Re: Extremely high interrupt count / load on recent CURRENT



On Fri, Jul 27, 2007 at 01:02:16PM -0600, Scott Long wrote:
Can you try removing drivers one at time until it goes away? Also,

I'll give this a try later. It's faily easy to reproduce - just use the
machine for longer than 1 hour and it'll waste 50% time on interrupts...

what kind of machine is this?

This is a 2x Opteron 846 on a Tyan K8SE motherboard, running i386. 2GB
RAM. Anything else you need to know? FWIW, the dmesg is available at
http://rink.nu/tmp/dmesg-pitchfork.txt

Regards,

--
Rink P.W. Springer - http://rink.nu
"root is always right" -- the kernel


Scott


Rink Springer wrote:
Hi everyone,

Since my last csup, I'm encountering very high interrupt counts (and
thus, a high associated interrupt load) on my workstation.

vmstat -i reports:

interrupt total rate
irq1: atkbd0 26087 1
irq14: ata0 58 0
irq16: emu10kx0 1051388503 78968
irq17: nvidia0+ 1189799 89
irq20: ohci0 16596 1
irq21: ehci0 1 0
irq22: atapci1 1387849 104
irq23: atapci2 905538 68
irq24: arcmsr0 176404 13
irq27: isp0 85 0
irq28: bge0 37642180 2827
cpu0: timer 26627670 1999
cpu1: timer 26595670 1997
Total 1145956440 86071

The problem is, when I remove the snd_emu10kx driver (I load it as a
module in loader(8)), then irq28: bge0 starts to exhibit the same
pattern, generating thousands of interrupts.

My kernel is GENERIC + ULE - WITNESS, but the problem also
occured using plain GENERIC. dmesg is available at
http://rink.nu/tmp/dmesg-pitchfork.txt

Can anyone help me track this down?

Regards,


_______________________________________________
freebsd-current@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@xxxxxxxxxxx"
_______________________________________________
freebsd-current@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@xxxxxxxxxxx"



Relevant Pages

  • Extremely high interrupt count / load on recent CURRENT
    ... I'm encountering very high interrupt counts (and ... a high associated interrupt load) on my workstation. ... vmstat -i reports: ...
    (freebsd-current)
  • Re: Extremely high interrupt count / load on recent CURRENT
    ... Can you try removing drivers one at time until it goes away? ... I'm encountering very high interrupt counts (and ... a high associated interrupt load) on my workstation. ...
    (freebsd-current)
  • Re: throughput and interrupts
    ... I have discovered some IRQ conflicts and removed them, now I have single device per IRQ, but still i have near 20-30 % of interrupt load, when i download something through this router. ... Some people advised me to use polling, so I've built kernel with HZ=1000, polling and tried it. ... However, I think the real problem here is the performance difference, which suggests a problem that needs resolving, rather than optimization being required up front. ...
    (freebsd-current)
  • 3.4 & VAIO Cardbus/USB/SCSI?
    ... floppy is on sd0, connected by a USB. ... From dmesg, it seems the scsi_cb ... destination (floppy), at the same time..... ... >: couldn't map interrupt ...
    (comp.unix.bsd.openbsd.misc)
  • Re: throughput and interrupts
    ... single device per IRQ, but still i have near 20-30 % of interrupt load, ... CPU). ... HZ=1000, polling and tried it. ...
    (freebsd-current)