[tulip-bug] BUG: Digital DS21143-xD Tulip rev 65 -- autoneg. takes 3 minutes

Donald Becker becker@scyld.com
Mon, 2 Jul 2001 16:33:03 -0400 (EDT)


On Mon, 2 Jul 2001, Till Immanuel Patzschke wrote:

> still having problems w/ my IntraServer Quad-Ethernet board (tulip
> based). Using driver version 0.92t I still see wierd behaviour: the
> autonegotiation takes very long (3 minutes) [looks kind of arbitrary].

> Linux: 2.2.18-... IntraServer ITI-5232E (Quad-Ethernet)
> Tulip driver: 0.92t
...
> Jul  2 21:35:05 ipat01 kernel: tulip.c:v0.92t 1/15/2001  Written by Donald
> Becker <becker@scyld.com>
> Jul  2 21:35:05 ipat01 kernel: eth4: Digital DS21143-xD Tulip rev 65 at
> 0xfc96f000, 00:06:2B:01:A7:D3, IRQ 19.
> Jul  2 21:35:05 ipat01 kernel: eth4:  EEPROM default media type Autosense.
...
> Jul  2 21:35:05 ipat01 kernel: eth4:  Index #2 - Media 100baseTx (#3) described
> by a 21143 SYM PHY (4) block.
> Jul  2 21:35:05 ipat01 kernel: eth4:  Index #3 - Media 100baseTx-FDX (#5)
> described by a 21143 SYM PHY (4) block.

OK, this looks like the reference design SROM from Digital.  That's
sometimes not a good sign -- I've encountered several boards where the
designers just copied the example, without understanding that a media
table has to be written for their specific board.

> eth4: 21143 link status interrupt 45e1d0ce, CSR5 f0668010, fffbffff.

Good.  We have a link partner that advertised 05e1 -- 10/100 half+full,
with flow control.

> eth4: Switching to 100baseTx-FDX based on link negotiation 01e0 & 45e1 = 01e0.

Good

> eth4:  Media table type 4.
> eth4: 21143 non-MII 100baseTx-FDX transceiver
> control 08af/0005.

This is verbatim from the example design SROM.

> eth4:  Setting CSR15 to 08af0008/00050008.
> eth4: Using media type 100baseTx-FDX, CSR12 is ce.
> eth4:  Setting CSR6 838e0200/b38e2202 CSR12
> 000000ce.
> eth4: exiting interrupt, csr5=0xf0660000.
> eth4: interrupt  csr5=0xf0668010 new csr5=0xf0660000.
> eth4: 21143 link status interrupt 0000020a, CSR5 f0668010, fffbff7f.
> eth4: 21143 100baseTx-FDX link beat failed.

Ohhh, bad.  Either the link partner didn't start generating link beat,
or this board doesn't report the link beat value as promised.

Could you send the output of
   tulip-diag -ee
for the Tulip interface with the EEPROM?

> eth4: 21143 link status interrupt 000002c2, CSR5 f0668010, fffbff
> eth4: 21143 100baseTx-FDX link beat failed.
> eth4: Restarting internal NWay autonegotiation, 0003ffff.

> 
> snip,snip----more-of-the-same-stuff----------snip,snip
> 
> After some time we finally succeed... (somehow)

> Jul  2 21:37:22 ipat01 kernel: eth4: 21143 link status interrupt 000002c8, CSR5
> f8668000, fffbff7f.
> Jul  2 21:37:22 ipat01 kernel: eth4: 21143 100baseTx-FDX link beat good.

Hmmm, that's strange.  What is your link partner?  I'm wondering about a
timing issue...

Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993