Thank you. If I am working on both the user side and the driver itself, why the buffer must contain a full image? Is that a limitation of v4l2 m2m framework? On Wed, Mar 19, 2014 at 12:22 AM, Hans Verkuil <hverkuil@xxxxxxxxx> wrote: > On 03/19/2014 12:18 AM, m silverstri wrote: >> I am using v4l2 m2m framework to develop a resize driver. I have an >> image , pass it to the driver and it generated a resize output image. >> >> My v4l2 sequence is >> 1. qbuf OUTPUT, CAPTURE >> 2. stream on OUTPUT, CAPTURE >> 3. dqbuf OUTPUT, CAPTURE >> 4. stream off OUTPUT, CAPTURE >> >> this works if i have a full frame of image before i start streaming. >> >> But what I only have partial buffers when I start streaming, how can I >> qbuf more buffer after I 'stream on' OUTPUT, > > You can't. Each buffer passed to the driver must contain a full image. > So your application needs some logic to keep filling a buffer until it > is complete and only then do you issue the QBUF. > > Regards, > > Hans > >> >> I try this, but this fail >> 1. qbuf OUTPUT, CAPTURE (I qbuf only partial OUTPUT) >> 2. stream on OUTPUT, CAPTURE >> >> // do this in a loop: >> 3. dqbuf OUTPUT (I want to queue more OUTPUT as they become available) >> 4. qbuf OUTPUT >> >> // now I am done, I want to dqbuf my output >> 5. dqbuf CAPTURE >> 6. stream off OUTPUT, CAPTURE >> >> I try to do dqbuf/qbuf OUTPUT in step #3, #4 above, but it just stuck >> in dqbuf OUTPUT. >> >> How can I queue more of my input data after I stream on? >> >> Thank you. >> -- >> To unsubscribe from this list: send the line "unsubscribe linux-media" in >> the body of a message to majordomo@xxxxxxxxxxxxxxx >> More majordomo info at http://vger.kernel.org/majordomo-info.html >> > -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
