Re: [PATCH] Btrfs: setup blank root and fs_info for mount time

[Li Zefan <lizf@cn.fujitsu.com>]

> Have a look in fs/super.c:generic_shutdown_super(), called by
> fs/super.c:kill_anon_super(), where the super method ->put_super() is
> called, setting the super s_fs_info to NULL, before taking the sb_lock
> and removing it from the list of supers.
> 
> Here's my patch.
> 
> btrfs - fix race between btrfs_get_sb() and umount
> 
> From: Ian Kent <raven@themaw.net>
> 
> When mounting a btrfs file system btrfs_test_super() may attempt to
> use sb->s_fs_info, the btrfs root, of a super block that is going away
> and that has had the btrfs root set to NULL in its ->put_super(). But
> if the super block is going away it cannot be an existing super block
> so we can return false in this case.

I think your analysis is right.

Actually I ran the test script (posted in an earlier email), and it
still crashed without your supplementary patch.

> 
> Signed-off-by: Ian Kent <raven@themaw.net>
> ---
> 
>  fs/btrfs/super.c |    6 ++++++
>  1 files changed, 6 insertions(+), 0 deletions(-)
> 
> 
> diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
> index 6b57da3..960b320 100644
> --- a/fs/btrfs/super.c
> +++ b/fs/btrfs/super.c
> @@ -565,6 +565,12 @@ static int btrfs_test_super(struct super_block *s, void *data)
>  	struct btrfs_fs_devices *test_fs_devices = data->fs_info->fs_devices;
>  	struct btrfs_root *root = btrfs_sb(s);
>  
> +	/*
> +	 * If this super block is going away, return false as it
> +	 * can't match as an existing super block.
> +	 */
> +	if (!atomic_read(&s->s_active))
> +		return 0;
>  	return root->fs_info->fs_devices == test_fs_devices;
>  }
>  
>