diff --git a/src/cmd/5l/5.out.h b/src/cmd/5l/5.out.h
index eda379c6f6..85dd17a8e5 100644
--- a/src/cmd/5l/5.out.h
+++ b/src/cmd/5l/5.out.h
@@ -31,12 +31,7 @@
 #define	NSNAME		8
 #define	NSYM		50
 #define	NREG		16
-
-#define NOPROF		(1<<0)
-#define DUPOK		(1<<1)
-#define NOSPLIT		(1<<2)
-#define RODATA	(1<<3)
-#define NOPTR	(1<<4)
+#include "../ld/textflag.h"
 
 #define	REGRET		0
 /* -1 disables use of REGARG */
diff --git a/src/cmd/6l/6.out.h b/src/cmd/6l/6.out.h
index b95b3fd13b..b96be60240 100644
--- a/src/cmd/6l/6.out.h
+++ b/src/cmd/6l/6.out.h
@@ -30,11 +30,7 @@
 
 #define	NSYM	50
 #define	NSNAME	8
-#define NOPROF	(1<<0)
-#define DUPOK	(1<<1)
-#define NOSPLIT	(1<<2)
-#define RODATA	(1<<3)
-#define NOPTR	(1<<4)
+#include "../ld/textflag.h"
 
 /*
  *	amd64
diff --git a/src/cmd/8l/8.out.h b/src/cmd/8l/8.out.h
index 7683d50ada..a804ae94ba 100644
--- a/src/cmd/8l/8.out.h
+++ b/src/cmd/8l/8.out.h
@@ -30,11 +30,7 @@
 
 #define	NSYM	50
 #define	NSNAME	8
-#define NOPROF	(1<<0)
-#define DUPOK	(1<<1)
-#define NOSPLIT	(1<<2)
-#define RODATA	(1<<3)
-#define NOPTR	(1<<4)
+#include "../ld/textflag.h"
 
 enum	as
 {
diff --git a/src/cmd/ld/textflag.h b/src/cmd/ld/textflag.h
new file mode 100644
index 0000000000..7b16865be4
--- /dev/null
+++ b/src/cmd/ld/textflag.h
@@ -0,0 +1,19 @@
+// 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.
+
+// This file defines flags attached to various functions
+// and data objects.  The compilers, assemblers, and linker must
+// all agree on these values.
+
+// Don't profile the marked routine.  This flag is deprecated.
+#define NOPROF	(1<<0)
+// It is ok for the linker to get multiple of these symbols.  It will
+// pick one of the duplicates to use.
+#define DUPOK	(1<<1)
+// Don't insert stack check preamble.
+#define NOSPLIT	(1<<2)
+// Put this data in a read-only section.
+#define RODATA	(1<<3)
+// This data contains no pointers.
+#define NOPTR	(1<<4)
diff --git a/src/pkg/runtime/asm_386.s b/src/pkg/runtime/asm_386.s
index a441afc208..904287e69e 100644
--- a/src/pkg/runtime/asm_386.s
+++ b/src/pkg/runtime/asm_386.s
@@ -4,8 +4,9 @@
 
 #include "zasm_GOOS_GOARCH.h"
 #include "funcdata.h"
+#include "../../cmd/ld/textflag.h"
 
-TEXT _rt0_go(SB),7,$0
+TEXT _rt0_go(SB),NOSPLIT,$0
 	// copy arguments forward on an even stack
 	MOVL	argc+0(FP), AX
 	MOVL	argv+4(FP), BX
@@ -109,13 +110,13 @@ ok:
 	RET
 
 DATA	runtime·main·f+0(SB)/4,$runtime·main(SB)
-GLOBL	runtime·main·f(SB),8,$4
+GLOBL	runtime·main·f(SB),RODATA,$4
 
-TEXT runtime·breakpoint(SB),7,$0-0
+TEXT runtime·breakpoint(SB),NOSPLIT,$0-0
 	INT $3
 	RET
 
-TEXT runtime·asminit(SB),7,$0-0
+TEXT runtime·asminit(SB),NOSPLIT,$0-0
 	// Linux and MinGW start the FPU in extended double precision.
 	// Other operating systems use double precision.
 	// Change to double precision to match them,
@@ -131,7 +132,7 @@ TEXT runtime·asminit(SB),7,$0-0
 
 // void gosave(Gobuf*)
 // save state in Gobuf; setjmp
-TEXT runtime·gosave(SB), 7, $0-4
+TEXT runtime·gosave(SB), NOSPLIT, $0-4
 	MOVL	4(SP), AX		// gobuf
 	LEAL	4(SP), BX		// caller's SP
 	MOVL	BX, gobuf_sp(AX)
@@ -146,7 +147,7 @@ TEXT runtime·gosave(SB), 7, $0-4
 
 // void gogo(Gobuf*)
 // restore state from Gobuf; longjmp
-TEXT runtime·gogo(SB), 7, $0-4
+TEXT runtime·gogo(SB), NOSPLIT, $0-4
 	MOVL	4(SP), BX		// gobuf
 	MOVL	gobuf_g(BX), DX
 	MOVL	0(DX), CX		// make sure g != nil
@@ -165,7 +166,7 @@ TEXT runtime·gogo(SB), 7, $0-4
 // Switch to m->g0's stack, call fn(g).
 // Fn must never return.  It should gogo(&g->sched)
 // to keep running g.
-TEXT runtime·mcall(SB), 7, $0-4
+TEXT runtime·mcall(SB), NOSPLIT, $0-4
 	MOVL	fn+0(FP), DI
 	
 	get_tls(CX)
@@ -200,7 +201,7 @@ TEXT runtime·mcall(SB), 7, $0-4
 // the top of a stack (for example, morestack calling newstack
 // calling the scheduler calling newm calling gc), so we must
 // record an argument size. For that purpose, it has no arguments.
-TEXT runtime·morestack(SB),7,$0-0
+TEXT runtime·morestack(SB),NOSPLIT,$0-0
 	// Cannot grow scheduler stack (m->g0).
 	get_tls(CX)
 	MOVL	m(CX), BX
@@ -249,7 +250,7 @@ TEXT runtime·morestack(SB),7,$0-0
 // with the desired args running the desired function.
 //
 // func call(fn *byte, arg *byte, argsize uint32).
-TEXT runtime·newstackcall(SB), 7, $0-12
+TEXT runtime·newstackcall(SB), NOSPLIT, $0-12
 	get_tls(CX)
 	MOVL	m(CX), BX
 
@@ -304,7 +305,7 @@ TEXT runtime·newstackcall(SB), 7, $0-12
 	JMP	AX
 // Note: can't just "JMP runtime·NAME(SB)" - bad inlining results.
 
-TEXT reflect·call(SB), 7, $0-12
+TEXT reflect·call(SB), NOSPLIT, $0-12
 	MOVL	argsize+8(FP), CX
 	DISPATCH(call16, 16)
 	DISPATCH(call32, 32)
@@ -385,7 +386,7 @@ CALLFN(call1073741824, 1073741824)
 //
 // Lessstack can appear in stack traces for the same reason
 // as morestack; in that context, it has 0 arguments.
-TEXT runtime·lessstack(SB), 7, $0-0
+TEXT runtime·lessstack(SB), NOSPLIT, $0-0
 	// Save return value in m->cret
 	get_tls(CX)
 	MOVL	m(CX), BX
@@ -407,7 +408,7 @@ TEXT runtime·lessstack(SB), 7, $0-0
 //		return 1;
 //	}else
 //		return 0;
-TEXT runtime·cas(SB), 7, $0-12
+TEXT runtime·cas(SB), NOSPLIT, $0-12
 	MOVL	4(SP), BX
 	MOVL	8(SP), AX
 	MOVL	12(SP), CX
@@ -427,7 +428,7 @@ TEXT runtime·cas(SB), 7, $0-12
 //	} else {
 //		return 0;
 //	}
-TEXT runtime·cas64(SB), 7, $0-20
+TEXT runtime·cas64(SB), NOSPLIT, $0-20
 	MOVL	4(SP), BP
 	MOVL	8(SP), AX
 	MOVL	12(SP), DX
@@ -449,7 +450,7 @@ cas64_fail:
 //		return 1;
 //	}else
 //		return 0;
-TEXT runtime·casp(SB), 7, $0-12
+TEXT runtime·casp(SB), NOSPLIT, $0-12
 	MOVL	4(SP), BX
 	MOVL	8(SP), AX
 	MOVL	12(SP), CX
@@ -465,7 +466,7 @@ TEXT runtime·casp(SB), 7, $0-12
 // Atomically:
 //	*val += delta;
 //	return *val;
-TEXT runtime·xadd(SB), 7, $0-8
+TEXT runtime·xadd(SB), NOSPLIT, $0-8
 	MOVL	4(SP), BX
 	MOVL	8(SP), AX
 	MOVL	AX, CX
@@ -474,13 +475,13 @@ TEXT runtime·xadd(SB), 7, $0-8
 	ADDL	CX, AX
 	RET
 
-TEXT runtime·xchg(SB), 7, $0-8
+TEXT runtime·xchg(SB), NOSPLIT, $0-8
 	MOVL	4(SP), BX
 	MOVL	8(SP), AX
 	XCHGL	AX, 0(BX)
 	RET
 
-TEXT runtime·procyield(SB),7,$0-0
+TEXT runtime·procyield(SB),NOSPLIT,$0-0
 	MOVL	4(SP), AX
 again:
 	PAUSE
@@ -488,13 +489,13 @@ again:
 	JNZ	again
 	RET
 
-TEXT runtime·atomicstorep(SB), 7, $0-8
+TEXT runtime·atomicstorep(SB), NOSPLIT, $0-8
 	MOVL	4(SP), BX
 	MOVL	8(SP), AX
 	XCHGL	AX, 0(BX)
 	RET
 
-TEXT runtime·atomicstore(SB), 7, $0-8
+TEXT runtime·atomicstore(SB), NOSPLIT, $0-8
 	MOVL	4(SP), BX
 	MOVL	8(SP), AX
 	XCHGL	AX, 0(BX)
@@ -503,7 +504,7 @@ TEXT runtime·atomicstore(SB), 7, $0-8
 // uint64 atomicload64(uint64 volatile* addr);
 // so actually
 // void atomicload64(uint64 *res, uint64 volatile *addr);
-TEXT runtime·atomicload64(SB), 7, $0-8
+TEXT runtime·atomicload64(SB), NOSPLIT, $0-8
 	MOVL	4(SP), BX
 	MOVL	8(SP), AX
 	// MOVQ (%EAX), %MM0
@@ -515,7 +516,7 @@ TEXT runtime·atomicload64(SB), 7, $0-8
 	RET
 
 // void runtime·atomicstore64(uint64 volatile* addr, uint64 v);
-TEXT runtime·atomicstore64(SB), 7, $0-12
+TEXT runtime·atomicstore64(SB), NOSPLIT, $0-12
 	MOVL	4(SP), AX
 	// MOVQ and EMMS were introduced on the Pentium MMX.
 	// MOVQ 0x8(%ESP), %MM0
@@ -536,7 +537,7 @@ TEXT runtime·atomicstore64(SB), 7, $0-12
 // 1. pop the caller
 // 2. sub 5 bytes from the callers return
 // 3. jmp to the argument
-TEXT runtime·jmpdefer(SB), 7, $0
+TEXT runtime·jmpdefer(SB), NOSPLIT, $0
 	MOVL	4(SP), DX	// fn
 	MOVL	8(SP), BX	// caller sp
 	LEAL	-4(BX), SP	// caller sp after CALL
@@ -545,7 +546,7 @@ TEXT runtime·jmpdefer(SB), 7, $0
 	JMP	BX	// but first run the deferred function
 
 // Save state of caller into g->sched.
-TEXT gosave<>(SB),7,$0
+TEXT gosave<>(SB),NOSPLIT,$0
 	PUSHL	AX
 	PUSHL	BX
 	get_tls(BX)
@@ -564,7 +565,7 @@ TEXT gosave<>(SB),7,$0
 // Call fn(arg) on the scheduler stack,
 // aligned appropriately for the gcc ABI.
 // See cgocall.c for more details.
-TEXT runtime·asmcgocall(SB),7,$0-8
+TEXT runtime·asmcgocall(SB),NOSPLIT,$0-8
 	MOVL	fn+0(FP), AX
 	MOVL	arg+4(FP), BX
 	MOVL	SP, DX
@@ -600,7 +601,7 @@ TEXT runtime·asmcgocall(SB),7,$0-8
 // cgocallback(void (*fn)(void*), void *frame, uintptr framesize)
 // Turn the fn into a Go func (by taking its address) and call
 // cgocallback_gofunc.
-TEXT runtime·cgocallback(SB),7,$12-12
+TEXT runtime·cgocallback(SB),NOSPLIT,$12-12
 	LEAL	fn+0(FP), AX
 	MOVL	AX, 0(SP)
 	MOVL	frame+4(FP), AX
@@ -613,7 +614,7 @@ TEXT runtime·cgocallback(SB),7,$12-12
 
 // cgocallback_gofunc(FuncVal*, void *frame, uintptr framesize)
 // See cgocall.c for more details.
-TEXT runtime·cgocallback_gofunc(SB),7,$12-12
+TEXT runtime·cgocallback_gofunc(SB),NOSPLIT,$12-12
 	// If m is nil, Go did not create the current thread.
 	// Call needm to obtain one for temporary use.
 	// In this case, we're running on the thread stack, so there's
@@ -704,7 +705,7 @@ havem:
 	RET
 
 // void setmg(M*, G*); set m and g. for use by needm.
-TEXT runtime·setmg(SB), 7, $0-8
+TEXT runtime·setmg(SB), NOSPLIT, $0-8
 #ifdef GOOS_windows
 	MOVL	mm+0(FP), AX
 	CMPL	AX, $0
@@ -724,7 +725,7 @@ settls:
 	RET
 
 // void setmg_gcc(M*, G*); set m and g. for use by gcc
-TEXT setmg_gcc<>(SB), 7, $0
+TEXT setmg_gcc<>(SB), NOSPLIT, $0
 	get_tls(AX)
 	MOVL	mm+0(FP), DX
 	MOVL	DX, m(AX)
@@ -733,7 +734,7 @@ TEXT setmg_gcc<>(SB), 7, $0
 	RET
 
 // check that SP is in range [g->stackbase, g->stackguard)
-TEXT runtime·stackcheck(SB), 7, $0-0
+TEXT runtime·stackcheck(SB), NOSPLIT, $0-0
 	get_tls(CX)
 	MOVL	g(CX), AX
 	CMPL	g_stackbase(AX), SP
@@ -744,7 +745,7 @@ TEXT runtime·stackcheck(SB), 7, $0-0
 	INT	$3
 	RET
 
-TEXT runtime·memclr(SB),7,$0-8
+TEXT runtime·memclr(SB),NOSPLIT,$0-8
 	MOVL	4(SP), DI		// arg 1 addr
 	MOVL	8(SP), CX		// arg 2 count
 	MOVL	CX, BX
@@ -759,31 +760,31 @@ TEXT runtime·memclr(SB),7,$0-8
 	STOSB
 	RET
 
-TEXT runtime·getcallerpc(SB),7,$0-4
+TEXT runtime·getcallerpc(SB),NOSPLIT,$0-4
 	MOVL	x+0(FP),AX		// addr of first arg
 	MOVL	-4(AX),AX		// get calling pc
 	RET
 
-TEXT runtime·setcallerpc(SB),7,$0-8
+TEXT runtime·setcallerpc(SB),NOSPLIT,$0-8
 	MOVL	x+0(FP),AX		// addr of first arg
 	MOVL	x+4(FP), BX
 	MOVL	BX, -4(AX)		// set calling pc
 	RET
 
-TEXT runtime·getcallersp(SB), 7, $0-4
+TEXT runtime·getcallersp(SB), NOSPLIT, $0-4
 	MOVL	sp+0(FP), AX
 	RET
 
 // int64 runtime·cputicks(void), so really
 // void runtime·cputicks(int64 *ticks)
-TEXT runtime·cputicks(SB),7,$0-4
+TEXT runtime·cputicks(SB),NOSPLIT,$0-4
 	RDTSC
 	MOVL	ret+0(FP), DI
 	MOVL	AX, 0(DI)
 	MOVL	DX, 4(DI)
 	RET
 
-TEXT runtime·ldt0setup(SB),7,$16-0
+TEXT runtime·ldt0setup(SB),NOSPLIT,$16-0
 	// set up ldt 7 to point at tls0
 	// ldt 1 would be fine on Linux, but on OS X, 7 is as low as we can go.
 	// the entry number is just a hint.  setldt will set up GS with what it used.
@@ -797,10 +798,10 @@ TEXT runtime·ldt0setup(SB),7,$16-0
 TEXT runtime·emptyfunc(SB),0,$0-0
 	RET
 
-TEXT runtime·abort(SB),7,$0-0
+TEXT runtime·abort(SB),NOSPLIT,$0-0
 	INT $0x3
 
-TEXT runtime·stackguard(SB),7,$0-8
+TEXT runtime·stackguard(SB),NOSPLIT,$0-8
 	MOVL	SP, DX
 	MOVL	DX, sp+0(FP)
 	get_tls(CX)
@@ -812,13 +813,13 @@ TEXT runtime·stackguard(SB),7,$0-8
 GLOBL runtime·tls0(SB), $32
 
 // hash function using AES hardware instructions
-TEXT runtime·aeshash(SB),7,$0-12
+TEXT runtime·aeshash(SB),NOSPLIT,$0-12
 	MOVL	4(SP), DX	// ptr to hash value
 	MOVL	8(SP), CX	// size
 	MOVL	12(SP), AX	// ptr to data
 	JMP	runtime·aeshashbody(SB)
 
-TEXT runtime·aeshashstr(SB),7,$0-12
+TEXT runtime·aeshashstr(SB),NOSPLIT,$0-12
 	MOVL	4(SP), DX	// ptr to hash value
 	MOVL	12(SP), AX	// ptr to string struct
 	MOVL	4(AX), CX	// length of string
@@ -828,7 +829,7 @@ TEXT runtime·aeshashstr(SB),7,$0-12
 // AX: data
 // CX: length
 // DX: ptr to seed input / hash output
-TEXT runtime·aeshashbody(SB),7,$0-12
+TEXT runtime·aeshashbody(SB),NOSPLIT,$0-12
 	MOVL	(DX), X0	// seed to low 32 bits of xmm0
 	PINSRD	$1, CX, X0	// size to next 32 bits of xmm0
 	MOVO	runtime·aeskeysched+0(SB), X2
@@ -883,7 +884,7 @@ finalize:
 	MOVL	X0, (DX)
 	RET
 
-TEXT runtime·aeshash32(SB),7,$0-12
+TEXT runtime·aeshash32(SB),NOSPLIT,$0-12
 	MOVL	4(SP), DX	// ptr to hash value
 	MOVL	12(SP), AX	// ptr to data
 	MOVL	(DX), X0	// seed
@@ -894,7 +895,7 @@ TEXT runtime·aeshash32(SB),7,$0-12
 	MOVL	X0, (DX)
 	RET
 
-TEXT runtime·aeshash64(SB),7,$0-12
+TEXT runtime·aeshash64(SB),NOSPLIT,$0-12
 	MOVL	4(SP), DX	// ptr to hash value
 	MOVL	12(SP), AX	// ptr to data
 	MOVQ	(AX), X0	// data
@@ -986,7 +987,7 @@ DATA masks<>+0xf4(SB)/4, $0xffffffff
 DATA masks<>+0xf8(SB)/4, $0xffffffff
 DATA masks<>+0xfc(SB)/4, $0x00ffffff
 
-GLOBL masks<>(SB),8,$256
+GLOBL masks<>(SB),RODATA,$256
 
 // these are arguments to pshufb.  They move data down from
 // the high bytes of the register to the low bytes of the register.
@@ -1071,15 +1072,15 @@ DATA shifts<>+0xf4(SB)/4, $0x08070605
 DATA shifts<>+0xf8(SB)/4, $0x0c0b0a09
 DATA shifts<>+0xfc(SB)/4, $0xff0f0e0d
 
-GLOBL shifts<>(SB),8,$256
+GLOBL shifts<>(SB),RODATA,$256
 
-TEXT runtime·memeq(SB),7,$0-12
+TEXT runtime·memeq(SB),NOSPLIT,$0-12
 	MOVL	a+0(FP), SI
 	MOVL	b+4(FP), DI
 	MOVL	count+8(FP), BX
 	JMP	runtime·memeqbody(SB)
 
-TEXT bytes·Equal(SB),7,$0-25
+TEXT bytes·Equal(SB),NOSPLIT,$0-25
 	MOVL	a_len+4(FP), BX
 	MOVL	b_len+16(FP), CX
 	XORL	AX, AX
@@ -1095,7 +1096,7 @@ eqret:
 // a in SI
 // b in DI
 // count in BX
-TEXT runtime·memeqbody(SB),7,$0-0
+TEXT runtime·memeqbody(SB),NOSPLIT,$0-0
 	XORL	AX, AX
 
 	CMPL	BX, $4
@@ -1188,7 +1189,7 @@ equal:
 	SETEQ	AX
 	RET
 
-TEXT runtime·cmpstring(SB),7,$0-20
+TEXT runtime·cmpstring(SB),NOSPLIT,$0-20
 	MOVL	s1+0(FP), SI
 	MOVL	s1+4(FP), BX
 	MOVL	s2+8(FP), DI
@@ -1197,7 +1198,7 @@ TEXT runtime·cmpstring(SB),7,$0-20
 	MOVL	AX, res+16(FP)
 	RET
 
-TEXT bytes·Compare(SB),7,$0-28
+TEXT bytes·Compare(SB),NOSPLIT,$0-28
 	MOVL	s1+0(FP), SI
 	MOVL	s1+4(FP), BX
 	MOVL	s2+12(FP), DI
@@ -1206,7 +1207,7 @@ TEXT bytes·Compare(SB),7,$0-28
 	MOVL	AX, res+24(FP)
 	RET
 
-TEXT bytes·IndexByte(SB),7,$0
+TEXT bytes·IndexByte(SB),NOSPLIT,$0
 	MOVL	s+0(FP), SI
 	MOVL	s_len+4(FP), CX
 	MOVB	c+12(FP), AL
@@ -1220,7 +1221,7 @@ TEXT bytes·IndexByte(SB),7,$0
 	MOVL	DI, ret+16(FP)
 	RET
 
-TEXT strings·IndexByte(SB),7,$0
+TEXT strings·IndexByte(SB),NOSPLIT,$0
 	MOVL	s+0(FP), SI
 	MOVL	s_len+4(FP), CX
 	MOVB	c+8(FP), AL
@@ -1241,7 +1242,7 @@ TEXT strings·IndexByte(SB),7,$0
 //   DX = blen
 // output:
 //   AX = 1/0/-1
-TEXT runtime·cmpbody(SB),7,$0-0
+TEXT runtime·cmpbody(SB),NOSPLIT,$0-0
 	CMPL	SI, DI
 	JEQ	cmp_allsame
 	CMPL	BX, DX
diff --git a/src/pkg/runtime/asm_amd64.s b/src/pkg/runtime/asm_amd64.s
index 8280ac02a9..391a1129d9 100644
--- a/src/pkg/runtime/asm_amd64.s
+++ b/src/pkg/runtime/asm_amd64.s
@@ -4,8 +4,9 @@
 
 #include "zasm_GOOS_GOARCH.h"
 #include "funcdata.h"
+#include "../../cmd/ld/textflag.h"
 
-TEXT _rt0_go(SB),7,$0
+TEXT _rt0_go(SB),NOSPLIT,$0
 	// copy arguments forward on an even stack
 	MOVQ	DI, AX		// argc
 	MOVQ	SI, BX		// argv
@@ -102,13 +103,13 @@ ok:
 	RET
 
 DATA	runtime·main·f+0(SB)/8,$runtime·main(SB)
-GLOBL	runtime·main·f(SB),8,$8
+GLOBL	runtime·main·f(SB),RODATA,$8
 
-TEXT runtime·breakpoint(SB),7,$0-0
+TEXT runtime·breakpoint(SB),NOSPLIT,$0-0
 	BYTE	$0xcc
 	RET
 
-TEXT runtime·asminit(SB),7,$0-0
+TEXT runtime·asminit(SB),NOSPLIT,$0-0
 	// No per-thread init.
 	RET
 
@@ -118,7 +119,7 @@ TEXT runtime·asminit(SB),7,$0-0
 
 // void gosave(Gobuf*)
 // save state in Gobuf; setjmp
-TEXT runtime·gosave(SB), 7, $0-8
+TEXT runtime·gosave(SB), NOSPLIT, $0-8
 	MOVQ	8(SP), AX		// gobuf
 	LEAQ	8(SP), BX		// caller's SP
 	MOVQ	BX, gobuf_sp(AX)
@@ -133,7 +134,7 @@ TEXT runtime·gosave(SB), 7, $0-8
 
 // void gogo(Gobuf*)
 // restore state from Gobuf; longjmp
-TEXT runtime·gogo(SB), 7, $0-8
+TEXT runtime·gogo(SB), NOSPLIT, $0-8
 	MOVQ	8(SP), BX		// gobuf
 	MOVQ	gobuf_g(BX), DX
 	MOVQ	0(DX), CX		// make sure g != nil
@@ -152,7 +153,7 @@ TEXT runtime·gogo(SB), 7, $0-8
 // Switch to m->g0's stack, call fn(g).
 // Fn must never return.  It should gogo(&g->sched)
 // to keep running g.
-TEXT runtime·mcall(SB), 7, $0-8
+TEXT runtime·mcall(SB), NOSPLIT, $0-8
 	MOVQ	fn+0(FP), DI
 	
 	get_tls(CX)
@@ -191,7 +192,7 @@ TEXT runtime·mcall(SB), 7, $0-8
 // the top of a stack (for example, morestack calling newstack
 // calling the scheduler calling newm calling gc), so we must
 // record an argument size. For that purpose, it has no arguments.
-TEXT runtime·morestack(SB),7,$0-0
+TEXT runtime·morestack(SB),NOSPLIT,$0-0
 	// Cannot grow scheduler stack (m->g0).
 	MOVQ	m_g0(BX), SI
 	CMPQ	g(CX), SI
@@ -230,7 +231,7 @@ TEXT runtime·morestack(SB),7,$0-0
 // with the desired args running the desired function.
 //
 // func call(fn *byte, arg *byte, argsize uint32).
-TEXT runtime·newstackcall(SB), 7, $0-20
+TEXT runtime·newstackcall(SB), NOSPLIT, $0-20
 	get_tls(CX)
 	MOVQ	m(CX), BX
 
@@ -285,7 +286,7 @@ TEXT runtime·newstackcall(SB), 7, $0-20
 	JMP	AX
 // Note: can't just "JMP runtime·NAME(SB)" - bad inlining results.
 
-TEXT reflect·call(SB), 7, $0-20
+TEXT reflect·call(SB), NOSPLIT, $0-20
 	MOVLQZX argsize+16(FP), CX
 	DISPATCH(call16, 16)
 	DISPATCH(call32, 32)
@@ -366,7 +367,7 @@ CALLFN(call1073741824, 1073741824)
 //
 // Lessstack can appear in stack traces for the same reason
 // as morestack; in that context, it has 0 arguments.
-TEXT runtime·lessstack(SB), 7, $0-0
+TEXT runtime·lessstack(SB), NOSPLIT, $0-0
 	// Save return value in m->cret
 	get_tls(CX)
 	MOVQ	m(CX), BX
@@ -381,7 +382,7 @@ TEXT runtime·lessstack(SB), 7, $0-0
 	RET
 
 // morestack trampolines
-TEXT runtime·morestack00(SB),7,$0
+TEXT runtime·morestack00(SB),NOSPLIT,$0
 	get_tls(CX)
 	MOVQ	m(CX), BX
 	MOVQ	$0, AX
@@ -389,7 +390,7 @@ TEXT runtime·morestack00(SB),7,$0
 	MOVQ	$runtime·morestack(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack01(SB),7,$0
+TEXT runtime·morestack01(SB),NOSPLIT,$0
 	get_tls(CX)
 	MOVQ	m(CX), BX
 	SHLQ	$32, AX
@@ -397,7 +398,7 @@ TEXT runtime·morestack01(SB),7,$0
 	MOVQ	$runtime·morestack(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack10(SB),7,$0
+TEXT runtime·morestack10(SB),NOSPLIT,$0
 	get_tls(CX)
 	MOVQ	m(CX), BX
 	MOVLQZX	AX, AX
@@ -405,7 +406,7 @@ TEXT runtime·morestack10(SB),7,$0
 	MOVQ	$runtime·morestack(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack11(SB),7,$0
+TEXT runtime·morestack11(SB),NOSPLIT,$0
 	get_tls(CX)
 	MOVQ	m(CX), BX
 	MOVQ	AX, m_moreframesize(BX)
@@ -414,37 +415,37 @@ TEXT runtime·morestack11(SB),7,$0
 
 // subcases of morestack01
 // with const of 8,16,...48
-TEXT runtime·morestack8(SB),7,$0
+TEXT runtime·morestack8(SB),NOSPLIT,$0
 	MOVQ	$1, R8
 	MOVQ	$morestack<>(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack16(SB),7,$0
+TEXT runtime·morestack16(SB),NOSPLIT,$0
 	MOVQ	$2, R8
 	MOVQ	$morestack<>(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack24(SB),7,$0
+TEXT runtime·morestack24(SB),NOSPLIT,$0
 	MOVQ	$3, R8
 	MOVQ	$morestack<>(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack32(SB),7,$0
+TEXT runtime·morestack32(SB),NOSPLIT,$0
 	MOVQ	$4, R8
 	MOVQ	$morestack<>(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack40(SB),7,$0
+TEXT runtime·morestack40(SB),NOSPLIT,$0
 	MOVQ	$5, R8
 	MOVQ	$morestack<>(SB), AX
 	JMP	AX
 
-TEXT runtime·morestack48(SB),7,$0
+TEXT runtime·morestack48(SB),NOSPLIT,$0
 	MOVQ	$6, R8
 	MOVQ	$morestack<>(SB), AX
 	JMP	AX
 
-TEXT morestack<>(SB),7,$0
+TEXT morestack<>(SB),NOSPLIT,$0
 	get_tls(CX)
 	MOVQ	m(CX), BX
 	SHLQ	$35, R8
@@ -459,7 +460,7 @@ TEXT morestack<>(SB),7,$0
 //		return 1;
 //	} else
 //		return 0;
-TEXT runtime·cas(SB), 7, $0-16
+TEXT runtime·cas(SB), NOSPLIT, $0-16
 	MOVQ	8(SP), BX
 	MOVL	16(SP), AX
 	MOVL	20(SP), CX
@@ -479,7 +480,7 @@ TEXT runtime·cas(SB), 7, $0-16
 //	} else {
 //		return 0;
 //	}
-TEXT runtime·cas64(SB), 7, $0-24
+TEXT runtime·cas64(SB), NOSPLIT, $0-24
 	MOVQ	8(SP), BX
 	MOVQ	16(SP), AX
 	MOVQ	24(SP), CX
@@ -499,7 +500,7 @@ cas64_fail:
 //		return 1;
 //	} else
 //		return 0;
-TEXT runtime·casp(SB), 7, $0-24
+TEXT runtime·casp(SB), NOSPLIT, $0-24
 	MOVQ	8(SP), BX
 	MOVQ	16(SP), AX
 	MOVQ	24(SP), CX
@@ -515,7 +516,7 @@ TEXT runtime·casp(SB), 7, $0-24
 // Atomically:
 //	*val += delta;
 //	return *val;
-TEXT runtime·xadd(SB), 7, $0-12
+TEXT runtime·xadd(SB), NOSPLIT, $0-12
 	MOVQ	8(SP), BX
 	MOVL	16(SP), AX
 	MOVL	AX, CX
@@ -524,7 +525,7 @@ TEXT runtime·xadd(SB), 7, $0-12
 	ADDL	CX, AX
 	RET
 
-TEXT runtime·xadd64(SB), 7, $0-16
+TEXT runtime·xadd64(SB), NOSPLIT, $0-16
 	MOVQ	8(SP), BX
 	MOVQ	16(SP), AX
 	MOVQ	AX, CX
@@ -533,19 +534,19 @@ TEXT runtime·xadd64(SB), 7, $0-16
 	ADDQ	CX, AX
 	RET
 
-TEXT runtime·xchg(SB), 7, $0-12
+TEXT runtime·xchg(SB), NOSPLIT, $0-12
 	MOVQ	8(SP), BX
 	MOVL	16(SP), AX
 	XCHGL	AX, 0(BX)
 	RET
 
-TEXT runtime·xchg64(SB), 7, $0-16
+TEXT runtime·xchg64(SB), NOSPLIT, $0-16
 	MOVQ	8(SP), BX
 	MOVQ	16(SP), AX
 	XCHGQ	AX, 0(BX)
 	RET
 
-TEXT runtime·procyield(SB),7,$0-0
+TEXT runtime·procyield(SB),NOSPLIT,$0-0
 	MOVL	8(SP), AX
 again:
 	PAUSE
@@ -553,19 +554,19 @@ again:
 	JNZ	again
 	RET
 
-TEXT runtime·atomicstorep(SB), 7, $0-16
+TEXT runtime·atomicstorep(SB), NOSPLIT, $0-16
 	MOVQ	8(SP), BX
 	MOVQ	16(SP), AX
 	XCHGQ	AX, 0(BX)
 	RET
 
-TEXT runtime·atomicstore(SB), 7, $0-12
+TEXT runtime·atomicstore(SB), NOSPLIT, $0-12
 	MOVQ	8(SP), BX
 	MOVL	16(SP), AX
 	XCHGL	AX, 0(BX)
 	RET
 
-TEXT runtime·atomicstore64(SB), 7, $0-16
+TEXT runtime·atomicstore64(SB), NOSPLIT, $0-16
 	MOVQ	8(SP), BX
 	MOVQ	16(SP), AX
 	XCHGQ	AX, 0(BX)
@@ -576,7 +577,7 @@ TEXT runtime·atomicstore64(SB), 7, $0-16
 // 1. pop the caller
 // 2. sub 5 bytes from the callers return
 // 3. jmp to the argument
-TEXT runtime·jmpdefer(SB), 7, $0
+TEXT runtime·jmpdefer(SB), NOSPLIT, $0
 	MOVQ	8(SP), DX	// fn
 	MOVQ	16(SP), BX	// caller sp
 	LEAQ	-8(BX), SP	// caller sp after CALL
@@ -585,7 +586,7 @@ TEXT runtime·jmpdefer(SB), 7, $0
 	JMP	BX	// but first run the deferred function
 
 // Save state of caller into g->sched. Smashes R8, R9.
-TEXT gosave<>(SB),7,$0
+TEXT gosave<>(SB),NOSPLIT,$0
 	get_tls(R8)
 	MOVQ	g(R8), R8
 	MOVQ	0(SP), R9
@@ -600,7 +601,7 @@ TEXT gosave<>(SB),7,$0
 // Call fn(arg) on the scheduler stack,
 // aligned appropriately for the gcc ABI.
 // See cgocall.c for more details.
-TEXT runtime·asmcgocall(SB),7,$0-16
+TEXT runtime·asmcgocall(SB),NOSPLIT,$0-16
 	MOVQ	fn+0(FP), AX
 	MOVQ	arg+8(FP), BX
 	MOVQ	SP, DX
@@ -639,7 +640,7 @@ TEXT runtime·asmcgocall(SB),7,$0-16
 // cgocallback(void (*fn)(void*), void *frame, uintptr framesize)
 // Turn the fn into a Go func (by taking its address) and call
 // cgocallback_gofunc.
-TEXT runtime·cgocallback(SB),7,$24-24
+TEXT runtime·cgocallback(SB),NOSPLIT,$24-24
 	LEAQ	fn+0(FP), AX
 	MOVQ	AX, 0(SP)
 	MOVQ	frame+8(FP), AX
@@ -652,7 +653,7 @@ TEXT runtime·cgocallback(SB),7,$24-24
 
 // cgocallback_gofunc(FuncVal*, void *frame, uintptr framesize)
 // See cgocall.c for more details.
-TEXT runtime·cgocallback_gofunc(SB),7,$8-24
+TEXT runtime·cgocallback_gofunc(SB),NOSPLIT,$8-24
 	// If m is nil, Go did not create the current thread.
 	// Call needm to obtain one for temporary use.
 	// In this case, we're running on the thread stack, so there's
@@ -741,7 +742,7 @@ havem:
 	RET
 
 // void setmg(M*, G*); set m and g. for use by needm.
-TEXT runtime·setmg(SB), 7, $0-16
+TEXT runtime·setmg(SB), NOSPLIT, $0-16
 	MOVQ	mm+0(FP), AX
 #ifdef GOOS_windows
 	CMPQ	AX, $0
@@ -760,14 +761,14 @@ settls:
 	RET
 
 // void setmg_gcc(M*, G*); set m and g called from gcc.
-TEXT setmg_gcc<>(SB),7,$0
+TEXT setmg_gcc<>(SB),NOSPLIT,$0
 	get_tls(AX)
 	MOVQ	DI, m(AX)
 	MOVQ	SI, g(AX)
 	RET
 
 // check that SP is in range [g->stackbase, g->stackguard)
-TEXT runtime·stackcheck(SB), 7, $0-0
+TEXT runtime·stackcheck(SB), NOSPLIT, $0-0
 	get_tls(CX)
 	MOVQ	g(CX), AX
 	CMPQ	g_stackbase(AX), SP
@@ -778,7 +779,7 @@ TEXT runtime·stackcheck(SB), 7, $0-0
 	INT	$3
 	RET
 
-TEXT runtime·memclr(SB),7,$0-16
+TEXT runtime·memclr(SB),NOSPLIT,$0-16
 	MOVQ	8(SP), DI		// arg 1 addr
 	MOVQ	16(SP), CX		// arg 2 count
 	MOVQ	CX, BX
@@ -793,29 +794,29 @@ TEXT runtime·memclr(SB),7,$0-16
 	STOSB
 	RET
 
-TEXT runtime·getcallerpc(SB),7,$0-8
+TEXT runtime·getcallerpc(SB),NOSPLIT,$0-8
 	MOVQ	x+0(FP),AX		// addr of first arg
 	MOVQ	-8(AX),AX		// get calling pc
 	RET
 
-TEXT runtime·setcallerpc(SB),7,$0-16
+TEXT runtime·setcallerpc(SB),NOSPLIT,$0-16
 	MOVQ	x+0(FP),AX		// addr of first arg
 	MOVQ	x+8(FP), BX
 	MOVQ	BX, -8(AX)		// set calling pc
 	RET
 
-TEXT runtime·getcallersp(SB),7,$0-8
+TEXT runtime·getcallersp(SB),NOSPLIT,$0-8
 	MOVQ	sp+0(FP), AX
 	RET
 
 // int64 runtime·cputicks(void)
-TEXT runtime·cputicks(SB),7,$0-0
+TEXT runtime·cputicks(SB),NOSPLIT,$0-0
 	RDTSC
 	SHLQ	$32, DX
 	ADDQ	DX, AX
 	RET
 
-TEXT runtime·stackguard(SB),7,$0-16
+TEXT runtime·stackguard(SB),NOSPLIT,$0-16
 	MOVQ	SP, DX
 	MOVQ	DX, sp+0(FP)
 	get_tls(CX)
@@ -827,13 +828,13 @@ TEXT runtime·stackguard(SB),7,$0-16
 GLOBL runtime·tls0(SB), $64
 
 // hash function using AES hardware instructions
-TEXT runtime·aeshash(SB),7,$0-24
+TEXT runtime·aeshash(SB),NOSPLIT,$0-24
 	MOVQ	8(SP), DX	// ptr to hash value
 	MOVQ	16(SP), CX	// size
 	MOVQ	24(SP), AX	// ptr to data
 	JMP	runtime·aeshashbody(SB)
 
-TEXT runtime·aeshashstr(SB),7,$0-24
+TEXT runtime·aeshashstr(SB),NOSPLIT,$0-24
 	MOVQ	8(SP), DX	// ptr to hash value
 	MOVQ	24(SP), AX	// ptr to string struct
 	MOVQ	8(AX), CX	// length of string
@@ -843,7 +844,7 @@ TEXT runtime·aeshashstr(SB),7,$0-24
 // AX: data
 // CX: length
 // DX: ptr to seed input / hash output
-TEXT runtime·aeshashbody(SB),7,$0-24
+TEXT runtime·aeshashbody(SB),NOSPLIT,$0-24
 	MOVQ	(DX), X0	// seed to low 64 bits of xmm0
 	PINSRQ	$1, CX, X0	// size to high 64 bits of xmm0
 	MOVO	runtime·aeskeysched+0(SB), X2
@@ -898,7 +899,7 @@ finalize:
 	MOVQ	X0, (DX)
 	RET
 
-TEXT runtime·aeshash32(SB),7,$0-24
+TEXT runtime·aeshash32(SB),NOSPLIT,$0-24
 	MOVQ	8(SP), DX	// ptr to hash value
 	MOVQ	24(SP), AX	// ptr to data
 	MOVQ	(DX), X0	// seed
@@ -909,7 +910,7 @@ TEXT runtime·aeshash32(SB),7,$0-24
 	MOVQ	X0, (DX)
 	RET
 
-TEXT runtime·aeshash64(SB),7,$0-24
+TEXT runtime·aeshash64(SB),NOSPLIT,$0-24
 	MOVQ	8(SP), DX	// ptr to hash value
 	MOVQ	24(SP), AX	// ptr to data
 	MOVQ	(DX), X0	// seed
@@ -953,7 +954,7 @@ DATA masks<>+0xe0(SB)/8, $0xffffffffffffffff
 DATA masks<>+0xe8(SB)/8, $0x0000ffffffffffff
 DATA masks<>+0xf0(SB)/8, $0xffffffffffffffff
 DATA masks<>+0xf8(SB)/8, $0x00ffffffffffffff
-GLOBL masks<>(SB),8,$256
+GLOBL masks<>(SB),RODATA,$256
 
 // these are arguments to pshufb.  They move data down from
 // the high bytes of the register to the low bytes of the register.
@@ -990,9 +991,9 @@ DATA shifts<>+0xe0(SB)/8, $0x0908070605040302
 DATA shifts<>+0xe8(SB)/8, $0xffff0f0e0d0c0b0a
 DATA shifts<>+0xf0(SB)/8, $0x0807060504030201
 DATA shifts<>+0xf8(SB)/8, $0xff0f0e0d0c0b0a09
-GLOBL shifts<>(SB),8,$256
+GLOBL shifts<>(SB),RODATA,$256
 
-TEXT runtime·memeq(SB),7,$0-24
+TEXT runtime·memeq(SB),NOSPLIT,$0-24
 	MOVQ	a+0(FP), SI
 	MOVQ	b+8(FP), DI
 	MOVQ	count+16(FP), BX
@@ -1001,7 +1002,7 @@ TEXT runtime·memeq(SB),7,$0-24
 // a in SI
 // b in DI
 // count in BX
-TEXT runtime·memeqbody(SB),7,$0-0
+TEXT runtime·memeqbody(SB),NOSPLIT,$0-0
 	XORQ	AX, AX
 
 	CMPQ	BX, $8
@@ -1090,7 +1091,7 @@ equal:
 	SETEQ	AX
 	RET
 
-TEXT runtime·cmpstring(SB),7,$0-40
+TEXT runtime·cmpstring(SB),NOSPLIT,$0-40
 	MOVQ	s1+0(FP), SI
 	MOVQ	s1+8(FP), BX
 	MOVQ	s2+16(FP), DI
@@ -1099,7 +1100,7 @@ TEXT runtime·cmpstring(SB),7,$0-40
 	MOVQ	AX, res+32(FP)
 	RET
 
-TEXT bytes·Compare(SB),7,$0-56
+TEXT bytes·Compare(SB),NOSPLIT,$0-56
 	MOVQ	s1+0(FP), SI
 	MOVQ	s1+8(FP), BX
 	MOVQ	s2+24(FP), DI
@@ -1115,7 +1116,7 @@ TEXT bytes·Compare(SB),7,$0-56
 //   DX = blen
 // output:
 //   AX = 1/0/-1
-TEXT runtime·cmpbody(SB),7,$0-0
+TEXT runtime·cmpbody(SB),NOSPLIT,$0-0
 	CMPQ	SI, DI
 	JEQ	cmp_allsame
 	CMPQ	BX, DX
@@ -1220,7 +1221,7 @@ cmp_allsame:
 	LEAQ	-1(CX)(AX*2), AX	// 1,0,-1 result
 	RET
 
-TEXT bytes·IndexByte(SB),7,$0
+TEXT bytes·IndexByte(SB),NOSPLIT,$0
 	MOVQ s+0(FP), SI
 	MOVQ s_len+8(FP), BX
 	MOVB c+24(FP), AL
@@ -1228,7 +1229,7 @@ TEXT bytes·IndexByte(SB),7,$0
 	MOVQ AX, ret+32(FP)
 	RET
 
-TEXT strings·IndexByte(SB),7,$0
+TEXT strings·IndexByte(SB),NOSPLIT,$0
 	MOVQ s+0(FP), SI
 	MOVQ s_len+8(FP), BX
 	MOVB c+16(FP), AL
@@ -1242,7 +1243,7 @@ TEXT strings·IndexByte(SB),7,$0
 //   AL: byte sought
 // output:
 //   AX
-TEXT runtime·indexbytebody(SB),7,$0
+TEXT runtime·indexbytebody(SB),NOSPLIT,$0
 	MOVQ SI, DI
 
 	CMPQ BX, $16
@@ -1327,7 +1328,7 @@ success:
 	MOVQ DI, AX
 	RET
 
-TEXT bytes·Equal(SB),7,$0-49
+TEXT bytes·Equal(SB),NOSPLIT,$0-49
 	MOVQ	a_len+8(FP), BX
 	MOVQ	b_len+32(FP), CX
 	XORQ	AX, AX
diff --git a/src/pkg/runtime/asm_arm.s b/src/pkg/runtime/asm_arm.s
index b029674a47..d02ba6b031 100644
--- a/src/pkg/runtime/asm_arm.s
+++ b/src/pkg/runtime/asm_arm.s
@@ -4,9 +4,10 @@
 
 #include "zasm_GOOS_GOARCH.h"
 #include "funcdata.h"
+#include "../../cmd/ld/textflag.h"
 
 // using frame size $-4 means do not save LR on stack.
-TEXT _rt0_go(SB),7,$-4
+TEXT _rt0_go(SB),NOSPLIT,$-4
 	MOVW	$0xcafebabe, R12
 
 	// copy arguments forward on an even stack
@@ -75,9 +76,9 @@ TEXT _rt0_go(SB),7,$-4
 	MOVW	R0, (R1)	// fail hard
 
 DATA	runtime·main·f+0(SB)/4,$runtime·main(SB)
-GLOBL	runtime·main·f(SB),8,$4
+GLOBL	runtime·main·f(SB),RODATA,$4
 
-TEXT runtime·breakpoint(SB),7,$0-0
+TEXT runtime·breakpoint(SB),NOSPLIT,$0-0
 	// gdb won't skip this breakpoint instruction automatically,
 	// so you must manually "set $pc+=4" to skip it and continue.
 	WORD    $0xe1200071 // BKPT 0x0001
@@ -85,7 +86,7 @@ TEXT runtime·breakpoint(SB),7,$0-0
 
 GLOBL runtime·goarm(SB), $4
 
-TEXT runtime·asminit(SB),7,$0-0
+TEXT runtime·asminit(SB),NOSPLIT,$0-0
 	// disable runfast (flush-to-zero) mode of vfp if runtime.goarm > 5
 	MOVW runtime·goarm(SB), R11
 	CMP $5, R11
@@ -101,7 +102,7 @@ TEXT runtime·asminit(SB),7,$0-0
 
 // void gosave(Gobuf*)
 // save state in Gobuf; setjmp
-TEXT runtime·gosave(SB), 7, $-4-4
+TEXT runtime·gosave(SB), NOSPLIT, $-4-4
 	MOVW	0(FP), R0		// gobuf
 	MOVW	SP, gobuf_sp(R0)
 	MOVW	LR, gobuf_pc(R0)
@@ -114,7 +115,7 @@ TEXT runtime·gosave(SB), 7, $-4-4
 
 // void gogo(Gobuf*)
 // restore state from Gobuf; longjmp
-TEXT runtime·gogo(SB), 7, $-4-4
+TEXT runtime·gogo(SB), NOSPLIT, $-4-4
 	MOVW	0(FP), R1		// gobuf
 	MOVW	gobuf_g(R1), g
 	MOVW	0(g), R2		// make sure g != nil
@@ -137,7 +138,7 @@ TEXT runtime·gogo(SB), 7, $-4-4
 // Switch to m->g0's stack, call fn(g).
 // Fn must never return.  It should gogo(&g->sched)
 // to keep running g.
-TEXT runtime·mcall(SB), 7, $-4-4
+TEXT runtime·mcall(SB), NOSPLIT, $-4-4
 	MOVW	fn+0(FP), R0
 
 	// Save caller state in g->sched.
@@ -175,7 +176,7 @@ TEXT runtime·mcall(SB), 7, $-4-4
 // the top of a stack (for example, morestack calling newstack
 // calling the scheduler calling newm calling gc), so we must
 // record an argument size. For that purpose, it has no arguments.
-TEXT runtime·morestack(SB),7,$-4-0
+TEXT runtime·morestack(SB),NOSPLIT,$-4-0
 	// Cannot grow scheduler stack (m->g0).
 	MOVW	m_g0(m), R4
 	CMP	g, R4
@@ -213,7 +214,7 @@ TEXT runtime·morestack(SB),7,$-4-0
 // with the desired args running the desired function.
 //
 // func call(fn *byte, arg *byte, argsize uint32).
-TEXT runtime·newstackcall(SB), 7, $-4-12
+TEXT runtime·newstackcall(SB), NOSPLIT, $-4-12
 	// Save our caller's state as the PC and SP to
 	// restore when returning from f.
 	MOVW	LR, (m_morebuf+gobuf_pc)(m)	// our caller's PC
@@ -260,7 +261,7 @@ TEXT runtime·newstackcall(SB), 7, $-4-12
 	MOVW	$runtime·NAME(SB), R1;	\
 	B	(R1)
 
-TEXT reflect·call(SB), 7, $-4-12
+TEXT reflect·call(SB), NOSPLIT, $-4-12
 	MOVW	argsize+8(FP), R0
 	DISPATCH(call16, 16)
 	DISPATCH(call32, 32)
@@ -352,7 +353,7 @@ CALLFN(call1073741824, 1073741824)
 //
 // Lessstack can appear in stack traces for the same reason
 // as morestack; in that context, it has 0 arguments.
-TEXT runtime·lessstack(SB), 7, $-4-0
+TEXT runtime·lessstack(SB), NOSPLIT, $-4-0
 	// Save return value in m->cret
 	MOVW	R0, m_cret(m)
 
@@ -366,7 +367,7 @@ TEXT runtime·lessstack(SB), 7, $-4-0
 // 1. grab stored LR for caller
 // 2. sub 4 bytes to get back to BL deferreturn
 // 3. B to fn
-TEXT runtime·jmpdefer(SB), 7, $0
+TEXT runtime·jmpdefer(SB), NOSPLIT, $0
 	MOVW	0(SP), LR
 	MOVW	$-4(LR), LR	// BL deferreturn
 	MOVW	fn+0(FP), R7
@@ -376,7 +377,7 @@ TEXT runtime·jmpdefer(SB), 7, $0
 	B	(R1)
 
 // Save state of caller into g->sched. Smashes R11.
-TEXT gosave<>(SB),7,$0
+TEXT gosave<>(SB),NOSPLIT,$0
 	MOVW	LR, (g_sched+gobuf_pc)(g)
 	MOVW	R13, (g_sched+gobuf_sp)(g)
 	MOVW	$0, R11
@@ -389,7 +390,7 @@ TEXT gosave<>(SB),7,$0
 // Call fn(arg) on the scheduler stack,
 // aligned appropriately for the gcc ABI.
 // See cgocall.c for more details.
-TEXT	runtime·asmcgocall(SB),7,$0-8
+TEXT	runtime·asmcgocall(SB),NOSPLIT,$0-8
 	MOVW	fn+0(FP), R1
 	MOVW	arg+4(FP), R0
 	MOVW	R13, R2
@@ -421,7 +422,7 @@ TEXT	runtime·asmcgocall(SB),7,$0-8
 // cgocallback(void (*fn)(void*), void *frame, uintptr framesize)
 // Turn the fn into a Go func (by taking its address) and call
 // cgocallback_gofunc.
-TEXT runtime·cgocallback(SB),7,$12-12
+TEXT runtime·cgocallback(SB),NOSPLIT,$12-12
 	MOVW	$fn+0(FP), R0
 	MOVW	R0, 4(R13)
 	MOVW	frame+4(FP), R0
@@ -434,7 +435,7 @@ TEXT runtime·cgocallback(SB),7,$12-12
 
 // cgocallback_gofunc(void (*fn)(void*), void *frame, uintptr framesize)
 // See cgocall.c for more details.
-TEXT	runtime·cgocallback_gofunc(SB),7,$8-12
+TEXT	runtime·cgocallback_gofunc(SB),NOSPLIT,$8-12
 	// Load m and g from thread-local storage.
 	MOVW	_cgo_load_gm(SB), R0
 	CMP	$0, R0
@@ -513,7 +514,7 @@ havem:
 	RET
 
 // void setmg(M*, G*); set m and g. for use by needm.
-TEXT runtime·setmg(SB), 7, $0-8
+TEXT runtime·setmg(SB), NOSPLIT, $0-8
 	MOVW	mm+0(FP), m
 	MOVW	gg+4(FP), g
 
@@ -524,16 +525,16 @@ TEXT runtime·setmg(SB), 7, $0-8
 
 	RET
 
-TEXT runtime·getcallerpc(SB),7,$-4-4
+TEXT runtime·getcallerpc(SB),NOSPLIT,$-4-4
 	MOVW	0(SP), R0
 	RET
 
-TEXT runtime·setcallerpc(SB),7,$-4-8
+TEXT runtime·setcallerpc(SB),NOSPLIT,$-4-8
 	MOVW	x+4(FP), R0
 	MOVW	R0, 0(SP)
 	RET
 
-TEXT runtime·getcallersp(SB),7,$-4-4
+TEXT runtime·getcallersp(SB),NOSPLIT,$-4-4
 	MOVW	0(FP), R0
 	MOVW	$-4(R0), R0
 	RET
@@ -541,7 +542,7 @@ TEXT runtime·getcallersp(SB),7,$-4-4
 TEXT runtime·emptyfunc(SB),0,$0-0
 	RET
 
-TEXT runtime·abort(SB),7,$-4-0
+TEXT runtime·abort(SB),NOSPLIT,$-4-0
 	MOVW	$0, R0
 	MOVW	(R0), R1
 
@@ -556,10 +557,10 @@ TEXT runtime·abort(SB),7,$-4-0
 // To implement runtime·cas in sys_$GOOS_arm.s
 // using the native instructions, use:
 //
-//	TEXT runtime·cas(SB),7,$0
+//	TEXT runtime·cas(SB),NOSPLIT,$0
 //		B	runtime·armcas(SB)
 //
-TEXT runtime·armcas(SB),7,$0-12
+TEXT runtime·armcas(SB),NOSPLIT,$0-12
 	MOVW	valptr+0(FP), R1
 	MOVW	old+4(FP), R2
 	MOVW	new+8(FP), R3
@@ -576,7 +577,7 @@ casfail:
 	MOVW	$0, R0
 	RET
 
-TEXT runtime·stackguard(SB),7,$0-8
+TEXT runtime·stackguard(SB),NOSPLIT,$0-8
 	MOVW	R13, R1
 	MOVW	g_stackguard(g), R2
 	MOVW	R1, sp+0(FP)
@@ -584,20 +585,20 @@ TEXT runtime·stackguard(SB),7,$0-8
 	RET
 
 // AES hashing not implemented for ARM
-TEXT runtime·aeshash(SB),7,$-4-0
+TEXT runtime·aeshash(SB),NOSPLIT,$-4-0
 	MOVW	$0, R0
 	MOVW	(R0), R1
-TEXT runtime·aeshash32(SB),7,$-4-0
+TEXT runtime·aeshash32(SB),NOSPLIT,$-4-0
 	MOVW	$0, R0
 	MOVW	(R0), R1
-TEXT runtime·aeshash64(SB),7,$-4-0
+TEXT runtime·aeshash64(SB),NOSPLIT,$-4-0
 	MOVW	$0, R0
 	MOVW	(R0), R1
-TEXT runtime·aeshashstr(SB),7,$-4-0
+TEXT runtime·aeshashstr(SB),NOSPLIT,$-4-0
 	MOVW	$0, R0
 	MOVW	(R0), R1
 
-TEXT runtime·memeq(SB),7,$-4-12
+TEXT runtime·memeq(SB),NOSPLIT,$-4-12
 	MOVW	a+0(FP), R1
 	MOVW	b+4(FP), R2
 	MOVW	n+8(FP), R3
@@ -615,7 +616,7 @@ _next:
 	RET
 
 // TODO: share code with memeq?
-TEXT bytes·Equal(SB),7,$0
+TEXT bytes·Equal(SB),NOSPLIT,$0
 	MOVW	a_len+4(FP), R1
 	MOVW	b_len+16(FP), R3
 	
@@ -644,7 +645,7 @@ _equal:
 	MOVBU	R0, ret+24(FP)
 	RET
 
-TEXT bytes·IndexByte(SB),7,$0
+TEXT bytes·IndexByte(SB),NOSPLIT,$0
 	MOVW	s+0(FP), R0
 	MOVW	s_len+4(FP), R1
 	MOVBU	c+12(FP), R2	// byte to find
@@ -668,7 +669,7 @@ _notfound:
 	MOVW	R0, ret+16(FP)
 	RET
 
-TEXT strings·IndexByte(SB),7,$0
+TEXT strings·IndexByte(SB),NOSPLIT,$0
 	MOVW	s+0(FP), R0
 	MOVW	s_len+4(FP), R1
 	MOVBU	c+8(FP), R2	// byte to find