kernel: eth0: card reports no resources.

I know, i read the archives, this has been beat to death. But please bear
with me.

I have 2 ibm X330 servers, 2*PIII 1.1GHZ, 2 inter eepro100 onboard NICs,
both of them running TUX (advanced kernel mode web server that puts all
other web servers in a back pocket if measuring raw power). the boxes run
24/7, loaded, outputting between 20 to 40 MB/sec, with several thousands of
concurrent connections each.

i have just noticed this error in my message log. One of the servers works
harder then the other, on that one i see the message several times on the
last few days. the other has it only once.

the last thread stated, that his problem was solved, alog with this :
----------------------------------------------------------------------------
-----------------
As an update, for the list archives, our problem has been found. We were
experiencing DDOS attacks on a tenant IP at the rate of 130,000 packets per
second, 50MB/sec. This was obviously overrunning the capacities of the card
and causing the error messages.
----------------------------------------------------------------------------
----------------

this worries me, since i expected to continue trusting this machine as
traffic grows continuosly. Is there a limit that the eepro can reach, that
after that it not stable ? it the limit in a sence of packats per second, or
bandwidth ? If this card has such limits, what card could you recommend that
can winstand continous 70-80MB throughput, with high packet rate ?

I am running the vanilla 2.4.18 kernel, patched with tux.

the driver there is

"eepro100.c:v1.09j-t 9/29/99 Donald Becker
http://www.scyld.com/network/eepro100.html\n"
"eepro100.c: $Revision: 1.36 $ 2000/11/17 Modified by Andrey V. Savochkin
<[email protected]> and others\n";

the card is 82557 based

I didn't try the vm proc settings change, but i don't believe that they
would help - these machines serve from memory, they don't access the disk at
all during normal operation.

now, i read something about sleep mode, didn't find exactly what it is, but
i assume it's bad :

for the test i shutdown eth1 :

[root@w4 linux]# ./eepro100-diag -e
eepro100-diag.c:v2.11 8/27/2002 Donald Becker ([email protected])
 http://www.scyld.com/diag/index.html
Index #1: Found a Intel i82557/8/9 EtherExpressPro100 adapter at 0x2200.
A potential i82557 chip has been found, but it appears to be active.
Either shutdown the network, or use the '-f' flag.
Index #2: Found a Intel i82557/8/9 EtherExpressPro100 adapter at 0x2240.
Intel EtherExpress Pro 10/100 EEPROM contents:
  Station address 00:06:29:1F:1E:BD.
  Board assembly 754338-001, Physical connectors present: RJ45
  Primary interface chip i82555 PHY #1.
   Sleep mode is enabled.  This is not recommended.
   Under high load the card may not respond to
   PCI requests, and thus cause a master abort.
   To clear sleep mode use the '-G 0 -w -w -f' options.

i will run this tomorrow, when i am phisically near the box. (G
0 -w -w -f. )

but anyway, i would really like to know what can i expect from this card
when the load goes up ...