A few notes on testing

Chris Lingard chris at stockwith.co.uk
Tue Jul 13 14:09:07 PDT 2004


Matthew Burgess wrote:

> On Tue, 13 Jul 2004 21:43:02 +0100
> Chris Lingard <chris at stockwith.co.uk> wrote:
> 
>> Matthew Burgess wrote:
>> 
>> > On Tue, 13 Jul 2004 11:27:49 -0700
>> > "Kevin P. Fleming" <kpfleming at linuxfromscratch.org> wrote:
>> > 
>> >> Matthew Burgess wrote:
>> >> 
>> >> > Then I wonder how hotplug will do it's coldplugging then?  I mean
>> >> > the*.rc scripts currently use /proc/[bus] (where bus is USB, PCI,
>> >> > etc.) to determine what hardware is present on boot.
>> >> 
>> >> Those same scripts know how to scan /sys/bus/pci as well, and will
>> >do> when /sys is present.
>> > 
>> > But I thought that /sys (as the mount point for sysfs) will only be
>> > populated by the information exported via the device drivers.  If
>> > those drivers are modules, then we've got a chicken & egg problem,
>> > no?  I
>> 
>> I use hotplug on an initramfs; but not straight away.
>> I mount /proc, /sys, /dev, run udevstart.
>> 
>> Then you can start populating /bin, /usr, /lib; either with
>> symbolic links or mount -bind, or whatever you like.
>> 
>> Once this is set up you can exec  /sbin/init
>> 
>> My start up scripts will run hotplug as one of the first.
>> 
>> And networking and USB modules are loaded and work fine.
> 
> It only works though because /proc is mounted, surely, and therefore
> the *.rc scripts can scan /proc/[bus] to see what devices are available?
> What happens if you *don't* mount /proc before hotplug starts doing it's
> thing?
> 
> I realise we're getting way ahead of ourselves here - hotplug's still
> just in unstable, and I'd imagine /proc/[bus] won't be deprecated until
> at least the 2.7.x series which means it won't affect LFS until the
> 2.8.x series, but still...I like a good technical discussion, even if it
> ends up demonstrating my constant need for a cluebat.


Never tried it; it would probably fail, you must mount /proc first.

/proc, /sys, /dev + shm and pts, /etc; udevstart

Yes you need an /etc with the things

BTW, you do not need a root; just use the one the kernel gives you.

Modules will only be available once you mount/link/bind  /lib




More information about the lfs-dev mailing list