Re: [linux-next] cpus stalls detected few hours after booting next kernel

[Paul E. McKenney]

On Tue, Jul 04, 2017 at 07:44:58PM +0530, Abdul Haleem wrote:
> On Fri, 2017-06-30 at 17:28 +1000, Nicholas Piggin wrote:
> > On Fri, 30 Jun 2017 10:52:18 +0530
> > Abdul Haleem <abdhalee@xxxxxxxxxxxxxxxxxx> wrote:
> > 
> > > On Fri, 2017-06-30 at 00:45 +1000, Nicholas Piggin wrote:
> > > > On Thu, 29 Jun 2017 20:23:05 +1000
> > > > Nicholas Piggin <npiggin@xxxxxxxxx> wrote:
> > > > 
> > > > > On Thu, 29 Jun 2017 19:36:14 +1000
> > > > > Nicholas Piggin <npiggin@xxxxxxxxx> wrote:
> > > > 
> > > > > > I don't *think* the replay-wakeup-interrupt patch is directly involved, but
> > > > > > it's likely to be one of the idle patches.  
> > > > 
> > > > Okay this turned out to be misconfigured sleep states I added for the
> > > > simulator, sorry for the false alarm.
> > > > 
> > > > > Although you have this in the backtrace. I wonder if that's a stuck
> > > > > lock in rcu_process_callbacks?
> > > > 
> > > > So this spinlock becomes top of the list of suspects. Can you try
> > > > enabling lockdep and try to reproduce it?
> > > 
> > > Yes, recreated again with CONFIG_LOCKDEP=y & CONFIG_DEBUG_LOCKDEP=y set.
> > > I do not see any difference in trace messages with and without LOCKDEP
> > > enabled.
> > > 
> > > Please find the attached log file.
> > 
> > Can you get an rcu_invoke_callback event trace that Paul suggested?
> 
> I could not reproduce the issue with the latest next kernel
> (4.12.0-rc7-next-20170703).
> 
> > Does this bug show up with just the powerpc next branch?
> 
> Perhaps, stress-ng test on today's mainline kernel (4.12.0) triggered a
> different trace message and this not reproducible all the time.
> 
> 
> stress-ng --io 100  --vm 10 --vm-bytes 100g --timeout 1h --oomable
> 
> INFO: rcu_sched self-detected stall on CPU
>         8-...: (2099 ticks this GP) idle=37e/140000000000001/0
> softirq=1404131/1404131 fqs=932 
>          (t=2100 jiffies g=394894 c=394893 q=21997)
> Task dump for CPU 8:
> kworker/u162:2  R  running task    11168 28468      2 0x00000884   
> Workqueue: writeback wb_workfn (flush-253:1)
> Call Trace:
> [c00000036e3eb340] [c000000000114480] sched_show_task+0xf0/0x160
> (unreliable)
> [c00000036e3eb3b0] [c0000000009adc2c] rcu_dump_cpu_stacks+0xd0/0x134
> [c00000036e3eb400] [c00000000015e4d0] rcu_check_callbacks+0x8f0/0xaf0
> [c00000036e3eb530] [c000000000165e2c] update_process_times+0x3c/0x90
> [c00000036e3eb560] [c00000000017b73c] tick_sched_handle.isra.13
> +0x2c/0xc0
> [c00000036e3eb590] [c00000000017b828] tick_sched_timer+0x58/0xb0
> [c00000036e3eb5d0] [c0000000001669e8] __hrtimer_run_queues+0xf8/0x330
> [c00000036e3eb650] [c000000000167744] hrtimer_interrupt+0xe4/0x280
> [c00000036e3eb710] [c000000000022620] __timer_interrupt+0x90/0x270
> [c00000036e3eb760] [c000000000022cf0] timer_interrupt+0xa0/0xe0
> [c00000036e3eb790] [c0000000000091e8] decrementer_common+0x158/0x160
> --- interrupt: 901 at move_expired_inodes+0x30/0x200
>     LR = queue_io+0x8c/0x190
> [c00000036e3eba80] [c00000036e3ebac0] 0xc00000036e3ebac0 (unreliable)
> [c00000036e3ebac0] [c000000000319b24] wb_writeback+0x2b4/0x420

The usual assumption would be that wb_writeback() is looping or...

> [c00000036e3ebb90] [c00000000031a980] wb_workfn+0xf0/0x4b0
> [c00000036e3ebca0] [c0000000000fa160] process_one_work+0x180/0x470
> [c00000036e3ebd30] [c0000000000fa6d4] worker_thread+0x284/0x500

...there is so much work being scheduled that worker_thread() cannot
keep up.  If tracing locates such a loop, the usual trick is to
place a cond_resched_rcu_qs() somewhere in it (but of course not
where interrupts or preemption are disabled).

Tracing can help detect this.  Or printk()s, for that matter.  ;-)

							Thanx, Paul

> [c00000036e3ebdc0] [c000000000101fc0] kthread+0x160/0x1a0
> [c00000036e3ebe30] [c00000000000bb60] ret_from_kernel_thread+0x5c/0x7c
> INFO: rcu_sched self-detected stall on CPU
>         51-...: (2099 ticks this GP) idle=7c2/140000000000001/0
> softirq=1272749/1272749 fqs=995 
>          (t=2100 jiffies g=394900 c=394899 q=32186)
> ~                                                   
> 
> stress-ng : http://kernel.ubuntu.com/git/cking/stress-ng.git
> 
> > 
> > Thanks,
> > Nick
> > 
> 
> 
> -- 
> Regard's
> 
> Abdul Haleem
> IBM Linux Technology Centre
> 
> 
>