[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#511165: Does not show up with standard clients



I tried to further locate when the bug appears. As it turned out the command line ftp client does not cause the kernel panic, neither in passive nor in standard mode. I also ran the ftpsync Perl script without problems.

In order to allow the systems to access ftp, I added the following rule to the applicable chain:

iptables -A _chain_ -p tcp -o ppp0 -s _client-system_ --syn -m state --state NEW -j ACCEPT

(yes, there is a state established, related rule somewhere at the top of the stack.) However, even after having performed the ftpsync, using gFTP through frox panicked the firewall.

However, during further searching for the point of no return, I found that my proxy did not even have permission to open a connection to port 21. After allowing this, I can browse ftp:// URLs through the same squid3 as used by frox without problems. The kernel panic using gFTP still persists.

If nf_nat_ftp and nf_conntrack_ftp are not loaded during browser access, the data connection is just dropped by the firewall. Nothing evil happens.

This is the frox configuration file. So far untested, but a very similar one ran for a couple of years on my Etch server.

/# grep -v '^#' /etc/frox.conf | grep -v '^ *$'
Listen proxy.mgr
Port 2121
ResolvLoadHack wontresolve.doesntexist.abc
User frox
Group frox
WorkingDir /srv/proxy/frox
LogLevel 25
LogFile /var/log/frox.log
PidFile /var/run/frox.pid
BounceDefend yes
CacheModule http
HTTPProxy 127.0.0.1:8080
MinCacheSize 5000
DoNTP yes
MaxForks 10
MaxForksPerHost 4
ACL Allow sleipnir - *

gFTP on sleipnir is configured to use:
Proxy: proxy.mgr:2121
Typ: user@host NOAUTH
USER %hu@%hh
PASS %hp

... and yes proxy.mgr resolves to the OpenVZ container homing the frox.



Reply to: