[EMAIL PROTECTED] wrote:
> Under Mandrake 6.1, on a laptop with a PCMCIA NE2000
> compatible network card, I get lots and lots of kernel messages
> complaining about dropped interrupts, such as like:
>
> kernel: eth0: interrupt(s) dropped!
>
> The card is recognized by the PCMCIA system (see the following
> messages when the card is inserted).
>
> cardmgr[336]: initializing socket 0
> cardmgr[336]: socket 0: RPTI EP401 Ethernet
> cardmgr[336]: executing: 'insmod /lib/modules/2.2.13-7mdk/net/8390.o'
> cardmgr[336]: executing: 'insmod /lib/modules/2.2.13-7mdk/pcmcia/pcnet_cs.o'
> kernel: eth0: NE2000 Compatible: io 0x300, irq 10, hw_addr 00:40:59:04:43:48
>
> The same card did not generate these problems under RedHat 5.1
> (albeit on a different laptop).
>
> Does anybody know what the problem is -- and what the solution
> is, more importantly :-)
>
> Thanks, Fabio
Hmmmm,
I do believe one theoretical possibility is that higher-level interrupts are
being serviced (in a masked state) when the eth0 interrupt occurs and the
resulting time delay in getting to the eth0 Interrupt Service Routine (ISR)
takes too long, so that the waiting message has "passed the frame" and data is
dropped. Since it is TCP/IP (most likely), the result is less efficiency in
communication rather than a catastrophic crash. Our old acquaintance, the
accidentally shared interrupt, would most likely cause an unresponsive system
with no error messages.
If your Hard Disk Drive is higher in priority, and if it is doing a lot of
swapping, this could happen. "Cures" would be as follows:
1. Make sure linux is recognizing and using memory properly
a. run # free in an xterm to see if the memory recognized is the same as the
memory you have.
b. setup a "mem=64M" in /etc/lilo.conf (includng the quotes and for your
amount of RAM)
But first, observe what is happening when the interrupt is dropped. If you have
a lot of HDD activity when these messages appear then this is a LIKELY culprit.
2. Add memory
3. Reconfigure network services to shorten the MTU and MRU you are using
3 a. If you are on a network using dhcp, try going to a static address (the one
leased to you) to eliminate a LOT of unnecessary traffic (esp with NT)
4. Go to
http://www.best.com/~cae/irqtune
and give eth0 higher or highest priority for ISRs. by "dynamically patching" the
kernel.
If it is not that, then I would begin to suspect the need for a full bug report,
with logs. Still, that seems unlikely. I am running 6.1 on a laptop using a
3C589 and it works without dropped interrupts.
Civileme
