Re: [PATCHv7 13/18] mm: Drop fake head checks

[Peter Zijlstra]

On Tue, Oct 27, 2026 at 01:47:17PM +0100, Peter Zijlstra wrote:
> On Tue, Oct 27, 2026 at 12:31:09PM +0000, Kiryl Shutsemau wrote:
> > On Tue, Oct 27, 2026 at 01:05:58PM +0100, David Hildenbrand (Arm) wrote:
> > > > Writing this, I just realized that we have another option -- map the
> > > > shared tail page writable. A failing LOCK CMPXCHG writes back the value
> > > > it read, and the get never changes a zero refcount, so the race becomes
> > > > harmless. The cost is the protection: the page is shared by every
> > > > hugetlb folio of that order in the zone, so a stray write to any tail
> > > > silently corrupts all of them instead of oopsing.
> > > 
> > > Right. Without HVO we would just corrupt pages in the single folio (which is
> > > also not nice).
> > > 
> > > I'm also questioning whether we need this RO protection all the time active.
> > 
> > What do you have in mind? RO only under a debug option, or switching the
> > tail page to RO some time after the remap?
> > 
> > If it is the delay, a grace period alone does not close the race. We
> > would need rcu_read_lock() back around the get, and then it is the old
> > scheme with the wait taken off the allocation path.
> 
> So it used to be; back when Nick introduced try_get_page, that it was
> only allowed with IRQs disabled.
> 
> And looking at gup.c now, I see gup_fast_pgd_range() still has
> local_irq_save/restore around it.
> 
> So what do you need that rcu crud for?

Also; as long as the synchronize_rcu() only lives when a debug option is
on, that sounds like a win. Debug stuff has costs, news at 11 and all
that.