Re: [RFC PATCH 5/6] ext4: Search for an inode to update under the RCU lock if we can

[Al Viro]

On Thu, Apr 18, 2019 at 03:06:44PM +0100, David Howells wrote:
> +	inode = find_inode_by_ino_rcu(sb, ino);
> +	if (!inode)
> +		return;
>  
>  	if ((inode->i_ino != ino) ||
>  	    (inode->i_state & (I_FREEING | I_WILL_FREE | I_NEW |
>  			       I_DIRTY_INODE)) ||
>  	    ((inode->i_state & I_DIRTY_TIME) == 0))

Umm...  The situation with memory barriers here needs to be documented,
IMO.  And I'm not sure if the thing is actually correct...