Hi Hans, On Wed, Jun 22, 2022 at 11:31:43AM +0200, Hans Verkuil wrote: > Add support for two new (un)prepare_streaming queue ops that are called > when the user calls VIDIOC_STREAMON and STREAMOFF (or closes the fh). > > This gives drivers a callback to validate the video pipeline and claim > or release streaming resources at the time that userspace indicates > that it wants to start streaming (VIDIOC_STREAMON) rather than when > the actual DMA engine kicks in (the start_streaming callback). This > is relevant for drivers that needs a minimum of X buffers before the > DMA can start. The actual pipeline validation needs to happen sooner > to avoid unnecessary errors at VIDIOC_QBUF time. > > As a bonus this allows us to move the horrible call to > v4l_vb2q_enable_media_source() in vb2_core_streamon() to the au0828 > driver (currently the only driver that uses this feature). Can we drop the horrible .enable_source() from media_device too ? :-) > That call never belonged in vb2, but it had the same purpose as the > new prepare_streaming op: validate the pipeline. > > This is a follow-up from my previous RFCv2: > > https://patchwork.linuxtv.org/project/linux-media/patch/ba4bca14-488f-94ea-48d9-e7343103c5aa@xxxxxxxxx/ > > I would like to get consensus for this series. I don't like it much. vb2 is already doing too much in my opinion, growing it isn't the right way to fix it. I'm still working on a new version of the V4L2 streams series that may conflict with this, I'd propose discussing the two together. > Hans Verkuil (2): > vb2: add (un)prepare_streaming queue ops > vb2/au0828: move the v4l_vb2q_enable_media_source to the au0828 driver > > .../media/common/videobuf2/videobuf2-core.c | 26 ++++++++++++++----- > drivers/media/usb/au0828/au0828-vbi.c | 2 ++ > drivers/media/usb/au0828/au0828-video.c | 1 + > include/media/videobuf2-core.h | 14 ++++++++++ > 4 files changed, 37 insertions(+), 6 deletions(-) -- Regards, Laurent Pinchart
