Messages in this thread | | | Date | Sat, 11 Feb 2012 18:30:42 +0000 | Subject | Re: exit_aio() hang after I/O failure | From | Bart Van Assche <> |
| |
On Mon, Jan 23, 2012 at 4:47 PM, Bart Van Assche <bvanassche@acm.org> wrote: > On Mon, Jan 23, 2012 at 4:15 PM, Jeff Moyer <jmoyer@redhat.com> wrote: > > Bart Van Assche <bvanassche@acm.org> writes: > > > Apparently processes can hang in exit_aio() with at least kernel 3.2.1 > > > after an I/O failure. Has anyone seen this before ? > > > > > > This occurred after a SCSI device had been removed entirely (and hence > > > after all I/O requests were killed by scsi_remove_host()). > > > > Fixed here: > > > > http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=69e4747ee9727d660b88d7e1efe0f4afcb35db1b > > Thanks a lot for the feedback - I'll give this patch a try.
Bad news: I've been able to reproduce exactly the same call stack with kernel 3.2.5. That kernel version includes the aforementioned commit.
# echo t >/proc/sysrq-trigger [ ... ] fio D 0000000000000001 0 25052 25008 0x00000004 ffff88001c32fb88 0000000000000046 ffff880000000000 ffff88007d949bc8 ffff88001c8e14d0 ffff88001c32ffd8 ffff88001c32ffd8 ffff88001c32ffd8 ffff880128b894d0 ffff88001c8e14d0 ffff88001c32fb88 000000018106f24d Call Trace: [<ffffffff813b683f>] schedule+0x3f/0x60 [<ffffffff813b68ef>] io_schedule+0x8f/0xd0 [<ffffffff81174410>] wait_for_all_aios+0xc0/0x100 [<ffffffff8103c3c0>] ? try_to_wake_up+0x270/0x270 [<ffffffff81175385>] exit_aio+0x55/0xc0 [<ffffffff810413cd>] mmput+0x2d/0x110 [<ffffffff81047c1d>] exit_mm+0x10d/0x130 [<ffffffff810482b1>] do_exit+0x671/0x860 [<ffffffff81033c1e>] ? finish_task_switch+0x4e/0xe0 [<ffffffff81048804>] do_group_exit+0x44/0xb0 [<ffffffff81058018>] get_signal_to_deliver+0x218/0x5a0 [<ffffffff81002065>] do_signal+0x65/0x700 [<ffffffff811740d0>] ? aio_read_evt+0x150/0x150 [<ffffffff8103c3c0>] ? try_to_wake_up+0x270/0x270 [<ffffffff81002785>] do_notify_resume+0x65/0x80 [<ffffffff811df84e>] ? trace_hardirqs_on_thunk+0x3a/0x3c [<ffffffff813c0333>] int_signal+0x12/0x17 [ ... ]
Bart.
| |