1/*
  2 * Read-Copy Update mechanism for mutual exclusion, the Bloatwatch edition.
  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 IBM Corporation, 2008
 19 *
 20 * Author: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
 21 *
 22 * For detailed explanation of Read-Copy Update mechanism see -
 23 *		Documentation/RCU
 24 */
 25#include <linux/moduleparam.h>
 26#include <linux/completion.h>
 27#include <linux/interrupt.h>
 28#include <linux/notifier.h>
 29#include <linux/rcupdate.h>
 30#include <linux/kernel.h>
 31#include <linux/module.h>
 32#include <linux/mutex.h>
 33#include <linux/sched.h>
 34#include <linux/types.h>
 35#include <linux/init.h>
 36#include <linux/time.h>
 37#include <linux/cpu.h>
 38#include <linux/prefetch.h>
 39
 40/* Controls for rcu_kthread() kthread, replacing RCU_SOFTIRQ used previously. */
 41static struct task_struct *rcu_kthread_task;
 42static DECLARE_WAIT_QUEUE_HEAD(rcu_kthread_wq);
 43static unsigned long have_rcu_kthread_work;
 44
 45/* Forward declarations for rcutiny_plugin.h. */
 46struct rcu_ctrlblk;
 47static void invoke_rcu_kthread(void);
 48static void rcu_process_callbacks(struct rcu_ctrlblk *rcp);
 49static int rcu_kthread(void *arg);
 50static void __call_rcu(struct rcu_head *head,
 51		       void (*func)(struct rcu_head *rcu),
 52		       struct rcu_ctrlblk *rcp);
 53
 54#include "rcutiny_plugin.h"
 55
 56#ifdef CONFIG_NO_HZ
 57
 58static long rcu_dynticks_nesting = 1;
 59
 60/*
 61 * Enter dynticks-idle mode, which is an extended quiescent state
 62 * if we have fully entered that mode (i.e., if the new value of
 63 * dynticks_nesting is zero).
 64 */
 65void rcu_enter_nohz(void)
 66{
 67	if (--rcu_dynticks_nesting == 0)
 68		rcu_sched_qs(0); /* implies rcu_bh_qsctr_inc(0) */
 69}
 70
 71/*
 72 * Exit dynticks-idle mode, so that we are no longer in an extended
 73 * quiescent state.
 74 */
 75void rcu_exit_nohz(void)
 76{
 77	rcu_dynticks_nesting++;
 78}
 79
 80#endif /* #ifdef CONFIG_NO_HZ */
 81
 82/*
 83 * Helper function for rcu_sched_qs() and rcu_bh_qs().
 84 * Also irqs are disabled to avoid confusion due to interrupt handlers
 85 * invoking call_rcu().
 86 */
 87static int rcu_qsctr_help(struct rcu_ctrlblk *rcp)
 88{
 89	if (rcp->rcucblist != NULL &&
 90	    rcp->donetail != rcp->curtail) {
 91		rcp->donetail = rcp->curtail;
 92		return 1;
 93	}
 94
 95	return 0;
 96}
 97
 98/*
 99 * Wake up rcu_kthread() to process callbacks now eligible for invocation
100 * or to boost readers.
101 */
102static void invoke_rcu_kthread(void)
103{
104	have_rcu_kthread_work = 1;
105	wake_up(&rcu_kthread_wq);
106}
107
108/*
109 * Record an rcu quiescent state.  And an rcu_bh quiescent state while we
110 * are at it, given that any rcu quiescent state is also an rcu_bh
111 * quiescent state.  Use "+" instead of "||" to defeat short circuiting.
112 */
113void rcu_sched_qs(int cpu)
114{
115	unsigned long flags;
116
117	local_irq_save(flags);
118	if (rcu_qsctr_help(&rcu_sched_ctrlblk) +
119	    rcu_qsctr_help(&rcu_bh_ctrlblk))
120		invoke_rcu_kthread();
121	local_irq_restore(flags);
122}
123
124/*
125 * Record an rcu_bh quiescent state.
126 */
127void rcu_bh_qs(int cpu)
128{
129	unsigned long flags;
130
131	local_irq_save(flags);
132	if (rcu_qsctr_help(&rcu_bh_ctrlblk))
133		invoke_rcu_kthread();
134	local_irq_restore(flags);
135}
136
137/*
138 * Check to see if the scheduling-clock interrupt came from an extended
139 * quiescent state, and, if so, tell RCU about it.
140 */
141void rcu_check_callbacks(int cpu, int user)
142{
143	if (user ||
144	    (idle_cpu(cpu) &&
145	     !in_softirq() &&
146	     hardirq_count() <= (1 << HARDIRQ_SHIFT)))
147		rcu_sched_qs(cpu);
148	else if (!in_softirq())
149		rcu_bh_qs(cpu);
150	rcu_preempt_check_callbacks();
151}
152
153/*
154 * Invoke the RCU callbacks on the specified rcu_ctrlkblk structure
155 * whose grace period has elapsed.
156 */
157static void rcu_process_callbacks(struct rcu_ctrlblk *rcp)
158{
159	struct rcu_head *next, *list;
160	unsigned long flags;
161	RCU_TRACE(int cb_count = 0);
162
163	/* If no RCU callbacks ready to invoke, just return. */
164	if (&rcp->rcucblist == rcp->donetail)
165		return;
166
167	/* Move the ready-to-invoke callbacks to a local list. */
168	local_irq_save(flags);
169	list = rcp->rcucblist;
170	rcp->rcucblist = *rcp->donetail;
171	*rcp->donetail = NULL;
172	if (rcp->curtail == rcp->donetail)
173		rcp->curtail = &rcp->rcucblist;
174	rcu_preempt_remove_callbacks(rcp);
175	rcp->donetail = &rcp->rcucblist;
176	local_irq_restore(flags);
177
178	/* Invoke the callbacks on the local list. */
179	while (list) {
180		next = list->next;
181		prefetch(next);
182		debug_rcu_head_unqueue(list);
183		local_bh_disable();
184		__rcu_reclaim(list);
185		local_bh_enable();
186		list = next;
187		RCU_TRACE(cb_count++);
188	}
189	RCU_TRACE(rcu_trace_sub_qlen(rcp, cb_count));
190}
191
192/*
193 * This kthread invokes RCU callbacks whose grace periods have
194 * elapsed.  It is awakened as needed, and takes the place of the
195 * RCU_SOFTIRQ that was used previously for this purpose.
196 * This is a kthread, but it is never stopped, at least not until
197 * the system goes down.
198 */
199static int rcu_kthread(void *arg)
200{
201	unsigned long work;
202	unsigned long morework;
203	unsigned long flags;
204
205	for (;;) {
206		wait_event_interruptible(rcu_kthread_wq,
207					 have_rcu_kthread_work != 0);
208		morework = rcu_boost();
209		local_irq_save(flags);
210		work = have_rcu_kthread_work;
211		have_rcu_kthread_work = morework;
212		local_irq_restore(flags);
213		if (work) {
214			rcu_process_callbacks(&rcu_sched_ctrlblk);
215			rcu_process_callbacks(&rcu_bh_ctrlblk);
216			rcu_preempt_process_callbacks();
217		}
218		schedule_timeout_interruptible(1); /* Leave CPU for others. */
219	}
220
221	return 0;  /* Not reached, but needed to shut gcc up. */
222}
223
224/*
225 * Wait for a grace period to elapse.  But it is illegal to invoke
226 * synchronize_sched() from within an RCU read-side critical section.
227 * Therefore, any legal call to synchronize_sched() is a quiescent
228 * state, and so on a UP system, synchronize_sched() need do nothing.
229 * Ditto for synchronize_rcu_bh().  (But Lai Jiangshan points out the
230 * benefits of doing might_sleep() to reduce latency.)
231 *
232 * Cool, huh?  (Due to Josh Triplett.)
233 *
234 * But we want to make this a static inline later.  The cond_resched()
235 * currently makes this problematic.
236 */
237void synchronize_sched(void)
238{
239	cond_resched();
240}
241EXPORT_SYMBOL_GPL(synchronize_sched);
242
243/*
244 * Helper function for call_rcu() and call_rcu_bh().
245 */
246static void __call_rcu(struct rcu_head *head,
247		       void (*func)(struct rcu_head *rcu),
248		       struct rcu_ctrlblk *rcp)
249{
250	unsigned long flags;
251
252	debug_rcu_head_queue(head);
253	head->func = func;
254	head->next = NULL;
255
256	local_irq_save(flags);
257	*rcp->curtail = head;
258	rcp->curtail = &head->next;
259	RCU_TRACE(rcp->qlen++);
260	local_irq_restore(flags);
261}
262
263/*
264 * Post an RCU callback to be invoked after the end of an RCU-sched grace
265 * period.  But since we have but one CPU, that would be after any
266 * quiescent state.
267 */
268void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
269{
270	__call_rcu(head, func, &rcu_sched_ctrlblk);
271}
272EXPORT_SYMBOL_GPL(call_rcu_sched);
273
274/*
275 * Post an RCU bottom-half callback to be invoked after any subsequent
276 * quiescent state.
277 */
278void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
279{
280	__call_rcu(head, func, &rcu_bh_ctrlblk);
281}
282EXPORT_SYMBOL_GPL(call_rcu_bh);
283
284void rcu_barrier_bh(void)
285{
286	struct rcu_synchronize rcu;
287
288	init_rcu_head_on_stack(&rcu.head);
289	init_completion(&rcu.completion);
290	/* Will wake me after RCU finished. */
291	call_rcu_bh(&rcu.head, wakeme_after_rcu);
292	/* Wait for it. */
293	wait_for_completion(&rcu.completion);
294	destroy_rcu_head_on_stack(&rcu.head);
295}
296EXPORT_SYMBOL_GPL(rcu_barrier_bh);
297
298void rcu_barrier_sched(void)
299{
300	struct rcu_synchronize rcu;
301
302	init_rcu_head_on_stack(&rcu.head);
303	init_completion(&rcu.completion);
304	/* Will wake me after RCU finished. */
305	call_rcu_sched(&rcu.head, wakeme_after_rcu);
306	/* Wait for it. */
307	wait_for_completion(&rcu.completion);
308	destroy_rcu_head_on_stack(&rcu.head);
309}
310EXPORT_SYMBOL_GPL(rcu_barrier_sched);
311
312/*
313 * Spawn the kthread that invokes RCU callbacks.
314 */
315static int __init rcu_spawn_kthreads(void)
316{
317	struct sched_param sp;
318
319	rcu_kthread_task = kthread_run(rcu_kthread, NULL, "rcu_kthread");
320	sp.sched_priority = RCU_BOOST_PRIO;
321	sched_setscheduler_nocheck(rcu_kthread_task, SCHED_FIFO, &sp);
322	return 0;
323}
324early_initcall(rcu_spawn_kthreads);