You are not logged in.

#1 2014-07-19 09:48:47

lynix
Member
From: Karlsruhe, Germany
Registered: 2008-04-23
Posts: 230

[upstream] kernel: kworker (btrfs) blocking since 3.15

I'm getting messages like the following from dmesg when running I/O intensive programs:

[ 4557.652716] INFO: task kworker/u8:5:28827 blocked for more than 120 seconds.
[ 4557.652721]       Tainted: P           O  3.15.5-2-ARCH #1
[ 4557.652723] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 4557.652725] kworker/u8:5    D 0000000000000000     0 28827      2 0x00000000
[ 4557.652734] Workqueue: writeback bdi_writeback_workfn (flush-btrfs-2)
[ 4557.652737]  ffff8801c85879b8 0000000000000046 ffff8800a2a11460 0000000000014700
[ 4557.652741]  ffff8801c8587fd8 0000000000014700 ffff8800a2a11460 ffffffffa16ed692
[ 4557.652744]  0000000000000000 ffffffff00000050 ffff8801c8587a34 ffff8801c8587938
[ 4557.652747] Call Trace:
[ 4557.652766]  [<ffffffffa16ed692>] ? run_delalloc_range+0x192/0x350 [btrfs]
[ 4557.652773]  [<ffffffff8113f880>] ? filemap_fdatawait+0x30/0x30
[ 4557.652778]  [<ffffffff81509fa9>] schedule+0x29/0x70
[ 4557.652781]  [<ffffffff8150a294>] io_schedule+0x94/0xf0
[ 4557.652785]  [<ffffffff8113f88e>] sleep_on_page+0xe/0x20
[ 4557.652789]  [<ffffffff8150a788>] __wait_on_bit_lock+0x48/0xb0
[ 4557.652793]  [<ffffffff8113f9e8>] __lock_page+0x78/0x90
[ 4557.652797]  [<ffffffff810b1c80>] ? autoremove_wake_function+0x40/0x40
[ 4557.652811]  [<ffffffffa1705c98>] extent_write_cache_pages.isra.28.constprop.45+0x258/0x3a0 [btrfs]
[ 4557.652824]  [<ffffffffa170720c>] extent_writepages+0x5c/0x90 [btrfs]
[ 4557.652835]  [<ffffffffa16e99d0>] ? __btrfs_submit_bio_start_direct_io+0x40/0x40 [btrfs]
[ 4557.652845]  [<ffffffffa16e86e8>] btrfs_writepages+0x28/0x30 [btrfs]
[ 4557.652849]  [<ffffffff8114d1de>] do_writepages+0x1e/0x30
[ 4557.652853]  [<ffffffff811df3a0>] __writeback_single_inode+0x40/0x2b0
[ 4557.652856]  [<ffffffff811e08aa>] writeback_sb_inodes+0x26a/0x430
[ 4557.652859]  [<ffffffff811e0b0f>] __writeback_inodes_wb+0x9f/0xd0
[ 4557.652863]  [<ffffffff811e0d6b>] wb_writeback+0x22b/0x360
[ 4557.652866]  [<ffffffff811ce451>] ? get_nr_inodes+0x51/0x70
[ 4557.652870]  [<ffffffff811e133f>] bdi_writeback_workfn+0x33f/0x4c0
[ 4557.652875]  [<ffffffff810861d8>] process_one_work+0x168/0x450
[ 4557.652878]  [<ffffffff81086c32>] worker_thread+0x132/0x3e0
[ 4557.652882]  [<ffffffff81086b00>] ? manage_workers.isra.23+0x2d0/0x2d0
[ 4557.652886]  [<ffffffff8108d43a>] kthread+0xea/0x100
[ 4557.652889]  [<ffffffff8108d350>] ? kthread_create_on_node+0x1b0/0x1b0
[ 4557.652895]  [<ffffffff81515efc>] ret_from_fork+0x7c/0xb0
[ 4557.652898]  [<ffffffff8108d350>] ? kthread_create_on_node+0x1b0/0x1b0

After the first occurance the corresponding program is blocked permanently, not even KILLable.

As there is always btrfs_writepages() in the call trace I conclude it must be some kind of btrfs-related problem.
Using the kernel from linux-lts (currently 3.14.12-1) avoids it, so I further conclude it must be specific to 3.15.

Can anyone confirm similar behaviour or might I be running into a hardware problem here?

Last edited by lynix (2014-07-19 18:47:25)

Offline

#2 2014-07-19 13:03:11

Scimmia
Fellow
Registered: 2012-09-01
Posts: 11,636

Re: [upstream] kernel: kworker (btrfs) blocking since 3.15

Offline

#3 2014-07-19 18:47:41

lynix
Member
From: Karlsruhe, Germany
Registered: 2008-04-23
Posts: 230

Re: [upstream] kernel: kworker (btrfs) blocking since 3.15

Thx, "closed".

Offline

Board footer

Powered by FluxBB