Re: r18730 - trunk/packages/rootskel/src/sbin

[Wouter Verhelst <wouter@grep.be>]

On Fri, Jul 30, 2004 at 10:35:24AM +0200, Bastian Blank wrote:
> On Thu, Jul 29, 2004 at 06:58:45PM -0600, Wouter Verhelst wrote:
> > Author: wouter
> > Date: Thu Jul 29 18:58:45 2004
> > New Revision: 18730
> > 
> > Modified:
> >    trunk/packages/rootskel/src/sbin/init
> > Log:
> > Whoops, forgot the most important bit.
> 
> What is the problem with accessing this devices?

Busybox init tries to access those devices over and over again, issueing
an error message every time it fails. Since a VME box doesn't have
anything but a serial console, it fails every time. Therefore, the
interface scrolls away because of the error messages, making d-i quite
useless. I'm not sure whether these error messages will appear when
using a serial console on a kernel which does have support for virtual
consoles (because I didn't test), but since the virtual devices aren't
useful there either, this shouldn't hurt.

There is code in busybox init which tries to access() the devices before
it tries to open() them, and which will avoid opening the devices if the
access() fails. However, this code is only executed once, at startup; if
access() finds the right device file, the error messages will appear.
The only way to make sure access() doesn't find the devices is to remove
them before init is started. I tried different approaches, even
rewriting inittab and sending HUP to init, but unfortunately none of
them was successful.

-- 
         EARTH
     smog  |   bricks
 AIR  --  mud  -- FIRE
soda water |   tequila
         WATER
 -- with thanks to fortune