forwarded 624549 http://thread.gmane.org/gmane.linux.kernel/1126934
quit

Hi,

Martin Teufel wrote:

> radeon 0000:01:05.0: HDMI-A-1: EDID block 0 invalid.
> [drm:radeon_dvi_detect] *ERROR* HDMI-A-1: probed a monitor but no|invalid EDID
>
> This error message is spamming to tty1 (and because of my
> configuration to ttyS0 too).
[...]
> I already reportet this on the LKML btw.

Thanks for letting us know. Could you try the patch from

https://bugs.freedesktop.org/show_bug.cgi?id=27708#c7

and boot with the drm.edid_strict=0 option?



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110429214608.GA399@elie">http://lists.debian.org/20110429214608.GA399@elie

Hi,

I applied this patch to the source tree of linux 2.6.38 with debian
patches (package linux-source-2.6.38 of wheezy), compiled the kernel
with the same config as of debians linux-image-2.6.38-2-amd64 (wheezy)
and set 'drm_kms_helper.poll=0', but it had absolutely no affect. Same
messages every 10 seconds.

Any further suggestions?

Best regards,
Martin


On Fri, Apr 29, 2011 at 04:46:08PM -0500, Jonathan Nieder wrote:
> Thanks for letting us know. Could you try the patch from
>
> https://bugs.freedesktop.org/show_bug.cgi?id=27708#c7
>
> and boot with the drm.edid_strict=0 option?



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110503020306.GA5231@hekate.rbi.informatik.uni-frankfurt.de">http://lists.debian.org/20110503020306.GA5231@hekate.rbi.informatik.uni-frankfurt.de

Hi Martin,

Martin Teufel wrote:
> On Fri, Apr 29, 2011 at 04:46:08PM -0500, Jonathan Nieder wrote:

>> Thanks for letting us know. Could you try the patch from
>>
>> https://bugs.freedesktop.org/show_bug.cgi?id=27708#c7
>>
>> and boot with the drm.edid_strict=0 option?
>
> I applied this patch to the source tree of linux 2.6.38 with debian
> patches (package linux-source-2.6.38 of wheezy), compiled the kernel
> with the same config as of debians linux-image-2.6.38-2-amd64 (wheezy)
> and set 'drm_kms_helper.poll=0', but it had absolutely no affect. Same
> messages every 10 seconds.
>
> Any further suggestions?

No, aside from bisecting or studying the code. But just to be sure:
drm_kms_helper.poll=0 != drm.edid_strict=0

Thanks and hope that helps,
Jonathan



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110503030234.GA25314@elie">http://lists.debian.org/20110503030234.GA25314@elie

Martin Teufel wrote:

> *g* - yes, I know that drm_kms_helper.poll=0 != drm.edid_strict=0.
> I temporarily removed drm_kms_helper.poll=0 and added drm.edid_strict=0
> to the linux kernel parameters (replaced it).
> Hm, I will try it a second time and see what happens (but don't think it
> will change something).
> Strange thing...

Thanks; cc-ing the bug log.

So, my only other suggestion is a few rounds of bisection. It works
like this:

apt-get install git build-essential

git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
cd linux-2.6
git bisect start
git bisect bad v2.6.37
git bisect good v2.6.32; # or whichever was known to be good

# now it checks out a version half-way between
make localmodconfig; # minimal configuration
make deb-pkg

... install the package, reboot, test ...
git bisect good; # if it doesn't spam dmesg
git bisect bad; # if it does
git bisect skip; # if some other bug makes it hard to test

# okay, next iteration...
make oldconfig
make deb-pkg

At the end of this process, we would know what the "first bad commit"
is, which can be very helpful for tracking it down. Just narrowing
down the regression range with a few rounds is already useful; in that
case, if you send "git bisect log" output at the end we should be able
to use it to figure something out. If you are the visual sort of
person like I am, "git bisect visualize" with gitk installed can show
what's going on between steps.

Thanks again for working on this, and sorry I don't have more useful
advice.

Regards,
Jonathan



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110503190108.GB1019@elie">http://lists.debian.org/20110503190108.GB1019@elie

Hi together,

does someone know if this patch has been applied already to
linux-image-2.6.38-bpo.2-amd64 (2.6.38-3~bpo60+1)?
I installed this image yesterday and had the idea to test the
drm.edid_strict=0 parameter on it today. - It works like a charm.
Dunno why my patch didn't work... (But I don't have the time nor the
mood to track this down atm.)

Nevertheless: Thank you for trying to help me. - And sorry for
bugreporting this twice and not cc-ing the last mail. Reason for the
first was that I never used reportbug on this machine before and
therefore had no .reportbugrc on it which caused in a nonexistent mail
address due to some circumstances.

Best regards,
Martin


On Tue, May 03, 2011 at 02:01:08PM -0500, Jonathan Nieder wrote:
> Thanks; cc-ing the bug log.
>
> So, my only other suggestion is a few rounds of bisection. It works
> like this:
>
> apt-get install git build-essential
>
> git clone git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
> cd linux-2.6
> git bisect start
> git bisect bad v2.6.37
> git bisect good v2.6.32; # or whichever was known to be good
>
> # now it checks out a version half-way between
> make localmodconfig; # minimal configuration
> make deb-pkg
>
> ... install the package, reboot, test ...
> git bisect good; # if it doesn't spam dmesg
> git bisect bad; # if it does
> git bisect skip; # if some other bug makes it hard to test
>
> # okay, next iteration...
> make oldconfig
> make deb-pkg
>
> At the end of this process, we would know what the "first bad commit"
> is, which can be very helpful for tracking it down. Just narrowing
> down the regression range with a few rounds is already useful; in that
> case, if you send "git bisect log" output at the end we should be able
> to use it to figure something out. If you are the visual sort of
> person like I am, "git bisect visualize" with gitk installed can show
> what's going on between steps.
>
> Thanks again for working on this, and sorry I don't have more useful
> advice.
>
> Regards,
> Jonathan

Hi,

I want and have to rectify my last mail:
This patch is obviously not applied in 2.6.38-bpo.2 because I get:
[ 5.924366] drm: Unknown parameter `edid_strict'
Could you get this applied to the next version of a linux backport?

Best regards,
Martin


On Tue, May 03, 2011 at 10:29:27PM +0200, Martin Teufel wrote:
> Hi together,
>
> does someone know if this patch has been applied already to
> linux-image-2.6.38-bpo.2-amd64 (2.6.38-3~bpo60+1)?
> I installed this image yesterday and had the idea to test the
> drm.edid_strict=0 parameter on it today. - It works like a charm.
> Dunno why my patch didn't work... (But I don't have the time nor the
> mood to track this down atm.)
>
> Nevertheless: Thank you for trying to help me. - And sorry for
> bugreporting this twice and not cc-ing the last mail. Reason for the
> first was that I never used reportbug on this machine before and
> therefore had no .reportbugrc on it which caused in a nonexistent mail
> address due to some circumstances.
>
> Best regards,
> Martin

Martin Teufel wrote:
> On Tue, May 03, 2011 at 10:29:27PM +0200, Martin Teufel wrote:

>> I installed this image yesterday and had the idea to test the
>> drm.edid_strict=0 parameter on it today. - It works like a charm.
[...]
> I want and have to rectify my last mail:
> This patch is obviously not applied in 2.6.38-bpo.2 because I get:
> [ 5.924366] drm: Unknown parameter `edid_strict'

So that leaves us with a puzzle: does that mean that 2.6.38-3~bpo60+1
fixed it (forgetting about the drm.edid_strict parameter for a moment)?

> Could you get this applied to the next version of a linux backport?

Ah, sorry I wasn't clearer before. The edid_strict patch is something
Alex had been thinking about but it is not in appliable shape (it does
not have strict enough buffer size checks to cope with e.g.
maliciously made hardware). And of course it's even better if it is
possible through a driver change to get an uncorrupted edid.



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110505205631.GA4758@elie">http://lists.debian.org/20110505205631.GA4758@elie

Hi,

On Thu, May 05, 2011 at 03:56:31PM -0500, Jonathan Nieder wrote:
> So that leaves us with a puzzle: does that mean that 2.6.38-3~bpo60+1
> fixed it (forgetting about the drm.edid_strict parameter for a moment)?
no. Without drm.edid_strict=0 the error occurs (every 10 seconds ...).
With drm.edid_strict=0 given, the error doesn't occur but KMS gets
disabled (don't know why and figured this out only today, when I got to
the machine (did this test via a serial line behind a ssh connection on
tuesday).

> Ah, sorry I wasn't clearer before. The edid_strict patch is something
> Alex had been thinking about but it is not in appliable shape (it does
> not have strict enough buffer size checks to cope with e.g.
> maliciously made hardware). And of course it's even better if it is
> possible through a driver change to get an uncorrupted edid.
Hm, I see (and agree).
So we/I'll wait for a fix from upstream.
I got back to drm_kms_helper.poll=0 for now.

Regards,
Martin

Martin Teufel wrote:

> Hm, I see (and agree).
> So we/I'll wait for a fix from upstream.
> I got back to drm_kms_helper.poll=0 for now.

Mm, I have read your report carefully enough yet, but: did you get any
response upstream? I would suggest filing a bug at [1], product DRI,
component DRM/Radeon, version unspecified (yes, this is the place for
kernel DRI bugs), mentioning which versions worked and did not work,
pointing to your lkml message and attaching logs.

That way one could be sure upstream knows about the problem.

If you do so, do please send us the bug number.

Thanks and regards,
Jonathan

[1] https://bugs.freedesktop.org/enter_bug.cgi?product=DRI



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110505220705.GB5147@elie">http://lists.debian.org/20110505220705.GB5147@elie

Jonathan Nieder wrote:
> Martin Teufel wrote:

>> Hm, I see (and agree).
>> So we/I'll wait for a fix from upstream.
>> I got back to drm_kms_helper.poll=0 for now.
>
> I have read your report carefully enough yet, but: did you get any

Erm, what I mean is that I have _not_ read and investigated your
report carefully. Sorry for the noise.



--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 20110505221037.GC5147@elie">http://lists.debian.org/20110505221037.GC5147@elie