all: merge NaCl branch (part 1)

See golang.org/s/go13nacl for design overview.

This CL is the mostly mechanical changes from rsc's Go 1.2 based NaCl branch, specifically 39cb35750369 to 500771b477cf from https://code.google.com/r/rsc-go13nacl. This CL does not include working NaCl support, there are probably two or three more large merges to come.

CL 15750044 is not included as it involves more invasive changes to the linker which will need to be merged separately.

The exact change lists included are

15050047: syscall: support for Native Client
15360044: syscall: unzip implementation for Native Client
15370044: syscall: Native Client SRPC implementation
15400047: cmd/dist, cmd/go, go/build, test: support for Native Client
15410048: runtime: support for Native Client
15410049: syscall: file descriptor table for Native Client
15410050: syscall: in-memory file system for Native Client
15440048: all: update +build lines for Native Client port
15540045: cmd/6g, cmd/8g, cmd/gc: support for Native Client
15570045: os: support for Native Client
15680044: crypto/..., hash/crc32, reflect, sync/atomic: support for amd64p32
15690044: net: support for Native Client
15690048: runtime: support for fake time like on Go Playground
15690051: build: disable various tests on Native Client

LGTM=rsc
R=rsc
CC=golang-codereviews
https://golang.org/cl/68150047

src/cmd/5g/galign.c

@@ -29,6 +29,7 @@ betypeinit(void)
 {
 	widthptr = 4;
 	widthint = 4;
+	widthreg = 4;
 
 	zprog.link = P;
 	zprog.as = AGOK;

src/cmd/6g/cgen.c

@@ -1417,23 +1417,23 @@ sgen(Node *n, Node *ns, int64 w)
 		// reverse direction
 		gins(ASTD, N, N);		// set direction flag
 		if(c > 0) {
-			gconreg(AADDQ, w-1, D_SI);
-			gconreg(AADDQ, w-1, D_DI);
+			gconreg(addptr, w-1, D_SI);
+			gconreg(addptr, w-1, D_DI);
 
-			gconreg(AMOVQ, c, D_CX);
+			gconreg(movptr, c, D_CX);
 			gins(AREP, N, N);	// repeat
 			gins(AMOVSB, N, N);	// MOVB *(SI)-,*(DI)-
 		}
 
 		if(q > 0) {
 			if(c > 0) {
-				gconreg(AADDQ, -7, D_SI);
-				gconreg(AADDQ, -7, D_DI);
+				gconreg(addptr, -7, D_SI);
+				gconreg(addptr, -7, D_DI);
 			} else {
-				gconreg(AADDQ, w-8, D_SI);
-				gconreg(AADDQ, w-8, D_DI);
+				gconreg(addptr, w-8, D_SI);
+				gconreg(addptr, w-8, D_DI);
 			}
-			gconreg(AMOVQ, q, D_CX);
+			gconreg(movptr, q, D_CX);
 			gins(AREP, N, N);	// repeat
 			gins(AMOVSQ, N, N);	// MOVQ *(SI)-,*(DI)-
 		}
@@ -1442,7 +1442,7 @@ sgen(Node *n, Node *ns, int64 w)
 	} else {
 		// normal direction
 		if(q >= 4) {
-			gconreg(AMOVQ, q, D_CX);
+			gconreg(movptr, q, D_CX);
 			gins(AREP, N, N);	// repeat
 			gins(AMOVSQ, N, N);	// MOVQ *(SI)+,*(DI)+
 		} else

src/cmd/6g/galign.c

@@ -12,6 +12,12 @@ LinkArch*	thelinkarch = &linkamd64;
 
 vlong MAXWIDTH = 1LL<<50;
 
+int	addptr = AADDQ;
+int	movptr = AMOVQ;
+int	leaptr = ALEAQ;
+int	stosptr = ASTOSQ;
+int	cmpptr = ACMPQ;
+
 /*
  * go declares several platform-specific type aliases:
  * int, uint, float, and uintptr
@@ -29,6 +35,20 @@ betypeinit(void)
 {
 	widthptr = 8;
 	widthint = 8;
+	widthreg = 8;
+	if(strcmp(getgoarch(), "amd64p32") == 0) {
+		widthptr = 4;
+		widthint = 4;
+		addptr = AADDL;
+		movptr = AMOVL;
+		leaptr = ALEAL;
+		stosptr = ASTOSL;
+		cmpptr = ACMPL;
+		typedefs[0].sameas = TINT32;
+		typedefs[1].sameas = TUINT32;
+		typedefs[2].sameas = TUINT32;
+		
+	}
 
 	zprog.link = P;
 	zprog.as = AGOK;

src/cmd/6g/gg.h

@@ -21,8 +21,14 @@ EXTERN	Node*	deferproc;
 EXTERN	Node*	deferreturn;
 EXTERN	Node*	panicindex;
 EXTERN	Node*	panicslice;
+EXTERN	Node*	panicdiv;
 EXTERN	Node*	throwreturn;
 extern	vlong	unmappedzero;
+extern	int	addptr;
+extern	int	cmpptr;
+extern	int	movptr;
+extern	int	leaptr;
+extern	int	stosptr;
 
 /*
  * ggen.c

src/cmd/6g/ggen.c

@@ -180,17 +180,21 @@ ginscall(Node *f, int proc)
 
 	case 1:	// call in new proc (go)
 	case 2:	// deferred call (defer)
-		nodreg(&reg, types[TINT64], D_CX);
-		if(flag_largemodel) {
-			regalloc(&r1, f->type, f);
+		nodconst(&con, types[TINT64], argsize(f->type));
+		if(widthptr == 4) {
+			nodreg(&r1, types[TINT32], D_CX);
 			gmove(f, &r1);
-			gins(APUSHQ, &r1, N);
-			regfree(&r1);
+			nodreg(&reg, types[TINT64], D_CX);
+			nodconst(&r1, types[TINT64], 32);
+			gins(ASHLQ, &r1, &reg);
+			gins(AORQ, &con, &reg);
+			gins(APUSHQ, &reg, N);
 		} else {
-			gins(APUSHQ, f, N);
+			nodreg(&reg, types[TINT64], D_CX);
+			gmove(f, &reg);
+			gins(APUSHQ, &reg, N);
+			gins(APUSHQ, &con, N);
 		}
-		nodconst(&con, types[TINT32], argsize(f->type));
-		gins(APUSHQ, &con, N);
 		if(proc == 1)
 			ginscall(newproc, 0);
 		else {
@@ -198,8 +202,10 @@ ginscall(Node *f, int proc)
 				fatal("hasdefer=0 but has defer");
 			ginscall(deferproc, 0);
 		}
+		nodreg(&reg, types[TINT64], D_CX);
 		gins(APOPQ, N, &reg);
-		gins(APOPQ, N, &reg);
+		if(widthptr == 8)
+			gins(APOPQ, N, &reg);
 		if(proc == 2) {
 			nodreg(&reg, types[TINT64], D_AX);
 			gins(ATESTQ, &reg, &reg);
@@ -387,11 +393,11 @@ cgen_aret(Node *n, Node *res)
 
 	if(res->op != OREGISTER) {
 		regalloc(&nod2, types[tptr], res);
-		gins(ALEAQ, &nod1, &nod2);
-		gins(AMOVQ, &nod2, res);
+		gins(leaptr, &nod1, &nod2);
+		gins(movptr, &nod2, res);
 		regfree(&nod2);
 	} else
-		gins(ALEAQ, &nod1, res);
+		gins(leaptr, &nod1, res);
 }
 
 /*
@@ -634,6 +640,18 @@ dodiv(int op, Node *nl, Node *nr, Node *res)
 	}
 
 	p2 = P;
+	if(nacl) {
+		// Native Client does not relay the divide-by-zero trap
+		// to the executing program, so we must insert a check
+		// for ourselves.
+		nodconst(&n4, t, 0);
+		gins(optoas(OCMP, t), &n3, &n4);
+		p1 = gbranch(optoas(ONE, t), T, +1);
+		if(panicdiv == N)
+			panicdiv = sysfunc("panicdivide");
+		ginscall(panicdiv, -1);
+		patch(p1, pc);
+	}
 	if(check) {
 		nodconst(&n4, t, -1);
 		gins(optoas(OCMP, t), &n3, &n4);
@@ -1045,10 +1063,10 @@ clearfat(Node *nl)
 	agen(nl, &n1);
 
 	savex(D_AX, &ax, &oldax, N, types[tptr]);
-	gconreg(AMOVQ, 0, D_AX);
+	gconreg(AMOVL, 0, D_AX);
 
 	if(q >= 4) {
-		gconreg(AMOVQ, q, D_CX);
+		gconreg(movptr, q, D_CX);
 		gins(AREP, N, N);	// repeat
 		gins(ASTOSQ, N, N);	// STOQ AL,*(DI)+
 	} else
@@ -1111,7 +1129,7 @@ expandchecks(Prog *firstp)
 		p2->lineno = p->lineno;
 		p1->pc = 9999;
 		p2->pc = 9999;
-		p->as = ACMPQ;
+		p->as = cmpptr;
 		p->to.type = D_CONST;
 		p->to.offset = 0;
 		p1->as = AJNE;

src/cmd/6g/gsubr.c

@@ -296,6 +296,11 @@ ginit(void)
 
 	for(i=0; i<nelem(resvd); i++)
 		reg[resvd[i]]++;
+	
+	if(nacl) {
+		reg[D_BP]++;
+		reg[D_R15]++;
+	}
 }
 
 void
@@ -305,6 +310,11 @@ gclean(void)
 
 	for(i=0; i<nelem(resvd); i++)
 		reg[resvd[i]]--;
+	if(nacl) {
+		reg[D_BP]--;
+		reg[D_R15]--;
+	}
+
 
 	for(i=D_AX; i<=D_R15; i++)
 		if(reg[i])
@@ -520,7 +530,16 @@ gconreg(int as, vlong c, int reg)
 {
 	Node nr;
 
-	nodreg(&nr, types[TINT64], reg);
+	switch(as) {
+	case AADDL:
+	case AMOVL:
+	case ALEAL:
+		nodreg(&nr, types[TINT32], reg);
+		break;
+	default:
+		nodreg(&nr, types[TINT64], reg);
+	}
+
 	ginscon(as, c, &nr);
 }
 
@@ -533,10 +552,18 @@ ginscon(int as, vlong c, Node *n2)
 {
 	Node n1, ntmp;
 
-	nodconst(&n1, types[TINT64], c);
+	switch(as) {
+	case AADDL:
+	case AMOVL:
+	case ALEAL:
+		nodconst(&n1, types[TINT32], c);
+		break;
+	default:
+		nodconst(&n1, types[TINT64], c);
+	}
 
 	if(as != AMOVQ && (c < -(1LL<<31) || c >= 1LL<<31)) {
-		// cannot have 64-bit immediokate in ADD, etc.
+		// cannot have 64-bit immediate in ADD, etc.
 		// instead, MOV into register first.
 		regalloc(&ntmp, types[TINT64], N);
 		gins(AMOVQ, &n1, &ntmp);
@@ -2040,7 +2067,7 @@ odot:
 	for(i=1; i<o; i++) {
 		if(oary[i] >= 0)
 			fatal("can't happen");
-		gins(AMOVQ, &n1, reg);
+		gins(movptr, &n1, reg);
 		cgen_checknil(reg);
 		n1.xoffset = -(oary[i]+1);
 	}
@@ -2252,7 +2279,7 @@ oindex_const_sudo:
 	if(reg->op == OEMPTY)
 		regalloc(reg, types[tptr], N);
 
-	p1 = gins(AMOVQ, N, reg);
+	p1 = gins(movptr, N, reg);
 	p1->from = *a;
 
 	n2 = *reg;

src/cmd/6g/prog.c

@@ -143,6 +143,7 @@ static ProgInfo progtable[ALAST] = {
 
 	[AJMP]=		{Jump | Break | KillCarry},
 
+	[ALEAL]=	{LeftAddr | RightWrite},
 	[ALEAQ]=	{LeftAddr | RightWrite},
 
 	[AMOVBLSX]=	{SizeL | LeftRead | RightWrite | Conv},

src/cmd/6g/reg.c

@@ -487,7 +487,7 @@ addmove(Reg *r, int bn, int rn, int f)
 	// need to clean this up with wptr and
 	// some of the defaults
 	p1->as = AMOVL;
-	switch(v->etype) {
+	switch(simtype[(uchar)v->etype]) {
 	default:
 		fatal("unknown type %E", v->etype);
 	case TINT8:
@@ -501,7 +501,6 @@ addmove(Reg *r, int bn, int rn, int f)
 		break;
 	case TINT64:
 	case TUINT64:
-	case TUINTPTR:
 	case TPTR64:
 		p1->as = AMOVQ;
 		break;
@@ -511,8 +510,6 @@ addmove(Reg *r, int bn, int rn, int f)
 	case TFLOAT64:
 		p1->as = AMOVSD;
 		break;
-	case TINT:
-	case TUINT:
 	case TINT32:
 	case TUINT32:
 	case TPTR32:
@@ -1088,6 +1085,8 @@ int
 BtoR(int32 b)
 {
 	b &= 0xffffL;
+	if(nacl)
+		b &= ~((1<<(D_BP-D_AX)) | (1<<(D_R15-D_AX)));
 	if(b == 0)
 		return 0;
 	return bitno(b) + D_AX;

src/cmd/8g/galign.c

@@ -29,6 +29,7 @@ betypeinit(void)
 {
 	widthptr = 4;
 	widthint = 4;
+	widthreg = 4;
 
 	zprog.link = P;
 	zprog.as = AGOK;

src/cmd/8g/gg.h

@@ -29,6 +29,7 @@ EXTERN	Node*	deferproc;
 EXTERN	Node*	deferreturn;
 EXTERN	Node*	panicindex;
 EXTERN	Node*	panicslice;
+EXTERN	Node*	panicdiv;
 EXTERN	Node*	throwreturn;
 EXTERN	int	maxstksize;
 extern	uint32	unmappedzero;

src/cmd/8g/ggen.c

@@ -664,6 +664,18 @@ dodiv(int op, Node *nl, Node *nr, Node *res, Node *ax, Node *dx)
 	gmove(&t2, &n1);
 	gmove(&t1, ax);
 	p2 = P;
+	if(nacl) {
+		// Native Client does not relay the divide-by-zero trap
+		// to the executing program, so we must insert a check
+		// for ourselves.
+		nodconst(&n4, t, 0);
+		gins(optoas(OCMP, t), &n1, &n4);
+		p1 = gbranch(optoas(ONE, t), T, +1);
+		if(panicdiv == N)
+			panicdiv = sysfunc("panicdivide");
+		ginscall(panicdiv, -1);
+		patch(p1, pc);
+	}
 	if(check) {
 		nodconst(&n4, t, -1);
 		gins(optoas(OCMP, t), &n1, &n4);

src/cmd/dist/build.c

@@ -38,13 +38,14 @@ static void dopack(char*, char*, char**, int);
 static char *findgoversion(void);
 
 // The known architecture letters.
-static char *gochars = "568";
+static char *gochars = "5668";
 
 // The known architectures.
 static char *okgoarch[] = {
 	// same order as gochars
 	"arm",
 	"amd64",
+	"amd64p32",
 	"386",
 };
 
@@ -55,6 +56,7 @@ static char *okgoos[] = {
 	"linux",
 	"solaris",
 	"freebsd",
+	"nacl",
 	"netbsd",
 	"openbsd",
 	"plan9",

src/cmd/gc/align.c

@@ -175,11 +175,11 @@ dowidth(Type *t)
 	case TFLOAT64:
 	case TCOMPLEX64:
 		w = 8;
-		t->align = widthptr;
+		t->align = widthreg;
 		break;
 	case TCOMPLEX128:
 		w = 16;
-		t->align = widthptr;
+		t->align = widthreg;
 		break;
 	case TPTR32:
 		w = 4;
@@ -288,10 +288,10 @@ dowidth(Type *t)
 		// compute their widths as side-effect.
 		t1 = t->type;
 		w = widstruct(t->type, *getthis(t1), 0, 0);
-		w = widstruct(t->type, *getinarg(t1), w, widthptr);
-		w = widstruct(t->type, *getoutarg(t1), w, widthptr);
+		w = widstruct(t->type, *getinarg(t1), w, widthreg);
+		w = widstruct(t->type, *getoutarg(t1), w, widthreg);
 		t1->argwid = w;
-		if(w%widthptr)
+		if(w%widthreg)
 			warn("bad type %T %d\n", t1, w);
 		t->align = 1;
 		break;

src/cmd/gc/builtin.c

@@ -5,6 +5,7 @@ char *runtimeimport =
 	"func @\"\".new (@\"\".typ·2 *byte) (? *any)\n"
 	"func @\"\".panicindex ()\n"
 	"func @\"\".panicslice ()\n"
+	"func @\"\".panicdivide ()\n"
 	"func @\"\".throwreturn ()\n"
 	"func @\"\".throwinit ()\n"
 	"func @\"\".panicwrap (? string, ? string, ? string)\n"

src/cmd/gc/go.h

@@ -950,6 +950,7 @@ EXTERN	Node*	curfn;
 
 EXTERN	int	widthptr;
 EXTERN	int	widthint;
+EXTERN	int	widthreg;
 
 EXTERN	Node*	typesw;
 EXTERN	Node*	nblank;
@@ -982,6 +983,8 @@ EXTERN	int	writearchive;
 
 EXTERN	Biobuf	bstdout;
 
+EXTERN	int	nacl;
+
 /*
  *	y.tab.c
  */

src/cmd/gc/lex.c

@@ -258,8 +258,18 @@ main(int argc, char *argv[])
 
 	goroot = getgoroot();
 	goos = getgoos();
-	goarch = thestring;
+
+	// Allow GOARCH=thestring or GOARCH=thestringsuffix,
+	// but not other values.	
+	p = getgoarch();
+	if(strncmp(p, thestring, strlen(thestring)) != 0)
+		fatal("cannot use %cg with GOARCH=%s", thechar, p);
+	goarch = p;
 	
+	nacl = strcmp(goos, "nacl") == 0;
+	if(nacl)
+		flag_largemodel = 1;
+
 	setexp();
 
 	outfile = nil;
@@ -779,7 +789,7 @@ importfile(Val *f, int line)
 			yyerror("import %s: not a go object file", file);
 			errorexit();
 		}
-		q = smprint("%s %s %s %s", getgoos(), thestring, getgoversion(), expstring());
+		q = smprint("%s %s %s %s", getgoos(), getgoarch(), getgoversion(), expstring());
 		if(strcmp(p+10, q) != 0) {
 			yyerror("import %s: object is [%s] expected [%s]", file, p+10, q);
 			errorexit();

src/cmd/gc/obj.c

@@ -47,7 +47,7 @@ dumpobj(void)
 		Bwrite(bout, arhdr, sizeof arhdr);
 		startobj = Boffset(bout);
 	}
-	Bprint(bout, "go object %s %s %s %s\n", getgoos(), thestring, getgoversion(), expstring());
+	Bprint(bout, "go object %s %s %s %s\n", getgoos(), getgoarch(), getgoversion(), expstring());
 	dumpexport();
 	
 	if(writearchive) {

src/cmd/gc/pgen.c

@@ -426,7 +426,7 @@ allocauto(Prog* ptxt)
 		}
 		n->stkdelta = -stksize - n->xoffset;
 	}
-	stksize = rnd(stksize, widthptr);
+	stksize = rnd(stksize, widthreg);
 	stkptrsize = rnd(stkptrsize, widthptr);
 	stkzerosize = rnd(stkzerosize, widthptr);
 

src/cmd/gc/reflect.c

@@ -143,6 +143,11 @@ mapbucket(Type *t)
 	overflowfield->sym = mal(sizeof(Sym)); // not important but needs to be set to give this type a name
 	overflowfield->sym->name = "overflow";
 	offset += widthptr;
+	
+	// The keys are padded to the native integer alignment.
+	// This is usually the same as widthptr; the exception (as usual) is nacl/amd64.
+	if(widthreg > widthptr)
+		offset += widthreg - widthptr;
 
 	keysfield = typ(TFIELD);
 	keysfield->type = typ(TARRAY);

src/cmd/gc/runtime.go

@@ -15,6 +15,7 @@ package PACKAGE
 func new(typ *byte) *any
 func panicindex()
 func panicslice()
+func panicdivide()
 func throwreturn()
 func throwinit()
 func panicwrap(string, string, string)

src/cmd/gc/walk.c

@@ -1059,7 +1059,7 @@ walkexpr(Node **np, NodeList **init)
 		switch(n->op) {
 		case OMOD:
 		case ODIV:
-			if(widthptr > 4 || (et != TUINT64 && et != TINT64))
+			if(widthreg >= 8 || (et != TUINT64 && et != TINT64))
 				goto ret;
 			if(et == TINT64)
 				strcpy(namebuf, "int64");

src/cmd/go/build.go

@@ -204,6 +204,9 @@ type stringsFlag []string
 func (v *stringsFlag) Set(s string) error {
 	var err error
 	*v, err = splitQuotedFields(s)
+	if *v == nil {
+		*v = []string{}
+	}
 	return err
 }
 

src/cmd/go/run.go

@@ -8,9 +8,27 @@ import (
 	"fmt"
 	"os"
 	"os/exec"
+	"runtime"
 	"strings"
 )
 
+var execCmd []string // -exec flag, for run and test
+
+func findExecCmd() []string {
+	if execCmd != nil {
+		return execCmd
+	}
+	execCmd = []string{} // avoid work the second time
+	if goos == runtime.GOOS && goarch == runtime.GOARCH {
+		return execCmd
+	}
+	path, err := exec.LookPath(fmt.Sprintf("go_%s_%s_exec", goos, goarch))
+	if err == nil {
+		execCmd = []string{path}
+	}
+	return execCmd
+}
+
 var cmdRun = &Command{
 	UsageLine: "run [build flags] gofiles... [arguments...]",
 	Short:     "compile and run Go program",
@@ -28,6 +46,7 @@ func init() {
 	cmdRun.Run = runRun // break init loop
 
 	addBuildFlags(cmdRun)
+	cmdRun.Flag.Var((*stringsFlag)(&execCmd), "exec", "")
 }
 
 func printStderr(args ...interface{}) (int, error) {
@@ -90,20 +109,20 @@ func runRun(cmd *Command, args []string) {
 // runProgram is the action for running a binary that has already
 // been compiled.  We ignore exit status.
 func (b *builder) runProgram(a *action) error {
+	cmdline := stringList(findExecCmd(), a.deps[0].target, a.args)
 	if buildN || buildX {
-		b.showcmd("", "%s %s", a.deps[0].target, strings.Join(a.args, " "))
+		b.showcmd("", "%s", strings.Join(cmdline, " "))
 		if buildN {
 			return nil
 		}
 	}
 
-	runStdin(a.deps[0].target, a.args)
+	runStdin(cmdline)
 	return nil
 }
 
 // runStdin is like run, but connects Stdin.
-func runStdin(cmdargs ...interface{}) {
-	cmdline := stringList(cmdargs...)
+func runStdin(cmdline []string) {
 	cmd := exec.Command(cmdline[0], cmdline[1:]...)
 	cmd.Stdin = os.Stdin
 	cmd.Stdout = os.Stdout

src/cmd/go/signal_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package main
 

src/cmd/go/test.go

@@ -293,6 +293,8 @@ func runTest(cmd *Command, args []string) {
 	var pkgArgs []string
 	pkgArgs, testArgs = testFlags(args)
 
+	findExecCmd() // initialize cached result
+
 	raceInit()
 	pkgs := packagesForBuild(pkgArgs)
 	if len(pkgs) == 0 {
@@ -835,7 +837,7 @@ func declareCoverVars(importPath string, files ...string) map[string]*CoverVar {
 
 // runTest is the action for running a test binary.
 func (b *builder) runTest(a *action) error {
-	args := stringList(a.deps[0].target, testArgs)
+	args := stringList(findExecCmd(), a.deps[0].target, testArgs)
 	a.testOutput = new(bytes.Buffer)
 
 	if buildN || buildX {

src/cmd/go/testflag.go

@@ -77,6 +77,7 @@ var testFlagDefn = []*testFlagSpec{
 	{name: "x", boolVar: &buildX},
 	{name: "work", boolVar: &buildWork},
 	{name: "gcflags"},
+	{name: "exec"},
 	{name: "ldflags"},
 	{name: "gccgoflags"},
 	{name: "tags"},
@@ -154,6 +155,11 @@ func testFlags(args []string) (packageNames, passToTest []string) {
 			setBoolFlag(f.boolVar, value)
 		case "p":
 			setIntFlag(&buildP, value)
+		case "exec":
+			execCmd, err = splitQuotedFields(value)
+			if err != nil {
+				fatalf("invalid flag argument for -%s: %v", f.name, err)
+			}
 		case "gcflags":
 			buildGcflags, err = splitQuotedFields(value)
 			if err != nil {

src/pkg/crypto/md5/md5block_amd64p32.s

+// Original source:
+//	http://www.zorinaq.com/papers/md5-amd64.html
+//	http://www.zorinaq.com/papers/md5-amd64.tar.bz2
+//
+// Translated from Perl generating GNU assembly into
+// #defines generating 6a assembly by the Go Authors.
+//
+// Restrictions to make code safe for Native Client:
+// replace BP with R11, reloaded before use at return.
+// replace R15 with R11.
+
+#include "../../../cmd/ld/textflag.h"
+
+// MD5 optimized for AMD64.
+//
+// Author: Marc Bevand <bevand_m (at) epita.fr>
+// Licence: I hereby disclaim the copyright on this code and place it
+// in the public domain.
+
+TEXT	·block(SB),NOSPLIT,$0-32
+	MOVL	dig+0(FP),	R11
+	MOVL	p+4(FP),	SI
+	MOVL	p_len+8(FP), DX
+	SHRQ	$6,		DX
+	SHLQ	$6,		DX
+
+	LEAQ	(SI)(DX*1),	DI
+	MOVL	(0*4)(R11),	AX
+	MOVL	(1*4)(R11),	BX
+	MOVL	(2*4)(R11),	CX
+	MOVL	(3*4)(R11),	DX
+
+	CMPQ	SI,		DI
+	JEQ	end
+
+loop:
+	MOVL	AX,		R12
+	MOVL	BX,		R13
+	MOVL	CX,		R14
+	MOVL	DX,		R11
+
+	MOVL	(0*4)(SI),	R8
+	MOVL	DX,		R9
+
+#define ROUND1(a, b, c, d, index, const, shift) \
+	XORL	c, R9; \
+	LEAL	const(a)(R8*1), a; \
+	ANDL	b, R9; \
+	XORL d, R9; \
+	MOVL (index*4)(SI), R8; \
+	ADDL R9, a; \
+	ROLL $shift, a; \
+	MOVL c, R9; \
+	ADDL b, a
+
+	ROUND1(AX,BX,CX,DX, 1,0xd76aa478, 7);
+	ROUND1(DX,AX,BX,CX, 2,0xe8c7b756,12);
+	ROUND1(CX,DX,AX,BX, 3,0x242070db,17);
+	ROUND1(BX,CX,DX,AX, 4,0xc1bdceee,22);
+	ROUND1(AX,BX,CX,DX, 5,0xf57c0faf, 7);
+	ROUND1(DX,AX,BX,CX, 6,0x4787c62a,12);
+	ROUND1(CX,DX,AX,BX, 7,0xa8304613,17);
+	ROUND1(BX,CX,DX,AX, 8,0xfd469501,22);
+	ROUND1(AX,BX,CX,DX, 9,0x698098d8, 7);
+	ROUND1(DX,AX,BX,CX,10,0x8b44f7af,12);
+	ROUND1(CX,DX,AX,BX,11,0xffff5bb1,17);
+	ROUND1(BX,CX,DX,AX,12,0x895cd7be,22);
+	ROUND1(AX,BX,CX,DX,13,0x6b901122, 7);
+	ROUND1(DX,AX,BX,CX,14,0xfd987193,12);
+	ROUND1(CX,DX,AX,BX,15,0xa679438e,17);
+	ROUND1(BX,CX,DX,AX, 0,0x49b40821,22);
+
+	MOVL	(1*4)(SI),	R8
+	MOVL	DX,		R9
+	MOVL	DX,		R10
+
+#define ROUND2(a, b, c, d, index, const, shift) \
+	NOTL	R9; \
+	LEAL	const(a)(R8*1),a; \
+	ANDL	b,		R10; \
+	ANDL	c,		R9; \
+	MOVL	(index*4)(SI),R8; \
+	ORL	R9,		R10; \
+	MOVL	c,		R9; \
+	ADDL	R10,		a; \
+	MOVL	c,		R10; \
+	ROLL	$shift,	a; \
+	ADDL	b,		a
+
+	ROUND2(AX,BX,CX,DX, 6,0xf61e2562, 5);
+	ROUND2(DX,AX,BX,CX,11,0xc040b340, 9);
+	ROUND2(CX,DX,AX,BX, 0,0x265e5a51,14);
+	ROUND2(BX,CX,DX,AX, 5,0xe9b6c7aa,20);
+	ROUND2(AX,BX,CX,DX,10,0xd62f105d, 5);
+	ROUND2(DX,AX,BX,CX,15, 0x2441453, 9);
+	ROUND2(CX,DX,AX,BX, 4,0xd8a1e681,14);
+	ROUND2(BX,CX,DX,AX, 9,0xe7d3fbc8,20);
+	ROUND2(AX,BX,CX,DX,14,0x21e1cde6, 5);
+	ROUND2(DX,AX,BX,CX, 3,0xc33707d6, 9);
+	ROUND2(CX,DX,AX,BX, 8,0xf4d50d87,14);
+	ROUND2(BX,CX,DX,AX,13,0x455a14ed,20);
+	ROUND2(AX,BX,CX,DX, 2,0xa9e3e905, 5);
+	ROUND2(DX,AX,BX,CX, 7,0xfcefa3f8, 9);
+	ROUND2(CX,DX,AX,BX,12,0x676f02d9,14);
+	ROUND2(BX,CX,DX,AX, 0,0x8d2a4c8a,20);
+ 
+	MOVL	(5*4)(SI),	R8
+	MOVL	CX,		R9
+
+#define ROUND3(a, b, c, d, index, const, shift) \
+	LEAL	const(a)(R8*1),a; \
+	MOVL	(index*4)(SI),R8; \
+	XORL	d,		R9; \
+	XORL	b,		R9; \
+	ADDL	R9,		a; \
+	ROLL	$shift,		a; \
+	MOVL	b,		R9; \
+	ADDL	b,		a
+
+	ROUND3(AX,BX,CX,DX, 8,0xfffa3942, 4);
+	ROUND3(DX,AX,BX,CX,11,0x8771f681,11);
+	ROUND3(CX,DX,AX,BX,14,0x6d9d6122,16);
+	ROUND3(BX,CX,DX,AX, 1,0xfde5380c,23);
+	ROUND3(AX,BX,CX,DX, 4,0xa4beea44, 4);
+	ROUND3(DX,AX,BX,CX, 7,0x4bdecfa9,11);
+	ROUND3(CX,DX,AX,BX,10,0xf6bb4b60,16);
+	ROUND3(BX,CX,DX,AX,13,0xbebfbc70,23);
+	ROUND3(AX,BX,CX,DX, 0,0x289b7ec6, 4);
+	ROUND3(DX,AX,BX,CX, 3,0xeaa127fa,11);
+	ROUND3(CX,DX,AX,BX, 6,0xd4ef3085,16);
+	ROUND3(BX,CX,DX,AX, 9, 0x4881d05,23);
+	ROUND3(AX,BX,CX,DX,12,0xd9d4d039, 4);
+	ROUND3(DX,AX,BX,CX,15,0xe6db99e5,11);
+	ROUND3(CX,DX,AX,BX, 2,0x1fa27cf8,16);
+	ROUND3(BX,CX,DX,AX, 0,0xc4ac5665,23);
+
+	MOVL	(0*4)(SI),	R8
+	MOVL	$0xffffffff,	R9
+	XORL	DX,		R9
+
+#define ROUND4(a, b, c, d, index, const, shift) \
+	LEAL	const(a)(R8*1),a; \
+	ORL	b,		R9; \
+	XORL	c,		R9; \
+	ADDL	R9,		a; \
+	MOVL	(index*4)(SI),R8; \
+	MOVL	$0xffffffff,	R9; \
+	ROLL	$shift,		a; \
+	XORL	c,		R9; \
+	ADDL	b,		a
+	
+	ROUND4(AX,BX,CX,DX, 7,0xf4292244, 6);
+	ROUND4(DX,AX,BX,CX,14,0x432aff97,10);
+	ROUND4(CX,DX,AX,BX, 5,0xab9423a7,15);
+	ROUND4(BX,CX,DX,AX,12,0xfc93a039,21);
+	ROUND4(AX,BX,CX,DX, 3,0x655b59c3, 6);
+	ROUND4(DX,AX,BX,CX,10,0x8f0ccc92,10);
+	ROUND4(CX,DX,AX,BX, 1,0xffeff47d,15);
+	ROUND4(BX,CX,DX,AX, 8,0x85845dd1,21);
+	ROUND4(AX,BX,CX,DX,15,0x6fa87e4f, 6);
+	ROUND4(DX,AX,BX,CX, 6,0xfe2ce6e0,10);
+	ROUND4(CX,DX,AX,BX,13,0xa3014314,15);
+	ROUND4(BX,CX,DX,AX, 4,0x4e0811a1,21);
+	ROUND4(AX,BX,CX,DX,11,0xf7537e82, 6);
+	ROUND4(DX,AX,BX,CX, 2,0xbd3af235,10);
+	ROUND4(CX,DX,AX,BX, 9,0x2ad7d2bb,15);
+	ROUND4(BX,CX,DX,AX, 0,0xeb86d391,21);
+
+	ADDL	R12,	AX
+	ADDL	R13,	BX
+	ADDL	R14,	CX
+	ADDL	R11,	DX
+
+	ADDQ	$64,		SI
+	CMPQ	SI,		DI
+	JB	loop
+
+end:
+	MOVL	dig+0(FP),	R11
+	MOVL	AX,		(0*4)(R11)
+	MOVL	BX,		(1*4)(R11)
+	MOVL	CX,		(2*4)(R11)
+	MOVL	DX,		(3*4)(R11)
+	RET

src/pkg/crypto/md5/md5block_decl.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64 386 arm
+// +build amd64 amd64p32 386 arm
 
 package md5
 

src/pkg/crypto/rand/rand_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd plan9 solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd plan9 solaris
 
 // Unix cryptographically secure pseudorandom number
 // generator.

src/pkg/crypto/rc4/rc4_amd64p32.s

+// Original source:
+//	http://www.zorinaq.com/papers/rc4-amd64.html
+//	http://www.zorinaq.com/papers/rc4-amd64.tar.bz2
+
+#include "../../../cmd/ld/textflag.h"
+
+// Local modifications:
+//
+// Transliterated from GNU to 6a assembly syntax by the Go authors.
+// The comments and spacing are from the original.
+//
+// The new EXTEND macros avoid a bad stall on some systems after 8-bit math.
+//
+// The original code accumulated 64 bits of key stream in an integer
+// register and then XOR'ed the key stream into the data 8 bytes at a time.
+// Modified to accumulate 128 bits of key stream into an XMM register
+// and then XOR the key stream into the data 16 bytes at a time.
+// Approximately doubles throughput.
+//
+// Converted to amd64p32.
+//
+// To make safe for Native Client, avoid use of BP, R15,
+// and two-register addressing modes.
+
+// NOTE: Changing EXTEND to a no-op makes the code run 1.2x faster on Core i5
+// but makes the code run 2.0x slower on Xeon.
+#define EXTEND(r) MOVBLZX r, r
+
+/*
+** RC4 implementation optimized for AMD64.
+**
+** Author: Marc Bevand <bevand_m (at) epita.fr>
+** Licence: I hereby disclaim the copyright on this code and place it
+** in the public domain.
+**
+** The code has been designed to be easily integrated into openssl:
+** the exported RC4() function can replace the actual implementations
+** openssl already contains. Please note that when linking with openssl,
+** it requires that sizeof(RC4_INT) == 8. So openssl must be compiled
+** with -DRC4_INT='unsigned long'.
+**
+** The throughput achieved by this code is about 320 MBytes/sec, on
+** a 1.8 GHz AMD Opteron (rev C0) processor.
+*/
+
+TEXT ·xorKeyStream(SB),NOSPLIT,$0
+	MOVL	n+8(FP),	BX		// rbx = ARG(len)
+	MOVL	src+4(FP),	SI		// in = ARG(in)
+	MOVL	dst+0(FP),	DI		// out = ARG(out)
+	MOVL	state+12(FP),	R10		// d = ARG(data)
+	MOVL	i+16(FP),	AX
+	MOVBQZX	0(AX),		CX		// x = *xp
+	MOVL	j+20(FP),	AX
+	MOVBQZX	0(AX),		DX		// y = *yp
+
+	LEAQ	(SI)(BX*1),	R9		// limit = in+len
+
+l1:	CMPQ	SI,		R9		// cmp in with in+len
+	JGE	finished			// jump if (in >= in+len)
+
+	INCB	CX
+	EXTEND(CX)
+	TESTL	$15,		CX
+	JZ	wordloop
+	LEAL	(R10)(CX*4), R12
+
+	MOVBLZX	(R12),	AX
+
+	ADDB	AX,		DX		// y += tx
+	EXTEND(DX)
+	LEAL (R10)(DX*4), R11
+	MOVBLZX	(R11),	BX		// ty = d[y]
+	MOVB	BX,		(R12)	// d[x] = ty
+	ADDB	AX,		BX		// val = ty+tx
+	EXTEND(BX)
+	LEAL (R10)(BX*4), R13
+	MOVB	AX,		(R11)	// d[y] = tx
+	MOVBLZX	(R13),	R8		// val = d[val]
+	XORB	(SI),		R8		// xor 1 byte
+	MOVB	R8,		(DI)
+	INCQ	SI				// in++
+	INCQ	DI				// out++
+	JMP l1
+
+wordloop:
+	SUBQ	$16,		R9
+	CMPQ	SI,		R9
+	JGT	end
+
+start:
+	ADDQ	$16,		SI		// increment in
+	ADDQ	$16,		DI		// increment out
+
+	// Each KEYROUND generates one byte of key and
+	// inserts it into an XMM register at the given 16-bit index.
+	// The key state array is uint32 words only using the bottom
+	// byte of each word, so the 16-bit OR only copies 8 useful bits.
+	// We accumulate alternating bytes into X0 and X1, and then at
+	// the end we OR X1<<8 into X0 to produce the actual key.
+	//
+	// At the beginning of the loop, CX%16 == 0, so the 16 loads
+	// at state[CX], state[CX+1], ..., state[CX+15] can precompute
+	// (state+CX) as R12 and then become R12[0], R12[1], ... R12[15],
+	// without fear of the byte computation CX+15 wrapping around.
+	//
+	// The first round needs R12[0], the second needs R12[1], and so on.
+	// We can avoid memory stalls by starting the load for round n+1
+	// before the end of round n, using the LOAD macro.
+	LEAQ	(R10)(CX*4),	R12
+
+#define KEYROUND(xmm, load, off, r1, r2, index) \
+	LEAL (R10)(DX*4), R11; \
+	MOVBLZX	(R11),	R8; \
+	MOVB	r1,		(R11); \
+	load((off+1), r2); \
+	MOVB	R8,		(off*4)(R12); \
+	ADDB	r1,		R8; \
+	EXTEND(R8); \
+	LEAL (R10)(R8*4), R14; \
+	PINSRW	$index, (R14), xmm
+
+#define LOAD(off, reg) \
+	MOVBLZX	(off*4)(R12),	reg; \
+	ADDB	reg,		DX; \
+	EXTEND(DX)
+
+#define SKIP(off, reg)
+
+	LOAD(0, AX)
+	KEYROUND(X0, LOAD, 0, AX, BX, 0)
+	KEYROUND(X1, LOAD, 1, BX, AX, 0)
+	KEYROUND(X0, LOAD, 2, AX, BX, 1)
+	KEYROUND(X1, LOAD, 3, BX, AX, 1)
+	KEYROUND(X0, LOAD, 4, AX, BX, 2)
+	KEYROUND(X1, LOAD, 5, BX, AX, 2)
+	KEYROUND(X0, LOAD, 6, AX, BX, 3)
+	KEYROUND(X1, LOAD, 7, BX, AX, 3)
+	KEYROUND(X0, LOAD, 8, AX, BX, 4)
+	KEYROUND(X1, LOAD, 9, BX, AX, 4)
+	KEYROUND(X0, LOAD, 10, AX, BX, 5)
+	KEYROUND(X1, LOAD, 11, BX, AX, 5)
+	KEYROUND(X0, LOAD, 12, AX, BX, 6)
+	KEYROUND(X1, LOAD, 13, BX, AX, 6)
+	KEYROUND(X0, LOAD, 14, AX, BX, 7)
+	KEYROUND(X1, SKIP, 15, BX, AX, 7)
+	
+	ADDB	$16,		CX
+
+	PSLLQ	$8,		X1
+	PXOR	X1,		X0
+	MOVOU	-16(SI),	X2
+	PXOR	X0,		X2
+	MOVOU	X2,		-16(DI)
+
+	CMPQ	SI,		R9		// cmp in with in+len-16
+	JLE	start				// jump if (in <= in+len-16)
+
+end:
+	DECB	CX
+	ADDQ	$16,		R9		// tmp = in+len
+
+	// handle the last bytes, one by one
+l2:	CMPQ	SI,		R9		// cmp in with in+len
+	JGE	finished			// jump if (in >= in+len)
+
+	INCB	CX
+	EXTEND(CX)
+	LEAL (R10)(CX*4), R12
+	MOVBLZX	(R12),	AX
+
+	ADDB	AX,		DX		// y += tx
+	EXTEND(DX)
+	LEAL (R10)(DX*4), R11
+	MOVBLZX	(R11),	BX		// ty = d[y]
+	MOVB	BX,		(R12)	// d[x] = ty
+	ADDB	AX,		BX		// val = ty+tx
+	EXTEND(BX)
+	LEAL (R10)(BX*4), R13
+	MOVB	AX,		(R11)	// d[y] = tx
+	MOVBLZX	(R13),	R8		// val = d[val]
+	XORB	(SI),		R8		// xor 1 byte
+	MOVB	R8,		(DI)
+	INCQ	SI				// in++
+	INCQ	DI				// out++
+	JMP l2
+
+finished:
+	MOVL	j+20(FP),	BX
+	MOVB	DX, 0(BX)
+	MOVL	i+16(FP),	AX
+	MOVB	CX, 0(AX)
+	RET

src/pkg/crypto/rc4/rc4_asm.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64 arm 386
+// +build amd64 amd64p32 arm 386
 
 package rc4
 

src/pkg/crypto/rc4/rc4_ref.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build !amd64,!arm,!386
+// +build !amd64,!amd64p32,!arm,!386
 
 package rc4
 

src/pkg/crypto/sha1/sha1block_amd64p32.s

+// Copyright 2013 The Go Authors.  All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#include "../../../cmd/ld/textflag.h"
+
+// SHA1 block routine. See sha1block.go for Go equivalent.
+//
+// There are 80 rounds of 4 types:
+//   - rounds 0-15 are type 1 and load data (ROUND1 macro).
+//   - rounds 16-19 are type 1 and do not load data (ROUND1x macro).
+//   - rounds 20-39 are type 2 and do not load data (ROUND2 macro).
+//   - rounds 40-59 are type 3 and do not load data (ROUND3 macro).
+//   - rounds 60-79 are type 4 and do not load data (ROUND4 macro).
+//
+// Each round loads or shuffles the data, then computes a per-round
+// function of b, c, d, and then mixes the result into and rotates the
+// five registers a, b, c, d, e holding the intermediate results.
+//
+// The register rotation is implemented by rotating the arguments to
+// the round macros instead of by explicit move instructions.
+//
+// amd64p32 version.
+// To ensure safety for Native Client, avoids use of BP and R15
+// as well as two-register addressing modes.
+
+#define LOAD(index) \
+	MOVL	(index*4)(SI), R10; \
+	BSWAPL	R10; \
+	MOVL	R10, (index*4)(SP)
+
+#define SHUFFLE(index) \
+	MOVL	(((index)&0xf)*4)(SP), R10; \
+	XORL	(((index-3)&0xf)*4)(SP), R10; \
+	XORL	(((index-8)&0xf)*4)(SP), R10; \
+	XORL	(((index-14)&0xf)*4)(SP), R10; \
+	ROLL	$1, R10; \
+	MOVL	R10, (((index)&0xf)*4)(SP)
+
+#define FUNC1(a, b, c, d, e) \
+	MOVL	d, R9; \
+	XORL	c, R9; \
+	ANDL	b, R9; \
+	XORL	d, R9
+
+#define FUNC2(a, b, c, d, e) \
+	MOVL	b, R9; \
+	XORL	c, R9; \
+	XORL	d, R9
+
+#define FUNC3(a, b, c, d, e) \
+	MOVL	b, R8; \
+	ORL	c, R8; \
+	ANDL	d, R8; \
+	MOVL	b, R9; \
+	ANDL	c, R9; \
+	ORL	R8, R9
+	
+#define FUNC4 FUNC2
+
+#define MIX(a, b, c, d, e, const) \
+	ROLL	$30, b; \
+	ADDL	R9, e; \
+	MOVL	a, R8; \
+	ROLL	$5, R8; \
+	LEAL	const(e)(R10*1), e; \
+	ADDL	R8, e
+
+#define ROUND1(a, b, c, d, e, index) \
+	LOAD(index); \
+	FUNC1(a, b, c, d, e); \
+	MIX(a, b, c, d, e, 0x5A827999)
+
+#define ROUND1x(a, b, c, d, e, index) \
+	SHUFFLE(index); \
+	FUNC1(a, b, c, d, e); \
+	MIX(a, b, c, d, e, 0x5A827999)
+
+#define ROUND2(a, b, c, d, e, index) \
+	SHUFFLE(index); \
+	FUNC2(a, b, c, d, e); \
+	MIX(a, b, c, d, e, 0x6ED9EBA1)
+
+#define ROUND3(a, b, c, d, e, index) \
+	SHUFFLE(index); \
+	FUNC3(a, b, c, d, e); \
+	MIX(a, b, c, d, e, 0x8F1BBCDC)
+
+#define ROUND4(a, b, c, d, e, index) \
+	SHUFFLE(index); \
+	FUNC4(a, b, c, d, e); \
+	MIX(a, b, c, d, e, 0xCA62C1D6)
+
+TEXT ·block(SB),NOSPLIT,$64-32
+	MOVL	dig+0(FP),	R14
+	MOVL	p_base+4(FP),	SI
+	MOVL	p_len+8(FP),	DX
+	SHRQ	$6,		DX
+	SHLQ	$6,		DX
+	
+	LEAQ	(SI)(DX*1),	DI
+	MOVL	(0*4)(R14),	AX
+	MOVL	(1*4)(R14),	BX
+	MOVL	(2*4)(R14),	CX
+	MOVL	(3*4)(R14),	DX
+	MOVL	(4*4)(R14),	R13
+
+	CMPQ	SI,		DI
+	JEQ	end
+
+loop:
+#define BP R13 /* keep diff from sha1block_amd64.s small */
+	ROUND1(AX, BX, CX, DX, BP, 0)
+	ROUND1(BP, AX, BX, CX, DX, 1)
+	ROUND1(DX, BP, AX, BX, CX, 2)
+	ROUND1(CX, DX, BP, AX, BX, 3)
+	ROUND1(BX, CX, DX, BP, AX, 4)
+	ROUND1(AX, BX, CX, DX, BP, 5)
+	ROUND1(BP, AX, BX, CX, DX, 6)
+	ROUND1(DX, BP, AX, BX, CX, 7)
+	ROUND1(CX, DX, BP, AX, BX, 8)
+	ROUND1(BX, CX, DX, BP, AX, 9)
+	ROUND1(AX, BX, CX, DX, BP, 10)
+	ROUND1(BP, AX, BX, CX, DX, 11)
+	ROUND1(DX, BP, AX, BX, CX, 12)
+	ROUND1(CX, DX, BP, AX, BX, 13)
+	ROUND1(BX, CX, DX, BP, AX, 14)
+	ROUND1(AX, BX, CX, DX, BP, 15)
+
+	ROUND1x(BP, AX, BX, CX, DX, 16)
+	ROUND1x(DX, BP, AX, BX, CX, 17)
+	ROUND1x(CX, DX, BP, AX, BX, 18)
+	ROUND1x(BX, CX, DX, BP, AX, 19)
+	
+	ROUND2(AX, BX, CX, DX, BP, 20)
+	ROUND2(BP, AX, BX, CX, DX, 21)
+	ROUND2(DX, BP, AX, BX, CX, 22)
+	ROUND2(CX, DX, BP, AX, BX, 23)
+	ROUND2(BX, CX, DX, BP, AX, 24)
+	ROUND2(AX, BX, CX, DX, BP, 25)
+	ROUND2(BP, AX, BX, CX, DX, 26)
+	ROUND2(DX, BP, AX, BX, CX, 27)
+	ROUND2(CX, DX, BP, AX, BX, 28)
+	ROUND2(BX, CX, DX, BP, AX, 29)
+	ROUND2(AX, BX, CX, DX, BP, 30)
+	ROUND2(BP, AX, BX, CX, DX, 31)
+	ROUND2(DX, BP, AX, BX, CX, 32)
+	ROUND2(CX, DX, BP, AX, BX, 33)
+	ROUND2(BX, CX, DX, BP, AX, 34)
+	ROUND2(AX, BX, CX, DX, BP, 35)
+	ROUND2(BP, AX, BX, CX, DX, 36)
+	ROUND2(DX, BP, AX, BX, CX, 37)
+	ROUND2(CX, DX, BP, AX, BX, 38)
+	ROUND2(BX, CX, DX, BP, AX, 39)
+	
+	ROUND3(AX, BX, CX, DX, BP, 40)
+	ROUND3(BP, AX, BX, CX, DX, 41)
+	ROUND3(DX, BP, AX, BX, CX, 42)
+	ROUND3(CX, DX, BP, AX, BX, 43)
+	ROUND3(BX, CX, DX, BP, AX, 44)
+	ROUND3(AX, BX, CX, DX, BP, 45)
+	ROUND3(BP, AX, BX, CX, DX, 46)
+	ROUND3(DX, BP, AX, BX, CX, 47)
+	ROUND3(CX, DX, BP, AX, BX, 48)
+	ROUND3(BX, CX, DX, BP, AX, 49)
+	ROUND3(AX, BX, CX, DX, BP, 50)
+	ROUND3(BP, AX, BX, CX, DX, 51)
+	ROUND3(DX, BP, AX, BX, CX, 52)
+	ROUND3(CX, DX, BP, AX, BX, 53)
+	ROUND3(BX, CX, DX, BP, AX, 54)
+	ROUND3(AX, BX, CX, DX, BP, 55)
+	ROUND3(BP, AX, BX, CX, DX, 56)
+	ROUND3(DX, BP, AX, BX, CX, 57)
+	ROUND3(CX, DX, BP, AX, BX, 58)
+	ROUND3(BX, CX, DX, BP, AX, 59)
+	
+	ROUND4(AX, BX, CX, DX, BP, 60)
+	ROUND4(BP, AX, BX, CX, DX, 61)
+	ROUND4(DX, BP, AX, BX, CX, 62)
+	ROUND4(CX, DX, BP, AX, BX, 63)
+	ROUND4(BX, CX, DX, BP, AX, 64)
+	ROUND4(AX, BX, CX, DX, BP, 65)
+	ROUND4(BP, AX, BX, CX, DX, 66)
+	ROUND4(DX, BP, AX, BX, CX, 67)
+	ROUND4(CX, DX, BP, AX, BX, 68)
+	ROUND4(BX, CX, DX, BP, AX, 69)
+	ROUND4(AX, BX, CX, DX, BP, 70)
+	ROUND4(BP, AX, BX, CX, DX, 71)
+	ROUND4(DX, BP, AX, BX, CX, 72)
+	ROUND4(CX, DX, BP, AX, BX, 73)
+	ROUND4(BX, CX, DX, BP, AX, 74)
+	ROUND4(AX, BX, CX, DX, BP, 75)
+	ROUND4(BP, AX, BX, CX, DX, 76)
+	ROUND4(DX, BP, AX, BX, CX, 77)
+	ROUND4(CX, DX, BP, AX, BX, 78)
+	ROUND4(BX, CX, DX, BP, AX, 79)
+#undef BP
+
+	ADDL	(0*4)(R14), AX
+	ADDL	(1*4)(R14), BX
+	ADDL	(2*4)(R14), CX
+	ADDL	(3*4)(R14), DX
+	ADDL	(4*4)(R14), R13
+
+	MOVL	AX, (0*4)(R14)
+	MOVL	BX, (1*4)(R14)
+	MOVL	CX, (2*4)(R14)
+	MOVL	DX, (3*4)(R14)
+	MOVL	R13, (4*4)(R14)
+
+	ADDQ	$64, SI
+	CMPQ	SI, DI
+	JB	loop
+
+end:
+	RET

src/pkg/crypto/sha1/sha1block_decl.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64 386 arm
+// +build amd64 amd64p32 386
 
 package sha1
 

src/pkg/crypto/x509/root_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build dragonfly freebsd linux openbsd netbsd solaris
+// +build dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package x509
 

src/pkg/go/build/build.go

@@ -1209,7 +1209,7 @@ func ArchChar(goarch string) (string, error) {
 	switch goarch {
 	case "386":
 		return "8", nil
-	case "amd64":
+	case "amd64", "amd64p32":
 		return "6", nil
 	case "arm":
 		return "5", nil

src/pkg/go/build/deps_test.go

@@ -8,6 +8,7 @@
 package build
 
 import (
+	"runtime"
 	"sort"
 	"testing"
 )
@@ -359,7 +360,7 @@ func allowed(pkg string) map[string]bool {
 }
 
 var bools = []bool{false, true}
-var geese = []string{"darwin", "dragonfly", "freebsd", "linux", "netbsd", "openbsd", "plan9", "solaris", "windows"}
+var geese = []string{"darwin", "dragonfly", "freebsd", "linux", "nacl", "netbsd", "openbsd", "plan9", "solaris", "windows"}
 var goarches = []string{"386", "amd64", "arm"}
 
 type osPkg struct {
@@ -374,6 +375,11 @@ var allowedErrors = map[osPkg]bool{
 }
 
 func TestDependencies(t *testing.T) {
+	if runtime.GOOS == "nacl" {
+		// NaCl tests run in a limited file system and we do not
+		// provide access to every source file.
+		t.Skip("skipping on NaCl")
+	}
 	var all []string
 
 	for k := range pkgDeps {
@@ -387,6 +393,9 @@ func TestDependencies(t *testing.T) {
 			if isMacro(pkg) {
 				continue
 			}
+			if pkg == "runtime/cgo" && !ctxt.CgoEnabled {
+				continue
+			}
 			p, err := ctxt.Import(pkg, "", 0)
 			if err != nil {
 				if allowedErrors[osPkg{ctxt.GOOS, pkg}] {

src/pkg/go/build/syslist.go

@@ -4,5 +4,5 @@
 
 package build
 
-const goosList = "darwin dragonfly freebsd linux netbsd openbsd plan9 solaris windows "
-const goarchList = "386 amd64 arm "
+const goosList = "darwin dragonfly freebsd linux nacl netbsd openbsd plan9 solaris windows "
+const goarchList = "386 amd64 amd64p32 arm "

src/pkg/hash/crc32/crc32_amd64p32.s

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#include "../../../cmd/ld/textflag.h"
+
+// func castagnoliSSE42(crc uint32, p []byte) uint32
+TEXT ·castagnoliSSE42(SB),NOSPLIT,$0
+	MOVL crc+0(FP), AX  // CRC value
+	MOVL p+4(FP), SI  // data pointer
+	MOVL p_len+8(FP), CX  // len(p)
+
+	NOTL AX
+
+	/* If there's less than 8 bytes to process, we do it byte-by-byte. */
+	CMPQ CX, $8
+	JL cleanup
+
+	/* Process individual bytes until the input is 8-byte aligned. */
+startup:
+	MOVQ SI, BX
+	ANDQ $7, BX
+	JZ aligned
+
+	CRC32B (SI), AX
+	DECQ CX
+	INCQ SI
+	JMP startup
+
+aligned:
+	/* The input is now 8-byte aligned and we can process 8-byte chunks. */
+	CMPQ CX, $8
+	JL cleanup
+
+	CRC32Q (SI), AX
+	ADDQ $8, SI
+	SUBQ $8, CX
+	JMP aligned
+
+cleanup:
+	/* We may have some bytes left over that we process one at a time. */
+	CMPQ CX, $0
+	JE done
+
+	CRC32B (SI), AX
+	INCQ SI
+	DECQ CX
+	JMP cleanup
+
+done:
+	NOTL AX
+	MOVL AX, ret+16(FP)
+	RET
+
+// func haveSSE42() bool
+TEXT ·haveSSE42(SB),NOSPLIT,$0
+	XORQ AX, AX
+	INCL AX
+	CPUID
+	SHRQ $20, CX
+	ANDQ $1, CX
+	MOVB CX, ret+0(FP)
+	RET
+

src/pkg/hash/crc32/crc32_amd64.go → src/pkg/hash/crc32/crc32_amd64x.go

@@ -2,6 +2,8 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
+// +build amd64 amd64p32
+
 package crc32
 
 // This file contains the code to call the SSE 4.2 version of the Castagnoli

src/pkg/log/syslog/syslog.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build !windows,!plan9
+// +build !windows,!nacl,!plan9
 
 // Package syslog provides a simple interface to the system log
 // service. It can send messages to the syslog daemon using UNIX

src/pkg/log/syslog/syslog_test.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build !windows,!plan9
+// +build !windows,!nacl,!plan9
 
 package syslog
 

src/pkg/log/syslog/syslog_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build !windows,!plan9
+// +build !windows,!nacl,!plan9
 
 package syslog
 

src/pkg/mime/type_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package mime
 

src/pkg/net/conn_test.go

@@ -33,7 +33,7 @@ func TestConnAndListener(t *testing.T) {
 		switch tt.net {
 		case "unix", "unixpacket":
 			switch runtime.GOOS {
-			case "plan9", "windows":
+			case "plan9", "windows", "nacl":
 				continue
 			}
 			if tt.net == "unixpacket" && runtime.GOOS != "linux" {

src/pkg/net/dnsclient_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 // DNS client: see RFC 1035.
 // Has to be linked into package net for Dial.

src/pkg/net/dnsconfig_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 // Read system DNS config from /etc/resolv.conf
 

src/pkg/net/fd_poll_nacl.go

+// Copyright 2013 The Go Authors.  All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package net
+
+import (
+	"syscall"
+	"time"
+)
+
+type pollDesc struct {
+	fd      *netFD
+	closing bool
+}
+
+func (pd *pollDesc) Init(fd *netFD) error { pd.fd = fd; return nil }
+
+func (pd *pollDesc) Close() {}
+
+func (pd *pollDesc) Lock() {}
+
+func (pd *pollDesc) Unlock() {}
+
+func (pd *pollDesc) Wakeup() {}
+
+func (pd *pollDesc) Evict() bool {
+	pd.closing = true
+	if pd.fd != nil {
+		syscall.StopIO(pd.fd.sysfd)
+	}
+	return false
+}
+
+func (pd *pollDesc) Prepare(mode int) error {
+	if pd.closing {
+		return errClosing
+	}
+	return nil
+}
+
+func (pd *pollDesc) PrepareRead() error { return pd.Prepare('r') }
+
+func (pd *pollDesc) PrepareWrite() error { return pd.Prepare('w') }
+
+func (pd *pollDesc) Wait(mode int) error {
+	if pd.closing {
+		return errClosing
+	}
+	return errTimeout
+}
+
+func (pd *pollDesc) WaitRead() error { return pd.Wait('r') }
+
+func (pd *pollDesc) WaitWrite() error { return pd.Wait('w') }
+
+func (pd *pollDesc) WaitCanceled(mode int) {}
+
+func (pd *pollDesc) WaitCanceledRead() {}
+
+func (pd *pollDesc) WaitCanceledWrite() {}
+
+func (fd *netFD) setDeadline(t time.Time) error {
+	return setDeadlineImpl(fd, t, 'r'+'w')
+}
+
+func (fd *netFD) setReadDeadline(t time.Time) error {
+	return setDeadlineImpl(fd, t, 'r')
+}
+
+func (fd *netFD) setWriteDeadline(t time.Time) error {
+	return setDeadlineImpl(fd, t, 'w')
+}
+
+func setDeadlineImpl(fd *netFD, t time.Time, mode int) error {
+	d := t.UnixNano()
+	if t.IsZero() {
+		d = 0
+	}
+	if err := fd.incref(); err != nil {
+		return err
+	}
+	switch mode {
+	case 'r':
+		syscall.SetReadDeadline(fd.sysfd, d)
+	case 'w':
+		syscall.SetWriteDeadline(fd.sysfd, d)
+	case 'r' + 'w':
+		syscall.SetReadDeadline(fd.sysfd, d)
+		syscall.SetWriteDeadline(fd.sysfd, d)
+	}
+	fd.decref()
+	return nil
+}

src/pkg/net/fd_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package net
 

src/pkg/net/file_test.go

@@ -181,7 +181,7 @@ var filePacketConnTests = []struct {
 
 func TestFilePacketConn(t *testing.T) {
 	switch runtime.GOOS {
-	case "plan9", "windows":
+	case "plan9", "windows", "nacl":
 		t.Skipf("skipping test on %q", runtime.GOOS)
 	}
 

src/pkg/net/file_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package net
 
@@ -24,6 +24,7 @@ func newFileFD(f *os.File) (*netFD, error) {
 
 	sotype, err := syscall.GetsockoptInt(fd, syscall.SOL_SOCKET, syscall.SO_TYPE)
 	if err != nil {
+		println("getsockopt failed", fd, err.Error())
 		closesocket(fd)
 		return nil, os.NewSyscallError("getsockopt", err)
 	}

src/pkg/net/interface_stub.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build plan9 solaris
+// +build nacl plan9 solaris
 
 package net
 

src/pkg/net/ipraw_test.go

@@ -247,7 +247,7 @@ var ipConnLocalNameTests = []struct {
 
 func TestIPConnLocalName(t *testing.T) {
 	switch runtime.GOOS {
-	case "plan9", "windows":
+	case "plan9", "windows", "nacl":
 		t.Skipf("skipping test on %q", runtime.GOOS)
 	default:
 		if os.Getuid() != 0 {

src/pkg/net/iprawsock_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/net/ipsock_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 // Internet protocol family sockets for POSIX
 

src/pkg/net/lookup_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package net
 

src/pkg/net/multicast_test.go

@@ -25,7 +25,7 @@ var ipv4MulticastListenerTests = []struct {
 // port.
 func TestIPv4MulticastListener(t *testing.T) {
 	switch runtime.GOOS {
-	case "plan9":
+	case "nacl", "plan9":
 		t.Skipf("skipping test on %q", runtime.GOOS)
 	case "solaris":
 		t.Skipf("skipping test on solaris, see issue 7399")

src/pkg/net/net_test.go

@@ -62,7 +62,7 @@ func TestShutdown(t *testing.T) {
 
 func TestShutdownUnix(t *testing.T) {
 	switch runtime.GOOS {
-	case "windows", "plan9":
+	case "nacl", "plan9", "windows":
 		t.Skipf("skipping test on %q", runtime.GOOS)
 	}
 	f, err := ioutil.TempFile("", "go_net_unixtest")

src/pkg/net/port_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 // Read system port mappings from /etc/services
 

src/pkg/net/sendfile_stub.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin netbsd openbsd solaris
+// +build darwin nacl netbsd openbsd solaris
 
 package net
 

src/pkg/net/sock_bsd.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd netbsd openbsd
+// +build darwin dragonfly freebsd nacl netbsd openbsd
 
 package net
 

src/pkg/net/sock_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/net/sockopt_bsd.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd netbsd openbsd
+// +build darwin dragonfly freebsd nacl netbsd openbsd
 
 package net
 

src/pkg/net/sockopt_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/net/sockoptip_bsd.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd netbsd openbsd solaris
+// +build darwin dragonfly freebsd nacl netbsd openbsd solaris
 
 package net
 

src/pkg/net/sockoptip_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/net/sys_cloexec.go

@@ -5,7 +5,7 @@
 // This file implements sysSocket and accept for platforms that do not
 // provide a fast path for setting SetNonblock and CloseOnExec.
 
-// +build darwin dragonfly freebsd netbsd openbsd solaris
+// +build darwin dragonfly freebsd nacl netbsd openbsd solaris
 
 package net
 

src/pkg/net/tcpsock_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/net/tcpsockopt_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/net/tcpsockopt_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build dragonfly freebsd linux netbsd
+// +build dragonfly freebsd linux nacl netbsd
 
 package net
 

src/pkg/net/udpsock_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/net/unixsock_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/os/dir_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package os
 

src/pkg/os/error_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package os
 

src/pkg/os/exec/lp_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package exec
 

src/pkg/os/exec_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package os
 

src/pkg/os/exec_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package os
 

src/pkg/os/file_posix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package os
 

src/pkg/os/file_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package os
 

src/pkg/os/path_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package os
 

src/pkg/os/pipe_bsd.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd netbsd openbsd solaris
+// +build darwin dragonfly freebsd nacl netbsd openbsd solaris
 
 package os
 

src/pkg/os/signal/sig.s

@@ -4,7 +4,7 @@
 
 // Assembly to get into package runtime without using exported symbols.
 
-// +build amd64 arm 386
+// +build amd64 amd64p32 arm 386
 
 #include "../../../cmd/ld/textflag.h"
 

src/pkg/os/signal/signal_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package signal
 

src/pkg/os/stat_nacl.go

+// Copyright 2009 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package os
+
+import (
+	"syscall"
+	"time"
+)
+
+func sameFile(fs1, fs2 *fileStat) bool {
+	stat1 := fs1.sys.(*syscall.Stat_t)
+	stat2 := fs2.sys.(*syscall.Stat_t)
+	return stat1.Dev == stat2.Dev && stat1.Ino == stat2.Ino
+}
+
+func fileInfoFromStat(st *syscall.Stat_t, name string) FileInfo {
+	fs := &fileStat{
+		name:    basename(name),
+		size:    int64(st.Size),
+		modTime: timespecToTime(st.Mtime, st.MtimeNsec),
+		sys:     st,
+	}
+	fs.mode = FileMode(st.Mode & 0777)
+	switch st.Mode & syscall.S_IFMT {
+	case syscall.S_IFBLK:
+		fs.mode |= ModeDevice
+	case syscall.S_IFCHR:
+		fs.mode |= ModeDevice | ModeCharDevice
+	case syscall.S_IFDIR:
+		fs.mode |= ModeDir
+	case syscall.S_IFIFO:
+		fs.mode |= ModeNamedPipe
+	case syscall.S_IFLNK:
+		fs.mode |= ModeSymlink
+	case syscall.S_IFREG:
+		// nothing to do
+	case syscall.S_IFSOCK:
+		fs.mode |= ModeSocket
+	}
+	if st.Mode&syscall.S_ISGID != 0 {
+		fs.mode |= ModeSetgid
+	}
+	if st.Mode&syscall.S_ISUID != 0 {
+		fs.mode |= ModeSetuid
+	}
+	if st.Mode&syscall.S_ISVTX != 0 {
+		fs.mode |= ModeSticky
+	}
+	return fs
+}
+
+func timespecToTime(sec, nsec int64) time.Time {
+	return time.Unix(sec, nsec)
+}
+
+// For testing.
+func atime(fi FileInfo) time.Time {
+	st := fi.Sys().(*syscall.Stat_t)
+	return timespecToTime(st.Atime, st.AtimeNsec)
+}

src/pkg/os/sys_bsd.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd netbsd openbsd
+// +build darwin dragonfly freebsd nacl netbsd openbsd
 
 // os code shared between *BSD systems including OS X (Darwin)
 // and FreeBSD.

src/pkg/path/filepath/path_unix.go

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris
 
 package filepath
 

src/pkg/reflect/asm_amd64p32.s

+// Copyright 2012 The Go Authors.  All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#include "../../cmd/ld/textflag.h"
+
+// makeFuncStub is the code half of the function returned by MakeFunc.
+// See the comment on the declaration of makeFuncStub in makefunc.go
+// for more details.
+// No argsize here, gc generates argsize info at call site.
+TEXT ·makeFuncStub(SB),(NOSPLIT|WRAPPER),$8
+	MOVL	DX, 0(SP)
+	LEAL	argframe+0(FP), CX
+	MOVL	CX, 4(SP)
+	CALL	·callReflect(SB)
+	RET
+
+// methodValueCall is the code half of the function returned by makeMethodValue.
+// See the comment on the declaration of methodValueCall in makefunc.go
+// for more details.
+// No argsize here, gc generates argsize info at call site.
+TEXT ·methodValueCall(SB),(NOSPLIT|WRAPPER),$8
+	MOVL	DX, 0(SP)
+	LEAL	argframe+0(FP), CX
+	MOVL	CX, 4(SP)
+	CALL	·callMethod(SB)
+	RET

src/pkg/reflect/type.go

@@ -16,6 +16,7 @@
 package reflect
 
 import (
+	"runtime"
 	"strconv"
 	"sync"
 	"unsafe"
@@ -1572,6 +1573,10 @@ func bucketOf(ktyp, etyp *rtype) *rtype {
 	gc = append(gc, _GC_PTR, offset, 0 /*self pointer set below*/) // overflow
 	offset += ptrsize
 
+	if runtime.GOARCH == "amd64p32" {
+		offset += 4
+	}
+
 	// keys
 	if ktyp.kind&kindNoPointers == 0 {
 		gc = append(gc, _GC_ARRAY_START, offset, _BUCKETSIZE, ktyp.size)

src/pkg/runtime/arch_386.h

@@ -7,5 +7,10 @@ enum {
 	BigEndian = 0,
 	CacheLineSize = 64,
 	RuntimeGogoBytes = 64,
+#ifdef GOOS_nacl
+	PhysPageSize = 65536,
+#else
+	PhysPageSize = 4096,
+#endif
 	PCQuantum = 1
 };

src/pkg/runtime/arch_amd64.h

@@ -10,6 +10,11 @@ enum {
 	RuntimeGogoBytes = 80,
 #else
 	RuntimeGogoBytes = 64,
+#endif
+#ifdef GOOS_nacl
+	PhysPageSize = 65536,
+#else
+	PhysPageSize = 4096,
 #endif
 	PCQuantum = 1
 };

src/pkg/runtime/arch_amd64p32.h

+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+enum {
+	thechar = '6',
+	BigEndian = 0,
+	CacheLineSize = 64,
+#ifdef GOOS_solaris
+	RuntimeGogoBytes = 80,
+#else
+	RuntimeGogoBytes = 64,
+#endif
+#ifdef GOOS_nacl
+	PhysPageSize = 65536,
+#else
+	PhysPageSize = 4096,
+#endif
+	PCQuantum = 1
+};

src/pkg/runtime/arch_arm.h

@@ -7,5 +7,6 @@ enum {
 	BigEndian = 0,
 	CacheLineSize = 32,
 	RuntimeGogoBytes = 80,
+	PhysPageSize = 4096,
 	PCQuantum = 4
 };

src/pkg/runtime/mem_linux.c

@@ -11,6 +11,7 @@
 enum
 {
 	_PAGE_SIZE = 4096,
+	EACCES = 13,
 };
 
 static int32

src/pkg/runtime/mem_nacl.c

+// Copyright 2010 The Go Authors.  All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#include "runtime.h"
+#include "arch_GOARCH.h"
+#include "defs_GOOS_GOARCH.h"
+#include "os_GOOS.h"
+#include "malloc.h"
+
+enum
+{
+	Debug = 0,
+};
+
+void*
+runtime·SysAlloc(uintptr n, uint64 *stat)
+{
+	void *v;
+
+	v = runtime·mmap(nil, n, PROT_READ|PROT_WRITE, MAP_ANON|MAP_PRIVATE, -1, 0);
+	if(v < (void*)4096) {
+		if(Debug)
+			runtime·printf("SysAlloc(%p): %p\n", n, v);
+		return nil;
+	}
+	runtime·xadd64(stat, n);
+	if(Debug)
+		runtime·printf("SysAlloc(%p) = %p\n", n, v);
+	return v;
+}
+
+void
+runtime·SysUnused(void *v, uintptr n)
+{
+	if(Debug)
+		runtime·printf("SysUnused(%p, %p)\n", v, n);
+}
+
+void
+runtime·SysUsed(void *v, uintptr n)
+{
+	USED(v);
+	USED(n);
+}
+
+void
+runtime·SysFree(void *v, uintptr n, uint64 *stat)
+{
+	if(Debug)
+		runtime·printf("SysFree(%p, %p)\n", v, n);
+	runtime·xadd64(stat, -(uint64)n);
+	runtime·munmap(v, n);
+}
+
+void*
+runtime·SysReserve(void *v, uintptr n)
+{
+	void *p;
+
+	// On 64-bit, people with ulimit -v set complain if we reserve too
+	// much address space.  Instead, assume that the reservation is okay
+	// and check the assumption in SysMap.
+	if(NaCl || sizeof(void*) == 8)
+		return v;
+	
+	p = runtime·mmap(v, n, PROT_NONE, MAP_ANON|MAP_PRIVATE, -1, 0);
+	if(p < (void*)4096)
+		return nil;
+	return p;
+}
+
+void
+runtime·SysMap(void *v, uintptr n, uint64 *stat)
+{
+	void *p;
+	
+	runtime·xadd64(stat, n);
+
+	// On 64-bit, we don't actually have v reserved, so tread carefully.
+	if(sizeof(void*) == 8) {
+		p = runtime·mmap(v, n, PROT_READ|PROT_WRITE, MAP_ANON|MAP_PRIVATE, -1, 0);
+		if(p == (void*)ENOMEM) {
+			runtime·printf("SysMap(%p, %p): %p\n", v, n, p);
+			runtime·throw("runtime: out of memory");
+		}
+		if(p != v) {
+			runtime·printf("SysMap(%p, %p): %p\n", v, n, p);
+			runtime·printf("runtime: address space conflict: map(%p) = %p\n", v, p);
+			runtime·throw("runtime: address space conflict");
+		}
+		if(Debug)
+			runtime·printf("SysMap(%p, %p) = %p\n", v, n, p);
+		return;
+	}
+
+	p = runtime·mmap(v, n, PROT_READ|PROT_WRITE, MAP_ANON|MAP_FIXED|MAP_PRIVATE, -1, 0);
+	if(p == (void*)ENOMEM) {
+		runtime·printf("SysMap(%p, %p): %p\n", v, n, p);
+		runtime·throw("runtime: out of memory");
+	}
+	if(p != v) {
+		runtime·printf("SysMap(%p, %p): %p\n", v, n, p);
+		runtime·printf("mmap MAP_FIXED %p returned %p\n", v, p);
+		runtime·throw("runtime: cannot map pages in arena address space");
+	}
+	if(Debug)
+		runtime·printf("SysMap(%p, %p) = %p\n", v, n, p);
+}

src/pkg/runtime/mgc0.c

@@ -2771,6 +2771,7 @@ runtime·MHeap_MapBits(MHeap *h)
 
 	n = (h->arena_used - h->arena_start) / wordsPerBitmapWord;
 	n = ROUND(n, bitmapChunk);
+	n = ROUND(n, PhysPageSize);
 	if(h->bitmap_mapped >= n)
 		return;
 

src/pkg/runtime/mheap.c

@@ -82,7 +82,7 @@ runtime·MHeap_MapSpans(MHeap *h)
 	n = (uintptr)h->arena_used;
 	n -= (uintptr)h->arena_start;
 	n = n / PageSize * sizeof(h->spans[0]);
-	n = ROUND(n, PageSize);
+	n = ROUND(n, PhysPageSize);
 	if(h->spans_mapped >= n)
 		return;
 	runtime·SysMap((byte*)h->spans + h->spans_mapped, n - h->spans_mapped, &mstats.other_sys);

src/pkg/runtime/mknacl.sh

+#!/bin/bash
+# Copyright 2013 The Go Authors.  All rights reserved.
+# Use of this source code is governed by a BSD-style
+# license that can be found in the LICENSE file.
+
+cat /Users/rsc/pub/native_client/src/trusted/service_runtime/include/bits/nacl_syscalls.h |
+	awk '
+	BEGIN {
+		printf("// generated by mknacl.sh - do not edit\n")
+	}
+	NF==3 && $1=="#define" && $2~/^NACL_sys_/ {
+		name=$2
+		sub(/^NACL_sys_/, "SYS_", name)
+		printf("#define %s %s\n", name, $3)
+	}' >syscall_nacl.h

src/pkg/runtime/netpoll.goc

@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build darwin dragonfly freebsd linux netbsd openbsd solaris windows
+// +build darwin dragonfly freebsd linux nacl netbsd openbsd solaris windows
 
 package net
 

src/pkg/runtime/netpoll_nacl.c

+// Copyright 2013 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#include "runtime.h"
+#include "defs_GOOS_GOARCH.h"
+#include "os_GOOS.h"
+
+// Fake network poller for NaCl.
+// Should never be used, because NaCl network connections do not honor "SetNonblock".
+
+void
+runtime·netpollinit(void)
+{
+}
+
+int32
+runtime·netpollopen(uintptr fd, PollDesc *pd)
+{
+	USED(fd);
+	USED(pd);
+	return 0;
+}
+
+int32
+runtime·netpollclose(uintptr fd)
+{
+	USED(fd);
+	return 0;
+}
+
+G*
+runtime·netpoll(bool block)
+{
+	USED(block);
+	return nil;
+}

src/pkg/runtime/os_nacl.c

+// Copyright 2010 The Go Authors.  All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+#include "runtime.h"
+#include "defs_GOOS_GOARCH.h"
+#include "os_GOOS.h"
+#include "arch_GOARCH.h"
+#include "../../cmd/ld/textflag.h"
+#include "stack.h"
+
+int8 *goos = "nacl";
+extern SigTab runtime·sigtab[];
+
+void runtime·sigtramp(void);
+
+// Called to initialize a new m (including the bootstrap m).
+// Called on the parent thread (main thread in case of bootstrap), can allocate memory.
+void
+runtime·mpreinit(M *mp)
+{
+	mp->gsignal = runtime·malg(32*1024);	// OS X wants >=8K, Linux >=2K
+}
+
+// Called to initialize a new m (including the bootstrap m).
+// Called on the new thread, can not allocate memory.
+void
+runtime·minit(void)
+{
+	int32 ret;
+
+	// Initialize signal handling
+	ret = runtime·nacl_exception_stack((byte*)m->gsignal->stackguard - StackGuard, 32*1024);
+	if(ret < 0)
+		runtime·printf("runtime: nacl_exception_stack: error %d\n", -ret);
+
+	ret = runtime·nacl_exception_handler(runtime·sigtramp, nil);
+	if(ret < 0)
+		runtime·printf("runtime: nacl_exception_handler: error %d\n", -ret);
+}
+
+// Called from dropm to undo the effect of an minit.
+void
+runtime·unminit(void)
+{
+}
+
+int8 runtime·sigtrampf[] = "runtime: signal at PC=%X AX=%X CX=%X DX=%X BX=%X DI=%X R15=%X *SP=%X\n";
+int8 runtime·sigtrampp[] = "runtime: sigtramp";
+
+extern byte runtime·tls0[];
+
+void
+runtime·osinit(void)
+{
+	runtime·ncpu = 1;
+	m->procid = 2;
+//runtime·nacl_exception_handler(runtime·sigtramp, nil);
+}
+
+void
+runtime·crash(void)
+{
+	*(int32*)0 = 0;
+}
+
+void
+runtime·get_random_data(byte **rnd, int32 *rnd_len)
+{
+	*rnd = nil;
+	*rnd_len = 0;
+}
+
+void
+runtime·goenvs(void)
+{
+	runtime·goenvs_unix();
+}
+
+void
+runtime·initsig(void)
+{
+}
+
+#pragma textflag NOSPLIT
+void
+runtime·usleep(uint32 us)
+{
+	Timespec ts;
+	
+	ts.tv_sec = us/1000000;
+	ts.tv_nsec = (us%1000000)*1000;
+	runtime·nacl_nanosleep(&ts, nil);
+}
+
+void runtime·mstart_nacl(void);
+
+void
+runtime·newosproc(M *mp, void *stk)
+{
+	int32 ret;
+	void **tls;
+
+	tls = (void**)mp->tls;
+	tls[0] = mp->g0;
+	tls[1] = mp;
+	ret = runtime·nacl_thread_create(runtime·mstart_nacl, stk, tls+2, 0);
+	if(ret < 0) {
+		runtime·printf("nacl_thread_create: error %d\n", -ret);
+		runtime·throw("newosproc");
+	}
+}
+
+uintptr
+runtime·semacreate(void)
+{
+	int32 mu, cond;
+	
+	mu = runtime·nacl_mutex_create(0);
+	if(mu < 0) {
+		runtime·printf("nacl_mutex_create: error %d\n", -mu);
+		runtime·throw("semacreate");
+	}
+	cond = runtime·nacl_cond_create(0);
+	if(cond < 0) {
+		runtime·printf("nacl_cond_create: error %d\n", -cond);
+		runtime·throw("semacreate");
+	}
+	m->waitsemalock = mu;
+	return cond; // assigned to m->waitsema
+}
+
+#pragma textflag NOSPLIT
+int32
+runtime·semasleep(int64 ns)
+{
+	int32 ret;
+	
+	ret = runtime·nacl_mutex_lock(m->waitsemalock);
+	if(ret < 0) {
+		//runtime·printf("nacl_mutex_lock: error %d\n", -ret);
+		runtime·throw("semasleep");
+	}
+	if(m->waitsemacount > 0) {
+		m->waitsemacount = 0;
+		runtime·nacl_mutex_unlock(m->waitsemalock);
+		return 0;
+	}
+
+	while(m->waitsemacount == 0) {
+		if(ns < 0) {
+			ret = runtime·nacl_cond_wait(m->waitsema, m->waitsemalock);
+			if(ret < 0) {
+				//runtime·printf("nacl_cond_wait: error %d\n", -ret);
+				runtime·throw("semasleep");
+			}
+		} else {
+			Timespec ts;
+			
+			ns += runtime·nanotime();
+			ts.tv_sec = runtime·timediv(ns, 1000000000, (int32*)&ts.tv_nsec);
+			ret = runtime·nacl_cond_timed_wait_abs(m->waitsema, m->waitsemalock, &ts);
+			if(ret == -ETIMEDOUT) {
+				runtime·nacl_mutex_unlock(m->waitsemalock);
+				return -1;
+			}
+			if(ret < 0) {
+				//runtime·printf("nacl_cond_timed_wait_abs: error %d\n", -ret);
+				runtime·throw("semasleep");
+			}
+		}
+	}
+			
+	m->waitsemacount = 0;
+	runtime·nacl_mutex_unlock(m->waitsemalock);
+	return 0;
+}
+
+void
+runtime·semawakeup(M *mp)
+{
+	int32 ret;
+	
+	ret = runtime·nacl_mutex_lock(mp->waitsemalock);
+	if(ret < 0) {
+		//runtime·printf("nacl_mutex_lock: error %d\n", -ret);
+		runtime·throw("semawakeup");
+	}
+	if(mp->waitsemacount != 0) {
+		//runtime·printf("semawakeup: double wakeup\n");
+		runtime·throw("semawakeup");
+	}
+	mp->waitsemacount = 1;
+	runtime·nacl_cond_signal(mp->waitsema);
+	runtime·nacl_mutex_unlock(mp->waitsemalock);
+}
+
+void
+os·sigpipe(void)
+{
+	runtime·throw("too many writes on closed pipe");
+}
+
+uintptr
+runtime·memlimit(void)
+{
+	runtime·printf("memlimit\n");
+	return 0;
+}
+
+#pragma dataflag NOPTR
+static int8 badsignal[] = "runtime: signal received on thread not created by Go.\n";
+
+// This runs on a foreign stack, without an m or a g.  No stack split.
+#pragma textflag NOSPLIT
+void
+runtime·badsignal2(void)
+{
+	runtime·write(2, badsignal, sizeof badsignal - 1);
+	runtime·exit(2);
+}
+
+void	runtime·madvise(byte*, uintptr, int32) { }
+void runtime·munmap(byte*, uintptr) {}
+
+void
+runtime·resetcpuprofiler(int32 hz)
+{
+	USED(hz);
+}
+
+void
+runtime·sigdisable(uint32)
+{
+}
+
+void
+runtime·sigenable(uint32)
+{
+}
+
+void
+runtime·closeonexec(int32)
+{
+}
+
+void
+runtime·sigpanic(void)
+{
+	// Native Client only invokes the exception handler for memory faults.
+	g->sig = SIGSEGV;
+	if(g->sigpc == 0)
+		runtime·panicstring("call of nil func value");
+	runtime·panicstring("invalid memory address or nil pointer dereference");
+}
+
+uint32 runtime·writelock; // test-and-set spin lock for runtime.write
+
+/*
+An attempt at IRT. Doesn't work. See end of sys_nacl_amd64.s.
+
+void (*runtime·nacl_irt_query)(void);
+
+int8 runtime·nacl_irt_basic_v0_1_str[] = "nacl-irt-basic-0.1";
+void *runtime·nacl_irt_basic_v0_1[6]; // exit, gettod, clock, nanosleep, sched_yield, sysconf
+int32 runtime·nacl_irt_basic_v0_1_size = sizeof(runtime·nacl_irt_basic_v0_1);
+
+int8 runtime·nacl_irt_memory_v0_3_str[] = "nacl-irt-memory-0.3";
+void *runtime·nacl_irt_memory_v0_3[3]; // mmap, munmap, mprotect
+int32 runtime·nacl_irt_memory_v0_3_size = sizeof(runtime·nacl_irt_memory_v0_3);
+
+int8 runtime·nacl_irt_thread_v0_1_str[] = "nacl-irt-thread-0.1";
+void *runtime·nacl_irt_thread_v0_1[3]; // thread_create, thread_exit, thread_nice
+int32 runtime·nacl_irt_thread_v0_1_size = sizeof(runtime·nacl_irt_thread_v0_1);
+*/
\ No newline at end of file