Re: [git pull] core fixes

[Nick Piggin <nickpiggin@yahoo.com.au>]

On Tuesday 12 August 2008 17:17, Peter Zijlstra wrote:
> On Tue, 2008-08-12 at 16:13 +1000, Nick Piggin wrote:
> > On Tuesday 12 August 2008 08:20, Ingo Molnar wrote:
> > > Nick Piggin (1):
> > >       generic-ipi: fix stack and rcu interaction bug in
> > > smp_call_function_mask()
> >
> > I'm still not 100% sure that I have this patch right... I might have seen
> > a lockup trace implicating the smp call function path... which may have
> > been due to some other problem or a different bug in the new call
> > function code, but if some more people can take a look at it before
> > merging?
>
> Right - so we cannot use synchronize_rcu() because the caller of
> smp_call_function_mask() might not be in a preemptible context.
>
> Therefore you implement this barrier like function, that uses the single
> call ipi to validate that all the cpus are done processing the
> call_function_queue - because that is with IRQs disabled, and this other
> IPI cannot interrupt. Thereby guaranteeing that there are no more
> references to any former elements on said list.
>
> Clever. But as you say, rather expensive.

Yeah... Aside, I think this is how some OSes implement a lot of
scalable read side primitives without RCU. Downside is that it
requires interrupt disabled read side, upside is much simpler and
potentially more deterministic to quiesce. (and fewer patents I
guess!)