On Tue, Jul 5, 2011 at 3:47 PM, Fabian Keil <[email protected]> wrote: > Ermal Luçi <[email protected]> wrote: > >> On Sat, Jul 2, 2011 at 5:33 PM, Pierre Lamy <[email protected]> wrote: >> > >> > >> > On 6/29/2011 1:22 PM, Fabian Keil wrote: >> >> >> >> "Bjoern A. Zeeb"<[email protected]> wrote: >> >> >> >>> Begin forwarded message: >> >>> >> >>>> From: "Bjoern A. Zeeb"<[email protected]> >> >>>> Date: June 28, 2011 11:57:25 AM GMT+00:00 >> >>>> To: [email protected], [email protected], >> >>>> [email protected] >> >>>> Subject: svn commit: r223637 - in head: . contrib/pf/authpf >> >>>> contrib/pf/ftp-proxy contrib/pf/man contrib/pf/pfctl contrib/pf/pflogd >> >>>> sbin/pflogd sys/conf sys/contrib/altq/altq sys/contrib/pf/net >> >>>> sys/modules >> >>>> s... >> >>>> >> >>>> Author: bz >> >>>> Date: Tue Jun 28 11:57:25 2011 >> >>>> New Revision: 223637 >> >>>> URL: http://svn.freebsd.org/changeset/base/223637 >> >>>> >> >>>> Log: >> >>>> Update packet filter (pf) code to OpenBSD 4.5. >> >> >> >> Thanks! >> >> >> >>> In short; please test! >> >> >> >> I didn't experience any real problems yet, but running >> >> Privoxy-Regression-Test, I reproducible got this log message >> >> for one of the tests: >> >> >> >> Jun 29 18:26:19 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo1, stored af=2, a0: 10.0.0.1:50722, a1: 10.0.0.1:12345, proto=6, >> >> found >> >> af=2, a0: 10.0.0.1:50722, a1: 10.0.0.1:12345, proto=6. >> >> >> >> This didn't happen with the previous pf version. >> >> >> >> I tracked it down to a test that does a connect() >> >> to a local unbound port. >> >> >> >> It's also reproducible for every address on the system with: >> >> >> >> ifconfig -a | awk '/inet / {system("telnet "$2" 12345")}' >> >> >> >> Jun 29 18:30:49 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo0, stored af=2, a0: 192.168.5.49:61512, a1: 192.168.5.49:12345, >> >> proto=6, found af=2, a0: 192.168.5.49:61512, a1: 192.168.5.49:12345, >> >> proto=6. >> >> Jun 29 18:30:49 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo0, stored af=2, a0: 127.0.0.1:44717, a1: 127.0.0.1:12345, proto=6, >> >> found af=2, a0: 127.0.0.1:44717, a1: 127.0.0.1:12345, proto=6. >> >> Jun 29 18:30:49 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo1, stored af=2, a0: 192.168.6.100:31600, a1: 192.168.6.100:12345, >> >> proto=6, found af=2, a0: 192.168.6.100:31600, a1: 192.168.6.100:12345, >> >> proto=6. >> >> Jun 29 18:30:49 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo1, stored af=2, a0: 10.0.0.1:20126, a1: 10.0.0.1:12345, proto=6, >> >> found >> >> af=2, a0: 10.0.0.1:20126, a1: 10.0.0.1:12345, proto=6. >> >> Jun 29 18:30:49 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo1, stored af=2, a0: 10.0.0.1:10895, a1: 10.0.0.2:12345, proto=6, >> >> found >> >> af=2, a0: 10.0.0.1:10895, a1: 10.0.0.2:12345, proto=6. >> >> Jun 29 18:30:49 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo1, stored af=2, a0: 10.0.0.1:25081, a1: 10.0.0.3:12345, proto=6, >> >> found >> >> af=2, a0: 10.0.0.1:25081, a1: 10.0.0.3:12345, proto=6. >> >> Jun 29 18:30:49 r500 kernel: pf: state key linking mismatch! dir=OUT, >> >> if=lo0, stored af=2, a0: 192.168.0.106:32448, a1: 192.168.0.106:12345, >> >> proto=6, found af=2, a0: 192.168.0.106:32448, a1: 192.168.0.106:12345, >> >> proto=6. >> >> >> >> 12345 can be replaced with any unbound port it seems. >> >> >> >> I'm additionally occasionally seeing the message for successfully >> >> established connections (both internal and outgoing) but don't >> >> know how to reproduce it. >> >> >> >> Fabian >> > >> > I also get the state key mismatch problem, it seems that pf is leaking >> > states (I assume this is the same problem). I also see a strange NAT issue, >> > internal IPs leak somewhat on the outside int. Eventually the system runs >> > out of state entry slots and connectivity is lost. This is on a -current >> > kernel from ~Jun 30, after the 4.5 import. >> > >> > tun0: flags=8151<UP,POINTOPOINT,RUNNING,PROMISC,MULTICAST> metric 0 mtu >> > 1492 >> > options=80000<LINKSTATE> >> > inet6 fe80::290:bff:fe1a:a674%tun0 prefixlen 64 scopeid 0xf >> > inet6 2607:f0b0:0:1:290:bff:fe1a:a674 prefixlen 64 autoconf >> > inet 216.106.102.33 --> 209.87.255.1 netmask 0xffffffff >> > nd6 options=23<PERFORMNUD,ACCEPT_RTADV,AUTO_LINKLOCAL> >> > Opened by PID 3446 >> > >> > em0 is on the 192.168.3/24 network >> > >> > <root.wheel@pyr7535> [/var/preserve/root] # tcpdump -i tun0 net 192.168.3.0 >> > mask 255.255.255.0 >> > tcpdump: verbose output suppressed, use -v or -vv for full protocol decode >> > listening on tun0, link-type NULL (BSD loopback), capture size 65535 bytes >> > 11:22:37.030244 IP 192.168.3.99 > 190.252.34.186: ICMP pandora.userid.org >> > udp port 16881 unreachable, length 134 >> > 11:24:03.137016 IP 192.168.3.99 > 190.252.34.186: ICMP pandora.userid.org >> > udp port 16881 unreachable, length 98 >> > >> > Relevant pf.conf lines: >> > int_if = "em0" >> > ext_if = "tun0" >> > # NAT >> > nat on $ext_if from $int_if:network to any -> ($ext_if) >> > >> > Here is the info about states leaking: >> > >> > State Table Total Rate >> > current entries 108488 >> > >> > <root.wheel@pyr7535> [/var/preserve/root] # pfctl -F states >> > 1003 states cleared >> > <root.wheel@pyr7535> [/var/preserve/root] # pfctl -s info >> > Status: Enabled for 0 days 02:21:18 Debug: Urgent >> > >> > Interface Stats for tun0 IPv4 IPv6 >> > Bytes In 1252327614 1907903 >> > Bytes Out 373783492 1429003 >> > Packets In >> > Passed 1341017 12360 >> > Blocked 45437 831 >> > Packets Out >> > Passed 1186359 13441 >> > Blocked 1641 3724 >> > >> > State Table Total Rate >> > current entries 125127 >> > >> > States aren't getting cleared properly. Below is a sample of the state key >> > linking mismatch problem: >> > >> > Jul 2 11:28:17 pyr7535 kernel: pf: state key linking mismatch! dir=OUT, >> > if=em0, stored af=2, a0: >> >> I just committed a fix for the state key linking mismatch issue. >> Can you test with the latest HEAD sources? > > Works for me, at least the error messages are gone. Thanks a lot. > > I never experienced the "state leak issue" Pierre described above, > but it does seem to take a while until cleared states are reported > as such:
It is normal the GC thread used for this takes only a bunch at a time. Though you want see those states in the output of pfctl -vss since they are masked out. > > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 1556 > fk@r500 ~ $sudo pfctl -F states > 1556 states cleared > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 1259 > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 1133 > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 1019 > fk@r500 ~ $sudo pfctl -F states > 0 states cleared > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 742 > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 667 > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 667 > fk@r500 ~ $sudo pfctl -F states > 0 states cleared > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 436 > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 436 > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 352 > fk@r500 ~ $sudo pfctl -F states > 0 states cleared > fk@r500 ~ $sudo pfctl -s all | grep current > current entries 185 > > I never looked at this before, so it might have always behaved that way. > > Fabian > -- Ermal _______________________________________________ [email protected] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-pf To unsubscribe, send any mail to "[email protected]"
