Re: [PATCH 0/3] Use Tegra's microsecond counter for udelay()

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Thu, Jun 12, 2014 at 01:38:28PM -0600, Stephen Warren wrote:
> On 06/12/2014 09:58 AM, Peter De Schrijver wrote:
> > This patchset introduces support for Tegra's microsecond counter as the
> > udelay() timer. This is useful on Tegra SoCs which do not have an arch timer
> > such as Tegra20 and Tegra30. Using the microsecond counter instead of a delay
> > based loop avoids potential problems during cpu frequency changes.
> > 
> > The set consists of 3 patches:
> > 
> > Patch 1 introduces a new call which is used by the ARM architecture delay
> > timer code to prevent changing the delay timer after calibration is finished
> > and thus can be in use.
> > 
> > Patch 2 adds logic to choose the delay timer with the highest resolution. This
> > allows the same registration code to be used on all Tegra SoCs and yet use the
> > higher resolution arch timer when available (eg on Tegra114 or Tegra124).
> > 
> > Patch 3 adds the actual delay timer code.
> > 
> > Patch set has been verified on ventana (Tegra20), beaver (Tegra30),
> > dalmore (Tegra114) and jetson TK1 (Tegra124).
> 
> Russell, Paul, do patches 1 and 2 look good to you? If so, if you can
> ack them, I'd be happy to queue this series in the Tegra git tree. If
> that doesn't work for you, please let me know who will apply these
> patches. Thanks.

When I saw them, I did have a question in the back of my mind about
the interlocking - whether we should just mandate that these are
always registered before delay calibration.  Then I remember the
mess that platforms made of the sched_clock() stuff.

So, I don't see any other way to ensure that platforms do the right
thing other than providing that kind of stick...  so I guess we'll
have to have them.

Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>

-- 
FTTC broadband for 0.8mile line: now at 9.7Mbps down 460kbps up... slowly
improving, and getting towards what was expected from it.