Re: [PATCH 2/4] locking: Apply contention tracepoints in the slow path

From: Namhyung Kim <namhyung@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>, Will Deacon <will@kernel.org>,
	Waiman Long <longman@redhat.com>,
	Boqun Feng <boqun.feng@gmail.com>,
	LKML <linux-kernel@vger.kernel.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Steven Rostedt <rostedt@goodmis.org>,
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
	Byungchul Park <byungchul.park@lge.com>,
	"Paul E. McKenney" <paulmck@kernel.org>,
	Arnd Bergmann <arnd@arndb.de>,
	linux-arch@vger.kernel.org, bpf <bpf@vger.kernel.org>,
	Radoslaw Burny <rburny@google.com>
Subject: Re: [PATCH 2/4] locking: Apply contention tracepoints in the slow path
Date: Tue, 1 Mar 2022 10:03:57 -0800	[thread overview]
Message-ID: <CAM9d7cguuyY74L9XK=z384bwO6nTCw4XOFNK1yaJhhmKYTBBoA@mail.gmail.com> (raw)
In-Reply-To: <Yh3cwBloddIGvCjU@hirez.programming.kicks-ass.net>

Hi Peter,

On Tue, Mar 1, 2022 at 12:44 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Mon, Feb 28, 2022 at 05:04:10PM -0800, Namhyung Kim wrote:
> > @@ -80,7 +81,9 @@ static inline void queued_read_lock(struct qrwlock *lock)
> >               return;
> >
> >       /* The slowpath will decrement the reader count, if necessary. */
> > +     LOCK_CONTENTION_BEGIN(lock, LCB_F_READ);
> >       queued_read_lock_slowpath(lock);
> > +     LOCK_CONTENTION_END(lock);
> >  }
> >
> >  /**
> > @@ -94,7 +97,9 @@ static inline void queued_write_lock(struct qrwlock *lock)
> >       if (likely(atomic_try_cmpxchg_acquire(&lock->cnts, &cnts, _QW_LOCKED)))
> >               return;
> >
> > +     LOCK_CONTENTION_BEGIN(lock, LCB_F_WRITE);
> >       queued_write_lock_slowpath(lock);
> > +     LOCK_CONTENTION_END(lock);
> >  }
>
> > @@ -82,7 +83,9 @@ static __always_inline void queued_spin_lock(struct qspinlock *lock)
> >       if (likely(atomic_try_cmpxchg_acquire(&lock->val, &val, _Q_LOCKED_VAL)))
> >               return;
> >
> > +     LOCK_CONTENTION_BEGIN(lock, 0);
> >       queued_spin_lock_slowpath(lock, val);
> > +     LOCK_CONTENTION_END(lock);
> >  }
>
> Can you please stick that _inside_ the slowpath? You really don't want
> to inline that.

I can move it into the slow path with caller ip.

Thanks,
Namhyung