Copyright (c) 2003, Sun Microsystems, Inc. All Rights Reserved.
The contents of this file are subject to the terms of the Common Development and Distribution License (the "License"). You may not use this file except in compliance with the License.
You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing. See the License for the specific language governing permissions and limitations under the License.
When distributing Covered Code, include this CDDL HEADER in each file and include the License file at usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your own identifying information: Portions Copyright [yyyy] [name of copyright owner]
rcm_scriptname scriptinfo
rcm_scriptname register
rcm_scriptname resourceinfo resourcename
rcm_scriptname queryremove resourcename
rcm_scriptname preremove resourcename
rcm_scriptname postremove resourcename
rcm_scriptname undoremove resourcename
Reconfiguration and Coordination Manager (RCM) is a framework designed to coordinate device consumers during Solaris Dynamic Reconfiguration (DR). The interfaces specified in this man page allow device consumers, such as application vendors or site administrators, to act before and after DR operations take place by providing RCM scripts. You can write your own RCM scripts to shut down your applications, or to cleanly release the devices from your applications during dynamic remove operations.
An RCM script is an executable perl script, a shell script or a binary. Perl is the recommended language. Each script is run in its own address space using the user-id of the script file owner.
An RCM script is invoked on demand in response to DR as follows:
<scriptname> <command> [args ...]
Every script must implement the following RCM commands: scriptinfo
Get script information.
Register devices the script handles.
Get resource information.
A script might include some or all the of the following commands: queryremove
Queries whether the resource can be released.
Releases the resource.
Provides post-resource removal notification.
Undo the actions done in preremove.
When a script's register command is run, the script should supply, in return data, all resource names the script or its application handles that could potentially be removed by DR. A resource name refers to a name in /dev path name.
Below is a high-level overview of the sequence of script invocations that occurs when dynamic removal of a script's registered resource is attempted. See the COMMANDS section for a detailed description of the commands.
1. Prior to removing the resource from the system during DR, the script's queryremove command is run:
<scriptname> queryremove <resourcename>The script should check for obvious reasons why the resource can not be removed from the perspective of its service or application.
2. If the script indicates that the resource can be removed in the queryremove command. The script's preremove command is run:
<scriptname> preremove <resourcename>The script releases the resource from the service or application represented by the script and prepares for the resource removal. Releasing the resource includes closing the resource if the resource is currently opened by its application.
3. The system then proceeds to remove the resource.
4. If the system has removed the resource successfully the script's postremove command is run:
<scriptname> postremove <resourcename>Otherwise the script's undoremove command is run:
<scriptname> undoremove <resourcename>
For any commands the script does not implement, it must exit with exit status of 2. RCM silently returns success for the script's unimplemented commands.
A script performs the following basic steps:
Takes RCM command and additional arguments from the command line and environment parameters.
Processes the command.
Writes the expected return data to stdout as name=value pairs delimited by newlines, where name is the name of the return data item that RCM expects and value is the value associated with the data item.
The initial environment of RCM scripts is set as follows:
Process UID is set to the UID of the script.
Process GID is set to the GID of the script.
Current working directory is set to:
/tmp for scripts not owned by root
Environment variable RCM_ENV_DEBUG_LEVEL is set to the debug level. Logging is discussed below.
The following environment variables are also set where possible:
LANG
LC_COLLATE
LC_CTYPE
LC_MESSAGES
LC_MONETARY
LC_NUMERIC
LC_TIME
LC_ALL
TZ
All environment variable names beginning with RCM_ENV_ are reserved for use by the RCM.
The character encoding used by the RCM and RCM scripts to exchange RCM commands, environment parameters, and name-value pairs is ASCII unless the controlling environment variables are specified otherwise.
The scriptinfo command is invoked to gather information about the script. Return data:
If successful, the script must write the following name-value pairs to stdout and exit with status 0:
rcm_script_version=1
rcm_script_func_info=script_func_info
rcm_cmd_timeout=command_timeout_value
The register command is invoked to allow a script to specify the resources that it or its application handles that could potentially be removed by DR. The script has to supply all its resource names to RCM using the name-value pair rcm_resource_name. Return Data:
If successful, the script must write the following name-value pairs to stdout and exit with status 0:
rcm_resource_name=resourcename rcm_resource_name=resourcename . . .where resourcename is the name of the resource the script is interested in. Upon failure, the script must specify the failure reason using the name-value pair rcm_failure_reason and exit with status 1.
The resourceinfo command is invoked to get the usage information about resourcename. Return Data:
If successful, the script must write the following name-value pair to stdout and exit with status 0:
rcm_resource_usage_info=resource_usage
where resource_usage is a localized human readable message describing the
usage of the resource by the script.
Upon failure, the script must specify the failure reason using the name-value
pair rcm_failure_reason and exit with status 1.
Prior to removing the resource from the system, the queryremove command is invoked to query the script to determine whether the script can release the given resource successfully from the service or application it represents. The script does not actually release the resource. The script might indicate that it is not able to release the resource if the resource is critical for its service or application.
Additional environment parameter: RCM_ENV_FORCE
Can be one of: FALSE
Normal request.
Request is urgent. The script should check whether the resource can be released successfully by force, such as by using the force option to unmount a file system.
If the command succeeds, the script must return no data and exit with status 0. If the script would not be able to release the resource, it must specify the reason using the name-value pair rcm_failure_reason and exit with status 3. Upon any other failure, the script must specify the failure reason using the name-value pair rcm_failure_reason and exit with status 1.
The preremove command is invoked prior to an attempt to remove the given resourcename. In response to this command the script can either release the resource (including closing the device if the device is currently opened) from the service or application it represents or indicate that it can not release the resource if the resource is critical for its service or application.
Additional environment parameter: RCM_ENV_FORCE
Can be one of: FALSE
Normal request.
Request is urgent. The script should make extra effort to release the resource, such as by using the force option to unmount a file system.
If the command succeeds, the script must return no data and exit with status 0. If the script cannot release the resource, it must specify the reason using the name-value pair rcm_failure_reason and exit with status 3. Upon any other failure, the script must specify the failure reason using the name-value pair rcm_failure_reason and exit with status 1.
The postremove command is invoked after the given resourcename has been removed. Return Data:
If the command succeeds, the script must return no data and exit with status 0. Upon failure, the script must specify the failure reason using the name-value pair rcm_failure_reason and exit with status 1.
undoremove resourcename
The undoremove command is invoked to undo what was done in the previous preremove command for the given resourcename. The script can bring the state of the resource to the same state it was in when the script received the preremove command for that resource. Return Data:
If the command succeeds, the script must return no data and exit with status 0. Upon failure, the script must specify the failure reason using the name-value pair rcm_failure_reason and exit with status 1.
A script must log all error and debug messages by writing to stdout the name-value pairs listed below. The logged messages go to syslogd(1M) with the syslog facility of LOG_DAEMON. See syslog.conf(4). rcm_log_err=message
Logs the message with the syslog level of LOG_ERR.
Logs the message with the syslog level of LOG_WARNING.
Logs the message with the syslog level of LOG_INFO.
Logs the message with the syslog level of LOG_DEBUG.
A script can use the environment variable RCM_ENV_DEBUG_LEVEL to control the amount of information to log. RCM_ENV_DEBUG_LEVEL is a numeric value ranging from 0 to 9, with 0 meaning log the least amount of information and 9 meaning log the most.
You must use the following format to name a script:
vendor,service
where vendor is the stock symbol (or any distinctive name) of the vendor providing the script and service is the name of service the script represents.
You must be a superuser (root) to install or remove an RCM script.
Select one of the following directories where you want to place the script: /etc/rcm/scripts
Scripts for specific systems
Scripts for specific hardware implementation
Scripts for specific hardware class
Scripts for any hardware
To install a script, copy the script to the appropriate directory from the list above, change the userid and the groupid of the script to the desired values, and send SIGHUP to rcm_daemon. For example:
# cp SUNW,sample.pl /usr/lib/rcm/scripts # chown user[:group] /usr/lib/rcm/scripts/SUNW,sample.pl # pkill -HUP -x -u root rcm_daemon
Remove the script from the appropriate directory from the list above and send SIGHUP to rcm_daemon. For example:
# rm /usr/lib/rcm/scripts/SUNW,sample.pl # pkill -HUP -x -u root rcm_daemon
Example 1 Site Customization RCM Script
#! /usr/bin/perl -w # # A sample site customization RCM script for a tape backup application. # # This script registers all tape drives in the system with RCM. # When the system attempts to remove a tape drive by DR the script # does the following: # - if the tape drive is not being used for backup, it allows the # DR to continue. # - if the tape drive is being used for backup, and when DR is not # forced (RCM_ENV_FORCE=FALSE) it indicates that it cannot release # the tape drive with appropriate error message. When forced # (RCM_ENV_FORCE=TRUE) it kills the tape backup application in # order to allow the DR to continue. # # This script does not implement the postremove and undoremove commands # since there is nothing to cleanup after DR remove operation is # completed or failed. If any cleanup is needed after the DR removal # completed, postremove command needs to implemented. If any cleanup is # needed in the event of DR removal failure, undoremove command needs # to be implemented. # use strict; my ($cmd, %dispatch); $cmd = shift(@ARGV); # dispatch table for RCM commands %dispatch = ( "scriptinfo" => do_scriptinfo, "register" => do_register, "resourceinfo" => do_resourceinfo, "queryremove" => do_preremove, "preremove" => do_preremove ); if (defined($dispatch{$cmd})) { &{$dispatch{$cmd}}; } else { exit (2); } sub do_scriptinfo { print "rcm_script_version=1\en"; print "rcm_script_func_info=Tape backup appl script for DR\en"; exit (0); } sub do_register { my ($dir, $f, $errmsg); $dir = opendir(RMT, "/dev/rmt"); if (!$dir) { $errmsg = "Unable to open /dev/rmt directory: $!"; print "rcm_failure_reason=$errmsg\en"; exit (1); } while ($f = readdir(RMT)) { # ignore hidden files and multiple names for the same device if (($f !~ /^\./) && ($f =~ /^[0-9]+$/)) { print "rcm_resource_name=/dev/rmt/$f\en"; } } closedir(RMT); exit (0); } sub do_resourceinfo { my ($rsrc, $unit); $rsrc = shift(@ARGV); if ($rsrc =~ /^\/dev\/rmt\/([0-9]+)$/) { $unit = $1; print "rcm_resource_usage_info=Backup Tape Unit Number $unit\en"; exit (0); } else { print "rcm_failure_reason=Unknown tape device!\en"; exit (1); } } sub do_preremove { my ($rsrc); $rsrc = shift(@ARGV); # check if backup application is using this resource # if (the backup application is not running on $rsrc) { # allow the DR to continue # exit (0); #} # # If RCM_ENV_FORCE is FALSE deny the operation. # If RCM_ENV_FORCE is TRUE kill the backup application in order # to allow the DR operation to proceed # if ($ENV{RCM_ENV_FORCE} eq 'TRUE') { if ($cmd eq 'preremove') { # kill the tape backup application } exit (0); } else { # # indicate that the tape drive can not be released # since the device is being used for backup by the # tape backup application # print "rcm_failure_reason=tape backup in progress pid=...\en"; exit (3); } }
A script must exit with following exit status values: 0
Operation specified by the given RCM command has been executed successfully by the script. For queryremove command it also means that the script can successfully release the resource.
An error occurred while processing the RCM command. The script should provide the error message to RCM using the name-value pair rcm_failure_reason before exiting.
The script does not support the given RCM command. A script must exit with this status if it cannot understand the given RCM command.
Indicates that the script cannot release the resource for preremove and queryremove commands. The script should provide a message to RCM specifying the reason for not being able to release the resource using the name-value pair rcm_failure_reason before exiting.
If a script cannot successfully process an RCM command, it must supply to the RCM a message indicating the reason for failure by writing a name-value pair, in the form shown below, to stdout and exiting with the appropriate exit status.
rcm_failure_reason=failure_reason
where failure_reason is a localized human readable message describing the reason for failure of the RCM command.
See attributes(5) for descriptions of the following attributes:
ATTRIBUTE TYPE ATTRIBUTE VALUE |
Interface Stability Evolving |
gettext(1), cfgadm(1M), cfgadm_scsi(1M), cfgadm_pci(1M), syslog(3C), signal.h(3HEAD), syslog.conf(4), attributes(5), environ(5)
RCM scripts are expected to properly handle all RCM commands that the script implements and to log all errors. Only root has permission to add or remove an RCM script. An ill-behaved RCM script can cause unexpected DR failures.
RCM commands are invoked only for the resources whose subsystems participate within the RCM framework. Currently, not all susbsystems participate within the RCM framework.