Re: [PATCH v5 2/7] media: v4l2: Add extended buffer operations

[Tomasz Figa <tfiga@chromium.org>]

On Tue, Nov 24, 2020 at 5:33 AM Helen Koike <helen.koike@collabora.com> wrote:
>
> Hi Tomasz,
>
>
> On 11/20/20 8:14 AM, Tomasz Figa wrote:
> > Hi Helen,
> >
> > On Tue, Aug 04, 2020 at 04:29:34PM -0300, Helen Koike wrote:
> >> From: Hans Verkuil <hans.verkuil@cisco.com>
> >>
> >> Those extended buffer ops have several purpose:
> >> 1/ Fix y2038 issues by converting the timestamp into an u64 counting
> >>    the number of ns elapsed since 1970
> >> 2/ Unify single/multiplanar handling
> >> 3/ Add a new start offset field to each v4l2 plane buffer info struct
> >>    to support the case where a single buffer object is storing all
> >>    planes data, each one being placed at a different offset
> >>
> >> New hooks are created in v4l2_ioctl_ops so that drivers can start using
> >> these new objects.
> >>
> >> The core takes care of converting new ioctls requests to old ones
> >> if the driver does not support the new hooks, and vice versa.
> >>
> >> Note that the timecode field is gone, since there doesn't seem to be
> >> in-kernel users. We can be added back in the reserved area if needed or
> >> use the Request API to collect more metadata information from the
> >> frame.
> >>
> >
> > Thanks for the patch. Please see my comments inline.
>
> Thank you for your detailed review, please see my comments below.
>
> >
> >> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
> >> Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
> >> Signed-off-by: Helen Koike <helen.koike@collabora.com>
> >> ---
> >> Changes in v5:
> >> - migrate memory from v4l2_ext_buffer to v4l2_ext_plane
> >> - return mem_offset to struct v4l2_ext_plane
> >> - change sizes and reorder fields to avoid holes in the struct and make
> >>   it the same for 32 and 64 bits
> >>
> >> Changes in v4:
> >> - Use v4l2_ext_pix_format directly in the ioctl, drop v4l2_ext_format,
> >> making V4L2_BUF_TYPE_VIDEO_[OUTPUT,CAPTURE] the only valid types.
> >> - Drop VIDIOC_EXT_EXPBUF, since the only difference from VIDIOC_EXPBUF
> >> was that with VIDIOC_EXT_EXPBUF we could export multiple planes at once.
> >> I think we can add this later, so I removed it from this RFC to simplify it.
> >> - Remove num_planes field from struct v4l2_ext_buffer
> >> - Add flags field to struct v4l2_ext_create_buffers
> >> - Reformulate struct v4l2_ext_plane
> >> - Fix some bugs caught by v4l2-compliance
> >> - Rebased on top of media/master (post 5.8-rc1)
> >>
> >> Changes in v3:
> >> - Rebased on top of media/master (post 5.4-rc1)
> >>
> >> Changes in v2:
> >> - Add reserved space to v4l2_ext_buffer so that new fields can be added
> >>   later on
> >> ---
> >>  drivers/media/v4l2-core/v4l2-dev.c   |  29 ++-
> >>  drivers/media/v4l2-core/v4l2-ioctl.c | 353 +++++++++++++++++++++++++--
> >>  include/media/v4l2-ioctl.h           |  26 ++
> >>  include/uapi/linux/videodev2.h       |  90 +++++++
> >>  4 files changed, 476 insertions(+), 22 deletions(-)
> >>
> >> diff --git a/drivers/media/v4l2-core/v4l2-dev.c b/drivers/media/v4l2-core/v4l2-dev.c
> >> index e1829906bc086..cb21ee8eb075c 100644
> >> --- a/drivers/media/v4l2-core/v4l2-dev.c
> >> +++ b/drivers/media/v4l2-core/v4l2-dev.c
> >> @@ -720,15 +720,34 @@ static void determine_valid_ioctls(struct video_device *vdev)
> >>              SET_VALID_IOCTL(ops, VIDIOC_TRY_FMT, vidioc_try_fmt_sdr_out);
> >>      }
> >>
> >> +    if (is_vid || is_tch) {
> >> +            /* ioctls valid for video and touch */
> >> +            if (ops->vidioc_querybuf || ops->vidioc_ext_querybuf)
> >> +                    set_bit(_IOC_NR(VIDIOC_EXT_QUERYBUF), valid_ioctls);
> >> +            if (ops->vidioc_qbuf || ops->vidioc_ext_qbuf)
> >> +                    set_bit(_IOC_NR(VIDIOC_EXT_QBUF), valid_ioctls);
> >> +            if (ops->vidioc_dqbuf || ops->vidioc_ext_dqbuf)
> >> +                    set_bit(_IOC_NR(VIDIOC_EXT_DQBUF), valid_ioctls);
> >> +            if (ops->vidioc_create_bufs || ops->vidioc_ext_create_bufs)
> >> +                    set_bit(_IOC_NR(VIDIOC_EXT_CREATE_BUFS), valid_ioctls);
> >> +            if (ops->vidioc_prepare_buf || ops->vidioc_ext_prepare_buf)
> >> +                    set_bit(_IOC_NR(VIDIOC_EXT_PREPARE_BUF), valid_ioctls);
> >
> > nit: Could we stick to the SET_VALID_IOCTL() macro and just call it twice,
> > once for the new and once for the legacy callback?
>
> Ack.
>
> >
> >> +    }
> >> +
> >>      if (is_vid || is_vbi || is_sdr || is_tch || is_meta) {
> >>              /* ioctls valid for video, vbi, sdr, touch and metadata */
> >>              SET_VALID_IOCTL(ops, VIDIOC_REQBUFS, vidioc_reqbufs);
> >> -            SET_VALID_IOCTL(ops, VIDIOC_QUERYBUF, vidioc_querybuf);
> >> -            SET_VALID_IOCTL(ops, VIDIOC_QBUF, vidioc_qbuf);
> >>              SET_VALID_IOCTL(ops, VIDIOC_EXPBUF, vidioc_expbuf);
> >> -            SET_VALID_IOCTL(ops, VIDIOC_DQBUF, vidioc_dqbuf);
> >> -            SET_VALID_IOCTL(ops, VIDIOC_CREATE_BUFS, vidioc_create_bufs);
> >> -            SET_VALID_IOCTL(ops, VIDIOC_PREPARE_BUF, vidioc_prepare_buf);
> >> +            if (ops->vidioc_querybuf || ops->vidioc_ext_querybuf)
> >> +                    set_bit(_IOC_NR(VIDIOC_QUERYBUF), valid_ioctls);
> >> +            if (ops->vidioc_qbuf || ops->vidioc_ext_qbuf)
> >> +                    set_bit(_IOC_NR(VIDIOC_QBUF), valid_ioctls);
> >> +            if (ops->vidioc_dqbuf || ops->vidioc_ext_dqbuf)
> >> +                    set_bit(_IOC_NR(VIDIOC_DQBUF), valid_ioctls);
> >> +            if (ops->vidioc_create_bufs || ops->vidioc_ext_create_bufs)
> >> +                    set_bit(_IOC_NR(VIDIOC_CREATE_BUFS), valid_ioctls);
> >> +            if (ops->vidioc_prepare_buf || ops->vidioc_ext_prepare_buf)
> >> +                    set_bit(_IOC_NR(VIDIOC_PREPARE_BUF), valid_ioctls);
> >
> > Is it valid to check the new callbacks for devices that the new API is not
> > valid for (e.g. vbi)? Perhaps we could call SET_VALID_IOCTL(ops, <ioctl>,
> > vidioc_ext_*) in the upper if added in this patch and keep the code above
> > as is?
>
> Just to be clear, the only valid type should be VFL_TYPE_VIDEO right?
> Ext but API won't support touch devices for instance, right?
>

Yes, at least at this point. If one needs, it could be added in the
future, but honestly I don't see much use of the other types these
days.

>
> >
> >>              SET_VALID_IOCTL(ops, VIDIOC_STREAMON, vidioc_streamon);
> >>              SET_VALID_IOCTL(ops, VIDIOC_STREAMOFF, vidioc_streamoff);
> >>      }
> >> diff --git a/drivers/media/v4l2-core/v4l2-ioctl.c b/drivers/media/v4l2-core/v4l2-ioctl.c
> >> index 14a0def50f8ea..7ecdd9cc1bf48 100644
> >> --- a/drivers/media/v4l2-core/v4l2-ioctl.c
> >> +++ b/drivers/media/v4l2-core/v4l2-ioctl.c
> >> @@ -527,6 +527,26 @@ static void v4l_print_buffer(const void *arg, bool write_only)
> >>                      tc->type, tc->flags, tc->frames, *(__u32 *)tc->userbits);
> >>  }
> >>
> >> +static void v4l_print_ext_buffer(const void *arg, bool write_only)
> >> +{
> >> +    const struct v4l2_ext_buffer *e = arg;
> >> +    const struct v4l2_ext_plane *plane;
> >> +    unsigned int i;
> >> +
> >> +    pr_cont("%lld index=%d, type=%s, flags=0x%08x, field=%s, sequence=%d\n",
> >> +            e->timestamp, e->index, prt_names(e->type, v4l2_type_names),
> >> +            e->flags, prt_names(e->field, v4l2_field_names), e->sequence);
> >
> > Should we also print the request FD?
>
> Yes, I'll update this for next version.
>
> >
> >> +
> >> +    for (i = 0; i < VIDEO_MAX_PLANES &&
> >> +                e->planes[i].buffer_length; i++) {
> >> +            plane = &e->planes[i];
> >> +            pr_debug("plane %d: buffer_length=%d, plane_length=%d offset=0x%08x, memory=%s\n",
> >> +                     i, plane->buffer_length, plane->plane_length,
> >> +                     plane->offset,
> >> +                     prt_names(plane->memory, v4l2_memory_names));
> >
> > Should we also print mem_offset/userptr/dmabuf_fd?
>
> I see they are not printed by v4l_print_buffer(),

offset/userptr are printed in v4l2_print_buffer:
https://elixir.bootlin.com/linux/latest/source/drivers/media/v4l2-core/v4l2-ioctl.c#L494

> since these fields are in an
> union, the value of two of them will be invalid (I wonder if this can bring
> confusion).
> I also wondered if printing them can't cause a security issue.
>
> But I can add those prints if you think it make sense.
>

We know the memory type, so we can interpret the union appropriately
and adjust the message printed.

I don't think this poses any security issue, as it prints things that
belong to the userspace already and are only meaningful in the context
of the given userspace process.

> >
> >> +    }
> >> +}
> >> +
> >>  static void v4l_print_exportbuffer(const void *arg, bool write_only)
> >>  {
> >>      const struct v4l2_exportbuffer *p = arg;
> >> @@ -546,6 +566,15 @@ static void v4l_print_create_buffers(const void *arg, bool write_only)
> >>      v4l_print_format(&p->format, write_only);
> >>  }
> >>
> >> +static void v4l_print_ext_create_buffers(const void *arg, bool write_only)
> >> +{
> >> +    const struct v4l2_ext_create_buffers *p = arg;
> >> +
> >> +    pr_cont("index=%d, count=%d, memory=%s, ", p->index, p->count,
> >> +            prt_names(p->memory, v4l2_memory_names));
> >> +    v4l_print_ext_pix_format(&p->format, write_only);
> >
> > Should we also print capabilities and flags?
>
> I just saw these prints are called after the ioctl handler, and not before,
> to I guess it make sense.
>
> It is not printed by v4l_print_create_buffers(), I think we can add in both then.
>
> >
> >> +}
> >> +
> >>  static void v4l_print_streamparm(const void *arg, bool write_only)
> >>  {
> >>      const struct v4l2_streamparm *p = arg;
> >> @@ -1220,6 +1249,143 @@ int v4l2_format_to_ext_pix_format(const struct v4l2_format *f,
> >>  }
> >>  EXPORT_SYMBOL_GPL(v4l2_format_to_ext_pix_format);
> >>
> >> +/*
> >> + * If mplane_cap is true, b->m.planes should have a valid pointer of a
> >> + * struct v4l2_plane array, and b->length with its size
> >> + */
> >> +int v4l2_ext_buffer_to_buffer(const struct v4l2_ext_buffer *e,
> >> +                          struct v4l2_buffer *b, bool mplane_cap)
> >> +{
> >> +    unsigned int planes_array_size = b->length;
> >> +    struct v4l2_plane *planes = b->m.planes;
> >> +    u64 nsecs;
> >> +
> >> +    if (!mplane_cap && e->planes[1].buffer_length != 0)
> >> +            return -EINVAL;
> >> +
> >> +    memset(b, 0, sizeof(*b));
> >> +
> >> +    b->index = e->index;
> >> +    b->flags = e->flags;
> >> +    b->field = e->field;
> >> +    b->sequence = e->sequence;
> >> +    b->memory = e->planes[0].memory;
> >> +    b->request_fd = e->request_fd;
> >> +    b->timestamp.tv_sec = div64_u64_rem(e->timestamp, NSEC_PER_SEC, &nsecs);
> >> +    b->timestamp.tv_usec = (u32)nsecs / NSEC_PER_USEC;
> >> +
> >> +    if (mplane_cap) {
> >> +            unsigned int i;
> >> +
> >> +            if (!planes || !planes_array_size)
> >> +                    return -EINVAL;
> >> +
> >> +            b->m.planes = planes;
> >
> > planes was initialized to b->m.planes at declaration time. Should we
> > perhaps move its declaration to within this block to make it more clear and
> > remove this assignment?
>
> The variable "planes" is saving the pointer of b->m.planes before we do
> memset(b, 0, sizeof(*b)), so I can reassing it back if it make sense.
>
> I can add a comment to make this more clear.
>

Right, I think the code is a bit confusing. Should we just pass the
planes array pointer as a separate argument to the function?

> >
> >> +
> >> +            if (e->type == V4L2_BUF_TYPE_VIDEO_CAPTURE)
> >> +                    b->type = V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE;
> >> +            else
> >> +                    b->type = V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE;
> >> +
> >> +            for (i = 0; i < VIDEO_MAX_PLANES && i < planes_array_size &&
> >> +                        e->planes[i].buffer_length; i++) {
> >> +
> >> +                    if (e->planes[0].memory != e->planes[i].memory)
> >> +                            return -EINVAL;
> >> +
> >> +                    if (e->planes[i].offset)
> >> +                            return -EINVAL;
> >
> > Is it really invalid to have a non-zero offset? Shouldn't the data_offset
> > field of the legacy struct be populated instead, in the cases where it was
> > defined to be valid?
>
> My understanding of data_offset, is that it is used when the hardware can
> write/read a header to/from the buffer.
>
> But this doesn't seem to be used by any driver

This is not true:
https://elixir.bootlin.com/linux/latest/source/drivers/media/platform/qcom/venus/venc.c#L963

>, so there is an attempt to
> repourpose it here:
>
>     https://patchwork.linuxtv.org/project/linux-media/patch/1429040689-23808-2-git-send-email-laurent.pinchart@ideasonboard.com/
>
> But this wans't merged.
>
> So in the current API, there is no way to specify an offset in the buffer.
>
> I guess we can repurpose data_offset first, what do you think?
>

We need to stick to the original behavior for data_offset, so that an
encoder has a way to tell the userspace how many bytes of the header
it needs to skip.

> >
> >> +
> >> +                    memset(&b->m.planes[i], 0, sizeof(b->m.planes[i]));
> >> +
> >> +                    if (b->memory == V4L2_MEMORY_MMAP)
> >> +                            b->m.planes[i].m.mem_offset = e->planes[i].m.mem_offset;
> >> +                    else if (b->memory == V4L2_MEMORY_DMABUF)
> >> +                            b->m.planes[i].m.fd = e->planes[i].m.dmabuf_fd;
> >> +                    else
> >> +                            b->m.planes[i].m.userptr = e->planes[i].m.userptr;
> >> +
> >> +                    b->m.planes[i].bytesused = e->planes[i].plane_length;
> >
> > I might be getting the meaning of plane_length wrong, but doesn't this
> > depend on the direction? If the userspace gives a CAPTURE buffer, it would
> > have bytesused = 0, but if the kernel returns it, it would have bytesused =
> > <size of the payload>.
>
> You are right, it depends on the direction, thanks for catching this.
>
> Also, in the Ext api, we don't have the <size of the payload>, so I'm using the
> plane_length instead, I'm not sure if there is a better way.

I thought about this for a while and it sounds like in the code added
by this series, plane_length is basically used as the old bytesused
and buffer_length as the old length. Would it make sense to just
preserve the old naming?

>
> >
> >> +                    b->m.planes[i].length = e->planes[i].buffer_length;
> >> +            }
> >> +            /* In multi-planar, length contain the number of planes */
> >> +            b->length = i;
> >> +    } else {
> >> +            b->type = e->type;
> >> +            b->bytesused = e->planes[0].plane_length;
> >> +            b->length = e->planes[0].buffer_length;
> >> +
> >> +            if (e->planes[0].offset)
> >> +                    return -EINVAL;
> >
> > Ditto.
>
> Ack.
>
> >
> >> +
> >> +            if (b->memory == V4L2_MEMORY_MMAP)
> >> +                    b->m.offset = e->planes[0].m.mem_offset;
> >> +            else if (b->memory == V4L2_MEMORY_DMABUF)
> >> +                    b->m.fd = e->planes[0].m.dmabuf_fd;
> >> +            else
> >> +                    b->m.userptr = e->planes[0].m.userptr;
> >> +    }
> >> +
> >> +    return 0;
> >> +}
> >> +EXPORT_SYMBOL_GPL(v4l2_ext_buffer_to_buffer);
> >> +
> >> +int v4l2_buffer_to_ext_buffer(const struct v4l2_buffer *b,
> >> +                          struct v4l2_ext_buffer *e)
> >> +{
> >> +    memset(e, 0, sizeof(*e));
> >> +
> >> +    e->index = b->index;
> >> +    e->flags = b->flags;
> >> +    e->field = b->field;
> >> +    e->sequence = b->sequence;
> >> +    e->request_fd = b->request_fd;
> >> +    e->timestamp = b->timestamp.tv_sec * NSEC_PER_SEC +
> >> +            b->timestamp.tv_usec * NSEC_PER_USEC;
> >> +    if (V4L2_TYPE_IS_MULTIPLANAR(b->type)) {
> >> +            unsigned int i;
> >> +
> >> +            if (!b->m.planes)
> >> +                    return -EINVAL;
> >> +
> >> +            if (b->type == V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE)
> >> +                    e->type = V4L2_BUF_TYPE_VIDEO_CAPTURE;
> >> +            else
> >> +                    e->type = V4L2_BUF_TYPE_VIDEO_OUTPUT;
> >> +
> >> +            /* In multi-planar, length contain the number of planes */
> >> +            for (i = 0; i < b->length; i++) {
> >
> > The design of the new struct implies that the planes describe color planes
> > and not memory planes, so this code is incorrect, because for non-M format
> > variants it would fill in only the first plane of the new struct.
>
> Yes, as discussed in 1/7, the handling of planes is wrong, I'll correct
> this for next version.
>
> >
> >> +                    if (b->memory == V4L2_MEMORY_MMAP)
> >> +                            e->planes[i].m.mem_offset = b->m.planes[i].m.mem_offset;
> >> +                    else if (b->memory == V4L2_MEMORY_DMABUF)
> >> +                            e->planes[i].m.dmabuf_fd = b->m.planes[i].m.fd;
> >> +                    else
> >> +                            e->planes[i].m.userptr = b->m.planes[i].m.userptr;
> >> +
> >> +                    e->planes[i].memory = b->memory;
> >> +                    e->planes[i].buffer_length = b->m.planes[i].length;
> >> +                    e->planes[i].plane_length = b->m.planes[i].bytesused;
> >> +                    if (b->m.planes[i].data_offset)
> >> +                            pr_warn("Ignoring data_offset value %d\n",
> >> +                                    b->m.planes[i].data_offset);
> >
> > Why? As per my comment above, there are valid use cases defined in the spec.
>
> Please see my comment about about data_offset.
>
> >
> >> +            }
> >> +    } else {
> >> +            e->type = b->type;
> >> +            e->planes[0].memory = b->memory;
> >> +            e->planes[0].plane_length = b->bytesused;
> >> +            e->planes[0].buffer_length = b->length;
> >> +            if (b->memory == V4L2_MEMORY_MMAP)
> >> +                    e->planes[0].m.mem_offset = b->m.offset;
> >> +            else if (b->memory == V4L2_MEMORY_DMABUF)
> >> +                    e->planes[0].m.dmabuf_fd = b->m.fd;
> >> +            else
> >> +                    e->planes[0].m.userptr = b->m.userptr;
> >
> > Similar to the MULTIPLANAR case, we should fill in the planes[] entries
> > corresponding to the number of color planes of the format, e.g. 2 for NV12.
>
> Ack.
>
> >
> >> +    }
> >> +
> >> +    return 0;
> >> +}
> >> +EXPORT_SYMBOL_GPL(v4l2_buffer_to_ext_buffer);
> >> +
> >>  static int v4l_querycap(const struct v4l2_ioctl_ops *ops,
> >>                              struct file *file, void *fh, void *arg)
> >>  {
> >> @@ -2473,31 +2639,112 @@ static int v4l_reqbufs(const struct v4l2_ioctl_ops *ops,
> >>      return ops->vidioc_reqbufs(file, fh, p);
> >>  }
> >>
> >> +static int v4l_do_buf_op(int (*op)(struct file *, void *,
> >> +                               struct v4l2_buffer *),
> >> +                     int (*ext_op)(struct file *, void *,
> >> +                                   struct v4l2_ext_buffer *),
> >> +                     struct file *file, void *fh, struct v4l2_buffer *b)
> >> +{
> >> +    struct v4l2_ext_buffer e;
> >> +    int ret;
> >> +
> >> +    ret = check_fmt(file, b->type);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    if (op)
> >> +            return op(file, fh, b);
> >> +
> >> +    ret = v4l2_buffer_to_ext_buffer(b, &e);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    ret = ext_op(file, fh, &e);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    v4l2_ext_buffer_to_buffer(&e, b, V4L2_TYPE_IS_MULTIPLANAR(b->type));
> >> +    return 0;
> >> +}
> >> +
> >> +static int v4l_do_ext_buf_op(int (*op)(struct file *, void *,
> >> +                                   struct v4l2_buffer *),
> >> +                         int (*ext_op)(struct file *, void *,
> >> +                                       struct v4l2_ext_buffer *),
> >> +                         struct file *file, void *fh,
> >> +                         struct v4l2_ext_buffer *e)
> >> +{
> >> +    struct video_device *vdev = video_devdata(file);
> >> +    struct v4l2_plane planes[VIDEO_MAX_PLANES];
> >> +    struct v4l2_buffer b;
> >> +    bool mplane_cap;
> >> +    int ret;
> >> +
> >> +    ret = check_fmt(file, e->type);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    if (ext_op)
> >> +            return ext_op(file, fh, e);
> >> +
> >> +    mplane_cap = !!(vdev->device_caps &
> >> +                    (V4L2_CAP_VIDEO_CAPTURE_MPLANE |
> >> +                     V4L2_CAP_VIDEO_OUTPUT_MPLANE |
> >> +                     V4L2_CAP_VIDEO_M2M_MPLANE));
> >> +    b.m.planes = planes;
> >> +    b.length = VIDEO_MAX_PLANES;
> >> +    ret = v4l2_ext_buffer_to_buffer(e, &b, mplane_cap);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    ret = op(file, fh, &b);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    v4l2_buffer_to_ext_buffer(&b, e);
> >> +    return 0;
> >> +}
> >> +
> >>  static int v4l_querybuf(const struct v4l2_ioctl_ops *ops,
> >> -                            struct file *file, void *fh, void *arg)
> >> +                    struct file *file, void *fh, void *arg)
> >>  {
> >> -    struct v4l2_buffer *p = arg;
> >> -    int ret = check_fmt(file, p->type);
> >> +    return v4l_do_buf_op(ops->vidioc_querybuf, ops->vidioc_ext_querybuf,
> >> +                         file, fh, arg);
> >> +}
> >>
> >> -    return ret ? ret : ops->vidioc_querybuf(file, fh, p);
> >> +static int v4l_ext_querybuf(const struct v4l2_ioctl_ops *ops,
> >> +                        struct file *file, void *fh, void *arg)
> >> +{
> >> +    return v4l_do_ext_buf_op(ops->vidioc_querybuf,
> >> +                             ops->vidioc_ext_querybuf, file, fh, arg);
> >>  }
> >>
> >>  static int v4l_qbuf(const struct v4l2_ioctl_ops *ops,
> >> -                            struct file *file, void *fh, void *arg)
> >> +                struct file *file, void *fh, void *arg)
> >>  {
> >> -    struct v4l2_buffer *p = arg;
> >> -    int ret = check_fmt(file, p->type);
> >> +    return v4l_do_buf_op(ops->vidioc_qbuf, ops->vidioc_ext_qbuf,
> >> +                         file, fh, arg);
> >> +}
> >>
> >> -    return ret ? ret : ops->vidioc_qbuf(file, fh, p);
> >> +static int v4l_ext_qbuf(const struct v4l2_ioctl_ops *ops,
> >> +                    struct file *file, void *fh, void *arg)
> >> +{
> >> +    return v4l_do_ext_buf_op(ops->vidioc_qbuf, ops->vidioc_ext_qbuf,
> >> +                             file, fh, arg);
> >>  }
> >>
> >>  static int v4l_dqbuf(const struct v4l2_ioctl_ops *ops,
> >> -                            struct file *file, void *fh, void *arg)
> >> +                 struct file *file, void *fh, void *arg)
> >>  {
> >> -    struct v4l2_buffer *p = arg;
> >> -    int ret = check_fmt(file, p->type);
> >> +    return v4l_do_buf_op(ops->vidioc_dqbuf, ops->vidioc_ext_dqbuf,
> >> +                         file, fh, arg);
> >> +}
> >>
> >> -    return ret ? ret : ops->vidioc_dqbuf(file, fh, p);
> >> +static int v4l_ext_dqbuf(const struct v4l2_ioctl_ops *ops,
> >> +                     struct file *file, void *fh, void *arg)
> >> +{
> >> +    return v4l_do_ext_buf_op(ops->vidioc_dqbuf, ops->vidioc_ext_dqbuf,
> >> +                             file, fh, arg);
> >>  }
> >>
> >>  static int v4l_create_bufs(const struct v4l2_ioctl_ops *ops,
> >> @@ -2513,7 +2760,27 @@ static int v4l_create_bufs(const struct v4l2_ioctl_ops *ops,
> >>
> >>      v4l_sanitize_format(&create->format);
> >>
> >> -    ret = ops->vidioc_create_bufs(file, fh, create);
> >> +    if (ops->vidioc_create_bufs) {
> >> +            ret = ops->vidioc_create_bufs(file, fh, create);
> >> +    } else {
> >> +            struct v4l2_ext_create_buffers ecreate = {
> >> +                    .count = create->count,
> >> +                    .memory = create->memory,
> >> +            };
> >> +
> >> +            ret = v4l2_format_to_ext_pix_format(&create->format,
> >> +                                                &ecreate.format, true);
> >> +            if (ret)
> >> +                    return ret;
> >> +
> >> +            ret = ops->vidioc_ext_create_bufs(file, fh, &ecreate);
> >> +            if (ret)
> >> +                    return ret;
> >> +
> >> +            create->index = ecreate.index;
> >> +            create->count = ecreate.count;
> >> +            create->capabilities = ecreate.capabilities;
> >> +    }
> >>
> >>      if (create->format.type == V4L2_BUF_TYPE_VIDEO_CAPTURE ||
> >>          create->format.type == V4L2_BUF_TYPE_VIDEO_OUTPUT)
> >> @@ -2522,13 +2789,60 @@ static int v4l_create_bufs(const struct v4l2_ioctl_ops *ops,
> >>      return ret;
> >>  }
> >>
> >> +static int v4l_ext_create_bufs(const struct v4l2_ioctl_ops *ops,
> >> +                           struct file *file, void *fh, void *arg)
> >> +{
> >> +    struct v4l2_ext_create_buffers *ecreate = arg;
> >> +    struct video_device *vdev = video_devdata(file);
> >> +    struct v4l2_create_buffers create = {
> >> +            .count = ecreate->count,
> >> +            .memory = ecreate->memory,
> >> +            .flags = ecreate->flags,
> >> +    };
> >> +    bool mplane_cap;
> >> +    int ret;
> >> +
> >> +    ret = check_fmt(file, ecreate->format.type);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    if (ops->vidioc_ext_create_bufs)
> >> +            return ops->vidioc_ext_create_bufs(file, fh, ecreate);
> >> +
> >> +    mplane_cap = !!(vdev->device_caps &
> >> +                    (V4L2_CAP_VIDEO_CAPTURE_MPLANE |
> >> +                     V4L2_CAP_VIDEO_OUTPUT_MPLANE |
> >> +                     V4L2_CAP_VIDEO_M2M_MPLANE));
> >> +    ret = v4l2_ext_pix_format_to_format(&ecreate->format,
> >> +                                        &create.format, mplane_cap, true);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    ret = v4l_create_bufs(ops, file, fh, &create);
> >> +    if (ret)
> >> +            return ret;
> >> +
> >> +    ecreate->index = create.index;
> >> +    ecreate->count = create.count;
> >> +    ecreate->capabilities = create.capabilities;
> >> +
> >> +    return 0;
> >> +}
> >> +
> >>  static int v4l_prepare_buf(const struct v4l2_ioctl_ops *ops,
> >> -                            struct file *file, void *fh, void *arg)
> >> +                       struct file *file, void *fh, void *arg)
> >>  {
> >> -    struct v4l2_buffer *b = arg;
> >> -    int ret = check_fmt(file, b->type);
> >> +    return v4l_do_buf_op(ops->vidioc_prepare_buf,
> >> +                         ops->vidioc_ext_prepare_buf,
> >> +                         file, fh, arg);
> >> +}
> >>
> >> -    return ret ? ret : ops->vidioc_prepare_buf(file, fh, b);
> >> +static int v4l_ext_prepare_buf(const struct v4l2_ioctl_ops *ops,
> >> +                           struct file *file, void *fh, void *arg)
> >> +{
> >> +    return v4l_do_ext_buf_op(ops->vidioc_prepare_buf,
> >> +                             ops->vidioc_ext_prepare_buf,
> >> +                             file, fh, arg);
> >>  }
> >>
> >>  static int v4l_g_parm(const struct v4l2_ioctl_ops *ops,
> >> @@ -3202,12 +3516,15 @@ static const struct v4l2_ioctl_info v4l2_ioctls[] = {
> >>      IOCTL_INFO(VIDIOC_S_EXT_PIX_FMT, v4l_s_ext_pix_fmt, v4l_print_ext_pix_format, INFO_FL_PRIO),
> >>      IOCTL_INFO(VIDIOC_REQBUFS, v4l_reqbufs, v4l_print_requestbuffers, INFO_FL_PRIO | INFO_FL_QUEUE),
> >>      IOCTL_INFO(VIDIOC_QUERYBUF, v4l_querybuf, v4l_print_buffer, INFO_FL_QUEUE | INFO_FL_CLEAR(v4l2_buffer, length)),
> >> +    IOCTL_INFO(VIDIOC_EXT_QUERYBUF, v4l_ext_querybuf, v4l_print_ext_buffer, INFO_FL_QUEUE | INFO_FL_CLEAR(v4l2_ext_buffer, planes)),
> >>      IOCTL_INFO(VIDIOC_G_FBUF, v4l_stub_g_fbuf, v4l_print_framebuffer, 0),
> >>      IOCTL_INFO(VIDIOC_S_FBUF, v4l_stub_s_fbuf, v4l_print_framebuffer, INFO_FL_PRIO),
> >>      IOCTL_INFO(VIDIOC_OVERLAY, v4l_overlay, v4l_print_u32, INFO_FL_PRIO),
> >>      IOCTL_INFO(VIDIOC_QBUF, v4l_qbuf, v4l_print_buffer, INFO_FL_QUEUE),
> >>      IOCTL_INFO(VIDIOC_EXPBUF, v4l_stub_expbuf, v4l_print_exportbuffer, INFO_FL_QUEUE | INFO_FL_CLEAR(v4l2_exportbuffer, flags)),
> >> +    IOCTL_INFO(VIDIOC_EXT_QBUF, v4l_ext_qbuf, v4l_print_ext_buffer, INFO_FL_QUEUE),
> >
> > Looking at the other entries, shouldn't this one be 1 line higher?
>
> Yes.
>
> >
> > That said, I wonder if it wouldn't look cleaner if we just put all the
> > EXT ioctls together at the bottom.
>
> I can move them and we can see if it is better or not.
>
> >
> >>      IOCTL_INFO(VIDIOC_DQBUF, v4l_dqbuf, v4l_print_buffer, INFO_FL_QUEUE),
> >> +    IOCTL_INFO(VIDIOC_EXT_DQBUF, v4l_ext_dqbuf, v4l_print_ext_buffer, INFO_FL_QUEUE),
> >>      IOCTL_INFO(VIDIOC_STREAMON, v4l_streamon, v4l_print_buftype, INFO_FL_PRIO | INFO_FL_QUEUE),
> >>      IOCTL_INFO(VIDIOC_STREAMOFF, v4l_streamoff, v4l_print_buftype, INFO_FL_PRIO | INFO_FL_QUEUE),
> >>      IOCTL_INFO(VIDIOC_G_PARM, v4l_g_parm, v4l_print_streamparm, INFO_FL_CLEAR(v4l2_streamparm, type)),
> >> @@ -3272,7 +3589,9 @@ static const struct v4l2_ioctl_info v4l2_ioctls[] = {
> >>      IOCTL_INFO(VIDIOC_SUBSCRIBE_EVENT, v4l_subscribe_event, v4l_print_event_subscription, 0),
> >>      IOCTL_INFO(VIDIOC_UNSUBSCRIBE_EVENT, v4l_unsubscribe_event, v4l_print_event_subscription, 0),
> >>      IOCTL_INFO(VIDIOC_CREATE_BUFS, v4l_create_bufs, v4l_print_create_buffers, INFO_FL_PRIO | INFO_FL_QUEUE),
> >> +    IOCTL_INFO(VIDIOC_EXT_CREATE_BUFS, v4l_ext_create_bufs, v4l_print_ext_create_buffers, INFO_FL_PRIO | INFO_FL_QUEUE),
> >>      IOCTL_INFO(VIDIOC_PREPARE_BUF, v4l_prepare_buf, v4l_print_buffer, INFO_FL_QUEUE),
> >> +    IOCTL_INFO(VIDIOC_EXT_PREPARE_BUF, v4l_ext_prepare_buf, v4l_print_ext_buffer, INFO_FL_QUEUE),
> >>      IOCTL_INFO(VIDIOC_ENUM_DV_TIMINGS, v4l_stub_enum_dv_timings, v4l_print_enum_dv_timings, INFO_FL_CLEAR(v4l2_enum_dv_timings, pad)),
> >>      IOCTL_INFO(VIDIOC_QUERY_DV_TIMINGS, v4l_stub_query_dv_timings, v4l_print_dv_timings, INFO_FL_ALWAYS_COPY),
> >>      IOCTL_INFO(VIDIOC_DV_TIMINGS_CAP, v4l_stub_dv_timings_cap, v4l_print_dv_timings_cap, INFO_FL_CLEAR(v4l2_dv_timings_cap, pad)),
> >> diff --git a/include/media/v4l2-ioctl.h b/include/media/v4l2-ioctl.h
> >> index 8bbcb74d8ee31..75996657ad1ba 100644
> >> --- a/include/media/v4l2-ioctl.h
> >> +++ b/include/media/v4l2-ioctl.h
> >> @@ -169,16 +169,26 @@ struct v4l2_fh;
> >>   *  :ref:`VIDIOC_REQBUFS <vidioc_reqbufs>` ioctl
> >>   * @vidioc_querybuf: pointer to the function that implements
> >>   *  :ref:`VIDIOC_QUERYBUF <vidioc_querybuf>` ioctl
> >> + * @vidioc_ext_querybuf: pointer to the function that implements
> >> + *  :ref:`VIDIOC_EXT_QUERYBUF <vidioc_ext_querybuf>` ioctl
> >>   * @vidioc_qbuf: pointer to the function that implements
> >>   *  :ref:`VIDIOC_QBUF <vidioc_qbuf>` ioctl
> >> + * @vidioc_ext_qbuf: pointer to the function that implements
> >> + *  :ref:`VIDIOC_EXT_QBUF <vidioc_ext_qbuf>` ioctl
> >>   * @vidioc_expbuf: pointer to the function that implements
> >>   *  :ref:`VIDIOC_EXPBUF <vidioc_expbuf>` ioctl
> >>   * @vidioc_dqbuf: pointer to the function that implements
> >>   *  :ref:`VIDIOC_DQBUF <vidioc_qbuf>` ioctl
> >> + * @vidioc_ext_dqbuf: pointer to the function that implements
> >> + *  :ref:`VIDIOC_EXT_DQBUF <vidioc_ext_qbuf>` ioctl
> >>   * @vidioc_create_bufs: pointer to the function that implements
> >>   *  :ref:`VIDIOC_CREATE_BUFS <vidioc_create_bufs>` ioctl
> >> + * @vidioc_ext_create_bufs: pointer to the function that implements
> >> + *  :ref:`VIDIOC_EXT_CREATE_BUFS <vidioc_ext_create_bufs>` ioctl
> >>   * @vidioc_prepare_buf: pointer to the function that implements
> >>   *  :ref:`VIDIOC_PREPARE_BUF <vidioc_prepare_buf>` ioctl
> >> + * @vidioc_ext_prepare_buf: pointer to the function that implements
> >> + *  :ref:`VIDIOC_EXT_PREPARE_BUF <vidioc_ext_prepare_buf>` ioctl
> >>   * @vidioc_overlay: pointer to the function that implements
> >>   *  :ref:`VIDIOC_OVERLAY <vidioc_overlay>` ioctl
> >>   * @vidioc_g_fbuf: pointer to the function that implements
> >> @@ -439,17 +449,27 @@ struct v4l2_ioctl_ops {
> >>                            struct v4l2_requestbuffers *b);
> >>      int (*vidioc_querybuf)(struct file *file, void *fh,
> >>                             struct v4l2_buffer *b);
> >> +    int (*vidioc_ext_querybuf)(struct file *file, void *fh,
> >> +                               struct v4l2_ext_buffer *b);
> >>      int (*vidioc_qbuf)(struct file *file, void *fh,
> >>                         struct v4l2_buffer *b);
> >> +    int (*vidioc_ext_qbuf)(struct file *file, void *fh,
> >> +                           struct v4l2_ext_buffer *b);
> >>      int (*vidioc_expbuf)(struct file *file, void *fh,
> >>                           struct v4l2_exportbuffer *e);
> >>      int (*vidioc_dqbuf)(struct file *file, void *fh,
> >>                          struct v4l2_buffer *b);
> >> +    int (*vidioc_ext_dqbuf)(struct file *file, void *fh,
> >> +                            struct v4l2_ext_buffer *b);
> >>
> >>      int (*vidioc_create_bufs)(struct file *file, void *fh,
> >>                                struct v4l2_create_buffers *b);
> >> +    int (*vidioc_ext_create_bufs)(struct file *file, void *fh,
> >> +                                  struct v4l2_ext_create_buffers *b);
> >>      int (*vidioc_prepare_buf)(struct file *file, void *fh,
> >>                                struct v4l2_buffer *b);
> >> +    int (*vidioc_ext_prepare_buf)(struct file *file, void *fh,
> >> +                                  struct v4l2_ext_buffer *b);
> >>
> >>      int (*vidioc_overlay)(struct file *file, void *fh, unsigned int i);
> >>      int (*vidioc_g_fbuf)(struct file *file, void *fh,
> >> @@ -758,6 +778,12 @@ int v4l2_ext_pix_format_to_format(const struct v4l2_ext_pix_format *e,
> >>                                struct v4l2_format *f,
> >>                                bool mplane_cap, bool strict);
> >>
> >> +int v4l2_ext_buffer_to_buffer(const struct v4l2_ext_buffer *e,
> >> +                          struct v4l2_buffer *b,
> >> +                          bool mplane_cap);
> >> +int v4l2_buffer_to_ext_buffer(const struct v4l2_buffer *b,
> >> +                          struct v4l2_ext_buffer *e);
> >> +
> >>  /*
> >>   * The user space interpretation of the 'v4l2_event' differs
> >>   * based on the 'time_t' definition on 32-bit architectures, so
> >> diff --git a/include/uapi/linux/videodev2.h b/include/uapi/linux/videodev2.h
> >> index 7123c6a4d9569..334cafdd2be97 100644
> >> --- a/include/uapi/linux/videodev2.h
> >> +++ b/include/uapi/linux/videodev2.h
> >> @@ -996,6 +996,41 @@ struct v4l2_plane {
> >>      __u32                   reserved[11];
> >>  };
> >>
> >> +/**
> >> + * struct v4l2_ext_plane - extended plane buffer info
> >> + * @buffer_length:  size of the entire buffer in bytes, should fit
> >> + *                  @offset + @plane_length
> >
> > Do we actually need this buffer_length at all? We have 3 memory types:
> >
> > 1) MMAP - here vb2 already knows the buffer size, because it created it.
> >
> > 2) DMABUF - the DMA-buf kAPI provides the information about buffer size.
> >
> > 3) USERPTR - this might actually benefit from buffer_length, because there
> >    are additional alignmnent requirements for the user memory, e.g. the
> >    offset and size must be cacheline aligned.
> >
> > Arguably, 1) and 2) are the main usage scenarios, while the user space that
> > uses them would have to suffer from added complexity, because of the
> > legacy/niche case 3).
> >
> > Could we make this field valid only for USERPTR?
>
> I think so, make sense, I'll implement this for next version.
>
> >
> >> + * @plane_length:   size of the plane in bytes.
> >> + * @mem_offset:             If V4L2_MEMORY_MMAP is used, then it can be a "cookie"
> >> + *                  that should be passed to mmap() called on the video node.
> >> + * @userptr:                when memory is V4L2_MEMORY_USERPTR, a userspace pointer pointing
> >> + *                  to this plane.
> >> + * @dmabuf_fd:              when memory is V4L2_MEMORY_DMABUF, a userspace file descriptor
> >> + *                  associated with this plane.
> >> + * @offset:         offset in the memory buffer where the plane starts.
> >> + * @memory:         enum v4l2_memory; the method, in which the actual video
> >> + *                  data is passed
> >> + * @reserved:               extra space reserved for future fields, must be set to 0.
> >> + *
> >> + *
> >> + * Buffers consist of one or more planes, e.g. an YCbCr buffer with two planes
> >> + * can have one plane for Y, and another for interleaved CbCr components.
> >> + * Each plane can reside in a separate memory buffer, or even in
> >> + * a completely separate memory node (e.g. in embedded devices).
> >> + */
> >> +struct v4l2_ext_plane {
> >> +    __u32 buffer_length;
> >> +    __u32 plane_length;
> >> +    union {
> >> +            __u32 mem_offset;
> >> +            __u64 userptr;
> >> +            __s32 dmabuf_fd;
> >> +    } m;
> >> +    __u32 offset;
> >> +    __u32 memory;
> >> +    __u32 reserved[4];
> >> +};
> >
> > Don't we also need bytesused? Or would plane_length essentially mean the
> > amount of space or payload, depending on the usage context?
>
> In my understanding, plane_length is the max amount of data the plane can
> occupy, othersize it can overflow the buffer or mess with another plane
> that is in the same buffer on a different offset.
>
> I'm probably wrong, but I don't really see why the payload size is usefull,
> unless if we set a plane_legth that is much bigger then the data it can carry,
> that can impact performance.
> Payload can also be calculated from the format.

It is required for non-2D-image formats, such as compressed bitstream.
In that case, the size of the image varies between frames and is
usually less than the size of the buffer.

>
> I can add it back if it is usefull. Please let me know your thoughts.
>

As I mentioned above, it looks like plane_length is used almost
exactly the same way bytesused was in the original code, so maybe it
could just stay this way?

> >
> > Similarly, the original data_offset was useful as a return field, which
> > some drivers use to indicate that the beginning of the plane is occupied by
> > some header or otherwise irrelevant data, which must be skipped. Would the
> > offset field be used for this purpose now?
>
> I didn't add an equivalent of the data_offset, since it seemed to be
> unused (please see my comments about this above).
>
> >
> >> +
> >>  /**
> >>   * struct v4l2_buffer - video buffer info
> >>   * @index:  id number of the buffer
> >> @@ -1057,6 +1092,33 @@ struct v4l2_buffer {
> >>      };
> >>  };
> >>
> >> +/**
> >> + * struct v4l2_ext_buffer - extended video buffer info
> >> + * @index:  id number of the buffer
> >> + * @type:   V4L2_BUF_TYPE_VIDEO_CAPTURE or V4L2_BUF_TYPE_VIDEO_OUTPUT
> >> + * @flags:  buffer informational flags
> >
> > nit: The order of comments doesn't match the order of fields in the struct.
>
> Ack.
>
> >
> >> + * @field:  enum v4l2_field; field order of the image in the buffer
> >> + * @timestamp:      frame timestamp
> >> + * @sequence:       sequence count of this frame
> >> + * @planes: per-plane buffer information
> >> + * @request_fd:     fd of the request that this buffer should use
> >> + * @reserved:       extra space reserved for future fields, must be set to 0
> >> + *
> >> + * Contains data exchanged by application and driver using one of the Streaming
> >> + * I/O methods.
> >> + */
> >> +struct v4l2_ext_buffer {
> >> +    __u32 index;
> >> +    __u32 type;
> >> +    __u32 field;
> >> +    __u32 sequence;
> >> +    __u64 flags;
> >> +    __u64 timestamp;
> >
> > What's the unit? How does this play with the other UAPI that the user space
> > may use, e.g. clock_gettime() which returns struct timespec?
>
> The unity is nsec:
>
>         e->timestamp = b->timestamp.tv_sec * NSEC_PER_SEC +
>                 b->timestamp.tv_usec * NSEC_PER_USEC;
>
> I can clarify in the docs, is this ok?
>

Yes, it definitely needs to be documented. That said, what's the
rationale for switching from the timeval representation to the flat
nsec-based one?

Best regards,
Tomasz