Re: [PATCH] ARM: convert arm/arm64 arch timer to use CLKSRC_OF init

[Catalin Marinas <catalin.marinas@arm.com>]

On Mon, Mar 25, 2013 at 09:28:10PM +0000, Rob Herring wrote:
> On 03/25/2013 12:26 PM, Russell King - ARM Linux wrote:
> > On Thu, Mar 21, 2013 at 11:06:47AM +0000, Mark Rutland wrote:
> >> On TC2 this series leads to using the vexpress 24MHz clock as the sched clock
> >> in preference to the architected timer:
> >>
> >>   Architected local timer running at 24.00MHz (virt).
> >>   Switching to timer-based delay loop
> >>   Registered arch_counter_get_cntvct+0x0/0x14 as sched_clock source
> >>   sched_clock: 32 bits at 24MHz, resolution 41ns, wraps every 178956ms
> >>   Registered versatile_read_sched_clock+0x0/0x28 as sched_clock source
> >>
> >> As they both have the same frequency, neither overrides the other, and
> >> whichever gets registered last is used as the sched_clock. As accesses
> >> to the architected timer are going to have a much lower overhead, this
> >> isn't very nice (and it could be better to use it even if it had a lower
> >> frequency).
> > 
> > I'll remind people that sched_clock() is supposed to be functional at
> > the point in the boot sequence where the call to sched_init() is called.
> > That is after setup_arch() and *before* time_init() is called.
> 
> I count integrator-cp, realview, versatile and non-DT VExpress that do
> this (not surprisingly) and 25 platforms or timer implementations plus
> arm64 that do sched_clock setup in time_init.

Before time_init(), sched_clock() currently returns 0 with the
architected timers (though I don't particularly like this for arm64).
Marc Rutland has patches to make arch_timer_read_counter() a function
which always returns the virtual counter. It requires the CNTVOFF
register to be set to 0 on AArch32 during boot. But this way
sched_clock() on arm64 would always return meaningful values as we have
the architected timers.

-- 
Catalin