Hello,
On Thu, Aug 11, 2022 at 11:59:55AM +0200, Alexander Bluhm wrote:
> Hi,
>
> ip_fragment() and ip6_fragment() do nearly the same thing, but they
> are implemented differently.
>
> The idea of this diff is to move things around. Then only differences
> from the IP standards but not in coding style remain. This allows
> to compare both functions easily.
don't mind at all if things will be more consistent.
>
> In IPv4 assume that m_pkthdr is correct and take the length from
> the, like in IPv6.
>
would it make sense to put asssert there at least for a while
just to discover any surprises? something like this:
> -ip_fragment(struct mbuf *m, struct mbuf_list *fml, struct ifnet *ifp,
> +ip_fragment(struct mbuf *m0, struct mbuf_list *fml, struct ifnet *ifp,
> u_long mtu)
> {
> - struct ip *ip, *mhip;
> - struct mbuf *m0;
> - int len, hlen, off;
> - int mhlen, firstlen;
> + struct mbuf *m;
> + struct ip *ip;
> + int firstlen, hlen, tlen, len, off;
> int error;
>
> ml_init(fml);
> - ml_enqueue(fml, m);
> + ml_enqueue(fml, m0);
>
> - ip = mtod(m, struct ip *);
> + ip = mtod(m0, struct ip *);
> hlen = ip->ip_hl << 2;
> + tlen = m0->m_pkthdr.len;
KASSERT(tlen == ntohs(ip->ip_len));
diff looks good to me (with or without assert).
OK sashan
