diff --git a/kernel/mem.c b/kernel/mem.c
index 6824e628..1fc11a10 100644
--- a/kernel/mem.c
+++ b/kernel/mem.c
@@ -265,6 +265,13 @@ void remote_flush_tlb_cpumask(struct process_vm *vm,
 		unsigned long tsc;
 		tsc = rdtsc() + 12884901888;  /* 1.2GHz =>10 sec */
 #endif
+		if (flush_entry->addr) {
+			flush_tlb_single(flush_entry->addr & PAGE_MASK);
+		}
+		/* Zero address denotes full TLB flush */
+		else {
+			flush_tlb();
+		}
 
 		/* Wait for all cores */
 		while (ihk_atomic_read(&flush_entry->pending) != 0) {
diff --git a/kernel/syscall.c b/kernel/syscall.c
index bd9eec9f..90ce79c1 100644
--- a/kernel/syscall.c
+++ b/kernel/syscall.c
@@ -917,8 +917,6 @@ static int do_munmap(void *addr, size_t len)
 	begin_free_pages_pending();
 	error = remove_process_memory_range(cpu_local_var(current)->vm,
 			(intptr_t)addr, (intptr_t)addr+len, &ro_freed);
-	// XXX: TLB flush
-	flush_tlb();
 	if (error || !ro_freed) {
 		clear_host_pte((uintptr_t)addr, len);
 	}