Add support for the frame_sync event, so user-space can become aware
earlier of new frames.
Suggested-by: Esker Wong <[email protected]>
Tested-by: Esker Wong <[email protected]>
Signed-off-by: Ricardo Ribalda <[email protected]>
---
We have measured a latency of around 30msecs between frame sync
and dqbuf.
---
drivers/media/usb/uvc/uvc_v4l2.c | 2 ++
drivers/media/usb/uvc/uvc_video.c | 8 +++++++-
2 files changed, 9 insertions(+), 1 deletion(-)
diff --git a/drivers/media/usb/uvc/uvc_v4l2.c b/drivers/media/usb/uvc/uvc_v4l2.c
index f4988f03640a..9f3fb5fd2375 100644
--- a/drivers/media/usb/uvc/uvc_v4l2.c
+++ b/drivers/media/usb/uvc/uvc_v4l2.c
@@ -1352,6 +1352,8 @@ static int uvc_ioctl_subscribe_event(struct v4l2_fh *fh,
switch (sub->type) {
case V4L2_EVENT_CTRL:
return v4l2_event_subscribe(fh, sub, 0, &uvc_ctrl_sub_ev_ops);
+ case V4L2_EVENT_FRAME_SYNC:
+ return v4l2_event_subscribe(fh, sub, 0, NULL);
default:
return -EINVAL;
}
diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
index 28dde08ec6c5..1d4b4807b005 100644
--- a/drivers/media/usb/uvc/uvc_video.c
+++ b/drivers/media/usb/uvc/uvc_video.c
@@ -1073,9 +1073,15 @@ static int uvc_video_decode_start(struct uvc_streaming *stream,
* that discontinuous sequence numbers always indicate lost frames.
*/
if (stream->last_fid != fid) {
- stream->sequence++;
+ struct v4l2_event event = {
+ .type = V4L2_EVENT_FRAME_SYNC,
+ .u.frame_sync.frame_sequence = ++stream->sequence,
+ };
+
if (stream->sequence)
uvc_video_stats_update(stream);
+
+ v4l2_event_queue(&stream->vdev, &event);
}
uvc_video_clock_decode(stream, buf, data, len);
---
base-commit: ce55c22ec8b223a90ff3e084d842f73cfba35588
change-id: 20231020-uvc-event-d3d1bbbdcb2f
Best regards,
--
Ricardo Ribalda <[email protected]>
Friendly ping (in text mode :P)
Hi Ricardo,
Thank you for the patch.
On Fri, Oct 20, 2023 at 06:41:45AM +0000, Ricardo Ribalda wrote:
> Add support for the frame_sync event, so user-space can become aware
> earlier of new frames.
>
> Suggested-by: Esker Wong <[email protected]>
> Tested-by: Esker Wong <[email protected]>
> Signed-off-by: Ricardo Ribalda <[email protected]>
> ---
> We have measured a latency of around 30msecs between frame sync
> and dqbuf.
Not surprising, especially for large resolutions. I'm curious though,
what do you use this event for ?
> ---
> drivers/media/usb/uvc/uvc_v4l2.c | 2 ++
> drivers/media/usb/uvc/uvc_video.c | 8 +++++++-
> 2 files changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/media/usb/uvc/uvc_v4l2.c b/drivers/media/usb/uvc/uvc_v4l2.c
> index f4988f03640a..9f3fb5fd2375 100644
> --- a/drivers/media/usb/uvc/uvc_v4l2.c
> +++ b/drivers/media/usb/uvc/uvc_v4l2.c
> @@ -1352,6 +1352,8 @@ static int uvc_ioctl_subscribe_event(struct v4l2_fh *fh,
> switch (sub->type) {
> case V4L2_EVENT_CTRL:
> return v4l2_event_subscribe(fh, sub, 0, &uvc_ctrl_sub_ev_ops);
> + case V4L2_EVENT_FRAME_SYNC:
> + return v4l2_event_subscribe(fh, sub, 0, NULL);
> default:
> return -EINVAL;
> }
> diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
> index 28dde08ec6c5..1d4b4807b005 100644
> --- a/drivers/media/usb/uvc/uvc_video.c
> +++ b/drivers/media/usb/uvc/uvc_video.c
> @@ -1073,9 +1073,15 @@ static int uvc_video_decode_start(struct uvc_streaming *stream,
> * that discontinuous sequence numbers always indicate lost frames.
> */
> if (stream->last_fid != fid) {
> - stream->sequence++;
> + struct v4l2_event event = {
> + .type = V4L2_EVENT_FRAME_SYNC,
> + .u.frame_sync.frame_sequence = ++stream->sequence,
Extra space before ++.
It's easy to miss the ++ there when reading the code, would the
following be more readable ?
struct v4l2_event event = {
.type = V4L2_EVENT_FRAME_SYNC,
};
stream->sequence++;
if (stream->sequence)
uvc_video_stats_update(stream);
.u.frame_sync.frame_sequence = stream->sequence;
v4l2_event_queue(&stream->vdev, &event);
> + };
> +
> if (stream->sequence)
> uvc_video_stats_update(stream);
> +
> + v4l2_event_queue(&stream->vdev, &event);
> }
>
> uvc_video_clock_decode(stream, buf, data, len);
>
> ---
> base-commit: ce55c22ec8b223a90ff3e084d842f73cfba35588
> change-id: 20231020-uvc-event-d3d1bbbdcb2f
--
Regards,
Laurent Pinchart
On Mon, Nov 06, 2023 at 12:39:26PM +0200, Laurent Pinchart wrote:
> On Fri, Oct 20, 2023 at 06:41:45AM +0000, Ricardo Ribalda wrote:
> > Add support for the frame_sync event, so user-space can become aware
> > earlier of new frames.
> >
> > Suggested-by: Esker Wong <[email protected]>
> > Tested-by: Esker Wong <[email protected]>
> > Signed-off-by: Ricardo Ribalda <[email protected]>
> > ---
> > We have measured a latency of around 30msecs between frame sync
> > and dqbuf.
>
> Not surprising, especially for large resolutions. I'm curious though,
> what do you use this event for ?
>
> > ---
> > drivers/media/usb/uvc/uvc_v4l2.c | 2 ++
> > drivers/media/usb/uvc/uvc_video.c | 8 +++++++-
> > 2 files changed, 9 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/media/usb/uvc/uvc_v4l2.c b/drivers/media/usb/uvc/uvc_v4l2.c
> > index f4988f03640a..9f3fb5fd2375 100644
> > --- a/drivers/media/usb/uvc/uvc_v4l2.c
> > +++ b/drivers/media/usb/uvc/uvc_v4l2.c
> > @@ -1352,6 +1352,8 @@ static int uvc_ioctl_subscribe_event(struct v4l2_fh *fh,
> > switch (sub->type) {
> > case V4L2_EVENT_CTRL:
> > return v4l2_event_subscribe(fh, sub, 0, &uvc_ctrl_sub_ev_ops);
> > + case V4L2_EVENT_FRAME_SYNC:
> > + return v4l2_event_subscribe(fh, sub, 0, NULL);
> > default:
> > return -EINVAL;
> > }
> > diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
> > index 28dde08ec6c5..1d4b4807b005 100644
> > --- a/drivers/media/usb/uvc/uvc_video.c
> > +++ b/drivers/media/usb/uvc/uvc_video.c
> > @@ -1073,9 +1073,15 @@ static int uvc_video_decode_start(struct uvc_streaming *stream,
> > * that discontinuous sequence numbers always indicate lost frames.
> > */
> > if (stream->last_fid != fid) {
> > - stream->sequence++;
> > + struct v4l2_event event = {
> > + .type = V4L2_EVENT_FRAME_SYNC,
> > + .u.frame_sync.frame_sequence = ++stream->sequence,
>
> Extra space before ++.
>
> It's easy to miss the ++ there when reading the code, would the
> following be more readable ?
>
> struct v4l2_event event = {
> .type = V4L2_EVENT_FRAME_SYNC,
> };
>
> stream->sequence++;
> if (stream->sequence)
> uvc_video_stats_update(stream);
>
> .u.frame_sync.frame_sequence = stream->sequence;
Obviously this should read
event.u.frame_sync.frame_sequence = stream->sequence;
> v4l2_event_queue(&stream->vdev, &event);
>
> > + };
> > +
> > if (stream->sequence)
> > uvc_video_stats_update(stream);
> > +
> > + v4l2_event_queue(&stream->vdev, &event);
> > }
> >
> > uvc_video_clock_decode(stream, buf, data, len);
> >
> > ---
> > base-commit: ce55c22ec8b223a90ff3e084d842f73cfba35588
> > change-id: 20231020-uvc-event-d3d1bbbdcb2f
--
Regards,
Laurent Pinchart
Hi Laurent
On Mon, 6 Nov 2023 at 11:42, Laurent Pinchart
<[email protected]> wrote:
>
> On Mon, Nov 06, 2023 at 12:39:26PM +0200, Laurent Pinchart wrote:
> > On Fri, Oct 20, 2023 at 06:41:45AM +0000, Ricardo Ribalda wrote:
> > > Add support for the frame_sync event, so user-space can become aware
> > > earlier of new frames.
> > >
> > > Suggested-by: Esker Wong <[email protected]>
> > > Tested-by: Esker Wong <[email protected]>
> > > Signed-off-by: Ricardo Ribalda <[email protected]>
> > > ---
> > > We have measured a latency of around 30msecs between frame sync
> > > and dqbuf.
> >
> > Not surprising, especially for large resolutions. I'm curious though,
> > what do you use this event for ?
I think Esker is using it to get more accurate power measurements of
the camera stack.
> > It's easy to miss the ++ there when reading the code, would the
> > following be more readable ?
Actually that was my original code, but I thought you would like this better :)
Thanks for the review, a v2 is on its way.
> --
> Regards,
>
> Laurent Pinchart
--
Ricardo Ribalda
On Mon, Nov 06, 2023 at 11:51:06AM +0100, Ricardo Ribalda wrote:
> On Mon, 6 Nov 2023 at 11:42, Laurent Pinchart wrote:
> > On Mon, Nov 06, 2023 at 12:39:26PM +0200, Laurent Pinchart wrote:
> > > On Fri, Oct 20, 2023 at 06:41:45AM +0000, Ricardo Ribalda wrote:
> > > > Add support for the frame_sync event, so user-space can become aware
> > > > earlier of new frames.
> > > >
> > > > Suggested-by: Esker Wong <[email protected]>
> > > > Tested-by: Esker Wong <[email protected]>
> > > > Signed-off-by: Ricardo Ribalda <[email protected]>
> > > > ---
> > > > We have measured a latency of around 30msecs between frame sync
> > > > and dqbuf.
> > >
> > > Not surprising, especially for large resolutions. I'm curious though,
> > > what do you use this event for ?
>
> I think Esker is using it to get more accurate power measurements of
> the camera stack.
Esker, would you be able to provide more information ?
> > > It's easy to miss the ++ there when reading the code, would the
> > > following be more readable ?
>
> Actually that was my original code, but I thought you would like this better :)
>
> Thanks for the review, a v2 is on its way.
Thank you.
--
Regards,
Laurent Pinchart