source: mainline/kernel/generic/include/atomic.h@ 6a0e568

Last change on this file since 6a0e568 was b2ec5cf, checked in by Jiří Zárevúcky <zarevucky.jiri@…>, 2 years ago

Implement atomic_time_stat_t for lockless timekeeping

We keep monotonically increasing temporal statistics in several places.
They are frequently written from the thread that owns them, and rarely
read from other threads in certain syscalls. This new code serves the
purpose of avoiding the need for synchronization on the writer side.
On 64b system, we can simply assume that 64b writes are indivisible,
and relaxed atomic read/writes simply serve to formally prevent C
undefined behavior from data races (they translate to regular memory
reads/writes in assembly).

On 32b systems, we use the same algorithm that's been used for userspace
clock access, using three fields and some memory barriers to maintain
consistency of reads when the upper half changes. Only readers always
synchronize though. For writers, barriers are avoided in the common case
when the upper half remains unchanged.

  • Property mode set to 100644
File size: 5.0 KB
Line 
1/*
2 * Copyright (c) 2006 Jakub Jermar
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 *
9 * - Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * - Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * - The name of the author may not be used to endorse or promote products
15 * derived from this software without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 */
28
29/** @addtogroup kernel_generic
30 * @{
31 */
32/** @file
33 */
34
35#ifndef KERN_ATOMIC_H_
36#define KERN_ATOMIC_H_
37
38#include <stdbool.h>
39#include <typedefs.h>
40#include <stdatomic.h>
41
42#define atomic_predec(val) \
43 (atomic_fetch_sub((val), 1) - 1)
44
45#define atomic_preinc(val) \
46 (atomic_fetch_add((val), 1) + 1)
47
48#define atomic_postdec(val) \
49 atomic_fetch_sub((val), 1)
50
51#define atomic_postinc(val) \
52 atomic_fetch_add((val), 1)
53
54#define atomic_dec(val) \
55 ((void) atomic_fetch_sub(val, 1))
56
57#define atomic_inc(val) \
58 ((void) atomic_fetch_add(val, 1))
59
60#define local_atomic_exchange(var_addr, new_val) \
61 atomic_exchange_explicit( \
62 (_Atomic typeof(*(var_addr)) *) (var_addr), \
63 (new_val), memory_order_relaxed)
64
65#if __64_BITS__
66
67typedef struct {
68 atomic_uint_fast64_t value;
69} atomic_time_stat_t;
70
71#define ATOMIC_TIME_INITIALIZER() (atomic_time_stat_t) {}
72
73static inline void atomic_time_increment(atomic_time_stat_t *time, int a)
74{
75 /*
76 * We require increments to be synchronized with each other, so we
77 * can use ordinary reads and writes instead of a more expensive atomic
78 * read-modify-write operations.
79 */
80 uint64_t v = atomic_load_explicit(&time->value, memory_order_relaxed);
81 atomic_store_explicit(&time->value, v + a, memory_order_relaxed);
82}
83
84static inline uint64_t atomic_time_read(atomic_time_stat_t *time)
85{
86 return atomic_load_explicit(&time->value, memory_order_relaxed);
87}
88
89#else
90
91/**
92 * A monotonically increasing 64b time statistic.
93 * Increments must be synchronized with each other (or limited to a single
94 * thread/CPU), but reads can be performed from any thread.
95 *
96 */
97typedef struct {
98 uint64_t true_value;
99 atomic_uint_fast32_t high1;
100 atomic_uint_fast32_t high2;
101 atomic_uint_fast32_t low;
102} atomic_time_stat_t;
103
104#define ATOMIC_TIME_INITIALIZER() (atomic_time_stat_t) {}
105
106static inline void atomic_time_increment(atomic_time_stat_t *time, int a)
107{
108 /*
109 * On 32b architectures, we can't rely on 64b memory reads/writes being
110 * architecturally atomic, but we also don't want to pay the cost of
111 * emulating atomic reads/writes, so instead we split value in half
112 * and perform some ordering magic to make sure readers always get
113 * consistent value.
114 */
115
116 /* true_value is only used by the writer, so this need not be atomic. */
117 uint64_t val = time->true_value;
118 uint32_t old_high = val >> 32;
119 val += a;
120 uint32_t new_high = val >> 32;
121 time->true_value = val;
122
123 /* Tell GCC that the first branch is far more likely than the second. */
124 if (__builtin_expect(old_high == new_high, 1)) {
125 /* If the high half didn't change, we need not bother with barriers. */
126 atomic_store_explicit(&time->low, (uint32_t) val, memory_order_relaxed);
127 } else {
128 /*
129 * If both halves changed, extra ordering is necessary.
130 * The idea is that if reader reads high1 and high2 with the same value,
131 * it is guaranteed that they read the correct low half for that value.
132 *
133 * This is the same sequence that is used by userspace to read clock.
134 */
135 atomic_store_explicit(&time->high1, new_high, memory_order_relaxed);
136 atomic_store_explicit(&time->low, (uint32_t) val, memory_order_release);
137 atomic_store_explicit(&time->high2, new_high, memory_order_release);
138 }
139}
140
141static inline uint64_t atomic_time_read(atomic_time_stat_t *time)
142{
143 uint32_t high2 = atomic_load_explicit(&time->high2, memory_order_acquire);
144 uint32_t low = atomic_load_explicit(&time->low, memory_order_acquire);
145 uint32_t high1 = atomic_load_explicit(&time->high1, memory_order_relaxed);
146
147 if (high1 != high2)
148 low = 0;
149
150 /* If the values differ, high1 is always the newer value. */
151 return (uint64_t) high1 << 32 | (uint64_t) low;
152}
153
154#endif /* __64_BITS__ */
155
156#endif
157
158/** @}
159 */
Note: See TracBrowser for help on using the repository browser.