Re: Time issues back?

[Greg Troxel <gdt%lexort.com@localhost>]

John Klos <john%ziaspace.com@localhost> writes:

> We used to have a time drift issue way back in NetBSD 6, I think. On
> my Mac mini I'm seeing it again (still?):
>
> 30 Jul 04:02:31 ntpdate[12515]: step time server
> 2001:470:a068:34:250:e4ff:fee0:c035 offset 0.801721 sec
>
> 30 Jul 04:23:38 ntpdate[883]: step time server
> 2001:470:a068:34:250:e4ff:fee0:c035 offset 2.515123 sec
>
> Is this expected? I thought this was fixed at some point. This is on
> netbsd-8 compiled from today's sources.

I have seen a timekeeping issue on a normal PC (Dell, 2014, i7),
starting around 8.1.

Looking at dmesg, it seems to have changed from ACPI-Safe to ACPI-Fast
and misestimated the TSC:

-timecounter: Timecounter "ACPI-Safe" frequency 3579545 Hz quality 900
+timecounter: Timecounter "ACPI-Fast" frequency 3579545 Hz quality 1000

-timecounter: Timecounter "TSC" frequency 2800071320 Hz quality 3000
+timecounter: Timecounter "TSC" frequency 3920113160 Hz quality 3000


Sometimes it gets the right value (2.8E9) on boot, sometimes not.
When wrong, if I set time to hpet0, it is then ok (but a slightly
different drift).

A few others have seen this, but it does not seem super common.

I suspect a bug someplace, either introduced or made apparent around
8.1.

Check your dmesg before/after and please let me know if you are seeing
TSC misestimation.