Skip to content

Latest commit

 

History

History
796 lines (556 loc) · 30 KB

config.md

File metadata and controls

796 lines (556 loc) · 30 KB

Configuration

Introduction

Finit can be configured using only the original /etc/finit.conf file or in combination with /etc/finit.d/*.conf. Useful for package-based Linux distributions -- each package can provide its own "script" file.

  • /etc/finit.conf: main configuration file
  • /etc/finit.d/*.conf: snippets, usually one service per file

Not all configuration directives are available in /etc/finit.d/*.conf and some directives are only available at bootstrap, runlevel S, see the section Limitations below for details.

To add a new service, drop a .conf file in /etc/finit.d and run initctl reload. (It is also possible to SIGHUP to PID 1, or call finit q, but that has been deprecated with the initctl tool). Finit monitors all known active .conf files, so if you want to force a restart of any service you can touch its corresponding .conf file in /etc/finit.d and call initctl reload. Finit handle any and all conditions and dependencies between services automatically.

It is also possible to drop .conf files in /etc/finit.d/available/ and use initctl enable to enable a service .conf file. This may be useful in particular to Linux distributions that may want to install all files for a package and let the user decide when to enable a service.

On initctl reload the following is checked for all services:

  • If a service's .conf file has been removed, or its conditions are no longer satisfied, the service is stopped.
  • If the file is modified, or a service it depends on has been reloaded, the service is reloaded (stopped and started).
  • If a new service is added it is automatically started — respecting runlevels and return values from any callbacks.

For more info on the different states of a service, see the separate document Finit Services.

When running make install no default /etc/finit.conf is installed since system requirements differ too much. There are some examples in the contrib/ directory, which can be used as a base.

Environment Variables

In Finit v4.3 support for setting environment variables in finit.conf, and any *.conf, was added. It is worth noting that these are global and shared with all services -- the only way to have service-local environment is detailed in the next section.

The syntax for global environment variables is straight forward:

foo=bar
baz="qux"

On reload of .conf files, all tracked environment variables are cleared so if foo=bar is removed from finit.conf, or any finit.d/*.conf file, it will no longer be used by Finit or any new (!) started run/tasks or services. The environment of already started processes can not be changed.

The only variables reset to sane defaults on .conf reload are:

PATH=_PATH_STDPATH
SHELL=_PATH_BSHELL
LOGNAME=root
USER=root

It is entirely possible to override these as well from the .conf files, but be careful. Changing SHELL changes the behavior of system() and a lot of other commands as well.

Service Environment

Finit supports sourcing environment variables from /etc/default/*, or similar --with-sysconfig=DIR. This is a common pattern from SysV init scripts, where the start-stop script is a generic script for the given service, foo, and the options for the service are sourced from the file /etc/default/foo. Like this:

  • /etc/default/foo:

      FOO_OPTIONS=--extra-arg="bar" -s -x

  • /etc/finit.conf:

      service [2345] env:-/etc/default/foo foo -n $FOO_OPTIONS -- Example foo daemon

Here the service foo is started with -n, to make sure it runs in the foreground, and the with the options found in the environment file. With the ps command we can see that the process is started with:

foo -n --extra-arg=bar -s -x

Note: the leading - determines if Finit should treat a missing environment file as blocking the start of the service or not. When - is used, a missing environment file does not block the start.

Service Wrapper Scripts

If your service requires to run additional commands, executed before the service is actually started, like the systemd ExecStartPre, you can use a wrapper shell script to start your service.

The Finit service .conf file can be put into /etc/finit.d/available, so you can control the service using initctl. Then use the path to the wrapper script in the Finit .conf service stanza. The following example employs a wrapper script in /etc/start.d.

Example:

  • /etc/finit.d/available/program.conf:

      service [235] <!> /etc/start.d/program -- Example Program

  • /etc/start.d/program:

      #!/bin/sh
  # Prepare the command line options
  OPTIONS="-u $(cat /etc/username)"

  # Execute the program
  exec /usr/bin/program $OPTIONS

Note: the example sets <!> to denote that it doesn't support SIGHUP. That way Finit will stop/start the service instead of sending SIGHUP at restart/reload events.

Cgroups

There are three major cgroup configuration directives:

  1. Global top-level group: init, system, user, or a custom group
  2. Selecting a top-level group for a set of run/task/services
  3. Per run/task/service limits

Top-level group configuration.

# Top-level cgroups and their default settings.  All groups mandatory
# but more can be added, max 8 groups in total currently.  The cgroup
# 'root' is also available, reserved for RT processes.  Settings are
# as-is, only one shorthand 'mem.' exists, other than that it's the
# cgroup v2 controller default names.
cgroup init   cpu.weight:100
cgroup user   cpu.weight:100
cgroup system cpu.weight:9800

Adding an extra cgroup maint/ will require you to adjust the weight of the above three. We leave init/ and user/ as-is reducing weight of system/ to 9700.

cgroup system cpu.weight:9700

# Example extra cgroup 'maint'
cgroup maint  cpu.weight:100

By default, the system/ cgroup is selected for almost everything. The init/ cgroup is reserved for PID 1 itself and its closest relatives. The user/ cgroup is for local TTY logins spawned by getty.

To select a different top-level cgroup, e.g. maint/, one can either define it for a group of run/task/service directives in a .conf or per each stanza:

cgroup.maint
service [...] <...> /path/to/foo args -- description
service [...] <...> /path/to/bar args -- description

or

service [...] <...> cgroup.maint /path/to/foo args -- description

The latter form also allows per-stanza limits on the form:

service [...] <...> cgroup.maint:cpu.max:10000,mem.max:655360 /path/to/foo args -- description

Notice the comma separation and the mem. exception to the rule: every cgroup setting maps directly to cgroup v2 syntax. I.e., cpu.max maps to the file /sys/fs/cgroup/maint/foo/cpu.max. There is no filtering, except for expanding the shorthand mem. to memory., if the file is not available, either the cgroup controller is not available in your Linux kernel, or the name is misspelled.

Linux cgroups and details surrounding values are not explained in the Finit documentation. The Linux admin-guide cover this well: https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html

Configuration File Syntax

Hostname

Syntax: host <NAME>, or hostname <NAME>

Set system hostname to NAME, unless /etc/hostname exists in which case the contents of that file is used.

Deprecated. We recommend using /etc/hostname instead.

Note: only read and executed in runlevel S (bootstrap).

Kernel Modules

Syntax: module <MODULE> [ARGS]

Load a kernel module, with optional arguments. Similar to insmod command line tool.

Deprecated, there is both a modules-load.so and a modprobe.so plugin that can handle module loading better. The former supports loading from /etc/modules-load.d/, the latter uses kernel modinfo to automatically load (or coldplug) every required module. For hotplug we recommend the BusyBox mdev tool, add to /etc/mdev.conf:

$MODALIAS=.*  root:root       0660    @modprobe -b "$MODALIAS"

Note: only read and executed in runlevel S (bootstrap).

Networking

Syntax: network <PATH>

Script or program to bring up networking, with optional arguments.

Deprecated. We recommend using dedicated task/run stanzas per runlevel, or /etc/network/interfaces if you have a system with ifupdown, like Debian, Ubuntu, Linux Mint, or an embedded BusyBox system.

Note: only read and executed in runlevel S (bootstrap).

Resource Limits

Syntax: rlimit [hard|soft] RESOURCE <LIMIT|unlimited>

Set the hard or soft limit for a resource, or both if that argument is omitted. RESOURCE is the lower-case RLIMIT_ string constants from setrlimit(2), without prefix. E.g. to set RLIMIT_CPU, use cpu.

LIMIT is an integer that depends on the resource being modified, see the man page, or the kernel /proc/PID/limits file, for details. Finit versions before v3.1 used infinity for unlimited, which is still supported, albeit deprecated.

# No process is allowed more than 8MB of address space
rlimit hard as 8388608

# Core dumps may be arbitrarily large
rlimit soft core infinity

# CPU limit for all services, soft & hard = 10 sec
rlimit cpu 10

rlimit can be set globally, in /etc/finit.conf, or locally per each /etc/finit.d/*.conf read. I.e., a set of task/run/service stanzas can share the same rlimits if they are in the same .conf.

Runlevels

Syntax: runlevel <N>

The system runlevel to go to after bootstrap (S) has completed. N is the runlevel number 0-9, where 6 is reserved for reboot and 0 for halt.

Default: 2

Note: only read and executed in runlevel S (bootstrap).

One-shot Commands (sequence)

Syntax: run [LVLS] <COND> /path/to/cmd ARGS -- Optional description

One-shot command to run in sequence when entering a runlevel, with optional arguments and description.

run commands are guaranteed to be completed before running the next command. Highly useful if true serialization is needed.

<COND> is described in the Services section.

One-shot Commands (parallel)

Syntax: task [LVLS] <COND> /path/to/cmd ARGS -- Optional description

One-shot like 'run', but starts in parallel with the next command.

Both run and task commands are run in a shell, so pipes and redirects can be freely used:

task [s] echo "foo" | cat >/tmp/bar

<COND> is described in the Services section.

SysV Init Scripts

Syntax: sysv [LVLS] <COND> /path/to/init-script -- Optional description

Similar to task is the sysv stanza, which can be used to call SysV style scripts. The primary intention for this command is to be able to re-use much of existing setup and init scripts in Linux distributions.

When entering an allowed runlevel, Finit calls init-script start, when entering a disallowed runlevel, Finit calls init-script stop, and if the Finit .conf, where sysv stanza is declared, is modified, Finit calls init-script restart on initctl reload. Similar to how service stanzas work.

Forking services started with sysv scripts can be monitored by Finit by declaring the PID file to look for: pid:!/path/to/pidfile.pid. Notice the leading !, it signifies Finit should not try to create the file, but rather watch that file for the resulting forked-off PID. This syntax also works for forking daemons that do not have a command line option to run it in the foreground, more on this below in service.

<COND> is described in the Services section.

Services

Syntax: service [LVLS] <COND> /path/to/daemon ARGS -- Optional description

Service, or daemon, to be monitored and automatically restarted if it exits prematurely. Finit tries to restart services that die, by default 10 times before giving up and marking them as crashed. After which they have to be manually restarted with initctl restart NAME. The limits controling this are configurable, see the options below.

Tip: to allow endless restarts, see below option respawn

For daemons that support it, we recommend appending --foreground, --no-background, -n, -F, or similar command line argument to prevent them from forking off a sub-process in the background. This is the most reliable way to monitor a service.

However, not all daemons support running in the foreground, or they may start logging to the foreground as well, these are forking daemons and are supported using the same syntax as forking sysv services, using the pid:!/path/to/pidfile.pid syntax. There is an alternative syntax that may be more intuitive, where Finit can also guess the PID file based on the daemon's command name:

service type:forking ntpd -- NTP daemon

This example lets BusyBox ntpd daemonize itself. Finit uses the basename of the binary to guess the PID file to watch for the PID: /var/run/ntpd.pid. If Finit guesses wrong, you have to submit the full pid:!/path/to/file.pid.

Example:

In the case of ospfd (below), we omit the -d flag (daemonize) to prevent it from forking to the background:

service [2345] <pid/zebra> /sbin/ospfd -- OSPF daemon

[2345] denote the runlevels ospfd is allowed to run in, they are optional and default to level 2-5 if omitted.

<...> is the condition for starting ospfd. In this example Finit waits for another service, zebra, to have created its PID file in /var/run/quagga/zebra.pid before starting ospfd. Finit watches all files in /var/run, for each file named *.pid, or */pid, Finit opens it and find the matching NAME:ID using the PID.

Some services do not maintain a PID file and rather than patching each application Finit provides a workaround. A pid keyword can be set to have Finit automatically create (when starting) and later remove (when stopping) the PID file. The file is created in the /var/run directory using the basename(1) of the service. The default can be modified with an optional pid:-argument:

pid[:[/path/to/]filename[.pid]]

For example, by adding pid:/run/foo.pid to the service /sbin/bar, that PID file will, not only be created and removed automatically, but also be used by the Finit condition subsystem. So a service/run/task can depend on <pid/bar>.

For a detailed description of conditions, and how to debug them, see the Finit Conditions document.

If a service should not be automatically started, it can be configured as manual with the optional manual argument. The service can then be started at any time by running initctl start <service>.

manual:yes

The name of a service, shown by the initctl tool, defaults to the basename of the service executable. It can be changed with the optional name argument:

name:<service-name>

As mentioned previously, services are automatically restarted, this is configurable with the following options:

  • restart:NUM -- number of times Finit tries to restart a crashing service, default: 10, max: 255. When this limit is reached the service is marked crashed and must be restarted manually with initctl restart NAME
  • restart_sec:SEC -- number of seconds before Finit tries to restart a crashing service, default: 2 seconds for the first five retries, then back-off to 5 seconds. The maximum of this configured value and the above (2 and 5) will be used
  • norestart -- dont restart on failures, same as restart:0
  • respawn -- bypasses the restart mechanism completely, allows endless restarts. Useful in many use-cases, but not what service was originally designed for so not the default behavior
  • oncrash:reboot -- when all retries have failed, and the service has crashed, if this option is set the system is rebooted. Note, future releases may include other oncrash: actions

When stopping a service (run/task/sysv/service), either manually or when moving to another runlevel, Finit starts by sending SIGTERM, to allow the process to shut down gracefully. If the process has not been collected within 3 seconds, Finit sends SIGKILL. To halt the process using a different signal, use the option halt:SIGNAL, e.g., halt:SIGPWR. To change the delay between your halt signal and KILL, use the option kill:SEC, e.g., kill:10 to wait 10 seconds before sending SIGKILL.

Services support pre:script and post:script actions as well. These run as the same @USER:GROUP as the service itself, with any env:file sourced. The scripts must use an absolute path, but are executed from the $HOME of the given user. The scripts are not called with any argument (currently), but both get the SERVICE_IDENT=foo environment variable set. Here foo denotes the identity of the service, which if there are multiple services named foo, may be foo:1, or any unique identifier specified in the .conf file. The post:script is called with an additional set of environment variables:

  • EXIT_CODE=[exited,signal]: set to one of exited or signal
  • EXIT_STATUS=[num,SIGNAME]: set to one of exit status code from the program, if it exited normally, or the signal name (HUP, TERM, etc.) if it exited due to signal

The scripts have a default execution time of 3 seconds before they are SIGKILLed, this can be adjusted using the above kill:SEC syntax.

Note: the pre:script must be idempotent, because a service can transition between READY and HALTED states any number of times before going to RUNNING.

Run-parts Scripts

Syntax: runparts <DIR>

Call run-parts(8) on DIR to run start scripts. All executable files, or scripts, in the directory are called, in alphabetic order. The scripts in this directory are executed at the very end of runlevel S, bootstrap.

It can be beneficial to use S01name, S02othername, etc. if there is a dependency order between the scripts. Symlinks to existing daemons can talso be used, but make sure they daemonize by default.

Similar to the /etc/rc.local shell script, make sure that all your services and programs either terminate or start in the background or you will block Finit.

Note: only read and executed in runlevel S (bootstrap).

Including Finit Configs

Syntax: include <CONF>

Include another configuration file. Absolute path required.

General Logging

Syntax: log size:200k count:5

Log rotation for run/task/services using the log sub-option with redirection to a log file. Global setting, applies to all services.

The size can be given as bytes, without a specifier, or in k, M, or G, e.g. size:10M, or size:3G. A value of size:0 disables log rotation. The default is 200k.

The count value is recommended to be between 1-5, with a default 5. Setting count to 0 means the logfile will be truncated when the MAX size limit is reached.

TTYs and Consoles

Syntax: tty [LVLS] <COND> DEV [BAUD] [noclear] [nowait] [nologin] [TERM]
tty [LVLS] <COND> CMD <ARGS> [noclear] [nowait]
tty [LVLS] <COND> [notty] [rescue]

The first variant of this option uses the built-in getty on the given TTY device DEV, in the given runlevels. DEV may be the special keyword @console, which is expanded from /sys/class/tty/console/active, useful on embedded systems.

The default baud rate is 0, i.e., keep kernel default.

The tty stanza inherits runlevel, condition (and other feature) parsing from the service stanza. So TTYs can run in one or many runlevels and depend on any condition supported by Finit. This is useful e.g. to depend on <pid/elogind> before starting a TTY.

Example:

tty [12345] /dev/ttyAMA0 115200 noclear vt220

The second tty syntax variant is for using an external getty, like agetty or the BusyBox getty.

The third variant is for board bringup and the rescue boot mode. No device node is required in this variant, the same output that the kernel uses is reused for stdio. If the rescue option is omitted, a shell is started (nologin, noclear, and nowait are implied), if the rescue option is set the bundled /libexec/finit/sulogin is started to present a bare-bones root login prompt. If the root (uid:0, gid:0) user does not have a password set, no rescue is possible. For more information, see the Rescue Mode section.

By default, the first two syntax variants clear the TTY and wait for the user to press enter before starting getty.

Example:

tty [12345] /sbin/getty  -L 115200 /dev/ttyAMA0 vt100
tty [12345] /sbin/agetty -L ttyAMA0 115200 vt100 nowait

The noclear option disables clearing the TTY after each session. Clearing the TTY when a user logs out is usually preferable.

The nowait option disables the press Enter to activate console message before actually starting the getty program. On small and embedded systems running multiple unused getty wastes both memory and CPU cycles, so wait is the preferred default.

The nologin option disables getty and /bin/login, and gives the user a root (login) shell on the given TTY <DEV> immediately. Needless to say, this is a rather insecure option, but can be very useful for developer builds, during board bringup, or similar.

Notice the ordering, the TERM option to the built-in getty must be the last argument.

Embedded systems may want to enable automatic DEV by supplying the special @console device. This works regardless weather the system uses ttyS0, ttyAMA0, ttyMXC0, or anything else. Finit figures it out by querying sysfs: /sys/class/tty/console/active. The speed can be omitted to keep the kernel default.

Most systems get by fine by just using console, which will evaluate to /dev/console. If you have to use @console to get any output, you may have some issue with your kernel config.

Example:

tty [12345] @console noclear vt220

On really bare bones systems, or for board bringup, Finit can give you a shell prompt as soon as bootstrap is done, without opening any device node:

tty [12345789] notty

This should of course not be enabled on production systems. Because it may give a user root access without having to log in. However, for board bringup and system debugging it can come in handy.

One can also use the service stanza to start a stand-alone shell:

service [12345] /bin/sh -l

Non-privileged Services

Every run, task, or service can also list the privileges the /path/to/cmd should be executed with. Prefix the command with @USR[:GRP], group is optional, like this:

run [2345] @joe:users logger "Hello world"

For multiple instances of the same command, e.g. a DHCP client or multiple web servers, add :ID somewhere between the run, task, service keyword and the command, like this:

service :80  [2345] httpd -f -h /http -p 80   -- Web server
service :8080[2345] httpd -f -h /http -p 8080 -- Old web server

Without the :ID to the service the latter will overwrite the former and only the old web server would be started and supervised.

Redirecting Output

The run, task, and service stanzas also allow the keyword log to redirect stderr and stdout of the application to a file or syslog using the native logit tool. This is useful for programs that do not support syslog on their own, which is sometimes the case when running in the foreground.

The full syntax is:

log:/path/to/file
log:prio:facility.level,tag:ident
log:console
log:null
log

Default prio is daemon.info and default tag is the basename of the service or run/task command.

Log rotation is controlled using the global log setting.

Example:

service log:prio:user.warn,tag:ntpd /sbin/ntpd pool.ntp.org -- NTP daemon

Rescue Mode

Finit supports a rescue mode which is activated by the rescue option on the kernel command line. See cmdline docs for how to activate it.

The rescue mode comes in two flavors; traditional and fallback.

Note: in this mode initctl will not work. Use the -f flag to force reboot, shutdown, or poweroff.

Traditional

This is what most users expect. A very early maintenance login prompt, served by the bundled /libexec/finit/sulogin program, or the standard sulogin from util-linux or BusyBox is searched for in the UNIX default $PATH. If a successful login is made, or the user exits (Ctrl-D), the rescue mode is ended and the system boots up normally.

Note: if the user (UID 0 and GID 0) does not have a password, or the account is locked, the user is presented with a password-less "Press enter to enter maintenance mode.", prompt which opens up a root shell.

Fallback

If no sulogin program is found, Finit tries to bring up as much of its own functionality as possible, yet limiting many aspects, meaning; no network, no fsck of file systems in /etc/fstab, no /etc/rc.local, no runparts, and most plugins are skipped (except those that provide functionality for the condition subsystem).

Instead of reading /etc/finit.conf et al, system configuration is read from /lib/finit/rescue.conf, which can be freely modified by the system administrator.

The bundled default rescue.conf contains nothing more than:

runlevel 1
tty [12345] rescue

The tty has the rescue option set, which works similar to the board bring-up tty option notty. The major difference being that sulogin is started to query for root/admin password. If sulogin is not found, rescue behaves like notty and gives a plain root shell prompt.

If Finit cannot find /lib/finit/rescue.conf it defaults to:

tty [12345] rescue

There is no way to exit the fallback rescue mode.

Limitations

As of Finit v4 there are no limitations to where .conf settings can be placed. Except for the system/global rlimit and cgroup top-level group declarations, which can only be set from /etc/finit.conf, since it is the first .conf file Finit reads.

Originally, /etc/finit.conf was the only way to set up a Finit system. Today it is mainly used for bootstrap settings like system hostname, early module loading for watchdogd, network bringup and system shutdown. These can now also be set in any .conf file in /etc/finit.d.

There is, however, nothing preventing you from having all configuration settings in /etc/finit.conf.

Note: The /etc/finit.d directory was previously the default Finit runparts directory. Finit no longer has a default runparts, make sure to update your setup, or the finit configuration, accordingly.

Watchdog

When built --with-watchdog a separate service is built and installed in /libexec/finit/watchdogd. If this exists at runtime, and the WDT device node exists, Finit will start it and treat it as the elected watchdog service to delegate its reboot to. See bootstrap for details. This delegation is to ensure that the system is rebooted by a hardware watchdog timer -- on many embedded systems this is crucial to ensure all circuits on the board are properly reset for the next boot, in effect ensuring the system works the same after both a power-on and reboot event.

The delegation is performed at the very last steps of system shutdown, if reboot has been selected and an elected watchdog is known, first a SIGPWR is sent to advise watchdogd of the pending reboot. Then, when the necessary steps of preparing the system for shutdown (umount etc.) are completed, Finit sends SIGTERM to watchdogd and puts itself in a 10 sec timeout loop waiting for the WDT to reset the board. If a reset is not done before the timeout, Finit falls back toreboot(RB_AUTOBOOT) which tells the kernel to do the reboot.

An external watchdog service can also be used. The more advanced cousin watchdogd is the recommended option here. It can register itself with Finit using the same IPC as initctl. If the bundled watchdogd is running a hand-over takes place, so it's safe to have both services installed on a system. For the hand-over to work it requires that the WDT driver supports the safe exit functionality where "V" is written to the device before closing the device descriptor. If the kernel driver has been built without this, the only option is to remove /libexec/finit/watchdogd or build without it at configure time.

keventd

The kernel event daemon bundled with Finit is a simple uevent monitor for /sys/class/power_supply. It provides the sys/pwr/ac condition, which can be useful to prevent power hungry services like anacron to run when a laptop is only running on battery, for instance.

Since keventd is not an integral part of Finit yet it is not enabled by default. Enable it using ./configuure --with-keventd. The bundled contrib build scripts for Debian, Alpine, and Void have this enabled.

This daemon is planned to be extended with monitoring of other uevents, patches and ideas are welcome in the issue tracker.