[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
(usagi-users 03109) Re: ipv6_addr_cmp performance patch
- To: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx>
- Subject: (usagi-users 03109) Re: ipv6_addr_cmp performance patch
- From: Henrik Nordstrom <hno@xxxxxxxxxxxxxxx>
- Date: Wed, 10 Nov 2004 00:30:03 +0100 (CET)
- Cc: "Craig, Dave" <dwcraig@xxxxxxxxxxxx>, usagi-users@xxxxxxxxxxxxxx, linux-net@xxxxxxxxxxxxxxx
- In-reply-to: <E1CRcet-0001Jc-00@gondolin.me.apana.org.au>
- References: <E1CRcet-0001Jc-00@gondolin.me.apana.org.au>
- Reply-to: usagi-users@xxxxxxxxxxxxxx
- Resent-date: Wed, 10 Nov 2004 08:52:29 +0900
- Resent-from: sekiya@xxxxxxxxxxxxxx
- Resent-message-id: <200411100852.FMLAAB27215.usagi-users@linux-ipv6.org>
- Resent-to: usagi-users@xxxxxxxxxxxxxx (moderated)
On Wed, 10 Nov 2004, Herbert Xu wrote:
What if the addresses are not aligned on a platform where it matters?
Can this at all happen with the sockaddr_in6/in6_addr alignment
requirements? It is already a union of uint32_t.
I suppose it can happen if a user is handcrafting his structures, not
letting the compiler align properly, but I would say this is a bug if
done.
With the size of the union a nice power of 2 (16 bytes) it is extremely
unlikely there is alignment padding holes which might eventually cause
different alignment of the different types of the union..
Regards
Henrik