Re: [PATCH] clocksource: Avoid selecting mult values that mightoverflow when adjusted

From: Yong Zhang
Date: Mon Nov 07 2011 - 22:11:53 EST

Next message: Axel Lin: "Re: [PATCH 2/2] mfd: include linux/module.h for ab5500-debugfs"
Previous message: Daniil Stolnikov: "Add IPSec IP Range in Linux kernel"
In reply to: John Stultz: "Re: [PATCH] clocksource: Avoid selecting mult values that mightoverflow when adjusted"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Mon, Nov 07, 2011 at 07:09:00PM -0800, John Stultz wrote:
> On Thu, 2011-11-03 at 22:10 +0100, Ingo Molnar wrote:
> > * John Stultz <john.stultz@xxxxxxxxxx> wrote:
> >
> > > For some frequqencies, the clocks_calc_mult_shift() function will
> > > unfortunately select mult values very close to 0xffffffff. This
> > > has the potential to overflow when NTP adjusts the clock, adding
> > > to the mult value.
> > >
> > > This patch adds a clocksource.maxadj value, which provides
> > > an approximation of an 11% adjustment(NTP limits adjustments to
> > > 500ppm and the tick adjustment is limited to 10%), which could
> > > be made to the clocksource.mult value. This is then used to both
> > > check that the current mult value won't overflow/underflow, as
> > > well as warning us if the timekeeping_adjust() code pushes over
> > > that 11% boundary.
> > >
> > > CC: Yong Zhang <yong.zhang0@xxxxxxxxx>
> > > CC: David Daney <ddaney.cavm@xxxxxxxxx>
> > > CC: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > > Reported-by: Chen Jie <chenj@xxxxxxxxxx>
> > > Reported-by: zhangfx <zhangfx@xxxxxxxxxx>
> > > Signed-off-by: John Stultz <john.stultz@xxxxxxxxxx>
> > > ---
> > > include/linux/clocksource.h | 3 +-
> > > kernel/time/clocksource.c | 53 ++++++++++++++++++++++++++++++++++--------
> > > kernel/time/timekeeping.c | 3 ++
> > > 3 files changed, 48 insertions(+), 11 deletions(-)
> >
> > This patch (included in tip:timers/urgent) causes the following boot
> > warning x86:
> >
> > [ 0.000000] Fast TSC calibration using PIT
> > [ 0.000000] ------------[ cut here ]------------
> > [ 0.000000] WARNING: at kernel/time/timekeeping.c:855 do_timer+0x47f/0x4c0()
> > [ 0.000000] Hardware name: System Product Name
> > [ 0.000000] Adjusting more then 11%
> > [ 0.000000] Modules linked in:
> > [ 0.000000] Pid: 0, comm: swapper Not tainted 3.1.0-tip+ #161792
> > [ 0.000000] Call Trace:
> > [ 0.000000] <IRQ> [<ffffffff81042d0a>] warn_slowpath_common+0x7a/0xb0
> > [ 0.000000] [<ffffffff81042de1>] warn_slowpath_fmt+0x41/0x50
> > [ 0.000000] [<ffffffff8106e78f>] do_timer+0x47f/0x4c0
> > [ 0.000000] [<ffffffff81073953>] tick_periodic+0x63/0x80
> > [ 0.000000] [<ffffffff810739f1>] tick_handle_periodic+0x21/0x70
> > [ 0.000000] [<ffffffff810046d8>] timer_interrupt+0x18/0x20
> > [ 0.000000] [<ffffffff8109ff9e>] handle_irq_event_percpu+0x5e/0x220
> > [ 0.000000] [<ffffffff810a019b>] handle_irq_event+0x3b/0x60
> > [ 0.000000] [<ffffffff810a295c>] handle_level_irq+0x6c/0xd0
> > [ 0.000000] [<ffffffff81003f34>] handle_irq+0x44/0xa0
> >
> > Full bootlog and config attached.
> >
> > i've excluded it from tip:master for now.
>
> Thanks again for the bug report. I was able to reproduce it using the
> jiffies clocksource. Looking at the code after a weekend of decent
> sleep, I see the max_adjustment calculation is simply wrong (it was
> proportional to the shift, not the mult - but adjustments are made on
> mult).
>
> This version simplifies the calculation and improves warn-on messages so
> we also catch any overflow potential on clocksources that don't use the
> clocksource_register_hz/khz interfaces.
>
> Ingo: could you give it a whirl on your test box and verify it doesn't
> have any trouble?
>
> Yong: Can you also give this a test run to make sure you don't see any
> problems?

Will do :)

Thanks,
Yong

>
> thanks
> -john
>
> >From 82c5b70fc5074b6bb6e05514afb6e9c73c740422 Mon Sep 17 00:00:00 2001
> From: John Stultz <john.stultz@xxxxxxxxxx>
> Date: Mon, 31 Oct 2011 17:06:35 -0400
> Subject: [PATCH] clocksource: Avoid selecting mult values that might overflow when adjusted
>
> For some frequqencies, the clocks_calc_mult_shift() function will
> unfortunately select mult values very close to 0xffffffff. This
> has the potential to overflow when NTP adjusts the clock, adding
> to the mult value.
>
> This patch adds a clocksource.maxadj value, which provides
> an approximation of an 11% adjustment(NTP limits adjustments to
> 500ppm and the tick adjustment is limited to 10%), which could
> be made to the clocksource.mult value. This is then used to both
> check that the current mult value won't overflow/underflow, as
> well as warning us if the timekeeping_adjust() code pushes over
> that 11% boundary.
>
> v2: Fix max_adjustment calculation, and improve WARN_ONCE
> messages.
>
> CC: Yong Zhang <yong.zhang0@xxxxxxxxx>
> CC: David Daney <ddaney.cavm@xxxxxxxxx>
> CC: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Reported-by: Chen Jie <chenj@xxxxxxxxxx>
> Reported-by: zhangfx <zhangfx@xxxxxxxxxx>
> Signed-off-by: John Stultz <john.stultz@xxxxxxxxxx>
> ---
> include/linux/clocksource.h | 3 +-
> kernel/time/clocksource.c | 58 +++++++++++++++++++++++++++++++++++-------
> kernel/time/timekeeping.c | 6 ++++
> 3 files changed, 56 insertions(+), 11 deletions(-)
>
> diff --git a/include/linux/clocksource.h b/include/linux/clocksource.h
> index 139c4db..c86c940 100644
> --- a/include/linux/clocksource.h
> +++ b/include/linux/clocksource.h
> @@ -156,6 +156,7 @@ extern u64 timecounter_cyc2time(struct timecounter *tc,
> * @mult: cycle to nanosecond multiplier
> * @shift: cycle to nanosecond divisor (power of two)
> * @max_idle_ns: max idle time permitted by the clocksource (nsecs)
> + * @maxadj maximum adjustment value to mult (~11%)
> * @flags: flags describing special properties
> * @archdata: arch-specific data
> * @suspend: suspend function for the clocksource, if necessary
> @@ -172,7 +173,7 @@ struct clocksource {
> u32 mult;
> u32 shift;
> u64 max_idle_ns;
> -
> + u32 maxadj;
> #ifdef CONFIG_ARCH_CLOCKSOURCE_DATA
> struct arch_clocksource_data archdata;
> #endif
> diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c
> index cf52fda..cfc65e1 100644
> --- a/kernel/time/clocksource.c
> +++ b/kernel/time/clocksource.c
> @@ -492,6 +492,22 @@ void clocksource_touch_watchdog(void)
> }
>
> /**
> + * clocksource_max_adjustment- Returns max adjustment amount
> + * @cs: Pointer to clocksource
> + *
> + */
> +static u32 clocksource_max_adjustment(struct clocksource *cs)
> +{
> + u64 ret;
> + /*
> + * We won't try to correct for more then 11% adjustments (110,000 ppm),
> + */
> + ret = (u64)cs->mult * 11;
> + do_div(ret,100);
> + return (u32)ret;
> +}
> +
> +/**
> * clocksource_max_deferment - Returns max time the clocksource can be deferred
> * @cs: Pointer to clocksource
> *
> @@ -503,25 +519,28 @@ static u64 clocksource_max_deferment(struct clocksource *cs)
> /*
> * Calculate the maximum number of cycles that we can pass to the
> * cyc2ns function without overflowing a 64-bit signed result. The
> - * maximum number of cycles is equal to ULLONG_MAX/cs->mult which
> - * is equivalent to the below.
> - * max_cycles < (2^63)/cs->mult
> - * max_cycles < 2^(log2((2^63)/cs->mult))
> - * max_cycles < 2^(log2(2^63) - log2(cs->mult))
> - * max_cycles < 2^(63 - log2(cs->mult))
> - * max_cycles < 1 << (63 - log2(cs->mult))
> + * maximum number of cycles is equal to ULLONG_MAX/(cs->mult+cs->maxadj)
> + * which is equivalent to the below.
> + * max_cycles < (2^63)/(cs->mult + cs->maxadj)
> + * max_cycles < 2^(log2((2^63)/(cs->mult + cs->maxadj)))
> + * max_cycles < 2^(log2(2^63) - log2(cs->mult + cs->maxadj))
> + * max_cycles < 2^(63 - log2(cs->mult + cs->maxadj))
> + * max_cycles < 1 << (63 - log2(cs->mult + cs->maxadj))
> * Please note that we add 1 to the result of the log2 to account for
> * any rounding errors, ensure the above inequality is satisfied and
> * no overflow will occur.
> */
> - max_cycles = 1ULL << (63 - (ilog2(cs->mult) + 1));
> + max_cycles = 1ULL << (63 - (ilog2(cs->mult + cs->maxadj) + 1));
>
> /*
> * The actual maximum number of cycles we can defer the clocksource is
> * determined by the minimum of max_cycles and cs->mask.
> + * Note: Here we subtract the maxadj to make sure we don't sleep for
> + * too long if there's a large negative adjustment.
> */
> max_cycles = min_t(u64, max_cycles, (u64) cs->mask);
> - max_nsecs = clocksource_cyc2ns(max_cycles, cs->mult, cs->shift);
> + max_nsecs = clocksource_cyc2ns(max_cycles, cs->mult - cs->maxadj,
> + cs->shift);
>
> /*
> * To ensure that the clocksource does not wrap whilst we are idle,
> @@ -640,7 +659,6 @@ static void clocksource_enqueue(struct clocksource *cs)
> void __clocksource_updatefreq_scale(struct clocksource *cs, u32 scale, u32 freq)
> {
> u64 sec;
> -
> /*
> * Calc the maximum number of seconds which we can run before
> * wrapping around. For clocksources which have a mask > 32bit
> @@ -661,6 +679,20 @@ void __clocksource_updatefreq_scale(struct clocksource *cs, u32 scale, u32 freq)
>
> clocks_calc_mult_shift(&cs->mult, &cs->shift, freq,
> NSEC_PER_SEC / scale, sec * scale);
> +
> + /*
> + * for clocksources that have large mults, to avoid overflow.
> + * Since mult may be adjusted by ntp, add an safety extra margin
> + *
> + */
> + cs->maxadj = clocksource_max_adjustment(cs);
> + while ((cs->mult + cs->maxadj < cs->mult)
> + || (cs->mult - cs->maxadj > cs->mult)) {
> + cs->mult >>= 1;
> + cs->shift--;
> + cs->maxadj = clocksource_max_adjustment(cs);
> + }
> +
> cs->max_idle_ns = clocksource_max_deferment(cs);
> }
> EXPORT_SYMBOL_GPL(__clocksource_updatefreq_scale);
> @@ -701,6 +733,12 @@ EXPORT_SYMBOL_GPL(__clocksource_register_scale);
> */
> int clocksource_register(struct clocksource *cs)
> {
> + /* calculate max adjustment for given mult/shift */
> + cs->maxadj = clocksource_max_adjustment(cs);
> + WARN_ONCE(cs->mult + cs->maxadj < cs->mult,
> + "Clocksource %s might overflow on 11%% adjustment\n",
> + cs->name);
> +
> /* calculate max idle time permitted for this clocksource */
> cs->max_idle_ns = clocksource_max_deferment(cs);
>
> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
> index 2b021b0e..2c04610 100644
> --- a/kernel/time/timekeeping.c
> +++ b/kernel/time/timekeeping.c
> @@ -820,6 +820,12 @@ static void timekeeping_adjust(s64 offset)
> } else
> return;
>
> + WARN_ONCE(timekeeper.mult+adj >
> + timekeeper.clock->mult + timekeeper.clock->maxadj,
> + "Adjusting %s more then 11%% (%ld vs %ld)\n",
> + timekeeper.clock->name, (long)timekeeper.mult+adj,
> + (long)timekeeper.clock->mult +
> + timekeeper.clock->maxadj);
> timekeeper.mult += adj;
> timekeeper.xtime_interval += interval;
> timekeeper.xtime_nsec -= offset;
> --
> 1.7.3.2.146.gca209
>
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/

--
Only stand for myself
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Axel Lin: "Re: [PATCH 2/2] mfd: include linux/module.h for ab5500-debugfs"
Previous message: Daniil Stolnikov: "Add IPSec IP Range in Linux kernel"
In reply to: John Stultz: "Re: [PATCH] clocksource: Avoid selecting mult values that mightoverflow when adjusted"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]