Subject: anybody got "ae0: ring buffer overrun"?
To: NetBSD/mac68k mailing list <port-mac68k@NetBSD.ORG>
From: Masami Nakata <masami@fa2.so-net.or.jp>
List: port-mac68k
Date: 01/10/1997 10:02:42
Hi, it's me again.
I've been seeing this "ae0: ring buffer overrun" error very often when
I send a chunk of data from my notebook PC running FreeBSD (*gasp*) to
my SE/30 running NetBSD, resulting in transfer rates ranging from
30KB/s to 80KB/s. The other way is much faster, ranging from 150KB/s to
300KB/s (still slow, I know).
The Ethernet card in the unaccelerated SE/30 is a Kinetics EtherPort
SE/30 which is currently unsupported officially. However, I've
patched if_aereg.h and if_ae.c to get it to work. Basically, the
EtherPort SE/30 is an ordinary NIC based card that has a byte-wide
shared buffer rather than a more popular word-wide buffer, so the
changes I've made are only to the initialization sequence.
Has anybody seen this, too? Or am I the only one?
BTW, the rights to EtherPort is now sold to Dayna, but their newer
DaynaPORT E/30s are much different creatures (SONIC based), so it
might take a while to support them. Yet, the stock ae driver
identifies them as a same card which could prevent my changes to the
driver from working. I need to know how to distinguish the two cards
before I can submit the changes to Scott. If anyone with a DaynaPORT
card could run Slots (a MacOS app written by Apple), write down its
output on the Ethernet card, and send it to me, I'd greatly appreciate
it.
ken