[tulip-bug] Problems with kne111tx

[Tom Parker]

Donald Becker wrote:

> On Sun, 19 Aug 2001, Tom Parker wrote:
>>I'm running a linux router/server off my cable modem, and I've been 
>>having a little problem with it.  The two network cards in it are both 
>>KNE111TX's, and it seems that if traffic flow goes anything higher than 
>>about 100 kb/s (rough guess, figure may be higher, but the value is 
>>significantly lower than what the 10mb network can do) that we get a 
>>collision light on the hub, and also tcpdump reports my computer waiting 
>>for a sequence number that the server does not ever give out until it 
>>times out for the server to respond, with no luck. The first few packets 
>>get through, but then everything messes up.
> This looks like a duplex mismatch problem, although it could also be a
> cable problem.
> 
> The following information is needed to track down the problem
>   the output the card detection message (use 'dmesg')
>   the error counts from /proc/net/dev
>   the output of 'tulip-diag -a' while the card is experiencing problems.

dmesg output below (with comments)
----------------------------------
Linux Tulip driver version 1.1.8 (June 16, 2001)
PCI: Found IRQ 10 for device 00:0b.0
PCI: Sharing IRQ 10 with 00:07.2 <- USB controller, on mainboard
PCI: Sharing IRQ 10 with 00:07.3 <- 2nd USB controller also on mainboard
eth0: Lite-On PNIC-II rev 37 at 0xd0800000, 00:C0:F0:76:66:8F, IRQ 10.
^^^^^ LAN Card

PCI: Found IRQ 11 for device 00:0f.0
PCI: Sharing IRQ 11 with 00:0d.0 <- Soundcard (SB Live)
eth1: Lite-On PNIC-II rev 37 at 0xd0802000, 00:C0:F0:76:66:93, IRQ 11.
^^^^^ Card to cable-modem
--------------------------------
Motherboard is an ABIT KT7A-RAID (if that's needed).

Error counts from /proc/net/dev are all zero, but the collision count is 
a different matter. Collsion count for the LAN interface (eth0) was 1218 
after my server had been up for a while and a number of tests done that 
day (compare to eth1 (external, same model network card) with a count of 
151). After an attempted file transfer by SMB (server->client) the 
collision count went up by 3 to 1221, with repeated tests upping the 
count by 3 each time. Attempted client->server SMB transfer adds 1 to 
collision count.

'tulip-diag -a' output is below, from during an attempted SMB file 
transfer (server->client). 1st card is LAN, 2nd is external.

--------------------------------------------------------------
tulip-diag.c:v2.08 5/15/2001 Donald Becker (becker@scyld.com)
  http://www.scyld.com/diag/index.html
Index #1: Found a Lite-On PNIC-II adapter at 0xdc00.
  * A potential Tulip chip has been found, but it appears to be active.
  * Either shutdown the network, or use the '-f' flag to see all values.
Lite-On PNIC-II chip registers at 0xdc00:
  0x00: fff88000 ffffffff ffffffff 0ff31000 0ff31200 e4660000 01002002 
effffbff
  Port selection is N-Way autonegotiation, half-duplex.
  Transmit started, Receive started, half-duplex.
   The Rx process state is 'Waiting for packets'.
   The Tx process state is 'Idle'.
   The transmit threshold is 72.
   The NWay status register is 000050ca.
  The current PNIC-II MAC address is 00:c0:f0:76:66:8f (c000c000 76f08f66).
  The current PNIC-II WOL address is 00:c0:f0:76:66:8f.
Index #2: Found a Lite-On PNIC-II adapter at 0xe800.
  * A potential Tulip chip has been found, but it appears to be active.
  * Either shutdown the network, or use the '-f' flag to see all values.
Lite-On PNIC-II chip registers at 0xe800:
  0x00: fff88000 ffffffff ffffffff 0ff2e000 0ff2e200 e4660000 01002002 
effffbff
  Port selection is N-Way autonegotiation, half-duplex.
  Transmit started, Receive started, half-duplex.
   The Rx process state is 'Waiting for packets'.
   The Tx process state is 'Idle'.
   The transmit threshold is 72.
   The NWay status register is 000050ca.
  The current PNIC-II MAC address is 00:c0:f0:76:66:93 (c000c000 76f09366).
  The current PNIC-II WOL address is 00:c0:f0:76:66:93.
--------------------------------------------------------------

On possible server/client duplex problems, the client network card (a 
'PRO120' - can't remember manufacturer. Other client had a 3com card, so 
I don't think that has any problems. It's been fine with every other 
network it's been on - 3 others, both 10mb and 100mb) is also set to 
auto-detect. Both cards at each end are 10/100, but the hub in the 
middle is only 10mb (getting replaced at some unknown point....). I 
doubt it's a cable problem, as all the cable is brand new, but I'm 
getting some new bits very soon anyway (the cable I'm using right now 
isn't long enough to properly go from room to room, it's all draped 
across my hallway right now).

Think that's everything. Hope it's of some help.

Tom Parker
-----------------------
palfrey@bits.bris.ac.uk
tp0945@bris.ac.uk