It was reported [1] that when Squid does a DNS lookup that results in a DNS response larger than 512 bytes, the server will try to query over TCP rather than UDP. Squid will start the TCP connection, and then erroneously thinks it is sending an IPv6 DNS query and will crash if no IPv6 resolver is present. This was reported to be evident in Squid version 3.1.6, but not 3.1.1. It is corrected upstream in 3.1.7 [2]. [1] http://bugs.squid-cache.org/show_bug.cgi?id=3021 [2] http://bazaar.launchpad.net/~squid/squid/3.1/revision/10072
Duplicate? https://bugzilla.redhat.com/show_bug.cgi?id=626927
This doesn't have any real bearing on Squid 2.x. Tested here in RHEL5 and connecting to a site that serves no content but has a large DNS response yields quite a timeout before getting an "unable to connect" message in elinks. Noticed no significant memory usage, same process IDs running before and after, so this would only affect >3.1.1 and <=3.1.6 (not sure of the exact version on the low end). As a result, this does not affect Red Hat Enterprise Linux 4 or 5 which provide squid 2.6.STABLE21 or earlier. It also does not affect Fedora which provides 3.1.8 across all supported versions.
(In reply to comment #2) > Duplicate? https://bugzilla.redhat.com/show_bug.cgi?id=626927 Yes, it is. Thanks for spotting that Kurt. *** This bug has been marked as a duplicate of bug 626927 ***