[deliverable/linux.git] / kernel / rcupdate.c

/*
 * Read-Copy Update mechanism for mutual exclusion
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
 *
 * Copyright (C) IBM Corporation, 2001
 *
 * Authors: Dipankar Sarma <dipankar@in.ibm.com>
 *	    Manfred Spraul <manfred@colorfullife.com>
 * 
 * Based on the original work by Paul McKenney <paulmck@us.ibm.com>
 * and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.
 * Papers:
 * http://www.rdrop.com/users/paulmck/paper/rclockpdcsproof.pdf
 * http://lse.sourceforge.net/locking/rclock_OLS.2001.05.01c.sc.pdf (OLS2001)
 *
 * For detailed explanation of Read-Copy Update mechanism see -
 * 		http://lse.sourceforge.net/locking/rcupdate.html
 *
 */
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/smp.h>
#include <linux/rcupdate.h>
#include <linux/interrupt.h>
#include <linux/sched.h>
#include <asm/atomic.h>
#include <linux/bitops.h>
#include <linux/module.h>
#include <linux/completion.h>
#include <linux/moduleparam.h>
#include <linux/percpu.h>
#include <linux/notifier.h>
#include <linux/rcupdate.h>
#include <linux/cpu.h>
#include <linux/mutex.h>

/* Definition for rcupdate control block. */
static struct rcu_ctrlblk rcu_ctrlblk = {
	.cur = -300,
	.completed = -300,
	.lock = SPIN_LOCK_UNLOCKED,
	.cpumask = CPU_MASK_NONE,
};
static struct rcu_ctrlblk rcu_bh_ctrlblk = {
	.cur = -300,
	.completed = -300,
	.lock = SPIN_LOCK_UNLOCKED,
	.cpumask = CPU_MASK_NONE,
};

DEFINE_PER_CPU(struct rcu_data, rcu_data) = { 0L };
DEFINE_PER_CPU(struct rcu_data, rcu_bh_data) = { 0L };

/* Fake initialization required by compiler */
static DEFINE_PER_CPU(struct tasklet_struct, rcu_tasklet) = {NULL};
static int blimit = 10;
static int qhimark = 10000;
static int qlowmark = 100;
#ifdef CONFIG_SMP
static int rsinterval = 1000;
#endif

static atomic_t rcu_barrier_cpu_count;
static DEFINE_MUTEX(rcu_barrier_mutex);
static struct completion rcu_barrier_completion;

#ifdef CONFIG_SMP
static void force_quiescent_state(struct rcu_data *rdp,
			struct rcu_ctrlblk *rcp)
{
	int cpu;
	cpumask_t cpumask;
	set_need_resched();
	if (unlikely(rdp->qlen - rdp->last_rs_qlen > rsinterval)) {
		rdp->last_rs_qlen = rdp->qlen;
		/*
		 * Don't send IPI to itself. With irqs disabled,
		 * rdp->cpu is the current cpu.
		 */
		cpumask = rcp->cpumask;
		cpu_clear(rdp->cpu, cpumask);
		for_each_cpu_mask(cpu, cpumask)
			smp_send_reschedule(cpu);
	}
}
#else
static inline void force_quiescent_state(struct rcu_data *rdp,
			struct rcu_ctrlblk *rcp)
{
	set_need_resched();
}
#endif

/**
 * call_rcu - Queue an RCU callback for invocation after a grace period.
 * @head: structure to be used for queueing the RCU updates.
 * @func: actual update function to be invoked after the grace period
 *
 * The update function will be invoked some time after a full grace
 * period elapses, in other words after all currently executing RCU
 * read-side critical sections have completed.  RCU read-side critical
 * sections are delimited by rcu_read_lock() and rcu_read_unlock(),
 * and may be nested.
 */
void fastcall call_rcu(struct rcu_head *head,
				void (*func)(struct rcu_head *rcu))
{
	unsigned long flags;
	struct rcu_data *rdp;

	head->func = func;
	head->next = NULL;
	local_irq_save(flags);
	rdp = &__get_cpu_var(rcu_data);
	*rdp->nxttail = head;
	rdp->nxttail = &head->next;
	if (unlikely(++rdp->qlen > qhimark)) {
		rdp->blimit = INT_MAX;
		force_quiescent_state(rdp, &rcu_ctrlblk);
	}
	local_irq_restore(flags);
}

/**
 * call_rcu_bh - Queue an RCU for invocation after a quicker grace period.
 * @head: structure to be used for queueing the RCU updates.
 * @func: actual update function to be invoked after the grace period
 *
 * The update function will be invoked some time after a full grace
 * period elapses, in other words after all currently executing RCU
 * read-side critical sections have completed. call_rcu_bh() assumes
 * that the read-side critical sections end on completion of a softirq
 * handler. This means that read-side critical sections in process
 * context must not be interrupted by softirqs. This interface is to be
 * used when most of the read-side critical sections are in softirq context.
 * RCU read-side critical sections are delimited by rcu_read_lock() and
 * rcu_read_unlock(), * if in interrupt context or rcu_read_lock_bh()
 * and rcu_read_unlock_bh(), if in process context. These may be nested.
 */
void fastcall call_rcu_bh(struct rcu_head *head,
				void (*func)(struct rcu_head *rcu))
{
	unsigned long flags;
	struct rcu_data *rdp;

	head->func = func;
	head->next = NULL;
	local_irq_save(flags);
	rdp = &__get_cpu_var(rcu_bh_data);
	*rdp->nxttail = head;
	rdp->nxttail = &head->next;

	if (unlikely(++rdp->qlen > qhimark)) {
		rdp->blimit = INT_MAX;
		force_quiescent_state(rdp, &rcu_bh_ctrlblk);
	}

	local_irq_restore(flags);
}

/*
 * Return the number of RCU batches processed thus far.  Useful
 * for debug and statistics.
 */
long rcu_batches_completed(void)
{
	return rcu_ctrlblk.completed;
}

static void rcu_barrier_callback(struct rcu_head *notused)
{
	if (atomic_dec_and_test(&rcu_barrier_cpu_count))
		complete(&rcu_barrier_completion);
}

/*
 * Called with preemption disabled, and from cross-cpu IRQ context.
 */
static void rcu_barrier_func(void *notused)
{
	int cpu = smp_processor_id();
	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
	struct rcu_head *head;

	head = &rdp->barrier;
	atomic_inc(&rcu_barrier_cpu_count);
	call_rcu(head, rcu_barrier_callback);
}

/**
 * rcu_barrier - Wait until all the in-flight RCUs are complete.
 */
void rcu_barrier(void)
{
	BUG_ON(in_interrupt());
	/* Take cpucontrol mutex to protect against CPU hotplug */
	mutex_lock(&rcu_barrier_mutex);
	init_completion(&rcu_barrier_completion);
	atomic_set(&rcu_barrier_cpu_count, 0);
	on_each_cpu(rcu_barrier_func, NULL, 0, 1);
	wait_for_completion(&rcu_barrier_completion);
	mutex_unlock(&rcu_barrier_mutex);
}
EXPORT_SYMBOL_GPL(rcu_barrier);

/*
 * Invoke the completed RCU callbacks. They are expected to be in
 * a per-cpu list.
 */
static void rcu_do_batch(struct rcu_data *rdp)
{
	struct rcu_head *next, *list;
	int count = 0;

	list = rdp->donelist;
	while (list) {
		next = rdp->donelist = list->next;
		list->func(list);
		list = next;
		rdp->qlen--;
		if (++count >= rdp->blimit)
			break;
	}
	if (rdp->blimit == INT_MAX && rdp->qlen <= qlowmark)
		rdp->blimit = blimit;
	if (!rdp->donelist)
		rdp->donetail = &rdp->donelist;
	else
		tasklet_schedule(&per_cpu(rcu_tasklet, rdp->cpu));
}

/*
 * Grace period handling:
 * The grace period handling consists out of two steps:
 * - A new grace period is started.
 *   This is done by rcu_start_batch. The start is not broadcasted to
 *   all cpus, they must pick this up by comparing rcp->cur with
 *   rdp->quiescbatch. All cpus are recorded  in the
 *   rcu_ctrlblk.cpumask bitmap.
 * - All cpus must go through a quiescent state.
 *   Since the start of the grace period is not broadcasted, at least two
 *   calls to rcu_check_quiescent_state are required:
 *   The first call just notices that a new grace period is running. The
 *   following calls check if there was a quiescent state since the beginning
 *   of the grace period. If so, it updates rcu_ctrlblk.cpumask. If
 *   the bitmap is empty, then the grace period is completed.
 *   rcu_check_quiescent_state calls rcu_start_batch(0) to start the next grace
 *   period (if necessary).
 */
/*
 * Register a new batch of callbacks, and start it up if there is currently no
 * active batch and the batch to be registered has not already occurred.
 * Caller must hold rcu_ctrlblk.lock.
 */
static void rcu_start_batch(struct rcu_ctrlblk *rcp)
{
	if (rcp->next_pending &&
			rcp->completed == rcp->cur) {
		rcp->next_pending = 0;
		/*
		 * next_pending == 0 must be visible in
		 * __rcu_process_callbacks() before it can see new value of cur.
		 */
		smp_wmb();
		rcp->cur++;

		/*
		 * Accessing nohz_cpu_mask before incrementing rcp->cur needs a
		 * Barrier  Otherwise it can cause tickless idle CPUs to be
		 * included in rcp->cpumask, which will extend graceperiods
		 * unnecessarily.
		 */
		smp_mb();
		cpus_andnot(rcp->cpumask, cpu_online_map, nohz_cpu_mask);

	}
}

/*
 * cpu went through a quiescent state since the beginning of the grace period.
 * Clear it from the cpu mask and complete the grace period if it was the last
 * cpu. Start another grace period if someone has further entries pending
 */
static void cpu_quiet(int cpu, struct rcu_ctrlblk *rcp)
{
	cpu_clear(cpu, rcp->cpumask);
	if (cpus_empty(rcp->cpumask)) {
		/* batch completed ! */
		rcp->completed = rcp->cur;
		rcu_start_batch(rcp);
	}
}

/*
 * Check if the cpu has gone through a quiescent state (say context
 * switch). If so and if it already hasn't done so in this RCU
 * quiescent cycle, then indicate that it has done so.
 */
static void rcu_check_quiescent_state(struct rcu_ctrlblk *rcp,
					struct rcu_data *rdp)
{
	if (rdp->quiescbatch != rcp->cur) {
		/* start new grace period: */
		rdp->qs_pending = 1;
		rdp->passed_quiesc = 0;
		rdp->quiescbatch = rcp->cur;
		return;
	}

	/* Grace period already completed for this cpu?
	 * qs_pending is checked instead of the actual bitmap to avoid
	 * cacheline trashing.
	 */
	if (!rdp->qs_pending)
		return;

	/* 
	 * Was there a quiescent state since the beginning of the grace
	 * period? If no, then exit and wait for the next call.
	 */
	if (!rdp->passed_quiesc)
		return;
	rdp->qs_pending = 0;

	spin_lock(&rcp->lock);
	/*
	 * rdp->quiescbatch/rcp->cur and the cpu bitmap can come out of sync
	 * during cpu startup. Ignore the quiescent state.
	 */
	if (likely(rdp->quiescbatch == rcp->cur))
		cpu_quiet(rdp->cpu, rcp);

	spin_unlock(&rcp->lock);
}


#ifdef CONFIG_HOTPLUG_CPU

/* warning! helper for rcu_offline_cpu. do not use elsewhere without reviewing
 * locking requirements, the list it's pulling from has to belong to a cpu
 * which is dead and hence not processing interrupts.
 */
static void rcu_move_batch(struct rcu_data *this_rdp, struct rcu_head *list,
				struct rcu_head **tail)
{
	local_irq_disable();
	*this_rdp->nxttail = list;
	if (list)
		this_rdp->nxttail = tail;
	local_irq_enable();
}

static void __rcu_offline_cpu(struct rcu_data *this_rdp,
				struct rcu_ctrlblk *rcp, struct rcu_data *rdp)
{
	/* if the cpu going offline owns the grace period
	 * we can block indefinitely waiting for it, so flush
	 * it here
	 */
	spin_lock_bh(&rcp->lock);
	if (rcp->cur != rcp->completed)
		cpu_quiet(rdp->cpu, rcp);
	spin_unlock_bh(&rcp->lock);
	rcu_move_batch(this_rdp, rdp->curlist, rdp->curtail);
	rcu_move_batch(this_rdp, rdp->nxtlist, rdp->nxttail);
	rcu_move_batch(this_rdp, rdp->donelist, rdp->donetail);
}

static void rcu_offline_cpu(int cpu)
{
	struct rcu_data *this_rdp = &get_cpu_var(rcu_data);
	struct rcu_data *this_bh_rdp = &get_cpu_var(rcu_bh_data);

	__rcu_offline_cpu(this_rdp, &rcu_ctrlblk,
					&per_cpu(rcu_data, cpu));
	__rcu_offline_cpu(this_bh_rdp, &rcu_bh_ctrlblk,
					&per_cpu(rcu_bh_data, cpu));
	put_cpu_var(rcu_data);
	put_cpu_var(rcu_bh_data);
	tasklet_kill_immediate(&per_cpu(rcu_tasklet, cpu), cpu);
}

#else

static void rcu_offline_cpu(int cpu)
{
}

#endif

/*
 * This does the RCU processing work from tasklet context. 
 */
static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp,
					struct rcu_data *rdp)
{
	if (rdp->curlist && !rcu_batch_before(rcp->completed, rdp->batch)) {
		*rdp->donetail = rdp->curlist;
		rdp->donetail = rdp->curtail;
		rdp->curlist = NULL;
		rdp->curtail = &rdp->curlist;
	}

	if (rdp->nxtlist && !rdp->curlist) {
		local_irq_disable();
		rdp->curlist = rdp->nxtlist;
		rdp->curtail = rdp->nxttail;
		rdp->nxtlist = NULL;
		rdp->nxttail = &rdp->nxtlist;
		local_irq_enable();

		/*
		 * start the next batch of callbacks
		 */

		/* determine batch number */
		rdp->batch = rcp->cur + 1;
		/* see the comment and corresponding wmb() in
		 * the rcu_start_batch()
		 */
		smp_rmb();

		if (!rcp->next_pending) {
			/* and start it/schedule start if it's a new batch */
			spin_lock(&rcp->lock);
			rcp->next_pending = 1;
			rcu_start_batch(rcp);
			spin_unlock(&rcp->lock);
		}
	}

	rcu_check_quiescent_state(rcp, rdp);
	if (rdp->donelist)
		rcu_do_batch(rdp);
}

static void rcu_process_callbacks(unsigned long unused)
{
	__rcu_process_callbacks(&rcu_ctrlblk, &__get_cpu_var(rcu_data));
	__rcu_process_callbacks(&rcu_bh_ctrlblk, &__get_cpu_var(rcu_bh_data));
}

static int __rcu_pending(struct rcu_ctrlblk *rcp, struct rcu_data *rdp)
{
	/* This cpu has pending rcu entries and the grace period
	 * for them has completed.
	 */
	if (rdp->curlist && !rcu_batch_before(rcp->completed, rdp->batch))
		return 1;

	/* This cpu has no pending entries, but there are new entries */
	if (!rdp->curlist && rdp->nxtlist)
		return 1;

	/* This cpu has finished callbacks to invoke */
	if (rdp->donelist)
		return 1;

	/* The rcu core waits for a quiescent state from the cpu */
	if (rdp->quiescbatch != rcp->cur || rdp->qs_pending)
		return 1;

	/* nothing to do */
	return 0;
}

/*
 * Check to see if there is any immediate RCU-related work to be done
 * by the current CPU, returning 1 if so.  This function is part of the
 * RCU implementation; it is -not- an exported member of the RCU API.
 */
int rcu_pending(int cpu)
{
	return __rcu_pending(&rcu_ctrlblk, &per_cpu(rcu_data, cpu)) ||
		__rcu_pending(&rcu_bh_ctrlblk, &per_cpu(rcu_bh_data, cpu));
}

/*
 * Check to see if any future RCU-related work will need to be done
 * by the current CPU, even if none need be done immediately, returning
 * 1 if so.  This function is part of the RCU implementation; it is -not-
 * an exported member of the RCU API.
 */
int rcu_needs_cpu(int cpu)
{
	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
	struct rcu_data *rdp_bh = &per_cpu(rcu_bh_data, cpu);

	return (!!rdp->curlist || !!rdp_bh->curlist || rcu_pending(cpu));
}

void rcu_check_callbacks(int cpu, int user)
{
	if (user || 
	    (idle_cpu(cpu) && !in_softirq() && 
				hardirq_count() <= (1 << HARDIRQ_SHIFT))) {
		rcu_qsctr_inc(cpu);
		rcu_bh_qsctr_inc(cpu);
	} else if (!in_softirq())
		rcu_bh_qsctr_inc(cpu);
	tasklet_schedule(&per_cpu(rcu_tasklet, cpu));
}

static void rcu_init_percpu_data(int cpu, struct rcu_ctrlblk *rcp,
						struct rcu_data *rdp)
{
	memset(rdp, 0, sizeof(*rdp));
	rdp->curtail = &rdp->curlist;
	rdp->nxttail = &rdp->nxtlist;
	rdp->donetail = &rdp->donelist;
	rdp->quiescbatch = rcp->completed;
	rdp->qs_pending = 0;
	rdp->cpu = cpu;
	rdp->blimit = blimit;
}

static void __devinit rcu_online_cpu(int cpu)
{
	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
	struct rcu_data *bh_rdp = &per_cpu(rcu_bh_data, cpu);

	rcu_init_percpu_data(cpu, &rcu_ctrlblk, rdp);
	rcu_init_percpu_data(cpu, &rcu_bh_ctrlblk, bh_rdp);
	tasklet_init(&per_cpu(rcu_tasklet, cpu), rcu_process_callbacks, 0UL);
}

static int rcu_cpu_notify(struct notifier_block *self,
				unsigned long action, void *hcpu)
{
	long cpu = (long)hcpu;
	switch (action) {
	case CPU_UP_PREPARE:
		rcu_online_cpu(cpu);
		break;
	case CPU_DEAD:
		rcu_offline_cpu(cpu);
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}

static struct notifier_block rcu_nb = {
	.notifier_call	= rcu_cpu_notify,
};

/*
 * Initializes rcu mechanism.  Assumed to be called early.
 * That is before local timer(SMP) or jiffie timer (uniproc) is setup.
 * Note that rcu_qsctr and friends are implicitly
 * initialized due to the choice of ``0'' for RCU_CTR_INVALID.
 */
void __init rcu_init(void)
{
	rcu_cpu_notify(&rcu_nb, CPU_UP_PREPARE,
			(void *)(long)smp_processor_id());
	/* Register notifier for non-boot CPUs */
	register_cpu_notifier(&rcu_nb);
}

struct rcu_synchronize {
	struct rcu_head head;
	struct completion completion;
};

/* Because of FASTCALL declaration of complete, we use this wrapper */
static void wakeme_after_rcu(struct rcu_head  *head)
{
	struct rcu_synchronize *rcu;

	rcu = container_of(head, struct rcu_synchronize, head);
	complete(&rcu->completion);
}

/**
 * synchronize_rcu - wait until a grace period has elapsed.
 *
 * Control will return to the caller some time after a full grace
 * period has elapsed, in other words after all currently executing RCU
 * read-side critical sections have completed.  RCU read-side critical
 * sections are delimited by rcu_read_lock() and rcu_read_unlock(),
 * and may be nested.
 *
 * If your read-side code is not protected by rcu_read_lock(), do -not-
 * use synchronize_rcu().
 */
void synchronize_rcu(void)
{
	struct rcu_synchronize rcu;

	init_completion(&rcu.completion);
	/* Will wake me after RCU finished */
	call_rcu(&rcu.head, wakeme_after_rcu);

	/* Wait for it */
	wait_for_completion(&rcu.completion);
}

/*
 * Deprecated, use synchronize_rcu() or synchronize_sched() instead.
 */
void synchronize_kernel(void)
{
	synchronize_rcu();
}

module_param(blimit, int, 0);
module_param(qhimark, int, 0);
module_param(qlowmark, int, 0);
#ifdef CONFIG_SMP
module_param(rsinterval, int, 0);
#endif
EXPORT_SYMBOL_GPL(rcu_batches_completed);
EXPORT_SYMBOL_GPL_FUTURE(call_rcu);	/* WARNING: GPL-only in April 2006. */
EXPORT_SYMBOL_GPL_FUTURE(call_rcu_bh);	/* WARNING: GPL-only in April 2006. */
EXPORT_SYMBOL_GPL(synchronize_rcu);
EXPORT_SYMBOL_GPL_FUTURE(synchronize_kernel); /* WARNING: GPL-only in April 2006. */
Commit	Line	Data
1da177e4 LT	1	/*
	2	* Read-Copy Update mechanism for mutual exclusion
	3	*
	4	* This program is free software; you can redistribute it and/or modify
	5	* it under the terms of the GNU General Public License as published by
	6	* the Free Software Foundation; either version 2 of the License, or
	7	* (at your option) any later version.
	8	*
	9	* This program is distributed in the hope that it will be useful,
	10	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	12	* GNU General Public License for more details.
	13	*
	14	* You should have received a copy of the GNU General Public License
	15	* along with this program; if not, write to the Free Software
	16	* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
	17	*
	18	* Copyright (C) IBM Corporation, 2001
	19	*
	20	* Authors: Dipankar Sarma <dipankar@in.ibm.com>
	21	* Manfred Spraul <manfred@colorfullife.com>
	22	*
	23	* Based on the original work by Paul McKenney <paulmck@us.ibm.com>
	24	* and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.
	25	* Papers:
	26	* http://www.rdrop.com/users/paulmck/paper/rclockpdcsproof.pdf
	27	* http://lse.sourceforge.net/locking/rclock_OLS.2001.05.01c.sc.pdf (OLS2001)
	28	*
	29	* For detailed explanation of Read-Copy Update mechanism see -
	30	* http://lse.sourceforge.net/locking/rcupdate.html
	31	*
	32	*/
	33	#include <linux/types.h>
	34	#include <linux/kernel.h>
	35	#include <linux/init.h>
	36	#include <linux/spinlock.h>
	37	#include <linux/smp.h>
e56d0903	38	#include <linux/rcupdate.h>
1da177e4 LT	39	#include <linux/interrupt.h>
	40	#include <linux/sched.h>
	41	#include <asm/atomic.h>
	42	#include <linux/bitops.h>
	43	#include <linux/module.h>
	44	#include <linux/completion.h>
	45	#include <linux/moduleparam.h>
	46	#include <linux/percpu.h>
	47	#include <linux/notifier.h>
	48	#include <linux/rcupdate.h>
	49	#include <linux/cpu.h>
9331b315	50	#include <linux/mutex.h>
1da177e4 LT	51
1da177e4 LT	52	/* Definition for rcupdate control block. */
2178426d	53	static struct rcu_ctrlblk rcu_ctrlblk = {
69a0b315 ON	54	.cur = -300,
	55	.completed = -300,
	56	.lock = SPIN_LOCK_UNLOCKED,
	57	.cpumask = CPU_MASK_NONE,
	58	};
2178426d	59	static struct rcu_ctrlblk rcu_bh_ctrlblk = {
69a0b315 ON	60	.cur = -300,
	61	.completed = -300,
	62	.lock = SPIN_LOCK_UNLOCKED,
	63	.cpumask = CPU_MASK_NONE,
1da177e4	64	};
1da177e4 LT	65
	66	DEFINE_PER_CPU(struct rcu_data, rcu_data) = { 0L };
	67	DEFINE_PER_CPU(struct rcu_data, rcu_bh_data) = { 0L };
	68
	69	/* Fake initialization required by compiler */
	70	static DEFINE_PER_CPU(struct tasklet_struct, rcu_tasklet) = {NULL};
21a1ea9e DS	71	static int blimit = 10;
	72	static int qhimark = 10000;
	73	static int qlowmark = 100;
	74	#ifdef CONFIG_SMP
	75	static int rsinterval = 1000;
	76	#endif
	77
	78	static atomic_t rcu_barrier_cpu_count;
9331b315	79	static DEFINE_MUTEX(rcu_barrier_mutex);
21a1ea9e DS	80	static struct completion rcu_barrier_completion;
	81
	82	#ifdef CONFIG_SMP
	83	static void force_quiescent_state(struct rcu_data *rdp,
	84	struct rcu_ctrlblk *rcp)
	85	{
	86	int cpu;
	87	cpumask_t cpumask;
	88	set_need_resched();
	89	if (unlikely(rdp->qlen - rdp->last_rs_qlen > rsinterval)) {
	90	rdp->last_rs_qlen = rdp->qlen;
	91	/*
	92	* Don't send IPI to itself. With irqs disabled,
	93	* rdp->cpu is the current cpu.
	94	*/
	95	cpumask = rcp->cpumask;
	96	cpu_clear(rdp->cpu, cpumask);
	97	for_each_cpu_mask(cpu, cpumask)
	98	smp_send_reschedule(cpu);
	99	}
	100	}
	101	#else
	102	static inline void force_quiescent_state(struct rcu_data *rdp,
	103	struct rcu_ctrlblk *rcp)
	104	{
	105	set_need_resched();
	106	}
	107	#endif
1da177e4 LT	108
	109	/**
	110	* call_rcu - Queue an RCU callback for invocation after a grace period.
	111	* @head: structure to be used for queueing the RCU updates.
	112	* @func: actual update function to be invoked after the grace period
	113	*
	114	* The update function will be invoked some time after a full grace
	115	* period elapses, in other words after all currently executing RCU
	116	* read-side critical sections have completed. RCU read-side critical
	117	* sections are delimited by rcu_read_lock() and rcu_read_unlock(),
	118	* and may be nested.
	119	*/
	120	void fastcall call_rcu(struct rcu_head *head,
	121	void (func)(struct rcu_head rcu))
	122	{
	123	unsigned long flags;
	124	struct rcu_data *rdp;
	125
	126	head->func = func;
	127	head->next = NULL;
	128	local_irq_save(flags);
	129	rdp = &__get_cpu_var(rcu_data);
	130	*rdp->nxttail = head;
	131	rdp->nxttail = &head->next;
21a1ea9e DS	132	if (unlikely(++rdp->qlen > qhimark)) {
	133	rdp->blimit = INT_MAX;
	134	force_quiescent_state(rdp, &rcu_ctrlblk);
	135	}
1da177e4 LT	136	local_irq_restore(flags);
	137	}
	138
	139	/**
	140	* call_rcu_bh - Queue an RCU for invocation after a quicker grace period.
	141	* @head: structure to be used for queueing the RCU updates.
	142	* @func: actual update function to be invoked after the grace period
	143	*
	144	* The update function will be invoked some time after a full grace
	145	* period elapses, in other words after all currently executing RCU
	146	* read-side critical sections have completed. call_rcu_bh() assumes
	147	* that the read-side critical sections end on completion of a softirq
	148	* handler. This means that read-side critical sections in process
	149	* context must not be interrupted by softirqs. This interface is to be
	150	* used when most of the read-side critical sections are in softirq context.
	151	* RCU read-side critical sections are delimited by rcu_read_lock() and
	152	* rcu_read_unlock(), * if in interrupt context or rcu_read_lock_bh()
	153	* and rcu_read_unlock_bh(), if in process context. These may be nested.
	154	*/
	155	void fastcall call_rcu_bh(struct rcu_head *head,
	156	void (func)(struct rcu_head rcu))
	157	{
	158	unsigned long flags;
	159	struct rcu_data *rdp;
	160
	161	head->func = func;
	162	head->next = NULL;
	163	local_irq_save(flags);
	164	rdp = &__get_cpu_var(rcu_bh_data);
	165	*rdp->nxttail = head;
	166	rdp->nxttail = &head->next;
21a1ea9e DS	167
	168	if (unlikely(++rdp->qlen > qhimark)) {
	169	rdp->blimit = INT_MAX;
	170	force_quiescent_state(rdp, &rcu_bh_ctrlblk);
	171	}
	172
1da177e4 LT	173	local_irq_restore(flags);
	174	}
	175
a241ec65 PM	176	/*
	177	* Return the number of RCU batches processed thus far. Useful
	178	* for debug and statistics.
	179	*/
	180	long rcu_batches_completed(void)
	181	{
	182	return rcu_ctrlblk.completed;
	183	}
	184
ab4720ec DS	185	static void rcu_barrier_callback(struct rcu_head *notused)
	186	{
	187	if (atomic_dec_and_test(&rcu_barrier_cpu_count))
	188	complete(&rcu_barrier_completion);
	189	}
	190
	191	/*
	192	* Called with preemption disabled, and from cross-cpu IRQ context.
	193	*/
	194	static void rcu_barrier_func(void *notused)
	195	{
	196	int cpu = smp_processor_id();
	197	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
	198	struct rcu_head *head;
	199
	200	head = &rdp->barrier;
	201	atomic_inc(&rcu_barrier_cpu_count);
	202	call_rcu(head, rcu_barrier_callback);
	203	}
	204
	205	/**
	206	* rcu_barrier - Wait until all the in-flight RCUs are complete.
	207	*/
	208	void rcu_barrier(void)
	209	{
	210	BUG_ON(in_interrupt());
9331b315 IM	211	/* Take cpucontrol mutex to protect against CPU hotplug */
9331b315 IM	212	mutex_lock(&rcu_barrier_mutex);
ab4720ec DS	213	init_completion(&rcu_barrier_completion);
	214	atomic_set(&rcu_barrier_cpu_count, 0);
	215	on_each_cpu(rcu_barrier_func, NULL, 0, 1);
	216	wait_for_completion(&rcu_barrier_completion);
9331b315	217	mutex_unlock(&rcu_barrier_mutex);
ab4720ec DS	218	}
	219	EXPORT_SYMBOL_GPL(rcu_barrier);
	220
1da177e4 LT	221	/*
	222	* Invoke the completed RCU callbacks. They are expected to be in
	223	* a per-cpu list.
	224	*/
	225	static void rcu_do_batch(struct rcu_data *rdp)
	226	{
	227	struct rcu_head next, list;
	228	int count = 0;
	229
	230	list = rdp->donelist;
	231	while (list) {
	232	next = rdp->donelist = list->next;
	233	list->func(list);
	234	list = next;
21a1ea9e DS	235	rdp->qlen--;
21a1ea9e DS	236	if (++count >= rdp->blimit)
1da177e4 LT	237	break;
1da177e4 LT	238	}
21a1ea9e DS	239	if (rdp->blimit == INT_MAX && rdp->qlen <= qlowmark)
21a1ea9e DS	240	rdp->blimit = blimit;
1da177e4 LT	241	if (!rdp->donelist)
	242	rdp->donetail = &rdp->donelist;
	243	else
	244	tasklet_schedule(&per_cpu(rcu_tasklet, rdp->cpu));
	245	}
	246
	247	/*
	248	* Grace period handling:
	249	* The grace period handling consists out of two steps:
	250	* - A new grace period is started.
	251	* This is done by rcu_start_batch. The start is not broadcasted to
	252	* all cpus, they must pick this up by comparing rcp->cur with
	253	* rdp->quiescbatch. All cpus are recorded in the
69a0b315	254	* rcu_ctrlblk.cpumask bitmap.
1da177e4 LT	255	* - All cpus must go through a quiescent state.
	256	* Since the start of the grace period is not broadcasted, at least two
	257	* calls to rcu_check_quiescent_state are required:
	258	* The first call just notices that a new grace period is running. The
	259	* following calls check if there was a quiescent state since the beginning
69a0b315	260	* of the grace period. If so, it updates rcu_ctrlblk.cpumask. If
1da177e4 LT	261	* the bitmap is empty, then the grace period is completed.
	262	* rcu_check_quiescent_state calls rcu_start_batch(0) to start the next grace
	263	* period (if necessary).
	264	*/
	265	/*
	266	* Register a new batch of callbacks, and start it up if there is currently no
	267	* active batch and the batch to be registered has not already occurred.
69a0b315	268	* Caller must hold rcu_ctrlblk.lock.
1da177e4	269	*/
69a0b315	270	static void rcu_start_batch(struct rcu_ctrlblk *rcp)
1da177e4	271	{
1da177e4 LT	272	if (rcp->next_pending &&
1da177e4 LT	273	rcp->completed == rcp->cur) {
1da177e4	274	rcp->next_pending = 0;
c3f59023 SV	275	/*
	276	* next_pending == 0 must be visible in
	277	* __rcu_process_callbacks() before it can see new value of cur.
1da177e4 LT	278	*/
	279	smp_wmb();
	280	rcp->cur++;
c3f59023 SV	281
	282	/*
	283	* Accessing nohz_cpu_mask before incrementing rcp->cur needs a
	284	* Barrier Otherwise it can cause tickless idle CPUs to be
69a0b315	285	* included in rcp->cpumask, which will extend graceperiods
c3f59023 SV	286	* unnecessarily.
	287	*/
	288	smp_mb();
69a0b315	289	cpus_andnot(rcp->cpumask, cpu_online_map, nohz_cpu_mask);
c3f59023	290
1da177e4 LT	291	}
	292	}
	293
	294	/*
	295	* cpu went through a quiescent state since the beginning of the grace period.
	296	* Clear it from the cpu mask and complete the grace period if it was the last
	297	* cpu. Start another grace period if someone has further entries pending
	298	*/
69a0b315	299	static void cpu_quiet(int cpu, struct rcu_ctrlblk *rcp)
1da177e4	300	{
69a0b315 ON	301	cpu_clear(cpu, rcp->cpumask);
69a0b315 ON	302	if (cpus_empty(rcp->cpumask)) {
1da177e4 LT	303	/* batch completed ! */
1da177e4 LT	304	rcp->completed = rcp->cur;
69a0b315	305	rcu_start_batch(rcp);
1da177e4 LT	306	}
	307	}
	308
	309	/*
	310	* Check if the cpu has gone through a quiescent state (say context
	311	* switch). If so and if it already hasn't done so in this RCU
	312	* quiescent cycle, then indicate that it has done so.
	313	*/
	314	static void rcu_check_quiescent_state(struct rcu_ctrlblk *rcp,
69a0b315	315	struct rcu_data *rdp)
1da177e4 LT	316	{
	317	if (rdp->quiescbatch != rcp->cur) {
	318	/* start new grace period: */
	319	rdp->qs_pending = 1;
	320	rdp->passed_quiesc = 0;
	321	rdp->quiescbatch = rcp->cur;
	322	return;
	323	}
	324
	325	/* Grace period already completed for this cpu?
	326	* qs_pending is checked instead of the actual bitmap to avoid
	327	* cacheline trashing.
	328	*/
	329	if (!rdp->qs_pending)
	330	return;
	331
	332	/*
	333	* Was there a quiescent state since the beginning of the grace
	334	* period? If no, then exit and wait for the next call.
	335	*/
	336	if (!rdp->passed_quiesc)
	337	return;
	338	rdp->qs_pending = 0;
	339
69a0b315	340	spin_lock(&rcp->lock);
1da177e4 LT	341	/*
	342	* rdp->quiescbatch/rcp->cur and the cpu bitmap can come out of sync
	343	* during cpu startup. Ignore the quiescent state.
	344	*/
	345	if (likely(rdp->quiescbatch == rcp->cur))
69a0b315	346	cpu_quiet(rdp->cpu, rcp);
1da177e4	347
69a0b315	348	spin_unlock(&rcp->lock);
1da177e4 LT	349	}
	350
	351
	352	#ifdef CONFIG_HOTPLUG_CPU
	353
	354	/* warning! helper for rcu_offline_cpu. do not use elsewhere without reviewing
	355	* locking requirements, the list it's pulling from has to belong to a cpu
	356	* which is dead and hence not processing interrupts.
	357	*/
	358	static void rcu_move_batch(struct rcu_data this_rdp, struct rcu_head list,
	359	struct rcu_head **tail)
	360	{
	361	local_irq_disable();
	362	*this_rdp->nxttail = list;
	363	if (list)
	364	this_rdp->nxttail = tail;
	365	local_irq_enable();
	366	}
	367
	368	static void __rcu_offline_cpu(struct rcu_data *this_rdp,
69a0b315	369	struct rcu_ctrlblk rcp, struct rcu_data rdp)
1da177e4 LT	370	{
	371	/* if the cpu going offline owns the grace period
	372	* we can block indefinitely waiting for it, so flush
	373	* it here
	374	*/
69a0b315	375	spin_lock_bh(&rcp->lock);
1da177e4	376	if (rcp->cur != rcp->completed)
69a0b315 ON	377	cpu_quiet(rdp->cpu, rcp);
69a0b315 ON	378	spin_unlock_bh(&rcp->lock);
1da177e4 LT	379	rcu_move_batch(this_rdp, rdp->curlist, rdp->curtail);
1da177e4 LT	380	rcu_move_batch(this_rdp, rdp->nxtlist, rdp->nxttail);
a9c82815	381	rcu_move_batch(this_rdp, rdp->donelist, rdp->donetail);
1da177e4	382	}
a9c82815	383
1da177e4 LT	384	static void rcu_offline_cpu(int cpu)
	385	{
	386	struct rcu_data *this_rdp = &get_cpu_var(rcu_data);
	387	struct rcu_data *this_bh_rdp = &get_cpu_var(rcu_bh_data);
	388
69a0b315	389	__rcu_offline_cpu(this_rdp, &rcu_ctrlblk,
1da177e4	390	&per_cpu(rcu_data, cpu));
69a0b315	391	__rcu_offline_cpu(this_bh_rdp, &rcu_bh_ctrlblk,
1da177e4 LT	392	&per_cpu(rcu_bh_data, cpu));
	393	put_cpu_var(rcu_data);
	394	put_cpu_var(rcu_bh_data);
	395	tasklet_kill_immediate(&per_cpu(rcu_tasklet, cpu), cpu);
	396	}
	397
	398	#else
	399
	400	static void rcu_offline_cpu(int cpu)
	401	{
	402	}
	403
	404	#endif
	405
	406	/*
	407	* This does the RCU processing work from tasklet context.
	408	*/
	409	static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp,
69a0b315	410	struct rcu_data *rdp)
1da177e4 LT	411	{
	412	if (rdp->curlist && !rcu_batch_before(rcp->completed, rdp->batch)) {
	413	*rdp->donetail = rdp->curlist;
	414	rdp->donetail = rdp->curtail;
	415	rdp->curlist = NULL;
	416	rdp->curtail = &rdp->curlist;
	417	}
	418
1da177e4	419	if (rdp->nxtlist && !rdp->curlist) {
caa9ee77	420	local_irq_disable();
1da177e4 LT	421	rdp->curlist = rdp->nxtlist;
	422	rdp->curtail = rdp->nxttail;
	423	rdp->nxtlist = NULL;
	424	rdp->nxttail = &rdp->nxtlist;
	425	local_irq_enable();
	426
	427	/*
	428	* start the next batch of callbacks
	429	*/
	430
	431	/* determine batch number */
	432	rdp->batch = rcp->cur + 1;
	433	/* see the comment and corresponding wmb() in
	434	* the rcu_start_batch()
	435	*/
	436	smp_rmb();
	437
	438	if (!rcp->next_pending) {
	439	/* and start it/schedule start if it's a new batch */
69a0b315	440	spin_lock(&rcp->lock);
dbc1651f	441	rcp->next_pending = 1;
69a0b315 ON	442	rcu_start_batch(rcp);
69a0b315 ON	443	spin_unlock(&rcp->lock);
1da177e4	444	}
1da177e4	445	}
caa9ee77	446
69a0b315	447	rcu_check_quiescent_state(rcp, rdp);
1da177e4 LT	448	if (rdp->donelist)
	449	rcu_do_batch(rdp);
	450	}
	451
	452	static void rcu_process_callbacks(unsigned long unused)
	453	{
69a0b315 ON	454	__rcu_process_callbacks(&rcu_ctrlblk, &__get_cpu_var(rcu_data));
69a0b315 ON	455	__rcu_process_callbacks(&rcu_bh_ctrlblk, &__get_cpu_var(rcu_bh_data));
1da177e4 LT	456	}
1da177e4 LT	457
67751777 ON	458	static int __rcu_pending(struct rcu_ctrlblk rcp, struct rcu_data rdp)
	459	{
	460	/* This cpu has pending rcu entries and the grace period
	461	* for them has completed.
	462	*/
	463	if (rdp->curlist && !rcu_batch_before(rcp->completed, rdp->batch))
	464	return 1;
	465
	466	/* This cpu has no pending entries, but there are new entries */
	467	if (!rdp->curlist && rdp->nxtlist)
	468	return 1;
	469
	470	/* This cpu has finished callbacks to invoke */
	471	if (rdp->donelist)
	472	return 1;
	473
	474	/* The rcu core waits for a quiescent state from the cpu */
	475	if (rdp->quiescbatch != rcp->cur \|\| rdp->qs_pending)
	476	return 1;
	477
	478	/* nothing to do */
	479	return 0;
	480	}
	481
986733e0 HC	482	/*
	483	* Check to see if there is any immediate RCU-related work to be done
	484	* by the current CPU, returning 1 if so. This function is part of the
	485	* RCU implementation; it is -not- an exported member of the RCU API.
	486	*/
67751777 ON	487	int rcu_pending(int cpu)
	488	{
	489	return __rcu_pending(&rcu_ctrlblk, &per_cpu(rcu_data, cpu)) \|\|
	490	__rcu_pending(&rcu_bh_ctrlblk, &per_cpu(rcu_bh_data, cpu));
	491	}
	492
986733e0 HC	493	/*
	494	* Check to see if any future RCU-related work will need to be done
	495	* by the current CPU, even if none need be done immediately, returning
	496	* 1 if so. This function is part of the RCU implementation; it is -not-
	497	* an exported member of the RCU API.
	498	*/
	499	int rcu_needs_cpu(int cpu)
	500	{
	501	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
	502	struct rcu_data *rdp_bh = &per_cpu(rcu_bh_data, cpu);
	503
	504	return (!!rdp->curlist \|\| !!rdp_bh->curlist \|\| rcu_pending(cpu));
	505	}
	506
1da177e4 LT	507	void rcu_check_callbacks(int cpu, int user)
	508	{
	509	if (user \|\|
	510	(idle_cpu(cpu) && !in_softirq() &&
	511	hardirq_count() <= (1 << HARDIRQ_SHIFT))) {
	512	rcu_qsctr_inc(cpu);
	513	rcu_bh_qsctr_inc(cpu);
	514	} else if (!in_softirq())
	515	rcu_bh_qsctr_inc(cpu);
	516	tasklet_schedule(&per_cpu(rcu_tasklet, cpu));
	517	}
	518
	519	static void rcu_init_percpu_data(int cpu, struct rcu_ctrlblk *rcp,
	520	struct rcu_data *rdp)
	521	{
	522	memset(rdp, 0, sizeof(*rdp));
	523	rdp->curtail = &rdp->curlist;
	524	rdp->nxttail = &rdp->nxtlist;
	525	rdp->donetail = &rdp->donelist;
	526	rdp->quiescbatch = rcp->completed;
	527	rdp->qs_pending = 0;
	528	rdp->cpu = cpu;
21a1ea9e	529	rdp->blimit = blimit;
1da177e4 LT	530	}
	531
	532	static void __devinit rcu_online_cpu(int cpu)
	533	{
	534	struct rcu_data *rdp = &per_cpu(rcu_data, cpu);
	535	struct rcu_data *bh_rdp = &per_cpu(rcu_bh_data, cpu);
	536
	537	rcu_init_percpu_data(cpu, &rcu_ctrlblk, rdp);
	538	rcu_init_percpu_data(cpu, &rcu_bh_ctrlblk, bh_rdp);
	539	tasklet_init(&per_cpu(rcu_tasklet, cpu), rcu_process_callbacks, 0UL);
	540	}
	541
83d722f7	542	static int rcu_cpu_notify(struct notifier_block *self,
1da177e4 LT	543	unsigned long action, void *hcpu)
	544	{
	545	long cpu = (long)hcpu;
	546	switch (action) {
	547	case CPU_UP_PREPARE:
	548	rcu_online_cpu(cpu);
	549	break;
	550	case CPU_DEAD:
	551	rcu_offline_cpu(cpu);
	552	break;
	553	default:
	554	break;
	555	}
	556	return NOTIFY_OK;
	557	}
	558
649bbaa4	559	static struct notifier_block rcu_nb = {
1da177e4 LT	560	.notifier_call = rcu_cpu_notify,
	561	};
	562
	563	/*
	564	* Initializes rcu mechanism. Assumed to be called early.
	565	* That is before local timer(SMP) or jiffie timer (uniproc) is setup.
	566	* Note that rcu_qsctr and friends are implicitly
	567	* initialized due to the choice of ``0'' for RCU_CTR_INVALID.
	568	*/
	569	void __init rcu_init(void)
	570	{
	571	rcu_cpu_notify(&rcu_nb, CPU_UP_PREPARE,
	572	(void *)(long)smp_processor_id());
	573	/* Register notifier for non-boot CPUs */
	574	register_cpu_notifier(&rcu_nb);
	575	}
	576
	577	struct rcu_synchronize {
	578	struct rcu_head head;
	579	struct completion completion;
	580	};
	581
	582	/* Because of FASTCALL declaration of complete, we use this wrapper */
	583	static void wakeme_after_rcu(struct rcu_head *head)
	584	{
	585	struct rcu_synchronize *rcu;
	586
	587	rcu = container_of(head, struct rcu_synchronize, head);
	588	complete(&rcu->completion);
	589	}
	590
	591	/**
9b06e818	592	* synchronize_rcu - wait until a grace period has elapsed.
1da177e4 LT	593	*
	594	* Control will return to the caller some time after a full grace
	595	* period has elapsed, in other words after all currently executing RCU
	596	* read-side critical sections have completed. RCU read-side critical
	597	* sections are delimited by rcu_read_lock() and rcu_read_unlock(),
	598	* and may be nested.
9b06e818 PM	599	*
	600	* If your read-side code is not protected by rcu_read_lock(), do -not-
	601	* use synchronize_rcu().
1da177e4	602	*/
9b06e818	603	void synchronize_rcu(void)
1da177e4 LT	604	{
	605	struct rcu_synchronize rcu;
	606
	607	init_completion(&rcu.completion);
	608	/* Will wake me after RCU finished */
	609	call_rcu(&rcu.head, wakeme_after_rcu);
	610
	611	/* Wait for it */
	612	wait_for_completion(&rcu.completion);
	613	}
	614
9b06e818 PM	615	/*
	616	* Deprecated, use synchronize_rcu() or synchronize_sched() instead.
	617	*/
	618	void synchronize_kernel(void)
	619	{
	620	synchronize_rcu();
	621	}
	622
21a1ea9e DS	623	module_param(blimit, int, 0);
	624	module_param(qhimark, int, 0);
	625	module_param(qlowmark, int, 0);
	626	#ifdef CONFIG_SMP
	627	module_param(rsinterval, int, 0);
	628	#endif
a241ec65	629	EXPORT_SYMBOL_GPL(rcu_batches_completed);
01ca70dc GKH	630	EXPORT_SYMBOL_GPL_FUTURE(call_rcu); /* WARNING: GPL-only in April 2006. */
01ca70dc GKH	631	EXPORT_SYMBOL_GPL_FUTURE(call_rcu_bh); /* WARNING: GPL-only in April 2006. */
9b06e818	632	EXPORT_SYMBOL_GPL(synchronize_rcu);
01ca70dc	633	EXPORT_SYMBOL_GPL_FUTURE(synchronize_kernel); /* WARNING: GPL-only in April 2006. */