Re: [PATCH RFC 01/10] rcu: Make CONFIG_SRCU unconditionally enabled

From: Michal Hocko
Date: Wed Aug 08 2018 - 06:27:42 EST

[CC Josh - the whole series is]

On Wed 08-08-18 13:17:44, Kirill Tkhai wrote:
> On 08.08.2018 10:20, Michal Hocko wrote:
> > On Tue 07-08-18 18:37:36, Kirill Tkhai wrote:
> >> This patch kills all CONFIG_SRCU defines and
> >> the code under !CONFIG_SRCU.
> >
> > The last time somebody tried to do this there was a pushback due to
> > kernel tinyfication. So this should really give some numbers about the
> > code size increase. Also why can't we make this depend on MMU. Is
> > anybody else than the reclaim asking for unconditional SRCU usage?
> I don't know one. The size numbers (sparc64) are:
> $ size image.srcu.disabled
> text data bss dec hex filename
> 5117546 8030506 1968104 15116156 e6a77c image.srcu.disabled
> $ size image.srcu.enabled
> text data bss dec hex filename
> 5126175 8064346 1968104 15158625 e74d61 image.srcu.enabled
> The difference is: 15158625-15116156 = 42469 ~41Kb
> Please, see the measurement details to my answer to Stephen.
> > Btw. I totaly agree with Steven. This is a very poor changelog. It is
> > trivial to see what the patch does but it is far from clear why it is
> > doing that and why we cannot go other ways.
> We possibly can go another way, and there is comment to [2/10] about this.
> Percpu rwsem may be used instead, the only thing, it is worse, is it will
> make shrink_slab() wait unregistering shrinkers, while srcu-based
> implementation does not require this.

Well, if unregisterring doesn't do anything subtle - e.g. an allocation
or take locks which depend on allocation - and we can guarantee that
then blocking shrink_slab shouldn't be a big deal. It is subtle though.
Maybe subtle enough to make unconditional SRCU worth it. This all should
be in the changelog though.

> This may be not a big problem.
> But, if SRCU is real problem for embedded people, I really don't want they
> hate me in the future because of this, so please CC someone if you know :)

I guess Josh was trying to pursue kernel tinification.

Michal Hocko