Verification done on Xenial. - server:
root@shuckle:~# uname -a Linux shuckle 4.4.0-142-generic #168-Ubuntu SMP Wed Jan 16 21:00:45 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux root@shuckle:~# iptables -F root@shuckle:~# iptables -A INPUT -p tcp -m tcp --syn --dport 7777 -m connlimit --connlimit-above 2000 --connlimit-mask 0 -j DROP root@shuckle:~# ulimit -SHn 65000 root@shuckle:~# ruby server.rb - client: root@dixie:~# ruby client.rb 10.230.56.116 7777 6000 3 Connecting to ["10.230.56.116"]:7777 6000 times with 3 1 2 3 ... 2000 <blocks> ** Tags removed: verification-needed-xenial ** Tags added: verification-done-xenial -- You received this bug notification because you are a member of Kernel Packages, which is subscribed to linux in Ubuntu. https://bugs.launchpad.net/bugs/1811094 Title: iptables connlimit allows more connections than the limit when using multiple CPUs Status in linux package in Ubuntu: Fix Committed Status in linux source package in Xenial: Fix Committed Status in linux source package in Bionic: Fix Committed Status in linux source package in Cosmic: Fix Committed Bug description: [Impact] * The iptables connection count/limit rules can be breached with multithreaded network driver/server/client (common) due to a race in the conncount/connlimit code. * For example: # iptables -A INPUT -p tcp -m tcp --syn --dport 7777 \ -m connlimit --connlimit-above 2000 --connlimit-mask 0 \ -j DROP * The fix is a backport from an upstream commit that resolves the problem (plus dependencies for a cleaner backport) that address the race condition: commit b36e4523d4d5 ("netfilter: nf_conncount: fix garbage collection confirm race"). [Test Case] * Server-side: (relevant kernel side) (limit TCP port 7777 to only 2000 connections) # iptables -A INPUT -p tcp -m tcp --syn --dport 7777 \ -m connlimit --connlimit-above 2000 --connlimit-mask 0 \ -j DROP # ulimit -SHn 65000 # increase number of open files # ruby server.rb # multi-threaded server * Client-side: # ulimit -SHn 65000 # ruby client.rb <server ip> <port> <target # connections> <# threads> <test output> * Results with Original kernel: (client achieves target of 6000 connections > limit of 2000 connections) # ruby client.rb 10.230.56.100 7777 6000 3 1 2 3 <...> 6000 Target reached. Thread finishing 6001 Target reached. Thread finishing 6002 Target reached. Thread finishing Threads done. 6002 connections press enter to exit * Results with Modified kernel: (client is limited to 2000 connections, and times out afterward) # ruby client.rb 10.230.56.100 7777 6000 3 1 2 3 <...> 2000 <... blocks for a few minutes ...> failed to create connection: Connection timed out - connect(2) for "10.230.56.100" port 7777 failed to create connection: Connection timed out - connect(2) for "10.230.56.100" port 7777 failed to create connection: Connection timed out - connect(2) for "10.230.56.100" port 7777 Threads done. 2000 connections press enter to exit * Test cases possibly available upon request, depending on original author's permission. [Regression Potential] * The patchset has been reviewed by a netfilter maintainer [1] in stable mailing list, and was considered OK for 4.14, and that's essentially the same backport for 4.15 and 4.4. * The changes are limited to netfilter connlimit/conncount (names change between older/newer kernel versions). [Other Info] * The backport for 4.14 [2] is applied as of 4.14.92. [1] https://www.spinics.net/lists/stable/msg276883.html [2] https://www.spinics.net/lists/stable/msg276910.html To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1811094/+subscriptions -- Mailing list: https://launchpad.net/~kernel-packages Post to : kernel-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~kernel-packages More help : https://help.launchpad.net/ListHelp
