Re: [PATCH v2 1/2] KVM: selftests: Provide generic way to read system counter

[Marc Zyngier <maz@kernel.org>]

On Tue, 28 Mar 2023 16:38:37 +0100,
Sean Christopherson <seanjc@google.com> wrote:
> 
> On Tue, Mar 28, 2023, Marc Zyngier wrote:
> > On Tue, 21 Mar 2023 19:10:04 +0000,
> > Colton Lewis <coltonlewis@google.com> wrote:
> > > In context, I'm trying to measure the time it takes to write to a buffer
> > > *with dirty memory logging enabled*. What do you mean by zero? I can
> > > confirm from running this code I am not measuring zero time.
> > 
> > See my earlier point: the counter tick is a few MHz, and the CPU
> > multiple GHz.
> 
> On x86, the system counter (TSC) counts at multiple GHz, so we
> should be able to continue with that approach for x86.
> 
> > So unless "whatever" is something that takes a significant time (several
> > thousands of CPU cycles), you'll measure nothing using the counter. Page
> > faults will probably show, but not a normal access.
> > 
> > The right tool for this job is to use PMU events, as they count at the CPU
> > frequency.
> 
> Out of curiosity, what does the kernel end up using for things like ndelay()?  I
> tried to follow the breadcrumbs for ARM and got as far as arm_arch_timer.c, but
> after that I'm more than a bit lost.

That's where it ends. We use the counter for everything. Even on
ARMv8.6+ HW that is supposed to give you a 1GHz counter,
implementations are allowed to perform "lumpy" increments (50MHz
counter with increments of 20 per tick, for example).

	M.

-- 
Without deviation from the norm, progress is not possible.