Messages in this thread | | | Subject | Re: [PATCH 1/3] stop_machine: kill __stop_machine() | From | Peter Zijlstra <> | Date | Thu, 16 Jun 2011 20:28:17 +0200 |
| |
On Thu, 2011-06-16 at 11:17 -0700, Suresh Siddha wrote:
> > > Peter, So it looks like we are allowing a new cpu to come online in > > > parallel, while we poke the text? Isn't it a problem? What am I missing? > > > > the caller already did get_online_cpus(), > > > > do_optimize_kprobes() > > get_online_cpus() > > arch_optimize_kprobes() > > text_poke_smp_batch() > > put_online_cpus() > > So the circular dependency reported is not possible in practice right?
why not? get_online_cpus() takes a mutex, that connects smp_alt and text_mutex and can cause a deadlock.
| |