Received-SPF: pass (google.com: domain of linux-nfs+bounces-3856-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Precedence: bulk
MIME-Version: 1.0
References: <20240614100329.1203579-1-hch@lst.de> <20240614100329.1203579-2-hch@lst.de>
 <20240614112148.cd1961e84b736060c54bdf26@linux-foundation.org>
In-Reply-To: <20240614112148.cd1961e84b736060c54bdf26@linux-foundation.org>
From: Barry Song <21cnbao@gmail.com>
Date: Sun, 16 Jun 2024 12:16:10 +1200
Message-ID: <CAGsJ_4wnWzoScqO9_NddHcDPbe_GbAiRFVm4w_H+QDmH=e=Rsw@mail.gmail.com>
Subject: Re: [PATCH] nfs: fix nfs_swap_rw for large-folio swap
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Christoph Hellwig <hch@lst.de>, Trond Myklebust <trondmy@kernel.org>, Anna Schumaker <anna@kernel.org>, 
	Steve French <sfrench@samba.org>, linux-nfs@vger.kernel.org, linux-cifs@vger.kernel.org, 
	linux-mm@kvack.org, Barry Song <v-songbaohua@oppo.com>, 
	Ryan Roberts <ryan.roberts@arm.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Sun, Jun 16, 2024 at 5:55=E2=80=AFAM Andrew Morton <akpm@linux-foundatio=
n.org> wrote:
>
> On Fri, 14 Jun 2024 12:03:25 +0200 Christoph Hellwig <hch@lst.de> wrote:
>
> > As of Linux 6.10-rc the MM can swap out larger than page size chunks.
> > NFS has all code ready to handle this, but has a VM_BUG_ON that
> > triggers when this happens.  Simply remove the VM_BUG_ON to fix this
> > use case.

As I understand it, this isn't happening because we don't support
mTHP swapping out to a swapfile, whether it's on NFS or any
other filesystem.
static int scan_swap_map_slots(struct swap_info_struct *si,
                               unsigned char usage, int nr,
                               swp_entry_t slots[], int order)
{
        ...
        if (order > 0) {
                ...

                /*
                 * Swapfile is not block device or not using clusters so un=
able
                 * to allocate large entries.
                 */
                if (!(si->flags & SWP_BLKDEV) || !si->cluster_info)
                        return 0;
        }
}

However, I'm pleased to see this patch, as we might need it someday.

> >
> > ...
> >
> > --- a/fs/nfs/direct.c
> > +++ b/fs/nfs/direct.c
> > @@ -141,8 +141,6 @@ int nfs_swap_rw(struct kiocb *iocb, struct iov_iter=
 *iter)
> >  {
> >       ssize_t ret;
> >
> > -     VM_BUG_ON(iov_iter_count(iter) !=3D PAGE_SIZE);
> > -
> >       if (iov_iter_rw(iter) =3D=3D READ)
> >               ret =3D nfs_file_direct_read(iocb, iter, true);
> >       else
>
> I'm thinking this should precede "mm: swap: entirely map large folios
> found in swapcache", or be a part of it.
>
> Barry/Chuanhua, any opinions?

The patch =E2=80=9Cmm: swap: entirely map large folios found in swapcache=
=E2=80=9D does not
perform any I/O operations; it only maps the accessed large folio within th=
e
swapcache. Therefore, this patch can be considered independent.

BTW, we haven't supported swapping out mTHP to nfs swapfile.

So, this patch is future-proof and can serve as a precedent if we
want to extend mthp swapout/swpin to non-blkdev systems.
Please correct me if I'm wrong, Ryan.

Thanks
Barry