I have a server at work that seems to have a rather serious problem with
IRQ routing. I noticed that it's clock was drifting rather heavily a while
back - then I spotted that if I flood pinged it it's clock would go mad
(drift a few days forward in a matter of seconds). At the time it had an
old rather heavily loaded (ipsec and god knows what all turned on) kernel
- so I rebuilt it a nice clean 2.4.16 install and restarted it.
A few weeks later the problem is now back! The system has 3 ethernet cards
- but ONLY packets to eth0 seem to trigger the problem.
eth0: OEM i82557/i82558 10/100 Ethernet, 00:D0:B7:20:6F:89, IRQ 11.
Board assembly 733470-004, Physical connectors present: RJ45
Primary interface chip i82555 PHY #1.
General self-test: passed.
Serial sub-system self-test: passed.
Internal registers self-test: passed.
ROM checksum self-test: passed (0x04f4518b).
# cat /proc/interrupts
CPU0 CPU1
0: 406778504 421411683 IO-APIC-edge timer
1: 1 1 IO-APIC-edge keyboard
2: 0 0 XT-PIC cascade
8: 1 0 IO-APIC-edge rtc
16: 67601276 67572131 IO-APIC-level eth2
17: 110005783 110266175 IO-APIC-level eth0
18: 1549893 1553272 IO-APIC-level megaraid
19: 186036215 186229458 IO-APIC-level eth1
NMI: 0 0
LOC: 826656995 826656993
ERR: 0
MIS: 0
If a flood ping the system (from a machine on eth0's subnet) I can do:
# date; date; date; date; date; date
Mon Mar 25 11:35:26 GMT 2002
Mon Mar 25 11:35:28 GMT 2002
Mon Mar 25 11:35:29 GMT 2002
Mon Mar 25 11:35:31 GMT 2002
Mon Mar 25 11:35:33 GMT 2002
Mon Mar 25 11:35:34 GMT 2002
I also see the timer interrupt count (from /proc/interrupts) incrementing
for each eth0 interrupt.
I've attached my .config file - the machine also runs a few CIPE tunnels,
so has it's kernel module loaded.
Anyone have any suggestions as to what might be causing this or how to fix
it? If there's any more info req then please email me.
Cheers,
Chris
-- Chris Wilson jakdaw@lists.jakdaw.org
This archive was generated by hypermail 2b29 : Sun Mar 31 2002 - 22:00:09 EST