"Ben Dibell" <thinkingrod...@gmail.com> wrote: > Hi, I've tried other resources, even reading the source for init, but I > can't seem to locate the magic that makes /sbin/init the approved init. > I'm porting my init system Epoch to BSD for personal reasons, and I'd like > it to work under OpenBSD, which I've been enjoying as of late. I come from > the linux world where init=/bin/sh is perfectly valid, so some aspects are > probably simpler in Linux. I am hoping there is a concise and clean > explanation as to how to write/port an init system to BSD. Is it signal > trickery? A checksum burned into the kernel? I'm lost. I'm given "init has > died, signal 0 exit 0" or something nearly identical to this. There are no > further useful debug messages and my keyboard becomes unresponsive on the > debugging prompt or the kernel locks up or something, so I can't do more > there. > > Thanks for your time. > > -Ben
The kernel doesn't support any command line syntax like that but I don't know why it wouldn't let you use any static binary. You should read kernel sources too if you're serious about this. But I don't see your point. Obviously you can do what makes you happy on your systems but have you thought about what OpenBSD init does vs. what your init does? init (among other things) exists to run getty and start /etc/rc. You mention status in a further message. You probably want to keep logs and manage daemon state like the other newfangled init systems. You're going to have to rewrite both init and the rc system. And you won't get any support for such a system. And if you want to monitor daemons you'll be better off monitoring the service the daemons are supposed to provide. It doesn't matter if httpd hasn't exited yet if you can't connect to it. - Martin Brandenburg
