2005-10-24 16:07:36

by Mark Knecht

[permalink] [raw]
Subject: 2.6.14-rc5-rt5 - softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103

Hi,
This is on an older machine. Athlon 1600+ / older Via chipset.


ACPI: Power Button (FF) [PWRF]
ACPI: Power Button (CM) [PWRB]
ACPI: CPU0 (power states: C1[C1] C2[C2])
ACPI: Processor [CPU0] (supports 16 throttling states)
Time: tsc clocksource has been installed.
WARNING: non-monotonic time!
... time warped from 151976744 to 147973105.
softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103
[<c0118e20>] __WARN_ON+0x60/0xc0 (8)
[<c013072d>] ktime_get+0xed/0x130 (40)
[<c01312de>] ktimer_run_queues+0x2e/0x130 (64)
[<c01217f9>] run_timer_softirq+0xc9/0x3e0 (60)
[<c03845ff>] schedule+0x4f/0x120 (16)
[<c011d837>] ksoftirqd+0xb7/0x110 (28)
[<c011d780>] ksoftirqd+0x0/0x110 (36)
[<c012d29a>] kthread+0xba/0xc0 (4)
[<c012d1e0>] kthread+0x0/0xc0 (28)
[<c01013d5>] kernel_thread_helper+0x5/0x10 (16)
WARNING: non-monotonic time!
... time warped from 151976744 to 148971665.
softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103
[<c0118e20>] __WARN_ON+0x60/0xc0 (8)
[<c013072d>] ktime_get+0xed/0x130 (40)
[<c01312de>] ktimer_run_queues+0x2e/0x130 (64)
[<c01217f9>] run_timer_softirq+0xc9/0x3e0 (60)
[<c03845ff>] schedule+0x4f/0x120 (16)
[<c011d837>] ksoftirqd+0xb7/0x110 (28)
[<c011d780>] ksoftirqd+0x0/0x110 (36)
[<c012d29a>] kthread+0xba/0xc0 (4)
[<c012d1e0>] kthread+0x0/0xc0 (28)
[<c01013d5>] kernel_thread_helper+0x5/0x10 (16)
WARNING: non-monotonic time!
... time warped from 151976744 to 149972579.
softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103
[<c0118e20>] __WARN_ON+0x60/0xc0 (8)
[<c013072d>] ktime_get+0xed/0x130 (40)
[<c01312de>] ktimer_run_queues+0x2e/0x130 (64)
[<c01217f9>] run_timer_softirq+0xc9/0x3e0 (60)
[<c03845ff>] schedule+0x4f/0x120 (16)
[<c011d837>] ksoftirqd+0xb7/0x110 (28)
[<c011d780>] ksoftirqd+0x0/0x110 (36)
[<c012d29a>] kthread+0xba/0xc0 (4)
[<c012d1e0>] kthread+0x0/0xc0 (28)
[<c01013d5>] kernel_thread_helper+0x5/0x10 (16)
lp: driver loaded but no devices found
PNP: PS/2 Controller [PNP0303:PS2K,PNP0f13:PS2M] at 0x60,0x64 irq 1,12
serio: i8042 AUX port at 0x60,0x64 irq 12


Attachments:
(No filename) (2.04 kB)
knecht.config-2.6.14-rc5-rt5.bz2 (6.97 kB)
Download all attachments

2005-10-24 17:06:05

by Steven Rostedt

[permalink] [raw]
Subject: Re: 2.6.14-rc5-rt5 - softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103

On Mon, 2005-10-24 at 09:07 -0700, Mark Knecht wrote:

> Time: tsc clocksource has been installed.
> WARNING: non-monotonic time!
> ... time warped from 151976744 to 147973105.
> softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103

Hi Mark,

Yeah, I saw this too, and it went through to Thomas Gleixner and then to
John Stultz, where he said that he may have fixed this in his latest
version. So we are now waiting on Thomas to pull John's work into the
ktimers code, and then onto Ingo's RT base.

So, until then, you may just ignore the messages. It should only happen
when the tsc clocksource changes.

Unless, of course -rt5 already has this (I just got back from Germany,
so I've been out of the loop).

Thomas, this hasn't been done yet, has it?

-- Steve


2005-10-24 18:13:17

by Ingo Molnar

[permalink] [raw]
Subject: Re: 2.6.14-rc5-rt5 - softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103


* Steven Rostedt <[email protected]> wrote:

> > Time: tsc clocksource has been installed.
> > WARNING: non-monotonic time!
> > ... time warped from 151976744 to 147973105.
> > softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103
>
> Hi Mark,
>
> Yeah, I saw this too, and it went through to Thomas Gleixner and then
> to John Stultz, where he said that he may have fixed this in his
> latest version. So we are now waiting on Thomas to pull John's work
> into the ktimers code, and then onto Ingo's RT base.
>
> So, until then, you may just ignore the messages. It should only
> happen when the tsc clocksource changes.
>
> Unless, of course -rt5 already has this (I just got back from Germany,
> so I've been out of the loop).
>
> Thomas, this hasn't been done yet, has it?

not yet, but it's in progress. Meanwhile i increased the number of times
the warning will be printed per bootup (from 1 to 3), so that if a time
warp happens outside of that clock-switch case it should be printed too.

Ingo

2005-10-24 18:40:31

by Steven Rostedt

[permalink] [raw]
Subject: Re: 2.6.14-rc5-rt5 - softirq-timer/0/3[CPU#0]: BUG in ktime_get at kernel/ktimers.c:103

On Mon, 2005-10-24 at 20:13 +0200, Ingo Molnar wrote:
> * Steven Rostedt <[email protected]> wrote:

> >
> > Thomas, this hasn't been done yet, has it?
>
> not yet, but it's in progress. Meanwhile i increased the number of times
> the warning will be printed per bootup (from 1 to 3), so that if a time
> warp happens outside of that clock-switch case it should be printed too.
>

Unfortunately, the problem doesn't go away until the clocks are
eventually updated. So as Mark got, we see three outputs for every time
this happens until the now catches up with the prev.

Perhaps we should also add another update of prev?

-- Steve

Index: kernels/linux-2.6.14-rc5-rt5/kernel/ktimers.c
===================================================================
--- kernels.orig/linux-2.6.14-rc5-rt5/kernel/ktimers.c 2005-10-24 14:34:32.000000000 -0400
+++ kernels/linux-2.6.14-rc5-rt5/kernel/ktimers.c 2005-10-24 14:34:53.000000000 -0400
@@ -102,6 +102,7 @@
ktime_to_ns(prev), ktime_to_ns(now));
WARN_ON(1);
}
+ per_cpu(prev_mono_time, cpu) = now;
return prev;
}
per_cpu(prev_mono_time, cpu) = now;