source: rtems-libbsd/freebsd/sys/contrib/ck/include/ck_epoch.h @ 1af372a

55-freebsd-126-freebsd-12
Last change on this file since 1af372a was bcdce02, checked in by Sebastian Huber <sebastian.huber@…>, on 08/21/18 at 11:47:02

Update to FreeBSD head 2018-06-01

Git mirror commit fb63610a69b0eb7f69a201ba05c4c1a7a2739cf9.

Update #3472.

  • Property mode set to 100644
File size: 7.8 KB
Line 
1/*
2 * Copyright 2011-2015 Samy Al Bahra.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#ifndef CK_EPOCH_H
28#define CK_EPOCH_H
29
30/*
31 * The implementation here is inspired from the work described in:
32 *   Fraser, K. 2004. Practical Lock-Freedom. PhD Thesis, University
33 *   of Cambridge Computing Laboratory.
34 */
35
36#include <ck_cc.h>
37#include <ck_md.h>
38#include <ck_pr.h>
39#include <ck_stack.h>
40#include <ck_stdbool.h>
41
42#ifndef CK_EPOCH_LENGTH
43#define CK_EPOCH_LENGTH 4
44#endif
45
46/*
47 * This is used for sense detection with-respect to concurrent
48 * epoch sections.
49 */
50#define CK_EPOCH_SENSE          (2)
51
52struct ck_epoch_entry;
53typedef struct ck_epoch_entry ck_epoch_entry_t;
54typedef void ck_epoch_cb_t(ck_epoch_entry_t *);
55
56/*
57 * This should be embedded into objects you wish to be the target of
58 * ck_epoch_cb_t functions (with ck_epoch_call).
59 */
60struct ck_epoch_entry {
61        ck_epoch_cb_t *function;
62        ck_stack_entry_t stack_entry;
63};
64
65/*
66 * A section object may be passed to every begin-end pair to allow for
67 * forward progress guarantees with-in prolonged active sections.
68 */
69struct ck_epoch_section {
70        unsigned int bucket;
71};
72typedef struct ck_epoch_section ck_epoch_section_t;
73
74/*
75 * Return pointer to ck_epoch_entry container object.
76 */
77#define CK_EPOCH_CONTAINER(T, M, N) \
78        CK_CC_CONTAINER(struct ck_epoch_entry, T, M, N)
79
80struct ck_epoch_ref {
81        unsigned int epoch;
82        unsigned int count;
83};
84
85struct ck_epoch_record {
86        ck_stack_entry_t record_next;
87        struct ck_epoch *global;
88        unsigned int state;
89        unsigned int epoch;
90        unsigned int active;
91        struct {
92                struct ck_epoch_ref bucket[CK_EPOCH_SENSE];
93        } local CK_CC_CACHELINE;
94        unsigned int n_pending;
95        unsigned int n_peak;
96        unsigned int n_dispatch;
97        void *ct;
98        ck_stack_t pending[CK_EPOCH_LENGTH];
99} CK_CC_CACHELINE;
100typedef struct ck_epoch_record ck_epoch_record_t;
101
102struct ck_epoch {
103        unsigned int epoch;
104        unsigned int n_free;
105        ck_stack_t records;
106};
107typedef struct ck_epoch ck_epoch_t;
108
109/*
110 * Internal functions.
111 */
112void _ck_epoch_addref(ck_epoch_record_t *, ck_epoch_section_t *);
113bool _ck_epoch_delref(ck_epoch_record_t *, ck_epoch_section_t *);
114
115CK_CC_FORCE_INLINE static void *
116ck_epoch_record_ct(const ck_epoch_record_t *record)
117{
118
119        return ck_pr_load_ptr(&record->ct);
120}
121
122/*
123 * Marks the beginning of an epoch-protected section.
124 */
125CK_CC_FORCE_INLINE static void
126ck_epoch_begin(ck_epoch_record_t *record, ck_epoch_section_t *section)
127{
128        struct ck_epoch *epoch = record->global;
129
130        /*
131         * Only observe new epoch if thread is not recursing into a read
132         * section.
133         */
134        if (record->active == 0) {
135                unsigned int g_epoch;
136
137                /*
138                 * It is possible for loads to be re-ordered before the store
139                 * is committed into the caller's epoch and active fields.
140                 * For this reason, store to load serialization is necessary.
141                 */
142#if defined(CK_MD_TSO)
143                ck_pr_fas_uint(&record->active, 1);
144                ck_pr_fence_atomic_load();
145#else
146                ck_pr_store_uint(&record->active, 1);
147                ck_pr_fence_memory();
148#endif
149
150                /*
151                 * This load is allowed to be re-ordered prior to setting
152                 * active flag due to monotonic nature of the global epoch.
153                 * However, stale values lead to measurable performance
154                 * degradation in some torture tests so we disallow early load
155                 * of global epoch.
156                 */
157                g_epoch = ck_pr_load_uint(&epoch->epoch);
158                ck_pr_store_uint(&record->epoch, g_epoch);
159        } else {
160                ck_pr_store_uint(&record->active, record->active + 1);
161        }
162
163        if (section != NULL)
164                _ck_epoch_addref(record, section);
165
166        return;
167}
168
169/*
170 * Marks the end of an epoch-protected section. Returns true if no more
171 * sections exist for the caller.
172 */
173CK_CC_FORCE_INLINE static bool
174ck_epoch_end(ck_epoch_record_t *record, ck_epoch_section_t *section)
175{
176
177        ck_pr_fence_release();
178        ck_pr_store_uint(&record->active, record->active - 1);
179
180        if (section != NULL)
181                return _ck_epoch_delref(record, section);
182
183        return record->active == 0;
184}
185
186/*
187 * Defers the execution of the function pointed to by the "cb"
188 * argument until an epoch counter loop. This allows for a
189 * non-blocking deferral.
190 *
191 * We can get away without a fence here due to the monotonic nature
192 * of the epoch counter. Worst case, this will result in some delays
193 * before object destruction.
194 */
195CK_CC_FORCE_INLINE static void
196ck_epoch_call(ck_epoch_record_t *record,
197              ck_epoch_entry_t *entry,
198              ck_epoch_cb_t *function)
199{
200        struct ck_epoch *epoch = record->global;
201        unsigned int e = ck_pr_load_uint(&epoch->epoch);
202        unsigned int offset = e & (CK_EPOCH_LENGTH - 1);
203
204        record->n_pending++;
205        entry->function = function;
206        ck_stack_push_spnc(&record->pending[offset], &entry->stack_entry);
207        return;
208}
209
210/*
211 * Same as ck_epoch_call, but allows for records to be shared and is reentrant.
212 */
213CK_CC_FORCE_INLINE static void
214ck_epoch_call_strict(ck_epoch_record_t *record,
215              ck_epoch_entry_t *entry,
216              ck_epoch_cb_t *function)
217{
218        struct ck_epoch *epoch = record->global;
219        unsigned int e = ck_pr_load_uint(&epoch->epoch);
220        unsigned int offset = e & (CK_EPOCH_LENGTH - 1);
221
222        ck_pr_inc_uint(&record->n_pending);
223        entry->function = function;
224
225        /* Store fence is implied by push operation. */
226        ck_stack_push_upmc(&record->pending[offset], &entry->stack_entry);
227        return;
228}
229
230/*
231 * This callback is used for synchronize_wait to allow for custom blocking
232 * behavior.
233 */
234typedef void ck_epoch_wait_cb_t(ck_epoch_t *, ck_epoch_record_t *,
235    void *);
236
237/*
238 * Return latest epoch value. This operation provides load ordering.
239 */
240CK_CC_FORCE_INLINE static unsigned int
241ck_epoch_value(const ck_epoch_t *ep)
242{
243
244        ck_pr_fence_load();
245        return ck_pr_load_uint(&ep->epoch);
246}
247
248void ck_epoch_init(ck_epoch_t *);
249
250/*
251 * Attempts to recycle an unused epoch record. If one is successfully
252 * allocated, the record context pointer is also updated.
253 */
254ck_epoch_record_t *ck_epoch_recycle(ck_epoch_t *, void *);
255
256/*
257 * Registers an epoch record. An optional context pointer may be passed that
258 * is retrievable with ck_epoch_record_ct.
259 */
260void ck_epoch_register(ck_epoch_t *, ck_epoch_record_t *, void *);
261
262/*
263 * Marks a record as available for re-use by a subsequent recycle operation.
264 * Note that the record cannot be physically destroyed.
265 */
266void ck_epoch_unregister(ck_epoch_record_t *);
267
268bool ck_epoch_poll(ck_epoch_record_t *);
269bool ck_epoch_poll_deferred(struct ck_epoch_record *record, ck_stack_t *deferred);
270void ck_epoch_synchronize(ck_epoch_record_t *);
271void ck_epoch_synchronize_wait(ck_epoch_t *, ck_epoch_wait_cb_t *, void *);
272void ck_epoch_barrier(ck_epoch_record_t *);
273void ck_epoch_barrier_wait(ck_epoch_record_t *, ck_epoch_wait_cb_t *, void *);
274
275/*
276 * Reclaim entries associated with a record. This is safe to call only on
277 * the caller's record or records that are using call_strict.
278 */
279void ck_epoch_reclaim(ck_epoch_record_t *);
280
281#endif /* CK_EPOCH_H */
Note: See TracBrowser for help on using the repository browser.