[ntp:hackers] FreeBSD 6.3/6.4 hiccup with ntp-dev and LDAP

Dave Hart davehart at gmail.com
Tue Sep 8 18:03:20 UTC 2009

On Wed, Sep 2, 2009 at 6:16 AM, Dave Hart wrote:
> On Wed, Sep 2, 2009 at 5:41 AM, Dave Hart wrote:
>> Neither of these
>> machines is happy running stock ntp-dev code, crashing during the
>> processing of ntp.conf in a call to getaddrinfo().
> I apologize, I lied, one of those machines is running 6.3, the other
> 6.4.  6.3 is known to crash with 4.2.5p200 while 6.4 crashed with
> 4.2.5p179.
> hart at psp-fb1> uname -a
> FreeBSD psp-fb1.ntp.org 6.3-RELEASE-ISC FreeBSD 6.3-RELEASE-ISC #0:
> Fri Jan 18 08:04:43 UTC 2008
> root at ops.lab.isc.org:/usr/obj/usr/src/sys/GENERIC  i386
> hart at psp-fb1> ssh psp-fb2 uname -a
> FreeBSD psp-fb2.ntp.org 6.4-STABLE FreeBSD 6.4-STABLE #0: Thu Jul 30
> 00:19:10 UTC 2009     root at psp-fb2.ntp.org:/usr/obj/usr/src/sys/FB2
> i386

I've been able to work around these crashes by adding a superfluous
call to getaddrinfo() early in ntpd's execution.  This has kept me
happy with testing multiple revisions of my pending bug 1243
workaround, but as I wrap that work up, I need to pull that local
patch out, so once again I'll be forced to mess with reapplying that
local patch on each of the two developer testing machines running
FreeBSD 6.x.

Any insight or assistance debugging or help reproducing the problem
would be appreciated.  To recap briefly, on both 6.3 and 6.4, with
ldap mentioned by /etc/nsswitch.conf, ntp-dev ntpd crashes in the
first call to getaddrinfo() while attempting to
dlopen("nss_ldap.so.1").  However, if I ensure the first call occurs
much earlier (I've been putting one in init_logging()) the crash does
not occur, hinting at some sort of corruption.

Dave Hart

More information about the hackers mailing list