losing 100megabit and/or full duplex setup on nics

Tue Jul 13 11:43:29 1999

This is a multi-part message in MIME format.

------=_NextPart_000_02F8_01BECD1C.7FCEF300
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 8bit

we have seen several system with different kind of switches have
a problem.

1. TCP runs extremely slow, here is a netperf result

Node wolf13
TCP STREAM TEST to wolf00
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

65535  65568   1472    60.41       0.78

udp seems to run fine though.

we are running kernel 2.2.2
we have seen the problem on the follwing cards

eth0: 3Com 3c905B Cyclone 100baseTx   driver - eepro100.c:v1.06
eth0: Intel EtherExpress Pro 10/100   driver - 3c59x.c:v0.99H

we have seen the problem with both types of cards above
with the following switches

Foundry Fast Iron 16 Port Switch set to autonegotiation on all ports
Extreme Summit48 image version 2.0.19

i have set the drivers in conf.modules to force 100meg/full duplex on
the
problem systems and everything seems to work again after a reboot.
however,
some systems set to autonegotiate don't seem to have this problem.
in fact in most cases we do not see this problem but every now and then
3 or 4 of our 96 nodes experience this problem.

i will be glad to provide any additional information that is required

here is some diag info from one system using 3com card from vortex-diag

vortex-diag.c:v1.08e 6/23/99 Donald Becker (becker@cesdis.gsfc.nasa.gov)
Index #1: Found a 3c905B Cyclone 100baseTx adapter at 0xec00.
Initial window 7, registers values by window:
  Window 0: 0000 0000 0000 0000 f5f5 00bf ffff 0000.
  Window 1: 0000 0000 0000 0000 0000 0000 0000 2000.
  Window 2: c000 684f 1f73 0000 0000 0000 000a 4000.
  Window 3: 0000 0180 05ea 0000 000a 0800 0800 6000.
  Window 4: 0000 00d0 0000 0cc0 0003 8880 6700 8000.
  Window 5: 1ffc 0000 0000 0600 0805 06de 06c6 a000.
  Window 6: 0000 0000 0000 e001 1100 f250 0957 c000.
  Window 7: 0000 0000 0000 0000 0000 0000 0000 e000.
Vortex chip registers at 0xec00
  0xEC10: 00000000 00000000 00000023 00000000
  0xEC20: 00000020 00000000 00080000 00000004
  0xEC30: 00000000 6a76958a 388ac190 00080004
Indication enable is 06c6, interrupt enable is 06de.
No interrupt sources are pending.
Transceiver/media interfaces available:  100baseTx 10baseT.
Transceiver type in use:  Autonegotiate.
MAC settings: half-duplex.
Station address set to 00:c0:4f:68:73:1f.
Configuration options 4000.
Parsing the EEPROM of a 3Com Vortex/Boomerang:
3Com Node Address 00:C0:4F:68:73:1F (used as a unique ID only).
OEM Station address 00:C0:4F:68:73:1F (used as the ethernet address).
Manufacture date (MM/DD/YYYY) 7/3/1997, division ·, product I .
Options: none.
  Vortex format checksum is incorrect (004b vs. 1028).
  Cyclone format checksum is correct (0xf9 vs. 0xf9).
  Hurricane format checksum is correct (0xf9 vs. 0xf9).
MII PHY found at address 24, status 7829.
MII PHY 0 at #24 transceiver registers:
   3100 782d 0000 0000 41e1 41e1 c003 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.

i do not have any diags for intel cards. notice how driver has dropped
to half-duplex for vortex diag. this came up full duplex

lots of errors in message files like the following

eth0: Transmit error, Tx status register 82.

------=_NextPart_000_02F8_01BECD1C.7FCEF300
Content-Type: text/x-vcard;
	name="Jeff Davis.vcf"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment;
	filename="Jeff Davis.vcf"

BEGIN:VCARD
N:Davis;Jeff
FN:Jeff Davis
ORG:Amerada Hess;MIS
ADR;WORK:;;;Houston;TX
LABEL;WORK:Houston, TX
X-WAB-GENDER:Unspecified
URL:http://www.jrdavis.net
URL:http://www.hess.com
EMAIL;PREF;INTERNET:work@jrdavis.net
EMAIL;INTERNET:jeff@jrdavis.net
END:VCARD

------=_NextPart_000_02F8_01BECD1C.7FCEF300--