2.0.1[15] hangs on Solaris 8 in test10-passwd.

[Nicholas Dronen <ndronen@frii.com>]

Hi,

I'm running 2.0.15 with gdbm.  I've seen what I believe is
the same problem running 2.0.11 as well.  Slapd hangs in
the following test:

>>>>> Starting test010-passwd ...
running defines.sh . ldbm
Cleaning up in ./test-db...
Starting slapd on TCP/IP port 9009...
Using ldapsearch to check that slapd is running...
Waiting 5 seconds for slapd to start...
Using ldapadd to populate the database...
Using ldapsearch to verify population ...
Using ldappasswd (PASS 1)  ...

Here's the truss output.

poll(0xFEE0B640, 3, -1)         (sleeping...)
signotifywait()                 (sleeping...)
door_return(0x00000000, 0, 0x00000000, 0) (sleeping...)
open("/dev/random", O_RDONLY)   (sleeping...)
lwp_cond_wait(0xFF0E55C8, 0xFF0E55D8, 0xFF083C48) (sleeping...)
lwp_cond_wait(0xFF0E55C8, 0xFF0E55D8, 0xFF083C48) Err#62 ETIME
time()                                          = 1001613374
poll(0xFEE0B640, 3, -1)         (sleeping...)
signotifywait()                 (sleeping...)
door_return(0x00000000, 0, 0x00000000, 0) (sleeping...)
open("/dev/random", O_RDONLY)   (sleeping...)
lwp_cond_wait(0xFF0E55C8, 0xFF0E55D8, 0xFF083C48) (sleeping...)

And the pstack output.

10285:  ../servers/slapd/slapd -f ./data/slapd-pw.conf -h
ldap://localhost:900
-----------------  lwp# 1 / thread# 4  --------------------
 ff19956c poll     (fee0b640, 3, ffffffff)
 ff14c798 select   (9, 0, ff1bb1e4, fee0bbf4, fee0bc74, fee0b640) + 2cc
 ff0cb470 select   (fee0b74e, ff0e4798, 0, 5, 1, fe401000) + 34
 ff0cbad0 _thread_start (0, 0, 0, 0, 0, 0) + 40
-----------------  lwp# 2 / thread# 2  --------------------
 ff19ad30 signotifywait ()
 ff0be8e0 _dynamiclwps (ff0de000, 5c, 0, 0, 0, 0) + 1c
 ff0c1b88 thr_yield (0, 0, 0, 0, 0, 0) + 8c
-----------------  lwp# 3  --------------------------------
 ff1988d4 door     (0, 0, 0, 0, ff0a5d18, 4)
 ff0ba474 _lwp_start (0, 0, 0, 0, 0, 0) + 18
-----------------  lwp# 4 / thread# 5  --------------------
 ff199448 open     (f010c, 0, 0)
 ff192714 _open    (f010c, 0, 0, ff1b8018, 0, 0) + 1c
 ff0caa98 open     (fed097a8, 4, 0, 105a72, 107d16, 7d) + 34
 0009ea0c hash_ssha1 (dc0e0, 114478, 109200, 7efefeff, 81010100, ff00) + 3c
 0009d75c lutil_passwd_hash (114478, 105a6c, ff1b8018, 0, ff1bbcf8, fed088f8) + 8c
 0005b6a8 slap_passwd_hash (114478, 4, efc14, 107cc8, 0, 0) + 70
 000951e4 ldbm_back_exop_passwd (128938, 1292e0, 114bd8, 110f28, 1144f8, fed09c10) + 314
 00089958 ldbm_back_extended (fed09c08, fed09c18, 94ed0, 110f28, 1144f8, fed09c10) + f0
 0005a8d4 passwd_extop (fed09c08, fed09c18, 89868, 1144f8, fed09c10, fed09c0c) + 2c4
 0005a0ec do_extended (fed09c18, 5a610, 10c478, ff1b8018, 0, 79fb0) + 764
 00024340 connection_operation (114328, 0, 0, 0, 0, 0) + 368
 000a69fc ldap_int_thread_pool_wrapper (10c470, ff093d18, 1, ff0eae04, 0, 2) + 19c
 ff0cbad0 _thread_start (10c470, 0, 0, 0, 0, 0) + 40
-----------------  lwp# 5  --------------------------------
 ff19b314 lwp_cond_wait (ff0e55c8, ff0e55d8, ff083c48)
 ff192cb0 _lwp_cond_timedwait (0, 3bb36a96, ff083cb0, ff0e55c8, ff0e55d8, 0) + 98
 ff0b8e24 _age     (ff0dedc0, ff0dedc4, ff0de000, 3, ff0de000, 1) + 94
 ff0ba474 _lwp_start (ff083d78, 0, 4000, ff00fc34, 0, 0) + 18
 ff0c1b88 thr_yield (0, 0, 0, 0, 0, 0) + 8c
--------------------------  thread# 1  --------------------
 ff0bd9a0 _reap_wait_cancel (ff0dee38, ff0de000, 80, ff0e59a8, fee0bd78, 1) + 40
 ff0bfc00 _thrp_join (ff0dee38, 4, 0, ff0de000, 0, 4) + 344
 000a6eec ldap_pvt_thread_join (4, 0, 1dbd8, 0, ffbef020, dc74d) + 1c
 0002067c slapd_daemon (0, 0, ff1bbe18, 109268, 0, 0) + 10c
 0001a81c main     (7, ffbeee3c, ffbeee5c, 104000, 0, 0) + bf4
 00019b94 _start   (0, 0, 0, 0, 0, 0) + dc
--------------------------  thread# 3  --------------------
 ff0bd948 _reap_wait (ff0e2a30, 20994, 0, ff0de000, 0, 0) + 38
 ff0bd6a0 _reaper  (ff0dee58, ff0e4798, ff0e2a30, ff0dee30, 1, fe400000) + 38
 ff0cbad0 _thread_start (0, 0, 0, 0, 0, 0) + 40

I don't know what's preventing slapd from continuing.  It could
be blocking while opening /dev/random, but it could also be blocking
in lwp_cond_wait (viz., pthread_cond_wait).  If the latter is the case,
I believe the problem is in libraries/libldap_r/tpool.c, either:

    369          if (pool->ltp_state == LDAP_INT_THREAD_POOL_RUNNING)
    370             ldap_pvt_thread_cond_wait(&pool->ltp_cond, &pool->ltp_mutex);

or

    379       (ctx->ltc_start_routine)(ctx->ltc_arg);

Has anyone seen this before?  Might it be a bug in the version
of libpthread.so on this Solaris machine?

Regards,

Nicholas Dronen: