Bug 11973 - Laptops with Nvidia MCP67 motherboard freeze during boot, unless a key is held down
Summary: Laptops with Nvidia MCP67 motherboard freeze during boot, unless a key is hel...
Status: CLOSED DOCUMENTED
Alias: None
Product: Platform Specific/Hardware
Classification: Unclassified
Component: x86-64 (show other bugs)
Hardware: All Linux
: P1 high
Assignee: platform_x86_64@kernel-bugs.osdl.org
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-11-07 09:36 UTC by Chris WIlson
Modified: 2012-05-22 15:15 UTC (History)
6 users (show)

See Also:
Kernel Version: 2.6.27
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments
dmesg.log (43.55 KB, text/plain)
2008-11-07 10:22 UTC, Chris WIlson
Details
lspci-vvnn.log (23.60 KB, text/plain)
2008-11-07 10:22 UTC, Chris WIlson
Details
uname-a.log (89 bytes, text/plain)
2008-11-07 10:23 UTC, Chris WIlson
Details
version.log (26 bytes, text/plain)
2008-11-07 10:23 UTC, Chris WIlson
Details
kernel bisect log (10.81 KB, text/plain)
2008-12-02 10:11 UTC, Benedetto Tagliavia
Details

Description Chris WIlson 2008-11-07 09:36:33 UTC
Latest working kernel version: 2.6.26
Earliest failing kernel version: 2.6.27 (All tested version display this behavior)
Distribution: Ubuntu 8.10 64bit
Hardware Environment: AMD64
Software Environment: Ubuntu 8.10 64bit
Problem Description:

Once the Ubuntu splash screen loads with the progress bar that moves back and forth, it freezes. Holding down any key unfreezes the system, and it continues
loading normally, until I release the key, at which point it freezes again. Holding a key down again unfreezes the system again. Inserting or removing a USB device at this point will also momentarily resume the boot.

Once X loads, it works perfectly fine though, and I don't have to keep holding a key down.

I have also personally seen this behavior with a Fedora 10 Beta live DVD using the same 2.6.27 kernel.

Steps to reproduce: Boot the computer. Once the kernel has loaded, it will consistently freeze every time, although you can continue booting by pressing a key.

Ubuntu bug:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/272247

Mandriva bug:
https://qa.mandriva.com/show_bug.cgi?id=44342
Comment 1 Chris WIlson 2008-11-07 10:22:33 UTC
Created attachment 18729 [details]
dmesg.log
Comment 2 Chris WIlson 2008-11-07 10:22:58 UTC
Created attachment 18730 [details]
lspci-vvnn.log
Comment 3 Chris WIlson 2008-11-07 10:23:17 UTC
Created attachment 18731 [details]
uname-a.log
Comment 4 Chris WIlson 2008-11-07 10:23:32 UTC
Created attachment 18732 [details]
version.log
Comment 5 Walter Ribeiro 2008-11-13 12:44:15 UTC
The problem happens with Ubuntu 8.10 i386 also
Comment 6 Benedetto Tagliavia 2008-11-25 11:45:59 UTC
I tried a kernel bisect.
Because it was my first bisect, I'm trying an other bisect to confirm the previous result (I read something more, now, about bisecting and I can understand better how does it work).

My result was:
6924d1ab8b7bbe5ab416713f5701b3316b2df85b is first bad commit

I do not know if the commit hash is good enough to recognize which is the bugged version of the kernel.

The only informations I have are the hash, and the following description:
commit 6924d1ab8b7bbe5ab416713f5701b3316b2df85b
Merge: 4e78c91... 25556c1... b764a15... 437a0a5... 41b3eae... 84e65b0... 684eb01... 9302213... 5cb04df... 44974c8... 48cf937... 205f932... c54f9da... 0ed368c... b478458... 2d144e6... 607baf1... 33af903... 3557b18... 63687a5... 009b9fc... f6477cc... e6b0ede... 400d349...
Author: Ingo Molnar <mingo@elte.hu>
Date: Tue Jul 8 09:16:56 2008 +0200

    Merge branches 'x86/numa-fixes', 'x86/apic', 'x86/apm', 'x86/bitops', 'x86/build', 'x86/cleanups', 'x86/cpa', 'x86/cpu', 'x86/defconfig', 'x86/gart', 'x86/i8259', 'x86/intel', 'x86/irqstats', 'x86/kconfig', 'x86/ldt', 'x86/mce', 'x86/memtest', 'x86/pat', 'x86/ptemask', 'x86/resumetrace', 'x86/threadinfo', 'x86/timers', 'x86/vdso' and 'x86/xen' into x86/devel

I hope, somebody will help the community solving this bug.

Thanks
Nitto
Comment 7 Benedetto Tagliavia 2008-12-02 10:10:49 UTC
Ok, I finished an other bisect, this time the result is a bit different because I skipped more revision (sometimes I could not compile them) but the last bad version is included as possible bad.

This is the message I got:

There are only 'skip'ped commit left to test.
The first bad commit could be any of:
1a750e0cd7a30c478723ecfa1df685efcdd38a90
7dbceaf9bb68919651901b101f44edd5391ee489
437a0a54eea7b101e8a5b70688009956f6522ed0
5136dea5734cfddbc6d7ccb7ead85a3ac7ce3de2
6924d1ab8b7bbe5ab416713f5701b3316b2df85b
We cannot bisect more!

The last one is the one I got the last time.

I hope this will be my last bisect for a while.... :-)

Thanks, I hope in a fast bug solving.
I also attach my bisect log.
Comment 8 Benedetto Tagliavia 2008-12-02 10:11:48 UTC
Created attachment 19104 [details]
kernel bisect log
Comment 9 Landon Baine 2009-02-18 05:48:33 UTC
This problem still exists in kernel 2.6.28.8.8 (from Ubuntu Daily Build Feb 18, 2009).  Is there anything I can provide to help with this?
Comment 10 John Zaitseff 2009-03-09 21:26:14 UTC
This behaviour (hang on boot) also occurs on an Asus M3N78 Pro motherboard with an AMD processor (running in 32-bit mode).  The chipset is "nVidia Corporation MCP78S [GeForce 8200]", according to lspci.
Comment 11 ykzhao 2009-04-13 07:58:04 UTC
Will you please try the boot option of "nolapic_timer" ?
     If the box can be booted successfully with the option of "nolapic_timer", maybe this issue is related with the AMD C1E C-state.
     Thanks.
Comment 12 grizmio 2009-05-01 02:58:33 UTC
I have/using a compaq f756la with the MCP67
I have tried using nolapic_timer at boot and it works!

I'm using sabayon 4.1
# uname -a
Linux lisa 2.6.29-sabayon #1 SMP Fri Apr 24 08:20:50 UTC 2009 x86_64 AMD Athlon(tm) 64 X2 Dual-Core Processor TK-57 AuthenticAMD GNU/Linux

I could also boot up a Ubuntu Intrepid or Jaunty, if booted with a 2.6.24-18 kernel.
But every time i updated the kernel, ubuntu didn't boot normaly(without a key been press), until i reboot and choose at grub the 2.6.24-18.
hoping it helps and i'm not to late.
thanks for the "nolapic_timer" trick!
Comment 13 dino99 2012-01-17 10:28:46 UTC
Outdated, should be set as "wont fix" now

Note You need to log in before you can comment on or make changes to this bug.