Re: select(), VIDIOC_DQBUF race

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



Abraham vd Merwe wrote:
Hi!

1. user does select() to wait for a frame to be ready.
2. select polls until it finds a frame that's done and return POLLIN
3. interrupt handler goes and unmark that frame as ready and start capturing
   a new frame into that same buffer.
4. user does VIDIOC_DQBUF ioctl which fails because that buffer is not ready
   anymore.

There's no way with the current API to avoid above scenario unless the
driver already dequeues the buffer in select() and then just confirms it in
the DQBUF ioctl. (Or if VIDIOC_DQBUF blocked).

How am I supposed to avoid the race without resorting to the methods
mentioned above?

This shouldn't be a problem. The driver isn't supposed to capture another image into the buffer until it is queued again. The buffer should have the following states:

1) Unused
2) Waiting to capture
3) Capturing
4) Capture completed
5) Dequeued (go back to 1)

There is no way to go from state 4 back to state 3 withou program intervention. If the driver handles it differently, then there is probably a problem with the driver.

-justin







[Index of Archives]     [Linux DVB]     [Video Disk Recorder]     [Asterisk]     [Photo]     [DCCP]     [Netdev]     [Xorg]     [Util Linux NG]     [Xfree86]     [Free Photo Albums]     [Fedora Users]     [Fedora Women]     [ALSA Users]     [ALSA Devel]     [Linux USB]

Powered by Linux