NFS mount from multi-homed server over TCP fails

Question

nico: Solaris 9 NFS server, multihomed into three different networks (A, B, C)
primary interface on A.
benno: Solaris 10, NFS client, on network
genie: RHEL5, NFS client, on network B
ilsco: Solaris 10, NFS client, on network B
I have immediate access to all servers except ilsco -- which, of course, is the one with the problem. =( If genie tries to mount a share from nico using its address on network A, everything works. If it tries to mount it using the address on network B, it fails.
In all cases, I can see the client make request to the server's portmapper daemon for the mountd port number. That request is issued to nico's address in network B. Nico replies, from its primary address in network A. When NFS works, I then see a null mount request issued by the client. Genie never issues that mount request.
Working request from benno:
[CLIENT] -> [SERVER, ADDR B] PORTMAP C GETPORT prog=100005 (MOUNT) vers=3 proto=UDP
[SERVER, ADDR A] -> [CLIENT] PORTMAP R GETPORT port=58352
[CLIENT] -> [SERVER, ADDR B] MOUNT3 C Null
[SERVER, ADDR A] -> [CLIENT] MOUNT3 R Null Non-working request from ilsco:
[CLIENT] -> [SERVER, ADDR B] PORTMAP C GETPORT prog=100005 (MOUNT) vers=3 proto=UDP
[SERVER, ADDR B] -> [CLIENT] TCP D=1019 S=2049 Ack=3278107669 Seq=2214943315 Len=0 Win=49640
[SERVER, ADDR B] -> [CLIENT] TCP D=1019 S=2049 Fin Ack=3278107669 Seq=2214943315 Len=0 Win=49640
[CLIENT] -> [SERVER, ADDR B] TCP D=2049 S=1019 Ack=2214943316 Seq=3278107669 Len=0 Win=21900
[SERVER, ADDR A] -> [CLIENT] PORTMAP R GETPORT port=32804
[CLIENT] -> [SERVER, ADDR B] PORTMAP C GETPORT prog=100005 (MOUNT) vers=3 proto=UDP (retransmit)
[SERVER, ADDR A] -> [CLIENT] PORTMAP R GETPORT port=32804
[CLIENT] -> [SERVER, ADDR B] PORTMAP C GETPORT prog=100005 (MOUNT) vers=3 proto=UDP (retransmit)
[SERVER, ADDR A] -> [CLIENT] PORTMAP R GETPORT port=32804Now, I've noted that, despite using the "-o proto=tcp" option, the portmap requests were pushed out over UDP. On benno, some of the traffic after the portmap request took place on the network A address; some of the traffic took place on the network B address.
On genie (the Linux box), I noted that the portmap requests were pushed out over the TCP and all traffic stays on the network B addresses.
Any suggestions, etc. as to what I am doing wrong or what else I should be looking at would be helpful.

Answer 1

Turned out it might just be a firewall issue... I didn't think about it first :(
Sorry for that :x