1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/percpu.h>
3 #include <linux/sched.h>
4 #include <linux/osq_lock.h>
5
6 /*
7 * An MCS like lock especially tailored for optimistic spinning for sleeping
8 * lock implementations (mutex, rwsem, etc).
9 *
10 * Using a single mcs node per CPU is safe because sleeping locks should not be
11 * called from interrupt context and we have preemption disabled while
12 * spinning.
13 */
14
15 struct optimistic_spin_node {
16 struct optimistic_spin_node *next, *prev;
17 int locked; /* 1 if lock acquired */
18 int cpu; /* encoded CPU # + 1 value */
19 };
20
21 static DEFINE_PER_CPU_SHARED_ALIGNED(struct optimistic_spin_node, osq_node);
22
23 /*
24 * We use the value 0 to represent "no CPU", thus the encoded value
25 * will be the CPU number incremented by 1.
26 */
encode_cpu(int cpu_nr)27 static inline int encode_cpu(int cpu_nr)
28 {
29 return cpu_nr + 1;
30 }
31
node_cpu(struct optimistic_spin_node * node)32 static inline int node_cpu(struct optimistic_spin_node *node)
33 {
34 return node->cpu - 1;
35 }
36
decode_cpu(int encoded_cpu_val)37 static inline struct optimistic_spin_node *decode_cpu(int encoded_cpu_val)
38 {
39 int cpu_nr = encoded_cpu_val - 1;
40
41 return per_cpu_ptr(&osq_node, cpu_nr);
42 }
43
44 /*
45 * Get a stable @node->next pointer, either for unlock() or unqueue() purposes.
46 * Can return NULL in case we were the last queued and we updated @lock instead.
47 *
48 * If osq_lock() is being cancelled there must be a previous node
49 * and 'old_cpu' is its CPU #.
50 * For osq_unlock() there is never a previous node and old_cpu is
51 * set to OSQ_UNLOCKED_VAL.
52 */
53 static inline struct optimistic_spin_node *
osq_wait_next(struct optimistic_spin_queue * lock,struct optimistic_spin_node * node,int old_cpu)54 osq_wait_next(struct optimistic_spin_queue *lock,
55 struct optimistic_spin_node *node,
56 int old_cpu)
57 {
58 int curr = encode_cpu(smp_processor_id());
59
60 for (;;) {
61 if (atomic_read(&lock->tail) == curr &&
62 atomic_cmpxchg_acquire(&lock->tail, curr, old_cpu) == curr) {
63 /*
64 * We were the last queued, we moved @lock back. @prev
65 * will now observe @lock and will complete its
66 * unlock()/unqueue().
67 */
68 return NULL;
69 }
70
71 /*
72 * We must xchg() the @node->next value, because if we were to
73 * leave it in, a concurrent unlock()/unqueue() from
74 * @node->next might complete Step-A and think its @prev is
75 * still valid.
76 *
77 * If the concurrent unlock()/unqueue() wins the race, we'll
78 * wait for either @lock to point to us, through its Step-B, or
79 * wait for a new @node->next from its Step-C.
80 */
81 if (node->next) {
82 struct optimistic_spin_node *next;
83
84 next = xchg(&node->next, NULL);
85 if (next)
86 return next;
87 }
88
89 cpu_relax();
90 }
91 }
92
osq_lock(struct optimistic_spin_queue * lock)93 bool osq_lock(struct optimistic_spin_queue *lock)
94 {
95 struct optimistic_spin_node *node = this_cpu_ptr(&osq_node);
96 struct optimistic_spin_node *prev, *next;
97 int curr = encode_cpu(smp_processor_id());
98 int old;
99
100 node->locked = 0;
101 node->next = NULL;
102 node->cpu = curr;
103
104 /*
105 * We need both ACQUIRE (pairs with corresponding RELEASE in
106 * unlock() uncontended, or fastpath) and RELEASE (to publish
107 * the node fields we just initialised) semantics when updating
108 * the lock tail.
109 */
110 old = atomic_xchg(&lock->tail, curr);
111 if (old == OSQ_UNLOCKED_VAL)
112 return true;
113
114 prev = decode_cpu(old);
115 node->prev = prev;
116
117 /*
118 * osq_lock() unqueue
119 *
120 * node->prev = prev osq_wait_next()
121 * WMB MB
122 * prev->next = node next->prev = prev // unqueue-C
123 *
124 * Here 'node->prev' and 'next->prev' are the same variable and we need
125 * to ensure these stores happen in-order to avoid corrupting the list.
126 */
127 smp_wmb();
128
129 WRITE_ONCE(prev->next, node);
130
131 /*
132 * Normally @prev is untouchable after the above store; because at that
133 * moment unlock can proceed and wipe the node element from stack.
134 *
135 * However, since our nodes are static per-cpu storage, we're
136 * guaranteed their existence -- this allows us to apply
137 * cmpxchg in an attempt to undo our queueing.
138 */
139
140 /*
141 * Wait to acquire the lock or cancellation. Note that need_resched()
142 * will come with an IPI, which will wake smp_cond_load_relaxed() if it
143 * is implemented with a monitor-wait. vcpu_is_preempted() relies on
144 * polling, be careful.
145 */
146 if (smp_cond_load_relaxed(&node->locked, VAL || need_resched() ||
147 vcpu_is_preempted(node_cpu(node->prev))))
148 return true;
149
150 /* unqueue */
151 /*
152 * Step - A -- stabilize @prev
153 *
154 * Undo our @prev->next assignment; this will make @prev's
155 * unlock()/unqueue() wait for a next pointer since @lock points to us
156 * (or later).
157 */
158
159 for (;;) {
160 /*
161 * cpu_relax() below implies a compiler barrier which would
162 * prevent this comparison being optimized away.
163 */
164 if (data_race(prev->next) == node &&
165 cmpxchg(&prev->next, node, NULL) == node)
166 break;
167
168 /*
169 * We can only fail the cmpxchg() racing against an unlock(),
170 * in which case we should observe @node->locked becoming
171 * true.
172 */
173 if (smp_load_acquire(&node->locked))
174 return true;
175
176 cpu_relax();
177
178 /*
179 * Or we race against a concurrent unqueue()'s step-B, in which
180 * case its step-C will write us a new @node->prev pointer.
181 */
182 prev = READ_ONCE(node->prev);
183 }
184
185 /*
186 * Step - B -- stabilize @next
187 *
188 * Similar to unlock(), wait for @node->next or move @lock from @node
189 * back to @prev.
190 */
191
192 next = osq_wait_next(lock, node, prev->cpu);
193 if (!next)
194 return false;
195
196 /*
197 * Step - C -- unlink
198 *
199 * @prev is stable because its still waiting for a new @prev->next
200 * pointer, @next is stable because our @node->next pointer is NULL and
201 * it will wait in Step-A.
202 */
203
204 WRITE_ONCE(next->prev, prev);
205 WRITE_ONCE(prev->next, next);
206
207 return false;
208 }
209
osq_unlock(struct optimistic_spin_queue * lock)210 void osq_unlock(struct optimistic_spin_queue *lock)
211 {
212 struct optimistic_spin_node *node, *next;
213 int curr = encode_cpu(smp_processor_id());
214
215 /*
216 * Fast path for the uncontended case.
217 */
218 if (likely(atomic_cmpxchg_release(&lock->tail, curr,
219 OSQ_UNLOCKED_VAL) == curr))
220 return;
221
222 /*
223 * Second most likely case.
224 */
225 node = this_cpu_ptr(&osq_node);
226 next = xchg(&node->next, NULL);
227 if (next) {
228 WRITE_ONCE(next->locked, 1);
229 return;
230 }
231
232 next = osq_wait_next(lock, node, OSQ_UNLOCKED_VAL);
233 if (next)
234 WRITE_ONCE(next->locked, 1);
235 }
236