Re: [patch 3/3] clockevents: Fix resume logic - updated version

[Andrew Morton <akpm@linux-foundation.org>]

On Sun, 13 May 2007 18:48:07 +0200 Thomas Gleixner <tglx@linutronix.de> wrote:

> On Sun, 2007-05-13 at 10:07 +0200, Thomas Gleixner wrote:
> > > > Can you send me your .config please ?
> > > > 
> > > 
> > > http://userweb.kernel.org/~akpm/config-sony.txt
> > 
> > No luck. 
> 
> I'm making progress. After fiddling with the config options I get a
> solid lockup of netconsole during boot :(

neat.  Sometimes this can be a device driver failure.

> But also lockdep itself is broken on my box, so I would not see warnings
> anyway:
> 
>      soft-irqs-on + irq-safe-A/21:  ok  |  ok  |  ok  |
>        sirq-safe-A => hirqs-on/12:  ok  |  ok  |irq event stamp: 472
> hardirqs last  enabled at (472): [<c01e8bba>] irqsafe2A_rlock_12+0x94/0xa1
> hardirqs last disabled at (471): [<c0109541>] native_sched_clock+0x55/0xe8
> softirqs last  enabled at (468): [<c01e8ba5>] irqsafe2A_rlock_12+0x7f/0xa1
> softirqs last disabled at (464): [<c01e8b31>] irqsafe2A_rlock_12+0xb/0xa1
> FAILED| [<c0104d21>] dump_trace+0x61/0x1e7
>  [<c0104ec1>] show_trace_log_lvl+0x1a/0x30
>  [<c0105a1b>] show_trace+0x12/0x14
>  [<c0105a73>] dump_stack+0x15/0x17
>  [<c01e6971>] dotest+0x6b/0x3ce
>  [<c01efc1f>] locking_selftest+0x915/0x1a5a
>  [<c04a0a0a>] start_kernel+0x1d5/0x2a7
>  =======================
> 
> .... more of those
> 
> -----------------------------------------------------------------
> BUG:   6 unexpected failures (out of 218) - debugging disabled! |
> -----------------------------------------------------------------
> 

Yeah, I expect quite a few people will start seeing that.  iirc it was
triggered by a couple of changes: a local_irq_save/local_irq_restore in
sched_clock() and a change Jeremy made to the softlockup detector.

There was no reasonable explanation why either of those changes should have
triggered the six failures so I just went ahead with them so we can flush
the real bug out.