Re: [PATCH v7 12/13] ext4: switch to multigrain timestamps

[Xi Ruoyao]

On Wed, 2023-09-20 at 10:41 +0200, Christian Brauner wrote:
> > > f1 was last written to *after* f2 was last written to. If the timestamp of f1
> > > is then lower than the timestamp of f2, timestamps are fundamentally broken.
> > > 
> > > Many things in user-space depend on timestamps, such as build system
> > > centered around 'make', but also 'find ... -newer ...'.
> > > 
> > 
> > 
> > What does breakage with make look like in this situation? The "fuzz"
> > here is going to be on the order of a jiffy. The typical case for make
> > timestamp comparisons is comparing source files vs. a build target. If
> > those are being written nearly simultaneously, then that could be an
> > issue, but is that a typical behavior? It seems like it would be hard to
> > rely on that anyway, esp. given filesystems like NFS that can do lazy
> > writeback.
> > 
> > One of the operating principles with this series is that timestamps can
> > be of varying granularity between different files. Note that Linux
> > already violates this assumption when you're working across filesystems
> > of different types.
> > 
> > As to potential fixes if this is a real problem:
> > 
> > I don't really want to put this behind a mount or mkfs option (a'la
> > relatime, etc.), but that is one possibility.
> > 
> > I wonder if it would be feasible to just advance the coarse-grained
> > current_time whenever we end up updating a ctime with a fine-grained
> > timestamp? It might produce some inode write amplification. Files that
> 
> Less than ideal imho.
> 
> If this risks breaking existing workloads by enabling it unconditionally
> and there isn't a clear way to detect and handle these situations
> without risk of regression then we should move this behind a mount
> option.
> 
> So how about the following:
> 
> From cb14add421967f6e374eb77c36cc4a0526b10d17 Mon Sep 17 00:00:00 2001
> From: Christian Brauner <brauner@xxxxxxxxxx>
> Date: Wed, 20 Sep 2023 10:00:08 +0200
> Subject: [PATCH] vfs: move multi-grain timestamps behind a mount option
> 
> While we initially thought we can do this unconditionally it turns out
> that this might break existing workloads that rely on timestamps in very
> specific ways and we always knew this was a possibility. Move
> multi-grain timestamps behind a vfs mount option.

I agree with this solution.

You can add some metainfo:

Reported-by: Ken Moffat <ken@xxxxxxxxxxxxxxxxxxxx>
Bisected-by: Xi Ruoyao <xry111@xxxxxxxxxxxxxxxxxxxx>
Link: https://lists.linuxfromscratch.org/sympa/arc/lfs-dev/2023-09/msg00036.html

> Signed-off-by: Christian Brauner <brauner@xxxxxxxxxx>
> ---
>  fs/fs_context.c     | 18 ++++++++++++++++++
>  fs/inode.c          |  4 ++--
>  fs/proc_namespace.c |  1 +
>  fs/stat.c           |  2 +-
>  include/linux/fs.h  |  4 +++-
>  5 files changed, 25 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/fs_context.c b/fs/fs_context.c
> index a0ad7a0c4680..dd4dade0bb9e 100644
> --- a/fs/fs_context.c
> +++ b/fs/fs_context.c
> @@ -44,6 +44,7 @@ static const struct constant_table common_set_sb_flag[] = {
>  	{ "mand",	SB_MANDLOCK },
>  	{ "ro",		SB_RDONLY },
>  	{ "sync",	SB_SYNCHRONOUS },
> +	{ "mgtime",	SB_MGTIME },
>  	{ },
>  };
>  
> @@ -52,18 +53,32 @@ static const struct constant_table common_clear_sb_flag[] = {
>  	{ "nolazytime",	SB_LAZYTIME },
>  	{ "nomand",	SB_MANDLOCK },
>  	{ "rw",		SB_RDONLY },
> +	{ "nomgtime",	SB_MGTIME },
>  	{ },
>  };
>  
> +static inline int check_mgtime(unsigned int token, const struct fs_context *fc)
> +{
> +	if (token != SB_MGTIME)
> +		return 0;
> +	if (!(fc->fs_type->fs_flags & FS_MGTIME))
> +		return invalf(fc, "Filesystem doesn't support multi-grain timestamps");
> +	return 0;
> +}
> +
>  /*
>   * Check for a common mount option that manipulates s_flags.
>   */
>  static int vfs_parse_sb_flag(struct fs_context *fc, const char *key)
>  {
>  	unsigned int token;
> +	int ret;
>  
>  	token = lookup_constant(common_set_sb_flag, key, 0);
>  	if (token) {
> +		ret = check_mgtime(token, fc);
> +		if (ret)
> +			return ret;
>  		fc->sb_flags |= token;
>  		fc->sb_flags_mask |= token;
>  		return 0;
> @@ -71,6 +86,9 @@ static int vfs_parse_sb_flag(struct fs_context *fc, const char *key)
>  
>  	token = lookup_constant(common_clear_sb_flag, key, 0);
>  	if (token) {
> +		ret = check_mgtime(token, fc);
> +		if (ret)
> +			return ret;
>  		fc->sb_flags &= ~token;
>  		fc->sb_flags_mask |= token;
>  		return 0;
> diff --git a/fs/inode.c b/fs/inode.c
> index 54237f4242ff..fd1a2390aaa3 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -2141,7 +2141,7 @@ EXPORT_SYMBOL(current_mgtime);
>  
>  static struct timespec64 current_ctime(struct inode *inode)
>  {
> -	if (is_mgtime(inode))
> +	if (IS_MGTIME(inode))
>  		return current_mgtime(inode);
>  	return current_time(inode);
>  }
> @@ -2588,7 +2588,7 @@ struct timespec64 inode_set_ctime_current(struct inode *inode)
>  		now = current_time(inode);
>  
>  		/* Just copy it into place if it's not multigrain */
> -		if (!is_mgtime(inode)) {
> +		if (!IS_MGTIME(inode)) {
>  			inode_set_ctime_to_ts(inode, now);
>  			return now;
>  		}
> diff --git a/fs/proc_namespace.c b/fs/proc_namespace.c
> index 250eb5bf7b52..08f5bf4d2c6c 100644
> --- a/fs/proc_namespace.c
> +++ b/fs/proc_namespace.c
> @@ -49,6 +49,7 @@ static int show_sb_opts(struct seq_file *m, struct super_block *sb)
>  		{ SB_DIRSYNC, ",dirsync" },
>  		{ SB_MANDLOCK, ",mand" },
>  		{ SB_LAZYTIME, ",lazytime" },
> +		{ SB_MGTIME, ",mgtime" },
>  		{ 0, NULL }
>  	};
>  	const struct proc_fs_opts *fs_infop;
> diff --git a/fs/stat.c b/fs/stat.c
> index 6e60389d6a15..2f18dd5de18b 100644
> --- a/fs/stat.c
> +++ b/fs/stat.c
> @@ -90,7 +90,7 @@ void generic_fillattr(struct mnt_idmap *idmap, u32 request_mask,
>  	stat->size = i_size_read(inode);
>  	stat->atime = inode->i_atime;
>  
> -	if (is_mgtime(inode)) {
> +	if (IS_MGTIME(inode)) {
>  		fill_mg_cmtime(stat, request_mask, inode);
>  	} else {
>  		stat->mtime = inode->i_mtime;
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 4aeb3fa11927..03e415fb3a7c 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -1114,6 +1114,7 @@ extern int send_sigurg(struct fown_struct *fown);
>  #define SB_NODEV        BIT(2)	/* Disallow access to device special files */
>  #define SB_NOEXEC       BIT(3)	/* Disallow program execution */
>  #define SB_SYNCHRONOUS  BIT(4)	/* Writes are synced at once */
> +#define SB_MGTIME	BIT(5)	/* Use multi-grain timestamps */
>  #define SB_MANDLOCK     BIT(6)	/* Allow mandatory locks on an FS */
>  #define SB_DIRSYNC      BIT(7)	/* Directory modifications are synchronous */
>  #define SB_NOATIME      BIT(10)	/* Do not update access times. */
> @@ -2105,6 +2106,7 @@ static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags
>  					((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
>  #define IS_MANDLOCK(inode)	__IS_FLG(inode, SB_MANDLOCK)
>  #define IS_NOATIME(inode)	__IS_FLG(inode, SB_RDONLY|SB_NOATIME)
> +#define IS_MGTIME(inode)	__IS_FLG(inode, SB_MGTIME)
>  #define IS_I_VERSION(inode)	__IS_FLG(inode, SB_I_VERSION)
>  
>  #define IS_NOQUOTA(inode)	((inode)->i_flags & S_NOQUOTA)
> @@ -2366,7 +2368,7 @@ struct file_system_type {
>   */
>  static inline bool is_mgtime(const struct inode *inode)
>  {
> -	return inode->i_sb->s_type->fs_flags & FS_MGTIME;
> +	return inode->i_sb->s_flags & SB_MGTIME;
>  }
>  
>  extern struct dentry *mount_bdev(struct file_system_type *fs_type,