1 /* ---------------------------------------------------------------------------
3 * (c) The GHC Team, 1998-2006
5 * Asynchronous exceptions
7 * --------------------------------------------------------------------------*/
9 #include "PosixSource.h"
13 #include "RaiseAsync.h"
16 #include "LdvProfile.h"
20 #if defined(mingw32_HOST_OS)
21 #include "win32/IOManager.h"
24 static void raiseAsync (Capability *cap,
26 StgClosure *exception,
27 rtsBool stop_at_atomically,
30 static void removeFromQueues(Capability *cap, StgTSO *tso);
32 static void blockedThrowTo (StgTSO *source, StgTSO *target);
34 static void performBlockedException (Capability *cap,
35 StgTSO *source, StgTSO *target);
37 /* -----------------------------------------------------------------------------
40 This version of throwTo is safe to use if and only if one of the
45 - all the other threads in the system are stopped (eg. during GC).
47 - we surely own the target TSO (eg. we just took it from the
48 run queue of the current capability, or we are running it).
50 It doesn't cater for blocking the source thread until the exception
52 -------------------------------------------------------------------------- */
55 throwToSingleThreaded(Capability *cap, StgTSO *tso, StgClosure *exception)
57 throwToSingleThreaded_(cap, tso, exception, rtsFalse, NULL);
61 throwToSingleThreaded_(Capability *cap, StgTSO *tso, StgClosure *exception,
62 rtsBool stop_at_atomically, StgPtr stop_here)
64 // Thread already dead?
65 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
69 // Remove it from any blocking queues
70 removeFromQueues(cap,tso);
72 raiseAsync(cap, tso, exception, stop_at_atomically, stop_here);
76 suspendComputation(Capability *cap, StgTSO *tso, StgPtr stop_here)
78 // Thread already dead?
79 if (tso->what_next == ThreadComplete || tso->what_next == ThreadKilled) {
83 // Remove it from any blocking queues
84 removeFromQueues(cap,tso);
86 raiseAsync(cap, tso, NULL, rtsFalse, stop_here);
89 /* -----------------------------------------------------------------------------
92 This function may be used to throw an exception from one thread to
93 another, during the course of normal execution. This is a tricky
94 task: the target thread might be running on another CPU, or it
95 may be blocked and could be woken up at any point by another CPU.
96 We have some delicate synchronisation to do.
98 There is a completely safe fallback scheme: it is always possible
99 to just block the source TSO on the target TSO's blocked_exceptions
100 queue. This queue is locked using lockTSO()/unlockTSO(). It is
101 checked at regular intervals: before and after running a thread
102 (schedule() and threadPaused() respectively), and just before GC
103 (scheduleDoGC()). Activating a thread on this queue should be done
104 using maybePerformBlockedException(): this is done in the context
105 of the target thread, so the exception can be raised eagerly.
107 This fallback scheme works even if the target thread is complete or
108 killed: scheduleDoGC() will discover the blocked thread before the
111 Blocking the source thread on the target thread's blocked_exception
112 queue is also employed when the target thread is currently blocking
113 exceptions (ie. inside Control.Exception.block).
115 We could use the safe fallback scheme exclusively, but that
116 wouldn't be ideal: most calls to throwTo would block immediately,
117 possibly until the next GC, which might require the deadlock
118 detection mechanism to kick in. So we try to provide promptness
121 We can promptly deliver the exception if the target thread is:
123 - runnable, on the same Capability as the source thread (because
124 we own the run queue and therefore the target thread).
126 - blocked, and we can obtain exclusive access to it. Obtaining
127 exclusive access to the thread depends on how it is blocked.
129 We must also be careful to not trip over threadStackOverflow(),
130 which might be moving the TSO to enlarge its stack.
131 lockTSO()/unlockTSO() are used here too.
135 THROWTO_SUCCESS exception was raised, ok to continue
137 THROWTO_BLOCKED exception was not raised; block the source
138 thread then call throwToReleaseTarget() when
139 the source thread is properly tidied away.
141 -------------------------------------------------------------------------- */
144 throwTo (Capability *cap, // the Capability we hold
145 StgTSO *source, // the TSO sending the exception
146 StgTSO *target, // the TSO receiving the exception
147 StgClosure *exception, // the exception closure
148 /*[out]*/ void **out USED_IF_THREADS)
152 // follow ThreadRelocated links in the target first
153 while (target->what_next == ThreadRelocated) {
154 target = target->link;
155 // No, it might be a WHITEHOLE:
156 // ASSERT(get_itbl(target)->type == TSO);
159 debugTrace(DEBUG_sched, "throwTo: from thread %lu to thread %lu",
160 (unsigned long)source->id, (unsigned long)target->id);
163 if (traceClass(DEBUG_sched)) {
164 debugTraceBegin("throwTo: target");
165 printThreadStatus(target);
172 debugTrace(DEBUG_sched, "throwTo: retrying...");
175 // Thread already dead?
176 if (target->what_next == ThreadComplete
177 || target->what_next == ThreadKilled) {
178 return THROWTO_SUCCESS;
181 status = target->why_blocked;
185 /* if status==NotBlocked, and target->cap == cap, then
186 we own this TSO and can raise the exception.
188 How do we establish this condition? Very carefully.
191 P = (status == NotBlocked)
192 Q = (tso->cap == cap)
194 Now, if P & Q are true, then the TSO is locked and owned by
195 this capability. No other OS thread can steal it.
197 If P==0 and Q==1: the TSO is blocked, but attached to this
198 capabilty, and it can be stolen by another capability.
200 If P==1 and Q==0: the TSO is runnable on another
201 capability. At any time, the TSO may change from runnable
202 to blocked and vice versa, while it remains owned by
205 Suppose we test like this:
211 this is defeated by another capability stealing a blocked
212 TSO from us to wake it up (Schedule.c:unblockOne()). The
213 other thread is doing
218 assuming arbitrary reordering, we could see this
228 so we need a memory barrier:
235 this avoids the problematic case. There are other cases
236 to consider, but this is the tricky one.
238 Note that we must be sure that unblockOne() does the
239 writes in the correct order: Q before P. The memory
240 barrier ensures that if we have seen the write to P, we
241 have also seen the write to Q.
244 Capability *target_cap;
247 target_cap = target->cap;
248 if (target_cap == cap && (target->flags & TSO_BLOCKEX) == 0) {
249 // It's on our run queue and not blocking exceptions
250 raiseAsync(cap, target, exception, rtsFalse, NULL);
251 return THROWTO_SUCCESS;
253 // Otherwise, just block on the blocked_exceptions queue
254 // of the target thread. The queue will get looked at
255 // soon enough: it is checked before and after running a
256 // thread, and during GC.
259 // Avoid race with threadStackOverflow, which may have
260 // just moved this TSO.
261 if (target->what_next == ThreadRelocated) {
263 target = target->link;
266 blockedThrowTo(source,target);
268 return THROWTO_BLOCKED;
275 To establish ownership of this TSO, we need to acquire a
276 lock on the MVar that it is blocked on.
279 StgInfoTable *info USED_IF_THREADS;
281 mvar = (StgMVar *)target->block_info.closure;
283 // ASSUMPTION: tso->block_info must always point to a
284 // closure. In the threaded RTS it does.
285 switch (get_itbl(mvar)->type) {
293 info = lockClosure((StgClosure *)mvar);
295 if (target->what_next == ThreadRelocated) {
296 target = target->link;
297 unlockClosure((StgClosure *)mvar,info);
300 // we have the MVar, let's check whether the thread
301 // is still blocked on the same MVar.
302 if (target->why_blocked != BlockedOnMVar
303 || (StgMVar *)target->block_info.closure != mvar) {
304 unlockClosure((StgClosure *)mvar, info);
308 if ((target->flags & TSO_BLOCKEX) &&
309 ((target->flags & TSO_INTERRUPTIBLE) == 0)) {
310 lockClosure((StgClosure *)target);
311 blockedThrowTo(source,target);
312 unlockClosure((StgClosure *)mvar, info);
314 return THROWTO_BLOCKED; // caller releases TSO
316 removeThreadFromMVarQueue(mvar, target);
317 raiseAsync(cap, target, exception, rtsFalse, NULL);
318 unblockOne(cap, target);
319 unlockClosure((StgClosure *)mvar, info);
320 return THROWTO_SUCCESS;
324 case BlockedOnBlackHole:
326 ACQUIRE_LOCK(&sched_mutex);
327 // double checking the status after the memory barrier:
328 if (target->why_blocked != BlockedOnBlackHole) {
329 RELEASE_LOCK(&sched_mutex);
333 if (target->flags & TSO_BLOCKEX) {
335 blockedThrowTo(source,target);
336 RELEASE_LOCK(&sched_mutex);
338 return THROWTO_BLOCKED; // caller releases TSO
340 removeThreadFromQueue(&blackhole_queue, target);
341 raiseAsync(cap, target, exception, rtsFalse, NULL);
342 unblockOne(cap, target);
343 RELEASE_LOCK(&sched_mutex);
344 return THROWTO_SUCCESS;
348 case BlockedOnException:
354 To obtain exclusive access to a BlockedOnException thread,
355 we must call lockClosure() on the TSO on which it is blocked.
356 Since the TSO might change underneath our feet, after we
357 call lockClosure() we must check that
359 (a) the closure we locked is actually a TSO
360 (b) the original thread is still BlockedOnException,
361 (c) the original thread is still blocked on the TSO we locked
362 and (d) the target thread has not been relocated.
364 We synchronise with threadStackOverflow() (which relocates
365 threads) using lockClosure()/unlockClosure().
367 target2 = target->block_info.tso;
369 info = lockClosure((StgClosure *)target2);
370 if (info != &stg_TSO_info) {
371 unlockClosure((StgClosure *)target2, info);
374 if (target->what_next == ThreadRelocated) {
375 target = target->link;
379 if (target2->what_next == ThreadRelocated) {
380 target->block_info.tso = target2->link;
384 if (target->why_blocked != BlockedOnException
385 || target->block_info.tso != target2) {
391 Now we have exclusive rights to the target TSO...
393 If it is blocking exceptions, add the source TSO to its
394 blocked_exceptions queue. Otherwise, raise the exception.
396 if ((target->flags & TSO_BLOCKEX) &&
397 ((target->flags & TSO_INTERRUPTIBLE) == 0)) {
399 blockedThrowTo(source,target);
402 return THROWTO_BLOCKED;
404 removeThreadFromQueue(&target2->blocked_exceptions, target);
405 raiseAsync(cap, target, exception, rtsFalse, NULL);
406 unblockOne(cap, target);
408 return THROWTO_SUCCESS;
414 // Unblocking BlockedOnSTM threads requires the TSO to be
415 // locked; see STM.c:unpark_tso().
416 if (target->why_blocked != BlockedOnSTM) {
419 if ((target->flags & TSO_BLOCKEX) &&
420 ((target->flags & TSO_INTERRUPTIBLE) == 0)) {
421 blockedThrowTo(source,target);
423 return THROWTO_BLOCKED;
425 raiseAsync(cap, target, exception, rtsFalse, NULL);
426 unblockOne(cap, target);
428 return THROWTO_SUCCESS;
432 case BlockedOnCCall_NoUnblockExc:
433 // I don't think it's possible to acquire ownership of a
434 // BlockedOnCCall thread. We just assume that the target
435 // thread is blocking exceptions, and block on its
436 // blocked_exception queue.
438 blockedThrowTo(source,target);
440 return THROWTO_BLOCKED;
442 #ifndef THREADEDED_RTS
446 #if defined(mingw32_HOST_OS)
447 case BlockedOnDoProc:
449 if ((target->flags & TSO_BLOCKEX) &&
450 ((target->flags & TSO_INTERRUPTIBLE) == 0)) {
451 blockedThrowTo(source,target);
452 return THROWTO_BLOCKED;
454 removeFromQueues(cap,target);
455 raiseAsync(cap, target, exception, rtsFalse, NULL);
456 return THROWTO_SUCCESS;
461 barf("throwTo: unrecognised why_blocked value");
466 // Block a TSO on another TSO's blocked_exceptions queue.
467 // Precondition: we hold an exclusive lock on the target TSO (this is
468 // complex to achieve as there's no single lock on a TSO; see
471 blockedThrowTo (StgTSO *source, StgTSO *target)
473 debugTrace(DEBUG_sched, "throwTo: blocking on thread %lu", (unsigned long)target->id);
474 source->link = target->blocked_exceptions;
475 target->blocked_exceptions = source;
476 dirtyTSO(target); // we modified the blocked_exceptions queue
478 source->block_info.tso = target;
479 write_barrier(); // throwTo_exception *must* be visible if BlockedOnException is.
480 source->why_blocked = BlockedOnException;
486 throwToReleaseTarget (void *tso)
488 unlockTSO((StgTSO *)tso);
492 /* -----------------------------------------------------------------------------
493 Waking up threads blocked in throwTo
495 There are two ways to do this: maybePerformBlockedException() will
496 perform the throwTo() for the thread at the head of the queue
497 immediately, and leave the other threads on the queue.
498 maybePerformBlockedException() also checks the TSO_BLOCKEX flag
499 before raising an exception.
501 awakenBlockedExceptionQueue() will wake up all the threads in the
502 queue, but not perform any throwTo() immediately. This might be
503 more appropriate when the target thread is the one actually running
506 Returns: non-zero if an exception was raised, zero otherwise.
507 -------------------------------------------------------------------------- */
510 maybePerformBlockedException (Capability *cap, StgTSO *tso)
514 if (tso->blocked_exceptions != END_TSO_QUEUE
515 && ((tso->flags & TSO_BLOCKEX) == 0
516 || ((tso->flags & TSO_INTERRUPTIBLE) && interruptible(tso)))) {
518 // Lock the TSO, this gives us exclusive access to the queue
521 // Check the queue again; it might have changed before we
523 if (tso->blocked_exceptions == END_TSO_QUEUE) {
528 // We unblock just the first thread on the queue, and perform
529 // its throw immediately.
530 source = tso->blocked_exceptions;
531 performBlockedException(cap, source, tso);
532 tso->blocked_exceptions = unblockOne_(cap, source,
533 rtsFalse/*no migrate*/);
541 awakenBlockedExceptionQueue (Capability *cap, StgTSO *tso)
543 if (tso->blocked_exceptions != END_TSO_QUEUE) {
545 awakenBlockedQueue(cap, tso->blocked_exceptions);
546 tso->blocked_exceptions = END_TSO_QUEUE;
552 performBlockedException (Capability *cap, StgTSO *source, StgTSO *target)
554 StgClosure *exception;
556 ASSERT(source->why_blocked == BlockedOnException);
557 ASSERT(source->block_info.tso->id == target->id);
558 ASSERT(source->sp[0] == (StgWord)&stg_block_throwto_info);
559 ASSERT(((StgTSO *)source->sp[1])->id == target->id);
560 // check ids not pointers, because the thread might be relocated
562 exception = (StgClosure *)source->sp[2];
563 throwToSingleThreaded(cap, target, exception);
567 /* -----------------------------------------------------------------------------
568 Remove a thread from blocking queues.
570 This is for use when we raise an exception in another thread, which
572 This has nothing to do with the UnblockThread event in GranSim. -- HWL
573 -------------------------------------------------------------------------- */
575 #if defined(GRAN) || defined(PARALLEL_HASKELL)
577 NB: only the type of the blocking queue is different in GranSim and GUM
578 the operations on the queue-elements are the same
579 long live polymorphism!
581 Locks: sched_mutex is held upon entry and exit.
585 removeFromQueues(Capability *cap, StgTSO *tso)
587 StgBlockingQueueElement *t, **last;
589 switch (tso->why_blocked) {
592 return; /* not blocked */
595 // Be careful: nothing to do here! We tell the scheduler that the thread
596 // is runnable and we leave it to the stack-walking code to abort the
597 // transaction while unwinding the stack. We should perhaps have a debugging
598 // test to make sure that this really happens and that the 'zombie' transaction
599 // does not get committed.
603 ASSERT(get_itbl(tso->block_info.closure)->type == MVAR);
605 StgBlockingQueueElement *last_tso = END_BQ_QUEUE;
606 StgMVar *mvar = (StgMVar *)(tso->block_info.closure);
608 last = (StgBlockingQueueElement **)&mvar->head;
609 for (t = (StgBlockingQueueElement *)mvar->head;
611 last = &t->link, last_tso = t, t = t->link) {
612 if (t == (StgBlockingQueueElement *)tso) {
613 *last = (StgBlockingQueueElement *)tso->link;
614 if (mvar->tail == tso) {
615 mvar->tail = (StgTSO *)last_tso;
620 barf("removeFromQueues (MVAR): TSO not found");
623 case BlockedOnBlackHole:
624 ASSERT(get_itbl(tso->block_info.closure)->type == BLACKHOLE_BQ);
626 StgBlockingQueue *bq = (StgBlockingQueue *)(tso->block_info.closure);
628 last = &bq->blocking_queue;
629 for (t = bq->blocking_queue;
631 last = &t->link, t = t->link) {
632 if (t == (StgBlockingQueueElement *)tso) {
633 *last = (StgBlockingQueueElement *)tso->link;
637 barf("removeFromQueues (BLACKHOLE): TSO not found");
640 case BlockedOnException:
642 StgTSO *target = tso->block_info.tso;
644 ASSERT(get_itbl(target)->type == TSO);
646 while (target->what_next == ThreadRelocated) {
647 target = target2->link;
648 ASSERT(get_itbl(target)->type == TSO);
651 last = (StgBlockingQueueElement **)&target->blocked_exceptions;
652 for (t = (StgBlockingQueueElement *)target->blocked_exceptions;
654 last = &t->link, t = t->link) {
655 ASSERT(get_itbl(t)->type == TSO);
656 if (t == (StgBlockingQueueElement *)tso) {
657 *last = (StgBlockingQueueElement *)tso->link;
661 barf("removeFromQueues (Exception): TSO not found");
666 #if defined(mingw32_HOST_OS)
667 case BlockedOnDoProc:
670 /* take TSO off blocked_queue */
671 StgBlockingQueueElement *prev = NULL;
672 for (t = (StgBlockingQueueElement *)blocked_queue_hd; t != END_BQ_QUEUE;
673 prev = t, t = t->link) {
674 if (t == (StgBlockingQueueElement *)tso) {
676 blocked_queue_hd = (StgTSO *)t->link;
677 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
678 blocked_queue_tl = END_TSO_QUEUE;
681 prev->link = t->link;
682 if ((StgBlockingQueueElement *)blocked_queue_tl == t) {
683 blocked_queue_tl = (StgTSO *)prev;
686 #if defined(mingw32_HOST_OS)
687 /* (Cooperatively) signal that the worker thread should abort
690 abandonWorkRequest(tso->block_info.async_result->reqID);
695 barf("removeFromQueues (I/O): TSO not found");
700 /* take TSO off sleeping_queue */
701 StgBlockingQueueElement *prev = NULL;
702 for (t = (StgBlockingQueueElement *)sleeping_queue; t != END_BQ_QUEUE;
703 prev = t, t = t->link) {
704 if (t == (StgBlockingQueueElement *)tso) {
706 sleeping_queue = (StgTSO *)t->link;
708 prev->link = t->link;
713 barf("removeFromQueues (delay): TSO not found");
717 barf("removeFromQueues");
721 tso->link = END_TSO_QUEUE;
722 tso->why_blocked = NotBlocked;
723 tso->block_info.closure = NULL;
724 pushOnRunQueue(cap,tso);
728 removeFromQueues(Capability *cap, StgTSO *tso)
730 switch (tso->why_blocked) {
736 // Be careful: nothing to do here! We tell the scheduler that the
737 // thread is runnable and we leave it to the stack-walking code to
738 // abort the transaction while unwinding the stack. We should
739 // perhaps have a debugging test to make sure that this really
740 // happens and that the 'zombie' transaction does not get
745 removeThreadFromMVarQueue((StgMVar *)tso->block_info.closure, tso);
748 case BlockedOnBlackHole:
749 removeThreadFromQueue(&blackhole_queue, tso);
752 case BlockedOnException:
754 StgTSO *target = tso->block_info.tso;
756 // NO: when called by threadPaused(), we probably have this
757 // TSO already locked (WHITEHOLEd) because we just placed
758 // ourselves on its queue.
759 // ASSERT(get_itbl(target)->type == TSO);
761 while (target->what_next == ThreadRelocated) {
762 target = target->link;
765 removeThreadFromQueue(&target->blocked_exceptions, tso);
769 #if !defined(THREADED_RTS)
772 #if defined(mingw32_HOST_OS)
773 case BlockedOnDoProc:
775 removeThreadFromDeQueue(&blocked_queue_hd, &blocked_queue_tl, tso);
776 #if defined(mingw32_HOST_OS)
777 /* (Cooperatively) signal that the worker thread should abort
780 abandonWorkRequest(tso->block_info.async_result->reqID);
785 removeThreadFromQueue(&sleeping_queue, tso);
790 barf("removeFromQueues");
794 tso->link = END_TSO_QUEUE;
795 tso->why_blocked = NotBlocked;
796 tso->block_info.closure = NULL;
797 appendToRunQueue(cap,tso);
799 // We might have just migrated this TSO to our Capability:
801 tso->bound->cap = cap;
807 /* -----------------------------------------------------------------------------
810 * The following function implements the magic for raising an
811 * asynchronous exception in an existing thread.
813 * We first remove the thread from any queue on which it might be
814 * blocked. The possible blockages are MVARs and BLACKHOLE_BQs.
816 * We strip the stack down to the innermost CATCH_FRAME, building
817 * thunks in the heap for all the active computations, so they can
818 * be restarted if necessary. When we reach a CATCH_FRAME, we build
819 * an application of the handler to the exception, and push it on
820 * the top of the stack.
822 * How exactly do we save all the active computations? We create an
823 * AP_STACK for every UpdateFrame on the stack. Entering one of these
824 * AP_STACKs pushes everything from the corresponding update frame
825 * upwards onto the stack. (Actually, it pushes everything up to the
826 * next update frame plus a pointer to the next AP_STACK object.
827 * Entering the next AP_STACK object pushes more onto the stack until we
828 * reach the last AP_STACK object - at which point the stack should look
829 * exactly as it did when we killed the TSO and we can continue
830 * execution by entering the closure on top of the stack.
832 * We can also kill a thread entirely - this happens if either (a) the
833 * exception passed to raiseAsync is NULL, or (b) there's no
834 * CATCH_FRAME on the stack. In either case, we strip the entire
835 * stack and replace the thread with a zombie.
837 * ToDo: in THREADED_RTS mode, this function is only safe if either
838 * (a) we hold all the Capabilities (eg. in GC, or if there is only
839 * one Capability), or (b) we own the Capability that the TSO is
840 * currently blocked on or on the run queue of.
842 * -------------------------------------------------------------------------- */
845 raiseAsync(Capability *cap, StgTSO *tso, StgClosure *exception,
846 rtsBool stop_at_atomically, StgPtr stop_here)
848 StgRetInfoTable *info;
852 debugTrace(DEBUG_sched,
853 "raising exception in thread %ld.", (long)tso->id);
855 // mark it dirty; we're about to change its stack.
860 // ASSUMES: the thread is not already complete or dead. Upper
861 // layers should deal with that.
862 ASSERT(tso->what_next != ThreadComplete && tso->what_next != ThreadKilled);
864 // The stack freezing code assumes there's a closure pointer on
865 // the top of the stack, so we have to arrange that this is the case...
867 if (sp[0] == (W_)&stg_enter_info) {
871 sp[0] = (W_)&stg_dummy_ret_closure;
875 while (stop_here == NULL || frame < stop_here) {
877 // 1. Let the top of the stack be the "current closure"
879 // 2. Walk up the stack until we find either an UPDATE_FRAME or a
882 // 3. If it's an UPDATE_FRAME, then make an AP_STACK containing the
883 // current closure applied to the chunk of stack up to (but not
884 // including) the update frame. This closure becomes the "current
885 // closure". Go back to step 2.
887 // 4. If it's a CATCH_FRAME, then leave the exception handler on
888 // top of the stack applied to the exception.
890 // 5. If it's a STOP_FRAME, then kill the thread.
892 // NB: if we pass an ATOMICALLY_FRAME then abort the associated
895 info = get_ret_itbl((StgClosure *)frame);
897 switch (info->i.type) {
904 // First build an AP_STACK consisting of the stack chunk above the
905 // current update frame, with the top word on the stack as the
908 words = frame - sp - 1;
909 ap = (StgAP_STACK *)allocateLocal(cap,AP_STACK_sizeW(words));
912 ap->fun = (StgClosure *)sp[0];
914 for(i=0; i < (nat)words; ++i) {
915 ap->payload[i] = (StgClosure *)*sp++;
918 SET_HDR(ap,&stg_AP_STACK_info,
919 ((StgClosure *)frame)->header.prof.ccs /* ToDo */);
920 TICK_ALLOC_UP_THK(words+1,0);
922 //IF_DEBUG(scheduler,
923 // debugBelch("sched: Updating ");
924 // printPtr((P_)((StgUpdateFrame *)frame)->updatee);
925 // debugBelch(" with ");
926 // printObj((StgClosure *)ap);
929 // Replace the updatee with an indirection
931 // Warning: if we're in a loop, more than one update frame on
932 // the stack may point to the same object. Be careful not to
933 // overwrite an IND_OLDGEN in this case, because we'll screw
934 // up the mutable lists. To be on the safe side, don't
935 // overwrite any kind of indirection at all. See also
936 // threadSqueezeStack in GC.c, where we have to make a similar
939 if (!closure_IND(((StgUpdateFrame *)frame)->updatee)) {
940 // revert the black hole
941 UPD_IND_NOLOCK(((StgUpdateFrame *)frame)->updatee,
944 sp += sizeofW(StgUpdateFrame) - 1;
945 sp[0] = (W_)ap; // push onto stack
947 continue; //no need to bump frame
952 // We've stripped the entire stack, the thread is now dead.
953 tso->what_next = ThreadKilled;
954 tso->sp = frame + sizeofW(StgStopFrame);
959 // If we find a CATCH_FRAME, and we've got an exception to raise,
960 // then build the THUNK raise(exception), and leave it on
961 // top of the CATCH_FRAME ready to enter.
965 StgCatchFrame *cf = (StgCatchFrame *)frame;
969 if (exception == NULL) break;
971 // we've got an exception to raise, so let's pass it to the
972 // handler in this frame.
974 raise = (StgThunk *)allocateLocal(cap,sizeofW(StgThunk)+1);
975 TICK_ALLOC_SE_THK(1,0);
976 SET_HDR(raise,&stg_raise_info,cf->header.prof.ccs);
977 raise->payload[0] = exception;
979 // throw away the stack from Sp up to the CATCH_FRAME.
983 /* Ensure that async excpetions are blocked now, so we don't get
984 * a surprise exception before we get around to executing the
987 tso->flags |= TSO_BLOCKEX | TSO_INTERRUPTIBLE;
989 /* Put the newly-built THUNK on top of the stack, ready to execute
990 * when the thread restarts.
993 sp[-1] = (W_)&stg_enter_info;
995 tso->what_next = ThreadRunGHC;
996 IF_DEBUG(sanity, checkTSO(tso));
1000 case ATOMICALLY_FRAME:
1001 if (stop_at_atomically) {
1002 ASSERT(stmGetEnclosingTRec(tso->trec) == NO_TREC);
1003 stmCondemnTransaction(cap, tso -> trec);
1007 // R1 is not a register: the return convention for IO in
1008 // this case puts the return value on the stack, so we
1009 // need to set up the stack to return to the atomically
1010 // frame properly...
1011 tso->sp = frame - 2;
1012 tso->sp[1] = (StgWord) &stg_NO_FINALIZER_closure; // why not?
1013 tso->sp[0] = (StgWord) &stg_ut_1_0_unreg_info;
1015 tso->what_next = ThreadRunGHC;
1018 // Not stop_at_atomically... fall through and abort the
1021 case CATCH_RETRY_FRAME:
1022 // IF we find an ATOMICALLY_FRAME then we abort the
1023 // current transaction and propagate the exception. In
1024 // this case (unlike ordinary exceptions) we do not care
1025 // whether the transaction is valid or not because its
1026 // possible validity cannot have caused the exception
1027 // and will not be visible after the abort.
1030 StgTRecHeader *trec = tso -> trec;
1031 StgTRecHeader *outer = stmGetEnclosingTRec(trec);
1032 debugTrace(DEBUG_stm,
1033 "found atomically block delivering async exception");
1034 stmAbortTransaction(cap, trec);
1035 stmFreeAbortedTRec(cap, trec);
1036 tso -> trec = outer;
1044 // move on to the next stack frame
1045 frame += stack_frame_sizeW((StgClosure *)frame);
1048 // if we got here, then we stopped at stop_here
1049 ASSERT(stop_here != NULL);