Hi there!

First, a bit of history about how I discovered this bug: I spent the
more than one week in tests to track it down, so I do not want to simply
trash this work, sorry.

While investigating bugs #412989 [1] and #500998 [2], I found out that
enabling LDAP lookup for all the primary four entries in
/etc/nsswitch.conf [3] causes a system lookup:

- in 5.0.8/lenny, libnss-ldap/261-2.1, network-manager/0.6.6-3
    kinit: No resume image, doing normal boot...
    INIT: version 2.86 booting
    Starting the hotplug events dispatcher: udevd[no-blinking cursor]

  To exit this situation a hard reboot is needed.

- in 6.0.0/squeeze, libnss-ldap/264-2.2, network-manager/0.8.1-6
    Starting periodic command scheduler: cron.
    CPUFreq Utilities: Setting ondemand CPUFreq gover...disabled,
     governor not available...done.
    Starting MTA:[blinking cursor]

  Ctrl-Alt-Del works as well as single-user mode, which means that debug
  is possible ;-)

[1] <http://bugs.debian.org/412989>
[2] <http://bugs.debian.org/500998>
[3] while I am still an LDAP newbie, this the most advised setup you can
    find on the net, e.g. <http://wiki.debian.org/LDAP/NSS>

Please note that as suggested by /usr/share/doc/udev/README.Debian.gz,
setting '[UNAVAIL=return]' in any ldap service does not help.

On lenny, surprisingly enough, this bug does not happen if the "unknown"
groups have been replaced by nobody/nogroup (because of #412989 [1]).
On squeeze, however, these "unknown" groups are no more unknown [4], but
this bug (or some incarnation of it) is still there.  Squeeze users can
simply install libpam-ldapd and libnss-ldapd, given that even the
squeeze Release Notes [5] suggests that for other reasons [6][7].

[4] <http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=412989#66>
[5] <http://www.debian.org/releases/stable/i386/release-notes/ch-information.en.html#ldap-gnutls>
[6] <http://bugs.debian.org/566351>
[7] <http://bugs.debian.org/545414>

Good, we have two bugs, in some sort related.  Given that I thought this
was a libnss-ldap problem, I looked in the Debian BTS founding two bugs
which could be linked to mine, even if I do not have the same symptoms:

  #143496 [8], libnss-ldap: Segfault when used for host resolution
  #218958 [9], libnss-ldap - host resolution hangs on Linux 2.6.0-test8/test9

[8] <http://bugs.debian.org/143496>
[9] <http://bugs.debian.org/218958>

As I wrote above, the situation on squeeze is a bit better, I have
access to single-user mode so I can debug, starting from dbus...
Unfortunately, it seems quite hard to have a dbus-daemon output (I read
the docs), so I was submitting this bug asking for hints.  OTOH, if
libnss-ldap maintainer would have thought that it is not worth it
(libnss-ldapd is the future), this bug should have simply been left as
it is or even closed.

Trying to debug dbus-daemon in single-user mode, I discovered that there
is no more network.  Because of #530024 [10], my /etc/network/interfaces
(generated by d-i for a wired interface, so #606268 is no more only
related to wireless [11]) is:

forcemerge 500998 614804
thanks

Am 23.02.2011 16:09, schrieb Luca Capello:

Looks like a duplicate of 500998.

Ultimately I think this is a bug in libnss-ldap (and you might want to try
libnss-ldapd and libpam-ldapd).

Michael

unmerge 614804
notfound 614804 0.6.6-2
notfound 614804 0.6.6-3
found 614804 0.8.1-6
severity 614804 important
block 614804 by 606268
thanks

Hi Michael!

Reverted, for the reasons explained below.  And please wait for my email
to #500998 before any action.

No, it is a different one:

1) what triggers them is different
   #500998 is triggered by "group ldap"
   #614804 is triggered by "passwd+group+shadow+hosts ldap"

2) the distributions they happen in is different
   #500998 is only in lenny
   #614804 is only in squeeze, it manifests in lenny because of the
    users/groups referenced by udev not present by default, which is
    now fixed, see #412989

FYI, I was closing #500998, given that I found the solution (in the end
it is a network-manager problem), but I waited until this bug got its
number to reference it there.  Now I will wait until the two bugs are
unmerged to finally send the solution for #500998, which closes it.

Have you read the *full* report I submitted?

First, I already know about libnss-ldapd and libpam-ldapd:

Second, I found even a more important bug than this one, for which I
left any future action to the network-manager maintainers:
[...]
[...]

Thx, bye,
Gismo / Luca