nmbd fork-bombed my server

[Ignacio Coupeau]

Hello,

I have the same problem with linux RedHat (2.0.30, pentium).
The nmbd forks a lot of child and figure as active processes (about 1
per minute in apentium 166 with 45 NT4 ws in the domain):
----

root      9620  0.3  0.5  1324   744  ?  S   19:55   0:00
/usr/local/etc/samba/bin/nmbd -D -d 5 -s /usr/local/etc/samba/lib/
-----

I found in the log that the following lines repeat cyclically in a loop:
-----
[1998/09/16 20:06:49, 4] nmbd/nmbd_workgroupdb.c:dump_workgroups(308)
  dump_workgroups()
   dump workgroup on subnet  UNICAST_SUBNET: netmask=  159.237.12.42:
        CTI-SMB(1) current master browser = UNKNOWN
                BILBO 40099b0b (Samba 2.0.0-prealpha)
[1998/09/16 20:06:49, 4]
nmbd/nmbd_workgroupdb.c:find_workgroup_on_subnet(181)
  find_workgroup_on_subnet: workgroup search for CTI-SMB on subnet
UNICAST_SUBNET: found.
[1998/09/16 20:06:49, 4]
nmbd/nmbd_workgroupdb.c:find_workgroup_on_subnet(181)
  find_workgroup_on_subnet: workgroup search for CTI-SMB on subnet
UNICAST_SUBNET: found.
[1998/09/16 20:06:49, 4] nmbd/nmbd_workgroupdb.c:dump_workgroups(308)
  dump_workgroups()
   dump workgroup on subnet   159.237.12.42: netmask=  255.255.255.0:
        CTI-SMB(1) current master browser = BILBO
                BILBO 400c9b0b (Samba 2.0.0-prealpha)
                POLAR 40019a03 (Samba 1.9.19-prealpha)
[1998/09/16 20:06:49, 4] nmbd/nmbd_workgroupdb.c:dump_workgroups(308)
  dump_workgroups()
   dump workgroup on subnet  UNICAST_SUBNET: netmask=  159.237.12.42:
        CTI-SMB(1) current master browser = UNKNOWN
                BILBO 40099b0b (Samba 2.0.0-prealpha)
[1998/09/16 20:06:49, 4]
nmbd/nmbd_workgroupdb.c:find_workgroup_on_subnet(181)
  find_workgroup_on_subnet: workgroup search for CTI-SMB on subnet
UNICAST_SUBNET: found.

-----



Jeremy Allison wrote:
> 
> Mike Black wrote:
> >
> > I trace this down to the ASYNC_DNS.  Didn't test commenting it out but I'll
> > bet that will solve the problem.  I noticed a new nmbd running for every DNS
> > request and reported it to the bug list.
> >
> 
> Well Andrew also added code that causes nmbd to fork when
> writing out a WINS database and also when doing a browse
> synchronisation.
> 
> When nmbd fork-bombs you are these processes in the zombie
> state or are they still flagged as running ?
> 
> If they're zombies then we need to look at the
> CatchChild() code in lib/signal.c.
> 
> Cheers,
> 
>         Jeremy Allison,
>         Samba Team.
> 
> --
> --------------------------------------------------------
> Buying an operating system without source is like buying
> a self-assembly Space Shuttle with no instructions.
> --------------------------------------------------------

-- 
____________________________________________________
Ignacio Coupeau, Ph.D.     e-mail: icoupeau at unav.es
CTI, Director              fax:    +48 425619
University of Navarra      voice:  +48 425600
Pamplona, SPAIN            http://www.unav.es/cti/