Re: misc/57300: Time system is not accurate on NetBSD VM

[mlelstv%serpens.de@localhost (Michael van Elst)]

The following reply was made to PR misc/57300; it has been noted by GNATS.

From: mlelstv%serpens.de@localhost (Michael van Elst)
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: misc/57300: Time system is not accurate on NetBSD VM
Date: Sun, 26 Mar 2023 10:44:17 -0000 (UTC)

 nan%chinadtrace.org@localhost writes:
 
 >-bash-5.1$ date; sleep 10; date
 >Sun Mar 26 18:59:44 +08 2023
 >Sun Mar 26 19:00:13 +08 2023
 
 >Actually the time just passes 10 seconds, but date command shows it has passes nearly half a minute. Not sure whether this could be accounted as NetBSD bug, or VirtualBox bug, thanks!
 
 
 "sleep 10" waits 1000 clock interrupts. If that lasts 10 seconds, the
 emulation of clock interrupts is correct. The exact delay can vary
 with host speed and load.
 
 "date" looks at the current timecounter. If that advances 30 seconds
 in 10 seconds, the emulation of the timecounter is bad. Fortunately we
 have multiple, e.g.:
 
 # sysctl kern.timecounter
 kern.timecounter.choice = TSC(q=3000, f=3292677360 Hz) clockinterrupt(q=0, f=100 Hz) ichlpcib0(q=1000, f=3579545 Hz) hpet0(q=2000, f=14318180 Hz) ACPI-Fast(q=1000, f=3579545 Hz) lapic(q=-100, f=99859155 Hz) i8254(q=100, f=1193182 Hz) dummy(q=-1000000, f=1000000 Hz)
 kern.timecounter.hardware = TSC
 kern.timecounter.timestepwarnings = 0
 
 By setting kern.timecounter.hardware to one of the choices you may get
 a timecounter that works better.
 
 While the choice "clockinterrupt" avoids discrepancies between "sleep"
 and "date", the result might not be accurate. For many cases a different
 choice (if there is any) is better.