Re: [PATCH 2/5] media: adv748x: csi2: Link AFE with TXA and TXB

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

Hi Jacopo,

Thank you for the patch.

On Wednesday, 12 December 2018 10:21:01 EET jacopo mondi wrote:
> On Tue, Dec 11, 2018 at 11:07:09PM +0000, Kieran Bingham wrote:
> > On 11/12/2018 15:16, Jacopo Mondi wrote:
> >> The ADV748x chip supports routing AFE output to either TXA or TXB.
> >> In order to support run-time configuration of video stream path, create
> >> an additional (not enabled) "AFE:8->TXA:0" link, and remove the
> >> IMMUTABLE flag from existing ones.
> >> 
> >> Signed-off-by: Jacopo Mondi <jacopo+renesas@jmondi.org>
> >> ---
> >> 
> >>  drivers/media/i2c/adv748x/adv748x-csi2.c | 48 ++++++++++++++++---------
> >>  1 file changed, 30 insertions(+), 18 deletions(-)
> >> 
> >> diff --git a/drivers/media/i2c/adv748x/adv748x-csi2.c
> >> b/drivers/media/i2c/adv748x/adv748x-csi2.c index
> >> 6ce21542ed48..4d1aefc2c8d0 100644
> >> --- a/drivers/media/i2c/adv748x/adv748x-csi2.c
> >> +++ b/drivers/media/i2c/adv748x/adv748x-csi2.c
> >> @@ -27,6 +27,7 @@ static int adv748x_csi2_set_virtual_channel(struct
> >> adv748x_csi2 *tx,> > 
> >>   * @v4l2_dev: Video registration device
> >>   * @src: Source subdevice to establish link
> >>   * @src_pad: Pad number of source to link to this @tx
> >> + * @flags: Flags for the newly created link

You may want to name this link_flags, up to you.

> >>   *
> >>   * Ensure that the subdevice is registered against the v4l2_device, and
> >>   link the
> >>   * source pad to the sink pad of the CSI2 bus entity.

[snip]

> >> @@ -68,24 +63,41 @@ static int adv748x_csi2_registered(struct
> >> v4l2_subdev *sd)
> >>  {
> >>  	struct adv748x_csi2 *tx = adv748x_sd_to_csi2(sd);
> >>  	struct adv748x_state *state = tx->state;
> >> +	int ret;
> >> 
> >>  	adv_dbg(state, "Registered %s (%s)", is_txa(tx) ? "TXA":"TXB",
> >>  			sd->name);
> >>  	
> >>  	/*
> >> -	 * The adv748x hardware allows the AFE to route through the TXA,
> >> however
> >> -	 * this is not currently supported in this driver.
> >> +	 * Link TXA to HDMI and AFE, and TXB to AFE only as TXB cannot output
> >> +	 * HDMI.
> >>  	 *
> >> -	 * Link HDMI->TXA, and AFE->TXB directly.
> >> +	 * The HDMI->TXA link is enabled by default, as the AFE->TXB is.
> >>  	 */
> >> -	if (is_txa(tx) && is_hdmi_enabled(state))
> >> -		return adv748x_csi2_register_link(tx, sd->v4l2_dev,
> >> -						  &state->hdmi.sd,
> >> -						  ADV748X_HDMI_SOURCE);
> >> -	if (!is_txa(tx) && is_afe_enabled(state))
> >> +	if (is_txa(tx)) {
> >> +		if (is_hdmi_enabled(state)) {
> >> +			ret = adv748x_csi2_register_link(tx, sd->v4l2_dev,
> >> +							 &state->hdmi.sd,
> >> +							 ADV748X_HDMI_SOURCE,
> >> +							 MEDIA_LNK_FL_ENABLED);
> >> +			if (ret)
> >> +				return ret;
> >> +		}
> >> +
> >> +		if (is_afe_enabled(state)) {
> >> +			ret = adv748x_csi2_register_link(tx, sd->v4l2_dev,
> >> +							 &state->afe.sd,
> >> +							 ADV748X_AFE_SOURCE,
> >> +							 0);
> >> +			if (ret)
> >> +				return ret;
> >> +		}

I wonder whether in the case where both HDMI and TXB are disabled we shouldn't 
create this link as immutable and enabled ? This would require a big 
refactoring, possibly including deep changes in the v4l2-async and MC code, so 
it's likely out of scope for this patch (unless you see an easy way to do so).

> >> +	} else if (is_afe_enabled(state))
> > 
> > I believe when adding braces to one side of an if statement, we are
> > supposed to add to the else clauses too ?
> 
> Correct

And I would write this

	} else {
		if (is_afe_enabled(state))
			...
	}

To more clearly visualize the TXA and TXB code.

> >>  		return adv748x_csi2_register_link(tx, sd->v4l2_dev,
> >>  						  &state->afe.sd,
> >> -						  ADV748X_AFE_SOURCE);
> >> +						  ADV748X_AFE_SOURCE,
> >> +						  MEDIA_LNK_FL_ENABLED);
> > 
> > Won't this enable the AFE link for both TXA and TXB ?
> > Which one will win? Just the last one ? the first one ?
> > Does it error?
> > 
> > (It might not be a problem ... I can't recall what the behaviour is)
> 
> The AFE->TXA link is created as not enabled (see the 0 as last
> argument in the adv748x_csi2_register_link() call above here, in the
> 'is_txa(tx)' case
> 
> >> +
> > 
> > There are a lot of nested if's above, and I think we can simplify
> > greatly if we move the logic for the flags inside
> > adv748x_csi2_register_link(), and adjust the checks on is_xxx_enabled()
> > 
> > What do you think about the following pseudo code?:
> > 
> > int adv748x_csi2_register_link(struct adv748x_csi2 *tx,
> >   				      struct v4l2_device *v4l2_dev,
> >   				      struct v4l2_subdev *src,
> > 				      unsigned int src_pad,
> > 				      bool enable)
> > {
> >   int flags = 0;
> >   int ret;
> >   
> >   if (!src->v4l2_dev) {
> > 	ret = v4l2_device_register_subdev(v4l2_dev, src)
> > 	if (ret) return ret;
> >   }
> >   
> >   if (enable)
> > 	flags = MEDIA_LNK_FL_ENABLED;
> > 	
> >    return media_create_pad_link(&src->entity, src_pad,
> >  			        &tx->sd.entity, ADV748X_CSI2_SINK,
> >  			        flags);
> > }
> > 
> > int adv748x_csi2_registered(struct v4l2_subdev *sd)
> > {
> >   int ret;
> >   

I would add

	/* The AFE can be routed to both TXA and TXB. */

> >   if (is_afe_enabled(state) {
> >       ret = adv748x_csi2_register_link(tx, sd->v4l2_dev, &state->afe.sd,
> > 				   ADV748X_AFE_SOURCE, !is_txa(tx));
> >       if (ret)
> > 	  return ret;
> >   }
> >   
> >   /* TX-B only supports AFE */
> >   if (!is_txa(tx) || !(is_hdmi_enabled(state))
> > 	return 0;
> > 	
> >   return adv748x_csi2_register_link(tx, sd->v4l2_dev, &state->hdmi.sd,
> > 				    ADV748X_HDMI_SOURCE, true);
> > }

Looks better than my above proposal :-)

> > The above will for TXA:
> > 	register_link(..., AFE_SOURCE, enable = false );
> > 	register_link(..., HDMI_SOURCE, enable = true );
> > 
> > then TXB:
> > 	register_link(..., AFE_SOURCE, enable = true );
> > 
> > Does that meet our needs?
> 
> Yes it does, and it looks better. Thanks!
> 
> >>  	return 0;
> >>  }

With Kieran's proposed change,

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

-- 
Regards,

Laurent Pinchart