From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: kernel test robot <xiaolong.ye@intel.com>,
LKML <linux-kernel@vger.kernel.org>,
lkp@01.org
Subject: Re: [lkp-robot] [torture] b151f93a71: INFO:rcu_preempt_detected_stalls_on_CPUs/tasks
Date: Tue, 28 Nov 2017 09:07:59 -0800 [thread overview]
Message-ID: <20171128170759.GU3624@linux.vnet.ibm.com> (raw)
In-Reply-To: <alpine.DEB.2.20.1711281745000.2222@nanos>
On Tue, Nov 28, 2017 at 05:47:35PM +0100, Thomas Gleixner wrote:
> On Tue, 28 Nov 2017, Paul E. McKenney wrote:
>
> > On Tue, Nov 28, 2017 at 03:16:30PM +0100, Thomas Gleixner wrote:
> > > On Mon, 27 Nov 2017, Paul E. McKenney wrote:
> > > > On Sun, Nov 26, 2017 at 04:42:03PM +0800, kernel test robot wrote:
> > > > > [ 116.353432] rcu_preempt kthread starved for 9974 jiffies! g4294967208 c4294967207 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x402 ->cpu=0
> > > >
> > > > So the immediate reason for the stall warning is that the RCU grace-period
> > > > kthread isn't being allowed to run.
> > > >
> > > > > [ 116.355517] rcu_preempt I 7464 8 2 0x80000000
> > > > > [ 116.356543] Call Trace:
> > > > > [ 116.357008] __schedule+0x493/0x620
> > > > > [ 116.357682] schedule+0x24/0x40
> > > > > [ 116.358291] schedule_timeout+0x330/0x3b0
> > > >
> > > > And the reason that it isn't being allowed to run is that its few-jiffy
> > > > schedule_timeout has extended for more than nine thousand jiffies.
> > > >
> > > > There was an odd combination of kernel parameters that Thomas Gleixner
> > > > came across that could cause writer-thread stalls (since fixed in -rcu
> > > > by the exact patch you call out here), but I don't see how this could
> > > > cause an RCU CPU stall warning.
> > >
> > > <Trimmed a lot of useless crap.....>
> > >
> > > The only reasonable explanation is that either a wakeup is missed or the
> > > timer is not expired. Hard to tell from that back trace, but it would be
> > > interesting to figure that out. Let me think about how that can be done.
> >
> > Please see below for my latest patch in service of futility. It leverages
> > the fact that the stall-warning code takes the desperation step of sending
> > an unsolicited wakeup to the grace-period kthread. This allows the patch
> > to use a stupid but effective heuristic to detect that the timer (or the
> > wakeup) has gone AWOL.
> >
> > I get the following "Waylayed timer" line from it:
> >
> > [ 8015.327554] rcu_sched kthread starved for 21096 jiffies! g12440 c12439 f0x0 RCU_GP_WAIT_FQS(3) ->state=0x402 ->cpu=3
> > [ 8015.327554] rcu_sched I14992 9 2 0x80000000
> > [ 8015.327554] Call Trace:
> > [ 8015.327554] ? __schedule+0x33c/0x6f0
> > [ 8015.327554] ? preempt_count_add+0x51/0x90
> > [ 8015.327554] schedule+0x37/0x90
> > [ 8015.327554] schedule_timeout+0x159/0x410
> > [ 8015.327554] ? __next_timer_interrupt+0xc0/0xc0
> > [ 8015.327554] rcu_gp_kthread+0xa42/0x1080
> > [ 8015.327554] kthread+0xf0/0x130
> > [ 8015.327554] ? rcu_barrier+0x10/0x10
> > [ 8015.327554] ? kthread_destroy_worker+0x40/0x40
> > [ 8015.327554] ret_from_fork+0x1f/0x30
> > [ 8015.421044] schedule_timeout: Waylayed timer idx: 56 idx_now: 222
> >
> > Is there something more useful that I could make this thing print?
>
> Please let it print base->clk, jiffies, base->next_expiry, timer->flags,
> and timer->expires.
Like this, on top of the previous patch?
Thanx, Paul
------------------------------------------------------------------------
commit 1085cc784fe1e9d4b5e3bb2f849c0a7fa47b7cb7
Author: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Date: Tue Nov 28 09:05:39 2017 -0800
EXP: timer: Dump more info if timer misses by huge amount
"Please let it print base->clk, jiffies, base->next_expiry, timer->flags,
and timer->expires."
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
diff --git a/kernel/time/timer.c b/kernel/time/timer.c
index db774b0f217e..a3321bb565db 100644
--- a/kernel/time/timer.c
+++ b/kernel/time/timer.c
@@ -1803,7 +1803,7 @@ signed long __sched schedule_timeout(signed long timeout)
idx = timer_get_idx(&timer.timer);
idx_now = calc_wheel_index(j, base->clk);
raw_spin_unlock_irqrestore(&base->lock, flags);
- pr_info("%s: Waylayed timer idx: %u idx_now: %u\n", __func__, idx, idx_now);
+ pr_info("%s: Waylayed timer base->clk: %#lx jiffies: %#lx base->next_expiry: %#lx timer->flags: %#x timer->expires %#lx idx: %u idx_now: %u\n", __func__, base->clk, j, base->next_expiry, timer.timer.flags, timer.timer.expires, idx, idx_now);
}
del_singleshot_timer_sync(&timer.timer);
next prev parent reply other threads:[~2017-11-28 17:09 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-26 8:42 [lkp-robot] [torture] b151f93a71: INFO:rcu_preempt_detected_stalls_on_CPUs/tasks kernel test robot
2017-11-27 21:57 ` Paul E. McKenney
2017-11-27 2:48 ` Ye Xiaolong
2017-11-28 14:16 ` Thomas Gleixner
2017-11-28 16:41 ` Paul E. McKenney
2017-11-28 16:47 ` Thomas Gleixner
2017-11-28 17:07 ` Paul E. McKenney [this message]
2017-11-28 17:10 ` Thomas Gleixner
2017-11-28 17:35 ` Paul E. McKenney
2017-11-28 20:46 ` Paul E. McKenney
2017-11-28 21:08 ` Paul E. McKenney
2017-11-29 19:08 ` Paul E. McKenney
2017-11-29 22:07 ` Paul E. McKenney
2017-11-29 22:38 ` Paul E. McKenney
2017-12-01 0:45 ` Paul E. McKenney
2017-11-28 16:52 ` Paul E. McKenney
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20171128170759.GU3624@linux.vnet.ibm.com \
--to=paulmck@linux.vnet.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=lkp@01.org \
--cc=tglx@linutronix.de \
--cc=xiaolong.ye@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).