I have upgraded from 6.2 to beta3. Leafnode was installed on my 6.2, and worked well. On upgrade beta3 didn't get the nntp-service from the old inetd.conf into the new xinet.d. I had to do it with inetdconvert. When trying to get in contact with my local nntpserver - nothing happens except that "syslogd" and "tcpd /usr/sbin/leafnode" consumes ca. 86% of my CPU! I don't get any error-messages, my computer is just working and working, and there is a lot of activity on my harddisc. And this is what "syslogd" is doing - 20 times every minute the last 4 hours - in /var/log/secure: (until I kill "tcpd /usr/sbin/leafnode" and restart syslogd) Jul 11 19:24:04 opuntia tcpd[2010]: warning: can't get client address: Invalid argument Jul 11 19:24:04 opuntia tcpd[2010]: connect from unknown Jul 11 19:24:04 opuntia tcpd[2010]: warning: can't get client address: Invalid argument Jul 11 19:24:04 opuntia tcpd[2010]: connect from unknown
The conversion problem is in the installer, which has since been fixed. Can you atttach the leafnode service file from /etc/xinetd.d? You don't need tcpd either... try without it, and see if the problem continues. Finally, if the problem continues, wait for the next beta and see if it disappears - I have a compiler bug suspected for this, which we have fixed since then (-foptimize-sibling-calls, part of -O2)(or try rebuilding xinetd and libwrappers on your computer yourself, with -O1) - please update this bug after the next release (still a problem or if it's fixed)
Created attachment 1087 [details] /etc/xinetd.d/nntp
Still a problem when updating the selected components from beta4? (BTW: no need to use tcpd, libwrapper support is compiled into xinetd)
Please test if the problem persists with beta 5
This defect is considered MUST-FIX for Winston Release-Candidate #1
The problem is the same in beta5. I have installed beta5 (not upgraded this time). I installed the same rpm of leafnode from redhat-contrib. Converted the nntp-service with inetdconvert. My /etc/xinetd.d/nntp file looks (still) like this: # Converted by inetdconvert service nntp { disable = no socket_type = stream protocol = tcp wait = yes user = news server = /usr/sbin/tcpd server_args = /usr/sbin/leafnode } To try "without tcpd", how? With server=/usr/sbin/leafnode and no server_args??
Yes, with server=/usr/sbin/leafnode.No server arguments. Can you try recompiling the leafnode source rpm (rpm --rebuild) and see if the problem persists? TIA.
I have now recompiled from SRPM: leafnode-1.9.14-1.src.rpm My /etc/xinetd.d/nntp looks like this: service nntp { socket_type = stream protocol = tcp wait = yes user = news server = /usr/sbin/leafnode disable = no } The result is a little bit different, but still problem. From /var/log/secure: Aug 1 19:55:18 opuntia xinetd[537]: START: nntp pid=919 from=<no address> Aug 1 19:55:24 opuntia xinetd[537]: START: nntp pid=920 from=<no address> Aug 1 19:55:29 opuntia xinetd[537]: START: nntp pid=921 from=<no address> Aug 1 19:55:34 opuntia xinetd[537]: START: nntp pid=922 from=<no address> again and again until the service stops, and /var/log/messages: Aug 1 19:55:18 opuntia xinetd[537]: warning: can't get client address: Invalid argument Aug 1 19:55:50 opuntia last message repeated 6 times Aug 1 19:56:52 opuntia last message repeated 12 times Aug 1 19:57:54 opuntia last message repeated 10 times Aug 1 19:58:55 opuntia last message repeated 142 times Aug 1 19:59:06 opuntia last message repeated 33 times Aug 1 19:59:06 opuntia xinetd[537]: nntp service was deactivated because of looping
OK - I can reproduce this. I'm looking into it now.
Ah... set wait=no. I'm making the convert script do that.