Re: Perl interface to syslog under Solaris

Balazs Scheidler <bazsi@xxxxxxxxxx> - Fri, Mar 05, 2004:

> Check the raw byte-stream as it is sent by the Perl module, I might help
> to recognize the culprit.

 I straced my standard syslogd on my Linux box and saw:
    recv(0, "<78>fm/logparse: F001|1078750293"..., 1022, 0) = 87
 on the unix socket /dev/log.

 You see "fm/logparse", my program's name, and "F001|1078...", the
 message.  I also captured some udp datagrams from the same Perl
 configured with 'udp' instead of unix, and got:
    <78>fm/logparse: F001|10785074...
 in the payload of the udp datagram.


 Then I've setup my syslogd to do udp broadcasting, and used "logger
 -p cron.info -t mytag tagada", captured with ethereal, and got:
    <78>mytag: tagada...
 in the udp datagram.

 The strace for a -p local7.info showed:
     recv(0, "<190>mar  8 14:02:44 mytag: taga"..., 1022, 0) = 34
     sendto(18, "<190>mytag: tagada\n", 19, 0, {sa_family=AF_INET,
         sin_port=htons(514),
         sin_addr=inet_addr("192.168.1.15")}, 16) = 19

 I attach the whole udp payload of a perl generated udp datagram as
 "perl-udp.raw".  I found no way to capture unix socket traffic, despite
 my attempts at using "socat", it seems it can't listen() to type 2
 sockets (datagram under Linux).  I also attach a syslogd generated
 datagram as "syslogd-udp.raw".

 AFAICT, it seems the beginning of an udp packet encodes the facility
 and priority with:
 - the '<' char;
 - the hexadecimal digits of the bits for facility and level of the
   message in ascii;
 - the '>' char.

 The only way I found to distinguish Perl generated datagrams from
 syslogd generated datagrams is the trailing zero.

 Do you think it is the cause of my troubles?  Where could I gather
 additional data?


   Regards,

-- 
Loïc Minier <lool@xxxxxxxx>

perl-udp.raw
Description: Text document

syslogd-udp.raw
Description: Text document