Regression: Backlight not coming on after resume.

Regression: Backlight not coming on after resume.

[James Courtier-Dutton]

Bisect done on tree:
git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git
Worked on 3.7.
Broken on 3.8

Problem: "Backlight not coming on after resume."
git finally bisected  for my Samsung Serial 7 Laptop.
Bisecting: 0 revisions left to test after this (roughly 0 steps)
[cf0a6584aa6d382f802f2c3cacac23ccbccde0cd] drm/i915: write backlight
harder     <- This is the problem commit. Seems that fixing some
machines breaks others, based on the existing comments on that patch.

git bisect log
git bisect start
# bad: [09884964335e85e897876d17783c2ad33cf8a2e0] mm: do not grow the
stack vma just because of an overrun on preceding vma
git bisect bad 09884964335e85e897876d17783c2ad33cf8a2e0
# good: [6c5096e5538b455bc3bea2b02588c380f070d8c6] Merge tag 'boards'
of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
git bisect good 6c5096e5538b455bc3bea2b02588c380f070d8c6
# good: [21fbd5809ad126b949206d78e0a0e07ec872ea11] Merge branch
'v4l_for_linus' of
git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media
git bisect good 21fbd5809ad126b949206d78e0a0e07ec872ea11
# bad: [fffddfd6c8e0c10c42c6e2cc54ba880fcc36ebbb] Merge branch
'drm-next' of git://people.freedesktop.org/~airlied/linux
git bisect bad fffddfd6c8e0c10c42c6e2cc54ba880fcc36ebbb
# bad: [b81e059ec5a7128622ab5d74d78e9b4f361b54ae] Merge branch
'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel
into drm-next
git bisect bad b81e059ec5a7128622ab5d74d78e9b4f361b54ae
# good: [d6dd9eb1d96d2b7345fe4664066c2b7ed86da898] drm/i915: dynamic
Haswell display power well support
git bisect good d6dd9eb1d96d2b7345fe4664066c2b7ed86da898
# good: [73ccd6962fff19e53a7d4baaa17cf9311284ac9c] Merge branch
'drm-next-3.9' of git://people.freedesktop.org/~agd5f/linux into
drm-next
git bisect good 73ccd6962fff19e53a7d4baaa17cf9311284ac9c
# good: [a4462f246c8821f625f45bce52c7ca7e0207dffe] staging/omapdrm:
Use kmemdup rather than duplicating its implementation
git bisect good a4462f246c8821f625f45bce52c7ca7e0207dffe
# bad: [d84f031bd230fdf9c3b7734940c859bf28b90219] drm: Use C8 instead
of RGB332 when determining the format from depth/bpp
git bisect bad d84f031bd230fdf9c3b7734940c859bf28b90219
# bad: [f73f760725636b9d0c3786273e185b053516d1eb] drm/i915/ctx: Remove
bad invariant
git bisect bad f73f760725636b9d0c3786273e185b053516d1eb
# bad: [26739f12cf210cb8df35969258a1f064e8e12b63] drm/i915: unify
HDMI/DP hpd definitions
git bisect bad 26739f12cf210cb8df35969258a1f064e8e12b63
# good: [b8efb17b3d687695b81485f606fc4e6c35a50f9a] i915: ignore lid
open event when resuming
git bisect good b8efb17b3d687695b81485f606fc4e6c35a50f9a
# bad: [07ea0d85ac8adb87b817913d9720e3c76171b1f6] drm/i915: Clarify HW
context size logic
git bisect bad 07ea0d85ac8adb87b817913d9720e3c76171b1f6

Re: Regression: Backlight not coming on after resume.

[Jani Nikula]

On Thu, 28 Feb 2013, James Courtier-Dutton <james.dutton@gmail.com> wrote:
> Bisect done on tree:
> git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git
> Worked on 3.7.
> Broken on 3.8
>
> Problem: "Backlight not coming on after resume."
> git finally bisected  for my Samsung Serial 7 Laptop.
> Bisecting: 0 revisions left to test after this (roughly 0 steps)
> [cf0a6584aa6d382f802f2c3cacac23ccbccde0cd] drm/i915: write backlight
> harder     <- This is the problem commit. Seems that fixing some
> machines breaks others, based on the existing comments on that patch.

The backlight haunts us again. Could you post the dmesg with
drm.debug=0xe module parameter, first running vanilla 3.8, and then with
the bad commit reverted, please?

Might not be a bad idea to file a bug on Product=DRI,
Component=DRM/Intel at https://bugs.freedesktop.org (see
https://01.org/linuxgraphics/documentation/how-report-bugs-0 for
details) and attach your bisect results and the dmesgs there.

Thanks,
Jani.

Re: Regression: Backlight not coming on after resume.

[Takashi Iwai]

At Thu, 28 Feb 2013 17:33:56 +0200,
Jani Nikula wrote:
> 
> On Thu, 28 Feb 2013, James Courtier-Dutton <james.dutton@gmail.com> wrote:
> > Bisect done on tree:
> > git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git
> > Worked on 3.7.
> > Broken on 3.8
> >
> > Problem: "Backlight not coming on after resume."
> > git finally bisected  for my Samsung Serial 7 Laptop.
> > Bisecting: 0 revisions left to test after this (roughly 0 steps)
> > [cf0a6584aa6d382f802f2c3cacac23ccbccde0cd] drm/i915: write backlight
> > harder     <- This is the problem commit. Seems that fixing some
> > machines breaks others, based on the existing comments on that patch.
> 
> The backlight haunts us again. Could you post the dmesg with
> drm.debug=0xe module parameter, first running vanilla 3.8, and then with
> the bad commit reverted, please?
> 
> Might not be a bad idea to file a bug on Product=DRI,
> Component=DRM/Intel at https://bugs.freedesktop.org (see
> https://01.org/linuxgraphics/documentation/how-report-bugs-0 for
> details) and attach your bisect results and the dmesgs there.

I'm afraid the patch doesn't do it right.  It fixes for Dell, but the
old problem on others back by that.  It was merged before actually
testing on the machines the original fix was intended for.

Now I can confirm a problem (partially) with the latest Linus tree on
a new laptop.  I write "partially" because it happens only during the
splash screen of S4 hibernation, so I've ignored, so far.

James, try to revert that commit.  Does it fix?

If so, you should check the output of intel_reg_dumper at broken and
working states.


Takashi

Re: Regression: Backlight not coming on after resume.

[James Courtier-Dutton]

[-- Attachment #1.1: Type: text/plain, Size: 1948 bytes --]

On Feb 28, 2013 5:20 PM, "Takashi Iwai" <tiwai@suse.de> wrote:
>
> At Thu, 28 Feb 2013 17:33:56 +0200,
> Jani Nikula wrote:
> >
> > On Thu, 28 Feb 2013, James Courtier-Dutton <james.dutton@gmail.com>
wrote:
> > > Bisect done on tree:
> > > git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git
> > > Worked on 3.7.
> > > Broken on 3.8
> > >
> > > Problem: "Backlight not coming on after resume."
> > > git finally bisected  for my Samsung Serial 7 Laptop.
> > > Bisecting: 0 revisions left to test after this (roughly 0 steps)
> > > [cf0a6584aa6d382f802f2c3cacac23ccbccde0cd] drm/i915: write backlight
> > > harder     <- This is the problem commit. Seems that fixing some
> > > machines breaks others, based on the existing comments on that patch.
> >
> > The backlight haunts us again. Could you post the dmesg with
> > drm.debug=0xe module parameter, first running vanilla 3.8, and then with
> > the bad commit reverted, please?
> >
> > Might not be a bad idea to file a bug on Product=DRI,
> > Component=DRM/Intel at https://bugs.freedesktop.org (see
> > https://01.org/linuxgraphics/documentation/how-report-bugs-0 for
> > details) and attach your bisect results and the dmesgs there.
>
> I'm afraid the patch doesn't do it right.  It fixes for Dell, but the
> old problem on others back by that.  It was merged before actually
> testing on the machines the original fix was intended for.
>
> Now I can confirm a problem (partially) with the latest Linus tree on
> a new laptop.  I write "partially" because it happens only during the
> splash screen of S4 hibernation, so I've ignored, so far.
>
> James, try to revert that commit.  Does it fix?
>
> If so, you should check the output of intel_reg_dumper at broken and
> working states.
>
Can someone point me to a datasheet that relates to this bit of code, so i
can understand what all the registers do.
I would like to understand the code in order to suggest a better fix.

[-- Attachment #1.2: Type: text/html, Size: 2735 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Re: Regression: Backlight not coming on after resume.

[Jani Nikula]

Hi James -

On Fri, 01 Mar 2013, James Courtier-Dutton <james.dutton@gmail.com> wrote:
>> James, try to revert that commit.  Does it fix?
>>
>> If so, you should check the output of intel_reg_dumper at broken and
>> working states.
>>
> Can someone point me to a datasheet that relates to this bit of code, so i
> can understand what all the registers do.
> I would like to understand the code in order to suggest a better fix.

https://01.org/linuxgraphics/documentation/driver-documentation-prms

Vol 3 is the relevant volume in each generation. I don't think any of
the docs back up the comments in the bad commit saying "BLC_PWM_CPU_CTL
is cleared to zero automatically when BLC_PWM_CPU_CTL2 and
BLC_PWM_PCH_CTL1 are written" though. I guess it could be some bad
ACPI/BIOS interaction too.

I'd still urge you to file the bug, and attach the dmesgs and register
dumps first. And let us know if the revert fixes it.


BR,
Jani.