Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 6 Jan 95 11:31:29 MST
From:      terry@cs.weber.edu (Terry Lambert)
To:        phk@ref.tfs.com (Poul-Henning Kamp)
Cc:        hackers@freebsd.org
Subject:   Re: sio.c, wd.c, if_ed.c & PCMCIA
Message-ID:  <9501061831.AA20067@cs.weber.edu>
In-Reply-To: <199501060944.BAA02736@ref.tfs.com> from "Poul-Henning Kamp" at Jan 6, 95 01:44:20 am

next in thread | previous in thread | raw e-mail | index | archive | help
> I have been studying rather heavily the last three weeks and belive I have
> the PCMCIA understood now.  I have looked at the Linux stuff and I have read
> a couple of books on the subject (All the ones I have been able to find: two).

You need to get a copy of the standards documents.  This will be $120 (I
think that's what I paid -- I may have it confused with the $80 I paid
for the PCI documentation) if the company you work for is a member of the
consortium, more (up to twice that) if not.  They also have dummy cards
and architectural overviews available.

You also need to get a copy of the Intel and IBM databooks.  There are
four different interface chip families -- the Intel and IBM books will
cover the most popular ones, and the largest number of chips.  There is
one european and one Japanese family that aren't just clones of the
Intel chips (the IBM is a clone of the Intel chip, but returns a different
ID than the Intel chips).  There is one obsolete laptop that uses the
European chipset -- I think it was nearly the first PCMCIA capable
machine, actually.

I'll need to bring the chip/family lists from home to give them to you,
and I'll try to dig up the 800 number for the consortium at the same time.

> I have a prototype running now, which will detect and configure my Megahertz
> modem or my Infomover correctly on boot.  The next step is to make it work
> when the cards are swapped.
>
> I think I have a workable architecture worked out.  I will present this to
> Soren in person this weekend, and if he doesn't send me to the "National
> Home Of The Recursively Bewildered", I will write a architectural paper
> on it and post it here during next week.
> It requires some architectural changes which may or may not come for free
> with the devfs, the main problems being the way the "softc" structure is
> identified from the dev_t and power-management.
> 
> If any major rewrites are planned for any of the drivers listed in the 
> subject, or for any other "PCMCIAble" device drivers, we should get the 
> PCMCIA support into it at the same time, so gimme a buzz before you
> redesign any of them.

I was considering this as a bus attach.  My money is still up in the air
as to whether it lands me an SMP'able PCI machine or a fast PCMCIA capable
portable.

Generally, I believe you want to probe and attach the devices as a bus
of the PCMCIA "bus" -- these are bridge chips to ISA, typically, and the
magic is the mapping of the device addresses into ISA addresses.

Each chip typically handles two devices.

If the architecture you have in mind is not considering the devices as
a seperate attached bus (at best) or a controller with subdevices (at
worst), then I think that it's perhaps the same trap that UnixWare has
fallen into on this (Kurt Mahon, who wrote the UnixWare code for the 2.0
UnixWare product agrees on this).


Currently, I don't have any PCMCIA devices, but I *do* have all of the
documentation available from the consortium and copies or originals of
all of the databooks.


					Terry Lambert
					terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9501061831.AA20067>