Hi,
I'm trying to get a PCI device (network) to work on one of my
boxes, and thus far i've only managed to get it to work with "noapic".
Without that kernel option the card stops processing interrupts (as can be
observed in /proc/interrupts) after about ~100 on each CPU. Are there any
things i should look out for when trying to track down the problem?
I've tried swapping the card around in different slots just in case it
was an IRQ routing problem (as suggested by various folks) to no avail.
The box is an IBM Netfinity 3500M20 SMP rig.
Any suggestions would be greatly appreciated,
Regards,
Zwane Mwaikambo
In article <[email protected]>,
Zwane Mwaikambo <[email protected]> wrote:
> I'm trying to get a PCI device (network) to work on one of my
>boxes, and thus far i've only managed to get it to work with "noapic".
>Without that kernel option the card stops processing interrupts (as can be
>observed in /proc/interrupts) after about ~100 on each CPU. Are there any
>things i should look out for when trying to track down the problem?
>I've tried swapping the card around in different slots just in case it
>was an IRQ routing problem (as suggested by various folks) to no avail.
>The box is an IBM Netfinity 3500M20 SMP rig.
What does cat /proc/interrupts say in APIC mode? Perhaps it thinks the
device uses edge-triggered interrupts instead of level-triggered ?
Mike.
--
"Only two things are infinite, the universe and human stupidity,
and I'm not sure about the former" -- Albert Einstein.
It shows up as level-triggered.
CPU0 CPU1
0: 4498447 0 local-APIC-edge timer
1: 465 6921 IO-APIC-edge keyboard
2: 0 0 XT-PIC cascade
8: 0 1 IO-APIC-edge rtc
14: 2 3 IO-APIC-edge ide0
16: 59755 52687 IO-APIC-level dpti0
18: 386054 576702 IO-APIC-level eth0
22: 0 0 IO-APIC-level pentanet0 <==
27: 25743 20344 IO-APIC-level eth1
NMI: 0 0
LOC: 4498376 4498509
Thanks,
Zwane Mwaikambo
According to Zwane Mwaikambo:
> It shows up as level-triggered.
>
> 22: 0 0 IO-APIC-level pentanet0 <==
OK. If it showed up as edge I would know what was wrong, but
level is correct.
Mike.
--
"Only two things are infinite, the universe and human stupidity,
and I'm not sure about the former" -- Albert Einstein.
On Fri, 30 Nov 2001, Zwane Mwaikambo wrote:
> 22: 0 0 IO-APIC-level pentanet0 <==
Do you have sources for the driver? Last time I looked at the driver, it
was binary-only (the distribution contained a copy of the GNU GPL, yet the
vendor refused to release sources I asked for) and it seemed to be broken
horribly. Be happy at least it works for you with interrupts routed
through the 8259A.
--
+ Maciej W. Rozycki, Technical University of Gdansk, Poland +
+--------------------------------------------------------------+
+ e-mail: [email protected], PGP key available +
On Fri, 30 Nov 2001, Maciej W. Rozycki wrote:
> Do you have sources for the driver? Last time I looked at the driver, it
> was binary-only (the distribution contained a copy of the GNU GPL, yet the
> vendor refused to release sources I asked for) and it seemed to be broken
> horribly. Be happy at least it works for you with interrupts routed
> through the 8259A.
Nope they provide some source as "glue" and do the final linking
against some object files. I've been hacking on the glue code, so far
i've managed to get multiple card detection working and with some
help removed some races but who knows what lurks in those .o's.
Cheers,
Zwane Mwaikambo
On Fri, 30 Nov 2001, Zwane Mwaikambo wrote:
> Nope they provide some source as "glue" and do the final linking
> against some object files. I've been hacking on the glue code, so far
> i've managed to get multiple card detection working and with some
> help removed some races but who knows what lurks in those .o's.
Good luck. I wouldn't bother myself.
--
+ Maciej W. Rozycki, Technical University of Gdansk, Poland +
+--------------------------------------------------------------+
+ e-mail: [email protected], PGP key available +
On Fri, 30 Nov 2001, Mark Hahn wrote:
> could your machine be one that contains
> a watchdog card? some IBM servers do,
> and as I recall, the card does something funky
> (perhaps just NMI) when it thinks there's a problem.
i don't believe it does, but i'm not 100% sure. This server is essentially
an entry level box so it might not have it
Cheers,
Zwane