Jeremy Jackson <[EMAIL PROTECTED]> writes:
> Ronald G Minnich wrote:
>
> > On 27 Feb 2001, Eric W. Biederman wrote:
> >
> > > I know windows can do it. I saw it last week. So I'm going to keep
> > > digging.
> >
> > This is a good point. Somehow windows knows when to ignore the bios, and
> > when not to. It's interesting to see it go, and it gives linux a bad
> > reputation. I've seen it happen.
> >
> > ron
>
> In the case of serial ports, windows may be accessing the superio as a PnP
> device
> or in "MB PNP" mode (thinking about wording used in ITE's superio docs).
> By doing
> it this way, (weither or not windows does) it doesn't matter what the bios
> did (or didn't
> do) other than to respect the settings (like bios ide geometry in bios) if
> they exist.
> I'm not sure that windows has code for every superio out there, so
> maybe it's using the PnP BIOS's pnp device list... linux currently ignores
> this
> and bangs the hardware... and i believe (have to reread ITE docs) when it's
>
> configured in MB pnp mode (by bios), it doesn't respond to pnp probe.
> (I've submitted idea to linux-kernel regarding this)
> if you look at 2.4's PCI, it can try direct hardware access first, but
> *can* fall
> back to BIOS access.... having the linux kernel able to do the same for PnP
>
> might catch things like the serial port, but would also enable things like
> laptop hotdock (and there's not much chance linux will do hot dock on *my*
> Compaq laptop any other
> way - unless someone has a 3D xray or spies who can get schematics)
PNP support in the kernel would be nice. However it avoids the
problem rather than tackling it. In particular I know the superio
on the L440GX motherboard is hardwared not to be isa PNP.
Eric
