Bug 166955 – Corrupted page table while generating UTF-8 locales.

• Product: Gentoo Linux
• Status: VERIFIED NEEDINFO

[Description Robert Gill 2007-02-15 03:26:58 UTC]

Since upgrading to glibc-2.5 my system has been crashing with a Corrupted page table error when trying to generate UTF-8 locales.  This does not occur with other locales.  Sometimes the system recovers, sometimes it locks up completely.  I've tried it with kernels gentoo-sources-2.6.17-r8 and gentoo-sources-2.6.18-r6 with the same results.  I'm going to try rebuild one of those kernels without SMP support.  I'll post the results in another comment.


Reproducible: Always

Steps to Reproduce:
1.run locale-gen with any UTF-8 locale specified in /etc/locale.gen



Actual Results:  
locale-gen segfaults or system locks up

Expected Results:  
locales should be generated.

error log:
Feb 14 19:08:27 onyx localedef: Corrupted page table at address 2b63e7a51000
Feb 14 19:08:27 onyx PGD 64a49067 PUD 683ea067 PMD 3fb78067 PTE 800062003f78a067
Feb 14 19:08:27 onyx Bad pagetable: 000f [1] PREEMPT SMP 
Feb 14 19:08:27 onyx CPU 1 
Feb 14 19:08:27 onyx Modules linked in: vmnet parport_pc parport vmmon nvidia
Feb 14 19:08:27 onyx Pid: 5791, comm: localedef Tainted: PF     2.6.17-gentoo-r8 #1
Feb 14 19:08:27 onyx RIP: 0033:[<00002b63e6a4566a>] [<00002b63e6a4566a>]
Feb 14 19:08:27 onyx RSP: 002b:00007fffc41f0f18  EFLAGS: 00010216
Feb 14 19:08:27 onyx RAX: 00002b63e6c0e040 RBX: 0000000000083afe RCX: 000000000000ca0f
Feb 14 19:08:27 onyx RDX: 00000000014137b0 RSI: 0000000000000000 RDI: 00002b63e7a50fc0
Feb 14 19:08:27 onyx RBP: 00002b63e842f040 R08: 0000000000080000 R09: 0000000000000000
Feb 14 19:08:27 onyx R10: 0000000000000022 R11: 0000000000000246 R12: 00000000005496e8
Feb 14 19:08:27 onyx R13: 00000000000019e7 R14: 0000000000041d73 R15: 00002b63e842f040
Feb 14 19:08:27 onyx FS:  00002b63e6c0d6d0(0000) GS:ffff810002f2a4c0(0000) knlGS:00000000f7c6d6b0
Feb 14 19:08:27 onyx CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
Feb 14 19:08:27 onyx CR2: 00002b63e7a51000 CR3: 000000006e473000 CR4: 00000000000006e0
Feb 14 19:08:27 onyx Process localedef (pid: 5791, threadinfo ffff810062c1e000, task ffff81007f475900)
Feb 14 19:08:27 onyx 
Feb 14 19:08:27 onyx RIP [<00002b63e6a4566a>] RSP <00007fffc41f0f18>

[Comment 1 Robert Gill 2007-02-15 03:28:05 UTC]

Created attachment 110222 [details]
emerge --info

[Comment 2 Robert Gill 2007-02-15 03:29:02 UTC]

Created attachment 110224 [details]
lspci

[Comment 3 Robert Gill 2007-02-15 03:29:37 UTC]

Created attachment 110225 [details]
kernel config

[Comment 4 Jakub Moc (RETIRED) 2007-02-15 07:49:00 UTC]

Try with 2.6.20 and report back please.

[Comment 5 Robert Gill 2007-02-23 06:36:23 UTC]

Same issue with 2.6.20.  I think it may be a hardware problem, though I'm not exactly sure what.  Memtest86 and a few hardware stress test utilities come up clean (app-benchmarks/stress and prime95).  I found it doesn't just occur with locale-gen though.  It's also happened while sorting or filtering large newsgroups in pan.

[Comment 6 Jakub Moc (RETIRED) 2007-02-23 08:05:27 UTC]

Hmmm...

[Comment 7 Daniel Drake (RETIRED) 2007-02-24 17:53:25 UTC]

Your kernel is tainted. Please reproduce this without the nvidia and vmware modules loaded (and make sure they weren't loaded then unloaded, it has to be clean) and post a new crash dump.

Also make sure your kernel is compiled with CONFIG_KALLSYMS and that you post the crash from dmesg as opposed to syslog.

[Comment 8 Daniel Drake (RETIRED) 2007-03-10 04:47:19 UTC]

see comment #7

[Comment 9 Robert Gill 2007-03-24 18:31:20 UTC]

Sorry, I was out of town for a while.  Anyways, I replaced my motherboard which solved the problem.