diff --git a/sys/SConscript b/sys/SConscript
index 6574f5b..f1545e5 100644
--- a/sys/SConscript
+++ b/sys/SConscript
@@ -47,7 +47,7 @@ src_arm64 = [
     "arm64/mp.c",
     "arm64/pci.c",
     "arm64/pmap.c",
-    "arm64/support.S",
+    "arm64/copy.c",
     "arm64/switch.S",
     "arm64/thread.c",
     "arm64/timer.c",
diff --git a/sys/arm64/copy.c b/sys/arm64/copy.c
new file mode 100644
index 0000000..2c29af8
--- /dev/null
+++ b/sys/arm64/copy.c
@@ -0,0 +1,14 @@
+#include <stdint.h>
+#include <string.h>
+
+int copy_unsafe(void *to_addr, void *from_addr, uintptr_t len)
+{
+    memcpy(to_addr, from_addr, len);
+    return 0;
+}
+
+int copystr_unsafe(void *to_addr, void *from_addr, uintptr_t len)
+{
+    memcpy(to_addr, from_addr, len);
+    return 0;
+}
\ No newline at end of file
diff --git a/sys/arm64/include/paging.h b/sys/arm64/include/paging.h
index eafd2c9..68f2d9f 100644
--- a/sys/arm64/include/paging.h
+++ b/sys/arm64/include/paging.h
@@ -68,7 +68,7 @@ _Static_assert(sizeof(struct vmpt) == PGSIZE);
 // #define VMPD_ATTR_CA_DEVICE (VMPD_ATTR_CA_MAKE(1ull))
 
 #define MTP_KERNEL (0b0010 | (0b0000 << 4) | (0b0000 << 8) | (0b0000 << 12))
-#define MTP_USER (0b0010 | (0b0000 << 4) | (0b0011 << 8) | (0b0010 << 12))
+#define MTP_USER (0b0010 | (0b0000 << 4) | (0b0000 << 8) | (0b0000 << 12))
 
 struct vmpd {
     paddr_t next; // the next hppte in the list
diff --git a/sys/arm64/include/trap.h b/sys/arm64/include/trap.h
index bcc1e62..9e1d3b7 100644
--- a/sys/arm64/include/trap.h
+++ b/sys/arm64/include/trap.h
@@ -121,5 +121,7 @@ typedef struct TrapFrame
 
 void Trap_Pop(struct TrapFrame * tf);
 
+#define SPSR_EL1H_DAIF0 (0b101)
+
 #endif /* __TRAP_H__ */
 
diff --git a/sys/arm64/machine.c b/sys/arm64/machine.c
index 31c0b05..eeb4fa2 100644
--- a/sys/arm64/machine.c
+++ b/sys/arm64/machine.c
@@ -148,11 +148,7 @@ void Machine_Init()
     Sched_SetRunnable(thr);
 
     Critical_Exit();
-    
-    while(1) {
-        kprintf("Main thread!\n");
-        hlt();
-    }
+
     /*
      * Load the init processor
      */
diff --git a/sys/arm64/mrt/priv.c b/sys/arm64/mrt/priv.c
index 1425f3b..6b28d72 100644
--- a/sys/arm64/mrt/priv.c
+++ b/sys/arm64/mrt/priv.c
@@ -19,9 +19,8 @@ IMPL_MRT(mrt_kenter)
         WMCR(MCREG_MTP, kmtp);
 
         // switch to kernel stack
-        struct CPUState * state = (struct CPUState *)SYSREG_GET(tpidr_el1);
-        const uint32_t id = state->id;
-        kstack_top = curProc[id]->kstack + PGSIZE;
+        const int coreid = SYSREG_GET(tpidr_el1);
+        kstack_top = curProc[coreid]->kstack + PGSIZE;
 
         const regval_t kcpl = MCPL_KERNEL;
         WMCR(MGREG_MCPL, kcpl);
diff --git a/sys/arm64/support.S b/sys/arm64/support.S
deleted file mode 100644
index 3949db0..0000000
--- a/sys/arm64/support.S
+++ /dev/null
@@ -1,26 +0,0 @@
-/*
- * Support Functions
- */
-
-#include <errno.h>
-#include <machine/asm.h>
-
-.text
-
-// copy_unsafe(to, from, len)
-FUNC_BEGIN(copy_unsafe)
-.globl copy_unsafe_done
-copy_unsafe_done:
-.globl copy_unsafe_fault
-copy_unsafe_fault:
-FUNC_END(copy_unsafe)
-
-// copystr_unsafe(to, from, len)
-FUNC_BEGIN(copystr_unsafe)
-.globl copystr_unsafe_done
-copystr_unsafe_done:
-.globl copystr_unsafe_fault
-copystr_unsafe_fault:
-.globl copystr_unsafe_toolong
-FUNC_END(copystr_unsafe)
-
diff --git a/sys/arm64/thread.c b/sys/arm64/thread.c
index 9a1df5b..529d5f3 100644
--- a/sys/arm64/thread.c
+++ b/sys/arm64/thread.c
@@ -27,12 +27,9 @@ void
 ThreadKThreadEntry(TrapFrame *tf) __NO_LOCK_ANALYSIS
 {
     Spinlock_Unlock(&schedLock);
-    kprintf("tf in entry = 0x%p, elr = 0x%lx\n", tf, tf->elr);
     Trap_Pop(tf);
 }
 
-#define SPSR_EL1H_DAIF0 (0b101)
-
 void
 Thread_SetupKThread(Thread *thr, kthread_entry f,
 		    uintptr_t arg1,  uintptr_t arg2,  uintptr_t arg3)
@@ -53,7 +50,6 @@ Thread_SetupKThread(Thread *thr, kthread_entry f,
     sf->lr = (uint64_t)&ThreadKThreadEntry;
     sf->r0 = (uint64_t)tf;
     sf->_r0 = (uint64_t)tf;
-    kprintf("tf before entry = 0x%p\n", tf);
 
     // interrupt enable
     tf->spsr = SPSR_EL1H_DAIF0; // el1h, enable interrupts
@@ -90,7 +86,7 @@ Thread_SetupUThread(Thread *thr, uintptr_t rip, uintptr_t arg)
 void
 Thread_SwitchArch(Thread *oldthr, Thread *newthr)
 {
-    kprintf("switch stack! oldthr sp = 0x%lx, newthr sp = 0x%lx\n", oldthr->arch.sp, newthr->arch.sp);
+    //kprintf("switch stack! oldthr sp = 0x%lx, newthr sp = 0x%lx\n", oldthr->arch.sp, newthr->arch.sp);
     // Jump to trapframe
     switchstack(&oldthr->arch.sp, newthr->arch.sp);
 }
diff --git a/sys/arm64/trap.c b/sys/arm64/trap.c
index 8bd7bfc..cf3e02c 100644
--- a/sys/arm64/trap.c
+++ b/sys/arm64/trap.c
@@ -4,6 +4,7 @@
 #include <sys/kassert.h>
 #include <sys/irq.h>
 #include <sys/kdebug.h>
+#include <sys/syscall.h>
 
 #include <machine/cpuop.h>
 #include <machine/gic.h>
@@ -104,9 +105,18 @@ trap_entry(TrapFrame * tf)
 {
     // kprintf("Trap: psp = 0x%lx, pcpl = %lu\n", tf->psp, tf->pcpl);
     if (tf->type == T_TYPE_EXC) {
-        Panic("Unexpected exception: Type = 0x%lx, IntID = 0x%lx, ESR = 0x%lx, FAR = 0x%lx, ELR = 0x%lx, SPSR = 0x%lx.\n", tf->type, tf->intid, tf->esr, tf->far, tf->elr, tf->spsr);
+        const uint64_t ec = (tf->esr >> 26) & 0b111111;
+        switch (ec) {
+            case 0b010101: {
+                tf->r0 = Syscall_Entry(tf->r0, tf->r1, tf->r2, tf->r3, tf->r4, tf->r5);
+                break;
+            }
+            default: {
+                Panic("Unexpected exception: Type = 0x%lx, IntID = 0x%lx, ESR = 0x%lx, FAR = 0x%lx, ELR = 0x%lx, SPSR = 0x%lx.\n", tf->type, tf->intid, tf->esr, tf->far, tf->elr, tf->spsr);
+            }
+        }    
     } else {
-        kprintf("IRQ: Type = 0x%lx, IntID = 0x%lx, ESR = 0x%lx, FAR = 0x%lx, ELR = 0x%lx, SPSR = 0x%lx.\n", tf->type, tf->intid, tf->esr, tf->far, tf->elr, tf->spsr);
+        // kprintf("IRQ: Type = 0x%lx, IntID = 0x%lx, ESR = 0x%lx, FAR = 0x%lx, ELR = 0x%lx, SPSR = 0x%lx.\n", tf->type, tf->intid, tf->esr, tf->far, tf->elr, tf->spsr);
         // irq
         IRQ_Handler(tf->intid);
     }
diff --git a/sys/include/elf_common.h b/sys/include/elf_common.h
index c9292e2..ee8de08 100644
--- a/sys/include/elf_common.h
+++ b/sys/include/elf_common.h
@@ -237,6 +237,7 @@ typedef struct {
 #define	EM_ARCA		109	/* Arca RISC Microprocessor. */
 #define	EM_UNICORE	110	/* Microprocessor series from PKU-Unity Ltd.
 				   and MPRC of Peking University */
+#define EM_AARCH64 	(0xb7)
 
 /* Non-standard or deprecated. */
 #define	EM_486		6	/* Intel i486. */
diff --git a/sys/kern/loader.c b/sys/kern/loader.c
index 3640c15..eaadbc1 100644
--- a/sys/kern/loader.c
+++ b/sys/kern/loader.c
@@ -17,6 +17,7 @@
 
 #include <machine/trap.h>
 #include <machine/pmap.h>
+#include <machine/metal.h>
 #include <sys/thread.h>
 #include <sys/spinlock.h>
 #include <sys/loader.h>
@@ -43,10 +44,15 @@ Loader_CheckHeader(const Elf64_Ehdr *ehdr)
 	return false;
     }
 
+#if defined (__x86_64__)
     if (ehdr->e_machine != EM_AMD64) {
-	return false;
+	    return false;
     }
-
+#elif defined(__aarch64__)
+    if (ehdr->e_machine != EM_AARCH64) {
+	    return false;
+    }
+#endif
     return true;
 }
 
@@ -227,7 +233,9 @@ Loader_LoadInit()
     /*
      * Load init binary
      */
-    Loader_Load(thr, initvn, pg, 1024);
+    if (!Loader_Load(thr, initvn, pg, 1024)) {
+        Panic("Failed to load init.\n");
+    }
 
     VFS_Close(initvn);
 
@@ -275,9 +283,12 @@ Loader_LoadInit()
     TrapFrame tf;
     memset(&tf, 0, sizeof(tf));
     tf.elr = thr->proc->entrypoint;
+    tf.spsr = SPSR_EL1H_DAIF0;
+    tf.psp = sp;
+    tf.pcpl = MCPL_USER;
     tf.r0 = sp;
 
-    //Trap_Pop(&tf);
+    Trap_Pop(&tf);
 #endif
 
     /*
diff --git a/sys/kern/sched.c b/sys/kern/sched.c
index 803eee5..269fd47 100644
--- a/sys/kern/sched.c
+++ b/sys/kern/sched.c
@@ -214,7 +214,6 @@ Sched_Scheduler()
     curProc[CPU()] = next;
     next->schedState = SCHED_STATE_RUNNING;
     next->ctxSwitches++;
-    kprintf("sched: cur = %lu next = %lu\n", prev->tid, next->tid);
 
     if (prev->schedState == SCHED_STATE_RUNNING) {
 	prev->schedState = SCHED_STATE_RUNNABLE;