Re: [PATCH v2 5/5] mm/page_alloc: Only call pgdat_set_deferred_range when the system boots

[Pavel Tatashin]

On Thu, Jul 19, 2018 at 10:03 AM Michal Hocko <mhocko@xxxxxxxxxx> wrote:
>
> On Thu 19-07-18 15:58:59, Oscar Salvador wrote:
> > On Thu, Jul 19, 2018 at 03:46:22PM +0200, Michal Hocko wrote:
> > > On Thu 19-07-18 15:27:40, osalvador@xxxxxxxxxxxxxxxxxx wrote:
> > > > From: Oscar Salvador <osalvador@xxxxxxx>
> > > >
> > > > We should only care about deferred initialization when booting.
> > >
> > > Again why is this worth doing?
> >
> > Well, it is not a big win if that is what you meant.
> >
> > Those two fields are only being used when dealing with deferred pages,
> > which only happens at boot time.
> >
> > If later on, free_area_init_node gets called from memhotplug code,
> > we will also set the fields, although they will not be used.
> >
> > Is this a problem? No, but I think it is more clear from the code if we
> > see when this is called.
> > So I would say it was only for code consistency.
>
> Then put it to the changelog.
>
> > If you think this this is not worth, I am ok with dropping it.
>
> I am not really sure. I am not a big fan of SYSTEM_BOOTING global
> thingy so I would rather not spread its usage.

I agree, I do not think this patch is necessary. Calling
pgdat_set_deferred_range() does not hurt in hotplug context, and it is
cheap too. SYSTEM_BOOTING sometimes useful, but it is better to use it
only where necessary, where without this "if" we will encounter some
bugs.

Thank you,
Pavel