Re: [PATCH 08/12] time: Add warnings when overflows or underflows are observed

[Ingo Molnar]

* John Stultz <john.stultz@xxxxxxxxxx> wrote:

> It was suggested that the underflow/overflow protection
> should probably throw some sort of warning out, rather
> then just silently fixing the issue.

Typo.

> So this patch adds some warnings here. The flag variables
> used are not protected by locks, but since we can't print
> from the reading functions, just being able to say we
> saw an issue in the update interval is useful enough,
> and can be slightly racy without real consequnece.

Typo.

> The big complication is that we're only under a read
> seqlock, so the data could shift under us during
> our calcualtion to see if there was a problem. This

Typo.

> patch avoids this issue by nesting another seqlock
> which allows us to snapshot the just required values
> atomically. So we shouldn't see false positives.
> 
> I also added some basic ratelimiting here, since
> on one build machine w/ skewed TSCs it was fairly
> noisy at bootup.

> +#define WARNINGFREQ (HZ*300) /* 5 minute rate-limiting */

Nit: so in general wereallytrytokeepwordsapart, so I'd suggest a 
name of WARNING_FREQ or so?

>  	cycle_t max_cycles = tk->tkr.clock->max_cycles;
>  	const char *name = tk->tkr.clock->name;
> +	static long last_warning; /* we always hold write on timekeeper lock */

So I'm not sure I ever heard the phrase 'to hold write', this doesn't 
parse for me.

Also, static global variables should really, really not be immersed 
amongst on-stack variables, they are so easy to overlook. Just put 
them in front of the function.

>  
>  	if (offset > max_cycles)
>  		printk_deferred("ERROR: cycle offset (%lld) is larger then"
> @@ -133,28 +145,60 @@ static void timekeeping_check_update(struct timekeeper *tk, cycle_t offset)
>  		printk_deferred("WARNING: cycle offset (%lld) is past"
>  			" the %s 50%% safety margin (%lld)\n",
>  			offset, name, max_cycles>>1);
> +
> +	if (timekeeping_underflow_seen) {
> +		if (jiffies - last_warning > WARNINGFREQ) {
> +			printk_deferred("WARNING: Clocksource underflow observed\n");
> +			last_warning = jiffies;
> +		}
> +		timekeeping_underflow_seen = 0;
> +	}
> +	if (timekeeping_overflow_seen) {
> +		if (jiffies - last_warning > WARNINGFREQ) {
> +			printk_deferred("WARNING: Clocksource overflow observed\n");

I think the warning should be more informative. If a distro turns this 
on and a user sees this value, what will he think? Is the kernel still 
OK? What can he do about it?

> +			last_warning = jiffies;
> +		}
> +		timekeeping_overflow_seen = 0;
> +	}
> +
>  }
>  
>  static inline cycle_t timekeeping_get_delta(struct tk_read_base *tkr)
>  {
> -	cycle_t cycle_now, delta;
> +	cycle_t now, last, mask, max, delta;
> +	unsigned int seq;
>  
> -	/* read clocksource */
> -	cycle_now = tkr->read(tkr->clock);
> +	/*
> +	 * Since we're called holding a seqlock, the data may shift
> +	 * under us while we're doign the calculation. This can cause

Typo...

> +	 * false positives, since we'd note a problem but throw the
> +	 * results away. So nest  another seqlock here to atomically

Spurious space. I know they are cheap, but still.

Thanks,

	Ingo
--
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/