sowmini.varadhan at sun.com wrote:
> not from looking at netbsd (http://tinyurl.com/be9kdv)
> rip_input() has:
>
> /*
> * XXX Compatibility: programs using raw IP expect ip_len
> * XXX to have the header length subtracted, and in host order.
> * XXX ip_off is also expected to be host order.
> */
> hlen = ip->ip_hl << 2;
> ip->ip_len = ntohs(ip->ip_len) - hlen;
> NTOHS(ip->ip_off);
>
> The problem with having icmp_bsd_compat as an ndd tunable is
> that we now need to have *all* apps running in bsd compat mode,
> or all apps running in the non-compat mode.
I guess you are going to add a new option for an app
to override this. Right? This is good :-)
> I don't think "interesting alignment" is something we should
> throw on the admin's shoulder to figure out.
Actually, it is probably not the sys admin who may be
interested in this. There are third parties who
configure a Solaris machine to do specific things and
ship the configured machine to customers. And customers
just use the machine as is. So I am wondering if this
is useful to those third party vendors.
> But what I was
> thinking was to have the *_wroff_extra set to the larger
> of (32, ill_phys_addr_length) with the appropriate rounding
> off to the beat alignment boundary.
--
K. Poon.
kacheong.poon at sun.com
