Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230630171052.985577-1-peter.hilber@opensynergy.com> <20230630171052.985577-4-peter.hilber@opensynergy.com>
In-Reply-To: <20230630171052.985577-4-peter.hilber@opensynergy.com>
From:   John Stultz <jstultz@google.com>
Date:   Fri, 7 Jul 2023 16:31:52 -0700
Message-ID: <CANDhNCoi4OrtXdqs4UErhM0ofecdg94-Yew2NETxhpKkDojt4A@mail.gmail.com>
Subject: Re: [RFC PATCH 3/7] timekeeping: Fix cross-timestamp interpolation
 for non-x86
To:     Peter Hilber <peter.hilber@opensynergy.com>
Cc:     linux-kernel@vger.kernel.org, Thomas Gleixner <tglx@linutronix.de>,
        Stephen Boyd <sboyd@kernel.org>,
        "Christopher S. Hall" <christopher.s.hall@intel.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, Jun 30, 2023 at 10:12=E2=80=AFAM Peter Hilber
<peter.hilber@opensynergy.com> wrote:
>
> So far, get_device_system_crosststamp() unconditionally passes
> system_counterval.cycles to timekeeping_cycles_to_ns(). But when
> interpolating system time (do_interp =3D=3D true), system_counterval.cycl=
es is
> before tkr_mono.cycle_last, contrary to the timekeeping_cycles_to_ns()
> expectations.
>
> On x86, CONFIG_CLOCKSOURCE_VALIDATE_LAST_CYCLE will mitigate on
> interpolating, setting delta to 0. With delta =3D=3D 0, xtstamp->sys_mono=
raw
> and xtstamp->sys_realtime are then set to the last update time, as
> implicitly expected by adjust_historical_crosststamp(). On other
> architectures, the resulting nonsense xtstamp->sys_monoraw and
> xtstamp->sys_realtime corrupt the xtstamp (ts) adjustment in
> adjust_historical_crosststamp().
>
> Fix this by always setting the delta to 0 when interpolating.
>
> Fixes: 2c756feb18d9 ("time: Add history to cross timestamp interface supp=
orting slower devices")
> Signed-off-by: Peter Hilber <peter.hilber@opensynergy.com>
> ---
>  kernel/time/timekeeping.c | 13 +++++++++----
>  1 file changed, 9 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
> index 7e86d5cd784d..7ccc2377c319 100644
> --- a/kernel/time/timekeeping.c
> +++ b/kernel/time/timekeeping.c
> @@ -1259,10 +1259,15 @@ int get_device_system_crosststamp(int (*get_time_=
fn)
>                                       tk_core.timekeeper.offs_real);
>                 base_raw =3D tk->tkr_raw.base;
>
> -               nsec_real =3D timekeeping_cycles_to_ns(&tk->tkr_mono,
> -                                                    system_counterval.cy=
cles);
> -               nsec_raw =3D timekeeping_cycles_to_ns(&tk->tkr_raw,
> -                                                   system_counterval.cyc=
les);
> +               if (do_interp) {
> +                       nsec_real =3D timekeeping_delta_to_ns(&tk->tkr_mo=
no, 0);
> +                       nsec_raw =3D timekeeping_delta_to_ns(&tk->tkr_raw=
, 0);
> +               } else {
> +                       nsec_real =3D timekeeping_cycles_to_ns(
> +                               &tk->tkr_mono, system_counterval.cycles);
> +                       nsec_raw =3D timekeeping_cycles_to_ns(
> +                               &tk->tkr_raw, system_counterval.cycles);
> +               }

Rather than adding another conditional branch here to go through, why
not just use "cycles" instead of system_counterval.cycles as it seems
to be set properly already?

thanks
-john