RE: [PATCH v5 4/6] drm: rcar-du: lvds: Allow for even and odd pixels swap

[Fabrizio Castro <fabrizio.castro@bp.renesas.com>]

Hi Laurent,

Thank you for your feedback!

> From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> Sent: 16 December 2019 21:35
> Subject: Re: [PATCH v5 4/6] drm: rcar-du: lvds: Allow for even and odd pixels swap
> 
> Hi Fabrizio,
> 
> Thank you for the patch.
> 
> On Mon, Dec 16, 2019 at 08:12:32PM +0000, Fabrizio Castro wrote:
> > DT properties dual-lvds-even-pixels and dual-lvds-odd-pixels
> > can be used to work out if the driver needs to swap even
> > and odd pixels around.
> >
> > This patch makes use of the return value from function
> > drm_of_lvds_get_dual_link_pixel_order to determine if we
> > need to swap odd and even pixels around for things to work
> > properly.
> >
> > Signed-off-by: Fabrizio Castro <fabrizio.castro@bp.renesas.com>
> >
> > ---
> > v4->v5:
> > * Addressed comments from Laurent's review
> >
> > v3->v4:
> > * New patch extracted from patch:
> >   "drm: rcar-du: lvds: Add dual-LVDS panels support"
> > ---
> >  drivers/gpu/drm/rcar-du/rcar_lvds.c | 67 +++++++++++++++++++++++++++++--------
> >  1 file changed, 53 insertions(+), 14 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/rcar-du/rcar_lvds.c b/drivers/gpu/drm/rcar-du/rcar_lvds.c
> > index 3eb208e..c6a38c3 100644
> > --- a/drivers/gpu/drm/rcar-du/rcar_lvds.c
> > +++ b/drivers/gpu/drm/rcar-du/rcar_lvds.c
> > @@ -37,6 +37,12 @@ enum rcar_lvds_mode {
> >  	RCAR_LVDS_MODE_VESA = 4,
> >  };
> >
> > +enum rcar_lvds_link_type {
> > +	RCAR_LVDS_SINGLE_LINK = 0,
> > +	RCAR_LVDS_DUAL_LINK_EVEN_ODD_PIXELS = 1,
> > +	RCAR_LVDS_DUAL_LINK_ODD_EVEN_PIXELS = 2,
> > +};
> > +
> >  #define RCAR_LVDS_QUIRK_LANES		BIT(0)	/* LVDS lanes 1 and 3 inverted */
> >  #define RCAR_LVDS_QUIRK_GEN3_LVEN	BIT(1)	/* LVEN bit needs to be set on R8A77970/R8A7799x */
> >  #define RCAR_LVDS_QUIRK_PWD		BIT(2)	/* PWD bit available (all of Gen3 but E3) */
> > @@ -67,7 +73,7 @@ struct rcar_lvds {
> >  	} clocks;
> >
> >  	struct drm_bridge *companion;
> > -	bool dual_link;
> > +	enum rcar_lvds_link_type dual_link;
> 
> Do you think we should rename this to link_type (and test for
> lvds->link_type != RCAR_LVDS_SINGLE_LINK) instead of lvds->dual_link) ?
> Apart from that,

I'll send v6 to address that. I will also rebase on top of:
https://patchwork.kernel.org/patch/11295991/
and will use that as a dependency for v6

Cheers,
Fab

> 
> Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> 
> >  };
> >
> >  #define bridge_to_rcar_lvds(b) \
> > @@ -484,12 +490,31 @@ static void rcar_lvds_atomic_enable(struct drm_bridge *bridge,
> >  	rcar_lvds_write(lvds, LVDCHCR, lvdhcr);
> >
> >  	if (lvds->info->quirks & RCAR_LVDS_QUIRK_DUAL_LINK) {
> > -		/*
> > -		 * Configure vertical stripe based on the mode of operation of
> > -		 * the connected device.
> > -		 */
> > -		rcar_lvds_write(lvds, LVDSTRIPE,
> > -				lvds->dual_link ? LVDSTRIPE_ST_ON : 0);
> > +		u32 lvdstripe = 0;
> > +
> > +		if (lvds->dual_link) {
> > +			/*
> > +			 * By default we generate even pixels from the primary
> > +			 * encoder and odd pixels from the companion encoder.
> > +			 * Swap pixels around if the sink requires odd pixels
> > +			 * from the primary encoder and even pixels from the
> > +			 * companion encoder.
> > +			 */
> > +			bool swap_pixels = lvds->dual_link ==
> > +				RCAR_LVDS_DUAL_LINK_ODD_EVEN_PIXELS;
> > +
> > +			/*
> > +			 * Configure vertical stripe since we are dealing with
> > +			 * an LVDS dual-link connection.
> > +			 *
> > +			 * ST_SWAP is reserved for the companion encoder, only
> > +			 * set it in the primary encoder.
> > +			 */
> > +			lvdstripe = LVDSTRIPE_ST_ON
> > +				  | (lvds->companion && swap_pixels ?
> > +				     LVDSTRIPE_ST_SWAP : 0);
> > +		}
> > +		rcar_lvds_write(lvds, LVDSTRIPE, lvdstripe);
> >  	}
> >
> >  	/*
> > @@ -716,15 +741,26 @@ static int rcar_lvds_parse_dt_companion(struct rcar_lvds *lvds)
> >  	of_node_put(port0);
> >  	of_node_put(port1);
> >
> > -	if (dual_link >= DRM_LVDS_DUAL_LINK_EVEN_ODD_PIXELS)
> > -		lvds->dual_link = true;
> > -	else if (lvds->next_bridge && lvds->next_bridge->timings)
> > +	switch (dual_link) {
> > +	case DRM_LVDS_DUAL_LINK_ODD_EVEN_PIXELS:
> > +		lvds->dual_link = RCAR_LVDS_DUAL_LINK_ODD_EVEN_PIXELS;
> > +		break;
> > +	case DRM_LVDS_DUAL_LINK_EVEN_ODD_PIXELS:
> > +		lvds->dual_link = RCAR_LVDS_DUAL_LINK_EVEN_ODD_PIXELS;
> > +		break;
> > +	default:
> >  		/*
> >  		 * Early dual-link bridge specific implementations populate the
> > -		 * timings field of drm_bridge, read the dual_link flag off the
> > -		 * bridge directly for backward compatibility.
> > +		 * timings field of drm_bridge. If the flag is set, we assume
> > +		 * that we are expected to generate even pixels from the primary
> > +		 * encoder, and odd pixels from the companion encoder.
> >  		 */
> > -		lvds->dual_link = lvds->next_bridge->timings->dual_link;
> > +		if (lvds->next_bridge && lvds->next_bridge->timings &&
> > +		    lvds->next_bridge->timings->dual_link)
> > +			lvds->dual_link = RCAR_LVDS_DUAL_LINK_EVEN_ODD_PIXELS;
> > +		else
> > +			lvds->dual_link = RCAR_LVDS_SINGLE_LINK;
> > +	}
> >
> >  	if (!lvds->dual_link) {
> >  		dev_dbg(dev, "Single-link configuration detected\n");
> > @@ -741,6 +777,9 @@ static int rcar_lvds_parse_dt_companion(struct rcar_lvds *lvds)
> >  		"Dual-link configuration detected (companion encoder %pOF)\n",
> >  		companion);
> >
> > +	if (lvds->dual_link == RCAR_LVDS_DUAL_LINK_ODD_EVEN_PIXELS)
> > +		dev_dbg(dev, "Data swapping required\n");
> > +
> >  	/*
> >  	 * FIXME: We should not be messing with the companion encoder private
> >  	 * data from the primary encoder, we should rather let the companion
> > @@ -751,7 +790,7 @@ static int rcar_lvds_parse_dt_companion(struct rcar_lvds *lvds)
> >  	 * for the time being.
> >  	 */
> >  	companion_lvds = bridge_to_rcar_lvds(lvds->companion);
> > -	companion_lvds->dual_link = true;
> > +	companion_lvds->dual_link = lvds->dual_link;
> >
> >  done:
> >  	of_node_put(companion);
> 
> --
> Regards,
> 
> Laurent Pinchart