c3249ae0c9
Some systems don't have a hardware real time clock and don't know the time after bootin. An obvious soulution for this is to use time from an ntp server. Unfortunately that requires domain name resolution, which resolvers like unbound won't do for us, if the DNSSEC certificates aren't valid, which they aren't if we start out with a time around 1970-1-1. The "software clock" service tries to provide a workaround by restoring a reasonably valid time from a backup file during boot, which we update when shuting down. If we wan't a more correct time, we have to update it from NTP in between. Signed-off-by: David Oberhollenzer <david.oberhollenzer@tele2.at> |
||
|---|---|---|
| cmd | ||
| docs | ||
| etc | ||
| initd | ||
| lib | ||
| m4 | ||
| netcfg | ||
| scripts | ||
| services | ||
| syslogd | ||
| .gitignore | ||
| autogen.sh | ||
| configure.ac | ||
| LICENSE | ||
| make_a_release.sh | ||
| Makefile.am | ||
| README.md | ||
About
This directory contains the source code for a tiny init devised for the Pygos system.
The main goal of this project is to create a simple framework for:
- system boot up and initialization
- service supervision
With the additional aims of having something that:
- simply works
- is easy to understand
- is easy to configure and maintain
The init process is intended to run on top of Linux and makes use of some Linux specific features (e.g. signalfd), but if sufficient interest exists, it should still be possible to make it run on some BSDs or whatever else.
The init system tries to mimic the concept of unit files from systemd as those were considered to be a good design choice.
Those parameterizeable service description files are stored in /usr/share/init
by default. Services are enabled by creating a symlink in /etc/init.d. This
can be done using the service command line tool.
See docs/services.md for more information on service description files.
See docs/bootup.md for more information on what the init daemon does during system boot.
Right now, the system is in a "basically works" proof of concept stage and needs some more work to become usable.
There are plans for maybe eventually adding support for Linux name spaces, seccomp filters and cgroups as needed in the medium future.
Why
There are already a bunch of similar projects out there that have been considered for use in the Pygos system. The reason for starting a new one was mainly dissatisfaction with the existing ones. Other Projects that have been considered include:
-
systemd
Contains a lot of good ideas, but it is HUGE. It has tons of dependencies. It implements tons of things that it simply shouldn't. It has a horrid, "modern", python based, hipster build system. It's simply too damn large and complex.
-
SystemV init
A bad combination of unnecessary complexity where it isn't needed and a complete lack of abstraction where it would be needed. Shell script copy and paste madness. There are reasons people started developing alternatives (other than "hurr-durr-parallel-boots").
-
upstart
Seems nice overall, but needlessly big and complex for the intended use case in Pygos. Would have needlessly added D-Bus to the system.
-
OpenRC
Was already integrated into Pygos. Things turned out to be broken. Upstream developers did not accept fixes (after ignoring them for weeks and preferring typo fixes instead). Complaints from other people who tried to contribute fixes were observed on GitHub. Complaints from package maintainers about deteriorating code quality were observed on the official IRC channel. Documentation is non-existent.
-
daemon tools and similar (runnit, s6, minit, ...)
The sixties are over. And even code from that era is more readable. The source code for those projects should better be tossed out the window and rewritten from scratch. If you are a first semester CS student and you hand something like this in as a homework, the best you might get is a well deserved slap on the back of your head.
-
busybox init
Nice and simple. Probably the best fit if the rest of your user space is busybox as well.