Bug #1182

intr 11 livelock with SMP

Added by mneumann over 5 years ago. Updated over 2 years ago.

Status:ClosedStart date:
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:-
Target version:-

Description

Hi,

UP works perfectly, SMP has this intr 11 livelock problem, which slows
it pretty down, so that network does not work. I post output of dmesg
for both UP and SMP as well as the output of vmstat -vi. Maybe those
information are helpful to someone :). I also tried SMP with a recent
kernel (the outputs below are of a kernel which is a few month old), but
it's even worse; it stops at "intr 7 livelocked limit engaged" during
ATA probing.

Looking at "vmstat -vi" it's obvious that the problem is with atapci1.
Also notably is the huge difference of "irq0: clk", which is ~1600 for
UP and ~13000 for SMP. I don't know exactly what rate means here, but
if it means interrupts/second and it's the clock interrupt which usually
fires at 1000 Hz, then something is seriously wrong.

Regards,

Michael

dmesg.smp (34.3 KB) mneumann, 12/16/2008 08:18 PM

dmesg.up (31.7 KB) mneumann, 12/16/2008 08:18 PM

vmstat.smp (1.04 KB) mneumann, 12/16/2008 08:18 PM

vmstat.up (1008 Bytes) mneumann, 12/16/2008 08:18 PM

History

#1 Updated by hasso over 5 years ago

Try to compile SMP kernel without IO_APIC option.

#2 Updated by mneumann over 5 years ago

This does not help.

Regards,

Michael

#3 Updated by sepherosa over 5 years ago

There are four cases could be tested:
compile a kernel w/ IO_APIC
boot w/ ACPI
boot w/o ACPI

compile a kernel w/o IO_APIC
boot w/ ACPI
boot w/o ACPI

Best Regards,
sephe

#4 Updated by mneumann over 5 years ago

Am 17.12.2008 06:52, schrieb Sepherosa Ziehau:
> On Wed, Dec 17, 2008 at 5:00 AM, Michael Neumann<>
wrote:
>> Am 16.12.2008 21:36, schrieb Hasso Tepper:
>>> Hasso Tepper<> added the comment:
>>>
>>> Try to compile SMP kernel without IO_APIC option.
>>>
>
> There are four cases could be tested:
> compile a kernel w/ IO_APIC
> boot w/ ACPI
> boot w/o ACPI
>
> compile a kernel w/o IO_APIC
> boot w/ ACPI
> boot w/o ACPI

Yes, I tried them all. But it didn't help.

Regards,

Michael

#5 Updated by corecode over 5 years ago

could you please post the output of mptable(1)?

#6 Updated by swildner over 2 years ago

  • Description updated (diff)
  • Status changed from New to Closed
  • Assignee deleted (0)

Assume it is fixed with all the recent interrupt work. If it still occurs, please reopen.

Also available in: Atom PDF