From: "Mkrtchyan, Tigran" <tigran.mkrtchyan@desy.de>
To: David Wysochanski <dwysocha@redhat.com>
Cc: trondmy <trondmy@hammerspace.com>,
Anna Schumaker <anna.schumaker@netapp.com>,
linux-nfs <linux-nfs@vger.kernel.org>
Subject: Re: [PATCH 11/11] NFS: Bring back nfs_dir_mapping_need_revalidate() in nfs_readdir()
Date: Mon, 2 Nov 2020 16:38:42 +0100 (CET) [thread overview]
Message-ID: <1976524196.5308081.1604331522179.JavaMail.zimbra@desy.de> (raw)
In-Reply-To: <1604325011-29427-12-git-send-email-dwysocha@redhat.com>
Hi Dave,
----- Original Message -----
> From: "David Wysochanski" <dwysocha@redhat.com>
> To: "trondmy" <trondmy@hammerspace.com>, "Anna Schumaker" <anna.schumaker@netapp.com>
> Cc: "linux-nfs" <linux-nfs@vger.kernel.org>
> Sent: Monday, 2 November, 2020 14:50:11
> Subject: [PATCH 11/11] NFS: Bring back nfs_dir_mapping_need_revalidate() in nfs_readdir()
> commit 79f687a3de9e ("NFS: Fix a performance regression in readdir")
> removed nfs_dir_mapping_need_revalidate() in an attempt to fix a
> performance regression, but had the effect that with NFSv3 the
> pagecache would never expire while a process was reading a directory.
> An earlier patch addressed this problem by allowing the pagecache
> to expire but the current process to continue using the last cookie
> returned by the server when searching the pagecache, rather than
> always starting at 0. Thus, bring back nfs_dir_mapping_need_revalidate()
> so that nfsi->cache_validity & NFS_INO_INVALID_DATA will be seen
> and nfs_revalidate_mapping() will be called with NFSv3 as well,
> making it behave the same as NFSv4.
>
> Signed-off-by: Dave Wysochanski <dwysocha@redhat.com>
> ---
> fs/nfs/dir.c | 13 ++++++++++++-
> 1 file changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c
> index cbd74cbdbb9f..aeb086fb3d0a 100644
> --- a/fs/nfs/dir.c
> +++ b/fs/nfs/dir.c
> @@ -872,6 +872,17 @@ int uncached_readdir(nfs_readdir_descriptor_t *desc)
> return status;
> }
>
> +static bool nfs_dir_mapping_need_revalidate(struct inode *dir)
> +{
> + struct nfs_inode *nfsi = NFS_I(dir);
> +
> + if (nfs_attribute_cache_expired(dir))
> + return true;
> + if (nfsi->cache_validity & NFS_INO_INVALID_DATA)
> + return true;
> + return false;
> +}
Is this the same as:
static bool nfs_dir_mapping_need_revalidate(struct inode *dir)
{
struct nfs_inode *nfsi = NFS_I(dir);
return nfs_attribute_cache_expired(dir) || nfsi->cache_validity & NFS_INO_INVALID_DATA);
}
Tigran.
> +
> /* The file offset position represents the dirent entry number. A
> last cookie cache takes care of the common case of reading the
> whole directory.
> @@ -903,7 +914,7 @@ static int nfs_readdir(struct file *file, struct dir_context
> *ctx)
> * to either find the entry with the appropriate number or
> * revalidate the cookie.
> */
> - if (ctx->pos == 0 || nfs_attribute_cache_expired(inode))
> + if (ctx->pos == 0 || nfs_dir_mapping_need_revalidate(inode))
> res = nfs_revalidate_mapping(inode, file->f_mapping);
> if (res < 0)
> goto out;
> --
> 1.8.3.1
next prev parent reply other threads:[~2020-11-02 15:48 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-11-02 13:50 [PATCH 00/11] Add NFS readdir tracepoints and improve performance of reading directories Dave Wysochanski
2020-11-02 13:50 ` [PATCH 01/11] NFSv4: Improve nfs4_readdir tracepoint by adding additional fields Dave Wysochanski
2020-11-02 13:50 ` [PATCH 02/11] NFS: Replace dfprintk statements with trace events in nfs_readdir Dave Wysochanski
2020-11-02 13:50 ` [PATCH 03/11] NFS: Move nfs_readdir_descriptor_t into internal header Dave Wysochanski
2020-11-02 13:50 ` [PATCH 04/11] NFS: Add tracepoints for functions involving nfs_readdir_descriptor_t Dave Wysochanski
2020-11-02 17:32 ` kernel test robot
2020-11-02 13:50 ` [PATCH 05/11] NFS: Add tracepoints for opendir, closedir, fsync_dir and llseek_dir Dave Wysochanski
2020-11-02 13:50 ` [PATCH 06/11] NFS: Add tracepoints for nfs_readdir_xdr_filler enter and exit Dave Wysochanski
2020-11-02 13:50 ` [PATCH 07/11] NFS: Add tracepoint to entry and exit of nfs_do_filldir Dave Wysochanski
2020-11-02 13:50 ` [PATCH 08/11] NFS: Replace LOOKUPCACHE dfprintk statements with tracepoints Dave Wysochanski
2020-11-02 13:50 ` [PATCH 09/11] NFS: Improve performance of listing directories being modified Dave Wysochanski
2020-11-02 16:21 ` Trond Myklebust
2020-11-02 16:26 ` David Wysochanski
2020-11-02 17:31 ` Trond Myklebust
2020-11-02 19:45 ` David Wysochanski
2020-11-02 21:30 ` Trond Myklebust
2020-11-02 22:05 ` David Wysochanski
2020-11-03 3:38 ` Trond Myklebust
2020-11-03 13:29 ` David Wysochanski
2020-11-03 0:09 ` Frank van der Linden
2020-11-03 17:49 ` David Wysochanski
2020-11-02 13:50 ` [PATCH 10/11] NFS: Add page_index to nfs_readdir enter and exit tracepoints Dave Wysochanski
2020-11-02 13:50 ` [PATCH 11/11] NFS: Bring back nfs_dir_mapping_need_revalidate() in nfs_readdir() Dave Wysochanski
2020-11-02 15:38 ` Mkrtchyan, Tigran [this message]
2020-11-02 16:16 ` David Wysochanski
2020-11-02 14:27 ` [PATCH 00/11] Add NFS readdir tracepoints and improve performance of reading directories Chuck Lever
2020-11-02 15:07 ` David Wysochanski
2020-11-02 15:13 ` Chuck Lever
2020-11-02 15:58 ` Mkrtchyan, Tigran
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1976524196.5308081.1604331522179.JavaMail.zimbra@desy.de \
--to=tigran.mkrtchyan@desy.de \
--cc=anna.schumaker@netapp.com \
--cc=dwysocha@redhat.com \
--cc=linux-nfs@vger.kernel.org \
--cc=trondmy@hammerspace.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox