Xen devel - Apr 2009 - [PATCH] xend: fix rtc_timeoffset when localtime=1

The rtc_timeoffset parameter in VM config is ignored when localtime=1.
Also it is not preserved across reboot.

NOTE:
This patch changes the meaning of the xenstore /vm/<uuid>/rtc/timeoffset
from utc offset to local offset. (it''s OK for the ioemu)

Signed-off-by: Kouya Shimura <kouya@jp.fujitsu.com>



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Kouya Shimura writes ("[Xen-devel] [PATCH] xend: fix rtc_timeoffset when
localtime=1"):
> The rtc_timeoffset parameter in VM config is ignored when localtime=1.
> Also it is not preserved across reboot.
> 
> NOTE:
> This patch changes the meaning of the xenstore
/vm/<uuid>/rtc/timeoffset
> from utc offset to local offset. (it''s OK for the ioemu)
Are you sure that this is an appropriate change ?  I''m not sure I
follow exactly and have only skimread your patch.  Can you explain why
it is right to change the meaning of a xenstore value like this ?

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Ian Jackson writes:
> Kouya Shimura writes ("[Xen-devel] [PATCH] xend: fix rtc_timeoffset
when localtime=1"):
> > The rtc_timeoffset parameter in VM config is ignored when localtime=1.
> > Also it is not preserved across reboot.
> > 
> > NOTE:
> > This patch changes the meaning of the xenstore
/vm/<uuid>/rtc/timeoffset
> > from utc offset to local offset. (it''s OK for the ioemu)
> 
> Are you sure that this is an appropriate change ?  I''m not sure I
> follow exactly and have only skimread your patch.  Can you explain why
> it is right to change the meaning of a xenstore value like this ?
Hi Ian,

I''m quite sure. From a ioemu point of view, the value is no concern
of the ioemu. The ioemu receives the ioreq, accumulates the value,
and then just passes it to the xenstore. 
You can apply the attached patch. :-)

When a guest changes the real time clock, the hypervisor sends the
difference between before and after via ioreq to the ioemu.
see rtc_set_time@xen/arch/x86/hvm/rtc.c.
Thus, /rtc/timeoffset keeps correct offset from any initial value.

And xend is the only one to refer /rtc/timeoffset, and xend knows
that it is a offset based on local time. So, my scheme works well.

In addition, PV domain doesn''t use the xenstore /rtc/timeoffset.

Thanks,
Kouya


Signed-off-by: Kouya Shimura <kouya@jp.fujitsu.com>



_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Thu, Apr 02, 2009 at 02:24:02PM +0900, Kouya Shimura wrote:
> In addition, PV domain doesn''t use the xenstore /rtc/timeoffset.
Are you positive about that? Solaris PV expects RTC in localtime.

The patch looked strange to me too but I haven''t had time to look at it
closely either.

regards
john

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

John Levon writes:
> On Thu, Apr 02, 2009 at 02:24:02PM +0900, Kouya Shimura wrote:
> 
> > In addition, PV domain doesn''t use the xenstore
/rtc/timeoffset.
> 
> Are you positive about that? Solaris PV expects RTC in localtime.
> 
> The patch looked strange to me too but I haven''t had time to look
at it
> closely either.
If solaris PV watches the xenstore /rtc/timeoffset, that''s a problem.
But I dont''t think so.

/rtc/timeoffset is an interface between HVM->hypervisor->ioemu->xend
(especially ioemu->xend), and xen hypervisor doesn''t prepare an
interface to change the xenstore /rtc/timeoffset for PV domain.

At least, xen PV domain even has no /rtc/timeoffset entry in xenstore.

Thanks,
Kouya

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Fri, Apr 03, 2009 at 09:24:41AM +0900, Kouya Shimura wrote:
> > > In addition, PV domain doesn''t use the xenstore
/rtc/timeoffset.
> > 
> > Are you positive about that? Solaris PV expects RTC in localtime.
> > 
> > The patch looked strange to me too but I haven''t had time to
look at it
> > closely either.
> 
> If solaris PV watches the xenstore /rtc/timeoffset, that''s a
problem.
> But I dont''t think so.
Right, it doesn''t, I''d missed you were talking about the
xenstore entry.

I think I understand the patch now.

regards
john

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel