]>
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 | |
16 | * percpu_ref_kill()/PCPU_COUNT_BIAS. | |
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 | |
32 | * calls percpu_ref_kill(), then hlist_del_rcu() and sychronize_rcu() to remove | |
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 KO |
56 | |
57 | struct percpu_ref { | |
58 | atomic_t count; | |
59 | /* | |
60 | * The low bit of the pointer indicates whether the ref is in percpu | |
9a1049da | 61 | * mode; if set, then get/put will manipulate the atomic_t. |
215e262f | 62 | */ |
7d742075 | 63 | unsigned long pcpu_count_ptr; |
ac899061 | 64 | percpu_ref_func_t *release; |
dbece3a0 | 65 | percpu_ref_func_t *confirm_kill; |
215e262f KO |
66 | struct rcu_head rcu; |
67 | }; | |
68 | ||
acac7883 | 69 | int __must_check percpu_ref_init(struct percpu_ref *ref, |
a34375ef | 70 | percpu_ref_func_t *release, gfp_t gfp); |
2d722782 | 71 | void percpu_ref_reinit(struct percpu_ref *ref); |
9a1049da | 72 | void percpu_ref_exit(struct percpu_ref *ref); |
dbece3a0 TH |
73 | void percpu_ref_kill_and_confirm(struct percpu_ref *ref, |
74 | percpu_ref_func_t *confirm_kill); | |
75 | ||
76 | /** | |
77 | * percpu_ref_kill - drop the initial ref | |
78 | * @ref: percpu_ref to kill | |
79 | * | |
80 | * Must be used to drop the initial ref on a percpu refcount; must be called | |
81 | * precisely once before shutdown. | |
82 | * | |
83 | * Puts @ref in non percpu mode, then does a call_rcu() before gathering up the | |
84 | * percpu counters and dropping the initial ref. | |
85 | */ | |
86 | static inline void percpu_ref_kill(struct percpu_ref *ref) | |
87 | { | |
88 | return percpu_ref_kill_and_confirm(ref, NULL); | |
89 | } | |
215e262f | 90 | |
215e262f KO |
91 | #define PCPU_REF_DEAD 1 |
92 | ||
eae7975d TH |
93 | /* |
94 | * Internal helper. Don't use outside percpu-refcount proper. The | |
95 | * function doesn't return the pointer and let the caller test it for NULL | |
96 | * because doing so forces the compiler to generate two conditional | |
97 | * branches as it can't assume that @ref->pcpu_count is not NULL. | |
98 | */ | |
99 | static inline bool __pcpu_ref_alive(struct percpu_ref *ref, | |
100 | unsigned __percpu **pcpu_countp) | |
101 | { | |
7d742075 | 102 | unsigned long pcpu_ptr = ACCESS_ONCE(ref->pcpu_count_ptr); |
eae7975d | 103 | |
2d722782 TH |
104 | /* paired with smp_store_release() in percpu_ref_reinit() */ |
105 | smp_read_barrier_depends(); | |
106 | ||
eae7975d TH |
107 | if (unlikely(pcpu_ptr & PCPU_REF_DEAD)) |
108 | return false; | |
109 | ||
110 | *pcpu_countp = (unsigned __percpu *)pcpu_ptr; | |
111 | return true; | |
112 | } | |
215e262f KO |
113 | |
114 | /** | |
115 | * percpu_ref_get - increment a percpu refcount | |
ac899061 | 116 | * @ref: percpu_ref to get |
215e262f KO |
117 | * |
118 | * Analagous to atomic_inc(). | |
119 | */ | |
120 | static inline void percpu_ref_get(struct percpu_ref *ref) | |
121 | { | |
122 | unsigned __percpu *pcpu_count; | |
123 | ||
a4244454 | 124 | rcu_read_lock_sched(); |
215e262f | 125 | |
eae7975d | 126 | if (__pcpu_ref_alive(ref, &pcpu_count)) |
0c36b390 | 127 | this_cpu_inc(*pcpu_count); |
215e262f KO |
128 | else |
129 | atomic_inc(&ref->count); | |
130 | ||
a4244454 | 131 | rcu_read_unlock_sched(); |
215e262f KO |
132 | } |
133 | ||
4fb6e250 TH |
134 | /** |
135 | * percpu_ref_tryget - try to increment a percpu refcount | |
136 | * @ref: percpu_ref to try-get | |
137 | * | |
138 | * Increment a percpu refcount unless its count already reached zero. | |
139 | * Returns %true on success; %false on failure. | |
140 | * | |
141 | * The caller is responsible for ensuring that @ref stays accessible. | |
142 | */ | |
143 | static inline bool percpu_ref_tryget(struct percpu_ref *ref) | |
144 | { | |
145 | unsigned __percpu *pcpu_count; | |
146 | int ret = false; | |
147 | ||
148 | rcu_read_lock_sched(); | |
149 | ||
eae7975d | 150 | if (__pcpu_ref_alive(ref, &pcpu_count)) { |
315c5554 | 151 | this_cpu_inc(*pcpu_count); |
4fb6e250 TH |
152 | ret = true; |
153 | } else { | |
154 | ret = atomic_inc_not_zero(&ref->count); | |
155 | } | |
156 | ||
157 | rcu_read_unlock_sched(); | |
158 | ||
159 | return ret; | |
160 | } | |
161 | ||
dbece3a0 | 162 | /** |
2070d50e | 163 | * percpu_ref_tryget_live - try to increment a live percpu refcount |
dbece3a0 TH |
164 | * @ref: percpu_ref to try-get |
165 | * | |
166 | * Increment a percpu refcount unless it has already been killed. Returns | |
167 | * %true on success; %false on failure. | |
168 | * | |
169 | * Completion of percpu_ref_kill() in itself doesn't guarantee that tryget | |
170 | * will fail. For such guarantee, percpu_ref_kill_and_confirm() should be | |
171 | * used. After the confirm_kill callback is invoked, it's guaranteed that | |
172 | * no new reference will be given out by percpu_ref_tryget(). | |
4fb6e250 TH |
173 | * |
174 | * The caller is responsible for ensuring that @ref stays accessible. | |
dbece3a0 | 175 | */ |
2070d50e | 176 | static inline bool percpu_ref_tryget_live(struct percpu_ref *ref) |
dbece3a0 TH |
177 | { |
178 | unsigned __percpu *pcpu_count; | |
179 | int ret = false; | |
180 | ||
a4244454 | 181 | rcu_read_lock_sched(); |
dbece3a0 | 182 | |
eae7975d | 183 | if (__pcpu_ref_alive(ref, &pcpu_count)) { |
0c36b390 | 184 | this_cpu_inc(*pcpu_count); |
dbece3a0 TH |
185 | ret = true; |
186 | } | |
187 | ||
a4244454 | 188 | rcu_read_unlock_sched(); |
dbece3a0 TH |
189 | |
190 | return ret; | |
191 | } | |
192 | ||
215e262f KO |
193 | /** |
194 | * percpu_ref_put - decrement a percpu refcount | |
ac899061 | 195 | * @ref: percpu_ref to put |
215e262f KO |
196 | * |
197 | * Decrement the refcount, and if 0, call the release function (which was passed | |
198 | * to percpu_ref_init()) | |
199 | */ | |
200 | static inline void percpu_ref_put(struct percpu_ref *ref) | |
201 | { | |
202 | unsigned __percpu *pcpu_count; | |
203 | ||
a4244454 | 204 | rcu_read_lock_sched(); |
215e262f | 205 | |
eae7975d | 206 | if (__pcpu_ref_alive(ref, &pcpu_count)) |
0c36b390 | 207 | this_cpu_dec(*pcpu_count); |
215e262f KO |
208 | else if (unlikely(atomic_dec_and_test(&ref->count))) |
209 | ref->release(ref); | |
210 | ||
a4244454 | 211 | rcu_read_unlock_sched(); |
215e262f KO |
212 | } |
213 | ||
2d722782 TH |
214 | /** |
215 | * percpu_ref_is_zero - test whether a percpu refcount reached zero | |
216 | * @ref: percpu_ref to test | |
217 | * | |
218 | * Returns %true if @ref reached zero. | |
219 | */ | |
220 | static inline bool percpu_ref_is_zero(struct percpu_ref *ref) | |
221 | { | |
222 | unsigned __percpu *pcpu_count; | |
223 | ||
224 | if (__pcpu_ref_alive(ref, &pcpu_count)) | |
225 | return false; | |
226 | return !atomic_read(&ref->count); | |
227 | } | |
228 | ||
215e262f | 229 | #endif |