cmd/gc: add GOEXPERIMENT=zerostack to clear stack on function entry

This is expensive but it might be useful in cases where
people are suffering from false positives during garbage
collection and are willing to trade the CPU time for getting
rid of the false positives.

On the other hand it only eliminates false positives caused
by other function calls, not false positives caused by dead
temporaries stored in the current function call.

The 5g/6g/8g changes were pulled out of the history, from
the last time we needed to do this (to work around a goto bug).
The code in go.h, lex.c, pgen.c is new but tiny.

R=ken2
CC=golang-dev
https://golang.org/cl/6938073

src/cmd/5g/gsubr.c

@@ -174,6 +174,64 @@ newplist(void)
 	return pl;
 }
 
+void
+clearstk(void)
+{
+	Plist *pl;
+	Prog *p, *p1, *p2, *p3;
+	Node dst, end, zero, con;
+
+	if(plast->firstpc->to.offset <= 0)
+		return;
+
+	// reestablish context for inserting code
+	// at beginning of function.
+	pl = plast;
+	p1 = pl->firstpc;
+	p2 = p1->link;
+	pc = mal(sizeof(*pc));
+	clearp(pc);
+	p1->link = pc;
+	
+	// zero stack frame
+
+	// MOVW $4(SP), R1
+	nodreg(&dst, types[tptr], 1);
+	p = gins(AMOVW, N, &dst);
+	p->from.type = D_CONST;
+	p->from.reg = REGSP;
+	p->from.offset = 4;
+
+	// MOVW $n(R1), R2
+	nodreg(&end, types[tptr], 2);
+	p = gins(AMOVW, N, &end);
+	p->from.type = D_CONST;
+	p->from.reg = 1;
+	p->from.offset = p1->to.offset;
+	
+	// MOVW $0, R3
+	nodreg(&zero, types[TUINT32], 3);
+	nodconst(&con, types[TUINT32], 0);
+	gmove(&con, &zero);
+
+	// L:
+	//	MOVW.P R3, 0(R1) +4
+	//	CMP R1, R2
+	//	BNE L
+	p = gins(AMOVW, &zero, &dst);
+	p->to.type = D_OREG;
+	p->to.offset = 4;
+	p->scond |= C_PBIT;
+	p3 = p;
+	p = gins(ACMP, &dst, N);
+	raddr(&end, p);
+	patch(gbranch(ABNE, T, 0), p3);
+
+	// continue with original code.
+	gins(ANOP, N, N)->link = p2;
+	pc = P;
+}	
+
 void
 gused(Node *n)
 {

src/cmd/6g/gsubr.c

@@ -172,6 +172,44 @@ newplist(void)
 	return pl;
 }
 
+void
+clearstk(void)
+{
+	Plist *pl;
+	Prog *p1, *p2;
+	Node sp, di, cx, con, ax;
+
+	if((uint32)plast->firstpc->to.offset <= 0)
+		return;
+
+	// reestablish context for inserting code
+	// at beginning of function.
+	pl = plast;
+	p1 = pl->firstpc;
+	p2 = p1->link;
+	pc = mal(sizeof(*pc));
+	clearp(pc);
+	p1->link = pc;
+	
+	// zero stack frame
+	nodreg(&sp, types[tptr], D_SP);
+	nodreg(&di, types[tptr], D_DI);
+	nodreg(&cx, types[TUINT64], D_CX);
+	nodconst(&con, types[TUINT64], (uint32)p1->to.offset / widthptr);
+	gins(ACLD, N, N);
+	gins(AMOVQ, &sp, &di);
+	gins(AMOVQ, &con, &cx);
+	nodconst(&con, types[TUINT64], 0);
+	nodreg(&ax, types[TUINT64], D_AX);
+	gins(AMOVQ, &con, &ax);
+	gins(AREP, N, N);
+	gins(ASTOSQ, N, N);
+
+	// continue with original code.
+	gins(ANOP, N, N)->link = p2;
+	pc = P;
+}	
+
 void
 gused(Node *n)
 {

src/cmd/8g/gsubr.c

@@ -173,6 +173,44 @@ newplist(void)
 	return pl;
 }
 
+void
+clearstk(void)
+{
+	Plist *pl;
+	Prog *p1, *p2;
+	Node sp, di, cx, con, ax;
+
+	if(plast->firstpc->to.offset <= 0)
+		return;
+
+	// reestablish context for inserting code
+	// at beginning of function.
+	pl = plast;
+	p1 = pl->firstpc;
+	p2 = p1->link;
+	pc = mal(sizeof(*pc));
+	clearp(pc);
+	p1->link = pc;
+	
+	// zero stack frame
+	nodreg(&sp, types[tptr], D_SP);
+	nodreg(&di, types[tptr], D_DI);
+	nodreg(&cx, types[TUINT32], D_CX);
+	nodconst(&con, types[TUINT32], p1->to.offset / widthptr);
+	gins(ACLD, N, N);
+	gins(AMOVL, &sp, &di);
+	gins(AMOVL, &con, &cx);
+	nodconst(&con, types[TUINT32], 0);
+	nodreg(&ax, types[TUINT32], D_AX);
+	gins(AMOVL, &con, &ax);
+	gins(AREP, N, N);
+	gins(ASTOSL, N, N);
+
+	// continue with original code.
+	gins(ANOP, N, N)->link = p2;
+	pc = P;
+}	
+
 void
 gused(Node *n)
 {

src/cmd/gc/go.h

@@ -937,6 +937,7 @@ EXTERN	int	funcdepth;
 EXTERN	int	typecheckok;
 EXTERN	int	compiling_runtime;
 EXTERN	int	compiling_wrappers;
+EXTERN	int	zerostack_enabled;
 
 EXTERN	int	nointerface;
 EXTERN	int	fieldtrack_enabled;
@@ -1092,6 +1093,7 @@ void	genlist(NodeList *l);
 Node*	sysfunc(char *name);
 void	tempname(Node *n, Type *t);
 Node*	temp(Type*);
+void	clearstk(void);
 
 /*
  *	init.c

src/cmd/gc/lex.c

@@ -41,6 +41,7 @@ static struct {
 } exper[] = {
 //	{"rune32", &rune32},
 	{"fieldtrack", &fieldtrack_enabled},
+	{"zerostack", &zerostack_enabled},
 	{nil, nil},
 };
 

src/cmd/gc/pgen.c

@@ -141,6 +141,9 @@ compile(Node *fn)
 	if(0)
 		frame(0);
 
+	if(zerostack_enabled)
+		clearstk();
+
 ret:
 	lineno = lno;
 }