Our x86 server (uptime of 117 days, running 1.3F -current from around
June) just started flaking on us.  /var/log/messages is full of
"fxp0: device timeout" and "nfsd send error 55" (No buffer
space available).  Besides rebooting, is there any other way to
fix this?  The machine is still up, so I might be able to do
some debugging.  Right now it is responsive, but occasionally
it stops responding for 10 minutes or more.

  systat mbufs shows a full line for data mbufs, with the number
"467" or thereabouts.  The sockets line is around 5.  nfsd has
accumulated 500 mins of CPU, but that might be reasonable for 4
months of heavy use.

  We have a Linux NFS file system mounted that won't go away (df
and umount hang, etc.), if that helps any.

  TIA!

  Brian