winbind (libnss_wins) calls openlog when it shouldn't
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
| samba (Ubuntu) |
Low
|
Unassigned |
Bug Description
Binary package hint: samba
I recently installed winbind and configured it in nssswitch.conf as follows:
hosts: files mdns4_minimal [NOTFOUND=return] wins dns mdns4
I had previously had fetchmail running, which regularly (due to an misconfigured external pop3 server) spat out errors like:
Oct 2 07:50:53 brooks fetchmail[6314]: awakened at Tue 02 Oct 2007 07:50:53 IST
Oct 2 07:50:53 brooks fetchmail[6314]: Server CommonName mismatch: localhost != mail.mounttemple.ie
Oct 2 07:50:53 brooks fetchmail[6314]: Server certificate verification error: self signed certificate
Oct 2 07:50:53 brooks fetchmail[6314]: Server certificate verification error: certificate has expired
Oct 2 07:50:58 brooks fetchmail[6314]: sleeping at Tue 02 Oct 2007 07:50:58 IST for 180 seconds
Now that nss_wins is configured, the error has changed to:
Oct 3 08:03:54 brooks nss_wins[6343]: awakened at Wed 03 Oct 2007 08:03:54 IST
Oct 3 08:03:57 brooks nss_wins[6343]: Server certificate verification error: self signed certificate
Oct 3 08:03:57 brooks nss_wins[6343]: Server certificate verification error: certificate has expired
Oct 3 08:04:02 brooks nss_wins[6343]: sleeping at Wed 03 Oct 2007 08:04:02 IST for 180 seconds
apparently because nss_wins is calling openlog, overwriting the fetchmail process's existing syslog entry with its own name and data.
gavinmc@brooks:~$ strings -a /lib/libnss_
gavinmc@brooks:~$ strings -a /lib/libnss_
openlog
This causes confusion as it appears nss_wins is the author of the error when in fact fetchmail is.
Gavin McCullagh (gmccullagh) wrote : | #1 |
Mathias Gug (mathiaz) wrote : | #2 |
It seems that there was a similar issue in pam_smbpass:
https:/
http://
To confirm the bug, a small testcase is needed.
Changed in samba: | |
importance: | Undecided → Medium |
status: | New → Incomplete |
Gavin McCullagh (gmccullagh) wrote : | #3 |
Do you mean I need to give you steps to repeat the problem?
I guess that would be:
1. Edit the hosts line in /etc/nsswitch.conf
hosts: files mdns4_minimal [NOTFOUND=return] wins dns mdns4
2. Set up samba, with /etc/samba/smb.conf including
wins support = yes
name resolve order = lmhosts host wins bcast dns
I can supply a full smb.conf if needed.
3. Set up fetchmail as a daemon, logging to syslog.
4. On receipt of mail or other output from fetchmail, the log will be from nss_wins instead of fetchmail.
I imagine you can substitute fetchmail with other things which do similar DNS lookups, mountd for example seems to work, although ssh doesn't seem to trigger it and I would imagine it does dns lookups. Perhaps ssh doesn't read nsswitch.conf.
Chuck Short (zulcss) wrote : | #4 |
Test case has been provided.
Changed in samba: | |
status: | Incomplete → Triaged |
Gavin McCullagh (gmccullagh) wrote : | #5 |
Does anyone know has this been fixed upstream?
Gavin McCullagh (gmccullagh) wrote : | #6 |
Has anyone any thoughts on this bug?
Was the test case repeatable?
Gavin
Thierry Carrez (ttx) wrote : | #7 |
I'm pretty sure this hasn't been fixed upstream yet, since they still call setup_logging() in nsswitch/wins.c and setup_logging still calls openlog().
If you confirm you can still experience the issue with Jaunty beta (3.3.2), I'll forward the bug to the https:/
Changed in samba: | |
importance: | Medium → Low |
status: | Triaged → Incomplete |
Gavin McCullagh (gmccullagh) wrote : | #8 |
The server I see it on regularly is running Hardy and is pretty mission-critical. I'll see if I can work up a test on a laptop or something.
Thanks,
Gavin
Thierry Carrez (ttx) wrote : | #9 |
@Gavin: any news ?
Gavin McCullagh (gmccullagh) wrote : | #10 |
I'd been having some trouble repeating it but I'll have a go again.
Sorry I've been a bit busy of late.
I also have this problem an can repeat it though.
I have set up samba the same way and also have the same values for the hosts: entry in /etc/nsswitch.conf
My syslog now shows the following:
Aug 5 13:39:44 gate nss_wins[3485]: awakened at Wed Aug 5 13:39:44 2009
Aug 5 13:44:47 gate nss_wins[3485]: 1 message for [...] at [...] (1861 octets).
Aug 5 13:44:48 gate nss_wins[3485]: reading message [...]:1 of 1 (1861 octets) flushed
Aug 5 13:39:45 gate nss_wins[3485]: sleeping at Wed Aug 5 13:39:45 2009 for 150 seconds
the process name fetchmail is replaced by nss_wins, resulting in the fact that the logcheck fetchmail rules do no longer filter out these messages from the logcheck generated mails.
If you need more information, don't hesitate to ask.
Chuck Short (zulcss) wrote : | #12 |
@Stefan,
Which version is this with?
REgards
chuck
Changed in samba (Ubuntu): | |
status: | Incomplete → Confirmed |
Hi,
I have the same problem in Lucid.
in "/etc/nsswitch.
===================
passwd: files ldap winbind
group: files ldap winbind
shadow: files ldap
hosts: files mdns4_minimal [NOTFOUND=return] dns mdns4 wins
networks: files dns
===================
my smb.conf (from testparm output)
=======
[global]
workgroup = MYDOMAIN
realm = MYDOMAIN.COM
server string = %h server (Servidor de Archivos - Samba %v, Ubuntu)
security = ADS
allow trusted domains = No
obey pam restrictions = Yes
password server = otherserver.
log level = 2
log file = /var/log/
max log size = 1000
name resolve order = host wins bcast
time server = Yes
unix extensions = No
printcap name = cups
dns proxy = No
wins server = 192.168.1.22
idmap uid = 10000-33554431
idmap gid = 10000-33554431
template shell = /bin/bash
winbind enum users = Yes
winbind enum groups = Yes
invalid users = "@MYDOMAIN\equipos del dominio"
admin users = admin_user
cups options = raw
wide links = Yes
delete readonly = Yes
dos filetime resolution = Yes
fake directory create times = Yes
=======
and in the syslog, I have many errors from different daemons (samba, snmpd and ntp), but with "nss_wins" ident, for example:
================
nss_wins[XXX]: call_nt_
nss_wins[XXX]: check_usershare
nss_wins[XXX]: Error writing 4 bytes to client. -1. (Transport endpoint is not connected)
nss_wins[XXX]: /etc/snmp/
nss_wins[XXX]: gbahamonde (192.168.1.188) couldn't find service abastecimiento
nss_wins[XXX]: getpeername failed. Error was Transport endpoint is not connected
nss_wins[XXX]: last message repeated 2 times
nss_wins[XXX]: last message repeated 3 times
nss_wins[XXX]: last message repeated 4 times
nss_wins[XXX]: libsmb/
nss_wins[XXX]: lib/util_
nss_wins[XXX]: lib/util_
nss_wins[XXX]: lib/util_
nss_wins[XXX]: nb-informatica (192.168.1.197) couldn't find service .
nss_wins[XXX]: net-snmp: 1 error(s) in config file(s)
nss_wins[XXX]: NET-SNMP version 5.4.2.1
nss_wins[XXX]: Packet send failed to 192.168.1.22(137) ERRNO=Network is unreachable
nss_wins[XXX]: param/loadparm.
nss_wins[XXX]: param/loadparm.
nss_wins[XXX]: param/loadparm.
nss_wins[XXX]: process_
nss_wins[XXX]: process_
nss_wins[XXX]: process_
nss_wins[XXX]: process_
nss_wins[XXX]: process_
nss_w...
Hello
Any news about this problem ??
Suse team publish a patch to resolv this problem
https:/
commentaries ??
bye
This has now started to affect some other programs which link against nss_wins:
Oct 19 10:57:55 brooks nss_wins[20797]: authenticated mount request from patella.mt:872 for /opt/ltsp/i386 (/opt/ltsp) kiosk/i386 (/opt/ltsp-kiosk)
Oct 19 10:58:44 brooks nss_wins[20797]: authenticated mount request from library5.mt:889 for /opt/ltsp-