On Sun, Oct 5, 2014 at 4:29 PM, Marc MERLIN <marc@xxxxxxxxxxx> wrote:
Deadlocks have been less frequent (good), but here is one.
An rsync from 5 days ago got stuck on btrfs it seems, and things just
started piling up on top until the system deadlocked
I see btrfs-transaction saying wait on page, but if it's RAM, I had
plenty left:
total used free shared buffers
cached
Mem: 7894580 6950108 944472 0 40
4816148
-/+ buffers/cache: 2133920 5760660
Swap: 15616764 767004 14849760
Here's the trace:
SysRq : Show Blocked State
task PC stack pid father
md8_raid5 D ffff88017028cb80 0 675 2 0x00000000
ffff88020fd67aa8 0000000000000046 ffffffff812f1799 ffff88020fd67fd8
ffff880037228410 00000000000140c0 ffff88021e3940c0 ffff880037228410
ffff8801f5579bf0 0000000000000004 ffff880211ad07c8 ffff88020fd67ab8
Call Trace:
[<ffffffff812f1799>] ? blk_flush_plug_list+0x1bc/0x1cb
[<ffffffff8163a889>] schedule+0x6e/0x70
[<ffffffff8163aa2b>] io_schedule+0x60/0x7a
[<ffffffff812ef8f6>] get_request+0x4b8/0x56a
[<ffffffff8130ba01>] ? cfq_merge+0x49/0x9e
[<ffffffff81087b70>] ? finish_wait+0x65/0x65
[<ffffffff812f1921>] blk_queue_bio+0x179/0x262
[<ffffffff812eec2c>] generic_make_request+0x9c/0xdb
[<ffffffffc042b31d>] handle_stripe+0x1e41/0x2166 [raid456]
[<ffffffff81017316>] ? ___preempt_schedule+0x56/0xa8
[<ffffffff8163d0d9>] ? _raw_spin_unlock_irqrestore+0x1f/0x32
[<ffffffffc042b925>] handle_active_stripes.isra.22+0x2e3/0x359
[raid456]
[<ffffffff8151d35e>] ? md_wakeup_thread+0x55/0x58
[<ffffffffc042bdd6>] raid5d+0x330/0x428 [raid456]
[<ffffffff810795b1>] ? get_parent_ip+0xd/0x3c
[<ffffffff815202f9>] md_thread+0x11c/0x13a
[<ffffffff81087b70>] ? finish_wait+0x65/0x65
[<ffffffff815201dd>] ? bb_store+0x55/0x55
[<ffffffff8106e3a9>] kthread+0xae/0xb6
[<ffffffff8106e2fb>] ? __kthread_parkme+0x61/0x61
[<ffffffff8163d8fc>] ret_from_fork+0x7c/0xb0
[<ffffffff8106e2fb>] ? __kthread_parkme+0x61/0x61
This trace shows we're stuck somewhere different from the 3.15 stalls.
md is waiting for a request, and unfortunately those are outside of
btrfs completely. It's likely that if you had let it sit, the box
would have eventually dig its way out.
-chris
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
