From: Davidlohr Bueso <dave@stgolabs.net>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: LKML <linux-kernel@vger.kernel.org>,
Peter Zijlstra <peterz@infradead.org>,
Ingo Molnar <mingo@kernel.org>,
Juri Lelli <juri.lelli@redhat.com>,
Steven Rostedt <rostedt@goodmis.org>,
Daniel Bristot de Oliveira <bristot@redhat.com>,
Will Deacon <will@kernel.org>, Waiman Long <longman@redhat.com>,
Boqun Feng <boqun.feng@gmail.com>,
Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
Mike Galbraith <efault@gmx.de>
Subject: Re: [patch V3 64/64] locking/rtmutex: Add adaptive spinwait mechanism
Date: Mon, 9 Aug 2021 13:28:06 -0700 [thread overview]
Message-ID: <20210809202806.whkrk6cvtejaxqcz@offworld> (raw)
In-Reply-To: <20210805153956.507589092@linutronix.de>
On Thu, 05 Aug 2021, Thomas Gleixner wrote:
>+#ifdef CONFIG_SMP
>+/*
>+ * Note that owner is a speculative pointer and dereferencing relies
>+ * on rcu_read_lock() and the check against the lock owner.
>+ */
I think the description we have in mutex.c is better.
>+static bool rtmutex_adaptive_spinwait(struct rt_mutex_base *lock,
>+ struct task_struct *owner)
I realize that adaptive spinning is the original term from Greg Haskins,
but as Peter suggested, rt_mutex_spin_on_owner() would probably be a
better name upstream considering all our other sleeping locks.
>+{
>+ bool res = true;
>+
>+ rcu_read_lock();
>+ for (;;) {
>+ /* Owner changed. Trylock again */
>+ if (owner != rt_mutex_owner(lock))
>+ break;
>+ /*
>+ * Ensure that owner->on_cpu is dereferenced _after_
>+ * checking the above to be valid.
>+ */
>+ barrier();
>+ if (!owner->on_cpu || need_resched() ||
>+ vcpu_is_preempted(task_cpu(owner))) {
I'm thinking we should also check for if the spinning waiter is no
longer the top-waiter, which could have changed while busy waiting.
Thanks,
Davidlohr
diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index 7522c3abacb6..c6925ebb3c9f 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -1284,12 +1284,9 @@ static __always_inline void __rt_mutex_unlock(struct rt_mutex_base *lock)
}
#ifdef CONFIG_SMP
-/*
- * Note that owner is a speculative pointer and dereferencing relies
- * on rcu_read_lock() and the check against the lock owner.
- */
static bool rtmutex_adaptive_spinwait(struct rt_mutex_base *lock,
- struct task_struct *owner)
+ struct rt_mutex_waiter *waiter,
+ struct task_struct *owner)
{
bool res = true;
@@ -1299,11 +1296,18 @@ static bool rtmutex_adaptive_spinwait(struct rt_mutex_base *lock,
if (owner != rt_mutex_owner(lock))
break;
/*
- * Ensure that owner->on_cpu is dereferenced _after_
- * checking the above to be valid.
+ * Ensure we emit the owner->on_cpu, dereference _after_
+ * checking lock->owner still matches owner. If that fails,
+ * owner might point to freed memory. If it still matches,
+ * the rcu_read_lock() ensures the memory stays valid.
+ *
+ * Also account for changes in the lock's top-waiter, if
+ * it's not us, it was updated while busy waiting.
*/
barrier();
+
if (!owner->on_cpu || need_resched() ||
+ waiter != rt_mutex_top_waiter(lock) ||
vcpu_is_preempted(task_cpu(owner))) {
res = false;
break;
@@ -1315,7 +1319,8 @@ static bool rtmutex_adaptive_spinwait(struct rt_mutex_base *lock,
}
#else
static bool rtmutex_adaptive_spinwait(struct rt_mutex_base *lock,
- struct task_struct *owner)
+ struct rt_mutex_waiter *waiter,
+ struct task_struct *owner)
{
return false;
}
@@ -1434,7 +1439,7 @@ static int __sched rt_mutex_slowlock_block(struct rt_mutex_base *lock,
owner = NULL;
raw_spin_unlock_irq(&lock->wait_lock);
- if (!owner || !rtmutex_adaptive_spinwait(lock, owner))
+ if (!owner || !rtmutex_adaptive_spinwait(lock, waiter, owner))
schedule();
raw_spin_lock_irq(&lock->wait_lock);
@@ -1616,7 +1621,7 @@ static void __sched rtlock_slowlock_locked(struct rt_mutex_base *lock)
owner = NULL;
raw_spin_unlock_irq(&lock->wait_lock);
- if (!owner || !rtmutex_adaptive_spinwait(lock, owner))
+ if (!owner || !rtmutex_adaptive_spinwait(lock, waiter, owner))
schedule_rtlock();
raw_spin_lock_irq(&lock->wait_lock);
prev parent reply other threads:[~2021-08-09 20:28 UTC|newest]
Thread overview: 79+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-05 15:13 [patch V3 00/64] locking, sched: The PREEMPT-RT locking infrastructure Thomas Gleixner
2021-08-05 15:13 ` [patch V3 01/64] sched: Split out the wakeup state check Thomas Gleixner
2021-08-05 15:13 ` [patch V3 02/64] sched: Introduce TASK_RTLOCK_WAIT Thomas Gleixner
2021-08-05 15:13 ` [patch V3 03/64] sched: Reorganize current::__state helpers Thomas Gleixner
2021-08-05 15:13 ` [patch V3 04/64] sched: Prepare for RT sleeping spin/rwlocks Thomas Gleixner
2021-08-05 15:13 ` [patch V3 05/64] sched: Rework the __schedule() preempt argument Thomas Gleixner
2021-08-05 15:13 ` [patch V3 06/64] sched: Provide schedule point for RT locks Thomas Gleixner
2021-08-05 15:13 ` [patch V3 07/64] sched/wake_q: Provide WAKE_Q_HEAD_INITIALIZER Thomas Gleixner
2021-08-05 15:13 ` [patch V3 08/64] media/atomisp: Use lockdep instead of *mutex_is_locked() Thomas Gleixner
2021-08-05 15:13 ` [patch V3 09/64] rtmutex: Remove rt_mutex_is_locked() Thomas Gleixner
2021-08-05 15:13 ` [patch V3 10/64] rtmutex: Convert macros to inlines Thomas Gleixner
2021-08-05 15:13 ` [patch V3 11/64] rtmutex: Switch to try_cmpxchg() Thomas Gleixner
2021-08-05 15:13 ` [patch V3 12/64] rtmutex: Split API and implementation Thomas Gleixner
2021-08-05 15:13 ` [patch V3 13/64] rtmutex: Split out the inner parts of struct rtmutex Thomas Gleixner
2021-08-05 15:13 ` [patch V3 14/64] locking/rtmutex: Provide rt_mutex_slowlock_locked() Thomas Gleixner
2021-08-05 15:13 ` [patch V3 15/64] rtmutex: Provide rt_mutex_base_is_locked() Thomas Gleixner
2021-08-08 20:41 ` Davidlohr Bueso
2021-08-09 10:18 ` Thomas Gleixner
2021-08-05 15:13 ` [patch V3 16/64] locking: Add base code for RT rw_semaphore and rwlock Thomas Gleixner
2021-08-05 15:13 ` [patch V3 17/64] locking/rwsem: Add rtmutex based R/W semaphore implementation Thomas Gleixner
2021-08-05 15:13 ` [patch V3 18/64] locking/rtmutex: Add wake_state to rt_mutex_waiter Thomas Gleixner
2021-08-05 15:13 ` [patch V3 19/64] locking/rtmutex: Provide rt_wake_q and helpers Thomas Gleixner
2021-08-05 15:13 ` [patch V3 20/64] locking/rtmutex: Use rt_mutex_wake_q_head Thomas Gleixner
2021-08-05 15:13 ` [patch V3 21/64] locking/rtmutex: Prepare RT rt_mutex_wake_q for RT locks Thomas Gleixner
2021-08-05 15:13 ` [patch V3 22/64] locking/rtmutex: Guard regular sleeping locks specific functions Thomas Gleixner
2021-08-05 15:13 ` [patch V3 23/64] locking/spinlock: Split the lock types header Thomas Gleixner
2021-08-05 15:13 ` [patch V3 24/64] locking/rtmutex: Prevent future include recursion hell Thomas Gleixner
2021-08-05 15:13 ` [patch V3 25/64] locking/lockdep: Reduce includes in debug_locks.h Thomas Gleixner
2021-08-05 15:13 ` [patch V3 26/64] rbtree: Split out the rbtree type definitions Thomas Gleixner
2021-08-05 15:13 ` [patch V3 27/64] locking/rtmutex: Include only rbtree types Thomas Gleixner
2021-08-05 15:13 ` [patch V3 28/64] locking/spinlock: Provide RT specific spinlock type Thomas Gleixner
2021-08-05 15:13 ` [patch V3 29/64] locking/spinlock: Provide RT variant header Thomas Gleixner
2021-08-05 15:13 ` [patch V3 30/64] locking/rtmutex: Provide the spin/rwlock core lock function Thomas Gleixner
2021-08-05 15:13 ` [patch V3 31/64] locking/spinlock: Provide RT variant Thomas Gleixner
2021-08-05 15:13 ` [patch V3 32/64] locking/rwlock: " Thomas Gleixner
2021-08-05 15:13 ` [patch V3 33/64] locking/mutex: Consolidate core headers Thomas Gleixner
2021-08-05 15:13 ` [patch V3 34/64] locking/mutex: Move waiter to core header Thomas Gleixner
2021-08-05 15:13 ` [patch V3 35/64] locking/ww_mutex: Move ww_mutex declarations into ww_mutex.h Thomas Gleixner
2021-08-05 15:13 ` [patch V3 36/64] locking/mutex: Make mutex::wait_lock raw Thomas Gleixner
2021-08-05 15:13 ` [patch V3 37/64] locking/ww_mutex: Simplify lockdep annotation Thomas Gleixner
2021-08-05 15:13 ` [patch V3 38/64] locking/ww_mutex: Gather mutex_waiter initialization Thomas Gleixner
2021-08-05 15:13 ` [patch V3 39/64] locking/ww_mutex: Split up ww_mutex_unlock() Thomas Gleixner
2021-08-05 15:13 ` [patch V3 40/64] locking/ww_mutex: Split W/W implementation logic Thomas Gleixner
2021-08-05 15:13 ` [patch V3 41/64] locking/ww_mutex: Remove __sched annotation Thomas Gleixner
2021-08-05 15:13 ` [patch V3 42/64] locking/ww_mutex: Abstract waiter iteration Thomas Gleixner
2021-08-05 15:13 ` [patch V3 43/64] locking/ww_mutex: Abstract waiter enqueueing Thomas Gleixner
2021-08-05 15:13 ` [patch V3 44/64] locking/ww_mutex: Abstract mutex accessors Thomas Gleixner
2021-08-05 15:13 ` [patch V3 45/64] locking/ww_mutex: Abstract mutex types Thomas Gleixner
2021-08-05 15:13 ` [patch V3 46/64] locking/ww_mutex: Abstract internal lock access Thomas Gleixner
2021-08-05 15:13 ` [patch V3 47/64] locking/ww_mutex: Implement rt_mutex accessors Thomas Gleixner
2021-08-05 15:13 ` [patch V3 48/64] locking/ww_mutex: Add RT priority to W/W order Thomas Gleixner
2021-08-06 10:48 ` Peter Zijlstra
2021-08-06 11:50 ` Thomas Gleixner
2021-08-05 15:13 ` [patch V3 49/64] locking/ww_mutex: Add rt_mutex based lock type and accessors Thomas Gleixner
2021-08-05 15:13 ` [patch V3 50/64] locking/rtmutex: Extend the rtmutex core to support ww_mutex Thomas Gleixner
2021-08-06 11:00 ` Peter Zijlstra
2021-08-06 11:19 ` Thomas Gleixner
2021-08-06 13:48 ` Peter Zijlstra
2021-08-07 20:07 ` Thomas Gleixner
2021-08-05 15:13 ` [patch V3 51/64] locking/ww_mutex: Implement rtmutex based ww_mutex API functions Thomas Gleixner
2021-08-05 15:13 ` [patch V3 52/64] locking/rtmutex: Add mutex variant for RT Thomas Gleixner
2021-08-05 15:13 ` [patch V3 53/64] lib/test_lockup: Adapt to changed variables Thomas Gleixner
2021-08-05 15:13 ` [patch V3 54/64] futex: Validate waiter correctly in futex_proxy_trylock_atomic() Thomas Gleixner
2021-08-05 15:13 ` [patch V3 55/64] futex: Cleanup stale comments Thomas Gleixner
2021-08-05 15:13 ` [patch V3 56/64] futex: Correct the number of requeued waiters for PI Thomas Gleixner
2021-08-08 17:05 ` Davidlohr Bueso
2021-08-09 8:18 ` Thomas Gleixner
2021-08-09 10:52 ` Thomas Gleixner
2021-08-05 15:13 ` [patch V3 57/64] futex: Restructure futex_requeue() Thomas Gleixner
2021-08-05 15:13 ` [patch V3 58/64] futex: Clarify comment in futex_requeue() Thomas Gleixner
2021-08-08 18:43 ` Davidlohr Bueso
2021-08-09 8:18 ` Thomas Gleixner
2021-08-05 15:13 ` [patch V3 59/64] futex: Simplify handle_early_requeue_pi_wakeup() Thomas Gleixner
2021-08-05 15:14 ` [patch V3 60/64] futex: Prevent requeue_pi() lock nesting issue on RT Thomas Gleixner
2021-08-05 15:14 ` [patch V3 61/64] rtmutex: Prevent lockdep false positive with PI futexes Thomas Gleixner
2021-08-05 15:14 ` [patch V3 62/64] preempt: Adjust PREEMPT_LOCK_OFFSET for RT Thomas Gleixner
2021-08-05 15:14 ` [patch V3 63/64] locking/rtmutex: Implement equal priority lock stealing Thomas Gleixner
2021-08-05 15:14 ` [patch V3 64/64] locking/rtmutex: Add adaptive spinwait mechanism Thomas Gleixner
2021-08-09 20:28 ` Davidlohr Bueso [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210809202806.whkrk6cvtejaxqcz@offworld \
--to=dave@stgolabs.net \
--cc=bigeasy@linutronix.de \
--cc=boqun.feng@gmail.com \
--cc=bristot@redhat.com \
--cc=efault@gmx.de \
--cc=juri.lelli@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=longman@redhat.com \
--cc=mingo@kernel.org \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
--cc=tglx@linutronix.de \
--cc=will@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).