Re: PnP in 2.2

Brian Gerst (bgerst@quark.vpplus.com)
Sun, 07 Dec 1997 21:59:16 -0500


Peter Svensson wrote:
>
> On Sun, 7 Dec 1997, Philippe Rivard wrote:
>
> > [SNIP] Not to say that the isapnp tools aren't good,
> > but they should at the very least be included in the kernel, to simplify
> > everyone's lives.
>
> Why put something that can be in userland in the kernel? My personal view
> is that the kernel should be as simple as possible and with as little
> policy as possible.

For devices that are needed at boot time (ie. ide interfaces, network
cards, etc.) It can be argued that these can be done via initrd, but
making an initrd is not for the faint of heart or the novice user, and
would have to be recreated anytime the configuration is changed. The
goal is to make the kernel work reliably on all hardware possible with
the least user intervention needed.

> How would you tell the kernel upon boot the things that it needs to know
> to initialize the pnp hardware? The microsoft way of having the kernel
> reading configuration files is not something I like the sound of.

If you go with automatic configuration (conflict resolution), then this
would be unnecesary. You would only need to pass info to the kernel if
you wished to override the auto config parameters. Tom Lees' PnP patch
allows such info to be preloaded by the bootloader (ie lilo or loadlin)
in the same manner as an initrd (but without the extra complication of
setting up a filesystem and scripts that using an initrd would need).

-- 

Brian Gerst