Red Hat Bugzilla – Bug 107613
Hang in malloc_consolidate() function from /lib/tls/libc.so.6
Last modified: 2016-11-24 10:14:36 EST
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; pl-PL; rv:1.4) Gecko/20030630
Description of problem:
I've downloaded the sources, compiled and run the jabberd-2.0b2 Jabber server.
All works well, with one exception: the c2s process (client to server
connections handler) hangs when the jabberd 2 server shuts down.
Process doesn't react to the TERM signal and can only be killed with the KILL
It's 100% reproducible.
Debugging info I've collected:
1) When I attach to the running c2s process with strace, I can see that it
doesn't do any syscalls - it does nothing.
2) When I attach to the running c2s process with gdb and run backtrace, I can
see it has hung in the malloc_consolidate() function from glibc:
Reading symbols from /usr/lib/sasl/libplain.so...done.
Loaded symbols for /usr/lib/sasl/libplain.so
0x42074bf4 in malloc_consolidate () from /lib/tls/libc.so.6
#0 0x42074bf4 in malloc_consolidate () from /lib/tls/libc.so.6
#1 0x42074249 in _int_malloc () from /lib/tls/libc.so.6
#2 0x4207360b in malloc () from /lib/tls/libc.so.6
#3 0x4017632f in ber_memalloc () from /usr/lib/liblber.so.2
#4 0x40176467 in ber_memrealloc () from /usr/lib/liblber.so.2
#5 0x40174e3c in ber_realloc () from /usr/lib/liblber.so.2
#6 0x40174d3b in ber_write () from /usr/lib/liblber.so.2
#7 0x40173431 in ber_scanf () from /usr/lib/liblber.so.2
#8 0x401736fe in ber_scanf () from /usr/lib/liblber.so.2
#9 0x40174908 in ber_printf () from /usr/lib/liblber.so.2
#10 0x40155faa in ldap_send_unbind () from /usr/lib/libldap.so.2
#11 0x4015c973 in ldap_free_connection () from /usr/lib/libldap.so.2
#12 0x40155eaa in ldap_ld_free () from /usr/lib/libldap.so.2
#13 0x40155c2b in ldap_unbind_ext () from /usr/lib/libldap.so.2
#14 0x40155f0f in ldap_unbind_s () from /usr/lib/libldap.so.2
#15 0x0804d047 in _ldap_free (ar=0x80b5718) at authreg_ldap.c:178
#16 0x0804a88a in authreg_free (ar=0x80b5718) at authreg.c:192
#17 0x08053f43 in main (argc=4, argv=0xbfffe664) at main.c:576
#18 0x420156a4 in __libc_start_main () from /lib/tls/libc.so.6
3) When I issue the "next" command, I can see that c2s cannot execute any
further - completely nothing happens as if the process was waiting for a
function to complete:
Single stepping until exit from function malloc_consolidate,
which has no line number information.
... here it hangs.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Create an account "jabber" and perform all the following steps using that account
2. Download jabberd2, apply the patch given in the bug description (additional
3. compile jabberd2 (./configure --prefix=/home/jabber --enable-authreg="anon
pipe mysql ldap" --enable-debug --enable-developer &&
make && make check && make install && echo OK)
4. configure jabberd2 to authenticate in LDAP
5. run it on a terminal with debug on
6. connect as a Jabber user
7. press CTRL-C on the server's terminal to shut it down
Actual Results: A c2s process (jabberd's child) stays running, hangs in
malloc_consolidate() and doesn't react to TERM signal.
Expected Results: All jabberd processes terminate cleanly.
a) This might be related: to successfully compile jabberd2, I had to add
/usr/kerberos/include path to GCC's include paths because it is in a
diff -urN jabberd-2.0b2/configure.in jabberd-2.0b2.rh9/configure.in
--- jabberd-2.0b2/configure.in 2003-10-08 00:45:26.000000000 +0200
+++ jabberd-2.0b2.rh9/configure.in 2003-10-20 14:04:49.000000000 +0200
@@ -258,6 +258,7 @@
J2_LIB_OPENSSL(0, 9, 6, 2)
b) the Jabberd 2 server is configured to use LDAP authentication.
Hangs in malloc_consolidate etc. are in 99.9% of cases bugs in memory handling
of the application (or its shared libraries).
So I'd suggest you to use some memory allocation debugger (ElectricFence,
MALLOC_CHECK_=3 or valgrind).
Ping. It's almost a month. If we don't get a reply we must assume
this is no real problem.
Sorry, I cannot test it - that system is no more, plus I'm too busy
Agree to close this bug.
Cannot reproduce it.