old smbd's block new connections

[Brad Greer]

Hi,

  We recenly upgraded to verion 1.9.18p4 on Digital Unix 4.0b (I believe our
previous smb version was 1.9.16).   Occasionally we have reports of one of
our users having their w95 logon script freeze at the point where it tries
to mount their samba drives.   I have found I can recreate this problem by:

   - login to W95.  mount a drive from the samba server
   - turn of the W95 pc.  smbd continues to exists because it doesn't know
the
pc has gone off the network.
   - reboot the W95 pc.  Try to mount the same share as before.  W95 system
'hangs'.  The old smbd is still on the server, no new smbd is started.

  If I try to mount a different share from the pc, that causes a new smbd to
start up and the mount succeeds (the old smbd is still there).   If I enable
SO_KEEPALIVE, the tcp stack will terminate the old smbd after about 2 hours.
I have turned down the tcp_keepintvl in the digital unix kernel so the
'unconnected' smbd's are now discovered after about 10 minutes, but this
doesn't really fix the underlying problem.   Shouldn't the 'old' smbd be
killed/reset when the same pc tries to create a new connection to an
existing smbd share ?  Did I miss a smb.conf parameter setting ?

-- brad