ++plane)
-		__vb2_plane_dmabuf_put(q, &vb->planes[plane]);
+		__vb2_plane_dmabuf_put(vb, &vb->planes[plane]);
 }
 
 /**
@@ -246,10 +301,11 @@ static int __vb2_queue_alloc(struct vb2_queue *q, enum v4l2_memory memory,
 			 * callback, if given. An error in initialization
 			 * results in queue setup failure.
 			 */
-			ret = call_qop(q, buf_init, vb);
+			ret = call_vb_qop(vb, buf_init, vb);
 			if (ret) {
 				dprintk(1, "Buffer %d %p initialization"
 					" failed\n", buffer, vb);
+				fail_vb_qop(vb, buf_init);
 				__vb2_buf_mem_free(vb);
 				kfree(vb);
 				break;
@@ -321,18 +377,79 @@ static int __vb2_queue_free(struct vb2_queue *q, unsigned int buffers)
 	}
 
 	/* Call driver-provided cleanup function for each buffer, if provided */
-	if (q->ops->buf_cleanup) {
-		for (buffer = q->num_buffers - buffers; buffer < q->num_buffers;
-		     ++buffer) {
-			if (NULL == q->bufs[buffer])
-				continue;
-			q->ops->buf_cleanup(q->bufs[buffer]);
-		}
+	for (buffer = q->num_buffers - buffers; buffer < q->num_buffers;
+	     ++buffer) {
+		struct vb2_buffer *vb = q->bufs[buffer];
+
+		if (vb && vb->planes[0].mem_priv)
+			call_vb_qop(vb, buf_cleanup, vb);
 	}
 
 	/* Release video buffer memory */
 	__vb2_free_mem(q, buffers);
 
+#ifdef CONFIG_VIDEO_ADV_DEBUG
+	/*
+	 * Check that all the calls were balances during the life-time of this
+	 * queue. If not (or if the debug level is 1 or up), then dump the
+	 * counters to the kernel log.
+	 */
+	if (q->num_buffers) {
+		bool unbalanced = q->cnt_start_streaming != q->cnt_stop_streaming ||
+				  q->cnt_wait_prepare != q->cnt_wait_finish;
+
+		if (unbalanced || debug) {
+			pr_info("vb2: counters for queue %p:%s\n", q,
+				unbalanced ? " UNBALANCED!" : "");
+			pr_info("vb2:     setup: %u start_streaming: %u stop_streaming: %u\n",
+				q->cnt_queue_setup, q->cnt_start_streaming,
+				q->cnt_stop_streaming);
+			pr_info("vb2:     wait_prepare: %u wait_finish: %u\n",
+				q->cnt_wait_prepare, q->cnt_wait_finish);
+		}
+		q->cnt_queue_setup = 0;
+		q->cnt_wait_prepare = 0;
+		q->cnt_wait_finish = 0;
+		q->cnt_start_streaming = 0;
+		q->cnt_stop_streaming = 0;
+	}
+	for (buffer = 0; buffer < q->num_buffers; ++buffer) {
+		struct vb2_buffer *vb = q->bufs[buffer];
+		bool unbalanced = vb->cnt_mem_alloc != vb->cnt_mem_put ||
+				  vb->cnt_mem_prepare != vb->cnt_mem_finish ||
+				  vb->cnt_mem_get_userptr != vb->cnt_mem_put_userptr ||
+				  vb->cnt_mem_attach_dmabuf != vb->cnt_mem_detach_dmabuf ||
+				  vb->cnt_mem_map_dmabuf != vb->cnt_mem_unmap_dmabuf ||
+				  vb->cnt_buf_queue != vb->cnt_buf_done ||
+				  vb->cnt_buf_prepare != vb->cnt_buf_finish ||
+				  vb->cnt_buf_init != vb->cnt_buf_cleanup;
+
+		if (unbalanced || debug) {
+			pr_info("vb2:   counters for queue %p, buffer %d:%s\n",
+				q, buffer, unbalanced ? " UNBALANCED!" : "");
+			pr_info("vb2:     buf_init: %u buf_cleanup: %u buf_prepare: %u buf_finish: %u\n",
+				vb->cnt_buf_init, vb->cnt_buf_cleanup,
+				vb->cnt_buf_prepare, vb->cnt_buf_finish);
+			pr_info("vb2:     buf_queue: %u buf_done: %u\n",
+				vb->cnt_buf_queue, vb->cnt_buf_done);
+			pr_info("vb2:     alloc: %u put: %u prepare: %u finish: %u mmap: %u\n",
+				vb->cnt_mem_alloc, vb->cnt_mem_put,
+				vb->cnt_mem_prepare, vb->cnt_mem_finish,
+				vb->cnt_mem_mmap);
+			pr_info("vb2:     get_userptr: %u put_userptr: %u\n",
+				vb->cnt_mem_get_userptr, vb->cnt_mem_put_userptr);
+			pr_info("vb2:     attach_dmabuf: %u detach_dmabuf: %u map_dmabuf: %u unmap_dmabuf: %u\n",
+				vb->cnt_mem_attach_dmabuf, vb->cnt_mem_detach_dmabuf,
+				vb->cnt_mem_map_dmabuf, vb->cnt_mem_unmap_dmabuf);
+			pr_info("vb2:     get_dmabuf: %u num_users: %u vaddr: %u cookie: %u\n",
+				vb->cnt_mem_get_dmabuf,
+				vb->cnt_mem_num_users,
+				vb->cnt_mem_vaddr,
+				vb->cnt_mem_cookie);
+		}
+	}
+#endif
+
 	/* Free videobuf buffers */
 	for (buffer = q->num_buffers - buffers; buffer < q->num_buffers;
 	     ++buffer) {
@@ -341,9 +458,10 @@ static int __vb2_queue_free(struct vb2_queue *q, unsigned int buffers)
 	}
 
 	q->num_buffers -= buffers;
-	if (!q->num_buffers)
+	if (!q->num_buffers) {
 		q->memory = 0;
-	INIT_LIST_HEAD(&q->queued_list);
+		INIT_LIST_HEAD(&q->queued_list);
+	}
 	return 0;
 }
 
@@ -424,7 +542,7 @@ static bool __buffer_in_use(struct vb2_queue *q, struct vb2_buffer *vb)
 		 * case anyway. If num_users() returns more than 1,
 		 * we are not the only user of the plane's memory.
 		 */
-		if (mem_priv && call_memop(q, num_users, mem_priv) > 1)
+		if (mem_priv && call_memop(vb, num_users, mem_priv) > 1)
 			return true;
 	}
 	return false;
@@ -484,7 +602,16 @@ static void __fill_v4l2_buffer(struct vb2_buffer *vb, struct v4l2_buffer *b)
 	 * Clear any buffer state related flags.
 	 */
 	b->flags &= ~V4L2_BUFFER_MASK_FLAGS;
-	b->flags |= q->timestamp_type;
+	b->flags |= q->timestamp_flags & V4L2_BUF_FLAG_TIMESTAMP_MASK;
+	if ((q->timestamp_flags & V4L2_BUF_FLAG_TIMESTAMP_MASK) !=
+	    V4L2_BUF_FLAG_TIMESTAMP_COPY) {
+		/*
+		 * For non-COPY timestamps, drop timestamp source bits
+		 * and obtain the timestamp source from the queue.
+		 */
+		b->flags &= ~V4L2_BUF_FLAG_TSTAMP_SRC_MASK;
+		b->flags |= q->timestamp_flags & V4L2_BUF_FLAG_TSTAMP_SRC_MASK;
+	}
 
 	switch (vb->state) {
 	case VB2_BUF_STATE_QUEUED:
@@ -677,6 +804,12 @@ static int __reqbufs(struct vb2_queue *q, struct v4l2_requestbuffers *req)
 			return -EBUSY;
 		}
 
+		/*
+		 * Call queue_cancel to clean up any buffers in the PREPARED or
+		 * QUEUED state which is possible if buffers were prepared or
+		 * queued without ever calling STREAMON.
+		 */
+		__vb2_queue_cancel(q);
 		ret = __vb2_queue_free(q, q->num_buffers);
 		if (ret)
 			return ret;
@@ -693,6 +826,7 @@ static int __reqbufs(struct vb2_queue *q, struct v4l2_requestbuffers *req)
 	 * Make sure the requested values and current defaults are sane.
 	 */
 	num_buffers = min_t(unsigned int, req->count, VIDEO_MAX_FRAME);
+	num_buffers = max_t(unsigned int, req->count, q->min_buffers_needed);
 	memset(q->plane_sizes, 0, sizeof(q->plane_sizes));
 	memset(q->alloc_ctx, 0, sizeof(q->alloc_ctx));
 	q->memory = req->memory;
@@ -703,26 +837,35 @@ static int __reqbufs(struct vb2_queue *q, struct v4l2_requestbuffers *req)
 	 */
 	ret = call_qop(q, queue_setup, q, NULL, &num_buffers, &num_planes,
 		       q->plane_sizes, q->alloc_ctx);
-	if (ret)
+	if (ret) {
+		fail_qop(q, queue_setup);
 		return ret;
+	}
 
 	/* Finally, allocate buffers and video memory */
-	ret = __vb2_queue_alloc(q, req->memory, num_buffers, num_planes);
-	if (ret == 0) {
+	allocated_buffers = __vb2_queue_alloc(q, req->memory, num_buffers, num_planes);
+	if (allocated_buffers == 0) {
 		dprintk(1, "Memory allocation failed\n");
 		return -ENOMEM;
 	}
 
-	allocated_buffers = ret;
+	/*
+	 * There is no point in continuing if we can't allocate the minimum
+	 * number of buffers needed by this vb2_queue.
+	 */
+	if (allocated_buffers < q->min_buffers_needed)
+		ret = -ENOMEM;
 
 	/*
 	 * Check if driver can handle the allocated number of buffers.
 	 */
-	if (allocated_buffers < num_buffers) {
+	if (!ret && allocated_buffers < num_buffers) {
 		num_buffers = allocated_buffers;
 
 		ret = call_qop(q, queue_setup, q, NULL, &num_buffers,
 			       &num_planes, q->plane_sizes, q->alloc_ctx);
+		if (ret)
+			fail_qop(q, queue_setup);
 
 		if (!ret && allocated_buffers < num_buffers)
 			ret = -ENOMEM;
@@ -736,6 +879,10 @@ static int __reqbufs(struct vb2_queue *q, struct v4l2_requestbuffers *req)
 	q->num_buffers = allocated_buffers;
 
 	if (ret < 0) {
+		/*
+		 * Note: __vb2_queue_free() will subtract 'allocated_buffers'
+		 * from q->num_buffers.
+		 */
 		__vb2_queue_free(q, allocated_buffers);
 		return ret;
 	}
@@ -803,24 +950,24 @@ static int __create_bufs(struct vb2_queue *q, struct v4l2_create_buffers *create
 	 */
 	ret = call_qop(q, queue_setup, q, &create->format, &num_buffers,
 		       &num_planes, q->plane_sizes, q->alloc_ctx);
-	if (ret)
+	if (ret) {
+		fail_qop(q, queue_setup);
 		return ret;
+	}
 
 	/* Finally, allocate buffers and video memory */
-	ret = __vb2_queue_alloc(q, create->memory, num_buffers,
+	allocated_buffers = __vb2_queue_alloc(q, create->memory, num_buffers,
 				num_planes);
-	if (ret == 0) {
+	if (allocated_buffers == 0) {
 		dprintk(1, "Memory allocation failed\n");
 		return -ENOMEM;
 	}
 
-	allocated_buffers = ret;
-
 	/*
 	 * Check if driver can handle the so far allocated number of buffers.
 	 */
-	if (ret < num_buffers) {
-		num_buffers = ret;
+	if (allocated_buffers < num_buffers) {
+		num_buffers = allocated_buffers;
 
 		/*
 		 * q->num_buffers contains the total number of buffers, that the
@@ -828,6 +975,8 @@ static int __create_bufs(struct vb2_queue *q, struct v4l2_create_buffers *create
 		 */
 		ret = call_qop(q, queue_setup, q, &create->format, &num_buffers,
 			       &num_planes, q->plane_sizes, q->alloc_ctx);
+		if (ret)
+			fail_qop(q, queue_setup);
 
 		if (!ret && allocated_buffers < num_buffers)
 			ret = -ENOMEM;
@@ -841,6 +990,10 @@ static int __create_bufs(struct vb2_queue *q, struct v4l2_create_buffers *create
 	q->num_buffers += allocated_buffers;
 
 	if (ret < 0) {
+		/*
+		 * Note: __vb2_queue_free() will subtract 'allocated_buffers'
+		 * from q->num_buffers.
+		 */
 		__vb2_queue_free(q, allocated_buffers);
 		return -ENOMEM;
 	}
@@ -882,12 +1035,10 @@ EXPORT_SYMBOL_GPL(vb2_create_bufs);
  */
 void *vb2_plane_vaddr(struct vb2_buffer *vb, unsigned int plane_no)
 {
-	struct vb2_queue *q = vb->vb2_queue;
-
 	if (plane_no > vb->num_planes || !vb->planes[plane_no].mem_priv)
 		return NULL;
 
-	return call_memop(q, vaddr, vb->planes[plane_no].mem_priv);
+	return call_memop(vb, vaddr, vb->planes[plane_no].mem_priv);
 
 }
 EXPORT_SYMBOL_GPL(vb2_plane_vaddr);
@@ -905,12 +1056,10 @@ EXPORT_SYMBOL_GPL(vb2_plane_vaddr);
  */
 void *vb2_plane_cookie(struct vb2_buffer *vb, unsigned int plane_no)
 {
-	struct vb2_queue *q = vb->vb2_queue;
-
 	if (plane_no > vb->num_planes || !vb->planes[plane_no].mem_priv)
 		return NULL;
 
-	return call_memop(q, cookie, vb->planes[plane_no].mem_priv);
+	return call_memop(vb, cookie, vb->planes[plane_no].mem_priv);
 }
 EXPORT_SYMBOL_GPL(vb2_plane_cookie);
 
@@ -918,13 +1067,20 @@ EXPORT_SYMBOL_GPL(vb2_plane_cookie);
  * vb2_buffer_done() - inform videobuf that an operation on a buffer is finished
  * @vb:		vb2_buffer returned from the driver
  * @state:	either VB2_BUF_STATE_DONE if the operation finished successfully
- *		or VB2_BUF_STATE_ERROR if the operation finished with an error
+ *		or VB2_BUF_STATE_ERROR if the operation finished with an error.
+ *		If start_streaming fails then it should return buffers with state
+ *		VB2_BUF_STATE_QUEUED to put them back into the queue.
  *
  * This function should be called by the driver after a hardware operation on
  * a buffer is finished and the buffer may be returned to userspace. The driver
  * cannot use this buffer anymore until it is queued back to it by videobuf
  * by the means of buf_queue callback. Only buffers previously queued to the
  * driver by buf_queue can be passed to this function.
+ *
+ * While streaming a buffer can only be returned in state DONE or ERROR.
+ * The start_streaming op can also return them in case the DMA engine cannot
+ * be started for some reason. In that case the buffers should be returned with
+ * state QUEUED.
  */
 void vb2_buffer_done(struct vb2_buffer *vb, enum vb2_buffer_state state)
 {
@@ -932,26 +1088,43 @@ void vb2_buffer_done(struct vb2_buffer *vb, enum vb2_buffer_state state)
 	unsigned long flags;
 	unsigned int plane;
 
-	if (vb->state != VB2_BUF_STATE_ACTIVE)
+	if (WARN_ON(vb->state != VB2_BUF_STATE_ACTIVE))
 		return;
 
-	if (state != VB2_BUF_STATE_DONE && state != VB2_BUF_STATE_ERROR)
-		return;
+	if (!q->start_streaming_called) {
+		if (WARN_ON(state != VB2_BUF_STATE_QUEUED))
+			state = VB2_BUF_STATE_QUEUED;
+	} else if (!WARN_ON(!q->start_streaming_called)) {
+		if (WARN_ON(state != VB2_BUF_STATE_DONE &&
+			    state != VB2_BUF_STATE_ERROR))
+			state = VB2_BUF_STATE_ERROR;
+	}
 
+#ifdef CONFIG_VIDEO_ADV_DEBUG
+	/*
+	 * Although this is not a callback, it still does have to balance
+	 * with the buf_queue op. So update this counter manually.
+	 */
+	vb->cnt_buf_done++;
+#endif
 	dprintk(4, "Done processing on buffer %d, state: %d\n",
 			vb->v4l2_buf.index, state);
 
 	/* sync buffers */
 	for (plane = 0; plane < vb->num_planes; ++plane)
-		call_memop(q, finish, vb->planes[plane].mem_priv);
+		call_memop(vb, finish, vb->planes[plane].mem_priv);
 
 	/* Add the buffer to the done buffers list */
 	spin_lock_irqsave(&q->done_lock, flags);
 	vb->state = state;
-	list_add_tail(&vb->done_entry, &q->done_list);
-	atomic_dec(&q->queued_count);
+	if (state != VB2_BUF_STATE_QUEUED)
+		list_add_tail(&vb->done_entry, &q->done_list);
+	atomic_dec(&q->owned_by_drv_count);
 	spin_unlock_irqrestore(&q->done_lock, flags);
 
+	if (state == VB2_BUF_STATE_QUEUED)
+		return;
+
 	/* Inform any processes that may be waiting for buffers */
 	wake_up(&q->done_wq);
 }
@@ -1025,9 +1198,31 @@ static void __fill_vb2_buffer(struct vb2_buffer *vb, const struct v4l2_buffer *b
 
 	}
 
-	vb->v4l2_buf.field = b->field;
-	vb->v4l2_buf.timestamp = b->timestamp;
+	/* Zero flags that the vb2 core handles */
 	vb->v4l2_buf.flags = b->flags & ~V4L2_BUFFER_MASK_FLAGS;
+	if ((vb->vb2_queue->timestamp_flags & V4L2_BUF_FLAG_TIMESTAMP_MASK) !=
+	    V4L2_BUF_FLAG_TIMESTAMP_COPY || !V4L2_TYPE_IS_OUTPUT(b->type)) {
+		/*
+		 * Non-COPY timestamps and non-OUTPUT queues will get
+		 * their timestamp and timestamp source flags from the
+		 * queue.
+		 */
+		vb->v4l2_buf.flags &= ~V4L2_BUF_FLAG_TSTAMP_SRC_MASK;
+	}
+
+	if (V4L2_TYPE_IS_OUTPUT(b->type)) {
+		/*
+		 * For output buffers mask out the timecode flag:
+		 * this will be handled later in vb2_internal_qbuf().
+		 * The 'field' is valid metadata for this output buffer
+		 * and so that needs to be copied here.
+		 */
+		vb->v4l2_buf.flags &= ~V4L2_BUF_FLAG_TIMECODE;
+		vb->v4l2_buf.field = b->field;
+	} else {
+		/* Zero any output buffer flags as this is a capture buffer */
+		vb->v4l2_buf.flags &= ~V4L2_BUFFER_OUT_FLAGS;
+	}
 }
 
 /**
@@ -1041,6 +1236,7 @@ static int __qbuf_userptr(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 	unsigned int plane;
 	int ret;
 	int write = !V4L2_TYPE_IS_OUTPUT(q->type);
+	bool reacquired = vb->planes[0].mem_priv == NULL;
 
 	/* Copy relevant information provided by the userspace */
 	__fill_vb2_buffer(vb, b, planes);
@@ -1066,20 +1262,25 @@ static int __qbuf_userptr(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 		}
 
 		/* Release previously acquired memory if present */
-		if (vb->planes[plane].mem_priv)
-			call_memop(q, put_userptr, vb->planes[plane].mem_priv);
+		if (vb->planes[plane].mem_priv) {
+			if (!reacquired) {
+				reacquired = true;
+				call_vb_qop(vb, buf_cleanup, vb);
+			}
+			call_memop(vb, put_userptr, vb->planes[plane].mem_priv);
+		}
 
 		vb->planes[plane].mem_priv = NULL;
-		vb->v4l2_planes[plane].m.userptr = 0;
-		vb->v4l2_planes[plane].length = 0;
+		memset(&vb->v4l2_planes[plane], 0, sizeof(struct v4l2_plane));
 
 		/* Acquire each plane's memory */
-		mem_priv = call_memop(q, get_userptr, q->alloc_ctx[plane],
+		mem_priv = call_memop(vb, get_userptr, q->alloc_ctx[plane],
 				      planes[plane].m.userptr,
 				      planes[plane].length, write);
 		if (IS_ERR_OR_NULL(mem_priv)) {
 			dprintk(1, "qbuf: failed acquiring userspace "
 						"memory for plane %d\n", plane);
+			fail_memop(vb, get_userptr);
 			ret = mem_priv ? PTR_ERR(mem_priv) : -EINVAL;
 			goto err;
 		}
@@ -1087,28 +1288,40 @@ static int __qbuf_userptr(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 	}
 
 	/*
-	 * Call driver-specific initialization on the newly acquired buffer,
-	 * if provided.
-	 */
-	ret = call_qop(q, buf_init, vb);
-	if (ret) {
-		dprintk(1, "qbuf: buffer initialization failed\n");
-		goto err;
-	}
-
-	/*
 	 * Now that everything is in order, copy relevant information
 	 * provided by userspace.
 	 */
 	for (plane = 0; plane < vb->num_planes; ++plane)
 		vb->v4l2_planes[plane] = planes[plane];
 
+	if (reacquired) {
+		/*
+		 * One or more planes changed, so we must call buf_init to do
+		 * the driver-specific initialization on the newly acquired
+		 * buffer, if provided.
+		 */
+		ret = call_vb_qop(vb, buf_init, vb);
+		if (ret) {
+			dprintk(1, "qbuf: buffer initialization failed\n");
+			fail_vb_qop(vb, buf_init);
+			goto err;
+		}
+	}
+
+	ret = call_vb_qop(vb, buf_prepare, vb);
+	if (ret) {
+		dprintk(1, "qbuf: buffer preparation failed\n");
+		fail_vb_qop(vb, buf_prepare);
+		call_vb_qop(vb, buf_cleanup, vb);
+		goto err;
+	}
+
 	return 0;
 err:
 	/* In case of errors, release planes that were already acquired */
 	for (plane = 0; plane < vb->num_planes; ++plane) {
 		if (vb->planes[plane].mem_priv)
-			call_memop(q, put_userptr, vb->planes[plane].mem_priv);
+			call_memop(vb, put_userptr, vb->planes[plane].mem_priv);
 		vb->planes[plane].mem_priv = NULL;
 		vb->v4l2_planes[plane].m.userptr = 0;
 		vb->v4l2_planes[plane].length = 0;
@@ -1122,8 +1335,13 @@ err:
  */
 static int __qbuf_mmap(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 {
+	int ret;
+
 	__fill_vb2_buffer(vb, b, vb->v4l2_planes);
-	return 0;
+	ret = call_vb_qop(vb, buf_prepare, vb);
+	if (ret)
+		fail_vb_qop(vb, buf_prepare);
+	return ret;
 }
 
 /**
@@ -1137,6 +1355,7 @@ static int __qbuf_dmabuf(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 	unsigned int plane;
 	int ret;
 	int write = !V4L2_TYPE_IS_OUTPUT(q->type);
+	bool reacquired = vb->planes[0].mem_priv == NULL;
 
 	/* Copy relevant information provided by the userspace */
 	__fill_vb2_buffer(vb, b, planes);
@@ -1172,15 +1391,21 @@ static int __qbuf_dmabuf(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 
 		dprintk(1, "qbuf: buffer for plane %d changed\n", plane);
 
+		if (!reacquired) {
+			reacquired = true;
+			call_vb_qop(vb, buf_cleanup, vb);
+		}
+
 		/* Release previously acquired memory if present */
-		__vb2_plane_dmabuf_put(q, &vb->planes[plane]);
+		__vb2_plane_dmabuf_put(vb, &vb->planes[plane]);
 		memset(&vb->v4l2_planes[plane], 0, sizeof(struct v4l2_plane));
 
 		/* Acquire each plane's memory */
-		mem_priv = call_memop(q, attach_dmabuf, q->alloc_ctx[plane],
+		mem_priv = call_memop(vb, attach_dmabuf, q->alloc_ctx[plane],
 			dbuf, planes[plane].length, write);
 		if (IS_ERR(mem_priv)) {
 			dprintk(1, "qbuf: failed to attach dmabuf\n");
+			fail_memop(vb, attach_dmabuf);
 			ret = PTR_ERR(mem_priv);
 			dma_buf_put(dbuf);
 			goto err;
@@ -1195,32 +1420,44 @@ static int __qbuf_dmabuf(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 	 * the buffer(s)..
 	 */
 	for (plane = 0; plane < vb->num_planes; ++plane) {
-		ret = call_memop(q, map_dmabuf, vb->planes[plane].mem_priv);
+		ret = call_memop(vb, map_dmabuf, vb->planes[plane].mem_priv);
 		if (ret) {
 			dprintk(1, "qbuf: failed to map dmabuf for plane %d\n",
 				plane);
+			fail_memop(vb, map_dmabuf);
 			goto err;
 		}
 		vb->planes[plane].dbuf_mapped = 1;
 	}
 
 	/*
-	 * Call driver-specific initialization on the newly acquired buffer,
-	 * if provided.
-	 */
-	ret = call_qop(q, buf_init, vb);
-	if (ret) {
-		dprintk(1, "qbuf: buffer initialization failed\n");
-		goto err;
-	}
-
-	/*
 	 * Now that everything is in order, copy relevant information
 	 * provided by userspace.
 	 */
 	for (plane = 0; plane < vb->num_planes; ++plane)
 		vb->v4l2_planes[plane] = planes[plane];
 
+	if (reacquired) {
+		/*
+		 * Call driver-specific initialization on the newly acquired buffer,
+		 * if provided.
+		 */
+		ret = call_vb_qop(vb, buf_init, vb);
+		if (ret) {
+			dprintk(1, "qbuf: buffer initialization failed\n");
+			fail_vb_qop(vb, buf_init);
+			goto err;
+		}
+	}
+
+	ret = call_vb_qop(vb, buf_prepare, vb);
+	if (ret) {
+		dprintk(1, "qbuf: buffer preparation failed\n");
+		fail_vb_qop(vb, buf_prepare);
+		call_vb_qop(vb, buf_cleanup, vb);
+		goto err;
+	}
+
 	return 0;
 err:
 	/* In case of errors, release planes that were already acquired */
@@ -1238,13 +1475,13 @@ static void __enqueue_in_driver(struct vb2_buffer *vb)
 	unsigned int plane;
 
 	vb->state = VB2_BUF_STATE_ACTIVE;
-	atomic_inc(&q->queued_count);
+	atomic_inc(&q->owned_by_drv_count);
 
 	/* sync buffers */
 	for (plane = 0; plane < vb->num_planes; ++plane)
-		call_memop(q, prepare, vb->planes[plane].mem_priv);
+		call_memop(vb, prepare, vb->planes[plane].mem_priv);
 
-	q->ops->buf_queue(vb);
+	call_vb_qop(vb, buf_queue, vb);
 }
 
 static int __buf_prepare(struct vb2_buffer *vb, const struct v4l2_buffer *b)
@@ -1261,6 +1498,10 @@ static int __buf_prepare(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 	}
 
 	vb->state = VB2_BUF_STATE_PREPARING;
+	vb->v4l2_buf.timestamp.tv_sec = 0;
+	vb->v4l2_buf.timestamp.tv_usec = 0;
+	vb->v4l2_buf.sequence = 0;
+
 	switch (q->memory) {
 	case V4L2_MEMORY_MMAP:
 		ret = __qbuf_mmap(vb, b);
@@ -1295,8 +1536,6 @@ static int __buf_prepare(struct vb2_buffer *vb, const struct v4l2_buffer *b)
 		ret = -EINVAL;
 	}
 
-	if (!ret)
-		ret = call_qop(q, buf_prepare, vb);
 	if (ret)
 		dprintk(1, "qbuf: buffer preparation failed: %d\n", ret);
 	vb->state = ret ? VB2_BUF_STATE_DEQUEUED : VB2_BUF_STATE_PREPARED;
@@ -1382,32 +1621,49 @@ EXPORT_SYMBOL_GPL(vb2_prepare_buf);
  * vb2_start_streaming() - Attempt to start streaming.
  * @q:		videobuf2 queue
  *
- * If there are not enough buffers, then retry_start_streaming is set to
- * 1 and 0 is returned. The next time a buffer is queued and
- * retry_start_streaming is 1, this function will be called again to
- * retry starting the DMA engine.
+ * Attempt to start streaming. When this function is called there must be
+ * at least q->min_buffers_needed buffers queued up (i.e. the minimum
+ * number of buffers required for the DMA engine to function). If the
+ * @start_streaming op fails it is supposed to return all the driver-owned
+ * buffers back to vb2 in state QUEUED. Check if that happened and if
+ * not warn and reclaim them forcefully.
  */
 static int vb2_start_streaming(struct vb2_queue *q)
 {
+	struct vb2_buffer *vb;
 	int ret;
 
-	/* Tell the driver to start streaming */
-	ret = call_qop(q, start_streaming, q, atomic_read(&q->queued_count));
-
 	/*
-	 * If there are not enough buffers queued to start streaming, then
-	 * the start_streaming operation will return -ENOBUFS and you have to
-	 * retry when the next buffer is queued.
+	 * If any buffers were queued before streamon,
+	 * we can now pass them to driver for processing.
 	 */
-	if (ret == -ENOBUFS) {
-		dprintk(1, "qbuf: not enough buffers, retry when more buffers are queued.\n");
-		q->retry_start_streaming = 1;
+	list_for_each_entry(vb, &q->queued_list, queued_entry)
+		__enqueue_in_driver(vb);
+
+	/* Tell the driver to start streaming */
+	ret = call_qop(q, start_streaming, q,
+		       atomic_read(&q->owned_by_drv_count));
+	q->start_streaming_called = ret == 0;
+	if (!ret)
 		return 0;
+
+	fail_qop(q, start_streaming);
+	dprintk(1, "qbuf: driver refused to start streaming\n");
+	if (WARN_ON(atomic_read(&q->owned_by_drv_count))) {
+		unsigned i;
+
+		/*
+		 * Forcefully reclaim buffers if the driver did not
+		 * correctly return them to vb2.
+		 */
+		for (i = 0; i < q->num_buffers; ++i) {
+			vb = q->bufs[i];
+			if (vb->state == VB2_BUF_STATE_ACTIVE)
+				vb2_buffer_done(vb, VB2_BUF_STATE_QUEUED);
+		}
+		/* Must be zero now */
+		WARN_ON(atomic_read(&q->owned_by_drv_count));
 	}
-	if (ret)
-		dprintk(1, "qbuf: driver refused to start streaming\n");
-	else
-		q->retry_start_streaming = 0;
 	return ret;
 }
 
@@ -1420,11 +1676,6 @@ static int vb2_internal_qbuf(struct vb2_queue *q, struct v4l2_buffer *b)
 		return ret;
 
 	vb = q->bufs[b->index];
-	if (vb->state != VB2_BUF_STATE_DEQUEUED) {
-		dprintk(1, "%s(): invalid buffer state %d\n", __func__,
-			vb->state);
-		return -EINVAL;
-	}
 
 	switch (vb->state) {
 	case VB2_BUF_STATE_DEQUEUED:
@@ -1438,7 +1689,8 @@ static int vb2_internal_qbuf(struct vb2_queue *q, struct v4l2_buffer *b)
 		dprintk(1, "qbuf: buffer still being prepared\n");
 		return -EINVAL;
 	default:
-		dprintk(1, "qbuf: buffer already in use\n");
+		dprintk(1, "%s(): invalid buffer state %d\n", __func__,
+			vb->state);
 		return -EINVAL;
 	}
 
@@ -1447,19 +1699,39 @@ static int vb2_internal_qbuf(struct vb2_queue *q, struct v4l2_buffer *b)
 	 * dequeued in dqbuf.
 	 */
 	list_add_tail(&vb->queued_entry, &q->queued_list);
+	q->queued_count++;
 	vb->state = VB2_BUF_STATE_QUEUED;
+	if (V4L2_TYPE_IS_OUTPUT(q->type)) {
+		/*
+		 * For output buffers copy the timestamp if needed,
+		 * and the timecode field and flag if needed.
+		 */
+		if ((q->timestamp_flags & V4L2_BUF_FLAG_TIMESTAMP_MASK) ==
+		    V4L2_BUF_FLAG_TIMESTAMP_COPY)
+			vb->v4l2_buf.timestamp = b->timestamp;
+		vb->v4l2_buf.flags |= b->flags & V4L2_BUF_FLAG_TIMECODE;
+		if (b->flags & V4L2_BUF_FLAG_TIMECODE)
+			vb->v4l2_buf.timecode = b->timecode;
+	}
 
 	/*
 	 * If already streaming, give the buffer to driver for processing.
 	 * If not, the buffer will be given to driver on next streamon.
 	 */
-	if (q->streaming)
+	if (q->start_streaming_called)
 		__enqueue_in_driver(vb);
 
 	/* Fill buffer information for the userspace */
 	__fill_v4l2_buffer(vb, b);
 
-	if (q->retry_start_streaming) {
+	/*
+	 * If streamon has been called, and we haven't yet called
+	 * start_streaming() since not enough buffers were queued, and
+	 * we now have reached the minimum number of queued buffers,
+	 * then we can finally call start_streaming().
+	 */
+	if (q->streaming && !q->start_streaming_called &&
+	    q->queued_count >= q->min_buffers_needed) {
 		ret = vb2_start_streaming(q);
 		if (ret)
 			return ret;
@@ -1614,8 +1886,8 @@ int vb2_wait_for_all_buffers(struct vb2_queue *q)
 		return -EINVAL;
 	}
 
-	if (!q->retry_start_streaming)
-		wait_event(q->done_wq, !atomic_read(&q->queued_count));
+	if (q->start_streaming_called)
+		wait_event(q->done_wq, !atomic_read(&q->owned_by_drv_count));
 	return 0;
 }
 EXPORT_SYMBOL_GPL(vb2_wait_for_all_buffers);
@@ -1639,7 +1911,7 @@ static void __vb2_dqbuf(struct vb2_buffer *vb)
 		for (i = 0; i < vb->num_planes; ++i) {
 			if (!vb->planes[i].dbuf_mapped)
 				continue;
-			call_memop(q, unmap_dmabuf, vb->planes[i].mem_priv);
+			call_memop(vb, unmap_dmabuf, vb->planes[i].mem_priv);
 			vb->planes[i].dbuf_mapped = 0;
 		}
 }
@@ -1657,12 +1929,6 @@ static int vb2_internal_dqbuf(struct vb2_queue *q, struct v4l2_buffer *b, bool n
 	if (ret < 0)
 		return ret;
 
-	ret = call_qop(q, buf_finish, vb);
-	if (ret) {
-		dprintk(1, "dqbuf: buffer finish failed\n");
-		return ret;
-	}
-
 	switch (vb->state) {
 	case VB2_BUF_STATE_DONE:
 		dprintk(3, "dqbuf: Returning done buffer\n");
@@ -1675,10 +1941,13 @@ static int vb2_internal_dqbuf(struct vb2_queue *q, struct v4l2_buffer *b, bool n
 		return -EINVAL;
 	}
 
+	call_vb_qop(vb, buf_finish, vb);
+
 	/* Fill buffer information for the userspace */
 	__fill_v4l2_buffer(vb, b);
 	/* Remove from videobuf queue */
 	list_del(&vb->queued_entry);
+	q->queued_count--;
 	/* go back to dequeued state */
 	__vb2_dqbuf(vb);
 
@@ -1729,18 +1998,23 @@ static void __vb2_queue_cancel(struct vb2_queue *q)
 {
 	unsigned int i;
 
-	if (q->retry_start_streaming) {
-		q->retry_start_streaming = 0;
-		q->streaming = 0;
-	}
-
 	/*
 	 * Tell driver to stop all transactions and release all queued
 	 * buffers.
 	 */
-	if (q->streaming)
+	if (q->start_streaming_called)
 		call_qop(q, stop_streaming, q);
 	q->streaming = 0;
+	q->start_streaming_called = 0;
+	q->queued_count = 0;
+
+	if (WARN_ON(atomic_read(&q->owned_by_drv_count))) {
+		for (i = 0; i < q->num_buffers; ++i)
+			if (q->bufs[i]->state == VB2_BUF_STATE_ACTIVE)
+				vb2_buffer_done(q->bufs[i], VB2_BUF_STATE_ERROR);
+		/* Must be zero now */
+		WARN_ON(atomic_read(&q->owned_by_drv_count));
+	}
 
 	/*
 	 * Remove all buffers from videobuf's list...
@@ -1751,19 +2025,31 @@ static void __vb2_queue_cancel(struct vb2_queue *q)
 	 * has not already dequeued before initiating cancel.
 	 */
 	INIT_LIST_HEAD(&q->done_list);
-	atomic_set(&q->queued_count, 0);
+	atomic_set(&q->owned_by_drv_count, 0);
 	wake_up_all(&q->done_wq);
 
 	/*
 	 * Reinitialize all buffers for next use.
+	 * Make sure to call buf_finish for any queued buffers. Normally
+	 * that's done in dqbuf, but that's not going to happen when we
+	 * cancel the whole queue. Note: this code belongs here, not in
+	 * __vb2_dqbuf() since in vb2_internal_dqbuf() there is a critical
+	 * call to __fill_v4l2_buffer() after buf_finish(). That order can't
+	 * be changed, so we can't move the buf_finish() to __vb2_dqbuf().
 	 */
-	for (i = 0; i < q->num_buffers; ++i)
-		__vb2_dqbuf(q->bufs[i]);
+	for (i = 0; i < q->num_buffers; ++i) {
+		struct vb2_buffer *vb = q->bufs[i];
+
+		if (vb->state != VB2_BUF_STATE_DEQUEUED) {
+			vb->state = VB2_BUF_STATE_PREPARED;
+			call_vb_qop(vb, buf_finish, vb);
+		}
+		__vb2_dqbuf(vb);
+	}
 }
 
 static int vb2_internal_streamon(struct vb2_queue *q, enum v4l2_buf_type type)
 {
-	struct vb2_buffer *vb;
 	int ret;
 
 	if (type != q->type) {
@@ -1781,18 +2067,26 @@ static int vb2_internal_streamon(struct vb2_queue *q, enum v4l2_buf_type type)
 		return -EINVAL;
 	}
 
+	if (!q->num_buffers) {
+		dprintk(1, "streamon: no buffers have been allocated\n");
+		return -EINVAL;
+	}
+	if (q->num_buffers < q->min_buffers_needed) {
+		dprintk(1, "streamon: need at least %u allocated buffers\n",
+				q->min_buffers_needed);
+		return -EINVAL;
+	}
+
 	/*
-	 * If any buffers were queued before streamon,
-	 * we can now pass them to driver for processing.
+	 * Tell driver to start streaming provided sufficient buffers
+	 * are available.
 	 */
-	list_for_each_entry(vb, &q->queued_list, queued_entry)
-		__enqueue_in_driver(vb);
-
-	/* Tell driver to start streaming. */
-	ret = vb2_start_streaming(q);
-	if (ret) {
-		__vb2_queue_cancel(q);
-		return ret;
+	if (q->queued_count >= q->min_buffers_needed) {
+		ret = vb2_start_streaming(q);
+		if (ret) {
+			__vb2_queue_cancel(q);
+			return ret;
+		}
 	}
 
 	q->streaming = 1;
@@ -1831,14 +2125,14 @@ static int vb2_internal_streamoff(struct vb2_queue *q, enum v4l2_buf_type type)
 		return -EINVAL;
 	}
 
-	if (!q->streaming) {
-		dprintk(3, "streamoff successful: not streaming\n");
-		return 0;
-	}
-
 	/*
 	 * Cancel will pause streaming and remove all buffers from the driver
 	 * and videobuf, effectively returning control over them to userspace.
+	 *
+	 * Note that we do this even if q->streaming == 0: if you prepare or
+	 * queue buffers, and then call streamoff without ever having called
+	 * streamon, you would still expect those buffers to be returned to
+	 * their normal dequeued state.
 	 */
 	__vb2_queue_cancel(q);
 
@@ -1950,10 +2244,11 @@ int vb2_expbuf(struct vb2_queue *q, struct v4l2_exportbuffer *eb)
 
 	vb_plane = &vb->planes[eb->plane];
 
-	dbuf = call_memop(q, get_dmabuf, vb_plane->mem_priv, eb->flags & O_ACCMODE);
+	dbuf = call_memop(vb, get_dmabuf, vb_plane->mem_priv, eb->flags & O_ACCMODE);
 	if (IS_ERR_OR_NULL(dbuf)) {
 		dprintk(1, "Failed to export buffer %d, plane %d\n",
 			eb->index, eb->plane);
+		fail_memop(vb, get_dmabuf);
 		return -EINVAL;
 	}
 
@@ -2045,9 +2340,11 @@ int vb2_mmap(struct vb2_queue *q, struct vm_area_struct *vma)
 		return -EINVAL;
 	}
 
-	ret = call_memop(q, mmap, vb->planes[plane].mem_priv, vma);
-	if (ret)
+	ret = call_memop(vb, mmap, vb->planes[plane].mem_priv, vma);
+	if (ret) {
+		fail_memop(vb, mmap);
 		return ret;
+	}
 
 	dprintk(3, "Buffer %d, plane %d successfully mapped\n", buffer, plane);
 	return 0;
@@ -2200,11 +2497,14 @@ int vb2_queue_init(struct vb2_queue *q)
 	    WARN_ON(!q->io_modes)	  ||
 	    WARN_ON(!q->ops->queue_setup) ||
 	    WARN_ON(!q->ops->buf_queue)   ||
-	    WARN_ON(q->timestamp_type & ~V4L2_BUF_FLAG_TIMESTAMP_MASK))
+	    WARN_ON(q->timestamp_flags &
+		    ~(V4L2_BUF_FLAG_TIMESTAMP_MASK |
+		      V4L2_BUF_FLAG_TSTAMP_SRC_MASK)))
 		return -EINVAL;
 
 	/* Warn that the driver should choose an appropriate timestamp type */
-	WARN_ON(q->timestamp_type == V4L2_BUF_FLAG_TIMESTAMP_UNKNOWN);
+	WARN_ON((q->timestamp_flags & V4L2_BUF_FLAG_TIMESTAMP_MASK) ==
+		V4L2_BUF_FLAG_TIMESTAMP_UNKNOWN);
 
 	INIT_LIST_HEAD(&q->queued_list);
 	INIT_LIST_HEAD(&q->done_list);
@@ -2251,6 +2551,22 @@ struct vb2_fileio_buf {
 /**
  * struct vb2_fileio_data - queue context used by file io emulator
  *
+ * @cur_index:	the index of the buffer currently being read from or
+ *		written to. If equal to q->num_buffers then a new buffer
+ *		must be dequeued.
+ * @initial_index: in the read() case all buffers are queued up immediately
+ *		in __vb2_init_fileio() and __vb2_perform_fileio() just cycles
+ *		buffers. However, in the write() case no buffers are initially
+ *		queued, instead whenever a buffer is full it is queued up by
+ *		__vb2_perform_fileio(). Only once all available buffers have
+ *		been queued up will __vb2_perform_fileio() start to dequeue
+ *		buffers. This means that initially __vb2_perform_fileio()
+ *		needs to know what buffer index to use when it is queuing up
+ *		the buffers for the first time. That initial index is stored
+ *		in this field. Once it is equal to q->num_buffers all
+ *		available buffers have been queued and __vb2_perform_fileio()
+ *		should start the normal dequeue/queue cycle.
+ *
  * vb2 provides a compatibility layer and emulator of file io (read and
  * write) calls on top of streaming API. For proper operation it required
  * this structure to save the driver state between each call of the read
@@ -2260,7 +2576,8 @@ struct vb2_fileio_data {
 	struct v4l2_requestbuffers req;
 	struct v4l2_buffer b;
 	struct vb2_fileio_buf bufs[VIDEO_MAX_FRAME];
-	unsigned int index;
+	unsigned int cur_index;
+	unsigned int initial_index;
 	unsigned int q_count;
 	unsigned int dq_count;
 	unsigned int flags;
@@ -2280,9 +2597,9 @@ static int __vb2_init_fileio(struct vb2_queue *q, int read)
 	/*
 	 * Sanity check
 	 */
-	if ((read && !(q->io_modes & VB2_READ)) ||
-	   (!read && !(q->io_modes & VB2_WRITE)))
-		BUG();
+	if (WARN_ON((read && !(q->io_modes & VB2_READ)) ||
+		    (!read && !(q->io_modes & VB2_WRITE))))
+		return -EINVAL;
 
 	/*
 	 * Check if device supports mapping buffers to kernel virtual space.
@@ -2360,7 +2677,12 @@ static int __vb2_init_fileio(struct vb2_queue *q, int read)
 				goto err_reqbufs;
 			fileio->bufs[i].queued = 1;
 		}
-		fileio->index = q->num_buffers;
+		/*
+		 * All buffers have been queued, so mark that by setting
+		 * initial_index to q->num_buffers
+		 */
+		fileio->initial_index = q->num_buffers;
+		fileio->cur_index = q->num_buffers;
 	}
 
 	/*
@@ -2439,7 +2761,7 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
 	/*
 	 * Check if we need to dequeue the buffer.
 	 */
-	index = fileio->index;
+	index = fileio->cur_index;
 	if (index >= q->num_buffers) {
 		/*
 		 * Call vb2_dqbuf to get buffer back.
@@ -2453,7 +2775,7 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
 			return ret;
 		fileio->dq_count += 1;
 
-		index = fileio->b.index;
+		fileio->cur_index = index = fileio->b.index;
 		buf = &fileio->bufs[index];
 
 		/*
@@ -2529,8 +2851,20 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
 		buf->queued = 1;
 		buf->size = vb2_plane_size(q->bufs[index], 0);
 		fileio->q_count += 1;
-		if (fileio->index < q->num_buffers)
-			fileio->index++;
+		/*
+		 * If we are queuing up buffers for the first time, then
+		 * increase initial_index by one.
+		 */
+		if (fileio->initial_index < q->num_buffers)
+			fileio->initial_index++;
+		/*
+		 * The next buffer to use is either a buffer that's going to be
+		 * queued for the first time (initial_index < q->num_buffers)
+		 * or it is equal to q->num_buffers, meaning that the next
+		 * time we need to dequeue a buffer since we've now queued up
+		 * all the 'first time' buffers.
+		 */
+		fileio->cur_index = fileio->initial_index;
 	}
 
 	/*