Re: [PATCH] mm/page_alloc: Add lockdep assertion for pageblock type change

[Brendan Jackman <jackmanb@google.com>]

On Thu, Feb 27, 2025 at 02:33:02PM -0800, Andrew Morton wrote:
> On Thu, 27 Feb 2025 16:15:47 +0000 Brendan Jackman <jackmanb@google.com> wrote:
> 
> > Since the migratetype hygiene patches [0], the locking here is
> > a bit more formalised, so write it down with an assert.
> > 
> > ...
> >
> > --- a/mm/page_alloc.c
> > +++ b/mm/page_alloc.c
> > @@ -417,6 +417,10 @@ void set_pfnblock_flags_mask(struct page *page, unsigned long flags,
> >  
> >  void set_pageblock_migratetype(struct page *page, int migratetype)
> >  {
> > +	lockdep_assert_once(system_state == SYSTEM_BOOTING ||
> > +		in_mem_hotplug() ||
> > +		lockdep_is_held(&page_zone(page)->lock));
> > +
> >  	if (unlikely(page_group_by_mobility_disabled &&
> >  		     migratetype < MIGRATE_PCPTYPES))
> >  		migratetype = MIGRATE_UNMOVABLE;
> > 
> 
> We could add such assertions all over the place.  Why this place in
> particular?

For the other stuff, it's pretty obvious that it would be protected by
the zone lock (or, I don't know about it!). But it didn't seem totally
self-evident to me that it should protect the pageblock type. So it
seems particularly helpful to have it written in the code.

I may be heavily biased about this though, because of the code I'm
working on for [0]. I use the pageblock type to remember whether it's
mapped in the ASI restricted address space.

[0] https://lore.kernel.org/linux-mm/CA+i-1C1gOBLxRxE5YFGzeayYWBYyE_X6oH4D=9eVePt4=ehTig@mail.gmail.com/T/