[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [cobalt-developers] Re: Replacing eepro100 with e100, need some .h source files
- Subject: Re: [cobalt-developers] Re: Replacing eepro100 with e100, need some .h source files
- From: "William L. Thomson Jr." <wlt@xxxxxxxxxxxxxxxxxxxx>
- Date: Fri Dec 7 20:29:46 2001
- Organization: Obsidian-Studios Inc.
- List-id: Discussion Forum for developers on Sun Cobalt Networks products <cobalt-developers.list.cobalt.com>
Tim,
FYI, on a different machine I had three nics, one of which used the
eepro100 mod. A guy from intel told me to replace the eepro100 with
e100, and I could not tell a difference.It was not any better, nor was
it worse. The only difference was my kernel loaded a single mod for all.
I have several machines now running e100, and have not had a problem.
Even during backup transfers of upwards of a gig. I noticed a big
difference when I grouped two nics together under a single IP using
IANS. And any time I have two or more available Intel nics I use IANS.
Once again, I am not a kernel hacker, and have no clue of other
experience with e100 other than my own. Which to date has been positive.
Also e100 provided some aditional features allowing me to dial it in
further with a Cisco switch, like disabling IFS. So all my experiences
to date have been positive. I imagine it will be the same with the
Cobalt, except for the IP alias issue which I know I will run into.
I will post results either way.
Tim Hockin wrote:
"William L. Thomson Jr." wrote:
So I would ask in return, any reason why Sun/Cobalt decided to go with
eepro100 instead of e100?
General consensus amongst kernel mailing list is that e100 sucks. Further,
when we started hacking on eepro100.c, e100 was not available, I believe.
Please, let me know how it works, but be sure to get actual numbers! I'd
like to see how our eepro100 compares to e100 on a bi-directional (1
interface) full duplex 100 MBit test.
Tim
Also, I have not fully looked into it, but when doing large transfer on
the local lan from the RaQ does take time, and has a ton of collisions
of the switch.
Without having looked into it myself, what is the setting of eepro100?
Is it set to auto detect duplex and speed? Or is it set to full duplex
100mps? Which I am sure is the reason for the collision, because by
default on the switch I force 100mps full duplex. Unless the device
requires autodetect. So far I have not had to force any of the e100 to
100mps full-duplex, not do I get large amounts or any collisions on them.
I am not blaming eepro100, I just do not think I have the switch dialed in properly.
--
Sincerely,
William L. Thomson Jr.
Obsidian-Studios, Inc.
439 Amber Way
Petaluma, Ca. 94952
Phone 707.766.9509
Fax
707.766.8989
http://www.obsidian-studios.com