1 /* ---------------------------------------------------------------------------
3 * (c) The GHC Team, 1998-2006
5 * The scheduler and thread-related functionality
7 * --------------------------------------------------------------------------*/
9 #include "PosixSource.h"
10 #define KEEP_LOCKCLOSURE
15 #include "OSThreads.h"
20 #include "StgMiscClosures.h"
21 #include "Interpreter.h"
23 #include "RtsSignals.h"
29 #include "ThreadLabels.h"
30 #include "LdvProfile.h"
32 #include "Proftimer.h"
37 /* PARALLEL_HASKELL includes go here */
40 #include "Capability.h"
42 #include "AwaitEvent.h"
43 #if defined(mingw32_HOST_OS)
44 #include "win32/IOManager.h"
47 #include "RaiseAsync.h"
49 #include "ThrIOManager.h"
51 #ifdef HAVE_SYS_TYPES_H
52 #include <sys/types.h>
66 // Turn off inlining when debugging - it obfuscates things
69 # define STATIC_INLINE static
72 /* -----------------------------------------------------------------------------
74 * -------------------------------------------------------------------------- */
76 #if !defined(THREADED_RTS)
77 // Blocked/sleeping thrads
78 StgTSO *blocked_queue_hd = NULL;
79 StgTSO *blocked_queue_tl = NULL;
80 StgTSO *sleeping_queue = NULL; // perhaps replace with a hash table?
83 /* Threads blocked on blackholes.
84 * LOCK: sched_mutex+capability, or all capabilities
86 StgTSO *blackhole_queue = NULL;
88 /* The blackhole_queue should be checked for threads to wake up. See
89 * Schedule.h for more thorough comment.
90 * LOCK: none (doesn't matter if we miss an update)
92 rtsBool blackholes_need_checking = rtsFalse;
94 /* Set to true when the latest garbage collection failed to reclaim
95 * enough space, and the runtime should proceed to shut itself down in
96 * an orderly fashion (emitting profiling info etc.)
98 rtsBool heap_overflow = rtsFalse;
100 /* flag that tracks whether we have done any execution in this time slice.
101 * LOCK: currently none, perhaps we should lock (but needs to be
102 * updated in the fast path of the scheduler).
104 * NB. must be StgWord, we do xchg() on it.
106 volatile StgWord recent_activity = ACTIVITY_YES;
108 /* if this flag is set as well, give up execution
109 * LOCK: none (changes monotonically)
111 volatile StgWord sched_state = SCHED_RUNNING;
113 /* This is used in `TSO.h' and gcc 2.96 insists that this variable actually
114 * exists - earlier gccs apparently didn't.
120 * Set to TRUE when entering a shutdown state (via shutdownHaskellAndExit()) --
121 * in an MT setting, needed to signal that a worker thread shouldn't hang around
122 * in the scheduler when it is out of work.
124 rtsBool shutting_down_scheduler = rtsFalse;
127 * This mutex protects most of the global scheduler data in
128 * the THREADED_RTS runtime.
130 #if defined(THREADED_RTS)
134 #if !defined(mingw32_HOST_OS)
135 #define FORKPROCESS_PRIMOP_SUPPORTED
138 /* -----------------------------------------------------------------------------
139 * static function prototypes
140 * -------------------------------------------------------------------------- */
142 static Capability *schedule (Capability *initialCapability, Task *task);
145 // These function all encapsulate parts of the scheduler loop, and are
146 // abstracted only to make the structure and control flow of the
147 // scheduler clearer.
149 static void schedulePreLoop (void);
150 static void scheduleFindWork (Capability *cap);
151 #if defined(THREADED_RTS)
152 static void scheduleYield (Capability **pcap, Task *task);
154 static void scheduleStartSignalHandlers (Capability *cap);
155 static void scheduleCheckBlockedThreads (Capability *cap);
156 static void scheduleCheckWakeupThreads(Capability *cap USED_IF_NOT_THREADS);
157 static void scheduleCheckBlackHoles (Capability *cap);
158 static void scheduleDetectDeadlock (Capability *cap, Task *task);
159 static void schedulePushWork(Capability *cap, Task *task);
160 #if defined(PARALLEL_HASKELL)
161 static rtsBool scheduleGetRemoteWork(Capability *cap);
162 static void scheduleSendPendingMessages(void);
164 #if defined(PARALLEL_HASKELL) || defined(THREADED_RTS)
165 static void scheduleActivateSpark(Capability *cap);
167 static void schedulePostRunThread(Capability *cap, StgTSO *t);
168 static rtsBool scheduleHandleHeapOverflow( Capability *cap, StgTSO *t );
169 static void scheduleHandleStackOverflow( Capability *cap, Task *task,
171 static rtsBool scheduleHandleYield( Capability *cap, StgTSO *t,
172 nat prev_what_next );
173 static void scheduleHandleThreadBlocked( StgTSO *t );
174 static rtsBool scheduleHandleThreadFinished( Capability *cap, Task *task,
176 static rtsBool scheduleNeedHeapProfile(rtsBool ready_to_gc);
177 static Capability *scheduleDoGC(Capability *cap, Task *task,
178 rtsBool force_major);
180 static rtsBool checkBlackHoles(Capability *cap);
182 static StgTSO *threadStackOverflow(Capability *cap, StgTSO *tso);
183 static StgTSO *threadStackUnderflow(Task *task, StgTSO *tso);
185 static void deleteThread (Capability *cap, StgTSO *tso);
186 static void deleteAllThreads (Capability *cap);
188 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
189 static void deleteThread_(Capability *cap, StgTSO *tso);
193 static char *whatNext_strs[] = {
203 /* -----------------------------------------------------------------------------
204 * Putting a thread on the run queue: different scheduling policies
205 * -------------------------------------------------------------------------- */
208 addToRunQueue( Capability *cap, StgTSO *t )
210 #if defined(PARALLEL_HASKELL)
211 if (RtsFlags.ParFlags.doFairScheduling) {
212 // this does round-robin scheduling; good for concurrency
213 appendToRunQueue(cap,t);
215 // this does unfair scheduling; good for parallelism
216 pushOnRunQueue(cap,t);
219 // this does round-robin scheduling; good for concurrency
220 appendToRunQueue(cap,t);
224 /* ---------------------------------------------------------------------------
225 Main scheduling loop.
227 We use round-robin scheduling, each thread returning to the
228 scheduler loop when one of these conditions is detected:
231 * timer expires (thread yields)
237 In a GranSim setup this loop iterates over the global event queue.
238 This revolves around the global event queue, which determines what
239 to do next. Therefore, it's more complicated than either the
240 concurrent or the parallel (GUM) setup.
241 This version has been entirely removed (JB 2008/08).
244 GUM iterates over incoming messages.
245 It starts with nothing to do (thus CurrentTSO == END_TSO_QUEUE),
246 and sends out a fish whenever it has nothing to do; in-between
247 doing the actual reductions (shared code below) it processes the
248 incoming messages and deals with delayed operations
249 (see PendingFetches).
250 This is not the ugliest code you could imagine, but it's bloody close.
252 (JB 2008/08) This version was formerly indicated by a PP-Flag PAR,
253 now by PP-flag PARALLEL_HASKELL. The Eden RTS (in GHC-6.x) uses it,
254 as well as future GUM versions. This file has been refurbished to
255 only contain valid code, which is however incomplete, refers to
256 invalid includes etc.
258 ------------------------------------------------------------------------ */
261 schedule (Capability *initialCapability, Task *task)
265 StgThreadReturnCode ret;
266 #if defined(PARALLEL_HASKELL)
267 rtsBool receivedFinish = rtsFalse;
271 #if defined(THREADED_RTS)
272 rtsBool first = rtsTrue;
275 cap = initialCapability;
277 // Pre-condition: this task owns initialCapability.
278 // The sched_mutex is *NOT* held
279 // NB. on return, we still hold a capability.
281 debugTrace (DEBUG_sched,
282 "### NEW SCHEDULER LOOP (task: %p, cap: %p)",
283 task, initialCapability);
285 if (running_finalizers) {
286 errorBelch("error: a C finalizer called back into Haskell.\n"
287 " use Foreign.Concurrent.newForeignPtr for Haskell finalizers.");
288 stg_exit(EXIT_FAILURE);
293 // -----------------------------------------------------------
294 // Scheduler loop starts here:
296 #if defined(PARALLEL_HASKELL)
297 #define TERMINATION_CONDITION (!receivedFinish)
299 #define TERMINATION_CONDITION rtsTrue
302 while (TERMINATION_CONDITION) {
304 // Check whether we have re-entered the RTS from Haskell without
305 // going via suspendThread()/resumeThread (i.e. a 'safe' foreign
307 if (cap->in_haskell) {
308 errorBelch("schedule: re-entered unsafely.\n"
309 " Perhaps a 'foreign import unsafe' should be 'safe'?");
310 stg_exit(EXIT_FAILURE);
313 // The interruption / shutdown sequence.
315 // In order to cleanly shut down the runtime, we want to:
316 // * make sure that all main threads return to their callers
317 // with the state 'Interrupted'.
318 // * clean up all OS threads assocated with the runtime
319 // * free all memory etc.
321 // So the sequence for ^C goes like this:
323 // * ^C handler sets sched_state := SCHED_INTERRUPTING and
324 // arranges for some Capability to wake up
326 // * all threads in the system are halted, and the zombies are
327 // placed on the run queue for cleaning up. We acquire all
328 // the capabilities in order to delete the threads, this is
329 // done by scheduleDoGC() for convenience (because GC already
330 // needs to acquire all the capabilities). We can't kill
331 // threads involved in foreign calls.
333 // * somebody calls shutdownHaskell(), which calls exitScheduler()
335 // * sched_state := SCHED_SHUTTING_DOWN
337 // * all workers exit when the run queue on their capability
338 // drains. All main threads will also exit when their TSO
339 // reaches the head of the run queue and they can return.
341 // * eventually all Capabilities will shut down, and the RTS can
344 // * We might be left with threads blocked in foreign calls,
345 // we should really attempt to kill these somehow (TODO);
347 switch (sched_state) {
350 case SCHED_INTERRUPTING:
351 debugTrace(DEBUG_sched, "SCHED_INTERRUPTING");
352 #if defined(THREADED_RTS)
353 discardSparksCap(cap);
355 /* scheduleDoGC() deletes all the threads */
356 cap = scheduleDoGC(cap,task,rtsFalse);
358 // after scheduleDoGC(), we must be shutting down. Either some
359 // other Capability did the final GC, or we did it above,
360 // either way we can fall through to the SCHED_SHUTTING_DOWN
362 ASSERT(sched_state == SCHED_SHUTTING_DOWN);
365 case SCHED_SHUTTING_DOWN:
366 debugTrace(DEBUG_sched, "SCHED_SHUTTING_DOWN");
367 // If we are a worker, just exit. If we're a bound thread
368 // then we will exit below when we've removed our TSO from
370 if (task->tso == NULL && emptyRunQueue(cap)) {
375 barf("sched_state: %d", sched_state);
378 scheduleFindWork(cap);
380 /* work pushing, currently relevant only for THREADED_RTS:
381 (pushes threads, wakes up idle capabilities for stealing) */
382 schedulePushWork(cap,task);
384 #if defined(PARALLEL_HASKELL)
385 /* since we perform a blocking receive and continue otherwise,
386 either we never reach here or we definitely have work! */
387 // from here: non-empty run queue
388 ASSERT(!emptyRunQueue(cap));
390 if (PacketsWaiting()) { /* now process incoming messages, if any
393 CAUTION: scheduleGetRemoteWork called
394 above, waits for messages as well! */
395 processMessages(cap, &receivedFinish);
397 #endif // PARALLEL_HASKELL: non-empty run queue!
399 scheduleDetectDeadlock(cap,task);
401 #if defined(THREADED_RTS)
402 cap = task->cap; // reload cap, it might have changed
405 // Normally, the only way we can get here with no threads to
406 // run is if a keyboard interrupt received during
407 // scheduleCheckBlockedThreads() or scheduleDetectDeadlock().
408 // Additionally, it is not fatal for the
409 // threaded RTS to reach here with no threads to run.
411 // win32: might be here due to awaitEvent() being abandoned
412 // as a result of a console event having been delivered.
414 #if defined(THREADED_RTS)
418 // // don't yield the first time, we want a chance to run this
419 // // thread for a bit, even if there are others banging at the
422 // ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
426 scheduleYield(&cap,task);
427 if (emptyRunQueue(cap)) continue; // look for work again
430 #if !defined(THREADED_RTS) && !defined(mingw32_HOST_OS)
431 if ( emptyRunQueue(cap) ) {
432 ASSERT(sched_state >= SCHED_INTERRUPTING);
437 // Get a thread to run
439 t = popRunQueue(cap);
441 // Sanity check the thread we're about to run. This can be
442 // expensive if there is lots of thread switching going on...
443 IF_DEBUG(sanity,checkTSO(t));
445 #if defined(THREADED_RTS)
446 // Check whether we can run this thread in the current task.
447 // If not, we have to pass our capability to the right task.
449 Task *bound = t->bound;
453 debugTrace(DEBUG_sched,
454 "### Running thread %lu in bound thread", (unsigned long)t->id);
455 // yes, the Haskell thread is bound to the current native thread
457 debugTrace(DEBUG_sched,
458 "### thread %lu bound to another OS thread", (unsigned long)t->id);
459 // no, bound to a different Haskell thread: pass to that thread
460 pushOnRunQueue(cap,t);
464 // The thread we want to run is unbound.
466 debugTrace(DEBUG_sched,
467 "### this OS thread cannot run thread %lu", (unsigned long)t->id);
468 // no, the current native thread is bound to a different
469 // Haskell thread, so pass it to any worker thread
470 pushOnRunQueue(cap,t);
477 // If we're shutting down, and this thread has not yet been
478 // killed, kill it now. This sometimes happens when a finalizer
479 // thread is created by the final GC, or a thread previously
480 // in a foreign call returns.
481 if (sched_state >= SCHED_INTERRUPTING &&
482 !(t->what_next == ThreadComplete || t->what_next == ThreadKilled)) {
486 /* context switches are initiated by the timer signal, unless
487 * the user specified "context switch as often as possible", with
490 if (RtsFlags.ConcFlags.ctxtSwitchTicks == 0
491 && !emptyThreadQueues(cap)) {
492 cap->context_switch = 1;
497 // CurrentTSO is the thread to run. t might be different if we
498 // loop back to run_thread, so make sure to set CurrentTSO after
500 cap->r.rCurrentTSO = t;
502 debugTrace(DEBUG_sched, "-->> running thread %ld %s ...",
503 (long)t->id, whatNext_strs[t->what_next]);
505 startHeapProfTimer();
507 // Check for exceptions blocked on this thread
508 maybePerformBlockedException (cap, t);
510 // ----------------------------------------------------------------------
511 // Run the current thread
513 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
514 ASSERT(t->cap == cap);
515 ASSERT(t->bound ? t->bound->cap == cap : 1);
517 prev_what_next = t->what_next;
519 errno = t->saved_errno;
521 SetLastError(t->saved_winerror);
524 cap->in_haskell = rtsTrue;
528 #if defined(THREADED_RTS)
529 if (recent_activity == ACTIVITY_DONE_GC) {
530 // ACTIVITY_DONE_GC means we turned off the timer signal to
531 // conserve power (see #1623). Re-enable it here.
533 prev = xchg((P_)&recent_activity, ACTIVITY_YES);
534 if (prev == ACTIVITY_DONE_GC) {
538 recent_activity = ACTIVITY_YES;
542 switch (prev_what_next) {
546 /* Thread already finished, return to scheduler. */
547 ret = ThreadFinished;
553 r = StgRun((StgFunPtr) stg_returnToStackTop, &cap->r);
554 cap = regTableToCapability(r);
559 case ThreadInterpret:
560 cap = interpretBCO(cap);
565 barf("schedule: invalid what_next field");
568 cap->in_haskell = rtsFalse;
570 // The TSO might have moved, eg. if it re-entered the RTS and a GC
571 // happened. So find the new location:
572 t = cap->r.rCurrentTSO;
574 // We have run some Haskell code: there might be blackhole-blocked
575 // threads to wake up now.
576 // Lock-free test here should be ok, we're just setting a flag.
577 if ( blackhole_queue != END_TSO_QUEUE ) {
578 blackholes_need_checking = rtsTrue;
581 // And save the current errno in this thread.
582 // XXX: possibly bogus for SMP because this thread might already
583 // be running again, see code below.
584 t->saved_errno = errno;
586 // Similarly for Windows error code
587 t->saved_winerror = GetLastError();
590 #if defined(THREADED_RTS)
591 // If ret is ThreadBlocked, and this Task is bound to the TSO that
592 // blocked, we are in limbo - the TSO is now owned by whatever it
593 // is blocked on, and may in fact already have been woken up,
594 // perhaps even on a different Capability. It may be the case
595 // that task->cap != cap. We better yield this Capability
596 // immediately and return to normaility.
597 if (ret == ThreadBlocked) {
598 debugTrace(DEBUG_sched,
599 "--<< thread %lu (%s) stopped: blocked",
600 (unsigned long)t->id, whatNext_strs[t->what_next]);
605 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
606 ASSERT(t->cap == cap);
608 // ----------------------------------------------------------------------
610 // Costs for the scheduler are assigned to CCS_SYSTEM
612 #if defined(PROFILING)
616 schedulePostRunThread(cap,t);
618 t = threadStackUnderflow(task,t);
620 ready_to_gc = rtsFalse;
624 ready_to_gc = scheduleHandleHeapOverflow(cap,t);
628 scheduleHandleStackOverflow(cap,task,t);
632 if (scheduleHandleYield(cap, t, prev_what_next)) {
633 // shortcut for switching between compiler/interpreter:
639 scheduleHandleThreadBlocked(t);
643 if (scheduleHandleThreadFinished(cap, task, t)) return cap;
644 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
648 barf("schedule: invalid thread return code %d", (int)ret);
651 if (ready_to_gc || scheduleNeedHeapProfile(ready_to_gc)) {
652 cap = scheduleDoGC(cap,task,rtsFalse);
654 } /* end of while() */
657 /* ----------------------------------------------------------------------------
658 * Setting up the scheduler loop
659 * ------------------------------------------------------------------------- */
662 schedulePreLoop(void)
664 // initialisation for scheduler - what cannot go into initScheduler()
667 /* -----------------------------------------------------------------------------
670 * Search for work to do, and handle messages from elsewhere.
671 * -------------------------------------------------------------------------- */
674 scheduleFindWork (Capability *cap)
676 scheduleStartSignalHandlers(cap);
678 // Only check the black holes here if we've nothing else to do.
679 // During normal execution, the black hole list only gets checked
680 // at GC time, to avoid repeatedly traversing this possibly long
681 // list each time around the scheduler.
682 if (emptyRunQueue(cap)) { scheduleCheckBlackHoles(cap); }
684 scheduleCheckWakeupThreads(cap);
686 scheduleCheckBlockedThreads(cap);
688 #if defined(THREADED_RTS) || defined(PARALLEL_HASKELL)
689 if (emptyRunQueue(cap)) { scheduleActivateSpark(cap); }
692 #if defined(PARALLEL_HASKELL)
693 // if messages have been buffered...
694 scheduleSendPendingMessages();
697 #if defined(PARALLEL_HASKELL)
698 if (emptyRunQueue(cap)) {
699 receivedFinish = scheduleGetRemoteWork(cap);
700 continue; // a new round, (hopefully) with new work
702 in GUM, this a) sends out a FISH and returns IF no fish is
704 b) (blocking) awaits and receives messages
706 in Eden, this is only the blocking receive, as b) in GUM.
712 #if defined(THREADED_RTS)
713 STATIC_INLINE rtsBool
714 shouldYieldCapability (Capability *cap, Task *task)
716 // we need to yield this capability to someone else if..
717 // - another thread is initiating a GC
718 // - another Task is returning from a foreign call
719 // - the thread at the head of the run queue cannot be run
720 // by this Task (it is bound to another Task, or it is unbound
721 // and this task it bound).
722 return (waiting_for_gc ||
723 cap->returning_tasks_hd != NULL ||
724 (!emptyRunQueue(cap) && (task->tso == NULL
725 ? cap->run_queue_hd->bound != NULL
726 : cap->run_queue_hd->bound != task)));
729 // This is the single place where a Task goes to sleep. There are
730 // two reasons it might need to sleep:
731 // - there are no threads to run
732 // - we need to yield this Capability to someone else
733 // (see shouldYieldCapability())
735 // Careful: the scheduler loop is quite delicate. Make sure you run
736 // the tests in testsuite/concurrent (all ways) after modifying this,
737 // and also check the benchmarks in nofib/parallel for regressions.
740 scheduleYield (Capability **pcap, Task *task)
742 Capability *cap = *pcap;
744 // if we have work, and we don't need to give up the Capability, continue.
745 if (!shouldYieldCapability(cap,task) &&
746 (!emptyRunQueue(cap) ||
747 blackholes_need_checking ||
748 sched_state >= SCHED_INTERRUPTING))
751 // otherwise yield (sleep), and keep yielding if necessary.
753 yieldCapability(&cap,task);
755 while (shouldYieldCapability(cap,task));
757 // note there may still be no threads on the run queue at this
758 // point, the caller has to check.
765 /* -----------------------------------------------------------------------------
768 * Push work to other Capabilities if we have some.
769 * -------------------------------------------------------------------------- */
772 schedulePushWork(Capability *cap USED_IF_THREADS,
773 Task *task USED_IF_THREADS)
775 /* following code not for PARALLEL_HASKELL. I kept the call general,
776 future GUM versions might use pushing in a distributed setup */
777 #if defined(THREADED_RTS)
779 Capability *free_caps[n_capabilities], *cap0;
782 // migration can be turned off with +RTS -qg
783 if (!RtsFlags.ParFlags.migrate) return;
785 // Check whether we have more threads on our run queue, or sparks
786 // in our pool, that we could hand to another Capability.
787 if (cap->run_queue_hd == END_TSO_QUEUE) {
788 if (sparkPoolSizeCap(cap) < 2) return;
790 if (cap->run_queue_hd->_link == END_TSO_QUEUE &&
791 sparkPoolSizeCap(cap) < 1) return;
794 // First grab as many free Capabilities as we can.
795 for (i=0, n_free_caps=0; i < n_capabilities; i++) {
796 cap0 = &capabilities[i];
797 if (cap != cap0 && tryGrabCapability(cap0,task)) {
798 if (!emptyRunQueue(cap0) || cap->returning_tasks_hd != NULL) {
799 // it already has some work, we just grabbed it at
800 // the wrong moment. Or maybe it's deadlocked!
801 releaseCapability(cap0);
803 free_caps[n_free_caps++] = cap0;
808 // we now have n_free_caps free capabilities stashed in
809 // free_caps[]. Share our run queue equally with them. This is
810 // probably the simplest thing we could do; improvements we might
811 // want to do include:
813 // - giving high priority to moving relatively new threads, on
814 // the gournds that they haven't had time to build up a
815 // working set in the cache on this CPU/Capability.
817 // - giving low priority to moving long-lived threads
819 if (n_free_caps > 0) {
820 StgTSO *prev, *t, *next;
821 rtsBool pushed_to_all;
823 debugTrace(DEBUG_sched,
824 "cap %d: %s and %d free capabilities, sharing...",
826 (!emptyRunQueue(cap) && cap->run_queue_hd->_link != END_TSO_QUEUE)?
827 "excess threads on run queue":"sparks to share (>=2)",
831 pushed_to_all = rtsFalse;
833 if (cap->run_queue_hd != END_TSO_QUEUE) {
834 prev = cap->run_queue_hd;
836 prev->_link = END_TSO_QUEUE;
837 for (; t != END_TSO_QUEUE; t = next) {
839 t->_link = END_TSO_QUEUE;
840 if (t->what_next == ThreadRelocated
841 || t->bound == task // don't move my bound thread
842 || tsoLocked(t)) { // don't move a locked thread
843 setTSOLink(cap, prev, t);
845 } else if (i == n_free_caps) {
846 pushed_to_all = rtsTrue;
849 setTSOLink(cap, prev, t);
852 debugTrace(DEBUG_sched, "pushing thread %lu to capability %d", (unsigned long)t->id, free_caps[i]->no);
853 appendToRunQueue(free_caps[i],t);
854 if (t->bound) { t->bound->cap = free_caps[i]; }
855 t->cap = free_caps[i];
859 cap->run_queue_tl = prev;
863 /* JB I left this code in place, it would work but is not necessary */
865 // If there are some free capabilities that we didn't push any
866 // threads to, then try to push a spark to each one.
867 if (!pushed_to_all) {
869 // i is the next free capability to push to
870 for (; i < n_free_caps; i++) {
871 if (emptySparkPoolCap(free_caps[i])) {
872 spark = tryStealSpark(cap->sparks);
874 debugTrace(DEBUG_sched, "pushing spark %p to capability %d", spark, free_caps[i]->no);
875 newSpark(&(free_caps[i]->r), spark);
880 #endif /* SPARK_PUSHING */
882 // release the capabilities
883 for (i = 0; i < n_free_caps; i++) {
884 task->cap = free_caps[i];
885 releaseAndWakeupCapability(free_caps[i]);
888 task->cap = cap; // reset to point to our Capability.
890 #endif /* THREADED_RTS */
894 /* ----------------------------------------------------------------------------
895 * Start any pending signal handlers
896 * ------------------------------------------------------------------------- */
898 #if defined(RTS_USER_SIGNALS) && !defined(THREADED_RTS)
900 scheduleStartSignalHandlers(Capability *cap)
902 if (RtsFlags.MiscFlags.install_signal_handlers && signals_pending()) {
903 // safe outside the lock
904 startSignalHandlers(cap);
909 scheduleStartSignalHandlers(Capability *cap STG_UNUSED)
914 /* ----------------------------------------------------------------------------
915 * Check for blocked threads that can be woken up.
916 * ------------------------------------------------------------------------- */
919 scheduleCheckBlockedThreads(Capability *cap USED_IF_NOT_THREADS)
921 #if !defined(THREADED_RTS)
923 // Check whether any waiting threads need to be woken up. If the
924 // run queue is empty, and there are no other tasks running, we
925 // can wait indefinitely for something to happen.
927 if ( !emptyQueue(blocked_queue_hd) || !emptyQueue(sleeping_queue) )
929 awaitEvent( emptyRunQueue(cap) && !blackholes_need_checking );
935 /* ----------------------------------------------------------------------------
936 * Check for threads woken up by other Capabilities
937 * ------------------------------------------------------------------------- */
940 scheduleCheckWakeupThreads(Capability *cap USED_IF_THREADS)
942 #if defined(THREADED_RTS)
943 // Any threads that were woken up by other Capabilities get
944 // appended to our run queue.
945 if (!emptyWakeupQueue(cap)) {
946 ACQUIRE_LOCK(&cap->lock);
947 if (emptyRunQueue(cap)) {
948 cap->run_queue_hd = cap->wakeup_queue_hd;
949 cap->run_queue_tl = cap->wakeup_queue_tl;
951 setTSOLink(cap, cap->run_queue_tl, cap->wakeup_queue_hd);
952 cap->run_queue_tl = cap->wakeup_queue_tl;
954 cap->wakeup_queue_hd = cap->wakeup_queue_tl = END_TSO_QUEUE;
955 RELEASE_LOCK(&cap->lock);
960 /* ----------------------------------------------------------------------------
961 * Check for threads blocked on BLACKHOLEs that can be woken up
962 * ------------------------------------------------------------------------- */
964 scheduleCheckBlackHoles (Capability *cap)
966 if ( blackholes_need_checking ) // check without the lock first
968 ACQUIRE_LOCK(&sched_mutex);
969 if ( blackholes_need_checking ) {
970 blackholes_need_checking = rtsFalse;
971 // important that we reset the flag *before* checking the
972 // blackhole queue, otherwise we could get deadlock. This
973 // happens as follows: we wake up a thread that
974 // immediately runs on another Capability, blocks on a
975 // blackhole, and then we reset the blackholes_need_checking flag.
976 checkBlackHoles(cap);
978 RELEASE_LOCK(&sched_mutex);
982 /* ----------------------------------------------------------------------------
983 * Detect deadlock conditions and attempt to resolve them.
984 * ------------------------------------------------------------------------- */
987 scheduleDetectDeadlock (Capability *cap, Task *task)
990 #if defined(PARALLEL_HASKELL)
991 // ToDo: add deadlock detection in GUM (similar to THREADED_RTS) -- HWL
996 * Detect deadlock: when we have no threads to run, there are no
997 * threads blocked, waiting for I/O, or sleeping, and all the
998 * other tasks are waiting for work, we must have a deadlock of
1001 if ( emptyThreadQueues(cap) )
1003 #if defined(THREADED_RTS)
1005 * In the threaded RTS, we only check for deadlock if there
1006 * has been no activity in a complete timeslice. This means
1007 * we won't eagerly start a full GC just because we don't have
1008 * any threads to run currently.
1010 if (recent_activity != ACTIVITY_INACTIVE) return;
1013 debugTrace(DEBUG_sched, "deadlocked, forcing major GC...");
1015 // Garbage collection can release some new threads due to
1016 // either (a) finalizers or (b) threads resurrected because
1017 // they are unreachable and will therefore be sent an
1018 // exception. Any threads thus released will be immediately
1020 cap = scheduleDoGC (cap, task, rtsTrue/*force major GC*/);
1021 // when force_major == rtsTrue. scheduleDoGC sets
1022 // recent_activity to ACTIVITY_DONE_GC and turns off the timer
1025 if ( !emptyRunQueue(cap) ) return;
1027 #if defined(RTS_USER_SIGNALS) && !defined(THREADED_RTS)
1028 /* If we have user-installed signal handlers, then wait
1029 * for signals to arrive rather then bombing out with a
1032 if ( RtsFlags.MiscFlags.install_signal_handlers && anyUserHandlers() ) {
1033 debugTrace(DEBUG_sched,
1034 "still deadlocked, waiting for signals...");
1038 if (signals_pending()) {
1039 startSignalHandlers(cap);
1042 // either we have threads to run, or we were interrupted:
1043 ASSERT(!emptyRunQueue(cap) || sched_state >= SCHED_INTERRUPTING);
1049 #if !defined(THREADED_RTS)
1050 /* Probably a real deadlock. Send the current main thread the
1051 * Deadlock exception.
1054 switch (task->tso->why_blocked) {
1056 case BlockedOnBlackHole:
1057 case BlockedOnException:
1059 throwToSingleThreaded(cap, task->tso,
1060 (StgClosure *)nonTermination_closure);
1063 barf("deadlock: main thread blocked in a strange way");
1072 /* ----------------------------------------------------------------------------
1073 * Send pending messages (PARALLEL_HASKELL only)
1074 * ------------------------------------------------------------------------- */
1076 #if defined(PARALLEL_HASKELL)
1078 scheduleSendPendingMessages(void)
1081 # if defined(PAR) // global Mem.Mgmt., omit for now
1082 if (PendingFetches != END_BF_QUEUE) {
1087 if (RtsFlags.ParFlags.BufferTime) {
1088 // if we use message buffering, we must send away all message
1089 // packets which have become too old...
1095 /* ----------------------------------------------------------------------------
1096 * Activate spark threads (PARALLEL_HASKELL and THREADED_RTS)
1097 * ------------------------------------------------------------------------- */
1099 #if defined(PARALLEL_HASKELL) || defined(THREADED_RTS)
1101 scheduleActivateSpark(Capability *cap)
1105 createSparkThread(cap);
1106 debugTrace(DEBUG_sched, "creating a spark thread");
1109 #endif // PARALLEL_HASKELL || THREADED_RTS
1111 /* ----------------------------------------------------------------------------
1112 * Get work from a remote node (PARALLEL_HASKELL only)
1113 * ------------------------------------------------------------------------- */
1115 #if defined(PARALLEL_HASKELL)
1116 static rtsBool /* return value used in PARALLEL_HASKELL only */
1117 scheduleGetRemoteWork (Capability *cap STG_UNUSED)
1119 #if defined(PARALLEL_HASKELL)
1120 rtsBool receivedFinish = rtsFalse;
1122 // idle() , i.e. send all buffers, wait for work
1123 if (RtsFlags.ParFlags.BufferTime) {
1124 IF_PAR_DEBUG(verbose,
1125 debugBelch("...send all pending data,"));
1128 for (i=1; i<=nPEs; i++)
1129 sendImmediately(i); // send all messages away immediately
1133 /* this would be the place for fishing in GUM...
1135 if (no-earlier-fish-around)
1136 sendFish(choosePe());
1139 // Eden:just look for incoming messages (blocking receive)
1140 IF_PAR_DEBUG(verbose,
1141 debugBelch("...wait for incoming messages...\n"));
1142 processMessages(cap, &receivedFinish); // blocking receive...
1145 return receivedFinish;
1146 // reenter scheduling look after having received something
1148 #else /* !PARALLEL_HASKELL, i.e. THREADED_RTS */
1150 return rtsFalse; /* return value unused in THREADED_RTS */
1152 #endif /* PARALLEL_HASKELL */
1154 #endif // PARALLEL_HASKELL || THREADED_RTS
1156 /* ----------------------------------------------------------------------------
1157 * After running a thread...
1158 * ------------------------------------------------------------------------- */
1161 schedulePostRunThread (Capability *cap, StgTSO *t)
1163 // We have to be able to catch transactions that are in an
1164 // infinite loop as a result of seeing an inconsistent view of
1168 // [a,b] <- mapM readTVar [ta,tb]
1169 // when (a == b) loop
1171 // and a is never equal to b given a consistent view of memory.
1173 if (t -> trec != NO_TREC && t -> why_blocked == NotBlocked) {
1174 if (!stmValidateNestOfTransactions (t -> trec)) {
1175 debugTrace(DEBUG_sched | DEBUG_stm,
1176 "trec %p found wasting its time", t);
1178 // strip the stack back to the
1179 // ATOMICALLY_FRAME, aborting the (nested)
1180 // transaction, and saving the stack of any
1181 // partially-evaluated thunks on the heap.
1182 throwToSingleThreaded_(cap, t, NULL, rtsTrue);
1184 ASSERT(get_itbl((StgClosure *)t->sp)->type == ATOMICALLY_FRAME);
1188 /* some statistics gathering in the parallel case */
1191 /* -----------------------------------------------------------------------------
1192 * Handle a thread that returned to the scheduler with ThreadHeepOverflow
1193 * -------------------------------------------------------------------------- */
1196 scheduleHandleHeapOverflow( Capability *cap, StgTSO *t )
1198 // did the task ask for a large block?
1199 if (cap->r.rHpAlloc > BLOCK_SIZE) {
1200 // if so, get one and push it on the front of the nursery.
1204 blocks = (lnat)BLOCK_ROUND_UP(cap->r.rHpAlloc) / BLOCK_SIZE;
1206 debugTrace(DEBUG_sched,
1207 "--<< thread %ld (%s) stopped: requesting a large block (size %ld)\n",
1208 (long)t->id, whatNext_strs[t->what_next], blocks);
1210 // don't do this if the nursery is (nearly) full, we'll GC first.
1211 if (cap->r.rCurrentNursery->link != NULL ||
1212 cap->r.rNursery->n_blocks == 1) { // paranoia to prevent infinite loop
1213 // if the nursery has only one block.
1216 bd = allocGroup( blocks );
1218 cap->r.rNursery->n_blocks += blocks;
1220 // link the new group into the list
1221 bd->link = cap->r.rCurrentNursery;
1222 bd->u.back = cap->r.rCurrentNursery->u.back;
1223 if (cap->r.rCurrentNursery->u.back != NULL) {
1224 cap->r.rCurrentNursery->u.back->link = bd;
1226 #if !defined(THREADED_RTS)
1227 ASSERT(g0s0->blocks == cap->r.rCurrentNursery &&
1228 g0s0 == cap->r.rNursery);
1230 cap->r.rNursery->blocks = bd;
1232 cap->r.rCurrentNursery->u.back = bd;
1234 // initialise it as a nursery block. We initialise the
1235 // step, gen_no, and flags field of *every* sub-block in
1236 // this large block, because this is easier than making
1237 // sure that we always find the block head of a large
1238 // block whenever we call Bdescr() (eg. evacuate() and
1239 // isAlive() in the GC would both have to do this, at
1243 for (x = bd; x < bd + blocks; x++) {
1244 x->step = cap->r.rNursery;
1250 // This assert can be a killer if the app is doing lots
1251 // of large block allocations.
1252 IF_DEBUG(sanity, checkNurserySanity(cap->r.rNursery));
1254 // now update the nursery to point to the new block
1255 cap->r.rCurrentNursery = bd;
1257 // we might be unlucky and have another thread get on the
1258 // run queue before us and steal the large block, but in that
1259 // case the thread will just end up requesting another large
1261 pushOnRunQueue(cap,t);
1262 return rtsFalse; /* not actually GC'ing */
1266 debugTrace(DEBUG_sched,
1267 "--<< thread %ld (%s) stopped: HeapOverflow",
1268 (long)t->id, whatNext_strs[t->what_next]);
1270 if (cap->context_switch) {
1271 // Sometimes we miss a context switch, e.g. when calling
1272 // primitives in a tight loop, MAYBE_GC() doesn't check the
1273 // context switch flag, and we end up waiting for a GC.
1274 // See #1984, and concurrent/should_run/1984
1275 cap->context_switch = 0;
1276 addToRunQueue(cap,t);
1278 pushOnRunQueue(cap,t);
1281 /* actual GC is done at the end of the while loop in schedule() */
1284 /* -----------------------------------------------------------------------------
1285 * Handle a thread that returned to the scheduler with ThreadStackOverflow
1286 * -------------------------------------------------------------------------- */
1289 scheduleHandleStackOverflow (Capability *cap, Task *task, StgTSO *t)
1291 debugTrace (DEBUG_sched,
1292 "--<< thread %ld (%s) stopped, StackOverflow",
1293 (long)t->id, whatNext_strs[t->what_next]);
1295 /* just adjust the stack for this thread, then pop it back
1299 /* enlarge the stack */
1300 StgTSO *new_t = threadStackOverflow(cap, t);
1302 /* The TSO attached to this Task may have moved, so update the
1305 if (task->tso == t) {
1308 pushOnRunQueue(cap,new_t);
1312 /* -----------------------------------------------------------------------------
1313 * Handle a thread that returned to the scheduler with ThreadYielding
1314 * -------------------------------------------------------------------------- */
1317 scheduleHandleYield( Capability *cap, StgTSO *t, nat prev_what_next )
1319 // Reset the context switch flag. We don't do this just before
1320 // running the thread, because that would mean we would lose ticks
1321 // during GC, which can lead to unfair scheduling (a thread hogs
1322 // the CPU because the tick always arrives during GC). This way
1323 // penalises threads that do a lot of allocation, but that seems
1324 // better than the alternative.
1325 cap->context_switch = 0;
1327 /* put the thread back on the run queue. Then, if we're ready to
1328 * GC, check whether this is the last task to stop. If so, wake
1329 * up the GC thread. getThread will block during a GC until the
1333 if (t->what_next != prev_what_next) {
1334 debugTrace(DEBUG_sched,
1335 "--<< thread %ld (%s) stopped to switch evaluators",
1336 (long)t->id, whatNext_strs[t->what_next]);
1338 debugTrace(DEBUG_sched,
1339 "--<< thread %ld (%s) stopped, yielding",
1340 (long)t->id, whatNext_strs[t->what_next]);
1345 //debugBelch("&& Doing sanity check on yielding TSO %ld.", t->id);
1347 ASSERT(t->_link == END_TSO_QUEUE);
1349 // Shortcut if we're just switching evaluators: don't bother
1350 // doing stack squeezing (which can be expensive), just run the
1352 if (t->what_next != prev_what_next) {
1356 addToRunQueue(cap,t);
1361 /* -----------------------------------------------------------------------------
1362 * Handle a thread that returned to the scheduler with ThreadBlocked
1363 * -------------------------------------------------------------------------- */
1366 scheduleHandleThreadBlocked( StgTSO *t
1367 #if !defined(GRAN) && !defined(DEBUG)
1373 // We don't need to do anything. The thread is blocked, and it
1374 // has tidied up its stack and placed itself on whatever queue
1375 // it needs to be on.
1377 // ASSERT(t->why_blocked != NotBlocked);
1378 // Not true: for example,
1379 // - in THREADED_RTS, the thread may already have been woken
1380 // up by another Capability. This actually happens: try
1381 // conc023 +RTS -N2.
1382 // - the thread may have woken itself up already, because
1383 // threadPaused() might have raised a blocked throwTo
1384 // exception, see maybePerformBlockedException().
1387 if (traceClass(DEBUG_sched)) {
1388 debugTraceBegin("--<< thread %lu (%s) stopped: ",
1389 (unsigned long)t->id, whatNext_strs[t->what_next]);
1390 printThreadBlockage(t);
1396 /* -----------------------------------------------------------------------------
1397 * Handle a thread that returned to the scheduler with ThreadFinished
1398 * -------------------------------------------------------------------------- */
1401 scheduleHandleThreadFinished (Capability *cap STG_UNUSED, Task *task, StgTSO *t)
1403 /* Need to check whether this was a main thread, and if so,
1404 * return with the return value.
1406 * We also end up here if the thread kills itself with an
1407 * uncaught exception, see Exception.cmm.
1409 debugTrace(DEBUG_sched, "--++ thread %lu (%s) finished",
1410 (unsigned long)t->id, whatNext_strs[t->what_next]);
1412 // blocked exceptions can now complete, even if the thread was in
1413 // blocked mode (see #2910). This unconditionally calls
1414 // lockTSO(), which ensures that we don't miss any threads that
1415 // are engaged in throwTo() with this thread as a target.
1416 awakenBlockedExceptionQueue (cap, t);
1419 // Check whether the thread that just completed was a bound
1420 // thread, and if so return with the result.
1422 // There is an assumption here that all thread completion goes
1423 // through this point; we need to make sure that if a thread
1424 // ends up in the ThreadKilled state, that it stays on the run
1425 // queue so it can be dealt with here.
1430 if (t->bound != task) {
1431 #if !defined(THREADED_RTS)
1432 // Must be a bound thread that is not the topmost one. Leave
1433 // it on the run queue until the stack has unwound to the
1434 // point where we can deal with this. Leaving it on the run
1435 // queue also ensures that the garbage collector knows about
1436 // this thread and its return value (it gets dropped from the
1437 // step->threads list so there's no other way to find it).
1438 appendToRunQueue(cap,t);
1441 // this cannot happen in the threaded RTS, because a
1442 // bound thread can only be run by the appropriate Task.
1443 barf("finished bound thread that isn't mine");
1447 ASSERT(task->tso == t);
1449 if (t->what_next == ThreadComplete) {
1451 // NOTE: return val is tso->sp[1] (see StgStartup.hc)
1452 *(task->ret) = (StgClosure *)task->tso->sp[1];
1454 task->stat = Success;
1457 *(task->ret) = NULL;
1459 if (sched_state >= SCHED_INTERRUPTING) {
1460 if (heap_overflow) {
1461 task->stat = HeapExhausted;
1463 task->stat = Interrupted;
1466 task->stat = Killed;
1470 removeThreadLabel((StgWord)task->tso->id);
1472 return rtsTrue; // tells schedule() to return
1478 /* -----------------------------------------------------------------------------
1479 * Perform a heap census
1480 * -------------------------------------------------------------------------- */
1483 scheduleNeedHeapProfile( rtsBool ready_to_gc STG_UNUSED )
1485 // When we have +RTS -i0 and we're heap profiling, do a census at
1486 // every GC. This lets us get repeatable runs for debugging.
1487 if (performHeapProfile ||
1488 (RtsFlags.ProfFlags.profileInterval==0 &&
1489 RtsFlags.ProfFlags.doHeapProfile && ready_to_gc)) {
1496 /* -----------------------------------------------------------------------------
1497 * Perform a garbage collection if necessary
1498 * -------------------------------------------------------------------------- */
1501 scheduleDoGC (Capability *cap, Task *task USED_IF_THREADS, rtsBool force_major)
1503 rtsBool heap_census;
1505 /* extern static volatile StgWord waiting_for_gc;
1506 lives inside capability.c */
1507 rtsBool gc_type, prev_pending_gc;
1511 if (sched_state == SCHED_SHUTTING_DOWN) {
1512 // The final GC has already been done, and the system is
1513 // shutting down. We'll probably deadlock if we try to GC
1519 if (sched_state < SCHED_INTERRUPTING
1520 && RtsFlags.ParFlags.parGcEnabled
1521 && N >= RtsFlags.ParFlags.parGcGen
1522 && ! oldest_gen->steps[0].mark)
1524 gc_type = PENDING_GC_PAR;
1526 gc_type = PENDING_GC_SEQ;
1529 // In order to GC, there must be no threads running Haskell code.
1530 // Therefore, the GC thread needs to hold *all* the capabilities,
1531 // and release them after the GC has completed.
1533 // This seems to be the simplest way: previous attempts involved
1534 // making all the threads with capabilities give up their
1535 // capabilities and sleep except for the *last* one, which
1536 // actually did the GC. But it's quite hard to arrange for all
1537 // the other tasks to sleep and stay asleep.
1540 /* Other capabilities are prevented from running yet more Haskell
1541 threads if waiting_for_gc is set. Tested inside
1542 yieldCapability() and releaseCapability() in Capability.c */
1544 prev_pending_gc = cas(&waiting_for_gc, 0, gc_type);
1545 if (prev_pending_gc) {
1547 debugTrace(DEBUG_sched, "someone else is trying to GC (%d)...",
1550 yieldCapability(&cap,task);
1551 } while (waiting_for_gc);
1552 return cap; // NOTE: task->cap might have changed here
1555 setContextSwitches();
1557 // The final shutdown GC is always single-threaded, because it's
1558 // possible that some of the Capabilities have no worker threads.
1560 if (gc_type == PENDING_GC_SEQ)
1562 // single-threaded GC: grab all the capabilities
1563 for (i=0; i < n_capabilities; i++) {
1564 debugTrace(DEBUG_sched, "ready_to_gc, grabbing all the capabilies (%d/%d)", i, n_capabilities);
1565 if (cap != &capabilities[i]) {
1566 Capability *pcap = &capabilities[i];
1567 // we better hope this task doesn't get migrated to
1568 // another Capability while we're waiting for this one.
1569 // It won't, because load balancing happens while we have
1570 // all the Capabilities, but even so it's a slightly
1571 // unsavoury invariant.
1573 waitForReturnCapability(&pcap, task);
1574 if (pcap != &capabilities[i]) {
1575 barf("scheduleDoGC: got the wrong capability");
1582 // multi-threaded GC: make sure all the Capabilities donate one
1584 debugTrace(DEBUG_sched, "ready_to_gc, grabbing GC threads");
1586 waitForGcThreads(cap);
1590 // so this happens periodically:
1591 if (cap) scheduleCheckBlackHoles(cap);
1593 IF_DEBUG(scheduler, printAllThreads());
1595 delete_threads_and_gc:
1597 * We now have all the capabilities; if we're in an interrupting
1598 * state, then we should take the opportunity to delete all the
1599 * threads in the system.
1601 if (sched_state == SCHED_INTERRUPTING) {
1602 deleteAllThreads(cap);
1603 sched_state = SCHED_SHUTTING_DOWN;
1606 heap_census = scheduleNeedHeapProfile(rtsTrue);
1608 if (recent_activity == ACTIVITY_INACTIVE && force_major)
1610 // We are doing a GC because the system has been idle for a
1611 // timeslice and we need to check for deadlock. Record the
1612 // fact that we've done a GC and turn off the timer signal;
1613 // it will get re-enabled if we run any threads after the GC.
1615 // Note: this is done before GC, because after GC there might
1616 // be threads already running (GarbageCollect() releases the
1617 // GC threads when it completes), so we risk turning off the
1618 // timer signal when it should really be on.
1619 recent_activity = ACTIVITY_DONE_GC;
1623 #if defined(THREADED_RTS)
1624 debugTrace(DEBUG_sched, "doing GC");
1625 // reset waiting_for_gc *before* GC, so that when the GC threads
1626 // emerge they don't immediately re-enter the GC.
1628 GarbageCollect(force_major || heap_census, gc_type, cap);
1630 GarbageCollect(force_major || heap_census, 0, cap);
1634 debugTrace(DEBUG_sched, "performing heap census");
1636 performHeapProfile = rtsFalse;
1639 if (heap_overflow && sched_state < SCHED_INTERRUPTING) {
1640 // GC set the heap_overflow flag, so we should proceed with
1641 // an orderly shutdown now. Ultimately we want the main
1642 // thread to return to its caller with HeapExhausted, at which
1643 // point the caller should call hs_exit(). The first step is
1644 // to delete all the threads.
1646 // Another way to do this would be to raise an exception in
1647 // the main thread, which we really should do because it gives
1648 // the program a chance to clean up. But how do we find the
1649 // main thread? It should presumably be the same one that
1650 // gets ^C exceptions, but that's all done on the Haskell side
1651 // (GHC.TopHandler).
1652 sched_state = SCHED_INTERRUPTING;
1653 goto delete_threads_and_gc;
1658 Once we are all together... this would be the place to balance all
1659 spark pools. No concurrent stealing or adding of new sparks can
1660 occur. Should be defined in Sparks.c. */
1661 balanceSparkPoolsCaps(n_capabilities, capabilities);
1664 #if defined(THREADED_RTS)
1665 if (gc_type == PENDING_GC_SEQ) {
1666 // release our stash of capabilities.
1667 for (i = 0; i < n_capabilities; i++) {
1668 if (cap != &capabilities[i]) {
1669 task->cap = &capabilities[i];
1670 releaseCapability(&capabilities[i]);
1684 /* ---------------------------------------------------------------------------
1685 * Singleton fork(). Do not copy any running threads.
1686 * ------------------------------------------------------------------------- */
1689 forkProcess(HsStablePtr *entry
1690 #ifndef FORKPROCESS_PRIMOP_SUPPORTED
1695 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
1702 #if defined(THREADED_RTS)
1703 if (RtsFlags.ParFlags.nNodes > 1) {
1704 errorBelch("forking not supported with +RTS -N<n> greater than 1");
1705 stg_exit(EXIT_FAILURE);
1709 debugTrace(DEBUG_sched, "forking!");
1711 // ToDo: for SMP, we should probably acquire *all* the capabilities
1714 // no funny business: hold locks while we fork, otherwise if some
1715 // other thread is holding a lock when the fork happens, the data
1716 // structure protected by the lock will forever be in an
1717 // inconsistent state in the child. See also #1391.
1718 ACQUIRE_LOCK(&sched_mutex);
1719 ACQUIRE_LOCK(&cap->lock);
1720 ACQUIRE_LOCK(&cap->running_task->lock);
1724 if (pid) { // parent
1726 RELEASE_LOCK(&sched_mutex);
1727 RELEASE_LOCK(&cap->lock);
1728 RELEASE_LOCK(&cap->running_task->lock);
1730 // just return the pid
1736 #if defined(THREADED_RTS)
1737 initMutex(&sched_mutex);
1738 initMutex(&cap->lock);
1739 initMutex(&cap->running_task->lock);
1742 // Now, all OS threads except the thread that forked are
1743 // stopped. We need to stop all Haskell threads, including
1744 // those involved in foreign calls. Also we need to delete
1745 // all Tasks, because they correspond to OS threads that are
1748 for (s = 0; s < total_steps; s++) {
1749 for (t = all_steps[s].threads; t != END_TSO_QUEUE; t = next) {
1750 if (t->what_next == ThreadRelocated) {
1753 next = t->global_link;
1754 // don't allow threads to catch the ThreadKilled
1755 // exception, but we do want to raiseAsync() because these
1756 // threads may be evaluating thunks that we need later.
1757 deleteThread_(cap,t);
1762 // Empty the run queue. It seems tempting to let all the
1763 // killed threads stay on the run queue as zombies to be
1764 // cleaned up later, but some of them correspond to bound
1765 // threads for which the corresponding Task does not exist.
1766 cap->run_queue_hd = END_TSO_QUEUE;
1767 cap->run_queue_tl = END_TSO_QUEUE;
1769 // Any suspended C-calling Tasks are no more, their OS threads
1771 cap->suspended_ccalling_tasks = NULL;
1773 // Empty the threads lists. Otherwise, the garbage
1774 // collector may attempt to resurrect some of these threads.
1775 for (s = 0; s < total_steps; s++) {
1776 all_steps[s].threads = END_TSO_QUEUE;
1779 // Wipe the task list, except the current Task.
1780 ACQUIRE_LOCK(&sched_mutex);
1781 for (task = all_tasks; task != NULL; task=task->all_link) {
1782 if (task != cap->running_task) {
1783 #if defined(THREADED_RTS)
1784 initMutex(&task->lock); // see #1391
1789 RELEASE_LOCK(&sched_mutex);
1791 #if defined(THREADED_RTS)
1792 // Wipe our spare workers list, they no longer exist. New
1793 // workers will be created if necessary.
1794 cap->spare_workers = NULL;
1795 cap->returning_tasks_hd = NULL;
1796 cap->returning_tasks_tl = NULL;
1799 // On Unix, all timers are reset in the child, so we need to start
1804 cap = rts_evalStableIO(cap, entry, NULL); // run the action
1805 rts_checkSchedStatus("forkProcess",cap);
1808 hs_exit(); // clean up and exit
1809 stg_exit(EXIT_SUCCESS);
1811 #else /* !FORKPROCESS_PRIMOP_SUPPORTED */
1812 barf("forkProcess#: primop not supported on this platform, sorry!\n");
1817 /* ---------------------------------------------------------------------------
1818 * Delete all the threads in the system
1819 * ------------------------------------------------------------------------- */
1822 deleteAllThreads ( Capability *cap )
1824 // NOTE: only safe to call if we own all capabilities.
1829 debugTrace(DEBUG_sched,"deleting all threads");
1830 for (s = 0; s < total_steps; s++) {
1831 for (t = all_steps[s].threads; t != END_TSO_QUEUE; t = next) {
1832 if (t->what_next == ThreadRelocated) {
1835 next = t->global_link;
1836 deleteThread(cap,t);
1841 // The run queue now contains a bunch of ThreadKilled threads. We
1842 // must not throw these away: the main thread(s) will be in there
1843 // somewhere, and the main scheduler loop has to deal with it.
1844 // Also, the run queue is the only thing keeping these threads from
1845 // being GC'd, and we don't want the "main thread has been GC'd" panic.
1847 #if !defined(THREADED_RTS)
1848 ASSERT(blocked_queue_hd == END_TSO_QUEUE);
1849 ASSERT(sleeping_queue == END_TSO_QUEUE);
1853 /* -----------------------------------------------------------------------------
1854 Managing the suspended_ccalling_tasks list.
1855 Locks required: sched_mutex
1856 -------------------------------------------------------------------------- */
1859 suspendTask (Capability *cap, Task *task)
1861 ASSERT(task->next == NULL && task->prev == NULL);
1862 task->next = cap->suspended_ccalling_tasks;
1864 if (cap->suspended_ccalling_tasks) {
1865 cap->suspended_ccalling_tasks->prev = task;
1867 cap->suspended_ccalling_tasks = task;
1871 recoverSuspendedTask (Capability *cap, Task *task)
1874 task->prev->next = task->next;
1876 ASSERT(cap->suspended_ccalling_tasks == task);
1877 cap->suspended_ccalling_tasks = task->next;
1880 task->next->prev = task->prev;
1882 task->next = task->prev = NULL;
1885 /* ---------------------------------------------------------------------------
1886 * Suspending & resuming Haskell threads.
1888 * When making a "safe" call to C (aka _ccall_GC), the task gives back
1889 * its capability before calling the C function. This allows another
1890 * task to pick up the capability and carry on running Haskell
1891 * threads. It also means that if the C call blocks, it won't lock
1894 * The Haskell thread making the C call is put to sleep for the
1895 * duration of the call, on the susepended_ccalling_threads queue. We
1896 * give out a token to the task, which it can use to resume the thread
1897 * on return from the C function.
1898 * ------------------------------------------------------------------------- */
1901 suspendThread (StgRegTable *reg)
1908 StgWord32 saved_winerror;
1911 saved_errno = errno;
1913 saved_winerror = GetLastError();
1916 /* assume that *reg is a pointer to the StgRegTable part of a Capability.
1918 cap = regTableToCapability(reg);
1920 task = cap->running_task;
1921 tso = cap->r.rCurrentTSO;
1923 debugTrace(DEBUG_sched,
1924 "thread %lu did a safe foreign call",
1925 (unsigned long)cap->r.rCurrentTSO->id);
1927 // XXX this might not be necessary --SDM
1928 tso->what_next = ThreadRunGHC;
1930 threadPaused(cap,tso);
1932 if ((tso->flags & TSO_BLOCKEX) == 0) {
1933 tso->why_blocked = BlockedOnCCall;
1934 tso->flags |= TSO_BLOCKEX;
1935 tso->flags &= ~TSO_INTERRUPTIBLE;
1937 tso->why_blocked = BlockedOnCCall_NoUnblockExc;
1940 // Hand back capability
1941 task->suspended_tso = tso;
1943 ACQUIRE_LOCK(&cap->lock);
1945 suspendTask(cap,task);
1946 cap->in_haskell = rtsFalse;
1947 releaseCapability_(cap,rtsFalse);
1949 RELEASE_LOCK(&cap->lock);
1951 #if defined(THREADED_RTS)
1952 /* Preparing to leave the RTS, so ensure there's a native thread/task
1953 waiting to take over.
1955 debugTrace(DEBUG_sched, "thread %lu: leaving RTS", (unsigned long)tso->id);
1958 errno = saved_errno;
1960 SetLastError(saved_winerror);
1966 resumeThread (void *task_)
1973 StgWord32 saved_winerror;
1976 saved_errno = errno;
1978 saved_winerror = GetLastError();
1982 // Wait for permission to re-enter the RTS with the result.
1983 waitForReturnCapability(&cap,task);
1984 // we might be on a different capability now... but if so, our
1985 // entry on the suspended_ccalling_tasks list will also have been
1988 // Remove the thread from the suspended list
1989 recoverSuspendedTask(cap,task);
1991 tso = task->suspended_tso;
1992 task->suspended_tso = NULL;
1993 tso->_link = END_TSO_QUEUE; // no write barrier reqd
1994 debugTrace(DEBUG_sched, "thread %lu: re-entering RTS", (unsigned long)tso->id);
1996 if (tso->why_blocked == BlockedOnCCall) {
1997 // avoid locking the TSO if we don't have to
1998 if (tso->blocked_exceptions != END_TSO_QUEUE) {
1999 awakenBlockedExceptionQueue(cap,tso);
2001 tso->flags &= ~(TSO_BLOCKEX | TSO_INTERRUPTIBLE);
2004 /* Reset blocking status */
2005 tso->why_blocked = NotBlocked;
2007 cap->r.rCurrentTSO = tso;
2008 cap->in_haskell = rtsTrue;
2009 errno = saved_errno;
2011 SetLastError(saved_winerror);
2014 /* We might have GC'd, mark the TSO dirty again */
2017 IF_DEBUG(sanity, checkTSO(tso));
2022 /* ---------------------------------------------------------------------------
2025 * scheduleThread puts a thread on the end of the runnable queue.
2026 * This will usually be done immediately after a thread is created.
2027 * The caller of scheduleThread must create the thread using e.g.
2028 * createThread and push an appropriate closure
2029 * on this thread's stack before the scheduler is invoked.
2030 * ------------------------------------------------------------------------ */
2033 scheduleThread(Capability *cap, StgTSO *tso)
2035 // The thread goes at the *end* of the run-queue, to avoid possible
2036 // starvation of any threads already on the queue.
2037 appendToRunQueue(cap,tso);
2041 scheduleThreadOn(Capability *cap, StgWord cpu USED_IF_THREADS, StgTSO *tso)
2043 #if defined(THREADED_RTS)
2044 tso->flags |= TSO_LOCKED; // we requested explicit affinity; don't
2045 // move this thread from now on.
2046 cpu %= RtsFlags.ParFlags.nNodes;
2047 if (cpu == cap->no) {
2048 appendToRunQueue(cap,tso);
2050 wakeupThreadOnCapability(cap, &capabilities[cpu], tso);
2053 appendToRunQueue(cap,tso);
2058 scheduleWaitThread (StgTSO* tso, /*[out]*/HaskellObj* ret, Capability *cap)
2062 // We already created/initialised the Task
2063 task = cap->running_task;
2065 // This TSO is now a bound thread; make the Task and TSO
2066 // point to each other.
2072 task->stat = NoStatus;
2074 appendToRunQueue(cap,tso);
2076 debugTrace(DEBUG_sched, "new bound thread (%lu)", (unsigned long)tso->id);
2078 cap = schedule(cap,task);
2080 ASSERT(task->stat != NoStatus);
2081 ASSERT_FULL_CAPABILITY_INVARIANTS(cap,task);
2083 debugTrace(DEBUG_sched, "bound thread (%lu) finished", (unsigned long)task->tso->id);
2087 /* ----------------------------------------------------------------------------
2089 * ------------------------------------------------------------------------- */
2091 #if defined(THREADED_RTS)
2092 void OSThreadProcAttr
2093 workerStart(Task *task)
2097 // See startWorkerTask().
2098 ACQUIRE_LOCK(&task->lock);
2100 RELEASE_LOCK(&task->lock);
2102 // set the thread-local pointer to the Task:
2105 // schedule() runs without a lock.
2106 cap = schedule(cap,task);
2108 // On exit from schedule(), we have a Capability, but possibly not
2109 // the same one we started with.
2111 // During shutdown, the requirement is that after all the
2112 // Capabilities are shut down, all workers that are shutting down
2113 // have finished workerTaskStop(). This is why we hold on to
2114 // cap->lock until we've finished workerTaskStop() below.
2116 // There may be workers still involved in foreign calls; those
2117 // will just block in waitForReturnCapability() because the
2118 // Capability has been shut down.
2120 ACQUIRE_LOCK(&cap->lock);
2121 releaseCapability_(cap,rtsFalse);
2122 workerTaskStop(task);
2123 RELEASE_LOCK(&cap->lock);
2127 /* ---------------------------------------------------------------------------
2130 * Initialise the scheduler. This resets all the queues - if the
2131 * queues contained any threads, they'll be garbage collected at the
2134 * ------------------------------------------------------------------------ */
2139 #if !defined(THREADED_RTS)
2140 blocked_queue_hd = END_TSO_QUEUE;
2141 blocked_queue_tl = END_TSO_QUEUE;
2142 sleeping_queue = END_TSO_QUEUE;
2145 blackhole_queue = END_TSO_QUEUE;
2147 sched_state = SCHED_RUNNING;
2148 recent_activity = ACTIVITY_YES;
2150 #if defined(THREADED_RTS)
2151 /* Initialise the mutex and condition variables used by
2153 initMutex(&sched_mutex);
2156 ACQUIRE_LOCK(&sched_mutex);
2158 /* A capability holds the state a native thread needs in
2159 * order to execute STG code. At least one capability is
2160 * floating around (only THREADED_RTS builds have more than one).
2166 #if defined(THREADED_RTS) || defined(PARALLEL_HASKELL)
2170 #if defined(THREADED_RTS)
2172 * Eagerly start one worker to run each Capability, except for
2173 * Capability 0. The idea is that we're probably going to start a
2174 * bound thread on Capability 0 pretty soon, so we don't want a
2175 * worker task hogging it.
2180 for (i = 1; i < n_capabilities; i++) {
2181 cap = &capabilities[i];
2182 ACQUIRE_LOCK(&cap->lock);
2183 startWorkerTask(cap, workerStart);
2184 RELEASE_LOCK(&cap->lock);
2189 trace(TRACE_sched, "start: %d capabilities", n_capabilities);
2191 RELEASE_LOCK(&sched_mutex);
2196 rtsBool wait_foreign
2197 #if !defined(THREADED_RTS)
2198 __attribute__((unused))
2201 /* see Capability.c, shutdownCapability() */
2205 #if defined(THREADED_RTS)
2206 ACQUIRE_LOCK(&sched_mutex);
2207 task = newBoundTask();
2208 RELEASE_LOCK(&sched_mutex);
2211 // If we haven't killed all the threads yet, do it now.
2212 if (sched_state < SCHED_SHUTTING_DOWN) {
2213 sched_state = SCHED_INTERRUPTING;
2214 #if defined(THREADED_RTS)
2215 waitForReturnCapability(&task->cap,task);
2216 scheduleDoGC(task->cap,task,rtsFalse);
2217 releaseCapability(task->cap);
2219 scheduleDoGC(&MainCapability,task,rtsFalse);
2222 sched_state = SCHED_SHUTTING_DOWN;
2224 #if defined(THREADED_RTS)
2228 for (i = 0; i < n_capabilities; i++) {
2229 shutdownCapability(&capabilities[i], task, wait_foreign);
2231 boundTaskExiting(task);
2237 freeScheduler( void )
2241 ACQUIRE_LOCK(&sched_mutex);
2242 still_running = freeTaskManager();
2243 // We can only free the Capabilities if there are no Tasks still
2244 // running. We might have a Task about to return from a foreign
2245 // call into waitForReturnCapability(), for example (actually,
2246 // this should be the *only* thing that a still-running Task can
2247 // do at this point, and it will block waiting for the
2249 if (still_running == 0) {
2251 if (n_capabilities != 1) {
2252 stgFree(capabilities);
2255 RELEASE_LOCK(&sched_mutex);
2256 #if defined(THREADED_RTS)
2257 closeMutex(&sched_mutex);
2261 /* -----------------------------------------------------------------------------
2264 This is the interface to the garbage collector from Haskell land.
2265 We provide this so that external C code can allocate and garbage
2266 collect when called from Haskell via _ccall_GC.
2267 -------------------------------------------------------------------------- */
2270 performGC_(rtsBool force_major)
2274 // We must grab a new Task here, because the existing Task may be
2275 // associated with a particular Capability, and chained onto the
2276 // suspended_ccalling_tasks queue.
2277 ACQUIRE_LOCK(&sched_mutex);
2278 task = newBoundTask();
2279 RELEASE_LOCK(&sched_mutex);
2281 waitForReturnCapability(&task->cap,task);
2282 scheduleDoGC(task->cap,task,force_major);
2283 releaseCapability(task->cap);
2284 boundTaskExiting(task);
2290 performGC_(rtsFalse);
2294 performMajorGC(void)
2296 performGC_(rtsTrue);
2299 /* -----------------------------------------------------------------------------
2302 If the thread has reached its maximum stack size, then raise the
2303 StackOverflow exception in the offending thread. Otherwise
2304 relocate the TSO into a larger chunk of memory and adjust its stack
2306 -------------------------------------------------------------------------- */
2309 threadStackOverflow(Capability *cap, StgTSO *tso)
2311 nat new_stack_size, stack_words;
2316 IF_DEBUG(sanity,checkTSO(tso));
2318 // don't allow throwTo() to modify the blocked_exceptions queue
2319 // while we are moving the TSO:
2320 lockClosure((StgClosure *)tso);
2322 if (tso->stack_size >= tso->max_stack_size && !(tso->flags & TSO_BLOCKEX)) {
2323 // NB. never raise a StackOverflow exception if the thread is
2324 // inside Control.Exceptino.block. It is impractical to protect
2325 // against stack overflow exceptions, since virtually anything
2326 // can raise one (even 'catch'), so this is the only sensible
2327 // thing to do here. See bug #767.
2329 debugTrace(DEBUG_gc,
2330 "threadStackOverflow of TSO %ld (%p): stack too large (now %ld; max is %ld)",
2331 (long)tso->id, tso, (long)tso->stack_size, (long)tso->max_stack_size);
2333 /* If we're debugging, just print out the top of the stack */
2334 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2337 // Send this thread the StackOverflow exception
2339 throwToSingleThreaded(cap, tso, (StgClosure *)stackOverflow_closure);
2343 /* Try to double the current stack size. If that takes us over the
2344 * maximum stack size for this thread, then use the maximum instead
2345 * (that is, unless we're already at or over the max size and we
2346 * can't raise the StackOverflow exception (see above), in which
2347 * case just double the size). Finally round up so the TSO ends up as
2348 * a whole number of blocks.
2350 if (tso->stack_size >= tso->max_stack_size) {
2351 new_stack_size = tso->stack_size * 2;
2353 new_stack_size = stg_min(tso->stack_size * 2, tso->max_stack_size);
2355 new_tso_size = (lnat)BLOCK_ROUND_UP(new_stack_size * sizeof(W_) +
2356 TSO_STRUCT_SIZE)/sizeof(W_);
2357 new_tso_size = round_to_mblocks(new_tso_size); /* Be MBLOCK-friendly */
2358 new_stack_size = new_tso_size - TSO_STRUCT_SIZEW;
2360 debugTrace(DEBUG_sched,
2361 "increasing stack size from %ld words to %d.",
2362 (long)tso->stack_size, new_stack_size);
2364 dest = (StgTSO *)allocateLocal(cap,new_tso_size);
2365 TICK_ALLOC_TSO(new_stack_size,0);
2367 /* copy the TSO block and the old stack into the new area */
2368 memcpy(dest,tso,TSO_STRUCT_SIZE);
2369 stack_words = tso->stack + tso->stack_size - tso->sp;
2370 new_sp = (P_)dest + new_tso_size - stack_words;
2371 memcpy(new_sp, tso->sp, stack_words * sizeof(W_));
2373 /* relocate the stack pointers... */
2375 dest->stack_size = new_stack_size;
2377 /* Mark the old TSO as relocated. We have to check for relocated
2378 * TSOs in the garbage collector and any primops that deal with TSOs.
2380 * It's important to set the sp value to just beyond the end
2381 * of the stack, so we don't attempt to scavenge any part of the
2384 tso->what_next = ThreadRelocated;
2385 setTSOLink(cap,tso,dest);
2386 tso->sp = (P_)&(tso->stack[tso->stack_size]);
2387 tso->why_blocked = NotBlocked;
2389 IF_PAR_DEBUG(verbose,
2390 debugBelch("@@ threadStackOverflow of TSO %d (now at %p): stack size increased to %ld\n",
2391 tso->id, tso, tso->stack_size);
2392 /* If we're debugging, just print out the top of the stack */
2393 printStackChunk(tso->sp, stg_min(tso->stack+tso->stack_size,
2399 IF_DEBUG(sanity,checkTSO(dest));
2401 IF_DEBUG(scheduler,printTSO(dest));
2408 threadStackUnderflow (Task *task STG_UNUSED, StgTSO *tso)
2410 bdescr *bd, *new_bd;
2411 lnat free_w, tso_size_w;
2414 tso_size_w = tso_sizeW(tso);
2416 if (tso_size_w < MBLOCK_SIZE_W ||
2417 (nat)(tso->stack + tso->stack_size - tso->sp) > tso->stack_size / 4)
2422 // don't allow throwTo() to modify the blocked_exceptions queue
2423 // while we are moving the TSO:
2424 lockClosure((StgClosure *)tso);
2426 // this is the number of words we'll free
2427 free_w = round_to_mblocks(tso_size_w/2);
2429 bd = Bdescr((StgPtr)tso);
2430 new_bd = splitLargeBlock(bd, free_w / BLOCK_SIZE_W);
2431 bd->free = bd->start + TSO_STRUCT_SIZEW;
2433 new_tso = (StgTSO *)new_bd->start;
2434 memcpy(new_tso,tso,TSO_STRUCT_SIZE);
2435 new_tso->stack_size = new_bd->free - new_tso->stack;
2437 debugTrace(DEBUG_sched, "thread %ld: reducing TSO size from %lu words to %lu",
2438 (long)tso->id, tso_size_w, tso_sizeW(new_tso));
2440 tso->what_next = ThreadRelocated;
2441 tso->_link = new_tso; // no write barrier reqd: same generation
2443 // The TSO attached to this Task may have moved, so update the
2445 if (task->tso == tso) {
2446 task->tso = new_tso;
2452 IF_DEBUG(sanity,checkTSO(new_tso));
2457 /* ---------------------------------------------------------------------------
2459 - usually called inside a signal handler so it mustn't do anything fancy.
2460 ------------------------------------------------------------------------ */
2463 interruptStgRts(void)
2465 sched_state = SCHED_INTERRUPTING;
2466 setContextSwitches();
2470 /* -----------------------------------------------------------------------------
2473 This function causes at least one OS thread to wake up and run the
2474 scheduler loop. It is invoked when the RTS might be deadlocked, or
2475 an external event has arrived that may need servicing (eg. a
2476 keyboard interrupt).
2478 In the single-threaded RTS we don't do anything here; we only have
2479 one thread anyway, and the event that caused us to want to wake up
2480 will have interrupted any blocking system call in progress anyway.
2481 -------------------------------------------------------------------------- */
2486 #if defined(THREADED_RTS)
2487 // This forces the IO Manager thread to wakeup, which will
2488 // in turn ensure that some OS thread wakes up and runs the
2489 // scheduler loop, which will cause a GC and deadlock check.
2494 /* -----------------------------------------------------------------------------
2497 * Check the blackhole_queue for threads that can be woken up. We do
2498 * this periodically: before every GC, and whenever the run queue is
2501 * An elegant solution might be to just wake up all the blocked
2502 * threads with awakenBlockedQueue occasionally: they'll go back to
2503 * sleep again if the object is still a BLACKHOLE. Unfortunately this
2504 * doesn't give us a way to tell whether we've actually managed to
2505 * wake up any threads, so we would be busy-waiting.
2507 * -------------------------------------------------------------------------- */
2510 checkBlackHoles (Capability *cap)
2513 rtsBool any_woke_up = rtsFalse;
2516 // blackhole_queue is global:
2517 ASSERT_LOCK_HELD(&sched_mutex);
2519 debugTrace(DEBUG_sched, "checking threads blocked on black holes");
2521 // ASSUMES: sched_mutex
2522 prev = &blackhole_queue;
2523 t = blackhole_queue;
2524 while (t != END_TSO_QUEUE) {
2525 ASSERT(t->why_blocked == BlockedOnBlackHole);
2526 type = get_itbl(UNTAG_CLOSURE(t->block_info.closure))->type;
2527 if (type != BLACKHOLE && type != CAF_BLACKHOLE) {
2528 IF_DEBUG(sanity,checkTSO(t));
2529 t = unblockOne(cap, t);
2531 any_woke_up = rtsTrue;
2541 /* -----------------------------------------------------------------------------
2544 This is used for interruption (^C) and forking, and corresponds to
2545 raising an exception but without letting the thread catch the
2547 -------------------------------------------------------------------------- */
2550 deleteThread (Capability *cap, StgTSO *tso)
2552 // NOTE: must only be called on a TSO that we have exclusive
2553 // access to, because we will call throwToSingleThreaded() below.
2554 // The TSO must be on the run queue of the Capability we own, or
2555 // we must own all Capabilities.
2557 if (tso->why_blocked != BlockedOnCCall &&
2558 tso->why_blocked != BlockedOnCCall_NoUnblockExc) {
2559 throwToSingleThreaded(cap,tso,NULL);
2563 #ifdef FORKPROCESS_PRIMOP_SUPPORTED
2565 deleteThread_(Capability *cap, StgTSO *tso)
2566 { // for forkProcess only:
2567 // like deleteThread(), but we delete threads in foreign calls, too.
2569 if (tso->why_blocked == BlockedOnCCall ||
2570 tso->why_blocked == BlockedOnCCall_NoUnblockExc) {
2571 unblockOne(cap,tso);
2572 tso->what_next = ThreadKilled;
2574 deleteThread(cap,tso);
2579 /* -----------------------------------------------------------------------------
2580 raiseExceptionHelper
2582 This function is called by the raise# primitve, just so that we can
2583 move some of the tricky bits of raising an exception from C-- into
2584 C. Who knows, it might be a useful re-useable thing here too.
2585 -------------------------------------------------------------------------- */
2588 raiseExceptionHelper (StgRegTable *reg, StgTSO *tso, StgClosure *exception)
2590 Capability *cap = regTableToCapability(reg);
2591 StgThunk *raise_closure = NULL;
2593 StgRetInfoTable *info;
2595 // This closure represents the expression 'raise# E' where E
2596 // is the exception raise. It is used to overwrite all the
2597 // thunks which are currently under evaluataion.
2600 // OLD COMMENT (we don't have MIN_UPD_SIZE now):
2601 // LDV profiling: stg_raise_info has THUNK as its closure
2602 // type. Since a THUNK takes at least MIN_UPD_SIZE words in its
2603 // payload, MIN_UPD_SIZE is more approprate than 1. It seems that
2604 // 1 does not cause any problem unless profiling is performed.
2605 // However, when LDV profiling goes on, we need to linearly scan
2606 // small object pool, where raise_closure is stored, so we should
2607 // use MIN_UPD_SIZE.
2609 // raise_closure = (StgClosure *)RET_STGCALL1(P_,allocate,
2610 // sizeofW(StgClosure)+1);
2614 // Walk up the stack, looking for the catch frame. On the way,
2615 // we update any closures pointed to from update frames with the
2616 // raise closure that we just built.
2620 info = get_ret_itbl((StgClosure *)p);
2621 next = p + stack_frame_sizeW((StgClosure *)p);
2622 switch (info->i.type) {
2625 // Only create raise_closure if we need to.
2626 if (raise_closure == NULL) {
2628 (StgThunk *)allocateLocal(cap,sizeofW(StgThunk)+1);
2629 SET_HDR(raise_closure, &stg_raise_info, CCCS);
2630 raise_closure->payload[0] = exception;
2632 UPD_IND(((StgUpdateFrame *)p)->updatee,(StgClosure *)raise_closure);
2636 case ATOMICALLY_FRAME:
2637 debugTrace(DEBUG_stm, "found ATOMICALLY_FRAME at %p", p);
2639 return ATOMICALLY_FRAME;
2645 case CATCH_STM_FRAME:
2646 debugTrace(DEBUG_stm, "found CATCH_STM_FRAME at %p", p);
2648 return CATCH_STM_FRAME;
2654 case CATCH_RETRY_FRAME:
2663 /* -----------------------------------------------------------------------------
2664 findRetryFrameHelper
2666 This function is called by the retry# primitive. It traverses the stack
2667 leaving tso->sp referring to the frame which should handle the retry.
2669 This should either be a CATCH_RETRY_FRAME (if the retry# is within an orElse#)
2670 or should be a ATOMICALLY_FRAME (if the retry# reaches the top level).
2672 We skip CATCH_STM_FRAMEs (aborting and rolling back the nested tx that they
2673 create) because retries are not considered to be exceptions, despite the
2674 similar implementation.
2676 We should not expect to see CATCH_FRAME or STOP_FRAME because those should
2677 not be created within memory transactions.
2678 -------------------------------------------------------------------------- */
2681 findRetryFrameHelper (StgTSO *tso)
2684 StgRetInfoTable *info;
2688 info = get_ret_itbl((StgClosure *)p);
2689 next = p + stack_frame_sizeW((StgClosure *)p);
2690 switch (info->i.type) {
2692 case ATOMICALLY_FRAME:
2693 debugTrace(DEBUG_stm,
2694 "found ATOMICALLY_FRAME at %p during retry", p);
2696 return ATOMICALLY_FRAME;
2698 case CATCH_RETRY_FRAME:
2699 debugTrace(DEBUG_stm,
2700 "found CATCH_RETRY_FRAME at %p during retrry", p);
2702 return CATCH_RETRY_FRAME;
2704 case CATCH_STM_FRAME: {
2705 StgTRecHeader *trec = tso -> trec;
2706 StgTRecHeader *outer = stmGetEnclosingTRec(trec);
2707 debugTrace(DEBUG_stm,
2708 "found CATCH_STM_FRAME at %p during retry", p);
2709 debugTrace(DEBUG_stm, "trec=%p outer=%p", trec, outer);
2710 stmAbortTransaction(tso -> cap, trec);
2711 stmFreeAbortedTRec(tso -> cap, trec);
2712 tso -> trec = outer;
2719 ASSERT(info->i.type != CATCH_FRAME);
2720 ASSERT(info->i.type != STOP_FRAME);
2727 /* -----------------------------------------------------------------------------
2728 resurrectThreads is called after garbage collection on the list of
2729 threads found to be garbage. Each of these threads will be woken
2730 up and sent a signal: BlockedOnDeadMVar if the thread was blocked
2731 on an MVar, or NonTermination if the thread was blocked on a Black
2734 Locks: assumes we hold *all* the capabilities.
2735 -------------------------------------------------------------------------- */
2738 resurrectThreads (StgTSO *threads)
2744 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
2745 next = tso->global_link;
2747 step = Bdescr((P_)tso)->step;
2748 tso->global_link = step->threads;
2749 step->threads = tso;
2751 debugTrace(DEBUG_sched, "resurrecting thread %lu", (unsigned long)tso->id);
2753 // Wake up the thread on the Capability it was last on
2756 switch (tso->why_blocked) {
2758 case BlockedOnException:
2759 /* Called by GC - sched_mutex lock is currently held. */
2760 throwToSingleThreaded(cap, tso,
2761 (StgClosure *)blockedOnDeadMVar_closure);
2763 case BlockedOnBlackHole:
2764 throwToSingleThreaded(cap, tso,
2765 (StgClosure *)nonTermination_closure);
2768 throwToSingleThreaded(cap, tso,
2769 (StgClosure *)blockedIndefinitely_closure);
2772 /* This might happen if the thread was blocked on a black hole
2773 * belonging to a thread that we've just woken up (raiseAsync
2774 * can wake up threads, remember...).
2778 barf("resurrectThreads: thread blocked in a strange way");
2783 /* -----------------------------------------------------------------------------
2784 performPendingThrowTos is called after garbage collection, and
2785 passed a list of threads that were found to have pending throwTos
2786 (tso->blocked_exceptions was not empty), and were blocked.
2787 Normally this doesn't happen, because we would deliver the
2788 exception directly if the target thread is blocked, but there are
2789 small windows where it might occur on a multiprocessor (see
2792 NB. we must be holding all the capabilities at this point, just
2793 like resurrectThreads().
2794 -------------------------------------------------------------------------- */
2797 performPendingThrowTos (StgTSO *threads)
2803 for (tso = threads; tso != END_TSO_QUEUE; tso = next) {
2804 next = tso->global_link;
2806 step = Bdescr((P_)tso)->step;
2807 tso->global_link = step->threads;
2808 step->threads = tso;
2810 debugTrace(DEBUG_sched, "performing blocked throwTo to thread %lu", (unsigned long)tso->id);
2813 maybePerformBlockedException(cap, tso);