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 std::vector< bug_message *, SnapshotAlloc<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 *, std::vector<action_list_t> *, uintptr_t, 4 >()),
89 promises(new std::vector< Promise *, SnapshotAlloc<Promise *> >()),
90 futurevalues(new std::vector< struct PendingFutureValue, SnapshotAlloc<struct PendingFutureValue> >()),
91 pending_rel_seqs(new std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >()),
92 thrd_last_action(new std::vector< ModelAction *, SnapshotAlloc<ModelAction *> >(1)),
93 thrd_last_fence_release(new std::vector< ModelAction *, SnapshotAlloc<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 std::vector<action_list_t> * get_safe_ptr_vect_action(HashTable<void *, std::vector<action_list_t> *, uintptr_t, 4> * hash, void * ptr)
142 std::vector<action_list_t> *tmp = hash->get(ptr);
144 tmp = new std::vector<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 Choose the next thread to execute.
210 * This function chooses the next thread that should execute. It can force the
211 * adjacency of read/write portions of a RMW action, force THREAD_CREATE to be
212 * followed by a THREAD_START, or it can enforce execution replay/backtracking.
213 * The model-checker may have no preference regarding the next thread (i.e.,
214 * when exploring a new execution ordering), in which case we defer to the
217 * @param curr Optional: The current ModelAction. Only used if non-NULL and it
218 * might guide the choice of next thread (i.e., THREAD_CREATE should be
219 * followed by THREAD_START, or ATOMIC_RMWR followed by ATOMIC_{RMW,RMWC})
220 * @return The next chosen thread to run, if any exist. Or else if no threads
221 * remain to be executed, return NULL.
223 Thread * ModelChecker::get_next_thread(ModelAction *curr)
228 /* Do not split atomic actions. */
230 return get_thread(curr);
231 else if (curr->get_type() == THREAD_CREATE)
232 return curr->get_thread_operand();
236 * Have we completed exploring the preselected path? Then let the
240 return scheduler->select_next_thread();
242 /* Else, we are trying to replay an execution */
243 ModelAction *next = node_stack->get_next()->get_action();
245 if (next == diverge) {
246 if (earliest_diverge == NULL || *diverge < *earliest_diverge)
247 earliest_diverge = diverge;
249 Node *nextnode = next->get_node();
250 Node *prevnode = nextnode->get_parent();
251 scheduler->update_sleep_set(prevnode);
253 /* Reached divergence point */
254 if (nextnode->increment_misc()) {
255 /* The next node will try to satisfy a different misc_index values. */
256 tid = next->get_tid();
257 node_stack->pop_restofstack(2);
258 } else if (nextnode->increment_promise()) {
259 /* The next node will try to satisfy a different set of promises. */
260 tid = next->get_tid();
261 node_stack->pop_restofstack(2);
262 } else if (nextnode->increment_read_from()) {
263 /* The next node will read from a different value. */
264 tid = next->get_tid();
265 node_stack->pop_restofstack(2);
266 } else if (nextnode->increment_future_value()) {
267 /* The next node will try to read from a different future value. */
268 tid = next->get_tid();
269 node_stack->pop_restofstack(2);
270 } else if (nextnode->increment_relseq_break()) {
271 /* The next node will try to resolve a release sequence differently */
272 tid = next->get_tid();
273 node_stack->pop_restofstack(2);
276 /* Make a different thread execute for next step */
277 scheduler->add_sleep(get_thread(next->get_tid()));
278 tid = prevnode->get_next_backtrack();
279 /* Make sure the backtracked thread isn't sleeping. */
280 node_stack->pop_restofstack(1);
281 if (diverge == earliest_diverge) {
282 earliest_diverge = prevnode->get_action();
285 /* Start the round robin scheduler from this thread id */
286 scheduler->set_scheduler_thread(tid);
287 /* The correct sleep set is in the parent node. */
290 DEBUG("*** Divergence point ***\n");
294 tid = next->get_tid();
296 DEBUG("*** ModelChecker chose next thread = %d ***\n", id_to_int(tid));
297 ASSERT(tid != THREAD_ID_T_NONE);
298 return thread_map->get(id_to_int(tid));
302 * We need to know what the next actions of all threads in the sleep
303 * set will be. This method computes them and stores the actions at
304 * the corresponding thread object's pending action.
307 void ModelChecker::execute_sleep_set()
309 for (unsigned int i = 0; i < get_num_threads(); i++) {
310 thread_id_t tid = int_to_id(i);
311 Thread *thr = get_thread(tid);
312 if (scheduler->is_sleep_set(thr) && thr->get_pending()) {
313 thr->get_pending()->set_sleep_flag();
319 * @brief Should the current action wake up a given thread?
321 * @param curr The current action
322 * @param thread The thread that we might wake up
323 * @return True, if we should wake up the sleeping thread; false otherwise
325 bool ModelChecker::should_wake_up(const ModelAction *curr, const Thread *thread) const
327 const ModelAction *asleep = thread->get_pending();
328 /* Don't allow partial RMW to wake anyone up */
331 /* Synchronizing actions may have been backtracked */
332 if (asleep->could_synchronize_with(curr))
334 /* All acquire/release fences and fence-acquire/store-release */
335 if (asleep->is_fence() && asleep->is_acquire() && curr->is_release())
337 /* Fence-release + store can awake load-acquire on the same location */
338 if (asleep->is_read() && asleep->is_acquire() && curr->same_var(asleep) && curr->is_write()) {
339 ModelAction *fence_release = get_last_fence_release(curr->get_tid());
340 if (fence_release && *(get_last_action(thread->get_id())) < *fence_release)
346 void ModelChecker::wake_up_sleeping_actions(ModelAction *curr)
348 for (unsigned int i = 0; i < get_num_threads(); i++) {
349 Thread *thr = get_thread(int_to_id(i));
350 if (scheduler->is_sleep_set(thr)) {
351 if (should_wake_up(curr, thr))
352 /* Remove this thread from sleep set */
353 scheduler->remove_sleep(thr);
358 /** @brief Alert the model-checker that an incorrectly-ordered
359 * synchronization was made */
360 void ModelChecker::set_bad_synchronization()
362 priv->bad_synchronization = true;
366 * Check whether the current trace has triggered an assertion which should halt
369 * @return True, if the execution should be aborted; false otherwise
371 bool ModelChecker::has_asserted() const
373 return priv->asserted;
377 * Trigger a trace assertion which should cause this execution to be halted.
378 * This can be due to a detected bug or due to an infeasibility that should
381 void ModelChecker::set_assert()
383 priv->asserted = true;
387 * Check if we are in a deadlock. Should only be called at the end of an
388 * execution, although it should not give false positives in the middle of an
389 * execution (there should be some ENABLED thread).
391 * @return True if program is in a deadlock; false otherwise
393 bool ModelChecker::is_deadlocked() const
395 bool blocking_threads = false;
396 for (unsigned int i = 0; i < get_num_threads(); i++) {
397 thread_id_t tid = int_to_id(i);
400 Thread *t = get_thread(tid);
401 if (!t->is_model_thread() && t->get_pending())
402 blocking_threads = true;
404 return blocking_threads;
408 * Check if this is a complete execution. That is, have all thread completed
409 * execution (rather than exiting because sleep sets have forced a redundant
412 * @return True if the execution is complete.
414 bool ModelChecker::is_complete_execution() const
416 for (unsigned int i = 0; i < get_num_threads(); i++)
417 if (is_enabled(int_to_id(i)))
423 * @brief Assert a bug in the executing program.
425 * Use this function to assert any sort of bug in the user program. If the
426 * current trace is feasible (actually, a prefix of some feasible execution),
427 * then this execution will be aborted, printing the appropriate message. If
428 * the current trace is not yet feasible, the error message will be stashed and
429 * printed if the execution ever becomes feasible.
431 * @param msg Descriptive message for the bug (do not include newline char)
432 * @return True if bug is immediately-feasible
434 bool ModelChecker::assert_bug(const char *msg)
436 priv->bugs.push_back(new bug_message(msg));
438 if (isfeasibleprefix()) {
446 * @brief Assert a bug in the executing program, asserted by a user thread
447 * @see ModelChecker::assert_bug
448 * @param msg Descriptive message for the bug (do not include newline char)
450 void ModelChecker::assert_user_bug(const char *msg)
452 /* If feasible bug, bail out now */
454 switch_to_master(NULL);
457 /** @return True, if any bugs have been reported for this execution */
458 bool ModelChecker::have_bug_reports() const
460 return priv->bugs.size() != 0;
463 /** @brief Print bug report listing for this execution (if any bugs exist) */
464 void ModelChecker::print_bugs() const
466 if (have_bug_reports()) {
467 model_print("Bug report: %zu bug%s detected\n",
469 priv->bugs.size() > 1 ? "s" : "");
470 for (unsigned int i = 0; i < priv->bugs.size(); i++)
471 priv->bugs[i]->print();
476 * @brief Record end-of-execution stats
478 * Must be run when exiting an execution. Records various stats.
479 * @see struct execution_stats
481 void ModelChecker::record_stats()
484 if (!isfeasibleprefix())
485 stats.num_infeasible++;
486 else if (have_bug_reports())
487 stats.num_buggy_executions++;
488 else if (is_complete_execution())
489 stats.num_complete++;
491 stats.num_redundant++;
494 /** @brief Print execution stats */
495 void ModelChecker::print_stats() const
497 model_print("Number of complete, bug-free executions: %d\n", stats.num_complete);
498 model_print("Number of redundant executions: %d\n", stats.num_redundant);
499 model_print("Number of buggy executions: %d\n", stats.num_buggy_executions);
500 model_print("Number of infeasible executions: %d\n", stats.num_infeasible);
501 model_print("Total executions: %d\n", stats.num_total);
502 model_print("Total nodes created: %d\n", node_stack->get_total_nodes());
506 * @brief End-of-exeuction print
507 * @param printbugs Should any existing bugs be printed?
509 void ModelChecker::print_execution(bool printbugs) const
511 print_program_output();
513 if (DBG_ENABLED() || params.verbose) {
514 model_print("Earliest divergence point since last feasible execution:\n");
515 if (earliest_diverge)
516 earliest_diverge->print();
518 model_print("(Not set)\n");
524 /* Don't print invalid bugs */
533 * Queries the model-checker for more executions to explore and, if one
534 * exists, resets the model-checker state to execute a new execution.
536 * @return If there are more executions to explore, return true. Otherwise,
539 bool ModelChecker::next_execution()
542 /* Is this execution a feasible execution that's worth bug-checking? */
543 bool complete = isfeasibleprefix() && (is_complete_execution() ||
546 /* End-of-execution bug checks */
549 assert_bug("Deadlock detected");
557 if (DBG_ENABLED() || params.verbose || (complete && have_bug_reports()))
558 print_execution(complete);
560 clear_program_output();
563 earliest_diverge = NULL;
565 if ((diverge = get_next_backtrack()) == NULL)
569 model_print("Next execution will diverge at:\n");
573 reset_to_initial_state();
578 * @brief Find the last fence-related backtracking conflict for a ModelAction
580 * This function performs the search for the most recent conflicting action
581 * against which we should perform backtracking, as affected by fence
582 * operations. This includes pairs of potentially-synchronizing actions which
583 * occur due to fence-acquire or fence-release, and hence should be explored in
584 * the opposite execution order.
586 * @param act The current action
587 * @return The most recent action which conflicts with act due to fences
589 ModelAction * ModelChecker::get_last_fence_conflict(ModelAction *act) const
591 /* Only perform release/acquire fence backtracking for stores */
592 if (!act->is_write())
595 /* Find a fence-release (or, act is a release) */
596 ModelAction *last_release;
597 if (act->is_release())
600 last_release = get_last_fence_release(act->get_tid());
604 /* Skip past the release */
605 action_list_t *list = action_trace;
606 action_list_t::reverse_iterator rit;
607 for (rit = list->rbegin(); rit != list->rend(); rit++)
608 if (*rit == last_release)
610 ASSERT(rit != list->rend());
615 * load --sb-> fence-acquire */
616 std::vector< ModelAction *, ModelAlloc<ModelAction *> > acquire_fences(get_num_threads(), NULL);
617 std::vector< ModelAction *, ModelAlloc<ModelAction *> > prior_loads(get_num_threads(), NULL);
618 bool found_acquire_fences = false;
619 for ( ; rit != list->rend(); rit++) {
620 ModelAction *prev = *rit;
621 if (act->same_thread(prev))
624 int tid = id_to_int(prev->get_tid());
626 if (prev->is_read() && act->same_var(prev)) {
627 if (prev->is_acquire()) {
628 /* Found most recent load-acquire, don't need
629 * to search for more fences */
630 if (!found_acquire_fences)
633 prior_loads[tid] = prev;
636 if (prev->is_acquire() && prev->is_fence() && !acquire_fences[tid]) {
637 found_acquire_fences = true;
638 acquire_fences[tid] = prev;
642 ModelAction *latest_backtrack = NULL;
643 for (unsigned int i = 0; i < acquire_fences.size(); i++)
644 if (acquire_fences[i] && prior_loads[i])
645 if (!latest_backtrack || *latest_backtrack < *acquire_fences[i])
646 latest_backtrack = acquire_fences[i];
647 return latest_backtrack;
651 * @brief Find the last backtracking conflict for a ModelAction
653 * This function performs the search for the most recent conflicting action
654 * against which we should perform backtracking. This primary includes pairs of
655 * synchronizing actions which should be explored in the opposite execution
658 * @param act The current action
659 * @return The most recent action which conflicts with act
661 ModelAction * ModelChecker::get_last_conflict(ModelAction *act) const
663 switch (act->get_type()) {
664 /* case ATOMIC_FENCE: fences don't directly cause backtracking */
668 ModelAction *ret = NULL;
670 /* linear search: from most recent to oldest */
671 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
672 action_list_t::reverse_iterator rit;
673 for (rit = list->rbegin(); rit != list->rend(); rit++) {
674 ModelAction *prev = *rit;
675 if (prev->could_synchronize_with(act)) {
681 ModelAction *ret2 = get_last_fence_conflict(act);
691 case ATOMIC_TRYLOCK: {
692 /* linear search: from most recent to oldest */
693 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
694 action_list_t::reverse_iterator rit;
695 for (rit = list->rbegin(); rit != list->rend(); rit++) {
696 ModelAction *prev = *rit;
697 if (act->is_conflicting_lock(prev))
702 case ATOMIC_UNLOCK: {
703 /* linear search: from most recent to oldest */
704 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
705 action_list_t::reverse_iterator rit;
706 for (rit = list->rbegin(); rit != list->rend(); rit++) {
707 ModelAction *prev = *rit;
708 if (!act->same_thread(prev) && prev->is_failed_trylock())
714 /* linear search: from most recent to oldest */
715 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
716 action_list_t::reverse_iterator rit;
717 for (rit = list->rbegin(); rit != list->rend(); rit++) {
718 ModelAction *prev = *rit;
719 if (!act->same_thread(prev) && prev->is_failed_trylock())
721 if (!act->same_thread(prev) && prev->is_notify())
727 case ATOMIC_NOTIFY_ALL:
728 case ATOMIC_NOTIFY_ONE: {
729 /* linear search: from most recent to oldest */
730 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
731 action_list_t::reverse_iterator rit;
732 for (rit = list->rbegin(); rit != list->rend(); rit++) {
733 ModelAction *prev = *rit;
734 if (!act->same_thread(prev) && prev->is_wait())
745 /** This method finds backtracking points where we should try to
746 * reorder the parameter ModelAction against.
748 * @param the ModelAction to find backtracking points for.
750 void ModelChecker::set_backtracking(ModelAction *act)
752 Thread *t = get_thread(act);
753 ModelAction *prev = get_last_conflict(act);
757 Node *node = prev->get_node()->get_parent();
759 int low_tid, high_tid;
760 if (node->enabled_status(t->get_id()) == THREAD_ENABLED) {
761 low_tid = id_to_int(act->get_tid());
762 high_tid = low_tid + 1;
765 high_tid = get_num_threads();
768 for (int i = low_tid; i < high_tid; i++) {
769 thread_id_t tid = int_to_id(i);
771 /* Make sure this thread can be enabled here. */
772 if (i >= node->get_num_threads())
775 /* Don't backtrack into a point where the thread is disabled or sleeping. */
776 if (node->enabled_status(tid) != THREAD_ENABLED)
779 /* Check if this has been explored already */
780 if (node->has_been_explored(tid))
783 /* See if fairness allows */
784 if (model->params.fairwindow != 0 && !node->has_priority(tid)) {
786 for (int t = 0; t < node->get_num_threads(); t++) {
787 thread_id_t tother = int_to_id(t);
788 if (node->is_enabled(tother) && node->has_priority(tother)) {
796 /* Cache the latest backtracking point */
797 set_latest_backtrack(prev);
799 /* If this is a new backtracking point, mark the tree */
800 if (!node->set_backtrack(tid))
802 DEBUG("Setting backtrack: conflict = %d, instead tid = %d\n",
803 id_to_int(prev->get_tid()),
804 id_to_int(t->get_id()));
813 * @brief Cache the a backtracking point as the "most recent", if eligible
815 * Note that this does not prepare the NodeStack for this backtracking
816 * operation, it only caches the action on a per-execution basis
818 * @param act The operation at which we should explore a different next action
819 * (i.e., backtracking point)
820 * @return True, if this action is now the most recent backtracking point;
823 bool ModelChecker::set_latest_backtrack(ModelAction *act)
825 if (!priv->next_backtrack || *act > *priv->next_backtrack) {
826 priv->next_backtrack = act;
833 * Returns last backtracking point. The model checker will explore a different
834 * path for this point in the next execution.
835 * @return The ModelAction at which the next execution should diverge.
837 ModelAction * ModelChecker::get_next_backtrack()
839 ModelAction *next = priv->next_backtrack;
840 priv->next_backtrack = NULL;
845 * Processes a read or rmw model action.
846 * @param curr is the read model action to process.
847 * @param second_part_of_rmw is boolean that is true is this is the second action of a rmw.
848 * @return True if processing this read updates the mo_graph.
850 bool ModelChecker::process_read(ModelAction *curr, bool second_part_of_rmw)
852 uint64_t value = VALUE_NONE;
853 bool updated = false;
855 const ModelAction *reads_from = curr->get_node()->get_read_from();
856 if (reads_from != NULL) {
857 mo_graph->startChanges();
859 value = reads_from->get_value();
860 bool r_status = false;
862 if (!second_part_of_rmw) {
863 check_recency(curr, reads_from);
864 r_status = r_modification_order(curr, reads_from);
867 if (!second_part_of_rmw && is_infeasible() && (curr->get_node()->increment_read_from() || curr->get_node()->increment_future_value())) {
868 mo_graph->rollbackChanges();
869 priv->too_many_reads = false;
873 read_from(curr, reads_from);
874 mo_graph->commitChanges();
875 mo_check_promises(curr, true);
878 } else if (!second_part_of_rmw) {
879 /* Read from future value */
880 struct future_value fv = curr->get_node()->get_future_value();
881 Promise *promise = new Promise(curr, fv);
883 curr->set_read_from_promise(promise);
884 promises->push_back(promise);
885 mo_graph->startChanges();
886 updated = r_modification_order(curr, promise);
887 mo_graph->commitChanges();
889 get_thread(curr)->set_return_value(value);
895 * Processes a lock, trylock, or unlock model action. @param curr is
896 * the read model action to process.
898 * The try lock operation checks whether the lock is taken. If not,
899 * it falls to the normal lock operation case. If so, it returns
902 * The lock operation has already been checked that it is enabled, so
903 * it just grabs the lock and synchronizes with the previous unlock.
905 * The unlock operation has to re-enable all of the threads that are
906 * waiting on the lock.
908 * @return True if synchronization was updated; false otherwise
910 bool ModelChecker::process_mutex(ModelAction *curr)
912 std::mutex *mutex = NULL;
913 struct std::mutex_state *state = NULL;
915 if (curr->is_trylock() || curr->is_lock() || curr->is_unlock()) {
916 mutex = (std::mutex *)curr->get_location();
917 state = mutex->get_state();
918 } else if (curr->is_wait()) {
919 mutex = (std::mutex *)curr->get_value();
920 state = mutex->get_state();
923 switch (curr->get_type()) {
924 case ATOMIC_TRYLOCK: {
925 bool success = !state->islocked;
926 curr->set_try_lock(success);
928 get_thread(curr)->set_return_value(0);
931 get_thread(curr)->set_return_value(1);
933 //otherwise fall into the lock case
935 if (curr->get_cv()->getClock(state->alloc_tid) <= state->alloc_clock)
936 assert_bug("Lock access before initialization");
937 state->islocked = true;
938 ModelAction *unlock = get_last_unlock(curr);
939 //synchronize with the previous unlock statement
940 if (unlock != NULL) {
941 curr->synchronize_with(unlock);
946 case ATOMIC_UNLOCK: {
948 state->islocked = false;
949 //wake up the other threads
950 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, curr->get_location());
951 //activate all the waiting threads
952 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
953 scheduler->wake(get_thread(*rit));
960 state->islocked = false;
961 //wake up the other threads
962 action_list_t *waiters = get_safe_ptr_action(lock_waiters_map, (void *) curr->get_value());
963 //activate all the waiting threads
964 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
965 scheduler->wake(get_thread(*rit));
968 //check whether we should go to sleep or not...simulate spurious failures
969 if (curr->get_node()->get_misc() == 0) {
970 get_safe_ptr_action(condvar_waiters_map, curr->get_location())->push_back(curr);
972 scheduler->sleep(get_thread(curr));
976 case ATOMIC_NOTIFY_ALL: {
977 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
978 //activate all the waiting threads
979 for (action_list_t::iterator rit = waiters->begin(); rit != waiters->end(); rit++) {
980 scheduler->wake(get_thread(*rit));
985 case ATOMIC_NOTIFY_ONE: {
986 action_list_t *waiters = get_safe_ptr_action(condvar_waiters_map, curr->get_location());
987 int wakeupthread = curr->get_node()->get_misc();
988 action_list_t::iterator it = waiters->begin();
989 advance(it, wakeupthread);
990 scheduler->wake(get_thread(*it));
1001 void ModelChecker::add_future_value(const ModelAction *writer, ModelAction *reader)
1003 /* Do more ambitious checks now that mo is more complete */
1004 if (mo_may_allow(writer, reader)) {
1005 Node *node = reader->get_node();
1007 /* Find an ancestor thread which exists at the time of the reader */
1008 Thread *write_thread = get_thread(writer);
1009 while (id_to_int(write_thread->get_id()) >= node->get_num_threads())
1010 write_thread = write_thread->get_parent();
1012 struct future_value fv = {
1013 writer->get_value(),
1014 writer->get_seq_number() + params.maxfuturedelay,
1015 write_thread->get_id(),
1017 if (node->add_future_value(fv))
1018 set_latest_backtrack(reader);
1023 * Process a write ModelAction
1024 * @param curr The ModelAction to process
1025 * @return True if the mo_graph was updated or promises were resolved
1027 bool ModelChecker::process_write(ModelAction *curr)
1029 bool updated_mod_order = w_modification_order(curr);
1030 bool updated_promises = resolve_promises(curr);
1032 if (promises->size() == 0) {
1033 for (unsigned int i = 0; i < futurevalues->size(); i++) {
1034 struct PendingFutureValue pfv = (*futurevalues)[i];
1035 add_future_value(pfv.writer, pfv.act);
1037 futurevalues->clear();
1040 mo_graph->commitChanges();
1041 mo_check_promises(curr, false);
1043 get_thread(curr)->set_return_value(VALUE_NONE);
1044 return updated_mod_order || updated_promises;
1048 * Process a fence ModelAction
1049 * @param curr The ModelAction to process
1050 * @return True if synchronization was updated
1052 bool ModelChecker::process_fence(ModelAction *curr)
1055 * fence-relaxed: no-op
1056 * fence-release: only log the occurence (not in this function), for
1057 * use in later synchronization
1058 * fence-acquire (this function): search for hypothetical release
1061 bool updated = false;
1062 if (curr->is_acquire()) {
1063 action_list_t *list = action_trace;
1064 action_list_t::reverse_iterator rit;
1065 /* Find X : is_read(X) && X --sb-> curr */
1066 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1067 ModelAction *act = *rit;
1070 if (act->get_tid() != curr->get_tid())
1072 /* Stop at the beginning of the thread */
1073 if (act->is_thread_start())
1075 /* Stop once we reach a prior fence-acquire */
1076 if (act->is_fence() && act->is_acquire())
1078 if (!act->is_read())
1080 /* read-acquire will find its own release sequences */
1081 if (act->is_acquire())
1084 /* Establish hypothetical release sequences */
1085 rel_heads_list_t release_heads;
1086 get_release_seq_heads(curr, act, &release_heads);
1087 for (unsigned int i = 0; i < release_heads.size(); i++)
1088 if (!curr->synchronize_with(release_heads[i]))
1089 set_bad_synchronization();
1090 if (release_heads.size() != 0)
1098 * @brief Process the current action for thread-related activity
1100 * Performs current-action processing for a THREAD_* ModelAction. Proccesses
1101 * may include setting Thread status, completing THREAD_FINISH/THREAD_JOIN
1102 * synchronization, etc. This function is a no-op for non-THREAD actions
1103 * (e.g., ATOMIC_{READ,WRITE,RMW,LOCK}, etc.)
1105 * @param curr The current action
1106 * @return True if synchronization was updated or a thread completed
1108 bool ModelChecker::process_thread_action(ModelAction *curr)
1110 bool updated = false;
1112 switch (curr->get_type()) {
1113 case THREAD_CREATE: {
1114 thrd_t *thrd = (thrd_t *)curr->get_location();
1115 struct thread_params *params = (struct thread_params *)curr->get_value();
1116 Thread *th = new Thread(thrd, params->func, params->arg, get_thread(curr));
1118 th->set_creation(curr);
1119 /* Promises can be satisfied by children */
1120 for (unsigned int i = 0; i < promises->size(); i++) {
1121 Promise *promise = (*promises)[i];
1122 if (promise->thread_is_available(curr->get_tid()))
1123 promise->add_thread(th->get_id());
1128 Thread *blocking = curr->get_thread_operand();
1129 ModelAction *act = get_last_action(blocking->get_id());
1130 curr->synchronize_with(act);
1131 updated = true; /* trigger rel-seq checks */
1134 case THREAD_FINISH: {
1135 Thread *th = get_thread(curr);
1136 while (!th->wait_list_empty()) {
1137 ModelAction *act = th->pop_wait_list();
1138 scheduler->wake(get_thread(act));
1141 /* Completed thread can't satisfy promises */
1142 for (unsigned int i = 0; i < promises->size(); i++) {
1143 Promise *promise = (*promises)[i];
1144 if (promise->thread_is_available(th->get_id()))
1145 if (promise->eliminate_thread(th->get_id()))
1146 priv->failed_promise = true;
1148 updated = true; /* trigger rel-seq checks */
1151 case THREAD_START: {
1152 check_promises(curr->get_tid(), NULL, curr->get_cv());
1163 * @brief Process the current action for release sequence fixup activity
1165 * Performs model-checker release sequence fixups for the current action,
1166 * forcing a single pending release sequence to break (with a given, potential
1167 * "loose" write) or to complete (i.e., synchronize). If a pending release
1168 * sequence forms a complete release sequence, then we must perform the fixup
1169 * synchronization, mo_graph additions, etc.
1171 * @param curr The current action; must be a release sequence fixup action
1172 * @param work_queue The work queue to which to add work items as they are
1175 void ModelChecker::process_relseq_fixup(ModelAction *curr, work_queue_t *work_queue)
1177 const ModelAction *write = curr->get_node()->get_relseq_break();
1178 struct release_seq *sequence = pending_rel_seqs->back();
1179 pending_rel_seqs->pop_back();
1181 ModelAction *acquire = sequence->acquire;
1182 const ModelAction *rf = sequence->rf;
1183 const ModelAction *release = sequence->release;
1187 ASSERT(release->same_thread(rf));
1189 if (write == NULL) {
1191 * @todo Forcing a synchronization requires that we set
1192 * modification order constraints. For instance, we can't allow
1193 * a fixup sequence in which two separate read-acquire
1194 * operations read from the same sequence, where the first one
1195 * synchronizes and the other doesn't. Essentially, we can't
1196 * allow any writes to insert themselves between 'release' and
1200 /* Must synchronize */
1201 if (!acquire->synchronize_with(release)) {
1202 set_bad_synchronization();
1205 /* Re-check all pending release sequences */
1206 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
1207 /* Re-check act for mo_graph edges */
1208 work_queue->push_back(MOEdgeWorkEntry(acquire));
1210 /* propagate synchronization to later actions */
1211 action_list_t::reverse_iterator rit = action_trace->rbegin();
1212 for (; (*rit) != acquire; rit++) {
1213 ModelAction *propagate = *rit;
1214 if (acquire->happens_before(propagate)) {
1215 propagate->synchronize_with(acquire);
1216 /* Re-check 'propagate' for mo_graph edges */
1217 work_queue->push_back(MOEdgeWorkEntry(propagate));
1221 /* Break release sequence with new edges:
1222 * release --mo--> write --mo--> rf */
1223 mo_graph->addEdge(release, write);
1224 mo_graph->addEdge(write, rf);
1227 /* See if we have realized a data race */
1232 * Initialize the current action by performing one or more of the following
1233 * actions, as appropriate: merging RMWR and RMWC/RMW actions, stepping forward
1234 * in the NodeStack, manipulating backtracking sets, allocating and
1235 * initializing clock vectors, and computing the promises to fulfill.
1237 * @param curr The current action, as passed from the user context; may be
1238 * freed/invalidated after the execution of this function, with a different
1239 * action "returned" its place (pass-by-reference)
1240 * @return True if curr is a newly-explored action; false otherwise
1242 bool ModelChecker::initialize_curr_action(ModelAction **curr)
1244 ModelAction *newcurr;
1246 if ((*curr)->is_rmwc() || (*curr)->is_rmw()) {
1247 newcurr = process_rmw(*curr);
1250 if (newcurr->is_rmw())
1251 compute_promises(newcurr);
1257 (*curr)->set_seq_number(get_next_seq_num());
1259 newcurr = node_stack->explore_action(*curr, scheduler->get_enabled_array());
1261 /* First restore type and order in case of RMW operation */
1262 if ((*curr)->is_rmwr())
1263 newcurr->copy_typeandorder(*curr);
1265 ASSERT((*curr)->get_location() == newcurr->get_location());
1266 newcurr->copy_from_new(*curr);
1268 /* Discard duplicate ModelAction; use action from NodeStack */
1271 /* Always compute new clock vector */
1272 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1275 return false; /* Action was explored previously */
1279 /* Always compute new clock vector */
1280 newcurr->create_cv(get_parent_action(newcurr->get_tid()));
1282 /* Assign most recent release fence */
1283 newcurr->set_last_fence_release(get_last_fence_release(newcurr->get_tid()));
1286 * Perform one-time actions when pushing new ModelAction onto
1289 if (newcurr->is_write())
1290 compute_promises(newcurr);
1291 else if (newcurr->is_relseq_fixup())
1292 compute_relseq_breakwrites(newcurr);
1293 else if (newcurr->is_wait())
1294 newcurr->get_node()->set_misc_max(2);
1295 else if (newcurr->is_notify_one()) {
1296 newcurr->get_node()->set_misc_max(get_safe_ptr_action(condvar_waiters_map, newcurr->get_location())->size());
1298 return true; /* This was a new ModelAction */
1303 * @brief Establish reads-from relation between two actions
1305 * Perform basic operations involved with establishing a concrete rf relation,
1306 * including setting the ModelAction data and checking for release sequences.
1308 * @param act The action that is reading (must be a read)
1309 * @param rf The action from which we are reading (must be a write)
1311 * @return True if this read established synchronization
1313 bool ModelChecker::read_from(ModelAction *act, const ModelAction *rf)
1315 act->set_read_from(rf);
1316 if (rf != NULL && act->is_acquire()) {
1317 rel_heads_list_t release_heads;
1318 get_release_seq_heads(act, act, &release_heads);
1319 int num_heads = release_heads.size();
1320 for (unsigned int i = 0; i < release_heads.size(); i++)
1321 if (!act->synchronize_with(release_heads[i])) {
1322 set_bad_synchronization();
1325 return num_heads > 0;
1331 * Check promises and eliminate potentially-satisfying threads when a thread is
1332 * blocked (e.g., join, lock). A thread which is waiting on another thread can
1333 * no longer satisfy a promise generated from that thread.
1335 * @param blocker The thread on which a thread is waiting
1336 * @param waiting The waiting thread
1338 void ModelChecker::thread_blocking_check_promises(Thread *blocker, Thread *waiting)
1340 for (unsigned int i = 0; i < promises->size(); i++) {
1341 Promise *promise = (*promises)[i];
1342 ModelAction *reader = promise->get_action();
1343 if (reader->get_tid() != blocker->get_id())
1345 if (!promise->thread_is_available(waiting->get_id()))
1347 if (promise->eliminate_thread(waiting->get_id())) {
1348 /* Promise has failed */
1349 priv->failed_promise = true;
1355 * @brief Check whether a model action is enabled.
1357 * Checks whether a lock or join operation would be successful (i.e., is the
1358 * lock already locked, or is the joined thread already complete). If not, put
1359 * the action in a waiter list.
1361 * @param curr is the ModelAction to check whether it is enabled.
1362 * @return a bool that indicates whether the action is enabled.
1364 bool ModelChecker::check_action_enabled(ModelAction *curr) {
1365 if (curr->is_lock()) {
1366 std::mutex *lock = (std::mutex *)curr->get_location();
1367 struct std::mutex_state *state = lock->get_state();
1368 if (state->islocked) {
1369 //Stick the action in the appropriate waiting queue
1370 get_safe_ptr_action(lock_waiters_map, curr->get_location())->push_back(curr);
1373 } else if (curr->get_type() == THREAD_JOIN) {
1374 Thread *blocking = (Thread *)curr->get_location();
1375 if (!blocking->is_complete()) {
1376 blocking->push_wait_list(curr);
1377 thread_blocking_check_promises(blocking, get_thread(curr));
1386 * This is the heart of the model checker routine. It performs model-checking
1387 * actions corresponding to a given "current action." Among other processes, it
1388 * calculates reads-from relationships, updates synchronization clock vectors,
1389 * forms a memory_order constraints graph, and handles replay/backtrack
1390 * execution when running permutations of previously-observed executions.
1392 * @param curr The current action to process
1393 * @return The ModelAction that is actually executed; may be different than
1394 * curr; may be NULL, if the current action is not enabled to run
1396 ModelAction * ModelChecker::check_current_action(ModelAction *curr)
1399 bool second_part_of_rmw = curr->is_rmwc() || curr->is_rmw();
1401 if (!check_action_enabled(curr)) {
1402 /* Make the execution look like we chose to run this action
1403 * much later, when a lock/join can succeed */
1404 get_thread(curr)->set_pending(curr);
1405 scheduler->sleep(get_thread(curr));
1409 bool newly_explored = initialize_curr_action(&curr);
1415 wake_up_sleeping_actions(curr);
1417 /* Add the action to lists before any other model-checking tasks */
1418 if (!second_part_of_rmw)
1419 add_action_to_lists(curr);
1421 /* Build may_read_from set for newly-created actions */
1422 if (newly_explored && curr->is_read())
1423 build_may_read_from(curr);
1425 /* Initialize work_queue with the "current action" work */
1426 work_queue_t work_queue(1, CheckCurrWorkEntry(curr));
1427 while (!work_queue.empty() && !has_asserted()) {
1428 WorkQueueEntry work = work_queue.front();
1429 work_queue.pop_front();
1431 switch (work.type) {
1432 case WORK_CHECK_CURR_ACTION: {
1433 ModelAction *act = work.action;
1434 bool update = false; /* update this location's release seq's */
1435 bool update_all = false; /* update all release seq's */
1437 if (process_thread_action(curr))
1440 if (act->is_read() && process_read(act, second_part_of_rmw))
1443 if (act->is_write() && process_write(act))
1446 if (act->is_fence() && process_fence(act))
1449 if (act->is_mutex_op() && process_mutex(act))
1452 if (act->is_relseq_fixup())
1453 process_relseq_fixup(curr, &work_queue);
1456 work_queue.push_back(CheckRelSeqWorkEntry(NULL));
1458 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1461 case WORK_CHECK_RELEASE_SEQ:
1462 resolve_release_sequences(work.location, &work_queue);
1464 case WORK_CHECK_MO_EDGES: {
1465 /** @todo Complete verification of work_queue */
1466 ModelAction *act = work.action;
1467 bool updated = false;
1469 if (act->is_read()) {
1470 const ModelAction *rf = act->get_reads_from();
1471 const Promise *promise = act->get_reads_from_promise();
1473 if (r_modification_order(act, rf))
1475 } else if (promise) {
1476 if (r_modification_order(act, promise))
1480 if (act->is_write()) {
1481 if (w_modification_order(act))
1484 mo_graph->commitChanges();
1487 work_queue.push_back(CheckRelSeqWorkEntry(act->get_location()));
1496 check_curr_backtracking(curr);
1497 set_backtracking(curr);
1501 void ModelChecker::check_curr_backtracking(ModelAction *curr)
1503 Node *currnode = curr->get_node();
1504 Node *parnode = currnode->get_parent();
1506 if ((parnode && !parnode->backtrack_empty()) ||
1507 !currnode->misc_empty() ||
1508 !currnode->read_from_empty() ||
1509 !currnode->future_value_empty() ||
1510 !currnode->promise_empty() ||
1511 !currnode->relseq_break_empty()) {
1512 set_latest_backtrack(curr);
1516 bool ModelChecker::promises_expired() const
1518 for (unsigned int i = 0; i < promises->size(); i++) {
1519 Promise *promise = (*promises)[i];
1520 if (promise->get_expiration() < priv->used_sequence_numbers)
1527 * This is the strongest feasibility check available.
1528 * @return whether the current trace (partial or complete) must be a prefix of
1531 bool ModelChecker::isfeasibleprefix() const
1533 return pending_rel_seqs->size() == 0 && is_feasible_prefix_ignore_relseq();
1537 * Print disagnostic information about an infeasible execution
1538 * @param prefix A string to prefix the output with; if NULL, then a default
1539 * message prefix will be provided
1541 void ModelChecker::print_infeasibility(const char *prefix) const
1545 if (mo_graph->checkForCycles())
1546 ptr += sprintf(ptr, "[mo cycle]");
1547 if (priv->failed_promise)
1548 ptr += sprintf(ptr, "[failed promise]");
1549 if (priv->too_many_reads)
1550 ptr += sprintf(ptr, "[too many reads]");
1551 if (priv->no_valid_reads)
1552 ptr += sprintf(ptr, "[no valid reads-from]");
1553 if (priv->bad_synchronization)
1554 ptr += sprintf(ptr, "[bad sw ordering]");
1555 if (promises_expired())
1556 ptr += sprintf(ptr, "[promise expired]");
1557 if (promises->size() != 0)
1558 ptr += sprintf(ptr, "[unresolved promise]");
1560 model_print("%s: %s\n", prefix ? prefix : "Infeasible", buf);
1564 * Returns whether the current completed trace is feasible, except for pending
1565 * release sequences.
1567 bool ModelChecker::is_feasible_prefix_ignore_relseq() const
1569 return !is_infeasible() && promises->size() == 0;
1573 * Check if the current partial trace is infeasible. Does not check any
1574 * end-of-execution flags, which might rule out the execution. Thus, this is
1575 * useful only for ruling an execution as infeasible.
1576 * @return whether the current partial trace is infeasible.
1578 bool ModelChecker::is_infeasible() const
1580 return mo_graph->checkForCycles() ||
1581 priv->no_valid_reads ||
1582 priv->failed_promise ||
1583 priv->too_many_reads ||
1584 priv->bad_synchronization ||
1588 /** Close out a RMWR by converting previous RMWR into a RMW or READ. */
1589 ModelAction * ModelChecker::process_rmw(ModelAction *act) {
1590 ModelAction *lastread = get_last_action(act->get_tid());
1591 lastread->process_rmw(act);
1592 if (act->is_rmw()) {
1593 if (lastread->get_reads_from())
1594 mo_graph->addRMWEdge(lastread->get_reads_from(), lastread);
1596 mo_graph->addRMWEdge(lastread->get_reads_from_promise(), lastread);
1597 mo_graph->commitChanges();
1603 * Checks whether a thread has read from the same write for too many times
1604 * without seeing the effects of a later write.
1607 * 1) there must a different write that we could read from that would satisfy the modification order,
1608 * 2) we must have read from the same value in excess of maxreads times, and
1609 * 3) that other write must have been in the reads_from set for maxreads times.
1611 * If so, we decide that the execution is no longer feasible.
1613 void ModelChecker::check_recency(ModelAction *curr, const ModelAction *rf)
1615 if (params.maxreads != 0) {
1616 if (curr->get_node()->get_read_from_size() <= 1)
1618 //Must make sure that execution is currently feasible... We could
1619 //accidentally clear by rolling back
1620 if (is_infeasible())
1622 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1623 int tid = id_to_int(curr->get_tid());
1626 if ((int)thrd_lists->size() <= tid)
1628 action_list_t *list = &(*thrd_lists)[tid];
1630 action_list_t::reverse_iterator rit = list->rbegin();
1631 /* Skip past curr */
1632 for (; (*rit) != curr; rit++)
1634 /* go past curr now */
1637 action_list_t::reverse_iterator ritcopy = rit;
1638 //See if we have enough reads from the same value
1640 for (; count < params.maxreads; rit++, count++) {
1641 if (rit == list->rend())
1643 ModelAction *act = *rit;
1644 if (!act->is_read())
1647 if (act->get_reads_from() != rf)
1649 if (act->get_node()->get_read_from_size() <= 1)
1652 for (int i = 0; i < curr->get_node()->get_read_from_size(); i++) {
1654 const ModelAction *write = curr->get_node()->get_read_from_at(i);
1656 /* Need a different write */
1660 /* Test to see whether this is a feasible write to read from */
1661 /** NOTE: all members of read-from set should be
1662 * feasible, so we no longer check it here **/
1666 bool feasiblewrite = true;
1667 //new we need to see if this write works for everyone
1669 for (int loop = count; loop > 0; loop--, rit++) {
1670 ModelAction *act = *rit;
1671 bool foundvalue = false;
1672 for (int j = 0; j < act->get_node()->get_read_from_size(); j++) {
1673 if (act->get_node()->get_read_from_at(j) == write) {
1679 feasiblewrite = false;
1683 if (feasiblewrite) {
1684 priv->too_many_reads = true;
1692 * Updates the mo_graph with the constraints imposed from the current
1695 * Basic idea is the following: Go through each other thread and find
1696 * the last action that happened before our read. Two cases:
1698 * (1) The action is a write => that write must either occur before
1699 * the write we read from or be the write we read from.
1701 * (2) The action is a read => the write that that action read from
1702 * must occur before the write we read from or be the same write.
1704 * @param curr The current action. Must be a read.
1705 * @param rf The ModelAction or Promise that curr reads from. Must be a write.
1706 * @return True if modification order edges were added; false otherwise
1708 template <typename rf_type>
1709 bool ModelChecker::r_modification_order(ModelAction *curr, const rf_type *rf)
1711 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1714 ASSERT(curr->is_read());
1716 /* Last SC fence in the current thread */
1717 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1719 /* Iterate over all threads */
1720 for (i = 0; i < thrd_lists->size(); i++) {
1721 /* Last SC fence in thread i */
1722 ModelAction *last_sc_fence_thread_local = NULL;
1723 if (int_to_id((int)i) != curr->get_tid())
1724 last_sc_fence_thread_local = get_last_seq_cst_fence(int_to_id(i), NULL);
1726 /* Last SC fence in thread i, before last SC fence in current thread */
1727 ModelAction *last_sc_fence_thread_before = NULL;
1728 if (last_sc_fence_local)
1729 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1731 /* Iterate over actions in thread, starting from most recent */
1732 action_list_t *list = &(*thrd_lists)[i];
1733 action_list_t::reverse_iterator rit;
1734 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1735 ModelAction *act = *rit;
1737 if (act->is_write() && !act->equals(rf) && act != curr) {
1738 /* C++, Section 29.3 statement 5 */
1739 if (curr->is_seqcst() && last_sc_fence_thread_local &&
1740 *act < *last_sc_fence_thread_local) {
1741 added = mo_graph->addEdge(act, rf) || added;
1744 /* C++, Section 29.3 statement 4 */
1745 else if (act->is_seqcst() && last_sc_fence_local &&
1746 *act < *last_sc_fence_local) {
1747 added = mo_graph->addEdge(act, rf) || added;
1750 /* C++, Section 29.3 statement 6 */
1751 else if (last_sc_fence_thread_before &&
1752 *act < *last_sc_fence_thread_before) {
1753 added = mo_graph->addEdge(act, rf) || added;
1759 * Include at most one act per-thread that "happens
1760 * before" curr. Don't consider reflexively.
1762 if (act->happens_before(curr) && act != curr) {
1763 if (act->is_write()) {
1764 if (!act->equals(rf)) {
1765 added = mo_graph->addEdge(act, rf) || added;
1768 const ModelAction *prevrf = act->get_reads_from();
1769 const Promise *prevrf_promise = act->get_reads_from_promise();
1771 if (!prevrf->equals(rf))
1772 added = mo_graph->addEdge(prevrf, rf) || added;
1773 } else if (!prevrf->equals(rf)) {
1774 added = mo_graph->addEdge(prevrf_promise, rf) || added;
1783 * All compatible, thread-exclusive promises must be ordered after any
1784 * concrete loads from the same thread
1786 for (unsigned int i = 0; i < promises->size(); i++)
1787 if ((*promises)[i]->is_compatible_exclusive(curr))
1788 added = mo_graph->addEdge(rf, (*promises)[i]) || added;
1794 * Updates the mo_graph with the constraints imposed from the current write.
1796 * Basic idea is the following: Go through each other thread and find
1797 * the lastest action that happened before our write. Two cases:
1799 * (1) The action is a write => that write must occur before
1802 * (2) The action is a read => the write that that action read from
1803 * must occur before the current write.
1805 * This method also handles two other issues:
1807 * (I) Sequential Consistency: Making sure that if the current write is
1808 * seq_cst, that it occurs after the previous seq_cst write.
1810 * (II) Sending the write back to non-synchronizing reads.
1812 * @param curr The current action. Must be a write.
1813 * @return True if modification order edges were added; false otherwise
1815 bool ModelChecker::w_modification_order(ModelAction *curr)
1817 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
1820 ASSERT(curr->is_write());
1822 if (curr->is_seqcst()) {
1823 /* We have to at least see the last sequentially consistent write,
1824 so we are initialized. */
1825 ModelAction *last_seq_cst = get_last_seq_cst_write(curr);
1826 if (last_seq_cst != NULL) {
1827 added = mo_graph->addEdge(last_seq_cst, curr) || added;
1831 /* Last SC fence in the current thread */
1832 ModelAction *last_sc_fence_local = get_last_seq_cst_fence(curr->get_tid(), NULL);
1834 /* Iterate over all threads */
1835 for (i = 0; i < thrd_lists->size(); i++) {
1836 /* Last SC fence in thread i, before last SC fence in current thread */
1837 ModelAction *last_sc_fence_thread_before = NULL;
1838 if (last_sc_fence_local && int_to_id((int)i) != curr->get_tid())
1839 last_sc_fence_thread_before = get_last_seq_cst_fence(int_to_id(i), last_sc_fence_local);
1841 /* Iterate over actions in thread, starting from most recent */
1842 action_list_t *list = &(*thrd_lists)[i];
1843 action_list_t::reverse_iterator rit;
1844 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1845 ModelAction *act = *rit;
1848 * 1) If RMW and it actually read from something, then we
1849 * already have all relevant edges, so just skip to next
1852 * 2) If RMW and it didn't read from anything, we should
1853 * whatever edge we can get to speed up convergence.
1855 * 3) If normal write, we need to look at earlier actions, so
1856 * continue processing list.
1858 if (curr->is_rmw()) {
1859 if (curr->get_reads_from() != NULL)
1867 /* C++, Section 29.3 statement 7 */
1868 if (last_sc_fence_thread_before && act->is_write() &&
1869 *act < *last_sc_fence_thread_before) {
1870 added = mo_graph->addEdge(act, curr) || added;
1875 * Include at most one act per-thread that "happens
1878 if (act->happens_before(curr)) {
1880 * Note: if act is RMW, just add edge:
1882 * The following edge should be handled elsewhere:
1883 * readfrom(act) --mo--> act
1885 if (act->is_write())
1886 added = mo_graph->addEdge(act, curr) || added;
1887 else if (act->is_read()) {
1888 //if previous read accessed a null, just keep going
1889 if (act->get_reads_from() == NULL)
1891 added = mo_graph->addEdge(act->get_reads_from(), curr) || added;
1894 } else if (act->is_read() && !act->could_synchronize_with(curr) &&
1895 !act->same_thread(curr)) {
1896 /* We have an action that:
1897 (1) did not happen before us
1898 (2) is a read and we are a write
1899 (3) cannot synchronize with us
1900 (4) is in a different thread
1902 that read could potentially read from our write. Note that
1903 these checks are overly conservative at this point, we'll
1904 do more checks before actually removing the
1908 if (thin_air_constraint_may_allow(curr, act)) {
1909 if (!is_infeasible())
1910 futurevalues->push_back(PendingFutureValue(curr, act));
1911 else if (curr->is_rmw() && act->is_rmw() && curr->get_reads_from() && curr->get_reads_from() == act->get_reads_from())
1912 add_future_value(curr, act);
1919 * All compatible, thread-exclusive promises must be ordered after any
1920 * concrete stores to the same thread, or else they can be merged with
1923 for (unsigned int i = 0; i < promises->size(); i++)
1924 if ((*promises)[i]->is_compatible_exclusive(curr))
1925 added = mo_graph->addEdge(curr, (*promises)[i]) || added;
1930 /** Arbitrary reads from the future are not allowed. Section 29.3
1931 * part 9 places some constraints. This method checks one result of constraint
1932 * constraint. Others require compiler support. */
1933 bool ModelChecker::thin_air_constraint_may_allow(const ModelAction *writer, const ModelAction *reader)
1935 if (!writer->is_rmw())
1938 if (!reader->is_rmw())
1941 for (const ModelAction *search = writer->get_reads_from(); search != NULL; search = search->get_reads_from()) {
1942 if (search == reader)
1944 if (search->get_tid() == reader->get_tid() &&
1945 search->happens_before(reader))
1953 * Arbitrary reads from the future are not allowed. Section 29.3 part 9 places
1954 * some constraints. This method checks one the following constraint (others
1955 * require compiler support):
1957 * If X --hb-> Y --mo-> Z, then X should not read from Z.
1959 bool ModelChecker::mo_may_allow(const ModelAction *writer, const ModelAction *reader)
1961 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, reader->get_location());
1963 /* Iterate over all threads */
1964 for (i = 0; i < thrd_lists->size(); i++) {
1965 const ModelAction *write_after_read = NULL;
1967 /* Iterate over actions in thread, starting from most recent */
1968 action_list_t *list = &(*thrd_lists)[i];
1969 action_list_t::reverse_iterator rit;
1970 for (rit = list->rbegin(); rit != list->rend(); rit++) {
1971 ModelAction *act = *rit;
1973 /* Don't disallow due to act == reader */
1974 if (!reader->happens_before(act) || reader == act)
1976 else if (act->is_write())
1977 write_after_read = act;
1978 else if (act->is_read() && act->get_reads_from() != NULL)
1979 write_after_read = act->get_reads_from();
1982 if (write_after_read && write_after_read != writer && mo_graph->checkReachable(write_after_read, writer))
1989 * Finds the head(s) of the release sequence(s) containing a given ModelAction.
1990 * The ModelAction under consideration is expected to be taking part in
1991 * release/acquire synchronization as an object of the "reads from" relation.
1992 * Note that this can only provide release sequence support for RMW chains
1993 * which do not read from the future, as those actions cannot be traced until
1994 * their "promise" is fulfilled. Similarly, we may not even establish the
1995 * presence of a release sequence with certainty, as some modification order
1996 * constraints may be decided further in the future. Thus, this function
1997 * "returns" two pieces of data: a pass-by-reference vector of @a release_heads
1998 * and a boolean representing certainty.
2000 * @param rf The action that might be part of a release sequence. Must be a
2002 * @param release_heads A pass-by-reference style return parameter. After
2003 * execution of this function, release_heads will contain the heads of all the
2004 * relevant release sequences, if any exists with certainty
2005 * @param pending A pass-by-reference style return parameter which is only used
2006 * when returning false (i.e., uncertain). Returns most information regarding
2007 * an uncertain release sequence, including any write operations that might
2008 * break the sequence.
2009 * @return true, if the ModelChecker is certain that release_heads is complete;
2012 bool ModelChecker::release_seq_heads(const ModelAction *rf,
2013 rel_heads_list_t *release_heads,
2014 struct release_seq *pending) const
2016 /* Only check for release sequences if there are no cycles */
2017 if (mo_graph->checkForCycles())
2020 for ( ; rf != NULL; rf = rf->get_reads_from()) {
2021 ASSERT(rf->is_write());
2023 if (rf->is_release())
2024 release_heads->push_back(rf);
2025 else if (rf->get_last_fence_release())
2026 release_heads->push_back(rf->get_last_fence_release());
2028 break; /* End of RMW chain */
2030 /** @todo Need to be smarter here... In the linux lock
2031 * example, this will run to the beginning of the program for
2033 /** @todo The way to be smarter here is to keep going until 1
2034 * thread has a release preceded by an acquire and you've seen
2037 /* acq_rel RMW is a sufficient stopping condition */
2038 if (rf->is_acquire() && rf->is_release())
2039 return true; /* complete */
2042 /* read from future: need to settle this later */
2044 return false; /* incomplete */
2047 if (rf->is_release())
2048 return true; /* complete */
2050 /* else relaxed write
2051 * - check for fence-release in the same thread (29.8, stmt. 3)
2052 * - check modification order for contiguous subsequence
2053 * -> rf must be same thread as release */
2055 const ModelAction *fence_release = rf->get_last_fence_release();
2056 /* Synchronize with a fence-release unconditionally; we don't need to
2057 * find any more "contiguous subsequence..." for it */
2059 release_heads->push_back(fence_release);
2061 int tid = id_to_int(rf->get_tid());
2062 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, rf->get_location());
2063 action_list_t *list = &(*thrd_lists)[tid];
2064 action_list_t::const_reverse_iterator rit;
2066 /* Find rf in the thread list */
2067 rit = std::find(list->rbegin(), list->rend(), rf);
2068 ASSERT(rit != list->rend());
2070 /* Find the last {write,fence}-release */
2071 for (; rit != list->rend(); rit++) {
2072 if (fence_release && *(*rit) < *fence_release)
2074 if ((*rit)->is_release())
2077 if (rit == list->rend()) {
2078 /* No write-release in this thread */
2079 return true; /* complete */
2080 } else if (fence_release && *(*rit) < *fence_release) {
2081 /* The fence-release is more recent (and so, "stronger") than
2082 * the most recent write-release */
2083 return true; /* complete */
2084 } /* else, need to establish contiguous release sequence */
2085 ModelAction *release = *rit;
2087 ASSERT(rf->same_thread(release));
2089 pending->writes.clear();
2091 bool certain = true;
2092 for (unsigned int i = 0; i < thrd_lists->size(); i++) {
2093 if (id_to_int(rf->get_tid()) == (int)i)
2095 list = &(*thrd_lists)[i];
2097 /* Can we ensure no future writes from this thread may break
2098 * the release seq? */
2099 bool future_ordered = false;
2101 ModelAction *last = get_last_action(int_to_id(i));
2102 Thread *th = get_thread(int_to_id(i));
2103 if ((last && rf->happens_before(last)) ||
2106 future_ordered = true;
2108 ASSERT(!th->is_model_thread() || future_ordered);
2110 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2111 const ModelAction *act = *rit;
2112 /* Reach synchronization -> this thread is complete */
2113 if (act->happens_before(release))
2115 if (rf->happens_before(act)) {
2116 future_ordered = true;
2120 /* Only non-RMW writes can break release sequences */
2121 if (!act->is_write() || act->is_rmw())
2124 /* Check modification order */
2125 if (mo_graph->checkReachable(rf, act)) {
2126 /* rf --mo--> act */
2127 future_ordered = true;
2130 if (mo_graph->checkReachable(act, release))
2131 /* act --mo--> release */
2133 if (mo_graph->checkReachable(release, act) &&
2134 mo_graph->checkReachable(act, rf)) {
2135 /* release --mo-> act --mo--> rf */
2136 return true; /* complete */
2138 /* act may break release sequence */
2139 pending->writes.push_back(act);
2142 if (!future_ordered)
2143 certain = false; /* This thread is uncertain */
2147 release_heads->push_back(release);
2148 pending->writes.clear();
2150 pending->release = release;
2157 * An interface for getting the release sequence head(s) with which a
2158 * given ModelAction must synchronize. This function only returns a non-empty
2159 * result when it can locate a release sequence head with certainty. Otherwise,
2160 * it may mark the internal state of the ModelChecker so that it will handle
2161 * the release sequence at a later time, causing @a acquire to update its
2162 * synchronization at some later point in execution.
2164 * @param acquire The 'acquire' action that may synchronize with a release
2166 * @param read The read action that may read from a release sequence; this may
2167 * be the same as acquire, or else an earlier action in the same thread (i.e.,
2168 * when 'acquire' is a fence-acquire)
2169 * @param release_heads A pass-by-reference return parameter. Will be filled
2170 * with the head(s) of the release sequence(s), if they exists with certainty.
2171 * @see ModelChecker::release_seq_heads
2173 void ModelChecker::get_release_seq_heads(ModelAction *acquire,
2174 ModelAction *read, rel_heads_list_t *release_heads)
2176 const ModelAction *rf = read->get_reads_from();
2177 struct release_seq *sequence = (struct release_seq *)snapshot_calloc(1, sizeof(struct release_seq));
2178 sequence->acquire = acquire;
2179 sequence->read = read;
2181 if (!release_seq_heads(rf, release_heads, sequence)) {
2182 /* add act to 'lazy checking' list */
2183 pending_rel_seqs->push_back(sequence);
2185 snapshot_free(sequence);
2190 * Attempt to resolve all stashed operations that might synchronize with a
2191 * release sequence for a given location. This implements the "lazy" portion of
2192 * determining whether or not a release sequence was contiguous, since not all
2193 * modification order information is present at the time an action occurs.
2195 * @param location The location/object that should be checked for release
2196 * sequence resolutions. A NULL value means to check all locations.
2197 * @param work_queue The work queue to which to add work items as they are
2199 * @return True if any updates occurred (new synchronization, new mo_graph
2202 bool ModelChecker::resolve_release_sequences(void *location, work_queue_t *work_queue)
2204 bool updated = false;
2205 std::vector< struct release_seq *, SnapshotAlloc<struct release_seq *> >::iterator it = pending_rel_seqs->begin();
2206 while (it != pending_rel_seqs->end()) {
2207 struct release_seq *pending = *it;
2208 ModelAction *acquire = pending->acquire;
2209 const ModelAction *read = pending->read;
2211 /* Only resolve sequences on the given location, if provided */
2212 if (location && read->get_location() != location) {
2217 const ModelAction *rf = read->get_reads_from();
2218 rel_heads_list_t release_heads;
2220 complete = release_seq_heads(rf, &release_heads, pending);
2221 for (unsigned int i = 0; i < release_heads.size(); i++) {
2222 if (!acquire->has_synchronized_with(release_heads[i])) {
2223 if (acquire->synchronize_with(release_heads[i]))
2226 set_bad_synchronization();
2231 /* Re-check all pending release sequences */
2232 work_queue->push_back(CheckRelSeqWorkEntry(NULL));
2233 /* Re-check read-acquire for mo_graph edges */
2234 if (acquire->is_read())
2235 work_queue->push_back(MOEdgeWorkEntry(acquire));
2237 /* propagate synchronization to later actions */
2238 action_list_t::reverse_iterator rit = action_trace->rbegin();
2239 for (; (*rit) != acquire; rit++) {
2240 ModelAction *propagate = *rit;
2241 if (acquire->happens_before(propagate)) {
2242 propagate->synchronize_with(acquire);
2243 /* Re-check 'propagate' for mo_graph edges */
2244 work_queue->push_back(MOEdgeWorkEntry(propagate));
2249 it = pending_rel_seqs->erase(it);
2250 snapshot_free(pending);
2256 // If we resolved promises or data races, see if we have realized a data race.
2263 * Performs various bookkeeping operations for the current ModelAction. For
2264 * instance, adds action to the per-object, per-thread action vector and to the
2265 * action trace list of all thread actions.
2267 * @param act is the ModelAction to add.
2269 void ModelChecker::add_action_to_lists(ModelAction *act)
2271 int tid = id_to_int(act->get_tid());
2272 ModelAction *uninit = NULL;
2274 action_list_t *list = get_safe_ptr_action(obj_map, act->get_location());
2275 if (list->empty() && act->is_atomic_var()) {
2276 uninit = new_uninitialized_action(act->get_location());
2277 uninit_id = id_to_int(uninit->get_tid());
2278 list->push_back(uninit);
2280 list->push_back(act);
2282 action_trace->push_back(act);
2284 action_trace->push_front(uninit);
2286 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, act->get_location());
2287 if (tid >= (int)vec->size())
2288 vec->resize(priv->next_thread_id);
2289 (*vec)[tid].push_back(act);
2291 (*vec)[uninit_id].push_front(uninit);
2293 if ((int)thrd_last_action->size() <= tid)
2294 thrd_last_action->resize(get_num_threads());
2295 (*thrd_last_action)[tid] = act;
2297 (*thrd_last_action)[uninit_id] = uninit;
2299 if (act->is_fence() && act->is_release()) {
2300 if ((int)thrd_last_fence_release->size() <= tid)
2301 thrd_last_fence_release->resize(get_num_threads());
2302 (*thrd_last_fence_release)[tid] = act;
2305 if (act->is_wait()) {
2306 void *mutex_loc = (void *) act->get_value();
2307 get_safe_ptr_action(obj_map, mutex_loc)->push_back(act);
2309 std::vector<action_list_t> *vec = get_safe_ptr_vect_action(obj_thrd_map, mutex_loc);
2310 if (tid >= (int)vec->size())
2311 vec->resize(priv->next_thread_id);
2312 (*vec)[tid].push_back(act);
2317 * @brief Get the last action performed by a particular Thread
2318 * @param tid The thread ID of the Thread in question
2319 * @return The last action in the thread
2321 ModelAction * ModelChecker::get_last_action(thread_id_t tid) const
2323 int threadid = id_to_int(tid);
2324 if (threadid < (int)thrd_last_action->size())
2325 return (*thrd_last_action)[id_to_int(tid)];
2331 * @brief Get the last fence release performed by a particular Thread
2332 * @param tid The thread ID of the Thread in question
2333 * @return The last fence release in the thread, if one exists; NULL otherwise
2335 ModelAction * ModelChecker::get_last_fence_release(thread_id_t tid) const
2337 int threadid = id_to_int(tid);
2338 if (threadid < (int)thrd_last_fence_release->size())
2339 return (*thrd_last_fence_release)[id_to_int(tid)];
2345 * Gets the last memory_order_seq_cst write (in the total global sequence)
2346 * performed on a particular object (i.e., memory location), not including the
2348 * @param curr The current ModelAction; also denotes the object location to
2350 * @return The last seq_cst write
2352 ModelAction * ModelChecker::get_last_seq_cst_write(ModelAction *curr) const
2354 void *location = curr->get_location();
2355 action_list_t *list = get_safe_ptr_action(obj_map, location);
2356 /* Find: max({i in dom(S) | seq_cst(t_i) && isWrite(t_i) && samevar(t_i, t)}) */
2357 action_list_t::reverse_iterator rit;
2358 for (rit = list->rbegin(); rit != list->rend(); rit++)
2359 if ((*rit)->is_write() && (*rit)->is_seqcst() && (*rit) != curr)
2365 * Gets the last memory_order_seq_cst fence (in the total global sequence)
2366 * performed in a particular thread, prior to a particular fence.
2367 * @param tid The ID of the thread to check
2368 * @param before_fence The fence from which to begin the search; if NULL, then
2369 * search for the most recent fence in the thread.
2370 * @return The last prior seq_cst fence in the thread, if exists; otherwise, NULL
2372 ModelAction * ModelChecker::get_last_seq_cst_fence(thread_id_t tid, const ModelAction *before_fence) const
2374 /* All fences should have NULL location */
2375 action_list_t *list = get_safe_ptr_action(obj_map, NULL);
2376 action_list_t::reverse_iterator rit = list->rbegin();
2379 for (; rit != list->rend(); rit++)
2380 if (*rit == before_fence)
2383 ASSERT(*rit == before_fence);
2387 for (; rit != list->rend(); rit++)
2388 if ((*rit)->is_fence() && (tid == (*rit)->get_tid()) && (*rit)->is_seqcst())
2394 * Gets the last unlock operation performed on a particular mutex (i.e., memory
2395 * location). This function identifies the mutex according to the current
2396 * action, which is presumed to perform on the same mutex.
2397 * @param curr The current ModelAction; also denotes the object location to
2399 * @return The last unlock operation
2401 ModelAction * ModelChecker::get_last_unlock(ModelAction *curr) const
2403 void *location = curr->get_location();
2404 action_list_t *list = get_safe_ptr_action(obj_map, location);
2405 /* Find: max({i in dom(S) | isUnlock(t_i) && samevar(t_i, t)}) */
2406 action_list_t::reverse_iterator rit;
2407 for (rit = list->rbegin(); rit != list->rend(); rit++)
2408 if ((*rit)->is_unlock() || (*rit)->is_wait())
2413 ModelAction * ModelChecker::get_parent_action(thread_id_t tid) const
2415 ModelAction *parent = get_last_action(tid);
2417 parent = get_thread(tid)->get_creation();
2422 * Returns the clock vector for a given thread.
2423 * @param tid The thread whose clock vector we want
2424 * @return Desired clock vector
2426 ClockVector * ModelChecker::get_cv(thread_id_t tid) const
2428 return get_parent_action(tid)->get_cv();
2432 * Resolve a set of Promises with a current write. The set is provided in the
2433 * Node corresponding to @a write.
2434 * @param write The ModelAction that is fulfilling Promises
2435 * @return True if promises were resolved; false otherwise
2437 bool ModelChecker::resolve_promises(ModelAction *write)
2439 bool haveResolved = false;
2440 std::vector< ModelAction *, ModelAlloc<ModelAction *> > actions_to_check;
2441 promise_list_t mustResolve, resolved;
2443 for (unsigned int i = 0, promise_index = 0; promise_index < promises->size(); i++) {
2444 Promise *promise = (*promises)[promise_index];
2445 if (write->get_node()->get_promise(i)) {
2446 ModelAction *read = promise->get_action();
2447 read_from(read, write);
2448 //Make sure the promise's value matches the write's value
2449 ASSERT(promise->is_compatible(write));
2450 mo_graph->resolvePromise(promise, write, &mustResolve);
2452 resolved.push_back(promise);
2453 promises->erase(promises->begin() + promise_index);
2454 actions_to_check.push_back(read);
2456 haveResolved = true;
2461 for (unsigned int i = 0; i < mustResolve.size(); i++) {
2462 if (std::find(resolved.begin(), resolved.end(), mustResolve[i])
2464 priv->failed_promise = true;
2466 for (unsigned int i = 0; i < resolved.size(); i++)
2468 //Check whether reading these writes has made threads unable to
2471 for (unsigned int i = 0; i < actions_to_check.size(); i++) {
2472 ModelAction *read = actions_to_check[i];
2473 mo_check_promises(read, true);
2476 return haveResolved;
2480 * Compute the set of promises that could potentially be satisfied by this
2481 * action. Note that the set computation actually appears in the Node, not in
2483 * @param curr The ModelAction that may satisfy promises
2485 void ModelChecker::compute_promises(ModelAction *curr)
2487 for (unsigned int i = 0; i < promises->size(); i++) {
2488 Promise *promise = (*promises)[i];
2489 const ModelAction *act = promise->get_action();
2490 ASSERT(act->is_read());
2491 if (!act->happens_before(curr) &&
2492 !act->could_synchronize_with(curr) &&
2493 promise->is_compatible(curr) &&
2494 promise->get_value() == curr->get_value()) {
2495 curr->get_node()->set_promise(i, act->is_rmw());
2500 /** Checks promises in response to change in ClockVector Threads. */
2501 void ModelChecker::check_promises(thread_id_t tid, ClockVector *old_cv, ClockVector *merge_cv)
2503 for (unsigned int i = 0; i < promises->size(); i++) {
2504 Promise *promise = (*promises)[i];
2505 const ModelAction *act = promise->get_action();
2506 if ((old_cv == NULL || !old_cv->synchronized_since(act)) &&
2507 merge_cv->synchronized_since(act)) {
2508 if (promise->eliminate_thread(tid)) {
2509 //Promise has failed
2510 priv->failed_promise = true;
2517 void ModelChecker::check_promises_thread_disabled()
2519 for (unsigned int i = 0; i < promises->size(); i++) {
2520 Promise *promise = (*promises)[i];
2521 if (promise->has_failed()) {
2522 priv->failed_promise = true;
2529 * @brief Checks promises in response to addition to modification order for
2532 * We test whether threads are still available for satisfying promises after an
2533 * addition to our modification order constraints. Those that are unavailable
2534 * are "eliminated". Once all threads are eliminated from satisfying a promise,
2535 * that promise has failed.
2537 * @param act The ModelAction which updated the modification order
2538 * @param is_read_check Should be true if act is a read and we must check for
2539 * updates to the store from which it read (there is a distinction here for
2540 * RMW's, which are both a load and a store)
2542 void ModelChecker::mo_check_promises(const ModelAction *act, bool is_read_check)
2544 const ModelAction *write = is_read_check ? act->get_reads_from() : act;
2546 for (unsigned int i = 0; i < promises->size(); i++) {
2547 Promise *promise = (*promises)[i];
2548 const ModelAction *pread = promise->get_action();
2550 // Is this promise on the same location?
2551 if (!pread->same_var(write))
2554 if (pread->happens_before(act) && mo_graph->checkPromise(write, promise)) {
2555 priv->failed_promise = true;
2559 // Don't do any lookups twice for the same thread
2560 if (!promise->thread_is_available(act->get_tid()))
2563 if (mo_graph->checkReachable(promise, write)) {
2564 if (mo_graph->checkPromise(write, promise)) {
2565 priv->failed_promise = true;
2573 * Compute the set of writes that may break the current pending release
2574 * sequence. This information is extracted from previou release sequence
2577 * @param curr The current ModelAction. Must be a release sequence fixup
2580 void ModelChecker::compute_relseq_breakwrites(ModelAction *curr)
2582 if (pending_rel_seqs->empty())
2585 struct release_seq *pending = pending_rel_seqs->back();
2586 for (unsigned int i = 0; i < pending->writes.size(); i++) {
2587 const ModelAction *write = pending->writes[i];
2588 curr->get_node()->add_relseq_break(write);
2591 /* NULL means don't break the sequence; just synchronize */
2592 curr->get_node()->add_relseq_break(NULL);
2596 * Build up an initial set of all past writes that this 'read' action may read
2597 * from, as well as any previously-observed future values that must still be valid.
2599 * @param curr is the current ModelAction that we are exploring; it must be a
2602 void ModelChecker::build_may_read_from(ModelAction *curr)
2604 std::vector<action_list_t> *thrd_lists = get_safe_ptr_vect_action(obj_thrd_map, curr->get_location());
2606 ASSERT(curr->is_read());
2608 ModelAction *last_sc_write = NULL;
2610 if (curr->is_seqcst())
2611 last_sc_write = get_last_seq_cst_write(curr);
2613 /* Iterate over all threads */
2614 for (i = 0; i < thrd_lists->size(); i++) {
2615 /* Iterate over actions in thread, starting from most recent */
2616 action_list_t *list = &(*thrd_lists)[i];
2617 action_list_t::reverse_iterator rit;
2618 for (rit = list->rbegin(); rit != list->rend(); rit++) {
2619 ModelAction *act = *rit;
2621 /* Only consider 'write' actions */
2622 if (!act->is_write() || act == curr)
2625 /* Don't consider more than one seq_cst write if we are a seq_cst read. */
2626 bool allow_read = true;
2628 if (curr->is_seqcst() && (act->is_seqcst() || (last_sc_write != NULL && act->happens_before(last_sc_write))) && act != last_sc_write)
2630 else if (curr->get_sleep_flag() && !curr->is_seqcst() && !sleep_can_read_from(curr, act))
2634 /* Only add feasible reads */
2635 mo_graph->startChanges();
2636 r_modification_order(curr, act);
2637 if (!is_infeasible())
2638 curr->get_node()->add_read_from(act);
2639 mo_graph->rollbackChanges();
2642 /* Include at most one act per-thread that "happens before" curr */
2643 if (act->happens_before(curr))
2648 /* Inherit existing, promised future values */
2649 for (i = 0; i < promises->size(); i++) {
2650 const Promise *promise = (*promises)[i];
2651 const ModelAction *promise_read = promise->get_action();
2652 if (promise_read->same_var(curr)) {
2653 /* Only add feasible future-values */
2654 mo_graph->startChanges();
2655 r_modification_order(curr, promise);
2656 if (!is_infeasible()) {
2657 const struct future_value fv = promise->get_fv();
2658 curr->get_node()->add_future_value(fv);
2660 mo_graph->rollbackChanges();
2664 /* We may find no valid may-read-from only if the execution is doomed */
2665 if (!curr->get_node()->get_read_from_size() && curr->get_node()->future_value_empty()) {
2666 priv->no_valid_reads = true;
2670 if (DBG_ENABLED()) {
2671 model_print("Reached read action:\n");
2673 model_print("Printing may_read_from\n");
2674 curr->get_node()->print_may_read_from();
2675 model_print("End printing may_read_from\n");
2679 bool ModelChecker::sleep_can_read_from(ModelAction *curr, const ModelAction *write)
2681 for ( ; write != NULL; write = write->get_reads_from()) {
2682 /* UNINIT actions don't have a Node, and they never sleep */
2683 if (write->is_uninitialized())
2685 Node *prevnode = write->get_node()->get_parent();
2687 bool thread_sleep = prevnode->enabled_status(curr->get_tid()) == THREAD_SLEEP_SET;
2688 if (write->is_release() && thread_sleep)
2690 if (!write->is_rmw())
2697 * @brief Create a new action representing an uninitialized atomic
2698 * @param location The memory location of the atomic object
2699 * @return A pointer to a new ModelAction
2701 ModelAction * ModelChecker::new_uninitialized_action(void *location) const
2703 ModelAction *act = (ModelAction *)snapshot_malloc(sizeof(class ModelAction));
2704 act = new (act) ModelAction(ATOMIC_UNINIT, std::memory_order_relaxed, location, 0, model_thread);
2705 act->create_cv(NULL);
2709 static void print_list(action_list_t *list)
2711 action_list_t::iterator it;
2713 model_print("---------------------------------------------------------------------\n");
2715 unsigned int hash = 0;
2717 for (it = list->begin(); it != list->end(); it++) {
2719 hash = hash^(hash<<3)^((*it)->hash());
2721 model_print("HASH %u\n", hash);
2722 model_print("---------------------------------------------------------------------\n");
2725 #if SUPPORT_MOD_ORDER_DUMP
2726 void ModelChecker::dumpGraph(char *filename) const
2729 sprintf(buffer, "%s.dot", filename);
2730 FILE *file = fopen(buffer, "w");
2731 fprintf(file, "digraph %s {\n", filename);
2732 mo_graph->dumpNodes(file);
2733 ModelAction **thread_array = (ModelAction **)model_calloc(1, sizeof(ModelAction *) * get_num_threads());
2735 for (action_list_t::iterator it = action_trace->begin(); it != action_trace->end(); it++) {
2736 ModelAction *action = *it;
2737 if (action->is_read()) {
2738 fprintf(file, "N%u [label=\"N%u, T%u\"];\n", action->get_seq_number(), action->get_seq_number(), action->get_tid());
2739 if (action->get_reads_from() != NULL)
2740 fprintf(file, "N%u -> N%u[label=\"rf\", color=red];\n", action->get_seq_number(), action->get_reads_from()->get_seq_number());
2742 if (thread_array[action->get_tid()] != NULL) {
2743 fprintf(file, "N%u -> N%u[label=\"sb\", color=blue];\n", thread_array[action->get_tid()]->get_seq_number(), action->get_seq_number());
2746 thread_array[action->get_tid()] = action;
2748 fprintf(file, "}\n");
2749 model_free(thread_array);
2754 /** @brief Prints an execution trace summary. */
2755 void ModelChecker::print_summary() const
2757 #if SUPPORT_MOD_ORDER_DUMP
2758 char buffername[100];
2759 sprintf(buffername, "exec%04u", stats.num_total);
2760 mo_graph->dumpGraphToFile(buffername);
2761 sprintf(buffername, "graph%04u", stats.num_total);
2762 dumpGraph(buffername);
2765 model_print("Execution %d:", stats.num_total);
2766 if (isfeasibleprefix())
2769 print_infeasibility(" INFEASIBLE");
2770 print_list(action_trace);
2775 * Add a Thread to the system for the first time. Should only be called once
2777 * @param t The Thread to add
2779 void ModelChecker::add_thread(Thread *t)
2781 thread_map->put(id_to_int(t->get_id()), t);
2782 scheduler->add_thread(t);
2786 * Removes a thread from the scheduler.
2787 * @param the thread to remove.
2789 void ModelChecker::remove_thread(Thread *t)
2791 scheduler->remove_thread(t);
2795 * @brief Get a Thread reference by its ID
2796 * @param tid The Thread's ID
2797 * @return A Thread reference
2799 Thread * ModelChecker::get_thread(thread_id_t tid) const
2801 return thread_map->get(id_to_int(tid));
2805 * @brief Get a reference to the Thread in which a ModelAction was executed
2806 * @param act The ModelAction
2807 * @return A Thread reference
2809 Thread * ModelChecker::get_thread(const ModelAction *act) const
2811 return get_thread(act->get_tid());
2815 * @brief Check if a Thread is currently enabled
2816 * @param t The Thread to check
2817 * @return True if the Thread is currently enabled
2819 bool ModelChecker::is_enabled(Thread *t) const
2821 return scheduler->is_enabled(t);
2825 * @brief Check if a Thread is currently enabled
2826 * @param tid The ID of the Thread to check
2827 * @return True if the Thread is currently enabled
2829 bool ModelChecker::is_enabled(thread_id_t tid) const
2831 return scheduler->is_enabled(tid);
2835 * Switch from a model-checker context to a user-thread context. This is the
2836 * complement of ModelChecker::switch_to_master and must be called from the
2837 * model-checker context
2839 * @param thread The user-thread to switch to
2841 void ModelChecker::switch_from_master(Thread *thread)
2843 scheduler->set_current_thread(thread);
2844 Thread::swap(&system_context, thread);
2848 * Switch from a user-context to the "master thread" context (a.k.a. system
2849 * context). This switch is made with the intention of exploring a particular
2850 * model-checking action (described by a ModelAction object). Must be called
2851 * from a user-thread context.
2853 * @param act The current action that will be explored. May be NULL only if
2854 * trace is exiting via an assertion (see ModelChecker::set_assert and
2855 * ModelChecker::has_asserted).
2856 * @return Return the value returned by the current action
2858 uint64_t ModelChecker::switch_to_master(ModelAction *act)
2861 Thread *old = thread_current();
2862 ASSERT(!old->get_pending());
2863 old->set_pending(act);
2864 if (Thread::swap(old, &system_context) < 0) {
2865 perror("swap threads");
2868 return old->get_return_value();
2872 * Takes the next step in the execution, if possible.
2873 * @param curr The current step to take
2874 * @return Returns the next Thread to run, if any; NULL if this execution
2877 Thread * ModelChecker::take_step(ModelAction *curr)
2879 Thread *curr_thrd = get_thread(curr);
2880 ASSERT(curr_thrd->get_state() == THREAD_READY);
2882 curr = check_current_action(curr);
2884 /* Infeasible -> don't take any more steps */
2885 if (is_infeasible())
2887 else if (isfeasibleprefix() && have_bug_reports()) {
2892 if (params.bound != 0 && priv->used_sequence_numbers > params.bound)
2895 if (curr_thrd->is_blocked() || curr_thrd->is_complete())
2896 scheduler->remove_thread(curr_thrd);
2898 Thread *next_thrd = get_next_thread(curr);
2900 DEBUG("(%d, %d)\n", curr_thrd ? id_to_int(curr_thrd->get_id()) : -1,
2901 next_thrd ? id_to_int(next_thrd->get_id()) : -1);
2906 /** Wrapper to run the user's main function, with appropriate arguments */
2907 void user_main_wrapper(void *)
2909 user_main(model->params.argc, model->params.argv);
2912 /** @brief Run ModelChecker for the user program */
2913 void ModelChecker::run()
2917 Thread *t = new Thread(&user_thread, &user_main_wrapper, NULL, NULL);
2922 * Stash next pending action(s) for thread(s). There
2923 * should only need to stash one thread's action--the
2924 * thread which just took a step--plus the first step
2925 * for any newly-created thread
2927 for (unsigned int i = 0; i < get_num_threads(); i++) {
2928 thread_id_t tid = int_to_id(i);
2929 Thread *thr = get_thread(tid);
2930 if (!thr->is_model_thread() && !thr->is_complete() && !thr->get_pending()) {
2931 switch_from_master(thr);
2935 /* Catch assertions from prior take_step or from
2936 * between-ModelAction bugs (e.g., data races) */
2940 /* Consume the next action for a Thread */
2941 ModelAction *curr = t->get_pending();
2942 t->set_pending(NULL);
2943 t = take_step(curr);
2944 } while (t && !t->is_model_thread());
2947 * Launch end-of-execution release sequence fixups only when
2948 * the execution is otherwise feasible AND there are:
2950 * (1) pending release sequences
2951 * (2) pending assertions that could be invalidated by a change
2952 * in clock vectors (i.e., data races)
2953 * (3) no pending promises
2955 while (!pending_rel_seqs->empty() &&
2956 is_feasible_prefix_ignore_relseq() &&
2957 !unrealizedraces.empty()) {
2958 model_print("*** WARNING: release sequence fixup action "
2959 "(%zu pending release seuqence(s)) ***\n",
2960 pending_rel_seqs->size());
2961 ModelAction *fixup = new ModelAction(MODEL_FIXUP_RELSEQ,
2962 std::memory_order_seq_cst, NULL, VALUE_NONE,
2966 } while (next_execution());
2968 model_print("******* Model-checking complete: *******\n");