Tarjei Skorgenes created AMQ-4249:
-------------------------------------
Summary: Race condition in PropertiesLoginModule
Key: AMQ-4249
URL: https://issues.apache.org/jira/browse/AMQ-4249
Project: ActiveMQ
Issue Type: Bug
Components: Broker
Affects Versions: 5.7.0, 5.6.0, 5.5.1
Environment: ActiveMQ 5.5.1 with JAAS authentication via
PropertiesLoginModule configured, Spring JMS 3.0.5, 64-bit, Intel quad core CPU
(i7 950), Windows 7
Reporter: Tarjei Skorgenes
Priority: Minor
h1. Problem
While setting up a connection pool towards ActiveMq 5.5.1 using Bitronix 2.1.3
I've been having some issues related to authentication and authorization of the
JMS connections.
When doing a clean restart of the JMS-clients JVM the connection pool has been
unable to connect successfully with one or more of the 14 connections it's been
set up to use.
The error messages I've been getting has usually been one of the following two:
# User system is not authorized to read from:
ActiveMQ.Advisory.TempQueue,ActiveMQ.Advisory.TempTopic
# User name or password is invalid.
The broker has been set up using a very simplistic configuration:
{code:xml}
<?xml version="1.0" encoding="UTF-8"?>
<beans xmlns="http://www.springframework.org/schema/beans"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xmlns:context="http://www.springframework.org/schema/context"
xsi:schemaLocation="http://activemq.apache.org/schema/core
http://activemq.apache.org/schema/core/activemq-core-5.5.0.xsd
http://www.springframework.org/schema/beans
http://www.springframework.org/schema/beans/spring-beans.xsd
http://www.springframework.org/schema/context
http://www.springframework.org/schema/context/spring-context.xsd">
<context:property-placeholder />
<broker brokerId="localhost"
xmlns="http://activemq.apache.org/schema/core"
dataDirectory="${datadir}" start="false">
<managementContext>
<managementContext connectorHost="0.0.0.0"
connectorPort="14522" createConnector="true" />
</managementContext>
<plugins>
<jaasAuthenticationPlugin
configuration="activemq-domain"
discoverLoginConfig="true" />
<authorizationPlugin>
<map>
<authorizationMap>
<authorizationEntries>
<authorizationEntry
queue=">" read="admins"
write="admins"
admin="admins" />
<authorizationEntry
topic=">" read="admins"
write="admins"
admin="admins" />
<authorizationEntry
topic="ActiveMQ.Advisory.>"
read="guests,users" write="guests,users" admin="guests,users" />
</authorizationEntries>
</authorizationMap>
</map>
</authorizationPlugin>
</plugins>
<transportConnectors>
<transportConnector id="openwire"
uri="tcp://0.0.0.0:61616?trace=true" />
</transportConnectors>
</broker>
</beans>
{code}
The JAAS-configuration has been verified to match username and password used by
the client when connecting (username = system):
h4. login.config
{code}
activemq-domain {
org.apache.activemq.jaas.PropertiesLoginModule required
debug=false
org.apache.activemq.jaas.properties.user="users.properties"
org.apache.activemq.jaas.properties.group="groups.properties";
};
{code}
h4.users.properties
{code}
system=manager
user=password
guest=password
{code}
h4. groups.properties
{code}
admins=system
users=user
guests=guest
{code}
h1. Cause
After debugging the problem it seems as if the problem is caused by a race
condition introduced in PropertiesLoginModule in revision
[1086219|https://fisheye6.atlassian.com/changelog/activemq?showid=1086219]
(AMQ-3244). When the reload-support was added the users- and groups-fields were
changed into static fields. But no additional synchronization was introduced,
thereby introducing a race condition when several threads are entering the
initialize- and commit-methods are the same time.
The following section of the initialize-method in PropertiesLoginModule
contains the first part of the race condition. Please note the unsynchronized
modification of both the users- and groups static fields:
{code:java}
if (reload || users == null || uf.lastModified() > usersReloadTime) {
if (debug) {
LOG.debug("Reloading users from " + uf.getAbsolutePath());
}
try {
users = new Properties(); // XXX Here be dragons
java.io.FileInputStream in = new java.io.FileInputStream(uf);
users.load(in);
in.close();
usersReloadTime = System.currentTimeMillis();
} catch (IOException ioe) {
LOG.warn("Unable to load user properties file " + uf);
}
}
groupsFile = (String) options.get(GROUP_FILE) + "";
File gf = baseDir != null ? new File(baseDir, groupsFile) : new
File(groupsFile);
if (reload || groups == null || gf.lastModified() > groupsReloadTime) {
if (debug) {
LOG.debug("Reloading groups from " + gf.getAbsolutePath());
}
try {
groups = new Properties(); // XXX Here be dragons
java.io.FileInputStream in = new java.io.FileInputStream(gf);
groups.load(in);
in.close();
groupsReloadTime = System.currentTimeMillis();
} catch (IOException ioe) {
LOG.warn("Unable to load group properties file " + gf);
}
}
{code}
The next part comes in the login-method when the users password is retrieved:
{code:java}
String password = users.getProperty(user);
{code}
The final part of the puzzle occurs in the commit-method:
{code:java}
for (Enumeration<?> enumeration = groups.keys();
enumeration.hasMoreElements();) {
String name = (String)enumeration.nextElement();
String[] userList = ((String)groups.getProperty(name) +
"").split(",");
for (int i = 0; i < userList.length; i++) {
if (user.equals(userList[i])) {
principals.add(new GroupPrincipal(name));
break;
}
}
}
{code}
The retrieval of the user password will fail if invoked by a thread immediately
after a different thread has assigned an empty Properties-object to the users
field in the initialize-method.
Similarly population of the GroupPrincipals into the Subject in the
commit-method will silently fail if executed by a thread immediately after a
different thread has assigned an empty Properties-object to the groups-field in
the initialize-method.
h1. Proposed solution
I've created a testcase that illustrates the problem and an additional patch
that introduces a wrapper around the Properties-objects for the users- and
groups-fields. The wrapper uses a read-write lock to synchronize access and
modification of the properties-objects containing the users and groups.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira