[comp.unix.i386] Sys V 3.2.2 and EISA bus

wtm@uhura.neoucom.EDU (Bill Mayhew) (06/05/90)

Sorry that the headers on my articles don't precisely match the
message thread.  The pnews program on this system is broken, and I
fixing it is beyond my control.

Now for the real topic.  I noticed that Under  AT&T Sys V 3.3.3,
the /etc/default/boot file is examined as the kernel is loaded.
The parameter MEMRANGE allows one to specify whether or not DMA is
premissible in a given range.  For instance, for a Compaq 386, the
entry is:

MEMRANGE=0-640K:0,1M-15M:0,16M-15M,1

The ranges are separated by commas.

The attribute is separated from the range by a colon.  A 0
means there are no special properties for the range, while a 1
means that DMA is not allowed in that range.

The first address is inclusive, while the second is exclusive.
That is why the last pair is given in decending order, such that
the two "15M"s do not overlap.


Now for my question about EISA.  I thought that EISA supported edge
triggered interrupts.  I suppose that only matters for EISA cards,
though.  Seems like it could lead to complications, however, when
unix kernel messes around with the 8259 interrupt controller and
there are fany boards in the system, such as an EISA SCSI disk
controller.

While on the topic of interrupts, I recall several people bemoaning
the fact that IBM ISA bus IRQs are level sensitive.  That wouldn't
be so bad if your software could poll boards to see which board
needed service.  The problem, however, is that a lot of board
designers seem to be Airco Technical School graduates, which makes
them one notch below the summer co-op students that designed the ISA
bus.  I've seen several boards that drive the IRQ lead with a totem
pole TTL gate!  Grrrrrr.  Makes sharing the IRQ lead pretty tough
if they do that.

While on the subject of ISA bus peculiarities, does anybody know
the purpose of the barrel shifter on the EGA board?  By the time
one goes through all the gyrations necessary to use the shifter, it
seems that it would be quite a bit easier just to do the shift in
the CPU in the first place.

==Bill==
-- 

Bill Mayhew  Northeastern Ohio Universities College of Medicine
Rootstown, OH  44272-9995  USA    216-325-2511
wtm@uhura.neoucom.edu   ....!uunet!aablue!neoucom!wtm