[CRIU] [PATCH] mount: restore cwd after creating a roots yard

[Pavel Emelyanov]

On 12/03/2015 10:58 PM, Andrey Vagin wrote:
> From: Andrew Vagin <avagin at virtuozzo.com>
> 
> Currently we see that a cgroup yard are not umounted
> with the ENOENT error, because cwd was changed.
> 
> Signed-off-by: Andrew Vagin <avagin at virtuozzo.com>

> @@ -2756,11 +2772,21 @@ int depopulate_roots_yard(void)
>  
>  void cleanup_mnt_ns(void)
>  {
> +	char path[PATH_MAX], *root = opts.root ? : "/";
> +	int fd;
> +
>  	if (mnt_roots == NULL)
>  		return;
>  
> -	if (rmdir(mnt_roots))
> +	fd = open(root, O_DIRECTORY | O_RDONLY);
> +	if (fd < 0) {
> +		pr_perror("Unable to open %s", opts.root);
> +		return;
> +	}
> +	snprintf(path, sizeof(path), "/proc/self/fd/%d/%s", fd, mnt_roots);
> +	if (rmdir(path))
>  		pr_perror("Can't remove the directory %s", mnt_roots);
> +	close(fd);

This hunk doesn't fit the patch description. What does it do?

>  }
>  
>  int prepare_mnt_ns(void)
>