linux-kernel - Re: [PATCH] drm: rcar_lvds: Fix dual link mode operations

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20190805075109.oe6slqetqoe5wqzr@uno.localdomain>
Date:   Mon, 5 Aug 2019 09:51:09 +0200
From:   Jacopo Mondi <jacopo@...ndi.org>
To:     Laurent Pinchart <laurent.pinchart@...asonboard.com>
Cc:     Jacopo Mondi <jacopo+renesas@...ndi.org>,
        Kieran Bingham <kieran.bingham+renesas@...asonboard.com>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>,
        fabrizio.castro@...renesas.com,
        "open list:DRM DRIVERS FOR RENESAS" <dri-devel@...ts.freedesktop.org>,
        "open list:DRM DRIVERS FOR RENESAS" 
        <linux-renesas-soc@...r.kernel.org>,
        open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] drm: rcar_lvds: Fix dual link mode operations

Hi Laurent,

On Thu, Aug 01, 2019 at 06:11:29PM +0300, Laurent Pinchart wrote:
> Hello Jacopo,
>
> Thank you for the patch.
>
> On Tue, Jul 23, 2019 at 06:57:00PM +0200, Jacopo Mondi wrote:
> > The R-Car LVDS encoder units support dual-link operations by splitting
> > the pixel output between the primary encoder and the companion one.
>
> s/the companion one/its companion/
>
> >
> > In order for the primary encoder to succesfully control the companion's
> > operations this should not fail at probe time and register itself its
> > associated drm bridge so that the primary one can find it.
>
> This is hard to parse.
>

Re-reading the whole commit message, I would actually drop it
completely, it's enough what we have here below.

> > Currently the companion encoder fails at probe time, causing the
> > registration of the primary to fail preventing the whole DU unit to be
> > registered correctly.
> >
> > Fixes: fa440d870358 ("drm: rcar-du: lvds: Add support for dual-link mode")
> > Reported-by: Fabrizio Castro <fabrizio.castro@...renesas.com>
> > Signed-off-by: Jacopo Mondi <jacopo+renesas@...ndi.org>
> >
> > ---
> > The "Fixes" tag refers to a patch currently part of the
> > renesas-drivers-2019-07-09-v5.2 branch of Geert's renesas-drivers tree.
> >
> >  drivers/gpu/drm/rcar-du/rcar_lvds.c | 31 +++++++++++++++++++++--------
> >  1 file changed, 23 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/rcar-du/rcar_lvds.c b/drivers/gpu/drm/rcar-du/rcar_lvds.c
> > index bada7ee98544..8b015ba95895 100644
> > --- a/drivers/gpu/drm/rcar-du/rcar_lvds.c
> > +++ b/drivers/gpu/drm/rcar-du/rcar_lvds.c
> > @@ -767,14 +767,29 @@ static int rcar_lvds_parse_dt(struct rcar_lvds *lvds)
> >  	of_node_put(remote_input);
> >  	of_node_put(remote);
> >
> > -	/*
> > -	 * On D3/E3 the LVDS encoder provides a clock to the DU, which can be
> > -	 * used for the DPAD output even when the LVDS output is not connected.
> > -	 * Don't fail probe in that case as the DU will need the bridge to
> > -	 * control the clock.
> > -	 */
> > -	if (lvds->info->quirks & RCAR_LVDS_QUIRK_EXT_PLL)
> > -		return ret == -ENODEV ? 0 : ret;
> > +	switch (ret) {
> > +	case -ENODEV:
> > +		/*
> > +		 * On D3/E3 the LVDS encoder provides a clock to the DU, which
> > +		 * can be used for the DPAD output even when the LVDS output is
> > +		 * not connected. Don't fail probe in that case as the DU will
> > +		 * need the bridge to control the clock.
> > +		 */
> > +		if (lvds->info->quirks & RCAR_LVDS_QUIRK_EXT_PLL)
> > +			ret = 0;
> > +		break;
> > +	case -ENXIO:
> > +		/*
> > +		 * When the LVDS output is used in dual link mode, the
> > +		 * companion encoder fails at
> > +		 * 'rcar_lvds_parse_dt_companion()'. Don't fail probe in
> > +		 * that case as the master encoder will need the companion's
> > +		 * bridge to control its operations.
> > +		 */
> > +		if (lvds->info->quirks & RCAR_LVDS_QUIRK_DUAL_LINK)
> > +			ret = 0;
>
> As -ENXIO can only be returned by rcar_lvds_parse_dt_companion(), and
> rcar_lvds_parse_dt_companion() is only called when the
> RCAR_LVDS_QUIRK_DUAL_LINK flag is set, this essentially means that you
> always zero the error returned from rcar_lvds_parse_dt_companion(). This

Not totally correct, as rcar_lvds_parse_dt_companion() might also
return EPROBE_DEFER, but...

> is both complicated and too drastic as the second -ENXIO error case
> shouldn't be ignored. It would be better and simpler to return 0 from
> rcar_lvds_parse_dt_companion() when the renesas,companion property can't
> be found.

I agree, returning 0 when when the property is not specified is enough
and much simpler. I got dragged away by the idea of centralizing error
handling at the end of the function, but it's ugly and also wrongly
zeroes the second -ENXIO error returned by the parse_companion
function.

I'll change to what you suggested!
Thanks
  j

>
> > +		break;
> > +	}
> >
> >  	return ret;
> >  }
>
> --
> Regards,
>
> Laurent Pinchart

Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)