[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20091207172747.GE16900@redhat.com>
Date: Mon, 7 Dec 2009 12:27:47 -0500
From: Vivek Goyal <vgoyal@...hat.com>
To: Jeff Moyer <jmoyer@...hat.com>
Cc: Jens Axboe <jens.axboe@...cle.com>,
Corrado Zoccolo <czoccolo@...il.com>,
Linux-Kernel <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] cfq-iosched: reduce write depth only if sync was
delayed
On Mon, Dec 07, 2009 at 11:45:43AM -0500, Jeff Moyer wrote:
> Jens Axboe <jens.axboe@...cle.com> writes:
>
> > On Mon, Dec 07 2009, Jeff Moyer wrote:
> >> Jens Axboe <jens.axboe@...cle.com> writes:
> >>
> >> > On Sun, Dec 06 2009, Corrado Zoccolo wrote:
> >> >> Hi Jeff,
> >> >> I remember you saw large performance drop on your SAN for sequential
> >> >> writes with low_latency=1. Can you test if Shaohua's and this patch
> >> >> fix allow to recover some bandwidth? I think that enabling the queue
> >> >> depth ramp up only if a sync request was delayed should disable it for
> >> >> fast hardware like yours, so you should not be seeing the slowdown any
> >> >> more.
> >> >
> >> > I queued this up for post inclusion into 2.6.33, with the time_after()
> >> > fixed.
> >> >
> >> > The patch was word-wrapped, btw.
> >>
> >> So in what branch can I find this fix? Once I know that I can queue up
> >> some tests.
> >
> > It's in next-2.6.33
>
> next-2.6.33 won't boot for me:
>
> general protection fault: 0000 [#1] SMP
> async/0 used greatest stack depth: 4256 bytes left
>
> last sysfs file: /sys/class/firmware/timeout
> CPU 1
> Modules linked in: ata_piix pata_acpi libata sd_mod scsi_mod ext3 jbd mbcache uh
> ci_hcd ohci_hcd ehci_hcd
> Pid: 729, comm: async/1 Not tainted 2.6.32 #1 ProLiant DL320 G5p
> RIP: 0010:[<ffffffff81199cee>] [<ffffffff81199cee>] cfq_put_cfqg+0x0/0x91
> RSP: 0018:ffff8801251b1d48 EFLAGS: 00010002
> RAX: ffff880126dcdd28 RBX: ffff8801251fa158 RCX: 0000000000170001
> RDX: ffff880125556700 RSI: ffff8801251fa158 RDI: 6b6b6b6b6b6b6b6b
> RBP: ffff8801251b1d70 R08: ffff8801255a0448 R09: 000000000000005a
> R10: ffff8801255a0448 R11: ffffffff818d6210 R12: ffff880126dcdb18
> R13: ffff880126dcdb50 R14: 0000000000000286 R15: ffff880125556760
> FS: 0000000000000000(0000) GS:ffff88002f200000(0000) knlGS:0000000000000000
> CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b
> CR2: 0000000000000000 CR3: 00000001256a5000 CR4: 00000000000006a0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> Process async/1 (pid: 729, threadinfo ffff8801251b0000, task ffff880125556770)
> Stack:
> ffffffff8119a856 0000000000000002 ffff880126dcdb18 ffff8801251fa158
> <0> ffff880125210000 ffff8801251b1d90 ffffffff8119a9e5 ffff8801251f80d0
> <0> ffff880126dcdb18 ffff8801251b1db0 ffffffff8119aa62 ffff880126dcdb18
> Call Trace:
> [<ffffffff8119a856>] ? cfq_put_queue+0xfa/0x102
> [<ffffffff8119a9e5>] cfq_exit_cfqq+0x99/0x9e
> [<ffffffff8119aa62>] __cfq_exit_single_io_context+0x78/0x85
> [<ffffffff8119aaa9>] cfq_exit_single_io_context+0x3a/0x52
> [<ffffffff8119aa6f>] ? cfq_exit_single_io_context+0x0/0x52
> [<ffffffff8119b27b>] call_for_each_cic+0x56/0x7c
> [<ffffffff8119b225>] ? call_for_each_cic+0x0/0x7c
> [<ffffffff8119b2b1>] cfq_exit_io_context+0x10/0x12
> [<ffffffff81192d3b>] exit_io_context+0x93/0xbc
> [<ffffffff81192d03>] ? exit_io_context+0x5b/0xbc
> [<ffffffff810474e5>] do_exit+0x71a/0x747
> [<ffffffff810628f1>] ? async_thread+0x0/0x1fa
> [<ffffffff8105cd9e>] kthread_stop+0x0/0xb3
> [<ffffffff81033fa6>] ? complete+0x1c/0x4b
> [<ffffffff8100cafa>] child_rip+0xa/0x20
> [<ffffffff8103d667>] ? finish_task_switch+0x0/0xe3
> [<ffffffff8100c4bc>] ? restore_args+0x0/0x30
> [<ffffffff8105ccf8>] ? kthreadd+0xdf/0x100
> [<ffffffff8105cd19>] ? kthread+0x0/0x85
> [<ffffffff8100caf0>] ? child_rip+0x0/0x20
> Code: 48 c7 43 38 00 00 00 00 48 c7 43 40 00 00 00 00 48 89 3e 48 8b 73 48 e8 fd 9e 00 00 eb 08 48 c7 43 48 00 00 00 00 5b 41 5c c9 c3 <8b> 87 d8 01 00 00 55 48 89 e5 85 c0 7f 04 0f 0b eb fe 48 8d 87
> RIP [<ffffffff81199cee>] cfq_put_cfqg+0x0/0x91
> RSP <ffff8801251b1d48>
Hm.., I seem to be accessing cfqq->orig_cfqg, after I have freed cfqq.
Following patch should fix it. Testing it now.
Thanks
Vivek
Index: linux-2.6-block/block/cfq-iosched.c
===================================================================
--- linux-2.6-block.orig/block/cfq-iosched.c
+++ linux-2.6-block/block/cfq-iosched.c
@@ -2368,7 +2368,7 @@ static int cfq_dispatch_requests(struct
static void cfq_put_queue(struct cfq_queue *cfqq)
{
struct cfq_data *cfqd = cfqq->cfqd;
- struct cfq_group *cfqg;
+ struct cfq_group *cfqg, *orig_cfqg;
BUG_ON(atomic_read(&cfqq->ref) <= 0);
@@ -2379,6 +2379,7 @@ static void cfq_put_queue(struct cfq_que
BUG_ON(rb_first(&cfqq->sort_list));
BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
cfqg = cfqq->cfqg;
+ orig_cfqg = cfqq->orig_cfqg;
if (unlikely(cfqd->active_queue == cfqq)) {
__cfq_slice_expired(cfqd, cfqq, 0);
@@ -2388,8 +2389,8 @@ static void cfq_put_queue(struct cfq_que
BUG_ON(cfq_cfqq_on_rr(cfqq));
kmem_cache_free(cfq_pool, cfqq);
cfq_put_cfqg(cfqg);
- if (cfqq->orig_cfqg)
- cfq_put_cfqg(cfqq->orig_cfqg);
+ if (orig_cfqg)
+ cfq_put_cfqg(orig_cfqg);
}
/*
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists