[ On Monday, March 22, 1999 at 20:06:38 (-0500), Michael Richardson wrote: ]
> Subject: questions re: Lance chips
>
>   Does anyone have any experience with the Lance chipset(s) failing to report
> missed packets, such as due to being out of receive descriptors? I'm not
> talking about specific lance.c issues, but generally with the chip? 
> 
>   I figure that NetBSD folk have the most experience now with lance chips
> due to Sparc/Sun3/etc. Unfortunately, I have to finish debugging our Linux/NT
> before I can pass it onwards, and focus on completing the more important
> driver... 

I've not seen any noticable problem on my sparcs (nor do I recall any on
the sun3).  I've always wanted to port the "lestats" program I used to
run on SunOS-4, but I haven't got around to whacking the driver to
return the right info in a user-visible structure.

I do get the following under load from the "le" driver on i386 though
(1.3.3 and 1.3I):  "le0: device timeout". On occasion I get "excessive
collisions" messages too, but only when there are too many.  I don't
think I've ever seen the latter on a busy squid cache that's connected
directly to a switch port though.

FYI the probe reports the following on the i386 machines (these are
10/100 interfaces too):

le0 at pci1 dev 5 function 0: PCnet-PCI Ethernet
le0: address 00:06:29:05:XX:XX
le0: 8 receive buffers, 2 transmit buffers
le0: interrupting at irq 14

On 1.3.3 ifconfig reports "media: manual status: active" with a 100baseT
connection, and on 1.3I it reports "media: Ethernet autoselect" with a
10baseT connection.  No "mediaopt" flags were specified in either case.


-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>