On Wed, 29 Jun 2011 16:10:45 +0200 Marek Szyprowski <m.szyprowski@xxxxxxxxxxx> wrote: > > I do still wonder why this is an issue - why not pass the buffers through > > to the driver at VIDIOC_QBUF time? I assume there must be a reason for > > doing things this way, I'd like to understand what it is. > > I want to delay giving the ownership of the buffers to the driver until it > is certain that start_streaming method will be called. This way I achieve > a well defined states of the queued buffers: > > 1. successful start_streaming() -> the driver is processing the queue buffers > 2. unsuccessful start_streaming() -> the driver is responsible to discard all > queued buffers > 3. stop_streaming() called -> the driver has finished or discarded all queued > buffers So it's a buffer ownership thing. I wonder if there would be value in adding a buf_give_them_all_back_now() callback? You have an implicit change of buffer ownership now that seems easy for drivers to mess up. It might be better to send an explicit signal at such times and, perhaps, even require the driver to explicitly hand each buffer back to vb2? That would make the rules clear and give some flexibility - stopping and starting streaming without needing to start over with buffers, for example. Dunno, I'm just sort of babbling as I think; what's there now clearly works. Thanks, jon -- 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
