Hi Hongju, On Mon, Nov 04, 2024 at 03:07:58AM +0000, Wang, Hongju wrote: > Hi Sakari > > > -----Original Message----- > > From: Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx> > > Sent: Thursday, October 24, 2024 6:31 PM > > To: Wang, Hongju <hongju.wang@xxxxxxxxx> > > Cc: linux-media@xxxxxxxxxxxxxxx; bingbu.cao@xxxxxxxxxxxxxxx > > Subject: Re: [PATCH] media: intel/ipu6: Set the VC of the stream as the > > SOF event id > > > > Hi Hongju, > > > > On Thu, Oct 24, 2024 at 11:22:01AM +0800, Hongju Wang wrote: > > > In the virtual channel case, they should use v4l2_event.id to > > > distinguish SOF events of different streams. Therefore, we set the > > > virtual channel number of the stream as the SOF event ID. This number is > > unique. > > > > > > Signed-off-by: Hongju Wang <hongju.wang@xxxxxxxxx> > > > --- > > > drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c | 1 + > > > 1 file changed, 1 insertion(+) > > > > > > diff --git a/drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c > > > b/drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c > > > index 051898ce53f4..5ad426afa0f0 100644 > > > --- a/drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c > > > +++ b/drivers/media/pci/intel/ipu6/ipu6-isys-csi2.c > > > @@ -578,6 +578,7 @@ void ipu6_isys_csi2_sof_event_by_stream(struct > > ipu6_isys_stream *stream) > > > .type = V4L2_EVENT_FRAME_SYNC, > > > }; > > > > > > + ev.id = stream->vc; > > > ev.u.frame_sync.frame_sequence = atomic_fetch_inc(&stream->sequence); > > > v4l2_event_queue(vdev, &ev); > > > > > > > The id field in struct v4l2_event isn't used for the FRAME_SYNC event and > > also the virtual channel isn't communicated to the user space currently. > > How to use the id field? User space should use SOF to do FRAME_SYNC, > but the v4l2_event doesn't distinguish different streams. Do you have > any good suggestions? Ah, now I think I do understand the problem. The events are produced by the CSI-2 receiver sub-device but there are multiple streams and the event isn't stream-aware? It's nice that the id field of the FRAME_SYNC event isn't in use for any other purpose. This should be taken into account in event subscription, too. I'll add a patch documenting this in the metadata series. Any thoughts? Cc Laurent and Hans. -- Kind regards, Sakari Ailus
