diff --git a/zion/loader/elf_loader.cpp b/zion/loader/elf_loader.cpp
index 2a7ed6a..5b4b5e6 100644
--- a/zion/loader/elf_loader.cpp
+++ b/zion/loader/elf_loader.cpp
@@ -46,17 +46,9 @@ typedef struct {
   uint64_t align;
 } Elf64ProgramHeader;
 
-void badmemcpy(uint64_t base, uint64_t offset, uint64_t dest) {
-  uint8_t* ptr = reinterpret_cast<uint8_t*>(base);
-  uint8_t* dest_ptr = reinterpret_cast<uint8_t*>(dest);
-  for (uint64_t i = 0; i < offset; i++) {
-    dest_ptr[i] = ptr[i];
-  }
-}
-
 }  // namespace
 
-uint64_t LoadElfProgram(uint64_t base, uint64_t offset) {
+uint64_t LoadElfProgram(uint64_t cr3, uint64_t base, uint64_t offset) {
   Elf64Header* header = reinterpret_cast<Elf64Header*>(base);
   dbgln("phoff: %u phnum: %u", header->phoff, header->phnum);
   Elf64ProgramHeader* programs =
@@ -68,8 +60,8 @@ uint64_t LoadElfProgram(uint64_t base, uint64_t offset) {
         "filesz: %u, memsz: %u, align: %u",
         program.type, program.flags, program.offset, program.vaddr,
         program.paddr, program.filesz, program.memsz, program.align);
-    EnsureResident(program.vaddr, program.memsz);
-    badmemcpy(base + program.offset, program.filesz, program.vaddr);
+    CopyIntoNonResidentProcess(base + program.offset, program.filesz, cr3,
+                               program.vaddr);
   }
   return header->entry;
 }
diff --git a/zion/loader/elf_loader.h b/zion/loader/elf_loader.h
index 1c6c7e3..f6ae328 100644
--- a/zion/loader/elf_loader.h
+++ b/zion/loader/elf_loader.h
@@ -3,4 +3,4 @@
 #include <stdint.h>
 
 // Loads the elf program and returns its entry point.
-uint64_t LoadElfProgram(uint64_t base, uint64_t length);
+uint64_t LoadElfProgram(uint64_t cr3, uint64_t base, uint64_t length);
diff --git a/zion/loader/init_loader.cpp b/zion/loader/init_loader.cpp
index a10da9c..79bb7cf 100644
--- a/zion/loader/init_loader.cpp
+++ b/zion/loader/init_loader.cpp
@@ -25,6 +25,11 @@ const limine_file& GetInitProgram() {
 void LoadInitProgram() {
   const limine_file& init_prog = GetInitProgram();
 
-  gProcMan->InsertProcess(
-      new Process(reinterpret_cast<uint64_t>(init_prog.address)));
+  SharedPtr<Process> proc = MakeShared<Process>();
+  gProcMan->InsertProcess(proc);
+
+  uint64_t entry =
+      LoadElfProgram(proc->cr3(), reinterpret_cast<uint64_t>(init_prog.address),
+                     init_prog.size);
+  proc->CreateThread(entry);
 }
diff --git a/zion/memory/paging_util.cpp b/zion/memory/paging_util.cpp
index c8d2c27..f77d5dd 100644
--- a/zion/memory/paging_util.cpp
+++ b/zion/memory/paging_util.cpp
@@ -69,7 +69,7 @@ uint64_t* PageTableEntry(uint64_t pt_phys, uint64_t addr) {
                                      ShiftForEntryIndexing(addr, PT_OFFSET));
 }
 
-bool IsPageResident(uint64_t cr3, uint64_t virt) {
+uint64_t PagePhysIfResident(uint64_t cr3, uint64_t virt) {
   uint64_t* pml4_entry = Pml4Entry(cr3, virt);
   if (!(*pml4_entry & PRESENT_BIT)) {
     return false;
@@ -86,10 +86,10 @@ bool IsPageResident(uint64_t cr3, uint64_t virt) {
   if (!(*pt_entry & PRESENT_BIT)) {
     return false;
   }
-  return true;
+  return *pt_entry & ~0xFFF;
 }
 
-void MapPage(uint64_t cr3, uint64_t virt) {
+uint64_t MapPage(uint64_t cr3, uint64_t virt) {
   uint64_t access_bits = PRESENT_BIT | READ_WRITE_BIT;
   uint64_t higher_half = 0xffff8000'00000000;
   if ((virt & higher_half) != higher_half) {
@@ -121,8 +121,10 @@ void MapPage(uint64_t cr3, uint64_t virt) {
     *pt_entry = PageAlign(phys) | access_bits;
     ZeroOutPage(reinterpret_cast<uint64_t*>(boot::GetHigherHalfDirectMap() +
                                             PageAlign(phys)));
+    return phys;
   } else {
     panic("Page already allocated.");
+    return 0;
   }
 }
 
@@ -152,14 +154,15 @@ void InitializePml4(uint64_t pml4_physical_addr) {
   pml4_virtual[Pml4Index(hhdm)] = *Pml4Entry(curr_cr3, hhdm);
 }
 
-void AllocatePageIfNecessary(uint64_t addr, uint64_t cr3) {
+uint64_t AllocatePageIfNecessary(uint64_t addr, uint64_t cr3) {
   if (cr3 == 0) {
     cr3 = CurrCr3();
   }
-  if (IsPageResident(cr3, addr)) {
-    return;
+  uint64_t phys = PagePhysIfResident(cr3, addr);
+  if (phys) {
+    return phys;
   }
-  MapPage(cr3, addr);
+  return MapPage(cr3, addr);
 }
 
 void EnsureResident(uint64_t addr, uint64_t size) {
@@ -170,3 +173,21 @@ void EnsureResident(uint64_t addr, uint64_t size) {
     addr += 0x1000;
   }
 }
+
+void CopyIntoNonResidentProcess(uint64_t base, uint64_t size, uint64_t dest_cr3,
+                                uint64_t dest_virt) {
+  if (size > 0x1000) {
+    panic("Unimplemented NR copy > 1 page");
+  }
+  if (dest_virt & 0xFFF) {
+    panic("Unimplemented NR copy to non page aligned");
+  }
+  uint64_t phys = AllocatePageIfNecessary(dest_virt, dest_cr3);
+  uint8_t* src = reinterpret_cast<uint8_t*>(base);
+  uint8_t* dest =
+      reinterpret_cast<uint8_t*>(phys + boot::GetHigherHalfDirectMap());
+
+  for (uint64_t i = 0; i < size; i++) {
+    dest[i] = src[i];
+  }
+}
diff --git a/zion/memory/paging_util.h b/zion/memory/paging_util.h
index 055eb3e..889234e 100644
--- a/zion/memory/paging_util.h
+++ b/zion/memory/paging_util.h
@@ -4,5 +4,8 @@
 
 void InitializePml4(uint64_t pml4_physical_addr);
 
-void AllocatePageIfNecessary(uint64_t addr, uint64_t cr3 = 0);
+uint64_t AllocatePageIfNecessary(uint64_t addr, uint64_t cr3 = 0);
 void EnsureResident(uint64_t addr, uint64_t size);
+
+void CopyIntoNonResidentProcess(uint64_t base, uint64_t size, uint64_t dest_cr3,
+                                uint64_t dest_virt);
diff --git a/zion/scheduler/process.cpp b/zion/scheduler/process.cpp
index 32f6a08..07c7282 100644
--- a/zion/scheduler/process.cpp
+++ b/zion/scheduler/process.cpp
@@ -22,14 +22,13 @@ SharedPtr<Process> Process::RootProcess() {
   return proc;
 }
 
-Process::Process(uint64_t elf_ptr) : id_(gNextId++), state_(RUNNING) {
+Process::Process() : id_(gNextId++), state_(RUNNING) {
   cr3_ = phys_mem::AllocatePage();
   InitializePml4(cr3_);
-  CreateThread(elf_ptr);
 }
 
-void Process::CreateThread(uint64_t elf_ptr) {
-  Thread* thread = new Thread(this, next_thread_id_++, elf_ptr);
+void Process::CreateThread(uint64_t entry) {
+  Thread* thread = new Thread(this, next_thread_id_++, entry);
   threads_.PushBack(thread);
   gScheduler->Enqueue(thread);
 }
diff --git a/zion/scheduler/process.h b/zion/scheduler/process.h
index 3014b3e..80a669d 100644
--- a/zion/scheduler/process.h
+++ b/zion/scheduler/process.h
@@ -17,12 +17,12 @@ class Process {
     FINISHED,
   };
   static SharedPtr<Process> RootProcess();
-  Process(uint64_t elf_ptr);
+  Process();
 
   uint64_t id() const { return id_; }
   uint64_t cr3() const { return cr3_; }
 
-  void CreateThread(uint64_t elf_ptr);
+  void CreateThread(uint64_t entry);
   SharedPtr<Thread> GetThread(uint64_t tid);
 
   // Checks the state of all child threads and transitions to
diff --git a/zion/scheduler/thread.cpp b/zion/scheduler/thread.cpp
index 4851478..b1f6af5 100644
--- a/zion/scheduler/thread.cpp
+++ b/zion/scheduler/thread.cpp
@@ -23,8 +23,8 @@ SharedPtr<Thread> Thread::RootThread(Process* root_proc) {
   return new Thread(root_proc);
 }
 
-Thread::Thread(const SharedPtr<Process>& proc, uint64_t tid, uint64_t elf_ptr)
-    : process_(proc), id_(tid), elf_ptr_(elf_ptr) {
+Thread::Thread(const SharedPtr<Process>& proc, uint64_t tid, uint64_t entry)
+    : process_(proc), id_(tid), rip_(entry) {
   uint64_t* stack = new uint64_t[512];
   uint64_t* stack_ptr = stack + 511;
   // 0: rip
@@ -42,12 +42,11 @@ Thread::Thread(const SharedPtr<Process>& proc, uint64_t tid, uint64_t elf_ptr)
 uint64_t Thread::pid() { return process_->id(); }
 
 void Thread::Init() {
-  dbgln("[%u.%u]", pid(), id_);
-  uint64_t rip = LoadElfProgram(elf_ptr_, 0);
+  dbgln("[%u.%u] thread start.", pid(), id_);
   uint64_t rsp = 0x80000000;
   EnsureResident(rsp - 1, 1);
   SetRsp0(rsp0_start_);
-  jump_user_space(rip, rsp);
+  jump_user_space(rip_, rsp);
 }
 
 void Thread::Exit() {
diff --git a/zion/scheduler/thread.h b/zion/scheduler/thread.h
index da8036b..2f5e102 100644
--- a/zion/scheduler/thread.h
+++ b/zion/scheduler/thread.h
@@ -11,16 +11,13 @@ class Thread {
  public:
   enum State {
     UNSPECIFIED,
-    CREATED,
     RUNNING,
     RUNNABLE,
-    BLOCKED,
     FINISHED,
   };
   static SharedPtr<Thread> RootThread(Process* root_proc);
 
-  explicit Thread(const SharedPtr<Process>& proc, uint64_t tid,
-                  uint64_t elf_ptr);
+  explicit Thread(const SharedPtr<Process>& proc, uint64_t tid, uint64_t entry);
 
   uint64_t tid() { return id_; };
   uint64_t pid();
@@ -45,7 +42,8 @@ class Thread {
   uint64_t id_;
   State state_ = RUNNABLE;
 
-  uint64_t elf_ptr_;
+  // Startup Context for the thread.
+  uint64_t rip_;
 
   // Stack pointer to take on resume.
   // Stack will contain the full thread context.