runtime: use gp->sched.sp for stack overflow check

On x86 it is a few words lower on the stack than m->morebuf.sp
so it is a more precise check. Enabling the check requires recording
a valid gp->sched in reflect.call too. This is a good thing in general,
since it will make stack traces during reflect.call work better, and it
may be useful for preemption too.

R=dvyukov
CC=golang-dev
https://golang.org/cl/10709043

src/pkg/runtime/asm_386.s

@@ -254,6 +254,11 @@ TEXT reflect·call(SB), 7, $0
 	MOVL	g(CX), AX
 	MOVL	AX, (m_morebuf+gobuf_g)(BX)
 
+	// Save our own state as the PC and SP to restore
+	// if this goroutine needs to be restarted.
+	MOVL	$reflect·call(SB), (g_sched+gobuf_pc)(AX)
+	MOVL	SP, (g_sched+gobuf_sp)(AX)
+
 	// Set up morestack arguments to call f on a new stack.
 	// We set f's frame size to 1, as a hint to newstack
 	// that this is a call from reflect·call.

src/pkg/runtime/asm_amd64.s

@@ -231,6 +231,11 @@ TEXT reflect·call(SB), 7, $0
 	MOVQ	AX, (m_morebuf+gobuf_sp)(BX)
 	MOVQ	g(CX), AX
 	MOVQ	AX, (m_morebuf+gobuf_g)(BX)
+	
+	// Save our own state as the PC and SP to restore
+	// if this goroutine needs to be restarted.
+	MOVQ	$reflect·call(SB), (g_sched+gobuf_pc)(AX)
+	MOVQ	SP, (g_sched+gobuf_sp)(AX)
 
 	// Set up morestack arguments to call f on a new stack.
 	// We set f's frame size to 1, as a hint to newstack

src/pkg/runtime/asm_arm.s

@@ -207,6 +207,13 @@ TEXT reflect·call(SB), 7, $-4
 	MOVW	SP, (m_morebuf+gobuf_sp)(m)	// our caller's SP
 	MOVW	g,  (m_morebuf+gobuf_g)(m)
 
+	// Save our own state as the PC and SP to restore
+	// if this goroutine needs to be restarted.
+	MOVW	$reflect·call(SB), R11
+	MOVW	R11, (g_sched+gobuf_pc)(g)
+	MOVW	LR, (g_sched+gobuf_lr)(g)
+	MOVW	SP, (g_sched+gobuf_sp)(g)
+
 	// Set up morestack arguments to call f on a new stack.
 	// We set f's frame size to 1, as a hint to newstack
 	// that this is a call from reflect·call.

src/pkg/runtime/stack.c

@@ -215,7 +215,7 @@ runtime·newstack(void)
 	if(!reflectcall)
 		runtime·rewindmorestack(&gp->sched);
 
-	sp = m->morebuf.sp;
+	sp = gp->sched.sp;
 	if(thechar == '6' || thechar == '8') {
 		// The call to morestack cost a word.
 		sp -= sizeof(uintptr);