Well spotted that this is the issue! On Saturday 21 July 2007 15:30, Nemui Ailin wrote: > by the way, the problem with this encryption is that d-i loads the > wrong loop module
Well, at the time iso-scan mounts the CD image, the loop-aes module is not
yet available, so "wrong" is a relative term here.
> I guess such a bug might not be noticed when the installation is not
> done from an usb media with an iso image that needs to be mounted with
> loop
Correct. For other installation methods the loop module will not already
be loaded when partman-crypto is started, so the USB-stick installation
method is the only one that has this issue.
I'll add an erratum for Etch for this issue.
> it can easily be remedied by unloading the wrong loop module and
> manually insmod loop-AES's loop.ko
The actual name of the module is loop-aes.ko. I've just verified that if
the regular loop.ko is already loaded, 'modprobe loop-aes' seems to
succeed, but does not do anything.
# modprobe loop
# lsmod | grep loop
loop 15240 0
# modprobe loop-aes
# echo $?
0
# lsmod | grep loop
loop 15240 0
# modprobe -r loop
# modprobe loop-aes
# lsmod | grep loop
loop 59532 0
As the CD image is mounted at that time (and so in theory could other
things), I think what we'd need to do when loading loop-aes is:
- test if loop is already loaded
- somehow test if it is the correct variant of the module or not
- if the wrong module is loaded
- get and save mount info for loop-mounted devices:
'grep "/dev/loop" /proc/mounts'
- umount the devices (if one fails: remount others and error)
- 'modprobe -r loop; modprobe loop-aes'
- remount the loop-mounted devices using the saved info
If someone can come up with a test that can be used for the second step,
coding this should be relatively straightforward.
Cheers,
FJP
pgpHKMDCkSwUx.pgp
Description: PGP signature
