On Mon, Apr 24, 2023 at 11:11:02AM -0400, Jeff Layton wrote:
> The VFS always uses coarse-grained timestamp updates for filling out the
> ctime and mtime after a change. This has the benefit of allowing
> filesystems to optimize away a lot metaupdates, to around once 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. 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 (e.g backup applications).
>
> Switching to always using fine-grained timestamps would improve the
> situation for NFS, but that becomes rather expensive, as the underlying
> filesystem will 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:
>
> Whenever the mtime changes, the ctime must also change since we're
> changing the metadata. When a superblock has a s_time_gran >1, we can
> use the lowest-order bit of the inode->i_ctime as a flag to indicate
> that the value has been queried. Then on the next write, we'll fetch a
> fine-grained timestamp instead of the usual coarse-grained one.
>
> We could enable this for any filesystem that has a s_time_gran >1, but
> for now, this patch adds a new SB_MULTIGRAIN_TS flag to allow filesystems
> to opt-in to this behavior.
>
> It then adds a new current_ctime function that acts like the
> current_time helper, but will conditionally grab fine-grained timestamps
> when the flag is set in the current ctime. Also, there is a new
> generic_fill_multigrain_cmtime for grabbing the c/mtime out of the inode
> and atomically marking the ctime as queried.
>
> Later patches will convert filesystems over to this new scheme.
>
> Signed-off-by: Jeff Layton <[email protected]>
> ---
> fs/inode.c | 57 +++++++++++++++++++++++++++++++++++++++---
> fs/stat.c | 24 ++++++++++++++++++
> include/linux/fs.h | 62 ++++++++++++++++++++++++++++++++--------------
> 3 files changed, 121 insertions(+), 22 deletions(-)
>
> diff --git a/fs/inode.c b/fs/inode.c
> index 4558dc2f1355..4bd11bdb46d4 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -2030,6 +2030,7 @@ EXPORT_SYMBOL(file_remove_privs);
> static int inode_needs_update_time(struct inode *inode, struct timespec64 *now)
> {
> int sync_it = 0;
> + struct timespec64 ctime = inode->i_ctime;
>
> /* First try to exhaust all avenues to not sync */
> if (IS_NOCMTIME(inode))
> @@ -2038,7 +2039,9 @@ static int inode_needs_update_time(struct inode *inode, struct timespec64 *now)
> if (!timespec64_equal(&inode->i_mtime, now))
> sync_it = S_MTIME;
>
> - if (!timespec64_equal(&inode->i_ctime, now))
> + if (is_multigrain_ts(inode))
> + ctime.tv_nsec &= ~I_CTIME_QUERIED;
> + if (!timespec64_equal(&ctime, now))
> sync_it |= S_CTIME;
>
> if (IS_I_VERSION(inode) && inode_iversion_need_inc(inode))
> @@ -2062,6 +2065,50 @@ static int __file_update_time(struct file *file, struct timespec64 *now,
> return ret;
> }
>
> +/**
> + * current_ctime - Return FS time (possibly high-res)
> + * @inode: inode.
> + *
> + * Return the current time truncated to the time granularity supported by
> + * the fs, as suitable for a ctime/mtime change.
> + *
> + * For a multigrain timestamp, if the timestamp is flagged as having been
> + * QUERIED, then get a fine-grained timestamp.
> + */
> +struct timespec64 current_ctime(struct inode *inode)
> +{
> + struct timespec64 now;
> + long nsec = 0;
> + bool multigrain = is_multigrain_ts(inode);
> +
> + if (multigrain) {
> + atomic_long_t *pnsec = (atomic_long_t *)&inode->i_ctime.tv_nsec;
> +
> + nsec = atomic_long_fetch_and(~I_CTIME_QUERIED, pnsec);
> + }
> +
> + if (nsec & I_CTIME_QUERIED) {
> + ktime_get_real_ts64(&now);
> + } else {
> + ktime_get_coarse_real_ts64(&now);
> +
> + if (multigrain) {
> + /*
> + * If we've recently fetched a fine-grained timestamp
> + * then the coarse-grained one may be earlier than the
> + * existing one. Just keep the existing ctime if so.
> + */
> + struct timespec64 ctime = inode->i_ctime;
> +
> + if (timespec64_compare(&ctime, &now) > 0)
> + now = ctime;
> + }
> + }
> +
> + return timestamp_truncate(now, inode);
> +}
> +EXPORT_SYMBOL(current_ctime);
> +
> /**
> * file_update_time - update mtime and ctime time
> * @file: file accessed
> @@ -2080,7 +2127,7 @@ int file_update_time(struct file *file)
> {
> int ret;
> struct inode *inode = file_inode(file);
> - struct timespec64 now = current_time(inode);
> + struct timespec64 now = current_ctime(inode);
>
> ret = inode_needs_update_time(inode, &now);
> if (ret <= 0)
> @@ -2109,7 +2156,7 @@ static int file_modified_flags(struct file *file, int flags)
> {
> int ret;
> struct inode *inode = file_inode(file);
> - struct timespec64 now = current_time(inode);
> + struct timespec64 now = current_ctime(inode);
>
> /*
> * Clear the security bits if the process is not being run by root.
> @@ -2419,9 +2466,11 @@ struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode)
> if (unlikely(t.tv_sec == sb->s_time_max || t.tv_sec == sb->s_time_min))
> t.tv_nsec = 0;
>
> - /* Avoid division in the common cases 1 ns and 1 s. */
> + /* Avoid division in the common cases 1 ns, 2 ns and 1 s. */
> if (gran == 1)
> ; /* nothing */
> + else if (gran == 2)
> + t.tv_nsec &= ~1L;
Is that trying to mask off I_CTIME_QUERIED?
If so, can we please use that constant as raw constants tend to be
confusing in the long run.