kern/155238: 20 minute time jumps in VirtualBox

[Martin Birgmeier]

>Number:         155238
>Category:       kern
>Synopsis:       20 minute time jumps in VirtualBox
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Thu Mar 03 20:30:08 UTC 2011
>Closed-Date:
>Last-Modified:
>Originator:     Martin Birgmeier
>Release:        8.2, HEAD
>Organization:
MBi at home
>Environment:
FreeBSD v902.xyzzy 8.2-RELEASE FreeBSD 8.2-RELEASE #1: Sun Feb 20 19:23:23 CET 2011     root at v902.xyzzy:/.../hal/z/OBJ/FreeBSD/i386/RELENG_8_2_0_RELEASE/src/sys/XYZZY_SMP  i386

FreeBSD v903.xyzzy 9.0-CURRENT FreeBSD 9.0-CURRENT #1: Sun Jan 30 13:12:33 CET 2011     root at v903.xyzzy:/usr/obj/.../hal/z/SRC/FreeBSD/HEAD/src/sys/XYZZY_SMP_NA  amd64
>Description:
I am running several versions of FreeBSD under VirtualBox, on a 4-core AMD box running FreeBSD 8.2 amd64.

Since recently, and I believe since installing 8.2, every few minutes the time in the vbox clients jumps by exactly 20 minutes like this (hal is the host, the v90x are the vbox clients):

[0]# rdate -p hal ; date
Thu Mar  3 19:36:27 2011
Thu Mar  3 19:36:27 CET 2011
[0]# rdate -p hal ; date
Thu Mar  3 19:44:06 2011
Thu Mar  3 19:44:06 CET 2011
[0]# rdate -p hal ; date
Thu Mar  3 20:09:14 2011
Thu Mar  3 20:29:14 CET 2011 <=== jump by 20 minutes
[0]# date ; rdate hal   
Thu Mar  3 20:29:16 CET 2011 <=== manual reset using rdate
Thu Mar  3 20:09:16 2011
[0]# date ; rdate hal
Thu Mar  3 20:43:58 CET 2011
Thu Mar  3 20:43:59 2011

or this:

[0]# rdate -p hal ; date
Thu Mar  3 20:57:44 2011
Thu Mar  3 20:57:43 CET 2011
[0]# rdate -p hal ; date
Thu Mar  3 20:58:31 2011
Thu Mar  3 21:18:30 CET 2011 <=== jump
[0]# date ; rdate hal
Thu Mar  3 21:18:34 CET 2011 <=== manual reset
Thu Mar  3 20:58:35 2011
[0]# rdate -p hal ; date
Thu Mar  3 21:00:49 2011
Thu Mar  3 21:00:49 CET 2011
[0]# rdate -p hal ; date
Thu Mar  3 21:05:31 2011
Thu Mar  3 21:25:31 CET 2011 <=== jump
[0]# date ; rdate hal   
Thu Mar  3 21:25:37 CET 2011 <=== manual reset
Thu Mar  3 21:05:37 2011
[0]# date ; rdate hal
Thu Mar  3 21:08:12 CET 2011
Thu Mar  3 21:08:13 2011

This does not seem to happen on the 7.4 client which is also running at the same time as the two others (HEAD, 8.2).

Previously (and I believe even now), if the host was very busy, the clients would simply lag behind the host, and I could use rdate to set them forward again. But these exact 20 minute jumps are a completely new experience and rather annoying.

Btw., if I remember correctly, they also accumulate: If I do not manually reset the time, after a longer while I will have the clock 40 minutes into the future, etc.
>How-To-Repeat:
Run HEAD or 8.2 in VirtualBox, observe the time.
>Fix:


>Release-Note:
>Audit-Trail:
>Unformatted: