Bug 296409

the init scripts are werners work. And if this is an blocker i eat a broom...

1) As Werner said, boot.crypto always required boot.localfs. I didn't check whether that's still really necessary. Does yast allow you to encrypt only /usr?

2) boot.crypto runs /etc/init.d/kbd to set up the keyboard mapping. Did you put /usr on a crypted partition?

I didn't put /usr on a crypted partition. The two partitions I created where only backup/data partitons and /boot and / are located in my fstab before the crypted devices.

I can only say: it worked before perfectly (beta 5/6 IIRC) and now no longer with alpha 7. I needed the changes from the initial description to not fall into the maintainance mode (I assume because boot.localfs tried to mount a not encrypted partition). Maybe something else cause the problem, but this was the only way for me to boot my system.

There were no changes to the cryptsetup package since mid june. Did you use yast to freshly create the encrypted volumes? What's the content of your /etc/fstab and /etc/crypttab resp /etc/cryptotab?

fs_freq and fs_passno are set for sda9 therefore fsck tries to check the mapped device which doesn't exist yet. The entry for sda8 is correct. Please attach yast logs and reassign to yast.

The needed logfiles are e.g. in bug #296393

I reassign it, but what about the keyboard layout problems?

I have another (but probably related) problem:
used /etc/fstab from 10.2:
/etc/vpn.file   /etc/openvpn    ext2    loop=/dev/loop1,encryption=twofish256,acl,noauto,user_xattr 0 0

copy /etc/vpn.file
and after mount /etc/openvpn (and write same password):
ioctl: LOOP_SET_STATUS: No such file or directory

from strace:
...
ioctl(6, SNDCTL_TMR_CONTINUE or TCSETSF, {B38400 opost isig icanon echo ...}) = 0
close(6)                                = 0
munmap(0xb7f0a000, 4096)                = 0
ioctl(4, 0x4c00, 0x3)                   = 0
close(3)                                = 0
ioctl(4, 0x4c04, 0xbfde8810)            = -1 ENOENT (No such file or directory)
ioctl(4, 0x4c02, 0xbfde88f8)            = -1 ENOENT (No such file or directory)
dup(2)                                  = 3
fcntl64(3, F_GETFL)                     = 0x402 (flags O_RDWR|O_APPEND)
fstat64(3, {st_mode=S_IFCHR|0600, st_rdev=makedev(136, 1), ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb7f0a000
_llseek(3, 0, 0xbfde865c, SEEK_CUR)     = -1 ESPIPE (Illegal seek)
...

I suppose I need only this modules (they are inserted):
cryptoloop              7040  0
loop                   21508  1 cryptoloop
twofish                10880  0
twofish_common         38912  1 twofish


btw - rpm -ql cryptsetup shows both /etc/cryptotab and /etc/crypttab but they're not included in rpm (cryptsetup-1.0.5-20)

wrt keyboard problem: boot.crypto runs before boot.cleanup. kbd checks for presence of /var/run/keymap and exits early. The file is at this point left over from the previous boot so kbd doesn't actually do anything useful. Fixed by rm -f /var/run/keymap in boot.crypto

Michal: Your problem is unrelated. The vanilla upstream losetup doesn't have the ability to parse strings like 'twofish256' which should be reintroduced with a (potentially suse specific) patch. Please open a separate bug report for util-linux.

Reassigning to Thomas to look at fstab problem (comment 7 and 8)

ad comment#10:
file based encryption is different in SL 10.3 and 10.2.
encrypted fstab entries from 10.2 will not work any more.

Still investigating, why the fstab line for sda9 is wrong and sda8 is correct.

So far I was not able to reproduce the problem with having non-zero values
for freq and passno fields of encrypted devices in fstab.
I did an installation with exactly the same disk layout than in the report
but the resulting fstab was correct.

Interestingly the resulting fstab differed also in other lines from the one
pasted in comment 7 (order of /boot and / entry was inverted and by-id name
was used for swap partition). So maybe there was some bug in Alpha7 fixed in
beta#2 that had this as a side effect.

Please retest with beta#3. I just added more debug code to fstab handling in
libstorage so maybe this will help debugging thus further.

Should be fixed meanwhile.
Reopen with y2log files if this still occurs.