On Fri, 2025-07-04 at 09:44 +1000, NeilBrown wrote:
> On Fri, 04 Jul 2025, Jeff Layton wrote:
> > Recent testing has shown that keeping pagecache pages around for too
> > long can be detrimental to performance with nfsd. Clients only rarely
> > revisit the same data, so the pages tend to just hang around.
> >
> > This patch changes the pc_release callbacks for NFSv3 READ, WRITE and
> > COMMIT to call generic_fadvise(..., POSIX_FADV_DONTNEED) on the accessed
> > range.
> >
> > Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > ---
> > fs/nfsd/debugfs.c | 2 ++
> > fs/nfsd/nfs3proc.c | 59 +++++++++++++++++++++++++++++++++++++++++++++---------
> > fs/nfsd/nfsd.h | 1 +
> > fs/nfsd/nfsproc.c | 4 ++--
> > fs/nfsd/vfs.c | 21 ++++++++++++++-----
> > fs/nfsd/vfs.h | 5 +++--
> > fs/nfsd/xdr3.h | 3 +++
> > 7 files changed, 77 insertions(+), 18 deletions(-)
> >
> > diff --git a/fs/nfsd/debugfs.c b/fs/nfsd/debugfs.c
> > index 84b0c8b559dc90bd5c2d9d5e15c8e0682c0d610c..b007718dd959bc081166ec84e06f577a8fc2b46b 100644
> > --- a/fs/nfsd/debugfs.c
> > +++ b/fs/nfsd/debugfs.c
> > @@ -44,4 +44,6 @@ void nfsd_debugfs_init(void)
> >
> > debugfs_create_file("disable-splice-read", S_IWUSR | S_IRUGO,
> > nfsd_top_dir, NULL, &nfsd_dsr_fops);
> > + debugfs_create_bool("enable-fadvise-dontneed", 0644,
> > + nfsd_top_dir, &nfsd_enable_fadvise_dontneed);
> > }
> > diff --git a/fs/nfsd/nfs3proc.c b/fs/nfsd/nfs3proc.c
> > index b6d03e1ef5f7a5e8dd111b0d56c061f1e91abff7..11261cf67ea817ec566626f08b733e09c9e121de 100644
> > --- a/fs/nfsd/nfs3proc.c
> > +++ b/fs/nfsd/nfs3proc.c
> > @@ -9,6 +9,7 @@
> > #include <linux/ext2_fs.h>
> > #include <linux/magic.h>
> > #include <linux/namei.h>
> > +#include <linux/fadvise.h>
> >
> > #include "cache.h"
> > #include "xdr3.h"
> > @@ -206,11 +207,25 @@ nfsd3_proc_read(struct svc_rqst *rqstp)
> >
> > fh_copy(&resp->fh, &argp->fh);
> > resp->status = nfsd_read(rqstp, &resp->fh, argp->offset,
> > - &resp->count, &resp->eof);
> > + &resp->count, &resp->eof, &resp->nf);
> > resp->status = nfsd3_map_status(resp->status);
> > return rpc_success;
> > }
> >
> > +static void
> > +nfsd3_release_read(struct svc_rqst *rqstp)
> > +{
> > + struct nfsd3_readargs *argp = rqstp->rq_argp;
> > + struct nfsd3_readres *resp = rqstp->rq_resp;
> > +
> > + if (nfsd_enable_fadvise_dontneed && resp->status == nfs_ok)
> > + generic_fadvise(nfsd_file_file(resp->nf), argp->offset, resp->count,
> > + POSIX_FADV_DONTNEED);
> > + if (resp->nf)
> > + nfsd_file_put(resp->nf);
> > + fh_put(&resp->fh);
>
> This looks wrong - testing resp->nf after assuming it was non-NULL.
> I don't think it *is* wrong because ->state == nfs_ok ensures
> ->nf is valid. But still....
>
That was my thinking, but I agree that it's a bit fragile.
> How about:
>
> fh_put(resp->fh);
> if (!resp->nf)
> return;
> if (nfsd_enable_fadvise_dontneed)
> generic_fadvise(nfsd_file_file(resp->nf), argp->offset, resp->count,
> POSIX_FADV_DONTNEED);
> nfsd_file_put(resp->nf);
>
> ??
> Note that we don't test ->status because that is identical to testing ->nf.
> Ditto for other release functions.
>
That looks good. I'll plan to do that in the next respin.
> Otherwise it makes sense for exploring how to optimise IO.
>
> Reviewed-by: NeilBrown <neil@xxxxxxxxxx>
>
> NeilBrown
Thanks for the reviews!
--
Jeff Layton <jlayton@xxxxxxxxxx>
