Commit | Line | Data |
---|---|---|
215e262f KO |
1 | /* |
2 | * Percpu refcounts: | |
3 | * (C) 2012 Google, Inc. | |
4 | * Author: Kent Overstreet <koverstreet@google.com> | |
5 | * | |
6 | * This implements a refcount with similar semantics to atomic_t - atomic_inc(), | |
7 | * atomic_dec_and_test() - but percpu. | |
8 | * | |
9 | * There's one important difference between percpu refs and normal atomic_t | |
10 | * refcounts; you have to keep track of your initial refcount, and then when you | |
11 | * start shutting down you call percpu_ref_kill() _before_ dropping the initial | |
12 | * refcount. | |
13 | * | |
14 | * The refcount will have a range of 0 to ((1U << 31) - 1), i.e. one bit less | |
15 | * than an atomic_t - this is because of the way shutdown works, see | |
eecc16ba | 16 | * percpu_ref_kill()/PERCPU_COUNT_BIAS. |
215e262f KO |
17 | * |
18 | * Before you call percpu_ref_kill(), percpu_ref_put() does not check for the | |
19 | * refcount hitting 0 - it can't, if it was in percpu mode. percpu_ref_kill() | |
20 | * puts the ref back in single atomic_t mode, collecting the per cpu refs and | |
21 | * issuing the appropriate barriers, and then marks the ref as shutting down so | |
22 | * that percpu_ref_put() will check for the ref hitting 0. After it returns, | |
23 | * it's safe to drop the initial ref. | |
24 | * | |
25 | * USAGE: | |
26 | * | |
27 | * See fs/aio.c for some example usage; it's used there for struct kioctx, which | |
28 | * is created when userspaces calls io_setup(), and destroyed when userspace | |
29 | * calls io_destroy() or the process exits. | |
30 | * | |
31 | * In the aio code, kill_ioctx() is called when we wish to destroy a kioctx; it | |
d7cdb968 | 32 | * calls percpu_ref_kill(), then hlist_del_rcu() and synchronize_rcu() to remove |
215e262f KO |
33 | * the kioctx from the proccess's list of kioctxs - after that, there can't be |
34 | * any new users of the kioctx (from lookup_ioctx()) and it's then safe to drop | |
35 | * the initial ref with percpu_ref_put(). | |
36 | * | |
37 | * Code that does a two stage shutdown like this often needs some kind of | |
38 | * explicit synchronization to ensure the initial refcount can only be dropped | |
39 | * once - percpu_ref_kill() does this for you, it returns true once and false if | |
40 | * someone else already called it. The aio code uses it this way, but it's not | |
41 | * necessary if the code has some other mechanism to synchronize teardown. | |
42 | * around. | |
43 | */ | |
44 | ||
45 | #ifndef _LINUX_PERCPU_REFCOUNT_H | |
46 | #define _LINUX_PERCPU_REFCOUNT_H | |
47 | ||
48 | #include <linux/atomic.h> | |
49 | #include <linux/kernel.h> | |
50 | #include <linux/percpu.h> | |
51 | #include <linux/rcupdate.h> | |
a34375ef | 52 | #include <linux/gfp.h> |
215e262f KO |
53 | |
54 | struct percpu_ref; | |
ac899061 | 55 | typedef void (percpu_ref_func_t)(struct percpu_ref *); |
215e262f | 56 | |
9e804d1f TH |
57 | /* flags set in the lower bits of percpu_ref->percpu_count_ptr */ |
58 | enum { | |
59 | __PERCPU_REF_ATOMIC = 1LU << 0, /* operating in atomic mode */ | |
27344a90 TH |
60 | __PERCPU_REF_DEAD = 1LU << 1, /* (being) killed */ |
61 | __PERCPU_REF_ATOMIC_DEAD = __PERCPU_REF_ATOMIC | __PERCPU_REF_DEAD, | |
62 | ||
63 | __PERCPU_REF_FLAG_BITS = 2, | |
9e804d1f TH |
64 | }; |
65 | ||
2aad2a86 TH |
66 | /* @flags for percpu_ref_init() */ |
67 | enum { | |
68 | /* | |
69 | * Start w/ ref == 1 in atomic mode. Can be switched to percpu | |
1cae13e7 TH |
70 | * operation using percpu_ref_switch_to_percpu(). If initialized |
71 | * with this flag, the ref will stay in atomic mode until | |
72 | * percpu_ref_switch_to_percpu() is invoked on it. | |
2aad2a86 TH |
73 | */ |
74 | PERCPU_REF_INIT_ATOMIC = 1 << 0, | |
75 | ||
76 | /* | |
77 | * Start dead w/ ref == 0 in atomic mode. Must be revived with | |
78 | * percpu_ref_reinit() before used. Implies INIT_ATOMIC. | |
79 | */ | |
80 | PERCPU_REF_INIT_DEAD = 1 << 1, | |
81 | }; | |
82 | ||
215e262f | 83 | struct percpu_ref { |
e625305b | 84 | atomic_long_t count; |
215e262f KO |
85 | /* |
86 | * The low bit of the pointer indicates whether the ref is in percpu | |
9a1049da | 87 | * mode; if set, then get/put will manipulate the atomic_t. |
215e262f | 88 | */ |
eecc16ba | 89 | unsigned long percpu_count_ptr; |
ac899061 | 90 | percpu_ref_func_t *release; |
9e804d1f | 91 | percpu_ref_func_t *confirm_switch; |
1cae13e7 | 92 | bool force_atomic:1; |
215e262f KO |
93 | struct rcu_head rcu; |
94 | }; | |
95 | ||
acac7883 | 96 | int __must_check percpu_ref_init(struct percpu_ref *ref, |
2aad2a86 TH |
97 | percpu_ref_func_t *release, unsigned int flags, |
98 | gfp_t gfp); | |
9a1049da | 99 | void percpu_ref_exit(struct percpu_ref *ref); |
490c79a6 TH |
100 | void percpu_ref_switch_to_atomic(struct percpu_ref *ref, |
101 | percpu_ref_func_t *confirm_switch); | |
f47ad457 | 102 | void percpu_ref_switch_to_percpu(struct percpu_ref *ref); |
dbece3a0 TH |
103 | void percpu_ref_kill_and_confirm(struct percpu_ref *ref, |
104 | percpu_ref_func_t *confirm_kill); | |
f47ad457 | 105 | void percpu_ref_reinit(struct percpu_ref *ref); |
dbece3a0 TH |
106 | |
107 | /** | |
108 | * percpu_ref_kill - drop the initial ref | |
109 | * @ref: percpu_ref to kill | |
110 | * | |
111 | * Must be used to drop the initial ref on a percpu refcount; must be called | |
112 | * precisely once before shutdown. | |
113 | * | |
114 | * Puts @ref in non percpu mode, then does a call_rcu() before gathering up the | |
115 | * percpu counters and dropping the initial ref. | |
116 | */ | |
117 | static inline void percpu_ref_kill(struct percpu_ref *ref) | |
118 | { | |
119 | return percpu_ref_kill_and_confirm(ref, NULL); | |
120 | } | |
215e262f | 121 | |
eae7975d TH |
122 | /* |
123 | * Internal helper. Don't use outside percpu-refcount proper. The | |
124 | * function doesn't return the pointer and let the caller test it for NULL | |
125 | * because doing so forces the compiler to generate two conditional | |
eecc16ba | 126 | * branches as it can't assume that @ref->percpu_count is not NULL. |
eae7975d | 127 | */ |
9e804d1f TH |
128 | static inline bool __ref_is_percpu(struct percpu_ref *ref, |
129 | unsigned long __percpu **percpu_countp) | |
eae7975d | 130 | { |
2d722782 | 131 | /* paired with smp_store_release() in percpu_ref_reinit() */ |
eadac03e | 132 | unsigned long percpu_ptr = lockless_dereference(ref->percpu_count_ptr); |
2d722782 | 133 | |
4aab3b5b TH |
134 | /* |
135 | * Theoretically, the following could test just ATOMIC; however, | |
136 | * then we'd have to mask off DEAD separately as DEAD may be | |
137 | * visible without ATOMIC if we race with percpu_ref_kill(). DEAD | |
138 | * implies ATOMIC anyway. Test them together. | |
139 | */ | |
140 | if (unlikely(percpu_ptr & __PERCPU_REF_ATOMIC_DEAD)) | |
eae7975d TH |
141 | return false; |
142 | ||
eecc16ba | 143 | *percpu_countp = (unsigned long __percpu *)percpu_ptr; |
eae7975d TH |
144 | return true; |
145 | } | |
215e262f KO |
146 | |
147 | /** | |
e8ea14cc | 148 | * percpu_ref_get_many - increment a percpu refcount |
ac899061 | 149 | * @ref: percpu_ref to get |
e8ea14cc | 150 | * @nr: number of references to get |
215e262f | 151 | * |
e8ea14cc | 152 | * Analogous to atomic_long_add(). |
6251f997 TH |
153 | * |
154 | * This function is safe to call as long as @ref is between init and exit. | |
155 | */ | |
e8ea14cc | 156 | static inline void percpu_ref_get_many(struct percpu_ref *ref, unsigned long nr) |
215e262f | 157 | { |
eecc16ba | 158 | unsigned long __percpu *percpu_count; |
215e262f | 159 | |
a4244454 | 160 | rcu_read_lock_sched(); |
215e262f | 161 | |
9e804d1f | 162 | if (__ref_is_percpu(ref, &percpu_count)) |
e8ea14cc | 163 | this_cpu_add(*percpu_count, nr); |
215e262f | 164 | else |
e8ea14cc | 165 | atomic_long_add(nr, &ref->count); |
215e262f | 166 | |
a4244454 | 167 | rcu_read_unlock_sched(); |
215e262f KO |
168 | } |
169 | ||
e8ea14cc JW |
170 | /** |
171 | * percpu_ref_get - increment a percpu refcount | |
172 | * @ref: percpu_ref to get | |
173 | * | |
174 | * Analagous to atomic_long_inc(). | |
175 | * | |
176 | * This function is safe to call as long as @ref is between init and exit. | |
177 | */ | |
178 | static inline void percpu_ref_get(struct percpu_ref *ref) | |
179 | { | |
180 | percpu_ref_get_many(ref, 1); | |
181 | } | |
182 | ||
4fb6e250 TH |
183 | /** |
184 | * percpu_ref_tryget - try to increment a percpu refcount | |
185 | * @ref: percpu_ref to try-get | |
186 | * | |
187 | * Increment a percpu refcount unless its count already reached zero. | |
188 | * Returns %true on success; %false on failure. | |
189 | * | |
6251f997 | 190 | * This function is safe to call as long as @ref is between init and exit. |
4fb6e250 TH |
191 | */ |
192 | static inline bool percpu_ref_tryget(struct percpu_ref *ref) | |
193 | { | |
eecc16ba | 194 | unsigned long __percpu *percpu_count; |
6251f997 | 195 | int ret; |
4fb6e250 TH |
196 | |
197 | rcu_read_lock_sched(); | |
198 | ||
9e804d1f | 199 | if (__ref_is_percpu(ref, &percpu_count)) { |
eecc16ba | 200 | this_cpu_inc(*percpu_count); |
4fb6e250 TH |
201 | ret = true; |
202 | } else { | |
e625305b | 203 | ret = atomic_long_inc_not_zero(&ref->count); |
4fb6e250 TH |
204 | } |
205 | ||
206 | rcu_read_unlock_sched(); | |
207 | ||
208 | return ret; | |
209 | } | |
210 | ||
dbece3a0 | 211 | /** |
2070d50e | 212 | * percpu_ref_tryget_live - try to increment a live percpu refcount |
dbece3a0 TH |
213 | * @ref: percpu_ref to try-get |
214 | * | |
215 | * Increment a percpu refcount unless it has already been killed. Returns | |
216 | * %true on success; %false on failure. | |
217 | * | |
6251f997 TH |
218 | * Completion of percpu_ref_kill() in itself doesn't guarantee that this |
219 | * function will fail. For such guarantee, percpu_ref_kill_and_confirm() | |
220 | * should be used. After the confirm_kill callback is invoked, it's | |
221 | * guaranteed that no new reference will be given out by | |
222 | * percpu_ref_tryget_live(). | |
4fb6e250 | 223 | * |
6251f997 | 224 | * This function is safe to call as long as @ref is between init and exit. |
dbece3a0 | 225 | */ |
2070d50e | 226 | static inline bool percpu_ref_tryget_live(struct percpu_ref *ref) |
dbece3a0 | 227 | { |
eecc16ba | 228 | unsigned long __percpu *percpu_count; |
dbece3a0 TH |
229 | int ret = false; |
230 | ||
a4244454 | 231 | rcu_read_lock_sched(); |
dbece3a0 | 232 | |
9e804d1f | 233 | if (__ref_is_percpu(ref, &percpu_count)) { |
eecc16ba | 234 | this_cpu_inc(*percpu_count); |
dbece3a0 | 235 | ret = true; |
490c79a6 TH |
236 | } else if (!(ACCESS_ONCE(ref->percpu_count_ptr) & __PERCPU_REF_DEAD)) { |
237 | ret = atomic_long_inc_not_zero(&ref->count); | |
dbece3a0 TH |
238 | } |
239 | ||
a4244454 | 240 | rcu_read_unlock_sched(); |
dbece3a0 TH |
241 | |
242 | return ret; | |
243 | } | |
244 | ||
215e262f | 245 | /** |
e8ea14cc | 246 | * percpu_ref_put_many - decrement a percpu refcount |
ac899061 | 247 | * @ref: percpu_ref to put |
e8ea14cc | 248 | * @nr: number of references to put |
215e262f KO |
249 | * |
250 | * Decrement the refcount, and if 0, call the release function (which was passed | |
251 | * to percpu_ref_init()) | |
6251f997 TH |
252 | * |
253 | * This function is safe to call as long as @ref is between init and exit. | |
215e262f | 254 | */ |
e8ea14cc | 255 | static inline void percpu_ref_put_many(struct percpu_ref *ref, unsigned long nr) |
215e262f | 256 | { |
eecc16ba | 257 | unsigned long __percpu *percpu_count; |
215e262f | 258 | |
a4244454 | 259 | rcu_read_lock_sched(); |
215e262f | 260 | |
9e804d1f | 261 | if (__ref_is_percpu(ref, &percpu_count)) |
e8ea14cc JW |
262 | this_cpu_sub(*percpu_count, nr); |
263 | else if (unlikely(atomic_long_sub_and_test(nr, &ref->count))) | |
215e262f KO |
264 | ref->release(ref); |
265 | ||
a4244454 | 266 | rcu_read_unlock_sched(); |
215e262f KO |
267 | } |
268 | ||
e8ea14cc JW |
269 | /** |
270 | * percpu_ref_put - decrement a percpu refcount | |
271 | * @ref: percpu_ref to put | |
272 | * | |
273 | * Decrement the refcount, and if 0, call the release function (which was passed | |
274 | * to percpu_ref_init()) | |
275 | * | |
276 | * This function is safe to call as long as @ref is between init and exit. | |
277 | */ | |
278 | static inline void percpu_ref_put(struct percpu_ref *ref) | |
279 | { | |
280 | percpu_ref_put_many(ref, 1); | |
281 | } | |
282 | ||
2d722782 TH |
283 | /** |
284 | * percpu_ref_is_zero - test whether a percpu refcount reached zero | |
285 | * @ref: percpu_ref to test | |
286 | * | |
287 | * Returns %true if @ref reached zero. | |
6251f997 TH |
288 | * |
289 | * This function is safe to call as long as @ref is between init and exit. | |
2d722782 TH |
290 | */ |
291 | static inline bool percpu_ref_is_zero(struct percpu_ref *ref) | |
292 | { | |
eecc16ba | 293 | unsigned long __percpu *percpu_count; |
2d722782 | 294 | |
9e804d1f | 295 | if (__ref_is_percpu(ref, &percpu_count)) |
2d722782 | 296 | return false; |
e625305b | 297 | return !atomic_long_read(&ref->count); |
2d722782 TH |
298 | } |
299 | ||
215e262f | 300 | #endif |