Bug 76661 - Crash on poweroff or reboot since Linux-3.13 - 32-bit core i3-4330T CPU
Summary: Crash on poweroff or reboot since Linux-3.13 - 32-bit core i3-4330T CPU
Status: VERIFIED CODE_FIX
Alias: None
Product: Platform Specific/Hardware
Classification: Unclassified
Component: i386 (show other bugs)
Hardware: x86-64 Linux
: P1 normal
Assignee: Fenghua Yu
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-05-21 18:04 UTC by Bernd Kuhls
Modified: 2016-02-20 09:53 UTC (History)
4 users (show)

See Also:
Kernel Version: 3.14.4
Tree: Mainline
Regression: Yes


Attachments
dmesg (39.73 KB, text/plain)
2014-05-21 18:16 UTC, Bernd Kuhls
Details
lsmod (2.31 KB, text/plain)
2014-05-21 18:17 UTC, Bernd Kuhls
Details
lspci (1.73 KB, text/plain)
2014-05-21 18:17 UTC, Bernd Kuhls
Details
.config (117.20 KB, text/plain)
2014-05-23 04:21 UTC, Bernd Kuhls
Details
screenshot of kernel oops (124.50 KB, image/jpeg)
2014-05-23 04:22 UTC, Bernd Kuhls
Details
patch used with 3.14.4, applies also to 3.15-rc8 (1.51 KB, patch)
2014-06-03 16:46 UTC, Bernd Kuhls
Details | Diff
/proc/cpuinfo (4.72 KB, text/plain)
2014-06-11 05:00 UTC, Bernd Kuhls
Details
Bug fix patch: Disable local irq before disabling IO APIC (1.88 KB, patch)
2014-06-11 22:26 UTC, Fenghua Yu
Details | Diff

Description Bernd Kuhls 2014-05-21 18:04:20 UTC
Overview:
until kernel 3.11.x poweroff or reboot worked fine, since 3.13 it crashes

Steps to Reproduce:
halt, poweroff or reboot

Actual Results:
the kernel crashed, screenshot attached

Expected Results:
halt, poweroff or reboot of the machine

Hardware:
Intel(R) Core(TM) i3-4330T CPU @ 3.00GHz
AsRock Z87 Extreme4, BIOS P2.70 01/21/2014

I already tried to backport some patches from 3.15-rc, without success
https://bugzilla.kernel.org/show_bug.cgi?id=69791#c12
Comment 1 Bernd Kuhls 2014-05-21 18:16:44 UTC
Created attachment 137011 [details]
dmesg
Comment 2 Bernd Kuhls 2014-05-21 18:17:14 UTC
Created attachment 137021 [details]
lsmod
Comment 3 Bernd Kuhls 2014-05-21 18:17:36 UTC
Created attachment 137031 [details]
lspci
Comment 4 Bernd Kuhls 2014-05-21 21:30:45 UTC
Adding "xhci_hcd.quirks=270336", as suggested here:
https://bugzilla.kernel.org/show_bug.cgi?id=66171#c67
did not help.

Using 3.15-rc5 shows the same problem.

Uploading larger attachments to this bug report, like the screenshot of the kernel oops and .config seems, seems to broken atm.
Comment 5 Bernd Kuhls 2014-05-23 04:21:48 UTC
Created attachment 137191 [details]
.config
Comment 6 Bernd Kuhls 2014-05-23 04:22:45 UTC
Created attachment 137201 [details]
screenshot of kernel oops
Comment 7 Len Brown 2014-06-02 20:28:56 UTC
What is the earliest kernel version that fails?
You mentioned 3.11 works and 3.13 fails -- does 3.12 work?

What do you see when you boot with maxcpus=1?

I expect this hardware supports 64-bit -- do you see the issue
also with a 64-bit kernel?
Comment 8 Bernd Kuhls 2014-06-02 21:02:27 UTC
The distro I am using (32bit userspace only, no 64bit kernel provided) bumped the kernel from 3.11 to 3.13, I did not use 3.12.

Setting up bisecting will not be easy, but I will try if no other option remains.

Using maxcpus=1 did not solve the bug, neither removing ?hci-*.ko usb modules.
Comment 9 Zhang Rui 2014-06-03 07:30:40 UTC
please check if reverting commit 522e66464467543c0d88d023336eec4df03ad40b helps?
Comment 10 Bernd Kuhls 2014-06-03 16:43:36 UTC
(In reply to Zhang Rui from comment #9)
> please check if reverting commit 522e66464467543c0d88d023336eec4df03ad40b
> helps?

Yes, reverting this commit fixes the crash during reboot and poweroff, thanks!
Comment 11 Bernd Kuhls 2014-06-03 16:46:06 UTC
Created attachment 137961 [details]
patch used with 3.14.4, applies also to 3.15-rc8
Comment 12 Bernd Kuhls 2014-06-03 16:54:40 UTC
An additional info: Another user using the same distro reported the same problem using an Intel D2500CC board: http://lists.spline.de/pipermail/fli4l_dev/2014-May/001746.html (German)
Comment 13 Fenghua Yu 2014-06-09 23:36:27 UTC
Is this issue reproducible on any board with 32-bit kernel? Is this issue reproducible every time on your board? I can not reproduce on a Thinkpad machine.
Comment 14 Bernd Kuhls 2014-06-10 06:33:23 UTC
Hi,

(In reply to Fenghua Yu from comment #13)
> Is this issue reproducible on any board with 32-bit kernel?

In total there were three reports of this bug, including mine:
http://lists.spline.de/pipermail/fli4l_dev/2014-May/001746.html (Intel D2500CC)
http://lists.spline.de/pipermail/fli4l_dev/2014-June/001815.html (ALIX APU1C)

> Is this issue reproducible every time on your board?

Yes, since Jan 31st, 2014, when the distro changed from 3.11 to 3.13, the crash happened always, 100% reproducible, until I applied the patch suggested by Zhang Rui, which also fixed the problem for the other two reporters.
Comment 15 Bernd Kuhls 2014-06-11 05:00:01 UTC
Created attachment 139031 [details]
/proc/cpuinfo
Comment 16 Bernd Kuhls 2014-06-11 05:01:53 UTC
Some infos about my CPU:
http://ark.intel.com/products/77770/Intel-Core-i3-4330T-Processor-4M-Cache-3_00-GHz
Comment 17 Fenghua Yu 2014-06-11 22:26:44 UTC
Created attachment 139351 [details]
Bug fix patch: Disable local irq before disabling IO APIC

This bug fix patch which disables local irq before disabling IO APIC during reboot is supposed to fix the issue here. It was tested and verified by Bernd. A new German 32-bit Linux distro will be released on this Friday to include this bug fix patch and will enable more German users to test it. After this patch is fully verified, I will publish this patch to lkml.
Comment 18 Fenghua Yu 2014-06-12 02:43:12 UTC
I really appreciate it if anyone can test this patch in any test enviornment to provide wider coverage.
Comment 19 Bernd Kuhls 2014-06-13 10:38:44 UTC
(In reply to Fenghua Yu from comment #17)
> Created attachment 139351 [details]
> Bug fix patch: Disable local irq before disabling IO APIC

Patch works fine and solves the problem, thanks!
Comment 20 Bernd Kuhls 2014-06-16 15:23:21 UTC
(In reply to Bernd Kuhls from comment #14)

> In total there were three reports of this bug, including mine:
> http://lists.spline.de/pipermail/fli4l_dev/2014-May/001746.html (Intel
> D2500CC)

The forementioned bug reporter writes that the patch from comment #17 works fine for him and solves the reboot problem:
http://lists.spline.de/pipermail/fli4l_dev/2014-June/001884.html
Comment 21 Glenn Williamson 2014-06-19 20:44:22 UTC
Has this patch been pushed to LKML.org?
Comment 22 Fenghua Yu 2014-06-21 04:59:13 UTC
The patch is in lkml now: https://lkml.org/lkml/2014/6/21/4
Comment 23 Bernd Kuhls 2014-10-17 21:30:57 UTC
ping, any update about the inclusion of the patch into the kernel?
Comment 24 Bernd Kuhls 2016-02-20 09:53:00 UTC
ping, any update about the inclusion of the patch into the kernel?

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