rmid

Section: User Commands (1)
Updated: 05 Aug 2006
Index Return to Main Contents
\ " Changed by: Ann Wollrath - JavaSoft East, 2-Mar-2000 \ " Changed by: Jen McGinn - Jini Technology Software, 6-Mar-2000

 

NAME

rmid - The Java RMI Activation System Daemon

rmid starts the activation system daemon that allows objects to be registered and activated in a virtual machine (VM).

 

SYNOPSIS


rmid [options]

 

DESCRIPTION

The rmid tool starts the activation system daemon. The activation system daemon must be started before activatable objects can be either registered with the activation system or activated in a VM. See the Java RMI Specification @ http://java.sun.com/javase/6/docs/platform/rmi/spec/rmitoc.html and Activation tutorials @ http://java.sun.com/javase/6/docs/technotes/guides/rmi/activation/overview.html for details on how to write programs that use activatable remote objects.

The daemon can be started by executing the rmid command, and specifying a security policy file, as follows:


    rmid -J-Djava.security.policy=rmid.policy

Note: When running Sun's implementation of rmid, by default you will need to specify a security policy file so that rmid can verify whether or not the information in each ActivationGroupDesc is allowed to be used to launch a VM for an activation group. Specifically, the command and options specified by the CommandEnvironment and any Properties passed to an ActivationGroupDesc's constructor must now be explicitly allowed in the security policy file for rmid. The value of the sun.rmi.activation.execPolicy property dictates the policy that rmid uses to determine whether or not the information in an ActivationGroupDesc may be used to launch a VM for an activation group.

Executing rmid by default

*
starts the Activator and an internal registry on the default port, 1098, and
*
binds an ActivationSystem to the name java.rmi.activation.ActivationSystem in this internal registry.

To specify an alternate port for the registry, you must specify the -port option when starting up rmid. For example,


    rmid -J-Djava.security.policy=rmid.policy -port 1099

starts the activation system daemon and a registry on the registry's default port, 1099.

 

Starting rmid from

inetd/xinetd

An alternative to starting rmid from the command line is to configure inetd (Solaris) or xinetd (Linux) to start rmid on demand.

When rmid starts up, it attempts to obtain an inherited channel (inherited from inetd/xinetd) by invoking the System.inheritedChannel method. If the inherited channel is null or not an instance of java.nio.channels.ServerSocketChannel, then rmid assumes that it was not started by inetd/xinetd, and it starts up as described above.

If the inherited channel is a ServerSocketChannel instance, then rmid uses the java.net.ServerSocket obtained from the ServerSocketChannel as the server socket that accepts requests for the remote objects it exports, namely the registry in which the java.rmi.activation.ActivationSystem is bound and the java.rmi.activation.Activator remote object. In this mode, rmid behaves the same as when it is started from the command line, except:

*
Output printed to System.err is redirected to a file. This file is located in the directory specified by the java.io.tmpdir system property (typically /var/tmp or /tmp) with the prefix "rmid-err" and the suffix "tmp".
*
The -port option is disallowed. If this option is specified, rmid will exit with an error message.
*
The -log option is required. If this option is not specified, rmid will exit with an error message.

See the man pages for inetd (Solaris) or xinetd (Linux) for details on how to configure services to be started on demand.

 

OPTIONS

-C<someCommandLineOption>
Specifies an option that is passed as a command-line argument to each child process (activation group) of rmid when that process is created. For example, you could pass a property to each virtual machine spawned by the activation system daemon:

    rmid -C-Dsome.property=value

This ability to pass command-line arguments to child processes can be useful for debugging. For example, the following command:

    rmid -C-Djava.rmi.server.logCalls=true

will enable server-call logging in all child VMs.

-J<someCommandLineOption>
Specifies an option that is passed to the java interpreter running rmid. For example, to specify that rmid use a policy file named rmid.policy, the -J option can be used to define the java.security.policy property on rmid's command line, for example:

    rmid -J-Djava.security.policy=rmid.policy

-J-Dsun.rmi.activation.execPolicy=<policy> Specifies the policy that rmid employs to check commands and command-line options used to launch the VM in which an activation group runs. Please note that this option exists only in Sun's implementation of the Java RMI activation daemon. If this property is not specified on the command line, the result is the same as if -J-Dsun.rmi.activation.execPolicy=default were specified. The possible values of <policy> can be default, <policyClassName>, or none:

*
default (or if this property is unspecified)

The default execPolicy allows rmid to execute commands with specific command-line options only if rmid has been granted permission to execute those commands and options in the security policy file that rmid uses. Only the default activation group implementation can be used with the default execution policy.

rmid launches a VM for an activation group using the information in the group's registered activation group descriptor, an ActivationGroupDesc. The group descriptor specifies an optional ActivationGroupDesc.CommandEnvironment which includes the command to execute to start the activation group as well as any command line options to be added to the command line. By default, rmid uses the java command found in java.home. The group descriptor also contains properties overrides that are added to the command line as options defined as:


    -D<property>=<value>

The permission com.sun.rmi.rmid.ExecPermission is used to grant rmid permission to execute a command, specified in the group descriptor's CommandEnvironment to launch an activation group. The permission com.sun.rmi.rmid.ExecOptionPermission is used to allow rmid to use command-line options, specified as properties overrides in the group descriptor or as options in the CommandEnvironment, when launching the activation group.

When granting rmid permission to execute various commands and options, the permissions ExecPermission and ExecOptionPermission need to be granted universally (i.e., granted to all code sources).

ExecPermission
The ExecPermission class represents permission for rmid to execute a specific command to launch an activation group.

Syntax
The name of an ExecPermission is the path name of a command to grant rmid permission to execute. A path name that ends in "/*" indicates all the files contained in that directory (where "/" is the file-separator character, File.separatorChar). A path name that ends with "/-" indicates all files and subdirectories contained in that directory (recursively). A path name consisting of the special token "<<ALL FILES>>" matches any file.

Note: A path name consisting of a single "*" indicates all the files in the current directory, while a path name consisting of a single "-" indicates all the files in the current directory and (recursively) all files and subdirectories contained in the current directory.

ExecOptionPermission
The ExecOptionPermission class represents permission for rmid to use a specific command-line option when launching an activation group. The name of an ExecOptionPermission is the value of a command line option.

Syntax
Options support a limited wildcard scheme. An asterisk signifies a wildcard match, and it may appear as the option name itself (i.e., it matches any option), or an asterisk may appear at the end of the option name only if the asterisk follows either a "." or "=".

For example: "*" or "-Dfoo.*" or "-Da.b.c=*" is valid, "*foo" or "-Da*b" or "ab*" is not.

Policy file for
rmid When granting rmid permission to execute various commands and options, the permissions ExecPermission and ExecOptionPermission need to be granted universally (i.e., granted to all code sources). It is safe to grant these permissions universally because only rmid checks these permissions.

An example policy file that grants various execute permissions to rmid is:


grant {
    permission com.sun.rmi.rmid.ExecPermission
        "/files/apps/java/jdk1.2.2/solaris/bin/java";

    permission com.sun.rmi.rmid.ExecPermission
        "/files/apps/java/jdk1.2.2/solaris/bin/java_g";

    permission com.sun.rmi.rmid.ExecPermission
        "/files/apps/rmidcmds/*";

    permission com.sun.rmi.rmid.ExecOptionPermission
        "-Djava.security.policy=/files/policies/group.policy";

    permission com.sun.rmi.rmid.ExecOptionPermission
        "-Djava.security.debug=*";

    permission com.sun.rmi.rmid.ExecOptionPermission
        "-Dsun.rmi.*";
};

The first two permissions granted allow rmid to execute the 1.2.2 version of the java and java_g commands, specified by their explicit path names. Note that by default, the version of the java command found in java.home is used (the same one that rmid uses), and does not need to be specified in the policy file. The third permission allows rmid to execute any command in the directory /files/apps/rmidcmds.

The fourth permission granted, an ExecOptionPermission, allows rmid to launch an activation group that defines the security policy file to be /files/policies/group.policy. The next permission allows the java.security.debug property to be used by an activation group. The last permission allows any property in the sun.rmi property name hierarchy to be used by activation groups.

To start rmid with a policy file, the java.security.policy property needs to be specified on rmid's command line, for example:

rmid -J-Djava.security.policy=rmid.policy

*
*
<policyClassName>

If the default behavior is not flexible enough, an administrator can provide, when starting rmid, the name of a class whose checkExecCommand method is executed in order to check commands to be executed by rmid.

The policyClassName specifies a public class with a public, no-argument constructor and an implementation of the following checkExecCommand method:


    public void checkExecCommand(ActivationGroupDesc desc,
                                 String[] command)
        throws SecurityException;

Before launching an activation group, rmid calls the policy's checkExecCommand method, passing it the activation group descriptor and an array containing the complete command to launch the activation group. If the checkExecCommand throws a SecurityException, rmid will not launch the activation group and an ActivationException will be thrown to the caller attempting to activate the object.

*
none

If the sun.rmi.activation.execPolicy property value is "none", then rmid will not perform any validation of commands to launch activation groups.

-log dir
Specifies the name of the directory the activation system daemon uses to write its database and associated information. The log directory defaults to creating a directory, log, in the directory in which the rmid command was executed.
-port port
Specifies the port rmid's registry uses. The activation system daemon binds the ActivationSystem, with the name java.rmi.activation.ActivationSystem, in this registry. Thus, the ActivationSystem on the local machine can be obtained using the following Naming.lookup method call:

    import java.rmi.*; 
    import java.rmi.activation.*;

    ActivationSystem system; system = (ActivationSystem)
    Naming.lookup("//:port/java.rmi.activation.ActivationSystem");
-stop
Stops the current invocation of rmid, for a port specified by the -port option. If no port is specified, it will stop the rmid running on port 1098.
 

ENVIRONMENT VARIABLES

CLASSPATH
Used to provide the system a path to user-defined classes. Directories are separated by colons. For example:

    .:/usr/local/java/classes

 

SEE ALSO

rmic, CLASSPATH, java