Re: [PATCH v2 5/7] fs: add an ATTR_CTIME_DLG flag

[Christian Brauner <brauner@kernel.org>]

On Mon, Aug 26, 2024 at 08:46:15AM GMT, Jeff Layton wrote:
> When updating the ctime on an inode for a setattr with a multigrain
> filesystem, we usually want to take the latest time we can get for the
> ctime. The exception to this rule is when there is a nfsd write
> delegation and the server is proxying timestamps from the client.
> 
> When nfsd gets a CB_GETATTR response, we want to update the timestamp
> value in the inode to the values that the client is tracking. The client
> doesn't send a ctime value (since that's always determined by the
> exported filesystem), but it does send a mtime value. In the case where
> it does, then we may also need to update the ctime to a value
> commensurate with that.
> 
> Add a ATTR_CTIME_DELEG flag, which tells the underlying setattr

Fwiw: disconnect between commit message and actually used ATTR_CTIME_DLG.

> machinery to respect that value and not to set it to the current time.
> 
> Signed-off-by: Jeff Layton <jlayton@kernel.org>
> ---

Are you set on sending us on a mission to free up ATTR_* bits after
freeing up FMODE_* bits? ;)

If there's going to be more ATTR_*DELEG* flags that modify the
behavior when delegation is in effect then we could consider adding
another unsigned int ia_deleg field to struct iattr so that you can check:

if (ia_valid & ATTR_CTIME) {
	if (unlikely(iattr->ia_deleg & ATTR_CTIME))
		// do some special stuff
	else
		// do the regular stuff
}

or some such variant.

>  fs/attr.c          | 10 +++++++++-
>  include/linux/fs.h |  1 +
>  2 files changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/attr.c b/fs/attr.c
> index 7144b207e715..0eb7b228b94d 100644
> --- a/fs/attr.c
> +++ b/fs/attr.c
> @@ -295,7 +295,15 @@ static void setattr_copy_mgtime(struct inode *inode, const struct iattr *attr)
>  		return;
>  	}
>  
> -	now = inode_set_ctime_current(inode);
> +	/*
> +	 * In the case of an update for a write delegation, we must respect
> +	 * the value in ia_ctime and not use the current time.
> +	 */
> +	if (ia_valid & ATTR_CTIME_DLG)
> +		inode_set_ctime_to_ts(inode, attr->ia_ctime);
> +	else
> +		now = inode_set_ctime_current(inode);
> +
>  	if (ia_valid & ATTR_ATIME_SET)
>  		inode_set_atime_to_ts(inode, attr->ia_atime);
>  	else if (ia_valid & ATTR_ATIME)
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 7c1da3c687bd..43a802b2cb0d 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -211,6 +211,7 @@ typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
>  #define ATTR_TIMES_SET	(1 << 16)
>  #define ATTR_TOUCH	(1 << 17)
>  #define ATTR_DELEG	(1 << 18) /* Delegated attrs (don't break) */
> +#define ATTR_CTIME_DLG	(1 << 19) /* Delegation in effect */

What's the interaction between ATTR_DELEG and ATTR_CTIME_DLG? I think
that's potentially confusing.