Re: [Xen-devel] [PATCH v11] tolerate jitter in cpu_khz calculation to avoid TSC emulation

[Olaf Hering <olaf@xxxxxxxxx>]

Am Thu, 13 Dec 2018 03:41:44 -0700
schrieb "Jan Beulich" <JBeulich@xxxxxxxx>:

> And further argumentation that everyone is using NTP anyway
> doesn't make it any better, when it's no-where written down that
> Xen is unusable with NTP running in all guests (I'm exaggerating
> here just to get the point over). Don't forget that e.g. with
> XenoLinux'es independent-wallclock setting defaulting to false, we've
> been suggesting that people _don't_ need to use NTP inside their
> (admittedly PV) guests. IOW - your change may not break people
> not using NTP. Hence I don't think the mode you introduce can be
> a default-on one, which in turn means a per-domain or at least
> global enable control is needed (as over previous iterations we
> seem to have been agreeing).

Regarding the possible unexpected drift if NTP is not used, and the domU
uses TSC as clocksource anyway: If one host is on one edge of the assumed
cpu_khz value and the other host is on the opposite edge, and the range
will be 200 as it is in my patch, the daily drift on a 2.3GHz host would
be 7.4seconds per day. This number is based on the fact that 200kz happen
within a time span of 86us, which I think is correct.

The question is, how much drift can be tolerated even without my patch.
Looking through 5 different hosts I use, the range is between -11 and +22,
and it happens to be +56 on my Laptop. So even that host with a calculated
drift of +22 would be off by 2 seconds per day if ntpd would not run.


Olaf

Attachment: pgp70KWzHMMQV.pgp
Description: Digitale Signatur von OpenPGP

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel