lkml.org 
[lkml]   [2008]   [Jun]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: FIO: kjournald blocked for more than 120 seconds
On Mon, Jun 16 2008, Lin Ming wrote:
> Hi, Jens
>
> When runnig FIO benchmark, kjournald blocked for more than 120 seconds.
> Detailed root cause analysis and proposed solutions as below.
>
> Any comment is appreciated.
>
> Hardware Environment
> ---------------------
> 13 SEAGATE ST373307FC disks in a JBOD, connected by a Qlogic ISP2312
> Fibe Channel HBA.
>
> Bug description
> ----------------
> fio vsync random read 4K in 13 disks, 4 processes per disk, fio global
> paramter as below,
> [global]
> direct=0
> ioengine=vsync
> iodepth=256
> iodepth_batch=32
> size=2G
> bs=4k
> numjobs=2
> loops=5
> runtime=1200
> group_reporting
>
> Tested 4 IO schedulers, issue is only seen in CFQ.
>
> INFO: task kjournald:20558 blocked for more than 120 seconds.
> "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this
> message.
> kjournald D ffff810010820978 6712 20558 2
> ffff81022ddb1d10 0000000000000046 ffff81022e7baa10 ffffffff803ba6f2
> ffff81022ecd0000 ffff8101e6dc9160 ffff81022ecd0348 000000008048b6cb
> 0000000000000086 ffff81022c4e8d30 0000000000000000 ffffffff80247537
> Call Trace:
> [<ffffffff803ba6f2>] kobject_get+0x12/0x17
> [<ffffffff80247537>] getnstimeofday+0x2f/0x83
> [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f
> [<ffffffff8066d195>] io_schedule+0x5d/0x9f
> [<ffffffff8029c1e7>] sync_buffer+0x3b/0x3f
> [<ffffffff8066d3f0>] __wait_on_bit+0x40/0x6f
> [<ffffffff8029c1ac>] sync_buffer+0x0/0x3f
> [<ffffffff8066d48b>] out_of_line_wait_on_bit+0x6c/0x78
> [<ffffffff80243909>] wake_bit_function+0x0/0x23
> [<ffffffff8029e3ad>] sync_dirty_buffer+0x98/0xcb
> [<ffffffff8030056b>] journal_commit_transaction+0x97d/0xcb6
> [<ffffffff8023a676>] lock_timer_base+0x26/0x4b
> [<ffffffff8030300a>] kjournald+0xc1/0x1fb
> [<ffffffff802438db>] autoremove_wake_function+0x0/0x2e
> [<ffffffff80302f49>] kjournald+0x0/0x1fb
> [<ffffffff802437bb>] kthread+0x47/0x74
> [<ffffffff8022de51>] schedule_tail+0x28/0x5d
> [<ffffffff8020cac8>] child_rip+0xa/0x12
> [<ffffffff80243774>] kthread+0x0/0x74
> [<ffffffff8020cabe>] child_rip+0x0/0x12
>
> Other people reported the similar issue on LKML.
> http://lkml.org/lkml/2008/3/26/494
>
> I did lots of kernel instrumentation and testing, and located the root
> cause.
>
> Root cause analysis
> -------------------
>
> kjournald is blocked at:
> journal_write_commit_record ->
> sync_dirty_buffer(bh) ->
> wait_on_buffer(bh)
>
> Buffer is submitted but it never finished.
>
> git bisect to below patch,
>
> commit cc19747977824ece6aa1c56a29e974fef5ec2b32
> Author: Jens Axboe <jens.axboe@oracle.com>
> Date: Fri Apr 20 20:45:39 2007 +0200
>
> cfq-iosched: tighten queue request overlap condition
>
> For tagged devices, allow overlap of requests if the idle window
> isn't enabled on the current active queue.
>
> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
>
> diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
> index a8237be..e859b49 100644
> --- a/block/cfq-iosched.c
> +++ b/block/cfq-iosched.c
> @@ -989,7 +989,8 @@ static struct cfq_queue *cfq_select_queue(struct
> cfq_data *cfqd)
> * flight or is idling for a new request, allow either of these
> * conditions to happen (or time out) before selecting a new queue.
> */
> - if (cfqq->dispatched || timer_pending(&cfqd->idle_slice_timer)) {
> + if (timer_pending(&cfqd->idle_slice_timer) ||
> + (cfqq->dispatched && cfq_cfqq_idle_window(cfqq))) {
> cfqq = NULL;
> goto keep_queue;
> }
>
> The disks of my testing machine are tagged devices, so the CFQ idle
> window is disabled. In other words, the active queue of tagged
> devices(cfqd->hw_tag=1) never idle for a new request.
>
> This causes active queue be expired immediately if it's empty, although
> it has not run out of time. CFQ will select next queue as active queue.
> In this testcase, there are thousands of FIO read requests in sync
> queues, only a few write requests by journal_write_commit_record in
> async queues.
>
> In the other hand, all processes use the default io class and priority.
> They share the async queue for the same device, but have their own sync
> queue, so the sync queue number is 4 while asyn queue number is just 1
> for the same device.
>
> So sync queue has much more chances be selected as new active queue than
> async queue.
>
> Sync queues do not idle and they are dispatched all the time. This leads
> to many unfinished requests in external queue,
> namely, cfqd->sync_flight > 0.
>
> static int cfq_dispatch_requests (...) {
> ....
> while ((cfqq = cfq_select_queue(cfqd)) != NULL) {
> ....
> if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq))
> break;
> ....
> __cfq_dispatch_requests(cfqq)
> }
> ....
> }
>
> When cfq_select_queue selects the async queue which includes kjournald's
> write request, this selected async queue will never be dispatched since
> cfqd->sync_flight > 0, so kjournald is blocked.
>
> Proposed 3 solutions
> ------------------
> 1. Do not check cfqd->sync_flight
>
> - if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq))
> - break;
>
> 2. If we do need to check cfqd->sync_flight, then for tagged devices, we
> should give a little more chances to async queue to be dispatched.
>
> @@ -1102,7 +1102,7 @@ static int cfq_dispatch_requests(struct
> request_queue *q, int force)
> break;
> }
>
> - if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq))
> + if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq) && !
> cfqd->hw_tag)
> break;
>
> 3. Force write request issued by journal_write_commit_record as sync
> request. As a matter of fact, it looks like most write requests
> submitted by kjournald is async request. We need convert them to sync
> requests.

Thanks for the very detailed analysis of the problem, complete with
suggestions. While I think that any code that does:

submit async io
wait for it

should be issuing sync IO (or, better, automatically upgrade the request
from async -> sync), we cannot rely on that.

This problem is similar in nature to device starvation, and a classic
solution to that problem is to issue occasional ordered tags to prevent
indefinite starvation. Perhaps we can apply some similar logic here.

For 2.6.26, the simple approach of just removing the sync_flight check
is probably the safest.

--
Jens Axboe



\
 
 \ /
  Last update: 2008-06-16 21:33    [from the cache]
©2003-2011 Jasper Spaans