linux-kernel - Re: [RFC v2 1/8] drm/i915/dp: Program source OUI on eDP panels

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: <53161495c2915ad60f749a0917bc606259a78e7c.camel@redhat.com>
Date:   Mon, 30 Nov 2020 18:06:00 -0500
From:   Lyude Paul <lyude@...hat.com>
To:     Jani Nikula <jani.nikula@...el.com>,
        intel-gfx@...ts.freedesktop.org
Cc:     thaytan@...aisin.net, David Airlie <airlied@...ux.ie>,
        open list <linux-kernel@...r.kernel.org>,
        Gwan-gyeong Mun <gwan-gyeong.mun@...el.com>,
        Vasily Khoruzhick <anarsoul@...il.com>,
        Uma Shankar <uma.shankar@...el.com>,
        Sean Paul <seanpaul@...omium.org>,
        dri-devel@...ts.freedesktop.org,
        Rodrigo Vivi <rodrigo.vivi@...el.com>,
        José Roberto de Souza 
        <jose.souza@...el.com>, Manasi Navare <manasi.d.navare@...el.com>,
        Wambui Karuga <wambui.karugax@...il.com>
Subject: Re: [RFC v2 1/8] drm/i915/dp: Program source OUI on eDP panels

On Thu, 2020-11-26 at 12:51 +0200, Jani Nikula wrote:
> On Wed, 16 Sep 2020, Lyude Paul <lyude@...hat.com> wrote:
> > Since we're about to start adding support for Intel's magic HDR
> > backlight interface over DPCD, we need to ensure we're properly
> > programming this field so that Intel specific sink services are exposed.
> > Otherwise, 0x300-0x3ff will just read zeroes.
> > 
> > We also take care not to reprogram the source OUI if it already matches
> > what we expect. This is just to be careful so that we don't accidentally
> > take the panel out of any backlight control modes we found it in.
> > 
> > v2:
> > * Add careful parameter to intel_edp_init_source_oui() to avoid
> >   re-writing the source OUI if it's already been set during driver
> >   initialization
> > 
> > Signed-off-by: Lyude Paul <lyude@...hat.com>
> > Cc: thaytan@...aisin.net
> > Cc: Vasily Khoruzhick <anarsoul@...il.com>
> > ---
> >  drivers/gpu/drm/i915/display/intel_dp.c | 33 +++++++++++++++++++++++++
> >  1 file changed, 33 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/display/intel_dp.c
> > b/drivers/gpu/drm/i915/display/intel_dp.c
> > index 4bd10456ad188..7db2b6a3cd52e 100644
> > --- a/drivers/gpu/drm/i915/display/intel_dp.c
> > +++ b/drivers/gpu/drm/i915/display/intel_dp.c
> > @@ -3424,6 +3424,29 @@ void intel_dp_sink_set_decompression_state(struct
> > intel_dp *intel_dp,
> >                             enable ? "enable" : "disable");
> >  }
> >  
> > +static void
> > +intel_edp_init_source_oui(struct intel_dp *intel_dp, bool careful)
> > +{
> > +       struct drm_i915_private *i915 = dp_to_i915(intel_dp);
> > +       u8 oui[] = { 0x00, 0xaa, 0x01 };
> 
> Nitpick, could be static const.
> 
> > +       u8 buf[3] = { 0 };
> > +
> > +       /*
> > +        * During driver init, we want to be careful and avoid changing
> > the source OUI if it's
> > +        * already set to what we want, so as to avoid clearing any state
> > by accident
> > +        */
> 
> Did you actually observe any ill behaviour with unconditionally writing
> the source OUI?
> 
> I mean it's easy to add the "careful" mode afterwards if there are
> concrete issues, but it'll be hard to remove because it'll be a
> functional change potentially causing regressions.

I haven't yet, although I'll give a test on some of the other machines I've
got lying around.

FWIW, relevant spec info:

   A Sink device that does not support the Source-device specified behavior
   specified by the owner of the IEEE OUI written to in DPCD Addresses 00300h
   through 00302h as being associated with the Source Identification shall
   AUX_ACK all writes, but take no other action, and shall respond to reads
   with an AUX_ACK and the value 00h.

> 
> > +       if (careful) {
> > +               if (drm_dp_dpcd_read(&intel_dp->aux, DP_SOURCE_OUI, buf,
> > sizeof(buf)) < 0)
> > +                       drm_err(&i915->drm, "Failed to read source
> > OUI\n");
> > +
> > +               if (memcmp(oui, buf, sizeof(oui)) == 0)
> > +                       return;
> > +       }
> > +
> > +       if (drm_dp_dpcd_write(&intel_dp->aux, DP_SOURCE_OUI, oui,
> > sizeof(oui)) < 0)
> > +               drm_err(&i915->drm, "Failed to write source OUI\n");
> > +}
> > +
> >  /* If the sink supports it, try to set the power state appropriately */
> >  void intel_dp_sink_dpms(struct intel_dp *intel_dp, int mode)
> >  {
> > @@ -3443,6 +3466,10 @@ void intel_dp_sink_dpms(struct intel_dp *intel_dp,
> > int mode)
> >         } else {
> >                 struct intel_lspcon *lspcon = dp_to_lspcon(intel_dp);
> >  
> > +               /* Write the source OUI as early as possible */
> > +               if (intel_dp_is_edp(intel_dp))
> > +                       intel_edp_init_source_oui(intel_dp, false);
> > +
> 
> This hunk conflicts, will need a rebase.
> 
> BR,
> Jani.
> 
> 
> >                 /*
> >                  * When turning on, we need to retry for 1ms to give the
> > sink
> >                  * time to wake up.
> > @@ -4607,6 +4634,12 @@ intel_edp_init_dpcd(struct intel_dp *intel_dp)
> >         if (INTEL_GEN(dev_priv) >= 10 || IS_GEMINILAKE(dev_priv))
> >                 intel_dp_get_dsc_sink_cap(intel_dp);
> >  
> > +       /*
> > +        * If needed, program our source OUI so we can make various Intel-
> > specific AUX services
> > +        * available (such as HDR backlight controls)
> > +        */
> > +       intel_edp_init_source_oui(intel_dp, true);
> > +
> >         return true;
> >  }
> 

-- 
Cheers,
 Lyude Paul (she/her)
 Software Engineer at Red Hat