Bug 16097

Summary: 3.0.4 on Samsung P460: reset after "Waiting for /dev to be fully populated"
Product: Platform Specific/Hardware Reporter: Maciej Rutecki (maciej.rutecki)
Component: x86-64Assignee: platform_x86_64 (platform_x86_64)
Status: ASSIGNED ---    
Severity: normal CC: acpi-bugzilla, alan, harald.dunkel, harri, lenb, maciej.rutecki, mjg59-kernel, rjw
Priority: P1    
Hardware: All   
OS: Linux   
Kernel Version: 3.0.4 Subsystem:
Regression: Yes Bisected commit-id:
Bug Depends on:    
Bug Blocks: 15310    
Attachments: acpidump
acpidump for P460 model AA01DE

Description Maciej Rutecki 2010-06-01 20:07:21 UTC
Subject    : 2.6.34 on Samsung P460: reset after "Waiting for /dev to be fully populated"
Submitter  : Harald Dunkel <harald.dunkel@aixigo.de>
Date       : 2010-05-25 9:12
Message-ID : 4BFB947E.9080509@aixigo.de
References : http://marc.info/?l=linux-kernel&m=127477877432254&w=2

This entry is being used for tracking a regression from 2.6.33.  Please don't
close it until the problem is fixed in the mainline.
Comment 1 Harald Dunkel 2010-08-03 08:41:54 UTC
How comes that this bug report has been closed? Is it fixed?

I just figured out that loading the "processor" acpi module triggers the problem (using 2.6.35).
Comment 2 Rafael J. Wysocki 2010-08-03 09:19:47 UTC
It was closed due to the lack of response.  Reopened now.
Comment 3 Harald Dunkel 2010-08-04 06:54:10 UTC
Sorry for the delay. Here is some more data:

Its a Samsung P460 model AA03DE, firmware version 07LQ. 

I actually tried 3 laptops of this kind. All show the problem, so I doubt its a hardware failure. 

An old P460 model AA01DE boots fine. Unfortunately I am not allowed to change the BIOS of this one.
Comment 4 Harald Dunkel 2010-08-04 07:24:00 UTC
Created attachment 27339 [details]
acpidump
Comment 5 Matthew Garrett 2010-08-04 13:32:07 UTC
Are you able to also provide the acpidump output for the AA01DE?
Comment 6 Harald Dunkel 2010-08-04 14:12:29 UTC
Created attachment 27348 [details]
acpidump for P460 model AA01DE
Comment 7 Harald Dunkel 2010-08-04 14:34:38 UTC
Attached. Hope this helps.
Comment 8 Matthew Garrett 2010-08-04 14:54:08 UTC
It looks like the AA03DE introduces an extra SMM trap related to the processor code, but it's not obvious to me why this would make a difference. I know it's a significant amount of work, but would it be possible for you to bisect between the last working kernel and 2.6.34? I suspect that you can limit it to drivers/acpi.
Comment 9 Harald Dunkel 2010-08-05 05:13:34 UTC
I will check. I found an old Knoppix CD of 2004 to start with.

Wasn't there a note somewhere about some laptops changing the processor state in SMM, bypassing the operating system? Unfortunately I haven't found it anymore.
Comment 10 Harald Dunkel 2010-08-05 08:39:28 UTC
Booting old CDs there is one thing I found out: Seems that this problem is restricted to amd64. The i386 kernel included in Debian Testing works fine on the AA03DE. The amd64 version doesn't.

The AA01DE laptop works with both amd64 and i386.
Comment 11 Harald Dunkel 2010-08-06 07:06:29 UTC
(In reply to comment #8)
> I know it's
> a significant amount of work, but would it be possible for you to bisect
> between the last working kernel and 2.6.34? I suspect that you can limit it
> to
> drivers/acpi.

2.6.24 (found on Ubuntu 8.04.4) still doesn't work. The CPU is of 2009, so I doubt that this bisect thing will reveal something new.
Comment 12 Len Brown 2010-09-28 21:00:21 UTC
is there any version of Linux that boots the processor module
on this machine?  If not, we should clear the regression flag.

on the failing system, please try booting with "idle=poll".
Comment 13 Harald Dunkel 2010-09-30 07:27:49 UTC
By now I haven't found any kernel without this problem yet.

"idle=poll" seems to work for 2.6.35.5 (amd64). The processor module is loaded. On the next reboot the laptop gets stuck in the bios, but I wouldn't blame Linux for this. I have to turn it off and on again.

If I omit the "idle=poll", then the laptop does a reset on loading the processor module, as before.

Hope this helps.
Comment 14 Harald Dunkel 2011-02-28 10:08:40 UTC
I provided the requested information 5 months ago. Any news about this?
Comment 15 Harald Dunkel 2011-09-05 10:47:51 UTC
Update: Using 3.0.4 "idle=poll" doesn't help anymore.