overruns under heavy load

[Charles 'Buck' Krasic]

Hi.

I am having performance problems with an eepro100 I have just started
to use.

My machine is a Toshiba Tecra 700 notebook.  The eepro is plugged into 
a PCI slot in a docking station.

I've been using ttcp to check performance, as well as some "dd"'ing to 
an NFS server. 

Numbers from ttcp are in the range of 2~3MB/s.  Other machine pairs on 
the same network, using eepro100's and 3Coms are seeing 10MB/s+.

The output from netstat on the Tecra is showing overruns:

[krasic@giove krasic]$ netstat -i
Kernel Interface table
Iface   MTU Met  RX-OK RX-ERR RX-DRP RX-OVR  TX-OK TX-ERR TX-DRP TX-OVR Flags
lo     3584   0  16374      0      0      0  16374      0      0      0 BLRU
eth0   1500   0  13358      0      0    258  15727      0      0 11492 BRU

None of the other hosts, using either 3Coms or eepro's, are showing
any overruns at all.

The "dd" tests on the Tecra are triggering "NFS server not responding" 
timeounts, which I suspect are due to overruns.  

The key differences between the Tecra and the other machines is that it
is very slow (120Mhz, no L2 cache), and the PCI slot in the dock is
accessed via a PCI bridge.

Are there any "tuning" knobs that I can adjust for the driver?

-- Buck

ps Here is the startup message from the driver:

eepro100.c:v1.05 10/16/98 Donald Becker http://cesdis.gsfc.nasa.gov/linux/drivers/eepro100.html
eth0: Intel EtherExpress Pro 10/100 at 0xff60, 00:A0:C9:C6:D2:0B, IRQ 11.
  Board assembly 678400-001, Physical connectors present: RJ45
  Primary interface chip i82555 PHY #1.
  Forcing 100Mbs full-duplex operation.
  General self-test: passed.
  Serial sub-system self-test: passed.
  Internal registers self-test: passed.
  ROM checksum self-test: passed (0x49caa8d6).
  Receiver lock-up workaround activated.
sysctl: ip forwarding enabled