On Tue, Sep 3, 2024 at 11:51 AM Kairui Song <ryncsn@xxxxxxxxx> wrote:
>
> On Wed, Sep 4, 2024 at 2:22 AM Yosry Ahmed <yosryahmed@xxxxxxxxxx> wrote:
> >
>
> Hi Yosry,
>
> > > > > diff --git a/mm/swapfile.c b/mm/swapfile.c
> > > > > index c1638a009113..8ff58be40544 100644
> > > > > --- a/mm/swapfile.c
> > > > > +++ b/mm/swapfile.c
> > > > > @@ -1514,6 +1514,8 @@ static bool __swap_entries_free(struct swap_info_struct *si,
> > > > > unlock_cluster_or_swap_info(si, ci);
> > > > >
> > > > > if (!has_cache) {
> > > > > + for (i = 0; i < nr; i++)
> > > > > + zswap_invalidate(swp_entry(si->type, offset + i));
> > > > > spin_lock(&si->lock);
> > > > > swap_entry_range_free(si, entry, nr);
> > > > > spin_unlock(&si->lock);
> > > > >
>
> This fix from Barry have been applied for mm-unstable and it's looking
> good so far.
Are we sure it fixes the problem? Unfortunately we don't have a
reproducer. The swap entry freeing paths are getting too tangled to
know for sure if zswap_invalidate() is being called in all cases,
which is why I prefer having a centralized callback for all swap entry
freeing hooks.
>
> > Kairui, Barry, any thoughts on this? Any preferences on how to make
> > sure zswap_invalidate() is being called in all possible swap freeing
> > paths?
>
> I have a set of patched that removed the si->lock around
> swap_entry_range_free, and the slot return cache is removed too. So we
> can move zswap_invalidate into it as locking or caching is no longer
> an issue, I will post it ASAP.
That sounds good, but I am not sure I understand what you mean by
removing the slot return cache. I will wait to see the patches.
