Re: [PATCH v2 4/4] kernel/time/clockevents: compensate for monotonic clock's dynamic frequency

From: Nicolai Stange <nicstange@gmail.com>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: Nicolai Stange <nicstange@gmail.com>,
	Ingo Molnar <mingo@redhat.com>, "H. Peter Anvin" <hpa@zytor.com>,
	x86@kernel.org, John Stultz <john.stultz@linaro.org>,
	Borislav Petkov <bp@suse.de>, Paolo Bonzini <pbonzini@redhat.com>,
	Viresh Kumar <viresh.kumar@linaro.org>,
	Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>,
	"Peter Zijlstra \(Intel\)" <peterz@infradead.org>,
	"Christopher S. Hall" <christopher.s.hall@intel.com>,
	Adrian Hunter <adrian.hunter@intel.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 4/4] kernel/time/clockevents: compensate for monotonic clock's dynamic frequency
Date: Wed, 13 Jul 2016 15:08:33 +0200	[thread overview]
Message-ID: <871t2xy9bi.fsf@gmail.com> (raw)
In-Reply-To: <alpine.DEB.2.11.1607121651580.4083@nanos> (Thomas Gleixner's message of "Tue, 12 Jul 2016 17:04:17 +0200 (CEST)")

Thomas Gleixner <tglx@linutronix.de> writes:

> On Tue, 12 Jul 2016, Nicolai Stange wrote:
>> Another issue is that ->min_delta_ns and ->max_delta_ns are measured in
>> raw clock time while the delta in clockevents_program_event() would now
>> be interpreted as being in monotonic clock time:
>>   clc = ((unsigned long long) delta * dev->mult_mono) >> dev->shift;
>
> Does that really matter much?
>
>> Ideally, I'd like to get rid of ->min_delta_ns and ->max_delta_ns
>> alltogether and consistently use the ->min_delta_ticks and
>> ->max_delta_ticks instead. AFAICS, ->min_delta_ns is really needed only
>> for setting dev->next_event in clockevents_program_min_delta().
>> dev->next_event is read only from __clockevents_update_freq() for
>> reprogramming purposes and thus, assuming 0 for ->delta_min_ns in
>> clockevents_program_min_delta() would probably work: a reprogramming
>> would invoke clockevents_program_min_delta() once again.
>
> I completely fail to parse the above paragraph. 
>  
>> The downside of this approach is that a quick grep reveals 40 clockevent
>> device drivers whose initialization code would need to get touched in
>> order to convert them from min_delta_ns/max_delta_ns to
>> min_delta_ticks/max_delta_ticks.
>> 
>> So, the question is whether I should do all of this or whether the
>> doubled timer interrupts aren't annoying enough to justify such a big
>> change?
>
> Can you provide an initial patch which does the adjustment w/o all the related
> churn so we can see how intrusive that gets?

Please see the RFC tagged series at
  http://lkml.kernel.org/g/20160713130017.8202-1-nicstange@gmail.com
I tried to answer/address your above questions in the cover letter.

Note that I split the x86 TSC related patches off:
  http://lkml.kernel.org/g/20160713130344.8319-1-nicstange@gmail.com

Thanks,

Nicolai