Re: [PATCH 1/3] xfs: Do not unshare ranges beyond EOF

[Christoph Hellwig <hch@infradead.org>]

On Fri, Sep 20, 2024 at 08:26:21PM +0800, Julian Sun wrote:
> Attempting to unshare extents beyond EOF will trigger
> the need zeroing case, which in turn triggers a warning.
> Therefore, let's skip the unshare process if extents are
> beyond EOF.
> 
> Reported-and-tested-by: syzbot+296b1c84b9cbf306e5a0@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=296b1c84b9cbf306e5a0
> Fixes: 32a38a499104 ("iomap: use write_begin to read pages to unshare")
> Inspired-by: Dave Chinner <david@fromorbit.com>
> Signed-off-by: Julian Sun <sunjunchao2870@gmail.com>
> ---
>  fs/xfs/xfs_reflink.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c
> index 6fde6ec8092f..65509ff6aba0 100644
> --- a/fs/xfs/xfs_reflink.c
> +++ b/fs/xfs/xfs_reflink.c
> @@ -3,6 +3,7 @@
>   * Copyright (C) 2016 Oracle.  All Rights Reserved.
>   * Author: Darrick J. Wong <darrick.wong@oracle.com>
>   */
> +#include "linux/fs.h"

This really should not be needed (and is the wrong way to include
non-local headers anyway).

>  #include "xfs.h"
>  #include "xfs_fs.h"
>  #include "xfs_shared.h"
> @@ -1669,6 +1670,9 @@ xfs_reflink_unshare(
>  
>  	if (!xfs_is_reflink_inode(ip))
>  		return 0;
> +	/* don't try to unshare any ranges beyond EOF. */
> +	if (offset + len > i_size_read(inode))
> +		len = i_size_read(inode) - offset;

So i_size is a byte granularity value, but later on iomap_file_unshare
operates on blocks.  If you reduce the value like this here this means
we can't ever unshare the last block of a file if the file size is
not block aligned, which feels odd.