Re: RCU?: Scheduling while atomic with 3.4.23-rt33

From: Tim Sander
Date: Mon Feb 18 2013 - 08:36:23 EST

This is a multi-part message in MIME format.
Am Montag, 18. Februar 2013, 12:16:41 schrieb Thomas Gleixner:
> On Mon, 18 Feb 2013, Tim Sander wrote:
> > Here is a transtribed backtrace for easier reading:
> > c0010fac t dump_backtrace
> > c022c188 T dump_stack
> > c022c71c t __schedule_bug
> > c022fb28 t __schedule
> > c0230644 t rt_spin_lock_slowlock
> > c0230ba8 T rt_spin_lock
> > c003a9b8 T __wake_up
> > c0052a44 t invoke_rcu_callbacks
> > c0053764 T rcu_check_callbacks
> > c0026060 T update_process_times
> That's due to CONFIG_TINY_PREEMPT_RCU=y. That's a known issue, which
> has been fixed in 3.6-rt.
> Steven, could you backport that to 3.2/3.4 please ?
Mh, i just saw that i forgot to add the other patch which adds the wait-simple
stuff. Its attached so thats hopefully all to get rid of that error?

Local smoke test is running... but as its a relative rare race i'm not holding
my breath.

Btw. is that wait-simple stuff safe from irq context?

Best regards

PS: sorry for the legalese in the first mail, its just for Steven to keep him
learning german legalese ;-).