[comp.unix.i386] FLAME ABOUT STUPID INTERRUPT SYSTEM DESIGNERS

baxter@zola.ICS.UCI.EDU (Ira Baxter) (05/29/90)

I recently flamed the entire PC world, including UNIX OS porters, for
stupid interrupt system design; I now understand that the problem
seems to be active drivers on each I/O board for each interrupt level.

I want to at least apologize to the UNIX OS porters, for making the best
of a bad thing. :-{{{

The flame to the idiot hardware designers... well, *that* still stands.
And malarkey about how the PC wasn't designed for big systems is no d---
excuse (comments about how it wasn't designed at all, well, now, those
make sense!).

Given all that... has anybody made the ISC 2.0.2 X5 ASY update operate
with more than one device on the same interrupt level?  The docs
explicitly say that some serial ports really do allow multiple devices
on the same level.  How can this possibly work?  What do serial ports
do differently on the bus to allow a logical or of interrupts?  I have
tried configuring in ths driver with a standard dual serial (COM1: and
COM2: for MSDOS) with an internal modem (COM4:), hacking ASY_2_VECT to
be level 3, but apparantly the output interrupts get mangled, and
stuff comes out at the rate of 1 ch/second.  Obviously... some weird
interrupt conflict.

IDB
(714) 856-6693  ICS Dept/ UC Irvine, Irvine CA 92717

haugen@bulus3.BMA.COM (John M. Haugen) (05/29/90)

In article <9005282014.aa08953@PARIS.ICS.UCI.EDU>, baxter@zola.ICS.UCI.EDU (Ira Baxter) writes:
. . .
> Given all that... has anybody made the ISC 2.0.2 X5 ASY update operate
> with more than one device on the same interrupt level?  The docs
> explicitly say that some serial ports really do allow multiple devices

Yes IBM goofed in their first attempt at the bus (ISA) by not allowing shared
interrupts. But there second attempt (Microchannel) does.  Also I believe that
the interrupts on the ISA bus are edge triggered.  It would be hard to share
interrupts that are edge triggered.  Microchannel is level sensitive.  That is
the interrupt line is not dropped until the card driving the interrupt is
referenced by reading or writing to it's I/O registers.  That would make it
much easier to share interrupts.

> IDB
> (714) 856-6693  ICS Dept/ UC Irvine, Irvine CA 92717

John M. Haugen               Domain:  haugen@bma.com
Bull Micral of America       UUCP:    ...!uunet!bulus3!haugen
900 Long Lake Road           ATT:  612-633-5660