svn commit: r186393 - head/sys/netinet6

[Bjoern A. Zeeb]

On Tue, 23 Dec 2008, Robert Watson wrote:

Hi,

> In IPv4 we've eliminated use of ipprotosw, although we still have 
> ipprotosw.h, which should go away as it's unused; I'd like to see ip6protosw 
> go away as well.  The domain registration code assumes that when an array of 
> protocols are registered, the size of the array entry is sizeof(struct 
> protosw), which happens currently to be true for ip6protosw.  At least, I 
> think. :-)

There is some assert like code under DIAGNOSTICS in there I think.


So from what I could see the main problems just changing to protosw
from ip6protosw are: the pr_input and pr_output routines as they
differ in either return code or arguemnts:

* pr_input v4     void (struct mbuf *, int)
            v6     int  (struct mbuf **, int *, int)

* pr_output v4    int (struct mbuf *, struct socket *)
             v6    int (struct mbuf *, ...)

All the other function pointers seem equal already.

I think the pr_output is solveable more easily as I can see - which
seemed strange only

rip6_output     int (struct mbuf *, struct socket *,
                      struct sockaddr_in6 *, struct mbuf *)

called from raw_usend(m, so), rip6_send(m, so, dst, control) are
defined. I might have missed a few because they weren't yet converted
to c99 style initializers.

pr_input is more of a problem and we may need to break the INET case
to be able to merge the INET6 case to use protosw. I'll further
investigate but this might become a bit disruptive.

/bz

-- 
Bjoern A. Zeeb                      The greatest risk is not taking one.