diff --git a/start.c b/start.c
index 7f72faf..d0b47c5 100644
--- a/start.c
+++ b/start.c
@@ -18,9 +18,14 @@ char stack0[4096 * NCPU] __attribute__((aligned(16)));
 /* Keep this here and sync on it until we have synchronized printf */
 spinlock_t   sl = {0};
 volatile int hold = 1;
+volatile int max_hart = 0;
 
 /* This is where entry.S drops us of. All cores land here */
 void start() {
+    // Do this first
+    __sync_fetch_and_add(&max_hart, 1);
+    __sync_synchronize();
+
     u64 id = read_mhartid();
 
     // Keep each CPU's hartid in its tp (thread pointer) register, for cpuid().
@@ -35,15 +40,29 @@ void start() {
         uart_puts("Hello Neptune!\n");
         spinlock_init(&sl);
         hold = 0;
+    } else {
+        while (hold);
     }
 
-    while (hold);
+    // spin_lock(&sl);
+    //
+    // uart_puts("Hart number: ");
+    // uart_putc(id + '0');
+    // uart_putc('\n');
+    //
+    // spin_unlock(&sl);
 
-    spin_lock(&sl);
-    uart_puts("Hart number: ");
-    uart_putc(id + '0');
-    uart_putc('\n');
-    spin_unlock(&sl);
+    if (id == 0) {
+        spin_lock(&sl);
+        uart_puts("Core count: ");
+        uart_putc(max_hart + '0');
+        uart_putc('\n');
+        if (max_hart == NCPU) {
+            uart_puts("All cores up!");
+            uart_putc('\n');
+        }
+        spin_unlock(&sl);
+    }
 
     // We should not arrive here, but if we do, hang in a while on wfi.
     while (1) __asm__ volatile("wfi"); // (Wait For Interrupt)