[9fans] PCNet and other short stories

[9fans] PCNet and other short stories

[Lucio De Re]

There will be working copies of etherpcnet.c on my web site in the
next hour or so:

	/sys/src/9/pc/etherpcnet.c
	http://wow.proxima.alt.za/src/kern.pcnet.c
and
	/sys/src/boot/pc/etherpcnet.c
	http://wow.proxima.alt.za/src/boot.pcnet.c

intended (and tested) to operate with the old Am79c960 but designed
for the entire range of controllers.  The list is in the file headers
and I'm also hoping to merge the two sources (despite Bell Labs's
aversion for #ifdefs) as they seem similar enough to warrant it.

As mentioned I've only tested these on the original controller chip,
I'll perform further tests on a VMware host (that will take some
putting together, unfortunately) and leave it to interested parties to
do additional testing - right now the untested portions are in a
potentially disastrous state.  At this point, if anyone wants to send
me some hardware to test against, I'll gladly return it when done
(that's probably more expensive that buying the stuff new - e-mail me
for details).

I was astounded at how well 9P handles broken device drivers (I ought
to have guessed so, but I didn't think about it).  Being a bit of an
embedded fan, I was very worried about damaging information on the
fileservers (et al.) on the network, but nothing of the sort ever took
place, thank goodness.

Lastly, the unsolved problem of the boot hang on slow hosts.  To all
who encountered the problem, I have one piece of advice: check your
floppy disks and drives.  All problems I had with such hangs have been
solved by making sure the diskettes were readable on the target host.

It's a bit of a chicken-and-egg situation, caused, I am almost
certain, by a rather unforgiving and uncommunicative floppy disk
driver, but it's worth keeping in mind.

++L

Re: [9fans] PCNet and other short stories

[Lucio De Re]

On Sat, Apr 13, 2002 at 04:16:56PM +0200, Lucio De Re wrote:
>
> There will be working copies of etherpcnet.c on my web site in the
> next hour or so:
>
The DMA management is a scary hack.  It is tempting to not do anything
at all, but then one gets very little choice of DMA channels  :-)

If anyone wants to contribute some intelligence, my knowledge stops at
a single controller on the 8088, cascaded DMA chips are a total
mystery, specially when the clever engineers at Intel made it
impossible to query the chip settings before one chooses to change
them.

The plan9.ini line I use is

	ether0=type=PCnet port=0x320 irq=9 dma=6 10BaseT

YMMV.  I have not included any acknowledgements in the source, which I
will correct as soon as I get a chance.  The plagiarised parties are
the ether79c970.c driver from the Plan 9 distribution and a similar
driver (ether79c960.c) from forsyth.  Anyone else who feels they
deserve acknowledgement, please just let me know :-)

++L

PS: I'll concoct an entry for the plan9.ini man page as soon as I have
managed a little broader testing.