Re: [PATCH 2/2] x86/idle: use dynamic halt poll
From: Thomas Gleixner
Date: Tue Jul 04 2017 - 10:54:36 EST
On Tue, 4 Jul 2017, Radim Krčmář wrote:
> I think there is a nicer solution to avoid the expensive timer rewrite:
> Linux uses one-shot APIC timers and getting the timer interrupt is about
> as expensive as programming the timer, so the guest can keep the timer
> armed, but not re-arm it after the expiration if the CPU is idle.
So much for the theory. The NOHZ logic has to reprogram the timer when the
next expiry is farther away than the next tick. Otherwise you wake up on
every idle entry once for nothing, which defeats the whole purpose of NOHZ
to not do that.