[coLinux-users] change in slirp behavior from 0.7.1 to 0.7.3 with udp forwarded ports

[Mike F. <vap...@gm...>]

i took a working system that is running on 0.7.1 and looked at
updating to 0.7.3, but my slirp networking seems to break (sort of).
i'm using dhcp on the slirp network device.  same rootfs, so same
userspace configs between the two setups ...

here's my network settings given to colinux:
eth0=slirp,,tcp:22:22/tcp:333:22/tcp:10000:10000/udp:69:69
eth1=tuntap

and here's my network settings in userspace (/etc/network/interfaces):
auto eth0 eth1
iface eth0 inet dhcp
iface eth1 inet static
  address 10.12.12.2
  netmask 255.255.255.0

i use colinux as a tftp server.  so the setup is:
 - colinux side of slirp is 10.0.2.15
 - windows side of slirp is 10.0.2.{2,3} (i guess dns/gateway have diff IPs?)
 - windows public ip is 192.168.0.35
 - running tftp client on 192.168.0.17 ("G5")

on the working system (0.7.1), i can see this communication when
running tcpdump from inside colinux:
01:34:47.410000 IP G5.60808 > 10.0.2.15.tftp:  15 RRQ "moo" netascii
01:34:47.413067 IP 10.0.2.15.32797 > 10.0.2.3.domain:  59766+ PTR?
15.2.0.10.in-addr.arpa. (40)
01:34:47.447703 IP 10.0.2.15.32798 > G5.60808: UDP, length 9

on the failing system (0.7.3), i see this:
01:44:45.645802 IP 10.0.2.2.33431 > 10.0.2.15.tftp:  15 RRQ "moo" netascii
01:44:46.105809 IP 10.0.2.15.32769 > 10.0.2.2.33431: UDP, length 9
01:44:45.675802 IP 10.0.2.2 > 10.0.2.15: ICMP 10.0.2.2 udp port 33431
unreachable, length 45

for some reason, the incoming packets from the tftp client as seen by
colinux have the gateway's ip (10.0.2.2) instead of the real ip
(192.168.0.17) ... any hints as to what's going wrong here ?

taking a step back and just hitting colinux with netcat shows the same
behavior.  connecting to a tcp forwarded port and things look good
(the source ip is the real client, not the gateway).  connecting to a
udp forwarded port and things look bad (the source ip is the gateway
instead of the real client).
-mike