On Sat, Mar 21, 2020 at 10:41 PM Rafael Aquini <aquini@xxxxxxxxxx> wrote:
>
> On Sat, Mar 21, 2020 at 09:31:42PM -0700, Andrew Morton wrote:
> > On Sat, 21 Mar 2020 22:03:26 -0400 Rafael Aquini <aquini@xxxxxxxxxx> wrote:
> >
> > > > > + * In order to sort out that race, and get the after fault checks consistent,
> > > > > + * the "quick and dirty" trick below is required in order to force a call to
> > > > > + * lru_add_drain_all() to get the recently MLOCK_ONFAULT pages moved to
> > > > > + * the unevictable LRU, as expected by the checks in this selftest.
> > > > > + */
> > > > > +static void force_lru_add_drain_all(void)
> > > > > +{
> > > > > + sched_yield();
> > > > > + system("echo 1 > /proc/sys/vm/compact_memory");
> > > > > +}
> > > >
> > > > What is the sched_yield() for?
> > > >
> > >
> > > Mostly it's there to provide a sleeping gap after the fault, whithout
> > > actually adding an arbitrary value with usleep().
> > >
> > > It's not a hard requirement, but, in some of the tests I performed
> > > (whithout that sleeping gap) I would still see around 1% chance
> > > of hitting the false-negative. After adding it I could not hit
> > > the issue anymore.
> >
> > It's concerning that such deep machinery as pagevec draining is visible
> > to userspace.
> >
> > I suppose that for consistency and correctness we should perform a
> > drain prior to each read from /proc/*/pagemap. Presumably this would
> > be far too expensive.
> >
> > Is there any other way? One such might be to make the MLOCK_ONFAULT
> > pages bypass the lru_add_pvecs?
> >
>
> Well,
>
> I admit I wasn't taking the approach of changing the kernel because I was
> thinking it would require a partial, or even full, revert of commit
> 9c4e6b1a7027f, and that would be increasing complexity, but on a
> second thought, it seems that we might just be missing:
>
> diff --git a/mm/swap.c b/mm/swap.c
> index cf39d24ada2a..b1601228ded4 100644
> --- a/mm/swap.c
> +++ b/mm/swap.c
> @@ -473,6 +473,7 @@ void lru_cache_add_active_or_unevictable(struct page *page,
> __mod_zone_page_state(page_zone(page), NR_MLOCK,
> hpage_nr_pages(page));
> count_vm_event(UNEVICTABLE_PGMLOCKED);
> + SetPageUnevictable(page);
No, this is not correct. Check __pagevec_lru_add_fn() for
TestClearPageUnevictable().
As I mentioned in the other email, I think the better solution would
be to introduce a sysctl to drain the pageves. That way there will not
be any dependency on compaction as was in your original patch.
Shakeel
