Milter-greylist

Hi,

On Mon, 2006-03-27 at 09:18 +0900, Hajimu UMEMOTO wrote:
> Hi,
> 
> >>>>> On Mon, 27 Mar 2006 02:01:20 +0300
> >>>>> Ranko Zivojnovic <ranko@...> said:
> 
> ranko> ! #define BUCKET_HASH_V6(sa, from, rcpt, bucket_count)
> \
> ranko> !   ((((uint32_t)(SADDR6(sa)->s6_addr)[3] &                  \
> ranko> !      (uint32_t)(((ipaddr
> *)&conf.c_match_mask)->in6.s6_addr)[3])\
> ranko> !     ^ F2B_SPICE(from, rcpt))                              \
> ranko> !    % bucket_count)
> 
> It seems this test is broken for an IPv6.  Why do you check only
> (uint32_t)(SADDR6(sa)->s6_addr)[3]?  We need to check whole address.
> 

I was taking only the last part of the v6 address because that was the
place I would expect the most "activity" - and we are only creating the
hash. But either way you're right - we should take the whole address
length into account. Especially now with the netmask in the game.

Please consider this version.

Best regards,

Ranko