Re: lockdep splat in CPU hotplug

[Peter Zijlstra <peterz@infradead.org>]

On Wed, Oct 22, 2014 at 11:53:49AM +0200, Jiri Kosina wrote:
> > The reason for CCing Ingo and Peter is that I can't make any sense of one 
> > of the stacktraces lockdep is providing.
> > 
> > Please have a look at the very first stacktrace in the dump, where lockdep 
> > is trying to explain where cpu_hotplug.lock#2 has been acquired. It seems 
> > to imply that cpuidle_pause() is taking cpu_hotplug.lock, but that's not 
> > the case at all.
> > 
> > What am I missing?

> Still, the lockdep stacktrace is bogus and didn't really help 
> understanding this. Any idea why it's wrong?

> > -> #1 (cpu_hotplug.lock#2){+.+.+.}:
> >         [<ffffffff81099fac>] lock_acquire+0xac/0x130
> >         [<ffffffff815b9f2c>] mutex_lock_nested+0x5c/0x3b0
> >         [<ffffffff81491892>] cpuidle_pause+0x12/0x30
> >         [<ffffffff81402314>] dpm_suspend_noirq+0x44/0x340
> >         [<ffffffff81402958>] dpm_suspend_end+0x38/0x80
> >         [<ffffffff810a07bd>] hibernation_snapshot+0xcd/0x370
> >         [<ffffffff810a1248>] hibernate+0x168/0x210
> >         [<ffffffff8109e9b4>] state_store+0xe4/0xf0
> >         [<ffffffff813003ef>] kobj_attr_store+0xf/0x20
> >         [<ffffffff8121e9a3>] sysfs_kf_write+0x43/0x60
> >         [<ffffffff8121e287>] kernfs_fop_write+0xe7/0x170
> >         [<ffffffff811a7342>] vfs_write+0xb2/0x1f0
> >         [<ffffffff811a7da4>] SyS_write+0x44/0xb0
> >         [<ffffffff815be856>] system_call_fastpath+0x16/0x1b

Right, so I've seen it more often, and I'm not sure I can explain
either.

Lockdep uses save_stack_trace() with trace->skip=3, typically if you
get ->skip wrong you'd not even see the lock_acquire, so that can't be
it.

The only thing I can come up with is that for some reason the
intermediate entries are !reliable, save_stack_trace() skips those for
CONFIG_FRAME_POINTER=y, see
arch/x86/kernel/stacktrace.c:__save_stack_address().