2 {-# OPTIONS_GHC -XNoImplicitPrelude #-}
3 {-# OPTIONS_GHC -fno-warn-missing-signatures #-}
4 {-# OPTIONS_HADDOCK not-home #-}
5 -----------------------------------------------------------------------------
8 -- Copyright : (c) The University of Glasgow, 1994-2002
9 -- License : see libraries/base/LICENSE
11 -- Maintainer : cvs-ghc@haskell.org
12 -- Stability : internal
13 -- Portability : non-portable (GHC extensions)
15 -- Basic concurrency stuff.
17 -----------------------------------------------------------------------------
19 -- No: #hide, because bits of this module are exposed by the stm package.
20 -- However, we don't want this module to be the home location for the
21 -- bits it exports, we'd rather have Control.Concurrent and the other
22 -- higher level modules be the home. Hence:
30 -- * Forking and suchlike
31 , forkIO -- :: IO a -> IO ThreadId
32 , forkOnIO -- :: Int -> IO a -> IO ThreadId
33 , numCapabilities -- :: Int
34 , childHandler -- :: Exception -> IO ()
35 , myThreadId -- :: IO ThreadId
36 , killThread -- :: ThreadId -> IO ()
37 , throwTo -- :: ThreadId -> Exception -> IO ()
38 , par -- :: a -> b -> b
39 , pseq -- :: a -> b -> b
41 , labelThread -- :: ThreadId -> String -> IO ()
43 , ThreadStatus(..), BlockReason(..)
44 , threadStatus -- :: ThreadId -> IO ThreadStatus
47 , threadDelay -- :: Int -> IO ()
48 , registerDelay -- :: Int -> IO (TVar Bool)
49 , threadWaitRead -- :: Int -> IO ()
50 , threadWaitWrite -- :: Int -> IO ()
54 , newMVar -- :: a -> IO (MVar a)
55 , newEmptyMVar -- :: IO (MVar a)
56 , takeMVar -- :: MVar a -> IO a
57 , putMVar -- :: MVar a -> a -> IO ()
58 , tryTakeMVar -- :: MVar a -> IO (Maybe a)
59 , tryPutMVar -- :: MVar a -> a -> IO Bool
60 , isEmptyMVar -- :: MVar a -> IO Bool
61 , addMVarFinalizer -- :: MVar a -> IO () -> IO ()
65 , atomically -- :: STM a -> IO a
67 , orElse -- :: STM a -> STM a -> STM a
68 , catchSTM -- :: STM a -> (Exception -> STM a) -> STM a
69 , alwaysSucceeds -- :: STM a -> STM ()
70 , always -- :: STM Bool -> STM ()
72 , newTVar -- :: a -> STM (TVar a)
73 , newTVarIO -- :: a -> STM (TVar a)
74 , readTVar -- :: TVar a -> STM a
75 , readTVarIO -- :: TVar a -> IO a
76 , writeTVar -- :: a -> TVar a -> STM ()
77 , unsafeIOToSTM -- :: IO a -> STM a
80 #ifdef mingw32_HOST_OS
81 , asyncRead -- :: Int -> Int -> Int -> Ptr a -> IO (Int, Int)
82 , asyncWrite -- :: Int -> Int -> Int -> Ptr a -> IO (Int, Int)
83 , asyncDoProc -- :: FunPtr (Ptr a -> IO Int) -> Ptr a -> IO Int
85 , asyncReadBA -- :: Int -> Int -> Int -> Int -> MutableByteArray# RealWorld -> IO (Int, Int)
86 , asyncWriteBA -- :: Int -> Int -> Int -> Int -> MutableByteArray# RealWorld -> IO (Int, Int)
89 #ifndef mingw32_HOST_OS
93 , ensureIOManagerIsRunning
95 #ifdef mingw32_HOST_OS
100 , setUncaughtExceptionHandler -- :: (Exception -> IO ()) -> IO ()
101 , getUncaughtExceptionHandler -- :: IO (Exception -> IO ())
103 , reportError, reportStackOverflow
106 import System.Posix.Types
107 #ifndef mingw32_HOST_OS
108 import System.Posix.Internals
116 import {-# SOURCE #-} GHC.Handle
118 import GHC.Num ( Num(..) )
119 import GHC.Real ( fromIntegral )
120 #ifdef mingw32_HOST_OS
121 import GHC.Real ( div )
122 import GHC.Ptr ( plusPtr, FunPtr(..) )
124 #ifdef mingw32_HOST_OS
125 import GHC.Read ( Read )
126 import GHC.Enum ( Enum )
128 import GHC.Exception ( SomeException(..), throw )
129 import GHC.Pack ( packCString# )
130 import GHC.Ptr ( Ptr(..) )
132 import GHC.Show ( Show(..), showString )
136 infixr 0 `par`, `pseq`
139 %************************************************************************
141 \subsection{@ThreadId@, @par@, and @fork@}
143 %************************************************************************
146 data ThreadId = ThreadId ThreadId# deriving( Typeable )
147 -- ToDo: data ThreadId = ThreadId (Weak ThreadId#)
148 -- But since ThreadId# is unlifted, the Weak type must use open
151 A 'ThreadId' is an abstract type representing a handle to a thread.
152 'ThreadId' is an instance of 'Eq', 'Ord' and 'Show', where
153 the 'Ord' instance implements an arbitrary total ordering over
154 'ThreadId's. The 'Show' instance lets you convert an arbitrary-valued
155 'ThreadId' to string form; showing a 'ThreadId' value is occasionally
156 useful when debugging or diagnosing the behaviour of a concurrent
159 /Note/: in GHC, if you have a 'ThreadId', you essentially have
160 a pointer to the thread itself. This means the thread itself can\'t be
161 garbage collected until you drop the 'ThreadId'.
162 This misfeature will hopefully be corrected at a later date.
164 /Note/: Hugs does not provide any operations on other threads;
165 it defines 'ThreadId' as a synonym for ().
168 instance Show ThreadId where
170 showString "ThreadId " .
171 showsPrec d (getThreadId (id2TSO t))
173 foreign import ccall unsafe "rts_getThreadId" getThreadId :: ThreadId# -> CInt
175 id2TSO :: ThreadId -> ThreadId#
176 id2TSO (ThreadId t) = t
178 foreign import ccall unsafe "cmp_thread" cmp_thread :: ThreadId# -> ThreadId# -> CInt
181 cmpThread :: ThreadId -> ThreadId -> Ordering
183 case cmp_thread (id2TSO t1) (id2TSO t2) of
188 instance Eq ThreadId where
190 case t1 `cmpThread` t2 of
194 instance Ord ThreadId where
198 Sparks off a new thread to run the 'IO' computation passed as the
199 first argument, and returns the 'ThreadId' of the newly created
202 The new thread will be a lightweight thread; if you want to use a foreign
203 library that uses thread-local storage, use 'Control.Concurrent.forkOS' instead.
205 GHC note: the new thread inherits the /blocked/ state of the parent
206 (see 'Control.Exception.block').
208 forkIO :: IO () -> IO ThreadId
209 forkIO action = IO $ \ s ->
210 case (fork# action_plus s) of (# s1, tid #) -> (# s1, ThreadId tid #)
212 action_plus = catchException action childHandler
215 Like 'forkIO', but lets you specify on which CPU the thread is
216 created. Unlike a `forkIO` thread, a thread created by `forkOnIO`
217 will stay on the same CPU for its entire lifetime (`forkIO` threads
218 can migrate between CPUs according to the scheduling policy).
219 `forkOnIO` is useful for overriding the scheduling policy when you
220 know in advance how best to distribute the threads.
222 The `Int` argument specifies the CPU number; it is interpreted modulo
223 'numCapabilities' (note that it actually specifies a capability number
224 rather than a CPU number, but to a first approximation the two are
227 forkOnIO :: Int -> IO () -> IO ThreadId
228 forkOnIO (I# cpu) action = IO $ \ s ->
229 case (forkOn# cpu action_plus s) of (# s1, tid #) -> (# s1, ThreadId tid #)
231 action_plus = catchException action childHandler
233 -- | the value passed to the @+RTS -N@ flag. This is the number of
234 -- Haskell threads that can run truly simultaneously at any given
235 -- time, and is typically set to the number of physical CPU cores on
237 numCapabilities :: Int
238 numCapabilities = unsafePerformIO $ do
239 n <- peek n_capabilities
240 return (fromIntegral n)
242 #if defined(mingw32_HOST_OS) && defined(__PIC__)
243 foreign import ccall "_imp__n_capabilities" n_capabilities :: Ptr CInt
245 foreign import ccall "&n_capabilities" n_capabilities :: Ptr CInt
247 childHandler :: SomeException -> IO ()
248 childHandler err = catchException (real_handler err) childHandler
250 real_handler :: SomeException -> IO ()
251 real_handler se@(SomeException ex) =
252 -- ignore thread GC and killThread exceptions:
254 Just BlockedOnDeadMVar -> return ()
256 Just BlockedIndefinitely -> return ()
258 Just ThreadKilled -> return ()
260 -- report all others:
261 Just StackOverflow -> reportStackOverflow
264 {- | 'killThread' terminates the given thread (GHC only).
265 Any work already done by the thread isn\'t
266 lost: the computation is suspended until required by another thread.
267 The memory used by the thread will be garbage collected if it isn\'t
268 referenced from anywhere. The 'killThread' function is defined in
271 > killThread tid = throwTo tid ThreadKilled
273 Killthread is a no-op if the target thread has already completed.
275 killThread :: ThreadId -> IO ()
276 killThread tid = throwTo tid ThreadKilled
278 {- | 'throwTo' raises an arbitrary exception in the target thread (GHC only).
280 'throwTo' does not return until the exception has been raised in the
282 The calling thread can thus be certain that the target
283 thread has received the exception. This is a useful property to know
284 when dealing with race conditions: eg. if there are two threads that
285 can kill each other, it is guaranteed that only one of the threads
286 will get to kill the other.
288 If the target thread is currently making a foreign call, then the
289 exception will not be raised (and hence 'throwTo' will not return)
290 until the call has completed. This is the case regardless of whether
291 the call is inside a 'block' or not.
293 Important note: the behaviour of 'throwTo' differs from that described in
294 the paper \"Asynchronous exceptions in Haskell\"
295 (<http://research.microsoft.com/~simonpj/Papers/asynch-exns.htm>).
296 In the paper, 'throwTo' is non-blocking; but the library implementation adopts
297 a more synchronous design in which 'throwTo' does not return until the exception
298 is received by the target thread. The trade-off is discussed in Section 8 of the paper.
299 Like any blocking operation, 'throwTo' is therefore interruptible (see Section 4.3 of
302 There is currently no guarantee that the exception delivered by 'throwTo' will be
303 delivered at the first possible opportunity. In particular, if a thread may
304 unblock and then re-block exceptions (using 'unblock' and 'block') without receiving
305 a pending 'throwTo'. This is arguably undesirable behaviour.
308 throwTo :: Exception e => ThreadId -> e -> IO ()
309 throwTo (ThreadId tid) ex = IO $ \ s ->
310 case (killThread# tid (toException ex) s) of s1 -> (# s1, () #)
312 -- | Returns the 'ThreadId' of the calling thread (GHC only).
313 myThreadId :: IO ThreadId
314 myThreadId = IO $ \s ->
315 case (myThreadId# s) of (# s1, tid #) -> (# s1, ThreadId tid #)
318 -- |The 'yield' action allows (forces, in a co-operative multitasking
319 -- implementation) a context-switch to any other currently runnable
320 -- threads (if any), and is occasionally useful when implementing
321 -- concurrency abstractions.
324 case (yield# s) of s1 -> (# s1, () #)
326 {- | 'labelThread' stores a string as identifier for this thread if
327 you built a RTS with debugging support. This identifier will be used in
328 the debugging output to make distinction of different threads easier
329 (otherwise you only have the thread state object\'s address in the heap).
331 Other applications like the graphical Concurrent Haskell Debugger
332 (<http://www.informatik.uni-kiel.de/~fhu/chd/>) may choose to overload
333 'labelThread' for their purposes as well.
336 labelThread :: ThreadId -> String -> IO ()
337 labelThread (ThreadId t) str = IO $ \ s ->
338 let ps = packCString# str
339 adr = byteArrayContents# ps in
340 case (labelThread# t adr s) of s1 -> (# s1, () #)
342 -- Nota Bene: 'pseq' used to be 'seq'
343 -- but 'seq' is now defined in PrelGHC
345 -- "pseq" is defined a bit weirdly (see below)
347 -- The reason for the strange "lazy" call is that
348 -- it fools the compiler into thinking that pseq and par are non-strict in
349 -- their second argument (even if it inlines pseq at the call site).
350 -- If it thinks pseq is strict in "y", then it often evaluates
351 -- "y" before "x", which is totally wrong.
355 pseq x y = x `seq` lazy y
359 par x y = case (par# x) of { _ -> lazy y }
364 -- ^blocked on on 'MVar'
366 -- ^blocked on a computation in progress by another thread
368 -- ^blocked in 'throwTo'
370 -- ^blocked in 'retry' in an STM transaction
371 | BlockedOnForeignCall
372 -- ^currently in a foreign call
374 -- ^blocked on some other resource. Without @-threaded@,
375 -- I\/O and 'threadDelay' show up as 'BlockedOnOther', with @-threaded@
376 -- they show up as 'BlockedOnMVar'.
377 deriving (Eq,Ord,Show)
379 -- | The current status of a thread
382 -- ^the thread is currently runnable or running
384 -- ^the thread has finished
385 | ThreadBlocked BlockReason
386 -- ^the thread is blocked on some resource
388 -- ^the thread received an uncaught exception
389 deriving (Eq,Ord,Show)
391 threadStatus :: ThreadId -> IO ThreadStatus
392 threadStatus (ThreadId t) = IO $ \s ->
393 case threadStatus# t s of
394 (# s', stat #) -> (# s', mk_stat (I# stat) #)
396 -- NB. keep these in sync with includes/Constants.h
397 mk_stat 0 = ThreadRunning
398 mk_stat 1 = ThreadBlocked BlockedOnMVar
399 mk_stat 2 = ThreadBlocked BlockedOnBlackHole
400 mk_stat 3 = ThreadBlocked BlockedOnException
401 mk_stat 7 = ThreadBlocked BlockedOnSTM
402 mk_stat 11 = ThreadBlocked BlockedOnForeignCall
403 mk_stat 12 = ThreadBlocked BlockedOnForeignCall
404 mk_stat 16 = ThreadFinished
405 mk_stat 17 = ThreadDied
406 mk_stat _ = ThreadBlocked BlockedOnOther
410 %************************************************************************
412 \subsection[stm]{Transactional heap operations}
414 %************************************************************************
416 TVars are shared memory locations which support atomic memory
420 -- |A monad supporting atomic memory transactions.
421 newtype STM a = STM (State# RealWorld -> (# State# RealWorld, a #))
423 unSTM :: STM a -> (State# RealWorld -> (# State# RealWorld, a #))
426 INSTANCE_TYPEABLE1(STM,stmTc,"STM")
428 instance Functor STM where
429 fmap f x = x >>= (return . f)
431 instance Monad STM where
432 {-# INLINE return #-}
436 return x = returnSTM x
437 m >>= k = bindSTM m k
439 bindSTM :: STM a -> (a -> STM b) -> STM b
440 bindSTM (STM m) k = STM ( \s ->
442 (# new_s, a #) -> unSTM (k a) new_s
445 thenSTM :: STM a -> STM b -> STM b
446 thenSTM (STM m) k = STM ( \s ->
448 (# new_s, _ #) -> unSTM k new_s
451 returnSTM :: a -> STM a
452 returnSTM x = STM (\s -> (# s, x #))
454 -- | Unsafely performs IO in the STM monad. Beware: this is a highly
455 -- dangerous thing to do.
457 -- * The STM implementation will often run transactions multiple
458 -- times, so you need to be prepared for this if your IO has any
461 -- * The STM implementation will abort transactions that are known to
462 -- be invalid and need to be restarted. This may happen in the middle
463 -- of `unsafeIOToSTM`, so make sure you don't acquire any resources
464 -- that need releasing (exception handlers are ignored when aborting
465 -- the transaction). That includes doing any IO using Handles, for
466 -- example. Getting this wrong will probably lead to random deadlocks.
468 -- * The transaction may have seen an inconsistent view of memory when
469 -- the IO runs. Invariants that you expect to be true throughout
470 -- your program may not be true inside a transaction, due to the
471 -- way transactions are implemented. Normally this wouldn't be visible
472 -- to the programmer, but using `unsafeIOToSTM` can expose it.
474 unsafeIOToSTM :: IO a -> STM a
475 unsafeIOToSTM (IO m) = STM m
477 -- |Perform a series of STM actions atomically.
479 -- You cannot use 'atomically' inside an 'unsafePerformIO' or 'unsafeInterleaveIO'.
480 -- Any attempt to do so will result in a runtime error. (Reason: allowing
481 -- this would effectively allow a transaction inside a transaction, depending
482 -- on exactly when the thunk is evaluated.)
484 -- However, see 'newTVarIO', which can be called inside 'unsafePerformIO',
485 -- and which allows top-level TVars to be allocated.
487 atomically :: STM a -> IO a
488 atomically (STM m) = IO (\s -> (atomically# m) s )
490 -- |Retry execution of the current memory transaction because it has seen
491 -- values in TVars which mean that it should not continue (e.g. the TVars
492 -- represent a shared buffer that is now empty). The implementation may
493 -- block the thread until one of the TVars that it has read from has been
494 -- udpated. (GHC only)
496 retry = STM $ \s# -> retry# s#
498 -- |Compose two alternative STM actions (GHC only). If the first action
499 -- completes without retrying then it forms the result of the orElse.
500 -- Otherwise, if the first action retries, then the second action is
501 -- tried in its place. If both actions retry then the orElse as a
503 orElse :: STM a -> STM a -> STM a
504 orElse (STM m) e = STM $ \s -> catchRetry# m (unSTM e) s
506 -- |Exception handling within STM actions.
507 catchSTM :: STM a -> (SomeException -> STM a) -> STM a
508 catchSTM (STM m) k = STM $ \s -> catchSTM# m (\ex -> unSTM (k ex)) s
510 -- | Low-level primitive on which always and alwaysSucceeds are built.
511 -- checkInv differs form these in that (i) the invariant is not
512 -- checked when checkInv is called, only at the end of this and
513 -- subsequent transcations, (ii) the invariant failure is indicated
514 -- by raising an exception.
515 checkInv :: STM a -> STM ()
516 checkInv (STM m) = STM (\s -> (check# m) s)
518 -- | alwaysSucceeds adds a new invariant that must be true when passed
519 -- to alwaysSucceeds, at the end of the current transaction, and at
520 -- the end of every subsequent transaction. If it fails at any
521 -- of those points then the transaction violating it is aborted
522 -- and the exception raised by the invariant is propagated.
523 alwaysSucceeds :: STM a -> STM ()
524 alwaysSucceeds i = do ( do i ; retry ) `orElse` ( return () )
527 -- | always is a variant of alwaysSucceeds in which the invariant is
528 -- expressed as an STM Bool action that must return True. Returning
529 -- False or raising an exception are both treated as invariant failures.
530 always :: STM Bool -> STM ()
531 always i = alwaysSucceeds ( do v <- i
532 if (v) then return () else ( error "Transacional invariant violation" ) )
534 -- |Shared memory locations that support atomic memory transactions.
535 data TVar a = TVar (TVar# RealWorld a)
537 INSTANCE_TYPEABLE1(TVar,tvarTc,"TVar")
539 instance Eq (TVar a) where
540 (TVar tvar1#) == (TVar tvar2#) = sameTVar# tvar1# tvar2#
542 -- |Create a new TVar holding a value supplied
543 newTVar :: a -> STM (TVar a)
544 newTVar val = STM $ \s1# ->
545 case newTVar# val s1# of
546 (# s2#, tvar# #) -> (# s2#, TVar tvar# #)
548 -- |@IO@ version of 'newTVar'. This is useful for creating top-level
549 -- 'TVar's using 'System.IO.Unsafe.unsafePerformIO', because using
550 -- 'atomically' inside 'System.IO.Unsafe.unsafePerformIO' isn't
552 newTVarIO :: a -> IO (TVar a)
553 newTVarIO val = IO $ \s1# ->
554 case newTVar# val s1# of
555 (# s2#, tvar# #) -> (# s2#, TVar tvar# #)
557 -- |Return the current value stored in a TVar.
558 -- This is equivalent to
560 -- > readTVarIO = atomically . readTVar
562 -- but works much faster, because it doesn't perform a complete
563 -- transaction, it just reads the current value of the 'TVar'.
564 readTVarIO :: TVar a -> IO a
565 readTVarIO (TVar tvar#) = IO $ \s# -> readTVarIO# tvar# s#
567 -- |Return the current value stored in a TVar
568 readTVar :: TVar a -> STM a
569 readTVar (TVar tvar#) = STM $ \s# -> readTVar# tvar# s#
571 -- |Write the supplied value into a TVar
572 writeTVar :: TVar a -> a -> STM ()
573 writeTVar (TVar tvar#) val = STM $ \s1# ->
574 case writeTVar# tvar# val s1# of
579 %************************************************************************
581 \subsection[mvars]{M-Structures}
583 %************************************************************************
585 M-Vars are rendezvous points for concurrent threads. They begin
586 empty, and any attempt to read an empty M-Var blocks. When an M-Var
587 is written, a single blocked thread may be freed. Reading an M-Var
588 toggles its state from full back to empty. Therefore, any value
589 written to an M-Var may only be read once. Multiple reads and writes
590 are allowed, but there must be at least one read between any two
594 --Defined in IOBase to avoid cycle: data MVar a = MVar (SynchVar# RealWorld a)
596 -- |Create an 'MVar' which is initially empty.
597 newEmptyMVar :: IO (MVar a)
598 newEmptyMVar = IO $ \ s# ->
600 (# s2#, svar# #) -> (# s2#, MVar svar# #)
602 -- |Create an 'MVar' which contains the supplied value.
603 newMVar :: a -> IO (MVar a)
605 newEmptyMVar >>= \ mvar ->
606 putMVar mvar value >>
609 -- |Return the contents of the 'MVar'. If the 'MVar' is currently
610 -- empty, 'takeMVar' will wait until it is full. After a 'takeMVar',
611 -- the 'MVar' is left empty.
613 -- There are two further important properties of 'takeMVar':
615 -- * 'takeMVar' is single-wakeup. That is, if there are multiple
616 -- threads blocked in 'takeMVar', and the 'MVar' becomes full,
617 -- only one thread will be woken up. The runtime guarantees that
618 -- the woken thread completes its 'takeMVar' operation.
620 -- * When multiple threads are blocked on an 'MVar', they are
621 -- woken up in FIFO order. This is useful for providing
622 -- fairness properties of abstractions built using 'MVar's.
624 takeMVar :: MVar a -> IO a
625 takeMVar (MVar mvar#) = IO $ \ s# -> takeMVar# mvar# s#
627 -- |Put a value into an 'MVar'. If the 'MVar' is currently full,
628 -- 'putMVar' will wait until it becomes empty.
630 -- There are two further important properties of 'putMVar':
632 -- * 'putMVar' is single-wakeup. That is, if there are multiple
633 -- threads blocked in 'putMVar', and the 'MVar' becomes empty,
634 -- only one thread will be woken up. The runtime guarantees that
635 -- the woken thread completes its 'putMVar' operation.
637 -- * When multiple threads are blocked on an 'MVar', they are
638 -- woken up in FIFO order. This is useful for providing
639 -- fairness properties of abstractions built using 'MVar's.
641 putMVar :: MVar a -> a -> IO ()
642 putMVar (MVar mvar#) x = IO $ \ s# ->
643 case putMVar# mvar# x s# of
646 -- |A non-blocking version of 'takeMVar'. The 'tryTakeMVar' function
647 -- returns immediately, with 'Nothing' if the 'MVar' was empty, or
648 -- @'Just' a@ if the 'MVar' was full with contents @a@. After 'tryTakeMVar',
649 -- the 'MVar' is left empty.
650 tryTakeMVar :: MVar a -> IO (Maybe a)
651 tryTakeMVar (MVar m) = IO $ \ s ->
652 case tryTakeMVar# m s of
653 (# s', 0#, _ #) -> (# s', Nothing #) -- MVar is empty
654 (# s', _, a #) -> (# s', Just a #) -- MVar is full
656 -- |A non-blocking version of 'putMVar'. The 'tryPutMVar' function
657 -- attempts to put the value @a@ into the 'MVar', returning 'True' if
658 -- it was successful, or 'False' otherwise.
659 tryPutMVar :: MVar a -> a -> IO Bool
660 tryPutMVar (MVar mvar#) x = IO $ \ s# ->
661 case tryPutMVar# mvar# x s# of
662 (# s, 0# #) -> (# s, False #)
663 (# s, _ #) -> (# s, True #)
665 -- |Check whether a given 'MVar' is empty.
667 -- Notice that the boolean value returned is just a snapshot of
668 -- the state of the MVar. By the time you get to react on its result,
669 -- the MVar may have been filled (or emptied) - so be extremely
670 -- careful when using this operation. Use 'tryTakeMVar' instead if possible.
671 isEmptyMVar :: MVar a -> IO Bool
672 isEmptyMVar (MVar mv#) = IO $ \ s# ->
673 case isEmptyMVar# mv# s# of
674 (# s2#, flg #) -> (# s2#, not (flg ==# 0#) #)
676 -- |Add a finalizer to an 'MVar' (GHC only). See "Foreign.ForeignPtr" and
677 -- "System.Mem.Weak" for more about finalizers.
678 addMVarFinalizer :: MVar a -> IO () -> IO ()
679 addMVarFinalizer (MVar m) finalizer =
680 IO $ \s -> case mkWeak# m () finalizer s of { (# s1, _ #) -> (# s1, () #) }
682 withMVar :: MVar a -> (a -> IO b) -> IO b
686 b <- catchAny (unblock (io a))
687 (\e -> do putMVar m a; throw e)
693 %************************************************************************
695 \subsection{Thread waiting}
697 %************************************************************************
700 #ifdef mingw32_HOST_OS
702 -- Note: threadWaitRead and threadWaitWrite aren't really functional
703 -- on Win32, but left in there because lib code (still) uses them (the manner
704 -- in which they're used doesn't cause problems on a Win32 platform though.)
706 asyncRead :: Int -> Int -> Int -> Ptr a -> IO (Int, Int)
707 asyncRead (I# fd) (I# isSock) (I# len) (Ptr buf) =
708 IO $ \s -> case asyncRead# fd isSock len buf s of
709 (# s', len#, err# #) -> (# s', (I# len#, I# err#) #)
711 asyncWrite :: Int -> Int -> Int -> Ptr a -> IO (Int, Int)
712 asyncWrite (I# fd) (I# isSock) (I# len) (Ptr buf) =
713 IO $ \s -> case asyncWrite# fd isSock len buf s of
714 (# s', len#, err# #) -> (# s', (I# len#, I# err#) #)
716 asyncDoProc :: FunPtr (Ptr a -> IO Int) -> Ptr a -> IO Int
717 asyncDoProc (FunPtr proc) (Ptr param) =
718 -- the 'length' value is ignored; simplifies implementation of
719 -- the async*# primops to have them all return the same result.
720 IO $ \s -> case asyncDoProc# proc param s of
721 (# s', _len#, err# #) -> (# s', I# err# #)
723 -- to aid the use of these primops by the IO Handle implementation,
724 -- provide the following convenience funs:
726 -- this better be a pinned byte array!
727 asyncReadBA :: Int -> Int -> Int -> Int -> MutableByteArray# RealWorld -> IO (Int,Int)
728 asyncReadBA fd isSock len off bufB =
729 asyncRead fd isSock len ((Ptr (byteArrayContents# (unsafeCoerce# bufB))) `plusPtr` off)
731 asyncWriteBA :: Int -> Int -> Int -> Int -> MutableByteArray# RealWorld -> IO (Int,Int)
732 asyncWriteBA fd isSock len off bufB =
733 asyncWrite fd isSock len ((Ptr (byteArrayContents# (unsafeCoerce# bufB))) `plusPtr` off)
737 -- -----------------------------------------------------------------------------
740 -- | Block the current thread until data is available to read on the
741 -- given file descriptor (GHC only).
742 threadWaitRead :: Fd -> IO ()
744 #ifndef mingw32_HOST_OS
745 | threaded = waitForReadEvent fd
747 | otherwise = IO $ \s ->
748 case fromIntegral fd of { I# fd# ->
749 case waitRead# fd# s of { s' -> (# s', () #)
752 -- | Block the current thread until data can be written to the
753 -- given file descriptor (GHC only).
754 threadWaitWrite :: Fd -> IO ()
756 #ifndef mingw32_HOST_OS
757 | threaded = waitForWriteEvent fd
759 | otherwise = IO $ \s ->
760 case fromIntegral fd of { I# fd# ->
761 case waitWrite# fd# s of { s' -> (# s', () #)
764 -- | Suspends the current thread for a given number of microseconds
767 -- There is no guarantee that the thread will be rescheduled promptly
768 -- when the delay has expired, but the thread will never continue to
769 -- run /earlier/ than specified.
771 threadDelay :: Int -> IO ()
773 | threaded = waitForDelayEvent time
774 | otherwise = IO $ \s ->
775 case fromIntegral time of { I# time# ->
776 case delay# time# s of { s' -> (# s', () #)
780 -- | Set the value of returned TVar to True after a given number of
781 -- microseconds. The caveats associated with threadDelay also apply.
783 registerDelay :: Int -> IO (TVar Bool)
785 | threaded = waitForDelayEventSTM usecs
786 | otherwise = error "registerDelay: requires -threaded"
788 foreign import ccall unsafe "rtsSupportsBoundThreads" threaded :: Bool
790 waitForDelayEvent :: Int -> IO ()
791 waitForDelayEvent usecs = do
793 target <- calculateTarget usecs
794 atomicModifyIORef pendingDelays (\xs -> (Delay target m : xs, ()))
798 -- Delays for use in STM
799 waitForDelayEventSTM :: Int -> IO (TVar Bool)
800 waitForDelayEventSTM usecs = do
801 t <- atomically $ newTVar False
802 target <- calculateTarget usecs
803 atomicModifyIORef pendingDelays (\xs -> (DelaySTM target t : xs, ()))
807 calculateTarget :: Int -> IO USecs
808 calculateTarget usecs = do
810 return $ now + (fromIntegral usecs)
813 -- ----------------------------------------------------------------------------
814 -- Threaded RTS implementation of threadWaitRead, threadWaitWrite, threadDelay
816 -- In the threaded RTS, we employ a single IO Manager thread to wait
817 -- for all outstanding IO requests (threadWaitRead,threadWaitWrite)
818 -- and delays (threadDelay).
820 -- We can do this because in the threaded RTS the IO Manager can make
821 -- a non-blocking call to select(), so we don't have to do select() in
822 -- the scheduler as we have to in the non-threaded RTS. We get performance
823 -- benefits from doing it this way, because we only have to restart the select()
824 -- when a new request arrives, rather than doing one select() each time
825 -- around the scheduler loop. Furthermore, the scheduler can be simplified
826 -- by not having to check for completed IO requests.
828 -- Issues, possible problems:
830 -- - we might want bound threads to just do the blocking
831 -- operation rather than communicating with the IO manager
832 -- thread. This would prevent simgle-threaded programs which do
833 -- IO from requiring multiple OS threads. However, it would also
834 -- prevent bound threads waiting on IO from being killed or sent
837 -- - Apprently exec() doesn't work on Linux in a multithreaded program.
838 -- I couldn't repeat this.
840 -- - How do we handle signal delivery in the multithreaded RTS?
842 -- - forkProcess will kill the IO manager thread. Let's just
843 -- hope we don't need to do any blocking IO between fork & exec.
845 #ifndef mingw32_HOST_OS
847 = Read {-# UNPACK #-} !Fd {-# UNPACK #-} !(MVar ())
848 | Write {-# UNPACK #-} !Fd {-# UNPACK #-} !(MVar ())
852 = Delay {-# UNPACK #-} !USecs {-# UNPACK #-} !(MVar ())
853 | DelaySTM {-# UNPACK #-} !USecs {-# UNPACK #-} !(TVar Bool)
855 #ifndef mingw32_HOST_OS
856 pendingEvents :: IORef [IOReq]
858 pendingDelays :: IORef [DelayReq]
859 -- could use a strict list or array here
860 {-# NOINLINE pendingEvents #-}
861 {-# NOINLINE pendingDelays #-}
862 (pendingEvents,pendingDelays) = unsafePerformIO $ do
867 -- the first time we schedule an IO request, the service thread
868 -- will be created (cool, huh?)
870 ensureIOManagerIsRunning :: IO ()
871 ensureIOManagerIsRunning
872 | threaded = seq pendingEvents $ return ()
873 | otherwise = return ()
875 insertDelay :: DelayReq -> [DelayReq] -> [DelayReq]
876 insertDelay d [] = [d]
877 insertDelay d1 ds@(d2 : rest)
878 | delayTime d1 <= delayTime d2 = d1 : ds
879 | otherwise = d2 : insertDelay d1 rest
881 delayTime :: DelayReq -> USecs
882 delayTime (Delay t _) = t
883 delayTime (DelaySTM t _) = t
887 -- XXX: move into GHC.IOBase from Data.IORef?
888 atomicModifyIORef :: IORef a -> (a -> (a,b)) -> IO b
889 atomicModifyIORef (IORef (STRef r#)) f = IO $ \s -> atomicModifyMutVar# r# f s
891 foreign import ccall unsafe "getUSecOfDay"
892 getUSecOfDay :: IO USecs
894 prodding :: IORef Bool
895 {-# NOINLINE prodding #-}
896 prodding = unsafePerformIO (newIORef False)
898 prodServiceThread :: IO ()
899 prodServiceThread = do
900 was_set <- atomicModifyIORef prodding (\a -> (True,a))
901 if (not (was_set)) then wakeupIOManager else return ()
903 #ifdef mingw32_HOST_OS
904 -- ----------------------------------------------------------------------------
905 -- Windows IO manager thread
907 startIOManagerThread :: IO ()
908 startIOManagerThread = do
909 wakeup <- c_getIOManagerEvent
910 forkIO $ service_loop wakeup []
913 service_loop :: HANDLE -- read end of pipe
914 -> [DelayReq] -- current delay requests
917 service_loop wakeup old_delays = do
918 -- pick up new delay requests
919 new_delays <- atomicModifyIORef pendingDelays (\a -> ([],a))
920 let delays = foldr insertDelay old_delays new_delays
923 (delays', timeout) <- getDelay now delays
925 r <- c_WaitForSingleObject wakeup timeout
927 0xffffffff -> do c_maperrno; throwErrno "service_loop"
929 r2 <- c_readIOManagerEvent
932 _ | r2 == io_MANAGER_WAKEUP -> return False
933 _ | r2 == io_MANAGER_DIE -> return True
934 0 -> return False -- spurious wakeup
935 _ -> do start_console_handler (r2 `shiftR` 1); return False
938 else service_cont wakeup delays'
940 _other -> service_cont wakeup delays' -- probably timeout
942 service_cont :: HANDLE -> [DelayReq] -> IO ()
943 service_cont wakeup delays = do
944 atomicModifyIORef prodding (\_ -> (False,False))
945 service_loop wakeup delays
947 -- must agree with rts/win32/ThrIOManager.c
948 io_MANAGER_WAKEUP, io_MANAGER_DIE :: Word32
949 io_MANAGER_WAKEUP = 0xffffffff
950 io_MANAGER_DIE = 0xfffffffe
956 -- these are sent to Services only.
959 deriving (Eq, Ord, Enum, Show, Read, Typeable)
961 start_console_handler :: Word32 -> IO ()
962 start_console_handler r =
963 case toWin32ConsoleEvent r of
964 Just x -> withMVar win32ConsoleHandler $ \handler -> do
969 toWin32ConsoleEvent :: Num a => a -> Maybe ConsoleEvent
970 toWin32ConsoleEvent ev =
972 0 {- CTRL_C_EVENT-} -> Just ControlC
973 1 {- CTRL_BREAK_EVENT-} -> Just Break
974 2 {- CTRL_CLOSE_EVENT-} -> Just Close
975 5 {- CTRL_LOGOFF_EVENT-} -> Just Logoff
976 6 {- CTRL_SHUTDOWN_EVENT-} -> Just Shutdown
979 win32ConsoleHandler :: MVar (ConsoleEvent -> IO ())
980 win32ConsoleHandler = unsafePerformIO (newMVar (error "win32ConsoleHandler"))
982 -- XXX Is this actually needed?
983 stick :: IORef HANDLE
984 {-# NOINLINE stick #-}
985 stick = unsafePerformIO (newIORef nullPtr)
987 wakeupIOManager :: IO ()
989 _hdl <- readIORef stick
990 c_sendIOManagerEvent io_MANAGER_WAKEUP
992 -- Walk the queue of pending delays, waking up any that have passed
993 -- and return the smallest delay to wait for. The queue of pending
994 -- delays is kept ordered.
995 getDelay :: USecs -> [DelayReq] -> IO ([DelayReq], DWORD)
996 getDelay _ [] = return ([], iNFINITE)
997 getDelay now all@(d : rest)
999 Delay time m | now >= time -> do
1002 DelaySTM time t | now >= time -> do
1003 atomically $ writeTVar t True
1006 -- delay is in millisecs for WaitForSingleObject
1007 let micro_seconds = delayTime d - now
1008 milli_seconds = (micro_seconds + 999) `div` 1000
1009 in return (all, fromIntegral milli_seconds)
1011 -- ToDo: this just duplicates part of System.Win32.Types, which isn't
1012 -- available yet. We should move some Win32 functionality down here,
1013 -- maybe as part of the grand reorganisation of the base package...
1014 type HANDLE = Ptr ()
1018 iNFINITE = 0xFFFFFFFF -- urgh
1020 foreign import ccall unsafe "getIOManagerEvent" -- in the RTS (ThrIOManager.c)
1021 c_getIOManagerEvent :: IO HANDLE
1023 foreign import ccall unsafe "readIOManagerEvent" -- in the RTS (ThrIOManager.c)
1024 c_readIOManagerEvent :: IO Word32
1026 foreign import ccall unsafe "sendIOManagerEvent" -- in the RTS (ThrIOManager.c)
1027 c_sendIOManagerEvent :: Word32 -> IO ()
1029 foreign import ccall unsafe "maperrno" -- in Win32Utils.c
1032 foreign import stdcall "WaitForSingleObject"
1033 c_WaitForSingleObject :: HANDLE -> DWORD -> IO DWORD
1036 -- ----------------------------------------------------------------------------
1037 -- Unix IO manager thread, using select()
1039 startIOManagerThread :: IO ()
1040 startIOManagerThread = do
1041 allocaArray 2 $ \fds -> do
1042 throwErrnoIfMinus1 "startIOManagerThread" (c_pipe fds)
1043 rd_end <- peekElemOff fds 0
1044 wr_end <- peekElemOff fds 1
1045 writeIORef stick (fromIntegral wr_end)
1046 c_setIOManagerPipe wr_end
1048 allocaBytes sizeofFdSet $ \readfds -> do
1049 allocaBytes sizeofFdSet $ \writefds -> do
1050 allocaBytes sizeofTimeVal $ \timeval -> do
1051 service_loop (fromIntegral rd_end) readfds writefds timeval [] []
1055 :: Fd -- listen to this for wakeup calls
1062 service_loop wakeup readfds writefds ptimeval old_reqs old_delays = do
1064 -- pick up new IO requests
1065 new_reqs <- atomicModifyIORef pendingEvents (\a -> ([],a))
1066 let reqs = new_reqs ++ old_reqs
1068 -- pick up new delay requests
1069 new_delays <- atomicModifyIORef pendingDelays (\a -> ([],a))
1070 let delays0 = foldr insertDelay old_delays new_delays
1072 -- build the FDSets for select()
1075 fdSet wakeup readfds
1076 maxfd <- buildFdSets 0 readfds writefds reqs
1078 -- perform the select()
1079 let do_select delays = do
1080 -- check the current time and wake up any thread in
1081 -- threadDelay whose timeout has expired. Also find the
1082 -- timeout value for the select() call.
1084 (delays', timeout) <- getDelay now ptimeval delays
1086 res <- c_select (fromIntegral ((max wakeup maxfd)+1)) readfds writefds
1092 _ | err == eINTR -> do_select delays'
1093 -- EINTR: just redo the select()
1094 _ | err == eBADF -> return (True, delays)
1095 -- EBADF: one of the file descriptors is closed or bad,
1096 -- we don't know which one, so wake everyone up.
1097 _ | otherwise -> throwErrno "select"
1098 -- otherwise (ENOMEM or EINVAL) something has gone
1099 -- wrong; report the error.
1101 return (False,delays')
1103 (wakeup_all,delays') <- do_select delays0
1106 if wakeup_all then return False
1108 b <- fdIsSet wakeup readfds
1111 else alloca $ \p -> do
1112 c_read (fromIntegral wakeup) p 1; return ()
1115 _ | s == io_MANAGER_WAKEUP -> return False
1116 _ | s == io_MANAGER_DIE -> return True
1117 _ -> withMVar signalHandlerLock $ \_ -> do
1118 handler_tbl <- peek handlers
1119 sp <- peekElemOff handler_tbl (fromIntegral s)
1120 io <- deRefStablePtr sp
1124 if exit then return () else do
1126 atomicModifyIORef prodding (\_ -> (False,False))
1128 reqs' <- if wakeup_all then do wakeupAll reqs; return []
1129 else completeRequests reqs readfds writefds []
1131 service_loop wakeup readfds writefds ptimeval reqs' delays'
1133 io_MANAGER_WAKEUP, io_MANAGER_DIE :: CChar
1134 io_MANAGER_WAKEUP = 0xff
1135 io_MANAGER_DIE = 0xfe
1138 {-# NOINLINE stick #-}
1139 stick = unsafePerformIO (newIORef 0)
1141 wakeupIOManager :: IO ()
1142 wakeupIOManager = do
1143 fd <- readIORef stick
1144 with io_MANAGER_WAKEUP $ \pbuf -> do
1145 c_write (fromIntegral fd) pbuf 1; return ()
1147 -- Lock used to protect concurrent access to signal_handlers. Symptom of
1148 -- this race condition is #1922, although that bug was on Windows a similar
1149 -- bug also exists on Unix.
1150 signalHandlerLock :: MVar ()
1151 signalHandlerLock = unsafePerformIO (newMVar ())
1153 foreign import ccall "&signal_handlers" handlers :: Ptr (Ptr (StablePtr (IO ())))
1155 foreign import ccall "setIOManagerPipe"
1156 c_setIOManagerPipe :: CInt -> IO ()
1158 -- -----------------------------------------------------------------------------
1161 buildFdSets :: Fd -> Ptr CFdSet -> Ptr CFdSet -> [IOReq] -> IO Fd
1162 buildFdSets maxfd _ _ [] = return maxfd
1163 buildFdSets maxfd readfds writefds (Read fd _ : reqs)
1164 | fd >= fD_SETSIZE = error "buildFdSets: file descriptor out of range"
1167 buildFdSets (max maxfd fd) readfds writefds reqs
1168 buildFdSets maxfd readfds writefds (Write fd _ : reqs)
1169 | fd >= fD_SETSIZE = error "buildFdSets: file descriptor out of range"
1172 buildFdSets (max maxfd fd) readfds writefds reqs
1174 completeRequests :: [IOReq] -> Ptr CFdSet -> Ptr CFdSet -> [IOReq]
1176 completeRequests [] _ _ reqs' = return reqs'
1177 completeRequests (Read fd m : reqs) readfds writefds reqs' = do
1178 b <- fdIsSet fd readfds
1180 then do putMVar m (); completeRequests reqs readfds writefds reqs'
1181 else completeRequests reqs readfds writefds (Read fd m : reqs')
1182 completeRequests (Write fd m : reqs) readfds writefds reqs' = do
1183 b <- fdIsSet fd writefds
1185 then do putMVar m (); completeRequests reqs readfds writefds reqs'
1186 else completeRequests reqs readfds writefds (Write fd m : reqs')
1188 wakeupAll :: [IOReq] -> IO ()
1189 wakeupAll [] = return ()
1190 wakeupAll (Read _ m : reqs) = do putMVar m (); wakeupAll reqs
1191 wakeupAll (Write _ m : reqs) = do putMVar m (); wakeupAll reqs
1193 waitForReadEvent :: Fd -> IO ()
1194 waitForReadEvent fd = do
1196 atomicModifyIORef pendingEvents (\xs -> (Read fd m : xs, ()))
1200 waitForWriteEvent :: Fd -> IO ()
1201 waitForWriteEvent fd = do
1203 atomicModifyIORef pendingEvents (\xs -> (Write fd m : xs, ()))
1207 -- -----------------------------------------------------------------------------
1210 -- Walk the queue of pending delays, waking up any that have passed
1211 -- and return the smallest delay to wait for. The queue of pending
1212 -- delays is kept ordered.
1213 getDelay :: USecs -> Ptr CTimeVal -> [DelayReq] -> IO ([DelayReq], Ptr CTimeVal)
1214 getDelay _ _ [] = return ([],nullPtr)
1215 getDelay now ptimeval all@(d : rest)
1217 Delay time m | now >= time -> do
1219 getDelay now ptimeval rest
1220 DelaySTM time t | now >= time -> do
1221 atomically $ writeTVar t True
1222 getDelay now ptimeval rest
1224 setTimevalTicks ptimeval (delayTime d - now)
1225 return (all,ptimeval)
1229 foreign import ccall unsafe "sizeofTimeVal"
1230 sizeofTimeVal :: Int
1232 foreign import ccall unsafe "setTimevalTicks"
1233 setTimevalTicks :: Ptr CTimeVal -> USecs -> IO ()
1236 On Win32 we're going to have a single Pipe, and a
1237 waitForSingleObject with the delay time. For signals, we send a
1238 byte down the pipe just like on Unix.
1241 -- ----------------------------------------------------------------------------
1242 -- select() interface
1244 -- ToDo: move to System.Posix.Internals?
1248 foreign import ccall safe "select"
1249 c_select :: CInt -> Ptr CFdSet -> Ptr CFdSet -> Ptr CFdSet -> Ptr CTimeVal
1252 foreign import ccall unsafe "hsFD_SETSIZE"
1253 c_fD_SETSIZE :: CInt
1256 fD_SETSIZE = fromIntegral c_fD_SETSIZE
1258 foreign import ccall unsafe "hsFD_ISSET"
1259 c_fdIsSet :: CInt -> Ptr CFdSet -> IO CInt
1261 fdIsSet :: Fd -> Ptr CFdSet -> IO CInt
1262 fdIsSet (Fd fd) fdset = c_fdIsSet fd fdset
1264 foreign import ccall unsafe "hsFD_SET"
1265 c_fdSet :: CInt -> Ptr CFdSet -> IO ()
1267 fdSet :: Fd -> Ptr CFdSet -> IO ()
1268 fdSet (Fd fd) fdset = c_fdSet fd fdset
1270 foreign import ccall unsafe "hsFD_ZERO"
1271 fdZero :: Ptr CFdSet -> IO ()
1273 foreign import ccall unsafe "sizeof_fd_set"
1278 reportStackOverflow :: IO a
1279 reportStackOverflow = do callStackOverflowHook; return undefined
1281 reportError :: SomeException -> IO a
1283 handler <- getUncaughtExceptionHandler
1287 -- SUP: Are the hooks allowed to re-enter Haskell land? If so, remove
1288 -- the unsafe below.
1289 foreign import ccall unsafe "stackOverflow"
1290 callStackOverflowHook :: IO ()
1292 {-# NOINLINE uncaughtExceptionHandler #-}
1293 uncaughtExceptionHandler :: IORef (SomeException -> IO ())
1294 uncaughtExceptionHandler = unsafePerformIO (newIORef defaultHandler)
1296 defaultHandler :: SomeException -> IO ()
1297 defaultHandler se@(SomeException ex) = do
1298 (hFlush stdout) `catchAny` (\ _ -> return ())
1299 let msg = case cast ex of
1300 Just Deadlock -> "no threads to run: infinite loop or deadlock?"
1301 _ -> case cast ex of
1302 Just (ErrorCall s) -> s
1303 _ -> showsPrec 0 se ""
1304 withCString "%s" $ \cfmt ->
1305 withCString msg $ \cmsg ->
1306 errorBelch cfmt cmsg
1308 -- don't use errorBelch() directly, because we cannot call varargs functions
1310 foreign import ccall unsafe "HsBase.h errorBelch2"
1311 errorBelch :: CString -> CString -> IO ()
1313 setUncaughtExceptionHandler :: (SomeException -> IO ()) -> IO ()
1314 setUncaughtExceptionHandler = writeIORef uncaughtExceptionHandler
1316 getUncaughtExceptionHandler :: IO (SomeException -> IO ())
1317 getUncaughtExceptionHandler = readIORef uncaughtExceptionHandler