smbd cpu spin & Failed to marshall NET_R_SAM_LOGON

[Peter Polkinghorne]

Apologies if this has already been dealt with, but I am over 3 weeks behind on 
Samba mailing list reading - but desperately skimming to catch up.

I have just upgraded to 2.2.2 from 2.2.1 in order to combat some runaway smbd 
daemons.  This upgrade does not seemed to have helped.  This is a Solaris 2.6 
environment. Every time the smbd goes into CPU spin, I notice the following 
things:

a) truss -p on the process report nothing happening ie looks like internal loop 
not stuck on some system resource).

b) there is a report that complains about sam_logon everytime I have checked 
loke the following (in log.smbd):

[2001/11/05 08:36:26, 0, pid=16184, effective(60001, 60001), real(0, 0)] 
rpc_server/srv_netlog.c:api_net_sam_logon(208)
  api_net_sam_logon: Failed to marshall NET_R_SAM_LOGON.
[2001/11/05 08:36:26, 0, pid=16184, effective(60001, 60001), real(0, 0)] 
rpc_server/srv_pipe.c:api_rpcTNP(1204)
  api_rpcTNP: api_netlog_rpc: NET_SAMLOGON failed.

... seen messages about this in the Samba mailing lists, but no answers.

c) Apart from one user (not the same each time), everyone else seems OK.


The other thing that may be unusual about us, is we are using W2K terminal 
servers - thus 1 Samba daemon is support multiple users with multiple shares 
open.  I have turned up the logging to 3 to try capturing more info.

Current workaround is to kill the offending process and on the whole the users 
seem not to complain!

-- 
Peter Polkinghorne, IT Manager		Kingsford Stacey Blackwell
ppolkinghorne at kingsfords.co.uk		14 Old Square, Lincoln's Inn
(44) 020 7447 1200			London WC2A 3UB