Re: [PATCH -v4] generic-ipi: remove kmalloc()

[Oleg Nesterov <oleg@redhat.com>]

On 02/17, Peter Zijlstra wrote:
>
> On Tue, 2009-02-17 at 21:16 +0100, Peter Zijlstra wrote:
> > On Tue, 2009-02-17 at 21:11 +0100, Peter Zijlstra wrote:
> > >
> > > > As for the previous version. I am almost sure I missed something
> > > > again, but do we really need both CSD_FLAG_WAIT and CSD_FLAG_LOCK
> > > > flags? smp_call_function_many(wait => 1) can just wait for
> > > > !CSD_FLAG_LOCK.
> > >
> > > I suppose we can.
> >
> > Ah, no, that would mean we cannot distinguish between the sync and async
> > case in generic_exec_single().
>
> Which can of course be solved like so, we could even drop the new .wait
> argument to __smp_call_function_single() as CSD_FLAG_WAIT wasn't exposed
> before and non of the current users use it.

Yes, I think this should work.

Or, if you prefer to stay with with 2 flags for now, then perhaps it makes
sense to at least clear both flags in xxx_interrupt() at once.

I guess you should re-send your previous patch in any case,

		} else if ((unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) {
	-               struct call_single_data *data;
	+               struct call_single_data *data = {
	+                       .flags = 0,
	+               };

this looks a bit suspicious ;)

Oleg.