[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y0dT0D3d/07mNHAK@google.com>
Date: Wed, 12 Oct 2022 23:54:56 +0000
From: Joel Fernandes <joel@...lfernandes.org>
To: Waiman Long <longman@...hat.com>
Cc: Connor O'Brien <connoro@...gle.com>, linux-kernel@...r.kernel.org,
kernel-team@...roid.com, John Stultz <jstultz@...gle.com>,
Qais Yousef <qais.yousef@....com>,
Ingo Molnar <mingo@...hat.com>,
Peter Zijlstra <peterz@...radead.org>,
Juri Lelli <juri.lelli@...hat.com>,
Vincent Guittot <vincent.guittot@...aro.org>,
Dietmar Eggemann <dietmar.eggemann@....com>,
Steven Rostedt <rostedt@...dmis.org>,
Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
Daniel Bristot de Oliveira <bristot@...hat.com>,
Valentin Schneider <vschneid@...hat.com>,
Will Deacon <will@...nel.org>,
Boqun Feng <boqun.feng@...il.com>,
"Paul E . McKenney" <paulmck@...nel.org>
Subject: Re: [RFC PATCH 01/11] locking/ww_mutex: Remove wakeups from under
mutex::wait_lock
On Tue, Oct 04, 2022 at 12:01:13PM -0400, Waiman Long wrote:
> On 10/3/22 17:44, Connor O'Brien wrote:
> > diff --git a/kernel/locking/ww_mutex.h b/kernel/locking/ww_mutex.h
> > index 56f139201f24..dfc174cd96c6 100644
> > --- a/kernel/locking/ww_mutex.h
> > +++ b/kernel/locking/ww_mutex.h
> > @@ -161,6 +161,11 @@ static inline void lockdep_assert_wait_lock_held(struct rt_mutex *lock)
> > #endif /* WW_RT */
> > +void ww_ctx_wake(struct ww_acquire_ctx *ww_ctx)
> > +{
> > + wake_up_q(&ww_ctx->wake_q);
> > +}
> > +
> > /*
> > * Wait-Die:
> > * The newer transactions are killed when:
> > @@ -284,7 +289,7 @@ __ww_mutex_die(struct MUTEX *lock, struct MUTEX_WAITER *waiter,
> > #ifndef WW_RT
> > debug_mutex_wake_waiter(lock, waiter);
> > #endif
> > - wake_up_process(waiter->task);
> > + wake_q_add(&ww_ctx->wake_q, waiter->task);
> > }
> > return true;
> > @@ -331,7 +336,7 @@ static bool __ww_mutex_wound(struct MUTEX *lock,
> > * wakeup pending to re-read the wounded state.
> > */
> > if (owner != current)
> > - wake_up_process(owner);
> > + wake_q_add(&ww_ctx->wake_q, owner);
> > return true;
> > }
> > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > index ee28253c9ac0..617e737392be 100644
> > --- a/kernel/sched/core.c
> > +++ b/kernel/sched/core.c
> > @@ -1013,6 +1013,13 @@ void wake_up_q(struct wake_q_head *head)
> > wake_up_process(task);
> > put_task_struct(task);
> > }
> > + /*
> > + * XXX connoro: seems this is needed now that ww_ctx_wake() passes in a
> > + * wake_q_head that is embedded in struct ww_acquire_ctx rather than
> > + * declared locally.
> > + */
> > + head->first = node;
> > + head->lastp = &head->first;
> > }
>
> You shouldn't do wake_q_init() here in wake_up_q(). Instead, you should do
> it in ww_ctx_wake() right after the wake_up_q() call.
> :
Exactly, it is also mentioned in the wake_q.h header:
* The DEFINE_WAKE_Q macro declares and initializes the list head.
* wake_up_q() does NOT reinitialize the list; it's expected to be
* called near the end of a function. Otherwise, the list can be
* re-initialized for later re-use by wake_q_init().
*
Thanks.
Powered by blists - more mailing lists