On Mon, Oct 02, 2000 at 10:20:09AM -0500, Eric Haszlakiewicz wrote:

 > 	As for not padding the final sockaddr: this seems like a not unreasonable
 > thing for something to do.  Therefore, instead of changing routed to pad
 > the last sockaddr I'm going to relax the check in the kernel a little
 > (the last addr may take up either sa_len or ALIGN(sa_len) bytes).
 > This also means that the COMPAT_15 ifdef can go away since on the architectures
 > where it would matter stuff didn't work anyway.

Doesn't this basically boil down to applications simply doing unsafe
things?  Shouldn't the applications just be changed to always copy
something from a socket into a known-aligned buffer if the data structure
needs strict alignment?

-- 
        -- Jason R. Thorpe <thorpej@zembu.com>