updated to new Firefox, then huge crash could they be related?

useradded · 2011-03-22 23:39:24

Hello

I just updated to the Firefox 4, however when I launch it there is a full system, or at least X, crash and the following appears on the screen:

Mar 22 23:12:08 localhost kernel: ------------[ cut here ]------------
Mar 22 23:12:08 localhost kernel: kernel BUG at fs/ext4/inode.c:2170!
Mar 22 23:12:08 localhost kernel: invalid opcode: 0000 [#1] PREEMPT SMP 
Mar 22 23:12:08 localhost kernel: last sysfs file: /sys/devices/pci0000:00/0000:00:1a.7/usb1/1-1/1-1:1.0/host6/target6:0:0/6:0:0:0/block/sdb/dev
Mar 22 23:12:08 localhost kernel: CPU 2 
Mar 22 23:12:08 localhost kernel: Modules linked in: nls_cp437 vfat fat usb_storage uas ipv6 ext2 snd_hda_codec_hdmi joydev snd_seq_dummy snd_hda_codec_realtek snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device snd_hda_intel snd_pcm_oss 
snd_mixer_oss snd_hda_codec vboxdrv firewire_ohci r8169 snd_hwdep floppy processor iTCO_wdt snd_pcm snd_timer sg firewire_core asus_atk0110 psmouse pcspkr serio_raw snd soundcore snd_page_alloc i2c_i801 evdev mii shpchp pci_hotplug iTCO_ven
dor_support crc_itu_t ext4 mbcache jbd2 crc16 hid_logitech ff_memless usbhid sr_mod cdrom hid sd_mod pata_acpi uhci_hcd ata_piix pata_marvell libata ehci_hcd scsi_mod usbcore i915 drm_kms_helper drm i2c_algo_bit button i2c_core video output
 intel_agp intel_gtt
Mar 22 23:12:08 localhost kernel: 
Mar 22 23:12:08 localhost kernel: Pid: 1511, comm: firefox Not tainted 2.6.37-ARCH #1 System manufacturer P5Q-EM/P5Q-EM
Mar 22 23:12:08 localhost kernel: RIP: 0010:[<ffffffffa01f1a48>]  [<ffffffffa01f1a48>] ext4_da_block_invalidatepages.clone.39+0x158/0x160 [ext4]
Mar 22 23:12:08 localhost kernel: RSP: 0018:ffff88021789dbf8  EFLAGS: 00010246
Mar 22 23:12:08 localhost kernel: RAX: 8000000000000024 RBX: 00000000000002ad RCX: 000000000000000e
Mar 22 23:12:08 localhost kernel: RDX: 000000000000000e RSI: 000000000000000e RDI: ffffea0006c10f88
Mar 22 23:12:08 localhost kernel: RBP: ffff88021789dcb8 R08: 0000000000000002 R09: 0000000000000002
Mar 22 23:12:08 localhost kernel: R10: ffffea0006c10f90 R11: 0000000000000001 R12: ffff88021789dc08
Mar 22 23:12:08 localhost kernel: R13: ffff8801f6b2b9f8 R14: ffffea0006c2f410 R15: ffff88021789dce8
Mar 22 23:12:08 localhost kernel: FS:  00007f9236d94720(0000) GS:ffff8800c7b00000(0000) knlGS:0000000000000000
Mar 22 23:12:08 localhost kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
Mar 22 23:12:08 localhost kernel: CR2: 00007f3eb9168034 CR3: 00000002185a1000 CR4: 00000000000406e0
Mar 22 23:12:08 localhost kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Mar 22 23:12:08 localhost kernel: DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Mar 22 23:12:08 localhost kernel: Process firefox (pid: 1511, threadinfo ffff88021789c000, task ffff880227256f00)
Mar 22 23:12:08 localhost kernel: Stack:
Mar 22 23:12:08 localhost kernel: ffffea0000000008 0000000e1789dc58 000000000000000e 00000000a77d0788
Mar 22 23:12:08 localhost kernel: ffffea0006c2f410 ffffea0006c2f608 ffffea0006c44fe0 ffffea0006c37018
Mar 22 23:12:08 localhost kernel: ffffea0006bb4f58 ffffea0006c36e90 ffffea0006c21218 ffffea0006c44f70
Mar 22 23:12:08 localhost kernel: Call Trace:
Mar 22 23:12:08 localhost kernel: [<ffffffffa01f7b4b>] mpage_da_map_and_submit+0x2bb/0x450 [ext4]
Mar 22 23:12:08 localhost kernel: [<ffffffffa01f860d>] ext4_da_writepages+0x3dd/0x760 [ext4]
Mar 22 23:12:08 localhost kernel: [<ffffffff810f020c>] do_writepages+0x1c/0x30
Mar 22 23:12:08 localhost kernel: [<ffffffff810e7573>] __filemap_fdatawrite_range+0x53/0x60
Mar 22 23:12:08 localhost kernel: [<ffffffff810e75da>] filemap_write_and_wait_range+0x5a/0x80
Mar 22 23:12:08 localhost kernel: [<ffffffff8115e252>] vfs_fsync_range+0x52/0x90
Mar 22 23:12:08 localhost kernel: [<ffffffff8115e2f7>] vfs_fsync+0x17/0x20
Mar 22 23:12:08 localhost kernel: [<ffffffff8115e335>] do_fsync+0x35/0x60
Mar 22 23:12:08 localhost kernel: [<ffffffff8115e62e>] sys_fdatasync+0xe/0x20
Mar 22 23:12:08 localhost kernel: [<ffffffff8100bed2>] system_call_fastpath+0x16/0x1b
Mar 22 23:12:08 localhost kernel: Code: 5c 41 5d 41 5e 41 5f c9 c3 66 2e 0f 1f 84 00 00 00 00 00 4c 89 e7 48 89 95 40 ff ff ff e8 c1 fe ef e0 48 8b 95 40 ff ff ff eb c3 <0f> 0b 0f 0b 0f 1f 40 00 55 48 89 e5 48 83 ec 20 48 89 5d e8 4c 
Mar 22 23:12:08 localhost kernel: RIP  [<ffffffffa01f1a48>] ext4_da_block_invalidatepages.clone.39+0x158/0x160 [ext4]
Mar 22 23:12:08 localhost kernel: RSP <ffff88021789dbf8>
Mar 22 23:12:08 localhost kernel: ---[ end trace 5e35ab4d81785e45 ]---

If I switch Ctl-Alt-F1 and then back to F7 the X session is restored but is unusable unless I reboot.

EDIT before I even post: this thread http://forums.gentoo.org/viewtopic-t-86 … t-125.html on the Gentoo forums seems to suggest its related to instability with regard to transparent huge pages (THP), not that I know what those are.

slacknatcher · 2011-03-23 01:29:51

Transparent huge pages are a kernel option to fit blocks of memory as big as 2Mb instead of the traditional 4kb blocks on the L2 cache memory. the only software that benefits from this are usually big RDBMS (oracle, Mysql) but if you aren't gonna move big volumes of memory on demand then is safe to disable. also this option appeared on 2.6.38 kernel so this is the first official version of it, one can wait bugs.

i knew about some error with the Nvidia driver for Linux (new Firefox 4 uses hardware rendering), disable that on options and see what happens

ataraxia · 2011-03-23 01:56:59

The crash on that gentoo forum thread is in memory management, which is what THP is about. Your crash is in ext4 filesystem code. You might have some kind of corruption on your disk (or I guess, something wrong with your disk). You might try to do one of these:

- reinstall the packages for xulrunner, firefox, and whatever else you installed with them
- force a full fsck of your filesystem(s) - you can do this easily by making a file called forcefsck in your root directory (touch /forcefsck) and rebooting

useradded · 2011-03-23 08:01:22

Thanks for the replies.

I forced fsck, but it didn't find anything except some non-contiguous files. I then reinstalled Firefox and exactly the same behaviour occurred. Running Firefox from the command line produces no output, apart from the crash described above.

Every time this happens there is the same reference to inode 2170 at the top of the error output, on my system this refers to three files:

2170    0 -rw-r--r--   1 root     root         4096 Mar 23 07:24 /sys/devices/LNXSYSTM:00/device:00/PNP0C0F:00/power/async

2170  716 -rw-r--r--   1 root     root       731230 Jan 28 15:10 /usr/share/locale/sr/LC_MESSAGES/gcc.mo

2170    4 -rw-r--r--   1 james    james        1081 Mar  6 10:29 /home/james/.config/smplayer/file_settings/a/a451014d6963d0e0.ini

Root and home are on different partitions.

EDIT: Also, this is the list of packages installed with Firefox:

[2011-03-22 22:27] Running 'pacman -Syu'
[2011-03-22 22:27] synchronizing package lists
[2011-03-22 22:27] starting full system upgrade
[2011-03-22 22:29] upgraded cairo (1.10.2-1 -> 1.10.2-2)
[2011-03-22 22:29] upgraded xulrunner (1.9.2.15-2 -> 2.0-1)
[2011-03-22 22:29] upgraded firefox (3.6.15-1 -> 4.0-1)
[2011-03-22 22:29] installed gstreamer0.10 (0.10.32-1)
[2011-03-22 22:29] installed gstreamer0.10-base (0.10.32-2)
[2011-03-22 22:29] installed libwebkit (1.2.7-1)
[2011-03-22 22:29] upgraded libproxy (0.4.6-2 -> 0.4.6-4)
[2011-03-22 22:29] upgraded linux-firmware (20110201-1 -> 20110227-1)
[2011-03-22 22:29] upgraded vlc (1.1.7-5 -> 1.1.7-6)

Last edited by useradded (2011-03-23 08:06:52)

useradded · 2011-03-23 14:24:59

Well, this is getting even weirder.

Using chromium, I downloaded a nightly build of Firefox to see if I'd have any more luck with that than with the Arch package. I click on the box in the damn awful chromium download bar and I get the same kernel bug output come up - just from clicking on the download bar! It seems I am cursed to have nothing to do with Firefox 4, on this machine at least.

On my laptop, also running Arch, Firefox 4 is working no problem whatsoever, although that is formatted ext3.

This is all quite distressing as I was really looking forward to playing around with it.

useradded · 2011-03-23 20:50:50

OK,

So now I've updated my netbook, which is ext4 too, and no problems at all. This leaves me very worried that my desktop HDD has a problem.

Oh dear.

From the look of the kernel bug trace in the first post does anyone have an idea what this problem might be?

thanks

stqn · 2011-03-23 21:14:43

I'm thinking memory corruption, which could be caused by anything but probably not by the hard disk...

Edit: Hmm wait... based on this old bug report and its answer, it looks like it could be an actual bug in ext4. (I know it's most probably not the same bug, but they look similar enough...)

Edit2: typo.

Last edited by stqn (2011-03-24 09:32:39)

useradded · 2011-03-24 09:24:33

Hi stqn,

Yeah that does look pretty similar, but why would a bug in the filesystem only be triggered by one particular application?

stqn · 2011-03-24 10:09:08

I don't know... I would try reporting the bug; however, I can't find any "ext4" or even "linux" bugtracker anywhere. I found an ext4 bug reported in the Arch bugtracker, but after a short followup (asking for a fsck) it died there.

Arch Linux

#1 2011-03-22 23:39:24

updated to new Firefox, then huge crash could they be related?

#2 2011-03-23 01:29:51

Re: updated to new Firefox, then huge crash could they be related?

#3 2011-03-23 01:56:59

Re: updated to new Firefox, then huge crash could they be related?

#4 2011-03-23 08:01:22

Re: updated to new Firefox, then huge crash could they be related?

#5 2011-03-23 14:24:59

Re: updated to new Firefox, then huge crash could they be related?

#6 2011-03-23 20:50:50

Re: updated to new Firefox, then huge crash could they be related?

#7 2011-03-23 21:14:43

Re: updated to new Firefox, then huge crash could they be related?

#8 2011-03-24 09:24:33

Re: updated to new Firefox, then huge crash could they be related?

#9 2011-03-24 10:09:08

Re: updated to new Firefox, then huge crash could they be related?

Board footer