Re: [RFC] New timeofday proposal (v.A1)

[Ulrich Windl]

On 8 Dec 2004 at 11:58, john stultz wrote:

[...]
> behavior of adjtimex().  We need to be able to implement the following
> adjustments within a single tick:
> 
> 1. Adjust the frequency by 500ppm for 10usecs 

What do you mean by "for 10usecs"?

> 2. After that adjust the frequency by 30ppm for the rest of the tick.

I'm not sure what you are taling about: plain old adjtime() or the NTP kernel 
interface?

[...]
> I may have asked this before, but w/ 32 bit mult and shifts, how
> granular can these adjustments be?

Independent of any bits, the precision should be up to 1ns for reading and setting 
the clock, and as a consequence you might provide internal fractional nanoseconds 
(if you want to have a truly stable nanosecond clock model). If we get this right, 
there will be peace in this area until the wires in a PC are significantly shorter 
than 30cm (I think this is how far the light goes in 1ns). ;-)

> 
> Also additional complications arise when we have multiple things (like
> cpufreq) playing with the timesource frequency values as well. 

I see a bug difference between precise time keeping and "playing" with 
timesources.

Regards,
Ulrich

-
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/