/* -----------------------------------------------------------------------------
- * $Id: Select.c,v 1.9 2000/03/16 17:24:08 simonmar Exp $
+ * $Id: Select.c,v 1.29 2003/06/26 12:22:59 stolz Exp $
*
- * (c) The GHC Team 1995-1999
+ * (c) The GHC Team 1995-2002
*
* Support for concurrent non-blocking I/O and thread waiting.
*
* ---------------------------------------------------------------------------*/
+
/* we're outside the realms of POSIX here... */
-#define NON_POSIX_SOURCE
+/* #include "PosixSource.h" */
#include "Rts.h"
#include "Schedule.h"
#include "RtsUtils.h"
#include "RtsFlags.h"
+#include "Timer.h"
#include "Itimer.h"
#include "Signals.h"
+#include "Capability.h"
-# if defined(HAVE_SYS_TYPES_H)
+# ifdef HAVE_SYS_TYPES_H
# include <sys/types.h>
# endif
# include <sys/time.h>
# endif
-nat ticks_since_select = 0;
+#include <errno.h>
+#include <string.h>
+
+#ifdef HAVE_UNISTD_H
+#include <unistd.h>
+#endif
+
+/* last timestamp */
+nat timestamp = 0;
+
+#ifdef RTS_SUPPORTS_THREADS
+static rtsBool isWorkerBlockedInAwaitEvent = rtsFalse;
+static rtsBool workerWakeupPending = rtsFalse;
+static int workerWakeupPipe[2];
+static rtsBool workerWakeupInited = rtsFalse;
+#endif
+
+/* There's a clever trick here to avoid problems when the time wraps
+ * around. Since our maximum delay is smaller than 31 bits of ticks
+ * (it's actually 31 bits of microseconds), we can safely check
+ * whether a timer has expired even if our timer will wrap around
+ * before the target is reached, using the following formula:
+ *
+ * (int)((uint)current_time - (uint)target_time) < 0
+ *
+ * if this is true, then our time has expired.
+ * (idea due to Andy Gill).
+ */
+rtsBool
+wakeUpSleepingThreads(nat ticks)
+{
+ StgTSO *tso;
+ rtsBool flag = rtsFalse;
+
+ while (sleeping_queue != END_TSO_QUEUE &&
+ (int)(ticks - sleeping_queue->block_info.target) > 0) {
+ tso = sleeping_queue;
+ sleeping_queue = tso->link;
+ tso->why_blocked = NotBlocked;
+ tso->link = END_TSO_QUEUE;
+ IF_DEBUG(scheduler,belch("Waking up sleeping thread %d\n", tso->id));
+ PUSH_ON_RUN_QUEUE(tso);
+ flag = rtsTrue;
+ }
+ return flag;
+}
/* Argument 'wait' says whether to wait for I/O to become available,
* or whether to just check and return immediately. If there are
* otherwise we wait (see Schedule.c).
*
* SMP note: must be called with sched_mutex locked.
+ *
+ * Windows: select only works on sockets, so this doesn't really work,
+ * though it makes things better than before. MsgWaitForMultipleObjects
+ * should really be used, though it only seems to work for read handles,
+ * not write handles.
+ *
*/
void
awaitEvent(rtsBool wait)
{
-#ifdef mingw32_TARGET_OS
-/*
- * Win32 doesn't support select(). ToDo: use MsgWaitForMultipleObjects()
- * to achieve (similar) effect.
- *
- */
- return;
-#else
-
StgTSO *tso, *prev, *next;
rtsBool ready;
fd_set rfd,wfd;
int numFound;
- nat min, delta;
int maxfd = -1;
-
+ rtsBool select_succeeded = rtsTrue;
+ rtsBool unblock_all = rtsFalse;
struct timeval tv;
-#ifndef linux_TARGET_OS
- struct timeval tv_before,tv_after;
-#endif
+ lnat min, ticks;
- IF_DEBUG(scheduler,belch("Checking for threads blocked on I/O...\n"));
+ tv.tv_sec = 0;
+ tv.tv_usec = 0;
+
+ IF_DEBUG(scheduler,
+ belch("scheduler: checking for threads blocked on I/O");
+ if (wait) {
+ belch(" (waiting)");
+ }
+ belch("\n");
+ );
/* loop until we've woken up some threads. This loop is needed
* because the select timing isn't accurate, we sometimes sleep
*/
do {
- /* see how long it's been since we last checked the blocked queue.
- * ToDo: make this check atomic, so we don't lose any ticks.
- */
- delta = ticks_since_select;
- ticks_since_select = 0;
- delta = delta * TICK_MILLISECS * 1000;
+ ticks = timestamp = getourtimeofday();
+ if (wakeUpSleepingThreads(ticks)) {
+ return;
+ }
- min = wait == rtsTrue ? 0x7fffffff : 0;
+ if (!wait) {
+ min = 0;
+ } else if (sleeping_queue != END_TSO_QUEUE) {
+ min = (sleeping_queue->block_info.target - ticks)
+ * TICK_MILLISECS * 1000;
+ } else {
+ min = 0x7ffffff;
+ }
/*
- * Collect all of the fd's that we're interested in, and capture
- * the minimum waiting time (in microseconds) for the delayed threads.
+ * Collect all of the fd's that we're interested in
*/
FD_ZERO(&rfd);
FD_ZERO(&wfd);
continue;
}
- case BlockedOnDelay:
- {
- if (tso->block_info.delay < min)
- min = tso->block_info.delay;
- continue;
- }
-
default:
barf("AwaitEvent");
}
}
+#ifdef RTS_SUPPORTS_THREADS
+ if(!workerWakeupInited) {
+ pipe(workerWakeupPipe);
+ workerWakeupInited = rtsTrue;
+ }
+ FD_SET(workerWakeupPipe[0], &rfd);
+ maxfd = workerWakeupPipe[0] > maxfd ? workerWakeupPipe[0] : maxfd;
+#endif
+
/* Release the scheduler lock while we do the poll.
* this means that someone might muck with the blocked_queue
* while we do this, but it shouldn't matter:
*
* I believe none of these cases lead to trouble --SDM.
*/
+
+#ifdef RTS_SUPPORTS_THREADS
+ isWorkerBlockedInAwaitEvent = rtsTrue;
+ workerWakeupPending = rtsFalse;
+#endif
RELEASE_LOCK(&sched_mutex);
/* Check for any interesting events */
-
- tv.tv_sec = min / 1000000;
+
+ tv.tv_sec = min / 1000000;
tv.tv_usec = min % 1000000;
-#ifndef linux_TARGET_OS
- gettimeofday(&tv_before, (struct timezone *) NULL);
+ while ((numFound = select(maxfd+1, &rfd, &wfd, NULL, &tv)) < 0) {
+ if (errno != EINTR) {
+ /* Handle bad file descriptors by unblocking all the
+ waiting threads. Why? Because a thread might have been
+ a bit naughty and closed a file descriptor while another
+ was blocked waiting. This is less-than-good programming
+ practice, but having the RTS as a result fall over isn't
+ acceptable, so we simply unblock all the waiting threads
+ should we see a bad file descriptor & give the threads
+ a chance to clean up their act.
+
+ Note: assume here that threads becoming unblocked
+ will try to read/write the file descriptor before trying
+ to issue a threadWaitRead/threadWaitWrite again (==> an
+ IOError will result for the thread that's got the bad
+ file descriptor.) Hence, there's no danger of a bad
+ file descriptor being repeatedly select()'ed on, so
+ the RTS won't loop.
+ */
+ if ( errno == EBADF ) {
+ unblock_all = rtsTrue;
+ break;
+ } else {
+ fprintf(stderr,"%d\n", errno);
+ fflush(stderr);
+ perror("select");
+ barf("select failed");
+ }
+ }
+ ACQUIRE_LOCK(&sched_mutex);
+#ifdef RTS_SUPPORTS_THREADS
+ isWorkerBlockedInAwaitEvent = rtsFalse;
#endif
- while (!interrupted &&
- (numFound = select(maxfd+1, &rfd, &wfd, NULL, &tv)) < 0) {
- if (errno != EINTR) {
- /* fflush(stdout); */
- perror("select");
- barf("select failed");
- }
- ACQUIRE_LOCK(&sched_mutex);
-
- /* We got a signal; could be one of ours. If so, we need
- * to start up the signal handler straight away, otherwise
- * we could block for a long time before the signal is
- * serviced.
- */
- if (signals_pending()) {
- RELEASE_LOCK(&sched_mutex);
- start_signal_handlers();
- break;
- }
-
- /* If new runnable threads have arrived, stop waiting for
- * I/O and run them.
- */
- if (run_queue_hd != END_TSO_QUEUE) {
- RELEASE_LOCK(&sched_mutex);
- break;
- }
-
- RELEASE_LOCK(&sched_mutex);
- }
-
-#ifdef linux_TARGET_OS
- /* on Linux, tv is set to indicate the amount of time not
- * slept, so we don't need to gettimeofday() to find out.
- */
- delta += min - (tv.tv_sec * 1000000 + tv.tv_usec);
-#else
- gettimeofday(&tv_after, (struct timezone *) NULL);
- delta += (tv_after.tv_sec - tv_before.tv_sec) * 1000000 +
- tv_after.tv_usec - tv_before.tv_usec;
+ /* We got a signal; could be one of ours. If so, we need
+ * to start up the signal handler straight away, otherwise
+ * we could block for a long time before the signal is
+ * serviced.
+ */
+#if defined(RTS_USER_SIGNALS)
+ if (signals_pending()) {
+ RELEASE_LOCK(&sched_mutex); /* ToDo: kill */
+ startSignalHandlers();
+ ACQUIRE_LOCK(&sched_mutex);
+ return; /* still hold the lock */
+ }
#endif
-#if 0
- if (delta != 0) { fprintf(stderr,"waited: %d %d %d\n", min, delta,
- interrupted); }
+ /* we were interrupted, return to the scheduler immediately.
+ */
+ if (interrupted) {
+ return; /* still hold the lock */
+ }
+
+ /* check for threads that need waking up
+ */
+ wakeUpSleepingThreads(getourtimeofday());
+
+ /* If new runnable threads have arrived, stop waiting for
+ * I/O and run them.
+ */
+ if (run_queue_hd != END_TSO_QUEUE) {
+ return; /* still hold the lock */
+ }
+
+#ifdef RTS_SUPPORTS_THREADS
+ /* If another worker thread wants to take over,
+ * return to the scheduler
+ */
+ if (needToYieldToReturningWorker()) {
+ return; /* still hold the lock */
+ }
#endif
+
+#ifdef RTS_SUPPORTS_THREADS
+ isWorkerBlockedInAwaitEvent = rtsTrue;
+#endif
+ RELEASE_LOCK(&sched_mutex);
+ }
ACQUIRE_LOCK(&sched_mutex);
/* Step through the waiting queue, unblocking every thread that now has
* a file descriptor in a ready state.
-
- * For the delayed threads, decrement the number of microsecs
- * we've been blocked for. Unblock the threads that have thusly expired.
*/
prev = NULL;
- for(tso = blocked_queue_hd; tso != END_TSO_QUEUE; tso = next) {
- next = tso->link;
- switch (tso->why_blocked) {
- case BlockedOnRead:
- ready = FD_ISSET(tso->block_info.fd, &rfd);
- break;
-
- case BlockedOnWrite:
- ready = FD_ISSET(tso->block_info.fd, &wfd);
- break;
-
- case BlockedOnDelay:
- if (tso->block_info.delay > delta) {
- tso->block_info.delay -= delta;
- ready = 0;
- } else {
- tso->block_info.delay = 0;
- ready = 1;
- }
- break;
-
- default:
- barf("awaitEvent");
- }
+ if (select_succeeded || unblock_all) {
+ for(tso = blocked_queue_hd; tso != END_TSO_QUEUE; tso = next) {
+ next = tso->link;
+ switch (tso->why_blocked) {
+ case BlockedOnRead:
+ ready = unblock_all || FD_ISSET(tso->block_info.fd, &rfd);
+ break;
+ case BlockedOnWrite:
+ ready = unblock_all || FD_ISSET(tso->block_info.fd, &wfd);
+ break;
+ default:
+ barf("awaitEvent");
+ }
- if (ready) {
- IF_DEBUG(scheduler,belch("Waking up thread %d\n", tso->id));
- tso->why_blocked = NotBlocked;
- tso->link = END_TSO_QUEUE;
- PUSH_ON_RUN_QUEUE(tso);
- } else {
+ if (ready) {
+ IF_DEBUG(scheduler,belch("Waking up blocked thread %d\n", tso->id));
+ tso->why_blocked = NotBlocked;
+ tso->link = END_TSO_QUEUE;
+ PUSH_ON_RUN_QUEUE(tso);
+ } else {
+ if (prev == NULL)
+ blocked_queue_hd = tso;
+ else
+ prev->link = tso;
+ prev = tso;
+ }
+ }
+
if (prev == NULL)
- blocked_queue_hd = tso;
- else
- prev->link = tso;
- prev = tso;
- }
+ blocked_queue_hd = blocked_queue_tl = END_TSO_QUEUE;
+ else {
+ prev->link = END_TSO_QUEUE;
+ blocked_queue_tl = prev;
+ }
}
-
- if (prev == NULL)
- blocked_queue_hd = blocked_queue_tl = END_TSO_QUEUE;
- else {
- prev->link = END_TSO_QUEUE;
- blocked_queue_tl = prev;
+
+#if defined(RTS_SUPPORTS_THREADS)
+ // if we were woken up by wakeBlockedWorkerThread,
+ // read the dummy byte from the pipe
+ if(select_succeeded && FD_ISSET(workerWakeupPipe[0], &rfd)) {
+ unsigned char dummy;
+ wait = rtsFalse;
+ read(workerWakeupPipe[0],&dummy,1);
}
-
- } while (wait && run_queue_hd == END_TSO_QUEUE);
#endif
+ } while (wait && !interrupted && run_queue_hd == END_TSO_QUEUE);
}
+
+
+#ifdef RTS_SUPPORTS_THREADS
+/* wakeBlockedWorkerThread
+ *
+ * If a worker thread is currently blocked within awaitEvent,
+ * wake it.
+ * Must be called with sched_mutex held.
+ */
+void
+wakeBlockedWorkerThread()
+{
+ if(isWorkerBlockedInAwaitEvent && !workerWakeupPending) {
+ unsigned char dummy = 42; // Any value will do here
+
+ // write something so that select() wakes up
+ write(workerWakeupPipe[1],&dummy,1);
+ workerWakeupPending = rtsTrue;
+ }
+}
+#endif