"Larry G. Linde" <[EMAIL PROTECTED]> writes:
> There are only a couple of changes to stage1. That which is required to be
> there. Such as:
>
> The memory sizing code has been changed, The original code had a few
> problems.
> Memory gets passed to kernel.
> Most of stage 1 has been changed to protected mode only. No real mode.
LinuxBIOS does this already, except when using the DOC devices
because every byte is precious before you initialize memory.
> The pci enumeration is complete now.
I'm not certain what the differences are here.
> The irq routing table gets passed to the kernel now.
LinuxBIOS also does this.
> Keyboard stuff.
> Lots of irq setup stuff
This is the only really big area of difference, but it is something
that we probably should have the option of doing in linuxBIOS so
it doesn't hurt.
> Enable access to full flash not just 64k.
Other ports do this also.
>
>
> I think it will merge back in ok, but it may not be as simple as you might
> think.
Sounds about right. It's a noticeable divergence but my ports tend
to get about as bad before I merge everything back in.
> Stage2 is quite different. It fully inits all the hw, display, 950, kbd/mse,
> disks, enet, nvram, rtc, everything.
Not a problem. Stage2/tiara shouldn't be a problem to compile
seperately, or if we have a sufficiently close relationship between the
two as a slightly seperate binary. The only really tricky bit here
is that stage2 seems to be built to use symbols from stage1. Which
could get interesting if these were maintained separately.
Eric
