EMAC OE Boot Process Customization
When designing an embedded system, it is often necessary to add or remove applications and tasks from the system initialization sequence. This guide describes the initialization method used for EMAC OE systems and provides information on customizing the boot process.
Contents
System V Initialization
EMAC OE uses the System V Initialization method. This is a simple method for system initialization using a set of scripts run in sequential order. When the kernel has finished loading, /sbin/init is started to initialize the userspace services in the system. init is process id 1 and the parent of all processes in the system.
Runlevels
System V Init uses different runlevels to control the boot process. Each runlevel has a set of scripts that are run sequentially to start various services on the system. The default runlevel on EMAC OE systems is 5. This is set in /etc/inittab with the line:
# The default runlevel. id:5:initdefault:
There are a total of seven runlevels available in System V Init, labeled 0 through 6. Runlevel 5 is the full user level in EMAC OE systems, regardless of whether a GUI is installed on the board or not. Runlevel 0 halts the system, and runlevel 6 is used for reboot. The other runlevels can be used for other purposes if desired, such as to configure different levels of user functionality in each runlevel.
During boot, the scripts /etc/init.d/rcS and /etc/init.d/rc are executed to run the scripts in /etc/rcS.d/ starting with an 'S' in lexicographic order followed by the scripts in /etc/rc5.d/ (assuming that 5 is the default runlevel). All of these startup scripts are passed the argument start. During halt or reboot, the scripts in /etc/rc0.d/ or /etc/rc6.d/ starting with a 'K' are run in lexicographic order with the argument stop. To control the order in which the scripts are run, each filename is prefixed with a number from 00-99. For example, the listing below illustrates the files in /etc/rcS.d/ that will be run in order before entering the default runlevel:
root@emac-oe:/etc/rcS.d# ls S02banner S20modutils.sh S40networking S03sysfs S30ramdisk S41ifplugd S03udev S35mountall.sh S45mountnfs.sh S05devices S37populate-volatile.sh S55bootmisc.sh S06alignment S38devpts.sh S98ipkg-configure S10checkroot S39hostname.sh S99finish.sh
The action to perform at each level is specified in /etc/inittab. For example, the following lines are used to trigger the execution of the /etc/init.d/rcS and /etc/init.d/rc scripts:
si::sysinit:/etc/init.d/rcS .... l5:5:wait:/etc/init.d/rc 5 ....
Initscripts
The directory /etc/init.d/ holds initialization scripts that are run by init during boot or shutdown. These scripts should be designed to accept at least three arguments: start, stop, or restart. The files in the /etc/rc*.d/ directories are symbolic links to the scripts in /etc/init.d/. This structure allows for easy modification of the boot process and the ability for a script to be run at different places in different runlevels. The detailed listing of the /etc/rcS.d/ directory is shown below:
root@emac-oe:/etc/rcS.d# ls -l lrwxrwxrwx 1 root root 16 Dec 31 1969 S02banner -> ../init.d/banner lrwxrwxrwx 1 root root 18 Dec 31 1969 S03sysfs -> ../init.d/sysfs.sh lrwxrwxrwx 1 root root 14 Dec 31 1969 S03udev -> ../init.d/udev lrwxrwxrwx 1 root root 17 Dec 31 1969 S05devices -> ../init.d/devices lrwxrwxrwx 1 root root 22 Dec 31 1969 S06alignment -> ../init.d/alignment.sh lrwxrwxrwx 1 root root 19 Dec 31 1969 S10checkroot -> ../init.d/checkroot lrwxrwxrwx 1 root root 21 Dec 31 1969 S20modutils.sh -> ../init.d/modutils.sh lrwxrwxrwx 1 root root 17 Dec 31 1969 S30ramdisk -> ../init.d/ramdisk lrwxrwxrwx 1 root root 21 Dec 31 1969 S35mountall.sh -> ../init.d/mountall.sh lrwxrwxrwx 1 root root 30 Dec 31 1969 S37populate-volatile.sh -> ../init.d/populate-volatile.sh lrwxrwxrwx 1 root root 19 Dec 31 1969 S38devpts.sh -> ../init.d/devpts.sh lrwxrwxrwx 1 root root 21 Dec 31 1969 S39hostname.sh -> ../init.d/hostname.sh lrwxrwxrwx 1 root root 20 Dec 31 1969 S40networking -> ../init.d/networking lrwxrwxrwx 1 root root 17 Dec 31 1969 S41ifplugd -> ../init.d/ifplugd lrwxrwxrwx 1 root root 21 Dec 31 1969 S45mountnfs.sh -> ../init.d/mountnfs.sh lrwxrwxrwx 1 root root 21 Dec 31 1969 S55bootmisc.sh -> ../init.d/bootmisc.sh lrwxrwxrwx 1 root root 24 Dec 31 1969 S98ipkg-configure -> ../init.d/ipkg-configure lrwxrwxrwx 1 root root 19 Dec 31 1969 S99finish.sh -> ../init.d/finish.sh
Use the other boot scripts on the system for examples when creating custom initscripts. The application that is being started should be stored in the system PATH, such as /usr/bin/, and started from the script. For example, the busybox-httpd initscript is shown below.
#!/bin/sh
DAEMON=/usr/sbin/httpd
NAME=httpd
DESC="Busybox HTTP Daemon"
#HTTPROOT="/srv/www"
HTTPROOT="/home/www"
ARGS="-h $HTTPROOT"
test -f $DAEMON || exit 0
set -e
case "$1" in
start)
echo -n "starting $DESC: $NAME... "
if [ ! -d $HTTPROOT ]; then
echo "$HTTPROOT is missing."
exit 1
fi
start-stop-daemon -S -b -n $NAME -a $DAEMON -- $ARGS
echo "done."
;;
stop)
echo -n "stopping $DESC: $NAME... "
start-stop-daemon -K -n $NAME
echo "done."
;;
restart)
echo "restarting $DESC: $NAME... "
$0 stop
$0 start
echo "done."
;;
reload)
echo -n "reloading $DESC: $NAME... "
killall -HUP $(basename ${DAEMON})
echo "done."
;;
*)
echo "Usage: $0 {start|stop|restart|reload}"
exit 1
;;
esac
exit 0
Notice the different actions for each argument in the case statement. The start and stop blocks will be run directly during the boot and shutdown processes based on the location of the 'S' and 'K' links to this script. The restart and reload options are additional controls that are provided by many init scripts to allow for easier restarting of the daemon or process manually from a running system. For example, to restart the busybox-httpd server, you would run the command:
root@emac-oe:~# /etc/init.d/busybox-httpd restart
This would effectively stop the server and then start it again. This is generally useful for forcing reconfiguration after configuration file changes are made without needing to reboot the system.
The start-stop-daemon can be used to control the initialization and termination of the application as illustrated in the busybox-httpd initscript above. The available options for the busybox version of the start-stop-daemon included with EMAC OE are shown below.
BusyBox v1.13.2 (2009-06-24 18:00:48 CDT) multi-call binary
Usage: start-stop-daemon [OPTIONS] [-S|-K] ... [-- arguments...]
Search for matching processes, and then
-K: stop all matching processes.
-S: start a process unless a matching process is found.
Process matching:
-u,--user USERNAME|UID Match only this user's processes
-n,--name NAME Match processes with NAME
in comm field in /proc/PID/stat
-x,--exec EXECUTABLE Match processes with this command
in /proc/PID/cmdline
-p,--pidfile FILE Match a process with PID from the file
All specified conditions must match
-S only:
-x,--exec EXECUTABLE Program to run
-a,--startas NAME Zeroth argument
-b,--background Background
-N,--nicelevel N Change nice level
-c,--chuid USER[:[GRP]] Change to user/group
-m,--make-pidfile Write PID to the pidfile specified by -p
-K only:
-s,--signal SIG Signal to send
-t,--test Match only, exit with 0 if a process is found
Other:
-o,--oknodo Exit with status 0 if nothing is done
-v,--verbose Verbose
-q,--quiet Quiet
Adding and Removing Scripts
A new or existing initscript can be added or removed from the start process by simply creating or destroying the symbolic links from the /etc/rc*.d/ directories. A utility application, update-rc.d is provided to automate this process. The usage of update-rc.d is shown below:
usage: update-rc.d [-n] [-f] [-r <root>] <basename> remove
update-rc.d [-n] [-r <root>] [-s] <basename> defaults [NN | sNN kNN]
update-rc.d [-n] [-r <root>] [-s] <basename> start|stop NN runlvl [runlvl] [...] .
-n: not really
-f: force
-r: alternate root path (default is /)
-s: invoke start methods if appropriate to current runlevel
The update-rc.d utility can be used to add or remove the startup links for any initscript in the system. For example, the /etc/init.d/busybox-httpd script has the following links to it in the boot process:
/etc/rc0.d/K20busybox-httpd /etc/rc1.d/K20busybox-httpd /etc/rc2.d/S20busybox-httpd /etc/rc3.d/S20busybox-httpd /etc/rc4.d/S20busybox-httpd /etc/rc5.d/S20busybox-httpd /etc/rc6.d/K20busybox-httpd
Removing the links from the system startup without actually removing the /etc/init.d/busybox-httpd file will effectively disable the daemon from running at system startup or shutdown. To accomplish this, use the remove argument to update-rc.d as illustrated below:
root@emac-oe:~# update-rc.d -f busybox-httpd remove update-rc.d: /etc/init.d/busybox-httpd exists during rc.d purge (continuing) Removing any system startup links for busybox-httpd ... /etc/rc0.d/K20busybox-httpd /etc/rc1.d/K20busybox-httpd /etc/rc2.d/S20busybox-httpd /etc/rc3.d/S20busybox-httpd /etc/rc4.d/S20busybox-httpd /etc/rc5.d/S20busybox-httpd /etc/rc6.d/K20busybox-httpd
 | The -f argument is required in this case to "force" the links to be removed while the /etc/init.d/busybox-httpd still exists. |
To add the links for the /etc/init.d/busybox-httpd script back into the system, the defaults argument may be used. Alternatively, the start and stop number for each runlevel may be specified explicitly as shown in the example below. Both methods will yield exactly the same results in this case; the runlevels are specified in this example simply to demonstrate the syntax. Using this method is required when it is important that a script starts at a specific point in the init process such as directly after networking is started.
root@emac-oe:~# update-rc.d busybox-httpd start 20 2 3 4 5 . stop 20 0 1 6 . Adding system startup for /etc/init.d/busybox-httpd ... /etc/rc2.d/S20busybox-httpd -> ../init.d/busybox-httpd /etc/rc3.d/S20busybox-httpd -> ../init.d/busybox-httpd /etc/rc4.d/S20busybox-httpd -> ../init.d/busybox-httpd /etc/rc5.d/S20busybox-httpd -> ../init.d/busybox-httpd /etc/rc0.d/K20busybox-httpd -> ../init.d/busybox-httpd /etc/rc1.d/K20busybox-httpd -> ../init.d/busybox-httpd /etc/rc6.d/K20busybox-httpd -> ../init.d/busybox-httpd
Custom Initialization
In some designs, the target application is so specific that only one main task will be performed. After a custom application has been developed and tested, it may be possible to simplify the boot process significantly by modifying the System V Init configuration or replacing init with a custom script or application. The end result is generally a must faster boot time.
The custom initialization script will need to do all requires system configuration and provide or start all required services. For example, the /proc and /sys virtual filesystems will need to be mounted, networking may need to be configured, devices and temporary filesystems may need to be created and mounted, and any other tasks that are required for the primary application to function.
 | Using this method will disable all of the standard system services unless they are explicitly started. It is possible to lock out all access to the system if you have no way of accessing the bootloader. |
Overriding Init
The Linux kernel accepts a boot argument init that specifies which process to execute for init. If not specified, the system will try a number of predefined locations, executing /sbin/init on EMAC OE systems. It is possible to create a custom script or application that runs instead of /sbin/init and have the kernel execute that immediately on boot. For example, if the custom init script is located at /sbin/custom-init.sh, specify init=/sbin/custom-init.sh on the kernel boot arguments.
Changing the Boot Arguments
Appending the custom init value is accomplished by changing the bootloader configuration. For U-Boot, the bootargs variable will need to be changed to add the init value. RedBoot stores this configuration in the boot_script_data configuration. The kernel arguments can be modified for LILO by entering them directly at the LILO prompt or in the LILO configuration file. Refer to the documentation for the bootloader on your system for more detailed information.
| Be very careful when making any changes to settings in the bootloader. It is possible to render a system unusable from incorrect configuration. |
Modifying the Init Configuration
While the method of overriding init described in the previous section will work well and offers absolute control over the system, it has several disadvantages, such as requiring bootloader configuration changes and not providing a system login console. It is possible to simply modify the init configuration in /etc/inittab and achieve a very similar effect. Since init is still being used, the serial console and other directives in the inittab file will still be valid.
