On Sat, Feb 01, 2025 at 04:01:43PM +0800, Kairui Song wrote:
> On Thu, Jan 30, 2025 at 6:02 PM Kirill A. Shutemov
> <kirill.shutemov@xxxxxxxxxxxxxxx> wrote:
> >
> > The recently introduced PG_dropbehind allows for freeing folios
> > immediately after writeback. Unlike PG_reclaim, it does not need vmscan
> > to be involved to get the folio freed.
> >
> > Instead of using folio_set_reclaim(), use folio_set_dropbehind() in
> > pageout().
> >
> > It is safe to leave PG_dropbehind on the folio if, for some reason
> > (bug?), the folio is not in a writeback state after ->writepage().
> > In these cases, the kernel had to clear PG_reclaim as it shared a page
> > flag bit with PG_readahead.
> >
> > Signed-off-by: Kirill A. Shutemov <kirill.shutemov@xxxxxxxxxxxxxxx>
> > Acked-by: David Hildenbrand <david@xxxxxxxxxx>
> > ---
> > mm/vmscan.c | 9 +++------
> > 1 file changed, 3 insertions(+), 6 deletions(-)
> >
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index bc1826020159..c97adb0fdaa4 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -692,19 +692,16 @@ static pageout_t pageout(struct folio *folio, struct address_space *mapping,
> > if (shmem_mapping(mapping) && folio_test_large(folio))
> > wbc.list = folio_list;
> >
> > - folio_set_reclaim(folio);
> > + folio_set_dropbehind(folio);
> > +
> > res = mapping->a_ops->writepage(&folio->page, &wbc);
> > if (res < 0)
> > handle_write_error(mapping, folio, res);
> > if (res == AOP_WRITEPAGE_ACTIVATE) {
> > - folio_clear_reclaim(folio);
> > + folio_clear_dropbehind(folio);
> > return PAGE_ACTIVATE;
> > }
> >
> > - if (!folio_test_writeback(folio)) {
> > - /* synchronous write or broken a_ops? */
> > - folio_clear_reclaim(folio);
> > - }
> > trace_mm_vmscan_write_folio(folio);
> > node_stat_add_folio(folio, NR_VMSCAN_WRITE);
> > return PAGE_SUCCESS;
> > --
> > 2.47.2
> >
>
> Hi, I'm seeing following panic with SWAP after this commit:
>
> [ 29.672319] Oops: general protection fault, probably for
> non-canonical address 0xffff88909a3be3: 0000 [#1] PREEMPT SMP NOPTI
> [ 29.675503] CPU: 82 UID: 0 PID: 5145 Comm: tar Kdump: loaded Not
> tainted 6.13.0.ptch-g1fe9ea48ec98 #917
> [ 29.677508] Hardware name: Red Hat KVM/RHEL-AV, BIOS 0.0.0 02/06/2015
> [ 29.678886] RIP: 0010:__lock_acquire+0x20/0x15d0
Ouch.
I failed to trigger it my setup. Could you share your reproducer?
> I'm testing with PROVE_LOCKING on. It seems folio_unmap_invalidate is
> called for swapcache folio and it doesn't work well, following PATCH
> on top of mm-unstable seems fix it well:
Right. I don't understand swapping good enough. I missed this.
> diff --git a/mm/filemap.c b/mm/filemap.c
> index 4fe551037bf7..98493443d120 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -1605,8 +1605,9 @@ static void folio_end_reclaim_write(struct folio *folio)
> * invalidation in that case.
> */
> if (in_task() && folio_trylock(folio)) {
> - if (folio->mapping)
> - folio_unmap_invalidate(folio->mapping, folio, 0);
> + struct address_space *mapping = folio_mapping(folio);
> + if (mapping)
> + folio_unmap_invalidate(mapping, folio, 0);
> folio_unlock(folio);
> }
> }
Once you do this, folio_unmap_invalidate() will never succeed for
swapcache as folio->mapping != mapping check will always be true and it
will fail with -EBUSY.
I guess we need to do something similar to what __remove_mapping() does
for swapcache folios.
--
Kiryl Shutsemau / Kirill A. Shutemov
