Eeek! page_mapcount(page) went negative! > (-1)

[agraham]

My machine just locked up, /var/log/messages show a single line message

Aug 1 20:01:32 agraham kernel: Eeek! page_mapcount(page) went negative! (-1)

uname -a
Linux agraham 2.6.17-1.2157_FC5smp #1 SMP Tue Jul 11 23:24:16 EDT 2006 i686 i686 i386 GNU/Linux

Also, I've started getting a lot of these "BUG: soft lockup detected on CPU#0!".

I've never had problems before and don't believe this to be a hardware issue.

I posted this to the Fedora User mailing list and they suggested that I send you a nvidia-bug-report.log which is attached.

Any Ideas ?

Albert.

The following is the log from a "BUG: soft lockup" I get today.


Aug 1 14:34:40 agraham kernel: BUG: soft lockup detected on CPU#0!
Aug 1 14:34:40 agraham kernel: <c044a982> softlockup_tick+0xad/0xc4 <c042d850> update_process_times+0x39/0x5c
Aug 1 14:34:40 agraham kernel: <c0418af3> smp_apic_timer_interrupt+0x5a/0x63 <c040490f> apic_timer_interrupt+0x1f/0x24
Aug 1 14:34:40 agraham kernel: <c044ef08> find_get_pages+0x51/0x59 <c0453dd8> pagevec_lookup+0x1c/0x22
Aug 1 14:34:40 agraham kernel: <c0454192> invalidate_mapping_pages+0xa1/0xb3 <c046cc6f> invalidate_bh_lru+0x0/0x3b
Aug 1 14:34:40 agraham kernel: <c0429aa5> on_each_cpu+0x1f/0x27 <c047234b> kill_bdev+0xd/0x20
Aug 1 14:34:40 agraham kernel: <c04723d2> set_blocksize+0x74/0x80 <c04723ef> sb_set_blocksize+0x11/0x34
Aug 1 14:34:40 agraham kernel: <f894698c> ext3_fill_super+0x11c/0x1556 [ext3] <c043891c> __mutex_init+0x30/0x48
Aug 1 14:34:40 agraham kernel: <c04e94e3> snprintf+0x1f/0x22 <c04a36f0> disk_name+0x30/0x88
Aug 1 14:34:40 agraham kernel: <c0471ad3> get_sb_bdev+0xcd/0x116 <c0484827> alloc_vfsmnt+0x97/0xbe
Aug 1 14:34:40 agraham kernel: <f8945160> ext3_get_sb+0x18/0x1c [ext3] <f8946870> ext3_fill_super+0x0/0x1556 [ext3]
Aug 1 14:34:40 agraham kernel: <c04718c3> do_kern_mount+0x8a/0x131 <c0485fb5> do_mount+0x6cc/0x726
Aug 1 14:34:40 agraham kernel: <c0451639> get_page_from_freelist+0x2a8/0x411 <c04049d7> error_code+0x4f/0x54
Aug 1 14:34:40 agraham kernel: <c0484c4e> copy_mount_options+0x26/0x109 <c0486086> sys_mount+0x77/0xae
Aug 1 14:34:40 agraham kernel: <c0403e3f> syscall_call+0x7/0xb

[netllama]

I don't see any references to the nvidia driver in the output that you posted. Was that the full output?

Is there any reliable way to trigger this crash?

Also, can you generate and post an nvidia-bug-report.log?

Thanks,
Lonni

[agraham]

I tried to upload and it says it's done, but it does not display anywhere!

Actually, I've just spotted the error:

nvidia-bug-report.log:
Your file of 141.6 KB bytes exceeds the forum's limit of 100.0 KB for this filetype.

You can download it from here:

http://www.g-b.net/nvidia-installer.log

http://www.g-b.net/nvidia-bug-report.log
Albert.

PS. I hate message boards.

[agraham]

Originally Posted by netllama

I don't see any references to the nvidia driver in the output that you posted. Was that the full output?

Is there any reliable way to trigger this crash?

Also, can you generate and post an nvidia-bug-report.log?

Thanks,
Lonni

There is no more to the log, it's a single line, it has occured 3 times:
and the machine locks up: see this log, nothing for 1 hour before and then reboots

Aug 1 18:42:28 agraham dhcpd: DHCPACK on 10.0.0.230 to 00:06:5b:d5:5a:42 via eth0
Aug 1 20:01:32 agraham kernel: Eeek! page_mapcount(page) went negative! (-1)
Aug 1 20:11:00 agraham syslogd 1.4.1: restart (remote reception).

-- Other times

Aug 1 20:01:32 agraham kernel: Eeek! page_mapcount(page) went negative! (-1)
Jul 20 14:39:41 agraham kernel: Eeek! page_mapcount(page) went negative! (-1)
Jul 21 18:10:57 agraham kernel: Eeek! page_mapcount(page) went negative! (-1)

Albert.

[netllama]

For future reference, you can zip the bug report to attach it here.

I have a few questions regarding the bug report though:
0) Just to confirm, this crash is random, with no clear steps to trigger it?
1) Does this also reproduce if you are using a kernel.org 2.6.17.x kernel?
2) Does this only reproduce with two displays (CRT + TV)? If so, does it reproduce if you are using TwinView instead of separate X screens?
3) Does this reproduce if you set NvAGP to 0?
4) Why are you using the "NoPowerConnectorCheck" and "NoBandWidthTest" options?
5) Does booting with the acpi=off and/or noapic kernel parameters have any impact?

Thanks,
Lonni