[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAOzc2pyZ9qHzEmNzYXjUu93wRpmW4af_e0eG=_AvNMzjU_V=CA@mail.gmail.com>
Date: Wed, 12 Oct 2022 09:02:36 -0700
From: Vishal Moola <vishal.moola@...il.com>
To: Matthew Wilcox <willy@...radead.org>
Cc: akpm@...ux-foundation.org, hughd@...gle.com,
linux-fsdevel@...r.kernel.org, linux-mm@...ck.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/4] filemap: find_lock_entries() now updates start offset
On Tue, Oct 11, 2022 at 7:10 PM Matthew Wilcox <willy@...radead.org> wrote:
>
> On Tue, Oct 11, 2022 at 02:56:31PM -0700, Vishal Moola (Oracle) wrote:
> > @@ -2116,7 +2118,16 @@ unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
> > folio_put(folio);
> > }
> > rcu_read_unlock();
> > + nr = folio_batch_count(fbatch);
> > +
> > + if (nr) {
> > + folio = fbatch->folios[nr - 1];
> > + nr = folio_nr_pages(folio);
> >
> > + if (folio_test_hugetlb(folio))
> > + nr = 1;
> > + *start = folio->index + nr;
> > + }
>
> Hmm ... this is going to go wrong if the folio is actually a shadow
> entry, isn't it?
You're right! I missed that.
> > +++ b/mm/shmem.c
> > @@ -922,21 +922,18 @@ static void shmem_undo_range(struct inode *inode, loff_t lstart, loff_t lend,
> >
> > folio_batch_init(&fbatch);
> > index = start;
> > - while (index < end && find_lock_entries(mapping, index, end - 1,
> > + while (index < end && find_lock_entries(mapping, &index, end - 1,
> > &fbatch, indices)) {
> > for (i = 0; i < folio_batch_count(&fbatch); i++) {
> > folio = fbatch.folios[i];
> >
> > - index = indices[i];
> > -
> > if (xa_is_value(folio)) {
> > if (unfalloc)
> > continue;
> > nr_swaps_freed += !shmem_free_swap(mapping,
> > - index, folio);
> > + folio->index, folio);
>
> We know this is a value entry, so we definitely can't look at
> folio->index. This should probably be:
>
> + indices[i], folio);
>
> > @@ -510,20 +509,18 @@ unsigned long invalidate_mapping_pagevec(struct address_space *mapping,
> > int i;
> >
> > folio_batch_init(&fbatch);
> > - while (find_lock_entries(mapping, index, end, &fbatch, indices)) {
> > + while (find_lock_entries(mapping, &index, end, &fbatch, indices)) {
> > for (i = 0; i < folio_batch_count(&fbatch); i++) {
> > struct folio *folio = fbatch.folios[i];
> >
> > /* We rely upon deletion not changing folio->index */
> > - index = indices[i];
> >
> > if (xa_is_value(folio)) {
> > count += invalidate_exceptional_entry(mapping,
> > - index,
> > - folio);
> > + folio->index,
> > + folio);
>
> Same here. I'd fix the indent while you're at it to get more on that
> second line and not need a third line.
>
Turns out I had misunderstood what a value entry was. I now understand
why we do in fact need the indices array. I'll fix the first 2 patches and
drop the last 2.
Powered by blists - more mailing lists