diff --git a/kernel/include/profile.h b/kernel/include/profile.h
index 35a0fa92..ceee4a63 100644
--- a/kernel/include/profile.h
+++ b/kernel/include/profile.h
@@ -38,6 +38,7 @@ enum profile_event_type {
 	PROFILE_page_fault_file,
 	PROFILE_page_fault_dev_file,
 	PROFILE_page_fault_file_clr,
+	PROFILE_remote_page_fault,
 	PROFILE_mpol_alloc_missed,
 	PROFILE_mmap_anon_contig_phys,
 	PROFILE_mmap_anon_no_contig_phys,
diff --git a/kernel/mem.c b/kernel/mem.c
index c501ddc2..63d66b29 100644
--- a/kernel/mem.c
+++ b/kernel/mem.c
@@ -558,7 +558,7 @@ static void *mckernel_allocate_aligned_pages_node(int npages, int p2align,
 			}
 			else {
 #ifdef PROFILE_ENABLE
-				//profile_event_add(PROFILE_numa_alloc_missed, npages * 4096);
+				profile_event_add(PROFILE_mpol_alloc_missed, npages * 4096);
 #endif
 				dkprintf("%s: couldn't fulfill explicit NUMA request for %d pages\n",
 						__FUNCTION__, npages);
@@ -983,7 +983,10 @@ void remote_flush_tlb_array_cpumask(struct process_vm *vm,
 void tlb_flush_handler(int vector)
 {
 #ifdef PROFILE_ENABLE
-	unsigned long t_s = rdtsc();
+	unsigned long t_s = 0;
+	if (cpu_local_var(current)->profile) {
+		t_s = rdtsc();
+	}
 #endif // PROFILE_ENABLE
 	int flags = cpu_disable_interrupt_save();
 
@@ -1011,11 +1014,12 @@ void tlb_flush_handler(int vector)
 	cpu_restore_interrupt(flags);
 #ifdef PROFILE_ENABLE
 	{
-		unsigned long t_e = rdtsc();
-		profile_event_add(PROFILE_tlb_invalidate, (t_e - t_s));
-		if (cpu_local_var(current)->profile)
+		if (cpu_local_var(current)->profile) {
+			unsigned long t_e = rdtsc();
+			profile_event_add(PROFILE_tlb_invalidate, (t_e - t_s));
 			cpu_local_var(current)->profile_elapsed_ts +=
 				(t_e - t_s);
+		}
 	}
 #endif // PROFILE_ENABLE
 }
@@ -1025,8 +1029,9 @@ static void page_fault_handler(void *fault_addr, uint64_t reason, void *regs)
 	struct thread *thread = cpu_local_var(current);
 	int error;
 #ifdef PROFILE_ENABLE
-	uint64_t t_s;
-	t_s = rdtsc();
+	uint64_t t_s = 0;
+	if (thread->profile)
+		t_s = rdtsc();
 #endif // PROFILE_ENABLE
 
 	set_cputime(interrupt_from_user(regs)? 1: 2);
@@ -1091,7 +1096,8 @@ out:
 	check_need_resched();
 	set_cputime(0);
 #ifdef PROFILE_ENABLE
-	profile_event_add(PROFILE_page_fault, (rdtsc() - t_s));
+	if (thread->profile)
+		profile_event_add(PROFILE_page_fault, (rdtsc() - t_s));
 #endif // PROFILE_ENABLE
 	return;
 }
diff --git a/kernel/profile.c b/kernel/profile.c
index da3c355c..f2c0cfc2 100644
--- a/kernel/profile.c
+++ b/kernel/profile.c
@@ -63,6 +63,7 @@ char *profile_event_names[] =
 	"page_fault_file",
 	"page_fault_dev_file",
 	"page_fault_file_clr_mem",
+	"remote_page_fault",
 	"mpol_alloc_missed",
 	"mmap_anon_contig_phys",
 	"mmap_anon_no_contig_phys",
diff --git a/kernel/syscall.c b/kernel/syscall.c
index 0bc6f224..c4a738ca 100644
--- a/kernel/syscall.c
+++ b/kernel/syscall.c
@@ -182,8 +182,12 @@ long do_syscall(struct syscall_request *req, int cpu, int pid)
 	int mstatus = 0;
 
 #ifdef PROFILE_ENABLE
-	uint64_t t_s;
-	t_s = rdtsc();
+	/* We cannot use thread->profile_start_ts here because the
+	 * caller may be utilizing it already */
+	unsigned long t_s = 0;
+	if (thread->profile) {
+		t_s = rdtsc();
+	}
 #endif // PROFILE_ENABLE
 
 	dkprintf("SC(%d)[%3d] sending syscall\n",
@@ -273,6 +277,15 @@ long do_syscall(struct syscall_request *req, int cpu, int pid)
 		}
 
 		if (res.status == STATUS_PAGE_FAULT) {
+#ifdef PROFILE_ENABLE
+			/* We cannot use thread->profile_start_ts here because the
+			 * caller may be utilizing it already */
+			unsigned long t_s = 0;
+			if (thread->profile) {
+				t_s = rdtsc();
+			}
+#endif // PROFILE_ENABLE
+
 			dkprintf("STATUS_PAGE_FAULT in syscall, pid: %d\n", 
 					cpu_local_var(current)->proc->pid);
 			error = page_fault_process_vm(thread->vm,
@@ -291,6 +304,10 @@ long do_syscall(struct syscall_request *req, int cpu, int pid)
 
 			res.req_thread_status = IHK_SCD_REQ_THREAD_SPINNING;
 			send_syscall(&req2, cpu, pid, &res);
+#ifdef PROFILE_ENABLE
+			profile_event_add(PROFILE_remote_page_fault,
+					(rdtsc() - t_s));
+#endif // PROFILE_ENABLE
 		}
 
 		if (res.status == STATUS_SYACALL) {
@@ -9382,7 +9399,7 @@ long syscall(int num, ihk_mc_user_context_t *ctx)
 	}
 
 #ifdef PROFILE_ENABLE
-	if (thread->profile) {
+	{
 		unsigned long ts = rdtsc();
 
 		/*