Re: OpenLDAP-2.0alpha2 bind problem

[Frédéric Poels <fpo@atos.be>]

At 15:45 26/08/99 -0700, you wrote:
>At 05:37 PM 8/26/99 +0200, Frédéric Poels wrote:
>>When I perform the following test on a 2.0-alpha2 slapd server, it crashes
>>(for those of you who know perl):
>
>running "ldapadd -D user -w passwd -p 9000 < /dev/null" (which
>does just a bind/unbind) a thousand times in a tight loop did not
>crash 2.0-alpha2 or -devel slapd on my test box system.

You're a lucky guy!
When I start my test, I can observe the following:

1) It starts fine

2) Most of the time, the server seems to loop (my test program waits for a
connection to terminate before initializing another one but the connection
never terminates and further connection attempts hangs). In that case,
slapd won't stop (kill -15 <slapd> fails, I have to kill -9 <slapd>). This
doesn't seem to happen with a decent trace level though (maybe because
traces slow down slapd) :o(

3) Sometimes I get:
do_unbind
ber_get_next
ber_get_next on fd 5 failed errno=0 (unknown)
        *** got 539137492 of 0 so far
connection_read(5): input error=-2 id=459, closing.
connection_closing: readying conn=459 sd=5 for close
connection_close: deferring conn=459 sd=5
connection_resched: attempting closing conn=459 sd=5
connection_close: conn=459 sd=5
daemon: removing 5
daemon: select: listen=4 active_threads=0 tvp=NULL
daemon: activity on 1 descriptors
daemon: new connection on 5
Assertion failed: c->c_struct_state != SLAP_C_USED, file  connection.c,
line 351
Abort (core dumped)

4) Let's have a look at that core file:
Program terminated with signal 6, IOT/Abort trap.
#0  0xd035dc6c in pthread_kill ()
Backtrace:
#7  0xd0353f40 in _pthread_body ()
#6  0x1000f5b8 in slapd_daemon_task ()
#5  0x10034a70 in connection_init ()
#4  0xd0272df4 in __assert ()
#3  0xd025cc74 in abort ()
#2  0xd035d8a8 in raise ()
#1  0xffffffff in ?? ()
#0  0xd035dc6c in pthread_kill ()


>Kurt

Fred.