From: Denis Plotnikov <dplotnikov@virtuozzo.com>
To: Paolo Bonzini <pbonzini@redhat.com>,
rkrcmar@redhat.com, kvm@vger.kernel.org
Cc: rkagan@virtuozzo.com, den@virtuozzo.com, svt-core@lists.sw.ru
Subject: Re: [PATCH v3 2/6] KVM: x86: switch to masterclock update using timekeeper functionality
Date: Tue, 1 Aug 2017 15:11:10 +0300 [thread overview]
Message-ID: <616b4427-49a9-7815-56c3-9171248d9673@virtuozzo.com> (raw)
In-Reply-To: <82c81e37-9c56-26b9-d8c5-cf87eeb1d470@redhat.com>
On 01.08.2017 13:03, Paolo Bonzini wrote:
> On 01/08/2017 11:30, Denis Plotnikov wrote:
>>> - implementing kvm_clock_read_with_cycles (can be merged with patch 6)
>>
>> Having a stable clocksource for kvmklock means making kvmclock stable.
>> The patch enables this functionality that's why I'd prefer to keep patch
>> 6 separate
>
> Ok, though it depends on how you deal with cs_stable.
>
>>> - using ktime_get_snapshot in KVM (can be merged with patch 4?)
>>
>> agree, but want to keep 4 separate. Just to make the changes done
>> logically consecutive in git tree.
>
> Fine by me.
>
>>> Maybe what we want is some kind of "bool cycles_valid", and then
>>> read_clock_and_systime can return it:
>>>
>>>
>>> if (clock->read_clock_and_systime) {
>>> systime_snapshot->cycles_valid =
>>> clock->read_clock_and_systime(
>>> &now, &systime_snapshot->cycles);
>>> } else {
>>> now = tk_clock_read(&tk->tkr_mono);
>>> systime_snapshot->cycles_valid = true;
>>> systime_snapshot->cycles = now;
>>> }
>>>
>>> ? (This is honestly just a suggestion, which may be wrong depedning
>>> on the answer to the two questions above).
>>
>> cs_stable means "there is no unexpected time jumps".
>
> But even for kvmclock there are no unexpected time jumps, the global
> accumulator in pvclock_clocksource_read ensures that. And the concept
> is different from CLOCK_SOURCE_UNSTABLE which will basically blacklist
> the clocksource for hrtimers.
>
> It seems a different concept to me, somewhat specific to
> ktime_get_snapshot. More precisely, the question is "is there a 1:1
> mapping from cycles to nanoseconds?"---but if there is no such mapping
> cycles is useless, hence my proposal of "cycles_valid".
>
> Thanks,
>
> Paolo
In fact, this "cycles_valid" is going to be used for deciding whether to
use KVM masterclock or not. And if it's not we still want to know
cycles_stamp value to use it in KVM.
So the cycles is valid, but clocksource is not reliable (why? let decide
to a clocksource, by default assume they are all not stable), thus we
must calculate time values for a guest each time its needed.
So, my proposal is to name the variable sightly differently: cs_reliable
and go like:
if (clock->read_clock_with_stamp) {
systime_snapshot->cs_reliable =
clock->read_clock_with_stamp(
&now, &systime_snapshot->cycles);
} else {
now = tk_clock_read(&tk->tkr_mono);
systime_snapshot->cs_reliable = false;
systime_snapshot->cycles = now;
}
What do you think?
Thanks!
Denis
>
>> I don't mind merging this "check stability" functionality with
>> read_clock_and_systime. Actually, I thought about having it there but
>> eventually split it to make the code more explicit
>> (detailed and understandable).
>> I'd like to use your approach but keep the variable name the same.
>>
>> Thanks for reviewing!
>>
>> Denis
>>>
>>> Paolo
>>>
>>>> systime_snapshot->cs_was_changed_seq = tk->cs_was_changed_seq;
>>>> systime_snapshot->clock_was_set_seq = tk->clock_was_set_seq;
>>>> base_real = ktime_add(tk->tkr_mono.base,
>>>> tk_core.timekeeper.offs_real);
>>>> base_raw = tk->tkr_raw.base;
>>>> + base_boot = ktime_add(tk->tkr_mono.base,
>>>> + tk_core.timekeeper.offs_boot);
>>>> nsec_real = timekeeping_cycles_to_ns(&tk->tkr_mono, now);
>>>> nsec_raw = timekeeping_cycles_to_ns(&tk->tkr_raw, now);
>>>> } while (read_seqcount_retry(&tk_core.seq, seq));
>>>> - systime_snapshot->cycles = now;
>>>> systime_snapshot->real = ktime_add_ns(base_real, nsec_real);
>>>> systime_snapshot->raw = ktime_add_ns(base_raw, nsec_raw);
>>>> + systime_snapshot->boot = ktime_add_ns(base_boot, nsec_real);
>>>> }
>>>> EXPORT_SYMBOL_GPL(ktime_get_snapshot);
>>>>
>>>
>>
>
--
Best,
Denis
next prev parent reply other threads:[~2017-08-01 12:11 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-07-29 12:35 [PATCH v3 0/6] make L2's kvm-clock stable, get rid of pvclock_gtod Denis Plotnikov
2017-07-29 12:35 ` [PATCH v3 1/6] pvclock: add parameter to store cycles stamp to pvclock reading function Denis Plotnikov
2017-07-31 14:08 ` Paolo Bonzini
2017-07-29 12:35 ` [PATCH v3 2/6] KVM: x86: switch to masterclock update using timekeeper functionality Denis Plotnikov
2017-07-31 14:20 ` Paolo Bonzini
2017-08-01 9:30 ` Denis Plotnikov
2017-08-01 10:03 ` Paolo Bonzini
2017-08-01 10:16 ` Paolo Bonzini
2017-08-01 12:11 ` Denis Plotnikov [this message]
2017-08-01 12:28 ` Denis Plotnikov
2017-08-01 12:41 ` Paolo Bonzini
2017-08-01 12:46 ` Denis Plotnikov
2017-08-01 17:47 ` Radim Krčmář
2017-07-29 12:35 ` [PATCH v3 3/6] timekeeper: add clocksource change notifier Denis Plotnikov
2017-07-29 12:35 ` [PATCH v3 4/6] KVM: x86: remove not used pvclock_gtod_copy Denis Plotnikov
2017-07-29 12:35 ` [PATCH v3 5/6] pvclock: add clocksource change notification on changing of tsc stable bit Denis Plotnikov
2017-07-31 14:21 ` Paolo Bonzini
2017-07-29 12:35 ` [PATCH v3 6/6] kvmclock: add the clocksource stability querying function Denis Plotnikov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=616b4427-49a9-7815-56c3-9171248d9673@virtuozzo.com \
--to=dplotnikov@virtuozzo.com \
--cc=den@virtuozzo.com \
--cc=kvm@vger.kernel.org \
--cc=pbonzini@redhat.com \
--cc=rkagan@virtuozzo.com \
--cc=rkrcmar@redhat.com \
--cc=svt-core@lists.sw.ru \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).