Optimize RC Scripts

From eLinux.org
Jump to: navigation, search

RC Scripts Speed-Up

This material was excerpted from a document with the following copyright statement:

  • Copyright 2002, 2003, 2004 Sony Corporation
  • Copyright 2002, 2003, 2004 Matsushita Electric Industrial Co., Ltd.
  • Copyright 2002-2004 by MontaVista Software.

It was submitted as input to the forum by Sony Corporation, on April 8, 2004.

Introduction

The init scripts of the existing Embedded Linux distribution are the shell scripts to be executed with ’bash’. To reduce system boot time, some modifications to be applied to the scripts and faster shell interpreter to be used. The document describes the BusyBox optimizations; the init scripts modifications; how to reduce the system boot time using the optimized BusyBox, and to optimize shell scripts for BusyBox. Also, the benchmark procedure and optimization results are described.

Purpose of Feature

The init scripts execution time, i.e. the time interval between the start of the init process and the start of user applications, must be reduced.

Feature requirements

  • The modified init scripts must be run with "bash" as well as the BusyBox "ash" shell.
  • The execution time of the init scripts and the total system boot time must be reduced.
  • The guide to speed-optimization of the init scripts must be provided.

Acceptance Criteria

  • The modified init scripts are able to successfully run with "bash" as well as the BusyBox "ash" shell.
  • The execution time of the init scripts with BusyBox is reduced in the comparison with the original init scripts with "bash"; the total system boot time is not greater then 5 sec.
  • The guide to speed-optimization of the init scripts is available.

BusyBox Optimization

Since ’bash’ and GNU utilities are very heavy applications, BusyBox is useful to reduce system boot time. BusyBox combines tiny versions of many common UNIX utilities (shellutils, fileutils, etc.) into a single small executable. The commands and utilities included in BusyBox are divided into the classes: built-ins and applets. The built-ins are simply invoked as functions, and applets are invoked by means of the ’fork/exec’ system calls. Also, BusyBox scripts can use external commands and utilities.

Usage of the BusyBox built-ins is rather than the applets and external commands by performance reasons, because the ’fork/exec’ system calls are very heavy and they make the main contribution to shell inefficiency. Because the original BusyBox is only size-optimized, the following features must be considered from the performance standpoint:

  • Each command including built-ins within pipes are forked.
  • Each back-quoted command is forked.
  • The ’echo’, ’test’, and ’[’ commands and other most frequent commands in the scripts are implemented as applets.

To avoid such drawbacks, we optimized BusyBox 1.00-pre3 in order to speed up script execution. The BusyBox optimizations for the ’ash’ shell are listed below:

  • The set of shell commands and utilities is implemented as built-ins.
  • The invoked ’cat’ command at the beginning of pipes is eliminated and file descriptors are passed only into the next command of the pipe.

The following shell commands and utilities are implemented as built-ins in the optimized BusyBox ’ash’ shell: ’,’, ’:’,’[’, ’alias’, ’break’, ’cd’, ’chdir’, ’continue’, ’echo’, ’eval’, ’exec’, ’exit’, ’export’, ’false’, ’kill’, ’let’, ’local’, ’pidof’, ’pwd’, ’read’, ’readonly’, ’return’, ’set’, ’shift’, ’test’, ’times’, ’trap’, ’true’, ’type’, ’ulimit’, ’umask’, ’unalias’, ’unset’, ’wait’.

NOTES:

  • This need seems to be addressed for all applets at least in Busybox 1.1.1 (and perhaps in earlier versions). The "Shells -> Standalone shell" configuration setting is supposed to address this need. See the menuconfig help for details.
  • In newer BusyBox releases (1.13.0 and maybe even earlier releases) the most frequently used applets, 'test', 'echo' can be configured as being 'built-in'. The newer BusyBox releases are also smaller in size and can save some extra milliseconds in execution of startup scripts.

Known Problems

The following BusyBox commands work in the different manner from the ’bash’ commands and GNU utilities: ’nice’, ’find’, ’mount’, ’umount’, ’init’, ’halt’, ’shutdown’, ’syslogd’, ’klogd’, ’hwclock’, ’cron’, ’anacron’, ’crontab’, 'pidof'.

If the different behaviour is unwanted or these BusyBox applets do not provide the necessary utility, use the external commands instead of them, or simply do not configure them as applets.

How to Optimize Init Scripts for BusyBox

Follow the rules listed below to reduce execution time for the init scripts:

  • Do not use unnecessary codes in the scripts.
  • Replace external commands and utilities with the BusyBox built-ins as far as possible.
  • Do not use the piped commands as far as possible.
  • Reduce the number of commands within a pipe.
  • Do not use the back-quoted commands as far as possible.

The main goal of such optimization is to reduce the number of the "fork/exec" calls during a script execution.

Examples of Init Scripts Optimization

The following examples demonstrate how the recommendations of shell scripts optimization can be applied to the init scripts.

Unnecessary codes elimination

This example demonstrates the elimination of duplicate codes from the "mountswap.sh" and "checkrootfs.sh" scripts: the command "swapon" runs one time in the modified scripts.

  • Before optimization:

checkrootfs.sh:

if [ "$FSCKSWAP" !== no ]
then
   if [ -x /sbin/swapon ]
   then
       mount -n /proc
       if ! grep -qs resync /proc/mdstat
       then
           [ "$VERBOSE" !== no ] && echo "Activating swap..."

           swapon -a 2> /dev/null
       fi
       umount -n /proc
    fi
fi

mountswap.sh:

grep -qs resync /proc/mdstat || swapon -a 2> /dev/null
  • After optimization:

checkrootfs.sh:


mountswap.sh:


if [ "$FSCKSWAP" !== "no" ]; then
    if [ ! ‘grep -qs resync /proc/mdstat‘ ]
    then
        log_status_msg "Starting $DESC: " -n
        log_status_msg "$BASENAME1" -n
        $DAEMON1 $ARGS1
        RET==$?
        if [ $RET -eq 0 ]; then
            log_success_msg ". " -n
        else
            log_failure_msg " failed ($RET: $ERROR)."
            return 1
        fi
    fi
fi

Built-in usage

This example demonstrates usage of the "echo" built-in instead of the external command "printf" in the "nfs-common" script.

  • Before optimization:

nfs-common:

   printf "Starting $DESC:"
   printf " statd"
  • After optimization: init-functions:
log_status_msg() {
   if [ "$1" !== "" ] && [ "$1" !== "-n" ]
   then

      if [ "$2" == "-n" ]
      then echo -n "$1"
      else echo "$1"
      fi

    fi
return 0
}


nfs-common.sh:

 log_status_msg "Starting $DESC: " -n
 log_status_msg "$BASENAME1" ?n

Piped command usage

This example demonstrates the elimination of the piped commands (the example is hypothetic, because the init scripts do not contain such inefficiencies).

  • Before optimization:
cat /proc/mounts | grep ext3 | cut -d’ ’ -f2,3
  • After optimization:
sed -n ’s/^[^ ]* \([^ ]*\) \(ext3\) .*$/\1 \2/p’ /proc/mounts

This example demonstrates the reduction of the commands in the pipe (the example is also hypothetic). Note, the optimized version does not invoke the "fork" call, because the "cat" optimization is used.

  • Before optimization:
cat /etc/passwd | grep user | wc -l | tr -d ’’ | sed ’s/ *//’
  • After optimization:
cat /etc/passwd | grep -c user

Back-quoted command usage

This example demonstrates the back-quoted command elimination (the example is hypothetic, because the init scripts do not contain such inefficiencies).

  • Before optimization:
if [ "‘grep rpcuser /etc/passwd‘" !== "" ]
then
    echo "rpcuser"
else
    echo "no rpcuser"
fi
  • After optimization:
if grep rpcuser /etc/passwd >/dev/null
then
    echo "rpcuser"
else
    echo "no rpcuser"
fi

Init Scripts Optimization

The existing init scripts were modified to reduce their execution time following the recommendations, which are described above.

Benchmark Environment and Procedure

To estimate the results of the init script optimization and the BusyBox usage, the TI OMAP 1510 Innovator platform is used. To measure the duration of the kernel loading, the KFI support is used (to measure the init script execution time, the KFI support is disabled). The measurements are performed on the systems with/without XIP support To take measurements without the XIP support, the following kernel configuration is used:

CONFIG_ARM==y
CONFIG_UID16==y
CONFIG_RWSEM_GENERIC_SPINLOCK==y
CONFIG_EXPERIMENTAL==y
CONFIG_ADVANCED_OPTIONS==y
CONFIG_MODULES==y
CONFIG_KMOD==y
CONFIG_ARCH_OMAP==y
CONFIG_OMAP_INNOVATOR==y
CONFIG_INNOVATOR_MISSED_IRQS==y
CONFIG_ARCH_OMAP1510==y
CONFIG_CLOCK_COUNTS_DOWN==y
CONFIG_CPU_32==y
CONFIG_CPU_ARM925T==y
CONFIG_CPU_ARM925_CPU_IDLE==y
CONFIG_CPU_ARM925_I_CACHE_ON==y
CONFIG_CPU_ARM925_NON_STREAMING_ON==y
CONFIG_CPU_ARM925_D_CACHE_ON==y
CONFIG_CPU_32v4==y
CONFIG_KERNEL_START==0xc0000000
CONFIG_ZBOOT_ROM_TEXT==0
CONFIG_ZBOOT_ROM_BSS==0
CONFIG_NET==y
CONFIG_SYSVIPC==y
CONFIG_SYSCTL==y

CONFIG_MAX_USER_RT_PRIO==100
CONFIG_MAX_RT_PRIO==0
CONFIG_FPE_NWFPE==m
CONFIG_KCORE_ELF==y
CONFIG_BINFMT_AOUT==m
CONFIG_BINFMT_ELF==y
CONFIG_OMAP1510_PM==y
CONFIG_DPM==y
CONFIG_BOOT_FREQ==y
CONFIG_OMAP_ARM_168MHZ==y
CONFIG_OMAP1510_DPM==y
CONFIG_INNOVATOR_DPM==y
CONFIG_CMDLINE=="mem==32M console==ttyS0,115200n8 noinitrd root==/dev/null rootflags==physaddr==0x0260000"
CONFIG_ALIGNMENT_TRAP==y
CONFIG_PREEMPT==y
CONFIG_LOCK_BREAK==y
CONFIG_MTD==y
CONFIG_MTD_PARTITIONS==y
CONFIG_MTD_CONCAT==y
CONFIG_MTD_CHAR==y
CONFIG_MTD_BLOCK==y
CONFIG_MTD_CFI==y
CONFIG_MTD_GEN_PROBE==y
CONFIG_MTD_CFI_ADV_OPTIONS==y
CONFIG_MTD_CFI_NOSWAP==y
CONFIG_MTD_CFI_GEOMETRY==y
CONFIG_MTD_CFI_B2==y
CONFIG_MTD_CFI_I1==y
CONFIG_MTD_CFI_INTELEXT==y
CONFIG_MTD_CFI_AMDSTD==y
CONFIG_MTD_OMAP==y
CONFIG_MTD_OMAP_0==y
CONFIG_MTD_OMAP_1==y
CONFIG_BLK_DEV_LOOP==m
CONFIG_BLK_DEV_RAM==y
CONFIG_BLK_DEV_RAM_SIZE==4096
CONFIG_BLK_DEV_INITRD==y
CONFIG_PACKET==m
CONFIG_NETFILTER==y
CONFIG_UNIX==y
CONFIG_INET==y
CONFIG_IP_MULTICAST==y
CONFIG_IP_PNP==y
CONFIG_NETDEVICES==y
CONFIG_NET_ETHERNET==y
CONFIG_NET_VENDOR_SMC==y
CONFIG_SMC9194==y
CONFIG_PPP==m
CONFIG_PPP_MULTILINK==y
CONFIG_PPP_ASYNC==m
CONFIG_PPP_DEFLATE==m
CONFIG_PPPOE==m
CONFIG_IRDA==m
CONFIG_IRLAN==m
CONFIG_IRNET==m
CONFIG_IRCOMM==m
CONFIG_OMAP_SIR==m
CONFIG_INPUT==m
CONFIG_INPUT_KEYBDEV==m
CONFIG_INPUT_MOUSEDEV==m
CONFIG_INPUT_MOUSEDEV_SCREEN_X==240
CONFIG_INPUT_MOUSEDEV_SCREEN_Y==320
CONFIG_INPUT_EVDEV==m
CONFIG_VT==y
CONFIG_VT_CONSOLE==y
CONFIG_SERIAL==y
CONFIG_SERIAL_CONSOLE==y
CONFIG_UNIX98_PTYS==y

CONFIG_UNIX98_PTY_COUNT==256
CONFIG_I2C==m
CONFIG_I2C_ALGOBIT==m
CONFIG_I2C_OMAP1510==m
CONFIG_I2C_CHARDEV==m
CONFIG_I2C_PROC==m
CONFIG_SENSORS==y
CONFIG_SENSORS_OTHER==y
CONFIG_SENSORS_EEPROM==m
CONFIG_WATCHDOG==y
CONFIG_OMAP_WATCHDOG==m
CONFIG_OMAP_RTC==m
CONFIG_RV5C387_RTC==m
CONFIG_RV5C387_RTC==m
CONFIG_VIDEO_DEV==m
CONFIG_VIDEO_PROC_FS==y
CONFIG_AUTOFS4_FS==m
CONFIG_EXT3_FS==m
CONFIG_JBD==m
CONFIG_FAT_FS==m
CONFIG_MSDOS_FS==m
CONFIG_VFAT_FS==m
CONFIG_JFFS_FS==m
CONFIG_JFFS_FS_VERBOSE==0
CONFIG_JFFS2_FS==y
CONFIG_JFFS2_FS_DEBUG==0
CONFIG_CRAMFS==y
CONFIG_CRAMFS_LINEAR==y
CONFIG_CRAMFS_LINEAR_XIP==y
CONFIG_ROOT_CRAMFS_LINEAR==y
CONFIG_TMPFS==y
CONFIG_RAMFS==y
CONFIG_PROC_FS==y
CONFIG_DEVPTS_FS==y
CONFIG_EXT2_FS==m
CONFIG_NFS_FS==y
CONFIG_NFS_V3==y
CONFIG_NFSD==m
CONFIG_NFSD_V3==y
CONFIG_SUNRPC==y
CONFIG_LOCKD==y
CONFIG_LOCKD_V4==y
CONFIG_SMB_FS==m
CONFIG_MSDOS_PARTITION==y
CONFIG_SMB_NLS==y
CONFIG_NLS==y
CONFIG_NLS_DEFAULT=="iso8859-1"
CONFIG_NLS_CODEPAGE_437==m
CONFIG_PC_KEYMAP==y
CONFIG_FB==y
CONFIG_DUMMY_CONSOLE==y
CONFIG_FB_OMAP==y
CONFIG_FBCON_ADVANCED==y
CONFIG_FBCON_CFB16==y
CONFIG_FBCON_FONTWIDTH8_ONLY==y
CONFIG_FBCON_FONTS==y
CONFIG_FONT_8x8==y
CONFIG_FONT_ACORN_8x8==y
CONFIG_SOUND==m
CONFIG_SOUND_OMAP==m
CONFIG_SOUND_OMAP_AIC23==m
CONFIG_INNOVATOR_TS==y
CONFIG_MMC==m
CONFIG_OMAP_MMC==m
CONFIG_INSTANT_ON==y
CONFIG_DEFAULT_LPJ==414720
CONFIG_INSTANT_ON_LPJ==414720
CONFIG_FRAME_POINTER==y

CONFIG_DEBUG_INFO==y
CONFIG_DEBUG_KERNEL==y
CONFIG_MAGIC_SYSRQ==y
CONFIG_DEBUG_BUGVERBOSE==y
CONFIG_DEBUG_ERRORS==y
CONFIG_ZLIB_INFLATE==y
CONFIG_ZLIB_DEFLATE==y

To take measurements with the XIP support, the following kernel configuration is used:

CONFIG_ARM==y
CONFIG_UID16==y
CONFIG_RWSEM_GENERIC_SPINLOCK==y
CONFIG_EXPERIMENTAL==y
CONFIG_ADVANCED_OPTIONS==y
CONFIG_MODULES==y
CONFIG_KMOD==y
CONFIG_ARCH_OMAP==y
CONFIG_OMAP_INNOVATOR==y
CONFIG_INNOVATOR_MISSED_IRQS==y
CONFIG_ARCH_OMAP1510==y
CONFIG_CLOCK_COUNTS_DOWN==y
CONFIG_CPU_32==y
CONFIG_CPU_ARM925T==y
CONFIG_CPU_ARM925_CPU_IDLE==y
CONFIG_CPU_ARM925_I_CACHE_ON==y
CONFIG_CPU_ARM925_NON_STREAMING_ON==y
CONFIG_CPU_ARM925_D_CACHE_ON==y
CONFIG_CPU_32v4==y
CONFIG_KERNEL_START==0xc0000000
CONFIG_ZBOOT_ROM_TEXT==0
CONFIG_ZBOOT_ROM_BSS==0
CONFIG_NET==y
CONFIG_SYSVIPC==y
CONFIG_SYSCTL==y
CONFIG_MAX_USER_RT_PRIO==100
CONFIG_MAX_RT_PRIO==0
CONFIG_XIP_ROM==y
CONFIG_XIP_PHYS_ADDR==60400
CONFIG_FPE_NWFPE==m
CONFIG_KCORE_ELF==y
CONFIG_BINFMT_AOUT==m
CONFIG_BINFMT_ELF==y
CONFIG_OMAP1510_PM==y
CONFIG_DPM==y
CONFIG_BOOT_FREQ==y
CONFIG_OMAP_ARM_168MHZ==y
CONFIG_OMAP1510_DPM==y
CONFIG_INNOVATOR_DPM==y
CONFIG_CMDLINE=="mem==32M console==ttyS0,115200n8 noinitrd root==/dev/null rootflags==physaddr==0x0260000"
CONFIG_ALIGNMENT_TRAP==y
CONFIG_PREEMPT==y
CONFIG_LOCK_BREAK==y
CONFIG_MTD==y
CONFIG_MTD_DEBUG==y
CONFIG_MTD_DEBUG_VERBOSE==0
CONFIG_MTD_PARTITIONS==y
CONFIG_MTD_CONCAT==y
CONFIG_MTD_CHAR==y
CONFIG_MTD_BLOCK==y
CONFIG_MTD_CFI==y
CONFIG_MTD_GEN_PROBE==y
CONFIG_MTD_CFI_ADV_OPTIONS==y
CONFIG_MTD_CFI_NOSWAP==y
CONFIG_MTD_CFI_GEOMETRY==y
CONFIG_MTD_CFI_B2==y
CONFIG_MTD_CFI_I1==y

CONFIG_MTD_CFI_INTELEXT==y
CONFIG_MTD_CFI_AMDSTD==y
CONFIG_MTD_OMAP==y
CONFIG_MTD_OMAP_1==y
CONFIG_BLK_DEV_LOOP==m
CONFIG_BLK_DEV_RAM==y
CONFIG_BLK_DEV_RAM_SIZE==4096
CONFIG_BLK_DEV_INITRD==y
CONFIG_PACKET==m
CONFIG_NETFILTER==y
CONFIG_UNIX==y
CONFIG_INET==y
CONFIG_IP_MULTICAST==y
CONFIG_IP_PNP==y
CONFIG_NETDEVICES==y
CONFIG_NET_ETHERNET==y
CONFIG_NET_VENDOR_SMC==y
CONFIG_SMC9194==y
CONFIG_PPP==m
CONFIG_PPP_MULTILINK==y
CONFIG_PPP_ASYNC==m
CONFIG_PPP_DEFLATE==m
CONFIG_PPPOE==m
CONFIG_IRDA==m
CONFIG_IRLAN==m
CONFIG_IRNET==m
CONFIG_IRCOMM==m
CONFIG_OMAP_SIR==m
CONFIG_INPUT==m
CONFIG_INPUT_KEYBDEV==m
CONFIG_INPUT_MOUSEDEV==m
CONFIG_INPUT_MOUSEDEV_SCREEN_X==240
CONFIG_INPUT_MOUSEDEV_SCREEN_Y==320
CONFIG_INPUT_EVDEV==m
CONFIG_VT==y
CONFIG_VT_CONSOLE==y
CONFIG_SERIAL==y
CONFIG_SERIAL_CONSOLE==y
CONFIG_UNIX98_PTYS==y
CONFIG_UNIX98_PTY_COUNT==256
CONFIG_I2C==m
CONFIG_I2C_ALGOBIT==m
CONFIG_I2C_OMAP1510==m
CONFIG_I2C_CHARDEV==m
CONFIG_I2C_PROC==m
CONFIG_SENSORS==y
CONFIG_SENSORS_OTHER==y
CONFIG_SENSORS_EEPROM==m
CONFIG_WATCHDOG==y
CONFIG_OMAP_WATCHDOG==m
CONFIG_OMAP_RTC==m
CONFIG_RV5C387_RTC==m
CONFIG_RV5C387_RTC==m
CONFIG_VIDEO_DEV==m
CONFIG_VIDEO_PROC_FS==y
CONFIG_AUTOFS4_FS==m
CONFIG_EXT3_FS==m
CONFIG_JBD==m
CONFIG_FAT_FS==m
CONFIG_MSDOS_FS==m
CONFIG_VFAT_FS==m
CONFIG_JFFS_FS==m
CONFIG_JFFS_FS_VERBOSE==0
CONFIG_JFFS2_FS==y
CONFIG_JFFS2_FS_DEBUG==0
CONFIG_CRAMFS==y
CONFIG_CRAMFS_LINEAR==y
CONFIG_CRAMFS_LINEAR_XIP==y

CONFIG_ROOT_CRAMFS_LINEAR==y
CONFIG_TMPFS==y
CONFIG_RAMFS==y
CONFIG_PROC_FS==y
CONFIG_DEVPTS_FS==y
CONFIG_EXT2_FS==m
CONFIG_NFS_FS==y
CONFIG_NFS_V3==y
CONFIG_NFSD==m
CONFIG_NFSD_V3==y
CONFIG_SUNRPC==y
CONFIG_LOCKD==y
CONFIG_LOCKD_V4==y
CONFIG_SMB_FS==m
CONFIG_MSDOS_PARTITION==y
CONFIG_SMB_NLS==y
CONFIG_NLS==y
CONFIG_NLS_DEFAULT=="iso8859-1"
CONFIG_NLS_CODEPAGE_437==m
CONFIG_PC_KEYMAP==y
CONFIG_FB==y
CONFIG_DUMMY_CONSOLE==y
CONFIG_FB_OMAP==y
CONFIG_FBCON_ADVANCED==y
CONFIG_FBCON_CFB16==y
CONFIG_FBCON_FONTWIDTH8_ONLY==y
CONFIG_FBCON_FONTS==y
CONFIG_FONT_8x8==y
CONFIG_FONT_ACORN_8x8==y
CONFIG_SOUND==m
CONFIG_SOUND_OMAP==m
CONFIG_SOUND_OMAP_AIC23==m
CONFIG_INNOVATOR_TS==y
CONFIG_MMC==m
CONFIG_OMAP_MMC==m
CONFIG_TRACE==y
CONFIG_TRACE_BOOT==y
CONFIG_INSTANT_ON==y
CONFIG_DEFAULT_LPJ==414720
CONFIG_INSTANT_ON_LPJ==414720
CONFIG_FRAME_POINTER==y
CONFIG_DEBUG_INFO==y
CONFIG_DEBUG_KERNEL==y
CONFIG_MAGIC_SYSRQ==y
CONFIG_DEBUG_BUGVERBOSE==y
CONFIG_DEBUG_ERRORS==y
CONFIG_ZLIB_INFLATE==y
CONFIG_ZLIB_DEFLATE==y

The set of the init scripts of the consumer packages is divided into the minimal and optional subsets. The following scripts belong to the minimal subset: ’bootmisc.sh’, ’checkfs.sh’, ’checkroot.sh’, ’hwclock.sh’, ’modutils.sh’, ’mountall.sh’, ’networking.sh’, ’urandom.sh’.

The following scripts belong to the optional subset: ’anacron.sh’, ’cron.sh’, ’devfsd.sh’, ’devshm.sh’, ’ifupdown.sh’, ’rmnologin.sh’, ’syslog.sh’.

Thus,

  • the minimal packages are: "initscripts", "util-linux", "modutils", and "netbase";
  • the optional packages are: "initscripts", "anacron", "cron", "devfsd", "ifupdown", "sysklogd", "util-linux", "modutils", and "netbase".

The init scripts are used with the "bash", BusyBox, 0.60.3 and the optimized BusyBox 1.00-pre3. The sizes of the shell executables are:

Since profiling tools slow up program execution, the Linux Trace Toolkit (LTT) is not used to measure the execution time of the init scripts.


Shell Size
bash 562 Kb
BusyBox 0.60.3 872 Kb
BusyBox 1.00-pre3 210 Kb
The optimized BusyBox 1.00-pre3 259 Kb

To obtain the start time of the init scripts, the modified "init" utility is used. The following patch for the "init" utility is applied:

diff -Naur sysvinit-2.78.orig/src/init.c sysvinit-2.78/src/init.c
--- sysvinit-2.78.orig/src/init.c 2000-02-11 14:17:02.000000000 +0300
+++ sysvinit-2.78/src/init.c 2003-12-08 20:20:15.000000000 +0300
@@ -684,7 +684,7 @@
 	(void) tcgetattr(fd, &tty);
 	tty.c_cflag &== CBAUD|CBAUDEX|CSIZE|CSTOPB|PARENB|PARODD;
-	tty.c_cflag |== HUPCL|CLOCAL;
+	tty.c_cflag |== CREAD|HUPCL|CLOCAL;
 	tty.c_cc[VINTR] == 3; /* ctrl(’c’) */
 	tty.c_cc[VQUIT] == 28; /* ctrl(’\\’) */
@@ -2443,6 +2443,13 @@
 	FILE *fp;
 	char *p;
 	int f, fd;
+
+#ifdef TIME_MEASURE
+ struct timeval current;
+ if (!gettimeofday (&current, 0))
+ printf ("INIT: started at %ld\n", (current.tv_sec * 1000) + (current.tv_usec / 1000));
+#endif
 /* Get my own name */
 if ((p == strrchr(argv[0], ’/’)) !== NULL)

To obtain the time of the end of the init script execution, the following program is used in the inittab:

#include <sys/time.h>
#include <stdio.h>
int main (int argc, char **argv)
{
	struct timeval current;
	if (!gettimeofday (&current, 0))
	printf ("printtime: current time %ld ms\n", (current.tv_sec * 1000) + (current.tv_usec / 1000));
	return 0;
}

Thus, we consider the sum of the init script execution time and the kernel loading time, which are measured in ways described above, is the total system boot time.

Optimization Results

Measurement results with the XIP support Kernel loading time: 0.6 sec. Init script execution time:
Script set Shell Time %
minimal bash 3.6 sec. 35%
minimal BusyBox 0.60.3 5.3 sec. 52%
minimal optimized BusyBox 1.00-pre3 2.6 sec. 25%
minimal + optional bash 7.6 sec. 74%
minimal + optional BusyBox 0.60.3 10.1 sec. 100%
minimal + optional optimized BusyBox 1.00-pre3 4.9 sec. 48%


Total system boot time (kernel loading time + init script execution time):
Script set Shell Time %
minimal bash 4.3 sec. 39%
minimal BusyBox 0.60.3 6.0 sec. 55%
minimal optimized BusyBox 1.00-pre3 3.2 sec. 29%
minimal + optional bash 8.2 sec. 76%
minimal + optional BusyBox 0.60.3 10.7 sec. 100%
minimal + optional optimized BusyBox 1.00-pre3 5.5 sec. 51%


Measurement results without the XIP support Kernel loading time: 0.4 sec. Init script execution time:
Script set Shell Time %
minimal bash 4.4 sec. 52%
minimal BusyBox 0.60.3 5.0 sec. 59%
minimal optimized BusyBox 1.00-pre3 2.9 sec. 34%
minimal + optional bash 6.8 sec. 82%
minimal + optional BusyBox 0.60.3 8.3 sec. 100%
minimal + optional optimized BusyBox 1.00-pre3 4.6 sec. 55%


*Total system boot time (kernel loading time + init script execution time):
Script set Shell Time %
minimal bash 4.8 sec. 55%
minimal BusyBox 0.60.3 5.4 sec. 61%
minimal optimized BusyBox 1.00-pre3 3.3 sec. 38%
minimal + optional bash 7.3 sec. 83%
minimal + optional BusyBox 0.60.3 8.7 sec. 100%
minimal + optional optimized BusyBox 1.00-pre3 5.0 sec. 57%

Downloads