Re: [Xen-devel] [PATCH] x86: detect CMOS aliasing on ports other then 0x70/0x71

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 29.09.15 at 15:25, <andrew.cooper3@xxxxxxxxxx> wrote:
> On 25/09/15 12:00, Jan Beulich wrote:
>>>>> On 25.09.15 at 11:55, <andrew.cooper3@xxxxxxxxxx> wrote:
>>> It is safe for the SMM handler to use CMOS if it returns the index 
>>> register back to how it found it.  Furthermore, I am willing to bet that 
>>> there real SMM handlers out there which do do this.
>> So what options do you see then here? Don't do anything (drop the
>> patch) seems the only one I can see not getting in conflict with SMI.
>> Yet using the patch as is would not make anything less safe, it would
>> just have the potential of not always making things as safe as they
>> could be.
> 
> Given some orthogonal interaction with cmos-rtc-probe recently, I am
> fairly sure that we can remove all real RTC interaction from Xen.
> 
> Dom0 is required to provide fine-grained time via XENPF_settime{32,64}
> which gives an absolute time to seed the other domain wallclocks with. 

Is it? If so, I suppose we should enforce this in some way, e.g. by
failing domain creation until done. Right now I guess we expect that
to happen, but don't really require it.

> The RTC was assumed always to be in UTC, which allows full date and time
> information to be derived.
> 
> It would certainly simplify things to leave all of the RTC in dom0's hand.

Yes, it would.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel