2
* Copyright © 2015 Intel
4
* Permission is hereby granted, free of charge, to any person obtaining a
5
* copy of this software and associated documentation files (the "Software"),
6
* to deal in the Software without restriction, including without limitation
7
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
8
* and/or sell copies of the Software, and to permit persons to whom the
9
* Software is furnished to do so, subject to the following conditions:
11
* The above copyright notice and this permission notice (including the next
12
* paragraph) shall be included in all copies or substantial portions of the
15
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
27
#include "util/futex.h"
28
#include "util/macros.h"
30
#include "c11/threads.h"
32
#if UTIL_FUTEX_SUPPORTED
34
#if defined(HAVE_VALGRIND) && !defined(NDEBUG)
35
# include <valgrind.h>
36
# include <helgrind.h>
42
/* mtx_t - Fast, simple mutex
44
* While modern pthread mutexes are very fast (implemented using futex), they
45
* still incur a call to an external DSO and overhead of the generality and
46
* features of pthread mutexes. Most mutexes in mesa only needs lock/unlock,
47
* and the idea here is that we can inline the atomic operation and make the
48
* fast case just two intructions. Mutexes are subtle and finicky to
49
* implement, so we carefully copy the implementation from Ulrich Dreppers
50
* well-written and well-reviewed paper:
52
* "Futexes Are Tricky"
53
* http://www.akkadia.org/drepper/futex.pdf
55
* We implement "mutex3", which gives us a mutex that has no syscalls on
56
* uncontended lock or unlock. Further, the uncontended case boils down to a
57
* locked cmpxchg and an untaken branch, the uncontended unlock is just a
58
* locked decr and an untaken branch. We use __builtin_expect() to indicate
59
* that contention is unlikely so that gcc will put the contention code out of
62
* A fast mutex only supports lock/unlock, can't be recursive or used with
63
* condition variables.
70
#define _SIMPLE_MTX_INITIALIZER_NP { 0 }
72
#define _SIMPLE_MTX_INVALID_VALUE 0xd0d0d0d0
75
simple_mtx_init(simple_mtx_t *mtx, ASSERTED int type)
77
assert(type == mtx_plain);
81
HG(ANNOTATE_RWLOCK_CREATE(mtx));
85
simple_mtx_destroy(ASSERTED simple_mtx_t *mtx)
87
HG(ANNOTATE_RWLOCK_DESTROY(mtx));
89
mtx->val = _SIMPLE_MTX_INVALID_VALUE;
94
simple_mtx_lock(simple_mtx_t *mtx)
98
c = __sync_val_compare_and_swap(&mtx->val, 0, 1);
100
assert(c != _SIMPLE_MTX_INVALID_VALUE);
102
if (__builtin_expect(c != 0, 0)) {
104
c = __sync_lock_test_and_set(&mtx->val, 2);
106
futex_wait(&mtx->val, 2, NULL);
107
c = __sync_lock_test_and_set(&mtx->val, 2);
111
HG(ANNOTATE_RWLOCK_ACQUIRED(mtx, 1));
115
simple_mtx_unlock(simple_mtx_t *mtx)
119
HG(ANNOTATE_RWLOCK_RELEASED(mtx, 1));
121
c = __sync_fetch_and_sub(&mtx->val, 1);
123
assert(c != _SIMPLE_MTX_INVALID_VALUE);
125
if (__builtin_expect(c != 1, 0)) {
127
futex_wake(&mtx->val, 1);
132
simple_mtx_assert_locked(simple_mtx_t *mtx)
139
typedef mtx_t simple_mtx_t;
141
#define _SIMPLE_MTX_INITIALIZER_NP _MTX_INITIALIZER_NP
144
simple_mtx_init(simple_mtx_t *mtx, int type)
150
simple_mtx_destroy(simple_mtx_t *mtx)
156
simple_mtx_lock(simple_mtx_t *mtx)
162
simple_mtx_unlock(simple_mtx_t *mtx)
168
simple_mtx_assert_locked(simple_mtx_t *mtx)
171
/* NOTE: this would not work for recursive mutexes, but
172
* mtx_t doesn't support those
174
int ret = mtx_trylock(mtx);
175
assert(ret == thrd_busy);
176
if (ret == thrd_success)