Re: btrfs partition converted from ext4 becomes read-only minutes after booting: WARNING: CPU: 2 PID: 2777 at ../fs/btrfs/super.c:260 __btrfs_abort_transaction+0x4b/0x120

[Qu Wenruo <quwenruo@cn.fujitsu.com>]

Robert Munteanu wrote on 2015/06/12 15:19 +0300:
> Hi,
>
> I have converted my root ext4 partition to btrfs. I used an USB stick
> to boot and used btrfs-convert.
>
> I also did a balance and defrag ( in that order ) , both when the fs
> was mounted.
>
> After logging in to KDE I quickly get a read-only filesystem. I've
> pasted the backtrace below
>
> Jun 11 23:13:08 mars kernel: WARNING: CPU: 2 PID: 2777 at
> ../fs/btrfs/super.c:260 __btrfs_abort_transaction+0x4b/0x120 [btrfs]()
> Jun 11 23:13:08 mars kernel: BTRFS: Transaction aborted (error -95)
> Jun 11 23:13:08 mars kernel: Modules linked in: bnep bluetooth rfkill
> fuse vboxpci(O) vboxnetadp(O) vboxnetflt(O) vboxdrv(O) af_packet
> nf_log_ipv6 xt_pkttype nf_log_ip
> v4 nf_log_common xt_LOG xt_limit ip6t_REJECT xt_tcpudp
> nf_conntrack_ipv6 nf_defrag_ipv6 ip6table_raw ipt_REJECT iptable_raw
> xt_CT iptable_filter ip6table_mangle nf_con
> ntrack_netbios_ns nf_conntrack_broadcast nf_conntrack_ipv4
> nf_defrag_ipv4 ip_tables xt_conntrack nf_conntrack ip6table_filter
> ip6_tables x_tables xfs libcrc32c snd_hda
> _codec_hdmi raid1 md_mod gpio_ich ppdev iTCO_wdt iTCO_vendor_support
> coretemp snd_hda_codec_realtek snd_hda_codec_generic kvm_intel
> snd_hda_intel dm_mod kvm snd_hda_co
> ntroller snd_hda_codec snd_hwdep serio_raw pcspkr snd_pcm i2c_i801
> snd_seq joydev snd_seq_device snd_timer snd 8250_fintek parport_pc
> parport acpi_cpufreq lpc_ich
> Jun 11 23:13:08 mars kernel:  soundcore mfd_core shpchp processor
> ata_generic btrfs hid_logitech_hidpp xor raid6_pq sr_mod cdrom
> nvidia_uvm(PO) nvidia(PO) firewire_ohc
> i firewire_core crc_itu_t uas usb_storage r8169 mii pata_jmicron
> hid_logitech_dj drm button sg
> Jun 11 23:13:08 mars kernel: CPU: 2 PID: 2777 Comm: kworker/u8:0
> Tainted: P           O    4.0.4-3-desktop #1
> Jun 11 23:13:08 mars kernel: Hardware name: Gigabyte Technology Co.,
> Ltd. EP35-DS4/EP35-DS4, BIOS F6d 01/08/2009
> Jun 11 23:13:08 mars kernel: Workqueue: btrfs-endio-write
> btrfs_endio_write_helper [btrfs]
> Jun 11 23:13:08 mars kernel:  0000000000000000 ffffffffa0a92832
> ffffffff8167c4aa ffff880128513ca8
> Jun 11 23:13:08 mars kernel:  ffffffff81063bb1 ffff880031929d28
> ffff880221e71800 00000000ffffffa1
> Jun 11 23:13:08 mars kernel:  ffffffffa0a914e0 0000000000000b50
> ffffffff81063c2a ffffffffa0a95928
> Jun 11 23:13:08 mars kernel: Call Trace:
> Jun 11 23:13:08 mars kernel:  [<ffffffff8100574c>] dump_trace+0x8c/0x340
> Jun 11 23:13:08 mars kernel:  [<ffffffff81005aa3>] show_stack_log_lvl+0xa3/0x190
> Jun 11 23:13:08 mars kernel:  [<ffffffff81007201>] show_stack+0x21/0x50
> Jun 11 23:13:08 mars kernel:  [<ffffffff8167c4aa>] dump_stack+0x47/0x67
> Jun 11 23:13:08 mars kernel:  [<ffffffff81063bb1>]
> warn_slowpath_common+0x81/0xb0
> Jun 11 23:13:08 mars kernel:  [<ffffffff81063c2a>] warn_slowpath_fmt+0x4a/0x50
> Jun 11 23:13:08 mars kernel:  [<ffffffffa09e598b>]
> __btrfs_abort_transaction+0x4b/0x120 [btrfs]
> Jun 11 23:13:08 mars kernel:  [<ffffffffa0a1d18a>]
> btrfs_finish_ordered_io+0x5aa/0x620 [btrfs]
> Jun 11 23:13:08 mars kernel:  [<ffffffffa0a43253>]
> normal_work_helper+0xc3/0x320 [btrfs]
> Jun 11 23:13:08 mars kernel:  [<ffffffff8107bcf2>] process_one_work+0x142/0x420
> Jun 11 23:13:08 mars kernel:  [<ffffffff8107c0e4>] worker_thread+0x114/0x460
> Jun 11 23:13:08 mars kernel:  [<ffffffff81081261>] kthread+0xc1/0xe0
> Jun 11 23:13:08 mars kernel:  [<ffffffff81682d58>] ret_from_fork+0x58/0x90
> Jun 11 23:13:08 mars kernel: ---[ end trace 4c4eb7d6e98afa91 ]---
> Jun 11 23:13:08 mars kernel: BTRFS: error (device sda1) in
> btrfs_finish_ordered_io:2896: errno=-95 unknown

IIRC some one in the mail-list has reported the same bug.
Still not sure the root cause but seems highly related to converted fs.

It would be much better if you have a clue to trigger the bug.
Like read/write which file(s) may cause the bug.

If it's OK for you, please upload the btrfs-debug-tree output.
WARNING: This output will not contain any data but all your filename/dir 
name.

My first guess is some btrfs codes can't handle the special extent 
converted from ext* well, but still quite hard to say even the 
errno(EOPNOTSUPP) is quite unique and easy to find the source.... :(

Thanks,
Qu

> Jun 11 23:13:08 mars kernel: BTRFS info (device sda1): forced readonly
>
> Some diagnostic info:
>
> - btrfs scrub reports no errors
> - on the host machine I'm running btrfs v4.0+20150429 and kernel 4.0.4-3-desktop
> - on the live medium, used to run btrfs-convert, I was running btrfs
> v4.0+20150429 and kernel 4.0.3-1-default
>
> # btrfs fi show
> Label: none  uuid: 54dea125-74cd-4bb2-86a2-f7bc645b76cf
>          Total devices 1 FS bytes used 90.22GiB
>          devid    1 size 223.57GiB used 92.03GiB path /dev/sda1
>
> btrfs-progs v4.0+20150429
>
> # btrfs fi df /
> Data, single: total=89.00GiB, used=88.17GiB
> System, single: total=32.00MiB, used=16.00KiB
> Metadata, single: total=3.00GiB, used=2.05GiB
> GlobalReserve, single: total=512.00MiB, used=0.00B
>
> Is there a way out? I still have the old ext4 image and can revert,
> but I'm keeping the btrfs one for now, in case I can extract some
> useful debugging information from it.
>
> Thanks,
>
> Robert
>
>