Re[2]: mkinitrd problem

[Andreas Steinmetz <ast@domdv.de>]

Reminds me that at least in 2.4.16 you have to expicitly release the loop
device after an unmount:

pcast2:/tmp # dd if=/dev/zero of=tst bs=1024 count=1440
1440+0 records in
1440+0 records out
pcast2:/tmp # mke2fs tst
mke2fs 1.26 (3-Feb-2002)
tst is not a block special device.
Proceed anyway? (y,n) y
Filesystem label=
OS type: Linux
Block size=1024 (log=0)
Fragment size=1024 (log=0)
184 inodes, 1440 blocks
72 blocks (5.00%) reserved for the super user
First data block=1
1 block group
8192 blocks per group, 8192 fragments per group
184 inodes per group

Writing inode tables: done
Writing superblocks and filesystem accounting information: done

This filesystem will be automatically checked every 29 mounts or
180 days, whichever comes first.  Use tune2fs -c or -i to override.
pcast2:/tmp # mount -o loop /tmp/tst /mnt/tmp
pcast2:/tmp # umount /mnt/tmp
pcast2:/tmp # losetup -d /dev/loop0
pcast2:/tmp # losetup -d /dev/loop0
ioctl: LOOP_CLR_FD: No such device or address
pcast2:/tmp #

Note the two "losetup -d" commands above. The first one succeeds so umount
didn't release the loop device.


On Tue, 23 Jul 2002 00:34:28 -0600 (MDT) Thunder from the hill <thunder@ngforever.de> wrote:

> Hi,
> 
> On Tue, 23 Jul 2002, Daniel Lim wrote:
> > # mkinitrd /boot/initrd-2.4.9-34.img 2.4.9-34
> > All of your loopback devices are in use!
> 
> Yes, if all your loopback devices are in use, you'll have to umount
> some. 
> cat /proc/mounts, and there umount some of the filesystems with the loop 
> option.
> 
>                                                         Regards,
>                                                         Thunder