SYSTEM(4) File Formats and Configurations SYSTEM(4)

system - system configuration information file

The system file is used for customizing the operation of the operating system kernel. The recommended procedure is to preserve the original system file before modifying it.
It is not recommended to edit the /etc/system file directly but rather to deliver configuration fragments into files under /etc/system.d; files in this directory are combined in alphabetical order and read by the kernel before /etc/system is processed. Directives in /etc/system therefore take precedence over any settings delivered in fragment files.
The recommended naming schema for the fragment files is to use the name of the package which is delivering the file with ' /' characters replaced by ' :'; file names that start with a dot (.) will be ignored.
If /etc/system.d/ exists and contains any fragment files, then the directory must also be writable or it will not be possible to create or update the system boot archive.
The system file contains commands which are read by the kernel during initialization and used to customize the operation of your system. These commands are useful for modifying the system's treatment of its loadable kernel modules.
The syntax of the system file consists of a list of keyword/value pairs which are recognized by the system as valid commands. Comment lines must begin with an asterisk ( *) or a hash mark (#) and end with a newline character. All commands are case-insensitive except where noted.
Commands that modify the system's operation with respect to loadable kernel modules require you to specify the module type by listing the module's namespace. The following namespaces are currently supported on all platforms:
drv
Modules in this namespace are device drivers.
exec
Modules in this namespace are execution format modules. The following exec modules are currently provided:
Only on SPARC systems: 

aoutexec

On SPARC and IA systems: 

elfexec
intpexec
javaexec

firmware
Raw firmware images in subdirectories, one for each device driver module using firmload(9F).
fs
These modules are filesystems.
sched
These modules implement a process scheduling algorithm.
strmod
These modules are STREAMS modules.
sys
These modules implement loadable system-call modules.
misc
These modules do not fit into any of the above categories, so are considered "miscellaneous" modules.
SPARC only:
dacf
These modules provide rules and actions for device auto-configuration.
tod
These modules provide support for the time of day hardware.
cpu
These modules provide CPU-specific kernel routines.
A description of each of the supported commands follows:
exclude: <namespace>/<modulename>
Do not allow the listed loadable kernel module to be loaded. exclude commands are cumulative; the list of modules to exclude is created by combining every exclude entry in the system file.
include: <namespace>/<modulename>
Include the listed loadable kernel module. This is the system's default, so using include does not modify the system's operation. include commands are cumulative.
forceload: <namespace>/<modulename>
Force this kernel module to be loaded during kernel initialization. The default action is to automatically load the kernel module when its services are first accessed. forceload commands are cumulative.
rootdev: <device name>
Set the root device to the listed value instead of using the default root device as supplied by the boot program.
rootfs: <root filesystem type>
Set the root filesystem type to the listed value.
moddir: <first module path>[[{:, }<second ...>]...]
Set the search path for loadable kernel modules. This command operates very much like the PATH shell variable. Multiple directories to search can be listed together, delimited either by blank spaces or colons.
set [<module>:]<symbol> {=, |, &} [~][-] <value>
Set an integer or character pointer in the kernel or in the selected kernel module to a new value. This command is used to change kernel and module parameters and thus modify the operation of your system. Assignment operations are not cumulative, whereas bitwise AND and OR operations are cumulative.
Operations that are supported for modifying integer variables are: simple assignment, inclusive bitwise OR, bitwise AND, one's complement, and negation. Variables in a specific loadable module can be targeted for modification by specifying the variable name prefixed with the kernel module name and a colon (:) separator. Values can be specified as hexadecimal (0x10), Octal (046), or Decimal (5).
The only operation supported for modifying character pointers is simple assignment. Static string data such as character arrays cannot be modified using the set command. Use care and ensure that the variable you are modifying is in fact a character pointer. The set command is very powerful, and will likely cause problems if used carelessly. The following escape sequences are supported within the quoted string: 

\n	(newline)
\t	(tab)
\b	(backspace)

Example 1 A sample system file.
The following is a sample system file. 

* Force the ELF exec kernel module to be loaded during kernel
* initialization. Execution type modules are in the exec namespace.
forceload: exec/elfexec
* Change the root device to /sbus@1,f8000000/esp@0,800000/sd@3,0:a.
* You can derive root device names from /devices.
* Root device names must be the fully expanded Open Boot Prom
* device name. This command is platform and configuration specific.
* This example uses the first partition (a) of the SCSI disk at
* SCSI target 3 on the esp host adapter in slot 0 (on board)
* of the SBus of the machine.
* Adapter unit-address 3,0 at sbus unit-address 0,800000.
rootdev: /sbus@1,f8000000/esp@0,800000/sd@3,0:a
* Set the filesystem type of the root to ufs. Note that
* the equal sign can be used instead of the colon.
rootfs:ufs
* Set the search path for kernel modules to look first in
* /usr/phil/mod_test for modules, then in /kernel/modules (the
* default) if not found. Useful for testing new modules.
* Note that you can delimit your module pathnames using
* colons instead of spaces: moddir:/newmodules:/kernel/modules
moddir:/usr/phil/mod_test /kernel/modules.
* Set the configuration option {_POSIX_CHOWN_RESTRICTED} :
* This configuration option is enabled by default.
set rstchown = 1
* Disable the configuration option {_POSIX_CHOWN_RESTRICTED} :
set rstchown = 0
* Turn on debugging messages in the modules mydriver. This is useful
* during driver development.
set mydriver:debug = 1
* Bitwise AND the kernel variable "moddebug" with the
* one's complement of the hex value 0x880, and set
* "moddebug" to this new value.
set moddebug & ~0x880
* Demonstrate the cumulative effect of the SET
* bitwise AND/OR operations by further modifying "moddebug"
* by ORing it with 0x40.
set moddebug | 0x40

boot(1M), init(1M), kernel(1M)

Use care when modifying the system file; it modifies the operation of the kernel. If you preserved the original system file, you can boot using boot -a, which will ask you to specify the path to the saved file. This should allow the system to boot correctly. If you cannot locate a system file that will work, you may specify /dev/null. This acts as an empty system file, and the system will attempt to boot using its default settings.

The system files are read only once, at boot time.
April 20, 2019