losetup -a shows different status depends on its version

[akry]

hva shows an error when launching an instance.

2015-09-29 12:51:51 JobContext thr=JobWorker[0/1] [ERROR]: Caught RuntimeError: Failed to find loop device from: /var/lib/wakame-vdc/instances/i-3rq0dr2o/metadata.img
        /opt/axsh/wakame-vdc/dcmgr/lib/dcmgr/helpers/block_device_helper.rb:77:in `detach_loop'
        /opt/axsh/wakame-vdc/dcmgr/lib/dcmgr/drivers/hypervisor/linux_hypervisor.rb:146:in `setup_metadata_drive'

It happens here:
https://github.com/axsh/wakame-vdc/blob/develop/dcmgr/lib/dcmgr/helpers/block_device_helper.rb#L55

if i =~ %r{^(/dev/loop\d+): \[([0-9a-f]+)\]:(\d+) } && $2.hex == stat.dev && $3.to_i == stat.ino

It assumes $2 as a hex value where stat and losetup -a show different results than expected.

"stat: #<File::Stat dev=0x803, ino=2382902, mode=0100644, nlink=1, uid=0, gid=0, rdev=0x0, size=10485760, blksize=4096, blocks=312, atime=2015-09-29 12:51:51 -0700, mtime=2015-09-29 12:51:51 -0700, ctime=2015-09-29 12:51:51 -0700>"

[root@localhost dcmgr]# losetup -a
/dev/loop0: [2051]:141745283 (/var/lib/wakame-vdc/instances/i-1wllauqo/metadata.img (deleted))

0x803 is 2051 in decimal, but "2051".hex is not 0x803 so it fails.

[akry]

on CentOS 6.4

[root@localhost i-zb241121]# which losetup
/sbin/losetup
[root@localhost i-zb241121]# rpm -qf /sbin/losetup
util-linux-ng-2.17.2-12.18.el6.x86_64

on CentOS 7

[root@localhost dcmgr]# losetup --version
losetup from util-linux 2.23.2