10 #include "snapshot-interface.h"
12 #include "clockvector.h"
13 #include "cyclegraph.h"
16 #include "threads-model.h"
19 #define INITIAL_THREAD_ID 0
24 bug_message(const char *str) {
25 const char *fmt = " [BUG] %s\n";
26 msg = (char *)snapshot_malloc(strlen(fmt) + strlen(str));
27 sprintf(msg, fmt, str);
29 ~bug_message() { if (msg) snapshot_free(msg); }
32 void print() { model_print("%s", msg); }
38 * Structure for holding small ModelChecker members that should be snapshotted
40 struct model_snapshot_members {
41 model_snapshot_members() :
42 /* First thread created will have id INITIAL_THREAD_ID */
43 next_thread_id(INITIAL_THREAD_ID),
44 used_sequence_numbers(0),
48 failed_promise(false),
49 too_many_reads(false),
50 no_valid_reads(false),
51 bad_synchronization(false),
55 ~model_snapshot_members() {
56 for (unsigned int i = 0; i < bugs.size(); i++)
61 unsigned int next_thread_id;
62 modelclock_t used_sequence_numbers;
63 ModelAction *next_backtrack;
64 SnapVector<bug_message *> bugs;
65 struct execution_stats stats;
69 /** @brief Incorrectly-ordered synchronization was made */
70 bool bad_synchronization;
76 /** @brief Constructor */
77 ModelChecker::ModelChecker(struct model_params params) :
78 /* Initialize default scheduler */
80 scheduler(new Scheduler()),
82 earliest_diverge(NULL),
83 action_trace(new action_list_t()),
84 thread_map(new HashTable<int, Thread *, int>()),
85 obj_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
86 lock_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
87 condvar_waiters_map(new HashTable<const void *, action_list_t *, uintptr_t, 4>()),
88 obj_thrd_map(new HashTable<void *, SnapVector<action_list_t> *, uintptr_t, 4 >()),
89 promises(new SnapVector<Promise *>()),
90 futurevalues(new SnapVector<struct PendingFutureValue>()),
91 pending_rel_seqs(new SnapVector<struct release_seq *>()),
92 thrd_last_action(new SnapVector<ModelAction *>(1)),
93 thrd_last_fence_release(new SnapVector<ModelAction *>()),
94 node_stack(new NodeStack()),
95 priv(new struct model_snapshot_members()),
96 mo_graph(new CycleGraph())
98 /* Initialize a model-checker thread, for special ModelActions */
99 model_thread = new Thread(get_next_id());
100 thread_map->put(id_to_int(model_thread->get_id()), model_thread);
103 /** @brief Destructor */
104 ModelChecker::~ModelChecker()
106 for (unsigned int i = 0; i < get_num_threads(); i++)
107 delete thread_map->get(i);
112 delete lock_waiters_map;
113 delete condvar_waiters_map;
116 for (unsigned int i = 0; i < promises->size(); i++)
117 delete (*promises)[i];
120 delete pending_rel_seqs;
122 delete thrd_last_action;
123 delete thrd_last_fence_release;
130 static action_list_t * get_safe_ptr_action(HashTable<const void *, action_list_t *, uintptr_t, 4> * hash, void * ptr)
132 action_list_t *tmp = hash->get(ptr);
134 tmp = new action_list_t();
140 static SnapVector<action_list_t> * get_safe_ptr_vect_action(HashTable<void *, SnapVector<action_list_t> *, uintptr_t, 4> * hash, void * ptr)
142 SnapVector<action_list_t> *tmp = hash->get(ptr);
144 tmp = new SnapVector<action_list_t>();
151 * Restores user program to initial state and resets all model-checker data
154 void ModelChecker::reset_to_initial_state()
156 DEBUG("+++ Resetting to initial state +++\n");
157 node_stack->reset_execution();
159 /* Print all model-checker output before rollback */
163 * FIXME: if we utilize partial rollback, we will need to free only
164 * those pending actions which were NOT pending before the rollback
167 for (unsigned int i = 0; i < get_num_threads(); i++)
168 delete get_thread(int_to_id(i))->get_pending();
170 snapshot_backtrack_before(0);
173 /** @return a thread ID for a new Thread */
174 thread_id_t ModelChecker::get_next_id()
176 return priv->next_thread_id++;
179 /** @return the number of user threads created during this execution */
180 unsigned int ModelChecker::get_num_threads() const
182 return priv->next_thread_id;
186 * Must be called from user-thread context (e.g., through the global
187 * thread_current() interface)
189 * @return The currently executing Thread.
191 Thread * ModelChecker::get_current_thread() const
193 return scheduler->get_current_thread();
196 /** @return a sequence number for a new ModelAction */
197 modelclock_t ModelChecker::get_next_seq_num()
199 return ++priv->used_sequence_numbers;
202 Node * ModelChecker::get_curr_node() const
204 return node_stack->get_head();
208 * @brief Select the next thread to execute based on the curren action
210 * RMW actions occur in two parts, and we cannot split them. And THREAD_CREATE
211 * actions should be followed by the execution of their child thread. In either
212 * case, the current action should determine the next thread schedule.
214 * @param curr The current action
215 * @return The next thread to run, if the current action will determine this
216 * selection; otherwise NULL
218 Thread * ModelChecker::action_select_next_thread(const ModelAction *curr) const
220 /* Do not split atomic RMW */
222 return get_thread(curr);
223 /* Follow CREATE with the created thread */
224 if (curr->get_type() == THREAD_CREATE)
225 return curr->get_thread_operand();
230 * @brief Choose the next thread to execute.
232 * This function chooses the next thread that should execute. It can enforce
233 * execution replay/backtracking or, if the model-checker has no preference
234 * regarding the next thread (i.e., when exploring a new execution ordering),
235 * we defer to the scheduler.
237 * @return The next chosen thread to run, if any exist. Or else if the current
238 * execution should terminate, return NULL.
240 Thread * ModelChecker::get_next_thread()
245 * Have we completed exploring the preselected path? Then let the
249 return scheduler->select_next_thread();
251 /* Else, we are trying to replay an execution */
252 ModelAction *next = node_stack->get_next()->get_action();
254 if (next == diverge) {
255 if (earliest_diverge == NULL || *diverge < *earliest_diverge)
256 earliest_diverge = diverge;
258 Node *nextnode = next->get_node();
259 Node *prevnode = nextnode->get_parent();
260 scheduler->update_sleep_set(prevnode);
262 /* Reached divergence point */
263 if (nextnode->increment_behaviors()) {
264 /* Execute the same thread with a new behavior */
265 tid = next->get_tid();
266 node_stack->pop_restofstack(2);
269 /* Make a different thread execute for next step */
270 scheduler->add_sleep(get_thread(next->get_tid()));
271 tid = prevnode->get_next_backtrack();
272 /* Make sure the backtracked thread isn't sleeping. */
273 node_stack->pop_restofstack(1);
274 if (diverge == earliest_diverge) {
275 earliest_diverge = prevnode->get_action();
278 /* Start the round robin scheduler from this thread id */
279 scheduler->set_scheduler_thread(tid);
280 /* The correct sleep set is in the parent node. */
283 DEBUG("*** Divergence point ***\n");
287 tid = next->get_tid();
289 DEBUG("*** ModelChecker chose next thread = %d ***\n", id_to_int(tid));
290 ASSERT(tid != THREAD_ID_T_NONE);
291 return thread_map->get(id_to_int(tid));
295 * We need to know what the next actions of all threads in the sleep
296 * set will be. This method computes them and stores the actions at
297 * the corresponding thread object's pending action.
300 void ModelChecker::execute_sleep_set()
302 for (unsigned int i = 0; i < get_num_threads(); i++) {
303 thread_id_t tid = int_to_id(i);
304 Thread *thr = get_thread(tid);
305 if (scheduler->is_sleep_set(thr) && thr->get_pending()) {
306 thr->get_pending()->set_sleep_flag();
312 * @brief Should the current action wake up a given thread?
314 * @param curr The current action
315 * @param thread The thread that we might wake up
316 * @return True, if we should wake up the sleeping thread; false otherwise
318 bool ModelChecker::should_wake_up(const ModelAction *curr, const Thread *thread) const
320 const ModelAction *asleep = thread->get_pending();
321 /* Don't allow partial RMW to wake anyone up */
324 /* Synchronizing actions may have been backtracked */
325 if (asleep->could_synchronize_with(curr))
327 /* All acquire/release fences and fence-acquire/store-release */
328 if (asleep->is_fence() && asleep->is_acquire() && curr->is_release())
330 /* Fence-release + store can awake load-acquire on the same location */
331 if (asleep->is_read() && asleep->is_acquire() && curr->same_var(asleep) && curr->is_write()) {
332 ModelAction *fence_release = get_last_fence_release(curr->get_tid());
333 if (fence_release && *(get_last_action(thread->get_id())) < *fence_release)
339 void ModelChecker::wake_up_sleeping_actions(ModelAction *curr)
341 for (unsigned int i = 0; i < get_num_threads(); i++) {
342 Thread *thr = get_thread(int_to_id(i));
343 if (scheduler->is_sleep_set(thr)) {
344 if (should_wake_up(curr, thr))
345 /* Remove this thread from sleep set */
346 scheduler->remove_sleep(thr);
351 /** @brief Alert the model-checker that an incorrectly-ordered
352 * synchronization was made */
353 void ModelChecker::set_bad_synchronization()
355 priv->bad_synchronization = true;
359 * Check whether the current trace has triggered an assertion which should halt
362 * @return True, if the execution should be aborted; false otherwise
364 bool ModelChecker::has_asserted() const
366 return priv->asserted;
370 * Trigger a trace assertion which should cause this execution to be halted.
371 * This can be due to a detected bug or due to an infeasibility that should
374 void ModelChecker::set_assert()
376 priv->asserted = true;
380 * Check if we are in a deadlock. Should only be called at the end of an
381 * execution, although it should not give false positives in the middle of an
382 * execution (there should be some ENABLED thread).
384 * @return True if program is in a deadlock; false otherwise
386 bool ModelChecker::is_deadlocked() const
388 bool blocking_threads = false;
389 for (unsigned int i = 0; i < get_num_threads(); i++) {
390 thread_id_t tid = int_to_id(i);
393 Thread *t = get_thread(tid);
394 if (!t->is_model_thread() && t->get_pending())
395 blocking_threads = true;
397 return blocking_threads;
401 * Check if this is a complete execution. That is, have all thread completed
402 * execution (rather than exiting because sleep sets have forced a redundant
405 * @return True if the execution is complete.
407 bool ModelChecker::is_complete_execution() const
409 for (unsigned int i = 0; i < get_num_threads(); i++)
410 if (is_enabled(int_to_id(i)))
416 * @brief Assert a bug in the executing program.
418 * Use this function to assert any sort of bug in the user program. If the
419 * current trace is feasible (actually, a prefix of some feasible execution),
420 * then this execution will be aborted, printing the appropriate message. If
421 * the current trace is not yet feasible, the error message will be stashed and
422 * printed if the execution ever becomes feasible.
424 * @param msg Descriptive message for the bug (do not include newline char)
425 * @return True if bug is immediately-feasible
427 bool ModelChecker::assert_bug(const char *msg)
429 priv->bugs.push_back(new bug_message(msg));
431 if (isfeasibleprefix()) {
439 * @brief Assert a bug in the executing program, asserted by a user thread
440 * @see ModelChecker::assert_bug
441 * @param msg Descriptive message for the bug (do not include newline char)
443 void ModelChecker::assert_user_bug(const char *msg)
445 /* If feasible bug, bail out now */
447 switch_to_master(NULL);
450 /** @return True, if any bugs have been reported for this execution */
451 bool ModelChecker::have_bug_reports() const
453 return priv->bugs.size() != 0;
456 /** @brief Print bug report listing for this execution (if any bugs exist) */
457 void ModelChecker::print_bugs() const
459 if (have_bug_reports()) {
460 model_print("Bug report: %zu bug%s detected\n",
462 priv->bugs.size() > 1 ? "s" : "");
463 for (unsigned int i = 0; i < priv->bugs.size(); i++)
464 priv->bugs[i]->print();
469 * @brief Record end-of-execution stats
471 * Must be run when exiting an execution. Records various stats.
472 * @see struct execution_stats
474 void ModelChecker::record_stats()
477 if (!isfeasibleprefix())
478 stats.num_infeasible++;
479 else if (have_bug_reports())
480 stats.num_buggy_executions++;
481 else if (is_complete_execution())
482 stats.num_complete++;
484 stats.num_redundant++;
487 * @todo We can violate this ASSERT() when fairness/sleep sets
488 * conflict to cause an execution to terminate, e.g. with:
489 * Scheduler: [0: disabled][1: disabled][2: sleep][3: current, enabled]
491 //ASSERT(scheduler->all_threads_sleeping());
495 /** @brief Print execution stats */
496 void ModelChecker::print_stats() const
498 model_print("Number of complete, bug-free executions: %d\n", stats.num_complete);
499 model_print("Number of redundant executions: %d\n", stats.num_redundant);
500 model_print("Number of buggy executions: %d\n", stats.num_buggy_executions);
501 model_print("Number of infeasible executions: %d\n", stats.num_infeasible);
502 model_print("Total executions: %d\n", stats.num_total);
503 model_print("Total nodes created: %d\n", node_stack->get_total_nodes());
507 * @brief End-of-exeuction print
508 * @param printbugs Should any existing bugs be printed?
510 void ModelChecker::print_execution(bool printbugs) const
512 print_program_output();
514 if (params.verbose) {
515 model_print("Earliest divergence point since last feasible execution:\n");
516 if (earliest_diverge)
517 earliest_diverge->print();
519 model_print("(Not set)\n");
525 /* Don't print invalid bugs */
534 * Queries the model-checker for more executions to explore and, if one
535 * exists, resets the model-checker state to execute a new execution.
537 * @return If there are more executions to explore, return true. Otherwise,
540 bool ModelChecker::next_execution()
543 /* Is this execution a feasible execution that's worth bug-checking? */
544 bool complete = isfeasibleprefix() && (is_complete_execution() ||
547 /* End-of-execution bug checks */
550 assert_bug("Deadlock detected");
558 if (params.verbose || (complete && have_bug_reports()))
559 print_execution(complete);
561 clear_program_output();
564 earliest_diverge = NULL;
566 if ((diverge = get_next_backtrack()) == NULL)
570 model_print("Next execution will diverge at:\n");
574 reset_to_initial_state();
579 * @brief Find the last fence-related backtracking conflict for a ModelAction
581 * This function performs the search for the most recent conflicting action
582 * against which we should perform backtracking, as affected by fence
583 * operations. This includes pairs of potentially-synchronizing actions which
584 * occur due to fence-acquire or fence-release, and hence should be explored in
585 * the opposite execution order.
587 * @param act The current action
588 * @return The most recent action which conflicts with act due to fences
590 ModelAction * ModelChecker::get_last_fence_conflict(ModelAction *act) const
592 /* Only perform release/acquire fence backtracking for stores */
593 if (!act->is_write())
596 /* Find a fence-release (or, act is a release) */
597 ModelAction *last_release;
598 if (act->is_release())
601 last_release = get_last_fence_release(act->get_tid());
605 /* Skip past the release */
606 action_list_t *list = action_trace;
607 action_list_t::reverse_iterator rit;
608 for (rit = list->rbegin(); rit != list->rend(); rit++)
609 if (*rit == last_release)
611 ASSERT(rit != list->rend());
616 * load --sb-> fence-acquire */
617 ModelVector<ModelAction *> acquire_fences(get_num_threads(), NULL);
618 ModelVector<ModelAction *> prior_loads(get_num_threads(), NULL);
619 bool found_acquire_fences = false;
620 for ( ; rit != list->rend(); rit++) {
621 ModelAction *prev = *rit;
622 if (act->same_thread(prev))
625 int tid = id_to_int(prev->get_tid());
627 if (prev->is_read() && act->same_var(prev)) {
628 if (prev->is_acquire()) {
629 /* Found most recent load-acquire, don't need
630 * to search for more fences */
631 if (!found_acquire_fences)
634 prior_loads[tid] = prev;
637 if (prev->is_acquire() && prev->is_fence() && !acquire_fences[tid]) {
638 found_acquire_fences = true;
639 acquire_fences[tid] = prev;
643 ModelAction *latest_backtrack = NULL;
644 for (unsigned int i = 0; i < acquire_fences.size(); i++)
645 if (acquire_fences[i] && prior_loads[i])
646 if (!latest_backtrack || *latest_backtrack < *acquire_fences[i])
647 latest_backtrack = acquire_fences[i];
648 return latest_backtrack;
652 * @brief Find the last backtracking conflict for a ModelAction
654 * This function performs the search for the most recent conflicting action
655 * against which we should perform backtracking. This primary includes pairs of
656 * synchronizing actions which should be explored in the opposite execution
659 * @param act The current action
660 * @return The most recent action which conflicts with act
662 ModelAction * ModelChecker::get_last_conflict(ModelAction *act) const
664 switch (act->get_type()) {
665 /* case ATOMIC_FENCE: fences don't directly cause backtracking */
669 ModelAction *ret = NULL;
671 /* linear search: from most recent to oldest */
672 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
673 action_list_t::reverse_iterator rit;
674 for (rit = list->rbegin(); rit != list->rend(); rit++) {
675 ModelAction *prev = *rit;
676 if (prev->could_synchronize_with(act)) {
682 ModelAction *ret2 = get_last_fence_conflict(act);
692 case ATOMIC_TRYLOCK: {
693 /* linear search: from most recent to oldest */
694 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
695 action_list_t::reverse_iterator rit;
696 for (rit = list->rbegin(); rit != list->rend(); rit++) {
697 ModelAction *prev = *rit;
698 if (act->is_conflicting_lock(prev))
703 case ATOMIC_UNLOCK: {
704 /* linear search: from most recent to oldest */
705 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
706 action_list_t::reverse_iterator rit;
707 for (rit = list->rbegin(); rit != list->rend(); rit++) {
708 ModelAction *prev = *rit;
709 if (!act->same_thread(prev) && prev->is_failed_trylock())
715 /* linear search: from most recent to oldest */
716 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
717 action_list_t::reverse_iterator rit;
718 for (rit = list->rbegin(); rit != list->rend(); rit++) {
719 ModelAction *prev = *rit;
720 if (!act->same_thread(prev) && prev->is_failed_trylock())
722 if (!act->same_thread(prev) && prev->is_notify())
728 case ATOMIC_NOTIFY_ALL:
729 case ATOMIC_NOTIFY_ONE: {
730 /* linear search: from most recent to oldest */
731 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
732 action_list_t::reverse_iterator rit;
733 for (rit = list->rbegin(); rit != list->rend(); rit++) {
734 ModelAction *prev = *rit;
735 if (!act->same_thread(prev) && prev->is_wait())
746 /** This method finds backtracking points where we should try to
747 * reorder the parameter ModelAction against.
749 * @param the ModelAction to find backtracking points for.
751 void ModelChecker::set_backtracking(ModelAction *act)
753 Thread *t = get_thread(act);
754 ModelAction *prev = get_last_conflict(act);
758 Node *node = prev->get_node()->get_parent();
760 int low_tid, high_tid;
761 if (node->enabled_status(t->get_id()) == THREAD_ENABLED) {
762 low_tid = id_to_int(act->get_tid());
763 high_tid = low_tid + 1;
766 high_tid = get_num_threads();
769 for (int i = low_tid; i < high_tid; i++) {
770 thread_id_t tid = int_to_id(i);
772 /* Make sure this thread can be enabled here. */
773 if (i >= node->get_num_threads())
776 /* Don't backtrack into a point where the thread is disabled or sleeping. */
777 if (node->enabled_status(tid) != THREAD_ENABLED)
780 /* Check if this has been explored already */
781 if (node->has_been_explored(tid))
784 /* See if fairness allows */
785 if (model->params.fairwindow != 0 && !node->has_priority(tid)) {
787 for (int t = 0; t < node->get_num_threads(); t++) {
788 thread_id_t tother = int_to_id(t);
789 if (node->is_enabled(tother) && node->has_priority(tother)) {
798 /* See if CHESS-like yield fairness allows */
799 if (model->params.yieldon) {
801 for (int t = 0; t < node->get_num_threads(); t++) {
802 thread_id_t tother = int_to_id(t);
803 if (node->is_enabled(tother) && node->has_priority_over(tid, tother)) {
812 /* Cache the latest backtracking point */
813 set_latest_backtrack(prev);
815 /* If this is a new backtracking point, mark the tree */
816 if (!node->set_backtrack(tid))
818 DEBUG("Setting backtrack: conflict = %d, instead tid = %d\n",
819 id_to_int(prev->get_tid()),
820 id_to_int(t->get_id()));
829 * @brief Cache the a backtracking point as the "most recent", if eligible
831 * Note that this does not prepare the NodeStack for this backtracking
832 * operation, it only caches the action on a per-execution basis
834 * @param act The operation at which we should explore a different next action
835 * (i.e., backtracking point)
836 * @return True, if this action is now the most recent backtracking point;
839 bool ModelChecker::set_latest_backtrack(ModelAction *act)
841 if (!priv->next_backtrack || *act > *priv->next_backtrack) {
842 priv->next_backtrack = act;
849 * Returns last backtracking point. The model checker will explore a different
850 * path for this point in the next execution.
851 * @return The ModelAction at which the next execution should diverge.
853 ModelAction * ModelChecker::get_next_backtrack()
855 ModelAction *next = priv->next_backtrack;
856 priv->next_backtrack = NULL;
861 * Processes a read model action.
862 * @param curr is the read model action to process.
863 * @return True if processing this read updates the mo_graph.
865 bool ModelChecker::process_read(ModelAction *curr)
867 Node *node = curr->get_node();
869 bool updated = false;
870 switch (node->get_read_from_status()) {
871 case READ_FROM_PAST: {
872 const ModelAction *rf = node->get_read_from_past();
875 mo_graph->startChanges();
877 ASSERT(!is_infeasible());
878 if (!check_recency(curr, rf)) {
879 if (node->increment_read_from()) {
880 mo_graph->rollbackChanges();
883 priv->too_many_reads = true;
887 updated = r_modification_order(curr, rf);
889 mo_graph->commitChanges();
890 mo_check_promises(curr, true);
893 case READ_FROM_PROMISE: {
894 Promise *promise = curr->get_node()->get_read_from_promise();
895 if (promise->add_reader(curr))
896 priv->failed_promise = true;
897 curr->set_read_from_promise(promise);
898 mo_graph->startChanges();
899 if (!check_recency(curr, promise))
900 priv->too_many_reads = true;
901 updated = r_modification_order(curr, promise);
902 mo_graph->commitChanges();
905 case READ_FROM_FUTURE: {
906 /* Read from future value */
907 struct future_value fv = node->get_future_value();
908 Promise *promise = new Promise(curr, fv);
909 curr->set_read_from_promise(promise);
910 promises->push_back(promise);
911 mo_graph->startChanges();
912 updated = r_modification_order(curr, promise);
913 mo_graph->commitChanges();
919 get_thread(curr)->set_return_value(curr->get_return_value());
925 * Processes a lock, trylock, or unlock model action. @param curr is
926 * the read model action to process.
928 * The try lock operation checks whether the lock is taken. If not,
929 * it falls to the normal lock operation case. If so, it returns
932 * The lock operation has already been checked that it is enabled, so
933 * it just grabs the lock and synchronizes with the previous unlock.
935 * The unlock operation has to re-enable all of the threads that are
936 * waiting on the lock.
938 * @return True if synchronization was updated; false otherwise
940 bool ModelChecker::process_mutex(ModelAction *curr)
942 std::mutex *mutex = curr->get_mutex();
943 struct std::mutex_state *state = NULL;
946 state = mutex->get_state();
948 switch (curr->get_type()) {
949 case ATOMIC_TRYLOCK: {
950 bool success = !state->locked;
951 curr->set_try_lock(success);
953 get_thread(curr)->set_return_value(0);
956 get_thread(curr)->set_return_value(1);
958 //otherwise fall into the lock case
960 if (curr->get_cv()->getClock(state->alloc_tid) <= state->alloc_clock)
961 assert_bug("Lock access before initialization");
962 state->locked = get_thread(curr);
963 ModelAction *unlock = get_last_unlock(curr);
964 //synchronize with the previous unlock statement
965 if (unlock != NULL) {
966 curr->synchronize_with(unlock);
971 case ATOMIC_UNLOCK: {
973 state->locked = NULL;
974 //wake up the other threads
975 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, curr->get_location());
976 //activate all the waiting threads
977 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
978 scheduler->wake(get_thread(*rit));
985 state->locked = NULL;
986 //wake up the other threads
987 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, (void *) curr->get_value());
988 //activate all the waiting threads
989 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
990 scheduler->wake(get_thread(*rit));
993 //check whether we should go to sleep or not...simulate spurious failures
994 if (curr->get_node()->get_misc() == 0) {
995 get_safe_ptr_action(condvar_waiters_map, curr->get_location())->push_back(curr);
997 scheduler->sleep(get_thread(curr));
1001 case ATOMIC_NOTIFY_ALL: {
1002 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
1003 //activate all the waiting threads
1004 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
1005 scheduler->wake(get_thread(*rit));
1010 case ATOMIC_NOTIFY_ONE: {
1011 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
1012 int wakeupthread = curr->get_node()->get_misc();
1013 action_list_t::iterator it = waiters->begin();
1014 advance(it, wakeupthread);
1015 scheduler->wake(get_thread(*it));
1027 * @brief Check if the current pending promises allow a future value to be sent
1029 * If one of the following is true:
1030 * (a) there are no pending promises
1031 * (b) the reader and writer do not cross any promises
1032 * Then, it is safe to pass a future value back now.
1034 * Otherwise, we must save the pending future value until (a) or (b) is true
1036 * @param writer The operation which sends the future value. Must be a write.
1037 * @param reader The operation which will observe the value. Must be a read.
1038 * @return True if the future value can be sent now; false if it must wait.
1040 bool ModelChecker::promises_may_allow(const ModelAction *writer,
1041 const ModelAction *reader) const
1043 if (promises->empty())
1045 for(int i=promises->size()-1;i>=0;i--) {
1046 ModelAction *pr=(*promises)[i]->get_reader(0);
1047 //reader is after promise...doesn't cross any promise
1050 //writer is after promise, reader before...bad...
1058 * @brief Add a future value to a reader
1060 * This function performs a few additional checks to ensure that the future
1061 * value can be feasibly observed by the reader
1063 * @param writer The operation whose value is sent. Must be a write.
1064 * @param reader The read operation which may read the future value. Must be a read.
1066 void ModelChecker::add_future_value(const ModelAction *writer, ModelAction *reader)
1068 /* Do more ambitious checks now that mo is more complete */
1069 if (!mo_may_allow(writer, reader))
1072 Node *node = reader->get_node();
1074 /* Find an ancestor thread which exists at the time of the reader */
1075 Thread *write_thread = get_thread(writer);
1076 while (id_to_int(write_thread->get_id()) >= node->get_num_threads())
1077 write_thread = write_thread->get_parent();
1079 struct future_value fv = {
1080 writer->get_write_value(),
1081 writer->get_seq_number() + params.maxfuturedelay,
1082 write_thread->get_id(),
1084 if (node->add_future_value(fv))
1085 set_latest_backtrack(reader);
1089 * Process a write ModelAction
1090 * @param curr The ModelAction to process
1091 * @return True if the mo_graph was updated or promises were resolved
1093 bool ModelChecker::process_write(ModelAction *curr)
1095 /* Readers to which we may send our future value */
1096 ModelVector<ModelAction *> send_fv;
1098 const ModelAction *earliest_promise_reader;
1099 bool updated_promises = false;
1101 bool updated_mod_order = w_modification_order(curr, &send_fv);
1102 Promise *promise = pop_promise_to_resolve(curr);
1105 earliest_promise_reader = promise->get_reader(0);
1106 updated_promises = resolve_promise(curr, promise);
1108 earliest_promise_reader = NULL;
1110 for (unsigned int i = 0; i < send_fv.size(); i++) {
1111 ModelAction *read = send_fv[i];
1113 /* Don't send future values to reads after the Promise we resolve */
1114 if (!earliest_promise_reader || *read < *earliest_promise_reader) {
1115 /* Check if future value can be sent immediately */
1116 if (promises_may_allow(curr, read)) {
1117 add_future_value(curr, read);
1119 futurevalues->push_back(PendingFutureValue(curr, read));
1124 /* Check the pending future values */
1125 for (int i = (int)futurevalues->size() - 1; i >= 0; i--) {
1126 struct PendingFutureValue pfv = (*futurevalues)[i];
1127 if (promises_may_allow(pfv.writer, pfv.reader)) {
1128 add_future_value(pfv.writer, pfv.reader);
1129 futurevalues->erase(futurevalues->begin() + i);
1133 mo_graph->commitChanges();
1134 mo_check_promises(curr, false);
1136 get_thread(curr)->set_return_value(VALUE_NONE);
1137 return updated_mod_order || updated_promises;
1141 * Process a fence ModelAction
1142 * @param curr The ModelAction to process
1143 * @return True if synchronization was updated
1145 bool ModelChecker::process_fence(ModelAction *curr)
1148 * fence-relaxed: no-op
1149 * fence-release: only log the occurence (not in this function), for
1150 * use in later synchronization
1151 * fence-acquire (this function): search for hypothetical release
1153 * fence-seq-cst: MO constraints formed in {r,w}_modification_order
1155 bool updated = false;
1156 if (curr->is_acquire()) {
1157 action_list_t *list = action_trace;
1158 action_list_t::reverse_iterator rit;
1159 /* Find X : is_read(X) && X --sb-> curr */
1160 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1161 ModelAction *act = *rit;
1164 if (act->get_tid() != curr->get_tid())
1166 /* Stop at the beginning of the thread */
1167 if (act->is_thread_start())
1169 /* Stop once we reach a prior fence-acquire */
1170 if (act->is_fence() && act->is_acquire())
1172 if (!act->is_read())
1174 /* read-acquire will find its own release sequences */
1175 if (act->is_acquire())
1178 /* Establish hypothetical release sequences */
1179 rel_heads_list_t release_heads;
1180 get_release_seq_heads(curr, act, &release_heads);
1181 for (unsigned int i = 0; i < release_heads.size(); i++)
1182 if (!curr->synchronize_with(release_heads[i]))
1183 set_bad_synchronization();
1184 if (release_heads.size() != 0)
1192 * @brief Process the current action for thread-related activity
1194 * Performs current-action processing for a THREAD_* ModelAction. Proccesses
1195 * may include setting Thread status, completing THREAD_FINISH/THREAD_JOIN
1196 * synchronization, etc. This function is a no-op for non-THREAD actions
1197 * (e.g., ATOMIC_{READ,WRITE,RMW,LOCK}, etc.)
1199 * @param curr The current action
1200 * @return True if synchronization was updated or a thread completed
1202 bool ModelChecker::process_thread_action(ModelAction *curr)
1204 bool updated = false;
1206 switch (curr->get_type()) {
1207 case THREAD_CREATE: {
1208 thrd_t *thrd = (thrd_t *)curr->get_location();
1209 struct thread_params *params = (struct thread_params *)curr->get_value();
1210 Thread *th = new Thread(thrd, params->func, params->arg, get_thread(curr));
1212 th->set_creation(curr);
1213 /* Promises can be satisfied by children */
1214 for (unsigned int i = 0; i < promises->size(); i++) {
1215 Promise *promise = (*promises)[i];
1216 if (promise->thread_is_available(curr->get_tid()))
1217 promise->add_thread(th->get_id());
1222 Thread *blocking = curr->get_thread_operand();
1223 ModelAction *act = get_last_action(blocking->get_id());
1224 curr->synchronize_with(act);
1225 updated = true; /* trigger rel-seq checks */
1228 case THREAD_FINISH: {
1229 Thread *th = get_thread(curr);
1230 while (!th->wait_list_empty()) {
1231 ModelAction *act = th->pop_wait_list();
1232 scheduler->wake(get_thread(act));
1235 /* Completed thread can't satisfy promises */
1236 for (unsigned int i = 0; i < promises->size(); i++) {
1237 Promise *promise = (*promises)[i];
1238 if (promise->thread_is_available(th->get_id()))
1239 if (promise->eliminate_thread(th->get_id()))
1240 priv->failed_promise = true;
1242 updated = true; /* trigger rel-seq checks */
1245 case THREAD_START: {
1246 check_promises(curr->get_tid(), NULL, curr->get_cv());
1257 * @brief Process the current action for release sequence fixup activity
1259 * Performs model-checker release sequence fixups for the current action,
1260 * forcing a single pending release sequence to break (with a given, potential
1261 * "loose" write) or to complete (i.e., synchronize). If a pending release
1262 * sequence forms a complete release sequence, then we must perform the fixup
1263 * synchronization, mo_graph additions, etc.
1265 * @param curr The current action; must be a release sequence fixup action
1266 * @param work_queue The work queue to which to add work items as they are
1269 void ModelChecker::process_relseq_fixup(ModelAction *curr, work_queue_t *work_queue)
1271 const ModelAction *write = curr->get_node()->get_relseq_break();
1272 struct release_seq *sequence = pending_rel_seqs->back();
1273 pending_rel_seqs->pop_back();
1275 ModelAction *acquire = sequence->acquire;
1276 const ModelAction *rf = sequence->rf;
1277 const ModelAction *release = sequence->release;
1281 ASSERT(release->same_thread(rf));
1283 if (write == NULL) {
1285 * @todo Forcing a synchronization requires that we set
1286 * modification order constraints. For instance, we can't allow
1287 * a fixup sequence in which two separate read-acquire
1288 * operations read from the same sequence, where the first one
1289 * synchronizes and the other doesn't. Essentially, we can't
1290 * allow any writes to insert themselves between 'release' and
1294 /* Must synchronize */
1295 if (!acquire->synchronize_with(release)) {
1296 set_bad_synchronization();
1299 /* Re-check all pending release sequences */
1300 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
1301 /* Re-check act for mo_graph edges */
1302 work_queue->push_back(MOEdgeWorkEntry(acquire));
1304 /* propagate synchronization to later actions */
1305 action_list_t::reverse_iterator rit = action_trace->rbegin();
1306 for (; (*rit) != acquire; rit++) {
1307 ModelAction *propagate = *rit;
1308 if (acquire->happens_before(propagate)) {
1309 propagate->synchronize_with(acquire);
1310 /* Re-check 'propagate' for mo_graph edges */
1311 work_queue->push_back(MOEdgeWorkEntry(propagate));
1315 /* Break release sequence with new edges:
1316 * release --mo--> write --mo--> rf */
1317 mo_graph->addEdge(release, write);
1318 mo_graph->addEdge(write, rf);
1321 /* See if we have realized a data race */
1326 * Initialize the current action by performing one or more of the following
1327 * actions, as appropriate: merging RMWR and RMWC/RMW actions, stepping forward
1328 * in the NodeStack, manipulating backtracking sets, allocating and
1329 * initializing clock vectors, and computing the promises to fulfill.
1331 * @param curr The current action, as passed from the user context; may be
1332 * freed/invalidated after the execution of this function, with a different
1333 * action "returned" its place (pass-by-reference)
1334 * @return True if curr is a newly-explored action; false otherwise
1336 bool ModelChecker::initialize_curr_action(ModelAction **curr)
1338 ModelAction *newcurr;
1340 if ((*curr)->is_rmwc() || (*curr)->is_rmw()) {
1341 newcurr = process_rmw(*curr);
1344 if (newcurr->is_rmw())
1345 compute_promises(newcurr);
1351 (*curr)->set_seq_number(get_next_seq_num());
1353 newcurr = node_stack->explore_action(*curr, scheduler->get_enabled_array());
1355 /* First restore type and order in case of RMW operation */
1356 if ((*curr)->is_rmwr())
1357 newcurr->copy_typeandorder(*curr);
1359 ASSERT((*curr)->get_location() == newcurr->get_location());
1360 newcurr->copy_from_new(*curr);
1362 /* Discard duplicate ModelAction; use action from NodeStack */
1365 /* Always compute new clock vector */
1366 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1369 return false; /* Action was explored previously */
1373 /* Always compute new clock vector */
1374 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1376 /* Assign most recent release fence */
1377 newcurr->set_last_fence_release(get_last_fence_release(newcurr->get_tid()));
1380 * Perform one-time actions when pushing new ModelAction onto
1383 if (newcurr->is_write())
1384 compute_promises(newcurr);
1385 else if (newcurr->is_relseq_fixup())
1386 compute_relseq_breakwrites(newcurr);
1387 else if (newcurr->is_wait())
1388 newcurr->get_node()->set_misc_max(2);
1389 else if (newcurr->is_notify_one()) {
1390 newcurr->get_node()->set_misc_max(get_safe_ptr_action(condvar_waiters_map, newcurr->get_location())->size());
1392 return true; /* This was a new ModelAction */
1397 * @brief Establish reads-from relation between two actions
1399 * Perform basic operations involved with establishing a concrete rf relation,
1400 * including setting the ModelAction data and checking for release sequences.
1402 * @param act The action that is reading (must be a read)
1403 * @param rf The action from which we are reading (must be a write)
1405 * @return True if this read established synchronization
1407 bool ModelChecker::read_from(ModelAction *act, const ModelAction *rf)
1410 ASSERT(rf->is_write());
1412 act->set_read_from(rf);
1413 if (act->is_acquire()) {
1414 rel_heads_list_t release_heads;
1415 get_release_seq_heads(act, act, &release_heads);
1416 int num_heads = release_heads.size();
1417 for (unsigned int i = 0; i < release_heads.size(); i++)
1418 if (!act->synchronize_with(release_heads[i])) {
1419 set_bad_synchronization();
1422 return num_heads > 0;
1428 * Check promises and eliminate potentially-satisfying threads when a thread is
1429 * blocked (e.g., join, lock). A thread which is waiting on another thread can
1430 * no longer satisfy a promise generated from that thread.
1432 * @param blocker The thread on which a thread is waiting
1433 * @param waiting The waiting thread
1435 void ModelChecker::thread_blocking_check_promises(Thread *blocker, Thread *waiting)
1437 for (unsigned int i = 0; i < promises->size(); i++) {
1438 Promise *promise = (*promises)[i];
1439 if (!promise->thread_is_available(waiting->get_id()))
1441 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
1442 ModelAction *reader = promise->get_reader(j);
1443 if (reader->get_tid() != blocker->get_id())
1445 if (promise->eliminate_thread(waiting->get_id())) {
1446 /* Promise has failed */
1447 priv->failed_promise = true;
1449 /* Only eliminate the 'waiting' thread once */
1457 * @brief Check whether a model action is enabled.
1459 * Checks whether a lock or join operation would be successful (i.e., is the
1460 * lock already locked, or is the joined thread already complete). If not, put
1461 * the action in a waiter list.
1463 * @param curr is the ModelAction to check whether it is enabled.
1464 * @return a bool that indicates whether the action is enabled.
1466 bool ModelChecker::check_action_enabled(ModelAction *curr) {
1467 if (curr->is_lock()) {
1468 std::mutex *lock = (std::mutex *)curr->get_location();
1469 struct std::mutex_state *state = lock->get_state();
1470 if (state->locked) {
1471 //Stick the action in the appropriate waiting queue
1472 get_safe_ptr_action(lock_waiters_map, curr->get_location())->push_back(curr);
1475 } else if (curr->is_thread_join()) {
1476 Thread *blocking = (Thread *)curr->get_location();
1477 if (!blocking->is_complete()) {
1478 blocking->push_wait_list(curr);
1479 thread_blocking_check_promises(blocking, get_thread(curr));
1488 * This is the heart of the model checker routine. It performs model-checking
1489 * actions corresponding to a given "current action." Among other processes, it
1490 * calculates reads-from relationships, updates synchronization clock vectors,
1491 * forms a memory_order constraints graph, and handles replay/backtrack
1492 * execution when running permutations of previously-observed executions.
1494 * @param curr The current action to process
1495 * @return The ModelAction that is actually executed; may be different than
1496 * curr; may be NULL, if the current action is not enabled to run
1498 ModelAction * ModelChecker::check_current_action(ModelAction *curr)
1501 bool second_part_of_rmw = curr->is_rmwc() || curr->is_rmw();
1503 if (!check_action_enabled(curr)) {
1504 /* Make the execution look like we chose to run this action
1505 * much later, when a lock/join can succeed */
1506 get_thread(curr)->set_pending(curr);
1507 scheduler->sleep(get_thread(curr));
1511 bool newly_explored = initialize_curr_action(&curr);
1517 wake_up_sleeping_actions(curr);
1519 /* Compute fairness information for CHESS yield algorithm */
1520 if (model->params.yieldon) {
1521 curr->get_node()->update_yield(scheduler);
1524 /* Add the action to lists before any other model-checking tasks */
1525 if (!second_part_of_rmw)
1526 add_action_to_lists(curr);
1528 /* Build may_read_from set for newly-created actions */
1529 if (newly_explored && curr->is_read())
1530 build_may_read_from(curr);
1532 /* Initialize work_queue with the "current action" work */
1533 work_queue_t work_queue(1, CheckCurrWorkEntry(curr));
1534 while (!work_queue.empty() && !has_asserted()) {
1535 WorkQueueEntry work = work_queue.front();
1536 work_queue.pop_front();
1538 switch (work.type) {
1539 case WORK_CHECK_CURR_ACTION: {
1540 ModelAction *act = work.action;
1541 bool update = false; /* update this location's release seq's */
1542 bool update_all = false; /* update all release seq's */
1544 if (process_thread_action(curr))
1547 if (act->is_read() && !second_part_of_rmw && process_read(act))
1550 if (act->is_write() && process_write(act))
1553 if (act->is_fence() && process_fence(act))
1556 if (act->is_mutex_op() && process_mutex(act))
1559 if (act->is_relseq_fixup())
1560 process_relseq_fixup(curr, &work_queue);
1563 work_queue.push_back(CheckRelSeqWorkEntry(NULL));
1565 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1568 case WORK_CHECK_RELEASE_SEQ:
1569 resolve_release_sequences(work.location, &work_queue);
1571 case WORK_CHECK_MO_EDGES: {
1572 /** @todo Complete verification of work_queue */
1573 ModelAction *act = work.action;
1574 bool updated = false;
1576 if (act->is_read()) {
1577 const ModelAction *rf = act->get_reads_from();
1578 const Promise *promise = act->get_reads_from_promise();
1580 if (r_modification_order(act, rf))
1582 } else if (promise) {
1583 if (r_modification_order(act, promise))
1587 if (act->is_write()) {
1588 if (w_modification_order(act, NULL))
1591 mo_graph->commitChanges();
1594 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1603 check_curr_backtracking(curr);
1604 set_backtracking(curr);
1608 void ModelChecker::check_curr_backtracking(ModelAction *curr)
1610 Node *currnode = curr->get_node();
1611 Node *parnode = currnode->get_parent();
1613 if ((parnode && !parnode->backtrack_empty()) ||
1614 !currnode->misc_empty() ||
1615 !currnode->read_from_empty() ||
1616 !currnode->promise_empty() ||
1617 !currnode->relseq_break_empty()) {
1618 set_latest_backtrack(curr);
1622 bool ModelChecker::promises_expired() const
1624 for (unsigned int i = 0; i < promises->size(); i++) {
1625 Promise *promise = (*promises)[i];
1626 if (promise->get_expiration() < priv->used_sequence_numbers)
1633 * This is the strongest feasibility check available.
1634 * @return whether the current trace (partial or complete) must be a prefix of
1637 bool ModelChecker::isfeasibleprefix() const
1639 return pending_rel_seqs->size() == 0 && is_feasible_prefix_ignore_relseq();
1643 * Print disagnostic information about an infeasible execution
1644 * @param prefix A string to prefix the output with; if NULL, then a default
1645 * message prefix will be provided
1647 void ModelChecker::print_infeasibility(const char *prefix) const
1651 if (mo_graph->checkForCycles())
1652 ptr += sprintf(ptr, "[mo cycle]");
1653 if (priv->failed_promise)
1654 ptr += sprintf(ptr, "[failed promise]");
1655 if (priv->too_many_reads)
1656 ptr += sprintf(ptr, "[too many reads]");
1657 if (priv->no_valid_reads)
1658 ptr += sprintf(ptr, "[no valid reads-from]");
1659 if (priv->bad_synchronization)
1660 ptr += sprintf(ptr, "[bad sw ordering]");
1661 if (promises_expired())
1662 ptr += sprintf(ptr, "[promise expired]");
1663 if (promises->size() != 0)
1664 ptr += sprintf(ptr, "[unresolved promise]");
1666 model_print("%s: %s\n", prefix ? prefix : "Infeasible", buf);
1670 * Returns whether the current completed trace is feasible, except for pending
1671 * release sequences.
1673 bool ModelChecker::is_feasible_prefix_ignore_relseq() const
1675 return !is_infeasible() && promises->size() == 0;
1679 * Check if the current partial trace is infeasible. Does not check any
1680 * end-of-execution flags, which might rule out the execution. Thus, this is
1681 * useful only for ruling an execution as infeasible.
1682 * @return whether the current partial trace is infeasible.
1684 bool ModelChecker::is_infeasible() const
1686 return mo_graph->checkForCycles() ||
1687 priv->no_valid_reads ||
1688 priv->failed_promise ||
1689 priv->too_many_reads ||
1690 priv->bad_synchronization ||
1694 /** Close out a RMWR by converting previous RMWR into a RMW or READ. */
1695 ModelAction * ModelChecker::process_rmw(ModelAction *act) {
1696 ModelAction *lastread = get_last_action(act->get_tid());
1697 lastread->process_rmw(act);
1698 if (act->is_rmw()) {
1699 if (lastread->get_reads_from())
1700 mo_graph->addRMWEdge(lastread->get_reads_from(), lastread);
1702 mo_graph->addRMWEdge(lastread->get_reads_from_promise(), lastread);
1703 mo_graph->commitChanges();
1709 * A helper function for ModelChecker::check_recency, to check if the current
1710 * thread is able to read from a different write/promise for 'params.maxreads'
1711 * number of steps and if that write/promise should become visible (i.e., is
1712 * ordered later in the modification order). This helps model memory liveness.
1714 * @param curr The current action. Must be a read.
1715 * @param rf The write/promise from which we plan to read
1716 * @param other_rf The write/promise from which we may read
1717 * @return True if we were able to read from other_rf for params.maxreads steps
1719 template <typename T, typename U>
1720 bool ModelChecker::should_read_instead(const ModelAction *curr, const T *rf, const U *other_rf) const
1722 /* Need a different write/promise */
1723 if (other_rf->equals(rf))
1726 /* Only look for "newer" writes/promises */
1727 if (!mo_graph->checkReachable(rf, other_rf))
1730 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1731 action_list_t *list = &(*thrd_lists)[id_to_int(curr->get_tid())];
1732 action_list_t::reverse_iterator rit = list->rbegin();
1733 ASSERT((*rit) == curr);
1734 /* Skip past curr */
1737 /* Does this write/promise work for everyone? */
1738 for (int i = 0; i < params.maxreads; i++, rit++) {
1739 ModelAction *act = *rit;
1740 if (!act->may_read_from(other_rf))
1747 * Checks whether a thread has read from the same write or Promise for too many
1748 * times without seeing the effects of a later write/Promise.
1751 * 1) there must a different write/promise that we could read from,
1752 * 2) we must have read from the same write/promise in excess of maxreads times,
1753 * 3) that other write/promise must have been in the reads_from set for maxreads times, and
1754 * 4) that other write/promise must be mod-ordered after the write/promise we are reading.
1756 * If so, we decide that the execution is no longer feasible.
1758 * @param curr The current action. Must be a read.
1759 * @param rf The ModelAction/Promise from which we might read.
1760 * @return True if the read should succeed; false otherwise
1762 template <typename T>
1763 bool ModelChecker::check_recency(ModelAction *curr, const T *rf) const
1765 if (!params.maxreads)
1768 //NOTE: Next check is just optimization, not really necessary....
1769 if (curr->get_node()->get_read_from_past_size() +
1770 curr->get_node()->get_read_from_promise_size() <= 1)
1773 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1774 int tid = id_to_int(curr->get_tid());
1775 ASSERT(tid < (int)thrd_lists->size());
1776 action_list_t *list = &(*thrd_lists)[tid];
1777 action_list_t::reverse_iterator rit = list->rbegin();
1778 ASSERT((*rit) == curr);
1779 /* Skip past curr */
1782 action_list_t::reverse_iterator ritcopy = rit;
1783 /* See if we have enough reads from the same value */
1784 for (int count = 0; count < params.maxreads; ritcopy++, count++) {
1785 if (ritcopy == list->rend())
1787 ModelAction *act = *ritcopy;
1788 if (!act->is_read())
1790 if (act->get_reads_from_promise() && !act->get_reads_from_promise()->equals(rf))
1792 if (act->get_reads_from() && !act->get_reads_from()->equals(rf))
1794 if (act->get_node()->get_read_from_past_size() +
1795 act->get_node()->get_read_from_promise_size() <= 1)
1798 for (int i = 0; i < curr->get_node()->get_read_from_past_size(); i++) {
1799 const ModelAction *write = curr->get_node()->get_read_from_past(i);
1800 if (should_read_instead(curr, rf, write))
1801 return false; /* liveness failure */
1803 for (int i = 0; i < curr->get_node()->get_read_from_promise_size(); i++) {
1804 const Promise *promise = curr->get_node()->get_read_from_promise(i);
1805 if (should_read_instead(curr, rf, promise))
1806 return false; /* liveness failure */
1812 * Updates the mo_graph with the constraints imposed from the current
1815 * Basic idea is the following: Go through each other thread and find
1816 * the last action that happened before our read. Two cases:
1818 * (1) The action is a write => that write must either occur before
1819 * the write we read from or be the write we read from.
1821 * (2) The action is a read => the write that that action read from
1822 * must occur before the write we read from or be the same write.
1824 * @param curr The current action. Must be a read.
1825 * @param rf The ModelAction or Promise that curr reads from. Must be a write.
1826 * @return True if modification order edges were added; false otherwise
1828 template <typename rf_type>
1829 bool ModelChecker::r_modification_order(ModelAction *curr, const rf_type *rf)
1831 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1834 ASSERT(curr->is_read());
1836 /* Last SC fence in the current thread */
1837 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1838 ModelAction *last_sc_write = NULL;
1839 if (curr->is_seqcst())
1840 last_sc_write = get_last_seq_cst_write(curr);
1842 /* Iterate over all threads */
1843 for (i = 0; i < thrd_lists->size(); i++) {
1844 /* Last SC fence in thread i */
1845 ModelAction *last_sc_fence_thread_local = NULL;
1846 if (int_to_id((int)i) != curr->get_tid())
1847 last_sc_fence_thread_local = get_last_seq_cst_fence(int_to_id(i), NULL);
1849 /* Last SC fence in thread i, before last SC fence in current thread */
1850 ModelAction *last_sc_fence_thread_before = NULL;
1851 if (last_sc_fence_local)
1852 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1854 /* Iterate over actions in thread, starting from most recent */
1855 action_list_t *list = &(*thrd_lists)[i];
1856 action_list_t::reverse_iterator rit;
1857 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1858 ModelAction *act = *rit;
1863 /* Don't want to add reflexive edges on 'rf' */
1864 if (act->equals(rf)) {
1865 if (act->happens_before(curr))
1871 if (act->is_write()) {
1872 /* C++, Section 29.3 statement 5 */
1873 if (curr->is_seqcst() && last_sc_fence_thread_local &&
1874 *act < *last_sc_fence_thread_local) {
1875 added = mo_graph->addEdge(act, rf) || added;
1878 /* C++, Section 29.3 statement 4 */
1879 else if (act->is_seqcst() && last_sc_fence_local &&
1880 *act < *last_sc_fence_local) {
1881 added = mo_graph->addEdge(act, rf) || added;
1884 /* C++, Section 29.3 statement 6 */
1885 else if (last_sc_fence_thread_before &&
1886 *act < *last_sc_fence_thread_before) {
1887 added = mo_graph->addEdge(act, rf) || added;
1892 /* C++, Section 29.3 statement 3 (second subpoint) */
1893 if (curr->is_seqcst() && last_sc_write && act == last_sc_write) {
1894 added = mo_graph->addEdge(act, rf) || added;
1899 * Include at most one act per-thread that "happens
1902 if (act->happens_before(curr)) {
1903 if (act->is_write()) {
1904 added = mo_graph->addEdge(act, rf) || added;
1906 const ModelAction *prevrf = act->get_reads_from();
1907 const Promise *prevrf_promise = act->get_reads_from_promise();
1909 if (!prevrf->equals(rf))
1910 added = mo_graph->addEdge(prevrf, rf) || added;
1911 } else if (!prevrf_promise->equals(rf)) {
1912 added = mo_graph->addEdge(prevrf_promise, rf) || added;
1921 * All compatible, thread-exclusive promises must be ordered after any
1922 * concrete loads from the same thread
1924 for (unsigned int i = 0; i < promises->size(); i++)
1925 if ((*promises)[i]->is_compatible_exclusive(curr))
1926 added = mo_graph->addEdge(rf, (*promises)[i]) || added;
1932 * Updates the mo_graph with the constraints imposed from the current write.
1934 * Basic idea is the following: Go through each other thread and find
1935 * the lastest action that happened before our write. Two cases:
1937 * (1) The action is a write => that write must occur before
1940 * (2) The action is a read => the write that that action read from
1941 * must occur before the current write.
1943 * This method also handles two other issues:
1945 * (I) Sequential Consistency: Making sure that if the current write is
1946 * seq_cst, that it occurs after the previous seq_cst write.
1948 * (II) Sending the write back to non-synchronizing reads.
1950 * @param curr The current action. Must be a write.
1951 * @param send_fv A vector for stashing reads to which we may pass our future
1952 * value. If NULL, then don't record any future values.
1953 * @return True if modification order edges were added; false otherwise
1955 bool ModelChecker::w_modification_order(ModelAction *curr, ModelVector<ModelAction *> *send_fv)
1957 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1960 ASSERT(curr->is_write());
1962 if (curr->is_seqcst()) {
1963 /* We have to at least see the last sequentially consistent write,
1964 so we are initialized. */
1965 ModelAction *last_seq_cst = get_last_seq_cst_write(curr);
1966 if (last_seq_cst != NULL) {
1967 added = mo_graph->addEdge(last_seq_cst, curr) || added;
1971 /* Last SC fence in the current thread */
1972 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1974 /* Iterate over all threads */
1975 for (i = 0; i < thrd_lists->size(); i++) {
1976 /* Last SC fence in thread i, before last SC fence in current thread */
1977 ModelAction *last_sc_fence_thread_before = NULL;
1978 if (last_sc_fence_local && int_to_id((int)i) != curr->get_tid())
1979 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1981 /* Iterate over actions in thread, starting from most recent */
1982 action_list_t *list = &(*thrd_lists)[i];
1983 action_list_t::reverse_iterator rit;
1984 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1985 ModelAction *act = *rit;
1988 * 1) If RMW and it actually read from something, then we
1989 * already have all relevant edges, so just skip to next
1992 * 2) If RMW and it didn't read from anything, we should
1993 * whatever edge we can get to speed up convergence.
1995 * 3) If normal write, we need to look at earlier actions, so
1996 * continue processing list.
1998 if (curr->is_rmw()) {
1999 if (curr->get_reads_from() != NULL)
2007 /* C++, Section 29.3 statement 7 */
2008 if (last_sc_fence_thread_before && act->is_write() &&
2009 *act < *last_sc_fence_thread_before) {
2010 added = mo_graph->addEdge(act, curr) || added;
2015 * Include at most one act per-thread that "happens
2018 if (act->happens_before(curr)) {
2020 * Note: if act is RMW, just add edge:
2022 * The following edge should be handled elsewhere:
2023 * readfrom(act) --mo--> act
2025 if (act->is_write())
2026 added = mo_graph->addEdge(act, curr) || added;
2027 else if (act->is_read()) {
2028 //if previous read accessed a null, just keep going
2029 if (act->get_reads_from() == NULL)
2031 added = mo_graph->addEdge(act->get_reads_from(), curr) || added;
2034 } else if (act->is_read() && !act->could_synchronize_with(curr) &&
2035 !act->same_thread(curr)) {
2036 /* We have an action that:
2037 (1) did not happen before us
2038 (2) is a read and we are a write
2039 (3) cannot synchronize with us
2040 (4) is in a different thread
2042 that read could potentially read from our write. Note that
2043 these checks are overly conservative at this point, we'll
2044 do more checks before actually removing the
2048 if (send_fv && thin_air_constraint_may_allow(curr, act)) {
2049 if (!is_infeasible())
2050 send_fv->push_back(act);
2051 else if (curr->is_rmw() && act->is_rmw() && curr->get_reads_from() && curr->get_reads_from() == act->get_reads_from())
2052 add_future_value(curr, act);
2059 * All compatible, thread-exclusive promises must be ordered after any
2060 * concrete stores to the same thread, or else they can be merged with
2063 for (unsigned int i = 0; i < promises->size(); i++)
2064 if ((*promises)[i]->is_compatible_exclusive(curr))
2065 added = mo_graph->addEdge(curr, (*promises)[i]) || added;
2070 /** Arbitrary reads from the future are not allowed. Section 29.3
2071 * part 9 places some constraints. This method checks one result of constraint
2072 * constraint. Others require compiler support. */
2073 bool ModelChecker::thin_air_constraint_may_allow(const ModelAction *writer, const ModelAction *reader) const
2075 if (!writer->is_rmw())
2078 if (!reader->is_rmw())
2081 for (const ModelAction *search = writer->get_reads_from(); search != NULL; search = search->get_reads_from()) {
2082 if (search == reader)
2084 if (search->get_tid() == reader->get_tid() &&
2085 search->happens_before(reader))
2093 * Arbitrary reads from the future are not allowed. Section 29.3 part 9 places
2094 * some constraints. This method checks one the following constraint (others
2095 * require compiler support):
2097 * If X --hb-> Y --mo-> Z, then X should not read from Z.
2099 bool ModelChecker::mo_may_allow(const ModelAction *writer, const ModelAction *reader)
2101 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, reader->get_location());
2103 /* Iterate over all threads */
2104 for (i = 0; i < thrd_lists->size(); i++) {
2105 const ModelAction *write_after_read = NULL;
2107 /* Iterate over actions in thread, starting from most recent */
2108 action_list_t *list = &(*thrd_lists)[i];
2109 action_list_t::reverse_iterator rit;
2110 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2111 ModelAction *act = *rit;
2113 /* Don't disallow due to act == reader */
2114 if (!reader->happens_before(act) || reader == act)
2116 else if (act->is_write())
2117 write_after_read = act;
2118 else if (act->is_read() && act->get_reads_from() != NULL)
2119 write_after_read = act->get_reads_from();
2122 if (write_after_read && write_after_read != writer && mo_graph->checkReachable(write_after_read, writer))
2129 * Finds the head(s) of the release sequence(s) containing a given ModelAction.
2130 * The ModelAction under consideration is expected to be taking part in
2131 * release/acquire synchronization as an object of the "reads from" relation.
2132 * Note that this can only provide release sequence support for RMW chains
2133 * which do not read from the future, as those actions cannot be traced until
2134 * their "promise" is fulfilled. Similarly, we may not even establish the
2135 * presence of a release sequence with certainty, as some modification order
2136 * constraints may be decided further in the future. Thus, this function
2137 * "returns" two pieces of data: a pass-by-reference vector of @a release_heads
2138 * and a boolean representing certainty.
2140 * @param rf The action that might be part of a release sequence. Must be a
2142 * @param release_heads A pass-by-reference style return parameter. After
2143 * execution of this function, release_heads will contain the heads of all the
2144 * relevant release sequences, if any exists with certainty
2145 * @param pending A pass-by-reference style return parameter which is only used
2146 * when returning false (i.e., uncertain). Returns most information regarding
2147 * an uncertain release sequence, including any write operations that might
2148 * break the sequence.
2149 * @return true, if the ModelChecker is certain that release_heads is complete;
2152 bool ModelChecker::release_seq_heads(const ModelAction *rf,
2153 rel_heads_list_t *release_heads,
2154 struct release_seq *pending) const
2156 /* Only check for release sequences if there are no cycles */
2157 if (mo_graph->checkForCycles())
2160 for ( ; rf != NULL; rf = rf->get_reads_from()) {
2161 ASSERT(rf->is_write());
2163 if (rf->is_release())
2164 release_heads->push_back(rf);
2165 else if (rf->get_last_fence_release())
2166 release_heads->push_back(rf->get_last_fence_release());
2168 break; /* End of RMW chain */
2170 /** @todo Need to be smarter here... In the linux lock
2171 * example, this will run to the beginning of the program for
2173 /** @todo The way to be smarter here is to keep going until 1
2174 * thread has a release preceded by an acquire and you've seen
2177 /* acq_rel RMW is a sufficient stopping condition */
2178 if (rf->is_acquire() && rf->is_release())
2179 return true; /* complete */
2182 /* read from future: need to settle this later */
2184 return false; /* incomplete */
2187 if (rf->is_release())
2188 return true; /* complete */
2190 /* else relaxed write
2191 * - check for fence-release in the same thread (29.8, stmt. 3)
2192 * - check modification order for contiguous subsequence
2193 * -> rf must be same thread as release */
2195 const ModelAction *fence_release = rf->get_last_fence_release();
2196 /* Synchronize with a fence-release unconditionally; we don't need to
2197 * find any more "contiguous subsequence..." for it */
2199 release_heads->push_back(fence_release);
2201 int tid = id_to_int(rf->get_tid());
2202 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, rf->get_location());
2203 action_list_t *list = &(*thrd_lists)[tid];
2204 action_list_t::const_reverse_iterator rit;
2206 /* Find rf in the thread list */
2207 rit = std::find(list->rbegin(), list->rend(), rf);
2208 ASSERT(rit != list->rend());
2210 /* Find the last {write,fence}-release */
2211 for (; rit != list->rend(); rit++) {
2212 if (fence_release && *(*rit) < *fence_release)
2214 if ((*rit)->is_release())
2217 if (rit == list->rend()) {
2218 /* No write-release in this thread */
2219 return true; /* complete */
2220 } else if (fence_release && *(*rit) < *fence_release) {
2221 /* The fence-release is more recent (and so, "stronger") than
2222 * the most recent write-release */
2223 return true; /* complete */
2224 } /* else, need to establish contiguous release sequence */
2225 ModelAction *release = *rit;
2227 ASSERT(rf->same_thread(release));
2229 pending->writes.clear();
2231 bool certain = true;
2232 for (unsigned int i = 0; i < thrd_lists->size(); i++) {
2233 if (id_to_int(rf->get_tid()) == (int)i)
2235 list = &(*thrd_lists)[i];
2237 /* Can we ensure no future writes from this thread may break
2238 * the release seq? */
2239 bool future_ordered = false;
2241 ModelAction *last = get_last_action(int_to_id(i));
2242 Thread *th = get_thread(int_to_id(i));
2243 if ((last && rf->happens_before(last)) ||
2246 future_ordered = true;
2248 ASSERT(!th->is_model_thread() || future_ordered);
2250 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2251 const ModelAction *act = *rit;
2252 /* Reach synchronization -> this thread is complete */
2253 if (act->happens_before(release))
2255 if (rf->happens_before(act)) {
2256 future_ordered = true;
2260 /* Only non-RMW writes can break release sequences */
2261 if (!act->is_write() || act->is_rmw())
2264 /* Check modification order */
2265 if (mo_graph->checkReachable(rf, act)) {
2266 /* rf --mo--> act */
2267 future_ordered = true;
2270 if (mo_graph->checkReachable(act, release))
2271 /* act --mo--> release */
2273 if (mo_graph->checkReachable(release, act) &&
2274 mo_graph->checkReachable(act, rf)) {
2275 /* release --mo-> act --mo--> rf */
2276 return true; /* complete */
2278 /* act may break release sequence */
2279 pending->writes.push_back(act);
2282 if (!future_ordered)
2283 certain = false; /* This thread is uncertain */
2287 release_heads->push_back(release);
2288 pending->writes.clear();
2290 pending->release = release;
2297 * An interface for getting the release sequence head(s) with which a
2298 * given ModelAction must synchronize. This function only returns a non-empty
2299 * result when it can locate a release sequence head with certainty. Otherwise,
2300 * it may mark the internal state of the ModelChecker so that it will handle
2301 * the release sequence at a later time, causing @a acquire to update its
2302 * synchronization at some later point in execution.
2304 * @param acquire The 'acquire' action that may synchronize with a release
2306 * @param read The read action that may read from a release sequence; this may
2307 * be the same as acquire, or else an earlier action in the same thread (i.e.,
2308 * when 'acquire' is a fence-acquire)
2309 * @param release_heads A pass-by-reference return parameter. Will be filled
2310 * with the head(s) of the release sequence(s), if they exists with certainty.
2311 * @see ModelChecker::release_seq_heads
2313 void ModelChecker::get_release_seq_heads(ModelAction *acquire,
2314 ModelAction *read, rel_heads_list_t *release_heads)
2316 const ModelAction *rf = read->get_reads_from();
2317 struct release_seq *sequence = (struct release_seq *)snapshot_calloc(1, sizeof(struct release_seq));
2318 sequence->acquire = acquire;
2319 sequence->read = read;
2321 if (!release_seq_heads(rf, release_heads, sequence)) {
2322 /* add act to 'lazy checking' list */
2323 pending_rel_seqs->push_back(sequence);
2325 snapshot_free(sequence);
2330 * Attempt to resolve all stashed operations that might synchronize with a
2331 * release sequence for a given location. This implements the "lazy" portion of
2332 * determining whether or not a release sequence was contiguous, since not all
2333 * modification order information is present at the time an action occurs.
2335 * @param location The location/object that should be checked for release
2336 * sequence resolutions. A NULL value means to check all locations.
2337 * @param work_queue The work queue to which to add work items as they are
2339 * @return True if any updates occurred (new synchronization, new mo_graph
2342 bool ModelChecker::resolve_release_sequences(void *location, work_queue_t *work_queue)
2344 bool updated = false;
2345 SnapVector<struct release_seq *>::iterator it = pending_rel_seqs->begin();
2346 while (it != pending_rel_seqs->end()) {
2347 struct release_seq *pending = *it;
2348 ModelAction *acquire = pending->acquire;
2349 const ModelAction *read = pending->read;
2351 /* Only resolve sequences on the given location, if provided */
2352 if (location && read->get_location() != location) {
2357 const ModelAction *rf = read->get_reads_from();
2358 rel_heads_list_t release_heads;
2360 complete = release_seq_heads(rf, &release_heads, pending);
2361 for (unsigned int i = 0; i < release_heads.size(); i++) {
2362 if (!acquire->has_synchronized_with(release_heads[i])) {
2363 if (acquire->synchronize_with(release_heads[i]))
2366 set_bad_synchronization();
2371 /* Re-check all pending release sequences */
2372 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
2373 /* Re-check read-acquire for mo_graph edges */
2374 if (acquire->is_read())
2375 work_queue->push_back(MOEdgeWorkEntry(acquire));
2377 /* propagate synchronization to later actions */
2378 action_list_t::reverse_iterator rit = action_trace->rbegin();
2379 for (; (*rit) != acquire; rit++) {
2380 ModelAction *propagate = *rit;
2381 if (acquire->happens_before(propagate)) {
2382 propagate->synchronize_with(acquire);
2383 /* Re-check 'propagate' for mo_graph edges */
2384 work_queue->push_back(MOEdgeWorkEntry(propagate));
2389 it = pending_rel_seqs->erase(it);
2390 snapshot_free(pending);
2396 // If we resolved promises or data races, see if we have realized a data race.
2403 * Performs various bookkeeping operations for the current ModelAction. For
2404 * instance, adds action to the per-object, per-thread action vector and to the
2405 * action trace list of all thread actions.
2407 * @param act is the ModelAction to add.
2409 void ModelChecker::add_action_to_lists(ModelAction *act)
2411 int tid = id_to_int(act->get_tid());
2412 ModelAction *uninit = NULL;
2414 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
2415 if (list->empty() && act->is_atomic_var()) {
2416 uninit = get_uninitialized_action(act);
2417 uninit_id = id_to_int(uninit->get_tid());
2418 list->push_front(uninit);
2420 list->push_back(act);
2422 action_trace->push_back(act);
2424 action_trace->push_front(uninit);
2426 SnapVector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, act->get_location());
2427 if (tid >= (int)vec->size())
2428 vec->resize(priv->next_thread_id);
2429 (*vec)[tid].push_back(act);
2431 (*vec)[uninit_id].push_front(uninit);
2433 if ((int)thrd_last_action->size() <= tid)
2434 thrd_last_action->resize(get_num_threads());
2435 (*thrd_last_action)[tid] = act;
2437 (*thrd_last_action)[uninit_id] = uninit;
2439 if (act->is_fence() && act->is_release()) {
2440 if ((int)thrd_last_fence_release->size() <= tid)
2441 thrd_last_fence_release->resize(get_num_threads());
2442 (*thrd_last_fence_release)[tid] = act;
2445 if (act->is_wait()) {
2446 void *mutex_loc = (void *) act->get_value();
2447 get_safe_ptr_action(obj_map, mutex_loc)->push_back(act);
2449 SnapVector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, mutex_loc);
2450 if (tid >= (int)vec->size())
2451 vec->resize(priv->next_thread_id);
2452 (*vec)[tid].push_back(act);
2457 * @brief Get the last action performed by a particular Thread
2458 * @param tid The thread ID of the Thread in question
2459 * @return The last action in the thread
2461 ModelAction * ModelChecker::get_last_action(thread_id_t tid) const
2463 int threadid = id_to_int(tid);
2464 if (threadid < (int)thrd_last_action->size())
2465 return (*thrd_last_action)[id_to_int(tid)];
2471 * @brief Get the last fence release performed by a particular Thread
2472 * @param tid The thread ID of the Thread in question
2473 * @return The last fence release in the thread, if one exists; NULL otherwise
2475 ModelAction * ModelChecker::get_last_fence_release(thread_id_t tid) const
2477 int threadid = id_to_int(tid);
2478 if (threadid < (int)thrd_last_fence_release->size())
2479 return (*thrd_last_fence_release)[id_to_int(tid)];
2485 * Gets the last memory_order_seq_cst write (in the total global sequence)
2486 * performed on a particular object (i.e., memory location), not including the
2488 * @param curr The current ModelAction; also denotes the object location to
2490 * @return The last seq_cst write
2492 ModelAction * ModelChecker::get_last_seq_cst_write(ModelAction *curr) const
2494 void *location = curr->get_location();
2495 action_list_t *list = get_safe_ptr_action(obj_map, location);
2496 /* Find: max({i in dom(S) | seq_cst(t_i) && isWrite(t_i) && samevar(t_i, t)}) */
2497 action_list_t::reverse_iterator rit;
2498 for (rit = list->rbegin(); (*rit) != curr; rit++)
2500 rit++; /* Skip past curr */
2501 for ( ; rit != list->rend(); rit++)
2502 if ((*rit)->is_write() && (*rit)->is_seqcst())
2508 * Gets the last memory_order_seq_cst fence (in the total global sequence)
2509 * performed in a particular thread, prior to a particular fence.
2510 * @param tid The ID of the thread to check
2511 * @param before_fence The fence from which to begin the search; if NULL, then
2512 * search for the most recent fence in the thread.
2513 * @return The last prior seq_cst fence in the thread, if exists; otherwise, NULL
2515 ModelAction * ModelChecker::get_last_seq_cst_fence(thread_id_t tid, const ModelAction *before_fence) const
2517 /* All fences should have NULL location */
2518 action_list_t *list = get_safe_ptr_action(obj_map, NULL);
2519 action_list_t::reverse_iterator rit = list->rbegin();
2522 for (; rit != list->rend(); rit++)
2523 if (*rit == before_fence)
2526 ASSERT(*rit == before_fence);
2530 for (; rit != list->rend(); rit++)
2531 if ((*rit)->is_fence() && (tid == (*rit)->get_tid()) && (*rit)->is_seqcst())
2537 * Gets the last unlock operation performed on a particular mutex (i.e., memory
2538 * location). This function identifies the mutex according to the current
2539 * action, which is presumed to perform on the same mutex.
2540 * @param curr The current ModelAction; also denotes the object location to
2542 * @return The last unlock operation
2544 ModelAction * ModelChecker::get_last_unlock(ModelAction *curr) const
2546 void *location = curr->get_location();
2547 action_list_t *list = get_safe_ptr_action(obj_map, location);
2548 /* Find: max({i in dom(S) | isUnlock(t_i) && samevar(t_i, t)}) */
2549 action_list_t::reverse_iterator rit;
2550 for (rit = list->rbegin(); rit != list->rend(); rit++)
2551 if ((*rit)->is_unlock() || (*rit)->is_wait())
2556 ModelAction * ModelChecker::get_parent_action(thread_id_t tid) const
2558 ModelAction *parent = get_last_action(tid);
2560 parent = get_thread(tid)->get_creation();
2565 * Returns the clock vector for a given thread.
2566 * @param tid The thread whose clock vector we want
2567 * @return Desired clock vector
2569 ClockVector * ModelChecker::get_cv(thread_id_t tid) const
2571 return get_parent_action(tid)->get_cv();
2575 * @brief Find the promise (if any) to resolve for the current action and
2576 * remove it from the pending promise vector
2577 * @param curr The current ModelAction. Should be a write.
2578 * @return The Promise to resolve, if any; otherwise NULL
2580 Promise * ModelChecker::pop_promise_to_resolve(const ModelAction *curr)
2582 for (unsigned int i = 0; i < promises->size(); i++)
2583 if (curr->get_node()->get_promise(i)) {
2584 Promise *ret = (*promises)[i];
2585 promises->erase(promises->begin() + i);
2592 * Resolve a Promise with a current write.
2593 * @param write The ModelAction that is fulfilling Promises
2594 * @param promise The Promise to resolve
2595 * @return True if the Promise was successfully resolved; false otherwise
2597 bool ModelChecker::resolve_promise(ModelAction *write, Promise *promise)
2599 ModelVector<ModelAction *> actions_to_check;
2601 for (unsigned int i = 0; i < promise->get_num_readers(); i++) {
2602 ModelAction *read = promise->get_reader(i);
2603 read_from(read, write);
2604 actions_to_check.push_back(read);
2606 /* Make sure the promise's value matches the write's value */
2607 ASSERT(promise->is_compatible(write) && promise->same_value(write));
2608 if (!mo_graph->resolvePromise(promise, write))
2609 priv->failed_promise = true;
2612 * @todo It is possible to end up in an inconsistent state, where a
2613 * "resolved" promise may still be referenced if
2614 * CycleGraph::resolvePromise() failed, so don't delete 'promise'.
2616 * Note that the inconsistency only matters when dumping mo_graph to
2622 //Check whether reading these writes has made threads unable to
2624 for (unsigned int i = 0; i < actions_to_check.size(); i++) {
2625 ModelAction *read = actions_to_check[i];
2626 mo_check_promises(read, true);
2633 * Compute the set of promises that could potentially be satisfied by this
2634 * action. Note that the set computation actually appears in the Node, not in
2636 * @param curr The ModelAction that may satisfy promises
2638 void ModelChecker::compute_promises(ModelAction *curr)
2640 for (unsigned int i = 0; i < promises->size(); i++) {
2641 Promise *promise = (*promises)[i];
2642 if (!promise->is_compatible(curr) || !promise->same_value(curr))
2645 bool satisfy = true;
2646 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
2647 const ModelAction *act = promise->get_reader(j);
2648 if (act->happens_before(curr) ||
2649 act->could_synchronize_with(curr)) {
2655 curr->get_node()->set_promise(i);
2659 /** Checks promises in response to change in ClockVector Threads. */
2660 void ModelChecker::check_promises(thread_id_t tid, ClockVector *old_cv, ClockVector *merge_cv)
2662 for (unsigned int i = 0; i < promises->size(); i++) {
2663 Promise *promise = (*promises)[i];
2664 if (!promise->thread_is_available(tid))
2666 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
2667 const ModelAction *act = promise->get_reader(j);
2668 if ((!old_cv || !old_cv->synchronized_since(act)) &&
2669 merge_cv->synchronized_since(act)) {
2670 if (promise->eliminate_thread(tid)) {
2671 /* Promise has failed */
2672 priv->failed_promise = true;
2680 void ModelChecker::check_promises_thread_disabled()
2682 for (unsigned int i = 0; i < promises->size(); i++) {
2683 Promise *promise = (*promises)[i];
2684 if (promise->has_failed()) {
2685 priv->failed_promise = true;
2692 * @brief Checks promises in response to addition to modification order for
2695 * We test whether threads are still available for satisfying promises after an
2696 * addition to our modification order constraints. Those that are unavailable
2697 * are "eliminated". Once all threads are eliminated from satisfying a promise,
2698 * that promise has failed.
2700 * @param act The ModelAction which updated the modification order
2701 * @param is_read_check Should be true if act is a read and we must check for
2702 * updates to the store from which it read (there is a distinction here for
2703 * RMW's, which are both a load and a store)
2705 void ModelChecker::mo_check_promises(const ModelAction *act, bool is_read_check)
2707 const ModelAction *write = is_read_check ? act->get_reads_from() : act;
2709 for (unsigned int i = 0; i < promises->size(); i++) {
2710 Promise *promise = (*promises)[i];
2712 // Is this promise on the same location?
2713 if (!promise->same_location(write))
2716 for (unsigned int j = 0; j < promise->get_num_readers(); j++) {
2717 const ModelAction *pread = promise->get_reader(j);
2718 if (!pread->happens_before(act))
2720 if (mo_graph->checkPromise(write, promise)) {
2721 priv->failed_promise = true;
2727 // Don't do any lookups twice for the same thread
2728 if (!promise->thread_is_available(act->get_tid()))
2731 if (mo_graph->checkReachable(promise, write)) {
2732 if (mo_graph->checkPromise(write, promise)) {
2733 priv->failed_promise = true;
2741 * Compute the set of writes that may break the current pending release
2742 * sequence. This information is extracted from previou release sequence
2745 * @param curr The current ModelAction. Must be a release sequence fixup
2748 void ModelChecker::compute_relseq_breakwrites(ModelAction *curr)
2750 if (pending_rel_seqs->empty())
2753 struct release_seq *pending = pending_rel_seqs->back();
2754 for (unsigned int i = 0; i < pending->writes.size(); i++) {
2755 const ModelAction *write = pending->writes[i];
2756 curr->get_node()->add_relseq_break(write);
2759 /* NULL means don't break the sequence; just synchronize */
2760 curr->get_node()->add_relseq_break(NULL);
2764 * Build up an initial set of all past writes that this 'read' action may read
2765 * from, as well as any previously-observed future values that must still be valid.
2767 * @param curr is the current ModelAction that we are exploring; it must be a
2770 void ModelChecker::build_may_read_from(ModelAction *curr)
2772 SnapVector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
2774 ASSERT(curr->is_read());
2776 ModelAction *last_sc_write = NULL;
2778 if (curr->is_seqcst())
2779 last_sc_write = get_last_seq_cst_write(curr);
2781 /* Iterate over all threads */
2782 for (i = 0; i < thrd_lists->size(); i++) {
2783 /* Iterate over actions in thread, starting from most recent */
2784 action_list_t *list = &(*thrd_lists)[i];
2785 action_list_t::reverse_iterator rit;
2786 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2787 ModelAction *act = *rit;
2789 /* Only consider 'write' actions */
2790 if (!act->is_write() || act == curr)
2793 /* Don't consider more than one seq_cst write if we are a seq_cst read. */
2794 bool allow_read = true;
2796 if (curr->is_seqcst() && (act->is_seqcst() || (last_sc_write != NULL && act->happens_before(last_sc_write))) && act != last_sc_write)
2798 else if (curr->get_sleep_flag() && !curr->is_seqcst() && !sleep_can_read_from(curr, act))
2802 /* Only add feasible reads */
2803 mo_graph->startChanges();
2804 r_modification_order(curr, act);
2805 if (!is_infeasible())
2806 curr->get_node()->add_read_from_past(act);
2807 mo_graph->rollbackChanges();
2810 /* Include at most one act per-thread that "happens before" curr */
2811 if (act->happens_before(curr))
2816 /* Inherit existing, promised future values */
2817 for (i = 0; i < promises->size(); i++) {
2818 const Promise *promise = (*promises)[i];
2819 const ModelAction *promise_read = promise->get_reader(0);
2820 if (promise_read->same_var(curr)) {
2821 /* Only add feasible future-values */
2822 mo_graph->startChanges();
2823 r_modification_order(curr, promise);
2824 if (!is_infeasible())
2825 curr->get_node()->add_read_from_promise(promise_read);
2826 mo_graph->rollbackChanges();
2830 /* We may find no valid may-read-from only if the execution is doomed */
2831 if (!curr->get_node()->read_from_size()) {
2832 priv->no_valid_reads = true;
2836 if (DBG_ENABLED()) {
2837 model_print("Reached read action:\n");
2839 model_print("Printing read_from_past\n");
2840 curr->get_node()->print_read_from_past();
2841 model_print("End printing read_from_past\n");
2845 bool ModelChecker::sleep_can_read_from(ModelAction *curr, const ModelAction *write)
2847 for ( ; write != NULL; write = write->get_reads_from()) {
2848 /* UNINIT actions don't have a Node, and they never sleep */
2849 if (write->is_uninitialized())
2851 Node *prevnode = write->get_node()->get_parent();
2853 bool thread_sleep = prevnode->enabled_status(curr->get_tid()) == THREAD_SLEEP_SET;
2854 if (write->is_release() && thread_sleep)
2856 if (!write->is_rmw())
2863 * @brief Get an action representing an uninitialized atomic
2865 * This function may create a new one or try to retrieve one from the NodeStack
2867 * @param curr The current action, which prompts the creation of an UNINIT action
2868 * @return A pointer to the UNINIT ModelAction
2870 ModelAction * ModelChecker::get_uninitialized_action(const ModelAction *curr) const
2872 Node *node = curr->get_node();
2873 ModelAction *act = node->get_uninit_action();
2875 act = new ModelAction(ATOMIC_UNINIT, std::memory_order_relaxed, curr->get_location(), model->params.uninitvalue, model_thread);
2876 node->set_uninit_action(act);
2878 act->create_cv(NULL);
2882 static void print_list(action_list_t *list)
2884 action_list_t::iterator it;
2886 model_print("---------------------------------------------------------------------\n");
2888 unsigned int hash = 0;
2890 for (it = list->begin(); it != list->end(); it++) {
2891 const ModelAction *act = *it;
2892 if (act->get_seq_number() > 0)
2894 hash = hash^(hash<<3)^((*it)->hash());
2896 model_print("HASH %u\n", hash);
2897 model_print("---------------------------------------------------------------------\n");
2900 #if SUPPORT_MOD_ORDER_DUMP
2901 void ModelChecker::dumpGraph(char *filename) const
2904 sprintf(buffer, "%s.dot", filename);
2905 FILE *file = fopen(buffer, "w");
2906 fprintf(file, "digraph %s {\n", filename);
2907 mo_graph->dumpNodes(file);
2908 ModelAction **thread_array = (ModelAction **)model_calloc(1, sizeof(ModelAction *) * get_num_threads());
2910 for (action_list_t::iterator it = action_trace->begin(); it != action_trace->end(); it++) {
2911 ModelAction *act = *it;
2912 if (act->is_read()) {
2913 mo_graph->dot_print_node(file, act);
2914 if (act->get_reads_from())
2915 mo_graph->dot_print_edge(file,
2916 act->get_reads_from(),
2918 "label=\"rf\", color=red, weight=2");
2920 mo_graph->dot_print_edge(file,
2921 act->get_reads_from_promise(),
2923 "label=\"rf\", color=red");
2925 if (thread_array[act->get_tid()]) {
2926 mo_graph->dot_print_edge(file,
2927 thread_array[id_to_int(act->get_tid())],
2929 "label=\"sb\", color=blue, weight=400");
2932 thread_array[act->get_tid()] = act;
2934 fprintf(file, "}\n");
2935 model_free(thread_array);
2940 /** @brief Prints an execution trace summary. */
2941 void ModelChecker::print_summary() const
2943 #if SUPPORT_MOD_ORDER_DUMP
2944 char buffername[100];
2945 sprintf(buffername, "exec%04u", stats.num_total);
2946 mo_graph->dumpGraphToFile(buffername);
2947 sprintf(buffername, "graph%04u", stats.num_total);
2948 dumpGraph(buffername);
2951 model_print("Execution %d:", stats.num_total);
2952 if (isfeasibleprefix()) {
2953 if (scheduler->all_threads_sleeping())
2954 model_print(" SLEEP-SET REDUNDANT");
2957 print_infeasibility(" INFEASIBLE");
2958 print_list(action_trace);
2960 if (!promises->empty()) {
2961 model_print("Pending promises:\n");
2962 for (unsigned int i = 0; i < promises->size(); i++) {
2963 model_print(" [P%u] ", i);
2964 (*promises)[i]->print();
2971 * Add a Thread to the system for the first time. Should only be called once
2973 * @param t The Thread to add
2975 void ModelChecker::add_thread(Thread *t)
2977 thread_map->put(id_to_int(t->get_id()), t);
2978 scheduler->add_thread(t);
2982 * @brief Get a Thread reference by its ID
2983 * @param tid The Thread's ID
2984 * @return A Thread reference
2986 Thread * ModelChecker::get_thread(thread_id_t tid) const
2988 return thread_map->get(id_to_int(tid));
2992 * @brief Get a reference to the Thread in which a ModelAction was executed
2993 * @param act The ModelAction
2994 * @return A Thread reference
2996 Thread * ModelChecker::get_thread(const ModelAction *act) const
2998 return get_thread(act->get_tid());
3002 * @brief Get a Promise's "promise number"
3004 * A "promise number" is an index number that is unique to a promise, valid
3005 * only for a specific snapshot of an execution trace. Promises may come and go
3006 * as they are generated an resolved, so an index only retains meaning for the
3009 * @param promise The Promise to check
3010 * @return The promise index, if the promise still is valid; otherwise -1
3012 int ModelChecker::get_promise_number(const Promise *promise) const
3014 for (unsigned int i = 0; i < promises->size(); i++)
3015 if ((*promises)[i] == promise)
3022 * @brief Check if a Thread is currently enabled
3023 * @param t The Thread to check
3024 * @return True if the Thread is currently enabled
3026 bool ModelChecker::is_enabled(Thread *t) const
3028 return scheduler->is_enabled(t);
3032 * @brief Check if a Thread is currently enabled
3033 * @param tid The ID of the Thread to check
3034 * @return True if the Thread is currently enabled
3036 bool ModelChecker::is_enabled(thread_id_t tid) const
3038 return scheduler->is_enabled(tid);
3042 * Switch from a model-checker context to a user-thread context. This is the
3043 * complement of ModelChecker::switch_to_master and must be called from the
3044 * model-checker context
3046 * @param thread The user-thread to switch to
3048 void ModelChecker::switch_from_master(Thread *thread)
3050 scheduler->set_current_thread(thread);
3051 Thread::swap(&system_context, thread);
3055 * Switch from a user-context to the "master thread" context (a.k.a. system
3056 * context). This switch is made with the intention of exploring a particular
3057 * model-checking action (described by a ModelAction object). Must be called
3058 * from a user-thread context.
3060 * @param act The current action that will be explored. May be NULL only if
3061 * trace is exiting via an assertion (see ModelChecker::set_assert and
3062 * ModelChecker::has_asserted).
3063 * @return Return the value returned by the current action
3065 uint64_t ModelChecker::switch_to_master(ModelAction *act)
3068 Thread *old = thread_current();
3069 scheduler->set_current_thread(NULL);
3070 ASSERT(!old->get_pending());
3071 old->set_pending(act);
3072 if (Thread::swap(old, &system_context) < 0) {
3073 perror("swap threads");
3076 return old->get_return_value();
3080 * Takes the next step in the execution, if possible.
3081 * @param curr The current step to take
3082 * @return Returns the next Thread to run, if any; NULL if this execution
3085 Thread * ModelChecker::take_step(ModelAction *curr)
3087 Thread *curr_thrd = get_thread(curr);
3088 ASSERT(curr_thrd->get_state() == THREAD_READY);
3090 curr = check_current_action(curr);
3092 /* Infeasible -> don't take any more steps */
3093 if (is_infeasible())
3095 else if (isfeasibleprefix() && have_bug_reports()) {
3100 if (params.bound != 0 && priv->used_sequence_numbers > params.bound)
3103 if (curr_thrd->is_blocked() || curr_thrd->is_complete())
3104 scheduler->remove_thread(curr_thrd);
3106 Thread *next_thrd = NULL;
3108 next_thrd = action_select_next_thread(curr);
3110 next_thrd = get_next_thread();
3112 DEBUG("(%d, %d)\n", curr_thrd ? id_to_int(curr_thrd->get_id()) : -1,
3113 next_thrd ? id_to_int(next_thrd->get_id()) : -1);
3118 /** Wrapper to run the user's main function, with appropriate arguments */
3119 void user_main_wrapper(void *)
3121 user_main(model->params.argc, model->params.argv);
3124 /** @brief Run ModelChecker for the user program */
3125 void ModelChecker::run()
3129 Thread *t = new Thread(&user_thread, &user_main_wrapper, NULL, NULL);
3134 * Stash next pending action(s) for thread(s). There
3135 * should only need to stash one thread's action--the
3136 * thread which just took a step--plus the first step
3137 * for any newly-created thread
3139 for (unsigned int i = 0; i < get_num_threads(); i++) {
3140 thread_id_t tid = int_to_id(i);
3141 Thread *thr = get_thread(tid);
3142 if (!thr->is_model_thread() && !thr->is_complete() && !thr->get_pending()) {
3143 switch_from_master(thr);
3144 if (thr->is_waiting_on(thr))
3145 assert_bug("Deadlock detected");
3149 /* Catch assertions from prior take_step or from
3150 * between-ModelAction bugs (e.g., data races) */
3154 /* Consume the next action for a Thread */
3155 ModelAction *curr = t->get_pending();
3156 t->set_pending(NULL);
3157 t = take_step(curr);
3158 } while (t && !t->is_model_thread());
3161 * Launch end-of-execution release sequence fixups only when
3162 * the execution is otherwise feasible AND there are:
3164 * (1) pending release sequences
3165 * (2) pending assertions that could be invalidated by a change
3166 * in clock vectors (i.e., data races)
3167 * (3) no pending promises
3169 while (!pending_rel_seqs->empty() &&
3170 is_feasible_prefix_ignore_relseq() &&
3171 !unrealizedraces.empty()) {
3172 model_print("*** WARNING: release sequence fixup action "
3173 "(%zu pending release seuqence(s)) ***\n",
3174 pending_rel_seqs->size());
3175 ModelAction *fixup = new ModelAction(MODEL_FIXUP_RELSEQ,
3176 std::memory_order_seq_cst, NULL, VALUE_NONE,
3180 } while (next_execution());
3182 model_print("******* Model-checking complete: *******\n");