Re: [PATCH 21/21] time: Rework debugging variables so they aren't global

[Peter Zijlstra]

On Wed, Apr 01, 2015 at 08:34:41PM -0700, John Stultz wrote:
> Ingo suggested that the timekeeping debugging variables
> recently added should not be global, and should be tied
> to the timekeeper's read_base.

But why? its the same hardware clock for both tkr's. Surely if one goes
funny the other will too.

It doesn't make sense to duplicate this.

> I'm a little hesitant here, since the tkr structure
> has been carefully designed to fit in a cacheline.
> However, these additions are all at the end of the
> structure and are conditionally compiled.

That.

>  include/linux/timekeeper_internal.h | 18 +++++++++++++++++-
>  kernel/time/timekeeping.c           | 33 ++++++++++-----------------------
>  2 files changed, 27 insertions(+), 24 deletions(-)
> 
> diff --git a/include/linux/timekeeper_internal.h b/include/linux/timekeeper_internal.h
> index fb86963..9b33027 100644
> --- a/include/linux/timekeeper_internal.h
> +++ b/include/linux/timekeeper_internal.h
> @@ -20,9 +20,13 @@
>   * @shift:	Shift value for scaled math conversion
>   * @xtime_nsec: Shifted (fractional) nano seconds offset for readout
>   * @base:	ktime_t (nanoseconds) base time for readout
> + * @last_warning:   Warning ratelimiter (DEBUG_TIMEKEEPING)
> + * @underflow_seen: Underflow warning flag (DEBUG_TIMEKEEPING)
> + * @overflow_seen:  Overflow warning flag (DEBUG_TIMEKEEPING)
>   *
>   * This struct has size 56 byte on 64 bit. Together with a seqcount it
> - * occupies a single 64byte cache line.
> + * occupies a single 64byte cache line (when DEBUG_TIMEKEEPING is not
> + * enabled).
>   *
>   * The struct is separate from struct timekeeper as it is also used
>   * for a fast NMI safe accessors.
> @@ -36,6 +40,18 @@ struct tk_read_base {
>  	u32			shift;
>  	u64			xtime_nsec;
>  	ktime_t			base;
> +#ifdef CONFIG_DEBUG_TIMEKEEPING
> +	long			last_warning;
> +	/*
> +	 * These simple flag variables are managed
> +	 * without locks, which is racy, but ok since
> +	 * we don't really care about being super
> +	 * precise about how many events were seen,
> +	 * just that a problem was observed.
> +	 */
> +	int			underflow_seen;
> +	int			overflow_seen;
> +#endif
>  };

Yeah, so what is the likelyhood of a distro blanked enabling that debug?

I really don't like this much.

Also, you're now lacking a call to timekeeping_check_update(&tkr_raw),
you update stats there but have no way to report on them.
--
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/