Problem with mass-collisions

Georg Koester georgk@bigfoot.de
Tue Aug 25 10:19:02 1998


Thank you very much for your fast reply! A pity that I forgot to mention
a possibly important fact. The EEPro100b is running at 10Mbit.
Donald Becker wrote:

> What kind of "hub" do you have?  With mixed speeds it's obviously not a
> repeater (the usual meaning of "hub" in Ethernet).  Is it a full switch
> or a bridged pair of repeaters?
Its a normal repeater. (see above)
 
> > - When using ftp between a ne2k and a 100Mbit I get massive collisions
> > on the hub.
> > - When doing so under winNT: normal operation; speed 360kb
> 
> This could be a either a duplex mismatch or the bridge dropping packets.

I think they´re both running at half-duplex. Windows told me:) And the
other card is a 16bit isa NE2k.
 
> If it's a switch, it may be that the driver that came with the NE2000 clone
> knows how to switch the card to full-duplex mode.  That's not part of the
> NE2000 interface -- the full-duplex setting is magic that is different on
> each card type.

It´s your linux-driver. This is what ne2k said:

Checking the ethercard at 0x300.
  Register 0x0d (0x30d) is ff
  Passed initial NE2000 probe, value 00.
8390 registers: 0a 3c 66 65 01 00 ff 00 01 00 30 07 01 00 00 00
SA PROM        0: 00 00 80 80 ad ad 71 71 b0 b0 8b 8b 04 04 22 22
SA PROM 0x10: 90 90 08 08 00 00 00 00 00 00 00 00 57 57 57 57

      NE2000 found at 0x300, using start page 0x40 and end page 0x80.

I don´ t see from that if the card is running full-duplex, but i don´ t
think so since the repeater cannot do it.

> There is usually a way to set full-duplex independent of the driver, usually
> through an EEPROM setup program.

Well, the ne2k hasn´ t got that option, and with windoze I can set the
EEPro100b to full-duplex, but that didn´ t work. I´ m afraid I can´ t
work with your eepro100-diag, cause -A didn´ t work for me and it doesn´
t talk to me in clear text. Here is the message when the card is up and
running:

eepro100-diag.c:v0.07 2/25/98 Donald Becker
(becker@cesdis.gsfc.nasa.gov)
Found Intel i82557/i82558 Speedo3 chip at 0x6100.
i82557 chip registers at 0x6100:
  00000050 02806904 00000000 00080002 1821782d 00000600
  No interrupt sources are pending.
   The transmit unit state is 'Suspended'.
   The receive unit state is 'Ready'.
EEPROM contents:
  a000 c5c9 20bd 0000 0000 0201 4701 0000
  6680 8104 4481 0009 8086 0000 0000 0000
  0000 0000 0000 0000 0000 0000 0000 0000
  0000 0000 0000 0000 0000 0000 0000 0000
  0000 0000 0000 0000 0000 0000 0000 0000
  0000 0000 0000 0000 0000 0000 0000 0000
  0000 0000 0000 0000 0000 0000 0000 0000
  0000 0000 0000 0000 0000 0000 0000 3e9e
 The EEPROM checksum (should be 0xbaba) is 0xbaba.
Intel EtherExpress Pro 10/100 EEPROM contents:
  Station address 00:A0:C9:C5:BD:20.
  Receiver lock-up bug exists. (The driver work-around *is*
implemented.)
  Board assembly 668081-004, Physical connectors present: RJ45
  Primary interface chip i82555 PHY #1.

> 
> > - When doing smb(both windoze and linux): massive collisions; speed 60kb
> > - When using nfs between two linux-boxes: no collisions; [speed about
> > 200kb]
> > - When using ttcp -t -s: collisions; speed 60kb
> 
> If it's a bridged repeater pair, faster transfer rates might cause the bridge
> to drop packets, although TCP should eventually throttle back in response to
> lost packets.  (This effect makes lower-performance protocols look good.)
> This same thing can happen with full switches, but they are likely to have
> more buffering.

And what if its just a repeater? 
 
> > I never have these collisions when using only two ne2k(I´m always using
> > only two interfaces) and never had a chance to test two 100Mbit-cards.
> 
> There is no speed mismatch in this case.  The bridge/switch never overflows.
> 
> > So please can somebody interprete this behavior to me? Is my hub shit? I
> > had as 100Mbit-cards: a tulip; and currently an Intel Pro/100b. My hub
> > is just a simple one. Nothing special, but it worked just fine when I
> > tried it in a bigger network with a mixed enviroment of bnc and
> > tp-cables and more computers. I tried some different settings in
> > eepro100.c as a module, but that didn´ t change anything.
> 
> Changing settings wouldn't have a good effect.
> Are you getting a full-duplex link from the EEPro100?  You can use mii-diag
> to find out:
>   http://cesdis.gsfc.nasa.gov/linux/diag/index.html
> 
> Donald Becker                                     becker@cesdis.gsfc.nasa.gov

I tried your tool, messages above. But I could not change the
autonegotiation things. It didn´ t know the option. 
Well, maybe the new information makes a difference, but to me it seems
that nobody ever thought about running a 100Mbit-card at 10... So nobody
I asked knew.
ciao Georg


> USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
> Code 930.5, Goddard Space Flight Center,  Greenbelt, MD.  20771
> 301-286-0882         http://cesdis.gsfc.nasa.gov/people/becker/whoiam.html

-- 

	From: Georg Koester e-Mail: georgk@bigfoot.de