[libcamera-devel] [PATCH 1/2] libcamera: v4l2_videodevice: Normalise multiPlanar configuration

[Laurent Pinchart]

Hi Kieran,

On Mon, Oct 21, 2019 at 11:40:24AM +0100, Kieran Bingham wrote:
> On 20/10/2019 06:23, Laurent Pinchart wrote:
> > On Fri, Oct 18, 2019 at 04:42:00PM +0100, Kieran Bingham wrote:
> >> The use of either single planar or multi planar API is constant
> >> throughout the usage of a device, based upon the buffer type of the
> >> stream, and once determined it does not change.
> >>
> >> Rather than interrogate the bufferType_ each time, set a device flag at
> >> open() time as to whether we are using planar or multi-planar buffer
> >> APIs.
> >>
> >> Signed-off-by: Kieran Bingham <kieran.bingham at ideasonboard.com>
> >> ---
> >>  src/libcamera/include/v4l2_videodevice.h |  1 +
> >>  src/libcamera/v4l2_videodevice.cpp       | 13 ++++++++-----
> >>  2 files changed, 9 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/src/libcamera/include/v4l2_videodevice.h b/src/libcamera/include/v4l2_videodevice.h
> >> index 4b8cf9394eb9..60fca01670c6 100644
> >> --- a/src/libcamera/include/v4l2_videodevice.h
> >> +++ b/src/libcamera/include/v4l2_videodevice.h
> >> @@ -182,6 +182,7 @@ private:
> >>  
> >>  	enum v4l2_buf_type bufferType_;
> >>  	enum v4l2_memory memoryType_;
> >> +	bool multiPlanar_;
> > 
> > I'm not really opposed to this, but does this change really add value to
> > offset for the larger memory usage ? That's of course a bit of a
> > rhetorical question as one bool isn't much :-)
> 
> Yes, I wasn't too worried about the single bool.
> 
> > If we decide to move forward, how should the method below react when
> > passing a buf.type that doesn't match multiPlanar_ ? Currently the
> > kernel will return an error, is it a good idea to ignore the issue
> > silently ?
> 
> Ugh, I hadn't considered that someone could give us a single planar
> buffer if we are configured in multiplanar.
> 
> That would indeed indicate someone had screwed up - so we shouldn't
> leave it to go silent.

Or we could improve the API by making the error impossible to trigger,
but that would require creating a custom structure that duplicates
v4l2_buffer without the type. I don't think that's worth it.

Another option is to ignore buf.type everywhere, using bufferType_
unconditionally, and remove all code that sets it manually.

> My main reason for this was that I later add another point of having to
> split code path for mplane/splane - and I felt having a single variable
> would make this clearer/simplified.
> 
> >>  	BufferPool *bufferPool_;
> >>  	std::map<unsigned int, Buffer *> queuedBuffers_;
> >> diff --git a/src/libcamera/v4l2_videodevice.cpp b/src/libcamera/v4l2_videodevice.cpp
> >> index 208ab54199b1..37f61b90ac46 100644
> >> --- a/src/libcamera/v4l2_videodevice.cpp
> >> +++ b/src/libcamera/v4l2_videodevice.cpp
> >> @@ -350,6 +350,8 @@ int V4L2VideoDevice::open()
> >>  		return -EINVAL;
> >>  	}
> >>  
> >> +	multiPlanar_ = V4L2_TYPE_IS_MULTIPLANAR(bufferType_);
> >> +
> >>  	fdEvent_->activated.connect(this, &V4L2VideoDevice::bufferAvailable);
> >>  	fdEvent_->setEnabled(false);
> >>  
> >> @@ -436,6 +438,8 @@ int V4L2VideoDevice::open(int handle, enum v4l2_buf_type type)
> >>  		return -EINVAL;
> >>  	}
> >>  
> >> +	multiPlanar_ = V4L2_TYPE_IS_MULTIPLANAR(bufferType_);
> >> +
> >>  	fdEvent_->activated.connect(this, &V4L2VideoDevice::bufferAvailable);
> >>  	fdEvent_->setEnabled(false);
> >>  
> >> @@ -870,7 +874,7 @@ int V4L2VideoDevice::exportBuffers(BufferPool *pool)
> >>  			break;
> >>  		}
> >>  
> >> -		if (V4L2_TYPE_IS_MULTIPLANAR(buf.type)) {
> >> +		if (multiPlanar_) {
> >>  			for (unsigned int p = 0; p < buf.length; ++p) {
> >>  				ret = createPlane(&buffer, i, p,
> >>  						  buf.m.planes[p].length);
> >> @@ -993,12 +997,11 @@ int V4L2VideoDevice::queueBuffer(Buffer *buffer)
> >>  	buf.memory = memoryType_;
> >>  	buf.field = V4L2_FIELD_NONE;
> >>  
> >> -	bool multiPlanar = V4L2_TYPE_IS_MULTIPLANAR(buf.type);
> >>  	BufferMemory *mem = &bufferPool_->buffers()[buf.index];
> >>  	const std::vector<Plane> &planes = mem->planes();
> >>  
> >>  	if (buf.memory == V4L2_MEMORY_DMABUF) {
> >> -		if (multiPlanar) {
> >> +		if (multiPlanar_) {
> >>  			for (unsigned int p = 0; p < planes.size(); ++p)
> >>  				v4l2Planes[p].m.fd = planes[p].dmabuf();
> >>  		} else {
> >> @@ -1006,7 +1009,7 @@ int V4L2VideoDevice::queueBuffer(Buffer *buffer)
> >>  		}
> >>  	}
> >>  
> >> -	if (multiPlanar) {
> >> +	if (multiPlanar_) {
> >>  		buf.length = planes.size();
> >>  		buf.m.planes = v4l2Planes;
> >>  	}
> >> @@ -1099,7 +1102,7 @@ Buffer *V4L2VideoDevice::dequeueBuffer()
> >>  	buf.type = bufferType_;
> >>  	buf.memory = memoryType_;
> >>  
> >> -	if (V4L2_TYPE_IS_MULTIPLANAR(buf.type)) {
> >> +	if (multiPlanar_) {
> >>  		buf.length = VIDEO_MAX_PLANES;
> >>  		buf.m.planes = planes;
> >>  	}

-- 
Regards,

Laurent Pinchart