Re: single user v multiuser boot

From: Warner Losh <imp_at_bsdimp.com>
Date: Fri, 18 Feb 2005 13:25:23 -0700 (MST)
From: John-Mark Gurney <gurney_j_at_resnet.uoregon.edu>
Subject: Re: single user v multiuser boot
Date: Fri, 18 Feb 2005 00:04:23 -0800

> Alex Burke wrote this message on Fri, Feb 18, 2005 at 00:33 +0000:
> > I was wondering what gives the kernel the ability to boot in multiuser
> > mode, and whether it is some code in the kernel or whether it is the
> > init process and associated tools?
> 
> single user usually means that you have enough system that you have
> a console running and possibly some disk device, though you can end
> up using md as part of the kernel, and not require any real disk
> devices to work...
> 
> multiuser usually implies that either disk or network is functional
> to some degree where /etc/rc can start executing...
> 
> single user pretty much only requires /sbin/init and /bin/sh to "get"
> to...

One can get to single user mode without a fully working vm system.
One can get to single user mode without having process termination
working or image rundown working.  Multiuser also requires that the
various synchronization primitives be fully functional.  So things
like fork/exec don't have to work completely.  Back when /sbin/init
and /bin/sh were statically linked, getting to single user didn't even
require mmap to work correctly.

Plus, in multi-user mode, jmg is right: networking needs to work,
along with generally a much higher load on the interrupt system, the
memory system, etc.   It also exposes races in the system more than
single user mode does (and sometimes a LOT more).  In some systems, it
can expose cache coherency problems, tlb lookup issues with multiple
processes, etc.

Warner
Received on Fri Feb 18 2005 - 19:28:11 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:28 UTC