[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Curious interrupt for ne2k-pci

I have one EB64+ alpha with "PCI NE2000 clone 'RealTek RTL-8029'" as a
second ethernet card (the first one is 3c509).  Ethernet interfaces
are on different networks. The host runs 2.0.33 kernel with ne2k-pci
for this card driver and the card simply does not work at all with the
the "other" ne2000 driver.

Both cards are functional but the problem is that a performance on
ne2000 is not overwhelming - to put it mildly.  Digging around for
possible reasons I noticed a curious thing.  Here are few fragments
from repeated 'cat /proc/interrupts':
17:   565794   eth1
22:   565794 + ncr53c8xx
17:   565820   eth1
22:   565820 + ncr53c8xx
17:   565838   eth1
22:   565838 + ncr53c8xx

Does anybody know if the above is a reporting artifact (or a bug in
"proc" interface?) or lines 17 and 22 are indeed the same interrupt
shared by the card and a SCSI controller?  Something like 2 and 9 lines
on (some?)  Intel boards?  I do not see a method of forcing one of
these two devices to use some other interrupt - I am afraid.  Whatever
is assigned by PCI is assigned.

Another interesting thing is that I was testing this particular card
in another, somewhat similar, EB64+ machine and I did not notice any
problems with performance or otherwise (excluding required use
of ne2k-pci.c to get anything at all).


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []