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 SnapVector<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 condvar_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
87 obj_thrd_map(new HashTable<void *, SnapVector<action_list_t> *, uintptr_t, 4 >()),
88 promises(new SnapVector<Promise *>()),
89 futurevalues(new SnapVector<struct PendingFutureValue>()),
90 pending_rel_seqs(new SnapVector<struct release_seq *>()),
91 thrd_last_action(new SnapVector<ModelAction *>(1)),
92 thrd_last_fence_release(new SnapVector<ModelAction *>()),
93 node_stack(new NodeStack()),
94 priv(new struct model_snapshot_members()),
95 mo_graph(new CycleGraph())
97 /* Initialize a model-checker thread, for special ModelActions */
98 model_thread = new Thread(get_next_id());
99 thread_map->put(id_to_int(model_thread->get_id()), model_thread);
102 /** @brief Destructor */
103 ModelChecker::~ModelChecker()
105 for (unsigned int i = 0; i < get_num_threads(); i++)
106 delete thread_map->get(i);
111 delete condvar_waiters_map;
114 for (unsigned int i = 0; i < promises->size(); i++)
115 delete (*promises)[i];
118 delete pending_rel_seqs;
120 delete thrd_last_action;
121 delete thrd_last_fence_release;
128 static action_list_t * get_safe_ptr_action(HashTable<const void *, action_list_t *, uintptr_t, 4> * hash, void * ptr)
130 action_list_t *tmp = hash->get(ptr);
132 tmp = new action_list_t();
138 static SnapVector<action_list_t> * get_safe_ptr_vect_action(HashTable<void *, SnapVector<action_list_t> *, uintptr_t, 4> * hash, void * ptr)
140 SnapVector<action_list_t> *tmp = hash->get(ptr);
142 tmp = new SnapVector<action_list_t>();
149 * Restores user program to initial state and resets all model-checker data
152 void ModelChecker::reset_to_initial_state()
154 DEBUG("+++ Resetting to initial state +++\n");
155 node_stack->reset_execution();
157 /* Print all model-checker output before rollback */
161 * FIXME: if we utilize partial rollback, we will need to free only
162 * those pending actions which were NOT pending before the rollback
165 for (unsigned int i = 0; i < get_num_threads(); i++)
166 delete get_thread(int_to_id(i))->get_pending();
168 snapshot_backtrack_before(0);
171 /** @return a thread ID for a new Thread */
172 thread_id_t ModelChecker::get_next_id()
174 return priv->next_thread_id++;
177 /** @return the number of user threads created during this execution */
178 unsigned int ModelChecker::get_num_threads() const
180 return priv->next_thread_id;
184 * Must be called from user-thread context (e.g., through the global
185 * thread_current() interface)
187 * @return The currently executing Thread.
189 Thread * ModelChecker::get_current_thread() const
191 return scheduler->get_current_thread();
194 /** @return a sequence number for a new ModelAction */
195 modelclock_t ModelChecker::get_next_seq_num()
197 return ++priv->used_sequence_numbers;
200 Node * ModelChecker::get_curr_node() const
202 return node_stack->get_head();
206 * @brief Select the next thread to execute based on the curren action
208 * RMW actions occur in two parts, and we cannot split them. And THREAD_CREATE
209 * actions should be followed by the execution of their child thread. In either
210 * case, the current action should determine the next thread schedule.
212 * @param curr The current action
213 * @return The next thread to run, if the current action will determine this
214 * selection; otherwise NULL
216 Thread * ModelChecker::action_select_next_thread(const ModelAction *curr) const
218 /* Do not split atomic RMW */
220 return get_thread(curr);
221 /* Follow CREATE with the created thread */
222 if (curr->get_type() == THREAD_CREATE)
223 return curr->get_thread_operand();
228 * @brief Choose the next thread to execute.
230 * This function chooses the next thread that should execute. It can enforce
231 * execution replay/backtracking or, if the model-checker has no preference
232 * regarding the next thread (i.e., when exploring a new execution ordering),
233 * we defer to the scheduler.
235 * @return The next chosen thread to run, if any exist. Or else if the current
236 * execution should terminate, return NULL.
238 Thread * ModelChecker::get_next_thread()
243 * Have we completed exploring the preselected path? Then let the
247 return scheduler->select_next_thread();
249 /* Else, we are trying to replay an execution */
250 ModelAction *next = node_stack->get_next()->get_action();
252 if (next == diverge) {
253 if (earliest_diverge == NULL || *diverge < *earliest_diverge)
254 earliest_diverge = diverge;
256 Node *nextnode = next->get_node();
257 Node *prevnode = nextnode->get_parent();
258 scheduler->update_sleep_set(prevnode);
260 /* Reached divergence point */
261 if (nextnode->increment_behaviors()) {
262 /* Execute the same thread with a new behavior */
263 tid = next->get_tid();
264 node_stack->pop_restofstack(2);
267 /* Make a different thread execute for next step */
268 scheduler->add_sleep(get_thread(next->get_tid()));
269 tid = prevnode->get_next_backtrack();
270 /* Make sure the backtracked thread isn't sleeping. */
271 node_stack->pop_restofstack(1);
272 if (diverge == earliest_diverge) {
273 earliest_diverge = prevnode->get_action();
276 /* Start the round robin scheduler from this thread id */
277 scheduler->set_scheduler_thread(tid);
278 /* The correct sleep set is in the parent node. */
281 DEBUG("*** Divergence point ***\n");
285 tid = next->get_tid();
287 DEBUG("*** ModelChecker chose next thread = %d ***\n", id_to_int(tid));
288 ASSERT(tid != THREAD_ID_T_NONE);
289 return get_thread(id_to_int(tid));
293 * We need to know what the next actions of all threads in the sleep
294 * set will be. This method computes them and stores the actions at
295 * the corresponding thread object's pending action.
298 void ModelChecker::execute_sleep_set()
300 for (unsigned int i = 0; i < get_num_threads(); i++) {
301 thread_id_t tid = int_to_id(i);
302 Thread *thr = get_thread(tid);
303 if (scheduler->is_sleep_set(thr) && thr->get_pending()) {
304 thr->get_pending()->set_sleep_flag();
310 * @brief Should the current action wake up a given thread?
312 * @param curr The current action
313 * @param thread The thread that we might wake up
314 * @return True, if we should wake up the sleeping thread; false otherwise
316 bool ModelChecker::should_wake_up(const ModelAction *curr, const Thread *thread) const
318 const ModelAction *asleep = thread->get_pending();
319 /* Don't allow partial RMW to wake anyone up */
322 /* Synchronizing actions may have been backtracked */
323 if (asleep->could_synchronize_with(curr))
325 /* All acquire/release fences and fence-acquire/store-release */
326 if (asleep->is_fence() && asleep->is_acquire() && curr->is_release())
328 /* Fence-release + store can awake load-acquire on the same location */
329 if (asleep->is_read() && asleep->is_acquire() && curr->same_var(asleep) && curr->is_write()) {
330 ModelAction *fence_release = get_last_fence_release(curr->get_tid());
331 if (fence_release && *(get_last_action(thread->get_id())) < *fence_release)
337 void ModelChecker::wake_up_sleeping_actions(ModelAction *curr)
339 for (unsigned int i = 0; i < get_num_threads(); i++) {
340 Thread *thr = get_thread(int_to_id(i));
341 if (scheduler->is_sleep_set(thr)) {
342 if (should_wake_up(curr, thr))
343 /* Remove this thread from sleep set */
344 scheduler->remove_sleep(thr);
349 /** @brief Alert the model-checker that an incorrectly-ordered
350 * synchronization was made */
351 void ModelChecker::set_bad_synchronization()
353 priv->bad_synchronization = true;
357 * Check whether the current trace has triggered an assertion which should halt
360 * @return True, if the execution should be aborted; false otherwise
362 bool ModelChecker::has_asserted() const
364 return priv->asserted;
368 * Trigger a trace assertion which should cause this execution to be halted.
369 * This can be due to a detected bug or due to an infeasibility that should
372 void ModelChecker::set_assert()
374 priv->asserted = true;
378 * Check if we are in a deadlock. Should only be called at the end of an
379 * execution, although it should not give false positives in the middle of an
380 * execution (there should be some ENABLED thread).
382 * @return True if program is in a deadlock; false otherwise
384 bool ModelChecker::is_deadlocked() const
386 bool blocking_threads = false;
387 for (unsigned int i = 0; i < get_num_threads(); i++) {
388 thread_id_t tid = int_to_id(i);
391 Thread *t = get_thread(tid);
392 if (!t->is_model_thread() && t->get_pending())
393 blocking_threads = true;
395 return blocking_threads;
399 * Check if this is a complete execution. That is, have all thread completed
400 * execution (rather than exiting because sleep sets have forced a redundant
403 * @return True if the execution is complete.
405 bool ModelChecker::is_complete_execution() const
407 for (unsigned int i = 0; i < get_num_threads(); i++)
408 if (is_enabled(int_to_id(i)))
414 * @brief Assert a bug in the executing program.
416 * Use this function to assert any sort of bug in the user program. If the
417 * current trace is feasible (actually, a prefix of some feasible execution),
418 * then this execution will be aborted, printing the appropriate message. If
419 * the current trace is not yet feasible, the error message will be stashed and
420 * printed if the execution ever becomes feasible.
422 * @param msg Descriptive message for the bug (do not include newline char)
423 * @return True if bug is immediately-feasible
425 bool ModelChecker::assert_bug(const char *msg)
427 priv->bugs.push_back(new bug_message(msg));
429 if (isfeasibleprefix()) {
437 * @brief Assert a bug in the executing program, asserted by a user thread
438 * @see ModelChecker::assert_bug
439 * @param msg Descriptive message for the bug (do not include newline char)
441 void ModelChecker::assert_user_bug(const char *msg)
443 /* If feasible bug, bail out now */
445 switch_to_master(NULL);
448 /** @return True, if any bugs have been reported for this execution */
449 bool ModelChecker::have_bug_reports() const
451 return priv->bugs.size() != 0;
454 /** @brief Print bug report listing for this execution (if any bugs exist) */
455 void ModelChecker::print_bugs() const
457 if (have_bug_reports()) {
458 model_print("Bug report: %zu bug%s detected\n",
460 priv->bugs.size() > 1 ? "s" : "");
461 for (unsigned int i = 0; i < priv->bugs.size(); i++)
462 priv->bugs[i]->print();
467 * @brief Record end-of-execution stats
469 * Must be run when exiting an execution. Records various stats.
470 * @see struct execution_stats
472 void ModelChecker::record_stats()
475 if (!isfeasibleprefix())
476 stats.num_infeasible++;
477 else if (have_bug_reports())
478 stats.num_buggy_executions++;
479 else if (is_complete_execution())
480 stats.num_complete++;
482 stats.num_redundant++;
485 * @todo We can violate this ASSERT() when fairness/sleep sets
486 * conflict to cause an execution to terminate, e.g. with:
487 * Scheduler: [0: disabled][1: disabled][2: sleep][3: current, enabled]
489 //ASSERT(scheduler->all_threads_sleeping());
493 /** @brief Print execution stats */
494 void ModelChecker::print_stats() const
496 model_print("Number of complete, bug-free executions: %d\n", stats.num_complete);
497 model_print("Number of redundant executions: %d\n", stats.num_redundant);
498 model_print("Number of buggy executions: %d\n", stats.num_buggy_executions);
499 model_print("Number of infeasible executions: %d\n", stats.num_infeasible);
500 model_print("Total executions: %d\n", stats.num_total);
501 model_print("Total nodes created: %d\n", node_stack->get_total_nodes());
505 * @brief End-of-exeuction print
506 * @param printbugs Should any existing bugs be printed?
508 void ModelChecker::print_execution(bool printbugs) const
510 print_program_output();
512 if (params.verbose) {
513 model_print("Earliest divergence point since last feasible execution:\n");
514 if (earliest_diverge)
515 earliest_diverge->print();
517 model_print("(Not set)\n");
523 /* Don't print invalid bugs */
532 * Queries the model-checker for more executions to explore and, if one
533 * exists, resets the model-checker state to execute a new execution.
535 * @return If there are more executions to explore, return true. Otherwise,
538 bool ModelChecker::next_execution()
541 /* Is this execution a feasible execution that's worth bug-checking? */
542 bool complete = isfeasibleprefix() && (is_complete_execution() ||
545 /* End-of-execution bug checks */
548 assert_bug("Deadlock detected");
556 if (params.verbose || (complete && have_bug_reports()))
557 print_execution(complete);
559 clear_program_output();
562 earliest_diverge = NULL;
564 if ((diverge = get_next_backtrack()) == NULL)
568 model_print("Next execution will diverge at:\n");
572 reset_to_initial_state();
577 * @brief Find the last fence-related backtracking conflict for a ModelAction
579 * This function performs the search for the most recent conflicting action
580 * against which we should perform backtracking, as affected by fence
581 * operations. This includes pairs of potentially-synchronizing actions which
582 * occur due to fence-acquire or fence-release, and hence should be explored in
583 * the opposite execution order.
585 * @param act The current action
586 * @return The most recent action which conflicts with act due to fences
588 ModelAction * ModelChecker::get_last_fence_conflict(ModelAction *act) const
590 /* Only perform release/acquire fence backtracking for stores */
591 if (!act->is_write())
594 /* Find a fence-release (or, act is a release) */
595 ModelAction *last_release;
596 if (act->is_release())
599 last_release = get_last_fence_release(act->get_tid());
603 /* Skip past the release */
604 action_list_t *list = action_trace;
605 action_list_t::reverse_iterator rit;
606 for (rit = list->rbegin(); rit != list->rend(); rit++)
607 if (*rit == last_release)
609 ASSERT(rit != list->rend());
614 * load --sb-> fence-acquire */
615 ModelVector<ModelAction *> acquire_fences(get_num_threads(), NULL);
616 ModelVector<ModelAction *> prior_loads(get_num_threads(), NULL);
617 bool found_acquire_fences = false;
618 for ( ; rit != list->rend(); rit++) {
619 ModelAction *prev = *rit;
620 if (act->same_thread(prev))
623 int tid = id_to_int(prev->get_tid());
625 if (prev->is_read() && act->same_var(prev)) {
626 if (prev->is_acquire()) {
627 /* Found most recent load-acquire, don't need
628 * to search for more fences */
629 if (!found_acquire_fences)
632 prior_loads[tid] = prev;
635 if (prev->is_acquire() && prev->is_fence() && !acquire_fences[tid]) {
636 found_acquire_fences = true;
637 acquire_fences[tid] = prev;
641 ModelAction *latest_backtrack = NULL;
642 for (unsigned int i = 0; i < acquire_fences.size(); i++)
643 if (acquire_fences[i] && prior_loads[i])
644 if (!latest_backtrack || *latest_backtrack < *acquire_fences[i])
645 latest_backtrack = acquire_fences[i];
646 return latest_backtrack;
650 * @brief Find the last backtracking conflict for a ModelAction
652 * This function performs the search for the most recent conflicting action
653 * against which we should perform backtracking. This primary includes pairs of
654 * synchronizing actions which should be explored in the opposite execution
657 * @param act The current action
658 * @return The most recent action which conflicts with act
660 ModelAction * ModelChecker::get_last_conflict(ModelAction *act) const
662 switch (act->get_type()) {
663 /* case ATOMIC_FENCE: fences don't directly cause backtracking */
667 ModelAction *ret = NULL;
669 /* linear search: from most recent to oldest */
670 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
671 action_list_t::reverse_iterator rit;
672 for (rit = list->rbegin(); rit != list->rend(); rit++) {
673 ModelAction *prev = *rit;
674 if (prev->could_synchronize_with(act)) {
680 ModelAction *ret2 = get_last_fence_conflict(act);
690 case ATOMIC_TRYLOCK: {
691 /* linear search: from most recent to oldest */
692 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
693 action_list_t::reverse_iterator rit;
694 for (rit = list->rbegin(); rit != list->rend(); rit++) {
695 ModelAction *prev = *rit;
696 if (act->is_conflicting_lock(prev))
701 case ATOMIC_UNLOCK: {
702 /* linear search: from most recent to oldest */
703 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
704 action_list_t::reverse_iterator rit;
705 for (rit = list->rbegin(); rit != list->rend(); rit++) {
706 ModelAction *prev = *rit;
707 if (!act->same_thread(prev) && prev->is_failed_trylock())
713 /* linear search: from most recent to oldest */
714 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
715 action_list_t::reverse_iterator rit;
716 for (rit = list->rbegin(); rit != list->rend(); rit++) {
717 ModelAction *prev = *rit;
718 if (!act->same_thread(prev) && prev->is_failed_trylock())
720 if (!act->same_thread(prev) && prev->is_notify())
726 case ATOMIC_NOTIFY_ALL:
727 case ATOMIC_NOTIFY_ONE: {
728 /* linear search: from most recent to oldest */
729 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
730 action_list_t::reverse_iterator rit;
731 for (rit = list->rbegin(); rit != list->rend(); rit++) {
732 ModelAction *prev = *rit;
733 if (!act->same_thread(prev) && prev->is_wait())
744 /** This method finds backtracking points where we should try to
745 * reorder the parameter ModelAction against.
747 * @param the ModelAction to find backtracking points for.
749 void ModelChecker::set_backtracking(ModelAction *act)
751 Thread *t = get_thread(act);
752 ModelAction *prev = get_last_conflict(act);
756 Node *node = prev->get_node()->get_parent();
758 /* See Dynamic Partial Order Reduction (addendum), POPL '05 */
759 int low_tid, high_tid;
760 if (node->enabled_status(t->get_id()) == THREAD_ENABLED) {
761 low_tid = id_to_int(act->get_tid());
762 high_tid = low_tid + 1;
765 high_tid = get_num_threads();
768 for (int i = low_tid; i < high_tid; i++) {
769 thread_id_t tid = int_to_id(i);
771 /* Make sure this thread can be enabled here. */
772 if (i >= node->get_num_threads())
775 /* See Dynamic Partial Order Reduction (addendum), POPL '05 */
776 /* Don't backtrack into a point where the thread is disabled or sleeping. */
777 if (node->enabled_status(tid) != THREAD_ENABLED)
780 /* Check if this has been explored already */
781 if (node->has_been_explored(tid))
784 /* See if fairness allows */
785 if (model->params.fairwindow != 0 && !node->has_priority(tid)) {
787 for (int t = 0; t < node->get_num_threads(); t++) {
788 thread_id_t tother = int_to_id(t);
789 if (node->is_enabled(tother) && node->has_priority(tother)) {
798 /* See if CHESS-like yield fairness allows */
799 if (model->params.yieldon) {
801 for (int t = 0; t < node->get_num_threads(); t++) {
802 thread_id_t tother = int_to_id(t);
803 if (node->is_enabled(tother) && node->has_priority_over(tid, tother)) {
812 /* Cache the latest backtracking point */
813 set_latest_backtrack(prev);
815 /* If this is a new backtracking point, mark the tree */
816 if (!node->set_backtrack(tid))
818 DEBUG("Setting backtrack: conflict = %d, instead tid = %d\n",
819 id_to_int(prev->get_tid()),
820 id_to_int(t->get_id()));
829 * @brief Cache the a backtracking point as the "most recent", if eligible
831 * Note that this does not prepare the NodeStack for this backtracking
832 * operation, it only caches the action on a per-execution basis
834 * @param act The operation at which we should explore a different next action
835 * (i.e., backtracking point)
836 * @return True, if this action is now the most recent backtracking point;
839 bool ModelChecker::set_latest_backtrack(ModelAction *act)
841 if (!priv->next_backtrack || *act > *priv->next_backtrack) {
842 priv->next_backtrack = act;
849 * Returns last backtracking point. The model checker will explore a different
850 * path for this point in the next execution.
851 * @return The ModelAction at which the next execution should diverge.
853 ModelAction * ModelChecker::get_next_backtrack()
855 ModelAction *next = priv->next_backtrack;
856 priv->next_backtrack = NULL;
861 * Processes a read model action.
862 * @param curr is the read model action to process.
863 * @return True if processing this read updates the mo_graph.
865 bool ModelChecker::process_read(ModelAction *curr)
867 Node *node = curr->get_node();
869 bool updated = false;
870 switch (node->get_read_from_status()) {
871 case READ_FROM_PAST: {
872 const ModelAction *rf = node->get_read_from_past();
875 mo_graph->startChanges();
877 ASSERT(!is_infeasible());
878 if (!check_recency(curr, rf)) {
879 if (node->increment_read_from()) {
880 mo_graph->rollbackChanges();
883 priv->too_many_reads = true;
887 updated = r_modification_order(curr, rf);
889 mo_graph->commitChanges();
890 mo_check_promises(curr, true);
893 case READ_FROM_PROMISE: {
894 Promise *promise = curr->get_node()->get_read_from_promise();
895 if (promise->add_reader(curr))
896 priv->failed_promise = true;
897 curr->set_read_from_promise(promise);
898 mo_graph->startChanges();
899 if (!check_recency(curr, promise))
900 priv->too_many_reads = true;
901 updated = r_modification_order(curr, promise);
902 mo_graph->commitChanges();
905 case READ_FROM_FUTURE: {
906 /* Read from future value */
907 struct future_value fv = node->get_future_value();
908 Promise *promise = new Promise(curr, fv);
909 curr->set_read_from_promise(promise);
910 promises->push_back(promise);
911 mo_graph->startChanges();
912 updated = r_modification_order(curr, promise);
913 mo_graph->commitChanges();
919 get_thread(curr)->set_return_value(curr->get_return_value());
925 * Processes a lock, trylock, or unlock model action. @param curr is
926 * the read model action to process.
928 * The try lock operation checks whether the lock is taken. If not,
929 * it falls to the normal lock operation case. If so, it returns
932 * The lock operation has already been checked that it is enabled, so
933 * it just grabs the lock and synchronizes with the previous unlock.
935 * The unlock operation has to re-enable all of the threads that are
936 * waiting on the lock.
938 * @return True if synchronization was updated; false otherwise
940 bool ModelChecker::process_mutex(ModelAction *curr)
942 std::mutex *mutex = curr->get_mutex();
943 struct std::mutex_state *state = NULL;
946 state = mutex->get_state();
948 switch (curr->get_type()) {
949 case ATOMIC_TRYLOCK: {
950 bool success = !state->locked;
951 curr->set_try_lock(success);
953 get_thread(curr)->set_return_value(0);
956 get_thread(curr)->set_return_value(1);
958 //otherwise fall into the lock case
960 if (curr->get_cv()->getClock(state->alloc_tid) <= state->alloc_clock)
961 assert_bug("Lock access before initialization");
962 state->locked = get_thread(curr);
963 ModelAction *unlock = get_last_unlock(curr);
964 //synchronize with the previous unlock statement
965 if (unlock != NULL) {
966 curr->synchronize_with(unlock);
972 case ATOMIC_UNLOCK: {
973 /* wake up the other threads */
974 for (unsigned int i = 0; i < get_num_threads(); i++) {
975 Thread *t = get_thread(int_to_id(i));
976 Thread *curr_thrd = get_thread(curr);
977 if (t->waiting_on() == curr_thrd && t->get_pending()->is_lock())
981 /* unlock the lock - after checking who was waiting on it */
982 state->locked = NULL;
984 if (!curr->is_wait())
985 break; /* The rest is only for ATOMIC_WAIT */
987 /* Should we go to sleep? (simulate spurious failures) */
988 if (curr->get_node()->get_misc() == 0) {
989 get_safe_ptr_action(condvar_waiters_map, curr->get_location())->push_back(curr);
991 scheduler->sleep(get_thread(curr));
995 case ATOMIC_NOTIFY_ALL: {
996 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
997 //activate all the waiting threads
998 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
999 scheduler->wake(get_thread(*rit));
1004 case ATOMIC_NOTIFY_ONE: {
1005 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
1006 int wakeupthread = curr->get_node()->get_misc();
1007 action_list_t::iterator it = waiters->begin();
1008 advance(it, wakeupthread);
1009 scheduler->wake(get_thread(*it));
1021 * @brief Check if the current pending promises allow a future value to be sent
1023 * If one of the following is true:
1024 * (a) there are no pending promises
1025 * (b) the reader and writer do not cross any promises
1026 * Then, it is safe to pass a future value back now.
1028 * Otherwise, we must save the pending future value until (a) or (b) is true
1030 * @param writer The operation which sends the future value. Must be a write.
1031 * @param reader The operation which will observe the value. Must be a read.
1032 * @return True if the future value can be sent now; false if it must wait.
1034 bool ModelChecker::promises_may_allow(const ModelAction *writer,
1035 const ModelAction *reader) const
1037 if (promises->empty())
1039 for(int i=promises->size()-1;i>=0;i--) {
1040 ModelAction *pr=(*promises)[i]->get_reader(0);
1041 //reader is after promise...doesn't cross any promise
1044 //writer is after promise, reader before...bad...
1052 * @brief Add a future value to a reader
1054 * This function performs a few additional checks to ensure that the future
1055 * value can be feasibly observed by the reader
1057 * @param writer The operation whose value is sent. Must be a write.
1058 * @param reader The read operation which may read the future value. Must be a read.
1060 void ModelChecker::add_future_value(const ModelAction *writer, ModelAction *reader)
1062 /* Do more ambitious checks now that mo is more complete */
1063 if (!mo_may_allow(writer, reader))
1066 Node *node = reader->get_node();
1068 /* Find an ancestor thread which exists at the time of the reader */
1069 Thread *write_thread = get_thread(writer);
1070 while (id_to_int(write_thread->get_id()) >= node->get_num_threads())
1071 write_thread = write_thread->get_parent();
1073 struct future_value fv = {
1074 writer->get_write_value(),
1075 writer->get_seq_number() + params.maxfuturedelay,
1076 write_thread->get_id(),
1078 if (node->add_future_value(fv))
1079 set_latest_backtrack(reader);
1083 * Process a write ModelAction
1084 * @param curr The ModelAction to process
1085 * @return True if the mo_graph was updated or promises were resolved
1087 bool ModelChecker::process_write(ModelAction *curr)
1089 /* Readers to which we may send our future value */
1090 ModelVector<ModelAction *> send_fv;
1092 const ModelAction *earliest_promise_reader;
1093 bool updated_promises = false;
1095 bool updated_mod_order = w_modification_order(curr, &send_fv);
1096 Promise *promise = pop_promise_to_resolve(curr);
1099 earliest_promise_reader = promise->get_reader(0);
1100 updated_promises = resolve_promise(curr, promise);
1102 earliest_promise_reader = NULL;
1104 for (unsigned int i = 0; i < send_fv.size(); i++) {
1105 ModelAction *read = send_fv[i];
1107 /* Don't send future values to reads after the Promise we resolve */
1108 if (!earliest_promise_reader || *read < *earliest_promise_reader) {
1109 /* Check if future value can be sent immediately */
1110 if (promises_may_allow(curr, read)) {
1111 add_future_value(curr, read);
1113 futurevalues->push_back(PendingFutureValue(curr, read));
1118 /* Check the pending future values */
1119 for (int i = (int)futurevalues->size() - 1; i >= 0; i--) {
1120 struct PendingFutureValue pfv = (*futurevalues)[i];
1121 if (promises_may_allow(pfv.writer, pfv.reader)) {
1122 add_future_value(pfv.writer, pfv.reader);
1123 futurevalues->erase(futurevalues->begin() + i);
1127 mo_graph->commitChanges();
1128 mo_check_promises(curr, false);
1130 get_thread(curr)->set_return_value(VALUE_NONE);
1131 return updated_mod_order || updated_promises;
1135 * Process a fence ModelAction
1136 * @param curr The ModelAction to process
1137 * @return True if synchronization was updated
1139 bool ModelChecker::process_fence(ModelAction *curr)
1142 * fence-relaxed: no-op
1143 * fence-release: only log the occurence (not in this function), for
1144 * use in later synchronization
1145 * fence-acquire (this function): search for hypothetical release
1147 * fence-seq-cst: MO constraints formed in {r,w}_modification_order
1149 bool updated = false;
1150 if (curr->is_acquire()) {
1151 action_list_t *list = action_trace;
1152 action_list_t::reverse_iterator rit;
1153 /* Find X : is_read(X) && X --sb-> curr */
1154 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1155 ModelAction *act = *rit;
1158 if (act->get_tid() != curr->get_tid())
1160 /* Stop at the beginning of the thread */
1161 if (act->is_thread_start())
1163 /* Stop once we reach a prior fence-acquire */
1164 if (act->is_fence() && act->is_acquire())
1166 if (!act->is_read())
1168 /* read-acquire will find its own release sequences */
1169 if (act->is_acquire())
1172 /* Establish hypothetical release sequences */
1173 rel_heads_list_t release_heads;
1174 get_release_seq_heads(curr, act, &release_heads);
1175 for (unsigned int i = 0; i < release_heads.size(); i++)
1176 if (!curr->synchronize_with(release_heads[i]))
1177 set_bad_synchronization();
1178 if (release_heads.size() != 0)
1186 * @brief Process the current action for thread-related activity
1188 * Performs current-action processing for a THREAD_* ModelAction. Proccesses
1189 * may include setting Thread status, completing THREAD_FINISH/THREAD_JOIN
1190 * synchronization, etc. This function is a no-op for non-THREAD actions
1191 * (e.g., ATOMIC_{READ,WRITE,RMW,LOCK}, etc.)
1193 * @param curr The current action
1194 * @return True if synchronization was updated or a thread completed
1196 bool ModelChecker::process_thread_action(ModelAction *curr)
1198 bool updated = false;
1200 switch (curr->get_type()) {
1201 case THREAD_CREATE: {
1202 thrd_t *thrd = (thrd_t *)curr->get_location();
1203 struct thread_params *params = (struct thread_params *)curr->get_value();
1204 Thread *th = new Thread(thrd, params->func, params->arg, get_thread(curr));
1206 th->set_creation(curr);
1207 /* Promises can be satisfied by children */
1208 for (unsigned int i = 0; i < promises->size(); i++) {
1209 Promise *promise = (*promises)[i];
1210 if (promise->thread_is_available(curr->get_tid()))
1211 promise->add_thread(th->get_id());
1216 Thread *blocking = curr->get_thread_operand();
1217 ModelAction *act = get_last_action(blocking->get_id());
1218 curr->synchronize_with(act);
1219 updated = true; /* trigger rel-seq checks */
1222 case THREAD_FINISH: {
1223 Thread *th = get_thread(curr);
1224 /* Wake up any joining threads */
1225 for (unsigned int i = 0; i < get_num_threads(); i++) {
1226 Thread *waiting = get_thread(int_to_id(i));
1227 if (waiting->waiting_on() == th &&
1228 waiting->get_pending()->is_thread_join())
1229 scheduler->wake(waiting);
1232 /* Completed thread can't satisfy promises */
1233 for (unsigned int i = 0; i < promises->size(); i++) {
1234 Promise *promise = (*promises)[i];
1235 if (promise->thread_is_available(th->get_id()))
1236 if (promise->eliminate_thread(th->get_id()))
1237 priv->failed_promise = true;
1239 updated = true; /* trigger rel-seq checks */
1242 case THREAD_START: {
1243 check_promises(curr->get_tid(), NULL, curr->get_cv());
1254 * @brief Process the current action for release sequence fixup activity
1256 * Performs model-checker release sequence fixups for the current action,
1257 * forcing a single pending release sequence to break (with a given, potential
1258 * "loose" write) or to complete (i.e., synchronize). If a pending release
1259 * sequence forms a complete release sequence, then we must perform the fixup
1260 * synchronization, mo_graph additions, etc.
1262 * @param curr The current action; must be a release sequence fixup action
1263 * @param work_queue The work queue to which to add work items as they are
1266 void ModelChecker::process_relseq_fixup(ModelAction *curr, work_queue_t *work_queue)
1268 const ModelAction *write = curr->get_node()->get_relseq_break();
1269 struct release_seq *sequence = pending_rel_seqs->back();
1270 pending_rel_seqs->pop_back();
1272 ModelAction *acquire = sequence->acquire;
1273 const ModelAction *rf = sequence->rf;
1274 const ModelAction *release = sequence->release;
1278 ASSERT(release->same_thread(rf));
1280 if (write == NULL) {
1282 * @todo Forcing a synchronization requires that we set
1283 * modification order constraints. For instance, we can't allow
1284 * a fixup sequence in which two separate read-acquire
1285 * operations read from the same sequence, where the first one
1286 * synchronizes and the other doesn't. Essentially, we can't
1287 * allow any writes to insert themselves between 'release' and
1291 /* Must synchronize */
1292 if (!acquire->synchronize_with(release)) {
1293 set_bad_synchronization();
1296 /* Re-check all pending release sequences */
1297 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
1298 /* Re-check act for mo_graph edges */
1299 work_queue->push_back(MOEdgeWorkEntry(acquire));
1301 /* propagate synchronization to later actions */
1302 action_list_t::reverse_iterator rit = action_trace->rbegin();
1303 for (; (*rit) != acquire; rit++) {
1304 ModelAction *propagate = *rit;
1305 if (acquire->happens_before(propagate)) {
1306 propagate->synchronize_with(acquire);
1307 /* Re-check 'propagate' for mo_graph edges */
1308 work_queue->push_back(MOEdgeWorkEntry(propagate));
1312 /* Break release sequence with new edges:
1313 * release --mo--> write --mo--> rf */
1314 mo_graph->addEdge(release, write);
1315 mo_graph->addEdge(write, rf);
1318 /* See if we have realized a data race */
1323 * Initialize the current action by performing one or more of the following
1324 * actions, as appropriate: merging RMWR and RMWC/RMW actions, stepping forward
1325 * in the NodeStack, manipulating backtracking sets, allocating and
1326 * initializing clock vectors, and computing the promises to fulfill.
1328 * @param curr The current action, as passed from the user context; may be
1329 * freed/invalidated after the execution of this function, with a different
1330 * action "returned" its place (pass-by-reference)
1331 * @return True if curr is a newly-explored action; false otherwise
1333 bool ModelChecker::initialize_curr_action(ModelAction **curr)
1335 ModelAction *newcurr;
1337 if ((*curr)->is_rmwc() || (*curr)->is_rmw()) {
1338 newcurr = process_rmw(*curr);
1341 if (newcurr->is_rmw())
1342 compute_promises(newcurr);
1348 (*curr)->set_seq_number(get_next_seq_num());
1350 newcurr = node_stack->explore_action(*curr, scheduler->get_enabled_array());
1352 /* First restore type and order in case of RMW operation */
1353 if ((*curr)->is_rmwr())
1354 newcurr->copy_typeandorder(*curr);
1356 ASSERT((*curr)->get_location() == newcurr->get_location());
1357 newcurr->copy_from_new(*curr);
1359 /* Discard duplicate ModelAction; use action from NodeStack */
1362 /* Always compute new clock vector */
1363 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1366 return false; /* Action was explored previously */
1370 /* Always compute new clock vector */
1371 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1373 /* Assign most recent release fence */
1374 newcurr->set_last_fence_release(get_last_fence_release(newcurr->get_tid()));
1377 * Perform one-time actions when pushing new ModelAction onto
1380 if (newcurr->is_write())
1381 compute_promises(newcurr);
1382 else if (newcurr->is_relseq_fixup())
1383 compute_relseq_breakwrites(newcurr);
1384 else if (newcurr->is_wait())
1385 newcurr->get_node()->set_misc_max(2);
1386 else if (newcurr->is_notify_one()) {
1387 newcurr->get_node()->set_misc_max(get_safe_ptr_action(condvar_waiters_map, newcurr->get_location())->size());
1389 return true; /* This was a new ModelAction */
1394 * @brief Establish reads-from relation between two actions
1396 * Perform basic operations involved with establishing a concrete rf relation,
1397 * including setting the ModelAction data and checking for release sequences.
1399 * @param act The action that is reading (must be a read)
1400 * @param rf The action from which we are reading (must be a write)
1402 * @return True if this read established synchronization
1404 bool ModelChecker::read_from(ModelAction *act, const ModelAction *rf)
1407 ASSERT(rf->is_write());
1409 act->set_read_from(rf);
1410 if (act->is_acquire()) {
1411 rel_heads_list_t release_heads;
1412 get_release_seq_heads(act, act, &release_heads);
1413 int num_heads = release_heads.size();
1414 for (unsigned int i = 0; i < release_heads.size(); i++)
1415 if (!act->synchronize_with(release_heads[i])) {
1416 set_bad_synchronization();
1419 return num_heads > 0;
1425 * Check promises and eliminate potentially-satisfying threads when a thread is
1426 * blocked (e.g., join, lock). A thread which is waiting on another thread can
1427 * no longer satisfy a promise generated from that thread.
1429 * @param blocker The thread on which a thread is waiting
1430 * @param waiting The waiting thread
1432 void ModelChecker::thread_blocking_check_promises(Thread *blocker, Thread *waiting)
1434 for (unsigned int i = 0; i < promises->size(); i++) {
1435 Promise *promise = (*promises)[i];
1436 if (!promise->thread_is_available(waiting->get_id()))
1438 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
1439 ModelAction *reader = promise->get_reader(j);
1440 if (reader->get_tid() != blocker->get_id())
1442 if (promise->eliminate_thread(waiting->get_id())) {
1443 /* Promise has failed */
1444 priv->failed_promise = true;
1446 /* Only eliminate the 'waiting' thread once */
1454 * @brief Check whether a model action is enabled.
1456 * Checks whether a lock or join operation would be successful (i.e., is the
1457 * lock already locked, or is the joined thread already complete). If not, put
1458 * the action in a waiter list.
1460 * @param curr is the ModelAction to check whether it is enabled.
1461 * @return a bool that indicates whether the action is enabled.
1463 bool ModelChecker::check_action_enabled(ModelAction *curr) {
1464 if (curr->is_lock()) {
1465 std::mutex *lock = curr->get_mutex();
1466 struct std::mutex_state *state = lock->get_state();
1469 } else if (curr->is_thread_join()) {
1470 Thread *blocking = curr->get_thread_operand();
1471 if (!blocking->is_complete()) {
1472 thread_blocking_check_promises(blocking, get_thread(curr));
1481 * This is the heart of the model checker routine. It performs model-checking
1482 * actions corresponding to a given "current action." Among other processes, it
1483 * calculates reads-from relationships, updates synchronization clock vectors,
1484 * forms a memory_order constraints graph, and handles replay/backtrack
1485 * execution when running permutations of previously-observed executions.
1487 * @param curr The current action to process
1488 * @return The ModelAction that is actually executed; may be different than
1489 * curr; may be NULL, if the current action is not enabled to run
1491 ModelAction * ModelChecker::check_current_action(ModelAction *curr)
1494 bool second_part_of_rmw = curr->is_rmwc() || curr->is_rmw();
1496 if (!check_action_enabled(curr)) {
1497 /* Make the execution look like we chose to run this action
1498 * much later, when a lock/join can succeed */
1499 get_thread(curr)->set_pending(curr);
1500 scheduler->sleep(get_thread(curr));
1504 bool newly_explored = initialize_curr_action(&curr);
1510 wake_up_sleeping_actions(curr);
1512 /* Compute fairness information for CHESS yield algorithm */
1513 if (model->params.yieldon) {
1514 curr->get_node()->update_yield(scheduler);
1517 /* Add the action to lists before any other model-checking tasks */
1518 if (!second_part_of_rmw)
1519 add_action_to_lists(curr);
1521 /* Build may_read_from set for newly-created actions */
1522 if (newly_explored && curr->is_read())
1523 build_may_read_from(curr);
1525 /* Initialize work_queue with the "current action" work */
1526 work_queue_t work_queue(1, CheckCurrWorkEntry(curr));
1527 while (!work_queue.empty() && !has_asserted()) {
1528 WorkQueueEntry work = work_queue.front();
1529 work_queue.pop_front();
1531 switch (work.type) {
1532 case WORK_CHECK_CURR_ACTION: {
1533 ModelAction *act = work.action;
1534 bool update = false; /* update this location's release seq's */
1535 bool update_all = false; /* update all release seq's */
1537 if (process_thread_action(curr))
1540 if (act->is_read() && !second_part_of_rmw && process_read(act))
1543 if (act->is_write() && process_write(act))
1546 if (act->is_fence() && process_fence(act))
1549 if (act->is_mutex_op() && process_mutex(act))
1552 if (act->is_relseq_fixup())
1553 process_relseq_fixup(curr, &work_queue);
1556 work_queue.push_back(CheckRelSeqWorkEntry(NULL));
1558 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1561 case WORK_CHECK_RELEASE_SEQ:
1562 resolve_release_sequences(work.location, &work_queue);
1564 case WORK_CHECK_MO_EDGES: {
1565 /** @todo Complete verification of work_queue */
1566 ModelAction *act = work.action;
1567 bool updated = false;
1569 if (act->is_read()) {
1570 const ModelAction *rf = act->get_reads_from();
1571 const Promise *promise = act->get_reads_from_promise();
1573 if (r_modification_order(act, rf))
1575 } else if (promise) {
1576 if (r_modification_order(act, promise))
1580 if (act->is_write()) {
1581 if (w_modification_order(act, NULL))
1584 mo_graph->commitChanges();
1587 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1596 check_curr_backtracking(curr);
1597 set_backtracking(curr);
1601 void ModelChecker::check_curr_backtracking(ModelAction *curr)
1603 Node *currnode = curr->get_node();
1604 Node *parnode = currnode->get_parent();
1606 if ((parnode && !parnode->backtrack_empty()) ||
1607 !currnode->misc_empty() ||
1608 !currnode->read_from_empty() ||
1609 !currnode->promise_empty() ||
1610 !currnode->relseq_break_empty()) {
1611 set_latest_backtrack(curr);
1615 bool ModelChecker::promises_expired() const
1617 for (unsigned int i = 0; i < promises->size(); i++) {
1618 Promise *promise = (*promises)[i];
1619 if (promise->get_expiration() < priv->used_sequence_numbers)
1626 * This is the strongest feasibility check available.
1627 * @return whether the current trace (partial or complete) must be a prefix of
1630 bool ModelChecker::isfeasibleprefix() const
1632 return pending_rel_seqs->size() == 0 && is_feasible_prefix_ignore_relseq();
1636 * Print disagnostic information about an infeasible execution
1637 * @param prefix A string to prefix the output with; if NULL, then a default
1638 * message prefix will be provided
1640 void ModelChecker::print_infeasibility(const char *prefix) const
1644 if (mo_graph->checkForCycles())
1645 ptr += sprintf(ptr, "[mo cycle]");
1646 if (priv->failed_promise)
1647 ptr += sprintf(ptr, "[failed promise]");
1648 if (priv->too_many_reads)
1649 ptr += sprintf(ptr, "[too many reads]");
1650 if (priv->no_valid_reads)
1651 ptr += sprintf(ptr, "[no valid reads-from]");
1652 if (priv->bad_synchronization)
1653 ptr += sprintf(ptr, "[bad sw ordering]");
1654 if (promises_expired())
1655 ptr += sprintf(ptr, "[promise expired]");
1656 if (promises->size() != 0)
1657 ptr += sprintf(ptr, "[unresolved promise]");
1659 model_print("%s: %s\n", prefix ? prefix : "Infeasible", buf);
1663 * Returns whether the current completed trace is feasible, except for pending
1664 * release sequences.
1666 bool ModelChecker::is_feasible_prefix_ignore_relseq() const
1668 return !is_infeasible() && promises->size() == 0;
1672 * Check if the current partial trace is infeasible. Does not check any
1673 * end-of-execution flags, which might rule out the execution. Thus, this is
1674 * useful only for ruling an execution as infeasible.
1675 * @return whether the current partial trace is infeasible.
1677 bool ModelChecker::is_infeasible() const
1679 return mo_graph->checkForCycles() ||
1680 priv->no_valid_reads ||
1681 priv->failed_promise ||
1682 priv->too_many_reads ||
1683 priv->bad_synchronization ||
1687 /** Close out a RMWR by converting previous RMWR into a RMW or READ. */
1688 ModelAction * ModelChecker::process_rmw(ModelAction *act) {
1689 ModelAction *lastread = get_last_action(act->get_tid());
1690 lastread->process_rmw(act);
1691 if (act->is_rmw()) {
1692 if (lastread->get_reads_from())
1693 mo_graph->addRMWEdge(lastread->get_reads_from(), lastread);
1695 mo_graph->addRMWEdge(lastread->get_reads_from_promise(), lastread);
1696 mo_graph->commitChanges();
1702 * A helper function for ModelChecker::check_recency, to check if the current
1703 * thread is able to read from a different write/promise for 'params.maxreads'
1704 * number of steps and if that write/promise should become visible (i.e., is
1705 * ordered later in the modification order). This helps model memory liveness.
1707 * @param curr The current action. Must be a read.
1708 * @param rf The write/promise from which we plan to read
1709 * @param other_rf The write/promise from which we may read
1710 * @return True if we were able to read from other_rf for params.maxreads steps
1712 template <typename T, typename U>
1713 bool ModelChecker::should_read_instead(const ModelAction *curr, const T *rf, const U *other_rf) const
1715 /* Need a different write/promise */
1716 if (other_rf->equals(rf))
1719 /* Only look for "newer" writes/promises */
1720 if (!mo_graph->checkReachable(rf, other_rf))
1723 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1724 action_list_t *list = &(*thrd_lists)[id_to_int(curr->get_tid())];
1725 action_list_t::reverse_iterator rit = list->rbegin();
1726 ASSERT((*rit) == curr);
1727 /* Skip past curr */
1730 /* Does this write/promise work for everyone? */
1731 for (int i = 0; i < params.maxreads; i++, rit++) {
1732 ModelAction *act = *rit;
1733 if (!act->may_read_from(other_rf))
1740 * Checks whether a thread has read from the same write or Promise for too many
1741 * times without seeing the effects of a later write/Promise.
1744 * 1) there must a different write/promise that we could read from,
1745 * 2) we must have read from the same write/promise in excess of maxreads times,
1746 * 3) that other write/promise must have been in the reads_from set for maxreads times, and
1747 * 4) that other write/promise must be mod-ordered after the write/promise we are reading.
1749 * If so, we decide that the execution is no longer feasible.
1751 * @param curr The current action. Must be a read.
1752 * @param rf The ModelAction/Promise from which we might read.
1753 * @return True if the read should succeed; false otherwise
1755 template <typename T>
1756 bool ModelChecker::check_recency(ModelAction *curr, const T *rf) const
1758 if (!params.maxreads)
1761 //NOTE: Next check is just optimization, not really necessary....
1762 if (curr->get_node()->get_read_from_past_size() +
1763 curr->get_node()->get_read_from_promise_size() <= 1)
1766 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1767 int tid = id_to_int(curr->get_tid());
1768 ASSERT(tid < (int)thrd_lists->size());
1769 action_list_t *list = &(*thrd_lists)[tid];
1770 action_list_t::reverse_iterator rit = list->rbegin();
1771 ASSERT((*rit) == curr);
1772 /* Skip past curr */
1775 action_list_t::reverse_iterator ritcopy = rit;
1776 /* See if we have enough reads from the same value */
1777 for (int count = 0; count < params.maxreads; ritcopy++, count++) {
1778 if (ritcopy == list->rend())
1780 ModelAction *act = *ritcopy;
1781 if (!act->is_read())
1783 if (act->get_reads_from_promise() && !act->get_reads_from_promise()->equals(rf))
1785 if (act->get_reads_from() && !act->get_reads_from()->equals(rf))
1787 if (act->get_node()->get_read_from_past_size() +
1788 act->get_node()->get_read_from_promise_size() <= 1)
1791 for (int i = 0; i < curr->get_node()->get_read_from_past_size(); i++) {
1792 const ModelAction *write = curr->get_node()->get_read_from_past(i);
1793 if (should_read_instead(curr, rf, write))
1794 return false; /* liveness failure */
1796 for (int i = 0; i < curr->get_node()->get_read_from_promise_size(); i++) {
1797 const Promise *promise = curr->get_node()->get_read_from_promise(i);
1798 if (should_read_instead(curr, rf, promise))
1799 return false; /* liveness failure */
1805 * Updates the mo_graph with the constraints imposed from the current
1808 * Basic idea is the following: Go through each other thread and find
1809 * the last action that happened before our read. Two cases:
1811 * (1) The action is a write => that write must either occur before
1812 * the write we read from or be the write we read from.
1814 * (2) The action is a read => the write that that action read from
1815 * must occur before the write we read from or be the same write.
1817 * @param curr The current action. Must be a read.
1818 * @param rf The ModelAction or Promise that curr reads from. Must be a write.
1819 * @return True if modification order edges were added; false otherwise
1821 template <typename rf_type>
1822 bool ModelChecker::r_modification_order(ModelAction *curr, const rf_type *rf)
1824 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1827 ASSERT(curr->is_read());
1829 /* Last SC fence in the current thread */
1830 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1831 ModelAction *last_sc_write = NULL;
1832 if (curr->is_seqcst())
1833 last_sc_write = get_last_seq_cst_write(curr);
1835 /* Iterate over all threads */
1836 for (i = 0; i < thrd_lists->size(); i++) {
1837 /* Last SC fence in thread i */
1838 ModelAction *last_sc_fence_thread_local = NULL;
1839 if (int_to_id((int)i) != curr->get_tid())
1840 last_sc_fence_thread_local = get_last_seq_cst_fence(int_to_id(i), NULL);
1842 /* Last SC fence in thread i, before last SC fence in current thread */
1843 ModelAction *last_sc_fence_thread_before = NULL;
1844 if (last_sc_fence_local)
1845 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1847 /* Iterate over actions in thread, starting from most recent */
1848 action_list_t *list = &(*thrd_lists)[i];
1849 action_list_t::reverse_iterator rit;
1850 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1851 ModelAction *act = *rit;
1856 /* Don't want to add reflexive edges on 'rf' */
1857 if (act->equals(rf)) {
1858 if (act->happens_before(curr))
1864 if (act->is_write()) {
1865 /* C++, Section 29.3 statement 5 */
1866 if (curr->is_seqcst() && last_sc_fence_thread_local &&
1867 *act < *last_sc_fence_thread_local) {
1868 added = mo_graph->addEdge(act, rf) || added;
1871 /* C++, Section 29.3 statement 4 */
1872 else if (act->is_seqcst() && last_sc_fence_local &&
1873 *act < *last_sc_fence_local) {
1874 added = mo_graph->addEdge(act, rf) || added;
1877 /* C++, Section 29.3 statement 6 */
1878 else if (last_sc_fence_thread_before &&
1879 *act < *last_sc_fence_thread_before) {
1880 added = mo_graph->addEdge(act, rf) || added;
1885 /* C++, Section 29.3 statement 3 (second subpoint) */
1886 if (curr->is_seqcst() && last_sc_write && act == last_sc_write) {
1887 added = mo_graph->addEdge(act, rf) || added;
1892 * Include at most one act per-thread that "happens
1895 if (act->happens_before(curr)) {
1896 if (act->is_write()) {
1897 added = mo_graph->addEdge(act, rf) || added;
1899 const ModelAction *prevrf = act->get_reads_from();
1900 const Promise *prevrf_promise = act->get_reads_from_promise();
1902 if (!prevrf->equals(rf))
1903 added = mo_graph->addEdge(prevrf, rf) || added;
1904 } else if (!prevrf_promise->equals(rf)) {
1905 added = mo_graph->addEdge(prevrf_promise, rf) || added;
1914 * All compatible, thread-exclusive promises must be ordered after any
1915 * concrete loads from the same thread
1917 for (unsigned int i = 0; i < promises->size(); i++)
1918 if ((*promises)[i]->is_compatible_exclusive(curr))
1919 added = mo_graph->addEdge(rf, (*promises)[i]) || added;
1925 * Updates the mo_graph with the constraints imposed from the current write.
1927 * Basic idea is the following: Go through each other thread and find
1928 * the lastest action that happened before our write. Two cases:
1930 * (1) The action is a write => that write must occur before
1933 * (2) The action is a read => the write that that action read from
1934 * must occur before the current write.
1936 * This method also handles two other issues:
1938 * (I) Sequential Consistency: Making sure that if the current write is
1939 * seq_cst, that it occurs after the previous seq_cst write.
1941 * (II) Sending the write back to non-synchronizing reads.
1943 * @param curr The current action. Must be a write.
1944 * @param send_fv A vector for stashing reads to which we may pass our future
1945 * value. If NULL, then don't record any future values.
1946 * @return True if modification order edges were added; false otherwise
1948 bool ModelChecker::w_modification_order(ModelAction *curr, ModelVector<ModelAction *> *send_fv)
1950 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1953 ASSERT(curr->is_write());
1955 if (curr->is_seqcst()) {
1956 /* We have to at least see the last sequentially consistent write,
1957 so we are initialized. */
1958 ModelAction *last_seq_cst = get_last_seq_cst_write(curr);
1959 if (last_seq_cst != NULL) {
1960 added = mo_graph->addEdge(last_seq_cst, curr) || added;
1964 /* Last SC fence in the current thread */
1965 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1967 /* Iterate over all threads */
1968 for (i = 0; i < thrd_lists->size(); i++) {
1969 /* Last SC fence in thread i, before last SC fence in current thread */
1970 ModelAction *last_sc_fence_thread_before = NULL;
1971 if (last_sc_fence_local && int_to_id((int)i) != curr->get_tid())
1972 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1974 /* Iterate over actions in thread, starting from most recent */
1975 action_list_t *list = &(*thrd_lists)[i];
1976 action_list_t::reverse_iterator rit;
1977 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1978 ModelAction *act = *rit;
1981 * 1) If RMW and it actually read from something, then we
1982 * already have all relevant edges, so just skip to next
1985 * 2) If RMW and it didn't read from anything, we should
1986 * whatever edge we can get to speed up convergence.
1988 * 3) If normal write, we need to look at earlier actions, so
1989 * continue processing list.
1991 if (curr->is_rmw()) {
1992 if (curr->get_reads_from() != NULL)
2000 /* C++, Section 29.3 statement 7 */
2001 if (last_sc_fence_thread_before && act->is_write() &&
2002 *act < *last_sc_fence_thread_before) {
2003 added = mo_graph->addEdge(act, curr) || added;
2008 * Include at most one act per-thread that "happens
2011 if (act->happens_before(curr)) {
2013 * Note: if act is RMW, just add edge:
2015 * The following edge should be handled elsewhere:
2016 * readfrom(act) --mo--> act
2018 if (act->is_write())
2019 added = mo_graph->addEdge(act, curr) || added;
2020 else if (act->is_read()) {
2021 //if previous read accessed a null, just keep going
2022 if (act->get_reads_from() == NULL)
2024 added = mo_graph->addEdge(act->get_reads_from(), curr) || added;
2027 } else if (act->is_read() && !act->could_synchronize_with(curr) &&
2028 !act->same_thread(curr)) {
2029 /* We have an action that:
2030 (1) did not happen before us
2031 (2) is a read and we are a write
2032 (3) cannot synchronize with us
2033 (4) is in a different thread
2035 that read could potentially read from our write. Note that
2036 these checks are overly conservative at this point, we'll
2037 do more checks before actually removing the
2041 if (send_fv && thin_air_constraint_may_allow(curr, act)) {
2042 if (!is_infeasible())
2043 send_fv->push_back(act);
2044 else if (curr->is_rmw() && act->is_rmw() && curr->get_reads_from() && curr->get_reads_from() == act->get_reads_from())
2045 add_future_value(curr, act);
2052 * All compatible, thread-exclusive promises must be ordered after any
2053 * concrete stores to the same thread, or else they can be merged with
2056 for (unsigned int i = 0; i < promises->size(); i++)
2057 if ((*promises)[i]->is_compatible_exclusive(curr))
2058 added = mo_graph->addEdge(curr, (*promises)[i]) || added;
2063 /** Arbitrary reads from the future are not allowed. Section 29.3
2064 * part 9 places some constraints. This method checks one result of constraint
2065 * constraint. Others require compiler support. */
2066 bool ModelChecker::thin_air_constraint_may_allow(const ModelAction *writer, const ModelAction *reader) const
2068 if (!writer->is_rmw())
2071 if (!reader->is_rmw())
2074 for (const ModelAction *search = writer->get_reads_from(); search != NULL; search = search->get_reads_from()) {
2075 if (search == reader)
2077 if (search->get_tid() == reader->get_tid() &&
2078 search->happens_before(reader))
2086 * Arbitrary reads from the future are not allowed. Section 29.3 part 9 places
2087 * some constraints. This method checks one the following constraint (others
2088 * require compiler support):
2090 * If X --hb-> Y --mo-> Z, then X should not read from Z.
2092 bool ModelChecker::mo_may_allow(const ModelAction *writer, const ModelAction *reader)
2094 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, reader->get_location());
2096 /* Iterate over all threads */
2097 for (i = 0; i < thrd_lists->size(); i++) {
2098 const ModelAction *write_after_read = NULL;
2100 /* Iterate over actions in thread, starting from most recent */
2101 action_list_t *list = &(*thrd_lists)[i];
2102 action_list_t::reverse_iterator rit;
2103 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2104 ModelAction *act = *rit;
2106 /* Don't disallow due to act == reader */
2107 if (!reader->happens_before(act) || reader == act)
2109 else if (act->is_write())
2110 write_after_read = act;
2111 else if (act->is_read() && act->get_reads_from() != NULL)
2112 write_after_read = act->get_reads_from();
2115 if (write_after_read && write_after_read != writer && mo_graph->checkReachable(write_after_read, writer))
2122 * Finds the head(s) of the release sequence(s) containing a given ModelAction.
2123 * The ModelAction under consideration is expected to be taking part in
2124 * release/acquire synchronization as an object of the "reads from" relation.
2125 * Note that this can only provide release sequence support for RMW chains
2126 * which do not read from the future, as those actions cannot be traced until
2127 * their "promise" is fulfilled. Similarly, we may not even establish the
2128 * presence of a release sequence with certainty, as some modification order
2129 * constraints may be decided further in the future. Thus, this function
2130 * "returns" two pieces of data: a pass-by-reference vector of @a release_heads
2131 * and a boolean representing certainty.
2133 * @param rf The action that might be part of a release sequence. Must be a
2135 * @param release_heads A pass-by-reference style return parameter. After
2136 * execution of this function, release_heads will contain the heads of all the
2137 * relevant release sequences, if any exists with certainty
2138 * @param pending A pass-by-reference style return parameter which is only used
2139 * when returning false (i.e., uncertain). Returns most information regarding
2140 * an uncertain release sequence, including any write operations that might
2141 * break the sequence.
2142 * @return true, if the ModelChecker is certain that release_heads is complete;
2145 bool ModelChecker::release_seq_heads(const ModelAction *rf,
2146 rel_heads_list_t *release_heads,
2147 struct release_seq *pending) const
2149 /* Only check for release sequences if there are no cycles */
2150 if (mo_graph->checkForCycles())
2153 for ( ; rf != NULL; rf = rf->get_reads_from()) {
2154 ASSERT(rf->is_write());
2156 if (rf->is_release())
2157 release_heads->push_back(rf);
2158 else if (rf->get_last_fence_release())
2159 release_heads->push_back(rf->get_last_fence_release());
2161 break; /* End of RMW chain */
2163 /** @todo Need to be smarter here... In the linux lock
2164 * example, this will run to the beginning of the program for
2166 /** @todo The way to be smarter here is to keep going until 1
2167 * thread has a release preceded by an acquire and you've seen
2170 /* acq_rel RMW is a sufficient stopping condition */
2171 if (rf->is_acquire() && rf->is_release())
2172 return true; /* complete */
2175 /* read from future: need to settle this later */
2177 return false; /* incomplete */
2180 if (rf->is_release())
2181 return true; /* complete */
2183 /* else relaxed write
2184 * - check for fence-release in the same thread (29.8, stmt. 3)
2185 * - check modification order for contiguous subsequence
2186 * -> rf must be same thread as release */
2188 const ModelAction *fence_release = rf->get_last_fence_release();
2189 /* Synchronize with a fence-release unconditionally; we don't need to
2190 * find any more "contiguous subsequence..." for it */
2192 release_heads->push_back(fence_release);
2194 int tid = id_to_int(rf->get_tid());
2195 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, rf->get_location());
2196 action_list_t *list = &(*thrd_lists)[tid];
2197 action_list_t::const_reverse_iterator rit;
2199 /* Find rf in the thread list */
2200 rit = std::find(list->rbegin(), list->rend(), rf);
2201 ASSERT(rit != list->rend());
2203 /* Find the last {write,fence}-release */
2204 for (; rit != list->rend(); rit++) {
2205 if (fence_release && *(*rit) < *fence_release)
2207 if ((*rit)->is_release())
2210 if (rit == list->rend()) {
2211 /* No write-release in this thread */
2212 return true; /* complete */
2213 } else if (fence_release && *(*rit) < *fence_release) {
2214 /* The fence-release is more recent (and so, "stronger") than
2215 * the most recent write-release */
2216 return true; /* complete */
2217 } /* else, need to establish contiguous release sequence */
2218 ModelAction *release = *rit;
2220 ASSERT(rf->same_thread(release));
2222 pending->writes.clear();
2224 bool certain = true;
2225 for (unsigned int i = 0; i < thrd_lists->size(); i++) {
2226 if (id_to_int(rf->get_tid()) == (int)i)
2228 list = &(*thrd_lists)[i];
2230 /* Can we ensure no future writes from this thread may break
2231 * the release seq? */
2232 bool future_ordered = false;
2234 ModelAction *last = get_last_action(int_to_id(i));
2235 Thread *th = get_thread(int_to_id(i));
2236 if ((last && rf->happens_before(last)) ||
2239 future_ordered = true;
2241 ASSERT(!th->is_model_thread() || future_ordered);
2243 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2244 const ModelAction *act = *rit;
2245 /* Reach synchronization -> this thread is complete */
2246 if (act->happens_before(release))
2248 if (rf->happens_before(act)) {
2249 future_ordered = true;
2253 /* Only non-RMW writes can break release sequences */
2254 if (!act->is_write() || act->is_rmw())
2257 /* Check modification order */
2258 if (mo_graph->checkReachable(rf, act)) {
2259 /* rf --mo--> act */
2260 future_ordered = true;
2263 if (mo_graph->checkReachable(act, release))
2264 /* act --mo--> release */
2266 if (mo_graph->checkReachable(release, act) &&
2267 mo_graph->checkReachable(act, rf)) {
2268 /* release --mo-> act --mo--> rf */
2269 return true; /* complete */
2271 /* act may break release sequence */
2272 pending->writes.push_back(act);
2275 if (!future_ordered)
2276 certain = false; /* This thread is uncertain */
2280 release_heads->push_back(release);
2281 pending->writes.clear();
2283 pending->release = release;
2290 * An interface for getting the release sequence head(s) with which a
2291 * given ModelAction must synchronize. This function only returns a non-empty
2292 * result when it can locate a release sequence head with certainty. Otherwise,
2293 * it may mark the internal state of the ModelChecker so that it will handle
2294 * the release sequence at a later time, causing @a acquire to update its
2295 * synchronization at some later point in execution.
2297 * @param acquire The 'acquire' action that may synchronize with a release
2299 * @param read The read action that may read from a release sequence; this may
2300 * be the same as acquire, or else an earlier action in the same thread (i.e.,
2301 * when 'acquire' is a fence-acquire)
2302 * @param release_heads A pass-by-reference return parameter. Will be filled
2303 * with the head(s) of the release sequence(s), if they exists with certainty.
2304 * @see ModelChecker::release_seq_heads
2306 void ModelChecker::get_release_seq_heads(ModelAction *acquire,
2307 ModelAction *read, rel_heads_list_t *release_heads)
2309 const ModelAction *rf = read->get_reads_from();
2310 struct release_seq *sequence = (struct release_seq *)snapshot_calloc(1, sizeof(struct release_seq));
2311 sequence->acquire = acquire;
2312 sequence->read = read;
2314 if (!release_seq_heads(rf, release_heads, sequence)) {
2315 /* add act to 'lazy checking' list */
2316 pending_rel_seqs->push_back(sequence);
2318 snapshot_free(sequence);
2323 * Attempt to resolve all stashed operations that might synchronize with a
2324 * release sequence for a given location. This implements the "lazy" portion of
2325 * determining whether or not a release sequence was contiguous, since not all
2326 * modification order information is present at the time an action occurs.
2328 * @param location The location/object that should be checked for release
2329 * sequence resolutions. A NULL value means to check all locations.
2330 * @param work_queue The work queue to which to add work items as they are
2332 * @return True if any updates occurred (new synchronization, new mo_graph
2335 bool ModelChecker::resolve_release_sequences(void *location, work_queue_t *work_queue)
2337 bool updated = false;
2338 SnapVector<struct release_seq *>::iterator it = pending_rel_seqs->begin();
2339 while (it != pending_rel_seqs->end()) {
2340 struct release_seq *pending = *it;
2341 ModelAction *acquire = pending->acquire;
2342 const ModelAction *read = pending->read;
2344 /* Only resolve sequences on the given location, if provided */
2345 if (location && read->get_location() != location) {
2350 const ModelAction *rf = read->get_reads_from();
2351 rel_heads_list_t release_heads;
2353 complete = release_seq_heads(rf, &release_heads, pending);
2354 for (unsigned int i = 0; i < release_heads.size(); i++) {
2355 if (!acquire->has_synchronized_with(release_heads[i])) {
2356 if (acquire->synchronize_with(release_heads[i]))
2359 set_bad_synchronization();
2364 /* Re-check all pending release sequences */
2365 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
2366 /* Re-check read-acquire for mo_graph edges */
2367 if (acquire->is_read())
2368 work_queue->push_back(MOEdgeWorkEntry(acquire));
2370 /* propagate synchronization to later actions */
2371 action_list_t::reverse_iterator rit = action_trace->rbegin();
2372 for (; (*rit) != acquire; rit++) {
2373 ModelAction *propagate = *rit;
2374 if (acquire->happens_before(propagate)) {
2375 propagate->synchronize_with(acquire);
2376 /* Re-check 'propagate' for mo_graph edges */
2377 work_queue->push_back(MOEdgeWorkEntry(propagate));
2382 it = pending_rel_seqs->erase(it);
2383 snapshot_free(pending);
2389 // If we resolved promises or data races, see if we have realized a data race.
2396 * Performs various bookkeeping operations for the current ModelAction. For
2397 * instance, adds action to the per-object, per-thread action vector and to the
2398 * action trace list of all thread actions.
2400 * @param act is the ModelAction to add.
2402 void ModelChecker::add_action_to_lists(ModelAction *act)
2404 int tid = id_to_int(act->get_tid());
2405 ModelAction *uninit = NULL;
2407 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
2408 if (list->empty() && act->is_atomic_var()) {
2409 uninit = get_uninitialized_action(act);
2410 uninit_id = id_to_int(uninit->get_tid());
2411 list->push_front(uninit);
2413 list->push_back(act);
2415 action_trace->push_back(act);
2417 action_trace->push_front(uninit);
2419 SnapVector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, act->get_location());
2420 if (tid >= (int)vec->size())
2421 vec->resize(priv->next_thread_id);
2422 (*vec)[tid].push_back(act);
2424 (*vec)[uninit_id].push_front(uninit);
2426 if ((int)thrd_last_action->size() <= tid)
2427 thrd_last_action->resize(get_num_threads());
2428 (*thrd_last_action)[tid] = act;
2430 (*thrd_last_action)[uninit_id] = uninit;
2432 if (act->is_fence() && act->is_release()) {
2433 if ((int)thrd_last_fence_release->size() <= tid)
2434 thrd_last_fence_release->resize(get_num_threads());
2435 (*thrd_last_fence_release)[tid] = act;
2438 if (act->is_wait()) {
2439 void *mutex_loc = (void *) act->get_value();
2440 get_safe_ptr_action(obj_map, mutex_loc)->push_back(act);
2442 SnapVector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, mutex_loc);
2443 if (tid >= (int)vec->size())
2444 vec->resize(priv->next_thread_id);
2445 (*vec)[tid].push_back(act);
2450 * @brief Get the last action performed by a particular Thread
2451 * @param tid The thread ID of the Thread in question
2452 * @return The last action in the thread
2454 ModelAction * ModelChecker::get_last_action(thread_id_t tid) const
2456 int threadid = id_to_int(tid);
2457 if (threadid < (int)thrd_last_action->size())
2458 return (*thrd_last_action)[id_to_int(tid)];
2464 * @brief Get the last fence release performed by a particular Thread
2465 * @param tid The thread ID of the Thread in question
2466 * @return The last fence release in the thread, if one exists; NULL otherwise
2468 ModelAction * ModelChecker::get_last_fence_release(thread_id_t tid) const
2470 int threadid = id_to_int(tid);
2471 if (threadid < (int)thrd_last_fence_release->size())
2472 return (*thrd_last_fence_release)[id_to_int(tid)];
2478 * Gets the last memory_order_seq_cst write (in the total global sequence)
2479 * performed on a particular object (i.e., memory location), not including the
2481 * @param curr The current ModelAction; also denotes the object location to
2483 * @return The last seq_cst write
2485 ModelAction * ModelChecker::get_last_seq_cst_write(ModelAction *curr) const
2487 void *location = curr->get_location();
2488 action_list_t *list = get_safe_ptr_action(obj_map, location);
2489 /* Find: max({i in dom(S) | seq_cst(t_i) && isWrite(t_i) && samevar(t_i, t)}) */
2490 action_list_t::reverse_iterator rit;
2491 for (rit = list->rbegin(); (*rit) != curr; rit++)
2493 rit++; /* Skip past curr */
2494 for ( ; rit != list->rend(); rit++)
2495 if ((*rit)->is_write() && (*rit)->is_seqcst())
2501 * Gets the last memory_order_seq_cst fence (in the total global sequence)
2502 * performed in a particular thread, prior to a particular fence.
2503 * @param tid The ID of the thread to check
2504 * @param before_fence The fence from which to begin the search; if NULL, then
2505 * search for the most recent fence in the thread.
2506 * @return The last prior seq_cst fence in the thread, if exists; otherwise, NULL
2508 ModelAction * ModelChecker::get_last_seq_cst_fence(thread_id_t tid, const ModelAction *before_fence) const
2510 /* All fences should have NULL location */
2511 action_list_t *list = get_safe_ptr_action(obj_map, NULL);
2512 action_list_t::reverse_iterator rit = list->rbegin();
2515 for (; rit != list->rend(); rit++)
2516 if (*rit == before_fence)
2519 ASSERT(*rit == before_fence);
2523 for (; rit != list->rend(); rit++)
2524 if ((*rit)->is_fence() && (tid == (*rit)->get_tid()) && (*rit)->is_seqcst())
2530 * Gets the last unlock operation performed on a particular mutex (i.e., memory
2531 * location). This function identifies the mutex according to the current
2532 * action, which is presumed to perform on the same mutex.
2533 * @param curr The current ModelAction; also denotes the object location to
2535 * @return The last unlock operation
2537 ModelAction * ModelChecker::get_last_unlock(ModelAction *curr) const
2539 void *location = curr->get_location();
2540 action_list_t *list = get_safe_ptr_action(obj_map, location);
2541 /* Find: max({i in dom(S) | isUnlock(t_i) && samevar(t_i, t)}) */
2542 action_list_t::reverse_iterator rit;
2543 for (rit = list->rbegin(); rit != list->rend(); rit++)
2544 if ((*rit)->is_unlock() || (*rit)->is_wait())
2549 ModelAction * ModelChecker::get_parent_action(thread_id_t tid) const
2551 ModelAction *parent = get_last_action(tid);
2553 parent = get_thread(tid)->get_creation();
2558 * Returns the clock vector for a given thread.
2559 * @param tid The thread whose clock vector we want
2560 * @return Desired clock vector
2562 ClockVector * ModelChecker::get_cv(thread_id_t tid) const
2564 return get_parent_action(tid)->get_cv();
2568 * @brief Find the promise (if any) to resolve for the current action and
2569 * remove it from the pending promise vector
2570 * @param curr The current ModelAction. Should be a write.
2571 * @return The Promise to resolve, if any; otherwise NULL
2573 Promise * ModelChecker::pop_promise_to_resolve(const ModelAction *curr)
2575 for (unsigned int i = 0; i < promises->size(); i++)
2576 if (curr->get_node()->get_promise(i)) {
2577 Promise *ret = (*promises)[i];
2578 promises->erase(promises->begin() + i);
2585 * Resolve a Promise with a current write.
2586 * @param write The ModelAction that is fulfilling Promises
2587 * @param promise The Promise to resolve
2588 * @return True if the Promise was successfully resolved; false otherwise
2590 bool ModelChecker::resolve_promise(ModelAction *write, Promise *promise)
2592 ModelVector<ModelAction *> actions_to_check;
2594 for (unsigned int i = 0; i < promise->get_num_readers(); i++) {
2595 ModelAction *read = promise->get_reader(i);
2596 read_from(read, write);
2597 actions_to_check.push_back(read);
2599 /* Make sure the promise's value matches the write's value */
2600 ASSERT(promise->is_compatible(write) && promise->same_value(write));
2601 if (!mo_graph->resolvePromise(promise, write))
2602 priv->failed_promise = true;
2605 * @todo It is possible to end up in an inconsistent state, where a
2606 * "resolved" promise may still be referenced if
2607 * CycleGraph::resolvePromise() failed, so don't delete 'promise'.
2609 * Note that the inconsistency only matters when dumping mo_graph to
2615 //Check whether reading these writes has made threads unable to
2617 for (unsigned int i = 0; i < actions_to_check.size(); i++) {
2618 ModelAction *read = actions_to_check[i];
2619 mo_check_promises(read, true);
2626 * Compute the set of promises that could potentially be satisfied by this
2627 * action. Note that the set computation actually appears in the Node, not in
2629 * @param curr The ModelAction that may satisfy promises
2631 void ModelChecker::compute_promises(ModelAction *curr)
2633 for (unsigned int i = 0; i < promises->size(); i++) {
2634 Promise *promise = (*promises)[i];
2635 if (!promise->is_compatible(curr) || !promise->same_value(curr))
2638 bool satisfy = true;
2639 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
2640 const ModelAction *act = promise->get_reader(j);
2641 if (act->happens_before(curr) ||
2642 act->could_synchronize_with(curr)) {
2648 curr->get_node()->set_promise(i);
2652 /** Checks promises in response to change in ClockVector Threads. */
2653 void ModelChecker::check_promises(thread_id_t tid, ClockVector *old_cv, ClockVector *merge_cv)
2655 for (unsigned int i = 0; i < promises->size(); i++) {
2656 Promise *promise = (*promises)[i];
2657 if (!promise->thread_is_available(tid))
2659 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
2660 const ModelAction *act = promise->get_reader(j);
2661 if ((!old_cv || !old_cv->synchronized_since(act)) &&
2662 merge_cv->synchronized_since(act)) {
2663 if (promise->eliminate_thread(tid)) {
2664 /* Promise has failed */
2665 priv->failed_promise = true;
2673 void ModelChecker::check_promises_thread_disabled()
2675 for (unsigned int i = 0; i < promises->size(); i++) {
2676 Promise *promise = (*promises)[i];
2677 if (promise->has_failed()) {
2678 priv->failed_promise = true;
2685 * @brief Checks promises in response to addition to modification order for
2688 * We test whether threads are still available for satisfying promises after an
2689 * addition to our modification order constraints. Those that are unavailable
2690 * are "eliminated". Once all threads are eliminated from satisfying a promise,
2691 * that promise has failed.
2693 * @param act The ModelAction which updated the modification order
2694 * @param is_read_check Should be true if act is a read and we must check for
2695 * updates to the store from which it read (there is a distinction here for
2696 * RMW's, which are both a load and a store)
2698 void ModelChecker::mo_check_promises(const ModelAction *act, bool is_read_check)
2700 const ModelAction *write = is_read_check ? act->get_reads_from() : act;
2702 for (unsigned int i = 0; i < promises->size(); i++) {
2703 Promise *promise = (*promises)[i];
2705 // Is this promise on the same location?
2706 if (!promise->same_location(write))
2709 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
2710 const ModelAction *pread = promise->get_reader(j);
2711 if (!pread->happens_before(act))
2713 if (mo_graph->checkPromise(write, promise)) {
2714 priv->failed_promise = true;
2720 // Don't do any lookups twice for the same thread
2721 if (!promise->thread_is_available(act->get_tid()))
2724 if (mo_graph->checkReachable(promise, write)) {
2725 if (mo_graph->checkPromise(write, promise)) {
2726 priv->failed_promise = true;
2734 * Compute the set of writes that may break the current pending release
2735 * sequence. This information is extracted from previou release sequence
2738 * @param curr The current ModelAction. Must be a release sequence fixup
2741 void ModelChecker::compute_relseq_breakwrites(ModelAction *curr)
2743 if (pending_rel_seqs->empty())
2746 struct release_seq *pending = pending_rel_seqs->back();
2747 for (unsigned int i = 0; i < pending->writes.size(); i++) {
2748 const ModelAction *write = pending->writes[i];
2749 curr->get_node()->add_relseq_break(write);
2752 /* NULL means don't break the sequence; just synchronize */
2753 curr->get_node()->add_relseq_break(NULL);
2757 * Build up an initial set of all past writes that this 'read' action may read
2758 * from, as well as any previously-observed future values that must still be valid.
2760 * @param curr is the current ModelAction that we are exploring; it must be a
2763 void ModelChecker::build_may_read_from(ModelAction *curr)
2765 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
2767 ASSERT(curr->is_read());
2769 ModelAction *last_sc_write = NULL;
2771 if (curr->is_seqcst())
2772 last_sc_write = get_last_seq_cst_write(curr);
2774 /* Iterate over all threads */
2775 for (i = 0; i < thrd_lists->size(); i++) {
2776 /* Iterate over actions in thread, starting from most recent */
2777 action_list_t *list = &(*thrd_lists)[i];
2778 action_list_t::reverse_iterator rit;
2779 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2780 ModelAction *act = *rit;
2782 /* Only consider 'write' actions */
2783 if (!act->is_write() || act == curr)
2786 /* Don't consider more than one seq_cst write if we are a seq_cst read. */
2787 bool allow_read = true;
2789 if (curr->is_seqcst() && (act->is_seqcst() || (last_sc_write != NULL && act->happens_before(last_sc_write))) && act != last_sc_write)
2791 else if (curr->get_sleep_flag() && !curr->is_seqcst() && !sleep_can_read_from(curr, act))
2795 /* Only add feasible reads */
2796 mo_graph->startChanges();
2797 r_modification_order(curr, act);
2798 if (!is_infeasible())
2799 curr->get_node()->add_read_from_past(act);
2800 mo_graph->rollbackChanges();
2803 /* Include at most one act per-thread that "happens before" curr */
2804 if (act->happens_before(curr))
2809 /* Inherit existing, promised future values */
2810 for (i = 0; i < promises->size(); i++) {
2811 const Promise *promise = (*promises)[i];
2812 const ModelAction *promise_read = promise->get_reader(0);
2813 if (promise_read->same_var(curr)) {
2814 /* Only add feasible future-values */
2815 mo_graph->startChanges();
2816 r_modification_order(curr, promise);
2817 if (!is_infeasible())
2818 curr->get_node()->add_read_from_promise(promise_read);
2819 mo_graph->rollbackChanges();
2823 /* We may find no valid may-read-from only if the execution is doomed */
2824 if (!curr->get_node()->read_from_size()) {
2825 priv->no_valid_reads = true;
2829 if (DBG_ENABLED()) {
2830 model_print("Reached read action:\n");
2832 model_print("Printing read_from_past\n");
2833 curr->get_node()->print_read_from_past();
2834 model_print("End printing read_from_past\n");
2838 bool ModelChecker::sleep_can_read_from(ModelAction *curr, const ModelAction *write)
2840 for ( ; write != NULL; write = write->get_reads_from()) {
2841 /* UNINIT actions don't have a Node, and they never sleep */
2842 if (write->is_uninitialized())
2844 Node *prevnode = write->get_node()->get_parent();
2846 bool thread_sleep = prevnode->enabled_status(curr->get_tid()) == THREAD_SLEEP_SET;
2847 if (write->is_release() && thread_sleep)
2849 if (!write->is_rmw())
2856 * @brief Get an action representing an uninitialized atomic
2858 * This function may create a new one or try to retrieve one from the NodeStack
2860 * @param curr The current action, which prompts the creation of an UNINIT action
2861 * @return A pointer to the UNINIT ModelAction
2863 ModelAction * ModelChecker::get_uninitialized_action(const ModelAction *curr) const
2865 Node *node = curr->get_node();
2866 ModelAction *act = node->get_uninit_action();
2868 act = new ModelAction(ATOMIC_UNINIT, std::memory_order_relaxed, curr->get_location(), model->params.uninitvalue, model_thread);
2869 node->set_uninit_action(act);
2871 act->create_cv(NULL);
2875 static void print_list(action_list_t *list)
2877 action_list_t::iterator it;
2879 model_print("---------------------------------------------------------------------\n");
2881 unsigned int hash = 0;
2883 for (it = list->begin(); it != list->end(); it++) {
2884 const ModelAction *act = *it;
2885 if (act->get_seq_number() > 0)
2887 hash = hash^(hash<<3)^((*it)->hash());
2889 model_print("HASH %u\n", hash);
2890 model_print("---------------------------------------------------------------------\n");
2893 #if SUPPORT_MOD_ORDER_DUMP
2894 void ModelChecker::dumpGraph(char *filename) const
2897 sprintf(buffer, "%s.dot", filename);
2898 FILE *file = fopen(buffer, "w");
2899 fprintf(file, "digraph %s {\n", filename);
2900 mo_graph->dumpNodes(file);
2901 ModelAction **thread_array = (ModelAction **)model_calloc(1, sizeof(ModelAction *) * get_num_threads());
2903 for (action_list_t::iterator it = action_trace->begin(); it != action_trace->end(); it++) {
2904 ModelAction *act = *it;
2905 if (act->is_read()) {
2906 mo_graph->dot_print_node(file, act);
2907 if (act->get_reads_from())
2908 mo_graph->dot_print_edge(file,
2909 act->get_reads_from(),
2911 "label=\"rf\", color=red, weight=2");
2913 mo_graph->dot_print_edge(file,
2914 act->get_reads_from_promise(),
2916 "label=\"rf\", color=red");
2918 if (thread_array[act->get_tid()]) {
2919 mo_graph->dot_print_edge(file,
2920 thread_array[id_to_int(act->get_tid())],
2922 "label=\"sb\", color=blue, weight=400");
2925 thread_array[act->get_tid()] = act;
2927 fprintf(file, "}\n");
2928 model_free(thread_array);
2933 /** @brief Prints an execution trace summary. */
2934 void ModelChecker::print_summary() const
2936 #if SUPPORT_MOD_ORDER_DUMP
2937 char buffername[100];
2938 sprintf(buffername, "exec%04u", stats.num_total);
2939 mo_graph->dumpGraphToFile(buffername);
2940 sprintf(buffername, "graph%04u", stats.num_total);
2941 dumpGraph(buffername);
2944 model_print("Execution %d:", stats.num_total);
2945 if (isfeasibleprefix()) {
2946 if (scheduler->all_threads_sleeping())
2947 model_print(" SLEEP-SET REDUNDANT");
2950 print_infeasibility(" INFEASIBLE");
2951 print_list(action_trace);
2953 if (!promises->empty()) {
2954 model_print("Pending promises:\n");
2955 for (unsigned int i = 0; i < promises->size(); i++) {
2956 model_print(" [P%u] ", i);
2957 (*promises)[i]->print();
2964 * Add a Thread to the system for the first time. Should only be called once
2966 * @param t The Thread to add
2968 void ModelChecker::add_thread(Thread *t)
2970 thread_map->put(id_to_int(t->get_id()), t);
2971 scheduler->add_thread(t);
2975 * @brief Get a Thread reference by its ID
2976 * @param tid The Thread's ID
2977 * @return A Thread reference
2979 Thread * ModelChecker::get_thread(thread_id_t tid) const
2981 return thread_map->get(id_to_int(tid));
2985 * @brief Get a reference to the Thread in which a ModelAction was executed
2986 * @param act The ModelAction
2987 * @return A Thread reference
2989 Thread * ModelChecker::get_thread(const ModelAction *act) const
2991 return get_thread(act->get_tid());
2995 * @brief Get a Promise's "promise number"
2997 * A "promise number" is an index number that is unique to a promise, valid
2998 * only for a specific snapshot of an execution trace. Promises may come and go
2999 * as they are generated an resolved, so an index only retains meaning for the
3002 * @param promise The Promise to check
3003 * @return The promise index, if the promise still is valid; otherwise -1
3005 int ModelChecker::get_promise_number(const Promise *promise) const
3007 for (unsigned int i = 0; i < promises->size(); i++)
3008 if ((*promises)[i] == promise)
3015 * @brief Check if a Thread is currently enabled
3016 * @param t The Thread to check
3017 * @return True if the Thread is currently enabled
3019 bool ModelChecker::is_enabled(Thread *t) const
3021 return scheduler->is_enabled(t);
3025 * @brief Check if a Thread is currently enabled
3026 * @param tid The ID of the Thread to check
3027 * @return True if the Thread is currently enabled
3029 bool ModelChecker::is_enabled(thread_id_t tid) const
3031 return scheduler->is_enabled(tid);
3035 * Switch from a model-checker context to a user-thread context. This is the
3036 * complement of ModelChecker::switch_to_master and must be called from the
3037 * model-checker context
3039 * @param thread The user-thread to switch to
3041 void ModelChecker::switch_from_master(Thread *thread)
3043 scheduler->set_current_thread(thread);
3044 Thread::swap(&system_context, thread);
3048 * Switch from a user-context to the "master thread" context (a.k.a. system
3049 * context). This switch is made with the intention of exploring a particular
3050 * model-checking action (described by a ModelAction object). Must be called
3051 * from a user-thread context.
3053 * @param act The current action that will be explored. May be NULL only if
3054 * trace is exiting via an assertion (see ModelChecker::set_assert and
3055 * ModelChecker::has_asserted).
3056 * @return Return the value returned by the current action
3058 uint64_t ModelChecker::switch_to_master(ModelAction *act)
3061 Thread *old = thread_current();
3062 scheduler->set_current_thread(NULL);
3063 ASSERT(!old->get_pending());
3064 old->set_pending(act);
3065 if (Thread::swap(old, &system_context) < 0) {
3066 perror("swap threads");
3069 return old->get_return_value();
3073 * Takes the next step in the execution, if possible.
3074 * @param curr The current step to take
3075 * @return Returns the next Thread to run, if any; NULL if this execution
3078 Thread * ModelChecker::take_step(ModelAction *curr)
3080 Thread *curr_thrd = get_thread(curr);
3081 ASSERT(curr_thrd->get_state() == THREAD_READY);
3083 curr = check_current_action(curr);
3085 /* Infeasible -> don't take any more steps */
3086 if (is_infeasible())
3088 else if (isfeasibleprefix() && have_bug_reports()) {
3093 if (params.bound != 0 && priv->used_sequence_numbers > params.bound)
3096 if (curr_thrd->is_blocked() || curr_thrd->is_complete())
3097 scheduler->remove_thread(curr_thrd);
3099 Thread *next_thrd = NULL;
3101 next_thrd = action_select_next_thread(curr);
3103 next_thrd = get_next_thread();
3105 DEBUG("(%d, %d)\n", curr_thrd ? id_to_int(curr_thrd->get_id()) : -1,
3106 next_thrd ? id_to_int(next_thrd->get_id()) : -1);
3111 /** Wrapper to run the user's main function, with appropriate arguments */
3112 void user_main_wrapper(void *)
3114 user_main(model->params.argc, model->params.argv);
3117 /** @brief Run ModelChecker for the user program */
3118 void ModelChecker::run()
3122 Thread *t = new Thread(&user_thread, &user_main_wrapper, NULL, NULL);
3127 * Stash next pending action(s) for thread(s). There
3128 * should only need to stash one thread's action--the
3129 * thread which just took a step--plus the first step
3130 * for any newly-created thread
3132 for (unsigned int i = 0; i < get_num_threads(); i++) {
3133 thread_id_t tid = int_to_id(i);
3134 Thread *thr = get_thread(tid);
3135 if (!thr->is_model_thread() && !thr->is_complete() && !thr->get_pending()) {
3136 switch_from_master(thr);
3137 if (thr->is_waiting_on(thr))
3138 assert_bug("Deadlock detected");
3142 /* Catch assertions from prior take_step or from
3143 * between-ModelAction bugs (e.g., data races) */
3147 /* Consume the next action for a Thread */
3148 ModelAction *curr = t->get_pending();
3149 t->set_pending(NULL);
3150 t = take_step(curr);
3151 } while (t && !t->is_model_thread());
3154 * Launch end-of-execution release sequence fixups only when
3155 * the execution is otherwise feasible AND there are:
3157 * (1) pending release sequences
3158 * (2) pending assertions that could be invalidated by a change
3159 * in clock vectors (i.e., data races)
3160 * (3) no pending promises
3162 while (!pending_rel_seqs->empty() &&
3163 is_feasible_prefix_ignore_relseq() &&
3164 !unrealizedraces.empty()) {
3165 model_print("*** WARNING: release sequence fixup action "
3166 "(%zu pending release seuqence(s)) ***\n",
3167 pending_rel_seqs->size());
3168 ModelAction *fixup = new ModelAction(MODEL_FIXUP_RELSEQ,
3169 std::memory_order_seq_cst, NULL, VALUE_NONE,
3173 } while (next_execution());
3175 model_print("******* Model-checking complete: *******\n");