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 bug%s detected\n",
378 priv->bugs.size() > 1 ? "s" : "");
379 for (unsigned int i = 0; i < priv->bugs.size(); i++)
380 priv->bugs[i]->print();
385 * @brief Record end-of-execution stats
387 * Must be run when exiting an execution. Records various stats.
388 * @see struct execution_stats
390 void ModelChecker::record_stats()
393 if (!isfinalfeasible())
394 stats.num_infeasible++;
395 else if (have_bug_reports())
396 stats.num_buggy_executions++;
397 else if (is_complete_execution())
398 stats.num_complete++;
401 /** @brief Print execution stats */
402 void ModelChecker::print_stats() const
404 printf("Number of complete, bug-free executions: %d\n", stats.num_complete);
405 printf("Number of buggy executions: %d\n", stats.num_buggy_executions);
406 printf("Number of infeasible executions: %d\n", stats.num_infeasible);
407 printf("Total executions: %d\n", stats.num_total);
408 printf("Total nodes created: %d\n", node_stack->get_total_nodes());
412 * Queries the model-checker for more executions to explore and, if one
413 * exists, resets the model-checker state to execute a new execution.
415 * @return If there are more executions to explore, return true. Otherwise,
418 bool ModelChecker::next_execution()
424 if (isfinalfeasible() && (is_complete_execution() || have_bug_reports())) {
425 printf("Earliest divergence point since last feasible execution:\n");
426 if (earliest_diverge)
427 earliest_diverge->print();
429 printf("(Not set)\n");
431 earliest_diverge = NULL;
434 assert_bug("Deadlock detected");
441 } else if (DBG_ENABLED()) {
446 if ((diverge = get_next_backtrack()) == NULL)
450 printf("Next execution will diverge at:\n");
454 reset_to_initial_state();
458 ModelAction * ModelChecker::get_last_conflict(ModelAction *act)
460 switch (act->get_type()) {
464 /* linear search: from most recent to oldest */
465 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
466 action_list_t::reverse_iterator rit;
467 for (rit = list->rbegin(); rit != list->rend(); rit++) {
468 ModelAction *prev = *rit;
469 if (prev->could_synchronize_with(act))
475 case ATOMIC_TRYLOCK: {
476 /* linear search: from most recent to oldest */
477 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
478 action_list_t::reverse_iterator rit;
479 for (rit = list->rbegin(); rit != list->rend(); rit++) {
480 ModelAction *prev = *rit;
481 if (act->is_conflicting_lock(prev))
486 case ATOMIC_UNLOCK: {
487 /* linear search: from most recent to oldest */
488 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
489 action_list_t::reverse_iterator rit;
490 for (rit = list->rbegin(); rit != list->rend(); rit++) {
491 ModelAction *prev = *rit;
492 if (!act->same_thread(prev)&&prev->is_failed_trylock())
498 /* linear search: from most recent to oldest */
499 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
500 action_list_t::reverse_iterator rit;
501 for (rit = list->rbegin(); rit != list->rend(); rit++) {
502 ModelAction *prev = *rit;
503 if (!act->same_thread(prev)&&prev->is_failed_trylock())
505 if (!act->same_thread(prev)&&prev->is_notify())
511 case ATOMIC_NOTIFY_ALL:
512 case ATOMIC_NOTIFY_ONE: {
513 /* linear search: from most recent to oldest */
514 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
515 action_list_t::reverse_iterator rit;
516 for (rit = list->rbegin(); rit != list->rend(); rit++) {
517 ModelAction *prev = *rit;
518 if (!act->same_thread(prev)&&prev->is_wait())
529 /** This method finds backtracking points where we should try to
530 * reorder the parameter ModelAction against.
532 * @param the ModelAction to find backtracking points for.
534 void ModelChecker::set_backtracking(ModelAction *act)
536 Thread *t = get_thread(act);
537 ModelAction * prev = get_last_conflict(act);
541 Node * node = prev->get_node()->get_parent();
543 int low_tid, high_tid;
544 if (node->is_enabled(t)) {
545 low_tid = id_to_int(act->get_tid());
546 high_tid = low_tid+1;
549 high_tid = get_num_threads();
552 for(int i = low_tid; i < high_tid; i++) {
553 thread_id_t tid = int_to_id(i);
555 /* Make sure this thread can be enabled here. */
556 if (i >= node->get_num_threads())
559 /* Don't backtrack into a point where the thread is disabled or sleeping. */
560 if (node->enabled_status(tid)!=THREAD_ENABLED)
563 /* Check if this has been explored already */
564 if (node->has_been_explored(tid))
567 /* See if fairness allows */
568 if (model->params.fairwindow != 0 && !node->has_priority(tid)) {
570 for(int t=0;t<node->get_num_threads();t++) {
571 thread_id_t tother=int_to_id(t);
572 if (node->is_enabled(tother) && node->has_priority(tother)) {
580 /* Cache the latest backtracking point */
581 if (!priv->next_backtrack || *prev > *priv->next_backtrack)
582 priv->next_backtrack = prev;
584 /* If this is a new backtracking point, mark the tree */
585 if (!node->set_backtrack(tid))
587 DEBUG("Setting backtrack: conflict = %d, instead tid = %d\n",
588 id_to_int(prev->get_tid()),
589 id_to_int(t->get_id()));
598 * Returns last backtracking point. The model checker will explore a different
599 * path for this point in the next execution.
600 * @return The ModelAction at which the next execution should diverge.
602 ModelAction * ModelChecker::get_next_backtrack()
604 ModelAction *next = priv->next_backtrack;
605 priv->next_backtrack = NULL;
610 * Processes a read or rmw model action.
611 * @param curr is the read model action to process.
612 * @param second_part_of_rmw is boolean that is true is this is the second action of a rmw.
613 * @return True if processing this read updates the mo_graph.
615 bool ModelChecker::process_read(ModelAction *curr, bool second_part_of_rmw)
617 uint64_t value = VALUE_NONE;
618 bool updated = false;
620 const ModelAction *reads_from = curr->get_node()->get_read_from();
621 if (reads_from != NULL) {
622 mo_graph->startChanges();
624 value = reads_from->get_value();
625 bool r_status = false;
627 if (!second_part_of_rmw) {
628 check_recency(curr, reads_from);
629 r_status = r_modification_order(curr, reads_from);
633 if (!second_part_of_rmw&&!isfeasible()&&(curr->get_node()->increment_read_from()||curr->get_node()->increment_future_value())) {
634 mo_graph->rollbackChanges();
635 too_many_reads = false;
639 curr->read_from(reads_from);
640 mo_graph->commitChanges();
641 mo_check_promises(curr->get_tid(), reads_from);
644 } else if (!second_part_of_rmw) {
645 /* Read from future value */
646 value = curr->get_node()->get_future_value();
647 modelclock_t expiration = curr->get_node()->get_future_value_expiration();
648 curr->read_from(NULL);
649 Promise *valuepromise = new Promise(curr, value, expiration);
650 promises->push_back(valuepromise);
652 get_thread(curr)->set_return_value(value);
658 * Processes a lock, trylock, or unlock model action. @param curr is
659 * the read model action to process.
661 * The try lock operation checks whether the lock is taken. If not,
662 * it falls to the normal lock operation case. If so, it returns
665 * The lock operation has already been checked that it is enabled, so
666 * it just grabs the lock and synchronizes with the previous unlock.
668 * The unlock operation has to re-enable all of the threads that are
669 * waiting on the lock.
671 * @return True if synchronization was updated; false otherwise
673 bool ModelChecker::process_mutex(ModelAction *curr) {
674 std::mutex *mutex=NULL;
675 struct std::mutex_state *state=NULL;
677 if (curr->is_trylock() || curr->is_lock() || curr->is_unlock()) {
678 mutex = (std::mutex *)curr->get_location();
679 state = mutex->get_state();
680 } else if(curr->is_wait()) {
681 mutex = (std::mutex *)curr->get_value();
682 state = mutex->get_state();
685 switch (curr->get_type()) {
686 case ATOMIC_TRYLOCK: {
687 bool success = !state->islocked;
688 curr->set_try_lock(success);
690 get_thread(curr)->set_return_value(0);
693 get_thread(curr)->set_return_value(1);
695 //otherwise fall into the lock case
697 if (curr->get_cv()->getClock(state->alloc_tid) <= state->alloc_clock)
698 assert_bug("Lock access before initialization");
699 state->islocked = true;
700 ModelAction *unlock = get_last_unlock(curr);
701 //synchronize with the previous unlock statement
702 if (unlock != NULL) {
703 curr->synchronize_with(unlock);
708 case ATOMIC_UNLOCK: {
710 state->islocked = false;
711 //wake up the other threads
712 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, curr->get_location());
713 //activate all the waiting threads
714 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
715 scheduler->wake(get_thread(*rit));
722 state->islocked = false;
723 //wake up the other threads
724 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, (void *) curr->get_value());
725 //activate all the waiting threads
726 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
727 scheduler->wake(get_thread(*rit));
730 //check whether we should go to sleep or not...simulate spurious failures
731 if (curr->get_node()->get_misc()==0) {
732 get_safe_ptr_action(condvar_waiters_map, curr->get_location())->push_back(curr);
734 scheduler->sleep(get_current_thread());
738 case ATOMIC_NOTIFY_ALL: {
739 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
740 //activate all the waiting threads
741 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
742 scheduler->wake(get_thread(*rit));
747 case ATOMIC_NOTIFY_ONE: {
748 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
749 int wakeupthread=curr->get_node()->get_misc();
750 action_list_t::iterator it = waiters->begin();
751 advance(it, wakeupthread);
752 scheduler->wake(get_thread(*it));
764 * Process a write ModelAction
765 * @param curr The ModelAction to process
766 * @return True if the mo_graph was updated or promises were resolved
768 bool ModelChecker::process_write(ModelAction *curr)
770 bool updated_mod_order = w_modification_order(curr);
771 bool updated_promises = resolve_promises(curr);
773 if (promises->size() == 0) {
774 for (unsigned int i = 0; i < futurevalues->size(); i++) {
775 struct PendingFutureValue pfv = (*futurevalues)[i];
776 //Do more ambitious checks now that mo is more complete
777 if (mo_may_allow(pfv.writer, pfv.act)&&
778 pfv.act->get_node()->add_future_value(pfv.writer->get_value(), pfv.writer->get_seq_number()+params.maxfuturedelay) &&
779 (!priv->next_backtrack || *pfv.act > *priv->next_backtrack))
780 priv->next_backtrack = pfv.act;
782 futurevalues->resize(0);
785 mo_graph->commitChanges();
786 mo_check_promises(curr->get_tid(), curr);
788 get_thread(curr)->set_return_value(VALUE_NONE);
789 return updated_mod_order || updated_promises;
793 * @brief Process the current action for thread-related activity
795 * Performs current-action processing for a THREAD_* ModelAction. Proccesses
796 * may include setting Thread status, completing THREAD_FINISH/THREAD_JOIN
797 * synchronization, etc. This function is a no-op for non-THREAD actions
798 * (e.g., ATOMIC_{READ,WRITE,RMW,LOCK}, etc.)
800 * @param curr The current action
801 * @return True if synchronization was updated or a thread completed
803 bool ModelChecker::process_thread_action(ModelAction *curr)
805 bool updated = false;
807 switch (curr->get_type()) {
808 case THREAD_CREATE: {
809 Thread *th = (Thread *)curr->get_location();
810 th->set_creation(curr);
814 Thread *blocking = (Thread *)curr->get_location();
815 ModelAction *act = get_last_action(blocking->get_id());
816 curr->synchronize_with(act);
817 updated = true; /* trigger rel-seq checks */
820 case THREAD_FINISH: {
821 Thread *th = get_thread(curr);
822 while (!th->wait_list_empty()) {
823 ModelAction *act = th->pop_wait_list();
824 scheduler->wake(get_thread(act));
827 updated = true; /* trigger rel-seq checks */
831 check_promises(curr->get_tid(), NULL, curr->get_cv());
842 * @brief Process the current action for release sequence fixup activity
844 * Performs model-checker release sequence fixups for the current action,
845 * forcing a single pending release sequence to break (with a given, potential
846 * "loose" write) or to complete (i.e., synchronize). If a pending release
847 * sequence forms a complete release sequence, then we must perform the fixup
848 * synchronization, mo_graph additions, etc.
850 * @param curr The current action; must be a release sequence fixup action
851 * @param work_queue The work queue to which to add work items as they are
854 void ModelChecker::process_relseq_fixup(ModelAction *curr, work_queue_t *work_queue)
856 const ModelAction *write = curr->get_node()->get_relseq_break();
857 struct release_seq *sequence = pending_rel_seqs->back();
858 pending_rel_seqs->pop_back();
860 ModelAction *acquire = sequence->acquire;
861 const ModelAction *rf = sequence->rf;
862 const ModelAction *release = sequence->release;
866 ASSERT(release->same_thread(rf));
870 * @todo Forcing a synchronization requires that we set
871 * modification order constraints. For instance, we can't allow
872 * a fixup sequence in which two separate read-acquire
873 * operations read from the same sequence, where the first one
874 * synchronizes and the other doesn't. Essentially, we can't
875 * allow any writes to insert themselves between 'release' and
879 /* Must synchronize */
880 if (!acquire->synchronize_with(release)) {
881 set_bad_synchronization();
884 /* Re-check all pending release sequences */
885 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
886 /* Re-check act for mo_graph edges */
887 work_queue->push_back(MOEdgeWorkEntry(acquire));
889 /* propagate synchronization to later actions */
890 action_list_t::reverse_iterator rit = action_trace->rbegin();
891 for (; (*rit) != acquire; rit++) {
892 ModelAction *propagate = *rit;
893 if (acquire->happens_before(propagate)) {
894 propagate->synchronize_with(acquire);
895 /* Re-check 'propagate' for mo_graph edges */
896 work_queue->push_back(MOEdgeWorkEntry(propagate));
900 /* Break release sequence with new edges:
901 * release --mo--> write --mo--> rf */
902 mo_graph->addEdge(release, write);
903 mo_graph->addEdge(write, rf);
906 /* See if we have realized a data race */
911 * Initialize the current action by performing one or more of the following
912 * actions, as appropriate: merging RMWR and RMWC/RMW actions, stepping forward
913 * in the NodeStack, manipulating backtracking sets, allocating and
914 * initializing clock vectors, and computing the promises to fulfill.
916 * @param curr The current action, as passed from the user context; may be
917 * freed/invalidated after the execution of this function, with a different
918 * action "returned" its place (pass-by-reference)
919 * @return True if curr is a newly-explored action; false otherwise
921 bool ModelChecker::initialize_curr_action(ModelAction **curr)
923 ModelAction *newcurr;
925 if ((*curr)->is_rmwc() || (*curr)->is_rmw()) {
926 newcurr = process_rmw(*curr);
929 if (newcurr->is_rmw())
930 compute_promises(newcurr);
936 (*curr)->set_seq_number(get_next_seq_num());
938 newcurr = node_stack->explore_action(*curr, scheduler->get_enabled());
940 /* First restore type and order in case of RMW operation */
941 if ((*curr)->is_rmwr())
942 newcurr->copy_typeandorder(*curr);
944 ASSERT((*curr)->get_location() == newcurr->get_location());
945 newcurr->copy_from_new(*curr);
947 /* Discard duplicate ModelAction; use action from NodeStack */
950 /* Always compute new clock vector */
951 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
954 return false; /* Action was explored previously */
958 /* Always compute new clock vector */
959 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
961 * Perform one-time actions when pushing new ModelAction onto
964 if (newcurr->is_write())
965 compute_promises(newcurr);
966 else if (newcurr->is_relseq_fixup())
967 compute_relseq_breakwrites(newcurr);
968 else if (newcurr->is_wait())
969 newcurr->get_node()->set_misc_max(2);
970 else if (newcurr->is_notify_one()) {
971 newcurr->get_node()->set_misc_max(get_safe_ptr_action(condvar_waiters_map, newcurr->get_location())->size());
973 return true; /* This was a new ModelAction */
978 * @brief Check whether a model action is enabled.
980 * Checks whether a lock or join operation would be successful (i.e., is the
981 * lock already locked, or is the joined thread already complete). If not, put
982 * the action in a waiter list.
984 * @param curr is the ModelAction to check whether it is enabled.
985 * @return a bool that indicates whether the action is enabled.
987 bool ModelChecker::check_action_enabled(ModelAction *curr) {
988 if (curr->is_lock()) {
989 std::mutex * lock = (std::mutex *)curr->get_location();
990 struct std::mutex_state * state = lock->get_state();
991 if (state->islocked) {
992 //Stick the action in the appropriate waiting queue
993 get_safe_ptr_action(lock_waiters_map, curr->get_location())->push_back(curr);
996 } else if (curr->get_type() == THREAD_JOIN) {
997 Thread *blocking = (Thread *)curr->get_location();
998 if (!blocking->is_complete()) {
999 blocking->push_wait_list(curr);
1008 * Stores the ModelAction for the current thread action. Call this
1009 * immediately before switching from user- to system-context to pass
1010 * data between them.
1011 * @param act The ModelAction created by the user-thread action
1013 void ModelChecker::set_current_action(ModelAction *act) {
1014 priv->current_action = act;
1018 * This is the heart of the model checker routine. It performs model-checking
1019 * actions corresponding to a given "current action." Among other processes, it
1020 * calculates reads-from relationships, updates synchronization clock vectors,
1021 * forms a memory_order constraints graph, and handles replay/backtrack
1022 * execution when running permutations of previously-observed executions.
1024 * @param curr The current action to process
1025 * @return The next Thread that must be executed. May be NULL if ModelChecker
1026 * makes no choice (e.g., according to replay execution, combining RMW actions,
1029 Thread * ModelChecker::check_current_action(ModelAction *curr)
1032 bool second_part_of_rmw = curr->is_rmwc() || curr->is_rmw();
1034 if (!check_action_enabled(curr)) {
1035 /* Make the execution look like we chose to run this action
1036 * much later, when a lock/join can succeed */
1037 get_current_thread()->set_pending(curr);
1038 scheduler->sleep(get_current_thread());
1039 return get_next_thread(NULL);
1042 bool newly_explored = initialize_curr_action(&curr);
1044 wake_up_sleeping_actions(curr);
1046 /* Add the action to lists before any other model-checking tasks */
1047 if (!second_part_of_rmw)
1048 add_action_to_lists(curr);
1050 /* Build may_read_from set for newly-created actions */
1051 if (newly_explored && curr->is_read())
1052 build_reads_from_past(curr);
1054 /* Initialize work_queue with the "current action" work */
1055 work_queue_t work_queue(1, CheckCurrWorkEntry(curr));
1056 while (!work_queue.empty() && !has_asserted()) {
1057 WorkQueueEntry work = work_queue.front();
1058 work_queue.pop_front();
1060 switch (work.type) {
1061 case WORK_CHECK_CURR_ACTION: {
1062 ModelAction *act = work.action;
1063 bool update = false; /* update this location's release seq's */
1064 bool update_all = false; /* update all release seq's */
1066 if (process_thread_action(curr))
1069 if (act->is_read() && process_read(act, second_part_of_rmw))
1072 if (act->is_write() && process_write(act))
1075 if (act->is_mutex_op() && process_mutex(act))
1078 if (act->is_relseq_fixup())
1079 process_relseq_fixup(curr, &work_queue);
1082 work_queue.push_back(CheckRelSeqWorkEntry(NULL));
1084 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1087 case WORK_CHECK_RELEASE_SEQ:
1088 resolve_release_sequences(work.location, &work_queue);
1090 case WORK_CHECK_MO_EDGES: {
1091 /** @todo Complete verification of work_queue */
1092 ModelAction *act = work.action;
1093 bool updated = false;
1095 if (act->is_read()) {
1096 const ModelAction *rf = act->get_reads_from();
1097 if (rf != NULL && r_modification_order(act, rf))
1100 if (act->is_write()) {
1101 if (w_modification_order(act))
1104 mo_graph->commitChanges();
1107 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1116 check_curr_backtracking(curr);
1117 set_backtracking(curr);
1118 return get_next_thread(curr);
1121 void ModelChecker::check_curr_backtracking(ModelAction * curr) {
1122 Node *currnode = curr->get_node();
1123 Node *parnode = currnode->get_parent();
1125 if ((!parnode->backtrack_empty() ||
1126 !currnode->misc_empty() ||
1127 !currnode->read_from_empty() ||
1128 !currnode->future_value_empty() ||
1129 !currnode->promise_empty() ||
1130 !currnode->relseq_break_empty())
1131 && (!priv->next_backtrack ||
1132 *curr > *priv->next_backtrack)) {
1133 priv->next_backtrack = curr;
1137 bool ModelChecker::promises_expired() const
1139 for (unsigned int promise_index = 0; promise_index < promises->size(); promise_index++) {
1140 Promise *promise = (*promises)[promise_index];
1141 if (promise->get_expiration()<priv->used_sequence_numbers) {
1148 /** @return whether the current partial trace must be a prefix of a
1149 * feasible trace. */
1150 bool ModelChecker::isfeasibleprefix() const
1152 return promises->size() == 0 && pending_rel_seqs->size() == 0 && isfeasible();
1155 /** @return whether the current partial trace is feasible. */
1156 bool ModelChecker::isfeasible() const
1158 if (DBG_ENABLED() && mo_graph->checkForRMWViolation())
1159 DEBUG("Infeasible: RMW violation\n");
1161 return !mo_graph->checkForRMWViolation() && isfeasibleotherthanRMW();
1164 /** @return whether the current partial trace is feasible other than
1165 * multiple RMW reading from the same store. */
1166 bool ModelChecker::isfeasibleotherthanRMW() const
1168 if (DBG_ENABLED()) {
1169 if (mo_graph->checkForCycles())
1170 DEBUG("Infeasible: modification order cycles\n");
1172 DEBUG("Infeasible: failed promise\n");
1174 DEBUG("Infeasible: too many reads\n");
1175 if (bad_synchronization)
1176 DEBUG("Infeasible: bad synchronization ordering\n");
1177 if (promises_expired())
1178 DEBUG("Infeasible: promises expired\n");
1180 return !mo_graph->checkForCycles() && !failed_promise && !too_many_reads && !bad_synchronization && !promises_expired();
1183 /** Returns whether the current completed trace is feasible. */
1184 bool ModelChecker::isfinalfeasible() const
1186 if (DBG_ENABLED() && promises->size() != 0)
1187 DEBUG("Infeasible: unrevolved promises\n");
1189 return isfeasible() && promises->size() == 0;
1192 /** Close out a RMWR by converting previous RMWR into a RMW or READ. */
1193 ModelAction * ModelChecker::process_rmw(ModelAction *act) {
1194 ModelAction *lastread = get_last_action(act->get_tid());
1195 lastread->process_rmw(act);
1196 if (act->is_rmw() && lastread->get_reads_from()!=NULL) {
1197 mo_graph->addRMWEdge(lastread->get_reads_from(), lastread);
1198 mo_graph->commitChanges();
1204 * Checks whether a thread has read from the same write for too many times
1205 * without seeing the effects of a later write.
1208 * 1) there must a different write that we could read from that would satisfy the modification order,
1209 * 2) we must have read from the same value in excess of maxreads times, and
1210 * 3) that other write must have been in the reads_from set for maxreads times.
1212 * If so, we decide that the execution is no longer feasible.
1214 void ModelChecker::check_recency(ModelAction *curr, const ModelAction *rf) {
1215 if (params.maxreads != 0) {
1217 if (curr->get_node()->get_read_from_size() <= 1)
1219 //Must make sure that execution is currently feasible... We could
1220 //accidentally clear by rolling back
1223 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1224 int tid = id_to_int(curr->get_tid());
1227 if ((int)thrd_lists->size() <= tid)
1229 action_list_t *list = &(*thrd_lists)[tid];
1231 action_list_t::reverse_iterator rit = list->rbegin();
1232 /* Skip past curr */
1233 for (; (*rit) != curr; rit++)
1235 /* go past curr now */
1238 action_list_t::reverse_iterator ritcopy = rit;
1239 //See if we have enough reads from the same value
1241 for (; count < params.maxreads; rit++,count++) {
1242 if (rit==list->rend())
1244 ModelAction *act = *rit;
1245 if (!act->is_read())
1248 if (act->get_reads_from() != rf)
1250 if (act->get_node()->get_read_from_size() <= 1)
1253 for (int i = 0; i<curr->get_node()->get_read_from_size(); i++) {
1255 const ModelAction * write = curr->get_node()->get_read_from_at(i);
1257 //Need a different write
1261 /* Test to see whether this is a feasible write to read from*/
1262 mo_graph->startChanges();
1263 r_modification_order(curr, write);
1264 bool feasiblereadfrom = isfeasible();
1265 mo_graph->rollbackChanges();
1267 if (!feasiblereadfrom)
1271 bool feasiblewrite = true;
1272 //new we need to see if this write works for everyone
1274 for (int loop = count; loop>0; loop--,rit++) {
1275 ModelAction *act=*rit;
1276 bool foundvalue = false;
1277 for (int j = 0; j<act->get_node()->get_read_from_size(); j++) {
1278 if (act->get_node()->get_read_from_at(j)==write) {
1284 feasiblewrite = false;
1288 if (feasiblewrite) {
1289 too_many_reads = true;
1297 * Updates the mo_graph with the constraints imposed from the current
1300 * Basic idea is the following: Go through each other thread and find
1301 * the lastest action that happened before our read. Two cases:
1303 * (1) The action is a write => that write must either occur before
1304 * the write we read from or be the write we read from.
1306 * (2) The action is a read => the write that that action read from
1307 * must occur before the write we read from or be the same write.
1309 * @param curr The current action. Must be a read.
1310 * @param rf The action that curr reads from. Must be a write.
1311 * @return True if modification order edges were added; false otherwise
1313 bool ModelChecker::r_modification_order(ModelAction *curr, const ModelAction *rf)
1315 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1318 ASSERT(curr->is_read());
1320 /* Iterate over all threads */
1321 for (i = 0; i < thrd_lists->size(); i++) {
1322 /* Iterate over actions in thread, starting from most recent */
1323 action_list_t *list = &(*thrd_lists)[i];
1324 action_list_t::reverse_iterator rit;
1325 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1326 ModelAction *act = *rit;
1329 * Include at most one act per-thread that "happens
1330 * before" curr. Don't consider reflexively.
1332 if (act->happens_before(curr) && act != curr) {
1333 if (act->is_write()) {
1335 mo_graph->addEdge(act, rf);
1339 const ModelAction *prevreadfrom = act->get_reads_from();
1340 //if the previous read is unresolved, keep going...
1341 if (prevreadfrom == NULL)
1344 if (rf != prevreadfrom) {
1345 mo_graph->addEdge(prevreadfrom, rf);
1357 /** This method fixes up the modification order when we resolve a
1358 * promises. The basic problem is that actions that occur after the
1359 * read curr could not property add items to the modification order
1362 * So for each thread, we find the earliest item that happens after
1363 * the read curr. This is the item we have to fix up with additional
1364 * constraints. If that action is write, we add a MO edge between
1365 * the Action rf and that action. If the action is a read, we add a
1366 * MO edge between the Action rf, and whatever the read accessed.
1368 * @param curr is the read ModelAction that we are fixing up MO edges for.
1369 * @param rf is the write ModelAction that curr reads from.
1372 void ModelChecker::post_r_modification_order(ModelAction *curr, const ModelAction *rf)
1374 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1376 ASSERT(curr->is_read());
1378 /* Iterate over all threads */
1379 for (i = 0; i < thrd_lists->size(); i++) {
1380 /* Iterate over actions in thread, starting from most recent */
1381 action_list_t *list = &(*thrd_lists)[i];
1382 action_list_t::reverse_iterator rit;
1383 ModelAction *lastact = NULL;
1385 /* Find last action that happens after curr that is either not curr or a rmw */
1386 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1387 ModelAction *act = *rit;
1388 if (curr->happens_before(act) && (curr != act || curr->is_rmw())) {
1394 /* Include at most one act per-thread that "happens before" curr */
1395 if (lastact != NULL) {
1396 if (lastact==curr) {
1397 //Case 1: The resolved read is a RMW, and we need to make sure
1398 //that the write portion of the RMW mod order after rf
1400 mo_graph->addEdge(rf, lastact);
1401 } else if (lastact->is_read()) {
1402 //Case 2: The resolved read is a normal read and the next
1403 //operation is a read, and we need to make sure the value read
1404 //is mod ordered after rf
1406 const ModelAction *postreadfrom = lastact->get_reads_from();
1407 if (postreadfrom != NULL&&rf != postreadfrom)
1408 mo_graph->addEdge(rf, postreadfrom);
1410 //Case 3: The resolved read is a normal read and the next
1411 //operation is a write, and we need to make sure that the
1412 //write is mod ordered after rf
1414 mo_graph->addEdge(rf, lastact);
1422 * Updates the mo_graph with the constraints imposed from the current write.
1424 * Basic idea is the following: Go through each other thread and find
1425 * the lastest action that happened before our write. Two cases:
1427 * (1) The action is a write => that write must occur before
1430 * (2) The action is a read => the write that that action read from
1431 * must occur before the current write.
1433 * This method also handles two other issues:
1435 * (I) Sequential Consistency: Making sure that if the current write is
1436 * seq_cst, that it occurs after the previous seq_cst write.
1438 * (II) Sending the write back to non-synchronizing reads.
1440 * @param curr The current action. Must be a write.
1441 * @return True if modification order edges were added; false otherwise
1443 bool ModelChecker::w_modification_order(ModelAction *curr)
1445 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1448 ASSERT(curr->is_write());
1450 if (curr->is_seqcst()) {
1451 /* We have to at least see the last sequentially consistent write,
1452 so we are initialized. */
1453 ModelAction *last_seq_cst = get_last_seq_cst(curr);
1454 if (last_seq_cst != NULL) {
1455 mo_graph->addEdge(last_seq_cst, curr);
1460 /* Iterate over all threads */
1461 for (i = 0; i < thrd_lists->size(); i++) {
1462 /* Iterate over actions in thread, starting from most recent */
1463 action_list_t *list = &(*thrd_lists)[i];
1464 action_list_t::reverse_iterator rit;
1465 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1466 ModelAction *act = *rit;
1469 * 1) If RMW and it actually read from something, then we
1470 * already have all relevant edges, so just skip to next
1473 * 2) If RMW and it didn't read from anything, we should
1474 * whatever edge we can get to speed up convergence.
1476 * 3) If normal write, we need to look at earlier actions, so
1477 * continue processing list.
1479 if (curr->is_rmw()) {
1480 if (curr->get_reads_from()!=NULL)
1489 * Include at most one act per-thread that "happens
1492 if (act->happens_before(curr)) {
1494 * Note: if act is RMW, just add edge:
1496 * The following edge should be handled elsewhere:
1497 * readfrom(act) --mo--> act
1499 if (act->is_write())
1500 mo_graph->addEdge(act, curr);
1501 else if (act->is_read()) {
1502 //if previous read accessed a null, just keep going
1503 if (act->get_reads_from() == NULL)
1505 mo_graph->addEdge(act->get_reads_from(), curr);
1509 } else if (act->is_read() && !act->could_synchronize_with(curr) &&
1510 !act->same_thread(curr)) {
1511 /* We have an action that:
1512 (1) did not happen before us
1513 (2) is a read and we are a write
1514 (3) cannot synchronize with us
1515 (4) is in a different thread
1517 that read could potentially read from our write. Note that
1518 these checks are overly conservative at this point, we'll
1519 do more checks before actually removing the
1523 if (thin_air_constraint_may_allow(curr, act)) {
1525 (curr->is_rmw() && act->is_rmw() && curr->get_reads_from() == act->get_reads_from() && isfeasibleotherthanRMW())) {
1526 struct PendingFutureValue pfv = {curr,act};
1527 futurevalues->push_back(pfv);
1537 /** Arbitrary reads from the future are not allowed. Section 29.3
1538 * part 9 places some constraints. This method checks one result of constraint
1539 * constraint. Others require compiler support. */
1540 bool ModelChecker::thin_air_constraint_may_allow(const ModelAction * writer, const ModelAction *reader) {
1541 if (!writer->is_rmw())
1544 if (!reader->is_rmw())
1547 for (const ModelAction *search = writer->get_reads_from(); search != NULL; search = search->get_reads_from()) {
1548 if (search == reader)
1550 if (search->get_tid() == reader->get_tid() &&
1551 search->happens_before(reader))
1559 * Arbitrary reads from the future are not allowed. Section 29.3 part 9 places
1560 * some constraints. This method checks one the following constraint (others
1561 * require compiler support):
1563 * If X --hb-> Y --mo-> Z, then X should not read from Z.
1565 bool ModelChecker::mo_may_allow(const ModelAction *writer, const ModelAction *reader)
1567 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, reader->get_location());
1569 /* Iterate over all threads */
1570 for (i = 0; i < thrd_lists->size(); i++) {
1571 const ModelAction *write_after_read = NULL;
1573 /* Iterate over actions in thread, starting from most recent */
1574 action_list_t *list = &(*thrd_lists)[i];
1575 action_list_t::reverse_iterator rit;
1576 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1577 ModelAction *act = *rit;
1579 if (!reader->happens_before(act))
1581 else if (act->is_write())
1582 write_after_read = act;
1583 else if (act->is_read() && act->get_reads_from() != NULL && act != reader) {
1584 write_after_read = act->get_reads_from();
1588 if (write_after_read && write_after_read!=writer && mo_graph->checkReachable(write_after_read, writer))
1595 * Finds the head(s) of the release sequence(s) containing a given ModelAction.
1596 * The ModelAction under consideration is expected to be taking part in
1597 * release/acquire synchronization as an object of the "reads from" relation.
1598 * Note that this can only provide release sequence support for RMW chains
1599 * which do not read from the future, as those actions cannot be traced until
1600 * their "promise" is fulfilled. Similarly, we may not even establish the
1601 * presence of a release sequence with certainty, as some modification order
1602 * constraints may be decided further in the future. Thus, this function
1603 * "returns" two pieces of data: a pass-by-reference vector of @a release_heads
1604 * and a boolean representing certainty.
1606 * @param rf The action that might be part of a release sequence. Must be a
1608 * @param release_heads A pass-by-reference style return parameter. After
1609 * execution of this function, release_heads will contain the heads of all the
1610 * relevant release sequences, if any exists with certainty
1611 * @param pending A pass-by-reference style return parameter which is only used
1612 * when returning false (i.e., uncertain). Returns most information regarding
1613 * an uncertain release sequence, including any write operations that might
1614 * break the sequence.
1615 * @return true, if the ModelChecker is certain that release_heads is complete;
1618 bool ModelChecker::release_seq_heads(const ModelAction *rf,
1619 rel_heads_list_t *release_heads,
1620 struct release_seq *pending) const
1622 /* Only check for release sequences if there are no cycles */
1623 if (mo_graph->checkForCycles())
1627 ASSERT(rf->is_write());
1629 if (rf->is_release())
1630 release_heads->push_back(rf);
1632 break; /* End of RMW chain */
1634 /** @todo Need to be smarter here... In the linux lock
1635 * example, this will run to the beginning of the program for
1637 /** @todo The way to be smarter here is to keep going until 1
1638 * thread has a release preceded by an acquire and you've seen
1641 /* acq_rel RMW is a sufficient stopping condition */
1642 if (rf->is_acquire() && rf->is_release())
1643 return true; /* complete */
1645 rf = rf->get_reads_from();
1648 /* read from future: need to settle this later */
1650 return false; /* incomplete */
1653 if (rf->is_release())
1654 return true; /* complete */
1656 /* else relaxed write; check modification order for contiguous subsequence
1657 * -> rf must be same thread as release */
1658 int tid = id_to_int(rf->get_tid());
1659 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, rf->get_location());
1660 action_list_t *list = &(*thrd_lists)[tid];
1661 action_list_t::const_reverse_iterator rit;
1663 /* Find rf in the thread list */
1664 rit = std::find(list->rbegin(), list->rend(), rf);
1665 ASSERT(rit != list->rend());
1667 /* Find the last write/release */
1668 for (; rit != list->rend(); rit++)
1669 if ((*rit)->is_release())
1671 if (rit == list->rend()) {
1672 /* No write-release in this thread */
1673 return true; /* complete */
1675 ModelAction *release = *rit;
1677 ASSERT(rf->same_thread(release));
1679 pending->writes.clear();
1681 bool certain = true;
1682 for (unsigned int i = 0; i < thrd_lists->size(); i++) {
1683 if (id_to_int(rf->get_tid()) == (int)i)
1685 list = &(*thrd_lists)[i];
1687 /* Can we ensure no future writes from this thread may break
1688 * the release seq? */
1689 bool future_ordered = false;
1691 ModelAction *last = get_last_action(int_to_id(i));
1692 Thread *th = get_thread(int_to_id(i));
1693 if ((last && rf->happens_before(last)) ||
1696 future_ordered = true;
1698 ASSERT(!th->is_model_thread() || future_ordered);
1700 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1701 const ModelAction *act = *rit;
1702 /* Reach synchronization -> this thread is complete */
1703 if (act->happens_before(release))
1705 if (rf->happens_before(act)) {
1706 future_ordered = true;
1710 /* Only non-RMW writes can break release sequences */
1711 if (!act->is_write() || act->is_rmw())
1714 /* Check modification order */
1715 if (mo_graph->checkReachable(rf, act)) {
1716 /* rf --mo--> act */
1717 future_ordered = true;
1720 if (mo_graph->checkReachable(act, release))
1721 /* act --mo--> release */
1723 if (mo_graph->checkReachable(release, act) &&
1724 mo_graph->checkReachable(act, rf)) {
1725 /* release --mo-> act --mo--> rf */
1726 return true; /* complete */
1728 /* act may break release sequence */
1729 pending->writes.push_back(act);
1732 if (!future_ordered)
1733 certain = false; /* This thread is uncertain */
1737 release_heads->push_back(release);
1738 pending->writes.clear();
1740 pending->release = release;
1747 * A public interface for getting the release sequence head(s) with which a
1748 * given ModelAction must synchronize. This function only returns a non-empty
1749 * result when it can locate a release sequence head with certainty. Otherwise,
1750 * it may mark the internal state of the ModelChecker so that it will handle
1751 * the release sequence at a later time, causing @a act to update its
1752 * synchronization at some later point in execution.
1753 * @param act The 'acquire' action that may read from a release sequence
1754 * @param release_heads A pass-by-reference return parameter. Will be filled
1755 * with the head(s) of the release sequence(s), if they exists with certainty.
1756 * @see ModelChecker::release_seq_heads
1758 void ModelChecker::get_release_seq_heads(ModelAction *act, rel_heads_list_t *release_heads)
1760 const ModelAction *rf = act->get_reads_from();
1761 struct release_seq *sequence = (struct release_seq *)snapshot_calloc(1, sizeof(struct release_seq));
1762 sequence->acquire = act;
1764 if (!release_seq_heads(rf, release_heads, sequence)) {
1765 /* add act to 'lazy checking' list */
1766 pending_rel_seqs->push_back(sequence);
1768 snapshot_free(sequence);
1773 * Attempt to resolve all stashed operations that might synchronize with a
1774 * release sequence for a given location. This implements the "lazy" portion of
1775 * determining whether or not a release sequence was contiguous, since not all
1776 * modification order information is present at the time an action occurs.
1778 * @param location The location/object that should be checked for release
1779 * sequence resolutions. A NULL value means to check all locations.
1780 * @param work_queue The work queue to which to add work items as they are
1782 * @return True if any updates occurred (new synchronization, new mo_graph
1785 bool ModelChecker::resolve_release_sequences(void *location, work_queue_t *work_queue)
1787 bool updated = false;
1788 std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >::iterator it = pending_rel_seqs->begin();
1789 while (it != pending_rel_seqs->end()) {
1790 struct release_seq *pending = *it;
1791 ModelAction *act = pending->acquire;
1793 /* Only resolve sequences on the given location, if provided */
1794 if (location && act->get_location() != location) {
1799 const ModelAction *rf = act->get_reads_from();
1800 rel_heads_list_t release_heads;
1802 complete = release_seq_heads(rf, &release_heads, pending);
1803 for (unsigned int i = 0; i < release_heads.size(); i++) {
1804 if (!act->has_synchronized_with(release_heads[i])) {
1805 if (act->synchronize_with(release_heads[i]))
1808 set_bad_synchronization();
1813 /* Re-check all pending release sequences */
1814 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
1815 /* Re-check act for mo_graph edges */
1816 work_queue->push_back(MOEdgeWorkEntry(act));
1818 /* propagate synchronization to later actions */
1819 action_list_t::reverse_iterator rit = action_trace->rbegin();
1820 for (; (*rit) != act; rit++) {
1821 ModelAction *propagate = *rit;
1822 if (act->happens_before(propagate)) {
1823 propagate->synchronize_with(act);
1824 /* Re-check 'propagate' for mo_graph edges */
1825 work_queue->push_back(MOEdgeWorkEntry(propagate));
1830 it = pending_rel_seqs->erase(it);
1831 snapshot_free(pending);
1837 // If we resolved promises or data races, see if we have realized a data race.
1844 * Performs various bookkeeping operations for the current ModelAction. For
1845 * instance, adds action to the per-object, per-thread action vector and to the
1846 * action trace list of all thread actions.
1848 * @param act is the ModelAction to add.
1850 void ModelChecker::add_action_to_lists(ModelAction *act)
1852 int tid = id_to_int(act->get_tid());
1853 action_trace->push_back(act);
1855 get_safe_ptr_action(obj_map, act->get_location())->push_back(act);
1857 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, act->get_location());
1858 if (tid >= (int)vec->size())
1859 vec->resize(priv->next_thread_id);
1860 (*vec)[tid].push_back(act);
1862 if ((int)thrd_last_action->size() <= tid)
1863 thrd_last_action->resize(get_num_threads());
1864 (*thrd_last_action)[tid] = act;
1866 if (act->is_wait()) {
1867 void *mutex_loc=(void *) act->get_value();
1868 get_safe_ptr_action(obj_map, mutex_loc)->push_back(act);
1870 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, mutex_loc);
1871 if (tid >= (int)vec->size())
1872 vec->resize(priv->next_thread_id);
1873 (*vec)[tid].push_back(act);
1875 if ((int)thrd_last_action->size() <= tid)
1876 thrd_last_action->resize(get_num_threads());
1877 (*thrd_last_action)[tid] = act;
1882 * @brief Get the last action performed by a particular Thread
1883 * @param tid The thread ID of the Thread in question
1884 * @return The last action in the thread
1886 ModelAction * ModelChecker::get_last_action(thread_id_t tid) const
1888 int threadid = id_to_int(tid);
1889 if (threadid < (int)thrd_last_action->size())
1890 return (*thrd_last_action)[id_to_int(tid)];
1896 * Gets the last memory_order_seq_cst write (in the total global sequence)
1897 * performed on a particular object (i.e., memory location), not including the
1899 * @param curr The current ModelAction; also denotes the object location to
1901 * @return The last seq_cst write
1903 ModelAction * ModelChecker::get_last_seq_cst(ModelAction *curr) const
1905 void *location = curr->get_location();
1906 action_list_t *list = get_safe_ptr_action(obj_map, location);
1907 /* Find: max({i in dom(S) | seq_cst(t_i) && isWrite(t_i) && samevar(t_i, t)}) */
1908 action_list_t::reverse_iterator rit;
1909 for (rit = list->rbegin(); rit != list->rend(); rit++)
1910 if ((*rit)->is_write() && (*rit)->is_seqcst() && (*rit) != curr)
1916 * Gets the last unlock operation performed on a particular mutex (i.e., memory
1917 * location). This function identifies the mutex according to the current
1918 * action, which is presumed to perform on the same mutex.
1919 * @param curr The current ModelAction; also denotes the object location to
1921 * @return The last unlock operation
1923 ModelAction * ModelChecker::get_last_unlock(ModelAction *curr) const
1925 void *location = curr->get_location();
1926 action_list_t *list = get_safe_ptr_action(obj_map, location);
1927 /* Find: max({i in dom(S) | isUnlock(t_i) && samevar(t_i, t)}) */
1928 action_list_t::reverse_iterator rit;
1929 for (rit = list->rbegin(); rit != list->rend(); rit++)
1930 if ((*rit)->is_unlock() || (*rit)->is_wait())
1935 ModelAction * ModelChecker::get_parent_action(thread_id_t tid)
1937 ModelAction *parent = get_last_action(tid);
1939 parent = get_thread(tid)->get_creation();
1944 * Returns the clock vector for a given thread.
1945 * @param tid The thread whose clock vector we want
1946 * @return Desired clock vector
1948 ClockVector * ModelChecker::get_cv(thread_id_t tid)
1950 return get_parent_action(tid)->get_cv();
1954 * Resolve a set of Promises with a current write. The set is provided in the
1955 * Node corresponding to @a write.
1956 * @param write The ModelAction that is fulfilling Promises
1957 * @return True if promises were resolved; false otherwise
1959 bool ModelChecker::resolve_promises(ModelAction *write)
1961 bool resolved = false;
1962 std::vector< thread_id_t, ModelAlloc<thread_id_t> > threads_to_check;
1964 for (unsigned int i = 0, promise_index = 0; promise_index < promises->size(); i++) {
1965 Promise *promise = (*promises)[promise_index];
1966 if (write->get_node()->get_promise(i)) {
1967 ModelAction *read = promise->get_action();
1968 if (read->is_rmw()) {
1969 mo_graph->addRMWEdge(write, read);
1971 read->read_from(write);
1972 //First fix up the modification order for actions that happened
1974 r_modification_order(read, write);
1975 //Next fix up the modification order for actions that happened
1977 post_r_modification_order(read, write);
1978 //Make sure the promise's value matches the write's value
1979 ASSERT(promise->get_value() == write->get_value());
1982 promises->erase(promises->begin() + promise_index);
1983 threads_to_check.push_back(read->get_tid());
1990 //Check whether reading these writes has made threads unable to
1993 for(unsigned int i=0;i<threads_to_check.size();i++)
1994 mo_check_promises(threads_to_check[i], write);
2000 * Compute the set of promises that could potentially be satisfied by this
2001 * action. Note that the set computation actually appears in the Node, not in
2003 * @param curr The ModelAction that may satisfy promises
2005 void ModelChecker::compute_promises(ModelAction *curr)
2007 for (unsigned int i = 0; i < promises->size(); i++) {
2008 Promise *promise = (*promises)[i];
2009 const ModelAction *act = promise->get_action();
2010 if (!act->happens_before(curr) &&
2012 !act->could_synchronize_with(curr) &&
2013 !act->same_thread(curr) &&
2014 act->get_location() == curr->get_location() &&
2015 promise->get_value() == curr->get_value()) {
2016 curr->get_node()->set_promise(i, act->is_rmw());
2021 /** Checks promises in response to change in ClockVector Threads. */
2022 void ModelChecker::check_promises(thread_id_t tid, ClockVector *old_cv, ClockVector *merge_cv)
2024 for (unsigned int i = 0; i < promises->size(); i++) {
2025 Promise *promise = (*promises)[i];
2026 const ModelAction *act = promise->get_action();
2027 if ((old_cv == NULL || !old_cv->synchronized_since(act)) &&
2028 merge_cv->synchronized_since(act)) {
2029 if (promise->increment_threads(tid)) {
2030 //Promise has failed
2031 failed_promise = true;
2038 void ModelChecker::check_promises_thread_disabled() {
2039 for (unsigned int i = 0; i < promises->size(); i++) {
2040 Promise *promise = (*promises)[i];
2041 if (promise->check_promise()) {
2042 failed_promise = true;
2048 /** Checks promises in response to addition to modification order for threads.
2050 * pthread is the thread that performed the read that created the promise
2052 * pread is the read that created the promise
2054 * pwrite is either the first write to same location as pread by
2055 * pthread that is sequenced after pread or the value read by the
2056 * first read to the same lcoation as pread by pthread that is
2057 * sequenced after pread..
2059 * 1. If tid=pthread, then we check what other threads are reachable
2060 * through the mode order starting with pwrite. Those threads cannot
2061 * perform a write that will resolve the promise due to modification
2062 * order constraints.
2064 * 2. If the tid is not pthread, we check whether pwrite can reach the
2065 * action write through the modification order. If so, that thread
2066 * cannot perform a future write that will resolve the promise due to
2067 * modificatin order constraints.
2069 * @parem tid The thread that either read from the model action
2070 * write, or actually did the model action write.
2072 * @parem write The ModelAction representing the relevant write.
2075 void ModelChecker::mo_check_promises(thread_id_t tid, const ModelAction *write) {
2076 void * location = write->get_location();
2077 for (unsigned int i = 0; i < promises->size(); i++) {
2078 Promise *promise = (*promises)[i];
2079 const ModelAction *act = promise->get_action();
2081 //Is this promise on the same location?
2082 if ( act->get_location() != location )
2085 //same thread as the promise
2086 if ( act->get_tid()==tid ) {
2088 //do we have a pwrite for the promise, if not, set it
2089 if (promise->get_write() == NULL ) {
2090 promise->set_write(write);
2091 //The pwrite cannot happen before the promise
2092 if (write->happens_before(act) && (write != act)) {
2093 failed_promise = true;
2097 if (mo_graph->checkPromise(write, promise)) {
2098 failed_promise = true;
2103 //Don't do any lookups twice for the same thread
2104 if (promise->has_sync_thread(tid))
2107 if (promise->get_write()&&mo_graph->checkReachable(promise->get_write(), write)) {
2108 if (promise->increment_threads(tid)) {
2109 failed_promise = true;
2117 * Compute the set of writes that may break the current pending release
2118 * sequence. This information is extracted from previou release sequence
2121 * @param curr The current ModelAction. Must be a release sequence fixup
2124 void ModelChecker::compute_relseq_breakwrites(ModelAction *curr)
2126 if (pending_rel_seqs->empty())
2129 struct release_seq *pending = pending_rel_seqs->back();
2130 for (unsigned int i = 0; i < pending->writes.size(); i++) {
2131 const ModelAction *write = pending->writes[i];
2132 curr->get_node()->add_relseq_break(write);
2135 /* NULL means don't break the sequence; just synchronize */
2136 curr->get_node()->add_relseq_break(NULL);
2140 * Build up an initial set of all past writes that this 'read' action may read
2141 * from. This set is determined by the clock vector's "happens before"
2143 * @param curr is the current ModelAction that we are exploring; it must be a
2146 void ModelChecker::build_reads_from_past(ModelAction *curr)
2148 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
2150 ASSERT(curr->is_read());
2152 ModelAction *last_seq_cst = NULL;
2154 /* Track whether this object has been initialized */
2155 bool initialized = false;
2157 if (curr->is_seqcst()) {
2158 last_seq_cst = get_last_seq_cst(curr);
2159 /* We have to at least see the last sequentially consistent write,
2160 so we are initialized. */
2161 if (last_seq_cst != NULL)
2165 /* Iterate over all threads */
2166 for (i = 0; i < thrd_lists->size(); i++) {
2167 /* Iterate over actions in thread, starting from most recent */
2168 action_list_t *list = &(*thrd_lists)[i];
2169 action_list_t::reverse_iterator rit;
2170 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2171 ModelAction *act = *rit;
2173 /* Only consider 'write' actions */
2174 if (!act->is_write() || act == curr)
2177 /* Don't consider more than one seq_cst write if we are a seq_cst read. */
2178 if (!curr->is_seqcst() || (!act->is_seqcst() && (last_seq_cst == NULL || !act->happens_before(last_seq_cst))) || act == last_seq_cst) {
2179 if (!curr->get_sleep_flag() || curr->is_seqcst() || sleep_can_read_from(curr, act)) {
2180 DEBUG("Adding action to may_read_from:\n");
2181 if (DBG_ENABLED()) {
2185 curr->get_node()->add_read_from(act);
2189 /* Include at most one act per-thread that "happens before" curr */
2190 if (act->happens_before(curr)) {
2198 assert_bug("May read from uninitialized atomic");
2200 if (DBG_ENABLED() || !initialized) {
2201 printf("Reached read action:\n");
2203 printf("Printing may_read_from\n");
2204 curr->get_node()->print_may_read_from();
2205 printf("End printing may_read_from\n");
2209 bool ModelChecker::sleep_can_read_from(ModelAction * curr, const ModelAction *write) {
2211 Node *prevnode=write->get_node()->get_parent();
2213 bool thread_sleep=prevnode->enabled_status(curr->get_tid())==THREAD_SLEEP_SET;
2214 if (write->is_release()&&thread_sleep)
2216 if (!write->is_rmw()) {
2219 if (write->get_reads_from()==NULL)
2221 write=write->get_reads_from();
2225 static void print_list(action_list_t *list)
2227 action_list_t::iterator it;
2229 printf("---------------------------------------------------------------------\n");
2231 unsigned int hash=0;
2233 for (it = list->begin(); it != list->end(); it++) {
2235 hash=hash^(hash<<3)^((*it)->hash());
2237 printf("HASH %u\n", hash);
2238 printf("---------------------------------------------------------------------\n");
2241 #if SUPPORT_MOD_ORDER_DUMP
2242 void ModelChecker::dumpGraph(char *filename) {
2244 sprintf(buffer, "%s.dot",filename);
2245 FILE *file=fopen(buffer, "w");
2246 fprintf(file, "digraph %s {\n",filename);
2247 mo_graph->dumpNodes(file);
2248 ModelAction ** thread_array=(ModelAction **)model_calloc(1, sizeof(ModelAction *)*get_num_threads());
2250 for (action_list_t::iterator it = action_trace->begin(); it != action_trace->end(); it++) {
2251 ModelAction *action=*it;
2252 if (action->is_read()) {
2253 fprintf(file, "N%u [label=\"%u, T%u\"];\n", action->get_seq_number(),action->get_seq_number(), action->get_tid());
2254 if (action->get_reads_from()!=NULL)
2255 fprintf(file, "N%u -> N%u[label=\"rf\", color=red];\n", action->get_seq_number(), action->get_reads_from()->get_seq_number());
2257 if (thread_array[action->get_tid()] != NULL) {
2258 fprintf(file, "N%u -> N%u[label=\"sb\", color=blue];\n", thread_array[action->get_tid()]->get_seq_number(), action->get_seq_number());
2261 thread_array[action->get_tid()]=action;
2263 fprintf(file,"}\n");
2264 model_free(thread_array);
2269 void ModelChecker::print_summary()
2271 #if SUPPORT_MOD_ORDER_DUMP
2273 char buffername[100];
2274 sprintf(buffername, "exec%04u", stats.num_total);
2275 mo_graph->dumpGraphToFile(buffername);
2276 sprintf(buffername, "graph%04u", stats.num_total);
2277 dumpGraph(buffername);
2280 if (!isfinalfeasible())
2281 printf("INFEASIBLE EXECUTION!\n");
2282 print_list(action_trace);
2287 * Add a Thread to the system for the first time. Should only be called once
2289 * @param t The Thread to add
2291 void ModelChecker::add_thread(Thread *t)
2293 thread_map->put(id_to_int(t->get_id()), t);
2294 scheduler->add_thread(t);
2298 * Removes a thread from the scheduler.
2299 * @param the thread to remove.
2301 void ModelChecker::remove_thread(Thread *t)
2303 scheduler->remove_thread(t);
2307 * @brief Get a Thread reference by its ID
2308 * @param tid The Thread's ID
2309 * @return A Thread reference
2311 Thread * ModelChecker::get_thread(thread_id_t tid) const
2313 return thread_map->get(id_to_int(tid));
2317 * @brief Get a reference to the Thread in which a ModelAction was executed
2318 * @param act The ModelAction
2319 * @return A Thread reference
2321 Thread * ModelChecker::get_thread(ModelAction *act) const
2323 return get_thread(act->get_tid());
2327 * @brief Check if a Thread is currently enabled
2328 * @param t The Thread to check
2329 * @return True if the Thread is currently enabled
2331 bool ModelChecker::is_enabled(Thread *t) const
2333 return scheduler->is_enabled(t);
2337 * @brief Check if a Thread is currently enabled
2338 * @param tid The ID of the Thread to check
2339 * @return True if the Thread is currently enabled
2341 bool ModelChecker::is_enabled(thread_id_t tid) const
2343 return scheduler->is_enabled(tid);
2347 * Switch from a user-context to the "master thread" context (a.k.a. system
2348 * context). This switch is made with the intention of exploring a particular
2349 * model-checking action (described by a ModelAction object). Must be called
2350 * from a user-thread context.
2352 * @param act The current action that will be explored. May be NULL only if
2353 * trace is exiting via an assertion (see ModelChecker::set_assert and
2354 * ModelChecker::has_asserted).
2355 * @return Return status from the 'swap' call (i.e., success/fail, 0/-1)
2357 int ModelChecker::switch_to_master(ModelAction *act)
2360 Thread *old = thread_current();
2361 set_current_action(act);
2362 old->set_state(THREAD_READY);
2363 return Thread::swap(old, &system_context);
2367 * Takes the next step in the execution, if possible.
2368 * @return Returns true (success) if a step was taken and false otherwise.
2370 bool ModelChecker::take_step() {
2374 Thread *curr = priv->current_action ? get_thread(priv->current_action) : NULL;
2376 if (curr->get_state() == THREAD_READY) {
2377 ASSERT(priv->current_action);
2379 priv->nextThread = check_current_action(priv->current_action);
2380 priv->current_action = NULL;
2382 if (curr->is_blocked() || curr->is_complete())
2383 scheduler->remove_thread(curr);
2388 Thread *next = scheduler->next_thread(priv->nextThread);
2390 /* Infeasible -> don't take any more steps */
2393 else if (isfeasibleprefix() && have_bug_reports()) {
2398 if (params.bound != 0) {
2399 if (priv->used_sequence_numbers > params.bound) {
2404 DEBUG("(%d, %d)\n", curr ? id_to_int(curr->get_id()) : -1,
2405 next ? id_to_int(next->get_id()) : -1);
2408 * Launch end-of-execution release sequence fixups only when there are:
2410 * (1) no more user threads to run (or when execution replay chooses
2411 * the 'model_thread')
2412 * (2) pending release sequences
2413 * (3) pending assertions (i.e., data races)
2414 * (4) no pending promises
2416 if (!pending_rel_seqs->empty() && (!next || next->is_model_thread()) &&
2417 isfinalfeasible() && !unrealizedraces.empty()) {
2418 printf("*** WARNING: release sequence fixup action (%zu pending release seuqences) ***\n",
2419 pending_rel_seqs->size());
2420 ModelAction *fixup = new ModelAction(MODEL_FIXUP_RELSEQ,
2421 std::memory_order_seq_cst, NULL, VALUE_NONE,
2423 set_current_action(fixup);
2427 /* next == NULL -> don't take any more steps */
2431 next->set_state(THREAD_RUNNING);
2433 if (next->get_pending() != NULL) {
2434 /* restart a pending action */
2435 set_current_action(next->get_pending());
2436 next->set_pending(NULL);
2437 next->set_state(THREAD_READY);
2441 /* Return false only if swap fails with an error */
2442 return (Thread::swap(&system_context, next) == 0);
2445 /** Runs the current execution until threre are no more steps to take. */
2446 void ModelChecker::finish_execution() {
2449 while (take_step());