On 6/20/25 10:37 AM, Herbert Xu wrote:
On Fri, Jun 20, 2025 at 10:10:53AM +0200, Milan Broz wrote:
Now I get wrong data instead of fail (both on 32bit and 64bit).
Patch just applied over today's Linus' tree
...
PBKDF vector 20 pbkdf2-whirlpool [FAILED]
got: 58 55 1e ef 29 40 d6 a2 f0 59 e0 d9 4a 50 c5 df 01 25 be ee 27 5b 35 47 6d 37 38 13 0f e0 da 29
want: 9c 1c 74 f5 88 26 e7 6a 53 58 f4 0c 39 e7 80 89 07 c0 31 19 9a 50 a2 48 f1 d9 fe 78 64 e5 84 50
PBKDF test failed.
(Whirlpool is translated to wp512 in the crypto backend and despite it is a quite rare use, some people
used if for LUKS PBKDF2. Actually the whole vector test was reaction to wrong Whirlpool implementation
in gcrypt years ago. It apparently can find breakage even today :-)
Oops, I forgot to increment the hash length for the final partial
update :)
It s still failing for me for that userspace crypt API.
Milan
