From mboxrd@z Thu Jan  1 00:00:00 1970
From: Hugo Tyson <hmt@cygnus.co.uk>
To: ecos-discuss@sources.redhat.com
Subject: Re: [ECOS] Loopback patch adopted OK
Date: Wed, 05 Sep 2001 00:10:00 -0000
Message-ID: <ptya3zgc69.fsf_-_@masala.cygnus.co.uk>
References: <000401bfd787$80e39ae0$d30110ac@netappi.com> <pt7lbjjnru.fsf@masala.cygnus.co.uk>
X-SW-Source: 2001-09/msg00087.html
Message-ID: <20010905001000.CwxSt7hCKYKgqrPFl8c-8gORmBDn0XhpN3xNyMMUgm0@z>


Hugo Tyson <hmt@cygnus.co.uk> writes:
> "Sorin Babeanu" <sorin@netappi.com> writes:
> > The attached patch contains a few modifications I have made in order to make
> > the loopback interface work on eCos without a real network device.
> > It also contains tests for ping, udp and tcp.
> 
> Just to keep folks up to date: Sorin, thanks for that.  I'm in the process
> of adopting it into the trunk - actually a variation on your patch which
> sets up the loopback dev in init_all_network_interfaces(), and is more
> configurable for 0,1,N loopback devs.
> 
> The tests are great, thanks, we can start to run meaningful network tests
> on all platforms with those in place, I'll just be adding PASS/FAIL
> messages so we can automate those in our test farm.
> 
> It'll all show up in anoncvs by the end of this week.
> 
> Is there any call for me to supply an alternative patch on this forum?
> I won't do so unless people ask.  Any replies to the list please, not me.

Phew, done that... ;-)

Sorin, I didn't see what the change in if_loop.c (pasted below) achieved,
so I didn't adopt it.  Can you explain more, convince me?  (I have got the
location it patches right haven't I, ie. looutput()?)

It seems to just do a load of checking twice, in a different order; I mean
the original looutput() checks for RTF_REJECT|RTF_BLACKHOLE &c, and returns
EAFNOSUPPORT with a bad AF.  Was this just experimentation while you were
testing the startup code, that leaked out?

Anyway, thanks again for the contribution; keep up the good work!

Watch anoncvs for our updates to emerge... end of the week sometime.

	- Huge


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 int
 looutput(ifp, m, dst, rt)
        struct ifnet *ifp;
        register struct mbuf *m;
 	struct sockaddr *dst;
 	register struct rtentry *rt;
 {
+        if ((m->m_flags & M_PKTHDR) == 0)
+                panic("looutput no HDR");
+
+        if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
+                m_freem(m);
+                return (rt->rt_flags & RTF_BLACKHOLE ? 0 :
+                        rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
+        }
+        ifp->if_opackets++;
+        ifp->if_obytes += m->m_pkthdr.len;
+#if 1   /* XXX */
+        switch (dst->sa_family) {
+        case AF_INET:
+        case AF_IPX:
+        case AF_NS:
+        case AF_ISO:
+        case AF_APPLETALK:
+                break;
+        default:
+                printf("looutput: af=%d unexpected", dst->sa_family);
+                m_freem(m);
+                return (EAFNOSUPPORT);
+        }
+#endif
+        return(if_simloop(ifp, m, dst, 0));
+}
+
+
+int
+if_simloop(ifp, m, dst, rt)
+	struct ifnet *ifp;
+	register struct mbuf *m;
+	struct sockaddr *dst;
+	register struct rtentry *rt;
+{
 	int s, isr;
 	register struct ifqueue *ifq = 0;
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%