[2.6.7] Ehci controller interrupts like crazy on nforce2

[2.6.7] Ehci controller interrupts like crazy on nforce2

[Jonathan Filiatrault]

Here it is: another nforce2 hardware bug. The ehci controller seems to
send a massive number of interrupts to the kernel (264379 per second).
This uses about 5 to 10% of the cpu. This shows up in top in the
"hi"(hard interrupts) indicator. Nothing unusual shows up in the kernel
log. My system has an Asus A7N8X Nforce2 Board with an Athlon XP 2800+
mounted on it.

[joe@omega3:~]$ cat /proc/interrupts ; sleep 1; cat /proc/interrupts
           CPU0
  0:     583513          XT-PIC  timer
  1:       1279    IO-APIC-edge  i8042
  7:     137293    IO-APIC-edge  parport0
  8:          0    IO-APIC-edge  rtc
  9:          0   IO-APIC-level  acpi
14:      41463    IO-APIC-edge  ide0
15:         23    IO-APIC-edge  ide1
17:          9   IO-APIC-level  EMU10K1
18:      18584   IO-APIC-level  eth0
20:  121541873   IO-APIC-level  ehci_hcd
21:          0   IO-APIC-level  ohci_hcd
22:          0   IO-APIC-level  ohci_hcd
NMI:          0
LOC:     583348
ERR:          0
MIS:          0
           CPU0
  0:     584760          XT-PIC  timer
  1:       1280    IO-APIC-edge  i8042
  7:     137578    IO-APIC-edge  parport0
  8:          0    IO-APIC-edge  rtc
  9:          0   IO-APIC-level  acpi
14:      41505    IO-APIC-edge  ide0
15:         23    IO-APIC-edge  ide1
17:          9   IO-APIC-level  EMU10K1
18:      18641   IO-APIC-level  eth0
20:  121806252   IO-APIC-level  ehci_hcd
21:          0   IO-APIC-level  ohci_hcd
22:          0   IO-APIC-level  ohci_hcd
NMI:          0
LOC:     584595
ERR:          0
MIS:          0
[joe@omega3:~]$

Any help or random thougths are welcome.

Happy Hacking,
Joe

Re: [2.6.7] Ehci controller interrupts like crazy on nforce2

[Kalin KOZHUHAROV]

Jonathan Filiatrault wrote:
> Here it is: another nforce2 hardware bug. The ehci controller seems to
> send a massive number of interrupts to the kernel (264379 per second).
> This uses about 5 to 10% of the cpu. This shows up in top in the
> "hi"(hard interrupts) indicator. Nothing unusual shows up in the kernel
> log. My system has an Asus A7N8X Nforce2 Board with an Athlon XP 2800+
> mounted on it.
Well, I have the same system (board is Deluxe version) and no such situation.

> [joe@omega3:~]$ cat /proc/interrupts ; sleep 1; cat /proc/interrupts
>           CPU0
>  0:     583513          XT-PIC  timer
Hmm, I don't have this though. "IO-APIC-edge  timer" is the closest in my box.

>  1:       1279    IO-APIC-edge  i8042
>  7:     137293    IO-APIC-edge  parport0
>  8:          0    IO-APIC-edge  rtc
>  9:          0   IO-APIC-level  acpi
> 14:      41463    IO-APIC-edge  ide0
> 15:         23    IO-APIC-edge  ide1
> 17:          9   IO-APIC-level  EMU10K1
> 18:      18584   IO-APIC-level  eth0
> 20:  121541873   IO-APIC-level  ehci_hcd
> 21:          0   IO-APIC-level  ohci_hcd
> 22:          0   IO-APIC-level  ohci_hcd
> NMI:          0
> LOC:     583348
> ERR:          0
> MIS:          0
[snip]
> 20:  121806252   IO-APIC-level  ehci_hcd
[snip]
> Any help or random thougths are welcome.

Well, the "usual answer" is try enable/disable PREEMPT, upgrade BIOS.
Also IO-APIC and APIC as a whole.

Here is my board:
$ cat /proc/interrupts 
           CPU0       
  0:  415667195    IO-APIC-edge  timer
  1:     184380    IO-APIC-edge  i8042
  8:          2    IO-APIC-edge  rtc
  9:          0   IO-APIC-level  acpi
 14:     865112    IO-APIC-edge  ide0
 15:       9195    IO-APIC-edge  ide1
 19:   31371344   IO-APIC-level  nvidia
 20:    8356973   IO-APIC-level  ehci_hcd, NVidia nForce2, eth0
 21:          0   IO-APIC-level  ohci_hcd
 22:     798076   IO-APIC-level  ohci_hcd
NMI:          0 
LOC:  415691931 
ERR:          0
MIS:          0
$ bash -c 'cat /proc/interrupts ; sleep 1; cat /proc/interrupts'|  grep ehci
 20:    8357517   IO-APIC-level  ehci_hcd, NVidia nForce2, eth0
 20:    8357530   IO-APIC-level  ehci_hcd, NVidia nForce2, eth0


-- 
 || ~~~~~~~~~~~~~~~~~~~~~~ ||
(  ) http://ThinRope.net/ (  )
 || ______________________ ||

Re: [2.6.7] Ehci controller interrupts like crazy on nforce2

[Daniel Schmitt]

On Friday 09 July 2004 03:16, Jonathan Filiatrault wrote:
> Here it is: another nforce2 hardware bug. The ehci controller seems to
> send a massive number of interrupts to the kernel (264379 per second).
> This uses about 5 to 10% of the cpu. This shows up in top in the
> "hi"(hard interrupts) indicator. Nothing unusual shows up in the kernel
> log. My system has an Asus A7N8X Nforce2 Board with an Athlon XP 2800+
> mounted on it.

I've seen this problem on my board (Epox 8RDA3+) as well. The root cause seems 
to be interrupt link devices enabled without regard for the actual device 
status. A recent patch that delayed IRQ assignment to device activiation time 
fixed this for me; you might want to try a fresh -mm or -bk kernel.

Hope that helps,

Daniel.

Re: [2.6.7] Ehci controller interrupts like crazy on nforce2

[Jonathan Filiatrault]

>I've seen this problem on my board (Epox 8RDA3+) as well. The root cause seems 
>
>to be interrupt link devices enabled without regard for the actual device 
>status. A recent patch that delayed IRQ assignment to device activiation time 
>fixed this for me; you might want to try a fresh -mm or -bk kernel.
>
2.6.7-bk21 fixed this problem for me, thanks

>
>Hope that helps,
>
it sure did.

Happy Hacking,
Joe

Re: [2.6.7] Ehci controller interrupts like crazy on nforce2

[Len Brown]

On Thu, 2004-07-08 at 21:16, Jonathan Filiatrault wrote:
> Here it is: another nforce2 hardware bug. The ehci controller seems to
> send a massive number of interrupts to the kernel (264379 per second).
> This uses about 5 to 10% of the cpu. This shows up in top in the
> "hi"(hard interrupts) indicator. Nothing unusual shows up in the
> kernel
> log. My system has an Asus A7N8X Nforce2 Board with an Athlon XP 2800+
> mounted on it.
> 
> [joe@omega3:~]$ cat /proc/interrupts ; sleep 1; cat /proc/interrupts
>            CPU0
>   0:     583513          XT-PIC  timer

Please boot with "acpi_skip_timer_override" to fix your IRQ0.
Yes, this workaround should be invoked automatically for you.
No, it probably will not help your EHCI problem.

-Len


>   1:       1279    IO-APIC-edge  i8042
>   7:     137293    IO-APIC-edge  parport0
>   8:          0    IO-APIC-edge  rtc
>   9:          0   IO-APIC-level  acpi
> 14:      41463    IO-APIC-edge  ide0
> 15:         23    IO-APIC-edge  ide1
> 17:          9   IO-APIC-level  EMU10K1
> 18:      18584   IO-APIC-level  eth0
> 20:  121541873   IO-APIC-level  ehci_hcd
> 21:          0   IO-APIC-level  ohci_hcd
> 22:          0   IO-APIC-level  ohci_hcd
> NMI:          0
> LOC:     583348
> ERR:          0
> MIS:          0
>            CPU0