Re: [PATCH] v4l2-compliance: call select before dequeuing event

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Jun 07, 2023 at 01:13:49PM +0200, Hans Verkuil wrote:
> On 07/06/2023 04:10, Deborah Brouwer wrote:
> > When the streaming option is called with a stateful decoder,
> > testUserPtr() and testDmaBuf() will hang indefinitely when attempting to
> > dequeue a source change event. To prevent this call select() with a
> > timeout.
> > 
> > Signed-off-by: Deborah Brouwer <deborah.brouwer@xxxxxxxxxxxxx>
> > ---
> > It looks like this was the same issue that was previously fixed for testMmap()
> > in commit f0a5b17d9 ("v4l2-compliance: add timeout when waiting for event").
> > 
> >  utils/v4l2-compliance/v4l2-test-buffers.cpp | 18 ++++++++++++++++++
> >  1 file changed, 18 insertions(+)
> > 
> > diff --git a/utils/v4l2-compliance/v4l2-test-buffers.cpp b/utils/v4l2-compliance/v4l2-test-buffers.cpp
> > index a097a0ff..0396e17e 100644
> > --- a/utils/v4l2-compliance/v4l2-test-buffers.cpp
> > +++ b/utils/v4l2-compliance/v4l2-test-buffers.cpp
> > @@ -1746,8 +1746,17 @@ int testUserPtr(struct node *node, struct node *node_m2m_cap, unsigned frame_cou
> >  
> >  		if (node->is_m2m) {
> >  			if (node->codec_mask & STATEFUL_DECODER) {
> > +				int fd_flags = fcntl(node->g_fd(), F_GETFL);
> > +				struct timeval tv = { 1, 0 };
> > +				fd_set efds;
> >  				v4l2_event ev;
> >  
> > +				fcntl(node->g_fd(), F_SETFL, fd_flags | O_NONBLOCK);
> > +				FD_ZERO(&efds);
> > +				FD_SET(node->g_fd(), &efds);
> > +				ret = select(node->g_fd() + 1, nullptr, nullptr, &efds, &tv);
> > +				fail_on_test_val(ret < 0, ret);
> > +				fail_on_test(ret == 0);
> >  				fail_on_test(node->dqevent(ev));
> 
> This code keeps the fd in non-blocking mode. You need to add a
> fcntl(node->g_fd(), F_SETFL, fd_flags) line here. See also testMmap().

Oh sorry yes I missed that line. I'll send v2 with it fixed.

Related question: would you ever expect this test to do anything other
than timeout? Probably I am missing something, but it seems like there
are no output buffers being queued, so there could never be a source change
event?
I was just testing it with vicodec and I always get the dmesg error
message "No buffer was provided with the request" and then it times out.

> 
> >  				fail_on_test(ev.type != V4L2_EVENT_SOURCE_CHANGE);
> >  				fail_on_test(!(ev.u.src_change.changes & V4L2_EVENT_SRC_CH_RESOLUTION));
> > @@ -1949,8 +1958,17 @@ int testDmaBuf(struct node *expbuf_node, struct node *node, struct node *node_m2
> >  
> >  		if (node->is_m2m) {
> >  			if (node->codec_mask & STATEFUL_DECODER) {
> > +				int fd_flags = fcntl(node->g_fd(), F_GETFL);
> > +				struct timeval tv = { 1, 0 };
> > +				fd_set efds;
> >  				v4l2_event ev;
> >  
> > +				fcntl(node->g_fd(), F_SETFL, fd_flags | O_NONBLOCK);
> > +				FD_ZERO(&efds);
> > +				FD_SET(node->g_fd(), &efds);
> > +				ret = select(node->g_fd() + 1, nullptr, nullptr, &efds, &tv);
> > +				fail_on_test_val(ret < 0, ret);
> > +				fail_on_test(ret == 0);
> >  				fail_on_test(node->dqevent(ev));
> 
> Ditto.
> 
> >  				fail_on_test(ev.type != V4L2_EVENT_SOURCE_CHANGE);
> >  				fail_on_test(!(ev.u.src_change.changes & V4L2_EVENT_SRC_CH_RESOLUTION));
> 
> Regards,
> 
> 	Hans



[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]

  Powered by Linux