Red Hat Bugzilla – Bug 432408
Bug in Replication aggrenment connections (SSL)
Last modified: 2015-01-04 18:30:42 EST
Description of problem:
There is a strange behavior in the replications initialization in at least 2 of our servers (by now).
The SEGFAULT happens randomly. This reproduction belongs to a server with FDS 1.1-CVS at the HEAD
revision (at the date of this post) compiled with debug symbols.
Starting program: /usr/sbin/ns-slapd -D /etc/dirsrv/slapd-grs
[Thread debugging using libthread_db enabled]
[New Thread -1208707392 (LWP 25645)]
warning: Lowest section in /usr/lib/libicudata.so.36 is .gnu.hash at 051120b4
[New LWP 25682]
[New LWP 25685]
Program received signal SIGSEGV, Segmentation fault.
[Switching to LWP 25685]
0x01af2429 in toupper_w (val=83) at lib/util_unistr.c:422
422 lib/util_unistr.c: No existe el fichero o el directorio.
#0 0x01af2429 in toupper_w (val=83) at lib/util_unistr.c:422
#1 0x01af30f6 in toupper_ascii (c=83) at lib/util_unistr.c:1062
#2 0x01af1041 in StrCaseCmp (s=0x1b32861 "SO_KEEPALIVE", t=0xafffc634 "SO_BROADCAST") at
#3 0x01af120f in strequal (s1=0x1b3b1ec "??\v", s2=0x0) at lib/util_str.c:290
#4 0x01afc928 in set_socket_options (fd=163, options=0x1b19e5d "") at lib/util_sock.c:233
#5 0x01a976a7 in _nss_wins_gethostbyname_r (hostname=0xae316b8
"xxxxxxxxxxxxxxx.yyyyy.zzzzz.tt", he=0xafffcc38, buffer=0xafffc900 "\177", buflen=512,
h_errnop=0xafffcc60) at nsswitch/wins.c:74
#6 0x01a9792c in _nss_wins_gethostbyname2_r (name=0xae316b8
"xxxxxxxxxxxxxxx.yyyyy.zzzzz.tt", af=2, he=0xafffcc38, buffer=0xafffc900 "\177", buflen=512,
h_errnop=0xafffcc60) at nsswitch/wins.c:380
#7 0x003377cd in gaih_inet () from /lib/libc.so.6
#8 0x00338860 in getaddrinfo () from /lib/libc.so.6
#9 0x001e5e00 in PR_GetAddrInfoByName () from /usr/lib/libnspr4.so
#10 0x0027ed9e in prldap_socket_arg_alloc () from /usr/lib/libprldap60.so
#11 0x00defcf4 in ldap_start_tls_s () from /usr/lib/libssldap60.so
#12 0x00249dd7 in nsldapi_connect_to_host () from /usr/lib/libldap60.so
#13 0x0024d7a7 in nsldapi_new_connection () from /usr/lib/libldap60.so
#14 0x002482b5 in nsldapi_open_ldap_defconn () from /usr/lib/libldap60.so
#15 0x0024e78b in nsldapi_send_server_request () from /usr/lib/libldap60.so
#16 0x0024eb8e in nsldapi_send_initial_request () from /usr/lib/libldap60.so
#17 0x00253af1 in ldap_simple_bind () from /usr/lib/libldap60.so
#18 0x06ed1a61 in do_simple_bind (conn=0xb0202958, ld=0x0, binddn=0x1b3b1ec "??\v",
#19 0x06ed3f6b in bind_and_check_pwp (conn=0xb0202958, binddn=0xae2ecd0 "cn=Replication
Manager,cn=config", password=0xae2db28 "xxxxxxxxxxxx")
#20 0x06ed4311 in conn_connect (conn=0xb0202958) at
#21 0x06edc386 in acquire_replica (prp=0xb0202a78, prot_oid=0x6f05057
#22 0x06ed5b67 in repl5_inc_run (prp=0xb0202a78) at
#23 0x06edb881 in prot_thread_main (arg=0xb02028d0) at
#24 0x001f319d in PR_JoinThread () from /usr/lib/libnspr4.so
#25 0x003c845b in start_thread () from /lib/libpthread.so.0
#26 0x0035124e in clone () from /lib/libc.so.6
78 Replication agreements. Most of them against non responding servers yet.
Are you using pam_ldap and nss_ldap? Can you post your /etc/nsswitch.conf?
Not in the server but clients.
nss_ldap-253-5.el5 (Centos 5.1)
passwd: files ldap
shadow: files ldap
group: files ldap
(No more LDAP related configurations)
/etc/ldap.conf: (libnss issues)
Samba Services are also LDAP-enabled.
I'm sorry so i didn't understand well that you asked me for my
/etc/nsswitch.conf cause of the wins hostname resolution. Initially i thought
that the problem was caused by a string overwritted or a buffer overflow in FDS,
but i was unable to trace it.
We had in /etc/nsswitch.conf:
hosts: files dns wins
Now, removing wins issue:
hosts: files dns
And Server seems to start ok all the times.
That's odd. I wonder what the problem is with using wins in the directory server?
I notice on my RHEL5 system:
rpm -q --whatprovides /lib/libnss_wins.so.2
I haven't heard of any problems with wins/samba code in the same process space
as directory server code. I know there is a problem with pam_ldap/nss_ldap
because it uses openldap client libraries, but I'm not aware of any problems
wins resolution brokes our systems.
O our system:
# rpm -q --whatprovides /lib/libnss_wins.so.2
Steps to reproduce (for help in other similar cases):
* Enable wins for host name resolution: /etc/nsswitch.conf. ex:
hosts: files dns wins
* Set up a replication agreenment against a non DNS nor wins resoluble hostname.
* Start/Stop dirsrv several times, try to initiallize the consummer or just wait.
It would be thankful that you recommended to us what to do exactly with this kind of non directly
related to dirsrv. Maybe we should post them to the fds-devel list?
(In reply to comment #5)
> That is.
> wins resolution brokes our systems.
> O our system:
> # rpm -q --whatprovides /lib/libnss_wins.so.2
> Steps to reproduce (for help in other similar cases):
> * Enable wins for host name resolution: /etc/nsswitch.conf. ex:
> hosts: files dns wins
> * Set up a replication agreenment against a non DNS nor wins resoluble hostname.
> * Start/Stop dirsrv several times, try to initiallize the consummer or just wait.
> It would be thankful that you recommended to us what to do exactly with this
kind of non directly
> related to dirsrv. Maybe we should post them to the fds-devel list?
Probably just the fedora-directory-users list to let people know there is a
problem mixing wins with fedora ds.
I don't think there is much we can do about this bug in the directory server itself.