Re: [GIT PULL RESEND] clocksource/clockevents fixes for 3.13

[Ingo Molnar <mingo@kernel.org>]

* Daniel Lezcano <daniel.lezcano@linaro.org> wrote:

> 
> Hi Thomas, Ingo,
> 
> here are the fixes for 3.13 based on timers/urgent
> 
>  * Axel Lin added a missing dependency on CLKSRC_MMIO in the Kconfig
> for the time-efm32.
> 
>  * Dinh Nguyen fixed read_sched_clock to return the right value for
> the dw_apb_timer.
> 
>  * Ezequiel Garcia registered the sched clock after the counter,
> thus preventing time jump in the traces for the armada-370-xp.
> 
>  * Marc Zyngier stopped the timer before enabling the irq in order
> to prevent it to be fired before the clockevent is registered for
> the sunxi.
> 
>  * Thierry Reding removed a of_node_put in clksrc-of because the
> reference is not held.

(For future reference, please post all patches not just a combo 
patch.)

> Dinh Nguyen (2):
>       clocksource: dw_apb_timer: Move timer defines to header file.
>       clocksource: dw_apb_timer_of: Fix read_sched_clock

Nit: we generally don't use periods at the end of titles, it looks 
inconsistent in shortlogs.

> diff --git a/drivers/clocksource/dw_apb_timer.c
> b/drivers/clocksource/dw_apb_timer.c
> index e54ca10..c3a8f52 100644
> --- a/drivers/clocksource/dw_apb_timer.c
> +++ b/drivers/clocksource/dw_apb_timer.c
> @@ -18,25 +18,6 @@
>  #include <linux/io.h>
>  #include <linux/slab.h>
> 
> -#define APBT_MIN_PERIOD			4
> -#define APBT_MIN_DELTA_USEC		200
> -
> -#define APBTMR_N_LOAD_COUNT		0x00
> -#define APBTMR_N_CURRENT_VALUE		0x04
> -#define APBTMR_N_CONTROL		0x08
> -#define APBTMR_N_EOI			0x0c
> -#define APBTMR_N_INT_STATUS		0x10
> -
> -#define APBTMRS_INT_STATUS		0xa0
> -#define APBTMRS_EOI			0xa4
> -#define APBTMRS_RAW_INT_STATUS		0xa8
> -#define APBTMRS_COMP_VERSION		0xac
> -
> -#define APBTMR_CONTROL_ENABLE		(1 << 0)
> -/* 1: periodic, 0:free running. */
> -#define APBTMR_CONTROL_MODE_PERIODIC	(1 << 1)
> -#define APBTMR_CONTROL_INT		(1 << 2)
> -
>  static inline struct dw_apb_clock_event_device *
>  ced_to_dw_apb_ced(struct clock_event_device *evt)
>  {
> diff --git a/drivers/clocksource/dw_apb_timer_of.c
> b/drivers/clocksource/dw_apb_timer_of.c
> index 45ba8ae..c3fe17e 100644
> --- a/drivers/clocksource/dw_apb_timer_of.c
> +++ b/drivers/clocksource/dw_apb_timer_of.c
> @@ -102,18 +102,17 @@ static void __init add_clocksource(struct
> device_node *source_timer)
>  	 * timer is found. sched_io_base then points to the current_value
>  	 * register of the clocksource timer.
>  	 */
> -	sched_io_base = iobase + 0x04;
> +	sched_io_base = iobase;
>  	sched_rate = rate;
>  }
> 
>  static u64 read_sched_clock(void)
>  {
> -	return __raw_readl(sched_io_base);
> +	return ~__raw_readl(sched_io_base + APBTMR_N_CURRENT_VALUE);

AFAICS the minimal fix here for v3.13 would be to just do:

> -	return __raw_readl(sched_io_base);
> +	return ~__raw_readl(sched_io_base);

Right? That way the fix is just a oneliner and we could delay the 
header file and magic-constants use cleanups to v3.14.

Thanks,

	Ingo