eepro100 frame errors with SMP

[Pedro Ribeiro]

Linux 2.0.36 SMP
MB N440BX DUAL PII333

eth0      Link encap:Ethernet  HWaddr 00:A0:C9:D3:49:5C
          inet addr:193.137.220.5  Bcast:193.137.220.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:7963391 errors:109 dropped:0 overruns:0 frame:0
          TX packets:7442105 errors:0 dropped:0 overruns:105 carrier:0
          collisions:0
          Interrupt:5 Base address:0x1060

Bus  0, device  15, function  0:
  Ethernet controller: Intel 82557 (rev 5).
    Medium devsel.  Fast back-to-back capable.  IRQ 5.  Master Capable.
Latency=64.  Min Gnt=8.Max Lat=56.
    Prefetchable 32 bit memory at 0xfa104000.
    I/O at 0x1060.
    Non-prefetchable 32 bit memory at 0xfa000000.

The switch (SSII3300/3COM) reports in the port of the machine:

Port:                9             Port Speed:          100Mbps FD Auto

Received Stats                     Transmit Stats
--------------                     --------------
Unicast Packets:     6515190       Unicast Packets:     6520983
Non Unicast Packets: 312224        Non Unicast Packets: 1180400
Octets:              2638317328    Octets:              2530995564
Fragments:           0             Collisions:          0

Errors
------
Undersize:           0             Oversize:            0
CRC Error:           0             Jabbers:             0

Packet Size Analysis
--------------------
64 Octets:           7033666       256 to 511 Octets:   472429
65 to 127 Octets:    2793557       512 to 1023 Octets:  618470
128 to 255 Octets:   1116457       1024 to 1518 Octets: 2494218

Hope this helps.

PS: Some time ago i'v also detected that the RTL8029 under WinNT4.0 with the
disk provided drivers don't enforce the minimum frame length.

[]---------------------------------------------------------------[]
  Pedro Ribeiro
  Online: http://www.isel.pt/~pribeiro/
  IRC(PTnet) Nick: PAntMaR
  e-Mail: Personal:  pribeiro@isel.pt
          Admin:     admin@isel.pt
[]---------------------------------------------------------------[]

-----Original Message-----
From: Alan Cox <alan@lxorguk.ukuu.org.uk>
To: Alan Curry <pacman-kernel@cqc.com>
Cc: linux-eepro100@cesdis1.gsfc.nasa.gov
<linux-eepro100@cesdis1.gsfc.nasa.gov>; linux-kernel@vger.rutgers.edu
<linux-kernel@vger.rutgers.edu>
Date: Fri Mar 12 07:25:04 1999
Subject: Re: eepro100 frame errors with SMP


>> kernels, leads us to believe that the problem only exists when more than
one
>> processor is being used. This smells like a driver bug to me.
>
>Assuming the eepro100 is counting tx underruns like normal drivers, and
these
>are TX then yes - the Linux box is sourcing short frames,  the CRC on the
cisco
>is exactly what you would expect too.
>
>The problem is - the short TX frames are almost certainly PCI bus access
>related. That is the chip can't get to main memory in time to get the bytes
>it needs for the next piece of the packet before it empties its internal
>FIFO.
>
>So what on your bus has a PCI latency set to 240/248 ? 8)
>
>Alan
>