[libcamera-devel] [PATCH v3 4/7] libcamera: Support passing ColorSpaces to V4L2 video devices
David Plowman
david.plowman at raspberrypi.com
Mon Oct 25 11:54:13 CEST 2021
Hi Naush
Thanks for asking another good question!
On Mon, 25 Oct 2021 at 09:20, Naushir Patuck <naush at raspberrypi.com> wrote:
>
> Hi David,
>
> Thank you for your work.
>
> On Wed, 20 Oct 2021 at 12:08, David Plowman <david.plowman at raspberrypi.com> wrote:
>>
>> The ColorSpace from the StreamConfiguration is now handled
>> appropriately in the V4L2VideoDevice.
>>
>> Signed-off-by: David Plowman <david.plowman at raspberrypi.com>
>>
>> ---
>> include/libcamera/internal/v4l2_videodevice.h | 2 +
>> src/libcamera/v4l2_videodevice.cpp | 69 +++++++++++++++++--
>> 2 files changed, 66 insertions(+), 5 deletions(-)
>>
>> diff --git a/include/libcamera/internal/v4l2_videodevice.h b/include/libcamera/internal/v4l2_videodevice.h
>> index efe34d47..34cc9cdd 100644
>> --- a/include/libcamera/internal/v4l2_videodevice.h
>> +++ b/include/libcamera/internal/v4l2_videodevice.h
>> @@ -20,6 +20,7 @@
>> #include <libcamera/base/log.h>
>> #include <libcamera/base/signal.h>
>>
>> +#include <libcamera/color_space.h>
>> #include <libcamera/framebuffer.h>
>> #include <libcamera/geometry.h>
>> #include <libcamera/pixel_format.h>
>> @@ -163,6 +164,7 @@ public:
>>
>> V4L2PixelFormat fourcc;
>> Size size;
>> + ColorSpace colorSpace;
>>
>> std::array<Plane, 3> planes;
>> unsigned int planesCount = 0;
>> diff --git a/src/libcamera/v4l2_videodevice.cpp b/src/libcamera/v4l2_videodevice.cpp
>> index ba5f88cd..9e57e257 100644
>> --- a/src/libcamera/v4l2_videodevice.cpp
>> +++ b/src/libcamera/v4l2_videodevice.cpp
>> @@ -364,11 +364,6 @@ bool V4L2BufferCache::Entry::operator==(const FrameBuffer &buffer) const
>> * \brief The plane line stride (in bytes)
>> */
>>
>> -/**
>> - * \var V4L2DeviceFormat::size
>> - * \brief The image size in pixels
>> - */
>> -
>> /**
>> * \var V4L2DeviceFormat::fourcc
>> * \brief The fourcc code describing the pixel encoding scheme
>> @@ -377,6 +372,23 @@ bool V4L2BufferCache::Entry::operator==(const FrameBuffer &buffer) const
>> * that identifies the image format pixel encoding scheme.
>> */
>>
>> +/**
>> + * \var V4L2DeviceFormat::size
>> + * \brief The image size in pixels
>> + */
>> +
>> +/**
>> + * \var V4L2DeviceFormat::colorSpace
>> + * \brief The color space of the pixels
>> + *
>> + * When setting or trying a format, passing in "Undefined" fields in the
>> + * ColorSpace is not permitted because the driver will then make an
>> + * arbitrary choice of its own. Choices made by the driver will be
>> + * passed back in the normal way, though note that "Undefined" values can
>> + * be returned if the device has chosen something that the ColorSpace
>> + * class cannot represent.
>> + */
>> +
>> /**
>> * \var V4L2DeviceFormat::planes
>> * \brief The per-plane memory size information
>> @@ -873,6 +885,12 @@ int V4L2VideoDevice::getFormatMultiplane(V4L2DeviceFormat *format)
>> format->fourcc = V4L2PixelFormat(pix->pixelformat);
>> format->planesCount = pix->num_planes;
>>
>> + format->colorSpace = v4l2ToColorSpace(*pix);
>> + if (!format->colorSpace.isFullyDefined())
>> + LOG(V4L2, Warning)
>> + << "Retrieved undefined color space: "
>> + << format->colorSpace.toString();
>> +
>> for (unsigned int i = 0; i < format->planesCount; ++i) {
>> format->planes[i].bpl = pix->plane_fmt[i].bytesperline;
>> format->planes[i].size = pix->plane_fmt[i].sizeimage;
>> @@ -887,6 +905,11 @@ int V4L2VideoDevice::trySetFormatMultiplane(V4L2DeviceFormat *format, bool set)
>> struct v4l2_pix_format_mplane *pix = &v4l2Format.fmt.pix_mp;
>> int ret;
>>
>> + if (!format->colorSpace.isFullyDefined())
>> + LOG(V4L2, Error)
>> + << "Trying to set undefined color space: "
>> + << format->colorSpace.toString();
>> +
>
>
> Should we return with an error code in the above if clause? Similar for the other setFormat() calls.
Good point. I was slightly nervous here about breaking pipeline
handlers that simply don't handle colour spaces yet. One might worry
about the ipu3/rkisp1 pipeline handlers but it might be even worse
than that. For example, if you bisected your tree and unluckily got
this as your top commit then nothing at all would run.
Perhaps there needs to be a further commit after all pipeline handlers
have been upgraded so that they never pass anything undefined? At that
point, failure here would be fine. Thoughts welcome on that!
Thanks
David
> Other than that:
>
> Reviewed-by: Naushir Patuck <naush at raspberrypi.com>
>
>
>>
>> v4l2Format.type = bufferType_;
>> pix->width = format->size.width;
>> pix->height = format->size.height;
>> @@ -894,6 +917,12 @@ int V4L2VideoDevice::trySetFormatMultiplane(V4L2DeviceFormat *format, bool set)
>> pix->num_planes = format->planesCount;
>> pix->field = V4L2_FIELD_NONE;
>>
>> + ret = colorSpaceToV4l2(format->colorSpace, *pix);
>> + if (ret < 0)
>> + LOG(V4L2, Warning)
>> + << "Setting color space unrecognised by V4L2: "
>> + << format->colorSpace.toString();
>> +
>> ASSERT(pix->num_planes <= std::size(pix->plane_fmt));
>>
>> for (unsigned int i = 0; i < pix->num_planes; ++i) {
>> @@ -922,6 +951,12 @@ int V4L2VideoDevice::trySetFormatMultiplane(V4L2DeviceFormat *format, bool set)
>> format->planes[i].size = pix->plane_fmt[i].sizeimage;
>> }
>>
>> + format->colorSpace = v4l2ToColorSpace(*pix);
>> + if (!format->colorSpace.isFullyDefined())
>> + LOG(V4L2, Warning)
>> + << "Undefined color space has been set: "
>> + << format->colorSpace.toString();
>> +
>> return 0;
>> }
>>
>> @@ -945,6 +980,12 @@ int V4L2VideoDevice::getFormatSingleplane(V4L2DeviceFormat *format)
>> format->planes[0].bpl = pix->bytesperline;
>> format->planes[0].size = pix->sizeimage;
>>
>> + format->colorSpace = v4l2ToColorSpace(*pix);
>> + if (!format->colorSpace.isFullyDefined())
>> + LOG(V4L2, Warning)
>> + << "Retrieved undefined color space: "
>> + << format->colorSpace.toString();
>> +
>> return 0;
>> }
>>
>> @@ -954,12 +995,24 @@ int V4L2VideoDevice::trySetFormatSingleplane(V4L2DeviceFormat *format, bool set)
>> struct v4l2_pix_format *pix = &v4l2Format.fmt.pix;
>> int ret;
>>
>> + if (!format->colorSpace.isFullyDefined())
>> + LOG(V4L2, Error)
>> + << "Trying to set undefined color space: "
>> + << format->colorSpace.toString();
>> +
>> v4l2Format.type = bufferType_;
>> pix->width = format->size.width;
>> pix->height = format->size.height;
>> pix->pixelformat = format->fourcc;
>> pix->bytesperline = format->planes[0].bpl;
>> pix->field = V4L2_FIELD_NONE;
>> +
>> + ret = colorSpaceToV4l2(format->colorSpace, *pix);
>> + if (ret < 0)
>> + LOG(V4L2, Warning)
>> + << "Set color space unrecognised by V4L2: "
>> + << format->colorSpace.toString();
>> +
>> ret = ioctl(set ? VIDIOC_S_FMT : VIDIOC_TRY_FMT, &v4l2Format);
>> if (ret) {
>> LOG(V4L2, Error)
>> @@ -979,6 +1032,12 @@ int V4L2VideoDevice::trySetFormatSingleplane(V4L2DeviceFormat *format, bool set)
>> format->planes[0].bpl = pix->bytesperline;
>> format->planes[0].size = pix->sizeimage;
>>
>> + format->colorSpace = v4l2ToColorSpace(*pix);
>> + if (!format->colorSpace.isFullyDefined())
>> + LOG(V4L2, Warning)
>> + << "Undefined color space has been set: "
>> + << format->colorSpace.toString();
>> +
>> return 0;
>> }
>>
>> --
>> 2.20.1
>>
More information about the libcamera-devel
mailing list