diff --git a/kernel-src/arch/x86-64/main.c b/kernel-src/arch/x86-64/main.c
index c6491f37..aa40a797 100644
--- a/kernel-src/arch/x86-64/main.c
+++ b/kernel-src/arch/x86-64/main.c
@@ -13,6 +13,7 @@
 #include <arch/apic.h>
 #include <arch/hpet.h>
 #include <kernel/timekeeper.h>
+#include <kernel/scheduler.h>
 
 static cpu_t bsp_cpu;
 
@@ -35,5 +36,6 @@ void kernel_entry() {
 	time_t u = arch_hpet_init();
 	timekeeper_init(arch_hpet_ticks, u);
 	arch_apic_timerinit();
+	sched_init();
 	__assert(!"kernel entry end");
 }
diff --git a/kernel-src/include/kernel/scheduler.h b/kernel-src/include/kernel/scheduler.h
new file mode 100644
index 00000000..12c8a3a2
--- /dev/null
+++ b/kernel-src/include/kernel/scheduler.h
@@ -0,0 +1,60 @@
+#ifndef _SCHEDULER_H
+#define _SCHEDULER_H
+
+#include <arch/context.h>
+#include <kernel/vmm.h>
+
+#define SCHED_THREAD_FLAGS_QUEUED 1
+#define SCHED_THREAD_FLAGS_RUNNING 2
+
+typedef int pid_t;
+typedef int tid_t;
+typedef int gid_t;
+typedef int uid_t;
+typedef unsigned int mode_t;
+
+struct proc_t;
+
+typedef struct thread_t {
+	struct thread_t *next;
+	struct thread_t *prev;
+	struct proc_t *proc;
+	struct cpu_t *cpu;
+	context_t context;
+	extracontext_t extracontext;
+	void *kernelstack;
+	void *kernelstacktop;
+	size_t kernelstacksize;
+	vmmcontext_t *vmmctx;
+	tid_t tid;
+	int flags;
+	long priority;
+} thread_t;
+
+typedef struct proc_t {
+	spinlock_t lock;
+	struct proc_t *sibling;
+	struct proc_t *parent;
+	struct proc_t *child;
+	pid_t pid;
+	gid_t gid;
+	uid_t uid;
+	thread_t **threads;
+	size_t threadtablesize;
+	size_t runningthreadcount;
+	mode_t umask;
+	int flags;
+} proc_t;
+
+#include <arch/cpu.h>
+
+void sched_init();
+void sched_runinit();
+void sched_threadexit();
+void sched_queue(thread_t *thread);
+void sched_stopcurrentthread();
+void sched_yield();
+proc_t *sched_newproc();
+thread_t *sched_newthread(void *ip, size_t kstacksize, int priority, proc_t *proc, void *ustack);
+
+#endif
diff --git a/kernel-src/include/x86-64/arch/cpu.h b/kernel-src/include/x86-64/arch/cpu.h
index 3890a86b..76c0c870 100644
--- a/kernel-src/include/x86-64/arch/cpu.h
+++ b/kernel-src/include/x86-64/arch/cpu.h
@@ -10,8 +10,9 @@
 #include <kernel/interrupt.h>
 #include <kernel/vmm.h>
 #include <kernel/timer.h>
+#include <kernel/scheduler.h>
 
-typedef struct {
+typedef struct cpu_t {
 	uint64_t gdt[7];
 	ist_t ist;
 	long id;
@@ -21,8 +22,13 @@ typedef struct {
 	timer_t *timer;
 	bool intstatus;
 	long ipl;
+	thread_t *thread;
+	thread_t *idlethread;
+	timerentry_t schedtimerentry;
 } cpu_t;
 
+#define CPU_HALT() asm volatile("hlt")
+
 static inline cpu_t *_cpu() {
 	return (cpu_t *)rdmsr(MSR_GSBASE);
 }
diff --git a/kernel-src/sys/scheduler.c b/kernel-src/sys/scheduler.c
new file mode 100644
index 00000000..7a7c0360
--- /dev/null
+++ b/kernel-src/sys/scheduler.c
@@ -0,0 +1,276 @@
+#include <kernel/scheduler.h>
+#include <kernel/slab.h>
+#include <logging.h>
+#include <arch/cpu.h>
+#include <spinlock.h>
+#include <kernel/vmm.h>
+#include <kernel/alloc.h>
+
+#define QUANTUM_US 100000
+
+static scache_t *threadcache;
+static scache_t *processcache;
+
+#define RUNQUEUE_COUNT 64
+
+typedef struct {
+	thread_t *list;
+	thread_t *last;
+} rqueue_t;
+
+static rqueue_t runqueue[RUNQUEUE_COUNT];
+static uint64_t runqueuebitmap;
+static spinlock_t runqueuelock;
+
+proc_t *sched_newproc() {
+	proc_t *proc = slab_allocate(processcache);
+	if (proc == NULL)
+		return NULL;
+
+	proc->threads = alloc(sizeof(thread_t *));
+	if (proc->threads == NULL) {
+		slab_free(processcache, proc);
+		return NULL;
+	}
+	// TODO move this to slab
+	proc->threadtablesize = 1;
+	proc->runningthreadcount = 1;
+	SPINLOCK_INIT(proc->lock);
+
+	return proc;
+}
+
+thread_t *sched_newthread(void *ip, size_t kstacksize, int priority, proc_t *proc, void *ustack) {
+	thread_t *thread = slab_allocate(threadcache);
+	if (thread == NULL)
+		return NULL;
+
+	thread->kernelstack = vmm_map(NULL, kstacksize, VMM_FLAGS_ALLOCATE, ARCH_MMU_FLAGS_WRITE | ARCH_MMU_FLAGS_READ | ARCH_MMU_FLAGS_NOEXEC, NULL);
+	if (thread->kernelstack == NULL) {
+		slab_free(threadcache, thread);
+		return NULL;
+	}
+
+	thread->kernelstacktop = (void *)((uintptr_t)thread->kernelstack + kstacksize);
+
+	// non kernel thread vmm contexts are handled by the caller
+	thread->vmmctx = proc ? NULL : &vmm_kernelctx;
+	thread->proc = proc;
+	thread->priority = priority;
+
+	CTX_INIT(&thread->context, proc != NULL);
+	CTX_SP(&thread->context) = proc ? (ctxreg_t)ustack : (ctxreg_t)thread->kernelstacktop;
+	CTX_IP(&thread->context) = (ctxreg_t)ip;
+
+	return thread;
+}
+
+void sched_destroythread(thread_t *thread) {
+	vmm_unmap(thread->kernelstack, thread->kernelstacksize, 0);
+	slab_free(threadcache, thread);
+}
+
+void sched_destroyproc(proc_t *proc) {
+	free(proc->threads);
+	slab_free(processcache, proc);
+}
+
+static thread_t *runqueuenext() {
+	bool intstate = interrupt_set(false);
+
+	thread_t *thread = NULL;
+
+	if (runqueuebitmap == 0)
+		goto leave;
+
+	// TODO use the bitmap for this
+	for (int i = 0; i < RUNQUEUE_COUNT; ++i) {
+		if (runqueue[i].list) {
+			thread = runqueue[i].list;
+			runqueue[i].list = thread->next;
+			if (runqueue[i].list == NULL) {
+				runqueue[i].last = NULL;
+				runqueuebitmap &= ~((uint64_t)1 << i);
+			}
+			break;
+		}
+	}
+
+	thread->flags &= ~SCHED_THREAD_FLAGS_QUEUED;
+
+	leave:
+	interrupt_set(intstate);
+	return thread;
+}
+
+static __attribute__((noreturn)) void switchthread(thread_t *thread) {
+	interrupt_set(false);
+	thread_t* current = _cpu()->thread;
+	
+	_cpu()->thread = thread;
+
+	if(current == NULL || thread->vmmctx != current->vmmctx)
+		vmm_switchcontext(thread->vmmctx);
+
+	ARCH_CONTEXT_SWITCHTHREAD(thread);
+
+	__builtin_unreachable();
+}
+
+static void runqueueinsert(thread_t *thread) {
+	thread->flags |= SCHED_THREAD_FLAGS_QUEUED;
+
+	runqueuebitmap |= ((uint64_t)1 << thread->priority);
+
+	thread->prev = runqueue[thread->priority].last;
+	if (thread->prev)
+		thread->prev->next = thread;
+	else
+		runqueue[thread->priority].list = thread;
+
+	thread->next = NULL;
+	runqueue[thread->priority].last = thread;
+}
+
+void sched_queue(thread_t *thread) {
+	bool intstate = interrupt_set(false);
+	spinlock_acquire(&runqueuelock);
+
+	// maybe instead of an assert, a simple return would suffice as the thread would already be queued anyways
+	__assert((thread->flags & SCHED_THREAD_FLAGS_QUEUED) == 0 && (thread->flags & SCHED_THREAD_FLAGS_RUNNING) == 0);
+
+	runqueueinsert(thread);
+
+	spinlock_release(&runqueuelock);
+	interrupt_set(intstate);
+	// TODO yield if higher priority than current thread (or send another CPU an IPI)
+}
+
+__attribute__((noreturn)) void sched_stopcurrentthread() {
+	interrupt_set(false);
+	if (_cpu()->thread)
+		_cpu()->thread->flags &= ~SCHED_THREAD_FLAGS_RUNNING;
+
+	spinlock_acquire(&runqueuelock);
+
+	thread_t *next = runqueuenext();
+	if (next == NULL)
+		next = _cpu()->idlethread;
+
+	spinlock_release(&runqueuelock);
+
+	switchthread(next);
+}
+
+void sched_threadexit() {
+	interrupt_set(false);
+	thread_t *thread = _cpu()->thread;
+	proc_t *proc = thread->proc;
+
+	if (proc) {
+		spinlock_acquire(&proc->lock);
+
+		int toffset = -1;
+
+		for (int i = 0; i < proc->threadtablesize; ++i) {
+			if (proc->threads[i] == thread) {
+				toffset = i;
+				break;
+			}
+		}
+
+		__assert(toffset != -1);
+		--proc->runningthreadcount;
+
+		if (proc->runningthreadcount == 0) {
+			// TODO prepare proc state to be left as a zombie
+		}
+		spinlock_release(&proc->lock);
+	}
+
+	_cpu()->thread = NULL;
+
+	// because a thread deallocating its own data is a nightmare, thread deallocation and such will be left to whoever frees the proc it's tied to
+	// (likely an exit(2) call)
+	// FIXME this doesn't apply to kernel threads and something should be figured out for them
+
+	sched_stopcurrentthread();
+}
+
+static void yield(context_t *context) {
+	thread_t *thread = _cpu()->thread;
+
+	spinlock_acquire(&runqueuelock);
+
+	thread_t *next = runqueuenext();
+
+	// TODO priority checking and stuff
+
+	if (next) {
+		ARCH_CONTEXT_THREADSAVE(thread, context);
+		runqueueinsert(thread);
+
+		spinlock_release(&runqueuelock);
+		switchthread(next);
+	}
+
+	spinlock_release(&runqueuelock);
+}
+
+void sched_yield() {
+	bool old = interrupt_set(false);
+	arch_context_saveandcall(yield);
+	interrupt_set(old);
+}
+
+static void timerhook(void *private, context_t *context) {
+	thread_t* current = _cpu()->thread;
+
+	spinlock_acquire(&runqueuelock);
+
+	thread_t *next = runqueuenext();
+
+	// TODO priority handling here
+
+	if (next) {
+		current->flags &= ~SCHED_THREAD_FLAGS_RUNNING;
+		ARCH_CONTEXT_THREADSAVE(current, context);
+
+		runqueueinsert(current);
+
+		ARCH_CONTEXT_THREADLOAD(next, context);
+		if (next->vmmctx != current->vmmctx)
+			vmm_switchcontext(next->vmmctx);
+		next->flags |= SCHED_THREAD_FLAGS_RUNNING;
+		_cpu()->thread = next;
+	}
+
+	spinlock_release(&runqueuelock);
+}
+
+static void cpuidlethread() {
+	// TODO check every interrupt rather than waiting for a scheduler timer interrupt
+	interrupt_set(true);
+	while (1)
+		CPU_HALT();
+}
+
+void sched_init() {
+	threadcache = slab_newcache(sizeof(thread_t), 0, NULL, NULL);
+	__assert(threadcache);
+	processcache = slab_newcache(sizeof(proc_t), 0, NULL, NULL);
+	__assert(processcache);
+
+	_cpu()->schedtimerentry.func = timerhook;
+	_cpu()->schedtimerentry.repeatus = QUANTUM_US;
+	SPINLOCK_INIT(runqueuelock);
+
+	_cpu()->idlethread = sched_newthread(cpuidlethread, PAGE_SIZE * 4, 0, NULL, NULL);
+	__assert(_cpu()->idlethread);
+	_cpu()->thread = sched_newthread(NULL, PAGE_SIZE * 32, 0, NULL, NULL);
+	__assert(_cpu()->thread);
+
+	timer_insert(_cpu()->timer, &_cpu()->schedtimerentry, QUANTUM_US);
+	// XXX move this resume to a more appropriate place
+	timer_resume(_cpu()->timer);
+}