1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2019 Arm Limited
3  */
4 
5 #ifndef _RTE_MCSLOCK_H_
6 #define _RTE_MCSLOCK_H_
7 
8 /**
9  * @file
10  *
11  * RTE MCS lock
12  *
13  * This file defines the main data structure and APIs for MCS queued lock.
14  *
15  * The MCS lock (proposed by John M. Mellor-Crummey and Michael L. Scott)
16  * provides scalability by spinning on a CPU/thread local variable which
17  * avoids expensive cache bouncings. It provides fairness by maintaining
18  * a list of acquirers and passing the lock to each CPU/thread in the order
19  * they acquired the lock.
20  */
21 
22 #include <rte_lcore.h>
23 #include <rte_common.h>
24 #include <rte_pause.h>
25 
26 /**
27  * The rte_mcslock_t type.
28  */
29 typedef struct rte_mcslock {
30 	struct rte_mcslock *next;
31 	int locked; /* 1 if the queue locked, 0 otherwise */
32 } rte_mcslock_t;
33 
34 /**
35  * Take the MCS lock.
36  *
37  * @param msl
38  *   A pointer to the pointer of a MCS lock.
39  *   When the lock is initialized or declared, the msl pointer should be
40  *   set to NULL.
41  * @param me
42  *   A pointer to a new node of MCS lock. Each CPU/thread acquiring the
43  *   lock should use its 'own node'.
44  */
45 static inline void
rte_mcslock_lock(rte_mcslock_t ** msl,rte_mcslock_t * me)46 rte_mcslock_lock(rte_mcslock_t **msl, rte_mcslock_t *me)
47 {
48 	rte_mcslock_t *prev;
49 
50 	/* Init me node */
51 	__atomic_store_n(&me->locked, 1, __ATOMIC_RELAXED);
52 	__atomic_store_n(&me->next, NULL, __ATOMIC_RELAXED);
53 
54 	/* If the queue is empty, the exchange operation is enough to acquire
55 	 * the lock. Hence, the exchange operation requires acquire semantics.
56 	 * The store to me->next above should complete before the node is
57 	 * visible to other CPUs/threads. Hence, the exchange operation requires
58 	 * release semantics as well.
59 	 */
60 	prev = __atomic_exchange_n(msl, me, __ATOMIC_ACQ_REL);
61 	if (likely(prev == NULL)) {
62 		/* Queue was empty, no further action required,
63 		 * proceed with lock taken.
64 		 */
65 		return;
66 	}
67 	/* The store to me->next above should also complete before the node is
68 	 * visible to predecessor thread releasing the lock. Hence, the store
69 	 * prev->next also requires release semantics. Note that, for example,
70 	 * on ARM, the release semantics in the exchange operation is not
71 	 * strong as a release fence and is not sufficient to enforce the
72 	 * desired order here.
73 	 */
74 	__atomic_store_n(&prev->next, me, __ATOMIC_RELEASE);
75 
76 	/* The while-load of me->locked should not move above the previous
77 	 * store to prev->next. Otherwise it will cause a deadlock. Need a
78 	 * store-load barrier.
79 	 */
80 	__atomic_thread_fence(__ATOMIC_ACQ_REL);
81 	/* If the lock has already been acquired, it first atomically
82 	 * places the node at the end of the queue and then proceeds
83 	 * to spin on me->locked until the previous lock holder resets
84 	 * the me->locked using mcslock_unlock().
85 	 */
86 	while (__atomic_load_n(&me->locked, __ATOMIC_ACQUIRE))
87 		rte_pause();
88 }
89 
90 /**
91  * Release the MCS lock.
92  *
93  * @param msl
94  *   A pointer to the pointer of a MCS lock.
95  * @param me
96  *   A pointer to the node of MCS lock passed in rte_mcslock_lock.
97  */
98 static inline void
rte_mcslock_unlock(rte_mcslock_t ** msl,rte_mcslock_t * me)99 rte_mcslock_unlock(rte_mcslock_t **msl, rte_mcslock_t *me)
100 {
101 	/* Check if there are more nodes in the queue. */
102 	if (likely(__atomic_load_n(&me->next, __ATOMIC_RELAXED) == NULL)) {
103 		/* No, last member in the queue. */
104 		rte_mcslock_t *save_me = __atomic_load_n(&me, __ATOMIC_RELAXED);
105 
106 		/* Release the lock by setting it to NULL */
107 		if (likely(__atomic_compare_exchange_n(msl, &save_me, NULL, 0,
108 				__ATOMIC_RELEASE, __ATOMIC_RELAXED)))
109 			return;
110 
111 		/* Speculative execution would be allowed to read in the
112 		 * while-loop first. This has the potential to cause a
113 		 * deadlock. Need a load barrier.
114 		 */
115 		__atomic_thread_fence(__ATOMIC_ACQUIRE);
116 		/* More nodes added to the queue by other CPUs.
117 		 * Wait until the next pointer is set.
118 		 */
119 		while (__atomic_load_n(&me->next, __ATOMIC_RELAXED) == NULL)
120 			rte_pause();
121 	}
122 
123 	/* Pass lock to next waiter. */
124 	__atomic_store_n(&me->next->locked, 0, __ATOMIC_RELEASE);
125 }
126 
127 /**
128  * Try to take the lock.
129  *
130  * @param msl
131  *   A pointer to the pointer of a MCS lock.
132  * @param me
133  *   A pointer to a new node of MCS lock.
134  * @return
135  *   1 if the lock is successfully taken; 0 otherwise.
136  */
137 static inline int
rte_mcslock_trylock(rte_mcslock_t ** msl,rte_mcslock_t * me)138 rte_mcslock_trylock(rte_mcslock_t **msl, rte_mcslock_t *me)
139 {
140 	/* Init me node */
141 	__atomic_store_n(&me->next, NULL, __ATOMIC_RELAXED);
142 
143 	/* Try to lock */
144 	rte_mcslock_t *expected = NULL;
145 
146 	/* The lock can be taken only when the queue is empty. Hence,
147 	 * the compare-exchange operation requires acquire semantics.
148 	 * The store to me->next above should complete before the node
149 	 * is visible to other CPUs/threads. Hence, the compare-exchange
150 	 * operation requires release semantics as well.
151 	 */
152 	return __atomic_compare_exchange_n(msl, &expected, me, 0,
153 			__ATOMIC_ACQ_REL, __ATOMIC_RELAXED);
154 }
155 
156 /**
157  * Test if the lock is taken.
158  *
159  * @param msl
160  *   A pointer to a MCS lock node.
161  * @return
162  *   1 if the lock is currently taken; 0 otherwise.
163  */
164 static inline int
rte_mcslock_is_locked(rte_mcslock_t * msl)165 rte_mcslock_is_locked(rte_mcslock_t *msl)
166 {
167 	return (__atomic_load_n(&msl, __ATOMIC_RELAXED) != NULL);
168 }
169 
170 #endif /* _RTE_MCSLOCK_H_ */
171