[PATCH] sched: generalize CONFIG_IRQ_TIME_ACCOUNTING for X86 and ARM

Venki Pallipadi venki at google.com
Fri Feb 10 17:28:08 EST 2012


> +       noirqtime       [X86,ARM] Used to run time disable IRQ_TIME_ACCOUNTING,
> +                       should give a negligible performance improvement.

Can you reword the above "negligible performance improvement" above to
something

> +       noirqtime       [X86,ARM] Run time disables IRQ_TIME_ACCOUNTING and
> +                       eliminates the timestamping on irq/softirq entry/exit.

Otherwise,
Acked-by: Venkatesh Pallipadi <venki at google.com>


On Thu, Feb 9, 2012 at 9:25 AM, Dmitry Antipov
<dmitry.antipov at linaro.org> wrote:
> Generalize CONFIG_IRQ_TIME_ACCOUNTING  between X86 and
> ARM, move "noirqtime=" option to common debugging code.
> For a bit of backward compatibility, X86-specific option
> "tsc=noirqtime" is preserved, but issues a warning.
>
> Suggested-by: Yong Zhang <yong.zhang0 at gmail.com>
> Suggested-by: Russell King <rmk+kernel at arm.linux.org.uk>
> Suggested-by: Venki Pallipadi <venki at google.com>
> Signed-off-by: Dmitry Antipov <dmitry.antipov at linaro.org>
> ---
>  Documentation/kernel-parameters.txt |    9 +++++----
>  arch/arm/kernel/sched_clock.c       |    2 ++
>  arch/x86/Kconfig                    |   11 -----------
>  arch/x86/kernel/tsc.c               |   12 ++++++------
>  include/linux/sched.h               |   17 ++++++++++-------
>  kernel/sched/core.c                 |   20 +++++++++++---------
>  lib/Kconfig.debug                   |   12 ++++++++++++
>  7 files changed, 46 insertions(+), 37 deletions(-)
>
> diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
> index 033d4e6..666d20e 100644
> --- a/Documentation/kernel-parameters.txt
> +++ b/Documentation/kernel-parameters.txt
> @@ -1790,6 +1790,9 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
>        noirqdebug      [X86-32] Disables the code which attempts to detect and
>                        disable unhandled interrupt sources.
>
> +       noirqtime       [X86,ARM] Used to run time disable IRQ_TIME_ACCOUNTING,
> +                       should give a negligible performance improvement.
> +
>        no_timer_check  [X86,APIC] Disables the code which tests for
>                        broken timer IRQ sources.
>
> @@ -2636,10 +2639,8 @@ bytes respectively. Such letter suffixes can also be entirely omitted.
>                        as the stability checks done at bootup. Used to enable
>                        high-resolution timer mode on older hardware, and in
>                        virtualized environment.
> -                       [x86] noirqtime: Do not use TSC to do irq accounting.
> -                       Used to run time disable IRQ_TIME_ACCOUNTING on any
> -                       platforms where RDTSC is slow and this accounting
> -                       can add overhead.
> +                       [x86] noirqtime: obsoleted by "noirqtime" generic option,
> +                       see it's documentation for details.
>
>        turbografx.map[2|3]=    [HW,JOY]
>                        TurboGraFX parallel port interface
> diff --git a/arch/arm/kernel/sched_clock.c b/arch/arm/kernel/sched_clock.c
> index 5416c7c..30b5f89 100644
> --- a/arch/arm/kernel/sched_clock.c
> +++ b/arch/arm/kernel/sched_clock.c
> @@ -144,6 +144,8 @@ void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate)
>         */
>        cd.epoch_ns = 0;
>
> +       enable_sched_clock_irqtime();
> +
>        pr_debug("Registered %pF as sched_clock source\n", read);
>  }
>
> diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
> index 5bed94e..4759676 100644
> --- a/arch/x86/Kconfig
> +++ b/arch/x86/Kconfig
> @@ -805,17 +805,6 @@ config SCHED_MC
>          making when dealing with multi-core CPU chips at a cost of slightly
>          increased overhead in some places. If unsure say N here.
>
> -config IRQ_TIME_ACCOUNTING
> -       bool "Fine granularity task level IRQ time accounting"
> -       default n
> -       ---help---
> -         Select this option to enable fine granularity task irq time
> -         accounting. This is done by reading a timestamp on each
> -         transitions between softirq and hardirq state, so there can be a
> -         small performance impact.
> -
> -         If in doubt, say N here.
> -
>  source "kernel/Kconfig.preempt"
>
>  config X86_UP_APIC
> diff --git a/arch/x86/kernel/tsc.c b/arch/x86/kernel/tsc.c
> index a62c201..f1b2b63 100644
> --- a/arch/x86/kernel/tsc.c
> +++ b/arch/x86/kernel/tsc.c
> @@ -103,14 +103,15 @@ int __init notsc_setup(char *str)
>
>  __setup("notsc", notsc_setup);
>
> -static int no_sched_irq_time;
> -
>  static int __init tsc_setup(char *str)
>  {
>        if (!strcmp(str, "reliable"))
>                tsc_clocksource_reliable = 1;
> -       if (!strncmp(str, "noirqtime", 9))
> -               no_sched_irq_time = 1;
> +       if (!strncmp(str, "noirqtime", 9)) {
> +               printk(KERN_WARNING "tsc: tsc=noirqtime is "
> +                      "obsolete, use noirqtime instead\n");
> +               disable_sched_clock_irqtime();
> +       }
>        return 1;
>  }
>
> @@ -978,8 +979,7 @@ void __init tsc_init(void)
>        /* now allow native_sched_clock() to use rdtsc */
>        tsc_disabled = 0;
>
> -       if (!no_sched_irq_time)
> -               enable_sched_clock_irqtime();
> +       enable_sched_clock_irqtime();
>
>        lpj = ((u64)tsc_khz * 1000);
>        do_div(lpj, HZ);
> diff --git a/include/linux/sched.h b/include/linux/sched.h
> index 7d379a6..9b13f79 100644
> --- a/include/linux/sched.h
> +++ b/include/linux/sched.h
> @@ -1961,13 +1961,16 @@ extern void sched_clock_idle_wakeup_event(u64 delta_ns);
>  #endif
>
>  #ifdef CONFIG_IRQ_TIME_ACCOUNTING
> -/*
> - * An i/f to runtime opt-in for irq time accounting based off of sched_clock.
> - * The reason for this explicit opt-in is not to have perf penalty with
> - * slow sched_clocks.
> - */
> -extern void enable_sched_clock_irqtime(void);
> -extern void disable_sched_clock_irqtime(void);
> +extern int sched_clock_irqtime;
> +static inline void enable_sched_clock_irqtime(void)
> +{
> +       if (sched_clock_irqtime == -1)
> +               sched_clock_irqtime = 1;
> +}
> +static inline void disable_sched_clock_irqtime(void)
> +{
> +       sched_clock_irqtime = 0;
> +}
>  #else
>  static inline void enable_sched_clock_irqtime(void) {}
>  static inline void disable_sched_clock_irqtime(void) {}
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 5255c9d..a7ec043 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -757,18 +757,20 @@ static DEFINE_PER_CPU(u64, cpu_hardirq_time);
>  static DEFINE_PER_CPU(u64, cpu_softirq_time);
>
>  static DEFINE_PER_CPU(u64, irq_start_time);
> -static int sched_clock_irqtime;
>
> -void enable_sched_clock_irqtime(void)
> -{
> -       sched_clock_irqtime = 1;
> -}
> +/* -1 if not initialized, 0 if disabled with "noirqtime" kernel option
> + * or after unstable clock was detected, 1 if enabled and active.
> + */
> +int sched_clock_irqtime = -1;
>
> -void disable_sched_clock_irqtime(void)
> +static int __init irqtime_setup(char *str)
>  {
>        sched_clock_irqtime = 0;
> +       return 1;
>  }
>
> +__setup("noirqtime", irqtime_setup);
> +
>  #ifndef CONFIG_64BIT
>  static DEFINE_PER_CPU(seqcount_t, irq_time_seq);
>
> @@ -822,7 +824,7 @@ void account_system_vtime(struct task_struct *curr)
>        s64 delta;
>        int cpu;
>
> -       if (!sched_clock_irqtime)
> +       if (sched_clock_irqtime < 1)
>                return;
>
>        local_irq_save(flags);
> @@ -2853,7 +2855,7 @@ void account_process_tick(struct task_struct *p, int user_tick)
>        cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
>        struct rq *rq = this_rq();
>
> -       if (sched_clock_irqtime) {
> +       if (sched_clock_irqtime > 0) {
>                irqtime_account_process_tick(p, user_tick, rq);
>                return;
>        }
> @@ -2887,7 +2889,7 @@ void account_steal_ticks(unsigned long ticks)
>  void account_idle_ticks(unsigned long ticks)
>  {
>
> -       if (sched_clock_irqtime) {
> +       if (sched_clock_irqtime > 0) {
>                irqtime_account_idle_ticks(ticks);
>                return;
>        }
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 8745ac7..236e814 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -299,6 +299,18 @@ config SCHEDSTATS
>          application, you can say N to avoid the very slight overhead
>          this adds.
>
> +config IRQ_TIME_ACCOUNTING
> +       bool "Fine granularity task level IRQ time accounting"
> +       depends on X86 || ARM
> +       default n
> +       ---help---
> +         Select this option to enable fine granularity task irq time
> +         accounting. This is done by reading a timestamp on each
> +         transitions between softirq and hardirq state, so there can be a
> +         small performance impact.
> +
> +         If in doubt, say N here.
> +
>  config TIMER_STATS
>        bool "Collect kernel timers statistics"
>        depends on DEBUG_KERNEL && PROC_FS
> --
> 1.7.7.6
>



More information about the linux-arm-kernel mailing list