Cisco VPN + EVDO

[russv]

When running a combination of the Cisco 4.9.01(0080) client on OS/X and a Verizon EVDO card, I find that my WinXP VM can no longer access resources across the tunnel via TCP/IP. My WinXP instance can ping (ICMP) remote sites and DNS resolution works fine but TCP is busted. A tcpdump on OS/X shows that, for some reason, the data isn't getting NAT'ed and tunneled correctly.

Ping resource:

23:32:06.488645 IP (tos 0x0, ttl 128, id 42, offset 0, flags [none], proto: UDP (17), length: 120) 75.214.144.46.4500 > 3.2.1.221.4500: [no cksum] UDP-encap: ESP(spi=0x439412ce,seq=0x2a), length 92
23:32:07.199595 IP (tos 0x0, ttl 46, id 24416, offset 0, flags [none], proto: UDP (17), length: 120) 3.2.1.221.4500 > 75.214.144.46.4500: [no cksum] UDP-encap: ESP(spi=0x17d8d097,seq=0x27), length 92
23:32:07.491412 IP (tos 0x0, ttl 128, id 43, offset 0, flags [none], proto: UDP (17), length: 120) 75.214.144.46.4500 > 3.2.1.221.4500: [no cksum] UDP-encap: ESP(spi=0x439412ce,seq=0x2b), length 92
23:32:07.654721 IP (tos 0x0, ttl 46, id 24441, offset 0, flags [none], proto: UDP (17), length: 120) 3.2.1.221.4500 > 75.214.144.46.4500: [no cksum] UDP-encap: ESP(spi=0x17d8d097,seq=0x28), length 92

Telnet resource Port 80:

23:32:08.937170 IP (tos 0x0, ttl 128, id 186, offset 0, flags [DF], proto: TCP (6), length: 48) 75.214.144.46.49178 > 172.16.0.17.80: S, cksum 0x4665 (correct), 1793362445:1793362445(0) win 65535 <mss 1260,nop,nop,sackOK>
23:32:11.887264 IP (tos 0x0, ttl 128, id 187, offset 0, flags [DF], proto: TCP (6), length: 48) 75.214.144.46.49178 > 172.16.0.17.80: S, cksum 0x4665 (correct), 1793362445:1793362445(0) win 65535 <mss 1260,nop,nop,sackOK>
23:32:17.795702 IP (tos 0x0, ttl 128, id 188, offset 0, flags [DF], proto: TCP (6), length: 48) 75.214.144.46.49178 > 172.16.0.17.80: S, cksum 0x4665 (correct), 1793362445:1793362445(0) win 65535 <mss 1260,nop,nop,sackOK>

As you can see, for some reason when I try to access our corporate server (in both tests I was trying 172.16.0.17) via TCP (in this case port 80 using telnet), it doesn't get translated and sent through the VPN. However, in the earlier ICMP example, the data is heading out via the VPN connection (where our VPN server is 3.2.1.221).

I know this worked back in Parallels 2.5 using shared networking. Now in 3.0 (build 4560 and 4124) it doesn't work.

Anyone know how to resolve this?