9 #include "snapshot-interface.h"
11 #include "clockvector.h"
12 #include "cyclegraph.h"
15 #include "threads-model.h"
17 #define INITIAL_THREAD_ID 0
22 bug_message(const char *str) {
23 const char *fmt = " [BUG] %s\n";
24 msg = (char *)snapshot_malloc(strlen(fmt) + strlen(str));
25 sprintf(msg, fmt, str);
27 ~bug_message() { if (msg) snapshot_free(msg); }
30 void print() { printf("%s", msg); }
36 * Structure for holding small ModelChecker members that should be snapshotted
38 struct model_snapshot_members {
39 ModelAction *current_action;
40 unsigned int next_thread_id;
41 modelclock_t used_sequence_numbers;
43 ModelAction *next_backtrack;
44 std::vector< bug_message *, SnapshotAlloc<bug_message *> > bugs;
45 struct execution_stats stats;
48 /** @brief Constructor */
49 ModelChecker::ModelChecker(struct model_params params) :
50 /* Initialize default scheduler */
52 scheduler(new Scheduler()),
54 earliest_diverge(NULL),
55 action_trace(new action_list_t()),
56 thread_map(new HashTable<int, Thread *, int>()),
57 obj_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
58 lock_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
59 condvar_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
60 obj_thrd_map(new HashTable<void *, std::vector<action_list_t> *, uintptr_t, 4 >()),
61 promises(new std::vector< Promise *, SnapshotAlloc<Promise *> >()),
62 futurevalues(new std::vector< struct PendingFutureValue, SnapshotAlloc<struct PendingFutureValue> >()),
63 pending_rel_seqs(new std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >()),
64 thrd_last_action(new std::vector< ModelAction *, SnapshotAlloc<ModelAction *> >(1)),
65 node_stack(new NodeStack()),
66 mo_graph(new CycleGraph()),
67 failed_promise(false),
68 too_many_reads(false),
70 bad_synchronization(false)
72 /* Allocate this "size" on the snapshotting heap */
73 priv = (struct model_snapshot_members *)snapshot_calloc(1, sizeof(*priv));
74 /* First thread created will have id INITIAL_THREAD_ID */
75 priv->next_thread_id = INITIAL_THREAD_ID;
77 /* Initialize a model-checker thread, for special ModelActions */
78 model_thread = new Thread(get_next_id());
79 thread_map->put(id_to_int(model_thread->get_id()), model_thread);
82 /** @brief Destructor */
83 ModelChecker::~ModelChecker()
85 for (unsigned int i = 0; i < get_num_threads(); i++)
86 delete thread_map->get(i);
91 delete lock_waiters_map;
92 delete condvar_waiters_map;
95 for (unsigned int i = 0; i < promises->size(); i++)
96 delete (*promises)[i];
99 delete pending_rel_seqs;
101 delete thrd_last_action;
106 for (unsigned int i = 0; i < priv->bugs.size(); i++)
107 delete priv->bugs[i];
112 static action_list_t * get_safe_ptr_action(HashTable<const void *, action_list_t *, uintptr_t, 4> * hash, void * ptr) {
113 action_list_t * tmp=hash->get(ptr);
115 tmp=new action_list_t();
121 static std::vector<action_list_t> * get_safe_ptr_vect_action(HashTable<void *, std::vector<action_list_t> *, uintptr_t, 4> * hash, void * ptr) {
122 std::vector<action_list_t> * tmp=hash->get(ptr);
124 tmp=new std::vector<action_list_t>();
131 * Restores user program to initial state and resets all model-checker data
134 void ModelChecker::reset_to_initial_state()
136 DEBUG("+++ Resetting to initial state +++\n");
137 node_stack->reset_execution();
138 failed_promise = false;
139 too_many_reads = false;
140 bad_synchronization = false;
142 snapshotObject->backTrackBeforeStep(0);
145 /** @return a thread ID for a new Thread */
146 thread_id_t ModelChecker::get_next_id()
148 return priv->next_thread_id++;
151 /** @return the number of user threads created during this execution */
152 unsigned int ModelChecker::get_num_threads() const
154 return priv->next_thread_id;
157 /** @return The currently executing Thread. */
158 Thread * ModelChecker::get_current_thread()
160 return scheduler->get_current_thread();
163 /** @return a sequence number for a new ModelAction */
164 modelclock_t ModelChecker::get_next_seq_num()
166 return ++priv->used_sequence_numbers;
169 Node * ModelChecker::get_curr_node() {
170 return node_stack->get_head();
174 * @brief Choose the next thread to execute.
176 * This function chooses the next thread that should execute. It can force the
177 * adjacency of read/write portions of a RMW action, force THREAD_CREATE to be
178 * followed by a THREAD_START, or it can enforce execution replay/backtracking.
179 * The model-checker may have no preference regarding the next thread (i.e.,
180 * when exploring a new execution ordering), in which case this will return
182 * @param curr The current ModelAction. This action might guide the choice of
184 * @return The next thread to run. If the model-checker has no preference, NULL.
186 Thread * ModelChecker::get_next_thread(ModelAction *curr)
191 /* Do not split atomic actions. */
193 return thread_current();
194 /* The THREAD_CREATE action points to the created Thread */
195 else if (curr->get_type() == THREAD_CREATE)
196 return (Thread *)curr->get_location();
199 /* Have we completed exploring the preselected path? */
203 /* Else, we are trying to replay an execution */
204 ModelAction *next = node_stack->get_next()->get_action();
206 if (next == diverge) {
207 if (earliest_diverge == NULL || *diverge < *earliest_diverge)
208 earliest_diverge=diverge;
210 Node *nextnode = next->get_node();
211 Node *prevnode = nextnode->get_parent();
212 scheduler->update_sleep_set(prevnode);
214 /* Reached divergence point */
215 if (nextnode->increment_misc()) {
216 /* The next node will try to satisfy a different misc_index values. */
217 tid = next->get_tid();
218 node_stack->pop_restofstack(2);
219 } else if (nextnode->increment_promise()) {
220 /* The next node will try to satisfy a different set of promises. */
221 tid = next->get_tid();
222 node_stack->pop_restofstack(2);
223 } else if (nextnode->increment_read_from()) {
224 /* The next node will read from a different value. */
225 tid = next->get_tid();
226 node_stack->pop_restofstack(2);
227 } else if (nextnode->increment_future_value()) {
228 /* The next node will try to read from a different future value. */
229 tid = next->get_tid();
230 node_stack->pop_restofstack(2);
231 } else if (nextnode->increment_relseq_break()) {
232 /* The next node will try to resolve a release sequence differently */
233 tid = next->get_tid();
234 node_stack->pop_restofstack(2);
236 /* Make a different thread execute for next step */
237 scheduler->add_sleep(thread_map->get(id_to_int(next->get_tid())));
238 tid = prevnode->get_next_backtrack();
239 /* Make sure the backtracked thread isn't sleeping. */
240 node_stack->pop_restofstack(1);
241 if (diverge==earliest_diverge) {
242 earliest_diverge=prevnode->get_action();
245 /* The correct sleep set is in the parent node. */
248 DEBUG("*** Divergence point ***\n");
252 tid = next->get_tid();
254 DEBUG("*** ModelChecker chose next thread = %d ***\n", id_to_int(tid));
255 ASSERT(tid != THREAD_ID_T_NONE);
256 return thread_map->get(id_to_int(tid));
260 * We need to know what the next actions of all threads in the sleep
261 * set will be. This method computes them and stores the actions at
262 * the corresponding thread object's pending action.
265 void ModelChecker::execute_sleep_set() {
266 for(unsigned int i=0;i<get_num_threads();i++) {
267 thread_id_t tid=int_to_id(i);
268 Thread *thr=get_thread(tid);
269 if ( scheduler->get_enabled(thr) == THREAD_SLEEP_SET &&
270 thr->get_pending() == NULL ) {
271 thr->set_state(THREAD_RUNNING);
272 scheduler->next_thread(thr);
273 Thread::swap(&system_context, thr);
274 priv->current_action->set_sleep_flag();
275 thr->set_pending(priv->current_action);
278 priv->current_action = NULL;
281 void ModelChecker::wake_up_sleeping_actions(ModelAction * curr) {
282 for(unsigned int i=0;i<get_num_threads();i++) {
283 thread_id_t tid=int_to_id(i);
284 Thread *thr=get_thread(tid);
285 if ( scheduler->get_enabled(thr) == THREAD_SLEEP_SET ) {
286 ModelAction *pending_act=thr->get_pending();
287 if ((!curr->is_rmwr())&&pending_act->could_synchronize_with(curr)) {
288 //Remove this thread from sleep set
289 scheduler->remove_sleep(thr);
296 * Check if we are in a deadlock. Should only be called at the end of an
297 * execution, although it should not give false positives in the middle of an
298 * execution (there should be some ENABLED thread).
300 * @return True if program is in a deadlock; false otherwise
302 bool ModelChecker::is_deadlocked() const
304 bool blocking_threads = false;
305 for (unsigned int i = 0; i < get_num_threads(); i++) {
306 thread_id_t tid = int_to_id(i);
309 Thread *t = get_thread(tid);
310 if (!t->is_model_thread() && t->get_pending())
311 blocking_threads = true;
313 return blocking_threads;
317 * Check if this is a complete execution. That is, have all thread completed
318 * execution (rather than exiting because sleep sets have forced a redundant
321 * @return True if the execution is complete.
323 bool ModelChecker::is_complete_execution() const
325 for (unsigned int i = 0; i < get_num_threads(); i++)
326 if (is_enabled(int_to_id(i)))
332 * @brief Assert a bug in the executing program.
334 * Use this function to assert any sort of bug in the user program. If the
335 * current trace is feasible (actually, a prefix of some feasible execution),
336 * then this execution will be aborted, printing the appropriate message. If
337 * the current trace is not yet feasible, the error message will be stashed and
338 * printed if the execution ever becomes feasible.
340 * @param msg Descriptive message for the bug (do not include newline char)
341 * @return True if bug is immediately-feasible
343 bool ModelChecker::assert_bug(const char *msg)
345 priv->bugs.push_back(new bug_message(msg));
347 if (isfeasibleprefix()) {
355 * @brief Assert a bug in the executing program, asserted by a user thread
356 * @see ModelChecker::assert_bug
357 * @param msg Descriptive message for the bug (do not include newline char)
359 void ModelChecker::assert_user_bug(const char *msg)
361 /* If feasible bug, bail out now */
363 switch_to_master(NULL);
366 /** @return True, if any bugs have been reported for this execution */
367 bool ModelChecker::have_bug_reports() const
369 return priv->bugs.size() != 0;
372 /** @brief Print bug report listing for this execution (if any bugs exist) */
373 void ModelChecker::print_bugs() const
375 if (have_bug_reports()) {
376 printf("Bug report: %zu bugs detected\n", priv->bugs.size());
377 for (unsigned int i = 0; i < priv->bugs.size(); i++)
378 priv->bugs[i]->print();
383 * @brief Record end-of-execution stats
385 * Must be run when exiting an execution. Records various stats.
386 * @see struct execution_stats
388 void ModelChecker::record_stats()
391 if (!isfinalfeasible())
392 stats.num_infeasible++;
393 else if (have_bug_reports())
394 stats.num_buggy_executions++;
395 else if (is_complete_execution())
396 stats.num_complete++;
399 /** @brief Print execution stats */
400 void ModelChecker::print_stats() const
402 printf("Number of complete, bug-free executions: %d\n", stats.num_complete);
403 printf("Number of buggy executions: %d\n", stats.num_buggy_executions);
404 printf("Number of infeasible executions: %d\n", stats.num_infeasible);
405 printf("Total executions: %d\n", stats.num_total);
406 printf("Total nodes created: %d\n", node_stack->get_total_nodes());
410 * Queries the model-checker for more executions to explore and, if one
411 * exists, resets the model-checker state to execute a new execution.
413 * @return If there are more executions to explore, return true. Otherwise,
416 bool ModelChecker::next_execution()
422 if (isfinalfeasible() && (is_complete_execution() || have_bug_reports())) {
423 printf("Earliest divergence point since last feasible execution:\n");
424 if (earliest_diverge)
425 earliest_diverge->print();
427 printf("(Not set)\n");
429 earliest_diverge = NULL;
432 assert_bug("Deadlock detected");
439 } else if (DBG_ENABLED()) {
444 if ((diverge = get_next_backtrack()) == NULL)
448 printf("Next execution will diverge at:\n");
452 reset_to_initial_state();
456 ModelAction * ModelChecker::get_last_conflict(ModelAction *act)
458 switch (act->get_type()) {
462 /* linear search: from most recent to oldest */
463 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
464 action_list_t::reverse_iterator rit;
465 for (rit = list->rbegin(); rit != list->rend(); rit++) {
466 ModelAction *prev = *rit;
467 if (prev->could_synchronize_with(act))
473 case ATOMIC_TRYLOCK: {
474 /* linear search: from most recent to oldest */
475 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
476 action_list_t::reverse_iterator rit;
477 for (rit = list->rbegin(); rit != list->rend(); rit++) {
478 ModelAction *prev = *rit;
479 if (act->is_conflicting_lock(prev))
484 case ATOMIC_UNLOCK: {
485 /* linear search: from most recent to oldest */
486 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
487 action_list_t::reverse_iterator rit;
488 for (rit = list->rbegin(); rit != list->rend(); rit++) {
489 ModelAction *prev = *rit;
490 if (!act->same_thread(prev)&&prev->is_failed_trylock())
496 /* linear search: from most recent to oldest */
497 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
498 action_list_t::reverse_iterator rit;
499 for (rit = list->rbegin(); rit != list->rend(); rit++) {
500 ModelAction *prev = *rit;
501 if (!act->same_thread(prev)&&prev->is_failed_trylock())
503 if (!act->same_thread(prev)&&prev->is_notify())
509 case ATOMIC_NOTIFY_ALL:
510 case ATOMIC_NOTIFY_ONE: {
511 /* linear search: from most recent to oldest */
512 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
513 action_list_t::reverse_iterator rit;
514 for (rit = list->rbegin(); rit != list->rend(); rit++) {
515 ModelAction *prev = *rit;
516 if (!act->same_thread(prev)&&prev->is_wait())
527 /** This method finds backtracking points where we should try to
528 * reorder the parameter ModelAction against.
530 * @param the ModelAction to find backtracking points for.
532 void ModelChecker::set_backtracking(ModelAction *act)
534 Thread *t = get_thread(act);
535 ModelAction * prev = get_last_conflict(act);
539 Node * node = prev->get_node()->get_parent();
541 int low_tid, high_tid;
542 if (node->is_enabled(t)) {
543 low_tid = id_to_int(act->get_tid());
544 high_tid = low_tid+1;
547 high_tid = get_num_threads();
550 for(int i = low_tid; i < high_tid; i++) {
551 thread_id_t tid = int_to_id(i);
553 /* Make sure this thread can be enabled here. */
554 if (i >= node->get_num_threads())
557 /* Don't backtrack into a point where the thread is disabled or sleeping. */
558 if (node->enabled_status(tid)!=THREAD_ENABLED)
561 /* Check if this has been explored already */
562 if (node->has_been_explored(tid))
565 /* See if fairness allows */
566 if (model->params.fairwindow != 0 && !node->has_priority(tid)) {
568 for(int t=0;t<node->get_num_threads();t++) {
569 thread_id_t tother=int_to_id(t);
570 if (node->is_enabled(tother) && node->has_priority(tother)) {
578 /* Cache the latest backtracking point */
579 if (!priv->next_backtrack || *prev > *priv->next_backtrack)
580 priv->next_backtrack = prev;
582 /* If this is a new backtracking point, mark the tree */
583 if (!node->set_backtrack(tid))
585 DEBUG("Setting backtrack: conflict = %d, instead tid = %d\n",
586 id_to_int(prev->get_tid()),
587 id_to_int(t->get_id()));
596 * Returns last backtracking point. The model checker will explore a different
597 * path for this point in the next execution.
598 * @return The ModelAction at which the next execution should diverge.
600 ModelAction * ModelChecker::get_next_backtrack()
602 ModelAction *next = priv->next_backtrack;
603 priv->next_backtrack = NULL;
608 * Processes a read or rmw model action.
609 * @param curr is the read model action to process.
610 * @param second_part_of_rmw is boolean that is true is this is the second action of a rmw.
611 * @return True if processing this read updates the mo_graph.
613 bool ModelChecker::process_read(ModelAction *curr, bool second_part_of_rmw)
615 uint64_t value = VALUE_NONE;
616 bool updated = false;
618 const ModelAction *reads_from = curr->get_node()->get_read_from();
619 if (reads_from != NULL) {
620 mo_graph->startChanges();
622 value = reads_from->get_value();
623 bool r_status = false;
625 if (!second_part_of_rmw) {
626 check_recency(curr, reads_from);
627 r_status = r_modification_order(curr, reads_from);
631 if (!second_part_of_rmw&&!isfeasible()&&(curr->get_node()->increment_read_from()||curr->get_node()->increment_future_value())) {
632 mo_graph->rollbackChanges();
633 too_many_reads = false;
637 curr->read_from(reads_from);
638 mo_graph->commitChanges();
639 mo_check_promises(curr->get_tid(), reads_from);
642 } else if (!second_part_of_rmw) {
643 /* Read from future value */
644 value = curr->get_node()->get_future_value();
645 modelclock_t expiration = curr->get_node()->get_future_value_expiration();
646 curr->read_from(NULL);
647 Promise *valuepromise = new Promise(curr, value, expiration);
648 promises->push_back(valuepromise);
650 get_thread(curr)->set_return_value(value);
656 * Processes a lock, trylock, or unlock model action. @param curr is
657 * the read model action to process.
659 * The try lock operation checks whether the lock is taken. If not,
660 * it falls to the normal lock operation case. If so, it returns
663 * The lock operation has already been checked that it is enabled, so
664 * it just grabs the lock and synchronizes with the previous unlock.
666 * The unlock operation has to re-enable all of the threads that are
667 * waiting on the lock.
669 * @return True if synchronization was updated; false otherwise
671 bool ModelChecker::process_mutex(ModelAction *curr) {
672 std::mutex *mutex=NULL;
673 struct std::mutex_state *state=NULL;
675 if (curr->is_trylock() || curr->is_lock() || curr->is_unlock()) {
676 mutex = (std::mutex *)curr->get_location();
677 state = mutex->get_state();
678 } else if(curr->is_wait()) {
679 mutex = (std::mutex *)curr->get_value();
680 state = mutex->get_state();
683 switch (curr->get_type()) {
684 case ATOMIC_TRYLOCK: {
685 bool success = !state->islocked;
686 curr->set_try_lock(success);
688 get_thread(curr)->set_return_value(0);
691 get_thread(curr)->set_return_value(1);
693 //otherwise fall into the lock case
695 if (curr->get_cv()->getClock(state->alloc_tid) <= state->alloc_clock)
696 assert_bug("Lock access before initialization");
697 state->islocked = true;
698 ModelAction *unlock = get_last_unlock(curr);
699 //synchronize with the previous unlock statement
700 if (unlock != NULL) {
701 curr->synchronize_with(unlock);
706 case ATOMIC_UNLOCK: {
708 state->islocked = false;
709 //wake up the other threads
710 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, curr->get_location());
711 //activate all the waiting threads
712 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
713 scheduler->wake(get_thread(*rit));
720 state->islocked = false;
721 //wake up the other threads
722 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, (void *) curr->get_value());
723 //activate all the waiting threads
724 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
725 scheduler->wake(get_thread(*rit));
728 //check whether we should go to sleep or not...simulate spurious failures
729 if (curr->get_node()->get_misc()==0) {
730 get_safe_ptr_action(condvar_waiters_map, curr->get_location())->push_back(curr);
732 scheduler->sleep(get_current_thread());
736 case ATOMIC_NOTIFY_ALL: {
737 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
738 //activate all the waiting threads
739 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
740 scheduler->wake(get_thread(*rit));
745 case ATOMIC_NOTIFY_ONE: {
746 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
747 int wakeupthread=curr->get_node()->get_misc();
748 action_list_t::iterator it = waiters->begin();
749 advance(it, wakeupthread);
750 scheduler->wake(get_thread(*it));
762 * Process a write ModelAction
763 * @param curr The ModelAction to process
764 * @return True if the mo_graph was updated or promises were resolved
766 bool ModelChecker::process_write(ModelAction *curr)
768 bool updated_mod_order = w_modification_order(curr);
769 bool updated_promises = resolve_promises(curr);
771 if (promises->size() == 0) {
772 for (unsigned int i = 0; i < futurevalues->size(); i++) {
773 struct PendingFutureValue pfv = (*futurevalues)[i];
774 //Do more ambitious checks now that mo is more complete
775 if (mo_may_allow(pfv.writer, pfv.act)&&
776 pfv.act->get_node()->add_future_value(pfv.writer->get_value(), pfv.writer->get_seq_number()+params.maxfuturedelay) &&
777 (!priv->next_backtrack || *pfv.act > *priv->next_backtrack))
778 priv->next_backtrack = pfv.act;
780 futurevalues->resize(0);
783 mo_graph->commitChanges();
784 mo_check_promises(curr->get_tid(), curr);
786 get_thread(curr)->set_return_value(VALUE_NONE);
787 return updated_mod_order || updated_promises;
791 * @brief Process the current action for thread-related activity
793 * Performs current-action processing for a THREAD_* ModelAction. Proccesses
794 * may include setting Thread status, completing THREAD_FINISH/THREAD_JOIN
795 * synchronization, etc. This function is a no-op for non-THREAD actions
796 * (e.g., ATOMIC_{READ,WRITE,RMW,LOCK}, etc.)
798 * @param curr The current action
799 * @return True if synchronization was updated or a thread completed
801 bool ModelChecker::process_thread_action(ModelAction *curr)
803 bool updated = false;
805 switch (curr->get_type()) {
806 case THREAD_CREATE: {
807 Thread *th = (Thread *)curr->get_location();
808 th->set_creation(curr);
812 Thread *blocking = (Thread *)curr->get_location();
813 ModelAction *act = get_last_action(blocking->get_id());
814 curr->synchronize_with(act);
815 updated = true; /* trigger rel-seq checks */
818 case THREAD_FINISH: {
819 Thread *th = get_thread(curr);
820 while (!th->wait_list_empty()) {
821 ModelAction *act = th->pop_wait_list();
822 scheduler->wake(get_thread(act));
825 updated = true; /* trigger rel-seq checks */
829 check_promises(curr->get_tid(), NULL, curr->get_cv());
840 * @brief Process the current action for release sequence fixup activity
842 * Performs model-checker release sequence fixups for the current action,
843 * forcing a single pending release sequence to break (with a given, potential
844 * "loose" write) or to complete (i.e., synchronize). If a pending release
845 * sequence forms a complete release sequence, then we must perform the fixup
846 * synchronization, mo_graph additions, etc.
848 * @param curr The current action; must be a release sequence fixup action
849 * @param work_queue The work queue to which to add work items as they are
852 void ModelChecker::process_relseq_fixup(ModelAction *curr, work_queue_t *work_queue)
854 const ModelAction *write = curr->get_node()->get_relseq_break();
855 struct release_seq *sequence = pending_rel_seqs->back();
856 pending_rel_seqs->pop_back();
858 ModelAction *acquire = sequence->acquire;
859 const ModelAction *rf = sequence->rf;
860 const ModelAction *release = sequence->release;
864 ASSERT(release->same_thread(rf));
868 * @todo Forcing a synchronization requires that we set
869 * modification order constraints. For instance, we can't allow
870 * a fixup sequence in which two separate read-acquire
871 * operations read from the same sequence, where the first one
872 * synchronizes and the other doesn't. Essentially, we can't
873 * allow any writes to insert themselves between 'release' and
877 /* Must synchronize */
878 if (!acquire->synchronize_with(release)) {
879 set_bad_synchronization();
882 /* Re-check all pending release sequences */
883 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
884 /* Re-check act for mo_graph edges */
885 work_queue->push_back(MOEdgeWorkEntry(acquire));
887 /* propagate synchronization to later actions */
888 action_list_t::reverse_iterator rit = action_trace->rbegin();
889 for (; (*rit) != acquire; rit++) {
890 ModelAction *propagate = *rit;
891 if (acquire->happens_before(propagate)) {
892 propagate->synchronize_with(acquire);
893 /* Re-check 'propagate' for mo_graph edges */
894 work_queue->push_back(MOEdgeWorkEntry(propagate));
898 /* Break release sequence with new edges:
899 * release --mo--> write --mo--> rf */
900 mo_graph->addEdge(release, write);
901 mo_graph->addEdge(write, rf);
904 /* See if we have realized a data race */
909 * Initialize the current action by performing one or more of the following
910 * actions, as appropriate: merging RMWR and RMWC/RMW actions, stepping forward
911 * in the NodeStack, manipulating backtracking sets, allocating and
912 * initializing clock vectors, and computing the promises to fulfill.
914 * @param curr The current action, as passed from the user context; may be
915 * freed/invalidated after the execution of this function, with a different
916 * action "returned" its place (pass-by-reference)
917 * @return True if curr is a newly-explored action; false otherwise
919 bool ModelChecker::initialize_curr_action(ModelAction **curr)
921 ModelAction *newcurr;
923 if ((*curr)->is_rmwc() || (*curr)->is_rmw()) {
924 newcurr = process_rmw(*curr);
927 if (newcurr->is_rmw())
928 compute_promises(newcurr);
934 (*curr)->set_seq_number(get_next_seq_num());
936 newcurr = node_stack->explore_action(*curr, scheduler->get_enabled());
938 /* First restore type and order in case of RMW operation */
939 if ((*curr)->is_rmwr())
940 newcurr->copy_typeandorder(*curr);
942 ASSERT((*curr)->get_location() == newcurr->get_location());
943 newcurr->copy_from_new(*curr);
945 /* Discard duplicate ModelAction; use action from NodeStack */
948 /* Always compute new clock vector */
949 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
952 return false; /* Action was explored previously */
956 /* Always compute new clock vector */
957 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
959 * Perform one-time actions when pushing new ModelAction onto
962 if (newcurr->is_write())
963 compute_promises(newcurr);
964 else if (newcurr->is_relseq_fixup())
965 compute_relseq_breakwrites(newcurr);
966 else if (newcurr->is_wait())
967 newcurr->get_node()->set_misc_max(2);
968 else if (newcurr->is_notify_one()) {
969 newcurr->get_node()->set_misc_max(get_safe_ptr_action(condvar_waiters_map, newcurr->get_location())->size());
971 return true; /* This was a new ModelAction */
976 * @brief Check whether a model action is enabled.
978 * Checks whether a lock or join operation would be successful (i.e., is the
979 * lock already locked, or is the joined thread already complete). If not, put
980 * the action in a waiter list.
982 * @param curr is the ModelAction to check whether it is enabled.
983 * @return a bool that indicates whether the action is enabled.
985 bool ModelChecker::check_action_enabled(ModelAction *curr) {
986 if (curr->is_lock()) {
987 std::mutex * lock = (std::mutex *)curr->get_location();
988 struct std::mutex_state * state = lock->get_state();
989 if (state->islocked) {
990 //Stick the action in the appropriate waiting queue
991 get_safe_ptr_action(lock_waiters_map, curr->get_location())->push_back(curr);
994 } else if (curr->get_type() == THREAD_JOIN) {
995 Thread *blocking = (Thread *)curr->get_location();
996 if (!blocking->is_complete()) {
997 blocking->push_wait_list(curr);
1006 * Stores the ModelAction for the current thread action. Call this
1007 * immediately before switching from user- to system-context to pass
1008 * data between them.
1009 * @param act The ModelAction created by the user-thread action
1011 void ModelChecker::set_current_action(ModelAction *act) {
1012 priv->current_action = act;
1016 * This is the heart of the model checker routine. It performs model-checking
1017 * actions corresponding to a given "current action." Among other processes, it
1018 * calculates reads-from relationships, updates synchronization clock vectors,
1019 * forms a memory_order constraints graph, and handles replay/backtrack
1020 * execution when running permutations of previously-observed executions.
1022 * @param curr The current action to process
1023 * @return The next Thread that must be executed. May be NULL if ModelChecker
1024 * makes no choice (e.g., according to replay execution, combining RMW actions,
1027 Thread * ModelChecker::check_current_action(ModelAction *curr)
1030 bool second_part_of_rmw = curr->is_rmwc() || curr->is_rmw();
1032 if (!check_action_enabled(curr)) {
1033 /* Make the execution look like we chose to run this action
1034 * much later, when a lock/join can succeed */
1035 get_current_thread()->set_pending(curr);
1036 scheduler->sleep(get_current_thread());
1037 return get_next_thread(NULL);
1040 bool newly_explored = initialize_curr_action(&curr);
1042 wake_up_sleeping_actions(curr);
1044 /* Add the action to lists before any other model-checking tasks */
1045 if (!second_part_of_rmw)
1046 add_action_to_lists(curr);
1048 /* Build may_read_from set for newly-created actions */
1049 if (newly_explored && curr->is_read())
1050 build_reads_from_past(curr);
1052 /* Initialize work_queue with the "current action" work */
1053 work_queue_t work_queue(1, CheckCurrWorkEntry(curr));
1054 while (!work_queue.empty() && !has_asserted()) {
1055 WorkQueueEntry work = work_queue.front();
1056 work_queue.pop_front();
1058 switch (work.type) {
1059 case WORK_CHECK_CURR_ACTION: {
1060 ModelAction *act = work.action;
1061 bool update = false; /* update this location's release seq's */
1062 bool update_all = false; /* update all release seq's */
1064 if (process_thread_action(curr))
1067 if (act->is_read() && process_read(act, second_part_of_rmw))
1070 if (act->is_write() && process_write(act))
1073 if (act->is_mutex_op() && process_mutex(act))
1076 if (act->is_relseq_fixup())
1077 process_relseq_fixup(curr, &work_queue);
1080 work_queue.push_back(CheckRelSeqWorkEntry(NULL));
1082 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1085 case WORK_CHECK_RELEASE_SEQ:
1086 resolve_release_sequences(work.location, &work_queue);
1088 case WORK_CHECK_MO_EDGES: {
1089 /** @todo Complete verification of work_queue */
1090 ModelAction *act = work.action;
1091 bool updated = false;
1093 if (act->is_read()) {
1094 const ModelAction *rf = act->get_reads_from();
1095 if (rf != NULL && r_modification_order(act, rf))
1098 if (act->is_write()) {
1099 if (w_modification_order(act))
1102 mo_graph->commitChanges();
1105 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1114 check_curr_backtracking(curr);
1115 set_backtracking(curr);
1116 return get_next_thread(curr);
1119 void ModelChecker::check_curr_backtracking(ModelAction * curr) {
1120 Node *currnode = curr->get_node();
1121 Node *parnode = currnode->get_parent();
1123 if ((!parnode->backtrack_empty() ||
1124 !currnode->misc_empty() ||
1125 !currnode->read_from_empty() ||
1126 !currnode->future_value_empty() ||
1127 !currnode->promise_empty() ||
1128 !currnode->relseq_break_empty())
1129 && (!priv->next_backtrack ||
1130 *curr > *priv->next_backtrack)) {
1131 priv->next_backtrack = curr;
1135 bool ModelChecker::promises_expired() const
1137 for (unsigned int promise_index = 0; promise_index < promises->size(); promise_index++) {
1138 Promise *promise = (*promises)[promise_index];
1139 if (promise->get_expiration()<priv->used_sequence_numbers) {
1146 /** @return whether the current partial trace must be a prefix of a
1147 * feasible trace. */
1148 bool ModelChecker::isfeasibleprefix() const
1150 return promises->size() == 0 && pending_rel_seqs->size() == 0 && isfeasible();
1153 /** @return whether the current partial trace is feasible. */
1154 bool ModelChecker::isfeasible() const
1156 if (DBG_ENABLED() && mo_graph->checkForRMWViolation())
1157 DEBUG("Infeasible: RMW violation\n");
1159 return !mo_graph->checkForRMWViolation() && isfeasibleotherthanRMW();
1162 /** @return whether the current partial trace is feasible other than
1163 * multiple RMW reading from the same store. */
1164 bool ModelChecker::isfeasibleotherthanRMW() const
1166 if (DBG_ENABLED()) {
1167 if (mo_graph->checkForCycles())
1168 DEBUG("Infeasible: modification order cycles\n");
1170 DEBUG("Infeasible: failed promise\n");
1172 DEBUG("Infeasible: too many reads\n");
1173 if (bad_synchronization)
1174 DEBUG("Infeasible: bad synchronization ordering\n");
1175 if (promises_expired())
1176 DEBUG("Infeasible: promises expired\n");
1178 return !mo_graph->checkForCycles() && !failed_promise && !too_many_reads && !bad_synchronization && !promises_expired();
1181 /** Returns whether the current completed trace is feasible. */
1182 bool ModelChecker::isfinalfeasible() const
1184 if (DBG_ENABLED() && promises->size() != 0)
1185 DEBUG("Infeasible: unrevolved promises\n");
1187 return isfeasible() && promises->size() == 0;
1190 /** Close out a RMWR by converting previous RMWR into a RMW or READ. */
1191 ModelAction * ModelChecker::process_rmw(ModelAction *act) {
1192 ModelAction *lastread = get_last_action(act->get_tid());
1193 lastread->process_rmw(act);
1194 if (act->is_rmw() && lastread->get_reads_from()!=NULL) {
1195 mo_graph->addRMWEdge(lastread->get_reads_from(), lastread);
1196 mo_graph->commitChanges();
1202 * Checks whether a thread has read from the same write for too many times
1203 * without seeing the effects of a later write.
1206 * 1) there must a different write that we could read from that would satisfy the modification order,
1207 * 2) we must have read from the same value in excess of maxreads times, and
1208 * 3) that other write must have been in the reads_from set for maxreads times.
1210 * If so, we decide that the execution is no longer feasible.
1212 void ModelChecker::check_recency(ModelAction *curr, const ModelAction *rf) {
1213 if (params.maxreads != 0) {
1215 if (curr->get_node()->get_read_from_size() <= 1)
1217 //Must make sure that execution is currently feasible... We could
1218 //accidentally clear by rolling back
1221 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1222 int tid = id_to_int(curr->get_tid());
1225 if ((int)thrd_lists->size() <= tid)
1227 action_list_t *list = &(*thrd_lists)[tid];
1229 action_list_t::reverse_iterator rit = list->rbegin();
1230 /* Skip past curr */
1231 for (; (*rit) != curr; rit++)
1233 /* go past curr now */
1236 action_list_t::reverse_iterator ritcopy = rit;
1237 //See if we have enough reads from the same value
1239 for (; count < params.maxreads; rit++,count++) {
1240 if (rit==list->rend())
1242 ModelAction *act = *rit;
1243 if (!act->is_read())
1246 if (act->get_reads_from() != rf)
1248 if (act->get_node()->get_read_from_size() <= 1)
1251 for (int i = 0; i<curr->get_node()->get_read_from_size(); i++) {
1253 const ModelAction * write = curr->get_node()->get_read_from_at(i);
1255 //Need a different write
1259 /* Test to see whether this is a feasible write to read from*/
1260 mo_graph->startChanges();
1261 r_modification_order(curr, write);
1262 bool feasiblereadfrom = isfeasible();
1263 mo_graph->rollbackChanges();
1265 if (!feasiblereadfrom)
1269 bool feasiblewrite = true;
1270 //new we need to see if this write works for everyone
1272 for (int loop = count; loop>0; loop--,rit++) {
1273 ModelAction *act=*rit;
1274 bool foundvalue = false;
1275 for (int j = 0; j<act->get_node()->get_read_from_size(); j++) {
1276 if (act->get_node()->get_read_from_at(j)==write) {
1282 feasiblewrite = false;
1286 if (feasiblewrite) {
1287 too_many_reads = true;
1295 * Updates the mo_graph with the constraints imposed from the current
1298 * Basic idea is the following: Go through each other thread and find
1299 * the lastest action that happened before our read. Two cases:
1301 * (1) The action is a write => that write must either occur before
1302 * the write we read from or be the write we read from.
1304 * (2) The action is a read => the write that that action read from
1305 * must occur before the write we read from or be the same write.
1307 * @param curr The current action. Must be a read.
1308 * @param rf The action that curr reads from. Must be a write.
1309 * @return True if modification order edges were added; false otherwise
1311 bool ModelChecker::r_modification_order(ModelAction *curr, const ModelAction *rf)
1313 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1316 ASSERT(curr->is_read());
1318 /* Iterate over all threads */
1319 for (i = 0; i < thrd_lists->size(); i++) {
1320 /* Iterate over actions in thread, starting from most recent */
1321 action_list_t *list = &(*thrd_lists)[i];
1322 action_list_t::reverse_iterator rit;
1323 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1324 ModelAction *act = *rit;
1327 * Include at most one act per-thread that "happens
1328 * before" curr. Don't consider reflexively.
1330 if (act->happens_before(curr) && act != curr) {
1331 if (act->is_write()) {
1333 mo_graph->addEdge(act, rf);
1337 const ModelAction *prevreadfrom = act->get_reads_from();
1338 //if the previous read is unresolved, keep going...
1339 if (prevreadfrom == NULL)
1342 if (rf != prevreadfrom) {
1343 mo_graph->addEdge(prevreadfrom, rf);
1355 /** This method fixes up the modification order when we resolve a
1356 * promises. The basic problem is that actions that occur after the
1357 * read curr could not property add items to the modification order
1360 * So for each thread, we find the earliest item that happens after
1361 * the read curr. This is the item we have to fix up with additional
1362 * constraints. If that action is write, we add a MO edge between
1363 * the Action rf and that action. If the action is a read, we add a
1364 * MO edge between the Action rf, and whatever the read accessed.
1366 * @param curr is the read ModelAction that we are fixing up MO edges for.
1367 * @param rf is the write ModelAction that curr reads from.
1370 void ModelChecker::post_r_modification_order(ModelAction *curr, const ModelAction *rf)
1372 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1374 ASSERT(curr->is_read());
1376 /* Iterate over all threads */
1377 for (i = 0; i < thrd_lists->size(); i++) {
1378 /* Iterate over actions in thread, starting from most recent */
1379 action_list_t *list = &(*thrd_lists)[i];
1380 action_list_t::reverse_iterator rit;
1381 ModelAction *lastact = NULL;
1383 /* Find last action that happens after curr that is either not curr or a rmw */
1384 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1385 ModelAction *act = *rit;
1386 if (curr->happens_before(act) && (curr != act || curr->is_rmw())) {
1392 /* Include at most one act per-thread that "happens before" curr */
1393 if (lastact != NULL) {
1394 if (lastact==curr) {
1395 //Case 1: The resolved read is a RMW, and we need to make sure
1396 //that the write portion of the RMW mod order after rf
1398 mo_graph->addEdge(rf, lastact);
1399 } else if (lastact->is_read()) {
1400 //Case 2: The resolved read is a normal read and the next
1401 //operation is a read, and we need to make sure the value read
1402 //is mod ordered after rf
1404 const ModelAction *postreadfrom = lastact->get_reads_from();
1405 if (postreadfrom != NULL&&rf != postreadfrom)
1406 mo_graph->addEdge(rf, postreadfrom);
1408 //Case 3: The resolved read is a normal read and the next
1409 //operation is a write, and we need to make sure that the
1410 //write is mod ordered after rf
1412 mo_graph->addEdge(rf, lastact);
1420 * Updates the mo_graph with the constraints imposed from the current write.
1422 * Basic idea is the following: Go through each other thread and find
1423 * the lastest action that happened before our write. Two cases:
1425 * (1) The action is a write => that write must occur before
1428 * (2) The action is a read => the write that that action read from
1429 * must occur before the current write.
1431 * This method also handles two other issues:
1433 * (I) Sequential Consistency: Making sure that if the current write is
1434 * seq_cst, that it occurs after the previous seq_cst write.
1436 * (II) Sending the write back to non-synchronizing reads.
1438 * @param curr The current action. Must be a write.
1439 * @return True if modification order edges were added; false otherwise
1441 bool ModelChecker::w_modification_order(ModelAction *curr)
1443 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1446 ASSERT(curr->is_write());
1448 if (curr->is_seqcst()) {
1449 /* We have to at least see the last sequentially consistent write,
1450 so we are initialized. */
1451 ModelAction *last_seq_cst = get_last_seq_cst(curr);
1452 if (last_seq_cst != NULL) {
1453 mo_graph->addEdge(last_seq_cst, curr);
1458 /* Iterate over all threads */
1459 for (i = 0; i < thrd_lists->size(); i++) {
1460 /* Iterate over actions in thread, starting from most recent */
1461 action_list_t *list = &(*thrd_lists)[i];
1462 action_list_t::reverse_iterator rit;
1463 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1464 ModelAction *act = *rit;
1467 * 1) If RMW and it actually read from something, then we
1468 * already have all relevant edges, so just skip to next
1471 * 2) If RMW and it didn't read from anything, we should
1472 * whatever edge we can get to speed up convergence.
1474 * 3) If normal write, we need to look at earlier actions, so
1475 * continue processing list.
1477 if (curr->is_rmw()) {
1478 if (curr->get_reads_from()!=NULL)
1487 * Include at most one act per-thread that "happens
1490 if (act->happens_before(curr)) {
1492 * Note: if act is RMW, just add edge:
1494 * The following edge should be handled elsewhere:
1495 * readfrom(act) --mo--> act
1497 if (act->is_write())
1498 mo_graph->addEdge(act, curr);
1499 else if (act->is_read()) {
1500 //if previous read accessed a null, just keep going
1501 if (act->get_reads_from() == NULL)
1503 mo_graph->addEdge(act->get_reads_from(), curr);
1507 } else if (act->is_read() && !act->could_synchronize_with(curr) &&
1508 !act->same_thread(curr)) {
1509 /* We have an action that:
1510 (1) did not happen before us
1511 (2) is a read and we are a write
1512 (3) cannot synchronize with us
1513 (4) is in a different thread
1515 that read could potentially read from our write. Note that
1516 these checks are overly conservative at this point, we'll
1517 do more checks before actually removing the
1521 if (thin_air_constraint_may_allow(curr, act)) {
1523 (curr->is_rmw() && act->is_rmw() && curr->get_reads_from() == act->get_reads_from() && isfeasibleotherthanRMW())) {
1524 struct PendingFutureValue pfv = {curr,act};
1525 futurevalues->push_back(pfv);
1535 /** Arbitrary reads from the future are not allowed. Section 29.3
1536 * part 9 places some constraints. This method checks one result of constraint
1537 * constraint. Others require compiler support. */
1538 bool ModelChecker::thin_air_constraint_may_allow(const ModelAction * writer, const ModelAction *reader) {
1539 if (!writer->is_rmw())
1542 if (!reader->is_rmw())
1545 for (const ModelAction *search = writer->get_reads_from(); search != NULL; search = search->get_reads_from()) {
1546 if (search == reader)
1548 if (search->get_tid() == reader->get_tid() &&
1549 search->happens_before(reader))
1557 * Arbitrary reads from the future are not allowed. Section 29.3 part 9 places
1558 * some constraints. This method checks one the following constraint (others
1559 * require compiler support):
1561 * If X --hb-> Y --mo-> Z, then X should not read from Z.
1563 bool ModelChecker::mo_may_allow(const ModelAction *writer, const ModelAction *reader)
1565 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, reader->get_location());
1567 /* Iterate over all threads */
1568 for (i = 0; i < thrd_lists->size(); i++) {
1569 const ModelAction *write_after_read = NULL;
1571 /* Iterate over actions in thread, starting from most recent */
1572 action_list_t *list = &(*thrd_lists)[i];
1573 action_list_t::reverse_iterator rit;
1574 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1575 ModelAction *act = *rit;
1577 if (!reader->happens_before(act))
1579 else if (act->is_write())
1580 write_after_read = act;
1581 else if (act->is_read() && act->get_reads_from() != NULL && act != reader) {
1582 write_after_read = act->get_reads_from();
1586 if (write_after_read && write_after_read!=writer && mo_graph->checkReachable(write_after_read, writer))
1593 * Finds the head(s) of the release sequence(s) containing a given ModelAction.
1594 * The ModelAction under consideration is expected to be taking part in
1595 * release/acquire synchronization as an object of the "reads from" relation.
1596 * Note that this can only provide release sequence support for RMW chains
1597 * which do not read from the future, as those actions cannot be traced until
1598 * their "promise" is fulfilled. Similarly, we may not even establish the
1599 * presence of a release sequence with certainty, as some modification order
1600 * constraints may be decided further in the future. Thus, this function
1601 * "returns" two pieces of data: a pass-by-reference vector of @a release_heads
1602 * and a boolean representing certainty.
1604 * @param rf The action that might be part of a release sequence. Must be a
1606 * @param release_heads A pass-by-reference style return parameter. After
1607 * execution of this function, release_heads will contain the heads of all the
1608 * relevant release sequences, if any exists with certainty
1609 * @param pending A pass-by-reference style return parameter which is only used
1610 * when returning false (i.e., uncertain). Returns most information regarding
1611 * an uncertain release sequence, including any write operations that might
1612 * break the sequence.
1613 * @return true, if the ModelChecker is certain that release_heads is complete;
1616 bool ModelChecker::release_seq_heads(const ModelAction *rf,
1617 rel_heads_list_t *release_heads,
1618 struct release_seq *pending) const
1620 /* Only check for release sequences if there are no cycles */
1621 if (mo_graph->checkForCycles())
1625 ASSERT(rf->is_write());
1627 if (rf->is_release())
1628 release_heads->push_back(rf);
1630 break; /* End of RMW chain */
1632 /** @todo Need to be smarter here... In the linux lock
1633 * example, this will run to the beginning of the program for
1635 /** @todo The way to be smarter here is to keep going until 1
1636 * thread has a release preceded by an acquire and you've seen
1639 /* acq_rel RMW is a sufficient stopping condition */
1640 if (rf->is_acquire() && rf->is_release())
1641 return true; /* complete */
1643 rf = rf->get_reads_from();
1646 /* read from future: need to settle this later */
1648 return false; /* incomplete */
1651 if (rf->is_release())
1652 return true; /* complete */
1654 /* else relaxed write; check modification order for contiguous subsequence
1655 * -> rf must be same thread as release */
1656 int tid = id_to_int(rf->get_tid());
1657 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, rf->get_location());
1658 action_list_t *list = &(*thrd_lists)[tid];
1659 action_list_t::const_reverse_iterator rit;
1661 /* Find rf in the thread list */
1662 rit = std::find(list->rbegin(), list->rend(), rf);
1663 ASSERT(rit != list->rend());
1665 /* Find the last write/release */
1666 for (; rit != list->rend(); rit++)
1667 if ((*rit)->is_release())
1669 if (rit == list->rend()) {
1670 /* No write-release in this thread */
1671 return true; /* complete */
1673 ModelAction *release = *rit;
1675 ASSERT(rf->same_thread(release));
1677 pending->writes.clear();
1679 bool certain = true;
1680 for (unsigned int i = 0; i < thrd_lists->size(); i++) {
1681 if (id_to_int(rf->get_tid()) == (int)i)
1683 list = &(*thrd_lists)[i];
1685 /* Can we ensure no future writes from this thread may break
1686 * the release seq? */
1687 bool future_ordered = false;
1689 ModelAction *last = get_last_action(int_to_id(i));
1690 Thread *th = get_thread(int_to_id(i));
1691 if ((last && rf->happens_before(last)) ||
1694 future_ordered = true;
1696 ASSERT(!th->is_model_thread() || future_ordered);
1698 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1699 const ModelAction *act = *rit;
1700 /* Reach synchronization -> this thread is complete */
1701 if (act->happens_before(release))
1703 if (rf->happens_before(act)) {
1704 future_ordered = true;
1708 /* Only non-RMW writes can break release sequences */
1709 if (!act->is_write() || act->is_rmw())
1712 /* Check modification order */
1713 if (mo_graph->checkReachable(rf, act)) {
1714 /* rf --mo--> act */
1715 future_ordered = true;
1718 if (mo_graph->checkReachable(act, release))
1719 /* act --mo--> release */
1721 if (mo_graph->checkReachable(release, act) &&
1722 mo_graph->checkReachable(act, rf)) {
1723 /* release --mo-> act --mo--> rf */
1724 return true; /* complete */
1726 /* act may break release sequence */
1727 pending->writes.push_back(act);
1730 if (!future_ordered)
1731 certain = false; /* This thread is uncertain */
1735 release_heads->push_back(release);
1736 pending->writes.clear();
1738 pending->release = release;
1745 * A public interface for getting the release sequence head(s) with which a
1746 * given ModelAction must synchronize. This function only returns a non-empty
1747 * result when it can locate a release sequence head with certainty. Otherwise,
1748 * it may mark the internal state of the ModelChecker so that it will handle
1749 * the release sequence at a later time, causing @a act to update its
1750 * synchronization at some later point in execution.
1751 * @param act The 'acquire' action that may read from a release sequence
1752 * @param release_heads A pass-by-reference return parameter. Will be filled
1753 * with the head(s) of the release sequence(s), if they exists with certainty.
1754 * @see ModelChecker::release_seq_heads
1756 void ModelChecker::get_release_seq_heads(ModelAction *act, rel_heads_list_t *release_heads)
1758 const ModelAction *rf = act->get_reads_from();
1759 struct release_seq *sequence = (struct release_seq *)snapshot_calloc(1, sizeof(struct release_seq));
1760 sequence->acquire = act;
1762 if (!release_seq_heads(rf, release_heads, sequence)) {
1763 /* add act to 'lazy checking' list */
1764 pending_rel_seqs->push_back(sequence);
1766 snapshot_free(sequence);
1771 * Attempt to resolve all stashed operations that might synchronize with a
1772 * release sequence for a given location. This implements the "lazy" portion of
1773 * determining whether or not a release sequence was contiguous, since not all
1774 * modification order information is present at the time an action occurs.
1776 * @param location The location/object that should be checked for release
1777 * sequence resolutions. A NULL value means to check all locations.
1778 * @param work_queue The work queue to which to add work items as they are
1780 * @return True if any updates occurred (new synchronization, new mo_graph
1783 bool ModelChecker::resolve_release_sequences(void *location, work_queue_t *work_queue)
1785 bool updated = false;
1786 std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >::iterator it = pending_rel_seqs->begin();
1787 while (it != pending_rel_seqs->end()) {
1788 struct release_seq *pending = *it;
1789 ModelAction *act = pending->acquire;
1791 /* Only resolve sequences on the given location, if provided */
1792 if (location && act->get_location() != location) {
1797 const ModelAction *rf = act->get_reads_from();
1798 rel_heads_list_t release_heads;
1800 complete = release_seq_heads(rf, &release_heads, pending);
1801 for (unsigned int i = 0; i < release_heads.size(); i++) {
1802 if (!act->has_synchronized_with(release_heads[i])) {
1803 if (act->synchronize_with(release_heads[i]))
1806 set_bad_synchronization();
1811 /* Re-check all pending release sequences */
1812 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
1813 /* Re-check act for mo_graph edges */
1814 work_queue->push_back(MOEdgeWorkEntry(act));
1816 /* propagate synchronization to later actions */
1817 action_list_t::reverse_iterator rit = action_trace->rbegin();
1818 for (; (*rit) != act; rit++) {
1819 ModelAction *propagate = *rit;
1820 if (act->happens_before(propagate)) {
1821 propagate->synchronize_with(act);
1822 /* Re-check 'propagate' for mo_graph edges */
1823 work_queue->push_back(MOEdgeWorkEntry(propagate));
1828 it = pending_rel_seqs->erase(it);
1829 snapshot_free(pending);
1835 // If we resolved promises or data races, see if we have realized a data race.
1842 * Performs various bookkeeping operations for the current ModelAction. For
1843 * instance, adds action to the per-object, per-thread action vector and to the
1844 * action trace list of all thread actions.
1846 * @param act is the ModelAction to add.
1848 void ModelChecker::add_action_to_lists(ModelAction *act)
1850 int tid = id_to_int(act->get_tid());
1851 action_trace->push_back(act);
1853 get_safe_ptr_action(obj_map, act->get_location())->push_back(act);
1855 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, act->get_location());
1856 if (tid >= (int)vec->size())
1857 vec->resize(priv->next_thread_id);
1858 (*vec)[tid].push_back(act);
1860 if ((int)thrd_last_action->size() <= tid)
1861 thrd_last_action->resize(get_num_threads());
1862 (*thrd_last_action)[tid] = act;
1864 if (act->is_wait()) {
1865 void *mutex_loc=(void *) act->get_value();
1866 get_safe_ptr_action(obj_map, mutex_loc)->push_back(act);
1868 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, mutex_loc);
1869 if (tid >= (int)vec->size())
1870 vec->resize(priv->next_thread_id);
1871 (*vec)[tid].push_back(act);
1873 if ((int)thrd_last_action->size() <= tid)
1874 thrd_last_action->resize(get_num_threads());
1875 (*thrd_last_action)[tid] = act;
1880 * @brief Get the last action performed by a particular Thread
1881 * @param tid The thread ID of the Thread in question
1882 * @return The last action in the thread
1884 ModelAction * ModelChecker::get_last_action(thread_id_t tid) const
1886 int threadid = id_to_int(tid);
1887 if (threadid < (int)thrd_last_action->size())
1888 return (*thrd_last_action)[id_to_int(tid)];
1894 * Gets the last memory_order_seq_cst write (in the total global sequence)
1895 * performed on a particular object (i.e., memory location), not including the
1897 * @param curr The current ModelAction; also denotes the object location to
1899 * @return The last seq_cst write
1901 ModelAction * ModelChecker::get_last_seq_cst(ModelAction *curr) const
1903 void *location = curr->get_location();
1904 action_list_t *list = get_safe_ptr_action(obj_map, location);
1905 /* Find: max({i in dom(S) | seq_cst(t_i) && isWrite(t_i) && samevar(t_i, t)}) */
1906 action_list_t::reverse_iterator rit;
1907 for (rit = list->rbegin(); rit != list->rend(); rit++)
1908 if ((*rit)->is_write() && (*rit)->is_seqcst() && (*rit) != curr)
1914 * Gets the last unlock operation performed on a particular mutex (i.e., memory
1915 * location). This function identifies the mutex according to the current
1916 * action, which is presumed to perform on the same mutex.
1917 * @param curr The current ModelAction; also denotes the object location to
1919 * @return The last unlock operation
1921 ModelAction * ModelChecker::get_last_unlock(ModelAction *curr) const
1923 void *location = curr->get_location();
1924 action_list_t *list = get_safe_ptr_action(obj_map, location);
1925 /* Find: max({i in dom(S) | isUnlock(t_i) && samevar(t_i, t)}) */
1926 action_list_t::reverse_iterator rit;
1927 for (rit = list->rbegin(); rit != list->rend(); rit++)
1928 if ((*rit)->is_unlock() || (*rit)->is_wait())
1933 ModelAction * ModelChecker::get_parent_action(thread_id_t tid)
1935 ModelAction *parent = get_last_action(tid);
1937 parent = get_thread(tid)->get_creation();
1942 * Returns the clock vector for a given thread.
1943 * @param tid The thread whose clock vector we want
1944 * @return Desired clock vector
1946 ClockVector * ModelChecker::get_cv(thread_id_t tid)
1948 return get_parent_action(tid)->get_cv();
1952 * Resolve a set of Promises with a current write. The set is provided in the
1953 * Node corresponding to @a write.
1954 * @param write The ModelAction that is fulfilling Promises
1955 * @return True if promises were resolved; false otherwise
1957 bool ModelChecker::resolve_promises(ModelAction *write)
1959 bool resolved = false;
1960 std::vector< thread_id_t, ModelAlloc<thread_id_t> > threads_to_check;
1962 for (unsigned int i = 0, promise_index = 0; promise_index < promises->size(); i++) {
1963 Promise *promise = (*promises)[promise_index];
1964 if (write->get_node()->get_promise(i)) {
1965 ModelAction *read = promise->get_action();
1966 if (read->is_rmw()) {
1967 mo_graph->addRMWEdge(write, read);
1969 read->read_from(write);
1970 //First fix up the modification order for actions that happened
1972 r_modification_order(read, write);
1973 //Next fix up the modification order for actions that happened
1975 post_r_modification_order(read, write);
1976 //Make sure the promise's value matches the write's value
1977 ASSERT(promise->get_value() == write->get_value());
1980 promises->erase(promises->begin() + promise_index);
1981 threads_to_check.push_back(read->get_tid());
1988 //Check whether reading these writes has made threads unable to
1991 for(unsigned int i=0;i<threads_to_check.size();i++)
1992 mo_check_promises(threads_to_check[i], write);
1998 * Compute the set of promises that could potentially be satisfied by this
1999 * action. Note that the set computation actually appears in the Node, not in
2001 * @param curr The ModelAction that may satisfy promises
2003 void ModelChecker::compute_promises(ModelAction *curr)
2005 for (unsigned int i = 0; i < promises->size(); i++) {
2006 Promise *promise = (*promises)[i];
2007 const ModelAction *act = promise->get_action();
2008 if (!act->happens_before(curr) &&
2010 !act->could_synchronize_with(curr) &&
2011 !act->same_thread(curr) &&
2012 act->get_location() == curr->get_location() &&
2013 promise->get_value() == curr->get_value()) {
2014 curr->get_node()->set_promise(i, act->is_rmw());
2019 /** Checks promises in response to change in ClockVector Threads. */
2020 void ModelChecker::check_promises(thread_id_t tid, ClockVector *old_cv, ClockVector *merge_cv)
2022 for (unsigned int i = 0; i < promises->size(); i++) {
2023 Promise *promise = (*promises)[i];
2024 const ModelAction *act = promise->get_action();
2025 if ((old_cv == NULL || !old_cv->synchronized_since(act)) &&
2026 merge_cv->synchronized_since(act)) {
2027 if (promise->increment_threads(tid)) {
2028 //Promise has failed
2029 failed_promise = true;
2036 void ModelChecker::check_promises_thread_disabled() {
2037 for (unsigned int i = 0; i < promises->size(); i++) {
2038 Promise *promise = (*promises)[i];
2039 if (promise->check_promise()) {
2040 failed_promise = true;
2046 /** Checks promises in response to addition to modification order for threads.
2048 * pthread is the thread that performed the read that created the promise
2050 * pread is the read that created the promise
2052 * pwrite is either the first write to same location as pread by
2053 * pthread that is sequenced after pread or the value read by the
2054 * first read to the same lcoation as pread by pthread that is
2055 * sequenced after pread..
2057 * 1. If tid=pthread, then we check what other threads are reachable
2058 * through the mode order starting with pwrite. Those threads cannot
2059 * perform a write that will resolve the promise due to modification
2060 * order constraints.
2062 * 2. If the tid is not pthread, we check whether pwrite can reach the
2063 * action write through the modification order. If so, that thread
2064 * cannot perform a future write that will resolve the promise due to
2065 * modificatin order constraints.
2067 * @parem tid The thread that either read from the model action
2068 * write, or actually did the model action write.
2070 * @parem write The ModelAction representing the relevant write.
2073 void ModelChecker::mo_check_promises(thread_id_t tid, const ModelAction *write) {
2074 void * location = write->get_location();
2075 for (unsigned int i = 0; i < promises->size(); i++) {
2076 Promise *promise = (*promises)[i];
2077 const ModelAction *act = promise->get_action();
2079 //Is this promise on the same location?
2080 if ( act->get_location() != location )
2083 //same thread as the promise
2084 if ( act->get_tid()==tid ) {
2086 //do we have a pwrite for the promise, if not, set it
2087 if (promise->get_write() == NULL ) {
2088 promise->set_write(write);
2089 //The pwrite cannot happen before the promise
2090 if (write->happens_before(act) && (write != act)) {
2091 failed_promise = true;
2095 if (mo_graph->checkPromise(write, promise)) {
2096 failed_promise = true;
2101 //Don't do any lookups twice for the same thread
2102 if (promise->has_sync_thread(tid))
2105 if (promise->get_write()&&mo_graph->checkReachable(promise->get_write(), write)) {
2106 if (promise->increment_threads(tid)) {
2107 failed_promise = true;
2115 * Compute the set of writes that may break the current pending release
2116 * sequence. This information is extracted from previou release sequence
2119 * @param curr The current ModelAction. Must be a release sequence fixup
2122 void ModelChecker::compute_relseq_breakwrites(ModelAction *curr)
2124 if (pending_rel_seqs->empty())
2127 struct release_seq *pending = pending_rel_seqs->back();
2128 for (unsigned int i = 0; i < pending->writes.size(); i++) {
2129 const ModelAction *write = pending->writes[i];
2130 curr->get_node()->add_relseq_break(write);
2133 /* NULL means don't break the sequence; just synchronize */
2134 curr->get_node()->add_relseq_break(NULL);
2138 * Build up an initial set of all past writes that this 'read' action may read
2139 * from. This set is determined by the clock vector's "happens before"
2141 * @param curr is the current ModelAction that we are exploring; it must be a
2144 void ModelChecker::build_reads_from_past(ModelAction *curr)
2146 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
2148 ASSERT(curr->is_read());
2150 ModelAction *last_seq_cst = NULL;
2152 /* Track whether this object has been initialized */
2153 bool initialized = false;
2155 if (curr->is_seqcst()) {
2156 last_seq_cst = get_last_seq_cst(curr);
2157 /* We have to at least see the last sequentially consistent write,
2158 so we are initialized. */
2159 if (last_seq_cst != NULL)
2163 /* Iterate over all threads */
2164 for (i = 0; i < thrd_lists->size(); i++) {
2165 /* Iterate over actions in thread, starting from most recent */
2166 action_list_t *list = &(*thrd_lists)[i];
2167 action_list_t::reverse_iterator rit;
2168 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2169 ModelAction *act = *rit;
2171 /* Only consider 'write' actions */
2172 if (!act->is_write() || act == curr)
2175 /* Don't consider more than one seq_cst write if we are a seq_cst read. */
2176 if (!curr->is_seqcst() || (!act->is_seqcst() && (last_seq_cst == NULL || !act->happens_before(last_seq_cst))) || act == last_seq_cst) {
2177 if (!curr->get_sleep_flag() || curr->is_seqcst() || sleep_can_read_from(curr, act)) {
2178 DEBUG("Adding action to may_read_from:\n");
2179 if (DBG_ENABLED()) {
2183 curr->get_node()->add_read_from(act);
2187 /* Include at most one act per-thread that "happens before" curr */
2188 if (act->happens_before(curr)) {
2196 assert_bug("May read from uninitialized atomic");
2198 if (DBG_ENABLED() || !initialized) {
2199 printf("Reached read action:\n");
2201 printf("Printing may_read_from\n");
2202 curr->get_node()->print_may_read_from();
2203 printf("End printing may_read_from\n");
2207 bool ModelChecker::sleep_can_read_from(ModelAction * curr, const ModelAction *write) {
2209 Node *prevnode=write->get_node()->get_parent();
2211 bool thread_sleep=prevnode->enabled_status(curr->get_tid())==THREAD_SLEEP_SET;
2212 if (write->is_release()&&thread_sleep)
2214 if (!write->is_rmw()) {
2217 if (write->get_reads_from()==NULL)
2219 write=write->get_reads_from();
2223 static void print_list(action_list_t *list)
2225 action_list_t::iterator it;
2227 printf("---------------------------------------------------------------------\n");
2229 unsigned int hash=0;
2231 for (it = list->begin(); it != list->end(); it++) {
2233 hash=hash^(hash<<3)^((*it)->hash());
2235 printf("HASH %u\n", hash);
2236 printf("---------------------------------------------------------------------\n");
2239 #if SUPPORT_MOD_ORDER_DUMP
2240 void ModelChecker::dumpGraph(char *filename) {
2242 sprintf(buffer, "%s.dot",filename);
2243 FILE *file=fopen(buffer, "w");
2244 fprintf(file, "digraph %s {\n",filename);
2245 mo_graph->dumpNodes(file);
2246 ModelAction ** thread_array=(ModelAction **)model_calloc(1, sizeof(ModelAction *)*get_num_threads());
2248 for (action_list_t::iterator it = action_trace->begin(); it != action_trace->end(); it++) {
2249 ModelAction *action=*it;
2250 if (action->is_read()) {
2251 fprintf(file, "N%u [label=\"%u, T%u\"];\n", action->get_seq_number(),action->get_seq_number(), action->get_tid());
2252 if (action->get_reads_from()!=NULL)
2253 fprintf(file, "N%u -> N%u[label=\"rf\", color=red];\n", action->get_seq_number(), action->get_reads_from()->get_seq_number());
2255 if (thread_array[action->get_tid()] != NULL) {
2256 fprintf(file, "N%u -> N%u[label=\"sb\", color=blue];\n", thread_array[action->get_tid()]->get_seq_number(), action->get_seq_number());
2259 thread_array[action->get_tid()]=action;
2261 fprintf(file,"}\n");
2262 model_free(thread_array);
2267 void ModelChecker::print_summary()
2269 #if SUPPORT_MOD_ORDER_DUMP
2271 char buffername[100];
2272 sprintf(buffername, "exec%04u", stats.num_total);
2273 mo_graph->dumpGraphToFile(buffername);
2274 sprintf(buffername, "graph%04u", stats.num_total);
2275 dumpGraph(buffername);
2278 if (!isfinalfeasible())
2279 printf("INFEASIBLE EXECUTION!\n");
2280 print_list(action_trace);
2285 * Add a Thread to the system for the first time. Should only be called once
2287 * @param t The Thread to add
2289 void ModelChecker::add_thread(Thread *t)
2291 thread_map->put(id_to_int(t->get_id()), t);
2292 scheduler->add_thread(t);
2296 * Removes a thread from the scheduler.
2297 * @param the thread to remove.
2299 void ModelChecker::remove_thread(Thread *t)
2301 scheduler->remove_thread(t);
2305 * @brief Get a Thread reference by its ID
2306 * @param tid The Thread's ID
2307 * @return A Thread reference
2309 Thread * ModelChecker::get_thread(thread_id_t tid) const
2311 return thread_map->get(id_to_int(tid));
2315 * @brief Get a reference to the Thread in which a ModelAction was executed
2316 * @param act The ModelAction
2317 * @return A Thread reference
2319 Thread * ModelChecker::get_thread(ModelAction *act) const
2321 return get_thread(act->get_tid());
2325 * @brief Check if a Thread is currently enabled
2326 * @param t The Thread to check
2327 * @return True if the Thread is currently enabled
2329 bool ModelChecker::is_enabled(Thread *t) const
2331 return scheduler->is_enabled(t);
2335 * @brief Check if a Thread is currently enabled
2336 * @param tid The ID of the Thread to check
2337 * @return True if the Thread is currently enabled
2339 bool ModelChecker::is_enabled(thread_id_t tid) const
2341 return scheduler->is_enabled(tid);
2345 * Switch from a user-context to the "master thread" context (a.k.a. system
2346 * context). This switch is made with the intention of exploring a particular
2347 * model-checking action (described by a ModelAction object). Must be called
2348 * from a user-thread context.
2350 * @param act The current action that will be explored. May be NULL only if
2351 * trace is exiting via an assertion (see ModelChecker::set_assert and
2352 * ModelChecker::has_asserted).
2353 * @return Return status from the 'swap' call (i.e., success/fail, 0/-1)
2355 int ModelChecker::switch_to_master(ModelAction *act)
2358 Thread *old = thread_current();
2359 set_current_action(act);
2360 old->set_state(THREAD_READY);
2361 return Thread::swap(old, &system_context);
2365 * Takes the next step in the execution, if possible.
2366 * @return Returns true (success) if a step was taken and false otherwise.
2368 bool ModelChecker::take_step() {
2372 Thread *curr = priv->current_action ? get_thread(priv->current_action) : NULL;
2374 if (curr->get_state() == THREAD_READY) {
2375 ASSERT(priv->current_action);
2377 priv->nextThread = check_current_action(priv->current_action);
2378 priv->current_action = NULL;
2380 if (curr->is_blocked() || curr->is_complete())
2381 scheduler->remove_thread(curr);
2386 Thread *next = scheduler->next_thread(priv->nextThread);
2388 /* Infeasible -> don't take any more steps */
2391 else if (isfeasibleprefix() && have_bug_reports()) {
2396 if (params.bound != 0) {
2397 if (priv->used_sequence_numbers > params.bound) {
2402 DEBUG("(%d, %d)\n", curr ? id_to_int(curr->get_id()) : -1,
2403 next ? id_to_int(next->get_id()) : -1);
2406 * Launch end-of-execution release sequence fixups only when there are:
2408 * (1) no more user threads to run (or when execution replay chooses
2409 * the 'model_thread')
2410 * (2) pending release sequences
2411 * (3) pending assertions (i.e., data races)
2412 * (4) no pending promises
2414 if (!pending_rel_seqs->empty() && (!next || next->is_model_thread()) &&
2415 isfinalfeasible() && !unrealizedraces.empty()) {
2416 printf("*** WARNING: release sequence fixup action (%zu pending release seuqences) ***\n",
2417 pending_rel_seqs->size());
2418 ModelAction *fixup = new ModelAction(MODEL_FIXUP_RELSEQ,
2419 std::memory_order_seq_cst, NULL, VALUE_NONE,
2421 set_current_action(fixup);
2425 /* next == NULL -> don't take any more steps */
2429 next->set_state(THREAD_RUNNING);
2431 if (next->get_pending() != NULL) {
2432 /* restart a pending action */
2433 set_current_action(next->get_pending());
2434 next->set_pending(NULL);
2435 next->set_state(THREAD_READY);
2439 /* Return false only if swap fails with an error */
2440 return (Thread::swap(&system_context, next) == 0);
2443 /** Runs the current execution until threre are no more steps to take. */
2444 void ModelChecker::finish_execution() {
2447 while (take_step());