linux-kernel - Re: [PATCH] Revert "ARM: dts: rockchip: add startup delay to rk3288-veyron panel-regulators"

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

Date:   Tue, 2 Jul 2019 21:54:58 -0700
From:   Doug Anderson <dianders@...omium.org>
To:     Heiko Stübner <heiko@...ech.de>
Cc:     Enric Balletbo i Serra <enric.balletbo@...labora.com>,
        Matthias Kaehlcke <mka@...omium.org>,
        devicetree@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>,
        "open list:ARM/Rockchip SoC..." <linux-rockchip@...ts.infradead.org>,
        Rob Herring <robh+dt@...nel.org>,
        Mark Rutland <mark.rutland@....com>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH] Revert "ARM: dts: rockchip: add startup delay to
 rk3288-veyron panel-regulators"

Hi,

On Thu, Jun 20, 2019 at 1:31 PM Doug Anderson <dianders@...omium.org> wrote:
>
> Hi,
>
> On Thu, Jun 20, 2019 at 11:21 AM Douglas Anderson <dianders@...omium.org> wrote:
> >
> > This reverts commit 1f45e8c6d0161f044d679f242fe7514e2625af4a.
> >
> > This 100 ms mystery delay is not on downstream kernels and no longer
> > seems needed on upstream kernels either [1].  Presumably something in the
> > meantime has made things better.  A few possibilities for patches that
> > have landed in the meantime that could have made this better are
> > commit 3157694d8c7f ("pwm-backlight: Add support for PWM delays
> > proprieties."), commit 5fb5caee92ba ("pwm-backlight: Enable/disable
> > the PWM before/after LCD enable toggle."), and commit 6d5922dd0d60
> > ("ARM: dts: rockchip: set PWM delay backlight settings for Veyron")
> >
> > Let's revert and get our 100 ms back.
> >
> > [1] https://lkml.kernel.org/r/2226970.BAPq4liE1j@diego
> >
> > Signed-off-by: Douglas Anderson <dianders@...omium.org>
> > ---
> >
> >  arch/arm/boot/dts/rk3288-veyron-jaq.dts    | 1 -
> >  arch/arm/boot/dts/rk3288-veyron-jerry.dts  | 1 -
> >  arch/arm/boot/dts/rk3288-veyron-minnie.dts | 1 -
> >  arch/arm/boot/dts/rk3288-veyron-speedy.dts | 1 -
> >  4 files changed, 4 deletions(-)
>
> Maybe wait before applying.  I've been running reboot tests now with
> this patch applied (among others) and with enough reboots I managed to
> see:
>
> [    5.682418] rockchip-dp ff970000.dp: eDP link training failed (-5)
>
> I'll see if I can confirm that it's this patch and why things are
> different compared to downstream.

OK, I finally got back to this and confirmed:

1. The above error is actually somewhat harmless.  The eDP failure
will be retried automatically despite the scary message.  Specifically
see the loop in analogix_dp_bridge_enable().  I confirmed that after
seeing the error the screen came up just fine (I looked at the screen
in two actual instances but I believe it's pretty much always fine).

2. I haven't seen any evidence that the eDP link training happens any
more often with this revert in place.  Specifically, I see the same
message in the logs (at what appears to be the same rate) with or
without this revert.

3. Probably the link-training failures here are the same ones we
debugged for PSR for rk3399-gru-kevin that we fixed by making the eDP
PCLK rate exactly 24 MHz.  See <https://crrev.com/c/433393> for
details.  On rk3399-gru-kevin it was super important to resolve the
root cause of these errors because we had PSR (which meant we were
constantly taking to the eDP controller).  On rk3288-veyron devices
with no PSR the retry should be a fine solution and it doesn't seem
like a good idea to fully rejigger our clock plan to fix the root
cause.

NOTE: I saw _one_ case on rk3288-veyron-minnie where the screen looked
wonky at bootup and I saw the eDP link training error in the logs.
That's what originally made me cautious.  I haven't been able to
reproduce this, but presumably I just got super unlucky in that one
case.  I've left devices rebooting all day at work and haven't seen
the wonky screen since then.

Summary: I think this revert is just fine.

-Doug