[ntp:hackers] u-blox reference clock driver

[MLewis]

Repeat. Didn't go out in text format.

On 20/08/2018 10:22 AM, MLewis via devel wrote:
> Udo,
>
> On 20/08/2018 9:43 AM, Udo van den Heuvel wrote:
>> Michael,
>> What else could we do with the timemark functionality?
>> See what is suggested at 
>> https://lists.ntpsec.org/pipermail/devel/2018-August/006447.html ?
> With kernel level timestamps back in user space, paired with the GPS 
> Timemark message, you can do whatever you want with them.
>
>>> I'm held up deciding on the method of sending messages for each 
>>> Trigger Timestamp from the kernel module back to user space.
>>> Looks like it should be a socket.
>>
>> This is because the ublox doesn't know the time of origin I guess?
> Correct. You need both sides.
> With PPS, you get one offset from system time, but with incoming 
> latency including interrupt lag.
> Using Timemarks give you FOUR kernal level latency ports going 
> high/low with kernel timestamps.
> You don't need the higher latency PPS for calculating offsets and 
> disciplining. See the thesis results.
>
>> Maybe make it an elaborate PPS device that can also be read?
>> Udo
>>
> Define "elaborate PPS device".
> Once this LKM is working, you can make a copy and play with it. 
> Perhaps with using the greater precision of CPU Time - if you can find 
> a way that will keep the LKM in the same core for the duration of your 
> measurements.
>
>
> Michael
>
>
>
>
> _______________________________________________
> devel mailing list
> devel at ntpsec.org
> http://lists.ntpsec.org/mailman/listinfo/devel