This issue has had me scratching my head for a few days now. For sites on a certain network (which happens to be a DDOS migitation provider), security management checkConnect calls seem to be taking a very, very long time in comparison to other sites.
Is there any way that sites on this network can be thrawting the security manager's access checks? Is there a configuration that I am not aware of? Am I going (more) insane?
Here is a test case that demonstrates...
package com.test;
import java.net.Socket;
import java.util.ArrayList;
import java.util.List;
public class SSCCE
{
static class StatCounter
{
boolean security;
String host;
long avg;
long total;
int iterations;
StatCounter(String host)
{
this.host = host;
}
@Override
public String toString()
{
return host + "\t\titerations (" + iterations + ")\t\tavg (" + avg + ")\t\tsecurity (" + security + ")";
}
void inc(long time)
{
++iterations;
total += time;
}
void avg()
{
avg = total / (long)iterations;
}
void reset()
{
total = 0;
iterations = 0;
}
}
static String[] hosts = new String[]
{
"google.com",
"youtube.com",
"oracle.com",
"random.org",
"phpbb.com",
"staminus.net",
// MUCH Higher Latency with site below (only with security manager enabled?)
"blacklotus.net"
};
public static void main(String[] argv) throws Throwable
{
int iterations = Integer.parseInt(argv[0]);
List<StatCounter> counters = new ArrayList<StatCounter>(hosts.length);
for(String host : hosts)
{
counters.add(new StatCounter(host));
}
System.out.println("Running Without Security");
for(int i = 0; i < iterations; ++i)
{
for(StatCounter counter : counters)
{
long then = System.currentTimeMillis();
new Socket(counter.host, 80).close();
counter.inc(System.currentTimeMillis() - then);
}
}
for(StatCounter counter : counters)
{
counter.avg();
System.out.println(counter);
counter.reset();
counter.security = true;
}
System.setProperty("java.security.policy", "sscce.policy");
System.setSecurityManager(new SecurityManager());
System.out.println("\n\nRunning With Security");
for(int i = 0; i < iterations; ++i)
{
for(StatCounter counter : counters)
{
long then = System.currentTimeMillis();
new Socket(counter.host, 80).close();
counter.inc(System.currentTimeMillis() - then);
}
}
for(StatCounter counter : counters)
{
counter.avg();
System.out.println(counter);
}
}
}
Policy File
grant
{
permission java.net.SocketPermission "google.com:80", "connect";
permission java.net.SocketPermission "youtube.com:80", "connect";
permission java.net.SocketPermission "oracle.com:80", "connect";
permission java.net.SocketPermission "random.org:80", "connect";
permission java.net.SocketPermission "phpbb.com:80", "connect";
permission java.net.SocketPermission "staminus.net:80", "connect";
permission java.net.SocketPermission "blacklotus.net:80", "connect";
};
Run with java com.test.SSCCE
Example Output
Running Without Security
google.com iterations (4) avg (65) security (false)
youtube.com iterations (4) avg (61) security (false)
oracle.com iterations (4) avg (104) security (false)
random.org iterations (4) avg (101) security (false)
phpbb.com iterations (4) avg (143) security (false)
staminus.net iterations (4) avg (137) security (false)
blacklotus.net iterations (4) avg (137) security (false)
Running With Security
google.com iterations (4) avg (261) security (true)
youtube.com iterations (4) avg (64) security (true)
oracle.com iterations (4) avg (103) security (true)
random.org iterations (4) avg (100) security (true)
phpbb.com iterations (4) avg (882) security (true)
staminus.net iterations (4) avg (303) security (true)
blacklotus.net iterations (4) avg (4669) security (true)
I'd really appreciate any input, thanks.
As far as I can see, everything looks good (except for what is seemingly network latency!)
I've finally tracked down the root cause of this phenomenon by reverse engineering the JRE system libraries with additional debugging information. I was able to trace the cause of the delay down to the native method
java.net.Inet4AddressImpl.getHostByAddr([B)Ljava.lang.String
It turns out resolving the host in question from its IP address fails for an unknown reason and after timing out, the calling method tries another route to return a valid hostname. I have checked this with several web-based reverse lookup tools and it seems that there is most likely an issue with blacklotus.net (or perhaps a feature to prevent reverse lookups?)
I hope this helps anyone else that encounters an odd-ball case like this. T-minus 36 hours, case closed.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With