Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754124AbbGWUXv (ORCPT ); Thu, 23 Jul 2015 16:23:51 -0400 Received: from mx2.suse.de ([195.135.220.15]:48031 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753544AbbGWUXt (ORCPT ); Thu, 23 Jul 2015 16:23:49 -0400 Message-ID: <1437683022.3298.41.camel@stgolabs.net> Subject: Re: [PATCH 1/2] locktorture: Support rtmutex torturing From: Davidlohr Bueso To: paulmck@linux.vnet.ibm.com Cc: Peter Zijlstra , Ingo Molnar , Thomas Gleixner , linux-kernel@vger.kernel.org Date: Thu, 23 Jul 2015 13:23:42 -0700 In-Reply-To: <20150723001736.GI3717@linux.vnet.ibm.com> References: <1437599248-12035-1-git-send-email-dave@stgolabs.net> <20150723001736.GI3717@linux.vnet.ibm.com> Content-Type: text/plain; charset="UTF-8" X-Mailer: Evolution 3.12.11 Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 12295 Lines: 347 On Wed, 2015-07-22 at 17:17 -0700, Paul E. McKenney wrote: > On Wed, Jul 22, 2015 at 02:07:27PM -0700, Davidlohr Bueso wrote: > > Real time mutexes is one of the few general primitives > > that we do not have in locktorture. Address this -- a few > > considerations: > > > > o To spice things up, enable competing thread(s) to become > > rt, such that we can stress different prio boosting paths > > in the rtmutex code. Introduce a ->task_boost callback, > > only used by rtmutex-torturer. Tasks will boost/deboost > > around every 50k (arbitrarily) lock/unlock operations. > > > > o Hold times are similar to what we have for other locks: > > only occasionally having longer hold times (per ~200k ops). > > So we roughly do two full rt boost+deboosting ops with > > short hold times. > > > > Signed-off-by: Davidlohr Bueso > > I have queued this one for testing, and by default would push it into > the 4.4 merge window (the one after next). Please let me know if you > want it sooner. Thanks, although here's a v2 with some small fixes: - comment when reseting prio s/500k/50k. - when reseting prio, there's a chance a non-rt task can be converted (which we certainly don't want). Check trsp for nil. While I was planning for 4.3, I can certainly wait a few more weeks for 4.4. Either way is fine. Thanks, Davidlohr 8<------------------------------------------------------------------- [PATCH v2] locktorture: Support rtmutex torturing Real time mutexes is one of the few general primitives that we do not have in locktorture. Address this -- a few considerations: o To spice things up, enable competing thread(s) to become rt, such that we can stress different prio boosting paths in the rtmutex code. Introduce a ->task_boost callback, only used by rtmutex-torturer. Tasks will boost/deboost around every 50k (arbitrarily) lock/unlock operations. o Hold times are similar to what we have for other locks: only occasionally having longer hold times (per ~200k ops). So we roughly do two full rt boost+deboosting ops with short hold times. Signed-off-by: Davidlohr Bueso --- Documentation/locking/locktorture.txt | 3 + kernel/locking/locktorture.c | 114 ++++++++++++++++++++- .../selftests/rcutorture/configs/lock/CFLIST | 3 +- .../selftests/rcutorture/configs/lock/LOCK05 | 6 ++ .../selftests/rcutorture/configs/lock/LOCK05.boot | 1 + 5 files changed, 124 insertions(+), 3 deletions(-) create mode 100644 tools/testing/selftests/rcutorture/configs/lock/LOCK05 create mode 100644 tools/testing/selftests/rcutorture/configs/lock/LOCK05.boot diff --git a/Documentation/locking/locktorture.txt b/Documentation/locking/locktorture.txt index 619f2bb..a2ef3a9 100644 --- a/Documentation/locking/locktorture.txt +++ b/Documentation/locking/locktorture.txt @@ -52,6 +52,9 @@ torture_type Type of lock to torture. By default, only spinlocks will o "mutex_lock": mutex_lock() and mutex_unlock() pairs. + o "rtmutex_lock": rtmutex_lock() and rtmutex_unlock() + pairs. Kernel must have CONFIG_RT_MUTEX=y. + o "rwsem_lock": read/write down() and up() semaphore pairs. torture_runnable Start locktorture at boot time in the case where the diff --git a/kernel/locking/locktorture.c b/kernel/locking/locktorture.c index 3224418..26ddd63 100644 --- a/kernel/locking/locktorture.c +++ b/kernel/locking/locktorture.c @@ -17,12 +17,14 @@ * * Copyright (C) IBM Corporation, 2014 * - * Author: Paul E. McKenney + * Authors: Paul E. McKenney + * Davidlohr Bueso * Based on kernel/rcu/torture.c. */ #include #include #include +#include #include #include #include @@ -91,11 +93,13 @@ struct lock_torture_ops { void (*init)(void); int (*writelock)(void); void (*write_delay)(struct torture_random_state *trsp); + void (*task_boost)(struct torture_random_state *trsp); void (*writeunlock)(void); int (*readlock)(void); void (*read_delay)(struct torture_random_state *trsp); void (*readunlock)(void); - unsigned long flags; + + unsigned long flags; /* for irq spinlocks */ const char *name; }; @@ -139,9 +143,15 @@ static void torture_lock_busted_write_unlock(void) /* BUGGY, do not use in real life!!! */ } +static void torture_boost_dummy(struct torture_random_state *trsp) +{ + /* Only rtmutexes care about priority */ +} + static struct lock_torture_ops lock_busted_ops = { .writelock = torture_lock_busted_write_lock, .write_delay = torture_lock_busted_write_delay, + .task_boost = torture_boost_dummy, .writeunlock = torture_lock_busted_write_unlock, .readlock = NULL, .read_delay = NULL, @@ -185,6 +195,7 @@ static void torture_spin_lock_write_unlock(void) __releases(torture_spinlock) static struct lock_torture_ops spin_lock_ops = { .writelock = torture_spin_lock_write_lock, .write_delay = torture_spin_lock_write_delay, + .task_boost = torture_boost_dummy, .writeunlock = torture_spin_lock_write_unlock, .readlock = NULL, .read_delay = NULL, @@ -211,6 +222,7 @@ __releases(torture_spinlock) static struct lock_torture_ops spin_lock_irq_ops = { .writelock = torture_spin_lock_write_lock_irq, .write_delay = torture_spin_lock_write_delay, + .task_boost = torture_boost_dummy, .writeunlock = torture_lock_spin_write_unlock_irq, .readlock = NULL, .read_delay = NULL, @@ -275,6 +287,7 @@ static void torture_rwlock_read_unlock(void) __releases(torture_rwlock) static struct lock_torture_ops rw_lock_ops = { .writelock = torture_rwlock_write_lock, .write_delay = torture_rwlock_write_delay, + .task_boost = torture_boost_dummy, .writeunlock = torture_rwlock_write_unlock, .readlock = torture_rwlock_read_lock, .read_delay = torture_rwlock_read_delay, @@ -315,6 +328,7 @@ __releases(torture_rwlock) static struct lock_torture_ops rw_lock_irq_ops = { .writelock = torture_rwlock_write_lock_irq, .write_delay = torture_rwlock_write_delay, + .task_boost = torture_boost_dummy, .writeunlock = torture_rwlock_write_unlock_irq, .readlock = torture_rwlock_read_lock_irq, .read_delay = torture_rwlock_read_delay, @@ -354,6 +368,7 @@ static void torture_mutex_unlock(void) __releases(torture_mutex) static struct lock_torture_ops mutex_lock_ops = { .writelock = torture_mutex_lock, .write_delay = torture_mutex_delay, + .task_boost = torture_boost_dummy, .writeunlock = torture_mutex_unlock, .readlock = NULL, .read_delay = NULL, @@ -361,6 +376,90 @@ static struct lock_torture_ops mutex_lock_ops = { .name = "mutex_lock" }; +#ifdef CONFIG_RT_MUTEXES +static DEFINE_RT_MUTEX(torture_rtmutex); + +static int torture_rtmutex_lock(void) __acquires(torture_rtmutex) +{ + rt_mutex_lock(&torture_rtmutex); + return 0; +} + +static void torture_rtmutex_boost(struct torture_random_state *trsp) +{ + int policy; + struct sched_param param; + const unsigned int factor = 50000; /* yes, quite arbitrary */ + + if (!rt_task(current)) { + /* + * (1) Boost priority once every ~50k operations. When the + * task tries to take the lock, the rtmutex it will account + * for the new priority, and do any corresponding pi-dance. + */ + if (trsp && !(torture_random(trsp) % + (cxt.nrealwriters_stress * factor))) { + policy = SCHED_FIFO; + param.sched_priority = MAX_RT_PRIO - 1; + } else /* common case, do nothing */ + return; + } else { + /* + * (2) The task will remain boosted for another ~50k operations, + * then restored back to its original prio, and so forth. + * + * When @trsp is nil, we want to force-reset the task for + * stopping the kthread. + */ + if (!trsp || !(torture_random(trsp) % + (cxt.nrealwriters_stress * factor * 2))) { + policy = SCHED_NORMAL; + param.sched_priority = 0; + } else /* common case, do nothing */ + return; + } + + sched_setscheduler_nocheck(current, policy, ¶m); +} + +static void torture_rtmutex_delay(struct torture_random_state *trsp) +{ + const unsigned long shortdelay_us = 2; + const unsigned long longdelay_ms = 100; + + /* + * We want a short delay mostly to emulate likely code, and + * we want a long delay occasionally to force massive contention. + */ + if (!(torture_random(trsp) % + (cxt.nrealwriters_stress * 2000 * longdelay_ms))) + mdelay(longdelay_ms); + if (!(torture_random(trsp) % + (cxt.nrealwriters_stress * 2 * shortdelay_us))) + udelay(shortdelay_us); +#ifdef CONFIG_PREEMPT + if (!(torture_random(trsp) % (cxt.nrealwriters_stress * 20000))) + preempt_schedule(); /* Allow test to be preempted. */ +#endif +} + +static void torture_rtmutex_unlock(void) __releases(torture_rtmutex) +{ + rt_mutex_unlock(&torture_rtmutex); +} + +static struct lock_torture_ops rtmutex_lock_ops = { + .writelock = torture_rtmutex_lock, + .write_delay = torture_rtmutex_delay, + .task_boost = torture_rtmutex_boost, + .writeunlock = torture_rtmutex_unlock, + .readlock = NULL, + .read_delay = NULL, + .readunlock = NULL, + .name = "rtmutex_lock" +}; +#endif + static DECLARE_RWSEM(torture_rwsem); static int torture_rwsem_down_write(void) __acquires(torture_rwsem) { @@ -419,6 +518,7 @@ static void torture_rwsem_up_read(void) __releases(torture_rwsem) static struct lock_torture_ops rwsem_lock_ops = { .writelock = torture_rwsem_down_write, .write_delay = torture_rwsem_write_delay, + .task_boost = torture_boost_dummy, .writeunlock = torture_rwsem_up_write, .readlock = torture_rwsem_down_read, .read_delay = torture_rwsem_read_delay, @@ -442,6 +542,7 @@ static int lock_torture_writer(void *arg) if ((torture_random(&rand) & 0xfffff) == 0) schedule_timeout_uninterruptible(1); + cxt.cur_ops->task_boost(&rand); cxt.cur_ops->writelock(); if (WARN_ON_ONCE(lock_is_write_held)) lwsp->n_lock_fail++; @@ -456,6 +557,8 @@ static int lock_torture_writer(void *arg) stutter_wait("lock_torture_writer"); } while (!torture_must_stop()); + + cxt.cur_ops->task_boost(NULL); /* reset prio */ torture_kthread_stopping("lock_torture_writer"); return 0; } @@ -642,6 +745,9 @@ static int __init lock_torture_init(void) &spin_lock_ops, &spin_lock_irq_ops, &rw_lock_ops, &rw_lock_irq_ops, &mutex_lock_ops, +#ifdef CONFIG_RT_MUTEXES + &rtmutex_lock_ops, +#endif &rwsem_lock_ops, }; @@ -676,6 +782,10 @@ static int __init lock_torture_init(void) if (strncmp(torture_type, "mutex", 5) == 0) cxt.debug_lock = true; #endif +#ifdef CONFIG_DEBUG_RT_MUTEXES + if (strncmp(torture_type, "rtmutex", 7) == 0) + cxt.debug_lock = true; +#endif #ifdef CONFIG_DEBUG_SPINLOCK if ((strncmp(torture_type, "spin", 4) == 0) || (strncmp(torture_type, "rw_lock", 7) == 0)) diff --git a/tools/testing/selftests/rcutorture/configs/lock/CFLIST b/tools/testing/selftests/rcutorture/configs/lock/CFLIST index 6910b73..6ed3279 100644 --- a/tools/testing/selftests/rcutorture/configs/lock/CFLIST +++ b/tools/testing/selftests/rcutorture/configs/lock/CFLIST @@ -1,4 +1,5 @@ LOCK01 LOCK02 LOCK03 -LOCK04 \ No newline at end of file +LOCK04 +LOCK05 diff --git a/tools/testing/selftests/rcutorture/configs/lock/LOCK05 b/tools/testing/selftests/rcutorture/configs/lock/LOCK05 new file mode 100644 index 0000000..1d1da14 --- /dev/null +++ b/tools/testing/selftests/rcutorture/configs/lock/LOCK05 @@ -0,0 +1,6 @@ +CONFIG_SMP=y +CONFIG_NR_CPUS=4 +CONFIG_HOTPLUG_CPU=y +CONFIG_PREEMPT_NONE=n +CONFIG_PREEMPT_VOLUNTARY=n +CONFIG_PREEMPT=y diff --git a/tools/testing/selftests/rcutorture/configs/lock/LOCK05.boot b/tools/testing/selftests/rcutorture/configs/lock/LOCK05.boot new file mode 100644 index 0000000..8ac3730 --- /dev/null +++ b/tools/testing/selftests/rcutorture/configs/lock/LOCK05.boot @@ -0,0 +1 @@ +locktorture.torture_type=rtmutex_lock -- 2.1.4 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/