Re: [PATCH] f2fs: add support for an i_version counter

[Chao Yu]

On 2023/11/20 17:54, Yangtao Li wrote:
> NFSv4 mandates a change attribute to avoid problems with timestamp
> granularity, which Linux implements using the i_version counter. This is
> particularly important when the underlying filesystem is fast.
>
> Signed-off-by: Yangtao Li <frank.li@xxxxxxxx>
> ---
>   fs/f2fs/f2fs.h  | 1 +
>   fs/f2fs/inode.c | 1 +
>   fs/f2fs/super.c | 1 +
>   3 files changed, 3 insertions(+)
>
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index 9043cedfa12b..68fd2ef35104 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -28,6 +28,7 @@
>   
>   #include <linux/fscrypt.h>
>   #include <linux/fsverity.h>
> +#include <linux/iversion.h>
>   
>   struct pagevec;
>   
> diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
> index 560bfcad1af2..2604fa4a0704 100644
> --- a/fs/f2fs/inode.c
> +++ b/fs/f2fs/inode.c
> @@ -32,6 +32,7 @@ void f2fs_mark_inode_dirty_sync(struct inode *inode, bool sync)
>   	if (f2fs_inode_dirtied(inode, sync))
>   		return;
>   
> +	inode_inc_iversion(inode);

For the case f2fs is not used by nfs as backend storage, there will be overhead for
iversion update logic.

So what about introducing a new mount option to enable the iversion functionality,
and disabling it by default?

Thanks,

>   	mark_inode_dirty_sync(inode);
>   }
>   
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 033af907c3b1..2bddccd47e3a 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -2204,6 +2204,7 @@ static void default_options(struct f2fs_sb_info *sbi, bool remount)
>   	set_opt(sbi, MERGE_CHECKPOINT);
>   	F2FS_OPTION(sbi).unusable_cap = 0;
>   	sbi->sb->s_flags |= SB_LAZYTIME;
> +	sbi->sb->s_flags |= SB_I_VERSION;
>   	if (!f2fs_is_readonly(sbi))
>   		set_opt(sbi, FLUSH_MERGE);
>   	if (f2fs_sb_has_blkzoned(sbi))