Message-ID: <ccc52562305bd1a1affb14e94a1cc08433eb8316.camel@kernel.org>
Date:   Wed, 02 Aug 2023 16:54:09 -0400
From:   Jeff Layton <jlayton@...nel.org>
To:     Jan Kara <jack@...e.cz>
Cc:     Eric Van Hensbergen <ericvh@...nel.org>,
        Latchesar Ionkov <lucho@...kov.net>,
        Dominique Martinet <asmadeus@...ewreck.org>,
        Christian Schoenebeck <linux_oss@...debyte.com>,
        David Howells <dhowells@...hat.com>,
        Marc Dionne <marc.dionne@...istor.com>,
        Chris Mason <clm@...com>, Josef Bacik <josef@...icpanda.com>,
        David Sterba <dsterba@...e.com>, Xiubo Li <xiubli@...hat.com>,
        Ilya Dryomov <idryomov@...il.com>,
        Jan Harkes <jaharkes@...cmu.edu>, coda@...cmu.edu,
        Tyler Hicks <code@...icks.com>, Gao Xiang <xiang@...nel.org>,
        Chao Yu <chao@...nel.org>, Yue Hu <huyue2@...lpad.com>,
        Jeffle Xu <jefflexu@...ux.alibaba.com>,
        Namjae Jeon <linkinjeon@...nel.org>,
        Sungjong Seo <sj1557.seo@...sung.com>,
        Jan Kara <jack@...e.com>, Theodore Ts'o <tytso@....edu>,
        Andreas Dilger <adilger.kernel@...ger.ca>,
        Jaegeuk Kim <jaegeuk@...nel.org>,
        OGAWA Hirofumi <hirofumi@...l.parknet.co.jp>,
        Miklos Szeredi <miklos@...redi.hu>,
        Bob Peterson <rpeterso@...hat.com>,
        Andreas Gruenbacher <agruenba@...hat.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Tejun Heo <tj@...nel.org>,
        Alexander Viro <viro@...iv.linux.org.uk>,
        Christian Brauner <brauner@...nel.org>,
        Trond Myklebust <trond.myklebust@...merspace.com>,
        Anna Schumaker <anna@...nel.org>,
        Konstantin Komarov <almaz.alexandrovich@...agon-software.com>,
        Mark Fasheh <mark@...heh.com>,
        Joel Becker <jlbec@...lplan.org>,
        Joseph Qi <joseph.qi@...ux.alibaba.com>,
        Mike Marshall <hubcap@...ibond.com>,
        Martin Brandenburg <martin@...ibond.com>,
        Luis Chamberlain <mcgrof@...nel.org>,
        Kees Cook <keescook@...omium.org>,
        Iurii Zaikin <yzaikin@...gle.com>,
        Steve French <sfrench@...ba.org>,
        Paulo Alcantara <pc@...guebit.com>,
        Ronnie Sahlberg <lsahlber@...hat.com>,
        Shyam Prasad N <sprasad@...rosoft.com>,
        Tom Talpey <tom@...pey.com>,
        Sergey Senozhatsky <senozhatsky@...omium.org>,
        Richard Weinberger <richard@....at>,
        Hans de Goede <hdegoede@...hat.com>,
        Hugh Dickins <hughd@...gle.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        "Darrick J. Wong" <djwong@...nel.org>,
        Dave Chinner <david@...morbit.com>,
        Anthony Iliopoulos <ailiop@...e.com>, v9fs@...ts.linux.dev,
        linux-kernel@...r.kernel.org, linux-afs@...ts.infradead.org,
        linux-btrfs@...r.kernel.org, ceph-devel@...r.kernel.org,
        codalist@...a.cs.cmu.edu, ecryptfs@...r.kernel.org,
        linux-erofs@...ts.ozlabs.org, linux-fsdevel@...r.kernel.org,
        linux-ext4@...r.kernel.org, linux-f2fs-devel@...ts.sourceforge.net,
        cluster-devel@...hat.com, linux-nfs@...r.kernel.org,
        ntfs3@...ts.linux.dev, ocfs2-devel@...ts.linux.dev,
        devel@...ts.orangefs.org, linux-cifs@...r.kernel.org,
        samba-technical@...ts.samba.org, linux-mtd@...ts.infradead.org,
        linux-mm@...ck.org, linux-xfs@...r.kernel.org
Subject: Re: [PATCH v6 2/7] fs: add infrastructure for multigrain timestamps

On Wed, 2023-08-02 at 21:35 +0200, Jan Kara wrote:
> On Tue 25-07-23 10:58:15, Jeff Layton wrote:
> > The VFS always uses coarse-grained timestamps when updating the ctime
> > and mtime after a change. This has the benefit of allowing filesystems
> > to optimize away a lot metadata updates, down to around 1 per jiffy,
> > even when a file is under heavy writes.
> > 
> > Unfortunately, this has always been an issue when we're exporting via
> > NFSv3, which relies on timestamps to validate caches. A lot of changes
> > can happen in a jiffy, so timestamps aren't sufficient to help the
> > client decide to invalidate the cache. Even with NFSv4, a lot of
> > exported filesystems don't properly support a change attribute and are
> > subject to the same problems with timestamp granularity. Other
> > applications have similar issues with timestamps (e.g backup
> > applications).
> > 
> > If we were to always use fine-grained timestamps, that would improve the
> > situation, but that becomes rather expensive, as the underlying
> > filesystem would have to log a lot more metadata updates.
> > 
> > What we need is a way to only use fine-grained timestamps when they are
> > being actively queried.
> > 
> > POSIX generally mandates that when the the mtime changes, the ctime must
> > also change. The kernel always stores normalized ctime values, so only
> > the first 30 bits of the tv_nsec field are ever used.
> > 
> > Use the 31st bit of the ctime tv_nsec field to indicate that something
> > has queried the inode for the mtime or ctime. When this flag is set,
> > on the next mtime or ctime update, the kernel will fetch a fine-grained
> > timestamp instead of the usual coarse-grained one.
> > 
> > Filesytems can opt into this behavior by setting the FS_MGTIME flag in
> > the fstype. Filesystems that don't set this flag will continue to use
> > coarse-grained timestamps.
> > 
> > Later patches will convert individual filesystems to use the new
> > infrastructure.
> > 
> > Signed-off-by: Jeff Layton <jlayton@...nel.org>
> > ---
> >  fs/inode.c         | 98 ++++++++++++++++++++++++++++++++++++++----------------
> >  fs/stat.c          | 41 +++++++++++++++++++++--
> >  include/linux/fs.h | 45 +++++++++++++++++++++++--
> >  3 files changed, 151 insertions(+), 33 deletions(-)
> > 
> > diff --git a/fs/inode.c b/fs/inode.c
> > index d4ab92233062..369621e7faf5 100644
> > --- a/fs/inode.c
> > +++ b/fs/inode.c
> > @@ -1919,6 +1919,21 @@ int inode_update_time(struct inode *inode, struct timespec64 *time, int flags)
> >  }
> >  EXPORT_SYMBOL(inode_update_time);
> >  
> > +/**
> > + * current_coarse_time - Return FS time
> > + * @inode: inode.
> > + *
> > + * Return the current coarse-grained time truncated to the time
> > + * granularity supported by the fs.
> > + */
> > +static struct timespec64 current_coarse_time(struct inode *inode)
> > +{
> > +	struct timespec64 now;
> > +
> > +	ktime_get_coarse_real_ts64(&now);
> > +	return timestamp_truncate(now, inode);
> > +}
> > +
> >  /**
> >   *	atime_needs_update	-	update the access time
> >   *	@path: the &struct path to update
> > @@ -1952,7 +1967,7 @@ bool atime_needs_update(const struct path *path, struct inode *inode)
> >  	if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
> >  		return false;
> >  
> > -	now = current_time(inode);
> > +	now = current_coarse_time(inode);
> >  
> >  	if (!relatime_need_update(mnt, inode, now))
> >  		return false;
> > @@ -1986,7 +2001,7 @@ void touch_atime(const struct path *path)
> >  	 * We may also fail on filesystems that have the ability to make parts
> >  	 * of the fs read only, e.g. subvolumes in Btrfs.
> >  	 */
> > -	now = current_time(inode);
> > +	now = current_coarse_time(inode);
> >  	inode_update_time(inode, &now, S_ATIME);
> >  	__mnt_drop_write(mnt);
> >  skip_update:
> 
> There are also calls in fs/smb/client/file.c:cifs_readpage_worker() and in
> fs/ocfs2/file.c:ocfs2_update_inode_atime() that should probably use
> current_coarse_time() to avoid needless querying of fine grained
> timestamps. But see below...
> 

Technically, they already devolve to current_coarse_time anyway, but
changing them would allow them to skip the fstype flag check, but I like
your idea below better anyway.

> > @@ -2072,6 +2087,56 @@ int file_remove_privs(struct file *file)
> >  }
> >  EXPORT_SYMBOL(file_remove_privs);
> >  
> > +/**
> > + * current_mgtime - Return FS time (possibly fine-grained)
> > + * @inode: inode.
> > + *
> > + * Return the current time truncated to the time granularity supported by
> > + * the fs, as suitable for a ctime/mtime change. If the ctime is flagged
> > + * as having been QUERIED, get a fine-grained timestamp.
> > + */
> > +static struct timespec64 current_mgtime(struct inode *inode)
> > +{
> > +	struct timespec64 now;
> > +	atomic_long_t *pnsec = (atomic_long_t *)&inode->__i_ctime.tv_nsec;
> > +	long nsec = atomic_long_read(pnsec);
> > +
> > +	if (nsec & I_CTIME_QUERIED) {
> > +		ktime_get_real_ts64(&now);
> > +	} else {
> > +		struct timespec64 ctime;
> > +
> > +		ktime_get_coarse_real_ts64(&now);
> > +
> > +		/*
> > +		 * If we've recently fetched a fine-grained timestamp
> > +		 * then the coarse-grained one may still be earlier than the
> > +		 * existing one. Just keep the existing ctime if so.
> > +		 */
> > +		ctime = inode_get_ctime(inode);
> > +		if (timespec64_compare(&ctime, &now) > 0)
> > +			now = ctime;
> > +	}
> > +
> > +	return timestamp_truncate(now, inode);
> > +}
> > +
> > +/**
> > + * current_time - Return timestamp suitable for ctime update
> > + * @inode: inode to eventually be updated
> > + *
> > + * Return the current time, which is usually coarse-grained but may be fine
> > + * grained if the filesystem uses multigrain timestamps and the existing
> > + * ctime was queried since the last update.
> > + */
> > +struct timespec64 current_time(struct inode *inode)
> > +{
> > +	if (is_mgtime(inode))
> > +		return current_mgtime(inode);
> > +	return current_coarse_time(inode);
> > +}
> > +EXPORT_SYMBOL(current_time);
> > +
> 
> So if you modify current_time() to handle multigrain timestamps the code
> will be still racy. In particular fill_mg_cmtime() can race with
> inode_set_ctime_current() like:
> 
> fill_mg_cmtime()				inode_set_ctime_current()
>   stat->mtime = inode->i_mtime;
>   stat->ctime.tv_sec = inode->__i_ctime.tv_sec;
> 						  now = current_time();
> 							/* fetches coarse
> 							 * grained timestamp */
>   stat->ctime.tv_nsec = atomic_long_fetch_or(I_CTIME_QUERIED, pnsec) &
> 				~I_CTIME_QUERIED;
> 						  inode_set_ctime(inode, now.tv_sec, now.tv_nsec);
> 
> and the information about a need for finegrained timestamp update gets
> lost. So what I'd propose is to leave current_time() alone (just always
> reporting coarse grained timestamps) and put all the magic into
> inode_set_ctime_current() only. There we need something like:
> 
> struct timespec64 inode_set_ctime_current(struct inode *inode)
> {
> 	... variables ...
> 
> 	nsec = READ_ONCE(inode->__i_ctime.tv_nsec);
>  	if (!(nsec & I_CTIME_QUERIED)) {
> 		now = current_time(inode);
> 
> 		if (!is_gmtime(inode)) {
> 			inode_set_ctime_to_ts(inode, now);
> 		} else {
> 			/*
> 			 * If we've recently fetched a fine-grained
> 			 * timestamp then the coarse-grained one may still
> 			 * be earlier than the existing one. Just keep the
> 			 * existing ctime if so.
> 			 */
> 			ctime = inode_get_ctime(inode);
> 			if (timespec64_compare(&ctime, &now) > 0)
> 				now = ctime;
> 
> 			/*
> 			 * Ctime updates are generally protected by inode
> 			 * lock but we could have raced with setting of
> 			 * I_CTIME_QUERIED flag.
> 			 */
> 			if (cmpxchg(&inode->__i_ctime.tv_nsec, nsec,
> 				    now.tv_nsec) != nsec)
> 				goto fine_grained;
> 			inode->__i_ctime.tv_sec = now.tv_sec;
> 		}
> 		return now;
> 	}
> fine_grained:
> 	ktime_get_real_ts64(&now);
> 	inode_set_ctime_to_ts(inode, now);
> 
> 	return now;
> }
> 
> 								Honza
> 

This is a great idea. I'll rework the series along the lines you
suggest. That also answers my earlier question to Christian:

I'll just resend the whole series (it's not very big anyway), and I'll
include the fill_mg_cmtime prototype change.

Cheers,

> >  static int inode_needs_update_time(struct inode *inode, struct timespec64 *now)
> >  {
> >  	int sync_it = 0;
> > @@ -2480,37 +2545,12 @@ struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode)
> >  }
> >  EXPORT_SYMBOL(timestamp_truncate);
> >  
> > -/**
> > - * current_time - Return FS time
> > - * @inode: inode.
> > - *
> > - * Return the current time truncated to the time granularity supported by
> > - * the fs.
> > - *
> > - * Note that inode and inode->sb cannot be NULL.
> > - * Otherwise, the function warns and returns time without truncation.
> > - */
> > -struct timespec64 current_time(struct inode *inode)
> > -{
> > -	struct timespec64 now;
> > -
> > -	ktime_get_coarse_real_ts64(&now);
> > -
> > -	if (unlikely(!inode->i_sb)) {
> > -		WARN(1, "current_time() called with uninitialized super_block in the inode");
> > -		return now;
> > -	}
> > -
> > -	return timestamp_truncate(now, inode);
> > -}
> > -EXPORT_SYMBOL(current_time);
> > -
> >  /**
> >   * inode_set_ctime_current - set the ctime to current_time
> >   * @inode: inode
> >   *
> > - * Set the inode->i_ctime to the current value for the inode. Returns
> > - * the current value that was assigned to i_ctime.
> > + * Set the inode->__i_ctime to the current value for the inode. Returns
> > + * the current value that was assigned to __i_ctime.
> >   */
> >  struct timespec64 inode_set_ctime_current(struct inode *inode)
> >  {
> > diff --git a/fs/stat.c b/fs/stat.c
> > index 062f311b5386..51effd1c2bc2 100644
> > --- a/fs/stat.c
> > +++ b/fs/stat.c
> > @@ -26,6 +26,37 @@
> >  #include "internal.h"
> >  #include "mount.h"
> >  
> > +/**
> > + * fill_mg_cmtime - Fill in the mtime and ctime and flag ctime as QUERIED
> > + * @request_mask: STATX_* values requested
> > + * @inode: inode from which to grab the c/mtime
> > + * @stat: where to store the resulting values
> > + *
> > + * Given @inode, grab the ctime and mtime out if it and store the result
> > + * in @stat. When fetching the value, flag it as queried so the next write
> > + * will use a fine-grained timestamp.
> > + */
> > +void fill_mg_cmtime(u32 request_mask, struct inode *inode, struct kstat *stat)
> > +{
> > +	atomic_long_t *pnsec = (atomic_long_t *)&inode->__i_ctime.tv_nsec;
> > +
> > +	/* If neither time was requested, then don't report them */
> > +	if (!(request_mask & (STATX_CTIME|STATX_MTIME))) {
> > +		stat->result_mask &= ~(STATX_CTIME|STATX_MTIME);
> > +		return;
> > +	}
> > +
> > +	stat->mtime = inode->i_mtime;
> > +	stat->ctime.tv_sec = inode->__i_ctime.tv_sec;
> > +	/*
> > +	 * Atomically set the QUERIED flag and fetch the new value with
> > +	 * the flag masked off.
> > +	 */
> > +	stat->ctime.tv_nsec = atomic_long_fetch_or(I_CTIME_QUERIED, pnsec) &
> > +					~I_CTIME_QUERIED;
> > +}
> > +EXPORT_SYMBOL(fill_mg_cmtime);
> > +
> >  /**
> >   * generic_fillattr - Fill in the basic attributes from the inode struct
> >   * @idmap:	idmap of the mount the inode was found from
> > @@ -58,8 +89,14 @@ void generic_fillattr(struct mnt_idmap *idmap, u32 request_mask,
> >  	stat->rdev = inode->i_rdev;
> >  	stat->size = i_size_read(inode);
> >  	stat->atime = inode->i_atime;
> > -	stat->mtime = inode->i_mtime;
> > -	stat->ctime = inode_get_ctime(inode);
> > +
> > +	if (is_mgtime(inode)) {
> > +		fill_mg_cmtime(request_mask, inode, stat);
> > +	} else {
> > +		stat->mtime = inode->i_mtime;
> > +		stat->ctime = inode_get_ctime(inode);
> > +	}
> > +
> >  	stat->blksize = i_blocksize(inode);
> >  	stat->blocks = inode->i_blocks;
> >  
> > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > index 42d1434cc427..a0bdbefbf293 100644
> > --- a/include/linux/fs.h
> > +++ b/include/linux/fs.h
> > @@ -1477,15 +1477,43 @@ static inline bool fsuidgid_has_mapping(struct super_block *sb,
> >  struct timespec64 current_time(struct inode *inode);
> >  struct timespec64 inode_set_ctime_current(struct inode *inode);
> >  
> > +/*
> > + * Multigrain timestamps
> > + *
> > + * Conditionally use fine-grained ctime and mtime timestamps when there
> > + * are users actively observing them via getattr. The primary use-case
> > + * for this is NFS clients that use the ctime to distinguish between
> > + * different states of the file, and that are often fooled by multiple
> > + * operations that occur in the same coarse-grained timer tick.
> > + *
> > + * The kernel always keeps normalized struct timespec64 values in the ctime,
> > + * which means that only the first 30 bits of the value are used. Use the
> > + * 31st bit of the ctime's tv_nsec field as a flag to indicate that the value
> > + * has been queried since it was last updated.
> > + */
> > +#define I_CTIME_QUERIED		(1L<<30)
> > +
> >  /**
> >   * inode_get_ctime - fetch the current ctime from the inode
> >   * @inode: inode from which to fetch ctime
> >   *
> > - * Grab the current ctime from the inode and return it.
> > + * Grab the current ctime tv_nsec field from the inode, mask off the
> > + * I_CTIME_QUERIED flag and return it. This is mostly intended for use by
> > + * internal consumers of the ctime that aren't concerned with ensuring a
> > + * fine-grained update on the next change (e.g. when preparing to store
> > + * the value in the backing store for later retrieval).
> > + *
> > + * This is safe to call regardless of whether the underlying filesystem
> > + * is using multigrain timestamps.
> >   */
> >  static inline struct timespec64 inode_get_ctime(const struct inode *inode)
> >  {
> > -	return inode->__i_ctime;
> > +	struct timespec64 ctime;
> > +
> > +	ctime.tv_sec = inode->__i_ctime.tv_sec;
> > +	ctime.tv_nsec = inode->__i_ctime.tv_nsec & ~I_CTIME_QUERIED;
> > +
> > +	return ctime;
> >  }
> >  
> >  /**
> > @@ -2261,6 +2289,7 @@ struct file_system_type {
> >  #define FS_USERNS_MOUNT		8	/* Can be mounted by userns root */
> >  #define FS_DISALLOW_NOTIFY_PERM	16	/* Disable fanotify permission events */
> >  #define FS_ALLOW_IDMAP         32      /* FS has been updated to handle vfs idmappings. */
> > +#define FS_MGTIME		64	/* FS uses multigrain timestamps */
> >  #define FS_RENAME_DOES_D_MOVE	32768	/* FS will handle d_move() during rename() internally. */
> >  	int (*init_fs_context)(struct fs_context *);
> >  	const struct fs_parameter_spec *parameters;
> > @@ -2284,6 +2313,17 @@ struct file_system_type {
> >  
> >  #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
> >  
> > +/**
> > + * is_mgtime: is this inode using multigrain timestamps
> > + * @inode: inode to test for multigrain timestamps
> > + *
> > + * Return true if the inode uses multigrain timestamps, false otherwise.
> > + */
> > +static inline bool is_mgtime(const struct inode *inode)
> > +{
> > +	return inode->i_sb->s_type->fs_flags & FS_MGTIME;
> > +}
> > +
> >  extern struct dentry *mount_bdev(struct file_system_type *fs_type,
> >  	int flags, const char *dev_name, void *data,
> >  	int (*fill_super)(struct super_block *, void *, int));
> > @@ -2919,6 +2959,7 @@ extern void page_put_link(void *);
> >  extern int page_symlink(struct inode *inode, const char *symname, int len);
> >  extern const struct inode_operations page_symlink_inode_operations;
> >  extern void kfree_link(void *);
> > +void fill_mg_cmtime(u32 request_mask, struct inode *inode, struct kstat *stat);
> >  void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *);
> >  void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
> >  extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
> > 
> > -- 
> > 2.41.0
> > 

-- 
Jeff Layton <jlayton@...nel.org>

Powered by blists - more mailing lists