Timo Sirainen wrote:
On Sun, 2007-09-30 at 16:01 +0200, Bernd Kuhls wrote:
unless you can provide another way to restrict access to IMAP and POP3 to our local IP range this is not an option for us.
iptables?
Until now no new crash in the logs. I will try to reproduce the bug.
I suppose this fixes it: http://hg.dovecot.org/dovecot-1.0/rev/8d6433652a33
Hi,
first the good news, since using Dovecot including the patch mentioned above there were no new crashes this week, well done!
/var/log/dovecot# grep Panic * dovecot.main:dovecot: 2007-09-25 16:54:08 Panic: Lost login process PID 420 dovecot.main:dovecot: 2007-09-27 12:13:02 Panic: Lost login process PID 5160 dovecot.main:dovecot: 2007-09-30 15:26:34 Panic: Lost login process PID 9091
But now my syslog is filled with this once every 10 minutes:
Oct 6 14:11:42 server inetd[16124]: pop3s/tcp: bind: Address already in use Oct 6 14:11:42 server inetd[16124]: pop3/tcp: bind: Address already in use Oct 6 14:11:42 server inetd[16124]: imaps/tcp: bind: Address already in use Oct 6 14:11:42 server inetd[16124]: imap/tcp: bind: Address already in use
But Dovecots behaviour has changed somehow. When updating dovecot.conf all I had to do to activate the new config was to kill the master process. Client-side connections let inetd restart dovecot, this is not the case anymore. Now I have to start /usr/sbin/dovecot on the shell to reactivate the IMAP/POP service. Debians /etc/init.d/dovecot has to be changed to reflect your patch, right? Currently it contains
# The init script should do nothing if dovecot or another imap/pop3 server # is being run from inetd, and dovecot is configured to run as an imap or # pop3 service for p in
sed -r "s/^ *(([^:]+|\[[^]]+]|\*):)?(pop3s?|imaps?)[ \t].*/\3/;t;d" \ /etc/inetd.confdo for q insed -r "s/^[ \t]*protocols[ \t]*=[ \t]*(([^\"]*)|\"(.*)\")/\2\3/;t;d" \ ${CONF}do if [ $p = $q ]; then exit 0 fi done done
Greetings, spiralvoice