Re: [PATCH] Dynamic tick, version 050127-1

[Tony Lindgren <tony@atomide.com>]

* Tony Lindgren <tony@atomide.com> [050204 11:14]:
> * Zwane Mwaikambo <zwane@arm.linux.org.uk> [050204 09:54]:
> > On Fri, 4 Feb 2005, Tony Lindgren wrote:
> > 
> > > * Zwane Mwaikambo <zwane@arm.linux.org.uk> [050204 09:31]:
> > > > On Fri, 4 Feb 2005, Tony Lindgren wrote:
> > > > 
> > > > > Yes, it's safer to keep the timer periodic, although it's
> > > > > used for oneshot purposes for the skips. If the timer interrupt
> > > > > got missed for some reason, the system would be able to recover when
> > > > > it's in periodic mode.
> > > > > 
> > > > > And with some timers, we can do the reprogramming faster, as we just
> > > > > need to load the new value.
> > > > > 
> > > > > I could not figure out how to disable the interrupts for PIT
> > > > > when local APIC is used and the ticks to skip is longer than PIT
> > > > > would allow. So I just changed the mode temporarily to disable it.
> > > > >
> > > > > Does anybody know if there's a way to stop PIT interrupts while
> > > > > keeping it in the periodic mode?
> > > > 
> > > > disable_irq(0) ?
> > > 
> > > Then the problem is that the CPU does not stay in sleep but wakes to
> > > the first PIT interrupt AFAIK.
> > 
> > I do not understand, do you want to disable the PIT from interrupting the 
> > processor and enable it interrupting at a later time?
> 
> Yes, that right. PIT max skip ticks = 54 and local APIC timer > 1000.
> PIT interrupt needs to be disabled to stay in sleep for over 54 ticks.
> 
> But I think you're right, disable_irq(0) should do the trick :)
> 
> Hmmm, we should be able to keep PIT irq disabled all the time when using
> local APIC timer. I'll play with it a bit.

Oops, no, PIT must be running at least when the system is busy.
Otherwise time won't get updated during load, as we never get to the
idle loop.

Tony