Re: [patch 6/13] signal/timer/event fds v9 - timerfd core ...

[Thomas Gleixner]

On Mon, 2007-04-02 at 10:30 -0700, Davide Libenzi wrote:

> > There is no inaccuracy when you rearm the timer on read: hrtimer_forward
> > takes care, that the period is accurate. It does not start the timer out
> > of the periodic order, i.e. on a different time frame.
> > 
> > Where is the win of keeping the timer running, when nobody cares about
> > the expiry at all ? It just generates interrupts and events for nothing.
> 
> Then you'd lose the ability to know if you lost one or more (yes, you 
> could figure it out by reading the time and with a few calculations). I 
> think that the capping (to a sane value) idea solves the DoS issue and at 
> the same time have the ability to report you missed ticks. What are your 
> strong points against that solution?

Err, the read function 

	ticks = hrtimer_forward(&ctx->tmr, ktime_get(),
                                ctx->tintv);

does give you the number of (lost) ticks.

tmr->expires holds the absolute expiry time of the last event.
hrtimer_forward() adds N intervals to tmr->expires, so that the new
tmr->expires value is greater than now (ktime_get()). It returns N.

So the number of lost ticks is N - 1. No time reading and no magic
math :)

	tglx


-
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/