1 | //
|
---|
2 | // Cforall Version 1.0.0 Copyright (C) 2021 University of Waterloo
|
---|
3 | //
|
---|
4 | // The contents of this file are covered under the licence agreement in the
|
---|
5 | // file "LICENCE" distributed with Cforall.
|
---|
6 | //
|
---|
7 | // locks.hfa -- PUBLIC
|
---|
8 | // Runtime locks that used with the runtime thread system.
|
---|
9 | //
|
---|
10 | // Author : Colby Alexander Parsons
|
---|
11 | // Created On : Thu Jan 21 19:46:50 2021
|
---|
12 | // Last Modified By :
|
---|
13 | // Last Modified On :
|
---|
14 | // Update Count :
|
---|
15 | //
|
---|
16 |
|
---|
17 | #pragma once
|
---|
18 |
|
---|
19 | #include <stdbool.h>
|
---|
20 | #include <stdio.h>
|
---|
21 |
|
---|
22 | #include "bits/weakso_locks.hfa"
|
---|
23 | #include "containers/lockfree.hfa"
|
---|
24 | #include "containers/list.hfa"
|
---|
25 |
|
---|
26 | #include "limits.hfa"
|
---|
27 | #include "thread.hfa"
|
---|
28 |
|
---|
29 | #include "time_t.hfa"
|
---|
30 | #include "time.hfa"
|
---|
31 |
|
---|
32 | #include "select.hfa"
|
---|
33 |
|
---|
34 | #include <fstream.hfa>
|
---|
35 |
|
---|
36 | // futex headers
|
---|
37 | #include <linux/futex.h> /* Definition of FUTEX_* constants */
|
---|
38 | #include <sys/syscall.h> /* Definition of SYS_* constants */
|
---|
39 | #include <unistd.h>
|
---|
40 |
|
---|
41 | typedef void (*__cfa_pre_park)( void * );
|
---|
42 |
|
---|
43 | static inline void pre_park_noop( void * ) {}
|
---|
44 |
|
---|
45 | //-----------------------------------------------------------------------------
|
---|
46 | // is_blocking_lock
|
---|
47 | forall( L & | sized(L) )
|
---|
48 | trait is_blocking_lock {
|
---|
49 | // For synchronization locks to use when acquiring
|
---|
50 | void on_notify( L &, struct thread$ * );
|
---|
51 |
|
---|
52 | // For synchronization locks to use when releasing
|
---|
53 | size_t on_wait( L &, __cfa_pre_park pp_fn, void * pp_datum );
|
---|
54 |
|
---|
55 | // to set recursion count after getting signalled;
|
---|
56 | void on_wakeup( L &, size_t recursion );
|
---|
57 | };
|
---|
58 |
|
---|
59 | static inline void pre_park_then_park( __cfa_pre_park pp_fn, void * pp_datum ) {
|
---|
60 | pp_fn( pp_datum );
|
---|
61 | park();
|
---|
62 | }
|
---|
63 |
|
---|
64 | // macros for default routine impls for is_blocking_lock trait that do not wait-morph
|
---|
65 |
|
---|
66 | #define DEFAULT_ON_NOTIFY( lock_type ) \
|
---|
67 | static inline void on_notify( lock_type & this, thread$ * t ){ unpark(t); }
|
---|
68 |
|
---|
69 | #define DEFAULT_ON_WAIT( lock_type ) \
|
---|
70 | static inline size_t on_wait( lock_type & this, __cfa_pre_park pp_fn, void * pp_datum ) { \
|
---|
71 | unlock( this ); \
|
---|
72 | pre_park_then_park( pp_fn, pp_datum ); \
|
---|
73 | return 0; \
|
---|
74 | }
|
---|
75 |
|
---|
76 | // on_wakeup impl if lock should be reacquired after waking up
|
---|
77 | #define DEFAULT_ON_WAKEUP_REACQ( lock_type ) \
|
---|
78 | static inline void on_wakeup( lock_type & this, size_t recursion ) { lock( this ); }
|
---|
79 |
|
---|
80 | // on_wakeup impl if lock will not be reacquired after waking up
|
---|
81 | #define DEFAULT_ON_WAKEUP_NO_REACQ( lock_type ) \
|
---|
82 | static inline void on_wakeup( lock_type & this, size_t recursion ) {}
|
---|
83 |
|
---|
84 |
|
---|
85 |
|
---|
86 | //-----------------------------------------------------------------------------
|
---|
87 | // Semaphore
|
---|
88 | struct semaphore {
|
---|
89 | __spinlock_t lock;
|
---|
90 | int count;
|
---|
91 | __queue_t(thread$) waiting;
|
---|
92 | };
|
---|
93 |
|
---|
94 | void ?{}(semaphore & this, int count = 1);
|
---|
95 | void ^?{}(semaphore & this);
|
---|
96 | bool P (semaphore & this);
|
---|
97 | bool V (semaphore & this);
|
---|
98 | bool V (semaphore & this, unsigned count);
|
---|
99 | thread$ * V (semaphore & this, bool );
|
---|
100 |
|
---|
101 | //----------
|
---|
102 | struct single_acquisition_lock {
|
---|
103 | inline blocking_lock;
|
---|
104 | };
|
---|
105 |
|
---|
106 | static inline void ?{}( single_acquisition_lock & this ) {((blocking_lock &)this){ false, false };}
|
---|
107 | static inline void ^?{}( single_acquisition_lock & this ) {}
|
---|
108 | static inline void lock ( single_acquisition_lock & this ) { lock ( (blocking_lock &)this ); }
|
---|
109 | static inline bool try_lock ( single_acquisition_lock & this ) { return try_lock( (blocking_lock &)this ); }
|
---|
110 | static inline void unlock ( single_acquisition_lock & this ) { unlock ( (blocking_lock &)this ); }
|
---|
111 | static inline size_t on_wait ( single_acquisition_lock & this, __cfa_pre_park pp_fn, void * pp_datum ) { return on_wait ( (blocking_lock &)this, pp_fn, pp_datum ); }
|
---|
112 | static inline void on_wakeup( single_acquisition_lock & this, size_t v ) { on_wakeup ( (blocking_lock &)this, v ); }
|
---|
113 | static inline void on_notify( single_acquisition_lock & this, struct thread$ * t ) { on_notify( (blocking_lock &)this, t ); }
|
---|
114 | static inline bool register_select( single_acquisition_lock & this, select_node & node ) { return register_select( (blocking_lock &)this, node ); }
|
---|
115 | static inline bool unregister_select( single_acquisition_lock & this, select_node & node ) { return unregister_select( (blocking_lock &)this, node ); }
|
---|
116 | static inline bool on_selected( single_acquisition_lock & this, select_node & node ) { return on_selected( (blocking_lock &)this, node ); }
|
---|
117 |
|
---|
118 | //----------
|
---|
119 | struct owner_lock {
|
---|
120 | inline blocking_lock;
|
---|
121 | };
|
---|
122 |
|
---|
123 | static inline void ?{}( owner_lock & this ) {((blocking_lock &)this){ true, true };}
|
---|
124 | static inline void ^?{}( owner_lock & this ) {}
|
---|
125 | static inline void lock ( owner_lock & this ) { lock ( (blocking_lock &)this ); }
|
---|
126 | static inline bool try_lock ( owner_lock & this ) { return try_lock( (blocking_lock &)this ); }
|
---|
127 | static inline void unlock ( owner_lock & this ) { unlock ( (blocking_lock &)this ); }
|
---|
128 | static inline size_t on_wait ( owner_lock & this, __cfa_pre_park pp_fn, void * pp_datum ) { return on_wait ( (blocking_lock &)this, pp_fn, pp_datum ); }
|
---|
129 | static inline void on_wakeup( owner_lock & this, size_t v ) { on_wakeup ( (blocking_lock &)this, v ); }
|
---|
130 | static inline void on_notify( owner_lock & this, struct thread$ * t ) { on_notify( (blocking_lock &)this, t ); }
|
---|
131 | static inline bool register_select( owner_lock & this, select_node & node ) { return register_select( (blocking_lock &)this, node ); }
|
---|
132 | static inline bool unregister_select( owner_lock & this, select_node & node ) { return unregister_select( (blocking_lock &)this, node ); }
|
---|
133 | static inline bool on_selected( owner_lock & this, select_node & node ) { return on_selected( (blocking_lock &)this, node ); }
|
---|
134 |
|
---|
135 | //-----------------------------------------------------------------------------
|
---|
136 | // MCS Lock
|
---|
137 | struct mcs_node {
|
---|
138 | mcs_node * volatile next;
|
---|
139 | single_sem sem;
|
---|
140 | };
|
---|
141 |
|
---|
142 | static inline void ?{}(mcs_node & this) { this.next = 0p; }
|
---|
143 |
|
---|
144 | static inline mcs_node * volatile & ?`next ( mcs_node * node ) {
|
---|
145 | return node->next;
|
---|
146 | }
|
---|
147 |
|
---|
148 | struct mcs_lock {
|
---|
149 | mcs_queue(mcs_node) queue;
|
---|
150 | };
|
---|
151 |
|
---|
152 | static inline void lock(mcs_lock & l, mcs_node & n) {
|
---|
153 | if(push(l.queue, &n))
|
---|
154 | wait(n.sem);
|
---|
155 | }
|
---|
156 |
|
---|
157 | static inline void unlock(mcs_lock & l, mcs_node & n) {
|
---|
158 | mcs_node * next = advance(l.queue, &n);
|
---|
159 | if(next) post(next->sem);
|
---|
160 | }
|
---|
161 |
|
---|
162 | //-----------------------------------------------------------------------------
|
---|
163 | // MCS Spin Lock
|
---|
164 | // - No recursive acquisition
|
---|
165 | // - Needs to be released by owner
|
---|
166 |
|
---|
167 | struct mcs_spin_node {
|
---|
168 | mcs_spin_node * volatile next;
|
---|
169 | volatile bool locked;
|
---|
170 | };
|
---|
171 |
|
---|
172 | struct mcs_spin_queue {
|
---|
173 | mcs_spin_node * volatile tail;
|
---|
174 | };
|
---|
175 |
|
---|
176 | static inline void ?{}(mcs_spin_node & this) { this.next = 0p; this.locked = true; }
|
---|
177 |
|
---|
178 | struct mcs_spin_lock {
|
---|
179 | mcs_spin_queue queue;
|
---|
180 | };
|
---|
181 |
|
---|
182 | static inline void lock(mcs_spin_lock & l, mcs_spin_node & n) {
|
---|
183 | n.locked = true;
|
---|
184 | mcs_spin_node * prev = __atomic_exchange_n(&l.queue.tail, &n, __ATOMIC_SEQ_CST);
|
---|
185 | if( prev == 0p ) return;
|
---|
186 | prev->next = &n;
|
---|
187 | while( __atomic_load_n(&n.locked, __ATOMIC_RELAXED) ) Pause();
|
---|
188 | }
|
---|
189 |
|
---|
190 | static inline void unlock(mcs_spin_lock & l, mcs_spin_node & n) {
|
---|
191 | mcs_spin_node * n_ptr = &n;
|
---|
192 | if (__atomic_compare_exchange_n(&l.queue.tail, &n_ptr, 0p, false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) return;
|
---|
193 | while (__atomic_load_n(&n.next, __ATOMIC_RELAXED) == 0p) Pause();
|
---|
194 | n.next->locked = false;
|
---|
195 | }
|
---|
196 |
|
---|
197 | //-----------------------------------------------------------------------------
|
---|
198 | // futex_mutex
|
---|
199 |
|
---|
200 | // - Kernel thd blocking alternative to the spinlock
|
---|
201 | // - No ownership (will deadlock on reacq)
|
---|
202 | // - no reacq on wakeup
|
---|
203 | struct futex_mutex {
|
---|
204 | // lock state any state other than UNLOCKED is locked
|
---|
205 | // enum LockState { UNLOCKED = 0, UNCONTENDED = 1, CONTENDED = 2 };
|
---|
206 |
|
---|
207 | // stores a lock state
|
---|
208 | int val;
|
---|
209 | };
|
---|
210 |
|
---|
211 | // to use for FUTEX_WAKE and FUTEX_WAIT (other futex calls will need more params)
|
---|
212 | static inline int futex(int *uaddr, int futex_op, int val) {
|
---|
213 | return syscall(SYS_futex, uaddr, futex_op, val, NULL, NULL, 0);
|
---|
214 | }
|
---|
215 |
|
---|
216 | static inline void ?{}( futex_mutex & this ) with(this) { val = 0; }
|
---|
217 |
|
---|
218 | static inline bool internal_try_lock( futex_mutex & this, int & compare_val) with(this) {
|
---|
219 | return __atomic_compare_exchange_n((int*)&val, (int*)&compare_val, 1, false, __ATOMIC_ACQUIRE, __ATOMIC_ACQUIRE);
|
---|
220 | }
|
---|
221 |
|
---|
222 | static inline int internal_exchange( futex_mutex & this ) with(this) {
|
---|
223 | return __atomic_exchange_n((int*)&val, 2, __ATOMIC_ACQUIRE);
|
---|
224 | }
|
---|
225 |
|
---|
226 | // if this is called recursively IT WILL DEADLOCK!!!!!
|
---|
227 | static inline void lock( futex_mutex & this ) with(this) {
|
---|
228 | int state;
|
---|
229 |
|
---|
230 | for( int spin = 4; spin < 1024; spin += spin) {
|
---|
231 | state = 0;
|
---|
232 | // if unlocked, lock and return
|
---|
233 | if (internal_try_lock(this, state)) return;
|
---|
234 | if (2 == state) break;
|
---|
235 | for (int i = 0; i < spin; i++) Pause();
|
---|
236 | }
|
---|
237 |
|
---|
238 | // if not in contended state, set to be in contended state
|
---|
239 | if (state != 2) state = internal_exchange(this);
|
---|
240 |
|
---|
241 | // block and spin until we win the lock
|
---|
242 | while (state != 0) {
|
---|
243 | futex((int*)&val, FUTEX_WAIT, 2); // if val is not 2 this returns with EWOULDBLOCK
|
---|
244 | state = internal_exchange(this);
|
---|
245 | }
|
---|
246 | }
|
---|
247 |
|
---|
248 | static inline void unlock(futex_mutex & this) with(this) {
|
---|
249 | // if uncontended do atomic unlock and then return
|
---|
250 | if (__atomic_exchange_n(&val, 0, __ATOMIC_RELEASE) == 1) return;
|
---|
251 |
|
---|
252 | // otherwise threads are blocked so we must wake one
|
---|
253 | futex((int *)&val, FUTEX_WAKE, 1);
|
---|
254 | }
|
---|
255 |
|
---|
256 | DEFAULT_ON_NOTIFY( futex_mutex )
|
---|
257 | DEFAULT_ON_WAIT( futex_mutex )
|
---|
258 | DEFAULT_ON_WAKEUP_NO_REACQ( futex_mutex )
|
---|
259 |
|
---|
260 | //-----------------------------------------------------------------------------
|
---|
261 | // go_mutex
|
---|
262 |
|
---|
263 | // - Kernel thd blocking alternative to the spinlock
|
---|
264 | // - No ownership (will deadlock on reacq)
|
---|
265 | // - Golang's flavour of mutex
|
---|
266 | // - Impl taken from Golang: src/runtime/lock_futex.go
|
---|
267 | struct go_mutex {
|
---|
268 | // lock state any state other than UNLOCKED is locked
|
---|
269 | // enum LockState { UNLOCKED = 0, LOCKED = 1, SLEEPING = 2 };
|
---|
270 |
|
---|
271 | // stores a lock state
|
---|
272 | int val;
|
---|
273 | };
|
---|
274 | static inline void ?{}( go_mutex & this ) with(this) { val = 0; }
|
---|
275 | // static inline void ?{}( go_mutex & this, go_mutex this2 ) = void; // these don't compile correctly at the moment so they should be omitted
|
---|
276 | // static inline void ?=?( go_mutex & this, go_mutex this2 ) = void;
|
---|
277 |
|
---|
278 | static inline bool internal_try_lock(go_mutex & this, int & compare_val, int new_val ) with(this) {
|
---|
279 | return __atomic_compare_exchange_n((int*)&val, (int*)&compare_val, new_val, false, __ATOMIC_ACQUIRE, __ATOMIC_ACQUIRE);
|
---|
280 | }
|
---|
281 |
|
---|
282 | static inline int internal_exchange(go_mutex & this, int swap ) with(this) {
|
---|
283 | return __atomic_exchange_n((int*)&val, swap, __ATOMIC_ACQUIRE);
|
---|
284 | }
|
---|
285 |
|
---|
286 | // if this is called recursively IT WILL DEADLOCK!!!!!
|
---|
287 | static inline void lock( go_mutex & this ) with( this ) {
|
---|
288 | int state, init_state;
|
---|
289 |
|
---|
290 | // speculative grab
|
---|
291 | state = internal_exchange(this, 1);
|
---|
292 | if ( !state ) return; // state == 0
|
---|
293 | init_state = state;
|
---|
294 | for (;;) {
|
---|
295 | for( int i = 0; i < 4; i++ ) {
|
---|
296 | while( !val ) { // lock unlocked
|
---|
297 | state = 0;
|
---|
298 | if ( internal_try_lock( this, state, init_state ) ) return;
|
---|
299 | }
|
---|
300 | for (int i = 0; i < 30; i++) Pause();
|
---|
301 | }
|
---|
302 |
|
---|
303 | while( !val ) { // lock unlocked
|
---|
304 | state = 0;
|
---|
305 | if ( internal_try_lock( this, state, init_state ) ) return;
|
---|
306 | }
|
---|
307 | sched_yield();
|
---|
308 |
|
---|
309 | // if not in contended state, set to be in contended state
|
---|
310 | state = internal_exchange( this, 2 );
|
---|
311 | if ( !state ) return; // state == 0
|
---|
312 | init_state = 2;
|
---|
313 | futex( (int*)&val, FUTEX_WAIT, 2 ); // if val is not 2 this returns with EWOULDBLOCK
|
---|
314 | }
|
---|
315 | }
|
---|
316 |
|
---|
317 | static inline void unlock( go_mutex & this ) with(this) {
|
---|
318 | // if uncontended do atomic unlock and then return
|
---|
319 | if ( __atomic_exchange_n(&val, 0, __ATOMIC_RELEASE) == 1 ) return;
|
---|
320 |
|
---|
321 | // otherwise threads are blocked so we must wake one
|
---|
322 | futex( (int *)&val, FUTEX_WAKE, 1 );
|
---|
323 | }
|
---|
324 |
|
---|
325 | DEFAULT_ON_NOTIFY( go_mutex )
|
---|
326 | DEFAULT_ON_WAIT( go_mutex )
|
---|
327 | DEFAULT_ON_WAKEUP_NO_REACQ( go_mutex )
|
---|
328 |
|
---|
329 | //-----------------------------------------------------------------------------
|
---|
330 | // Exponential backoff then block lock
|
---|
331 | struct exp_backoff_then_block_lock {
|
---|
332 | // Spin lock used for mutual exclusion
|
---|
333 | __spinlock_t spinlock;
|
---|
334 |
|
---|
335 | // List of blocked threads
|
---|
336 | dlist( thread$ ) blocked_threads;
|
---|
337 |
|
---|
338 | // Used for comparing and exchanging
|
---|
339 | volatile size_t lock_value;
|
---|
340 | };
|
---|
341 |
|
---|
342 | static inline void ?{}( exp_backoff_then_block_lock & this ) {
|
---|
343 | this.spinlock{};
|
---|
344 | this.blocked_threads{};
|
---|
345 | this.lock_value = 0;
|
---|
346 | }
|
---|
347 | static inline void ?{}( exp_backoff_then_block_lock & this, exp_backoff_then_block_lock this2 ) = void;
|
---|
348 | static inline void ?=?( exp_backoff_then_block_lock & this, exp_backoff_then_block_lock this2 ) = void;
|
---|
349 |
|
---|
350 | static inline void ^?{}( exp_backoff_then_block_lock & this ){}
|
---|
351 |
|
---|
352 | static inline bool internal_try_lock( exp_backoff_then_block_lock & this, size_t & compare_val ) with(this) {
|
---|
353 | return __atomic_compare_exchange_n(&lock_value, &compare_val, 1, false, __ATOMIC_ACQUIRE, __ATOMIC_RELAXED);
|
---|
354 | }
|
---|
355 |
|
---|
356 | static inline bool try_lock( exp_backoff_then_block_lock & this ) { size_t compare_val = 0; return internal_try_lock( this, compare_val ); }
|
---|
357 |
|
---|
358 | static inline bool try_lock_contention( exp_backoff_then_block_lock & this ) with(this) {
|
---|
359 | return !__atomic_exchange_n( &lock_value, 2, __ATOMIC_ACQUIRE );
|
---|
360 | }
|
---|
361 |
|
---|
362 | static inline bool block( exp_backoff_then_block_lock & this ) with(this) {
|
---|
363 | lock( spinlock __cfaabi_dbg_ctx2 );
|
---|
364 | if (__atomic_load_n( &lock_value, __ATOMIC_SEQ_CST) != 2) {
|
---|
365 | unlock( spinlock );
|
---|
366 | return true;
|
---|
367 | }
|
---|
368 | insert_last( blocked_threads, *active_thread() );
|
---|
369 | unlock( spinlock );
|
---|
370 | park( );
|
---|
371 | return true;
|
---|
372 | }
|
---|
373 |
|
---|
374 | static inline void lock( exp_backoff_then_block_lock & this ) with(this) {
|
---|
375 | size_t compare_val = 0;
|
---|
376 | int spin = 4;
|
---|
377 |
|
---|
378 | // linear backoff
|
---|
379 | for( ;; ) {
|
---|
380 | compare_val = 0;
|
---|
381 | if (internal_try_lock(this, compare_val)) return;
|
---|
382 | if (2 == compare_val) break;
|
---|
383 | for (int i = 0; i < spin; i++) Pause();
|
---|
384 | if (spin >= 1024) break;
|
---|
385 | spin += spin;
|
---|
386 | }
|
---|
387 |
|
---|
388 | if(2 != compare_val && try_lock_contention(this)) return;
|
---|
389 | // block until signalled
|
---|
390 | while (block(this)) if(try_lock_contention(this)) return;
|
---|
391 | }
|
---|
392 |
|
---|
393 | static inline void unlock( exp_backoff_then_block_lock & this ) with(this) {
|
---|
394 | if (__atomic_exchange_n(&lock_value, 0, __ATOMIC_RELEASE) == 1) return;
|
---|
395 | lock( spinlock __cfaabi_dbg_ctx2 );
|
---|
396 | thread$ * t = &try_pop_front( blocked_threads );
|
---|
397 | unlock( spinlock );
|
---|
398 | unpark( t );
|
---|
399 | }
|
---|
400 |
|
---|
401 | DEFAULT_ON_NOTIFY( exp_backoff_then_block_lock )
|
---|
402 | DEFAULT_ON_WAIT( exp_backoff_then_block_lock )
|
---|
403 | DEFAULT_ON_WAKEUP_REACQ( exp_backoff_then_block_lock )
|
---|
404 |
|
---|
405 | //-----------------------------------------------------------------------------
|
---|
406 | // Fast Block Lock
|
---|
407 |
|
---|
408 | // minimal blocking lock
|
---|
409 | // - No reacquire for cond var
|
---|
410 | // - No recursive acquisition
|
---|
411 | // - No ownership
|
---|
412 | struct fast_block_lock {
|
---|
413 | // List of blocked threads
|
---|
414 | dlist( thread$ ) blocked_threads;
|
---|
415 |
|
---|
416 | // Spin lock used for mutual exclusion
|
---|
417 | __spinlock_t lock;
|
---|
418 |
|
---|
419 | // flag showing if lock is held
|
---|
420 | bool held:1;
|
---|
421 | };
|
---|
422 |
|
---|
423 | static inline void ?{}( fast_block_lock & this ) with(this) {
|
---|
424 | lock{};
|
---|
425 | blocked_threads{};
|
---|
426 | held = false;
|
---|
427 | }
|
---|
428 | static inline void ^?{}( fast_block_lock & this ) {}
|
---|
429 | static inline void ?{}( fast_block_lock & this, fast_block_lock this2 ) = void;
|
---|
430 | static inline void ?=?( fast_block_lock & this, fast_block_lock this2 ) = void;
|
---|
431 |
|
---|
432 | // if this is called recursively IT WILL DEADLOCK!!!!!
|
---|
433 | static inline void lock( fast_block_lock & this ) with(this) {
|
---|
434 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
435 | if ( held ) {
|
---|
436 | insert_last( blocked_threads, *active_thread() );
|
---|
437 | unlock( lock );
|
---|
438 | park( );
|
---|
439 | return;
|
---|
440 | }
|
---|
441 | held = true;
|
---|
442 | unlock( lock );
|
---|
443 | }
|
---|
444 |
|
---|
445 | static inline void unlock( fast_block_lock & this ) with(this) {
|
---|
446 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
447 | /* paranoid */ verifyf( held != false, "Attempt to release lock %p that isn't held", &this );
|
---|
448 | thread$ * t = &try_pop_front( blocked_threads );
|
---|
449 | held = ( t ? true : false );
|
---|
450 | unpark( t );
|
---|
451 | unlock( lock );
|
---|
452 | }
|
---|
453 |
|
---|
454 | static inline void on_notify( fast_block_lock & this, struct thread$ * t ) with(this) {
|
---|
455 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
456 | insert_last( blocked_threads, *t );
|
---|
457 | unlock( lock );
|
---|
458 | }
|
---|
459 | DEFAULT_ON_WAIT( fast_block_lock )
|
---|
460 | DEFAULT_ON_WAKEUP_NO_REACQ( fast_block_lock )
|
---|
461 |
|
---|
462 | //-----------------------------------------------------------------------------
|
---|
463 | // simple_owner_lock
|
---|
464 |
|
---|
465 | // pthread owner lock
|
---|
466 | // - reacquire for cond var
|
---|
467 | // - recursive acquisition
|
---|
468 | // - ownership
|
---|
469 | struct simple_owner_lock {
|
---|
470 | // List of blocked threads
|
---|
471 | dlist( select_node ) blocked_threads;
|
---|
472 |
|
---|
473 | // Spin lock used for mutual exclusion
|
---|
474 | __spinlock_t lock;
|
---|
475 |
|
---|
476 | // owner showing if lock is held
|
---|
477 | struct thread$ * owner;
|
---|
478 |
|
---|
479 | size_t recursion_count;
|
---|
480 | };
|
---|
481 |
|
---|
482 | static inline void ?{}( simple_owner_lock & this ) with(this) {
|
---|
483 | lock{};
|
---|
484 | blocked_threads{};
|
---|
485 | owner = 0p;
|
---|
486 | recursion_count = 0;
|
---|
487 | }
|
---|
488 | static inline void ^?{}( simple_owner_lock & this ) {}
|
---|
489 | static inline void ?{}( simple_owner_lock & this, simple_owner_lock this2 ) = void;
|
---|
490 | static inline void ?=?( simple_owner_lock & this, simple_owner_lock this2 ) = void;
|
---|
491 |
|
---|
492 | static inline void lock( simple_owner_lock & this ) with(this) {
|
---|
493 | if ( owner == active_thread() ) {
|
---|
494 | recursion_count++;
|
---|
495 | return;
|
---|
496 | }
|
---|
497 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
498 |
|
---|
499 | if ( owner != 0p ) {
|
---|
500 | select_node node;
|
---|
501 | insert_last( blocked_threads, node );
|
---|
502 | unlock( lock );
|
---|
503 | park( );
|
---|
504 | return;
|
---|
505 | }
|
---|
506 | owner = active_thread();
|
---|
507 | recursion_count = 1;
|
---|
508 | unlock( lock );
|
---|
509 | }
|
---|
510 |
|
---|
511 | static inline void pop_node( simple_owner_lock & this ) with(this) {
|
---|
512 | __handle_waituntil_OR( blocked_threads );
|
---|
513 | select_node * node = &try_pop_front( blocked_threads );
|
---|
514 | if ( node ) {
|
---|
515 | owner = node->blocked_thread;
|
---|
516 | recursion_count = 1;
|
---|
517 | // if ( !node->clause_status || __make_select_node_available( *node ) ) unpark( node->blocked_thread );
|
---|
518 | wake_one( blocked_threads, *node );
|
---|
519 | } else {
|
---|
520 | owner = 0p;
|
---|
521 | recursion_count = 0;
|
---|
522 | }
|
---|
523 | }
|
---|
524 |
|
---|
525 | static inline void unlock( simple_owner_lock & this ) with(this) {
|
---|
526 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
527 | /* paranoid */ verifyf( owner != 0p, "Attempt to release lock %p that isn't held", &this );
|
---|
528 | /* paranoid */ verifyf( owner == active_thread(), "Thread %p other than the owner %p attempted to release owner lock %p", owner, active_thread(), &this );
|
---|
529 | // if recursion count is zero release lock and set new owner if one is waiting
|
---|
530 | recursion_count--;
|
---|
531 | if ( recursion_count == 0 ) {
|
---|
532 | pop_node( this );
|
---|
533 | }
|
---|
534 | unlock( lock );
|
---|
535 | }
|
---|
536 |
|
---|
537 | static inline void on_notify( simple_owner_lock & this, thread$ * t ) with(this) {
|
---|
538 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
539 | // lock held
|
---|
540 | if ( owner != 0p ) {
|
---|
541 | insert_last( blocked_threads, *(select_node *)t->link_node );
|
---|
542 | }
|
---|
543 | // lock not held
|
---|
544 | else {
|
---|
545 | owner = t;
|
---|
546 | recursion_count = 1;
|
---|
547 | unpark( t );
|
---|
548 | }
|
---|
549 | unlock( lock );
|
---|
550 | }
|
---|
551 |
|
---|
552 | static inline size_t on_wait( simple_owner_lock & this, __cfa_pre_park pp_fn, void * pp_datum ) with(this) {
|
---|
553 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
554 | /* paranoid */ verifyf( owner != 0p, "Attempt to release lock %p that isn't held", &this );
|
---|
555 | /* paranoid */ verifyf( owner == active_thread(), "Thread %p other than the owner %p attempted to release owner lock %p", owner, active_thread(), &this );
|
---|
556 |
|
---|
557 | size_t ret = recursion_count;
|
---|
558 |
|
---|
559 | pop_node( this );
|
---|
560 |
|
---|
561 | select_node node;
|
---|
562 | active_thread()->link_node = (void *)&node;
|
---|
563 | unlock( lock );
|
---|
564 |
|
---|
565 | pre_park_then_park( pp_fn, pp_datum );
|
---|
566 |
|
---|
567 | return ret;
|
---|
568 | }
|
---|
569 |
|
---|
570 | static inline void on_wakeup( simple_owner_lock & this, size_t recursion ) with(this) { recursion_count = recursion; }
|
---|
571 |
|
---|
572 | // waituntil() support
|
---|
573 | static inline bool register_select( simple_owner_lock & this, select_node & node ) with(this) {
|
---|
574 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
575 |
|
---|
576 | // check if we can complete operation. If so race to establish winner in special OR case
|
---|
577 | if ( !node.park_counter && ( owner == active_thread() || owner == 0p ) ) {
|
---|
578 | if ( !__make_select_node_available( node ) ) { // we didn't win the race so give up on registering
|
---|
579 | unlock( lock );
|
---|
580 | return false;
|
---|
581 | }
|
---|
582 | }
|
---|
583 |
|
---|
584 | if ( owner == active_thread() ) {
|
---|
585 | recursion_count++;
|
---|
586 | if ( node.park_counter ) __make_select_node_available( node );
|
---|
587 | unlock( lock );
|
---|
588 | return true;
|
---|
589 | }
|
---|
590 |
|
---|
591 | if ( owner != 0p ) {
|
---|
592 | insert_last( blocked_threads, node );
|
---|
593 | unlock( lock );
|
---|
594 | return false;
|
---|
595 | }
|
---|
596 |
|
---|
597 | owner = active_thread();
|
---|
598 | recursion_count = 1;
|
---|
599 |
|
---|
600 | if ( node.park_counter ) __make_select_node_available( node );
|
---|
601 | unlock( lock );
|
---|
602 | return true;
|
---|
603 | }
|
---|
604 |
|
---|
605 | static inline bool unregister_select( simple_owner_lock & this, select_node & node ) with(this) {
|
---|
606 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
607 | if ( node`isListed ) {
|
---|
608 | remove( node );
|
---|
609 | unlock( lock );
|
---|
610 | return false;
|
---|
611 | }
|
---|
612 |
|
---|
613 | if ( owner == active_thread() ) {
|
---|
614 | recursion_count--;
|
---|
615 | if ( recursion_count == 0 ) {
|
---|
616 | pop_node( this );
|
---|
617 | }
|
---|
618 | }
|
---|
619 | unlock( lock );
|
---|
620 | return false;
|
---|
621 | }
|
---|
622 |
|
---|
623 | static inline bool on_selected( simple_owner_lock & this, select_node & node ) { return true; }
|
---|
624 |
|
---|
625 |
|
---|
626 | //-----------------------------------------------------------------------------
|
---|
627 | // Spin Queue Lock
|
---|
628 |
|
---|
629 | // - No reacquire for cond var
|
---|
630 | // - No recursive acquisition
|
---|
631 | // - No ownership
|
---|
632 | // - spin lock with no locking/atomics in unlock
|
---|
633 | struct spin_queue_lock {
|
---|
634 | // Spin lock used for mutual exclusion
|
---|
635 | mcs_spin_lock lock;
|
---|
636 |
|
---|
637 | // flag showing if lock is held
|
---|
638 | volatile bool held;
|
---|
639 | };
|
---|
640 |
|
---|
641 | static inline void ?{}( spin_queue_lock & this ) with(this) {
|
---|
642 | lock{};
|
---|
643 | held = false;
|
---|
644 | }
|
---|
645 | static inline void ^?{}( spin_queue_lock & this ) {}
|
---|
646 | static inline void ?{}( spin_queue_lock & this, spin_queue_lock this2 ) = void;
|
---|
647 | static inline void ?=?( spin_queue_lock & this, spin_queue_lock this2 ) = void;
|
---|
648 |
|
---|
649 | // if this is called recursively IT WILL DEADLOCK!
|
---|
650 | static inline void lock( spin_queue_lock & this ) with(this) {
|
---|
651 | mcs_spin_node node;
|
---|
652 | lock( lock, node );
|
---|
653 | while(__atomic_load_n(&held, __ATOMIC_SEQ_CST)) Pause();
|
---|
654 | __atomic_store_n(&held, true, __ATOMIC_SEQ_CST);
|
---|
655 | unlock( lock, node );
|
---|
656 | }
|
---|
657 |
|
---|
658 | static inline void unlock( spin_queue_lock & this ) with(this) {
|
---|
659 | __atomic_store_n(&held, false, __ATOMIC_RELEASE);
|
---|
660 | }
|
---|
661 |
|
---|
662 | DEFAULT_ON_NOTIFY( spin_queue_lock )
|
---|
663 | DEFAULT_ON_WAIT( spin_queue_lock )
|
---|
664 | DEFAULT_ON_WAKEUP_REACQ( spin_queue_lock )
|
---|
665 |
|
---|
666 | //-----------------------------------------------------------------------------
|
---|
667 | // MCS Block Spin Lock
|
---|
668 |
|
---|
669 | // - No reacquire for cond var
|
---|
670 | // - No recursive acquisition
|
---|
671 | // - No ownership
|
---|
672 | // - Blocks but first node spins (like spin queue but blocking for not first thd)
|
---|
673 | struct mcs_block_spin_lock {
|
---|
674 | // Spin lock used for mutual exclusion
|
---|
675 | mcs_lock lock;
|
---|
676 |
|
---|
677 | // flag showing if lock is held
|
---|
678 | volatile bool held;
|
---|
679 | };
|
---|
680 |
|
---|
681 | static inline void ?{}( mcs_block_spin_lock & this ) with(this) {
|
---|
682 | lock{};
|
---|
683 | held = false;
|
---|
684 | }
|
---|
685 | static inline void ^?{}( mcs_block_spin_lock & this ) {}
|
---|
686 | static inline void ?{}( mcs_block_spin_lock & this, mcs_block_spin_lock this2 ) = void;
|
---|
687 | static inline void ?=?( mcs_block_spin_lock & this, mcs_block_spin_lock this2 ) = void;
|
---|
688 |
|
---|
689 | // if this is called recursively IT WILL DEADLOCK!!!!!
|
---|
690 | static inline void lock( mcs_block_spin_lock & this ) with(this) {
|
---|
691 | mcs_node node;
|
---|
692 | lock( lock, node );
|
---|
693 | while(__atomic_load_n(&held, __ATOMIC_SEQ_CST)) Pause();
|
---|
694 | __atomic_store_n(&held, true, __ATOMIC_SEQ_CST);
|
---|
695 | unlock( lock, node );
|
---|
696 | }
|
---|
697 |
|
---|
698 | static inline void unlock(mcs_block_spin_lock & this) with(this) {
|
---|
699 | __atomic_store_n(&held, false, __ATOMIC_SEQ_CST);
|
---|
700 | }
|
---|
701 |
|
---|
702 | DEFAULT_ON_NOTIFY( mcs_block_spin_lock )
|
---|
703 | DEFAULT_ON_WAIT( mcs_block_spin_lock )
|
---|
704 | DEFAULT_ON_WAKEUP_REACQ( mcs_block_spin_lock )
|
---|
705 |
|
---|
706 | //-----------------------------------------------------------------------------
|
---|
707 | // Block Spin Lock
|
---|
708 |
|
---|
709 | // - No reacquire for cond var
|
---|
710 | // - No recursive acquisition
|
---|
711 | // - No ownership
|
---|
712 | // - Blocks but first node spins (like spin queue but blocking for not first thd)
|
---|
713 | struct block_spin_lock {
|
---|
714 | // Spin lock used for mutual exclusion
|
---|
715 | fast_block_lock lock;
|
---|
716 |
|
---|
717 | // flag showing if lock is held
|
---|
718 | volatile bool held;
|
---|
719 | };
|
---|
720 |
|
---|
721 | static inline void ?{}( block_spin_lock & this ) with(this) {
|
---|
722 | lock{};
|
---|
723 | held = false;
|
---|
724 | }
|
---|
725 | static inline void ^?{}( block_spin_lock & this ) {}
|
---|
726 | static inline void ?{}( block_spin_lock & this, block_spin_lock this2 ) = void;
|
---|
727 | static inline void ?=?( block_spin_lock & this, block_spin_lock this2 ) = void;
|
---|
728 |
|
---|
729 | // if this is called recursively IT WILL DEADLOCK!!!!!
|
---|
730 | static inline void lock( block_spin_lock & this ) with(this) {
|
---|
731 | lock( lock );
|
---|
732 | while(__atomic_load_n(&held, __ATOMIC_SEQ_CST)) Pause();
|
---|
733 | __atomic_store_n(&held, true, __ATOMIC_RELEASE);
|
---|
734 | unlock( lock );
|
---|
735 | }
|
---|
736 |
|
---|
737 | static inline void unlock( block_spin_lock & this ) with(this) {
|
---|
738 | __atomic_store_n(&held, false, __ATOMIC_RELEASE);
|
---|
739 | }
|
---|
740 |
|
---|
741 | static inline void on_notify( block_spin_lock & this, struct thread$ * t ) with(this.lock) {
|
---|
742 | // first we acquire internal fast_block_lock
|
---|
743 | lock( lock __cfaabi_dbg_ctx2 );
|
---|
744 | if ( held ) { // if internal fast_block_lock is held
|
---|
745 | insert_last( blocked_threads, *t );
|
---|
746 | unlock( lock );
|
---|
747 | return;
|
---|
748 | }
|
---|
749 | // if internal fast_block_lock is not held
|
---|
750 | held = true;
|
---|
751 | unlock( lock );
|
---|
752 |
|
---|
753 | unpark(t);
|
---|
754 | }
|
---|
755 | DEFAULT_ON_WAIT( block_spin_lock )
|
---|
756 | static inline void on_wakeup( block_spin_lock & this, size_t recursion ) with(this) {
|
---|
757 | // now we acquire the entire block_spin_lock upon waking up
|
---|
758 | while(__atomic_load_n(&held, __ATOMIC_SEQ_CST)) Pause();
|
---|
759 | __atomic_store_n(&held, true, __ATOMIC_RELEASE);
|
---|
760 | unlock( lock ); // Now we release the internal fast_spin_lock
|
---|
761 | }
|
---|
762 |
|
---|
763 | //-----------------------------------------------------------------------------
|
---|
764 | // // info_thread
|
---|
765 | // // the info thread is a wrapper around a thread used
|
---|
766 | // // to store extra data for use in the condition variable
|
---|
767 | forall(L & | is_blocking_lock(L)) {
|
---|
768 | struct info_thread;
|
---|
769 | }
|
---|
770 |
|
---|
771 | //-----------------------------------------------------------------------------
|
---|
772 | // Synchronization Locks
|
---|
773 | forall(L & | is_blocking_lock(L)) {
|
---|
774 |
|
---|
775 | //-----------------------------------------------------------------------------
|
---|
776 | // condition_variable
|
---|
777 |
|
---|
778 | // The multi-tool condition variable
|
---|
779 | // - can pass timeouts to wait for either a signal or timeout
|
---|
780 | // - can wait without passing a lock
|
---|
781 | // - can have waiters reacquire different locks while waiting on the same cond var
|
---|
782 | // - has shadow queue
|
---|
783 | // - can be signalled outside of critical sections with no locks held
|
---|
784 | struct condition_variable {
|
---|
785 | // Spin lock used for mutual exclusion
|
---|
786 | __spinlock_t lock;
|
---|
787 |
|
---|
788 | // List of blocked threads
|
---|
789 | dlist( info_thread(L) ) blocked_threads;
|
---|
790 |
|
---|
791 | // Count of current blocked threads
|
---|
792 | int count;
|
---|
793 | };
|
---|
794 |
|
---|
795 |
|
---|
796 | void ?{}( condition_variable(L) & this );
|
---|
797 | void ^?{}( condition_variable(L) & this );
|
---|
798 |
|
---|
799 | bool notify_one( condition_variable(L) & this );
|
---|
800 | bool notify_all( condition_variable(L) & this );
|
---|
801 |
|
---|
802 | uintptr_t front( condition_variable(L) & this );
|
---|
803 |
|
---|
804 | bool empty ( condition_variable(L) & this );
|
---|
805 | int counter( condition_variable(L) & this );
|
---|
806 |
|
---|
807 | void wait( condition_variable(L) & this );
|
---|
808 | void wait( condition_variable(L) & this, uintptr_t info );
|
---|
809 | bool wait( condition_variable(L) & this, Duration duration );
|
---|
810 | bool wait( condition_variable(L) & this, uintptr_t info, Duration duration );
|
---|
811 |
|
---|
812 | void wait( condition_variable(L) & this, L & l );
|
---|
813 | void wait( condition_variable(L) & this, L & l, uintptr_t info );
|
---|
814 | bool wait( condition_variable(L) & this, L & l, Duration duration );
|
---|
815 | bool wait( condition_variable(L) & this, L & l, uintptr_t info, Duration duration );
|
---|
816 |
|
---|
817 | //-----------------------------------------------------------------------------
|
---|
818 | // fast_cond_var
|
---|
819 |
|
---|
820 | // The trimmed and slim condition variable
|
---|
821 | // - no internal lock so you must hold a lock while using this cond var
|
---|
822 | // - signalling without holding branded lock is UNSAFE!
|
---|
823 | // - only allows usage of one lock, cond var is branded after usage
|
---|
824 |
|
---|
825 | struct fast_cond_var {
|
---|
826 | // List of blocked threads
|
---|
827 | dlist( info_thread(L) ) blocked_threads;
|
---|
828 | #ifdef __CFA_DEBUG__
|
---|
829 | L * lock_used;
|
---|
830 | #endif
|
---|
831 | };
|
---|
832 |
|
---|
833 | void ?{}( fast_cond_var(L) & this );
|
---|
834 | void ^?{}( fast_cond_var(L) & this );
|
---|
835 |
|
---|
836 | bool notify_one( fast_cond_var(L) & this );
|
---|
837 | bool notify_all( fast_cond_var(L) & this );
|
---|
838 |
|
---|
839 | uintptr_t front( fast_cond_var(L) & this );
|
---|
840 | bool empty ( fast_cond_var(L) & this );
|
---|
841 |
|
---|
842 | void wait( fast_cond_var(L) & this, L & l );
|
---|
843 | void wait( fast_cond_var(L) & this, L & l, uintptr_t info );
|
---|
844 |
|
---|
845 |
|
---|
846 | //-----------------------------------------------------------------------------
|
---|
847 | // pthread_cond_var
|
---|
848 | //
|
---|
849 | // - cond var with minimal footprint
|
---|
850 | // - supports operations needed for phthread cond
|
---|
851 |
|
---|
852 | struct pthread_cond_var {
|
---|
853 | dlist( info_thread(L) ) blocked_threads;
|
---|
854 | __spinlock_t lock;
|
---|
855 | };
|
---|
856 |
|
---|
857 | void ?{}( pthread_cond_var(L) & this );
|
---|
858 | void ^?{}( pthread_cond_var(L) & this );
|
---|
859 |
|
---|
860 | bool notify_one( pthread_cond_var(L) & this );
|
---|
861 | bool notify_all( pthread_cond_var(L) & this );
|
---|
862 |
|
---|
863 | uintptr_t front( pthread_cond_var(L) & this );
|
---|
864 | bool empty ( pthread_cond_var(L) & this );
|
---|
865 |
|
---|
866 | void wait( pthread_cond_var(L) & this, L & l );
|
---|
867 | void wait( pthread_cond_var(L) & this, L & l, uintptr_t info );
|
---|
868 | bool wait( pthread_cond_var(L) & this, L & l, timespec t );
|
---|
869 | bool wait( pthread_cond_var(L) & this, L & l, uintptr_t info, timespec t );
|
---|
870 | }
|
---|