ethernet driver questions

[john andersson]

i made a posting a couple of days ago to no avail about
/dev/biodd/net0 on the RT... we just finished a conference call with
some ibm folks in austin but it was sketchy. we are attempting to get
our rs6000 to communicate to a DEC PDP11 running RSTS (don't ask) via
ethernet; for the past few years the RT has been doing this
successfully, apparently ibm supplied some test programs to illustrate
using the block i/o device driver and we took it from there.

on the rs6000, /dev/biodd is not there, for standard ethernet stuff,
there are ent0, lo0, and en0 devices. with biodd on the RT, we could
simply use system calls to manipulate the net with, say
open(/dev/biodd/net0, O_RDWR), and some ioctl() calls. with ioctl
there was a structure used that  contained among other things, local
address, remote address, and net id. for the net id, it was a
consensus among us & ibm to set this to a process id (or port number)
on the PDP11 instead of protocol type. we are trying this in AIX 3.1
(using open(/dev/ent0, O_RDWR) instead) and it appears that this net
id gets overwritten by the net device driver perhaps? using dbx we see
that read() hangs or when a packet is received from the PDP11 it's
apparently been flushed. i'll stop here - don't want to spew too much.
any ideas? how about /dev/lo0 or /dev/en0? ibm says they will get back
to us, perhaps with a suggestion to a hook to the device driver to
leave the net id field alone or with some "edl_" system call examples
(edl_open(), edl_ioctl(), etc - don't know if it was edl_ or not, the
conference call line was static ridden).

any ideas? thanks in advance.
-- 
John Anderson 
jra at bally.com   
Bally Mfg. Reno NV