Re: Freeze due to performance_cx_lowest=LOW

From: John Baldwin <jhb_at_freebsd.org>
Date: Fri, 16 Jun 2006 15:09:28 -0400
On Friday 16 June 2006 07:39, Andriy Gapon wrote:
> on 01/06/2006 22:04 John Baldwin said the following:
> > On Tuesday 30 May 2006 23:19, Nate Lawson wrote:
> >> Andriy Gapon wrote:
> >>> on 29/05/2006 23:10 Nate Lawson said the following:
> >>>> disable apic (hint.apic.0.disabled="1").  I isolated this a little 
while 
> >>>> ago to the change to enable LAPIC timer.  However, there is currently 
no 
> >>>> easy way to disable the LAPIC timer with APIC enabled so you have to 
> >>>> disable APIC.  jhb_at_ and I have been discussing how to do this better 
but 
> >>>> no easy answers apparently.
> >>> I am not sure what I am talking about, but is it potentially possible to
> >>> drive timer system by more than one clock, actively using the most
> >>> precise of them at any particular moment ? So that if LAPIC timer stops
> >>> i8254 can be used instead.
> >> That requires mixed mode delivery -- i8254 on legacy PIC irq and APIC 
> >> mode for other interrupts.  jhb_at_ just killed this and isn't eager to add 
> >> it back.  I'll let him explain if he has more to add.
> > 
> > Namely that it is unreliable and expressly forbidden by the ACPI spec.
> > 
> 
> Maybe I will say something too ignorant for this list, but is it
> possible to drive hardclock with two interrupts (I haven't thought yet
> how, though) and use RTC as the second interrupt source ?
> I think that RTC/IRQ8 (usually) doesn't have problems associated with
> 8254 timer/IRQ0 and can be used without mixed mode.

Well, you could use the RTC instead of the lapic timer on CPU 0 and then use 
IPIs to forward clock interrupts to all the other CPUs just as we did before 
we used the lapic timer.

-- 
John Baldwin
Received on Fri Jun 16 2006 - 19:00:30 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:57 UTC