Re: segfault on kernel v2.4.0

[Michael Driscoll <fenris@ulf.edgemail.com>]

On Wed, Jan 31, 2001 at 01:01:35AM +0100, Dejan Muhamedagic wrote:
> I'll try to clarify:
> 
> 1.  It is a segfault (in losetup).  No oopses.
>
> 2.  The kernel (2.4.0) seems to be OK.  loop devices work fine.
> Other stuff as well, i.e. sound, pcmcia network.  Furthermore, I
> tested the IDEA cipher and it seems to be OK.  So, it is only the
> Blowfish code being troubled.
> 
> 3.  With or without other modules, or in single-user mode, the
> same thing happens.  I agree that the log is weird, but I'm no
> kernel expert either (what on earth has a cipher module to do with
> the pcmcia?).  Also, when I try the same by removing all other
> modules previously, in syslog there is a referrence to the
> ipchains module :)
> 
> 4.  All ciphers are compiled as modules.
> 
> So, I can switch to IDEA.  Is it advisable to use this cipher in a
> production?  Any reports about its stability?  About the
> international patch?  Or should I stay put with the v2.2.x kernels
> for the time being?

I get the same problem (using blowfish with a loopback ext2 filesystem).

sudo losetup -e blowfish -k 160 /dev/loop0 ~/encrypted-fs
(losetup segfaults here)
(/dev/loop0 becomes unusable, further attempts get an EBUSY)

Then I try:
sudo losetup -e blowfish -k 160 /dev/loop1 ~/encrypted-fs
                                ^^^^^^^^^^

And it works fine (ie. only the first attempt to losetup segfaults).

Everything is modular in this kernel, except for some things like ext2 which
I need to boot.

And now that I look at my logs, I am also getting an oops when I try
to load blowfish:

Jan 30 17:14:07 diva kernel: loop: enabling 8 loop devices
Jan 30 17:14:08 diva modprobe: modprobe: Can't locate module cipher-blowfish-cbcJan 30 17:14:08 diva kernel: cipher-blowfish-cbccipher-blowfish<6>cryptoapi: Registered blowfish-ecb (0)
Jan 30 17:14:08 diva kernel: cryptoapi: Registered blowfish-cbc (65536)
Jan 30 17:14:08 diva kernel: divide error: 0000
Jan 30 17:14:08 diva kernel: CPU:    0
Jan 30 17:14:08 diva kernel: EIP:    0010:[snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+138251/14622137]
Jan 30 17:14:08 diva kernel: EFLAGS: 00010246
Jan 30 17:14:08 diva kernel: eax: 00000001   ebx: c618bf44   ecx: 00000000   edx: 00000000
Jan 30 17:14:08 diva kernel: esi: 00000000   edi: 00000400   ebp: 00000000   esp: c618bdf8
Jan 30 17:14:08 diva kernel: ds: 0018   es: 0018   ss: 0018
Jan 30 17:14:08 diva kernel: Process losetup (pid: 431, stackpage=c618b000)
Jan 30 17:14:08 diva kernel: Stack: c88d2100 c745e000 c88d2100 fffffff4 c618be18 c0231180 c745e07c c745e034
Jan 30 17:14:08 diva kernel:        c7f87480 c1236800 c88cb459 c745e000 c618bf44 00000000 00000000 c618bee4
Jan 30 17:14:08 diva kernel:        c88cd1c8 c745e000 c618bf44 00000000 c745e000 c88ced5c 00000012 00000000
Jan 30 17:14:08 diva kernel: Call Trace: [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+144288/14616100] [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+144288/14616100] [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+116473/14643915] [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+124008/14636380] [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+131068/14629320] [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+126662/14633726] [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+127209/14633179]
Jan 30 17:14:08 diva kernel:        [snd-seq-midi:__insmod_snd-seq-midi_S.bss_L32+127914/14632474] [blkdev_ioctl+40/56] [sys_ioctl+363/388] [system_call+51/56]
Jan 30 17:14:08 diva kernel:
Jan 30 17:14:08 diva kernel: Code: f7 fd 0f b6 1c 19 c1 e3 18 89 5c 24 14 8b 5c 24 30 0f b6 04

My sound driver is ALSA's YMFPCI.

So it looks like there is a problem in the blowfish init code?

Now that I know this is a blowfish problem I can reformat using AES or
something, I'm glad to know there is a workaround.

-- 
Michael Driscoll
fenris@ulf.edgemail.com

Linux-crypto:  cryptography in and on the Linux system
Archive:       http://mail.nl.linux.org/linux-crypto/