diff --git a/src/pkg/runtime/386/asm.s b/src/pkg/runtime/386/asm.s
index 40cef0c1c2..bafef01c7e 100644
--- a/src/pkg/runtime/386/asm.s
+++ b/src/pkg/runtime/386/asm.s
@@ -78,7 +78,7 @@ ok:
 	CALL	runtime·schedinit(SB)
 
 	// create a new goroutine to start program
-	PUSHL	$runtime·mainstart(SB)	// entry
+	PUSHL	$runtime·main(SB)	// entry
 	PUSHL	$0	// arg size
 	CALL	runtime·newproc(SB)
 	POPL	AX
@@ -90,15 +90,6 @@ ok:
 	INT $3
 	RET
 
-TEXT runtime·mainstart(SB),7,$0
-	CALL	main·init(SB)
-	CALL	main·main(SB)
-	PUSHL	$0
-	CALL	runtime·exit(SB)
-	POPL	AX
-	INT $3
-	RET
-
 TEXT runtime·breakpoint(SB),7,$0
 	INT $3
 	RET
diff --git a/src/pkg/runtime/amd64/asm.s b/src/pkg/runtime/amd64/asm.s
index f34ce4a883..866e083106 100644
--- a/src/pkg/runtime/amd64/asm.s
+++ b/src/pkg/runtime/amd64/asm.s
@@ -60,7 +60,7 @@ ok:
 	CALL	runtime·schedinit(SB)
 
 	// create a new goroutine to start program
-	PUSHQ	$runtime·mainstart(SB)		// entry
+	PUSHQ	$runtime·main(SB)		// entry
 	PUSHQ	$0			// arg size
 	CALL	runtime·newproc(SB)
 	POPQ	AX
@@ -72,15 +72,6 @@ ok:
 	CALL	runtime·notok(SB)		// never returns
 	RET
 
-TEXT runtime·mainstart(SB),7,$0
-	CALL	main·init(SB)
-	CALL	main·main(SB)
-	PUSHQ	$0
-	CALL	runtime·exit(SB)
-	POPQ	AX
-	CALL	runtime·notok(SB)
-	RET
-
 TEXT runtime·breakpoint(SB),7,$0
 	BYTE	$0xcc
 	RET
diff --git a/src/pkg/runtime/arm/asm.s b/src/pkg/runtime/arm/asm.s
index a427a465f9..c03422f10c 100644
--- a/src/pkg/runtime/arm/asm.s
+++ b/src/pkg/runtime/arm/asm.s
@@ -43,7 +43,7 @@ TEXT _rt0_arm(SB),7,$-4
 	BL	runtime·schedinit(SB)
 
 	// create a new goroutine to start program
-	MOVW	$runtime·mainstart(SB), R0
+	MOVW	$runtime·main(SB), R0
 	MOVW.W	R0, -4(R13)
 	MOVW	$8, R0
 	MOVW.W	R0, -4(R13)
@@ -60,20 +60,6 @@ TEXT _rt0_arm(SB),7,$-4
 	MOVW	R0, (R1)	// fail hard
 	B	runtime·_dep_dummy(SB)	// Never reached
 
-
-TEXT runtime·mainstart(SB),7,$4
-	BL	main·init(SB)
-	EOR	R0, R0
-	MOVW	R0, 0(R13)
-	BL	main·main(SB)
-	MOVW	$0, R0
-	MOVW	R0, 4(SP)
-	BL	runtime·exit(SB)
-	MOVW	$1234, R0
-	MOVW	$1001, R1
-	MOVW	R0, (R1)	// fail hard
-	RET
-
 // TODO(kaib): remove these once i actually understand how the linker removes symbols
 // pull in dummy dependencies
 TEXT runtime·_dep_dummy(SB),7,$0
diff --git a/src/pkg/runtime/debug.go b/src/pkg/runtime/debug.go
index 6370a57d80..124370384c 100644
--- a/src/pkg/runtime/debug.go
+++ b/src/pkg/runtime/debug.go
@@ -10,7 +10,6 @@ func Breakpoint()
 // LockOSThread wires the calling goroutine to its current operating system thread.
 // Until the calling goroutine exits or calls UnlockOSThread, it will always
 // execute in that thread, and no other goroutine can.
-// LockOSThread cannot be used during init functions.
 func LockOSThread()
 
 // UnlockOSThread unwires the calling goroutine from its fixed operating system thread.
diff --git a/src/pkg/runtime/proc.c b/src/pkg/runtime/proc.c
index 3408c06f69..d51e3d3522 100644
--- a/src/pkg/runtime/proc.c
+++ b/src/pkg/runtime/proc.c
@@ -72,6 +72,9 @@ struct Sched {
 	volatile uint32 atomic;	// atomic scheduling word (see below)
 
 	int32 profilehz;	// cpu profiling rate
+	
+	bool init;  // running initialization
+	bool lockmain;  // init called runtime.LockOSThread
 
 	Note	stopped;	// one g can set waitstop and wait here for m's to stop
 };
@@ -171,11 +174,7 @@ setmcpumax(uint32 n)
 //	make & queue new G
 //	call runtime·mstart
 //
-// The new G does:
-//
-//	call main·init_function
-//	call initdone
-//	call main·main
+// The new G calls runtime·main.
 void
 runtime·schedinit(void)
 {
@@ -212,6 +211,32 @@ runtime·schedinit(void)
 	m->nomemprof--;
 }
 
+extern void main·init(void);
+extern void main·main(void);
+
+// The main goroutine.
+void
+runtime·main(void)
+{
+	// Lock the main goroutine onto this, the main OS thread,
+	// during initialization.  Most programs won't care, but a few
+	// do require certain calls to be made by the main thread.
+	// Those can arrange for main.main to run in the main thread
+	// by calling runtime.LockOSThread during initialization
+	// to preserve the lock.
+	runtime·LockOSThread();
+	runtime·sched.init = true;
+	main·init();
+	runtime·sched.init = false;
+	if(!runtime·sched.lockmain)
+		runtime·UnlockOSThread();
+
+	main·main();
+	runtime·exit(0);
+	for(;;)
+		*(int32*)runtime·main = 0;
+}
+
 // Lock the scheduler.
 static void
 schedlock(void)
@@ -1494,13 +1519,6 @@ runtime·Gosched(void)
 	runtime·gosched();
 }
 
-void
-runtime·LockOSThread(void)
-{
-	m->lockedg = g;
-	g->lockedm = m;
-}
-
 // delete when scheduler is stronger
 int32
 runtime·gomaxprocsfunc(int32 n)
@@ -1540,9 +1558,24 @@ runtime·gomaxprocsfunc(int32 n)
 	return ret;
 }
 
+void
+runtime·LockOSThread(void)
+{
+	if(m == &runtime·m0 && runtime·sched.init) {
+		runtime·sched.lockmain = true;
+		return;
+	}
+	m->lockedg = g;
+	g->lockedm = m;
+}
+
 void
 runtime·UnlockOSThread(void)
 {
+	if(m == &runtime·m0 && runtime·sched.init) {
+		runtime·sched.lockmain = false;
+		return;
+	}
 	m->lockedg = nil;
 	g->lockedm = nil;
 }
diff --git a/src/pkg/runtime/runtime.h b/src/pkg/runtime/runtime.h
index e3ec197757..e45808f8e0 100644
--- a/src/pkg/runtime/runtime.h
+++ b/src/pkg/runtime/runtime.h
@@ -637,6 +637,8 @@ String	runtime·signame(int32 sig);
 int32	runtime·gomaxprocsfunc(int32 n);
 void	runtime·procyield(uint32);
 void	runtime·osyield(void);
+void	runtime·LockOSThread(void);
+void	runtime·UnlockOSThread(void);
 
 void	runtime·mapassign(MapType*, Hmap*, byte*, byte*);
 void	runtime·mapaccess(MapType*, Hmap*, byte*, byte*, bool*);