Duncan Coutts pushed to branch wip/dcoutts/issue-27105-stopTicker at Glasgow Haskell Compiler / GHC
Commits:
9421fe08 by Duncan Coutts at 2026-06-06T23:51:30+02:00
Promote HAVE_PREEMPTION from Timer.c to OSThreads.h
We will want to know about HAVE_PREEMPTION in more places.
HAVE_PREEMPTION tells us that we do have OS threads available,
irrespective of whether THREADED is defined. In particular,
HAVE_PREEMPTION is defined on all proper OSs, but not on WASM (and
hyopthetically may not be true on some other platforms like
micro-controllers, RTOSs, VM hypervisors etc).
- - - - -
542e0742 by Duncan Coutts at 2026-06-07T09:19:31+02:00
Define ACQUIRE_LOCK_ALWAYS and friends
Fix issue #27335
Like the atomic _ALWAYS variants, these lock actions are always defined,
rather than being dependent on whether we are in the THREADED case. All
the "normal" LOCK macros are defined to be no-ops when !THREADED.
The use case for the _ALWAYS variants is where we are using OS threads
even in the non-threaded RTS. This includes everything to do with the
timer/ticker thread, which is used in the non-threaded RTS too.
In particular, we will want to use this for eventlog things, because the
timer thread performs eventlogging concurrently with the main
capability, even in the non-threaded RTS.
- - - - -
5d90f476 by Duncan Coutts at 2026-06-07T09:19:31+02:00
Use ACQUIRE/RELEASE_LOCK_ALWAYS with eventBufMutex
Even in the non-threaded RTS the eventBufMutex is needed by both the
main capability and the timer/ticker thread, so always use the mutex.
This should fix #25165 which is about the main capability and the timer
thread posting events to the eventlog buffer concurrently and thereby
corrupting the buffer data.
- - - - -
233e6c16 by Duncan Coutts at 2026-06-07T09:19:31+02:00
Expose eventBufMutex in the EventLog interface/header
We will need it in forkProcess to ensure we don't write to the global
eventlog buffer concurrently with trying to flush eventlog buffers and
do the fork().
- - - - -
baba3e89 by Duncan Coutts at 2026-06-07T09:19:31+02:00
Split flushAllCapsEventsBufs into safe and unlocked version
Following the convention that unlocked versions have a trailing _
underscore in their name. This one requires the caller to hold the
eventlog global buffer mutex. We will need this in forkProcess.
- - - - -
57d166ae by Duncan Coutts at 2026-06-07T09:19:31+02:00
Remove redundant use of stopTimer in setNumCapabilities
Historically, the comment here was:
We must stop the interval timer while we are changing the
capabilities array lest handle_tick may try to context switch
an old capability. See #17289.
and
We must disable the timer while we do this since the tick handler may
call contextSwitchAllCapabilities, which may see the capabilities array
as we free it.
What this refers to is that historically, when changing the number of
capabilities, the array of capabilities was reallocated to a new size,
allocating new ones and freeing the old ones, thus invalidating all
existing capbility pointers.
Strangely, for good measure the code used to call stopTimer twice (hence
the two similar comments above).
However, since commit a3eccf06292dd666b24606251a52da2b466a9612, the
capabilities array is no longer reallocated. Instead the array is
allcoated once on RTS startup to the maximum size it could ever be
allowed to be, and then capabilities get enabled/disabled at runtime. So
the capability pointers never become invalid anymore. At worst, they may
point to capabilities that are disabled.
Thus we no longer need to stop the timer (twice) while we change the
number of enabled capabilities. This also partially solves issue #27105,
which notes that stopTimer is being used as if it were synchronous, when
it is not. At least for this case, the solution is that stopTimer is not
needed at all!
- - - - -
e2c8b108 by Duncan Coutts at 2026-06-07T09:19:31+02:00
Remove redundant use of stopTimer in forkProcess
but replace it with taking the eventlog buffer lock during the fork.
Fixes issue #27105
The original reason to block the timer during a fork was that
historically the timer was implemented using a periodic timer signal,
and the signal itself would interrupt the fork system call (returning
EINTR). For large processes (where fork() takes a while) this could
permanently livelock: the timer always would go off before the fork
could complete, which got retried in a loop forever.
The timer is no longer implemented as a unix signal, but uses threads.
Thus the original problem no longer exists. The only remaining reason to
block the timer tick is to prevent actions taken by the tick from
interfering with the delicate process involved in fork (taking a load of
locks and pausing everything).
The only thing we need to do is to prevent the eventlog from being
written to or flushed while the fork is taking place. To achieve this
all we need to do is hold the mutex for the global eventlog buffer.
This removes the last use of stopTimer that expects stopTimer to work
synchronously (which it was not) and thus solves issue #27105. To be
clear, we solve issue #27105 not by making stopTimer synchronous, but by
eliminating the use sites that expected it to be synchronous.
- - - - -
5a26a760 by Duncan Coutts at 2026-06-07T11:17:48+02:00
Add a test for thread scheduler fairness
It also tests that the interval timer and context switching works.
We also test that fairness is lost when the context switching interval
is too coarse for the duration of the test.
We add this test before doing surgery on the interval timer, so we have
decent coverage.
- - - - -
bcab4fff by Duncan Coutts at 2026-06-07T13:47:21+02:00
Improve and clarify ticker implementation(s)
We remove stop/startTicker and replace them with pause/unpauseTicker.
In the past the stop/startTicker actions were used incorrectly as if
they were synchronous, which they are not. See issue #27105. We now
document pause/unpackTicker as being async and not to be used for the
purpose of concurrency safety.
The existing stop/startTimer (note Timer not Ticker, the Timer calls the
Ticker!) are also exported from the RTS as a public API. This was
historically because the ticker used signals and it was important to
suspend the timer signel over a process fork. So these functions were
exported to be used by the process and unix libraries. This is no longer
necessary. Add a changelog entry to explain this.
We also take the opportunity to improve the ticker implementations
somewhat.
In the posix implementation, we take care to keep the tick action cheap.
Since pause/unpause are used very infrequently, we make them more
expensive and complicated to allow the normal hot path in the tick
action to be cheap. In particular we are able to eliminate the use of a
mutex in the tick action. We avoid locks and atomic memory ops in the
hot path. We use message passing via an eventfd or pipe.
In the win32 implementation, we continue to use the TimerQueue API, and
we make use of its ability to delete timers synchronously or
asynchronously. We improve the locking to ensure that it is safe to call
pause/unpause from any thread.
- - - - -
efa24f6a by Duncan Coutts at 2026-06-07T14:06:12+02:00
Make win32 ticker wait interval for initial tick too
There is no need to tick immediately. This is consistent with the
posix implementation.
- - - - -
55547c28 by Duncan Coutts at 2026-06-07T14:07:49+02:00
Remove now-unnecessary layer of RTS ticker enable/disable
There was an atomic variable used to block *part* of the actions of the
tick handler. This still did not make stopTimer synchronous, even for
the part of the the handle_tick actions it covered. It also added a more
expensive (sequentuially consistent) atomic operation in the hot path
for the handle_tick action, whereas our new design requires no atomic
ops at all.
Now that we have eliminate the need for synchronous stop/startTicker,
we don't need this not-quite-working-anyway atomic protocol. The new
pause/unpauseTicker is explicitly asynchronous and idempotent.
- - - - -
6a667450 by Duncan Coutts at 2026-06-07T14:12:09+02:00
Add TODOs about issue #27250: too much being done from handle_tick
The handle_tick should not perform I/O, block, perform long-running
operations or call arbitrary user code. Unfortunately, everything to
do with the eventlog (at the moment) falls into all those categories.
- - - - -
15 changed files:
- + changelog.d/T27105
- rts/Capability.c
- rts/RtsStartup.c
- rts/Schedule.c
- rts/Ticker.h
- rts/Timer.c
- rts/Timer.h
- rts/eventlog/EventLog.c
- rts/eventlog/EventLog.h
- rts/include/rts/OSThreads.h
- rts/include/rts/Timer.h
- rts/posix/Ticker.c
- rts/win32/Ticker.c
- + testsuite/tests/concurrent/should_run/T27105.hs
- testsuite/tests/concurrent/should_run/all.T
Changes:
=====================================
changelog.d/T27105
=====================================
@@ -0,0 +1,14 @@
+section: rts
+issues: #27105 #25165 #27335
+mrs: !16147 !16023
+synopsis: The RTS public APIs `stopTimer` and `startTimer` are now no-ops
+description: {
+ As a result of fixing some timer/ticker concurrency bugs, the exported RTS
+ APIs `stopTimer` and `startTimer` are now no-ops and are deprecated. They
+ were called at least by the process and unix libraries. No replacement is
+ needed.
+
+ They were used by libraries to temporarily block the RTS's use of the timer
+ signal. These functions no longer have a purpose since the RTS interval
+ timer no longer uses signals.
+}
=====================================
rts/Capability.c
=====================================
@@ -443,13 +443,6 @@ void
moreCapabilities (uint32_t from USED_IF_THREADS, uint32_t to USED_IF_THREADS)
{
#if defined(THREADED_RTS)
- // We must disable the timer while we do this since the tick handler may
- // call contextSwitchAllCapabilities, which may see the capabilities array
- // as we free it. The alternative would be to protect the capabilities
- // array with a lock but this seems more expensive than necessary.
- // See #17289.
- stopTimer();
-
if (to == 1) {
// THREADED_RTS must work on builds that don't have a mutable
// BaseReg (eg. unregisterised), so in this case
@@ -470,8 +463,6 @@ moreCapabilities (uint32_t from USED_IF_THREADS, uint32_t to USED_IF_THREADS)
}
debugTrace(DEBUG_sched, "allocated %d more capabilities", to - from);
-
- startTimer();
#endif
}
=====================================
rts/RtsStartup.c
=====================================
@@ -415,8 +415,8 @@ hs_init_ghc(int *argc, char **argv[], RtsConfig rts_config)
traceInitEvent(dumpIPEToEventLog);
initHeapProfiling();
- /* start the virtual timer 'subsystem'. */
- startTimer();
+ /* start the timer (after initTimer above) */
+ unpauseTimer();
#if defined(RTS_USER_SIGNALS)
if (RtsFlags.MiscFlags.install_signal_handlers) {
@@ -512,14 +512,12 @@ hs_exit_(bool wait_foreign)
}
#endif
- /* stop the ticker */
- stopTimer();
- /*
- * it is quite important that we wait here as some timer implementations
- * (e.g. pthread) may fire even after we exit, which may segfault as we've
- * already freed the capabilities.
+ /* We rely on the guarantee that exitTimer stops the timer synchronously,
+ * which ensures the timer handler does not get run again after this point.
+ * We are about to start freeing resources used by the timer handler (like
+ * the capabilities, eventlog and profiling data structures).
*/
- exitTimer(true);
+ exitTimer();
/*
* Dump the ticky counter definitions
=====================================
rts/Schedule.c
=====================================
@@ -37,6 +37,7 @@
#include "win32/AsyncWinIO.h"
#endif
#include "Trace.h"
+#include "eventlog/EventLog.h"
#include "RaiseAsync.h"
#include "Threads.h"
#include "Timer.h"
@@ -454,7 +455,7 @@ run_thread:
prev = setRecentActivity(ACTIVITY_YES);
if (prev == ACTIVITY_DONE_GC) {
#if !defined(PROFILING)
- startTimer();
+ unpauseTimer();
#endif
}
break;
@@ -1935,7 +1936,7 @@ delete_threads_and_gc:
// it will get re-enabled if we run any threads after the GC.
setRecentActivity(ACTIVITY_DONE_GC);
#if !defined(PROFILING)
- stopTimer();
+ pauseTimer();
#endif
break;
}
@@ -2100,24 +2101,31 @@ forkProcess(HsStablePtr *entry
ACQUIRE_LOCK(&all_tasks_mutex);
#endif
- stopTimer(); // See #4074
-
#if defined(TRACING)
- flushAllCapsEventsBufs(); // so that child won't inherit dirty file buffers
+#if defined(HAVE_PREEMPTION)
+ // We must hold the eventlog global mutex over the fork to prevent the
+ // timer thread from trying to post events. While holding the mutex we need
+ // to flush the eventlogs (global and per-cap) so that child won't inherit
+ // dirty eventlog buffers or file buffers.
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
+#endif
+ flushAllCapsEventsBufs_();
#endif
pid = fork();
if (pid) { // parent
- startTimer(); // #4074
-
RELEASE_LOCK(&sched_mutex);
RELEASE_LOCK(&sm_mutex);
RELEASE_LOCK(&stable_ptr_mutex);
RELEASE_LOCK(&stable_name_mutex);
RELEASE_LOCK(&task->lock);
+#if defined(TRACING) && defined(HAVE_PREEMPTION)
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
+#endif
+
#if defined(THREADED_RTS)
/* N.B. releaseCapability_ below may need to take all_tasks_mutex */
RELEASE_LOCK(&all_tasks_mutex);
@@ -2224,8 +2232,8 @@ forkProcess(HsStablePtr *entry
generations[g].threads = END_TSO_QUEUE;
}
- // On Unix, all timers are reset in the child, so we need to start
- // the timer again.
+ // The timer thread is not present in the child process, so we need
+ // to initialise the timer again.
initTimer();
// TODO: need to trace various other things in the child
@@ -2236,7 +2244,7 @@ forkProcess(HsStablePtr *entry
// start timer after the IOManager is initialized
// (the idle GC may wake up the IOManager)
- startTimer();
+ unpauseTimer();
// Install toplevel exception handlers, so interruption
// signal will be sent to the main thread.
@@ -2303,12 +2311,6 @@ setNumCapabilities (uint32_t new_n_capabilities USED_IF_THREADS)
cap = rts_lock();
task = cap->running_task;
-
- // N.B. We must stop the interval timer while we are changing the
- // capabilities array lest handle_tick may try to context switch
- // an old capability. See #17289.
- stopTimer();
-
stopAllCapabilities(&cap, task);
if (new_n_capabilities < enabled_capabilities)
@@ -2364,9 +2366,7 @@ setNumCapabilities (uint32_t new_n_capabilities USED_IF_THREADS)
tracingAddCapabilities(n_capabilities, new_n_capabilities);
#endif
- // Resize the capabilities array
- // NB. after this, capabilities points somewhere new. Any pointers
- // of type (Capability *) are now invalid.
+ // Allocate and initialise the extra capabilities
moreCapabilities(n_capabilities, new_n_capabilities);
// Resize and update storage manager data structures
@@ -2394,8 +2394,6 @@ setNumCapabilities (uint32_t new_n_capabilities USED_IF_THREADS)
// Notify IO manager that the number of capabilities has changed.
notifyIOManagerCapabilitiesChanged(&cap);
- startTimer();
-
rts_unlock(cap);
#endif // THREADED_RTS
=====================================
rts/Ticker.h
=====================================
@@ -12,9 +12,44 @@
typedef void (*TickProc)(int);
-void initTicker (Time interval, TickProc handle_tick);
-void startTicker (void);
-void stopTicker (void);
-void exitTicker (bool wait);
+/* The ticker is initialised in a paused state. Use unpauseTicker to start. */
+void initTicker(Time interval, TickProc handle_tick);
+
+/* Stop and terminate the ticker. It does not need to be stopped first.
+ * The exitTicker action is *synchronous*. When it returns the caller is
+ * guaranteed that the tick action is blocked.
+ */
+void exitTicker(void);
+
+/* Pause and unpause (resume) the ticker.
+ *
+ * The pauseTicker and unpauseTicker actions are *asynchronous*. After calling
+ * pauseTicker, the ticker will pause eventually, but there may be another tick
+ * action before it does pause (and theoretically there could be several but
+ * in practice this is unlikely). Similarly, after calling unpauseTicker the
+ * ticker will start up again eventually, but there is an unspecified delay
+ * between the unpause and the next tick action (but in practice it is short).
+ *
+ * This should be used for the purpose of *efficiency*: to avoid unnecessary
+ * OS thread wakeups caused by the ticker.
+ *
+ * These should *not* be used for the purpose of *concurrency safety*: to
+ * prevent the tick action from running concurrently with some other critical
+ * section. The synchronous case is not provided because it is not currently
+ * needed (and proper locking is often a better solution anyway).
+ *
+ * The pairing of unpauseTicker and the handle_tick action form a
+ * synchonises-with relation: values written before unpauseTicker can be
+ * read from the resulting handle_tick action.
+ *
+ * It *is* safe to call these functions from within the tick handler itself.
+ *
+ * It is safe to use these functions concurrently from multiple threads, but
+ * note that they *are* idempotent. This means it is not appropriate to use
+ * paired pause/unpause calls concurrently. They can be used by threads based
+ * on consistent use of some shared state or observation.
+ */
+void pauseTicker(void);
+void unpauseTicker(void);
#include "EndPrivate.h"
=====================================
rts/Timer.c
=====================================
@@ -28,20 +28,6 @@
#include "RtsSignals.h"
#include "rts/EventLogWriter.h"
-// See Note [No timer on wasm32]
-#if !defined(wasm32_HOST_ARCH)
-#define HAVE_PREEMPTION
-#endif
-
-// This global counter is used to allow multiple threads to stop the
-// timer temporarily with a stopTimer()/startTimer() pair. If
-// timer_enabled == 0 timer is enabled
-// timer_disabled == N, N > 0 timer is disabled by N threads
-// When timer_enabled makes a transition to 0, we enable the timer,
-// and when it makes a transition to non-0 we disable it.
-
-static StgWord timer_disabled;
-
/* ticks left before next pre-emptive context switch */
static int ticks_to_ctxt_switch = 0;
@@ -112,9 +98,9 @@ static
void
handle_tick(int unused STG_UNUSED)
{
- handleProfTick();
- if (RtsFlags.ConcFlags.ctxtSwitchTicks > 0
- && SEQ_CST_LOAD_ALWAYS(&timer_disabled) == 0)
+ handleProfTick(); // Bad or worse: see issue #27250.
+
+ if (RtsFlags.ConcFlags.ctxtSwitchTicks > 0)
{
ticks_to_ctxt_switch--;
if (ticks_to_ctxt_switch <= 0) {
@@ -128,7 +114,7 @@ handle_tick(int unused STG_UNUSED)
ticks_to_eventlog_flush--;
if (ticks_to_eventlog_flush <= 0) {
ticks_to_eventlog_flush = RtsFlags.TraceFlags.eventlogFlushTicks;
- flushEventLog(NULL);
+ flushEventLog(NULL); // Bad or worse: see issue #27250.
}
}
#endif
@@ -153,7 +139,7 @@ handle_tick(int unused STG_UNUSED)
RtsFlags.MiscFlags.tickInterval;
#if defined(THREADED_RTS)
wakeUpRts();
- // The scheduler will call stopTimer() when it has done
+ // The scheduler will call pauseTimer() when it has done
// the GC.
#endif
} else {
@@ -165,10 +151,10 @@ handle_tick(int unused STG_UNUSED)
#if defined(PROFILING)
if (!(RtsFlags.ProfFlags.doHeapProfile
|| RtsFlags.CcFlags.doCostCentres)) {
- stopTimer();
+ pauseTimer();
}
#else
- stopTimer();
+ pauseTimer();
#endif
}
} else {
@@ -181,48 +167,49 @@ handle_tick(int unused STG_UNUSED)
}
}
-void
-initTimer(void)
+void initTimer(void)
{
#if defined(HAVE_PREEMPTION)
initProfTimer();
if (RtsFlags.MiscFlags.tickInterval != 0) {
initTicker(RtsFlags.MiscFlags.tickInterval, handle_tick);
}
- SEQ_CST_STORE_ALWAYS(&timer_disabled, 1);
#endif
}
-void
-startTimer(void)
+/* Deprecated exported functions. Now no-ops.
+ * Historically they were used by the process and unix libraries to disable
+ * the signal-based interval timer, since otherwise the timer signal would
+ * keep going off in the child process and confusing everything. The interval
+ * timer no longer uses signals, so there is no need any more for libraries to
+ * disable the timer. Also, the timer internal API has changed.
+ */
+void stopTimer(void) { /* no-op */ }
+void startTimer(void) { /* no-op */ }
+
+void pauseTimer(void)
{
#if defined(HAVE_PREEMPTION)
- if (SEQ_CST_SUB_ALWAYS(&timer_disabled, 1) == 0) {
- if (RtsFlags.MiscFlags.tickInterval != 0) {
- startTicker();
- }
+ if (RtsFlags.MiscFlags.tickInterval != 0) {
+ pauseTicker();
}
#endif
}
-void
-stopTimer(void)
+void unpauseTimer(void)
{
#if defined(HAVE_PREEMPTION)
- if (SEQ_CST_ADD_ALWAYS(&timer_disabled, 1) == 1) {
- if (RtsFlags.MiscFlags.tickInterval != 0) {
- stopTicker();
- }
+ if (RtsFlags.MiscFlags.tickInterval != 0) {
+ unpauseTicker();
}
#endif
}
-void
-exitTimer (bool wait)
+void exitTimer (void)
{
#if defined(HAVE_PREEMPTION)
if (RtsFlags.MiscFlags.tickInterval != 0) {
- exitTicker(wait);
+ exitTicker();
}
#endif
}
=====================================
rts/Timer.h
=====================================
@@ -8,5 +8,12 @@
#pragma once
-RTS_PRIVATE void initTimer (void);
-RTS_PRIVATE void exitTimer (bool wait);
+#include "BeginPrivate.h"
+
+void initTimer(void);
+void exitTimer(void);
+
+void pauseTimer(void);
+void unpauseTimer(void);
+
+#include "EndPrivate.h"
=====================================
rts/eventlog/EventLog.c
=====================================
@@ -129,8 +129,11 @@ typedef struct _EventsBuf {
static EventsBuf *capEventBuf; // one EventsBuf for each Capability
static EventsBuf eventBuf; // an EventsBuf not associated with any Capability
-#if defined(THREADED_RTS)
-static Mutex eventBufMutex; // protected by this mutex
+#if defined(HAVE_PREEMPTION)
+// Note that this mutex is used even in the non-threaded RTS, since the timer
+// thread posts events and flushes. So _all_ uses of this mutex must use
+// ACQUIRE_LOCK_ALWAYS/RELEASE_LOCK_ALWAYS.
+Mutex eventBufMutex; // protects eventBuf above
#endif
// Event type
@@ -393,8 +396,10 @@ initEventLogging(void)
moreCapEventBufs(0, get_n_capabilities());
initEventsBuf(&eventBuf, EVENT_LOG_SIZE, (EventCapNo)(-1));
-#if defined(THREADED_RTS)
+#if defined(HAVE_PREEMPTION)
initMutex(&eventBufMutex);
+#endif
+#if defined(THREADED_RTS)
initMutex(&state_change_mutex);
#endif
}
@@ -416,7 +421,7 @@ startEventLogging_(void)
{
initEventLogWriter();
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
postHeaderEvents();
/*
@@ -425,7 +430,7 @@ startEventLogging_(void)
*/
printAndClearEventBuf(&eventBuf);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
return true;
}
@@ -495,7 +500,7 @@ endEventLogging(void)
flushEventLog_(NULL);
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
// Mark end of events (data).
postEventTypeNum(&eventBuf, EVENT_DATA_END);
@@ -503,7 +508,7 @@ endEventLogging(void)
// Flush the end of data marker.
printAndClearEventBuf(&eventBuf);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
stopEventLogWriter();
event_log_writer = NULL;
@@ -666,7 +671,7 @@ void
postCapEvent (EventTypeNum tag,
EventCapNo capno)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, tag);
postEventHeader(&eventBuf, tag);
@@ -685,14 +690,14 @@ postCapEvent (EventTypeNum tag,
barf("postCapEvent: unknown event tag %d", tag);
}
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postCapsetEvent (EventTypeNum tag,
EventCapsetID capset,
StgWord info)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, tag);
postEventHeader(&eventBuf, tag);
@@ -726,7 +731,7 @@ void postCapsetEvent (EventTypeNum tag,
barf("postCapsetEvent: unknown event tag %d", tag);
}
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postCapsetStrEvent (EventTypeNum tag,
@@ -740,14 +745,14 @@ void postCapsetStrEvent (EventTypeNum tag,
return;
}
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
if (!hasRoomForVariableEvent(&eventBuf, size)){
printAndClearEventBuf(&eventBuf);
if (!hasRoomForVariableEvent(&eventBuf, size)){
errorBelch("Event size exceeds buffer size, bail out");
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
return;
}
}
@@ -758,7 +763,7 @@ void postCapsetStrEvent (EventTypeNum tag,
postBuf(&eventBuf, (StgWord8*) msg, strsize);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postCapsetVecEvent (EventTypeNum tag,
@@ -783,14 +788,14 @@ void postCapsetVecEvent (EventTypeNum tag,
}
}
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
if (!hasRoomForVariableEvent(&eventBuf, size)){
printAndClearEventBuf(&eventBuf);
if(!hasRoomForVariableEvent(&eventBuf, size)){
errorBelch("Event size exceeds buffer size, bail out");
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
return;
}
}
@@ -804,7 +809,7 @@ void postCapsetVecEvent (EventTypeNum tag,
postBuf(&eventBuf, (StgWord8*) argv[i], 1 + strlen(argv[i]));
}
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postWallClockTime (EventCapsetID capset)
@@ -813,7 +818,7 @@ void postWallClockTime (EventCapsetID capset)
StgWord64 sec;
StgWord32 nsec;
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
/* The EVENT_WALL_CLOCK_TIME event is intended to allow programs
reading the eventlog to match up the event timestamps with wall
@@ -846,7 +851,7 @@ void postWallClockTime (EventCapsetID capset)
postWord64(&eventBuf, sec);
postWord32(&eventBuf, nsec);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
/*
@@ -885,7 +890,7 @@ void postEventHeapInfo (EventCapsetID heap_capset,
W_ mblockSize,
W_ blockSize)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_HEAP_INFO_GHC);
postEventHeader(&eventBuf, EVENT_HEAP_INFO_GHC);
@@ -899,7 +904,7 @@ void postEventHeapInfo (EventCapsetID heap_capset,
postWord64(&eventBuf, mblockSize);
postWord64(&eventBuf, blockSize);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postEventGcStats (Capability *cap,
@@ -952,7 +957,7 @@ void postTaskCreateEvent (EventTaskId taskId,
EventCapNo capno,
EventKernelThreadId tid)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_TASK_CREATE);
postEventHeader(&eventBuf, EVENT_TASK_CREATE);
@@ -961,14 +966,14 @@ void postTaskCreateEvent (EventTaskId taskId,
postCapNo(&eventBuf, capno);
postKernelThreadId(&eventBuf, tid);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postTaskMigrateEvent (EventTaskId taskId,
EventCapNo capno,
EventCapNo new_capno)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_TASK_MIGRATE);
postEventHeader(&eventBuf, EVENT_TASK_MIGRATE);
@@ -977,28 +982,28 @@ void postTaskMigrateEvent (EventTaskId taskId,
postCapNo(&eventBuf, capno);
postCapNo(&eventBuf, new_capno);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postTaskDeleteEvent (EventTaskId taskId)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_TASK_DELETE);
postEventHeader(&eventBuf, EVENT_TASK_DELETE);
/* EVENT_TASK_DELETE (taskID) */
postTaskId(&eventBuf, taskId);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void
postEventNoCap (EventTypeNum tag)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, tag);
postEventHeader(&eventBuf, tag);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void
@@ -1042,9 +1047,9 @@ void postLogMsg(EventsBuf *eb, EventTypeNum type, char *msg, va_list ap)
void postMsg(char *msg, va_list ap)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
postLogMsg(&eventBuf, EVENT_LOG_MSG, msg, ap);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postCapMsg(Capability *cap, char *msg, va_list ap)
@@ -1138,32 +1143,32 @@ void postConcUpdRemSetFlush(Capability *cap)
void postConcMarkEnd(StgWord32 marked_obj_count)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_CONC_MARK_END);
postEventHeader(&eventBuf, EVENT_CONC_MARK_END);
postWord32(&eventBuf, marked_obj_count);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postNonmovingHeapCensus(uint16_t blk_size,
const struct NonmovingAllocCensus *census)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
postEventHeader(&eventBuf, EVENT_NONMOVING_HEAP_CENSUS);
postWord16(&eventBuf, blk_size);
postWord32(&eventBuf, census->n_active_segs);
postWord32(&eventBuf, census->n_filled_segs);
postWord32(&eventBuf, census->n_live_blocks);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postNonmovingPrunedSegments(uint32_t pruned_segments, uint32_t free_segments)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
postEventHeader(&eventBuf, EVENT_NONMOVING_PRUNED_SEGMENTS);
postWord32(&eventBuf, pruned_segments);
postWord32(&eventBuf, free_segments);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void closeBlockMarker (EventsBuf *ebuf)
@@ -1224,7 +1229,7 @@ static HeapProfBreakdown getHeapProfBreakdown(void)
void postHeapProfBegin(void)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
PROFILING_FLAGS *flags = &RtsFlags.ProfFlags;
StgWord modSelector_len =
flags->modSelector ? strlen(flags->modSelector) : 0;
@@ -1258,42 +1263,42 @@ void postHeapProfBegin(void)
postStringLen(&eventBuf, flags->ccsSelector, ccsSelector_len);
postStringLen(&eventBuf, flags->retainerSelector, retainerSelector_len);
postStringLen(&eventBuf, flags->bioSelector, bioSelector_len);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postHeapProfSampleBegin(StgInt era)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_HEAP_PROF_SAMPLE_BEGIN);
postEventHeader(&eventBuf, EVENT_HEAP_PROF_SAMPLE_BEGIN);
postWord64(&eventBuf, era);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postHeapBioProfSampleBegin(StgInt era, StgWord64 time)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_HEAP_BIO_PROF_SAMPLE_BEGIN);
postEventHeader(&eventBuf, EVENT_HEAP_BIO_PROF_SAMPLE_BEGIN);
postWord64(&eventBuf, era);
postWord64(&eventBuf, time);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postHeapProfSampleEnd(StgInt era)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_HEAP_PROF_SAMPLE_END);
postEventHeader(&eventBuf, EVENT_HEAP_PROF_SAMPLE_END);
postWord64(&eventBuf, era);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postHeapProfSampleString(const char *label,
StgWord64 residency)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
StgWord label_len = strlen(label);
StgWord len = 1+8+label_len+1;
CHECK(!ensureRoomForVariableEvent(&eventBuf, len));
@@ -1303,7 +1308,7 @@ void postHeapProfSampleString(const char *label,
postWord8(&eventBuf, 0);
postWord64(&eventBuf, residency);
postStringLen(&eventBuf, label, label_len);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
#if defined(PROFILING)
@@ -1313,7 +1318,7 @@ void postHeapProfCostCentre(StgWord32 ccID,
const char *srcloc,
StgBool is_caf)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
StgWord label_len = strlen(label);
StgWord module_len = strlen(module);
StgWord srcloc_len = strlen(srcloc);
@@ -1326,13 +1331,13 @@ void postHeapProfCostCentre(StgWord32 ccID,
postStringLen(&eventBuf, module, module_len);
postStringLen(&eventBuf, srcloc, srcloc_len);
postWord8(&eventBuf, is_caf);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void postHeapProfSampleCostCentre(CostCentreStack *stack,
StgWord64 residency)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
StgWord depth = 0;
CostCentreStack *ccs;
for (ccs = stack; ccs != NULL && ccs != CCS_MAIN; ccs = ccs->prevStack)
@@ -1351,7 +1356,7 @@ void postHeapProfSampleCostCentre(CostCentreStack *stack,
depth>0 && ccs != NULL && ccs != CCS_MAIN;
ccs = ccs->prevStack, depth--)
postWord32(&eventBuf, ccs->cc->ccID);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
@@ -1359,7 +1364,7 @@ void postProfSampleCostCentre(Capability *cap,
CostCentreStack *stack,
StgWord64 tick)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
StgWord depth = 0;
CostCentreStack *ccs;
for (ccs = stack; ccs != NULL && ccs != CCS_MAIN; ccs = ccs->prevStack)
@@ -1377,7 +1382,7 @@ void postProfSampleCostCentre(Capability *cap,
depth>0 && ccs != NULL && ccs != CCS_MAIN;
ccs = ccs->prevStack, depth--)
postWord32(&eventBuf, ccs->cc->ccID);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
// This event is output at the start of profiling so the tick interval can
@@ -1385,11 +1390,11 @@ void postProfSampleCostCentre(Capability *cap,
// can be calculated from how many samples there are.
void postProfBegin(void)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
postEventHeader(&eventBuf, EVENT_PROF_BEGIN);
// The interval that each tick was sampled, in nanoseconds
postWord64(&eventBuf, TimeToNS(RtsFlags.MiscFlags.tickInterval));
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
#endif /* PROFILING */
@@ -1415,11 +1420,11 @@ static void postTickyCounterDef(EventsBuf *eb, StgEntCounter *p)
void postTickyCounterDefs(StgEntCounter *counters)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
for (StgEntCounter *p = counters; p != NULL; p = p->link) {
postTickyCounterDef(&eventBuf, p);
}
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
static void postTickyCounterSample(EventsBuf *eb, StgEntCounter *p)
@@ -1443,13 +1448,13 @@ static void postTickyCounterSample(EventsBuf *eb, StgEntCounter *p)
void postTickyCounterSamples(StgEntCounter *counters)
{
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
ensureRoomForEvent(&eventBuf, EVENT_TICKY_COUNTER_SAMPLE);
postEventHeader(&eventBuf, EVENT_TICKY_COUNTER_BEGIN_SAMPLE);
for (StgEntCounter *p = counters; p != NULL; p = p->link) {
postTickyCounterSample(&eventBuf, p);
}
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
#endif /* TICKY_TICKY */
void postIPE(const InfoProvEnt *ipe)
@@ -1459,7 +1464,7 @@ void postIPE(const InfoProvEnt *ipe)
// See Note [Maximum event length].
const StgWord MAX_IPE_STRING_LEN = 65535;
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
StgWord table_name_len = MIN(strlen(ipe->prov.table_name), MAX_IPE_STRING_LEN);
StgWord closure_desc_len = MIN(strlen(closure_desc_buf), MAX_IPE_STRING_LEN);
StgWord ty_desc_len = MIN(strlen(ipe->prov.ty_desc), MAX_IPE_STRING_LEN);
@@ -1489,7 +1494,7 @@ void postIPE(const InfoProvEnt *ipe)
postBuf(&eventBuf, &colon, 1);
postStringLen(&eventBuf, ipe->prov.src_span, src_span_len);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
}
void printAndClearEventBuf (EventsBuf *ebuf)
@@ -1601,14 +1606,21 @@ void flushLocalEventsBuf(Capability *cap)
// Flush all capabilities' event buffers when we already hold all capabilities.
// Used during forkProcess.
void flushAllCapsEventsBufs(void)
+{
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
+ flushAllCapsEventsBufs_();
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
+}
+
+// Unafe version that does not acquire/release eventBufMutex. You must
+// already hold the eventBufMutex, which you must do with ACQUIRE_LOCK_ALWAYS!
+void flushAllCapsEventsBufs_(void)
{
if (!event_log_writer) {
return;
}
- ACQUIRE_LOCK(&eventBufMutex);
printAndClearEventBuf(&eventBuf);
- RELEASE_LOCK(&eventBufMutex);
for (unsigned int i=0; i < getNumCapabilities(); i++) {
flushLocalEventsBuf(getCapability(i));
@@ -1641,9 +1653,9 @@ static void flushEventLog_(Capability **cap USED_IF_THREADS)
return;
}
- ACQUIRE_LOCK(&eventBufMutex);
+ ACQUIRE_LOCK_ALWAYS(&eventBufMutex);
printAndClearEventBuf(&eventBuf);
- RELEASE_LOCK(&eventBufMutex);
+ RELEASE_LOCK_ALWAYS(&eventBufMutex);
#if defined(THREADED_RTS)
Task *task = newBoundTask();
=====================================
rts/eventlog/EventLog.h
=====================================
@@ -18,6 +18,13 @@
#if defined(TRACING)
extern bool eventlog_enabled;
+#if defined(HAVE_PREEMPTION)
+// Avoid using this mutex directly if at all possible. It is needed in the
+// implementation of forkProcess.
+//
+// All uses of this mutex must use ACQUIRE_LOCK_ALWAYS/RELEASE_LOCK_ALWAYS.
+extern Mutex eventBufMutex;
+#endif
void initEventLogging(void);
void restartEventLogging(void);
@@ -27,6 +34,7 @@ void abortEventLogging(void); // #4512 - after fork child needs to abort
void moreCapEventBufs (uint32_t from, uint32_t to);
void flushLocalEventsBuf(Capability *cap);
void flushAllCapsEventsBufs(void);
+void flushAllCapsEventsBufs_(void);
void flushAllEventsBufs(Capability *cap);
typedef void (*EventlogInitPost)(void);
=====================================
rts/include/rts/OSThreads.h
=====================================
@@ -14,6 +14,40 @@
#pragma once
+/* Note [Threads and preemption]
+ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ All full-fat OSs that GHC works on have OS threads, and we use them even in
+ the non-threaded RTS for a few features:
+ * Haskell thread preemption;
+ * sample-based profiling;
+ * idle GC;
+ * periodic eventlog flushing.
+
+ We use defined(HAVE_PREEMPTION) to decide if these features are implemented
+ via OS threads.
+
+ On platforms like WASM/js we do not have OS threads in any conventional
+ sense, and the features above are either not available or are implemented
+ differently. See Note [No timer on wasm32].
+
+ In future if GHC is ported to platforms like bare-metal micro-controllers,
+ RTOSs or to run directly under hypervisors then such platforms may also not
+ have threads available and they should not define HAVE_PREEMPTION here. Or
+ for some micro-controller RTOSs like Zeypher one may have a choice about
+ whether to use threads or not (at a size cost). Here would be the right
+ place to control whether the feature list above is supported.
+ */
+#if defined(wasm32_HOST_ARCH)
+ // See Note [No timer on wasm32]
+ // To confuse matters, WASM _does_ have pthread.h but it doesnt work.
+#elif defined(HAVE_PTHREAD_H) || defined(HAVE_WINDOWS_H)
+#define HAVE_PREEMPTION
+#else
+#error Decide if this platform has threads and pre-emption or not.
+#endif
+// And JS does all of this differently, without using this bit of the RTS.
+
+
#if defined(HAVE_PTHREAD_H) && !defined(mingw32_HOST_OS)
#if defined(CMINUSMINUS)
@@ -210,9 +244,29 @@ extern bool timedWaitCondition ( Condition* pCond, Mutex* pMut, Time timeout)
//
// Mutexes
//
+// Even in the non-threaded RTS we use threads and mutexes! In particular the
+// timer/ticker is implemented using a thread. And using threads needs locks.
+// In particular we need locks for the data shared between the timer/ticker
+// thread and the thread running the main capability.
+#if defined(HAVE_PREEMPTION)
extern void initMutex ( Mutex* pMut );
extern void closeMutex ( Mutex* pMut );
+// The "always" variants do locking in the threaded and non-threaded RTS.
+// The normal variants below are no-ops in the non-threaded RTS.
+#define ACQUIRE_LOCK_ALWAYS(l) OS_ACQUIRE_LOCK(l)
+#define TRY_ACQUIRE_LOCK_ALWAYS(l) OS_TRY_ACQUIRE_LOCK(l)
+#define RELEASE_LOCK_ALWAYS(l) OS_RELEASE_LOCK(l)
+#define ASSERT_LOCK_HELD_ALWAYS(l) OS_ASSERT_LOCK_HELD(l)
+#else
+// And just to be a bit confusing, the always variants are still no-ops when we
+// do not HAVE_PREEMPTION, since then we don't have threads or mutexes at all.
+#define ACQUIRE_LOCK_ALWAYS(l)
+#define TRY_ACQUIRE_LOCK_ALWAYS(l) 0
+#define RELEASE_LOCK_ALWAYS(l)
+#define ASSERT_LOCK_HELD_ALWAYS(l)
+#endif
+
// Processors and affinity
void setThreadAffinity (uint32_t n, uint32_t m);
void setThreadNode (uint32_t node);
@@ -228,6 +282,7 @@ void releaseThreadNode (void);
#else
+// No-ops in the non-threaded RTS. See also the _ALWAYS variants above.
#define ACQUIRE_LOCK(l)
#define TRY_ACQUIRE_LOCK(l) 0
#define RELEASE_LOCK(l)
=====================================
rts/include/rts/Timer.h
=====================================
@@ -13,6 +13,6 @@
#pragma once
-void startTimer (void);
-void stopTimer (void);
+void startTimer (void); // Deprecated: see issue #27086
+void stopTimer (void); // Deprecated: see issue #27086
int rtsTimerSignal (void); // Deprecated: see issue #27073
=====================================
rts/posix/Ticker.c
=====================================
@@ -103,120 +103,139 @@
#include