Re: [PATCH 0/4] RCU: various merge candidates

[Dipankar Sarma <dipankar@in.ibm.com>]

On Mon, Aug 28, 2006 at 12:40:58PM -0700, Andrew Morton wrote:
> On Tue, 29 Aug 2006 00:46:42 +0530
> Dipankar Sarma <dipankar@in.ibm.com> wrote:
> 
> >
> > rcutorture fix patches independent of rcu implementation changes
> > in this patchset.
> 
> So this patchset is largely orthogonal to the presently-queued stuff?

Yes, it should be.

> > > Now what?
> > 
> > Heh. I can always re-submit against -mm after I wait for a day or two
> > for comments :)
> 
> That would be good, thanks.  We were seriously considering merging all the
> SRCU stuff for 2.6.18, because

I think non-srcu rcutorture patches can be merged in 2.6.19. srcu
is a tossup. Perhaps srcu and this patchset may be merge candidates
for 2.6.20 should things go well in review and testing. Should I re-submit
against 2.6.18-mm1 or so (after your patchset reduces in size) ?
What is a convenient time ?

> cpufreq-make-the-transition_notifier-chain-use-srcu.patch fixes a cpufreq
> down()-in-irq-disabled warning at suspend time.
> 
> But that's a lot of new stuff just to fix a warning about something which
> won't actually cause any misbehaviour.  We could just as well do
> 
> 	if (irqs_disabled())
> 		down_read_trylock(...);	/* suspend */
> 	else
> 		down_read(...);
> 
> in cpufreq to temporarily shut the thing up.

GAh! cpufreq. Already I am having to look at all of cpufreq and the
cpufreq drivers, change notifiers for the whole locking model for
the other (hotplug) cleanup. I will keep this in mind.

Thanks
Dipankar