On 2019/5/16 下午9:12, Nikolay Borisov wrote:
> When btrfs' 'filesystem' subcommand is passed path to an image file it
> currently fails since the code expects the image file is going to be
> recognised by libblkid (called from btrfs_scan_devices()). This is not
> the case since libblkid only scan well-known locations under /dev.
>
> Fix this by explicitly calling open_ctree which will correctly open
> the image and add it to the correct btrfs_fs_devices struct. This allows
> subsequent cmd_filesystem_show logic to correctly show requested
> information.
>
> Signed-off-by: Nikolay Borisov <nbori...@suse.com>
> ---
> cmds-filesystem.c | 13 ++++++++++++-
> 1 file changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/cmds-filesystem.c b/cmds-filesystem.c
> index b8beec13f0e5..f55ce9b4ab85 100644
> --- a/cmds-filesystem.c
> +++ b/cmds-filesystem.c
> @@ -771,7 +771,18 @@ static int cmd_filesystem_show(int argc, char **argv)
> goto out;
>
> devs_only:
> - ret = btrfs_scan_devices();
> + if (type == BTRFS_ARG_REG) {
> + /*
> + * We don't close the fs_info because it will free the device,
> + * this is not a long-running process so it's fine
> + */
The comment makes sense, but I'm pretty sure we still prefer to clean it up.
Just something like:
struct btrfs_root *root = NULL;
root = open_ctree();
if (root)
ret = 0;
else
ret = 1;
close_ctree(root);
Despite that, I think the patch looks good to me.
Thanks,
Qu
> + if (open_ctree(search, btrfs_sb_offset(0), 0))
> + ret = 0;
> + else
> + ret = 1;
> + } else {
> + ret = btrfs_scan_devices();
> + }
>
> if (ret) {
> error("blkid device scan returned %d", ret);
>
