deamonize.c and the option DSYSLOG_ASYNC Kamailio 1.5
Hi all, I have a problem with the DSYSLOG_ASYNC option in the Makefile.defs If I have set this option I could not longer receive log messages in the "normal" running mode (forking, debug level=2 stderr = no). If I set the debug to 9 I receive logging messages, then via $kamctl debug 2 is everything fine. If the debug is set to 2 and I run the proxy in foreground I also receive log messages. Any way, I have tried to find out what is going wrong here and found something in : daemonize.c At the end of the daemonize function is a for loop int daemonize(int * own_pgid) . . /* close any open file descriptors 32 is the maximum number of inherited open file descriptors */ for (r=3; r < 32; r++){ close(r); } . . The variable "r" is used for the return value of fprintf (Upon successful return, these functions return the number of characters printed ..). At the moment I have no idea why this triggering this error and only for the DSYSLOG_ASYNC option. Furthermore why or for what is this for-loop good ? Maybe I am totally wrong or off the track but I really want this async logging. best regards Torben
Hi all, OK sorry, the variable "r" is re-set in the for-loop, my fault(better think before write ;-) ). But the loop is the problem in my opinion. If I put a LM_* log direct in fron of the close(line208 in daemonize.c), everything seems fine. Which means the close command fails (maybe at the right fd). regards Torben Am Freitag, den 20.11.2009, 12:14 +0100 schrieb Torben Friese:
Hi all,
I have a problem with the DSYSLOG_ASYNC option in the Makefile.defs
If I have set this option I could not longer receive log messages in the "normal" running mode (forking, debug level=2 stderr = no).
If I set the debug to 9 I receive logging messages, then via $kamctl debug 2 is everything fine.
If the debug is set to 2 and I run the proxy in foreground I also receive log messages.
Any way, I have tried to find out what is going wrong here and found something in :
daemonize.c
At the end of the daemonize function is a for loop
int daemonize(int * own_pgid) . . /* close any open file descriptors 32 is the maximum number of inherited open file descriptors */ for (r=3; r < 32; r++){ close(r); } . .
The variable "r" is used for the return value of fprintf (Upon successful return, these functions return the number of characters printed ..).
At the moment I have no idea why this triggering this error and only for the DSYSLOG_ASYNC option.
Furthermore why or for what is this for-loop good ?
Maybe I am totally wrong or off the track but I really want this async logging.
best regards
Torben
_______________________________________________ sr-dev mailing list sr-dev@lists.sip-router.org http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev
On Freitag, 20. November 2009, Torben Friese wrote:
OK sorry, the variable "r" is re-set in the for-loop, my fault(better think before write ;-) ).
But the loop is the problem in my opinion.
If I put a LM_* log direct in fron of the close(line208 in daemonize.c), everything seems fine. Which means the close command fails (maybe at the right fd).
Hi Torben, did not dig in the code in question so far, but is this still the same problem you reported on the user list, that it work on etch but not on lenny? Regards, Henning
Hi Henning, I have found out that this NOT a lenny problem. It is general. Have you seen the other post : #### If I set the debug(in cfg) to 4 the proxy write logging messages, then via $kamctl debug 2 is everything fine(still logging). If the debug is set to 2 and I run the proxy in foreground I also receive log messages. ##### Another interesting thing is if you bring in a log at line 205 (daemonize.c) it is working: LM_ERR("what ever \n"); for (r=3; r < 32; r++){ close(r); } It seems that close() is not able to close the fd for the log (return value from mksock in syslog_async.c). But If I let the for-loop counter start at 4 it is still not working. After that I tried to set up log_fd as global( int log_fd=-1) and include the headers, to check the "r" against the log_fd before execute close(). But the same result. Also a correct close(closelog_async()) in deamonize do not work. I am open for any ideas, maybe I have oversight something. Could you reproduce this ? best regards Torben Am Donnerstag, den 26.11.2009, 12:40 +0100 schrieb Henning Westerholt:
On Freitag, 20. November 2009, Torben Friese wrote:
OK sorry, the variable "r" is re-set in the for-loop, my fault(better think before write ;-) ).
But the loop is the problem in my opinion.
If I put a LM_* log direct in fron of the close(line208 in daemonize.c), everything seems fine. Which means the close command fails (maybe at the right fd).
Hi Torben,
did not dig in the code in question so far, but is this still the same problem you reported on the user list, that it work on etch but not on lenny?
Regards,
Henning
On Donnerstag, 26. November 2009, Torben Friese wrote:
I have found out that this NOT a lenny problem. It is general.
Hello Torben, ok, understand.
Have you seen the other post :
####
If I set the debug(in cfg) to 4 the proxy write logging messages, then via $kamctl debug 2 is everything fine(still logging).
This sounds fine.
If the debug is set to 2 and I run the proxy in foreground I also receive log messages.
You don't expect to receive log messages here? Do you've set fork=no, or log_stderr=no?
#####
Another interesting thing is if you bring in a log at line 205 (daemonize.c) it is working:
LM_ERR("what ever \n"); for (r=3; r < 32; r++){ close(r); }
It seems that close() is not able to close the fd for the log (return value from mksock in syslog_async.c).
But If I let the for-loop counter start at 4 it is still not working.
After that I tried to set up log_fd as global( int log_fd=-1) and include the headers, to check the "r" against the log_fd before execute close(). But the same result.
Also a correct close(closelog_async()) in deamonize do not work.
I am open for any ideas, maybe I have oversight something. Could you reproduce this ?
I did not found time to setup a test scenario today, should be able to do so on monday. Regards, Henning
On Donnerstag, 26. November 2009, Torben Friese wrote:
[..] Another interesting thing is if you bring in a log at line 205 (daemonize.c) it is working:
LM_ERR("what ever \n"); for (r=3; r < 32; r++){ close(r); }
It seems that close() is not able to close the fd for the log (return value from mksock in syslog_async.c).
But If I let the for-loop counter start at 4 it is still not working.
After that I tried to set up log_fd as global( int log_fd=-1) and include the headers, to check the "r" against the log_fd before execute close(). But the same result.
Also a correct close(closelog_async()) in deamonize do not work.
I am open for any ideas, maybe I have oversight something. Could you reproduce this ?
Hello Torben, i've just was able to reproduce this on 1.5 branch. henning@ca:~/projects/openser/branches/1.5$ ./kamailio -V version: kamailio 1.5.3-notls (i386/linux) flags: STATISTICS, USE_IPV6, USE_TCP, DISABLE_NAGLE, USE_MCAST, SHM_MMAP, PKG_MALLOC, F_MALLOC, FAST_LOCK-ADAPTIVE_WAIT ADAPTIVE_WAIT_LOOPS=1024, MAX_RECV_BUFFER_SIZE 262144, MAX_LISTEN 16, MAX_URI_SIZE 1024, BUF_SIZE 65535, PKG_SIZE 4194304 poll method support: poll, epoll_lt, epoll_et, sigio_rt, select. svnrevision: 2:5958M @(#) $Id: main.c 5608 2009-02-13 16:48:17Z henningw $ main.c compiled on 00:29:27 Dec 1 2009 with gcc 4.3.3 Here i get log file output, regardsless if i use log_stderror or not, but only with debug=3. Dec 1 00:33:47 [7985] INFO:core:init_tcp: using epoll_lt as the TCP io watch method (auto detected) Dec 1 00:33:47 [7987] ERROR:core:main: install asynchronous syslog handler.. Dec 1 00:33:47 [7987] NOTICE:core:main: version: kamailio 1.5.3-notls (i386/linux) The second log msg was added from me, to see if the correct function is called. This is on a kubuntu 9.04, with version 1.5-5ubuntu3 of sysklogd. When i use debug=2 in the cfg, aparently it not works.. Interesting, i'll investigate more. Regards, Henning
participants (2)
-
Henning Westerholt -
Torben Friese