[Bug 267301] nfs uses udp for portmapper even when proto=tcp specified

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=267301

            Bug ID: 267301
           Summary: nfs uses udp for portmapper even when proto=tcp
                    specified
           Product: Base System
           Version: 13.1-STABLE
          Hardware: amd64
                OS: Any
            Status: New
          Severity: Affects Some People
          Priority: ---
         Component: kern
          Assignee: bugs@FreeBSD.org
          Reporter: brad@facefault.org

When attempting to mount a remote file share using NFSv3, portmapper calls use
UDP even when TCP is explicitly specified. Using the port= option would be a
satisfactory workaround but that doesn't eliminate portmapper calls.

Expected behavior: TCP explicitly specified means TCP everywhere.

Repro:

1. Start tcpdump with capture filter "port 111".
2. Attempt to mount an NFS share with proto=tcp.
3. The tcpdump contains UDP traffic to the remote portmapper when TCP is
expected.
4. Block outgoing UDP/111 from the client. (I guess this one's not strictly
necessary.)
5. Repeat #2 but add port=2049 to the mount options.
6. The tcpdump contains a successful NULL call but the client immediately tried
contacting the portmapper on UDP to get the mount port.

This prevents using Azure Blob Storage NFS, and I suspect there are some bad
and hard-to-get-fixed firewall configurations out there since Linux's behavior
is to use TCP for portmapper in this case.

-- 
You are receiving this mail because:
You are the assignee for the bug.