Re: [PATCH 2/5] x86/time: implement tsc as clocksource

[Joao Martins <joao.m.martins@oracle.com>]

On 03/22/2016 04:02 PM, Jan Beulich wrote:
>>>> On 22.03.16 at 16:51, <joao.m.martins@oracle.com> wrote:
> 
>>
>> On 03/22/2016 12:46 PM, Jan Beulich wrote:
>>>>>> On 22.03.16 at 13:41, <joao.m.martins@oracle.com> wrote:
>>>
>>>>
>>>> On 03/18/2016 08:21 PM, Andrew Cooper wrote:
>>>>> On 17/03/16 16:12, Joao Martins wrote:
>>>>>> Introduce support for using TSC as platform time which is the highest
>>>>>> resolution time and most performant to get (~20 nsecs).  Though there
>>>>>> are also several problems associated with its usage, and there isn't a
>>>>>> complete (and architecturally defined) guarantee that all machines
>>>>>> will provide reliable and monotonic TSC across all CPUs, on different
>>>>>> sockets and different P/C states.  I believe Intel to be the only that
>>>>>> can guarantee that. For this reason it's set with less priority when
>>>>>> compared to HPET unless adminstrator changes "clocksource" boot option
>>>>>> to "tsc". Upon initializing it, we also check for time warps and
>>>>>> appropriate CPU features i.e.  TSC_RELIABLE, CONSTANT_TSC and
>>>>>> NONSTOP_TSC. And in case none of these conditions are met, we fail in
>>>>>> order to fallback to the next available clocksource.
>>>>>>
>>>>>> It is also worth noting that with clocksource=tsc there isn't any
>>>>>> need to synchronise with another clocksource, and I could verify that
>>>>>> great portion the time skew was eliminated and seeing much less time
>>>>>> warps happening. With HPET I observed ~500 warps in the period
>>>>>> of 1h of around 27 us, and with TSC down to 50 warps in the same
>>>>>> period having each warp < 100 ns. The warps still exist though but
>>>>>> are only related to cross CPU calibration (being how much it takes to
>>>>>> rendezvous with master), in which a later patch in this series aims to
>>>>>> solve.
>>>>>>
>>>>>> Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
>>>>>> ---
>>>>>> Cc: Keir Fraser <keir@xen.org>
>>>>>> Cc: Jan Beulich <jbeulich@suse.com>
>>>>>> Cc: Andrew Cooper <andrew.cooper3@citrix.com>
>>>>>
>>>>> Some style corrections, but no functional problems.
>>>>>
>>>>> Reviewed-by Andrew Cooper <andrew.cooper3@citrix.com>
>>>>>
>>>> I found out one issue in the tsc clocksource initialization path with respect to
>>>> the reliability check. This check is running when initializing platform timer,
>>>> but not all CPUS are up at that point (init_xen_time()) which means that the
>>>> check will always succeed. So for clocksource=tsc I need to defer initialization
>>>> to a later point (on verify_tsc_reliability()) and if successful switch to that
>>>> clocksource. Unless you disagree, v2 would have this and just requires one
>>>> additional preparatory change prior to this patch.
>>>
>>> Hmm, that's suspicious when thinking about CPU hotplug: What
>>> do you intend to do when a CPU comes online later, failing the
>>> check?
>> Good point, but I am not sure whether that would happen. The initcall
>> verify_tsc_reliability seems to be called only for the boot processor. Or are
>> you saying that it's case that initcalls are called too when hotplugging cpus
>> later on? If that's the case then my suggestion wouldn't work as you point out -
>> or rather without having runtime switching support as you point out below.
> 
> Looks like I didn't express myself clearly enough: "failing the check"
> wasn't meant to imply the failure would actually occur, but rather
> that failure would occur if the check was run on that CPU. IOW the
> case of a CPU getting hotplugged which doesn't satisfy the needs
> for selecting TSC as the clock source.
Ah, I see. I believe this wouldn't be an issue with the current rendezvous
mechanism (std_rendezvous), as the delta is computed between local_tsc_stamp
taken in that (hotplugged) CPU in the calibration and the rdtsc() in the guest
same CPU, even though having CPU0 TSC (master) as system_time.

However it can be a problem with PVCLOCK_TSC_STABLE_BIT as the hotplugged CPU
could have its TSC drifted, and since setting this flag relies on
synchronization of TSCs we would need to clear the flag enterily.

>>> So far we don't do runtime switching of the clock source,
>>> and I'm not sure that's a good idea to do when there are running
>>> guests.
>> Totally agree, but I would be proposing to be at initialization phase where
>> there wouldn't be guests running. We would start with HPET, then only switch 
>> to
>> TSC if that check has passed (and would happen once).
>>
>> Simpler alternative would be to call init_xen_time after all CPUs are 
>> brought up
>> (and would also keep this patch as is), but I am not sure about the
>> repercussions of that.
> 
> I don't see how either would help with the hotplug case.
This was in response to what I thought you meant with your earlier question
(which I misunderstood). But my question is still valid I believe. The reason
for choosing between my suggested approaches is that tsc_check_reliability()
requires all CPUs up so that the check is correctly performed.

Joao

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