Bug 38675 - Recompiled kernel crash on KMS loading
Summary: Recompiled kernel crash on KMS loading
Status: RESOLVED NOTABUG
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Radeon (show other bugs)
Version: unspecified
Hardware: x86 (IA32) Linux (All)
: medium critical
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-06-25 11:31 UTC by muziofg
Modified: 2011-07-15 14:32 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
archive containing .config and dmesg output of my custom kernel (28.84 KB, application/x-gzip)
2011-06-25 11:31 UTC, muziofg
no flags Details
old and new config file of my custom kernel (32.05 KB, application/x-gzip)
2011-06-27 05:24 UTC, muziofg
no flags Details

Description muziofg 2011-06-25 11:31:47 UTC
Created attachment 48405 [details]
archive containing .config and dmesg output of my custom kernel

With my custom kernel (config and dmesg attached at this message in a tar.gz archive) I have a problem with radeon DRM module. 

When the KMS is loading the system is freezed, the keyboard is dead, the machine doesn't answer at the ping and I can only reset the machine.
I haven't any debug information because this crash show a black screen 
and no file (dmesg, kern.log, messages, ecc) are written.

This problem doesn't happen always, but often so it is also difficult to determine when a change in the configuration in the kernel may have solved the problem

I actually use a linux kernel version 2.6.39 on a debian testing. I compile a personalized kernel from the sources packaged by debian.

The problem seems don't happen with the prebuild kernel (I have tried the package  linux-image-2.6.39-2-686-pae) but I'm not sure, simply for now, with this, the machine was started without problem

I never had problems with kernel 2.6.38 using a very similar .config file

The radeon card is a HD6670 PCI Express (chip TURKS)

See here http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=631567 for other details
Comment 2 muziofg 2011-06-25 16:05:20 UTC
I tried to apply the patch and it did not work, the system freezes with a kernel panic. It seems correct because if a precompiled kernel (hence not patched) works and my custom kernel works (sometimes) could not be this the problem.

What do you mean with "bisect" ?
Comment 3 muziofg 2011-06-27 05:23:46 UTC
Problem solved

My config have many years, each time "make menuconfig" add the new options but in the original configuration I have exagerated with the hardening :) 

I have analysed, compared and merged the config of a custom kernel on my old laptop (wich is full amd) and this .config. Now the kernel boot and KMS is loaded without problem

I have attached a new archive with old and new confing files (configOldAndNewForDiff.tar.gz). The answer is in the diff

If you consider interesting the detailed analisys of the problem I can try to test more combination between this two configuration to find what is requested by KMS for work without problem
Comment 4 muziofg 2011-06-27 05:24:28 UTC
Created attachment 48463 [details]
old and new config file of my custom kernel
Comment 5 Ben Hutchings 2011-07-15 14:32:10 UTC
Although the 'crash' configuration is not very sensible, any valid configuration should not result in a crash so long as the hardware requirements are met. Comparing the 'ok' and 'crash' configurations, I think the important difference is probably:

-CONFIG_VIDEO_OUTPUT_CONTROL=y
+# CONFIG_VIDEO_OUTPUT_CONTROL is not set


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.