john-dev - formats failing test on big-endian

Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130206143508.GA25579@openwall.com>
Date: Wed, 6 Feb 2013 18:35:08 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: formats failing test on big-endian

magnum, all -

I built unstable-jumbo on debian_etch_sparc_small.qcow2 from
http://people.debian.org/~aurel32/qemu/

This took 72 minutes on bull (using one core):

real    72m16.424s
user    69m18.960s
sys     2m40.630s

For comparison, 1.7.9-jumbo-7 took 40 minutes.  Looks like we've almost
doubled the amount of code, or at least complexity for the compiler.

After fixing the [21] vs. [22] issue for the revised MSCHAPv2 and
NETNTLM formats, there are still plenty of formats failing tests -
although none are crashing the program anymore:

Benchmarking: dynamic_17: phpass ($P$ or $H$) [32/32 X2  (MD5_body)]... FAILED (get_hash[0](1))
Benchmarking: Eggdrop Blowfish [32/32]... FAILED (cmp_all(1))
Benchmarking: OpenVMS Purdy [32/32]... FAILED (get_hash[0](0))
Benchmarking: Apple DMG PBKDF2-HMAC-SHA-1 3DES / AES [32/32]... FAILED (cmp_all(1))
Benchmarking: EncFS PBKDF2 AES / Blowfish [32/32]... FAILED (cmp_all(1))
Benchmarking: Mac OS X Keychain PBKDF2-HMAC-SHA-1 3DES [32/32]... FAILED (cmp_all(1))
Benchmarking: Kerberos 5 AS-REQ Pre-Auth etype 17/18 aes-cts-hmac-sha1-96 [32/32]... FAILED (get_hash[0](0))
Benchmarking: KDE KWallet SHA-1 [32/32]... FAILED (cmp_all(1))
Benchmarking: MS SQL 2005 SHA-1 [32/32]... FAILED (get_hash[0](0))
Benchmarking: MS SQL 2012 SHA512 [32/32 OpenSSL]... FAILED (get_hash[0](0))
Benchmarking: ODF SHA-1 Blowfish / SHA-256 AES [32/32 OpenSSL]... FAILED (get_hash[0](0))
Benchmarking: Office 2007/2010 (SHA-1) / 2013 (SHA-512), with AES [32/32 OpenSSL]... FAILED (max. length in index 0: wrote 32, got 0 back)
Benchmarking: Office <= 2003 MD5/SHA-1, RC4 [32/32]... FAILED (max. length in index 0: wrote 64, got 0 back)
Benchmarking: PBKDF2-HMAC-SHA512 GRUB2 / OS X 10.8 [32/32 OpenSSL]... FAILED (get_hash[0](0))
Benchmarking: Password Safe SHA-256 [32/32]... FAILED (get_hash[0](0))
Benchmarking: STRIP Password Manager PBKDF2-SHA1 [32/32]... FAILED (cmp_all(1))
Benchmarking: SXC SHA-1 Blowfish [32/32]... FAILED (get_hash[0](0))
Benchmarking: WPA-PSK PBKDF2-HMAC-SHA-1 [32/32]... FAILED (valid)
Benchmarking: RAR3 SHA-1 AES (4 characters) [32/32]... FAILED (max. length in index 0: wrote 16, got 0 back)
Benchmarking: WinZip PBKDF2-HMAC-SHA-1 [32/32]... FAILED (cmp_all(1))

20 out of 192 tests have FAILED

This was a default build - I did not enable any of the formats requiring
extra libs and a change to the Makefile.

Perhaps we'll want to get at least some of these fixed so that they'll
work on big-endian before the next jumbo release, and we'll need to
document the rest in doc/BUGS.

Some of these formats were present in 1.7.9-jumbo-7 and passed test -
e.g., phpass.  Only these four formats failed test in 1.7.9-jumbo-7:

Benchmarking: MS SQL 2005 SHA-1 [32/32]... FAILED (get_hash[0](0))
Benchmarking: Office 2007/2010 SHA-1/AES [32/32]... FAILED (cmp_all(1))
Benchmarking: WPA-PSK PBKDF2-HMAC-SHA-1 [32/32]... FAILED (valid)
Benchmarking: RAR3 SHA-1 AES (4 characters) [32/32]... FAILED (cmp_all(1))

(tested on the same system as above, with the 40-minute build).

So it looks like we introduced bugs to previously existing code.

1.7.9-jumbo-7 required a minor change to get it to build, though: in
gost_plug.c the rhash_u32_swap_copy() should be wrapped in #ifdef
CPU_BIG_ENDIAN ... #endif, instead of the #ifndef __GLIBC__ condition
that is present in 1.7.9-jumbo-7's revision of that code.
unstable-jumbo does not have this issue (it built right away).

All of these builds were with "make linux-sparc".

Alexander
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.