2010-11-18 00:21:12

by Colin Cross

[permalink] [raw]
Subject: [PATCH] ARM: tegra: timer: Separate clocksource and sched_clock

tegra_clocksource_read should not use cnt32_to_63, wrapping is
already handled in the clocksource code. Move the cnt32_to_63
into the sched_clock function, and replace the use of clocksource
mult and shift with a multiplication by 1000 to convert us to ns.

Change-Id: Ica1158c3fbed0ba53391f7db61180129a957511d
Signed-off-by: Colin Cross <[email protected]>
---
arch/arm/mach-tegra/timer.c | 7 +++----
1 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/arch/arm/mach-tegra/timer.c b/arch/arm/mach-tegra/timer.c
index 9057d6f..b49f2f5 100644
--- a/arch/arm/mach-tegra/timer.c
+++ b/arch/arm/mach-tegra/timer.c
@@ -91,7 +91,7 @@ static void tegra_timer_set_mode(enum clock_event_mode mode,

static cycle_t tegra_clocksource_read(struct clocksource *cs)
{
- return cnt32_to_63(timer_readl(TIMERUS_CNTR_1US));
+ return timer_readl(TIMERUS_CNTR_1US);
}

static struct clock_event_device tegra_clockevent = {
@@ -106,14 +106,13 @@ static struct clocksource tegra_clocksource = {
.name = "timer_us",
.rating = 300,
.read = tegra_clocksource_read,
- .mask = 0x7FFFFFFFFFFFFFFFULL,
+ .mask = CLOCKSOURCE_MASK(32),
.flags = CLOCK_SOURCE_IS_CONTINUOUS,
};

unsigned long long sched_clock(void)
{
- return clocksource_cyc2ns(tegra_clocksource.read(&tegra_clocksource),
- tegra_clocksource.mult, tegra_clocksource.shift);
+ return cnt32_to_63(timer_readl(TIMERUS_CNTR_1US)) * 1000;
}

static irqreturn_t tegra_timer_interrupt(int irq, void *dev_id)
--
1.7.3.1


2010-11-18 00:27:17

by john stultz

[permalink] [raw]
Subject: Re: [PATCH] ARM: tegra: timer: Separate clocksource and sched_clock

On Wed, 2010-11-17 at 16:20 -0800, Colin Cross wrote:
> tegra_clocksource_read should not use cnt32_to_63, wrapping is
> already handled in the clocksource code. Move the cnt32_to_63
> into the sched_clock function, and replace the use of clocksource
> mult and shift with a multiplication by 1000 to convert us to ns.
>
> Change-Id: Ica1158c3fbed0ba53391f7db61180129a957511d
> Signed-off-by: Colin Cross <[email protected]>

Acked-by: John Stultz <[email protected]>

2010-11-19 19:09:23

by Linus Walleij

[permalink] [raw]
Subject: Re: [PATCH] ARM: tegra: timer: Separate clocksource and sched_clock

2010/11/18 Colin Cross <[email protected]>:

> tegra_clocksource_read should not use cnt32_to_63, wrapping is
> already handled in the clocksource code. ?Move the cnt32_to_63
> into the sched_clock function, and replace the use of clocksource
> mult and shift with a multiplication by 1000 to convert us to ns.
>
> Change-Id: Ica1158c3fbed0ba53391f7db61180129a957511d

The Linux kernel does not use Gerrit...

> Signed-off-by: Colin Cross <[email protected]>

Apart from that:
Acked-by: Linus Walleij <[email protected]>

Yours,
Linus Walleij

2010-12-13 13:51:56

by Russell King - ARM Linux

[permalink] [raw]
Subject: Re: [PATCH] ARM: tegra: timer: Separate clocksource and sched_clock

On Wed, Nov 17, 2010 at 04:20:15PM -0800, Colin Cross wrote:
> tegra_clocksource_read should not use cnt32_to_63, wrapping is
> already handled in the clocksource code. Move the cnt32_to_63
> into the sched_clock function, and replace the use of clocksource
> mult and shift with a multiplication by 1000 to convert us to ns.

Colin,

Mind if I pick this up into my tree - I'm doing other work with
clocksources and sched_clock() at the moment which will clash with
this patch.

Thanks.

2010-12-13 22:55:52

by Colin Cross

[permalink] [raw]
Subject: Re: [PATCH] ARM: tegra: timer: Separate clocksource and sched_clock

On Mon, Dec 13, 2010 at 5:51 AM, Russell King - ARM Linux
<[email protected]> wrote:
> On Wed, Nov 17, 2010 at 04:20:15PM -0800, Colin Cross wrote:
>> tegra_clocksource_read should not use cnt32_to_63, wrapping is
>> already handled in the clocksource code. ?Move the cnt32_to_63
>> into the sched_clock function, and replace the use of clocksource
>> mult and shift with a multiplication by 1000 to convert us to ns.
>
> Colin,
>
> Mind if I pick this up into my tree - I'm doing other work with
> clocksources and sched_clock() at the moment which will clash with
> this patch.

Sure, I'll need to coordinate the Tegra tree with your tree already
because of the GIC and SMP changes.