Re: [PATCH v9 06/19] media: ti: j721e-csi2rx: add a subdev for the core device

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

On Wed, Jan 21, 2026 at 09:38:29AM +0200, Tomi Valkeinen wrote:
> On 21/01/2026 01:25, Laurent Pinchart wrote:
> > On Thu, Jan 15, 2026 at 02:56:21PM +0200, Tomi Valkeinen wrote:
> >> On 15/01/2026 08:36, Jai Luthra wrote:
> >>> Quoting Tomi Valkeinen (2026-01-14 20:51:49)
> >>>> On 30/12/2025 10:32, Rishikesh Donadkar wrote:
> >>>>> From: Jai Luthra <j-luthra@ti.com>
> >>>>>
> >>>>> With single stream capture, it was simpler to use the video device as
> >>>>> the media entity representing the main TI CSI2RX device. Now with multi
> >>>>> stream capture coming into the picture, the model has shifted to each
> >>>>> video device having a link to the main device's subdev. The routing
> >>>>> would then be set on this subdev.
> >>>>>
> >>>>> Add this subdev, link each context to this subdev's entity and link the
> >>>>> subdev's entity to the source. Also add an array of media pads. It will
> >>>>> have one sink pad and source pads equal to the number of contexts.
> >>>>>
> >>>>> Support the new enable_stream()/disable_stream() APIs in the subdev
> >>>>> instead of s_stream() hook.
> >>>>>
> >>>>> Reviewed-by: Yemike Abhilash Chandra <y-abhilashchandra@ti.com>
> >>>>> Co-developed-by: Pratyush Yadav <p.yadav@ti.com>
> >>>>> Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
> >>>>> Signed-off-by: Jai Luthra <j-luthra@ti.com>
> >>>>> Signed-off-by: Rishikesh Donadkar <r-donadkar@ti.com>
> >>>>> ---
> >>>
> >>> [...]
> >>>
> >>>>> @@ -981,48 +1138,52 @@ static int ti_csi2rx_link_validate(struct media_link *link)
> >>>>>       struct ti_csi2rx_ctx *ctx = container_of(vdev, struct ti_csi2rx_ctx, vdev);
> >>>>>       struct ti_csi2rx_dev *csi = ctx->csi;
> >>>>>       struct v4l2_pix_format *csi_fmt = &ctx->v_fmt.fmt.pix;
> >>>>> -     struct v4l2_subdev_format source_fmt = {
> >>>>> -             .which  = V4L2_SUBDEV_FORMAT_ACTIVE,
> >>>>> -             .pad    = link->source->index,
> >>>>> -     };
> >>>>> +     struct v4l2_mbus_framefmt *format;
> >>>>> +     struct v4l2_subdev_state *state;
> >>>>>       const struct ti_csi2rx_fmt *ti_fmt;
> >>>>> -     int ret;
> >>>>>  
> >>>>> -     ret = v4l2_subdev_call_state_active(csi->source, pad,
> >>>>> -                                         get_fmt, &source_fmt);
> >>>>> -     if (ret)
> >>>>> -             return ret;
> >>>>> +     state = v4l2_subdev_lock_and_get_active_state(&csi->subdev);
> >>>>> +     format = v4l2_subdev_state_get_format(state, link->source->index, 0);
> >>>>> +     v4l2_subdev_unlock_state(state);
> >>>>>  
> >>>>> -     if (source_fmt.format.width != csi_fmt->width) {
> >>>>> +     if (!format) {
> >>>>> +             dev_dbg(csi->dev,
> >>>>> +                     "Skipping validation as no format present on \"%s\":%u:0\n",
> >>>>> +                     link->source->entity->name, link->source->index);
> >>>>> +             return 0;
> >>>>
> >>>> Isn't this an error?
> >>>
> >>> Well, the j7 shim subdev introduced here has immutable and active links to
> >>> all the video nodes, for each DMA channel (taken from DT), many of which
> >>> may be unused for certain setups, and thus there might not be any valid
> >>> format on the subdev source pad corresponding to an unused video node.
> >>>
> >>> Jacopo had a similar comment on v2, see this discussion (grep for Mali):
> >>> https://lore.kernel.org/linux-media/4mnlnsj4co3agvln4qsasmgvgwiyoo7yu2h5wyh4rmzzafhm5u@avhnbw7iknms/
> >>>
> >>> I know other drivers use a different approach with mutable links, so it
> >>> would be good if you/Laurent/Sakari can give your opinions on if only one
> >>> of these two approaches should be taken for multi-stream pipelines.
> >>
> >> I see.
> >>
> >> Well, I don't have a definite answer. With some thinking both options
> >> make certain sense. It makes sense to keep the links immutable and
> >> always enabled, as there's no configuration that can be done. On the
> >> other hand, it makes sense to require the unused links to be disabled,
> >> as, well, they are not used.
> > 
> > I'm not familiar with the implications this would have on this driver,
> > but generally speaking, if a stream is added to the media pipeline by
> > the pipeline build algorithm, then it is expected that applications
> > would have configured it correctly. Streams that are not used are
> > expected to be disabled if they would otherwise be added to the
> > pipeline.
> 
> I think the thing here is that the driver creates immutable
> always-enabled media links between the videodevs and the first subdev.
> Then, say, if only one stream is being used, only one of those links is
> actually used, and for every other link the above check fails as there's
> no stream, so no format.
> 
> In TI CAL driver the links were mutable, and unused links had to be
> disabled. There it made sense as the links had to be configurable (there
> were two PHYs). Here, there's no configuration needed, so immutable
> links make sense, but then they're enabled even when actually not used.

If the routing table in the subdev does not contain any route that goes
towards a video node, then that video node should not be added to the
pipeline by the validation code, and no validation will be attempted. At
least that's the theory.

I see that this driver implements .link_validate() as a
media_entity_operations, not a subdev operation. I wonder if that could
explain the issue.

-- 
Regards,

Laurent Pinchart