Path: gmdzi!unido!mcsun!news.funet.fi!hydra!klaava!torvalds
From: torva...@klaava.Helsinki.FI (Linus Benedict Torvalds)
Newsgroups: comp.sys.ibm.pc.hardware,comp.sys.ibm.pc.misc
Subject: 387 on a 386-AT
Keywords: protected mode, 386, 387
Message-ID: <1991Sep22.185745.22256@klaava.Helsinki.FI>
Date: 22 Sep 91 18:57:45 GMT
Organization: University of Helsinki
Lines: 36

Hello netters,

As I've gotten nice replies from you before, I'll try the net.wisdom
once more. I've got this problem with a 387 on an AT-clone running in
protected mode on this unix-clone of mine:

It seems the 387 #ERROR-pin is wired through some electronics to the
IRQ13 pin on the slave 8259A, and latching among others the PEREQ and
BUSY pins on the 386. Correct?

It also seems the #ERROR pin on the 386 is grounded. Tell me it isn't
so, please...

That's how far I've gotten with the meager docs I've got. The latching
of the PEREQ/BUSY pins seem to be cleared by writing 0 to port 0xf0, but
I'd be very interested to hear all the gory details:

Can you activate the #ERROR pin on the 386 side by some magic
incantation (write xx to port yy?), disconnecting the IRQ13/latching
hardware if possible? What I'd want is the "normal" 386/387 binding,
with #error pins just wired together. Is this at all possible on an
AT(386)?

What exactly does port 0xf0 do? Can you do anything other than just
clear the latch? Bitpatterns anyone?

Linus Torvalds (torva...@kruuna.helsinki.fi)

PS. As always, I'm a poor student, and while pointers to nice books (for
just $499.95, sigh) are wellcome, free advice is even more so.

PPS. Just get a 486, and set/reset the right bit in cr0 isn't the
answer either, as that costs even more than the books :-(

PPPS. Notice how I don't have a 10-line .sig, but I still get to clutter
up your screen with these never-ending post-scripts :-) Wow.

Path: gmdzi!unido!math.fu-berlin.de!ira.uka.de!sol.ctr.columbia.edu!
zaphod.mps.ohio-state.edu!qt.cs.utexas.edu!cs.utexas.edu!sun-barr!
olivea!uunet!mcsun!news.funet.fi!hydra!klaava!torvalds
From: torva...@klaava.Helsinki.FI (Linus Benedict Torvalds)
Newsgroups: comp.sys.intel
Subject: 387 in AT-386
Message-ID: <1991Sep26.070301.17611@klaava.Helsinki.FI>
Date: 26 Sep 91 07:03:01 GMT
Organization: University of Helsinki
Lines: 40


Ok, I posted this on c.s.ibm.pc.hardware some time ago, but got no
answers. Maybe you can help:

>Hello netters,
>
>As I've gotten nice replies from you before, I'll try the net.wisdom
>once more. I've got this problem with a 387 on an AT-clone running in
>protected mode on this unix-clone of mine:
>
>It seems the 387 #ERROR-pin is wired through some electronics to the
>IRQ13 pin on the slave 8259A, and latching among others the PEREQ and
>BUSY pins on the 386. Correct?
>
>It also seems the #ERROR pin on the 386 is grounded. Tell me it isn't
>so, please...
>
>That's how far I've gotten with the meager docs I've got. The latching
>of the PEREQ/BUSY pins seem to be cleared by writing 0 to port 0xf0, but
>I'd be very interested to hear all the gory details:
>
>Can you activate the #ERROR pin on the 386 side by some magic
>incantation (write xx to port yy?), disconnecting the IRQ13/latching
>hardware if possible? What I'd want is the "normal" 386/387 binding,
>with #error pins just wired together. Is this at all possible on an
>AT(386)?
>
>What exactly does port 0xf0 do? Can you do anything other than just
>clear the latch? Bitpatterns anyone?
>
> Linus Torvalds (torva...@kruuna.helsinki.fi)
>
>PS. As always, I'm a poor student, and while pointers to nice books (for
>just $499.95, sigh) are wellcome, free advice is even more so.
>
>PPS. Just get a 486, and set/reset the right bit in cr0 isn't the
>answer either, as that costs even more than the books :-(
>
>PPPS. Notice how I don't have a 10-line .sig, but I still get to clutter
>up your screen with these never-ending post-scripts :-) Wow.