public inbox for linux-media@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH RFC] videobuf2: Improve file I/O emulation to handle buffers in any order
@ 2013-11-21  1:05 Andy Walls
  2013-11-21  7:56 ` Hans Verkuil
  2013-11-21 13:52 ` Hans Verkuil
  0 siblings, 2 replies; 6+ messages in thread
From: Andy Walls @ 2013-11-21  1:05 UTC (permalink / raw)
  To: linux-media; +Cc: Hans Verkuil, Marek Szyprowski, Kyungmin Park, PawelOsciak

(This patch is RFC, because it was compiled and tested against kernel
v3.5)

videobuf2 file I/O emulation assumed that buffers dequeued from the
driver would return in the order they were enqueued in the driver. 

Improve the file I/O emulator's book-keeping to remove this assumption.

Also remove the, AFAICT, assumption that only read() calls would need to
dequeue a buffer from the driver.

Also set the buf->size properly, if a write() dequeues a buffer.


Signed-off-by: Andy Walls <awalls@md.metrocast.net>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: PawelOsciak<pawel@osciak.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>


diff --git a/drivers/media/video/videobuf2-core.c b/drivers/media/video/videobuf2-core.c
index 9d4e9ed..f330aa4 100644
--- a/drivers/media/video/videobuf2-core.c
+++ b/drivers/media/video/videobuf2-core.c
@@ -1796,6 +1796,7 @@ struct vb2_fileio_data {
 	unsigned int dq_count;
 	unsigned int flags;
 };
+#define FILEIO_INDEX_NOT_SET	((unsigned int) INT_MAX)
 
 /**
  * __vb2_init_fileio() - initialize file io emulator
@@ -1889,6 +1890,7 @@ static int __vb2_init_fileio(struct vb2_queue *q, int read)
 				goto err_reqbufs;
 			fileio->bufs[i].queued = 1;
 		}
+		fileio->index = FILEIO_INDEX_NOT_SET;
 
 		/*
 		 * Start streaming.
@@ -1975,15 +1977,11 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
 	 */
 	q->fileio = NULL;
 
-	index = fileio->index;
-	buf = &fileio->bufs[index];
-
 	/*
 	 * Check if we need to dequeue the buffer.
 	 */
-	if (buf->queued) {
-		struct vb2_buffer *vb;
-
+	index = fileio->index;
+	if (index == FILEIO_INDEX_NOT_SET) {
 		/*
 		 * Call vb2_dqbuf to get buffer back.
 		 */
@@ -1997,12 +1995,19 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
 			goto end;
 		fileio->dq_count += 1;
 
+		fileio->index = fileio->b.index;
+		index = fileio->index;
+		buf = &fileio->bufs[index];
+		
 		/*
 		 * Get number of bytes filled by the driver
 		 */
-		vb = q->bufs[index];
-		buf->size = vb2_get_plane_payload(vb, 0);
+		buf->pos = 0;
 		buf->queued = 0;
+		buf->size = read ? vb2_get_plane_payload(q->bufs[index], 0)
+				 : vb2_plane_size(q->bufs[index], 0);
+	} else {
+		buf = &fileio->bufs[index];
 	}
 
 	/*
@@ -2070,13 +2075,28 @@ static size_t __vb2_perform_fileio(struct vb2_queue *q, char __user *data, size_
 		 */
 		buf->pos = 0;
 		buf->queued = 1;
-		buf->size = q->bufs[0]->v4l2_planes[0].length;
+		buf->size = vb2_plane_size(q->bufs[index], 0);
 		fileio->q_count += 1;
 
 		/*
-		 * Switch to the next buffer
+		 * Decide on the next buffer
 		 */
-		fileio->index = (index + 1) % q->num_buffers;
+		if (read || (q->num_buffers == 1)) {
+			/* Use the next buffer the driver provides back */
+			fileio->index = FILEIO_INDEX_NOT_SET;
+		} else {
+			/* Prefer a buffer that is not quequed in the driver */
+			int initial_index = fileio->index;
+			fileio->index = FILEIO_INDEX_NOT_SET;
+			do {
+				if (++index == q->num_buffers)
+					index = 0;
+				if (!fileio->bufs[index].queued) {
+					fileio->index = index;
+					break;
+				}
+			} while (index != initial_index);
+		}
 
 		/*
 		 * Start streaming if required.



^ permalink raw reply related	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2013-11-22 11:48 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-11-21  1:05 [PATCH RFC] videobuf2: Improve file I/O emulation to handle buffers in any order Andy Walls
2013-11-21  7:56 ` Hans Verkuil
2013-11-21 11:56   ` Andy Walls
2013-11-21 13:52 ` Hans Verkuil
2013-11-21 14:52   ` Andy Walls
2013-11-22 11:50   ` Andy Walls

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox