kernel: add syscalls for thread affinity

scheduler modes adapted to take into account the thread affinity mask.
when no cpu in the affinity mask is enabled, the mask is ignored.

Change-Id: I577737441ab073941a4c5e06f94f7825cffdc2c4
Reviewed-on: https://review.haiku-os.org/c/haiku/+/7674
Reviewed-by: waddlesplash <[email protected]>
Tested-by: Commit checker robot <[email protected]>

headers/private/kernel/cpu.h

@@ -90,6 +90,7 @@ typedef struct CACHE_LINE_ALIGN cpu_ent {
 
 extern cpu_ent gCPU[];
 extern uint32 gCPUCacheLevelCount;
+extern CPUSet gCPUEnabled;
 
 
 #ifdef __cplusplus

headers/private/kernel/thread.h

@@ -160,6 +160,9 @@ thread_id _user_find_thread(const char *name);
 status_t _user_get_thread_info(thread_id id, thread_info *info);
 status_t _user_get_next_thread_info(team_id team, int32 *cookie, thread_info *info);
 int _user_get_cpu();
+status_t _user_get_thread_affinity(thread_id id, void* userMask, size_t size);
+status_t _user_set_thread_affinity(thread_id id, const void* userMask, size_t size);
+
 
 status_t _user_block_thread(uint32 flags, bigtime_t timeout);
 status_t _user_unblock_thread(thread_id thread, status_t status);

headers/private/kernel/thread_types.h

@@ -449,6 +449,7 @@ struct Thread : TeamThreadIteratorEntry<thread_id>, KernelReferenceable {
 	int32			state;			// protected by scheduler lock
 	struct cpu_ent	*cpu;			// protected by scheduler lock
 	struct cpu_ent	*previous_cpu;	// protected by scheduler lock
+	CPUSet			cpumask;
 	int32			pinned_to_cpu;	// only accessed by this thread or in the
 									// scheduler, when thread is not running
 	spinlock		scheduler_lock;

headers/private/kernel/util/Heap.h

@@ -78,7 +78,7 @@ public:
 						Heap(int initialSize);
 						~Heap();
 
-	inline	Element*	PeekRoot() const;
+	inline	Element*	PeekRoot(int32 index = 0) const;
 
 	static	const Key&	GetKey(Element* element);
 
@@ -188,10 +188,10 @@ HEAP_CLASS_NAME::~Heap()
 
 HEAP_TEMPLATE_LIST
 Element*
-HEAP_CLASS_NAME::PeekRoot() const
+HEAP_CLASS_NAME::PeekRoot(int32 index) const
 {
-	if (fLastElement > 0)
-		return fElements[0];
+	if (index < fLastElement)
+		return fElements[index];
 	return NULL;
 }
 

headers/private/kernel/util/MinMaxHeap.h

@@ -73,8 +73,8 @@ public:
 						MinMaxHeap(int initialSize);
 						~MinMaxHeap();
 
-	inline	Element*	PeekMinimum() const;
-	inline	Element*	PeekMaximum() const;
+	inline	Element*	PeekMinimum(int32 index = 0) const;
+	inline	Element*	PeekMaximum(int32 index = 0) const;
 
 	static	const Key&	GetKey(Element* element);
 
@@ -190,13 +190,12 @@ MIN_MAX_HEAP_CLASS_NAME::~MinMaxHeap()
 
 MIN_MAX_HEAP_TEMPLATE_LIST
 Element*
-MIN_MAX_HEAP_CLASS_NAME::PeekMinimum() const
+MIN_MAX_HEAP_CLASS_NAME::PeekMinimum(int32 index) const
 {
-	if (fMinLastElement > 0)
-		return fMinElements[0];
-	else if (fMaxLastElement > 0) {
-		ASSERT(fMaxLastElement == 1);
-		return fMaxElements[0];
+	if (index < fMinLastElement)
+		return fMinElements[index];
+	else if (index - fMinLastElement < fMaxLastElement) {
+		return fMaxElements[fMaxLastElement - (index - fMinLastElement) - 1];
 	}
 
 	return NULL;
@@ -205,13 +204,12 @@ MIN_MAX_HEAP_CLASS_NAME::PeekMinimum() const
 
 MIN_MAX_HEAP_TEMPLATE_LIST
 Element*
-MIN_MAX_HEAP_CLASS_NAME::PeekMaximum() const
+MIN_MAX_HEAP_CLASS_NAME::PeekMaximum(int32 index) const
 {
-	if (fMaxLastElement > 0)
-		return fMaxElements[0];
-	else if (fMinLastElement > 0) {
-		ASSERT(fMinLastElement == 1);
-		return fMinElements[0];
+	if (index < fMaxLastElement)
+		return fMaxElements[index];
+	else if (index - fMaxLastElement < fMinLastElement) {
+		return fMinElements[fMinLastElement - (index - fMaxLastElement) - 1];
 	}
 
 	return NULL;

headers/private/system/syscalls.h

@@ -197,6 +197,8 @@ extern status_t		_kern_get_team_usage_info(team_id team, int32 who,
 extern status_t		_kern_get_extended_team_info(team_id teamID, uint32 flags,
 						void* buffer, size_t size, size_t* _sizeNeeded);
 extern int			_kern_get_cpu();
+extern status_t		_kern_get_thread_affinity(thread_id id, void* userMask, size_t size);
+extern status_t		_kern_set_thread_affinity(thread_id id, const void* userMask, size_t size);
 
 extern status_t		_kern_start_watching_system(int32 object, uint32 flags,
 						port_id port, int32 token);

src/system/kernel/cpu.cpp

@@ -28,6 +28,7 @@
 
 /* global per-cpu structure */
 cpu_ent gCPU[SMP_MAX_CPUS];
+CPUSet gCPUEnabled;
 
 uint32 gCPUCacheLevelCount;
 static cpu_topology_node sCPUTopology;
@@ -149,6 +150,7 @@ cpu_preboot_init_percpu(kernel_args *args, int curr_cpu)
 	// we can use it for get_current_cpu
 	memset(&gCPU[curr_cpu], 0, sizeof(gCPU[curr_cpu]));
 	gCPU[curr_cpu].cpu_num = curr_cpu;
+	gCPUEnabled.SetBit(curr_cpu);
 
 	list_init(&gCPU[curr_cpu].irqs);
 	B_INITIALIZE_SPINLOCK(&gCPU[curr_cpu].irqs_lock);

src/system/kernel/scheduler/low_latency.cpp

@@ -44,7 +44,7 @@ has_cache_expired(const ThreadData* threadData)
 
 
 static CoreEntry*
-choose_core(const ThreadData* /* threadData */)
+choose_core(const ThreadData* threadData)
 {
 	SCHEDULER_ENTER_FUNCTION();
 
@@ -55,16 +55,31 @@ choose_core(const ThreadData* /* threadData */)
 		package = PackageEntry::GetMostIdlePackage();
 	}
 
+	int32 index = 0;
+	CPUSet mask = threadData->GetCPUMask();
+	if (mask.IsEmpty()) {
+		// ignore when empty
+		mask.SetAll();
+	}
 	CoreEntry* core = NULL;
-	if (package != NULL)
-		core = package->GetIdleCore();
-
+	if (package != NULL) {
+		do {
+			core = package->GetIdleCore(index++);
+		} while (core != NULL && !core->CPUMask().Matches(mask));
+	}
 	if (core == NULL) {
 		ReadSpinLocker coreLocker(gCoreHeapsLock);
+		index = 0;
 		// no idle cores, use least occupied core
-		core = gCoreLoadHeap.PeekMinimum();
-		if (core == NULL)
-			core = gCoreHighLoadHeap.PeekMinimum();
+		do {
+			core = gCoreLoadHeap.PeekMinimum(index++);
+		} while (core != NULL && !core->CPUMask().Matches(mask));
+		if (core == NULL) {
+			index = 0;
+			do {
+				core = gCoreHighLoadHeap.PeekMinimum(index++);
+			} while (core != NULL && !core->CPUMask().Matches(mask));
+		}
 	}
 
 	ASSERT(core != NULL);
@@ -82,9 +97,25 @@ rebalance(const ThreadData* threadData)
 
 	// Get the least loaded core.
 	ReadSpinLocker coreLocker(gCoreHeapsLock);
-	CoreEntry* other = gCoreLoadHeap.PeekMinimum();
-	if (other == NULL)
-		other = gCoreHighLoadHeap.PeekMinimum();
+	CPUSet mask = threadData->GetCPUMask();
+	if (mask.IsEmpty()) {
+		// ignore when empty
+		mask.SetAll();
+	}
+	int32 index = 0;
+	CoreEntry* other;
+	do {
+		other = gCoreLoadHeap.PeekMinimum(index++);
+		if (other != NULL && other->CPUMask().IsEmpty())
+			panic("other->CPUMask().IsEmpty()\n");
+	} while (other != NULL && !other->CPUMask().Matches(mask));
+
+	if (other == NULL) {
+		index = 0;
+		do {
+			other = gCoreHighLoadHeap.PeekMinimum(index++);
+		} while (other != NULL && !other->CPUMask().Matches(mask));
+	}
 	coreLocker.Unlock();
 	ASSERT(other != NULL);
 

src/system/kernel/scheduler/power_saving.cpp

@@ -77,7 +77,6 @@ choose_idle_core()
 
 	if (package != NULL)
 		return package->GetIdleCore();
-
 	return NULL;
 }
 
@@ -89,22 +88,38 @@ choose_core(const ThreadData* threadData)
 
 	CoreEntry* core = NULL;
 
+	CPUSet mask = threadData->GetCPUMask();
+	if (mask.IsEmpty()) {
+		// ignore when empty
+		mask.SetAll();
+	}
+
 	// try to pack all threads on one core
 	core = choose_small_task_core();
+	if (!core->CPUMask().Matches(mask))
+		core = NULL;
 
 	if (core == NULL || core->GetLoad() + threadData->GetLoad() >= kHighLoad) {
 		ReadSpinLocker coreLocker(gCoreHeapsLock);
 
 		// run immediately on already woken core
-		core = gCoreLoadHeap.PeekMinimum();
+		int32 index = 0;
+		do {
+			core = gCoreLoadHeap.PeekMinimum(index++);
+		} while (core != NULL && !core->CPUMask().Matches(mask));
 		if (core == NULL) {
 			coreLocker.Unlock();
 
 			core = choose_idle_core();
+			if (!core->CPUMask().Matches(mask))
+				core = NULL;
 
 			if (core == NULL) {
 				coreLocker.Lock();
-				core = gCoreHighLoadHeap.PeekMinimum();
+				index = 0;
+				do {
+					core = gCoreHighLoadHeap.PeekMinimum(index++);
+				} while (core != NULL && !core->CPUMask().Matches(mask));
 			}
 		}
 	}
@@ -121,6 +136,11 @@ rebalance(const ThreadData* threadData)
 
 	ASSERT(!gSingleCore);
 
+	CPUSet mask = threadData->GetCPUMask();
+	if (mask.IsEmpty()) {
+		// ignore when empty
+		mask.SetAll();
+	}
 	CoreEntry* core = threadData->Core();
 
 	int32 coreLoad = core->GetLoad();
@@ -130,7 +150,7 @@ rebalance(const ThreadData* threadData)
 			sSmallTaskCore = NULL;
 			CoreEntry* smallTaskCore = choose_small_task_core();
 
-			if (threadLoad > coreLoad / 3)
+			if (threadLoad > coreLoad / 3 || !smallTaskCore->CPUMask().Matches(mask))
 				return core;
 			return coreLoad > kVeryHighLoad ? smallTaskCore : core;
 		}
@@ -139,9 +159,17 @@ rebalance(const ThreadData* threadData)
 			return core;
 
 		ReadSpinLocker coreLocker(gCoreHeapsLock);
-		CoreEntry* other = gCoreLoadHeap.PeekMaximum();
-		if (other == NULL)
-			other = gCoreHighLoadHeap.PeekMinimum();
+		CoreEntry* other;
+		int32 index = 0;
+		do {
+			other = gCoreLoadHeap.PeekMaximum(index++);
+		} while (other != NULL && !core->CPUMask().Matches(mask));
+		if (other == NULL) {
+			index = 0;
+			do {
+				other = gCoreHighLoadHeap.PeekMinimum(index++);
+			} while (other != NULL && !core->CPUMask().Matches(mask));
+		}
 		coreLocker.Unlock();
 		ASSERT(other != NULL);
 
@@ -154,7 +182,7 @@ rebalance(const ThreadData* threadData)
 		return core;
 
 	CoreEntry* smallTaskCore = choose_small_task_core();
-	if (smallTaskCore == NULL)
+	if (smallTaskCore == NULL || !smallTaskCore->CPUMask().Matches(mask))
 		return core;
 	return smallTaskCore->GetLoad() + threadLoad < kHighLoad
 		? smallTaskCore : core;

src/system/kernel/scheduler/scheduler.cpp

@@ -117,7 +117,7 @@ enqueue(Thread* thread, bool newOne)
 
 	const bool rescheduleNeeded = threadData->ChooseCoreAndCPU(targetCore, targetCPU);
 
-	TRACE("enqueueing thread %ld with priority %ld on CPU %ld (core %ld)\n",
+	TRACE("enqueueing thread %" B_PRId32 " with priority %" B_PRId32 " on CPU %" B_PRId32 " (core %" B_PRId32 ")\n",
 		thread->id, threadPriority, targetCPU->ID(), targetCore->ID());
 
 	bool wasRunQueueEmpty = false;
@@ -153,7 +153,7 @@ scheduler_enqueue_in_run_queue(Thread *thread)
 
 	SchedulerModeLocker _;
 
-	TRACE("enqueueing new thread %ld with static priority %ld\n", thread->id,
+	TRACE("enqueueing new thread %" B_PRId32 " with static priority %" B_PRId32 "\n", thread->id,
 		thread->priority);
 
 	ThreadData* threadData = thread->scheduler_data;
@@ -180,7 +180,7 @@ scheduler_set_thread_priority(Thread *thread, int32 priority)
 	ThreadData* threadData = thread->scheduler_data;
 	int32 oldPriority = thread->priority;
 
-	TRACE("changing thread %ld priority to %ld (old: %ld, effective: %ld)\n",
+	TRACE("changing thread %" B_PRId32 " priority to %" B_PRId32 " (old: %" B_PRId32 ", effective: %" B_PRId32 ")\n",
 		thread->id, priority, oldPriority, threadData->GetEffectivePriority());
 
 	thread->priority = priority;
@@ -333,7 +333,7 @@ reschedule(int32 nextState)
 
 	SchedulerModeLocker modeLocker;
 
-	TRACE("reschedule(): cpu %ld, current thread = %ld\n", thisCPU,
+	TRACE("reschedule(): cpu %" B_PRId32 ", current thread = %" B_PRId32 "\n", thisCPU,
 		oldThread->id);
 
 	oldThread->state = nextState;
@@ -348,7 +348,7 @@ reschedule(int32 nextState)
 		case B_THREAD_READY:
 			enqueueOldThread = true;
 
-			if (!oldThreadData->IsIdle()) {
+			if (!oldThreadData->IsIdle() && oldThreadData->GetCPUMask().GetBit(thisCPU)) {
 				oldThreadData->Continues();
 				if (oldThreadData->HasQuantumEnded(oldThread->cpu->preempted,
 						oldThread->has_yielded)) {
@@ -394,9 +394,21 @@ reschedule(int32 nextState)
 		} else
 			nextThreadData = oldThreadData;
 	} else {
+		CPUSet mask = oldThreadData->GetCPUMask();
+		if (mask.IsEmpty())
+			mask.SetAll();
+		bool oldThreadShouldMigrate = !mask.GetBit(thisCPU);
+		if (oldThreadShouldMigrate)
+			enqueueOldThread = false;
 		nextThreadData
 			= cpu->ChooseNextThread(enqueueOldThread ? oldThreadData : NULL,
 				putOldThreadAtBack);
+		if (oldThreadShouldMigrate) {
+			enqueue(oldThread, true);
+			// replace with the idle thread, if no other thread could be found
+			if (oldThreadData == nextThreadData)
+				nextThreadData = cpu->PeekIdleThread();
+		}
 
 		// update CPU heap
 		CoreCPUHeapLocker cpuLocker(core);
@@ -417,7 +429,7 @@ reschedule(int32 nextState)
 		acquire_spinlock(&nextThread->scheduler_lock);
 	}
 
-	TRACE("reschedule(): cpu %ld, next thread = %ld\n", thisCPU,
+	TRACE("reschedule(): cpu %" B_PRId32 ", next thread = %" B_PRId32 "\n", thisCPU,
 		nextThread->id);
 
 	T(ScheduleThread(nextThread, oldThread));
@@ -572,6 +584,10 @@ scheduler_set_cpu_enabled(int32 cpuID, bool enabled)
 	}
 
 	gCPU[cpuID].disabled = !enabled;
+	if (enabled)
+		gCPUEnabled.SetBitAtomic(cpuID);
+	else
+		gCPUEnabled.ClearBitAtomic(cpuID);
 
 	if (!enabled) {
 		cpu->Stop();

src/system/kernel/scheduler/scheduler_cpu.cpp

@@ -233,7 +233,8 @@ CPUEntry::ChooseNextThread(ThreadData* oldThread, bool putAtBack)
 	CoreRunQueueLocker coreLocker(fCore);
 
 	ThreadData* sharedThread = fCore->PeekThread();
-	ASSERT(sharedThread != NULL || pinnedThread != NULL || oldThread != NULL);
+	if (sharedThread == NULL && pinnedThread == NULL && oldThread == NULL)
+		return NULL;
 
 	int32 sharedPriority = -1;
 	if (sharedThread != NULL)
@@ -477,6 +478,7 @@ CoreEntry::AddCPU(CPUEntry* cpu)
 
 		fPackage->AddIdleCore(this);
 	}
+	fCPUSet.SetBit(cpu->ID());
 
 	fCPUHeap.Insert(cpu, B_IDLE_PRIORITY);
 }
@@ -489,6 +491,7 @@ CoreEntry::RemoveCPU(CPUEntry* cpu, ThreadProcessing& threadPostProcessing)
 	ASSERT(fIdleCPUCount > 0);
 
 	fIdleCPUCount--;
+	fCPUSet.ClearBit(cpu->ID());
 	if (--fCPUCount == 0) {
 		// unassign threads
 		thread_map(CoreEntry::_UnassignThread, this);

src/system/kernel/scheduler/scheduler_cpu.h

@@ -8,6 +8,7 @@
 
 #include <OS.h>
 
+#include <smp.h>
 #include <thread.h>
 #include <util/AutoLock.h>
 #include <util/Heap.h>
@@ -131,6 +132,8 @@ public:
 	inline				PackageEntry*	Package() const	{ return fPackage; }
 	inline				int32			CPUCount() const
 											{ return fCPUCount; }
+	inline				const CPUSet&	CPUMask() const
+											{ return fCPUSet; }
 
 	inline				void			LockCPUHeap();
 	inline				void			UnlockCPUHeap();
@@ -182,6 +185,7 @@ private:
 						PackageEntry*	fPackage;
 
 						int32			fCPUCount;
+						CPUSet			fCPUSet;
 						int32			fIdleCPUCount;
 						CPUPriorityHeap	fCPUHeap;
 						spinlock		fCPULock;
@@ -229,7 +233,7 @@ public:
 	inline				void				CoreGoesIdle(CoreEntry* core);
 	inline				void				CoreWakesUp(CoreEntry* core);
 
-	inline				CoreEntry*			GetIdleCore() const;
+	inline				CoreEntry*			GetIdleCore(int32 index = 0) const;
 
 						void				AddIdleCore(CoreEntry* core);
 						void				RemoveIdleCore(CoreEntry* core);
@@ -538,10 +542,14 @@ CoreEntry::GetCore(int32 cpu)
 
 
 inline CoreEntry*
-PackageEntry::GetIdleCore() const
+PackageEntry::GetIdleCore(int32 index) const
 {
 	SCHEDULER_ENTER_FUNCTION();
-	return fIdleCores.Last();
+	CoreEntry* element = fIdleCores.Last();
+	for (int32 i = 0; element != NULL && i < index; i++)
+		element = fIdleCores.GetPrevious(element);
+
+	return element;
 }
 
 

src/system/kernel/scheduler/scheduler_thread.cpp

@@ -59,10 +59,16 @@ ThreadData::_ChooseCPU(CoreEntry* core, bool& rescheduleNeeded) const
 
 	int32 threadPriority = GetEffectivePriority();
 
-	if (fThread->previous_cpu != NULL) {
+	CPUSet mask = GetCPUMask();
+	if (mask.IsEmpty())
+		mask.SetAll();
+	ASSERT(mask.Matches(core->CPUMask()));
+
+	if (fThread->previous_cpu != NULL && !fThread->previous_cpu->disabled
+		&& mask.GetBit(fThread->previous_cpu->cpu_num)) {
 		CPUEntry* previousCPU
 			= CPUEntry::GetCPU(fThread->previous_cpu->cpu_num);
-		if (previousCPU->Core() == core && !fThread->previous_cpu->disabled) {
+		if (previousCPU->Core() == core) {
 			CoreCPUHeapLocker _(core);
 			if (CPUPriorityHeap::GetKey(previousCPU) < threadPriority) {
 				previousCPU->UpdatePriority(threadPriority);
@@ -73,7 +79,11 @@ ThreadData::_ChooseCPU(CoreEntry* core, bool& rescheduleNeeded) const
 	}
 
 	CoreCPUHeapLocker _(core);
-	CPUEntry* cpu = core->CPUHeap()->PeekRoot();
+	int32 index = 0;
+	CPUEntry* cpu;
+	do {
+		cpu = core->CPUHeap()->PeekRoot(index++);
+	} while (cpu != NULL && !mask.GetBit(cpu->ID()));
 	ASSERT(cpu != NULL);
 
 	if (CPUPriorityHeap::GetKey(cpu) < threadPriority) {
@@ -156,12 +166,21 @@ ThreadData::ChooseCoreAndCPU(CoreEntry*& targetCore, CPUEntry*& targetCPU)
 
 	bool rescheduleNeeded = false;
 
+	if (targetCore != NULL && !targetCore->CPUMask().Matches(GetCPUMask()))
+		targetCore = NULL;
+	if (targetCPU != NULL && !GetCPUMask().GetBit(targetCPU->ID()))
+		targetCPU = NULL;
+
 	if (targetCore == NULL && targetCPU != NULL)
 		targetCore = targetCPU->Core();
 	else if (targetCore != NULL && targetCPU == NULL)
 		targetCPU = _ChooseCPU(targetCore, rescheduleNeeded);
 	else if (targetCore == NULL && targetCPU == NULL) {
 		targetCore = _ChooseCore();
+		CPUSet mask = GetCPUMask();
+		if (mask.IsEmpty())
+			mask.SetAll();
+		ASSERT(mask.Matches(targetCore->CPUMask()));
 		targetCPU = _ChooseCPU(targetCore, rescheduleNeeded);
 	}
 

src/system/kernel/scheduler/scheduler_thread.h

@@ -39,6 +39,7 @@ public:
 
 	inline	int32		GetPriority() const	{ return fThread->priority; }
 	inline	Thread*		GetThread() const	{ return fThread; }
+	inline	CPUSet		GetCPUMask() const	{ return fThread->cpumask.And(gCPUEnabled); }
 
 	inline	bool		IsRealTime() const;
 	inline	bool		IsIdle() const;

src/system/kernel/thread.cpp

@@ -174,6 +174,7 @@ Thread::Thread(const char* name, thread_id threadID, struct cpu_ent* cpu)
 	io_priority(-1),
 	cpu(cpu),
 	previous_cpu(NULL),
+	cpumask(),
 	pinned_to_cpu(0),
 	sig_block_mask(0),
 	sigsuspend_original_unblocked_mask(0),
@@ -203,6 +204,7 @@ Thread::Thread(const char* name, thread_id threadID, struct cpu_ent* cpu)
 {
 	id = threadID >= 0 ? threadID : allocate_thread_id();
 	visible = false;
+	cpumask.SetAll();
 
 	// init locks
 	char lockName[32];
@@ -1778,6 +1780,7 @@ _dump_thread_info(Thread *thread, bool shortInfo)
 		kprintf("(%d)\n", thread->cpu->cpu_num);
 	else
 		kprintf("\n");
+	kprintf("cpumask:            %#" B_PRIx32 "\n", thread->cpumask.Bits(0));
 	kprintf("sig_pending:        %#" B_PRIx64 " (blocked: %#" B_PRIx64
 		", before sigsuspend(): %#" B_PRIx64 ")\n",
 		(int64)thread->ThreadPendingSignals(),
@@ -2752,6 +2755,8 @@ thread_init(kernel_args *args)
 		thread->team = team_get_kernel_team();
 		thread->priority = B_IDLE_PRIORITY;
 		thread->state = B_THREAD_RUNNING;
+		thread->cpumask.SetAll();
+
 		sprintf(name, "idle thread %" B_PRIu32 " kstack", i + 1);
 		thread->kernel_stack_area = find_area(name);
 
@@ -3923,3 +3928,69 @@ _user_get_cpu()
 	Thread* thread = thread_get_current_thread();
 	return thread->cpu->cpu_num;
 }
+
+
+status_t
+_user_get_thread_affinity(thread_id id, void* userMask, size_t size)
+{
+	if (userMask == NULL || id < B_OK)
+		return B_BAD_VALUE;
+
+	if (!IS_USER_ADDRESS(userMask))
+		return B_BAD_ADDRESS;
+
+	CPUSet mask;
+
+	if (id == 0)
+		id = thread_get_current_thread_id();
+	// get the thread
+	Thread* thread = Thread::GetAndLock(id);
+	if (thread == NULL)
+		return B_BAD_THREAD_ID;
+	BReference<Thread> threadReference(thread, true);
+	ThreadLocker threadLocker(thread, true);
+	memcpy(&mask, &thread->cpumask, sizeof(mask));
+
+	if (user_memcpy(userMask, &mask, min_c(sizeof(mask), size)) < B_OK)
+		return B_BAD_ADDRESS;
+
+	return B_OK;
+}
+
+status_t
+_user_set_thread_affinity(thread_id id, const void* userMask, size_t size)
+{
+	if (userMask == NULL || id < B_OK || size < sizeof(CPUSet))
+		return B_BAD_VALUE;
+
+	if (!IS_USER_ADDRESS(userMask))
+		return B_BAD_ADDRESS;
+
+	CPUSet mask;
+	if (user_memcpy(&mask, userMask, min_c(sizeof(CPUSet), size)) < B_OK)
+		return B_BAD_ADDRESS;
+
+	CPUSet cpus;
+	cpus.SetAll();
+	for (int i = 0; i < smp_get_num_cpus(); i++)
+		cpus.ClearBit(i);
+	if (mask.Matches(cpus))
+		return B_BAD_VALUE;
+
+	if (id == 0)
+		id = thread_get_current_thread_id();
+
+	// get the thread
+	Thread* thread = Thread::GetAndLock(id);
+	if (thread == NULL)
+		return B_BAD_THREAD_ID;
+	BReference<Thread> threadReference(thread, true);
+	ThreadLocker threadLocker(thread, true);
+	memcpy(&thread->cpumask, &mask, sizeof(mask));
+
+	// check if running on masked cpu
+	if (!thread->cpumask.GetBit(thread->cpu->cpu_num))
+		thread_yield();
+
+	return B_OK;
+}