From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.2.1) Gecko/20010901 Description of problem: I have a couple of machines running squid under fairly heavy loads, which are crashing every day or so (system freezes completely - requiring a power cycle). The systems have 2 1Ghz processors, running 2.4.9-13enterprise using iptables. I have now obtained a full stack trace from a serial console, and attach the output from ksymoops. This seems to point to a problem in ip_conntrack. Version-Release number of selected component (if applicable): 2.4.9-13enterprise How reproducible: Always Steps to Reproduce: 1.run squid under a heavy load 2.wait Additional info:
Created attachment 43034 [details] Output from ksymoops
The problem is still in 2.4.9-21enterprise, though the crash was slightly different this time (see new attachment). If I change the firewall so it doesn't use connection tracking (and make sure ip_conntrack and dependent modules aren't loaded) then 2.4.9-13enterprise seems to be stable. I also tried a mainline kernel 2.4.18-pre6 (based on the enterprise config from 2.4.9-13) and this seems to be stable with connection tracking active.
Created attachment 43980 [details] ksymoops from 2.4.9-21enterprise
This problem went away in the 2.4.18-* kernels in 7.3