First, thanks to Alexey for explaining some basic misunderstandings I
had about the networking code - I had got myself quite confused. Being
able to discuss things with the developers makes a very refeshing
change from commercial Un*x.
This time my question is about ip_rt_redirect. I'm pretty sure that it
doesn't actually work. I believe that the structure of the code is roughly:
For all possible keys,
{
Search through the route cache linked list for the matching key
if the key exists,
{
Allocate memory for a new route entry dst_alloc
Copy the old route cache entry *rt=*rth
Fix the gateway value rt->rt_gateway = new_gw
Check the gateway is responding to arp (?) arp_bind_neighbour
Delete the old route cache entry rt_del
Insert the new route cache entry rt_intern_hash
}
}
rt_intern_hash adds a new entry to the hash table provided that there
is no existing entry with a matching key. There usually won't be a
match because we have only just deleted the old route. Well, that's
what I think it meant to happen. We actually delete the new route so
packets continue to be sent to the old gateway. Here's a patch:
--- linux-2.3.18/net/ipv4/route.c.orig Thu Oct 7 05:34:08 1999
+++ linux-2.3.18/net/ipv4/route.c Thu Oct 7 05:34:49 1999
@@ -735,7 +735,7 @@
goto do_next;
}
- rt_del(hash, rt);
+ rt_del(hash, rth);
if (!rt_intern_hash(hash, rt, &rt))
ip_rt_put(rt);
goto do_next;
Peter
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
