click 1.0.5 -- patch

[Richard Mortier]

Hi,

Another patch, this time for EtherEncap and UDPEncap.  It appears
that EtherEncap, when "pushing" the sk_buff to get space for the
14 byte Ethernet header, forgets to pull the data pointer into
the sk_buff up to the head pointer.  This results in the Ethernet
header getting smacked over the first 14 bytes of the IP Header.
In UDPEncap, there is a missing htons() around the length field,
and it also doesn't seem to use the _cksum flag to decide whether
to checksum or not.  I've assumed that the flag decides whether
the checksum is the calculated value, or 0 for a "lazy" host.

Patch appended to this mail.

--
Richard Mortier                 --------------------
-----------------        ------ Computer Laboratory,
rmm1002 at cam.ac.uk  ----- University of Cambridge, UK
mort at ieee.org ---- http://www.cl.cam.ac.uk/~rmm1002/

diff -ur /usr/groups/srgnet/click-1.0.5-pristine/elements/standard/udpencap.cc click-1.0.5/elements/standard/udpencap.cc
--- /usr/groups/srgnet/click-1.0.5-pristine/elements/standard/udpencap.cc	Wed Mar 29 16:38:31 2000
+++ click-1.0.5/elements/standard/udpencap.cc	Wed May  3 16:35:18 2000
@@ -62,8 +62,8 @@
   
   udp->uh_sport = htons(_sport);
   udp->uh_dport = htons(_dport);
-  udp->uh_ulen = p->length();
-  udp->uh_sum = in_cksum(p->data(), p->length());
+  udp->uh_ulen  = htons(p->length());
+  udp->uh_sum   = (_cksum? in_cksum(p->data(), p->length()) : 0);
   return p;
 }
 
diff -ur /usr/groups/srgnet/click-1.0.5-pristine/lib/packet.cc click-1.0.5/lib/packet.cc
--- /usr/groups/srgnet/click-1.0.5-pristine/lib/packet.cc	Sat Apr 29 04:09:32 2000
+++ click-1.0.5/lib/packet.cc	Wed May  3 17:30:51 2000
@@ -196,6 +196,7 @@
   click_chatter("expensive Packet::push");
 #ifdef __KERNEL__
   Packet *q = Packet::make(skb_realloc_headroom(skb(), nbytes));
+  (void)skb_push(q->skb(), nbytes);
 #else
   Packet *q = Packet::make(length() + nbytes);
   memcpy(q->data() + nbytes, data(), length());