percpu-refcount.h 10.3 KB
Newer Older
1
/* SPDX-License-Identifier: GPL-2.0 */
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/*
 * Percpu refcounts:
 * (C) 2012 Google, Inc.
 * Author: Kent Overstreet <koverstreet@google.com>
 *
 * This implements a refcount with similar semantics to atomic_t - atomic_inc(),
 * atomic_dec_and_test() - but percpu.
 *
 * There's one important difference between percpu refs and normal atomic_t
 * refcounts; you have to keep track of your initial refcount, and then when you
 * start shutting down you call percpu_ref_kill() _before_ dropping the initial
 * refcount.
 *
 * The refcount will have a range of 0 to ((1U << 31) - 1), i.e. one bit less
 * than an atomic_t - this is because of the way shutdown works, see
17
 * percpu_ref_kill()/PERCPU_COUNT_BIAS.
18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
 *
 * Before you call percpu_ref_kill(), percpu_ref_put() does not check for the
 * refcount hitting 0 - it can't, if it was in percpu mode. percpu_ref_kill()
 * puts the ref back in single atomic_t mode, collecting the per cpu refs and
 * issuing the appropriate barriers, and then marks the ref as shutting down so
 * that percpu_ref_put() will check for the ref hitting 0.  After it returns,
 * it's safe to drop the initial ref.
 *
 * USAGE:
 *
 * See fs/aio.c for some example usage; it's used there for struct kioctx, which
 * is created when userspaces calls io_setup(), and destroyed when userspace
 * calls io_destroy() or the process exits.
 *
 * In the aio code, kill_ioctx() is called when we wish to destroy a kioctx; it
33 34 35 36 37 38 39 40
 * removes the kioctx from the proccess's table of kioctxs and kills percpu_ref.
 * After that, there can't be any new users of the kioctx (from lookup_ioctx())
 * and it's then safe to drop the initial ref with percpu_ref_put().
 *
 * Note that the free path, free_ioctx(), needs to go through explicit call_rcu()
 * to synchronize with RCU protected lookup_ioctx().  percpu_ref operations don't
 * imply RCU grace periods of any kind and if a user wants to combine percpu_ref
 * with RCU protection, it must be done explicitly.
41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56
 *
 * Code that does a two stage shutdown like this often needs some kind of
 * explicit synchronization to ensure the initial refcount can only be dropped
 * once - percpu_ref_kill() does this for you, it returns true once and false if
 * someone else already called it. The aio code uses it this way, but it's not
 * necessary if the code has some other mechanism to synchronize teardown.
 * around.
 */

#ifndef _LINUX_PERCPU_REFCOUNT_H
#define _LINUX_PERCPU_REFCOUNT_H

#include <linux/atomic.h>
#include <linux/kernel.h>
#include <linux/percpu.h>
#include <linux/rcupdate.h>
57
#include <linux/gfp.h>
58 59

struct percpu_ref;
T
Tejun Heo 已提交
60
typedef void (percpu_ref_func_t)(struct percpu_ref *);
61

62 63 64
/* flags set in the lower bits of percpu_ref->percpu_count_ptr */
enum {
	__PERCPU_REF_ATOMIC	= 1LU << 0,	/* operating in atomic mode */
T
Tejun Heo 已提交
65 66 67 68
	__PERCPU_REF_DEAD	= 1LU << 1,	/* (being) killed */
	__PERCPU_REF_ATOMIC_DEAD = __PERCPU_REF_ATOMIC | __PERCPU_REF_DEAD,

	__PERCPU_REF_FLAG_BITS	= 2,
69 70
};

71 72 73 74
/* @flags for percpu_ref_init() */
enum {
	/*
	 * Start w/ ref == 1 in atomic mode.  Can be switched to percpu
75 76 77
	 * operation using percpu_ref_switch_to_percpu().  If initialized
	 * with this flag, the ref will stay in atomic mode until
	 * percpu_ref_switch_to_percpu() is invoked on it.
78 79 80 81 82 83 84 85 86 87
	 */
	PERCPU_REF_INIT_ATOMIC	= 1 << 0,

	/*
	 * Start dead w/ ref == 0 in atomic mode.  Must be revived with
	 * percpu_ref_reinit() before used.  Implies INIT_ATOMIC.
	 */
	PERCPU_REF_INIT_DEAD	= 1 << 1,
};

88
struct percpu_ref {
89
	atomic_long_t		count;
90 91
	/*
	 * The low bit of the pointer indicates whether the ref is in percpu
92
	 * mode; if set, then get/put will manipulate the atomic_t.
93
	 */
94
	unsigned long		percpu_count_ptr;
T
Tejun Heo 已提交
95
	percpu_ref_func_t	*release;
96
	percpu_ref_func_t	*confirm_switch;
97
	bool			force_atomic:1;
98 99 100
	struct rcu_head		rcu;
};

101
int __must_check percpu_ref_init(struct percpu_ref *ref,
102 103
				 percpu_ref_func_t *release, unsigned int flags,
				 gfp_t gfp);
104
void percpu_ref_exit(struct percpu_ref *ref);
105 106
void percpu_ref_switch_to_atomic(struct percpu_ref *ref,
				 percpu_ref_func_t *confirm_switch);
107
void percpu_ref_switch_to_atomic_sync(struct percpu_ref *ref);
108
void percpu_ref_switch_to_percpu(struct percpu_ref *ref);
109 110
void percpu_ref_kill_and_confirm(struct percpu_ref *ref,
				 percpu_ref_func_t *confirm_kill);
111
void percpu_ref_resurrect(struct percpu_ref *ref);
112
void percpu_ref_reinit(struct percpu_ref *ref);
113 114 115 116 117 118 119 120

/**
 * percpu_ref_kill - drop the initial ref
 * @ref: percpu_ref to kill
 *
 * Must be used to drop the initial ref on a percpu refcount; must be called
 * precisely once before shutdown.
 *
121 122 123 124
 * Switches @ref into atomic mode before gathering up the percpu counters
 * and dropping the initial ref.
 *
 * There are no implied RCU grace periods between kill and release.
125 126 127
 */
static inline void percpu_ref_kill(struct percpu_ref *ref)
{
128
	percpu_ref_kill_and_confirm(ref, NULL);
129
}
130

131 132 133 134
/*
 * Internal helper.  Don't use outside percpu-refcount proper.  The
 * function doesn't return the pointer and let the caller test it for NULL
 * because doing so forces the compiler to generate two conditional
135
 * branches as it can't assume that @ref->percpu_count is not NULL.
136
 */
137 138
static inline bool __ref_is_percpu(struct percpu_ref *ref,
					  unsigned long __percpu **percpu_countp)
139
{
140 141 142 143 144 145 146 147
	unsigned long percpu_ptr;

	/*
	 * The value of @ref->percpu_count_ptr is tested for
	 * !__PERCPU_REF_ATOMIC, which may be set asynchronously, and then
	 * used as a pointer.  If the compiler generates a separate fetch
	 * when using it as a pointer, __PERCPU_REF_ATOMIC may be set in
	 * between contaminating the pointer value, meaning that
148
	 * READ_ONCE() is required when fetching it.
149 150 151
	 *
	 * The smp_read_barrier_depends() implied by READ_ONCE() pairs
	 * with smp_store_release() in __percpu_ref_switch_to_percpu().
152
	 */
153 154
	percpu_ptr = READ_ONCE(ref->percpu_count_ptr);

155 156 157 158 159 160 161
	/*
	 * Theoretically, the following could test just ATOMIC; however,
	 * then we'd have to mask off DEAD separately as DEAD may be
	 * visible without ATOMIC if we race with percpu_ref_kill().  DEAD
	 * implies ATOMIC anyway.  Test them together.
	 */
	if (unlikely(percpu_ptr & __PERCPU_REF_ATOMIC_DEAD))
162 163
		return false;

164
	*percpu_countp = (unsigned long __percpu *)percpu_ptr;
165 166
	return true;
}
167 168

/**
169
 * percpu_ref_get_many - increment a percpu refcount
T
Tejun Heo 已提交
170
 * @ref: percpu_ref to get
171
 * @nr: number of references to get
172
 *
173
 * Analogous to atomic_long_add().
174 175 176
 *
 * This function is safe to call as long as @ref is between init and exit.
 */
177
static inline void percpu_ref_get_many(struct percpu_ref *ref, unsigned long nr)
178
{
179
	unsigned long __percpu *percpu_count;
180

181
	rcu_read_lock_sched();
182

183
	if (__ref_is_percpu(ref, &percpu_count))
184
		this_cpu_add(*percpu_count, nr);
185
	else
186
		atomic_long_add(nr, &ref->count);
187

188
	rcu_read_unlock_sched();
189 190
}

191 192 193 194 195 196 197 198 199 200 201 202 203
/**
 * percpu_ref_get - increment a percpu refcount
 * @ref: percpu_ref to get
 *
 * Analagous to atomic_long_inc().
 *
 * This function is safe to call as long as @ref is between init and exit.
 */
static inline void percpu_ref_get(struct percpu_ref *ref)
{
	percpu_ref_get_many(ref, 1);
}

204 205 206 207 208 209 210
/**
 * percpu_ref_tryget - try to increment a percpu refcount
 * @ref: percpu_ref to try-get
 *
 * Increment a percpu refcount unless its count already reached zero.
 * Returns %true on success; %false on failure.
 *
211
 * This function is safe to call as long as @ref is between init and exit.
212 213 214
 */
static inline bool percpu_ref_tryget(struct percpu_ref *ref)
{
215
	unsigned long __percpu *percpu_count;
216
	bool ret;
217 218 219

	rcu_read_lock_sched();

220
	if (__ref_is_percpu(ref, &percpu_count)) {
221
		this_cpu_inc(*percpu_count);
222 223
		ret = true;
	} else {
224
		ret = atomic_long_inc_not_zero(&ref->count);
225 226 227 228 229 230 231
	}

	rcu_read_unlock_sched();

	return ret;
}

232
/**
233
 * percpu_ref_tryget_live - try to increment a live percpu refcount
234 235 236 237 238
 * @ref: percpu_ref to try-get
 *
 * Increment a percpu refcount unless it has already been killed.  Returns
 * %true on success; %false on failure.
 *
239 240 241 242 243
 * Completion of percpu_ref_kill() in itself doesn't guarantee that this
 * function will fail.  For such guarantee, percpu_ref_kill_and_confirm()
 * should be used.  After the confirm_kill callback is invoked, it's
 * guaranteed that no new reference will be given out by
 * percpu_ref_tryget_live().
244
 *
245
 * This function is safe to call as long as @ref is between init and exit.
246
 */
247
static inline bool percpu_ref_tryget_live(struct percpu_ref *ref)
248
{
249
	unsigned long __percpu *percpu_count;
250
	bool ret = false;
251

252
	rcu_read_lock_sched();
253

254
	if (__ref_is_percpu(ref, &percpu_count)) {
255
		this_cpu_inc(*percpu_count);
256
		ret = true;
257
	} else if (!(ref->percpu_count_ptr & __PERCPU_REF_DEAD)) {
258
		ret = atomic_long_inc_not_zero(&ref->count);
259 260
	}

261
	rcu_read_unlock_sched();
262 263 264 265

	return ret;
}

266
/**
267
 * percpu_ref_put_many - decrement a percpu refcount
T
Tejun Heo 已提交
268
 * @ref: percpu_ref to put
269
 * @nr: number of references to put
270 271 272
 *
 * Decrement the refcount, and if 0, call the release function (which was passed
 * to percpu_ref_init())
273 274
 *
 * This function is safe to call as long as @ref is between init and exit.
275
 */
276
static inline void percpu_ref_put_many(struct percpu_ref *ref, unsigned long nr)
277
{
278
	unsigned long __percpu *percpu_count;
279

280
	rcu_read_lock_sched();
281

282
	if (__ref_is_percpu(ref, &percpu_count))
283 284
		this_cpu_sub(*percpu_count, nr);
	else if (unlikely(atomic_long_sub_and_test(nr, &ref->count)))
285 286
		ref->release(ref);

287
	rcu_read_unlock_sched();
288 289
}

290 291 292 293 294 295 296 297 298 299 300 301 302 303
/**
 * percpu_ref_put - decrement a percpu refcount
 * @ref: percpu_ref to put
 *
 * Decrement the refcount, and if 0, call the release function (which was passed
 * to percpu_ref_init())
 *
 * This function is safe to call as long as @ref is between init and exit.
 */
static inline void percpu_ref_put(struct percpu_ref *ref)
{
	percpu_ref_put_many(ref, 1);
}

304 305 306 307 308 309 310 311 312 313 314 315 316 317
/**
 * percpu_ref_is_dying - test whether a percpu refcount is dying or dead
 * @ref: percpu_ref to test
 *
 * Returns %true if @ref is dying or dead.
 *
 * This function is safe to call as long as @ref is between init and exit
 * and the caller is responsible for synchronizing against state changes.
 */
static inline bool percpu_ref_is_dying(struct percpu_ref *ref)
{
	return ref->percpu_count_ptr & __PERCPU_REF_DEAD;
}

318 319 320 321 322
/**
 * percpu_ref_is_zero - test whether a percpu refcount reached zero
 * @ref: percpu_ref to test
 *
 * Returns %true if @ref reached zero.
323 324
 *
 * This function is safe to call as long as @ref is between init and exit.
325 326 327
 */
static inline bool percpu_ref_is_zero(struct percpu_ref *ref)
{
328
	unsigned long __percpu *percpu_count;
329

330
	if (__ref_is_percpu(ref, &percpu_count))
331
		return false;
332
	return !atomic_long_read(&ref->count);
333 334
}

335
#endif