MIME-Version: 1.0
References: <20230314110043.2139111-1-treapking@chromium.org> <CAD=FV=W=jVsvD620wWKfniRYQNJnb7goDUKb_HhL_qVxLYGZOA@mail.gmail.com>
In-Reply-To: <CAD=FV=W=jVsvD620wWKfniRYQNJnb7goDUKb_HhL_qVxLYGZOA@mail.gmail.com>
From:   Pin-yen Lin <treapking@chromium.org>
Date:   Wed, 15 Mar 2023 11:28:42 +0800
Message-ID: <CAEXTbpe6EyukjKfgaVtHdMK2Ppw715kUUnOqvFa+tEX913p9aQ@mail.gmail.com>
Subject: Re: [PATCH 1/2] drm/bridge: ps8640: Skip redundant bridge enable
To:     Doug Anderson <dianders@chromium.org>
Cc:     Andrzej Hajda <andrzej.hajda@intel.com>,
        Neil Armstrong <neil.armstrong@linaro.org>,
        Robert Foss <rfoss@kernel.org>,
        Laurent Pinchart <Laurent.pinchart@ideasonboard.com>,
        Jonas Karlman <jonas@kwiboo.se>,
        Jernej Skrabec <jernej.skrabec@gmail.com>,
        David Airlie <airlied@gmail.com>,
        Daniel Vetter <daniel@ffwll.ch>,
        Enric Balletbo i Serra <enric.balletbo@collabora.com>,
        linux-kernel@vger.kernel.org, dri-devel@lists.freedesktop.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Hi Doug,

On Wed, Mar 15, 2023 at 5:31=E2=80=AFAM Doug Anderson <dianders@chromium.or=
g> wrote:
>
> Hi,
>
> On Tue, Mar 14, 2023 at 4:00=E2=80=AFAM Pin-yen Lin <treapking@chromium.o=
rg> wrote:
> >
> > Skip the drm_bridge_chain_pre_enable call when the bridge is already
> > pre_enabled. This make pre_enable and post_disable (thus
> > pm_runtime_get/put) symmetric.
> >
> > Fixes: 46f206304db0 ("drm/bridge: ps8640: Rework power state handling")
> > Signed-off-by: Pin-yen Lin <treapking@chromium.org>
> > ---
> >
> >  drivers/gpu/drm/bridge/parade-ps8640.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/gpu/drm/bridge/parade-ps8640.c b/drivers/gpu/drm/b=
ridge/parade-ps8640.c
> > index 4b361d7d5e44..08de501c436e 100644
> > --- a/drivers/gpu/drm/bridge/parade-ps8640.c
> > +++ b/drivers/gpu/drm/bridge/parade-ps8640.c
> > @@ -557,7 +557,8 @@ static struct edid *ps8640_bridge_get_edid(struct d=
rm_bridge *bridge,
> >          * EDID, for this chip, we need to do a full poweron, otherwise=
 it will
> >          * fail.
> >          */
> > -       drm_atomic_bridge_chain_pre_enable(bridge, connector->state->st=
ate);
> > +       if (poweroff)
> > +               drm_atomic_bridge_chain_pre_enable(bridge, connector->s=
tate->state);
>
> It always seemed weird to me that this function was asymmetric, so I
> like this change, thanks!
>
> I also remember wondering before how this function was safe, though.
> The callpath for getting here from the ioctl is documented in the
> function and when I look at it I wonder if anything is preventing the
> bridge from being enabled / disabled through normal means at the same
> time your function is running. That could cause all sorts of badness
> if it is indeed possible. Does anyone reading this know if that's
> indeed a problem?

If the "normal mean" is disabling the bridge, then we are probably
disabling the whole display pipeline. If so, is the EDID still
relevant in this case?

drm_get_edid returns NULL whenever error happens, and the helpers seem
to handle this case properly.
>
> I suppose that, if this is unsafe, it's no more unsafe now than it was
> before your patch, so I guess:
>
> Reviewed-by: Douglas Anderson <dianders@chromium.org>
>
> If there are no issues, I'll plan to land this patch and the next one
> to drm-misc-next sometime late-ish next week.

Thanks for the review. I'll submit a v2 to collect the review tags and
fix up the nit in patch 2/2.

Best regards,
Pin-yen