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: <20230803083352.1585-1-zegao@tencent.com> <20230803083352.1585-4-zegao@tencent.com>
 <20230803051826.25ba7973@gandalf.local.home>
In-Reply-To: <20230803051826.25ba7973@gandalf.local.home>
From:   Ze Gao <zegao2021@gmail.com>
Date:   Thu, 3 Aug 2023 20:54:16 +0800
Message-ID: <CAD8CoPBkK64H0UE+_K9yxmiq-kMTvuM=N3S_H1bhdQuJ2JV2Jg@mail.gmail.com>
Subject: Re: [RFC PATCH v6 3/5] sched, tracing: reorganize fields of switch
 event struct
To:     Steven Rostedt <rostedt@goodmis.org>
Cc:     Adrian Hunter <adrian.hunter@intel.com>,
        Alexander Shishkin <alexander.shishkin@linux.intel.com>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Ian Rogers <irogers@google.com>,
        Ingo Molnar <mingo@redhat.com>, Jiri Olsa <jolsa@kernel.org>,
        Mark Rutland <mark.rutland@arm.com>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        Namhyung Kim <namhyung@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org,
        linux-trace-kernel@vger.kernel.org, Ze Gao <zegao@tencent.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Aug 3, 2023 at 5:18=E2=80=AFPM Steven Rostedt <rostedt@goodmis.org>=
 wrote:
>
> On Thu,  3 Aug 2023 04:33:50 -0400
> Ze Gao <zegao2021@gmail.com> wrote:
>
> > Report prioritiy and prev_state in 'short' to save some buffer
> > space. And also reorder the fields so that we take struct
> > alignment into consideration to make the record compact.
>
> If I were to write this, I would have wrote:
>
>   The prev_state field in the sched_switch event is assigned by
>   __trace_sched_switch_state(). The largest number that function will ret=
urn
>   is TASK_REPORT_MAX which is just 0x100. There's no reason that the
>   prev_state field is a full 32 bits when it is using just 9 bits max. In
>   order to save space on the ring buffer, shrink the prev_state to 16 bit=
s
>   (short).
>
>   Also, change the positions of the other fields to accommodate the short
>   value of prev_state to eliminate any holes that were created in the
>   structure.
>
> See the difference?
>
> >
> >  #ifdef CREATE_TRACE_POINTS
> > -static inline long __trace_sched_switch_state(bool preempt,
> > +static inline short __trace_sched_switch_state(bool preempt,
> >                                             unsigned int prev_state,
> >                                             struct task_struct *p)
> >  {
> >       unsigned int state;
> >
> >  #ifdef CONFIG_SCHED_DEBUG
> > -     BUG_ON(p !=3D current);
> > +     WARN_ON_ONCE(p !=3D current);
> >  #endif /* CONFIG_SCHED_DEBUG */
>
> The above needs to be a separate patch.

I've moved this to a new patch, and this is the changelog:

    sched, tracing: change BUG_ON to WARN_ON_ONCE in __trace_sched_switch_s=
tate

    BUG_ON() was introduced in 2014 and old, and we
    switch it to WARN_ON_ONCE() to not to crash the
    kernel when the sched-out task is unexpected than
    the current, as suggested by Steven.

    Signed-off-by: Ze Gao <zegao@tencent.com>

Regards,
Ze

> >
> >       /*
> > @@ -229,23 +229,23 @@ TRACE_EVENT(sched_switch,
> >       TP_ARGS(preempt, prev, next, prev_state),
> >
> >       TP_STRUCT__entry(
> > -             __array(        char,   prev_comm,      TASK_COMM_LEN   )
> >               __field(        pid_t,  prev_pid                        )
> > -             __field(        int,    prev_prio                       )
> > -             __field(        long,   prev_state                      )
> > -             __array(        char,   next_comm,      TASK_COMM_LEN   )
> >               __field(        pid_t,  next_pid                        )
> > -             __field(        int,    next_prio                       )
> > +             __field(        short,  prev_prio                       )
> > +             __field(        short,  next_prio                       )
> > +             __array(        char,   prev_comm,      TASK_COMM_LEN   )
> > +             __array(        char,   next_comm,      TASK_COMM_LEN   )
> > +             __field(        short,  prev_state                      )
> >       ),
> >
> >       TP_fast_assign(
> > -             memcpy(__entry->next_comm, next->comm, TASK_COMM_LEN);
> > -             __entry->prev_pid       =3D prev->pid;
> > -             __entry->prev_prio      =3D prev->prio;
> > -             __entry->prev_state     =3D __trace_sched_switch_state(pr=
eempt, prev_state, prev);
> > +             __entry->prev_pid               =3D prev->pid;
> > +             __entry->next_pid               =3D next->pid;
> > +             __entry->prev_prio              =3D (short) prev->prio;
> > +             __entry->next_prio              =3D (short) next->prio;
> >               memcpy(__entry->prev_comm, prev->comm, TASK_COMM_LEN);
> > -             __entry->next_pid       =3D next->pid;
> > -             __entry->next_prio      =3D next->prio;
> > +             memcpy(__entry->next_comm, next->comm, TASK_COMM_LEN);
> > +             __entry->prev_state             =3D __trace_sched_switch_=
state(preempt, prev_state, prev);
> >               /* XXX SCHED_DEADLINE */
> >       ),
> >
>