Bug 10757 (crash) - intermittant crash related to keyboard use unless acpi=off -- 2.6.24 regression - LG LW65 P4XP laptop (alviso)
Summary: intermittant crash related to keyboard use unless acpi=off -- 2.6.24 regressi...
Status: REJECTED INSUFFICIENT_DATA
Alias: crash
Product: ACPI
Classification: Unclassified
Component: Other (show other bugs)
Hardware: All Linux
: P1 blocking
Assignee: ykzhao
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-05-20 07:12 UTC by João Valente
Modified: 2009-05-05 06:45 UTC (History)
4 users (show)

See Also:
Kernel Version: 2.6.24
Subsystem:
Regression: Yes
Bisected commit-id:


Attachments
Dmesg output running kernel 2.6.22 from arch linux live cd. (19.19 KB, text/x-log)
2008-05-20 10:10 UTC, João Valente
Details
Dmesg output running updated arch system (kernel 2.6.24) (21.04 KB, text/x-log)
2008-05-20 10:11 UTC, João Valente
Details
Result of running acpidump (kernel 2.6.24) (169.34 KB, text/x-log)
2008-05-21 06:23 UTC, João Valente
Details
Result of running lspci -vxxx (kernel 2.6.24) (25.48 KB, text/x-log)
2008-05-21 06:24 UTC, João Valente
Details

Description João Valente 2008-05-20 07:12:01 UTC
Latest working kernel version: 2.6.23
Earliest failing kernel version: ?
Distribution: Arch
Hardware Environment: Laptop LG LW65 P4XP
Software Environment: Standard Arch Installation
Problem Description: Recent kernel versions (2.6.24 or up) cause my computer to crash a little while after boot-up, with no error or no apparent reason while 2.6.23 causes no trouble at all.

At the start I thought it was related to using the Tab key, but then I noticed it happened randomly while using the keyboard. Same thing happenned using latest Ubuntu's livecd.

There seems to be more people with this problem, but I can't really find it again to specify the hardware.

Steps to reproduce: Install Arch on a system like mine, get the latest software on pacman and start using the computer.
Comment 1 Adrian Bunk 2008-05-20 07:20:03 UTC
Please make two attachments containing the output of "dmesg -s 1000000" for 2.6.23 and 2.6.24 after booting.
Comment 2 João Valente 2008-05-20 10:10:06 UTC
Created attachment 16216 [details]
Dmesg output running kernel 2.6.22 from arch linux live cd.
Comment 3 João Valente 2008-05-20 10:11:43 UTC
Created attachment 16217 [details]
Dmesg output running updated arch system (kernel 2.6.24)
Comment 4 ykzhao 2008-05-20 20:30:32 UTC
Will you please capture the capture when the system crashes?

Please attach the output of acpidump and lspci -vxxx.
Thanks.
Comment 5 João Valente 2008-05-21 06:23:55 UTC
Created attachment 16228 [details]
Result of running acpidump (kernel 2.6.24)
Comment 6 João Valente 2008-05-21 06:24:55 UTC
Created attachment 16229 [details]
Result of running lspci -vxxx (kernel 2.6.24)
Comment 7 João Valente 2008-05-21 06:26:07 UTC
I don't understand what you mean with:

"Will you please capture the capture when the system crashes?"
Comment 8 ykzhao 2008-05-25 18:39:55 UTC
Sorry. 
What I mean is : please capture the screen when the system crashes.
Comment 9 ykzhao 2008-05-25 19:11:44 UTC
Hi, Valente
    From the log in comment #3 it seems that there is no error message on the kernel of 2.6.24 when the system is booted from the Live CD.
    Will you please capture the picture of the screen when the system crashes? It will be helpful to identify the root cause.
    Will you please try the boot option of "hpet=disable" and see whether the problem still exists?
    Will you please deselect cpuidle in kernel configuration and see whether the problem still exists?
    Thanks.
Comment 10 João Valente 2008-05-26 04:32:31 UTC
I don't think that a picture of the screen will help because there is no error message, the system simply stops.
It still the same way with "hpet=disable" and i now understand better when the system crashes:
1) When i hit tab;
2) When i backspace on an empty line;
3) When i press down (browsing the shell history) but ONLY if there are no commands left in history;
4) When i press the right arrow key.

I'll now do those modifications in kernel configuration.
Comment 11 ykzhao 2008-05-26 19:10:08 UTC
Does the system stil crash if you don't do the following actions:
>1) When i hit tab;
>2) When i backspace on an empty line;
>3) When i press down (browsing the shell history) but ONLY if there are no
commands left in history;
>4) When i press the right arrow key

Please login the system using ssh after the system is booted and then do the above actions to confirm whether the system still crashes.

Thanks.
Comment 12 João Valente 2008-05-28 08:29:52 UTC
Everything is doing great with ssh.
Comment 13 João Valente 2008-05-28 08:32:51 UTC
Sorry ykzhao, i didn't see your first question. No, if i don't do that actions the system works fine.
Comment 14 João Valente 2008-05-28 16:56:32 UTC
I've compiled kernel v2.6.25.4 with no CPU_IDLE and it still the same. :(
Comment 15 João Valente 2008-05-28 17:06:35 UTC
"acpi=off" boot option solved my problem, but this is not a solution, is it?
Comment 16 João Valente 2008-05-28 17:32:14 UTC
"acpi=off" boot option solved my problem, but this is not a solution, is it?
Comment 17 ykzhao 2008-05-29 00:40:25 UTC
From the acpidump it seems that one GPE is shared by the several devices(LID0, PWRB, PS2K). At the same time there exists the _PSW object for the LID0/PS2K device. In such case the _PSW object had better be called to disable the wakeup functionality of the device.

Will you please try the latest kernel (2.6.26-rc3/4) and see whether the problem still exists?
Thanks.
Comment 18 João Valente 2008-05-29 08:14:23 UTC
Do you want me to add or remove something in special to that kernel configuration?
Comment 19 ykzhao 2008-06-01 01:01:13 UTC
Please try the latest kernel(2.6.26-rc3/4). (It is unnecessary to add or remove something in kernel configuration).
Comment 20 João Valente 2008-06-01 09:13:42 UTC
It stills the same way with the new kernel (2.6.26-rc4), with cpu-idle activated. The only way to get this thing to work is with "acpi=off" boot option.
Comment 21 Len Brown 2008-06-02 18:51:37 UTC
rather than "acpi=off", is "acpi=noirq" sufficient to
make the issue go away?

how about booting with just "pci=nommconf"
Comment 22 João Valente 2008-06-08 16:32:29 UTC
None of that boot options have solved my problem.
Comment 23 Len Brown 2008-06-13 18:57:02 UTC
> 1) When i hit tab;
> 2) When i backspace on an empty line;
> 3) When i press down (browsing the shell history) but ONLY if there are no
> commands left in history;
> 4) When i press the right arrow key.

the system hangs every time you issue this sequence on the keyboard?
Does it still happen if you boot only to single user mode
and type on a VGA console?

if yes, then right before provoking the crash, please
enable verbose ACPI debugging per this page
http://www.lesswatts.org/projects/acpi/debug.php
and report the last thing printed before the hang.
Comment 24 Zhang Rui 2008-07-14 20:30:04 UTC
joao, any updates?
Comment 25 ykzhao 2008-07-22 01:57:52 UTC
Hi, Joao
    From the dmesg in comment #3 it seems that there exists the following message on the 2.6.24 kernel:
>sdhci: SDHCI controller found at 0000:06:00.4 [104c:8034] (rev 0)
>ACPI: PCI Interrupt 0000:06:00.4[A] -> GSI 16 (level, low) -> IRQ 16
>sdhci:slot0: Will use DMA mode even though HW doesn't fully claim to support
>it.
>sdhci:slot1: Will use DMA mode even though HW doesn't fully claim to support
>it.
>sdhci:slot2: Will use DMA mode even though HW doesn't fully claim to support
>it.
   >ACPI: PCI Interrupt 0000:00:1f.3[B] -> GSI 19 (level, low) -> IRQ 20 
   But on the 2.6.22 kernel there is no such message. Maybe on the 2.6.22 kernel the drivers are not loaded for the above PCI devices.

   Will you please check whether the above device can be disabled in BIOS option?(For example: SD controller, Wireless network , SMbus 1f.3 and so on).  
   If yes, please disable them and try the 2.6.26-rc3 kernel to see whether the problem still exists. 
   If no, please not load the device driver for them and see whether the problem still exists.

   Thanks.
Comment 26 ykzhao 2008-08-28 01:53:16 UTC
As there is no response for more than one month, the bug will be rejected.
If the problem still exists, please reopen it.
Comment 27 Gonçalo Silva 2009-03-16 07:16:02 UTC
Confirming this bug on my LG LW65 laptop, with every kernel after 2.6.23 and below 2.6.27 (haven't tested with newer versions).
Comment 28 ykzhao 2009-03-18 00:41:20 UTC
It seems that this issue is very strange.
   When login using SSH, there is no crash. And only when the following command is executed, it will be crashed.
   >1) When i hit tab;
   >2) When i backspace on an empty line;
   >3) When i press down (browsing the shell history) but ONLY if there are no
commands left in history;
   >4) When i press the right arrow key.
   
    Will you please try the latest kernel and do the following test?
    a. kill the process which is using /proc/acpi/event(lsof /proc/acpi/event to get the process)
    b. hit the keyboard as mentioned in comment #10 and see whether it still crashes.
    Thanks.
Comment 29 Zhang Rui 2009-03-30 07:36:47 UTC
ping Goncalo.
Comment 30 Len Brown 2009-04-01 02:05:40 UTC
Gonçalo,  João,
Is "acpi=off" still necessary to boot, even with 2.6.29?
Comment 31 Zhang Rui 2009-05-05 06:45:09 UTC
close this bug as there is no response from the bug reporter.
please reopen it if the problem still exists in the latest kernel release.

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