Burton M. Strauss III wrote:
i've not figured it was a thread about openbsd, so i just discover it. after few test with openbsd (3.5), i can only confirm many problems. seems compiling was only a small step, much stay to do (with threads at least)


Could well be that OpenBSD has the same userland threads as FreeBSD.

well, it would be strange as FreeBSD use thread in kernel.

The only work-around we have for that is the -set-pcap-nonblock option (converts the select() operation to a poll). That is dependent upon a version of libpcap which offers the call - so you may have a dependency with the very latest (0.8.x has it, 0.7.2 MAY -
I don't remember).
well, there is some lag here ... as 3.5 has lipbcap 0.5 (last import
from openbsd is 2.7; libpcap is in base system, not ports)
i ask on misc@ why.

after the test discussed at the end of the mail, i try to compile with
an extra libpcap (0.8.3) in /opt (setnonblock was detected by configure)
but where libpcap is binded ?
# ldd /export/tmp/ports-obj/ntop3-3.0/fake-i386/usr/local/bin/ntop |grep
pcap
#
-> test
ntop doesn't respond more

-> after a reboot
# ntop -i xl0 -t 5 -M -m 192.168.2.0/24 --ipv4 --skip-version-check -u _ntop
ntop doesn't respond
# lsof|grep ntop|grep IPv
ntop      21834    _ntop   11u  IPv4 0xd0fbb2e8      0t0      TCP *:3000
(LISTEN)

-> with -set-pcap-nonblock, doesn't seem to help more

-> same as -u root
respond but very slowly (telnet + firefox)
after some time, seems it runs normally

Report created on Mon Apr 12 19:57:09 2004 [ntop uptime: 24:19]
Generated by ntop v.3.0 SourceForge .tgz MT (SSL) [i386-unknown-openbsd3.5]
Build: Apr 12 2004 19:06:39. Listening on [xl0] without a kernel
(libpcap) filtering expression
Web report active on interface xl0

-> retry as _ntop
same here after ~15 sec, normal

(note also than it seems you have to way before rebooting ntop and the
port is free again)
-> another try with ssl
# ntop -i xl0 -t 5 -M -m 192.168.2.0/24 --ipv4 --skip-version-check -u
_ntop -w 0 -W 3000

Mon Apr 12 20:02:14 2004 [MSGID0986275] WEB: ntop's web server is now
processing requests
Mon Apr 12 20:02:14 2004 [MSGID0393976] THREADMGMT: pcap dispatch thread
running...
Mon Apr 12 20:02:16 2004 [MSGID0587829] **ERROR**
SSL(ssl_init_connection)ERROR [Thread 23462]: error:1407609C:SSL
routines:SSL23_GET_CLIENT_HELLO:http request at
/usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:02:16 2004 [MSGID0587829] **ERROR** SSL(read)ERROR [Thread
23462]: error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
at /usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:02:16 2004 [MSGID8807438] SECURITY: Loading items table
Mon Apr 12 20:02:43 2004 [MSGID0587829] **ERROR**
SSL(ssl_init_connection)ERROR [Thread 23462]: error:1407609C:SSL
routines:SSL23_GET_CLIENT_HELLO:http request at
/usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:02:43 2004 [MSGID0587829] **ERROR** SSL(read)ERROR [Thread
23462]: error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
at /usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:02:43 2004 [MSGID0587829] **ERROR**
SSL(ssl_init_connection)ERROR [Thread 23462]: error:1407609C:SSL
routines:SSL23_GET_CLIENT_HELLO:http request at
/usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:02:43 2004 [MSGID0587829] **ERROR** SSL(read)ERROR [Thread
23462]: error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
at /usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:03:11 2004 [MSGID0587829] **ERROR**
SSL(ssl_init_connection)ERROR [Thread 23462]: error:1407609C:SSL
routines:SSL23_GET_CLIENT_HELLO:http request at
/usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:03:11 2004 [MSGID0587829] **ERROR** SSL(read)ERROR [Thread
23462]: error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
at /usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:03:54 2004 [MSGID0587829] **ERROR**
SSL(ssl_init_connection)ERROR [Thread 23462]: error:1407609C:SSL
routines:SSL23_GET_CLIENT_HELLO:http request at
/usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)
Mon Apr 12 20:03:54 2004 [MSGID0587829] **ERROR** SSL(read)ERROR [Thread
23462]: error:1407609C:SSL routines:SSL23_GET_CLIENT_HELLO:http request
at /usr/src/lib/libssl/ssl/../src/ssl/s23_srvr.c(400)

# lsof|grep ntop|grep IPv
ntop      23462    _ntop   11u  IPv4 0xd0fbb180       0t0      TCP
*:3000 (LISTEN)

telnet connects but returns nothing
$ telnet localhost 3000
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
GET / HTTP/1.0
Connection closed by foreign host.

-> retry without ssl
slow response (telnet+firefox)
after 7 mins, it becomes normal
with some zombies
# ps ax|grep ntop|grep ZW|wc -l
      11
and big cpu load

very strange behavior

two problems: * Sun Apr 11 17:11:05 2004 [MSGID8837990] Loading plugin '/usr/local/lib/ntop/plugins/icmpPlugin.so' Sun Apr 11 17:11:05 2004 [MSGID0592890] **WARNING** Unable to locate plugin '/usr/local/lib/ntop/plugins/icmpPlugin.so' entry function [Unable to resolve symbol] + double-check perms: ok $ file /usr/local/lib/ntop/plugins/icmpPlugin.so /usr/local/lib/ntop/plugins/icmpPlugin.so: ELF 32-bit LSB shared object, Intel 80386, version 1, not stripped what is the "symbol problem" ?


Not a clue - it's an OpenBSD message, just passed through ntop. I did post a smidge from a google search yesterday - might be relevant.



seems not a ldconfig issue, so how other thing are needed/loaded ? i've
made a turn with ldd/objdump/nm and find nothing particular

* web server in all case, lsof report socket as LISTEN but telnet return conn refused with gdb, some responses for others (but very/slow/nothing under normal browser)


Yup - that's the select()/bpf problem.

this mealage seems to vary as after sending my last mail, i have ntop turning for 3 hours without a hitch as daemon or not (with libpcap 0.5) (ntop -i xl0 -t 5 -M -m 192.168.2.0/24 --ipv4 --skip-version-check -u _ntop -d)

Dropped (libpcap)       1.5%    14,114
Dropped (ntop)  0.0%    0
Total Received (ntop)   924,539
Total Packets Processed 924,539

Report created on Mon Apr 12 14:48:46 2004 [ntop uptime: 3:36:15]
Generated by ntop v.3.0 SourceForge .tgz MT (SSL) [i386-unknown-openbsd3.5]
Build: Apr 11 2004 19:22:43. Listening on [xl0] without a kernel
(libpcap) filtering expression

but next when i try to play with -w/-W, i return on the problem ...
don't know if it's a randow problem or a cleaning problem, or whatever else

Sounds like there may be more than a few missed tags for pre1. No real issue, IMHO, since I don't WANT people going back to a pre release w/o a really really good reason. In your case, the .tgz is available @ SourceForge (the full archives are available at http://osdn.dl.sourceforge.net/sourceforge/ntop - or pick any of the mirrors instead of osdn). So 3.0pre1 is at http://osdn.dl.sourceforge.net/sourceforge/ntop/ntop-3.0pre1.tgz.

ok, i've only check ntop sf project page where there is only 3.0.

else on the on the old zombie problem, could this be here ?
http.c:19930 => wait() for parent ?
when ntop was working, i quickly finish to sysctl limit of procs
syslog:
Apr 12 13:50:03 etenemanki /bsd: proc: table is full

$ ps ax|grep ntop|grep ZW| wc -l
        413

* i joined last port revision

so for now, at least, it doesn't seem the problem comes from libpcap but
problem stays mostly unidentified and random ...

i will probably not have much time during week, so no big advance before
next week-end.

Regards

Julien

note: request option for configure: --user/--group (so we could define
the default user without extra options on run)(on openbsd policy is port
user begin with '_')





Attachment: ntop3-ob.tar.gz
Description: application/gzip

Reply via email to