Debug:Qutex: Add deadlock detection based on elapsed time

We now detect that a deadlock is likely when
CONFIG_DEBUG_QUTEX_DEADLOCK_TIMEOUT_MS has elapsed. This is the
preliminary work required to do a backtrace through the call
stack and figure out if a deadlock has really occured.

To do this, we'd have to go through the async call chain and
search for a previous caller which acquired the same qutex as
the one that first failed during this Lockvoker LockSet acquisition
attempt.

CMakeLists.txt

@@ -26,6 +26,14 @@ math(EXPR MIND_VOSCILLATOR_FREQ_MS "1000 / ${MIND_VOSCILLATOR_PERIOD_MS}")
 # World thread configuration
 option(WORLD_USE_BODY_THREAD
 	"Use body thread for world component instead of separate world thread" OFF)
+
+# Qutex deadlock detection configuration
+set(DEBUG_CONFIG_QUTEX_DEADLOCK_TIMEOUT_MS 500 CACHE STRING
+	"Timeout in milliseconds for deadlock detection in qutex system")
+if(NOT DEBUG_CONFIG_QUTEX_DEADLOCK_TIMEOUT_MS GREATER 0)
+    message(FATAL_ERROR "DEBUG_CONFIG_QUTEX_DEADLOCK_TIMEOUT_MS must be a positive integer > 0")
+endif()
+
 # Test configuration
 option(ENABLE_TESTS "Enable building tests" OFF)
 

include/config.h.in

@@ -12,6 +12,9 @@
 /* World thread configuration */
 #cmakedefine WORLD_USE_BODY_THREAD
 
+/* Qutex deadlock detection configuration */
+#define DEBUG_CONFIG_QUTEX_DEADLOCK_TIMEOUT_MS @DEBUG_CONFIG_QUTEX_DEADLOCK_TIMEOUT_MS@
+
 /* Cross-compilation configuration */
 #cmakedefine CMAKE_CROSSCOMPILING
 

include/lockSet.h

@@ -81,9 +81,12 @@ public:
 	/**
 	 * @brief Try to acquire all locks in order; back off if acquisition fails
 	 * @param lockvoker The LockerAndInvoker attempting to acquire the locks
+	 * @param firstFailedQutex Output parameter to receive the first Qutex that
+	 * 	failed acquisition (can be nullptr)
 	 * @return true if all locks were acquired, false otherwise
 	 */
-	bool tryAcquireOrBackOff(LockerAndInvokerBase &lockvoker);
+	bool tryAcquireOrBackOff(
+		LockerAndInvokerBase &lockvoker, Qutex *firstFailedQutex = nullptr);
 	void unregisterFromQutexQueues();
 
 	/**

include/qutex.h

@@ -24,8 +24,8 @@ public:
 	/**
 	 * @brief Constructor
 	 */
-	Qutex()
-	: isOwned(false)
+	Qutex(const std::string &_name)
+	: isOwned(false), name(_name)
 	{}
 
 	/**
@@ -88,6 +88,7 @@ public:
 	SpinLock lock;
 	LockerAndInvokerBase::List queue;
 	bool isOwned;
+	std::string name;
 };
 
 } // namespace smo

include/serializedAsynchronousContinuation.h

@@ -4,6 +4,8 @@
 #include <functional>
 #include <memory>
 #include <atomic>
+#include <chrono>
+#include <iostream>
 #include <componentThread.h>
 #include <lockSet.h>
 #include <asynchronousContinuation.h>
@@ -63,7 +65,8 @@ public:
 		:	LockerAndInvokerBase(&serializedContinuation),
 		serializedContinuation(serializedContinuation),
 		target(target),
-		invocationTarget(std::move(invocationTarget))
+		invocationTarget(std::move(invocationTarget)),
+		creationTimestamp(std::chrono::steady_clock::now())
 		{
 			firstWake();
 		}
@@ -81,11 +84,28 @@ public:
 					"executing on wrong ComponentThread");
 			}
 
+			Qutex *firstFailedQutexPtr = nullptr;
+			bool isDeadlockLikely = isDeadlockLikely();
+
 			if (!serializedContinuation.requiredLocks.tryAcquireOrBackOff(
-				*this))
+				*this, (isDeadlockLikely ? &firstFailedQutexPtr : nullptr)))
 			{
 				// Just allow this lockvoker to be dropped from its io_service.
 				allowAwakening();
+
+				if (!isDeadlockLikely)
+					{ return; }
+
+				Qutex &firstFailedQutex = *firstFailedQutexPtr;
+
+				std::cerr << __func__ << ": Deadlock likely: "
+					<< "Lockvoker has been waiting for "
+					<< std::chrono::duration_cast<std::chrono::milliseconds>(
+						std::chrono::steady_clock::now() - creationTimestamp)
+						.count()
+					<< "ms, failed on qutex @" << &firstFailedQutex
+					<< " (" << firstFailedQutex.name << ")" << std::endl;
+
 				return;
 			}
 
@@ -137,9 +157,8 @@ public:
 			target->getIoService().post(*this);
 		}
 
-		/**
-		 * @brief Allow awakening by resetting the awake flag
-		 */
+	private:
+		// Allow awakening by resetting the awake flag
 		void allowAwakening()
 			{ serializedContinuation.isAwakeOrBeingAwakened.store(false); }
 
@@ -158,11 +177,21 @@ public:
 			awaken(true);
 		}
 
+		// Check if CONFIG_QUTEX_DEADLOCK_TIMEOUT_MS has elapsed since creation
+		bool isDeadlockLikely() const
+		{
+			auto now = std::chrono::steady_clock::now();
+			auto elapsed = std::chrono::duration_cast<std::chrono::milliseconds>(
+				now - creationTimestamp);
+			return elapsed.count() >= DEBUG_CONFIG_QUTEX_DEADLOCK_TIMEOUT_MS;
+		}
+
 	private:
 		SerializedAsynchronousContinuation<OriginalCbFnT>
 			&serializedContinuation;
 		InvocationTargetT invocationTarget;
 		std::shared_ptr<ComponentThread> target;
+		std::chrono::steady_clock::time_point creationTimestamp;
 	};
 };
 

smocore/lockSet.cpp

@@ -47,7 +47,9 @@ void LockSet<OriginalCbFnT>::registerInQutexQueues(
 }
 
 template <class OriginalCbFnT>
-bool LockSet<OriginalCbFnT>::tryAcquireOrBackOff(LockerAndInvokerBase &lockvoker)
+bool LockSet<OriginalCbFnT>::tryAcquireOrBackOff(
+	LockerAndInvokerBase &lockvoker, Qutex *firstFailedQutex
+	)
 {
 	if (!registeredInQutexQueues)
 	{
@@ -72,6 +74,10 @@ bool LockSet<OriginalCbFnT>::tryAcquireOrBackOff(LockerAndInvokerBase &lockvoker
 		if (!lockUsageDesc.first.get().tryAcquire(
 			lockvoker, nRequiredLocks))
 		{
+			// Set the first failed qutex for debugging
+			if (firstFailedQutex) {
+				*firstFailedQutex = &lockUsageDesc.first.get();
+			}
 			break;
 		}