Errors in dmesg after removing a btrfs flash disk

[Radomír Černoch]

When I remove an USB disk formatted with Btrfs, there is an error in 'dmesg', which mentions a kernel bug.

Description: Ubuntu 10.04 LTS
Release: 10.04

[80337.138374] usb 2-2: USB disconnect, address 2
[80338.337003] __ratelimit: 9 callbacks suppressed
[80338.337009] lost page write due to I/O error on sdc1
[80338.337057] end_request: I/O error, dev sdc, sector 131072
[80338.337062] lost page write due to I/O error on sdc1
[80338.337069] btrfs: 1 errors while writing supers
[80338.337107] ------------[ cut here ]------------
[80338.337111] kernel BUG at /build/buildd/linux-2.6.32/fs/btrfs/disk-io.c:2286!
[80338.337115] invalid opcode: 0000 [#1] SMP
[80338.337119] last sysfs file: /sys/devices/pci0000:00/0000:00:1d.7/class
[80338.337123] CPU 3
[80338.337126] Modules linked in: usb_storage binfmt_misc snd_hda_codec_intelhdmi snd_hda_codec_realtek ppdev fbcon tileblit font bitblit softcursor vga16fb vgastate snd_hda_intel snd_hda_codec snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_dummy snd_seq_oss arc4 snd_seq_midi snd_rawmidi snd_seq_midi_event snd_seq rt61pci crc_itu_t rt2x00pci rt2x00lib snd_timer led_class snd_seq_device i915 mac80211 drm_kms_helper snd drm i2c_algo_bit video output cfg80211 eeprom_93cx6 serio_raw xhci intel_agp soundcore snd_page_alloc lp parport usbhid hid btrfs zlib_deflate ohci1394 ieee1394 ahci crc32c r8169 mii libcrc32c
[80338.337188] Pid: 2844, comm: umount Not tainted 2.6.32-21-generic #32-Ubuntu H55M-USB3
[80338.337192] RIP: 0010:[<ffffffffa0089e64>] [<ffffffffa0089e64>] write_all_supers+0x274/0x280 [btrfs]
[80338.337215] RSP: 0018:ffff88020495dd78 EFLAGS: 00010292
[80338.337219] RAX: 000000000000003a RBX: ffff8801f012b068 RCX: 000000000000001e
[80338.337223] RDX: 0000000000000000 RSI: 0000000000000082 RDI: 0000000000000246
[80338.337226] RBP: ffff88020495ddd8 R08: 00000000ffffffff R09: 0000000000000000
[80338.337230] R10: 0000000000000005 R11: 0000000000000000 R12: 0000000000000001
[80338.337233] R13: ffff8801f012b068 R14: 0000000000000000 R15: ffff880209517800
[80338.337238] FS: 00007f3d85c50740(0000) GS:ffff8800282c0000(0000) knlGS:0000000000000000
[80338.337242] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[80338.337245] CR2: 00007f42584f06a8 CR3: 00000001dad9f000 CR4: 00000000000006e0
[80338.337249] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[80338.337253] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[80338.337257] Process umount (pid: 2844, threadinfo ffff88020495c000, task ffff8801924e44d0)
[80338.337260] Stack:
[80338.337262] ffff8801f01c0dce ffff8801f012b000 ffff880209517800 0000000000000000
[80338.337268] <0> ffff8801f01c0cb3 0000000100000001 ffff88020495ddd8 ffff880209517800
[80338.337274] <0> ffff880212fbf3f0 ffff88020495de28 ffff880212fbf460 0000000000000001
[80338.337280] Call Trace:
[80338.337299] [<ffffffffa0089e83>] write_ctree_super+0x13/0x20 [btrfs]
[80338.337318] [<ffffffffa008f259>] btrfs_commit_transaction+0x4c9/0x6f0 [btrfs]
[80338.337328] [<ffffffff81085320>] ? autoremove_wake_function+0x0/0x40
[80338.337346] [<ffffffffa008e699>] ? start_transaction+0x129/0x180 [btrfs]
[80338.337359] [<ffffffffa006eb23>] btrfs_sync_fs+0x43/0x80 [btrfs]
[80338.337366] [<ffffffff81197d9e>] sync_quota_sb+0x5e/0x130
[80338.337373] [<ffffffff81169f0a>] __sync_filesystem+0x7a/0x90
[80338.337378] [<ffffffff81169ff9>] sync_filesystems+0xd9/0x130
[80338.337384] [<ffffffff8115ece1>] sys_umount+0xb1/0xd0
[80338.337392] [<ffffffff810131b2>] system_call_fastpath+0x16/0x1b
[80338.337395] Code: 5e 41 5f c9 c3 44 89 f6 48 c7 c7 50 f1 0c a0 31 c0 e8 50 42 4b e1 0f 0b eb fe 44 89 e6 48 c7 c7 50 f1 0c a0 31 c0 e8 3b 42 4b e1 <0f> 0b eb fe 0f 1f 84 00 00 00 00 00 55 48 89 e5 0f 1f 44 00 00
[80338.337441] RIP [<ffffffffa0089e64>] write_all_supers+0x274/0x280 [btrfs]
[80338.337458] RSP <ffff88020495dd78>
[80338.337463] ---[ end trace 5dad66d8edaeb8a5 ]---

[Rocko (rockorequin)]

Does it lock the PC up? It tends to lock mine if I remove the drive before unmounting it. 2.6.34 performs a bit better, but removing the drive still makes the kernel hang eventually, requiring a reboot. It might also be related to https://bugs.launchpad.net/bugs/518638 - if you remount the btrfs drive on a different device (eg /dev/sdc), the kernel can't see it because it thinks it's still mounted on the old device (eg /dev/sdb).

I reported this kernel bug at https://bugzilla.kernel.org/show_bug.cgi?id=15325 a few months ago, but there's been no response so far.

[Radomír Černoch (radomir-cernoch)]

No, it does not. I was able to continue working flawlessly. Apart from the flash drive, I am using Btrfs on the root fs and no error has occured. Therefore I assume that the kernel was able to recover fully from this error.

[Rolf Leggewie (r0lf)]

Ubuntu has not shipped a btrfs package for a long time, moving to generic Ubuntu for triage.

This bug is pretty old. Did anyone affected by this experience it with trusty or later?

[Rolf Leggewie (r0lf)]

We'd like to figure out what's causing this bug for you, but we haven't heard back from you in a while. Could you please provide the requested information? Thanks!

[Rolf Leggewie (r0lf)]

closing as obsolete