Hi Laurent,
On Sat, Sep 7, 2024 at 10:09 PM Lad, Prabhakar
<prabhakar.csengg@xxxxxxxxx> wrote:
>
> Hi Laurent,
>
> Thank you for the review.
>
> On Sat, Sep 7, 2024 at 12:10 AM Laurent Pinchart
> <laurent.pinchart@xxxxxxxxxxxxxxxx> wrote:
<snip>
> > > +
> > > int rzg2l_cru_start_image_processing(struct rzg2l_cru_dev *cru)
> > > {
> > > struct v4l2_mbus_framefmt *fmt = rzg2l_cru_ip_get_src_fmt(cru);
> > > unsigned long flags;
> > > int ret;
> > >
> > > + ret = rzg2l_cru_get_virtual_channel(cru);
> > > + if (ret < 0)
> > > + return ret;
> > > + cru->csi.channel = ret;
> >
> > How about passing the value to the function that needs it, instead of
> > storing it in cru->csi.channel ? You can do that on top and drop the
> > csi.channel field.
> >
> OK, let me check if this can be done.
>
The virtual channel number is programmed in rzg2l_cru_csi2_setup() [0]
call, below is the code flow of the call. This code flow is called by
spinlock held.
rzg2l_cru_start_image_processing()
rzg2l_cru_initialize_image_conv()
rzg2l_cru_csi2_setup()
When rzg2l_cru_get_virtual_channel() is called directly in
rzg2l_cru_csi2_setup() function we get "BUG: Invalid wait context"
(when PROVE_LOCKING is enabled) this is due to we do a mutex lock as
part of v4l2_subdev_lock_and_get_active_state() in get_frame_desc.
So probably I'll leave this as it is now.
[0] https://elixir.bootlin.com/linux/v6.10.8/source/drivers/media/platform/renesas/rzg2l-cru/rzg2l-video.c#L311
Cheers,
Prabhakar
