Add QutexAcquisitionHistoryTracker; integrate plumbing

We add the new Qutex acquisision history tracker that allows us
to dynamically detect qutex gridlocks. We've integrated it into
LockerAndInvoker::operator() in a preliminary way.

We also moved all of the trace*ForGridlockOn() methods into the
new QutexAcquisitionHistoryTracker singleton class. They're
more appropriately located there. They're still unimplemented
though.
This commit is contained in:
2025-09-29 19:27:02 -04:00
parent 8123ec1227
commit 71564b4d83
6 changed files with 325 additions and 106 deletions
+54 -6
View File
@@ -12,6 +12,7 @@
#include <asynchronousContinuation.h>
#include <lockerAndInvokerBase.h>
#include <callback.h>
#include <qutexAcquisitionHistoryTracker.h>
namespace smo {
@@ -36,6 +37,10 @@ public:
std::forward<Args>(args)...);
}
// Return list of all qutexes in predecessors' LockSets; excludes self.
std::unique_ptr<std::forward_list<std::reference_wrapper<Qutex>>>
getAcquiredQutexHistory() const;
public:
LockSet<OriginalCbFnT> requiredLocks;
std::atomic<bool> isAwakeOrBeingAwakened{false};
@@ -116,8 +121,33 @@ public:
bool isDeadlock = traceContinuationHistoryForDeadlockOn(
firstFailedQutex);
bool isGridlock = heuristicallyTraceContinuationHistoryForGridlockOn(
firstFailedQutex);
bool gridlockIsHeuristicallyLikely = false;
bool gridlockIsAlgorithmicallyLikely = false;
if (gridlockLikely)
{
auto tracker = QutexAcquisitionHistoryTracker
::getInstance();
auto heldLocks = serializedContinuation
.getAcquiredQutexHistory();
// Add this continuation to the tracker
tracker.addIfNotExists(
serializedContinuation.shared_from_this(),
firstFailedQutex, std::move(heldLocks));
gridlockIsHeuristicallyLikely = tracker
.heuristicallyTraceContinuationHistoryForGridlockOn(
firstFailedQutex);
gridlockIsAlgorithmicallyLikely = tracker
.completelyTraceContinuationHistoryForGridlockOn(
firstFailedQutex);
}
bool isGridlock = (gridlockIsHeuristicallyLikely
|| gridlockIsAlgorithmicallyLikely);
if (!isDeadlock && !isGridlock)
{ return; }
@@ -145,6 +175,28 @@ public:
* can't acquire the locks anyway.
*/
serializedContinuation.requiredLocks.unregisterFromQutexQueues();
#ifdef CONFIG_ENABLE_DEBUG_LOCKS
/** EXPLANATION:
* If we were being tracked for gridlock detection but successfully
* acquired all locks, it was a false positive due to timed delay,
* long-running operation, or I/O delay
*/
if (gridlockLikely)
{
bool removed = QutexAcquisitionHistoryTracker::getInstance()
.remove(serializedContinuation.shared_from_this());
if (removed)
{
std::cerr << "LockerAndInvoker::operator(): False positive gridlock "
"detection - continuation was being tracked but successfully "
"acquired all locks. This was likely due to timed delay, "
"long-running operation, or I/O delay." << std::endl;
}
}
#endif
invocationTarget();
}
@@ -227,10 +279,6 @@ public:
};
bool traceContinuationHistoryForDeadlockOn(Qutex &firstFailedQutex);
bool heuristicallyTraceContinuationHistoryForGridlockOn(
Qutex &firstFailedQutex);
bool completelyTraceContinuationHistoryForGridlockOn(
Qutex &firstFailedQutex);
bool traceContinuationHistoryForDeadlock(void)
{
for (auto& lockUsageDesc