9 #include "snapshot-interface.h"
11 #include "clockvector.h"
12 #include "cyclegraph.h"
15 #include "threads-model.h"
18 #define INITIAL_THREAD_ID 0
23 bug_message(const char *str) {
24 const char *fmt = " [BUG] %s\n";
25 msg = (char *)snapshot_malloc(strlen(fmt) + strlen(str));
26 sprintf(msg, fmt, str);
28 ~bug_message() { if (msg) snapshot_free(msg); }
31 void print() { model_print("%s", msg); }
37 * Structure for holding small ModelChecker members that should be snapshotted
39 struct model_snapshot_members {
40 ModelAction *current_action;
41 unsigned int next_thread_id;
42 modelclock_t used_sequence_numbers;
44 ModelAction *next_backtrack;
45 std::vector< bug_message *, SnapshotAlloc<bug_message *> > bugs;
46 struct execution_stats stats;
49 /** @brief Constructor */
50 ModelChecker::ModelChecker(struct model_params params) :
51 /* Initialize default scheduler */
53 scheduler(new Scheduler()),
55 earliest_diverge(NULL),
56 action_trace(new action_list_t()),
57 thread_map(new HashTable<int, Thread *, int>()),
58 obj_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
59 lock_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
60 condvar_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
61 obj_thrd_map(new HashTable<void *, std::vector<action_list_t> *, uintptr_t, 4 >()),
62 promises(new std::vector< Promise *, SnapshotAlloc<Promise *> >()),
63 futurevalues(new std::vector< struct PendingFutureValue, SnapshotAlloc<struct PendingFutureValue> >()),
64 pending_rel_seqs(new std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >()),
65 thrd_last_action(new std::vector< ModelAction *, SnapshotAlloc<ModelAction *> >(1)),
66 node_stack(new NodeStack()),
67 mo_graph(new CycleGraph()),
68 failed_promise(false),
69 too_many_reads(false),
71 bad_synchronization(false)
73 /* Allocate this "size" on the snapshotting heap */
74 priv = (struct model_snapshot_members *)snapshot_calloc(1, sizeof(*priv));
75 /* First thread created will have id INITIAL_THREAD_ID */
76 priv->next_thread_id = INITIAL_THREAD_ID;
78 /* Initialize a model-checker thread, for special ModelActions */
79 model_thread = new Thread(get_next_id());
80 thread_map->put(id_to_int(model_thread->get_id()), model_thread);
83 /** @brief Destructor */
84 ModelChecker::~ModelChecker()
86 for (unsigned int i = 0; i < get_num_threads(); i++)
87 delete thread_map->get(i);
92 delete lock_waiters_map;
93 delete condvar_waiters_map;
96 for (unsigned int i = 0; i < promises->size(); i++)
97 delete (*promises)[i];
100 delete pending_rel_seqs;
102 delete thrd_last_action;
107 for (unsigned int i = 0; i < priv->bugs.size(); i++)
108 delete priv->bugs[i];
113 static action_list_t * get_safe_ptr_action(HashTable<const void *, action_list_t *, uintptr_t, 4> * hash, void * ptr) {
114 action_list_t * tmp=hash->get(ptr);
116 tmp=new action_list_t();
122 static std::vector<action_list_t> * get_safe_ptr_vect_action(HashTable<void *, std::vector<action_list_t> *, uintptr_t, 4> * hash, void * ptr) {
123 std::vector<action_list_t> * tmp=hash->get(ptr);
125 tmp=new std::vector<action_list_t>();
132 * Restores user program to initial state and resets all model-checker data
135 void ModelChecker::reset_to_initial_state()
137 DEBUG("+++ Resetting to initial state +++\n");
138 node_stack->reset_execution();
139 failed_promise = false;
140 too_many_reads = false;
141 bad_synchronization = false;
144 /* Print all model-checker output before rollback */
147 snapshotObject->backTrackBeforeStep(0);
150 /** @return a thread ID for a new Thread */
151 thread_id_t ModelChecker::get_next_id()
153 return priv->next_thread_id++;
156 /** @return the number of user threads created during this execution */
157 unsigned int ModelChecker::get_num_threads() const
159 return priv->next_thread_id;
162 /** @return The currently executing Thread. */
163 Thread * ModelChecker::get_current_thread()
165 return scheduler->get_current_thread();
168 /** @return a sequence number for a new ModelAction */
169 modelclock_t ModelChecker::get_next_seq_num()
171 return ++priv->used_sequence_numbers;
174 Node * ModelChecker::get_curr_node() {
175 return node_stack->get_head();
179 * @brief Choose the next thread to execute.
181 * This function chooses the next thread that should execute. It can force the
182 * adjacency of read/write portions of a RMW action, force THREAD_CREATE to be
183 * followed by a THREAD_START, or it can enforce execution replay/backtracking.
184 * The model-checker may have no preference regarding the next thread (i.e.,
185 * when exploring a new execution ordering), in which case this will return
187 * @param curr The current ModelAction. This action might guide the choice of
189 * @return The next thread to run. If the model-checker has no preference, NULL.
191 Thread * ModelChecker::get_next_thread(ModelAction *curr)
196 /* Do not split atomic actions. */
198 return thread_current();
199 /* The THREAD_CREATE action points to the created Thread */
200 else if (curr->get_type() == THREAD_CREATE)
201 return (Thread *)curr->get_location();
204 /* Have we completed exploring the preselected path? */
208 /* Else, we are trying to replay an execution */
209 ModelAction *next = node_stack->get_next()->get_action();
211 if (next == diverge) {
212 if (earliest_diverge == NULL || *diverge < *earliest_diverge)
213 earliest_diverge=diverge;
215 Node *nextnode = next->get_node();
216 Node *prevnode = nextnode->get_parent();
217 scheduler->update_sleep_set(prevnode);
219 /* Reached divergence point */
220 if (nextnode->increment_misc()) {
221 /* The next node will try to satisfy a different misc_index values. */
222 tid = next->get_tid();
223 node_stack->pop_restofstack(2);
224 } else if (nextnode->increment_promise()) {
225 /* The next node will try to satisfy a different set of promises. */
226 tid = next->get_tid();
227 node_stack->pop_restofstack(2);
228 } else if (nextnode->increment_read_from()) {
229 /* The next node will read from a different value. */
230 tid = next->get_tid();
231 node_stack->pop_restofstack(2);
232 } else if (nextnode->increment_future_value()) {
233 /* The next node will try to read from a different future value. */
234 tid = next->get_tid();
235 node_stack->pop_restofstack(2);
236 } else if (nextnode->increment_relseq_break()) {
237 /* The next node will try to resolve a release sequence differently */
238 tid = next->get_tid();
239 node_stack->pop_restofstack(2);
241 /* Make a different thread execute for next step */
242 scheduler->add_sleep(thread_map->get(id_to_int(next->get_tid())));
243 tid = prevnode->get_next_backtrack();
244 /* Make sure the backtracked thread isn't sleeping. */
245 node_stack->pop_restofstack(1);
246 if (diverge==earliest_diverge) {
247 earliest_diverge=prevnode->get_action();
250 /* The correct sleep set is in the parent node. */
253 DEBUG("*** Divergence point ***\n");
257 tid = next->get_tid();
259 DEBUG("*** ModelChecker chose next thread = %d ***\n", id_to_int(tid));
260 ASSERT(tid != THREAD_ID_T_NONE);
261 return thread_map->get(id_to_int(tid));
265 * We need to know what the next actions of all threads in the sleep
266 * set will be. This method computes them and stores the actions at
267 * the corresponding thread object's pending action.
270 void ModelChecker::execute_sleep_set() {
271 for(unsigned int i=0;i<get_num_threads();i++) {
272 thread_id_t tid=int_to_id(i);
273 Thread *thr=get_thread(tid);
274 if ( scheduler->get_enabled(thr) == THREAD_SLEEP_SET &&
275 thr->get_pending() == NULL ) {
276 thr->set_state(THREAD_RUNNING);
277 scheduler->next_thread(thr);
278 Thread::swap(&system_context, thr);
279 priv->current_action->set_sleep_flag();
280 thr->set_pending(priv->current_action);
283 priv->current_action = NULL;
286 void ModelChecker::wake_up_sleeping_actions(ModelAction * curr) {
287 for(unsigned int i=0;i<get_num_threads();i++) {
288 thread_id_t tid=int_to_id(i);
289 Thread *thr=get_thread(tid);
290 if ( scheduler->get_enabled(thr) == THREAD_SLEEP_SET ) {
291 ModelAction *pending_act=thr->get_pending();
292 if ((!curr->is_rmwr())&&pending_act->could_synchronize_with(curr)) {
293 //Remove this thread from sleep set
294 scheduler->remove_sleep(thr);
301 * Check if we are in a deadlock. Should only be called at the end of an
302 * execution, although it should not give false positives in the middle of an
303 * execution (there should be some ENABLED thread).
305 * @return True if program is in a deadlock; false otherwise
307 bool ModelChecker::is_deadlocked() const
309 bool blocking_threads = false;
310 for (unsigned int i = 0; i < get_num_threads(); i++) {
311 thread_id_t tid = int_to_id(i);
314 Thread *t = get_thread(tid);
315 if (!t->is_model_thread() && t->get_pending())
316 blocking_threads = true;
318 return blocking_threads;
322 * Check if this is a complete execution. That is, have all thread completed
323 * execution (rather than exiting because sleep sets have forced a redundant
326 * @return True if the execution is complete.
328 bool ModelChecker::is_complete_execution() const
330 for (unsigned int i = 0; i < get_num_threads(); i++)
331 if (is_enabled(int_to_id(i)))
337 * @brief Assert a bug in the executing program.
339 * Use this function to assert any sort of bug in the user program. If the
340 * current trace is feasible (actually, a prefix of some feasible execution),
341 * then this execution will be aborted, printing the appropriate message. If
342 * the current trace is not yet feasible, the error message will be stashed and
343 * printed if the execution ever becomes feasible.
345 * @param msg Descriptive message for the bug (do not include newline char)
346 * @return True if bug is immediately-feasible
348 bool ModelChecker::assert_bug(const char *msg)
350 priv->bugs.push_back(new bug_message(msg));
352 if (isfeasibleprefix()) {
360 * @brief Assert a bug in the executing program, asserted by a user thread
361 * @see ModelChecker::assert_bug
362 * @param msg Descriptive message for the bug (do not include newline char)
364 void ModelChecker::assert_user_bug(const char *msg)
366 /* If feasible bug, bail out now */
368 switch_to_master(NULL);
371 /** @return True, if any bugs have been reported for this execution */
372 bool ModelChecker::have_bug_reports() const
374 return priv->bugs.size() != 0;
377 /** @brief Print bug report listing for this execution (if any bugs exist) */
378 void ModelChecker::print_bugs() const
380 if (have_bug_reports()) {
381 model_print("Bug report: %zu bug%s detected\n",
383 priv->bugs.size() > 1 ? "s" : "");
384 for (unsigned int i = 0; i < priv->bugs.size(); i++)
385 priv->bugs[i]->print();
390 * @brief Record end-of-execution stats
392 * Must be run when exiting an execution. Records various stats.
393 * @see struct execution_stats
395 void ModelChecker::record_stats()
398 if (!isfinalfeasible())
399 stats.num_infeasible++;
400 else if (have_bug_reports())
401 stats.num_buggy_executions++;
402 else if (is_complete_execution())
403 stats.num_complete++;
406 /** @brief Print execution stats */
407 void ModelChecker::print_stats() const
409 model_print("Number of complete, bug-free executions: %d\n", stats.num_complete);
410 model_print("Number of buggy executions: %d\n", stats.num_buggy_executions);
411 model_print("Number of infeasible executions: %d\n", stats.num_infeasible);
412 model_print("Total executions: %d\n", stats.num_total);
413 model_print("Total nodes created: %d\n", node_stack->get_total_nodes());
417 * @brief End-of-exeuction print
418 * @param printbugs Should any existing bugs be printed?
420 void ModelChecker::print_execution(bool printbugs) const
422 print_program_output();
424 if (DBG_ENABLED() || params.verbose) {
425 model_print("Earliest divergence point since last feasible execution:\n");
426 if (earliest_diverge)
427 earliest_diverge->print();
429 model_print("(Not set)\n");
435 /* Don't print invalid bugs */
444 * Queries the model-checker for more executions to explore and, if one
445 * exists, resets the model-checker state to execute a new execution.
447 * @return If there are more executions to explore, return true. Otherwise,
450 bool ModelChecker::next_execution()
453 /* Is this execution a feasible execution that's worth bug-checking? */
454 bool complete = isfinalfeasible() && (is_complete_execution() ||
457 /* End-of-execution bug checks */
460 assert_bug("Deadlock detected");
468 if (DBG_ENABLED() || params.verbose || have_bug_reports())
469 print_execution(complete);
471 clear_program_output();
474 earliest_diverge = NULL;
476 if ((diverge = get_next_backtrack()) == NULL)
480 model_print("Next execution will diverge at:\n");
484 reset_to_initial_state();
488 ModelAction * ModelChecker::get_last_conflict(ModelAction *act)
490 switch (act->get_type()) {
494 /* linear search: from most recent to oldest */
495 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
496 action_list_t::reverse_iterator rit;
497 for (rit = list->rbegin(); rit != list->rend(); rit++) {
498 ModelAction *prev = *rit;
499 if (prev->could_synchronize_with(act))
505 case ATOMIC_TRYLOCK: {
506 /* linear search: from most recent to oldest */
507 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
508 action_list_t::reverse_iterator rit;
509 for (rit = list->rbegin(); rit != list->rend(); rit++) {
510 ModelAction *prev = *rit;
511 if (act->is_conflicting_lock(prev))
516 case ATOMIC_UNLOCK: {
517 /* linear search: from most recent to oldest */
518 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
519 action_list_t::reverse_iterator rit;
520 for (rit = list->rbegin(); rit != list->rend(); rit++) {
521 ModelAction *prev = *rit;
522 if (!act->same_thread(prev)&&prev->is_failed_trylock())
528 /* linear search: from most recent to oldest */
529 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
530 action_list_t::reverse_iterator rit;
531 for (rit = list->rbegin(); rit != list->rend(); rit++) {
532 ModelAction *prev = *rit;
533 if (!act->same_thread(prev)&&prev->is_failed_trylock())
535 if (!act->same_thread(prev)&&prev->is_notify())
541 case ATOMIC_NOTIFY_ALL:
542 case ATOMIC_NOTIFY_ONE: {
543 /* linear search: from most recent to oldest */
544 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
545 action_list_t::reverse_iterator rit;
546 for (rit = list->rbegin(); rit != list->rend(); rit++) {
547 ModelAction *prev = *rit;
548 if (!act->same_thread(prev)&&prev->is_wait())
559 /** This method finds backtracking points where we should try to
560 * reorder the parameter ModelAction against.
562 * @param the ModelAction to find backtracking points for.
564 void ModelChecker::set_backtracking(ModelAction *act)
566 Thread *t = get_thread(act);
567 ModelAction * prev = get_last_conflict(act);
571 Node * node = prev->get_node()->get_parent();
573 int low_tid, high_tid;
574 if (node->is_enabled(t)) {
575 low_tid = id_to_int(act->get_tid());
576 high_tid = low_tid+1;
579 high_tid = get_num_threads();
582 for(int i = low_tid; i < high_tid; i++) {
583 thread_id_t tid = int_to_id(i);
585 /* Make sure this thread can be enabled here. */
586 if (i >= node->get_num_threads())
589 /* Don't backtrack into a point where the thread is disabled or sleeping. */
590 if (node->enabled_status(tid)!=THREAD_ENABLED)
593 /* Check if this has been explored already */
594 if (node->has_been_explored(tid))
597 /* See if fairness allows */
598 if (model->params.fairwindow != 0 && !node->has_priority(tid)) {
600 for(int t=0;t<node->get_num_threads();t++) {
601 thread_id_t tother=int_to_id(t);
602 if (node->is_enabled(tother) && node->has_priority(tother)) {
610 /* Cache the latest backtracking point */
611 if (!priv->next_backtrack || *prev > *priv->next_backtrack)
612 priv->next_backtrack = prev;
614 /* If this is a new backtracking point, mark the tree */
615 if (!node->set_backtrack(tid))
617 DEBUG("Setting backtrack: conflict = %d, instead tid = %d\n",
618 id_to_int(prev->get_tid()),
619 id_to_int(t->get_id()));
628 * Returns last backtracking point. The model checker will explore a different
629 * path for this point in the next execution.
630 * @return The ModelAction at which the next execution should diverge.
632 ModelAction * ModelChecker::get_next_backtrack()
634 ModelAction *next = priv->next_backtrack;
635 priv->next_backtrack = NULL;
640 * Processes a read or rmw model action.
641 * @param curr is the read model action to process.
642 * @param second_part_of_rmw is boolean that is true is this is the second action of a rmw.
643 * @return True if processing this read updates the mo_graph.
645 bool ModelChecker::process_read(ModelAction *curr, bool second_part_of_rmw)
647 uint64_t value = VALUE_NONE;
648 bool updated = false;
650 const ModelAction *reads_from = curr->get_node()->get_read_from();
651 if (reads_from != NULL) {
652 mo_graph->startChanges();
654 value = reads_from->get_value();
655 bool r_status = false;
657 if (!second_part_of_rmw) {
658 check_recency(curr, reads_from);
659 r_status = r_modification_order(curr, reads_from);
663 if (!second_part_of_rmw&&!isfeasible()&&(curr->get_node()->increment_read_from()||curr->get_node()->increment_future_value())) {
664 mo_graph->rollbackChanges();
665 too_many_reads = false;
669 curr->read_from(reads_from);
670 mo_graph->commitChanges();
671 mo_check_promises(curr->get_tid(), reads_from);
674 } else if (!second_part_of_rmw) {
675 /* Read from future value */
676 value = curr->get_node()->get_future_value();
677 modelclock_t expiration = curr->get_node()->get_future_value_expiration();
678 curr->read_from(NULL);
679 Promise *valuepromise = new Promise(curr, value, expiration);
680 promises->push_back(valuepromise);
682 get_thread(curr)->set_return_value(value);
688 * Processes a lock, trylock, or unlock model action. @param curr is
689 * the read model action to process.
691 * The try lock operation checks whether the lock is taken. If not,
692 * it falls to the normal lock operation case. If so, it returns
695 * The lock operation has already been checked that it is enabled, so
696 * it just grabs the lock and synchronizes with the previous unlock.
698 * The unlock operation has to re-enable all of the threads that are
699 * waiting on the lock.
701 * @return True if synchronization was updated; false otherwise
703 bool ModelChecker::process_mutex(ModelAction *curr) {
704 std::mutex *mutex=NULL;
705 struct std::mutex_state *state=NULL;
707 if (curr->is_trylock() || curr->is_lock() || curr->is_unlock()) {
708 mutex = (std::mutex *)curr->get_location();
709 state = mutex->get_state();
710 } else if(curr->is_wait()) {
711 mutex = (std::mutex *)curr->get_value();
712 state = mutex->get_state();
715 switch (curr->get_type()) {
716 case ATOMIC_TRYLOCK: {
717 bool success = !state->islocked;
718 curr->set_try_lock(success);
720 get_thread(curr)->set_return_value(0);
723 get_thread(curr)->set_return_value(1);
725 //otherwise fall into the lock case
727 if (curr->get_cv()->getClock(state->alloc_tid) <= state->alloc_clock)
728 assert_bug("Lock access before initialization");
729 state->islocked = true;
730 ModelAction *unlock = get_last_unlock(curr);
731 //synchronize with the previous unlock statement
732 if (unlock != NULL) {
733 curr->synchronize_with(unlock);
738 case ATOMIC_UNLOCK: {
740 state->islocked = false;
741 //wake up the other threads
742 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, curr->get_location());
743 //activate all the waiting threads
744 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
745 scheduler->wake(get_thread(*rit));
752 state->islocked = false;
753 //wake up the other threads
754 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, (void *) curr->get_value());
755 //activate all the waiting threads
756 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
757 scheduler->wake(get_thread(*rit));
760 //check whether we should go to sleep or not...simulate spurious failures
761 if (curr->get_node()->get_misc()==0) {
762 get_safe_ptr_action(condvar_waiters_map, curr->get_location())->push_back(curr);
764 scheduler->sleep(get_current_thread());
768 case ATOMIC_NOTIFY_ALL: {
769 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
770 //activate all the waiting threads
771 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
772 scheduler->wake(get_thread(*rit));
777 case ATOMIC_NOTIFY_ONE: {
778 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
779 int wakeupthread=curr->get_node()->get_misc();
780 action_list_t::iterator it = waiters->begin();
781 advance(it, wakeupthread);
782 scheduler->wake(get_thread(*it));
794 * Process a write ModelAction
795 * @param curr The ModelAction to process
796 * @return True if the mo_graph was updated or promises were resolved
798 bool ModelChecker::process_write(ModelAction *curr)
800 bool updated_mod_order = w_modification_order(curr);
801 bool updated_promises = resolve_promises(curr);
803 if (promises->size() == 0) {
804 for (unsigned int i = 0; i < futurevalues->size(); i++) {
805 struct PendingFutureValue pfv = (*futurevalues)[i];
806 //Do more ambitious checks now that mo is more complete
807 if (mo_may_allow(pfv.writer, pfv.act)&&
808 pfv.act->get_node()->add_future_value(pfv.writer->get_value(), pfv.writer->get_seq_number()+params.maxfuturedelay) &&
809 (!priv->next_backtrack || *pfv.act > *priv->next_backtrack))
810 priv->next_backtrack = pfv.act;
812 futurevalues->resize(0);
815 mo_graph->commitChanges();
816 mo_check_promises(curr->get_tid(), curr);
818 get_thread(curr)->set_return_value(VALUE_NONE);
819 return updated_mod_order || updated_promises;
823 * @brief Process the current action for thread-related activity
825 * Performs current-action processing for a THREAD_* ModelAction. Proccesses
826 * may include setting Thread status, completing THREAD_FINISH/THREAD_JOIN
827 * synchronization, etc. This function is a no-op for non-THREAD actions
828 * (e.g., ATOMIC_{READ,WRITE,RMW,LOCK}, etc.)
830 * @param curr The current action
831 * @return True if synchronization was updated or a thread completed
833 bool ModelChecker::process_thread_action(ModelAction *curr)
835 bool updated = false;
837 switch (curr->get_type()) {
838 case THREAD_CREATE: {
839 Thread *th = (Thread *)curr->get_location();
840 th->set_creation(curr);
844 Thread *blocking = (Thread *)curr->get_location();
845 ModelAction *act = get_last_action(blocking->get_id());
846 curr->synchronize_with(act);
847 updated = true; /* trigger rel-seq checks */
850 case THREAD_FINISH: {
851 Thread *th = get_thread(curr);
852 while (!th->wait_list_empty()) {
853 ModelAction *act = th->pop_wait_list();
854 scheduler->wake(get_thread(act));
857 updated = true; /* trigger rel-seq checks */
861 check_promises(curr->get_tid(), NULL, curr->get_cv());
872 * @brief Process the current action for release sequence fixup activity
874 * Performs model-checker release sequence fixups for the current action,
875 * forcing a single pending release sequence to break (with a given, potential
876 * "loose" write) or to complete (i.e., synchronize). If a pending release
877 * sequence forms a complete release sequence, then we must perform the fixup
878 * synchronization, mo_graph additions, etc.
880 * @param curr The current action; must be a release sequence fixup action
881 * @param work_queue The work queue to which to add work items as they are
884 void ModelChecker::process_relseq_fixup(ModelAction *curr, work_queue_t *work_queue)
886 const ModelAction *write = curr->get_node()->get_relseq_break();
887 struct release_seq *sequence = pending_rel_seqs->back();
888 pending_rel_seqs->pop_back();
890 ModelAction *acquire = sequence->acquire;
891 const ModelAction *rf = sequence->rf;
892 const ModelAction *release = sequence->release;
896 ASSERT(release->same_thread(rf));
900 * @todo Forcing a synchronization requires that we set
901 * modification order constraints. For instance, we can't allow
902 * a fixup sequence in which two separate read-acquire
903 * operations read from the same sequence, where the first one
904 * synchronizes and the other doesn't. Essentially, we can't
905 * allow any writes to insert themselves between 'release' and
909 /* Must synchronize */
910 if (!acquire->synchronize_with(release)) {
911 set_bad_synchronization();
914 /* Re-check all pending release sequences */
915 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
916 /* Re-check act for mo_graph edges */
917 work_queue->push_back(MOEdgeWorkEntry(acquire));
919 /* propagate synchronization to later actions */
920 action_list_t::reverse_iterator rit = action_trace->rbegin();
921 for (; (*rit) != acquire; rit++) {
922 ModelAction *propagate = *rit;
923 if (acquire->happens_before(propagate)) {
924 propagate->synchronize_with(acquire);
925 /* Re-check 'propagate' for mo_graph edges */
926 work_queue->push_back(MOEdgeWorkEntry(propagate));
930 /* Break release sequence with new edges:
931 * release --mo--> write --mo--> rf */
932 mo_graph->addEdge(release, write);
933 mo_graph->addEdge(write, rf);
936 /* See if we have realized a data race */
941 * Initialize the current action by performing one or more of the following
942 * actions, as appropriate: merging RMWR and RMWC/RMW actions, stepping forward
943 * in the NodeStack, manipulating backtracking sets, allocating and
944 * initializing clock vectors, and computing the promises to fulfill.
946 * @param curr The current action, as passed from the user context; may be
947 * freed/invalidated after the execution of this function, with a different
948 * action "returned" its place (pass-by-reference)
949 * @return True if curr is a newly-explored action; false otherwise
951 bool ModelChecker::initialize_curr_action(ModelAction **curr)
953 ModelAction *newcurr;
955 if ((*curr)->is_rmwc() || (*curr)->is_rmw()) {
956 newcurr = process_rmw(*curr);
959 if (newcurr->is_rmw())
960 compute_promises(newcurr);
966 (*curr)->set_seq_number(get_next_seq_num());
968 newcurr = node_stack->explore_action(*curr, scheduler->get_enabled());
970 /* First restore type and order in case of RMW operation */
971 if ((*curr)->is_rmwr())
972 newcurr->copy_typeandorder(*curr);
974 ASSERT((*curr)->get_location() == newcurr->get_location());
975 newcurr->copy_from_new(*curr);
977 /* Discard duplicate ModelAction; use action from NodeStack */
980 /* Always compute new clock vector */
981 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
984 return false; /* Action was explored previously */
988 /* Always compute new clock vector */
989 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
991 * Perform one-time actions when pushing new ModelAction onto
994 if (newcurr->is_write())
995 compute_promises(newcurr);
996 else if (newcurr->is_relseq_fixup())
997 compute_relseq_breakwrites(newcurr);
998 else if (newcurr->is_wait())
999 newcurr->get_node()->set_misc_max(2);
1000 else if (newcurr->is_notify_one()) {
1001 newcurr->get_node()->set_misc_max(get_safe_ptr_action(condvar_waiters_map, newcurr->get_location())->size());
1003 return true; /* This was a new ModelAction */
1008 * @brief Check whether a model action is enabled.
1010 * Checks whether a lock or join operation would be successful (i.e., is the
1011 * lock already locked, or is the joined thread already complete). If not, put
1012 * the action in a waiter list.
1014 * @param curr is the ModelAction to check whether it is enabled.
1015 * @return a bool that indicates whether the action is enabled.
1017 bool ModelChecker::check_action_enabled(ModelAction *curr) {
1018 if (curr->is_lock()) {
1019 std::mutex * lock = (std::mutex *)curr->get_location();
1020 struct std::mutex_state * state = lock->get_state();
1021 if (state->islocked) {
1022 //Stick the action in the appropriate waiting queue
1023 get_safe_ptr_action(lock_waiters_map, curr->get_location())->push_back(curr);
1026 } else if (curr->get_type() == THREAD_JOIN) {
1027 Thread *blocking = (Thread *)curr->get_location();
1028 if (!blocking->is_complete()) {
1029 blocking->push_wait_list(curr);
1038 * Stores the ModelAction for the current thread action. Call this
1039 * immediately before switching from user- to system-context to pass
1040 * data between them.
1041 * @param act The ModelAction created by the user-thread action
1043 void ModelChecker::set_current_action(ModelAction *act) {
1044 priv->current_action = act;
1048 * This is the heart of the model checker routine. It performs model-checking
1049 * actions corresponding to a given "current action." Among other processes, it
1050 * calculates reads-from relationships, updates synchronization clock vectors,
1051 * forms a memory_order constraints graph, and handles replay/backtrack
1052 * execution when running permutations of previously-observed executions.
1054 * @param curr The current action to process
1055 * @return The next Thread that must be executed. May be NULL if ModelChecker
1056 * makes no choice (e.g., according to replay execution, combining RMW actions,
1059 Thread * ModelChecker::check_current_action(ModelAction *curr)
1062 bool second_part_of_rmw = curr->is_rmwc() || curr->is_rmw();
1064 if (!check_action_enabled(curr)) {
1065 /* Make the execution look like we chose to run this action
1066 * much later, when a lock/join can succeed */
1067 get_current_thread()->set_pending(curr);
1068 scheduler->sleep(get_current_thread());
1069 return get_next_thread(NULL);
1072 bool newly_explored = initialize_curr_action(&curr);
1074 wake_up_sleeping_actions(curr);
1076 /* Add the action to lists before any other model-checking tasks */
1077 if (!second_part_of_rmw)
1078 add_action_to_lists(curr);
1080 /* Build may_read_from set for newly-created actions */
1081 if (newly_explored && curr->is_read())
1082 build_reads_from_past(curr);
1084 /* Initialize work_queue with the "current action" work */
1085 work_queue_t work_queue(1, CheckCurrWorkEntry(curr));
1086 while (!work_queue.empty() && !has_asserted()) {
1087 WorkQueueEntry work = work_queue.front();
1088 work_queue.pop_front();
1090 switch (work.type) {
1091 case WORK_CHECK_CURR_ACTION: {
1092 ModelAction *act = work.action;
1093 bool update = false; /* update this location's release seq's */
1094 bool update_all = false; /* update all release seq's */
1096 if (process_thread_action(curr))
1099 if (act->is_read() && process_read(act, second_part_of_rmw))
1102 if (act->is_write() && process_write(act))
1105 if (act->is_mutex_op() && process_mutex(act))
1108 if (act->is_relseq_fixup())
1109 process_relseq_fixup(curr, &work_queue);
1112 work_queue.push_back(CheckRelSeqWorkEntry(NULL));
1114 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1117 case WORK_CHECK_RELEASE_SEQ:
1118 resolve_release_sequences(work.location, &work_queue);
1120 case WORK_CHECK_MO_EDGES: {
1121 /** @todo Complete verification of work_queue */
1122 ModelAction *act = work.action;
1123 bool updated = false;
1125 if (act->is_read()) {
1126 const ModelAction *rf = act->get_reads_from();
1127 if (rf != NULL && r_modification_order(act, rf))
1130 if (act->is_write()) {
1131 if (w_modification_order(act))
1134 mo_graph->commitChanges();
1137 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1146 check_curr_backtracking(curr);
1147 set_backtracking(curr);
1148 return get_next_thread(curr);
1151 void ModelChecker::check_curr_backtracking(ModelAction * curr) {
1152 Node *currnode = curr->get_node();
1153 Node *parnode = currnode->get_parent();
1155 if ((!parnode->backtrack_empty() ||
1156 !currnode->misc_empty() ||
1157 !currnode->read_from_empty() ||
1158 !currnode->future_value_empty() ||
1159 !currnode->promise_empty() ||
1160 !currnode->relseq_break_empty())
1161 && (!priv->next_backtrack ||
1162 *curr > *priv->next_backtrack)) {
1163 priv->next_backtrack = curr;
1167 bool ModelChecker::promises_expired() const
1169 for (unsigned int promise_index = 0; promise_index < promises->size(); promise_index++) {
1170 Promise *promise = (*promises)[promise_index];
1171 if (promise->get_expiration()<priv->used_sequence_numbers) {
1178 /** @return whether the current partial trace must be a prefix of a
1179 * feasible trace. */
1180 bool ModelChecker::isfeasibleprefix() const
1182 return promises->size() == 0 && pending_rel_seqs->size() == 0 && isfeasible();
1185 /** @return whether the current partial trace is feasible. */
1186 bool ModelChecker::isfeasible() const
1188 if (DBG_ENABLED() && mo_graph->checkForRMWViolation())
1189 DEBUG("Infeasible: RMW violation\n");
1191 return !mo_graph->checkForRMWViolation() && isfeasibleotherthanRMW();
1194 /** @return whether the current partial trace is feasible other than
1195 * multiple RMW reading from the same store. */
1196 bool ModelChecker::isfeasibleotherthanRMW() const
1198 if (DBG_ENABLED()) {
1199 if (mo_graph->checkForCycles())
1200 DEBUG("Infeasible: modification order cycles\n");
1202 DEBUG("Infeasible: failed promise\n");
1204 DEBUG("Infeasible: too many reads\n");
1205 if (bad_synchronization)
1206 DEBUG("Infeasible: bad synchronization ordering\n");
1207 if (promises_expired())
1208 DEBUG("Infeasible: promises expired\n");
1210 return !mo_graph->checkForCycles() && !failed_promise && !too_many_reads && !bad_synchronization && !promises_expired();
1213 /** Returns whether the current completed trace is feasible. */
1214 bool ModelChecker::isfinalfeasible() const
1216 if (DBG_ENABLED() && promises->size() != 0)
1217 DEBUG("Infeasible: unrevolved promises\n");
1219 return isfeasible() && promises->size() == 0;
1222 /** Close out a RMWR by converting previous RMWR into a RMW or READ. */
1223 ModelAction * ModelChecker::process_rmw(ModelAction *act) {
1224 ModelAction *lastread = get_last_action(act->get_tid());
1225 lastread->process_rmw(act);
1226 if (act->is_rmw() && lastread->get_reads_from()!=NULL) {
1227 mo_graph->addRMWEdge(lastread->get_reads_from(), lastread);
1228 mo_graph->commitChanges();
1234 * Checks whether a thread has read from the same write for too many times
1235 * without seeing the effects of a later write.
1238 * 1) there must a different write that we could read from that would satisfy the modification order,
1239 * 2) we must have read from the same value in excess of maxreads times, and
1240 * 3) that other write must have been in the reads_from set for maxreads times.
1242 * If so, we decide that the execution is no longer feasible.
1244 void ModelChecker::check_recency(ModelAction *curr, const ModelAction *rf) {
1245 if (params.maxreads != 0) {
1247 if (curr->get_node()->get_read_from_size() <= 1)
1249 //Must make sure that execution is currently feasible... We could
1250 //accidentally clear by rolling back
1253 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1254 int tid = id_to_int(curr->get_tid());
1257 if ((int)thrd_lists->size() <= tid)
1259 action_list_t *list = &(*thrd_lists)[tid];
1261 action_list_t::reverse_iterator rit = list->rbegin();
1262 /* Skip past curr */
1263 for (; (*rit) != curr; rit++)
1265 /* go past curr now */
1268 action_list_t::reverse_iterator ritcopy = rit;
1269 //See if we have enough reads from the same value
1271 for (; count < params.maxreads; rit++,count++) {
1272 if (rit==list->rend())
1274 ModelAction *act = *rit;
1275 if (!act->is_read())
1278 if (act->get_reads_from() != rf)
1280 if (act->get_node()->get_read_from_size() <= 1)
1283 for (int i = 0; i<curr->get_node()->get_read_from_size(); i++) {
1285 const ModelAction * write = curr->get_node()->get_read_from_at(i);
1287 //Need a different write
1291 /* Test to see whether this is a feasible write to read from*/
1292 mo_graph->startChanges();
1293 r_modification_order(curr, write);
1294 bool feasiblereadfrom = isfeasible();
1295 mo_graph->rollbackChanges();
1297 if (!feasiblereadfrom)
1301 bool feasiblewrite = true;
1302 //new we need to see if this write works for everyone
1304 for (int loop = count; loop>0; loop--,rit++) {
1305 ModelAction *act=*rit;
1306 bool foundvalue = false;
1307 for (int j = 0; j<act->get_node()->get_read_from_size(); j++) {
1308 if (act->get_node()->get_read_from_at(j)==write) {
1314 feasiblewrite = false;
1318 if (feasiblewrite) {
1319 too_many_reads = true;
1327 * Updates the mo_graph with the constraints imposed from the current
1330 * Basic idea is the following: Go through each other thread and find
1331 * the lastest action that happened before our read. Two cases:
1333 * (1) The action is a write => that write must either occur before
1334 * the write we read from or be the write we read from.
1336 * (2) The action is a read => the write that that action read from
1337 * must occur before the write we read from or be the same write.
1339 * @param curr The current action. Must be a read.
1340 * @param rf The action that curr reads from. Must be a write.
1341 * @return True if modification order edges were added; false otherwise
1343 bool ModelChecker::r_modification_order(ModelAction *curr, const ModelAction *rf)
1345 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1348 ASSERT(curr->is_read());
1350 /* Iterate over all threads */
1351 for (i = 0; i < thrd_lists->size(); i++) {
1352 /* Iterate over actions in thread, starting from most recent */
1353 action_list_t *list = &(*thrd_lists)[i];
1354 action_list_t::reverse_iterator rit;
1355 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1356 ModelAction *act = *rit;
1359 * Include at most one act per-thread that "happens
1360 * before" curr. Don't consider reflexively.
1362 if (act->happens_before(curr) && act != curr) {
1363 if (act->is_write()) {
1365 mo_graph->addEdge(act, rf);
1369 const ModelAction *prevreadfrom = act->get_reads_from();
1370 //if the previous read is unresolved, keep going...
1371 if (prevreadfrom == NULL)
1374 if (rf != prevreadfrom) {
1375 mo_graph->addEdge(prevreadfrom, rf);
1387 /** This method fixes up the modification order when we resolve a
1388 * promises. The basic problem is that actions that occur after the
1389 * read curr could not property add items to the modification order
1392 * So for each thread, we find the earliest item that happens after
1393 * the read curr. This is the item we have to fix up with additional
1394 * constraints. If that action is write, we add a MO edge between
1395 * the Action rf and that action. If the action is a read, we add a
1396 * MO edge between the Action rf, and whatever the read accessed.
1398 * @param curr is the read ModelAction that we are fixing up MO edges for.
1399 * @param rf is the write ModelAction that curr reads from.
1402 void ModelChecker::post_r_modification_order(ModelAction *curr, const ModelAction *rf)
1404 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1406 ASSERT(curr->is_read());
1408 /* Iterate over all threads */
1409 for (i = 0; i < thrd_lists->size(); i++) {
1410 /* Iterate over actions in thread, starting from most recent */
1411 action_list_t *list = &(*thrd_lists)[i];
1412 action_list_t::reverse_iterator rit;
1413 ModelAction *lastact = NULL;
1415 /* Find last action that happens after curr that is either not curr or a rmw */
1416 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1417 ModelAction *act = *rit;
1418 if (curr->happens_before(act) && (curr != act || curr->is_rmw())) {
1424 /* Include at most one act per-thread that "happens before" curr */
1425 if (lastact != NULL) {
1426 if (lastact==curr) {
1427 //Case 1: The resolved read is a RMW, and we need to make sure
1428 //that the write portion of the RMW mod order after rf
1430 mo_graph->addEdge(rf, lastact);
1431 } else if (lastact->is_read()) {
1432 //Case 2: The resolved read is a normal read and the next
1433 //operation is a read, and we need to make sure the value read
1434 //is mod ordered after rf
1436 const ModelAction *postreadfrom = lastact->get_reads_from();
1437 if (postreadfrom != NULL&&rf != postreadfrom)
1438 mo_graph->addEdge(rf, postreadfrom);
1440 //Case 3: The resolved read is a normal read and the next
1441 //operation is a write, and we need to make sure that the
1442 //write is mod ordered after rf
1444 mo_graph->addEdge(rf, lastact);
1452 * Updates the mo_graph with the constraints imposed from the current write.
1454 * Basic idea is the following: Go through each other thread and find
1455 * the lastest action that happened before our write. Two cases:
1457 * (1) The action is a write => that write must occur before
1460 * (2) The action is a read => the write that that action read from
1461 * must occur before the current write.
1463 * This method also handles two other issues:
1465 * (I) Sequential Consistency: Making sure that if the current write is
1466 * seq_cst, that it occurs after the previous seq_cst write.
1468 * (II) Sending the write back to non-synchronizing reads.
1470 * @param curr The current action. Must be a write.
1471 * @return True if modification order edges were added; false otherwise
1473 bool ModelChecker::w_modification_order(ModelAction *curr)
1475 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1478 ASSERT(curr->is_write());
1480 if (curr->is_seqcst()) {
1481 /* We have to at least see the last sequentially consistent write,
1482 so we are initialized. */
1483 ModelAction *last_seq_cst = get_last_seq_cst(curr);
1484 if (last_seq_cst != NULL) {
1485 mo_graph->addEdge(last_seq_cst, curr);
1490 /* Iterate over all threads */
1491 for (i = 0; i < thrd_lists->size(); i++) {
1492 /* Iterate over actions in thread, starting from most recent */
1493 action_list_t *list = &(*thrd_lists)[i];
1494 action_list_t::reverse_iterator rit;
1495 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1496 ModelAction *act = *rit;
1499 * 1) If RMW and it actually read from something, then we
1500 * already have all relevant edges, so just skip to next
1503 * 2) If RMW and it didn't read from anything, we should
1504 * whatever edge we can get to speed up convergence.
1506 * 3) If normal write, we need to look at earlier actions, so
1507 * continue processing list.
1509 if (curr->is_rmw()) {
1510 if (curr->get_reads_from()!=NULL)
1519 * Include at most one act per-thread that "happens
1522 if (act->happens_before(curr)) {
1524 * Note: if act is RMW, just add edge:
1526 * The following edge should be handled elsewhere:
1527 * readfrom(act) --mo--> act
1529 if (act->is_write())
1530 mo_graph->addEdge(act, curr);
1531 else if (act->is_read()) {
1532 //if previous read accessed a null, just keep going
1533 if (act->get_reads_from() == NULL)
1535 mo_graph->addEdge(act->get_reads_from(), curr);
1539 } else if (act->is_read() && !act->could_synchronize_with(curr) &&
1540 !act->same_thread(curr)) {
1541 /* We have an action that:
1542 (1) did not happen before us
1543 (2) is a read and we are a write
1544 (3) cannot synchronize with us
1545 (4) is in a different thread
1547 that read could potentially read from our write. Note that
1548 these checks are overly conservative at this point, we'll
1549 do more checks before actually removing the
1553 if (thin_air_constraint_may_allow(curr, act)) {
1555 (curr->is_rmw() && act->is_rmw() && curr->get_reads_from() == act->get_reads_from() && isfeasibleotherthanRMW())) {
1556 struct PendingFutureValue pfv = {curr,act};
1557 futurevalues->push_back(pfv);
1567 /** Arbitrary reads from the future are not allowed. Section 29.3
1568 * part 9 places some constraints. This method checks one result of constraint
1569 * constraint. Others require compiler support. */
1570 bool ModelChecker::thin_air_constraint_may_allow(const ModelAction * writer, const ModelAction *reader) {
1571 if (!writer->is_rmw())
1574 if (!reader->is_rmw())
1577 for (const ModelAction *search = writer->get_reads_from(); search != NULL; search = search->get_reads_from()) {
1578 if (search == reader)
1580 if (search->get_tid() == reader->get_tid() &&
1581 search->happens_before(reader))
1589 * Arbitrary reads from the future are not allowed. Section 29.3 part 9 places
1590 * some constraints. This method checks one the following constraint (others
1591 * require compiler support):
1593 * If X --hb-> Y --mo-> Z, then X should not read from Z.
1595 bool ModelChecker::mo_may_allow(const ModelAction *writer, const ModelAction *reader)
1597 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, reader->get_location());
1599 /* Iterate over all threads */
1600 for (i = 0; i < thrd_lists->size(); i++) {
1601 const ModelAction *write_after_read = NULL;
1603 /* Iterate over actions in thread, starting from most recent */
1604 action_list_t *list = &(*thrd_lists)[i];
1605 action_list_t::reverse_iterator rit;
1606 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1607 ModelAction *act = *rit;
1609 if (!reader->happens_before(act))
1611 else if (act->is_write())
1612 write_after_read = act;
1613 else if (act->is_read() && act->get_reads_from() != NULL && act != reader) {
1614 write_after_read = act->get_reads_from();
1618 if (write_after_read && write_after_read!=writer && mo_graph->checkReachable(write_after_read, writer))
1625 * Finds the head(s) of the release sequence(s) containing a given ModelAction.
1626 * The ModelAction under consideration is expected to be taking part in
1627 * release/acquire synchronization as an object of the "reads from" relation.
1628 * Note that this can only provide release sequence support for RMW chains
1629 * which do not read from the future, as those actions cannot be traced until
1630 * their "promise" is fulfilled. Similarly, we may not even establish the
1631 * presence of a release sequence with certainty, as some modification order
1632 * constraints may be decided further in the future. Thus, this function
1633 * "returns" two pieces of data: a pass-by-reference vector of @a release_heads
1634 * and a boolean representing certainty.
1636 * @param rf The action that might be part of a release sequence. Must be a
1638 * @param release_heads A pass-by-reference style return parameter. After
1639 * execution of this function, release_heads will contain the heads of all the
1640 * relevant release sequences, if any exists with certainty
1641 * @param pending A pass-by-reference style return parameter which is only used
1642 * when returning false (i.e., uncertain). Returns most information regarding
1643 * an uncertain release sequence, including any write operations that might
1644 * break the sequence.
1645 * @return true, if the ModelChecker is certain that release_heads is complete;
1648 bool ModelChecker::release_seq_heads(const ModelAction *rf,
1649 rel_heads_list_t *release_heads,
1650 struct release_seq *pending) const
1652 /* Only check for release sequences if there are no cycles */
1653 if (mo_graph->checkForCycles())
1657 ASSERT(rf->is_write());
1659 if (rf->is_release())
1660 release_heads->push_back(rf);
1662 break; /* End of RMW chain */
1664 /** @todo Need to be smarter here... In the linux lock
1665 * example, this will run to the beginning of the program for
1667 /** @todo The way to be smarter here is to keep going until 1
1668 * thread has a release preceded by an acquire and you've seen
1671 /* acq_rel RMW is a sufficient stopping condition */
1672 if (rf->is_acquire() && rf->is_release())
1673 return true; /* complete */
1675 rf = rf->get_reads_from();
1678 /* read from future: need to settle this later */
1680 return false; /* incomplete */
1683 if (rf->is_release())
1684 return true; /* complete */
1686 /* else relaxed write; check modification order for contiguous subsequence
1687 * -> rf must be same thread as release */
1688 int tid = id_to_int(rf->get_tid());
1689 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, rf->get_location());
1690 action_list_t *list = &(*thrd_lists)[tid];
1691 action_list_t::const_reverse_iterator rit;
1693 /* Find rf in the thread list */
1694 rit = std::find(list->rbegin(), list->rend(), rf);
1695 ASSERT(rit != list->rend());
1697 /* Find the last write/release */
1698 for (; rit != list->rend(); rit++)
1699 if ((*rit)->is_release())
1701 if (rit == list->rend()) {
1702 /* No write-release in this thread */
1703 return true; /* complete */
1705 ModelAction *release = *rit;
1707 ASSERT(rf->same_thread(release));
1709 pending->writes.clear();
1711 bool certain = true;
1712 for (unsigned int i = 0; i < thrd_lists->size(); i++) {
1713 if (id_to_int(rf->get_tid()) == (int)i)
1715 list = &(*thrd_lists)[i];
1717 /* Can we ensure no future writes from this thread may break
1718 * the release seq? */
1719 bool future_ordered = false;
1721 ModelAction *last = get_last_action(int_to_id(i));
1722 Thread *th = get_thread(int_to_id(i));
1723 if ((last && rf->happens_before(last)) ||
1726 future_ordered = true;
1728 ASSERT(!th->is_model_thread() || future_ordered);
1730 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1731 const ModelAction *act = *rit;
1732 /* Reach synchronization -> this thread is complete */
1733 if (act->happens_before(release))
1735 if (rf->happens_before(act)) {
1736 future_ordered = true;
1740 /* Only non-RMW writes can break release sequences */
1741 if (!act->is_write() || act->is_rmw())
1744 /* Check modification order */
1745 if (mo_graph->checkReachable(rf, act)) {
1746 /* rf --mo--> act */
1747 future_ordered = true;
1750 if (mo_graph->checkReachable(act, release))
1751 /* act --mo--> release */
1753 if (mo_graph->checkReachable(release, act) &&
1754 mo_graph->checkReachable(act, rf)) {
1755 /* release --mo-> act --mo--> rf */
1756 return true; /* complete */
1758 /* act may break release sequence */
1759 pending->writes.push_back(act);
1762 if (!future_ordered)
1763 certain = false; /* This thread is uncertain */
1767 release_heads->push_back(release);
1768 pending->writes.clear();
1770 pending->release = release;
1777 * A public interface for getting the release sequence head(s) with which a
1778 * given ModelAction must synchronize. This function only returns a non-empty
1779 * result when it can locate a release sequence head with certainty. Otherwise,
1780 * it may mark the internal state of the ModelChecker so that it will handle
1781 * the release sequence at a later time, causing @a act to update its
1782 * synchronization at some later point in execution.
1783 * @param act The 'acquire' action that may read from a release sequence
1784 * @param release_heads A pass-by-reference return parameter. Will be filled
1785 * with the head(s) of the release sequence(s), if they exists with certainty.
1786 * @see ModelChecker::release_seq_heads
1788 void ModelChecker::get_release_seq_heads(ModelAction *act, rel_heads_list_t *release_heads)
1790 const ModelAction *rf = act->get_reads_from();
1791 struct release_seq *sequence = (struct release_seq *)snapshot_calloc(1, sizeof(struct release_seq));
1792 sequence->acquire = act;
1794 if (!release_seq_heads(rf, release_heads, sequence)) {
1795 /* add act to 'lazy checking' list */
1796 pending_rel_seqs->push_back(sequence);
1798 snapshot_free(sequence);
1803 * Attempt to resolve all stashed operations that might synchronize with a
1804 * release sequence for a given location. This implements the "lazy" portion of
1805 * determining whether or not a release sequence was contiguous, since not all
1806 * modification order information is present at the time an action occurs.
1808 * @param location The location/object that should be checked for release
1809 * sequence resolutions. A NULL value means to check all locations.
1810 * @param work_queue The work queue to which to add work items as they are
1812 * @return True if any updates occurred (new synchronization, new mo_graph
1815 bool ModelChecker::resolve_release_sequences(void *location, work_queue_t *work_queue)
1817 bool updated = false;
1818 std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >::iterator it = pending_rel_seqs->begin();
1819 while (it != pending_rel_seqs->end()) {
1820 struct release_seq *pending = *it;
1821 ModelAction *act = pending->acquire;
1823 /* Only resolve sequences on the given location, if provided */
1824 if (location && act->get_location() != location) {
1829 const ModelAction *rf = act->get_reads_from();
1830 rel_heads_list_t release_heads;
1832 complete = release_seq_heads(rf, &release_heads, pending);
1833 for (unsigned int i = 0; i < release_heads.size(); i++) {
1834 if (!act->has_synchronized_with(release_heads[i])) {
1835 if (act->synchronize_with(release_heads[i]))
1838 set_bad_synchronization();
1843 /* Re-check all pending release sequences */
1844 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
1845 /* Re-check act for mo_graph edges */
1846 work_queue->push_back(MOEdgeWorkEntry(act));
1848 /* propagate synchronization to later actions */
1849 action_list_t::reverse_iterator rit = action_trace->rbegin();
1850 for (; (*rit) != act; rit++) {
1851 ModelAction *propagate = *rit;
1852 if (act->happens_before(propagate)) {
1853 propagate->synchronize_with(act);
1854 /* Re-check 'propagate' for mo_graph edges */
1855 work_queue->push_back(MOEdgeWorkEntry(propagate));
1860 it = pending_rel_seqs->erase(it);
1861 snapshot_free(pending);
1867 // If we resolved promises or data races, see if we have realized a data race.
1874 * Performs various bookkeeping operations for the current ModelAction. For
1875 * instance, adds action to the per-object, per-thread action vector and to the
1876 * action trace list of all thread actions.
1878 * @param act is the ModelAction to add.
1880 void ModelChecker::add_action_to_lists(ModelAction *act)
1882 int tid = id_to_int(act->get_tid());
1883 action_trace->push_back(act);
1885 get_safe_ptr_action(obj_map, act->get_location())->push_back(act);
1887 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, act->get_location());
1888 if (tid >= (int)vec->size())
1889 vec->resize(priv->next_thread_id);
1890 (*vec)[tid].push_back(act);
1892 if ((int)thrd_last_action->size() <= tid)
1893 thrd_last_action->resize(get_num_threads());
1894 (*thrd_last_action)[tid] = act;
1896 if (act->is_wait()) {
1897 void *mutex_loc=(void *) act->get_value();
1898 get_safe_ptr_action(obj_map, mutex_loc)->push_back(act);
1900 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, mutex_loc);
1901 if (tid >= (int)vec->size())
1902 vec->resize(priv->next_thread_id);
1903 (*vec)[tid].push_back(act);
1905 if ((int)thrd_last_action->size() <= tid)
1906 thrd_last_action->resize(get_num_threads());
1907 (*thrd_last_action)[tid] = act;
1912 * @brief Get the last action performed by a particular Thread
1913 * @param tid The thread ID of the Thread in question
1914 * @return The last action in the thread
1916 ModelAction * ModelChecker::get_last_action(thread_id_t tid) const
1918 int threadid = id_to_int(tid);
1919 if (threadid < (int)thrd_last_action->size())
1920 return (*thrd_last_action)[id_to_int(tid)];
1926 * Gets the last memory_order_seq_cst write (in the total global sequence)
1927 * performed on a particular object (i.e., memory location), not including the
1929 * @param curr The current ModelAction; also denotes the object location to
1931 * @return The last seq_cst write
1933 ModelAction * ModelChecker::get_last_seq_cst(ModelAction *curr) const
1935 void *location = curr->get_location();
1936 action_list_t *list = get_safe_ptr_action(obj_map, location);
1937 /* Find: max({i in dom(S) | seq_cst(t_i) && isWrite(t_i) && samevar(t_i, t)}) */
1938 action_list_t::reverse_iterator rit;
1939 for (rit = list->rbegin(); rit != list->rend(); rit++)
1940 if ((*rit)->is_write() && (*rit)->is_seqcst() && (*rit) != curr)
1946 * Gets the last unlock operation performed on a particular mutex (i.e., memory
1947 * location). This function identifies the mutex according to the current
1948 * action, which is presumed to perform on the same mutex.
1949 * @param curr The current ModelAction; also denotes the object location to
1951 * @return The last unlock operation
1953 ModelAction * ModelChecker::get_last_unlock(ModelAction *curr) const
1955 void *location = curr->get_location();
1956 action_list_t *list = get_safe_ptr_action(obj_map, location);
1957 /* Find: max({i in dom(S) | isUnlock(t_i) && samevar(t_i, t)}) */
1958 action_list_t::reverse_iterator rit;
1959 for (rit = list->rbegin(); rit != list->rend(); rit++)
1960 if ((*rit)->is_unlock() || (*rit)->is_wait())
1965 ModelAction * ModelChecker::get_parent_action(thread_id_t tid)
1967 ModelAction *parent = get_last_action(tid);
1969 parent = get_thread(tid)->get_creation();
1974 * Returns the clock vector for a given thread.
1975 * @param tid The thread whose clock vector we want
1976 * @return Desired clock vector
1978 ClockVector * ModelChecker::get_cv(thread_id_t tid)
1980 return get_parent_action(tid)->get_cv();
1984 * Resolve a set of Promises with a current write. The set is provided in the
1985 * Node corresponding to @a write.
1986 * @param write The ModelAction that is fulfilling Promises
1987 * @return True if promises were resolved; false otherwise
1989 bool ModelChecker::resolve_promises(ModelAction *write)
1991 bool resolved = false;
1992 std::vector< thread_id_t, ModelAlloc<thread_id_t> > threads_to_check;
1994 for (unsigned int i = 0, promise_index = 0; promise_index < promises->size(); i++) {
1995 Promise *promise = (*promises)[promise_index];
1996 if (write->get_node()->get_promise(i)) {
1997 ModelAction *read = promise->get_action();
1998 if (read->is_rmw()) {
1999 mo_graph->addRMWEdge(write, read);
2001 read->read_from(write);
2002 //First fix up the modification order for actions that happened
2004 r_modification_order(read, write);
2005 //Next fix up the modification order for actions that happened
2007 post_r_modification_order(read, write);
2008 //Make sure the promise's value matches the write's value
2009 ASSERT(promise->get_value() == write->get_value());
2012 promises->erase(promises->begin() + promise_index);
2013 threads_to_check.push_back(read->get_tid());
2020 //Check whether reading these writes has made threads unable to
2023 for(unsigned int i=0;i<threads_to_check.size();i++)
2024 mo_check_promises(threads_to_check[i], write);
2030 * Compute the set of promises that could potentially be satisfied by this
2031 * action. Note that the set computation actually appears in the Node, not in
2033 * @param curr The ModelAction that may satisfy promises
2035 void ModelChecker::compute_promises(ModelAction *curr)
2037 for (unsigned int i = 0; i < promises->size(); i++) {
2038 Promise *promise = (*promises)[i];
2039 const ModelAction *act = promise->get_action();
2040 if (!act->happens_before(curr) &&
2042 !act->could_synchronize_with(curr) &&
2043 !act->same_thread(curr) &&
2044 act->get_location() == curr->get_location() &&
2045 promise->get_value() == curr->get_value()) {
2046 curr->get_node()->set_promise(i, act->is_rmw());
2051 /** Checks promises in response to change in ClockVector Threads. */
2052 void ModelChecker::check_promises(thread_id_t tid, ClockVector *old_cv, ClockVector *merge_cv)
2054 for (unsigned int i = 0; i < promises->size(); i++) {
2055 Promise *promise = (*promises)[i];
2056 const ModelAction *act = promise->get_action();
2057 if ((old_cv == NULL || !old_cv->synchronized_since(act)) &&
2058 merge_cv->synchronized_since(act)) {
2059 if (promise->increment_threads(tid)) {
2060 //Promise has failed
2061 failed_promise = true;
2068 void ModelChecker::check_promises_thread_disabled() {
2069 for (unsigned int i = 0; i < promises->size(); i++) {
2070 Promise *promise = (*promises)[i];
2071 if (promise->check_promise()) {
2072 failed_promise = true;
2078 /** Checks promises in response to addition to modification order for threads.
2080 * pthread is the thread that performed the read that created the promise
2082 * pread is the read that created the promise
2084 * pwrite is either the first write to same location as pread by
2085 * pthread that is sequenced after pread or the value read by the
2086 * first read to the same lcoation as pread by pthread that is
2087 * sequenced after pread..
2089 * 1. If tid=pthread, then we check what other threads are reachable
2090 * through the mode order starting with pwrite. Those threads cannot
2091 * perform a write that will resolve the promise due to modification
2092 * order constraints.
2094 * 2. If the tid is not pthread, we check whether pwrite can reach the
2095 * action write through the modification order. If so, that thread
2096 * cannot perform a future write that will resolve the promise due to
2097 * modificatin order constraints.
2099 * @parem tid The thread that either read from the model action
2100 * write, or actually did the model action write.
2102 * @parem write The ModelAction representing the relevant write.
2105 void ModelChecker::mo_check_promises(thread_id_t tid, const ModelAction *write) {
2106 void * location = write->get_location();
2107 for (unsigned int i = 0; i < promises->size(); i++) {
2108 Promise *promise = (*promises)[i];
2109 const ModelAction *act = promise->get_action();
2111 //Is this promise on the same location?
2112 if ( act->get_location() != location )
2115 //same thread as the promise
2116 if ( act->get_tid()==tid ) {
2118 //do we have a pwrite for the promise, if not, set it
2119 if (promise->get_write() == NULL ) {
2120 promise->set_write(write);
2121 //The pwrite cannot happen before the promise
2122 if (write->happens_before(act) && (write != act)) {
2123 failed_promise = true;
2127 if (mo_graph->checkPromise(write, promise)) {
2128 failed_promise = true;
2133 //Don't do any lookups twice for the same thread
2134 if (promise->has_sync_thread(tid))
2137 if (promise->get_write()&&mo_graph->checkReachable(promise->get_write(), write)) {
2138 if (promise->increment_threads(tid)) {
2139 failed_promise = true;
2147 * Compute the set of writes that may break the current pending release
2148 * sequence. This information is extracted from previou release sequence
2151 * @param curr The current ModelAction. Must be a release sequence fixup
2154 void ModelChecker::compute_relseq_breakwrites(ModelAction *curr)
2156 if (pending_rel_seqs->empty())
2159 struct release_seq *pending = pending_rel_seqs->back();
2160 for (unsigned int i = 0; i < pending->writes.size(); i++) {
2161 const ModelAction *write = pending->writes[i];
2162 curr->get_node()->add_relseq_break(write);
2165 /* NULL means don't break the sequence; just synchronize */
2166 curr->get_node()->add_relseq_break(NULL);
2170 * Build up an initial set of all past writes that this 'read' action may read
2171 * from. This set is determined by the clock vector's "happens before"
2173 * @param curr is the current ModelAction that we are exploring; it must be a
2176 void ModelChecker::build_reads_from_past(ModelAction *curr)
2178 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
2180 ASSERT(curr->is_read());
2182 ModelAction *last_seq_cst = NULL;
2184 /* Track whether this object has been initialized */
2185 bool initialized = false;
2187 if (curr->is_seqcst()) {
2188 last_seq_cst = get_last_seq_cst(curr);
2189 /* We have to at least see the last sequentially consistent write,
2190 so we are initialized. */
2191 if (last_seq_cst != NULL)
2195 /* Iterate over all threads */
2196 for (i = 0; i < thrd_lists->size(); i++) {
2197 /* Iterate over actions in thread, starting from most recent */
2198 action_list_t *list = &(*thrd_lists)[i];
2199 action_list_t::reverse_iterator rit;
2200 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2201 ModelAction *act = *rit;
2203 /* Only consider 'write' actions */
2204 if (!act->is_write() || act == curr)
2207 /* Don't consider more than one seq_cst write if we are a seq_cst read. */
2208 if (!curr->is_seqcst() || (!act->is_seqcst() && (last_seq_cst == NULL || !act->happens_before(last_seq_cst))) || act == last_seq_cst) {
2209 if (!curr->get_sleep_flag() || curr->is_seqcst() || sleep_can_read_from(curr, act)) {
2210 DEBUG("Adding action to may_read_from:\n");
2211 if (DBG_ENABLED()) {
2215 curr->get_node()->add_read_from(act);
2219 /* Include at most one act per-thread that "happens before" curr */
2220 if (act->happens_before(curr)) {
2228 assert_bug("May read from uninitialized atomic");
2230 if (DBG_ENABLED() || !initialized) {
2231 model_print("Reached read action:\n");
2233 model_print("Printing may_read_from\n");
2234 curr->get_node()->print_may_read_from();
2235 model_print("End printing may_read_from\n");
2239 bool ModelChecker::sleep_can_read_from(ModelAction * curr, const ModelAction *write) {
2241 Node *prevnode=write->get_node()->get_parent();
2243 bool thread_sleep=prevnode->enabled_status(curr->get_tid())==THREAD_SLEEP_SET;
2244 if (write->is_release()&&thread_sleep)
2246 if (!write->is_rmw()) {
2249 if (write->get_reads_from()==NULL)
2251 write=write->get_reads_from();
2255 static void print_list(action_list_t *list, int exec_num = -1)
2257 action_list_t::iterator it;
2259 model_print("---------------------------------------------------------------------\n");
2261 model_print("Execution %d:\n", exec_num);
2263 unsigned int hash=0;
2265 for (it = list->begin(); it != list->end(); it++) {
2267 hash=hash^(hash<<3)^((*it)->hash());
2269 model_print("HASH %u\n", hash);
2270 model_print("---------------------------------------------------------------------\n");
2273 #if SUPPORT_MOD_ORDER_DUMP
2274 void ModelChecker::dumpGraph(char *filename) {
2276 sprintf(buffer, "%s.dot",filename);
2277 FILE *file=fopen(buffer, "w");
2278 fprintf(file, "digraph %s {\n",filename);
2279 mo_graph->dumpNodes(file);
2280 ModelAction ** thread_array=(ModelAction **)model_calloc(1, sizeof(ModelAction *)*get_num_threads());
2282 for (action_list_t::iterator it = action_trace->begin(); it != action_trace->end(); it++) {
2283 ModelAction *action=*it;
2284 if (action->is_read()) {
2285 fprintf(file, "N%u [label=\"%u, T%u\"];\n", action->get_seq_number(),action->get_seq_number(), action->get_tid());
2286 if (action->get_reads_from()!=NULL)
2287 fprintf(file, "N%u -> N%u[label=\"rf\", color=red];\n", action->get_seq_number(), action->get_reads_from()->get_seq_number());
2289 if (thread_array[action->get_tid()] != NULL) {
2290 fprintf(file, "N%u -> N%u[label=\"sb\", color=blue];\n", thread_array[action->get_tid()]->get_seq_number(), action->get_seq_number());
2293 thread_array[action->get_tid()]=action;
2295 fprintf(file,"}\n");
2296 model_free(thread_array);
2301 /** @brief Prints an execution trace summary. */
2302 void ModelChecker::print_summary() const
2304 #if SUPPORT_MOD_ORDER_DUMP
2306 char buffername[100];
2307 sprintf(buffername, "exec%04u", stats.num_total);
2308 mo_graph->dumpGraphToFile(buffername);
2309 sprintf(buffername, "graph%04u", stats.num_total);
2310 dumpGraph(buffername);
2313 if (!isfinalfeasible())
2314 model_print("INFEASIBLE EXECUTION!\n");
2315 print_list(action_trace, stats.num_total);
2320 * Add a Thread to the system for the first time. Should only be called once
2322 * @param t The Thread to add
2324 void ModelChecker::add_thread(Thread *t)
2326 thread_map->put(id_to_int(t->get_id()), t);
2327 scheduler->add_thread(t);
2331 * Removes a thread from the scheduler.
2332 * @param the thread to remove.
2334 void ModelChecker::remove_thread(Thread *t)
2336 scheduler->remove_thread(t);
2340 * @brief Get a Thread reference by its ID
2341 * @param tid The Thread's ID
2342 * @return A Thread reference
2344 Thread * ModelChecker::get_thread(thread_id_t tid) const
2346 return thread_map->get(id_to_int(tid));
2350 * @brief Get a reference to the Thread in which a ModelAction was executed
2351 * @param act The ModelAction
2352 * @return A Thread reference
2354 Thread * ModelChecker::get_thread(ModelAction *act) const
2356 return get_thread(act->get_tid());
2360 * @brief Check if a Thread is currently enabled
2361 * @param t The Thread to check
2362 * @return True if the Thread is currently enabled
2364 bool ModelChecker::is_enabled(Thread *t) const
2366 return scheduler->is_enabled(t);
2370 * @brief Check if a Thread is currently enabled
2371 * @param tid The ID of the Thread to check
2372 * @return True if the Thread is currently enabled
2374 bool ModelChecker::is_enabled(thread_id_t tid) const
2376 return scheduler->is_enabled(tid);
2380 * Switch from a user-context to the "master thread" context (a.k.a. system
2381 * context). This switch is made with the intention of exploring a particular
2382 * model-checking action (described by a ModelAction object). Must be called
2383 * from a user-thread context.
2385 * @param act The current action that will be explored. May be NULL only if
2386 * trace is exiting via an assertion (see ModelChecker::set_assert and
2387 * ModelChecker::has_asserted).
2388 * @return Return status from the 'swap' call (i.e., success/fail, 0/-1)
2390 int ModelChecker::switch_to_master(ModelAction *act)
2393 Thread *old = thread_current();
2394 set_current_action(act);
2395 old->set_state(THREAD_READY);
2396 return Thread::swap(old, &system_context);
2400 * Takes the next step in the execution, if possible.
2401 * @return Returns true (success) if a step was taken and false otherwise.
2403 bool ModelChecker::take_step() {
2407 Thread *curr = priv->current_action ? get_thread(priv->current_action) : NULL;
2409 if (curr->get_state() == THREAD_READY) {
2410 ASSERT(priv->current_action);
2412 priv->nextThread = check_current_action(priv->current_action);
2413 priv->current_action = NULL;
2415 if (curr->is_blocked() || curr->is_complete())
2416 scheduler->remove_thread(curr);
2421 Thread *next = scheduler->next_thread(priv->nextThread);
2423 /* Infeasible -> don't take any more steps */
2426 else if (isfeasibleprefix() && have_bug_reports()) {
2431 if (params.bound != 0) {
2432 if (priv->used_sequence_numbers > params.bound) {
2437 DEBUG("(%d, %d)\n", curr ? id_to_int(curr->get_id()) : -1,
2438 next ? id_to_int(next->get_id()) : -1);
2441 * Launch end-of-execution release sequence fixups only when there are:
2443 * (1) no more user threads to run (or when execution replay chooses
2444 * the 'model_thread')
2445 * (2) pending release sequences
2446 * (3) pending assertions (i.e., data races)
2447 * (4) no pending promises
2449 if (!pending_rel_seqs->empty() && (!next || next->is_model_thread()) &&
2450 isfinalfeasible() && !unrealizedraces.empty()) {
2451 model_print("*** WARNING: release sequence fixup action (%zu pending release seuqences) ***\n",
2452 pending_rel_seqs->size());
2453 ModelAction *fixup = new ModelAction(MODEL_FIXUP_RELSEQ,
2454 std::memory_order_seq_cst, NULL, VALUE_NONE,
2456 set_current_action(fixup);
2460 /* next == NULL -> don't take any more steps */
2464 next->set_state(THREAD_RUNNING);
2466 if (next->get_pending() != NULL) {
2467 /* restart a pending action */
2468 set_current_action(next->get_pending());
2469 next->set_pending(NULL);
2470 next->set_state(THREAD_READY);
2474 /* Return false only if swap fails with an error */
2475 return (Thread::swap(&system_context, next) == 0);
2478 /** Runs the current execution until threre are no more steps to take. */
2479 void ModelChecker::finish_execution() {
2482 while (take_step());