Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1656752 > unrolled thread

Re: [PATCH 2/2] tick: Make sure tick timer is active when bypassing reprogramming

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH 2/2] tick: Make sure tick timer is active when bypassing  reprogramming "Levin, Alexander (Sasha Levin)" <alexander.levin@verizon.com> - 2017-06-03 10:10 +0200
    Re: [PATCH 2/2] tick: Make sure tick timer is active when bypassing  reprogramming Frederic Weisbecker <fweisbec@gmail.com> - 2017-06-03 14:50 +0200
      Re: [PATCH 2/2] tick: Make sure tick timer is active when bypassing  reprogramming Frederic Weisbecker <fweisbec@gmail.com> - 2017-06-06 17:00 +0200
        Re: [PATCH 2/2] tick: Make sure tick timer is active when bypassing  reprogramming Frederic Weisbecker <fweisbec@gmail.com> - 2017-06-07 16:20 +0200

#1656752 — Re: [PATCH 2/2] tick: Make sure tick timer is active when bypassing reprogramming

On Fri, Apr 21, 2017 at 04:00:55PM +0200, Frederic Weisbecker wrote:
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index 502b320..be7ca4d 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -783,8 +783,13 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
>  	tick = expires;
>  
>  	/* Skip reprogram of event if its not changed */
> -	if (ts->tick_stopped && (expires == ts->next_tick))
> -		goto out;
> +	if (ts->tick_stopped && (expires == ts->next_tick)) {
> +		/* Sanity check: make sure clockevent is actually programmed */
> +		if (likely(dev->next_event <= ts->next_tick))
> +			goto out;
> +
> +		WARN_ON_ONCE(1);
> +	}

I seem to be hitting that in a KVM vm, even without load (sometimes
right after boot):

------------[ cut here ]------------
WARNING: CPU: 5 PID: 0 at kernel/time/tick-sched.c:794 tick_nohz_stop_sched_tick kernel/time/tick-sched.c:791 [inline]
WARNING: CPU: 5 PID: 0 at kernel/time/tick-sched.c:794 __tick_nohz_idle_enter+0x11f4/0x1ab0 kernel/time/tick-sched.c:950
Kernel panic - not syncing: panic_on_warn set ...

CPU: 5 PID: 0 Comm: swapper/5 Not tainted 4.12.0-rc3-next-20170601+ #53
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.1-1ubuntu1 04/01/2014
Call Trace:
 <IRQ>
 __dump_stack lib/dump_stack.c:16 [inline]
 dump_stack+0x115/0x1d1 lib/dump_stack.c:52
 panic+0x1d8/0x3c4 kernel/panic.c:180
 __warn+0x1d6/0x220 kernel/panic.c:541
 report_bug+0x223/0x310 lib/bug.c:183
 fixup_bug arch/x86/kernel/traps.c:190 [inline]
 do_trap_no_signal arch/x86/kernel/traps.c:224 [inline]
 do_trap+0x34c/0x490 arch/x86/kernel/traps.c:273
 do_error_trap+0x12f/0x390 arch/x86/kernel/traps.c:310
 do_invalid_op+0x1b/0x20 arch/x86/kernel/traps.c:323
 invalid_op+0x1e/0x30 arch/x86/entry/entry_64.S:844
RIP: 0010:tick_nohz_stop_sched_tick kernel/time/tick-sched.c:791 [inline]
RIP: 0010:__tick_nohz_idle_enter+0x11f4/0x1ab0 kernel/time/tick-sched.c:950
RSP: 0018:ffff88005f407dd8 EFLAGS: 00010082
RAX: 000000000000007e RBX: ffff88005f5dcfe0 RCX: 0000000000000000
RDX: 000000000000007e RSI: 1ffff1000be80f78 RDI: ffffed000be80faf
RBP: ffff88005f407f88 R08: ffff88005f416300 R09: 1ffff1000be82c76
R10: ffff88007ffd915d R11: 0000000000000001 R12: ffff88005f5dcff8
R13: 00000061d944ea40 R14: ffff88005f407ea0 R15: 1ffff1000be80fc8
 tick_nohz_irq_exit+0xac/0x120 kernel/time/tick-sched.c:1009
 tick_irq_exit kernel/softirq.c:386 [inline]
 irq_exit+0x141/0x1b0 kernel/softirq.c:407
 exiting_irq arch/x86/include/asm/apic.h:652 [inline]
 smp_call_function_interrupt+0x70/0xa0 arch/x86/kernel/smp.c:302
 smp_call_function_single_interrupt+0x9/0x10 include/linux/compiler.h:250
 call_function_single_interrupt+0x9d/0xb0 arch/x86/entry/entry_64.S:722
RIP: 0010:native_safe_halt+0x6/0x10 arch/x86/include/asm/irqflags.h:53
RSP: 0018:ffff88005ee7fdb8 EFLAGS: 00000246 ORIG_RAX: ffffffffffffff04
RAX: 0000000000000007 RBX: 1ffff1000bdcffba RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffffffff91672020 RDI: ffff88005ee708f4
RBP: ffff88005ee7fdb8 R08: ffff88005f416060 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: ffff88005ee70040
R13: ffff88005ee7fe70 R14: 0000000000000000 R15: ffff88005ee70040
 </IRQ>
 arch_safe_halt arch/x86/include/asm/paravirt.h:98 [inline]
 default_idle+0x8d/0x600 arch/x86/kernel/process.c:341
 arch_cpu_idle+0xa/0x10 arch/x86/kernel/process.c:332
 default_idle_call+0x4c/0xa0 kernel/sched/idle.c:98
 cpuidle_idle_call kernel/sched/idle.c:156 [inline]
 do_idle+0x287/0x410 kernel/sched/idle.c:245
 cpu_startup_entry+0x18/0x20 kernel/sched/idle.c:350
 start_secondary+0x2d3/0x400 arch/x86/kernel/smpboot.c:275
 secondary_startup_64+0x9f/0x9f arch/x86/kernel/head_64.S:304
Dumping ftrace buffer:
   (ftrace buffer empty)
Kernel Offset: 0xb200000 from 0xffffffff81000000 (relocation range: 0xffffffff80000000-0xffffffffbfffffff)
Rebooting in 86400 seconds..

-- 

Thanks,
Sasha

[toc] | [next] | [standalone]

#1656789

On Sat, Jun 03, 2017 at 08:06:41AM +0000, Levin, Alexander (Sasha Levin) wrote:
> On Fri, Apr 21, 2017 at 04:00:55PM +0200, Frederic Weisbecker wrote:
> > diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> > index 502b320..be7ca4d 100644
> > --- a/kernel/time/tick-sched.c
> > +++ b/kernel/time/tick-sched.c
> > @@ -783,8 +783,13 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
> >  	tick = expires;
> >  
> >  	/* Skip reprogram of event if its not changed */
> > -	if (ts->tick_stopped && (expires == ts->next_tick))
> > -		goto out;
> > +	if (ts->tick_stopped && (expires == ts->next_tick)) {
> > +		/* Sanity check: make sure clockevent is actually programmed */
> > +		if (likely(dev->next_event <= ts->next_tick))
> > +			goto out;
> > +
> > +		WARN_ON_ONCE(1);
> > +	}
> 
> I seem to be hitting that in a KVM vm, even without load (sometimes
> right after boot):

Ah, can you tell me which tree you were using? Is it tip/master?
Can you give me its HEAD and your config file?

Thanks

[toc] | [prev] | [next] | [standalone]

#1658845

On Sat, Jun 03, 2017 at 01:00:53PM +0000, Levin, Alexander (Sasha Levin) wrote:
> On Sat, Jun 03, 2017 at 02:42:43PM +0200, Frederic Weisbecker wrote:
> > On Sat, Jun 03, 2017 at 08:06:41AM +0000, Levin, Alexander (Sasha Levin) wrote:
> > > On Fri, Apr 21, 2017 at 04:00:55PM +0200, Frederic Weisbecker wrote:
> > > > diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> > > > index 502b320..be7ca4d 100644
> > > > --- a/kernel/time/tick-sched.c
> > > > +++ b/kernel/time/tick-sched.c
> > > > @@ -783,8 +783,13 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
> > > >  	tick = expires;
> > > >  
> > > >  	/* Skip reprogram of event if its not changed */
> > > > -	if (ts->tick_stopped && (expires == ts->next_tick))
> > > > -		goto out;
> > > > +	if (ts->tick_stopped && (expires == ts->next_tick)) {
> > > > +		/* Sanity check: make sure clockevent is actually programmed */
> > > > +		if (likely(dev->next_event <= ts->next_tick))
> > > > +			goto out;
> > > > +
> > > > +		WARN_ON_ONCE(1);
> > > > +	}
> > > 
> > > I seem to be hitting that in a KVM vm, even without load (sometimes
> > > right after boot):
> > 
> > Ah, can you tell me which tree you were using? Is it tip/master?
> 
> Its next-20170601: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?h=next-20170601&id=3ab334ebe84e0dfd1cc3ea2fe77f5ce4406f7370
> 
> > Can you give me its HEAD and your config file?
> 
> Attached config.

Thanks Sasha!

I couldn't reproduce it, that config boots fine on my kvm.
Would you have the time to dump some traces for me?

I'd need you to add this boot option: trace_event=hrtimer_cancel,hrtimer_start,hrtimer_expire_entry
And boot your kernel with the below patch. This will dump the timer traces to your console.
I would be very interested in the resulting console dump file.

Thanks!

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 2de9c55..ad1de28 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -156,6 +156,7 @@ static void tick_sched_handle(struct tick_sched *ts, struct pt_regs *regs)
 		 * to the same deadline.
 		 */
 		ts->next_tick = 0;
+		trace_printk("ts->next_tick reset (tick)\n");
 	}
 #endif
 	update_process_times(user_mode(regs));
@@ -672,6 +673,7 @@ static void tick_nohz_restart(struct tick_sched *ts, ktime_t now)
 	 * cached clock deadline.
 	 */
 	ts->next_tick = 0;
+	trace_printk("ts->next_tick reset (tick restart)\n");
 }
 
 static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
@@ -788,6 +790,11 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 		if (likely(dev->next_event <= ts->next_tick))
 			goto out;
 
+		trace_printk("basemono: %llu ts->next_tick: %llu dev->next_event: %llu timer->active: %d timer->expires: %llu\n",
+			    basemono, ts->next_tick, dev->next_event,
+			    hrtimer_active(&ts->sched_timer), hrtimer_get_expires(&ts->sched_timer));
+		tracing_stop();
+		ftrace_dump(DUMP_ORIG);
 		WARN_ON_ONCE(1);
 		printk_once("basemono: %llu ts->next_tick: %llu dev->next_event: %llu timer->active: %d timer->expires: %llu\n",
 			    basemono, ts->next_tick, dev->next_event,
@@ -812,6 +819,7 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 	}
 
 	ts->next_tick = tick;
+	trace_printk("ts->next_tick = %llu\n", ts->next_tick);
 
 	/*
 	 * If the expiration time == KTIME_MAX, then we simply stop
@@ -894,6 +902,7 @@ static bool can_stop_idle_tick(int cpu, struct tick_sched *ts)
 		 * deadline if it comes back online later.
 		 */
 		ts->next_tick = 0;
+		trace_printk("ts->next_tick reset (offline)\n");
 		return false;
 	}
 
@@ -1202,8 +1211,10 @@ static enum hrtimer_restart tick_sched_timer(struct hrtimer *timer)
 	 */
 	if (regs)
 		tick_sched_handle(ts, regs);
-	else
+	else {
+		trace_printk("ts->next_tick reset (tick)\n");
 		ts->next_tick = 0;
+	}
 
 	/* No need to reprogram if we are in idle or full dynticks mode */
 	if (unlikely(ts->tick_stopped))

[toc] | [prev] | [next] | [standalone]

#1659841

On Wed, Jun 07, 2017 at 04:17:41AM +0000, Levin, Alexander (Sasha Levin) wrote:
> > Thanks Sasha!
> > 
> > I couldn't reproduce it, that config boots fine on my kvm.
> > Would you have the time to dump some traces for me?
> > 
> > I'd need you to add this boot option: trace_event=hrtimer_cancel,hrtimer_start,hrtimer_expire_entry
> > And boot your kernel with the below patch. This will dump the timer traces to your console.
> > I would be very interested in the resulting console dump file.
> 
> Attached. Let me know if you need anything else.

Great! So now I can deduce that the problem doesn't come from the nohz code as
ts->next_tick matches the hrtimer deadline. But the dev->next_event from the
clockevent seems to be out of line.

Sorry to bother you again, but I'm chasing this bug for several weeks now and
you're one of the rare person who can reproduce it. So I may need some more
tracing details.

Here is another version of the debugging patch (not a delta), I added more trace_printk,
namely the places where we set this dev->next_event. Can you please apply the below and do
the dump again?

I'm adding a boot option as well for the stacktrace:

trace_event=hrtimer_cancel,hrtimer_start,hrtimer_expire_entry trace_options=stacktrace

Thanks a lot!

diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c
index 4237e07..e21e929 100644
--- a/kernel/time/clockevents.c
+++ b/kernel/time/clockevents.c
@@ -180,6 +180,7 @@ void clockevents_shutdown(struct clock_event_device *dev)
 {
 	clockevents_switch_state(dev, CLOCK_EVT_STATE_SHUTDOWN);
 	dev->next_event = KTIME_MAX;
+	trace_printk("dev->next_event: %llu\n", dev->next_event);
 }
 
 /**
@@ -214,6 +215,7 @@ static int clockevents_increase_min_delta(struct clock_event_device *dev)
 		printk_deferred(KERN_WARNING
 				"CE: Reprogramming failure. Giving up\n");
 		dev->next_event = KTIME_MAX;
+		trace_printk("dev->next_event: %llu\n", dev->next_event);
 		return -ETIME;
 	}
 
@@ -247,6 +249,7 @@ static int clockevents_program_min_delta(struct clock_event_device *dev)
 	for (i = 0;;) {
 		delta = dev->min_delta_ns;
 		dev->next_event = ktime_add_ns(ktime_get(), delta);
+		trace_printk("dev->next_event: %llu\n", dev->next_event);
 
 		if (clockevent_state_shutdown(dev))
 			return 0;
@@ -284,6 +287,7 @@ static int clockevents_program_min_delta(struct clock_event_device *dev)
 
 	delta = dev->min_delta_ns;
 	dev->next_event = ktime_add_ns(ktime_get(), delta);
+	trace_printk("dev->next_event: %llu\n", dev->next_event);
 
 	if (clockevent_state_shutdown(dev))
 		return 0;
@@ -316,6 +320,7 @@ int clockevents_program_event(struct clock_event_device *dev, ktime_t expires,
 	}
 
 	dev->next_event = expires;
+	trace_printk("dev->next_event: %llu\n", dev->next_event);
 
 	if (clockevent_state_shutdown(dev))
 		return 0;
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 2de9c55..ad1de28 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -156,6 +156,7 @@ static void tick_sched_handle(struct tick_sched *ts, struct pt_regs *regs)
 		 * to the same deadline.
 		 */
 		ts->next_tick = 0;
+		trace_printk("ts->next_tick reset (tick)\n");
 	}
 #endif
 	update_process_times(user_mode(regs));
@@ -672,6 +673,7 @@ static void tick_nohz_restart(struct tick_sched *ts, ktime_t now)
 	 * cached clock deadline.
 	 */
 	ts->next_tick = 0;
+	trace_printk("ts->next_tick reset (tick restart)\n");
 }
 
 static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
@@ -788,6 +790,11 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 		if (likely(dev->next_event <= ts->next_tick))
 			goto out;
 
+		trace_printk("basemono: %llu ts->next_tick: %llu dev->next_event: %llu timer->active: %d timer->expires: %llu\n",
+			    basemono, ts->next_tick, dev->next_event,
+			    hrtimer_active(&ts->sched_timer), hrtimer_get_expires(&ts->sched_timer));
+		tracing_stop();
+		ftrace_dump(DUMP_ORIG);
 		WARN_ON_ONCE(1);
 		printk_once("basemono: %llu ts->next_tick: %llu dev->next_event: %llu timer->active: %d timer->expires: %llu\n",
 			    basemono, ts->next_tick, dev->next_event,
@@ -812,6 +819,7 @@ static ktime_t tick_nohz_stop_sched_tick(struct tick_sched *ts,
 	}
 
 	ts->next_tick = tick;
+	trace_printk("ts->next_tick = %llu\n", ts->next_tick);
 
 	/*
 	 * If the expiration time == KTIME_MAX, then we simply stop
@@ -894,6 +902,7 @@ static bool can_stop_idle_tick(int cpu, struct tick_sched *ts)
 		 * deadline if it comes back online later.
 		 */
 		ts->next_tick = 0;
+		trace_printk("ts->next_tick reset (offline)\n");
 		return false;
 	}
 
@@ -1202,8 +1211,10 @@ static enum hrtimer_restart tick_sched_timer(struct hrtimer *timer)
 	 */
 	if (regs)
 		tick_sched_handle(ts, regs);
-	else
+	else {
+		trace_printk("ts->next_tick reset (tick)\n");
 		ts->next_tick = 0;
+	}
 
 	/* No need to reprogram if we are in idle or full dynticks mode */
 	if (unlikely(ts->tick_stopped))

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web