On 13.03.24 12:18, Alexandre Belloni wrote:
> On 13/03/2024 10:45:54+0100, Peter Hilber wrote:
>>> Exposing UTC as the only clock reference is bad enough; when leap
>>> seconds happen there's a whole second during which you don't *know*
>>> which second it is. It seems odd to me, for a precision clock to be
>>> deliberately ambiguous about what the time is!
>>
>> Just to be clear, the device can perfectly expose only a TAI reference
>> clock (or both UTC and TAI), the spec is just completely open about this,
>> as it tries to work for diverse use cases.
>>
>>>
>>> But if the virtio-rtc clock is defined as UTC and then expose something
>>> *different* in it, that's even worse. You potentially end up providing
>>> inaccurate time for a whole *day* leading up to the leap second.
>>>
>>> I think you're right that leap second smearing should be addressed. At
>>> the very least, by making it clear that the virtio-rtc clock which
>>> advertises UTC shall be used *only* for UTC, never UTC-SLS or any other
>>> yet-to-be-defined variant.
>>>
>>
>> Agreed.
>>
>>> Please make it explicit that any hypervisor which wants to advertise a
>>> smeared clock shall define a new type which specifies the precise
>>> smearing algorithm and cannot be conflated with the one you're defining
>>> here.
>>>
>>
>> I will add a requirement that the UTC clock can never have smeared/smoothed
>> leap seconds.
>>
>> I think that not every vendor would bother to first add a definition of a
>> smearing algorithm. Also, I think in some cases knowing the precise
>> smearing algorithm might not be important (when having the same time as the
>> hypervisor is enough and accuracy w.r.t. actual time is less important).
>>
>> So maybe I should add a VIRTIO_RTC_CLOCK_UTC_SMEARED clock type, which for
>> now could catch every UTC-like clock which smears/smoothes leap seconds,
>> where the vendor cannot be bothered to add the smearing algorithm to spec
>> and implementations.
>>
>
> I still don't know anything about virtio but under Linux, an RTC is
> always UTC (or localtime when dual booting but let's not care) and never
> accounts for leap seconds. Having an RTC and RTC driver behaving
> differently would be super inconvenient. Why don't you leave this to
> userspace?
>
> I guess I'm still questioning whether this is the correct interface to
> expose the host system time instead of an actual RTC.
>
virtio_rtc only registers RTC class devices for virtio_rtc clock type UTC,
so adding an UTC_SMEARED clock type would avoid leap seconds for the RTC
class.
But I understand that there are more concerns and I will re-consider. From
my POV CLOCK_{REALTIME,MONOTONIC}_ALARM support is very important however.
So the only alternative to me seems to be adding an alternative alarmtimer
backend (and time synchronization through user space).
Thanks for the comment,
Peter
