- A denial of service flaw was found in the way the Squid Cache Manager processed certain requests. A remote attacker who is able to access the Cache Manager CGI could use this flaw to cause Squid to consume an excessive amount of memory.
- Due to a bug in the
ConnStateData::noteMoreBodySpaceAvailable()function, child processes of Squid terminated upon encountering a failed assertion. An upstream patch has been provided and Squid child processes no longer terminate.
- Due to an upstream patch, which renamed the HTTP header controlling persistent connections from
Connection, the NTLM pass-through authentication does not work, thus preventing login. This update adds the new
http10option to the
squid.conffile, which can be used to enable the change in the patch. This option is set to
offby default. When set to
on, the NTLM pass-through authentication works properly, thus allowing login attempts to succeed.
- When the IPv6 protocol was disabled and Squid tried to handle an HTTP GET request containing an IPv6 address, the Squid child process terminated due to signal
6. This bug has been fixed and such requests are now handled as expected.
- The old "stale if hit" logic did not account for cases where the stored stale response became fresh due to a successful re-validation with the origin server. Consequently, incorrect warning messages were returned. Now, Squid no longer marks elements as stale in the described scenario.
- When squid packages were installed before samba-winbind, the
wbprivgroup did not include Squid. Consequently, NTLM authentication calls failed. Now, Squid correctly adds itself into the wbpriv group if samba-winbind is installed before Squid, thus fixing this bug.
- In FIPS mode, Squid was using private MD5 hash functions for user authentication and network access. As MD5 is incompatible with FIPS mode, Squid could fail to start. This update limits the use of the private MD5 functions to local disk file hash identifiers, thus allowing Squid to work in FIPS mode.
- Under the high system load, the squid process could terminate unexpectedly with a segmentation fault during reboot. This update provides better memory handling during reboot, thus fixing this bug.
- Squid incorrectly set the timeout limit for client HTTP connections with the value for server-side connections, which is much higher, thus creating unnecessary delays. With this update, Squid uses a proper value for the client timeout limit.
- When the GET method requested a fully-qualified domain name that did not contain the
AAAArecord, Squid delayed due to long DNS requesting time. This update introduces the
squid.conf. If the
dns_timeoutvalue of this option is properly set, Squid sends the
AAAAqueries in parallel and the delays no longer occur.
- Squid did not properly release allocated memory when generating error page contents, which caused memory leaks. Consequently, the Squid proxy server consumed a lot of memory within a short time period. This update fixes this memory leak.
- Squid did not pass the
identvalue to a URL rewriter that was configured using the
url_rewrite_programdirective. Consequently, the URL rewriter received the dash character (
-) as the user value instead of the correct user name. Now, the URL rewriter receives the correct user name in the described scenario.
- Squid, used as a transparent proxy, can only handle the HTTP protocol. Previously, it was possible to define a URL in which the access protocol contained the asterisk character (
*) or an unknown protocol namespace URI. Consequently, an
Invalid URLerror message was logged to
access.logduring reload. This update ensures that
http://is always used in transparent proxy URLs, and the error message is no longer logged in this scenario.