There is a TODO comment about the dma-buf being mapped in VIDIOC_QBUF
instead of doing it closer to when the actual DMA is going to happen
when the buffers are queued in the driver (i.e: __enqueue_in_driver).
But there is a reason to do it earlier in QBUF, and is that userspace
has no way to know if a exported dma-buf can be imported successfully
and so relies on QBUF succeeding as indication that the dma-buf mapped.
If QBUF fails, the application can fallback to another streaming I/O
method. But moving the dma-buf mapping later when queueing the buffers
can be too late for userspace to recover, since it may had dropped the
buffer(s) already when it knows that the dma-buf mapping failed.
So remove the TODO instead and change the comment to explain this.
Suggested-by: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
vb->planes[plane].mem_priv = mem_priv;
}
- /* TODO: This pins the buffer(s) with dma_buf_map_attachment()).. but
- * really we want to do this just before the DMA, not while queueing
- * the buffer(s)..
+ /*
+ * This pins the buffer(s) with dma_buf_map_attachment()). It's done
+ * here instead just before the DMA, while queueing the buffer(s) so
+ * userspace knows sooner rather than later if the dma-buf map fails.
*/
for (plane = 0; plane < vb->num_planes; ++plane) {
ret = call_memop(vb, map_dmabuf, vb->planes[plane].mem_priv);