Re: [PATCH net-next] ptp: add vclock timestamp conversion IOCTL

[Sebastien Laveze]

On Thu, 2021-10-07 at 13:19 -0700, Richard Cochran wrote:
> On Thu, Oct 07, 2021 at 03:31:43PM +0200, Sebastien Laveze wrote:
> > For example:
> > Let's take a worst case, the PHC is adjusted every 10 ms, vclock every
> > 1 sec with 1 ppm error.
> 
> You can't justify adjusting the HW clock and the vclocks
> asynchronously with a single, isolated example.

This was a single worst case example, many asynchronous PHC adjustments
impacting vclocks.

> If two independent processes are both adjusting a clock concurrently,
> asynchronously, and at different control rates, the result will be
> chaos.

This is especially what we want to prove feasible and we think it's
posssible with the following conditions:
-limited frequency adjustments
-offset adjustment in software

> It does not matter that it *might* work for some random setup of
> yours.
> 
> The kernel has to function correctly in general, not just in some
> special circumstances.

Of course, so what tests and measurements can we bring on the table to
convince you that it doesn't lead to chaos ?

Thanks,
Sebastien