Linksys PNIC II glitches

[Bob]

jay salmonson wrote:

> Hi Robin,
>
> Bravo!  I just read your diatribe on the PNIC II chipset as sent to
> the linux-tulip mailing list (and LinkSys):
> http://www.tux.org/hypermail/linux-tulip/1999-Aug/0036.html

Robin loves the Linksys 5-port 10/100 switch, and so do I.

We both got the Linksys pnic cards with that switch, and
Robin hates them. I have tried every driver from the
4/14/99 to latest 91u 10/15/99 and they either bumble
through or work fine, even though PNIC is a stupid
idea compared to 21143.

ftp://cesdis.gsfc.nasa.gov/pub/linux/drivers/kern-2.3

Use the README there. I've used that source with
kernel 2.2.10 and 2.2.13, so don't be put off by "kern-2.3".

I downloaded a whole linux installation and newer tulip
source through the default tulip.c even though it was crippled
with unresolved symbols. That's some kind of credit to Donald
Becker. "depmod -e tulip.o" to list unresolved symbols. That
was from a symbolic link problem with bin86 asm. With 91u
any unresolved symbols would probably have another cause,
because of new files, so--

#pushd /usr/src/linux/drivers/net
cat Makefile | gawk --source '{
  if ( $3 ~ /auto_irq.o/ print $1 "  " $2 "  " $3 " pci-netif.o ;
  else print $0}' > /tmp/gawk.out
#mv /tmp/gawk.out Makefile
#popd

I might replace my pnic cards with real 21143 tulip
cards, but before you stomp your pnic's, try 91u and
learn how to compile it with all the dependency files.
Then let us know if you can do real time sound.
Maybe you have an mp3 server on your net, like
icecast shoutcast shout listen, and see if you get
Robin's strange pauses even with 91u. "bing" does
a rudimentary first test and I can watch the lights
for the phenomenon Robin saw with 91g. Seems OK.

-Bob