BTRFS kernel errors in 3.15.5

Hi,

This happens systematically these days, while trying to backup using rsync two of my laptops on an external HD - BTRFS formatted.

First at mount :

juil. 19 09:08:32 zafu kernel: BTRFS info (device dm-3): disk space caching is enabled
juil. 19 09:08:38 zafu kernel: pool[7215]: segfault at 58252fd0 ip 00007fd06e466ac7 sp 00007fd062ffbf90 error 4 in libc-2.19.so[7fd06e41e000+1a4000]

Happening on 2 different machines (both Arch Linux with latest kernel), I suspected the external HD, but it spent the night passing all long SMART tests plus a "btrfs scrub", all with flying colors...

After the following error, all accesses to the external HD lock until a forced power off of the whole system...

juil. 19 09:12:57 zafu kernel: INFO: task kworker/u4:7:6677 blocked for more than 120 seconds.
juil. 19 09:12:57 zafu kernel: Tainted: G O 3.15.5-2-ARCH #1
juil. 19 09:12:57 zafu kernel: "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
juil. 19 09:12:57 zafu kernel: kworker/u4:7 D 0000000000000000 0 6677
2 0x00000000
juil. 19 09:12:57 zafu kernel: Workqueue: btrfs-flush_delalloc normal_work_helper [btrfs]
juil. 19 09:12:57 zafu kernel: ffff88001e9dfb48 0000000000000046 ffff8800879f1460 0000000000014700
juil. 19 09:12:57 zafu kernel: ffff88001e9dffd8 0000000000014700 ffff8800879f1460 ffffffffa0262692
juil. 19 09:12:57 zafu kernel: 0000000000000000 ffffffff00000050 ffff88001e9dfbc4 ffff88001e9dfac8
juil. 19 09:12:57 zafu kernel: Call Trace:
juil. 19 09:12:57 zafu kernel: [<ffffffffa0262692>] ?

run_delalloc_range+0x192/0x350 [btrfs]
juil. 19 09:12:57 zafu kernel:  [<ffffffff8113f880>] ? filemap_fdatawait+0x30/0x30
juil. 19 09:12:57 zafu kernel:  [<ffffffff81509fa9>] schedule+0x29/0x70
juil. 19 09:12:57 zafu kernel:  [<ffffffff8150a294>] io_schedule+0x94/0xf0
juil. 19 09:12:57 zafu kernel:  [<ffffffff8113f88e>] sleep_on_page+0xe/0x20
juil. 19 09:12:57 zafu kernel:  [<ffffffff8150a788>] __wait_on_bit_lock+0x48/0xb0
juil. 19 09:12:57 zafu kernel:  [<ffffffff8113f9e8>] __lock_page+0x78/0x90
juil. 19 09:12:57 zafu kernel:  [<ffffffff810b1c80>] ? 
autoremove_wake_function+0x40/0x40
juil. 19 09:12:57 zafu kernel:  [<ffffffffa027ac98>] 
extent_write_cache_pages.isra.28.constprop.45+0x258/0x3a0 [btrfs]
juil. 19 09:12:57 zafu kernel:  [<ffffffff810a714c>] ? update_curr+0xec/0x1b0
juil. 19 09:12:57 zafu kernel:  [<ffffffff810a3068>] ? __enqueue_entity+0x78/0x80
juil. 19 09:12:57 zafu kernel:  [<ffffffff810a902e>] ? enqueue_entity+0x24e/0xaa0
juil. 19 09:12:57 zafu kernel:  [<ffffffffa027c20c>] extent_writepages+0x5c/0x90 
[btrfs]
juil. 19 09:12:57 zafu kernel:  [<ffffffffa025e9d0>] ? 
__btrfs_submit_bio_start_direct_io+0x40/0x40 [btrfs]
juil. 19 09:12:57 zafu kernel:  [<ffffffffa025d6e8>] btrfs_writepages+0x28/0x30 
[btrfs]
juil. 19 09:12:57 zafu kernel:  [<ffffffff8114d1de>] do_writepages+0x1e/0x30
juil. 19 09:12:57 zafu kernel:  [<ffffffff8114139d>] 
__filemap_fdatawrite_range+0x5d/0x80
juil. 19 09:12:57 zafu kernel:  [<ffffffff8114140c>] filemap_flush+0x1c/0x20
juil. 19 09:12:57 zafu kernel:  [<ffffffffa026046a>] 
btrfs_run_delalloc_work+0x5a/0xa0 [btrfs]
juil. 19 09:12:57 zafu kernel:  [<ffffffffa028aa77>] normal_work_helper+0x77/0x350 
[btrfs]
juil. 19 09:12:57 zafu kernel:  [<ffffffff810861d8>] process_one_work+0x168/0x450
juil. 19 09:12:57 zafu kernel:  [<ffffffff81086c32>] worker_thread+0x132/0x3e0
juil. 19 09:12:57 zafu kernel:  [<ffffffff81086b00>] ? 
manage_workers.isra.23+0x2d0/0x2d0
juil. 19 09:12:57 zafu kernel:  [<ffffffff8108d43a>] kthread+0xea/0x100
juil. 19 09:12:57 zafu kernel:  [<ffffffff8108d350>] ? 
kthread_create_on_node+0x1b0/0x1b0
juil. 19 09:12:57 zafu kernel:  [<ffffffff81515efc>] ret_from_fork+0x7c/0xb0

juil. 19 09:12:57 zafu kernel: [<ffffffff8108d350>] ? kthread_create_on_node+0x1b0/0x1b0

Swâmi Petaramesh <[email protected]> http://petaramesh.org PGP 9076E32E

Swâmi Petaramesh posted on Sat, 19 Jul 2014 09:19:13 +0200 as excerpted:

While I've not seen it, based on reports here on the list, there's apparently a known issue with rsync on btrfs for 3.15 so far. It seems to trigger most often on btrfs' with compression enabled, tho I'm not sure if compression is required to trigger it or if it simply happens more often with it.

I'm not sure of the status, but Chris Mason mentioned that he was investigating and trying to reproduce, so they're definitely actively working on it.

People seem to have better luck either reverting to the latest 3.14 series kernel for the moment, or updating to the latest 3.16-rc or live- git build, with 3.16 of course getting close to release now. I'd move forward as I know they're always fixing stuff and I routinely test kernel rcs anyway, but Marc Merlin (one of the people reporting this bug), at least, has moved back to 3.14 instead, as he says that works well for him.

Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman