Hi,
Is it normal for nmbd to react to its own broadcasts? This is what seems
to be happening to me.
I'm just installed SAMBA 2.2.8a on a Gentoo box on my home network; a
3-node network using NetBIOS over TCP/IP--all B-nodes. Looking at the
logs I saw lines such as the following each time I started SAMBA.
[2004/01/07 16:47:07, 0]
nmbd/nmbd_responserecordsdb.c:find_response_record(235)
find_response_record: response packet id 32385 received with no
matching record.
[2004/01/07 16:47:07, 0]
nmbd/nmbd_responserecordsdb.c:find_response_record(235)
find_response_record: response packet id 32386 received with no
matching record.
Sniffing the network I found that each packet id referred to a series of
4 NBNS Registration queries--BUT NO REGISTRATION RESPONSES! The time of
the log entry was the time of the last of the series of packets. But the
2nd and 3rd packets arrived only 0.0005 s apart. Although all 5 names I
was registering had similar series of packets and timings, I would only
receive log messages for 2 of them.
I suspected that nmbd was reacting to a packet it originated itself. So
I set up iptables to drop all incoming trafic from my own ip address.
When I started SAMBA, I reveived hits on my new IP filter, the SAMBA log
messages stopped, and the delays between the registration packets for
each name became 2, 1, and 1 s. Problem solved!
But is this normal behavior? Should I be receiving my own UDP broacasts?
Is nmbd's reacting to them a bug? Is there a better way to avoid these
errors? Is this an artifact of my network (somewhat ancient: 350MHz
Pentium II server, one 10Mbps hub, but less than 10 m span)?
Sincerely,
Rick
