Kris Kennaway wrote: > A few months ago I posted about rpc.lockd interop problems I am having > between my 5.0 NFS client and a Redhat 7.1 server. Both are running > rpc.lockd, but when I send a lock request to the server it hangs > forever blocked on the /var/run/lock socket. > > tcpdump shows that the lock RPC request is being sent, and answered by > the server. I have the complete tcpdump trace if anyone is > interested.
This might be a long standing problem that FreeBSD has serving UDP NFS client requests. You need to look at the packets. Is the IP on the response the same as the one on the request? The cache for pending rpc requests in the kernel that is used to handle responses does not recognize aliases being used by the response sender, if they are not the same as the one the request went out on. This is a particular problem with UDP, because it's connectionless. The fix is to open a seperate UDP per IP address, and respond on the UIDP socket associated with the IP, rather than a global UDP on INADDR_ANY. -- Terry To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
