This file will contain a list of notable changes for each release. Note the information in this file is in reverse order.
This release drops the support for the split-usr build option. Also, it drops the support for ncurses/termcap and uses ansi codes directly for color support on terminals that support them.
The names of cgroups for services started by OpenRC are now prefixed by "openrc." This is done because some services, like docker, create their own cgroups.
It is now possible to override the mount options used to mount the cgroups filesystems.
This release drops the "runscript" and "rc" binaries. These binaries have been deprecated for multiple years, so it should be fine to remove them at this point.
There was an issue before this release with the default setting for cgroups being inconsistent. This is fixed.
Start-stop-daemon did not work correctly on Linux 6.6 systems. This has been fixed in this release as well.
The default RC_CGROUP_MODE has been updated to unified. This benefits users since it will assign each service to its own cgroup, making resource nanagement better over all.
OUTPUT_LOGGER and ERROR_LOGGER have been implemented for supervise-daemon. For mor einformation on these settings, please check the man page.
This is a bug fix release which fixes a significant performance issue on musl libc systems.
This release adds support for glibc's builtin strlcpy, strlcat etc functions, which will be in posix next. Also, it fixes completions.
This release is a maintenance release; it has no user-facing changes.
This release is primarily an internal cleanup release. The only user-visible difference is that unicode is now on by default.
The path for the reference file for the swclock service is now configurable in conf.d/swclock.
In the past, if supervise_daemon_args was not set or empty, it defaulted to
start_stop_daemon_args. This was bad because supervise-daemon doesn't
accept the same options as start-stop-daemon. So if we set e.g.
start_stop_daemon_args="--wait 50", but not supervise_daemon_args,
and the user adds supervisor=supervise-daemon to the corresponding
conf.d/ file, the service will fail to start due to
unrecognized option "wait".
It would be best to remove this fallback, but that might break some
existing scripts that depend on it. So we are changing it to
use start_stop_daemon_args as the default for supervise_daemon_args
only if supervise_daemon_args is not set at all, but not if it's
empty.
This fallback will be dropped in a future release.
The old make-based build system is removed in this release.
The killprocs service now has a KILL_DELAY setting to allow OpenRC based containers to close all TCP/IP connections before they are shut down.
The --oom-score-adj option has been added to start-stop-daemon and supervise-daemon. This sets the appropriate setting for the Linux kernel. for more info, see the man pages.
Support for Linux capabilities has been added to both supervise-daemon and start-stop-daemon using the --capabilities option. As a result, the user can specify the inheritable, ambient and bounding set by defining capabilities in the service script.
noexec has been added to the default mount options for /dev. If you need to remove this option, add an entry to fstab.
A --secbits option has been added to start-stop-daemon and supervise-daemon. This sets the security bits option which should be applied to the daemon.
A no-new-privs option has been added to start-stop-daemon and supervise-daemon. This sets the NO_NEW_PRIVS flag to apply to the daemon.
On Linux, the service that seeds the random number generator has been renamed from urandom to seedrng. This means that when you upgrade to this version, if urandom is in your boot runlevel, you must add seedrng. This can be done by issuing the following commands:
# rc-update add seedrng boot
# rc-update del urandom boot
This version is the first to use a meson-based build system. The gnu make build system should now be considered deprecated; it will be removed sometime after 0.44.x.
I have opened a discussion where you can voice concerns about removing it.
This version changes the behavior of the checkpath helper to address CVE-2018-21269. on Linux systems, We require non-terminal symbolic links to be owned by root. Since we can't do this on non-linux systems, we do not dereference non-terminal symbolic links by default. If you need them dereferenced, you should add the "-s" switch to the appropriate checkpath calls. For more information, see http://github.com/openrc/openrc/issues/201.
The SHLIBDIR variable has been removed from the makefiles to make them more consistent with most common makefiles. All libraries are now in LIBDIR, so if you need to put them in /, override the LIBDIR variable when you run make.
openrc-shutdown now has the ability to shut down sysvinit-based systems.
A guide has been added for migrating systems using another init system to openrc-init.
This version adds the ability to format the output of rc-status when showing the status of services in a runlevel so that it may be parsed. Currently, the -f switch only accepts ini as an argument which causes the output to be in the .ini format.
This version adds an experimental build time switch to allow setting the default shell to use for service scripts. By default, this is set to /bin/sh if it is changed, the new shell must be able to understand posix-compatible syntax.
In this version, the keymaps and termencoding services on Linux needed to be modified so they do not write to the root file system. This was done so they can run earlier in the boot sequence. AS a result, you will need to add save-termencoding and save-keymaps to your boot runlevel. This can be done as follows:
# rc-update add save-keymaps boot
# rc-update add save-termencoding boot
This version removes the support for addons. The only place I know that this was used was Gentoo Baselayout 1.x, so it shouldn't affect anyone since baselayout-1 has been dead for a few years.
Since all supported Linux kernel versions now make efivarfs immutable and all of the tools that access efivarfs are aware of this, we no longer mount efivarfs read-only. See the following github issue for more information:
This version adds timed shutdown and cancelation of shutdown to openrc-shutdown. Shutdowns can now be delayed for a certain amount of time or scheduled for an exact time.
supervise-daemon supports health checks, which are a periodic way to make sure a service is healthy. For more information on setting this up, please see supervise-daemon-guide.md.
The --first-time switch has been added to all modprobe commands in the modules service. This means that, on Linux, you will see failures if a module was loaded by an initramfs or device manager before this service runs. These messages are harmless, but to clean them up, you should adjust your modules autoload configuration.
start-stop-daemon now supports logging stdout and stderr of daemons to processes instead of files. These processes are defined by the output_logger and error_logger variables in standard service scripts, or by the -3/--output-logger or -4/--error-logger switches if you use start-stop-daemon directly. For more information on this, see the start-stop-daemon man page.
In this release, the modules-load service has been combined into the modules service since there is no reason I know of to keep them separate. However, modules also provides modules-load in case you were using modules-load in your dependencies.
The consolefont, keymaps, numlock and procfs service scripts no longer have a dependency on localmount. If you are a linux user and are still separating / from /usr, you will need to add the following line to the appropriate conf.d files:
rc_need="localmount"
In this version, the cgroups mounting logic has been moved from the sysfs service to the cgroups service. This was done so cgroups can be mounted inside an lxc/lxd container without using the other parts of the sysfs service.
?As a result of this change, if you are upgrading, you need to add cgroups to your sysinit runlevel by running the following command as root:
# rc-update add cgroups sysinit
For more information, see the following issue:
Consider this your second notification with regard to /etc/mtab being a file instead of a symbolic link.
In this version, the mtab service will complain loudly if you have mtab_is_file set to yes and recommend that you change this to no and restart the mtab service to migrate /etc/mtab to a symbolic link.
If there is a valid technical reason to keep /etc/mtab as a flat file instead of a symbolic link to /proc/self/mounts, we are interested and we will keep the support in that case. Please open an issue and let us know however. Otherwise, consider this your final notice that the mtab service will be removed in the future.
This version removes the "service" binary which was just a copy of "rc-service" provided for compatibility.
If you still need the "service" binary, as opposed to "rc-service", it is recommended that you use something like Debian's init-system-helpers. Otherwise, just use "rc-service" in place of "service".
This version adds support for Control Groups version 2, which is considered stable as of Linux-4.13. Please see /etc/rc.conf for documentation on how to configure control groups.
This version mounts efivars read only due to concerns about changes in this file system making systems unbootable. If you need to change something in this path, you will need to re-mount it read-write, make the change and re-mount it read-only.
Also, you can override this behavior by adding a line for efivars to fstab if you want efivars mounted read-write.
For more information on this issue, see the following url:
This version contains an OpenRC-specific implementation of init for Linux which can be used in place of sysvinit or any other init process. For information on its usage, see the man pages for openrc-init (8) and openrc-shutdown (8).
This version starts cleaning up the dependencies so that rc_parallel will work correctly.
The first step in this process is to remove the 'before *' from the depend functions in the clock services. This means some services not controlled by OpenRC may now start before instead of after the clock service. If it is important for these services to start after the clock service, they need to have 'after clock' added to their depend functions.
Since the deptree2dot tool and the perl requirement are completely optional, the deptree2dot tool has been moved to the support directory. As a result, the MKTOOLS=yes/no switch has been removed from the makefiles.
This version adds the agetty service which can be used to spawn agetty on a specific terminal. This is currently documented in the agetty-guide.md file at the top level of this distribution.
The tmpfiles.d processing code, which was part of previous versions of OpenRC, has been separated into its own package [1]. If you need to process systemd style tmpfiles.d files, please install this package.
[1] https://github.com/openrc/opentmpfiles
In previous versions of OpenRC, configuration information was processed so that service-specific configuration stored in /etc/conf.d/* was overridden by global configuration stored in /etc/rc.conf. This release reverses that. Global configuration is now overridden by service-specific configuration.
The swapfiles service, which was basically a copy of the swap service, has been removed. If you are only using swap partitions, this change will not affect you. If you are using swap files, please adjust the dependencies of the swap service as shown in /etc/conf.d/swap.
This version adds a daemon supervisor which can start daemons and restart them if they crash. See supervise-daemon-guide.md in the distribution for details on its use.
It is now possible to mark certain mount points as critical. If these mount points are unable to be mounted, localmount or netmount will fail. This is handled in /etc/conf.d/localmount and /etc/conf.d/netmount. See these files for the setup.
The deprecation messages in 0.13.x for runscript and rc are now made visible in preparation for the removal of these binaries in 1.0.
The steps you should take to get rid of these warnings is to run openrc in initialization steps instead of rc and change the shebang lines in service scripts to refer to "openrc-run" instead of "runscript".
In 0.21.4, a modules-load service was added. This works like the equivalent service in systemd. It looks for files named *.conf first in /usr/lib/modules-load.d, then /run/modules-load.d, then /etc/modules-load.d. These files contain a list of modules, one per line, which should be loaded into the kernel. If a file name appears in /run/modules-load.d, it overrides a file of the same name in /usr/lib/modules-load.d. A file appearing in /etc/modules-load.d overrides a file of the same name in both previous directories.
This version adds a net-online service. By default, this service will check all known network interfaces for a configured interface or a carrier. It will register as started only when all interfaces are configured and there is at least a carrier on one interface. The behaviour of this service can be modified in /etc/conf.d/net-online.
Currently, this only works on Linux, but if anyone wants to port to *bsd, that would be welcomed.
Modern Linux systems expect /etc/mtab to be a symbolic link to /proc/self/mounts. Reasons for this change include support for mount namespaces, which will not work if /etc/mtab is a file. By default, the mtab service enforces this on each reboot.
If you find that this breaks your system in some way, please do the following:
-
Set mtab_is_file=yes in /etc/conf.d/mtab.
-
Restart mtab. This will recreate the /etc/mtab file.
-
Check for an issue on https://github.com/openrc/openrc/issues explaining why you need /etc/mtab to be a file. If there isn't one, please open one and explain in detail why you need this to be a file. If there is one, please add your comments to it. Please give concrete examples of why it is important that /etc/mtab be a file instead of a symbolic link. Those comments will be taken into consideration for how long to keep supporting mtab as a file or when the support can be removed.
The behaviour of localmount and netmount in this version is changing. In the past, these services always started successfully. In this version, they will be able to fail if file systems they mount fail to mount. If you have file systems listed in fstab which should not be mounted at boot time, make sure to add noauto to the mount options. If you have file systems that you want to attempt to mount at boot time but failure should be allowed, add nofail to the mount options for these file systems in fstab.
The binfmt service, which registers misc binary formats with the Linux kernel, has been separated from the procfs service. This service will be automatically added to the boot runlevel for new Linux installs. When you upgrade, you will need to use rc-update to add it to your boot runlevel.
The procfs service no longer automounts the deprecated usbfs and usbdevfs file systems. Nothing should be using usbdevfs any longer, and if you still need usbfs it can be added to fstab.
Related to the above change, the procfs service no longer attempts to modprobe the usbcore module. If your device manager does not load it, you will need to configure the modules service to do so.
The override order of binfmt.d and tmpfiles.d directories has been changed to match systemd. Files in /run/binfmt.d and /run/tmpfiles.d override their /usr/lib counterparts, and files in the /etc counterparts override both /usr/lib and /run.
A chroot variable has been added to the service script variables. This fixes the support for running a service in a chroot. This is documented in man 8 openrc-run.
The netmount service now mounts nfs file systems. This change was made to correct a fix for an earlier bug.
/sbin/rc was renamed to /sbin/openrc and /sbin/runscript was renamed to /sbin/openrc-run due to naming conflicts with other software.
Backward compatible symbolic links are currently in place so your system will keep working if you are using the old names; however, it is strongly advised that you migrate to the new names because the symbolic links will be removed in the future. Warnings have been added to assist with this migration; however, due to the level of noise they produce, they only appear in verbose mode in this release.
The devfs script now handles the initial mounting and setup of the /dev directory. If /dev has already been mounted by the kernel or an initramfs, devfs will remount /dev with the correct mount options instead of mounting a second /dev over the existing mount point.
It attempts to mount /dev from fstab first if an entry exists there. If it doesn't it attempts to mount devtmpfs if it is configured in the kernel. If not, it attempts to mount tmpfs. If none of these is available, an error message is displayed and static /dev is assumed.
The net.* scripts, originally from Gentoo Linux, have been removed. If you need these scripts, look for a package called netifrc, which is maintained by them.