Red Hat Bugzilla – Bug 10363
Invalid socket causes syslog to spin
Last modified: 2014-03-16 22:13:09 EDT
Using the -a option on syslogd to specify a log socket in a directory that
does not exist will cause syslogd to drive CPU usage to 100% and refuse to
listen to log requests on legit sockets. This means that if you aren't
logged in as root, you will have to hit the reset button, because su,
sshd, etc. all want to write to syslog before they let you become root.
/sbin/syslogd -m 0 -a /tmp/dir_that_doesnt_exist/log
causes CPU usage to go to 100% and local syslog requests to lock up.
Fixed in sysklogd-1.3.31-17.
*** Bug 17188 has been marked as a duplicate of this bug. ***