[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210714112530.GE2591@worktop.programming.kicks-ass.net>
Date: Wed, 14 Jul 2021 13:25:30 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: LKML <linux-kernel@...r.kernel.org>,
Ingo Molnar <mingo@...nel.org>,
Juri Lelli <juri.lelli@...hat.com>,
Steven Rostedt <rostedt@...dmis.org>,
Daniel Bristot de Oliveira <bristot@...hat.com>,
Will Deacon <will@...nel.org>,
Waiman Long <longman@...hat.com>,
Boqun Feng <boqun.feng@...il.com>,
Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
Davidlohr Bueso <dave@...olabs.net>
Subject: Re: [patch 27/50] locking/spinlock: Provide RT variant
On Tue, Jul 13, 2021 at 05:11:21PM +0200, Thomas Gleixner wrote:
> + * - Non RT spin/rw_locks disable preemption and evtl. interrupts.
> + * Disabling preemption has the side effect of disabling migration and
> + * preventing RCU grace periods.
> + *
> + * The RT substitutions explicitly disable migration and take
> + * rcu_read_lock() across the lock held section.
> +static __always_inline void __rt_spin_lock(spinlock_t *lock)
> +{
> + rtlock_lock(&lock->lock);
> + rcu_read_lock();
> + migrate_disable();
> +}
One notable difference is that regular spinlocks disable preemption (and
hence imply the other things) *before* they acquire the lock, while this
thing does the implied semantics *after* it acquires the lock.
The difference is of course that the acquisition period is not covered
and I don't think anybody actually relies on that, nor do I readily see
how one could.
Powered by blists - more mailing lists