10 #include "snapshot-interface.h"
12 #include "clockvector.h"
13 #include "cyclegraph.h"
16 #include "threads-model.h"
19 #define INITIAL_THREAD_ID 0
24 bug_message(const char *str) {
25 const char *fmt = " [BUG] %s\n";
26 msg = (char *)snapshot_malloc(strlen(fmt) + strlen(str));
27 sprintf(msg, fmt, str);
29 ~bug_message() { if (msg) snapshot_free(msg); }
32 void print() { model_print("%s", msg); }
38 * Structure for holding small ModelChecker members that should be snapshotted
40 struct model_snapshot_members {
41 model_snapshot_members() :
42 /* First thread created will have id INITIAL_THREAD_ID */
43 next_thread_id(INITIAL_THREAD_ID),
44 used_sequence_numbers(0),
48 failed_promise(false),
49 too_many_reads(false),
50 no_valid_reads(false),
51 bad_synchronization(false),
55 ~model_snapshot_members() {
56 for (unsigned int i = 0; i < bugs.size(); i++)
61 unsigned int next_thread_id;
62 modelclock_t used_sequence_numbers;
63 ModelAction *next_backtrack;
64 std::vector< bug_message *, SnapshotAlloc<bug_message *> > bugs;
65 struct execution_stats stats;
69 /** @brief Incorrectly-ordered synchronization was made */
70 bool bad_synchronization;
76 /** @brief Constructor */
77 ModelChecker::ModelChecker(struct model_params params) :
78 /* Initialize default scheduler */
80 scheduler(new Scheduler()),
82 earliest_diverge(NULL),
83 action_trace(new action_list_t()),
84 thread_map(new HashTable<int, Thread *, int>()),
85 obj_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
86 lock_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
87 condvar_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
88 obj_thrd_map(new HashTable<void *, std::vector<action_list_t> *, uintptr_t, 4 >()),
89 promises(new std::vector< Promise *, SnapshotAlloc<Promise *> >()),
90 futurevalues(new std::vector< struct PendingFutureValue, SnapshotAlloc<struct PendingFutureValue> >()),
91 pending_rel_seqs(new std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >()),
92 thrd_last_action(new std::vector< ModelAction *, SnapshotAlloc<ModelAction *> >(1)),
93 thrd_last_fence_release(new std::vector< ModelAction *, SnapshotAlloc<ModelAction *> >()),
94 node_stack(new NodeStack()),
95 priv(new struct model_snapshot_members()),
96 mo_graph(new CycleGraph())
98 /* Initialize a model-checker thread, for special ModelActions */
99 model_thread = new Thread(get_next_id());
100 thread_map->put(id_to_int(model_thread->get_id()), model_thread);
103 /** @brief Destructor */
104 ModelChecker::~ModelChecker()
106 for (unsigned int i = 0; i < get_num_threads(); i++)
107 delete thread_map->get(i);
112 delete lock_waiters_map;
113 delete condvar_waiters_map;
116 for (unsigned int i = 0; i < promises->size(); i++)
117 delete (*promises)[i];
120 delete pending_rel_seqs;
122 delete thrd_last_action;
123 delete thrd_last_fence_release;
130 static action_list_t * get_safe_ptr_action(HashTable<const void *, action_list_t *, uintptr_t, 4> * hash, void * ptr)
132 action_list_t *tmp = hash->get(ptr);
134 tmp = new action_list_t();
140 static std::vector<action_list_t> * get_safe_ptr_vect_action(HashTable<void *, std::vector<action_list_t> *, uintptr_t, 4> * hash, void * ptr)
142 std::vector<action_list_t> *tmp = hash->get(ptr);
144 tmp = new std::vector<action_list_t>();
151 * Restores user program to initial state and resets all model-checker data
154 void ModelChecker::reset_to_initial_state()
156 DEBUG("+++ Resetting to initial state +++\n");
157 node_stack->reset_execution();
159 /* Print all model-checker output before rollback */
163 * FIXME: if we utilize partial rollback, we will need to free only
164 * those pending actions which were NOT pending before the rollback
167 for (unsigned int i = 0; i < get_num_threads(); i++)
168 delete get_thread(int_to_id(i))->get_pending();
170 snapshot_backtrack_before(0);
173 /** @return a thread ID for a new Thread */
174 thread_id_t ModelChecker::get_next_id()
176 return priv->next_thread_id++;
179 /** @return the number of user threads created during this execution */
180 unsigned int ModelChecker::get_num_threads() const
182 return priv->next_thread_id;
186 * Must be called from user-thread context (e.g., through the global
187 * thread_current() interface)
189 * @return The currently executing Thread.
191 Thread * ModelChecker::get_current_thread() const
193 return scheduler->get_current_thread();
196 /** @return a sequence number for a new ModelAction */
197 modelclock_t ModelChecker::get_next_seq_num()
199 return ++priv->used_sequence_numbers;
202 Node * ModelChecker::get_curr_node() const
204 return node_stack->get_head();
208 * @brief Choose the next thread to execute.
210 * This function chooses the next thread that should execute. It can force the
211 * adjacency of read/write portions of a RMW action, force THREAD_CREATE to be
212 * followed by a THREAD_START, or it can enforce execution replay/backtracking.
213 * The model-checker may have no preference regarding the next thread (i.e.,
214 * when exploring a new execution ordering), in which case we defer to the
217 * @param curr Optional: The current ModelAction. Only used if non-NULL and it
218 * might guide the choice of next thread (i.e., THREAD_CREATE should be
219 * followed by THREAD_START, or ATOMIC_RMWR followed by ATOMIC_{RMW,RMWC})
220 * @return The next chosen thread to run, if any exist. Or else if no threads
221 * remain to be executed, return NULL.
223 Thread * ModelChecker::get_next_thread(ModelAction *curr)
228 /* Do not split atomic actions. */
230 return get_thread(curr);
231 else if (curr->get_type() == THREAD_CREATE)
232 return curr->get_thread_operand();
236 * Have we completed exploring the preselected path? Then let the
240 return scheduler->select_next_thread();
242 /* Else, we are trying to replay an execution */
243 ModelAction *next = node_stack->get_next()->get_action();
245 if (next == diverge) {
246 if (earliest_diverge == NULL || *diverge < *earliest_diverge)
247 earliest_diverge = diverge;
249 Node *nextnode = next->get_node();
250 Node *prevnode = nextnode->get_parent();
251 scheduler->update_sleep_set(prevnode);
253 /* Reached divergence point */
254 if (nextnode->increment_misc()) {
255 /* The next node will try to satisfy a different misc_index values. */
256 tid = next->get_tid();
257 node_stack->pop_restofstack(2);
258 } else if (nextnode->increment_promise()) {
259 /* The next node will try to satisfy a different set of promises. */
260 tid = next->get_tid();
261 node_stack->pop_restofstack(2);
262 } else if (nextnode->increment_read_from()) {
263 /* The next node will read from a different value. */
264 tid = next->get_tid();
265 node_stack->pop_restofstack(2);
266 } else if (nextnode->increment_future_value()) {
267 /* The next node will try to read from a different future value. */
268 tid = next->get_tid();
269 node_stack->pop_restofstack(2);
270 } else if (nextnode->increment_relseq_break()) {
271 /* The next node will try to resolve a release sequence differently */
272 tid = next->get_tid();
273 node_stack->pop_restofstack(2);
276 /* Make a different thread execute for next step */
277 scheduler->add_sleep(get_thread(next->get_tid()));
278 tid = prevnode->get_next_backtrack();
279 /* Make sure the backtracked thread isn't sleeping. */
280 node_stack->pop_restofstack(1);
281 if (diverge == earliest_diverge) {
282 earliest_diverge = prevnode->get_action();
285 /* The correct sleep set is in the parent node. */
288 DEBUG("*** Divergence point ***\n");
292 tid = next->get_tid();
294 DEBUG("*** ModelChecker chose next thread = %d ***\n", id_to_int(tid));
295 ASSERT(tid != THREAD_ID_T_NONE);
296 return thread_map->get(id_to_int(tid));
300 * We need to know what the next actions of all threads in the sleep
301 * set will be. This method computes them and stores the actions at
302 * the corresponding thread object's pending action.
305 void ModelChecker::execute_sleep_set()
307 for (unsigned int i = 0; i < get_num_threads(); i++) {
308 thread_id_t tid = int_to_id(i);
309 Thread *thr = get_thread(tid);
310 if (scheduler->is_sleep_set(thr) && thr->get_pending()) {
311 thr->get_pending()->set_sleep_flag();
317 * @brief Should the current action wake up a given thread?
319 * @param curr The current action
320 * @param thread The thread that we might wake up
321 * @return True, if we should wake up the sleeping thread; false otherwise
323 bool ModelChecker::should_wake_up(const ModelAction *curr, const Thread *thread) const
325 const ModelAction *asleep = thread->get_pending();
326 /* Don't allow partial RMW to wake anyone up */
329 /* Synchronizing actions may have been backtracked */
330 if (asleep->could_synchronize_with(curr))
332 /* All acquire/release fences and fence-acquire/store-release */
333 if (asleep->is_fence() && asleep->is_acquire() && curr->is_release())
335 /* Fence-release + store can awake load-acquire on the same location */
336 if (asleep->is_read() && asleep->is_acquire() && curr->same_var(asleep) && curr->is_write()) {
337 ModelAction *fence_release = get_last_fence_release(curr->get_tid());
338 if (fence_release && *(get_last_action(thread->get_id())) < *fence_release)
344 void ModelChecker::wake_up_sleeping_actions(ModelAction *curr)
346 for (unsigned int i = 0; i < get_num_threads(); i++) {
347 Thread *thr = get_thread(int_to_id(i));
348 if (scheduler->is_sleep_set(thr)) {
349 if (should_wake_up(curr, thr))
350 /* Remove this thread from sleep set */
351 scheduler->remove_sleep(thr);
356 /** @brief Alert the model-checker that an incorrectly-ordered
357 * synchronization was made */
358 void ModelChecker::set_bad_synchronization()
360 priv->bad_synchronization = true;
364 * Check whether the current trace has triggered an assertion which should halt
367 * @return True, if the execution should be aborted; false otherwise
369 bool ModelChecker::has_asserted() const
371 return priv->asserted;
375 * Trigger a trace assertion which should cause this execution to be halted.
376 * This can be due to a detected bug or due to an infeasibility that should
379 void ModelChecker::set_assert()
381 priv->asserted = true;
385 * Check if we are in a deadlock. Should only be called at the end of an
386 * execution, although it should not give false positives in the middle of an
387 * execution (there should be some ENABLED thread).
389 * @return True if program is in a deadlock; false otherwise
391 bool ModelChecker::is_deadlocked() const
393 bool blocking_threads = false;
394 for (unsigned int i = 0; i < get_num_threads(); i++) {
395 thread_id_t tid = int_to_id(i);
398 Thread *t = get_thread(tid);
399 if (!t->is_model_thread() && t->get_pending())
400 blocking_threads = true;
402 return blocking_threads;
406 * Check if this is a complete execution. That is, have all thread completed
407 * execution (rather than exiting because sleep sets have forced a redundant
410 * @return True if the execution is complete.
412 bool ModelChecker::is_complete_execution() const
414 for (unsigned int i = 0; i < get_num_threads(); i++)
415 if (is_enabled(int_to_id(i)))
421 * @brief Assert a bug in the executing program.
423 * Use this function to assert any sort of bug in the user program. If the
424 * current trace is feasible (actually, a prefix of some feasible execution),
425 * then this execution will be aborted, printing the appropriate message. If
426 * the current trace is not yet feasible, the error message will be stashed and
427 * printed if the execution ever becomes feasible.
429 * @param msg Descriptive message for the bug (do not include newline char)
430 * @return True if bug is immediately-feasible
432 bool ModelChecker::assert_bug(const char *msg)
434 priv->bugs.push_back(new bug_message(msg));
436 if (isfeasibleprefix()) {
444 * @brief Assert a bug in the executing program, asserted by a user thread
445 * @see ModelChecker::assert_bug
446 * @param msg Descriptive message for the bug (do not include newline char)
448 void ModelChecker::assert_user_bug(const char *msg)
450 /* If feasible bug, bail out now */
452 switch_to_master(NULL);
455 /** @return True, if any bugs have been reported for this execution */
456 bool ModelChecker::have_bug_reports() const
458 return priv->bugs.size() != 0;
461 /** @brief Print bug report listing for this execution (if any bugs exist) */
462 void ModelChecker::print_bugs() const
464 if (have_bug_reports()) {
465 model_print("Bug report: %zu bug%s detected\n",
467 priv->bugs.size() > 1 ? "s" : "");
468 for (unsigned int i = 0; i < priv->bugs.size(); i++)
469 priv->bugs[i]->print();
474 * @brief Record end-of-execution stats
476 * Must be run when exiting an execution. Records various stats.
477 * @see struct execution_stats
479 void ModelChecker::record_stats()
482 if (!isfeasibleprefix())
483 stats.num_infeasible++;
484 else if (have_bug_reports())
485 stats.num_buggy_executions++;
486 else if (is_complete_execution())
487 stats.num_complete++;
489 stats.num_redundant++;
492 /** @brief Print execution stats */
493 void ModelChecker::print_stats() const
495 model_print("Number of complete, bug-free executions: %d\n", stats.num_complete);
496 model_print("Number of redundant executions: %d\n", stats.num_redundant);
497 model_print("Number of buggy executions: %d\n", stats.num_buggy_executions);
498 model_print("Number of infeasible executions: %d\n", stats.num_infeasible);
499 model_print("Total executions: %d\n", stats.num_total);
500 model_print("Total nodes created: %d\n", node_stack->get_total_nodes());
504 * @brief End-of-exeuction print
505 * @param printbugs Should any existing bugs be printed?
507 void ModelChecker::print_execution(bool printbugs) const
509 print_program_output();
511 if (DBG_ENABLED() || params.verbose) {
512 model_print("Earliest divergence point since last feasible execution:\n");
513 if (earliest_diverge)
514 earliest_diverge->print();
516 model_print("(Not set)\n");
522 /* Don't print invalid bugs */
531 * Queries the model-checker for more executions to explore and, if one
532 * exists, resets the model-checker state to execute a new execution.
534 * @return If there are more executions to explore, return true. Otherwise,
537 bool ModelChecker::next_execution()
540 /* Is this execution a feasible execution that's worth bug-checking? */
541 bool complete = isfeasibleprefix() && (is_complete_execution() ||
544 /* End-of-execution bug checks */
547 assert_bug("Deadlock detected");
555 if (DBG_ENABLED() || params.verbose || (complete && have_bug_reports()))
556 print_execution(complete);
558 clear_program_output();
561 earliest_diverge = NULL;
563 if ((diverge = get_next_backtrack()) == NULL)
567 model_print("Next execution will diverge at:\n");
571 reset_to_initial_state();
576 * @brief Find the last fence-related backtracking conflict for a ModelAction
578 * This function performs the search for the most recent conflicting action
579 * against which we should perform backtracking, as affected by fence
580 * operations. This includes pairs of potentially-synchronizing actions which
581 * occur due to fence-acquire or fence-release, and hence should be explored in
582 * the opposite execution order.
584 * @param act The current action
585 * @return The most recent action which conflicts with act due to fences
587 ModelAction * ModelChecker::get_last_fence_conflict(ModelAction *act) const
589 /* Only perform release/acquire fence backtracking for stores */
590 if (!act->is_write())
593 /* Find a fence-release (or, act is a release) */
594 ModelAction *last_release;
595 if (act->is_release())
598 last_release = get_last_fence_release(act->get_tid());
602 /* Skip past the release */
603 action_list_t *list = action_trace;
604 action_list_t::reverse_iterator rit;
605 for (rit = list->rbegin(); rit != list->rend(); rit++)
606 if (*rit == last_release)
608 ASSERT(rit != list->rend());
613 * load --sb-> fence-acquire */
614 std::vector< ModelAction *, ModelAlloc<ModelAction *> > acquire_fences(get_num_threads(), NULL);
615 std::vector< ModelAction *, ModelAlloc<ModelAction *> > prior_loads(get_num_threads(), NULL);
616 bool found_acquire_fences = false;
617 for ( ; rit != list->rend(); rit++) {
618 ModelAction *prev = *rit;
619 if (act->same_thread(prev))
622 int tid = id_to_int(prev->get_tid());
624 if (prev->is_read() && act->same_var(prev)) {
625 if (prev->is_acquire()) {
626 /* Found most recent load-acquire, don't need
627 * to search for more fences */
628 if (!found_acquire_fences)
631 prior_loads[tid] = prev;
634 if (prev->is_acquire() && prev->is_fence() && !acquire_fences[tid]) {
635 found_acquire_fences = true;
636 acquire_fences[tid] = prev;
640 ModelAction *latest_backtrack = NULL;
641 for (unsigned int i = 0; i < acquire_fences.size(); i++)
642 if (acquire_fences[i] && prior_loads[i])
643 if (!latest_backtrack || *latest_backtrack < *acquire_fences[i])
644 latest_backtrack = acquire_fences[i];
645 return latest_backtrack;
649 * @brief Find the last backtracking conflict for a ModelAction
651 * This function performs the search for the most recent conflicting action
652 * against which we should perform backtracking. This primary includes pairs of
653 * synchronizing actions which should be explored in the opposite execution
656 * @param act The current action
657 * @return The most recent action which conflicts with act
659 ModelAction * ModelChecker::get_last_conflict(ModelAction *act) const
661 switch (act->get_type()) {
662 /* case ATOMIC_FENCE: fences don't directly cause backtracking */
666 ModelAction *ret = NULL;
668 /* linear search: from most recent to oldest */
669 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
670 action_list_t::reverse_iterator rit;
671 for (rit = list->rbegin(); rit != list->rend(); rit++) {
672 ModelAction *prev = *rit;
673 if (prev->could_synchronize_with(act)) {
679 ModelAction *ret2 = get_last_fence_conflict(act);
689 case ATOMIC_TRYLOCK: {
690 /* linear search: from most recent to oldest */
691 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
692 action_list_t::reverse_iterator rit;
693 for (rit = list->rbegin(); rit != list->rend(); rit++) {
694 ModelAction *prev = *rit;
695 if (act->is_conflicting_lock(prev))
700 case ATOMIC_UNLOCK: {
701 /* linear search: from most recent to oldest */
702 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
703 action_list_t::reverse_iterator rit;
704 for (rit = list->rbegin(); rit != list->rend(); rit++) {
705 ModelAction *prev = *rit;
706 if (!act->same_thread(prev) && prev->is_failed_trylock())
712 /* linear search: from most recent to oldest */
713 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
714 action_list_t::reverse_iterator rit;
715 for (rit = list->rbegin(); rit != list->rend(); rit++) {
716 ModelAction *prev = *rit;
717 if (!act->same_thread(prev) && prev->is_failed_trylock())
719 if (!act->same_thread(prev) && prev->is_notify())
725 case ATOMIC_NOTIFY_ALL:
726 case ATOMIC_NOTIFY_ONE: {
727 /* linear search: from most recent to oldest */
728 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
729 action_list_t::reverse_iterator rit;
730 for (rit = list->rbegin(); rit != list->rend(); rit++) {
731 ModelAction *prev = *rit;
732 if (!act->same_thread(prev) && prev->is_wait())
743 /** This method finds backtracking points where we should try to
744 * reorder the parameter ModelAction against.
746 * @param the ModelAction to find backtracking points for.
748 void ModelChecker::set_backtracking(ModelAction *act)
750 Thread *t = get_thread(act);
751 ModelAction *prev = get_last_conflict(act);
755 Node *node = prev->get_node()->get_parent();
757 int low_tid, high_tid;
758 if (node->enabled_status(t->get_id()) == THREAD_ENABLED) {
759 low_tid = id_to_int(act->get_tid());
760 high_tid = low_tid + 1;
763 high_tid = get_num_threads();
766 for (int i = low_tid; i < high_tid; i++) {
767 thread_id_t tid = int_to_id(i);
769 /* Make sure this thread can be enabled here. */
770 if (i >= node->get_num_threads())
773 /* Don't backtrack into a point where the thread is disabled or sleeping. */
774 if (node->enabled_status(tid) != THREAD_ENABLED)
777 /* Check if this has been explored already */
778 if (node->has_been_explored(tid))
781 /* See if fairness allows */
782 if (model->params.fairwindow != 0 && !node->has_priority(tid)) {
784 for (int t = 0; t < node->get_num_threads(); t++) {
785 thread_id_t tother = int_to_id(t);
786 if (node->is_enabled(tother) && node->has_priority(tother)) {
794 /* Cache the latest backtracking point */
795 set_latest_backtrack(prev);
797 /* If this is a new backtracking point, mark the tree */
798 if (!node->set_backtrack(tid))
800 DEBUG("Setting backtrack: conflict = %d, instead tid = %d\n",
801 id_to_int(prev->get_tid()),
802 id_to_int(t->get_id()));
811 * @brief Cache the a backtracking point as the "most recent", if eligible
813 * Note that this does not prepare the NodeStack for this backtracking
814 * operation, it only caches the action on a per-execution basis
816 * @param act The operation at which we should explore a different next action
817 * (i.e., backtracking point)
818 * @return True, if this action is now the most recent backtracking point;
821 bool ModelChecker::set_latest_backtrack(ModelAction *act)
823 if (!priv->next_backtrack || *act > *priv->next_backtrack) {
824 priv->next_backtrack = act;
831 * Returns last backtracking point. The model checker will explore a different
832 * path for this point in the next execution.
833 * @return The ModelAction at which the next execution should diverge.
835 ModelAction * ModelChecker::get_next_backtrack()
837 ModelAction *next = priv->next_backtrack;
838 priv->next_backtrack = NULL;
843 * Processes a read or rmw model action.
844 * @param curr is the read model action to process.
845 * @param second_part_of_rmw is boolean that is true is this is the second action of a rmw.
846 * @return True if processing this read updates the mo_graph.
848 bool ModelChecker::process_read(ModelAction *curr, bool second_part_of_rmw)
850 uint64_t value = VALUE_NONE;
851 bool updated = false;
853 const ModelAction *reads_from = curr->get_node()->get_read_from();
854 if (reads_from != NULL) {
855 mo_graph->startChanges();
857 value = reads_from->get_value();
858 bool r_status = false;
860 if (!second_part_of_rmw) {
861 check_recency(curr, reads_from);
862 r_status = r_modification_order(curr, reads_from);
865 if (!second_part_of_rmw && is_infeasible() && (curr->get_node()->increment_read_from() || curr->get_node()->increment_future_value())) {
866 mo_graph->rollbackChanges();
867 priv->too_many_reads = false;
871 read_from(curr, reads_from);
872 mo_graph->commitChanges();
873 mo_check_promises(curr, true);
876 } else if (!second_part_of_rmw) {
877 /* Read from future value */
878 struct future_value fv = curr->get_node()->get_future_value();
879 Promise *promise = new Promise(curr, fv);
881 curr->set_read_from_promise(promise);
882 promises->push_back(promise);
883 mo_graph->startChanges();
884 updated = r_modification_order(curr, promise);
885 mo_graph->commitChanges();
887 get_thread(curr)->set_return_value(value);
893 * Processes a lock, trylock, or unlock model action. @param curr is
894 * the read model action to process.
896 * The try lock operation checks whether the lock is taken. If not,
897 * it falls to the normal lock operation case. If so, it returns
900 * The lock operation has already been checked that it is enabled, so
901 * it just grabs the lock and synchronizes with the previous unlock.
903 * The unlock operation has to re-enable all of the threads that are
904 * waiting on the lock.
906 * @return True if synchronization was updated; false otherwise
908 bool ModelChecker::process_mutex(ModelAction *curr)
910 std::mutex *mutex = NULL;
911 struct std::mutex_state *state = NULL;
913 if (curr->is_trylock() || curr->is_lock() || curr->is_unlock()) {
914 mutex = (std::mutex *)curr->get_location();
915 state = mutex->get_state();
916 } else if (curr->is_wait()) {
917 mutex = (std::mutex *)curr->get_value();
918 state = mutex->get_state();
921 switch (curr->get_type()) {
922 case ATOMIC_TRYLOCK: {
923 bool success = !state->islocked;
924 curr->set_try_lock(success);
926 get_thread(curr)->set_return_value(0);
929 get_thread(curr)->set_return_value(1);
931 //otherwise fall into the lock case
933 if (curr->get_cv()->getClock(state->alloc_tid) <= state->alloc_clock)
934 assert_bug("Lock access before initialization");
935 state->islocked = true;
936 ModelAction *unlock = get_last_unlock(curr);
937 //synchronize with the previous unlock statement
938 if (unlock != NULL) {
939 curr->synchronize_with(unlock);
944 case ATOMIC_UNLOCK: {
946 state->islocked = false;
947 //wake up the other threads
948 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, curr->get_location());
949 //activate all the waiting threads
950 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
951 scheduler->wake(get_thread(*rit));
958 state->islocked = false;
959 //wake up the other threads
960 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, (void *) curr->get_value());
961 //activate all the waiting threads
962 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
963 scheduler->wake(get_thread(*rit));
966 //check whether we should go to sleep or not...simulate spurious failures
967 if (curr->get_node()->get_misc() == 0) {
968 get_safe_ptr_action(condvar_waiters_map, curr->get_location())->push_back(curr);
970 scheduler->sleep(get_thread(curr));
974 case ATOMIC_NOTIFY_ALL: {
975 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
976 //activate all the waiting threads
977 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
978 scheduler->wake(get_thread(*rit));
983 case ATOMIC_NOTIFY_ONE: {
984 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
985 int wakeupthread = curr->get_node()->get_misc();
986 action_list_t::iterator it = waiters->begin();
987 advance(it, wakeupthread);
988 scheduler->wake(get_thread(*it));
999 void ModelChecker::add_future_value(const ModelAction *writer, ModelAction *reader)
1001 /* Do more ambitious checks now that mo is more complete */
1002 if (mo_may_allow(writer, reader)) {
1003 Node *node = reader->get_node();
1005 /* Find an ancestor thread which exists at the time of the reader */
1006 Thread *write_thread = get_thread(writer);
1007 while (id_to_int(write_thread->get_id()) >= node->get_num_threads())
1008 write_thread = write_thread->get_parent();
1010 struct future_value fv = {
1011 writer->get_value(),
1012 writer->get_seq_number() + params.maxfuturedelay,
1013 write_thread->get_id(),
1015 if (node->add_future_value(fv))
1016 set_latest_backtrack(reader);
1021 * Process a write ModelAction
1022 * @param curr The ModelAction to process
1023 * @return True if the mo_graph was updated or promises were resolved
1025 bool ModelChecker::process_write(ModelAction *curr)
1027 bool updated_mod_order = w_modification_order(curr);
1028 bool updated_promises = resolve_promises(curr);
1030 if (promises->size() == 0) {
1031 for (unsigned int i = 0; i < futurevalues->size(); i++) {
1032 struct PendingFutureValue pfv = (*futurevalues)[i];
1033 add_future_value(pfv.writer, pfv.act);
1035 futurevalues->clear();
1038 mo_graph->commitChanges();
1039 mo_check_promises(curr, false);
1041 get_thread(curr)->set_return_value(VALUE_NONE);
1042 return updated_mod_order || updated_promises;
1046 * Process a fence ModelAction
1047 * @param curr The ModelAction to process
1048 * @return True if synchronization was updated
1050 bool ModelChecker::process_fence(ModelAction *curr)
1053 * fence-relaxed: no-op
1054 * fence-release: only log the occurence (not in this function), for
1055 * use in later synchronization
1056 * fence-acquire (this function): search for hypothetical release
1059 bool updated = false;
1060 if (curr->is_acquire()) {
1061 action_list_t *list = action_trace;
1062 action_list_t::reverse_iterator rit;
1063 /* Find X : is_read(X) && X --sb-> curr */
1064 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1065 ModelAction *act = *rit;
1068 if (act->get_tid() != curr->get_tid())
1070 /* Stop at the beginning of the thread */
1071 if (act->is_thread_start())
1073 /* Stop once we reach a prior fence-acquire */
1074 if (act->is_fence() && act->is_acquire())
1076 if (!act->is_read())
1078 /* read-acquire will find its own release sequences */
1079 if (act->is_acquire())
1082 /* Establish hypothetical release sequences */
1083 rel_heads_list_t release_heads;
1084 get_release_seq_heads(curr, act, &release_heads);
1085 for (unsigned int i = 0; i < release_heads.size(); i++)
1086 if (!curr->synchronize_with(release_heads[i]))
1087 set_bad_synchronization();
1088 if (release_heads.size() != 0)
1096 * @brief Process the current action for thread-related activity
1098 * Performs current-action processing for a THREAD_* ModelAction. Proccesses
1099 * may include setting Thread status, completing THREAD_FINISH/THREAD_JOIN
1100 * synchronization, etc. This function is a no-op for non-THREAD actions
1101 * (e.g., ATOMIC_{READ,WRITE,RMW,LOCK}, etc.)
1103 * @param curr The current action
1104 * @return True if synchronization was updated or a thread completed
1106 bool ModelChecker::process_thread_action(ModelAction *curr)
1108 bool updated = false;
1110 switch (curr->get_type()) {
1111 case THREAD_CREATE: {
1112 thrd_t *thrd = (thrd_t *)curr->get_location();
1113 struct thread_params *params = (struct thread_params *)curr->get_value();
1114 Thread *th = new Thread(thrd, params->func, params->arg, get_thread(curr));
1116 th->set_creation(curr);
1117 /* Promises can be satisfied by children */
1118 for (unsigned int i = 0; i < promises->size(); i++) {
1119 Promise *promise = (*promises)[i];
1120 if (promise->thread_is_available(curr->get_tid()))
1121 promise->add_thread(th->get_id());
1126 Thread *blocking = curr->get_thread_operand();
1127 ModelAction *act = get_last_action(blocking->get_id());
1128 curr->synchronize_with(act);
1129 updated = true; /* trigger rel-seq checks */
1132 case THREAD_FINISH: {
1133 Thread *th = get_thread(curr);
1134 while (!th->wait_list_empty()) {
1135 ModelAction *act = th->pop_wait_list();
1136 scheduler->wake(get_thread(act));
1139 /* Completed thread can't satisfy promises */
1140 for (unsigned int i = 0; i < promises->size(); i++) {
1141 Promise *promise = (*promises)[i];
1142 if (promise->thread_is_available(th->get_id()))
1143 if (promise->eliminate_thread(th->get_id()))
1144 priv->failed_promise = true;
1146 updated = true; /* trigger rel-seq checks */
1149 case THREAD_START: {
1150 check_promises(curr->get_tid(), NULL, curr->get_cv());
1161 * @brief Process the current action for release sequence fixup activity
1163 * Performs model-checker release sequence fixups for the current action,
1164 * forcing a single pending release sequence to break (with a given, potential
1165 * "loose" write) or to complete (i.e., synchronize). If a pending release
1166 * sequence forms a complete release sequence, then we must perform the fixup
1167 * synchronization, mo_graph additions, etc.
1169 * @param curr The current action; must be a release sequence fixup action
1170 * @param work_queue The work queue to which to add work items as they are
1173 void ModelChecker::process_relseq_fixup(ModelAction *curr, work_queue_t *work_queue)
1175 const ModelAction *write = curr->get_node()->get_relseq_break();
1176 struct release_seq *sequence = pending_rel_seqs->back();
1177 pending_rel_seqs->pop_back();
1179 ModelAction *acquire = sequence->acquire;
1180 const ModelAction *rf = sequence->rf;
1181 const ModelAction *release = sequence->release;
1185 ASSERT(release->same_thread(rf));
1187 if (write == NULL) {
1189 * @todo Forcing a synchronization requires that we set
1190 * modification order constraints. For instance, we can't allow
1191 * a fixup sequence in which two separate read-acquire
1192 * operations read from the same sequence, where the first one
1193 * synchronizes and the other doesn't. Essentially, we can't
1194 * allow any writes to insert themselves between 'release' and
1198 /* Must synchronize */
1199 if (!acquire->synchronize_with(release)) {
1200 set_bad_synchronization();
1203 /* Re-check all pending release sequences */
1204 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
1205 /* Re-check act for mo_graph edges */
1206 work_queue->push_back(MOEdgeWorkEntry(acquire));
1208 /* propagate synchronization to later actions */
1209 action_list_t::reverse_iterator rit = action_trace->rbegin();
1210 for (; (*rit) != acquire; rit++) {
1211 ModelAction *propagate = *rit;
1212 if (acquire->happens_before(propagate)) {
1213 propagate->synchronize_with(acquire);
1214 /* Re-check 'propagate' for mo_graph edges */
1215 work_queue->push_back(MOEdgeWorkEntry(propagate));
1219 /* Break release sequence with new edges:
1220 * release --mo--> write --mo--> rf */
1221 mo_graph->addEdge(release, write);
1222 mo_graph->addEdge(write, rf);
1225 /* See if we have realized a data race */
1230 * Initialize the current action by performing one or more of the following
1231 * actions, as appropriate: merging RMWR and RMWC/RMW actions, stepping forward
1232 * in the NodeStack, manipulating backtracking sets, allocating and
1233 * initializing clock vectors, and computing the promises to fulfill.
1235 * @param curr The current action, as passed from the user context; may be
1236 * freed/invalidated after the execution of this function, with a different
1237 * action "returned" its place (pass-by-reference)
1238 * @return True if curr is a newly-explored action; false otherwise
1240 bool ModelChecker::initialize_curr_action(ModelAction **curr)
1242 ModelAction *newcurr;
1244 if ((*curr)->is_rmwc() || (*curr)->is_rmw()) {
1245 newcurr = process_rmw(*curr);
1248 if (newcurr->is_rmw())
1249 compute_promises(newcurr);
1255 (*curr)->set_seq_number(get_next_seq_num());
1257 newcurr = node_stack->explore_action(*curr, scheduler->get_enabled_array());
1259 /* First restore type and order in case of RMW operation */
1260 if ((*curr)->is_rmwr())
1261 newcurr->copy_typeandorder(*curr);
1263 ASSERT((*curr)->get_location() == newcurr->get_location());
1264 newcurr->copy_from_new(*curr);
1266 /* Discard duplicate ModelAction; use action from NodeStack */
1269 /* Always compute new clock vector */
1270 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1273 return false; /* Action was explored previously */
1277 /* Always compute new clock vector */
1278 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1280 /* Assign most recent release fence */
1281 newcurr->set_last_fence_release(get_last_fence_release(newcurr->get_tid()));
1284 * Perform one-time actions when pushing new ModelAction onto
1287 if (newcurr->is_write())
1288 compute_promises(newcurr);
1289 else if (newcurr->is_relseq_fixup())
1290 compute_relseq_breakwrites(newcurr);
1291 else if (newcurr->is_wait())
1292 newcurr->get_node()->set_misc_max(2);
1293 else if (newcurr->is_notify_one()) {
1294 newcurr->get_node()->set_misc_max(get_safe_ptr_action(condvar_waiters_map, newcurr->get_location())->size());
1296 return true; /* This was a new ModelAction */
1301 * @brief Establish reads-from relation between two actions
1303 * Perform basic operations involved with establishing a concrete rf relation,
1304 * including setting the ModelAction data and checking for release sequences.
1306 * @param act The action that is reading (must be a read)
1307 * @param rf The action from which we are reading (must be a write)
1309 * @return True if this read established synchronization
1311 bool ModelChecker::read_from(ModelAction *act, const ModelAction *rf)
1313 act->set_read_from(rf);
1314 if (rf != NULL && act->is_acquire()) {
1315 rel_heads_list_t release_heads;
1316 get_release_seq_heads(act, act, &release_heads);
1317 int num_heads = release_heads.size();
1318 for (unsigned int i = 0; i < release_heads.size(); i++)
1319 if (!act->synchronize_with(release_heads[i])) {
1320 set_bad_synchronization();
1323 return num_heads > 0;
1329 * @brief Check whether a model action is enabled.
1331 * Checks whether a lock or join operation would be successful (i.e., is the
1332 * lock already locked, or is the joined thread already complete). If not, put
1333 * the action in a waiter list.
1335 * @param curr is the ModelAction to check whether it is enabled.
1336 * @return a bool that indicates whether the action is enabled.
1338 bool ModelChecker::check_action_enabled(ModelAction *curr) {
1339 if (curr->is_lock()) {
1340 std::mutex *lock = (std::mutex *)curr->get_location();
1341 struct std::mutex_state *state = lock->get_state();
1342 if (state->islocked) {
1343 //Stick the action in the appropriate waiting queue
1344 get_safe_ptr_action(lock_waiters_map, curr->get_location())->push_back(curr);
1347 } else if (curr->get_type() == THREAD_JOIN) {
1348 Thread *blocking = (Thread *)curr->get_location();
1349 if (!blocking->is_complete()) {
1350 blocking->push_wait_list(curr);
1359 * This is the heart of the model checker routine. It performs model-checking
1360 * actions corresponding to a given "current action." Among other processes, it
1361 * calculates reads-from relationships, updates synchronization clock vectors,
1362 * forms a memory_order constraints graph, and handles replay/backtrack
1363 * execution when running permutations of previously-observed executions.
1365 * @param curr The current action to process
1366 * @return The ModelAction that is actually executed; may be different than
1367 * curr; may be NULL, if the current action is not enabled to run
1369 ModelAction * ModelChecker::check_current_action(ModelAction *curr)
1372 bool second_part_of_rmw = curr->is_rmwc() || curr->is_rmw();
1374 if (!check_action_enabled(curr)) {
1375 /* Make the execution look like we chose to run this action
1376 * much later, when a lock/join can succeed */
1377 get_thread(curr)->set_pending(curr);
1378 scheduler->sleep(get_thread(curr));
1382 bool newly_explored = initialize_curr_action(&curr);
1388 wake_up_sleeping_actions(curr);
1390 /* Add the action to lists before any other model-checking tasks */
1391 if (!second_part_of_rmw)
1392 add_action_to_lists(curr);
1394 /* Build may_read_from set for newly-created actions */
1395 if (newly_explored && curr->is_read())
1396 build_may_read_from(curr);
1398 /* Initialize work_queue with the "current action" work */
1399 work_queue_t work_queue(1, CheckCurrWorkEntry(curr));
1400 while (!work_queue.empty() && !has_asserted()) {
1401 WorkQueueEntry work = work_queue.front();
1402 work_queue.pop_front();
1404 switch (work.type) {
1405 case WORK_CHECK_CURR_ACTION: {
1406 ModelAction *act = work.action;
1407 bool update = false; /* update this location's release seq's */
1408 bool update_all = false; /* update all release seq's */
1410 if (process_thread_action(curr))
1413 if (act->is_read() && process_read(act, second_part_of_rmw))
1416 if (act->is_write() && process_write(act))
1419 if (act->is_fence() && process_fence(act))
1422 if (act->is_mutex_op() && process_mutex(act))
1425 if (act->is_relseq_fixup())
1426 process_relseq_fixup(curr, &work_queue);
1429 work_queue.push_back(CheckRelSeqWorkEntry(NULL));
1431 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1434 case WORK_CHECK_RELEASE_SEQ:
1435 resolve_release_sequences(work.location, &work_queue);
1437 case WORK_CHECK_MO_EDGES: {
1438 /** @todo Complete verification of work_queue */
1439 ModelAction *act = work.action;
1440 bool updated = false;
1442 if (act->is_read()) {
1443 const ModelAction *rf = act->get_reads_from();
1444 const Promise *promise = act->get_reads_from_promise();
1446 if (r_modification_order(act, rf))
1448 } else if (promise) {
1449 if (r_modification_order(act, promise))
1453 if (act->is_write()) {
1454 if (w_modification_order(act))
1457 mo_graph->commitChanges();
1460 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1469 check_curr_backtracking(curr);
1470 set_backtracking(curr);
1474 void ModelChecker::check_curr_backtracking(ModelAction *curr)
1476 Node *currnode = curr->get_node();
1477 Node *parnode = currnode->get_parent();
1479 if ((parnode && !parnode->backtrack_empty()) ||
1480 !currnode->misc_empty() ||
1481 !currnode->read_from_empty() ||
1482 !currnode->future_value_empty() ||
1483 !currnode->promise_empty() ||
1484 !currnode->relseq_break_empty()) {
1485 set_latest_backtrack(curr);
1489 bool ModelChecker::promises_expired() const
1491 for (unsigned int i = 0; i < promises->size(); i++) {
1492 Promise *promise = (*promises)[i];
1493 if (promise->get_expiration() < priv->used_sequence_numbers)
1500 * This is the strongest feasibility check available.
1501 * @return whether the current trace (partial or complete) must be a prefix of
1504 bool ModelChecker::isfeasibleprefix() const
1506 return pending_rel_seqs->size() == 0 && is_feasible_prefix_ignore_relseq();
1510 * Print disagnostic information about an infeasible execution
1511 * @param prefix A string to prefix the output with; if NULL, then a default
1512 * message prefix will be provided
1514 void ModelChecker::print_infeasibility(const char *prefix) const
1518 if (mo_graph->checkForCycles())
1519 ptr += sprintf(ptr, "[mo cycle]");
1520 if (priv->failed_promise)
1521 ptr += sprintf(ptr, "[failed promise]");
1522 if (priv->too_many_reads)
1523 ptr += sprintf(ptr, "[too many reads]");
1524 if (priv->no_valid_reads)
1525 ptr += sprintf(ptr, "[no valid reads-from]");
1526 if (priv->bad_synchronization)
1527 ptr += sprintf(ptr, "[bad sw ordering]");
1528 if (promises_expired())
1529 ptr += sprintf(ptr, "[promise expired]");
1530 if (promises->size() != 0)
1531 ptr += sprintf(ptr, "[unresolved promise]");
1533 model_print("%s: %s\n", prefix ? prefix : "Infeasible", buf);
1537 * Returns whether the current completed trace is feasible, except for pending
1538 * release sequences.
1540 bool ModelChecker::is_feasible_prefix_ignore_relseq() const
1542 return !is_infeasible() && promises->size() == 0;
1546 * Check if the current partial trace is infeasible. Does not check any
1547 * end-of-execution flags, which might rule out the execution. Thus, this is
1548 * useful only for ruling an execution as infeasible.
1549 * @return whether the current partial trace is infeasible.
1551 bool ModelChecker::is_infeasible() const
1553 return mo_graph->checkForCycles() ||
1554 priv->no_valid_reads ||
1555 priv->failed_promise ||
1556 priv->too_many_reads ||
1557 priv->bad_synchronization ||
1561 /** Close out a RMWR by converting previous RMWR into a RMW or READ. */
1562 ModelAction * ModelChecker::process_rmw(ModelAction *act) {
1563 ModelAction *lastread = get_last_action(act->get_tid());
1564 lastread->process_rmw(act);
1565 if (act->is_rmw()) {
1566 if (lastread->get_reads_from())
1567 mo_graph->addRMWEdge(lastread->get_reads_from(), lastread);
1569 mo_graph->addRMWEdge(lastread->get_reads_from_promise(), lastread);
1570 mo_graph->commitChanges();
1576 * Checks whether a thread has read from the same write for too many times
1577 * without seeing the effects of a later write.
1580 * 1) there must a different write that we could read from that would satisfy the modification order,
1581 * 2) we must have read from the same value in excess of maxreads times, and
1582 * 3) that other write must have been in the reads_from set for maxreads times.
1584 * If so, we decide that the execution is no longer feasible.
1586 void ModelChecker::check_recency(ModelAction *curr, const ModelAction *rf)
1588 if (params.maxreads != 0) {
1589 if (curr->get_node()->get_read_from_size() <= 1)
1591 //Must make sure that execution is currently feasible... We could
1592 //accidentally clear by rolling back
1593 if (is_infeasible())
1595 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1596 int tid = id_to_int(curr->get_tid());
1599 if ((int)thrd_lists->size() <= tid)
1601 action_list_t *list = &(*thrd_lists)[tid];
1603 action_list_t::reverse_iterator rit = list->rbegin();
1604 /* Skip past curr */
1605 for (; (*rit) != curr; rit++)
1607 /* go past curr now */
1610 action_list_t::reverse_iterator ritcopy = rit;
1611 //See if we have enough reads from the same value
1613 for (; count < params.maxreads; rit++, count++) {
1614 if (rit == list->rend())
1616 ModelAction *act = *rit;
1617 if (!act->is_read())
1620 if (act->get_reads_from() != rf)
1622 if (act->get_node()->get_read_from_size() <= 1)
1625 for (int i = 0; i < curr->get_node()->get_read_from_size(); i++) {
1627 const ModelAction *write = curr->get_node()->get_read_from_at(i);
1629 /* Need a different write */
1633 /* Test to see whether this is a feasible write to read from */
1634 /** NOTE: all members of read-from set should be
1635 * feasible, so we no longer check it here **/
1639 bool feasiblewrite = true;
1640 //new we need to see if this write works for everyone
1642 for (int loop = count; loop > 0; loop--, rit++) {
1643 ModelAction *act = *rit;
1644 bool foundvalue = false;
1645 for (int j = 0; j < act->get_node()->get_read_from_size(); j++) {
1646 if (act->get_node()->get_read_from_at(j) == write) {
1652 feasiblewrite = false;
1656 if (feasiblewrite) {
1657 priv->too_many_reads = true;
1665 * Updates the mo_graph with the constraints imposed from the current
1668 * Basic idea is the following: Go through each other thread and find
1669 * the last action that happened before our read. Two cases:
1671 * (1) The action is a write => that write must either occur before
1672 * the write we read from or be the write we read from.
1674 * (2) The action is a read => the write that that action read from
1675 * must occur before the write we read from or be the same write.
1677 * @param curr The current action. Must be a read.
1678 * @param rf The ModelAction or Promise that curr reads from. Must be a write.
1679 * @return True if modification order edges were added; false otherwise
1681 template <typename rf_type>
1682 bool ModelChecker::r_modification_order(ModelAction *curr, const rf_type *rf)
1684 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1687 ASSERT(curr->is_read());
1689 /* Last SC fence in the current thread */
1690 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1692 /* Iterate over all threads */
1693 for (i = 0; i < thrd_lists->size(); i++) {
1694 /* Last SC fence in thread i */
1695 ModelAction *last_sc_fence_thread_local = NULL;
1696 if (int_to_id((int)i) != curr->get_tid())
1697 last_sc_fence_thread_local = get_last_seq_cst_fence(int_to_id(i), NULL);
1699 /* Last SC fence in thread i, before last SC fence in current thread */
1700 ModelAction *last_sc_fence_thread_before = NULL;
1701 if (last_sc_fence_local)
1702 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1704 /* Iterate over actions in thread, starting from most recent */
1705 action_list_t *list = &(*thrd_lists)[i];
1706 action_list_t::reverse_iterator rit;
1707 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1708 ModelAction *act = *rit;
1710 if (act->is_write() && !act->equals(rf) && act != curr) {
1711 /* C++, Section 29.3 statement 5 */
1712 if (curr->is_seqcst() && last_sc_fence_thread_local &&
1713 *act < *last_sc_fence_thread_local) {
1714 added = mo_graph->addEdge(act, rf) || added;
1717 /* C++, Section 29.3 statement 4 */
1718 else if (act->is_seqcst() && last_sc_fence_local &&
1719 *act < *last_sc_fence_local) {
1720 added = mo_graph->addEdge(act, rf) || added;
1723 /* C++, Section 29.3 statement 6 */
1724 else if (last_sc_fence_thread_before &&
1725 *act < *last_sc_fence_thread_before) {
1726 added = mo_graph->addEdge(act, rf) || added;
1732 * Include at most one act per-thread that "happens
1733 * before" curr. Don't consider reflexively.
1735 if (act->happens_before(curr) && act != curr) {
1736 if (act->is_write()) {
1737 if (!act->equals(rf)) {
1738 added = mo_graph->addEdge(act, rf) || added;
1741 const ModelAction *prevreadfrom = act->get_reads_from();
1742 //if the previous read is unresolved, keep going...
1743 if (prevreadfrom == NULL)
1746 if (!prevreadfrom->equals(rf)) {
1747 added = mo_graph->addEdge(prevreadfrom, rf) || added;
1756 * All compatible, thread-exclusive promises must be ordered after any
1757 * concrete loads from the same thread
1759 for (unsigned int i = 0; i < promises->size(); i++)
1760 if ((*promises)[i]->is_compatible_exclusive(curr))
1761 added = mo_graph->addEdge(rf, (*promises)[i]) || added;
1767 * Updates the mo_graph with the constraints imposed from the current write.
1769 * Basic idea is the following: Go through each other thread and find
1770 * the lastest action that happened before our write. Two cases:
1772 * (1) The action is a write => that write must occur before
1775 * (2) The action is a read => the write that that action read from
1776 * must occur before the current write.
1778 * This method also handles two other issues:
1780 * (I) Sequential Consistency: Making sure that if the current write is
1781 * seq_cst, that it occurs after the previous seq_cst write.
1783 * (II) Sending the write back to non-synchronizing reads.
1785 * @param curr The current action. Must be a write.
1786 * @return True if modification order edges were added; false otherwise
1788 bool ModelChecker::w_modification_order(ModelAction *curr)
1790 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1793 ASSERT(curr->is_write());
1795 if (curr->is_seqcst()) {
1796 /* We have to at least see the last sequentially consistent write,
1797 so we are initialized. */
1798 ModelAction *last_seq_cst = get_last_seq_cst_write(curr);
1799 if (last_seq_cst != NULL) {
1800 added = mo_graph->addEdge(last_seq_cst, curr) || added;
1804 /* Last SC fence in the current thread */
1805 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1807 /* Iterate over all threads */
1808 for (i = 0; i < thrd_lists->size(); i++) {
1809 /* Last SC fence in thread i, before last SC fence in current thread */
1810 ModelAction *last_sc_fence_thread_before = NULL;
1811 if (last_sc_fence_local && int_to_id((int)i) != curr->get_tid())
1812 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1814 /* Iterate over actions in thread, starting from most recent */
1815 action_list_t *list = &(*thrd_lists)[i];
1816 action_list_t::reverse_iterator rit;
1817 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1818 ModelAction *act = *rit;
1821 * 1) If RMW and it actually read from something, then we
1822 * already have all relevant edges, so just skip to next
1825 * 2) If RMW and it didn't read from anything, we should
1826 * whatever edge we can get to speed up convergence.
1828 * 3) If normal write, we need to look at earlier actions, so
1829 * continue processing list.
1831 if (curr->is_rmw()) {
1832 if (curr->get_reads_from() != NULL)
1840 /* C++, Section 29.3 statement 7 */
1841 if (last_sc_fence_thread_before && act->is_write() &&
1842 *act < *last_sc_fence_thread_before) {
1843 added = mo_graph->addEdge(act, curr) || added;
1848 * Include at most one act per-thread that "happens
1851 if (act->happens_before(curr)) {
1853 * Note: if act is RMW, just add edge:
1855 * The following edge should be handled elsewhere:
1856 * readfrom(act) --mo--> act
1858 if (act->is_write())
1859 added = mo_graph->addEdge(act, curr) || added;
1860 else if (act->is_read()) {
1861 //if previous read accessed a null, just keep going
1862 if (act->get_reads_from() == NULL)
1864 added = mo_graph->addEdge(act->get_reads_from(), curr) || added;
1867 } else if (act->is_read() && !act->could_synchronize_with(curr) &&
1868 !act->same_thread(curr)) {
1869 /* We have an action that:
1870 (1) did not happen before us
1871 (2) is a read and we are a write
1872 (3) cannot synchronize with us
1873 (4) is in a different thread
1875 that read could potentially read from our write. Note that
1876 these checks are overly conservative at this point, we'll
1877 do more checks before actually removing the
1881 if (thin_air_constraint_may_allow(curr, act)) {
1882 if (!is_infeasible())
1883 futurevalues->push_back(PendingFutureValue(curr, act));
1884 else if (curr->is_rmw() && act->is_rmw() && curr->get_reads_from() && curr->get_reads_from() == act->get_reads_from())
1885 add_future_value(curr, act);
1892 * All compatible, thread-exclusive promises must be ordered after any
1893 * concrete stores to the same thread, or else they can be merged with
1896 for (unsigned int i = 0; i < promises->size(); i++)
1897 if ((*promises)[i]->is_compatible_exclusive(curr))
1898 added = mo_graph->addEdge(curr, (*promises)[i]) || added;
1903 /** Arbitrary reads from the future are not allowed. Section 29.3
1904 * part 9 places some constraints. This method checks one result of constraint
1905 * constraint. Others require compiler support. */
1906 bool ModelChecker::thin_air_constraint_may_allow(const ModelAction *writer, const ModelAction *reader)
1908 if (!writer->is_rmw())
1911 if (!reader->is_rmw())
1914 for (const ModelAction *search = writer->get_reads_from(); search != NULL; search = search->get_reads_from()) {
1915 if (search == reader)
1917 if (search->get_tid() == reader->get_tid() &&
1918 search->happens_before(reader))
1926 * Arbitrary reads from the future are not allowed. Section 29.3 part 9 places
1927 * some constraints. This method checks one the following constraint (others
1928 * require compiler support):
1930 * If X --hb-> Y --mo-> Z, then X should not read from Z.
1932 bool ModelChecker::mo_may_allow(const ModelAction *writer, const ModelAction *reader)
1934 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, reader->get_location());
1936 /* Iterate over all threads */
1937 for (i = 0; i < thrd_lists->size(); i++) {
1938 const ModelAction *write_after_read = NULL;
1940 /* Iterate over actions in thread, starting from most recent */
1941 action_list_t *list = &(*thrd_lists)[i];
1942 action_list_t::reverse_iterator rit;
1943 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1944 ModelAction *act = *rit;
1946 /* Don't disallow due to act == reader */
1947 if (!reader->happens_before(act) || reader == act)
1949 else if (act->is_write())
1950 write_after_read = act;
1951 else if (act->is_read() && act->get_reads_from() != NULL)
1952 write_after_read = act->get_reads_from();
1955 if (write_after_read && write_after_read != writer && mo_graph->checkReachable(write_after_read, writer))
1962 * Finds the head(s) of the release sequence(s) containing a given ModelAction.
1963 * The ModelAction under consideration is expected to be taking part in
1964 * release/acquire synchronization as an object of the "reads from" relation.
1965 * Note that this can only provide release sequence support for RMW chains
1966 * which do not read from the future, as those actions cannot be traced until
1967 * their "promise" is fulfilled. Similarly, we may not even establish the
1968 * presence of a release sequence with certainty, as some modification order
1969 * constraints may be decided further in the future. Thus, this function
1970 * "returns" two pieces of data: a pass-by-reference vector of @a release_heads
1971 * and a boolean representing certainty.
1973 * @param rf The action that might be part of a release sequence. Must be a
1975 * @param release_heads A pass-by-reference style return parameter. After
1976 * execution of this function, release_heads will contain the heads of all the
1977 * relevant release sequences, if any exists with certainty
1978 * @param pending A pass-by-reference style return parameter which is only used
1979 * when returning false (i.e., uncertain). Returns most information regarding
1980 * an uncertain release sequence, including any write operations that might
1981 * break the sequence.
1982 * @return true, if the ModelChecker is certain that release_heads is complete;
1985 bool ModelChecker::release_seq_heads(const ModelAction *rf,
1986 rel_heads_list_t *release_heads,
1987 struct release_seq *pending) const
1989 /* Only check for release sequences if there are no cycles */
1990 if (mo_graph->checkForCycles())
1993 for ( ; rf != NULL; rf = rf->get_reads_from()) {
1994 ASSERT(rf->is_write());
1996 if (rf->is_release())
1997 release_heads->push_back(rf);
1998 else if (rf->get_last_fence_release())
1999 release_heads->push_back(rf->get_last_fence_release());
2001 break; /* End of RMW chain */
2003 /** @todo Need to be smarter here... In the linux lock
2004 * example, this will run to the beginning of the program for
2006 /** @todo The way to be smarter here is to keep going until 1
2007 * thread has a release preceded by an acquire and you've seen
2010 /* acq_rel RMW is a sufficient stopping condition */
2011 if (rf->is_acquire() && rf->is_release())
2012 return true; /* complete */
2015 /* read from future: need to settle this later */
2017 return false; /* incomplete */
2020 if (rf->is_release())
2021 return true; /* complete */
2023 /* else relaxed write
2024 * - check for fence-release in the same thread (29.8, stmt. 3)
2025 * - check modification order for contiguous subsequence
2026 * -> rf must be same thread as release */
2028 const ModelAction *fence_release = rf->get_last_fence_release();
2029 /* Synchronize with a fence-release unconditionally; we don't need to
2030 * find any more "contiguous subsequence..." for it */
2032 release_heads->push_back(fence_release);
2034 int tid = id_to_int(rf->get_tid());
2035 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, rf->get_location());
2036 action_list_t *list = &(*thrd_lists)[tid];
2037 action_list_t::const_reverse_iterator rit;
2039 /* Find rf in the thread list */
2040 rit = std::find(list->rbegin(), list->rend(), rf);
2041 ASSERT(rit != list->rend());
2043 /* Find the last {write,fence}-release */
2044 for (; rit != list->rend(); rit++) {
2045 if (fence_release && *(*rit) < *fence_release)
2047 if ((*rit)->is_release())
2050 if (rit == list->rend()) {
2051 /* No write-release in this thread */
2052 return true; /* complete */
2053 } else if (fence_release && *(*rit) < *fence_release) {
2054 /* The fence-release is more recent (and so, "stronger") than
2055 * the most recent write-release */
2056 return true; /* complete */
2057 } /* else, need to establish contiguous release sequence */
2058 ModelAction *release = *rit;
2060 ASSERT(rf->same_thread(release));
2062 pending->writes.clear();
2064 bool certain = true;
2065 for (unsigned int i = 0; i < thrd_lists->size(); i++) {
2066 if (id_to_int(rf->get_tid()) == (int)i)
2068 list = &(*thrd_lists)[i];
2070 /* Can we ensure no future writes from this thread may break
2071 * the release seq? */
2072 bool future_ordered = false;
2074 ModelAction *last = get_last_action(int_to_id(i));
2075 Thread *th = get_thread(int_to_id(i));
2076 if ((last && rf->happens_before(last)) ||
2079 future_ordered = true;
2081 ASSERT(!th->is_model_thread() || future_ordered);
2083 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2084 const ModelAction *act = *rit;
2085 /* Reach synchronization -> this thread is complete */
2086 if (act->happens_before(release))
2088 if (rf->happens_before(act)) {
2089 future_ordered = true;
2093 /* Only non-RMW writes can break release sequences */
2094 if (!act->is_write() || act->is_rmw())
2097 /* Check modification order */
2098 if (mo_graph->checkReachable(rf, act)) {
2099 /* rf --mo--> act */
2100 future_ordered = true;
2103 if (mo_graph->checkReachable(act, release))
2104 /* act --mo--> release */
2106 if (mo_graph->checkReachable(release, act) &&
2107 mo_graph->checkReachable(act, rf)) {
2108 /* release --mo-> act --mo--> rf */
2109 return true; /* complete */
2111 /* act may break release sequence */
2112 pending->writes.push_back(act);
2115 if (!future_ordered)
2116 certain = false; /* This thread is uncertain */
2120 release_heads->push_back(release);
2121 pending->writes.clear();
2123 pending->release = release;
2130 * An interface for getting the release sequence head(s) with which a
2131 * given ModelAction must synchronize. This function only returns a non-empty
2132 * result when it can locate a release sequence head with certainty. Otherwise,
2133 * it may mark the internal state of the ModelChecker so that it will handle
2134 * the release sequence at a later time, causing @a acquire to update its
2135 * synchronization at some later point in execution.
2137 * @param acquire The 'acquire' action that may synchronize with a release
2139 * @param read The read action that may read from a release sequence; this may
2140 * be the same as acquire, or else an earlier action in the same thread (i.e.,
2141 * when 'acquire' is a fence-acquire)
2142 * @param release_heads A pass-by-reference return parameter. Will be filled
2143 * with the head(s) of the release sequence(s), if they exists with certainty.
2144 * @see ModelChecker::release_seq_heads
2146 void ModelChecker::get_release_seq_heads(ModelAction *acquire,
2147 ModelAction *read, rel_heads_list_t *release_heads)
2149 const ModelAction *rf = read->get_reads_from();
2150 struct release_seq *sequence = (struct release_seq *)snapshot_calloc(1, sizeof(struct release_seq));
2151 sequence->acquire = acquire;
2152 sequence->read = read;
2154 if (!release_seq_heads(rf, release_heads, sequence)) {
2155 /* add act to 'lazy checking' list */
2156 pending_rel_seqs->push_back(sequence);
2158 snapshot_free(sequence);
2163 * Attempt to resolve all stashed operations that might synchronize with a
2164 * release sequence for a given location. This implements the "lazy" portion of
2165 * determining whether or not a release sequence was contiguous, since not all
2166 * modification order information is present at the time an action occurs.
2168 * @param location The location/object that should be checked for release
2169 * sequence resolutions. A NULL value means to check all locations.
2170 * @param work_queue The work queue to which to add work items as they are
2172 * @return True if any updates occurred (new synchronization, new mo_graph
2175 bool ModelChecker::resolve_release_sequences(void *location, work_queue_t *work_queue)
2177 bool updated = false;
2178 std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >::iterator it = pending_rel_seqs->begin();
2179 while (it != pending_rel_seqs->end()) {
2180 struct release_seq *pending = *it;
2181 ModelAction *acquire = pending->acquire;
2182 const ModelAction *read = pending->read;
2184 /* Only resolve sequences on the given location, if provided */
2185 if (location && read->get_location() != location) {
2190 const ModelAction *rf = read->get_reads_from();
2191 rel_heads_list_t release_heads;
2193 complete = release_seq_heads(rf, &release_heads, pending);
2194 for (unsigned int i = 0; i < release_heads.size(); i++) {
2195 if (!acquire->has_synchronized_with(release_heads[i])) {
2196 if (acquire->synchronize_with(release_heads[i]))
2199 set_bad_synchronization();
2204 /* Re-check all pending release sequences */
2205 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
2206 /* Re-check read-acquire for mo_graph edges */
2207 if (acquire->is_read())
2208 work_queue->push_back(MOEdgeWorkEntry(acquire));
2210 /* propagate synchronization to later actions */
2211 action_list_t::reverse_iterator rit = action_trace->rbegin();
2212 for (; (*rit) != acquire; rit++) {
2213 ModelAction *propagate = *rit;
2214 if (acquire->happens_before(propagate)) {
2215 propagate->synchronize_with(acquire);
2216 /* Re-check 'propagate' for mo_graph edges */
2217 work_queue->push_back(MOEdgeWorkEntry(propagate));
2222 it = pending_rel_seqs->erase(it);
2223 snapshot_free(pending);
2229 // If we resolved promises or data races, see if we have realized a data race.
2236 * Performs various bookkeeping operations for the current ModelAction. For
2237 * instance, adds action to the per-object, per-thread action vector and to the
2238 * action trace list of all thread actions.
2240 * @param act is the ModelAction to add.
2242 void ModelChecker::add_action_to_lists(ModelAction *act)
2244 int tid = id_to_int(act->get_tid());
2245 ModelAction *uninit = NULL;
2247 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
2248 if (list->empty() && act->is_atomic_var()) {
2249 uninit = new_uninitialized_action(act->get_location());
2250 uninit_id = id_to_int(uninit->get_tid());
2251 list->push_back(uninit);
2253 list->push_back(act);
2255 action_trace->push_back(act);
2257 action_trace->push_front(uninit);
2259 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, act->get_location());
2260 if (tid >= (int)vec->size())
2261 vec->resize(priv->next_thread_id);
2262 (*vec)[tid].push_back(act);
2264 (*vec)[uninit_id].push_front(uninit);
2266 if ((int)thrd_last_action->size() <= tid)
2267 thrd_last_action->resize(get_num_threads());
2268 (*thrd_last_action)[tid] = act;
2270 (*thrd_last_action)[uninit_id] = uninit;
2272 if (act->is_fence() && act->is_release()) {
2273 if ((int)thrd_last_fence_release->size() <= tid)
2274 thrd_last_fence_release->resize(get_num_threads());
2275 (*thrd_last_fence_release)[tid] = act;
2278 if (act->is_wait()) {
2279 void *mutex_loc = (void *) act->get_value();
2280 get_safe_ptr_action(obj_map, mutex_loc)->push_back(act);
2282 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, mutex_loc);
2283 if (tid >= (int)vec->size())
2284 vec->resize(priv->next_thread_id);
2285 (*vec)[tid].push_back(act);
2290 * @brief Get the last action performed by a particular Thread
2291 * @param tid The thread ID of the Thread in question
2292 * @return The last action in the thread
2294 ModelAction * ModelChecker::get_last_action(thread_id_t tid) const
2296 int threadid = id_to_int(tid);
2297 if (threadid < (int)thrd_last_action->size())
2298 return (*thrd_last_action)[id_to_int(tid)];
2304 * @brief Get the last fence release performed by a particular Thread
2305 * @param tid The thread ID of the Thread in question
2306 * @return The last fence release in the thread, if one exists; NULL otherwise
2308 ModelAction * ModelChecker::get_last_fence_release(thread_id_t tid) const
2310 int threadid = id_to_int(tid);
2311 if (threadid < (int)thrd_last_fence_release->size())
2312 return (*thrd_last_fence_release)[id_to_int(tid)];
2318 * Gets the last memory_order_seq_cst write (in the total global sequence)
2319 * performed on a particular object (i.e., memory location), not including the
2321 * @param curr The current ModelAction; also denotes the object location to
2323 * @return The last seq_cst write
2325 ModelAction * ModelChecker::get_last_seq_cst_write(ModelAction *curr) const
2327 void *location = curr->get_location();
2328 action_list_t *list = get_safe_ptr_action(obj_map, location);
2329 /* Find: max({i in dom(S) | seq_cst(t_i) && isWrite(t_i) && samevar(t_i, t)}) */
2330 action_list_t::reverse_iterator rit;
2331 for (rit = list->rbegin(); rit != list->rend(); rit++)
2332 if ((*rit)->is_write() && (*rit)->is_seqcst() && (*rit) != curr)
2338 * Gets the last memory_order_seq_cst fence (in the total global sequence)
2339 * performed in a particular thread, prior to a particular fence.
2340 * @param tid The ID of the thread to check
2341 * @param before_fence The fence from which to begin the search; if NULL, then
2342 * search for the most recent fence in the thread.
2343 * @return The last prior seq_cst fence in the thread, if exists; otherwise, NULL
2345 ModelAction * ModelChecker::get_last_seq_cst_fence(thread_id_t tid, const ModelAction *before_fence) const
2347 /* All fences should have NULL location */
2348 action_list_t *list = get_safe_ptr_action(obj_map, NULL);
2349 action_list_t::reverse_iterator rit = list->rbegin();
2352 for (; rit != list->rend(); rit++)
2353 if (*rit == before_fence)
2356 ASSERT(*rit == before_fence);
2360 for (; rit != list->rend(); rit++)
2361 if ((*rit)->is_fence() && (tid == (*rit)->get_tid()) && (*rit)->is_seqcst())
2367 * Gets the last unlock operation performed on a particular mutex (i.e., memory
2368 * location). This function identifies the mutex according to the current
2369 * action, which is presumed to perform on the same mutex.
2370 * @param curr The current ModelAction; also denotes the object location to
2372 * @return The last unlock operation
2374 ModelAction * ModelChecker::get_last_unlock(ModelAction *curr) const
2376 void *location = curr->get_location();
2377 action_list_t *list = get_safe_ptr_action(obj_map, location);
2378 /* Find: max({i in dom(S) | isUnlock(t_i) && samevar(t_i, t)}) */
2379 action_list_t::reverse_iterator rit;
2380 for (rit = list->rbegin(); rit != list->rend(); rit++)
2381 if ((*rit)->is_unlock() || (*rit)->is_wait())
2386 ModelAction * ModelChecker::get_parent_action(thread_id_t tid) const
2388 ModelAction *parent = get_last_action(tid);
2390 parent = get_thread(tid)->get_creation();
2395 * Returns the clock vector for a given thread.
2396 * @param tid The thread whose clock vector we want
2397 * @return Desired clock vector
2399 ClockVector * ModelChecker::get_cv(thread_id_t tid) const
2401 return get_parent_action(tid)->get_cv();
2405 * Resolve a set of Promises with a current write. The set is provided in the
2406 * Node corresponding to @a write.
2407 * @param write The ModelAction that is fulfilling Promises
2408 * @return True if promises were resolved; false otherwise
2410 bool ModelChecker::resolve_promises(ModelAction *write)
2412 bool haveResolved = false;
2413 std::vector< ModelAction *, ModelAlloc<ModelAction *> > actions_to_check;
2414 promise_list_t mustResolve, resolved;
2416 for (unsigned int i = 0, promise_index = 0; promise_index < promises->size(); i++) {
2417 Promise *promise = (*promises)[promise_index];
2418 if (write->get_node()->get_promise(i)) {
2419 ModelAction *read = promise->get_action();
2420 read_from(read, write);
2421 //Make sure the promise's value matches the write's value
2422 ASSERT(promise->is_compatible(write));
2423 mo_graph->resolvePromise(read, write, &mustResolve);
2425 resolved.push_back(promise);
2426 promises->erase(promises->begin() + promise_index);
2427 actions_to_check.push_back(read);
2429 haveResolved = true;
2434 for (unsigned int i = 0; i < mustResolve.size(); i++) {
2435 if (std::find(resolved.begin(), resolved.end(), mustResolve[i])
2437 priv->failed_promise = true;
2439 for (unsigned int i = 0; i < resolved.size(); i++)
2441 //Check whether reading these writes has made threads unable to
2444 for (unsigned int i = 0; i < actions_to_check.size(); i++) {
2445 ModelAction *read = actions_to_check[i];
2446 mo_check_promises(read, true);
2449 return haveResolved;
2453 * Compute the set of promises that could potentially be satisfied by this
2454 * action. Note that the set computation actually appears in the Node, not in
2456 * @param curr The ModelAction that may satisfy promises
2458 void ModelChecker::compute_promises(ModelAction *curr)
2460 for (unsigned int i = 0; i < promises->size(); i++) {
2461 Promise *promise = (*promises)[i];
2462 const ModelAction *act = promise->get_action();
2463 if (!act->happens_before(curr) &&
2465 !act->could_synchronize_with(curr) &&
2466 !act->same_thread(curr) &&
2467 act->get_location() == curr->get_location() &&
2468 promise->get_value() == curr->get_value()) {
2469 curr->get_node()->set_promise(i, act->is_rmw());
2474 /** Checks promises in response to change in ClockVector Threads. */
2475 void ModelChecker::check_promises(thread_id_t tid, ClockVector *old_cv, ClockVector *merge_cv)
2477 for (unsigned int i = 0; i < promises->size(); i++) {
2478 Promise *promise = (*promises)[i];
2479 const ModelAction *act = promise->get_action();
2480 if ((old_cv == NULL || !old_cv->synchronized_since(act)) &&
2481 merge_cv->synchronized_since(act)) {
2482 if (promise->eliminate_thread(tid)) {
2483 //Promise has failed
2484 priv->failed_promise = true;
2491 void ModelChecker::check_promises_thread_disabled()
2493 for (unsigned int i = 0; i < promises->size(); i++) {
2494 Promise *promise = (*promises)[i];
2495 if (promise->has_failed()) {
2496 priv->failed_promise = true;
2503 * @brief Checks promises in response to addition to modification order for
2506 * We test whether threads are still available for satisfying promises after an
2507 * addition to our modification order constraints. Those that are unavailable
2508 * are "eliminated". Once all threads are eliminated from satisfying a promise,
2509 * that promise has failed.
2511 * @param act The ModelAction which updated the modification order
2512 * @param is_read_check Should be true if act is a read and we must check for
2513 * updates to the store from which it read (there is a distinction here for
2514 * RMW's, which are both a load and a store)
2516 void ModelChecker::mo_check_promises(const ModelAction *act, bool is_read_check)
2518 const ModelAction *write = is_read_check ? act->get_reads_from() : act;
2520 for (unsigned int i = 0; i < promises->size(); i++) {
2521 Promise *promise = (*promises)[i];
2522 const ModelAction *pread = promise->get_action();
2524 // Is this promise on the same location?
2525 if (!pread->same_var(write))
2528 if (pread->happens_before(act) && mo_graph->checkPromise(write, promise)) {
2529 priv->failed_promise = true;
2533 // Don't do any lookups twice for the same thread
2534 if (!promise->thread_is_available(act->get_tid()))
2537 if (mo_graph->checkReachable(promise, write)) {
2538 if (mo_graph->checkPromise(write, promise)) {
2539 priv->failed_promise = true;
2547 * Compute the set of writes that may break the current pending release
2548 * sequence. This information is extracted from previou release sequence
2551 * @param curr The current ModelAction. Must be a release sequence fixup
2554 void ModelChecker::compute_relseq_breakwrites(ModelAction *curr)
2556 if (pending_rel_seqs->empty())
2559 struct release_seq *pending = pending_rel_seqs->back();
2560 for (unsigned int i = 0; i < pending->writes.size(); i++) {
2561 const ModelAction *write = pending->writes[i];
2562 curr->get_node()->add_relseq_break(write);
2565 /* NULL means don't break the sequence; just synchronize */
2566 curr->get_node()->add_relseq_break(NULL);
2570 * Build up an initial set of all past writes that this 'read' action may read
2571 * from, as well as any previously-observed future values that must still be valid.
2573 * @param curr is the current ModelAction that we are exploring; it must be a
2576 void ModelChecker::build_may_read_from(ModelAction *curr)
2578 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
2580 ASSERT(curr->is_read());
2582 ModelAction *last_sc_write = NULL;
2584 if (curr->is_seqcst())
2585 last_sc_write = get_last_seq_cst_write(curr);
2587 /* Iterate over all threads */
2588 for (i = 0; i < thrd_lists->size(); i++) {
2589 /* Iterate over actions in thread, starting from most recent */
2590 action_list_t *list = &(*thrd_lists)[i];
2591 action_list_t::reverse_iterator rit;
2592 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2593 ModelAction *act = *rit;
2595 /* Only consider 'write' actions */
2596 if (!act->is_write() || act == curr)
2599 /* Don't consider more than one seq_cst write if we are a seq_cst read. */
2600 bool allow_read = true;
2602 if (curr->is_seqcst() && (act->is_seqcst() || (last_sc_write != NULL && act->happens_before(last_sc_write))) && act != last_sc_write)
2604 else if (curr->get_sleep_flag() && !curr->is_seqcst() && !sleep_can_read_from(curr, act))
2608 /* Only add feasible reads */
2609 mo_graph->startChanges();
2610 r_modification_order(curr, act);
2611 if (!is_infeasible())
2612 curr->get_node()->add_read_from(act);
2613 mo_graph->rollbackChanges();
2616 /* Include at most one act per-thread that "happens before" curr */
2617 if (act->happens_before(curr))
2622 /* Inherit existing, promised future values */
2623 for (i = 0; i < promises->size(); i++) {
2624 const Promise *promise = (*promises)[i];
2625 const ModelAction *promise_read = promise->get_action();
2626 if (promise_read->same_var(curr)) {
2627 /* Only add feasible future-values */
2628 mo_graph->startChanges();
2629 r_modification_order(curr, promise);
2630 if (!is_infeasible()) {
2631 const struct future_value fv = promise->get_fv();
2632 curr->get_node()->add_future_value(fv);
2634 mo_graph->rollbackChanges();
2638 /* We may find no valid may-read-from only if the execution is doomed */
2639 if (!curr->get_node()->get_read_from_size() && curr->get_node()->future_value_empty()) {
2640 priv->no_valid_reads = true;
2644 if (DBG_ENABLED()) {
2645 model_print("Reached read action:\n");
2647 model_print("Printing may_read_from\n");
2648 curr->get_node()->print_may_read_from();
2649 model_print("End printing may_read_from\n");
2653 bool ModelChecker::sleep_can_read_from(ModelAction *curr, const ModelAction *write)
2655 for ( ; write != NULL; write = write->get_reads_from()) {
2656 /* UNINIT actions don't have a Node, and they never sleep */
2657 if (write->is_uninitialized())
2659 Node *prevnode = write->get_node()->get_parent();
2661 bool thread_sleep = prevnode->enabled_status(curr->get_tid()) == THREAD_SLEEP_SET;
2662 if (write->is_release() && thread_sleep)
2664 if (!write->is_rmw())
2671 * @brief Create a new action representing an uninitialized atomic
2672 * @param location The memory location of the atomic object
2673 * @return A pointer to a new ModelAction
2675 ModelAction * ModelChecker::new_uninitialized_action(void *location) const
2677 ModelAction *act = (ModelAction *)snapshot_malloc(sizeof(class ModelAction));
2678 act = new (act) ModelAction(ATOMIC_UNINIT, std::memory_order_relaxed, location, 0, model_thread);
2679 act->create_cv(NULL);
2683 static void print_list(action_list_t *list)
2685 action_list_t::iterator it;
2687 model_print("---------------------------------------------------------------------\n");
2689 unsigned int hash = 0;
2691 for (it = list->begin(); it != list->end(); it++) {
2693 hash = hash^(hash<<3)^((*it)->hash());
2695 model_print("HASH %u\n", hash);
2696 model_print("---------------------------------------------------------------------\n");
2699 #if SUPPORT_MOD_ORDER_DUMP
2700 void ModelChecker::dumpGraph(char *filename) const
2703 sprintf(buffer, "%s.dot", filename);
2704 FILE *file = fopen(buffer, "w");
2705 fprintf(file, "digraph %s {\n", filename);
2706 mo_graph->dumpNodes(file);
2707 ModelAction **thread_array = (ModelAction **)model_calloc(1, sizeof(ModelAction *) * get_num_threads());
2709 for (action_list_t::iterator it = action_trace->begin(); it != action_trace->end(); it++) {
2710 ModelAction *action = *it;
2711 if (action->is_read()) {
2712 fprintf(file, "N%u [label=\"N%u, T%u\"];\n", action->get_seq_number(), action->get_seq_number(), action->get_tid());
2713 if (action->get_reads_from() != NULL)
2714 fprintf(file, "N%u -> N%u[label=\"rf\", color=red];\n", action->get_seq_number(), action->get_reads_from()->get_seq_number());
2716 if (thread_array[action->get_tid()] != NULL) {
2717 fprintf(file, "N%u -> N%u[label=\"sb\", color=blue];\n", thread_array[action->get_tid()]->get_seq_number(), action->get_seq_number());
2720 thread_array[action->get_tid()] = action;
2722 fprintf(file, "}\n");
2723 model_free(thread_array);
2728 /** @brief Prints an execution trace summary. */
2729 void ModelChecker::print_summary() const
2731 #if SUPPORT_MOD_ORDER_DUMP
2732 char buffername[100];
2733 sprintf(buffername, "exec%04u", stats.num_total);
2734 mo_graph->dumpGraphToFile(buffername);
2735 sprintf(buffername, "graph%04u", stats.num_total);
2736 dumpGraph(buffername);
2739 model_print("Execution %d:", stats.num_total);
2740 if (isfeasibleprefix())
2743 print_infeasibility(" INFEASIBLE");
2744 print_list(action_trace);
2749 * Add a Thread to the system for the first time. Should only be called once
2751 * @param t The Thread to add
2753 void ModelChecker::add_thread(Thread *t)
2755 thread_map->put(id_to_int(t->get_id()), t);
2756 scheduler->add_thread(t);
2760 * Removes a thread from the scheduler.
2761 * @param the thread to remove.
2763 void ModelChecker::remove_thread(Thread *t)
2765 scheduler->remove_thread(t);
2769 * @brief Get a Thread reference by its ID
2770 * @param tid The Thread's ID
2771 * @return A Thread reference
2773 Thread * ModelChecker::get_thread(thread_id_t tid) const
2775 return thread_map->get(id_to_int(tid));
2779 * @brief Get a reference to the Thread in which a ModelAction was executed
2780 * @param act The ModelAction
2781 * @return A Thread reference
2783 Thread * ModelChecker::get_thread(const ModelAction *act) const
2785 return get_thread(act->get_tid());
2789 * @brief Check if a Thread is currently enabled
2790 * @param t The Thread to check
2791 * @return True if the Thread is currently enabled
2793 bool ModelChecker::is_enabled(Thread *t) const
2795 return scheduler->is_enabled(t);
2799 * @brief Check if a Thread is currently enabled
2800 * @param tid The ID of the Thread to check
2801 * @return True if the Thread is currently enabled
2803 bool ModelChecker::is_enabled(thread_id_t tid) const
2805 return scheduler->is_enabled(tid);
2809 * Switch from a model-checker context to a user-thread context. This is the
2810 * complement of ModelChecker::switch_to_master and must be called from the
2811 * model-checker context
2813 * @param thread The user-thread to switch to
2815 void ModelChecker::switch_from_master(Thread *thread)
2817 scheduler->set_current_thread(thread);
2818 Thread::swap(&system_context, thread);
2822 * Switch from a user-context to the "master thread" context (a.k.a. system
2823 * context). This switch is made with the intention of exploring a particular
2824 * model-checking action (described by a ModelAction object). Must be called
2825 * from a user-thread context.
2827 * @param act The current action that will be explored. May be NULL only if
2828 * trace is exiting via an assertion (see ModelChecker::set_assert and
2829 * ModelChecker::has_asserted).
2830 * @return Return the value returned by the current action
2832 uint64_t ModelChecker::switch_to_master(ModelAction *act)
2835 Thread *old = thread_current();
2836 ASSERT(!old->get_pending());
2837 old->set_pending(act);
2838 if (Thread::swap(old, &system_context) < 0) {
2839 perror("swap threads");
2842 return old->get_return_value();
2846 * Takes the next step in the execution, if possible.
2847 * @param curr The current step to take
2848 * @return Returns the next Thread to run, if any; NULL if this execution
2851 Thread * ModelChecker::take_step(ModelAction *curr)
2853 Thread *curr_thrd = get_thread(curr);
2854 ASSERT(curr_thrd->get_state() == THREAD_READY);
2856 curr = check_current_action(curr);
2858 /* Infeasible -> don't take any more steps */
2859 if (is_infeasible())
2861 else if (isfeasibleprefix() && have_bug_reports()) {
2866 if (params.bound != 0 && priv->used_sequence_numbers > params.bound)
2869 if (curr_thrd->is_blocked() || curr_thrd->is_complete())
2870 scheduler->remove_thread(curr_thrd);
2872 Thread *next_thrd = get_next_thread(curr);
2874 DEBUG("(%d, %d)\n", curr_thrd ? id_to_int(curr_thrd->get_id()) : -1,
2875 next_thrd ? id_to_int(next_thrd->get_id()) : -1);
2880 /** Wrapper to run the user's main function, with appropriate arguments */
2881 void user_main_wrapper(void *)
2883 user_main(model->params.argc, model->params.argv);
2886 /** @brief Run ModelChecker for the user program */
2887 void ModelChecker::run()
2891 Thread *t = new Thread(&user_thread, &user_main_wrapper, NULL);
2896 * Stash next pending action(s) for thread(s). There
2897 * should only need to stash one thread's action--the
2898 * thread which just took a step--plus the first step
2899 * for any newly-created thread
2901 for (unsigned int i = 0; i < get_num_threads(); i++) {
2902 thread_id_t tid = int_to_id(i);
2903 Thread *thr = get_thread(tid);
2904 if (!thr->is_model_thread() && !thr->is_complete() && !thr->get_pending()) {
2905 switch_from_master(thr);
2909 /* Catch assertions from prior take_step or from
2910 * between-ModelAction bugs (e.g., data races) */
2914 /* Consume the next action for a Thread */
2915 ModelAction *curr = t->get_pending();
2916 t->set_pending(NULL);
2917 t = take_step(curr);
2918 } while (t && !t->is_model_thread());
2921 * Launch end-of-execution release sequence fixups only when
2922 * the execution is otherwise feasible AND there are:
2924 * (1) pending release sequences
2925 * (2) pending assertions that could be invalidated by a change
2926 * in clock vectors (i.e., data races)
2927 * (3) no pending promises
2929 while (!pending_rel_seqs->empty() &&
2930 is_feasible_prefix_ignore_relseq() &&
2931 !unrealizedraces.empty()) {
2932 model_print("*** WARNING: release sequence fixup action "
2933 "(%zu pending release seuqence(s)) ***\n",
2934 pending_rel_seqs->size());
2935 ModelAction *fixup = new ModelAction(MODEL_FIXUP_RELSEQ,
2936 std::memory_order_seq_cst, NULL, VALUE_NONE,
2940 } while (next_execution());
2942 model_print("******* Model-checking complete: *******\n");