runtime: unify asmcgocall and systemstack traceback setup

Both asmcgocall and systemstack need to save the calling Go code's
context for use by traceback, but they do it differently.
Systemstack's appraoch is better, because it doesn't require a
special case in traceback.
So make them both use that.

While we are here, the fake mstart caller in systemstack is
no longer needed and can be removed.
(traceback knows to stop in systemstack because of the writes to SP.)

Also remove the fake mstarts in sys_windows_*.s.

And while we are there, fix the control flow guard code in sys_windows_arm.s.
The current code is using pointers to a stack frame that technically is gone
once we hit the RET instruction. Clearly it's working OK, but better not to depend
on data below SP being preserved, even for just a few instructions.
Store the value we need in other registers instead.
(This code is only used for pushing a sigpanic call, which does not
actually return to the site of the fault and therefore doesn't need to
preserve any of the registers.)

This CL is part of a stack adding windows/arm64
support (#36439), intended to land in the Go 1.17 cycle.
This CL is, however, not windows/arm64-specific.
It is cleanup meant to make the port (and future ports) easier.

Change-Id: Id1e3ef5e54f7ad786e4b87043f2626eba7c3bbd9
Reviewed-on: https://go-review.googlesource.com/c/go/+/288799
Trust: Russ Cox <rsc@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>

13 files changed, 137 insertions, 180 deletions

diff --git a/misc/cgo/test/callback.go b/misc/cgo/test/callback.go
index 814888e3ac..08dd9b39d8 100644
--- a/misc/cgo/test/callback.go
+++ b/misc/cgo/test/callback.go
@@ -182,7 +182,7 @@ func testCallbackCallers(t *testing.T) {
 		"runtime.cgocallbackg1",
 		"runtime.cgocallbackg",
 		"runtime.cgocallback",
-		"runtime.asmcgocall",
+		"runtime.systemstack_switch",
 		"runtime.cgocall",
 		"test._Cfunc_callback",
 		"test.nestedCall.func1",
diff --git a/src/runtime/asm_386.s b/src/runtime/asm_386.s
index 3030101f03..a59054226c 100644
--- a/src/runtime/asm_386.s
+++ b/src/runtime/asm_386.s
@@ -352,18 +352,12 @@ TEXT runtime·systemstack(SB), NOSPLIT, $0-4
 	// switch stacks
 	// save our state in g->sched. Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVL	$runtime·systemstack_switch(SB), (g_sched+gobuf_pc)(AX)
-	MOVL	SP, (g_sched+gobuf_sp)(AX)
-	MOVL	AX, (g_sched+gobuf_g)(AX)
+	CALL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	get_tls(CX)
 	MOVL	DX, g(CX)
 	MOVL	(g_sched+gobuf_sp)(DX), BX
-	// make it look like mstart called systemstack on g0, to stop traceback
-	SUBL	$4, BX
-	MOVL	$runtime·mstart(SB), DX
-	MOVL	DX, 0(BX)
 	MOVL	BX, SP
 
 	// call target function
@@ -601,15 +595,18 @@ TEXT runtime·jmpdefer(SB), NOSPLIT, $0-8
 	MOVL	0(DX), BX
 	JMP	BX	// but first run the deferred function
 
-// Save state of caller into g->sched.
-TEXT gosave<>(SB),NOSPLIT,$0
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT,$0
 	PUSHL	AX
 	PUSHL	BX
 	get_tls(BX)
 	MOVL	g(BX), BX
 	LEAL	arg+0(FP), AX
 	MOVL	AX, (g_sched+gobuf_sp)(BX)
-	MOVL	-4(AX), AX
+	MOVL	$runtime·systemstack_switch(SB), AX
 	MOVL	AX, (g_sched+gobuf_pc)(BX)
 	MOVL	$0, (g_sched+gobuf_ret)(BX)
 	// Assert ctxt is zero. See func save.
@@ -661,7 +658,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-12
 	JEQ	noswitch
 	CMPL	DI, m_gsignal(BP)
 	JEQ	noswitch
-	CALL	gosave<>(SB)
+	CALL	gosave_systemstack_switch<>(SB)
 	get_tls(CX)
 	MOVL	SI, g(CX)
 	MOVL	(g_sched+gobuf_sp)(SI), SP
diff --git a/src/runtime/asm_amd64.s b/src/runtime/asm_amd64.s
index 9362ce1213..cd5ce3effb 100644
--- a/src/runtime/asm_amd64.s
+++ b/src/runtime/asm_amd64.s
@@ -338,20 +338,12 @@ TEXT runtime·systemstack(SB), NOSPLIT, $0-8
 	// switch stacks
 	// save our state in g->sched. Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVQ	$runtime·systemstack_switch(SB), SI
-	MOVQ	SI, (g_sched+gobuf_pc)(AX)
-	MOVQ	SP, (g_sched+gobuf_sp)(AX)
-	MOVQ	AX, (g_sched+gobuf_g)(AX)
-	MOVQ	BP, (g_sched+gobuf_bp)(AX)
+	CALL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOVQ	DX, g(CX)
 	MOVQ	DX, R14 // set the g register
 	MOVQ	(g_sched+gobuf_sp)(DX), BX
-	// make it look like mstart called systemstack on g0, to stop traceback
-	SUBQ	$8, BX
-	MOVQ	$runtime·mstart(SB), DX
-	MOVQ	DX, 0(BX)
 	MOVQ	BX, SP
 
 	// call target function
@@ -660,13 +652,17 @@ TEXT runtime·jmpdefer(SB), NOSPLIT, $0-16
 	MOVQ	0(DX), BX
 	JMP	BX	// but first run the deferred function
 
-// Save state of caller into g->sched. Smashes R9.
-TEXT gosave<>(SB),NOSPLIT,$0
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes R9.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT,$0
 #ifndef GOEXPERIMENT_REGABI
 	get_tls(R14)
 	MOVQ	g(R14), R14
 #endif
-	MOVQ	0(SP), R9
+	MOVQ	$runtime·systemstack_switch(SB), R9
 	MOVQ	R9, (g_sched+gobuf_pc)(R14)
 	LEAQ	8(SP), R9
 	MOVQ	R9, (g_sched+gobuf_sp)(R14)
@@ -724,7 +720,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-20
 
 	// Switch to system stack.
 	MOVQ	m_g0(R8), SI
-	CALL	gosave<>(SB)
+	CALL	gosave_systemstack_switch<>(SB)
 	MOVQ	SI, g(CX)
 	MOVQ	(g_sched+gobuf_sp)(SI), SP
 
diff --git a/src/runtime/asm_arm.s b/src/runtime/asm_arm.s
index 109030aada..c8c53e70db 100644
--- a/src/runtime/asm_arm.s
+++ b/src/runtime/asm_arm.s
@@ -305,24 +305,14 @@ TEXT runtime·systemstack(SB),NOSPLIT,$0-4
 switch:
 	// save our state in g->sched. Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVW	$runtime·systemstack_switch(SB), R3
-	ADD	$4, R3, R3 // get past push {lr}
-	MOVW	R3, (g_sched+gobuf_pc)(g)
-	MOVW	R13, (g_sched+gobuf_sp)(g)
-	MOVW	LR, (g_sched+gobuf_lr)(g)
-	MOVW	g, (g_sched+gobuf_g)(g)
+	BL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOVW	R0, R5
 	MOVW	R2, R0
 	BL	setg<>(SB)
 	MOVW	R5, R0
-	MOVW	(g_sched+gobuf_sp)(R2), R3
-	// make it look like mstart called systemstack on g0, to stop traceback
-	SUB	$4, R3, R3
-	MOVW	$runtime·mstart(SB), R4
-	MOVW	R4, 0(R3)
-	MOVW	R3, R13
+	MOVW	(g_sched+gobuf_sp)(R2), R13
 
 	// call target function
 	MOVW	R0, R7
@@ -537,19 +527,25 @@ TEXT runtime·jmpdefer(SB),NOSPLIT,$0-8
 	MOVW	0(R7), R1
 	B	(R1)
 
-// Save state of caller into g->sched. Smashes R11.
-TEXT gosave<>(SB),NOSPLIT|NOFRAME,$0
-	MOVW	LR, (g_sched+gobuf_pc)(g)
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes R11.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT|NOFRAME,$0
+	MOVW	$runtime·systemstack_switch(SB), R11
+	ADD	$4, R11 // get past push {lr}
+	MOVW	R11, (g_sched+gobuf_pc)(g)
 	MOVW	R13, (g_sched+gobuf_sp)(g)
+	MOVW	g, (g_sched+gobuf_g)(g)
 	MOVW	$0, R11
 	MOVW	R11, (g_sched+gobuf_lr)(g)
 	MOVW	R11, (g_sched+gobuf_ret)(g)
-	MOVW	R11, (g_sched+gobuf_ctxt)(g)
 	// Assert ctxt is zero. See func save.
 	MOVW	(g_sched+gobuf_ctxt)(g), R11
-	CMP	$0, R11
+	TST	R11, R11
 	B.EQ	2(PC)
-	CALL	runtime·badctxt(SB)
+	BL	runtime·badctxt(SB)
 	RET
 
 // func asmcgocall_no_g(fn, arg unsafe.Pointer)
@@ -590,7 +586,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-12
 	MOVW	m_g0(R8), R3
 	CMP	R3, g
 	BEQ	nosave
-	BL	gosave<>(SB)
+	BL	gosave_systemstack_switch<>(SB)
 	MOVW	R0, R5
 	MOVW	R3, R0
 	BL	setg<>(SB)
diff --git a/src/runtime/asm_arm64.s b/src/runtime/asm_arm64.s
index 79efd4cb17..31a6fe57b9 100644
--- a/src/runtime/asm_arm64.s
+++ b/src/runtime/asm_arm64.s
@@ -205,24 +205,12 @@ TEXT runtime·systemstack(SB), NOSPLIT, $0-8
 switch:
 	// save our state in g->sched. Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVD	$runtime·systemstack_switch(SB), R6
-	ADD	$8, R6	// get past prologue
-	MOVD	R6, (g_sched+gobuf_pc)(g)
-	MOVD	RSP, R0
-	MOVD	R0, (g_sched+gobuf_sp)(g)
-	MOVD	R29, (g_sched+gobuf_bp)(g)
-	MOVD	$0, (g_sched+gobuf_lr)(g)
-	MOVD	g, (g_sched+gobuf_g)(g)
+	BL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOVD	R5, g
 	BL	runtime·save_g(SB)
 	MOVD	(g_sched+gobuf_sp)(g), R3
-	// make it look like mstart called systemstack on g0, to stop traceback
-	SUB	$16, R3
-	AND	$~15, R3
-	MOVD	$runtime·mstart(SB), R4
-	MOVD	R4, 0(R3)
 	MOVD	R3, RSP
 	MOVD	(g_sched+gobuf_bp)(g), R29
 
@@ -859,14 +847,21 @@ TEXT runtime·jmpdefer(SB), NOSPLIT|NOFRAME, $0-16
 	MOVD	0(R26), R3
 	B	(R3)
 
-// Save state of caller into g->sched. Smashes R0.
-TEXT gosave<>(SB),NOSPLIT|NOFRAME,$0
-	MOVD	LR, (g_sched+gobuf_pc)(g)
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes R0.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT|NOFRAME,$0
+	MOVD	$runtime·systemstack_switch(SB), R0
+	ADD	$8, R0	// get past prologue
+	MOVD	R0, (g_sched+gobuf_pc)(g)
 	MOVD	RSP, R0
 	MOVD	R0, (g_sched+gobuf_sp)(g)
 	MOVD	R29, (g_sched+gobuf_bp)(g)
 	MOVD	$0, (g_sched+gobuf_lr)(g)
 	MOVD	$0, (g_sched+gobuf_ret)(g)
+	MOVD	g, (g_sched+gobuf_g)(g)
 	// Assert ctxt is zero. See func save.
 	MOVD	(g_sched+gobuf_ctxt)(g), R0
 	CBZ	R0, 2(PC)
@@ -908,8 +903,8 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-20
 	BEQ	nosave
 
 	// Switch to system stack.
-	MOVD	R0, R9	// gosave<> and save_g might clobber R0
-	BL	gosave<>(SB)
+	MOVD	R0, R9	// gosave_systemstack_switch<> and save_g might clobber R0
+	BL	gosave_systemstack_switch<>(SB)
 	MOVD	R3, g
 	BL	runtime·save_g(SB)
 	MOVD	(g_sched+gobuf_sp)(g), R0
diff --git a/src/runtime/asm_mips64x.s b/src/runtime/asm_mips64x.s
index 6e1d25cd90..75bb223066 100644
--- a/src/runtime/asm_mips64x.s
+++ b/src/runtime/asm_mips64x.s
@@ -169,21 +169,12 @@ TEXT runtime·systemstack(SB), NOSPLIT, $0-8
 switch:
 	// save our state in g->sched. Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVV	$runtime·systemstack_switch(SB), R4
-	ADDV	$8, R4	// get past prologue
-	MOVV	R4, (g_sched+gobuf_pc)(g)
-	MOVV	R29, (g_sched+gobuf_sp)(g)
-	MOVV	R0, (g_sched+gobuf_lr)(g)
-	MOVV	g, (g_sched+gobuf_g)(g)
+	JAL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOVV	R3, g
 	JAL	runtime·save_g(SB)
 	MOVV	(g_sched+gobuf_sp)(g), R1
-	// make it look like mstart called systemstack on g0, to stop traceback
-	ADDV	$-8, R1
-	MOVV	$runtime·mstart(SB), R2
-	MOVV	R2, 0(R1)
 	MOVV	R1, R29
 
 	// call target function
@@ -401,12 +392,19 @@ TEXT runtime·jmpdefer(SB), NOSPLIT|NOFRAME, $0-16
 	MOVV	0(REGCTXT), R4
 	JMP	(R4)
 
-// Save state of caller into g->sched. Smashes R1.
-TEXT gosave<>(SB),NOSPLIT|NOFRAME,$0
-	MOVV	R31, (g_sched+gobuf_pc)(g)
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes R1.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT|NOFRAME,$0
+	MOVV	$runtime·systemstack_switch(SB), R1
+	ADDV	$8, R1	// get past prologue
+	MOVV	R1, (g_sched+gobuf_pc)(g)
 	MOVV	R29, (g_sched+gobuf_sp)(g)
 	MOVV	R0, (g_sched+gobuf_lr)(g)
 	MOVV	R0, (g_sched+gobuf_ret)(g)
+	MOVV	g, (g_sched+gobuf_g)(g)
 	// Assert ctxt is zero. See func save.
 	MOVV	(g_sched+gobuf_ctxt)(g), R1
 	BEQ	R1, 2(PC)
@@ -440,7 +438,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-20
 	MOVV	m_g0(R5), R6
 	BEQ	R6, g, g0
 
-	JAL	gosave<>(SB)
+	JAL	gosave_systemstack_switch<>(SB)
 	MOVV	R6, g
 	JAL	runtime·save_g(SB)
 	MOVV	(g_sched+gobuf_sp)(g), R29
diff --git a/src/runtime/asm_mipsx.s b/src/runtime/asm_mipsx.s
index 8e5753d255..341da8e8d7 100644
--- a/src/runtime/asm_mipsx.s
+++ b/src/runtime/asm_mipsx.s
@@ -170,21 +170,12 @@ TEXT runtime·systemstack(SB),NOSPLIT,$0-4
 switch:
 	// save our state in g->sched.  Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVW	$runtime·systemstack_switch(SB), R4
-	ADDU	$8, R4	// get past prologue
-	MOVW	R4, (g_sched+gobuf_pc)(g)
-	MOVW	R29, (g_sched+gobuf_sp)(g)
-	MOVW	R0, (g_sched+gobuf_lr)(g)
-	MOVW	g, (g_sched+gobuf_g)(g)
+	JAL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOVW	R3, g
 	JAL	runtime·save_g(SB)
 	MOVW	(g_sched+gobuf_sp)(g), R1
-	// make it look like mstart called systemstack on g0, to stop traceback
-	ADDU	$-4, R1
-	MOVW	$runtime·mstart(SB), R2
-	MOVW	R2, 0(R1)
 	MOVW	R1, R29
 
 	// call target function
@@ -401,12 +392,19 @@ TEXT runtime·jmpdefer(SB),NOSPLIT,$0-8
 	MOVW	0(REGCTXT), R4
 	JMP	(R4)
 
-// Save state of caller into g->sched. Smashes R1.
-TEXT gosave<>(SB),NOSPLIT|NOFRAME,$0
-	MOVW	R31, (g_sched+gobuf_pc)(g)
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes R1.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT|NOFRAME,$0
+	MOVW	$runtime·systemstack_switch(SB), R1
+	ADDU	$8, R1	// get past prologue
+	MOVW	R1, (g_sched+gobuf_pc)(g)
 	MOVW	R29, (g_sched+gobuf_sp)(g)
 	MOVW	R0, (g_sched+gobuf_lr)(g)
 	MOVW	R0, (g_sched+gobuf_ret)(g)
+	MOVW	g, (g_sched+gobuf_g)(g)
 	// Assert ctxt is zero. See func save.
 	MOVW	(g_sched+gobuf_ctxt)(g), R1
 	BEQ	R1, 2(PC)
@@ -431,7 +429,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-12
 	MOVW	m_g0(R5), R6
 	BEQ	R6, g, g0
 
-	JAL	gosave<>(SB)
+	JAL	gosave_systemstack_switch<>(SB)
 	MOVW	R6, g
 	JAL	runtime·save_g(SB)
 	MOVW	(g_sched+gobuf_sp)(g), R29
diff --git a/src/runtime/asm_ppc64x.s b/src/runtime/asm_ppc64x.s
index 834023cce1..a99a61fd88 100644
--- a/src/runtime/asm_ppc64x.s
+++ b/src/runtime/asm_ppc64x.s
@@ -229,22 +229,12 @@ TEXT runtime·systemstack(SB), NOSPLIT, $0-8
 switch:
 	// save our state in g->sched. Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVD	$runtime·systemstack_switch(SB), R6
-	ADD     $16, R6 // get past prologue (including r2-setting instructions when they're there)
-	MOVD	R6, (g_sched+gobuf_pc)(g)
-	MOVD	R1, (g_sched+gobuf_sp)(g)
-	MOVD	R0, (g_sched+gobuf_lr)(g)
-	MOVD	g, (g_sched+gobuf_g)(g)
+	BL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOVD	R5, g
 	BL	runtime·save_g(SB)
-	MOVD	(g_sched+gobuf_sp)(g), R3
-	// make it look like mstart called systemstack on g0, to stop traceback
-	SUB	$FIXED_FRAME, R3
-	MOVD	$runtime·mstart(SB), R4
-	MOVD	R4, 0(R3)
-	MOVD	R3, R1
+	MOVD	(g_sched+gobuf_sp)(g), R1
 
 	// call target function
 	MOVD	0(R11), R12	// code pointer
@@ -534,13 +524,19 @@ TEXT runtime·jmpdefer(SB), NOSPLIT|NOFRAME, $0-16
 	MOVD	R12, CTR
 	BR	(CTR)
 
-// Save state of caller into g->sched. Smashes R31.
-TEXT gosave<>(SB),NOSPLIT|NOFRAME,$0
-	MOVD	LR, R31
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes R31.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT|NOFRAME,$0
+	MOVD	$runtime·systemstack_switch(SB), R31
+	ADD     $16, R31 // get past prologue (including r2-setting instructions when they're there)
 	MOVD	R31, (g_sched+gobuf_pc)(g)
 	MOVD	R1, (g_sched+gobuf_sp)(g)
 	MOVD	R0, (g_sched+gobuf_lr)(g)
 	MOVD	R0, (g_sched+gobuf_ret)(g)
+	MOVD	g, (g_sched+gobuf_g)(g)
 	// Assert ctxt is zero. See func save.
 	MOVD	(g_sched+gobuf_ctxt)(g), R31
 	CMP	R0, R31
@@ -577,7 +573,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-20
 	MOVD	m_g0(R8), R6
 	CMP	R6, g
 	BEQ	g0
-	BL	gosave<>(SB)
+	BL	gosave_systemstack_switch<>(SB)
 	MOVD	R6, g
 	BL	runtime·save_g(SB)
 	MOVD	(g_sched+gobuf_sp)(g), R1
diff --git a/src/runtime/asm_riscv64.s b/src/runtime/asm_riscv64.s
index 3d0349471a..fb7c6530dc 100644
--- a/src/runtime/asm_riscv64.s
+++ b/src/runtime/asm_riscv64.s
@@ -114,21 +114,12 @@ TEXT runtime·systemstack(SB), NOSPLIT, $0-8
 switch:
 	// save our state in g->sched. Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOV	$runtime·systemstack_switch(SB), T2
-	ADD	$8, T2	// get past prologue
-	MOV	T2, (g_sched+gobuf_pc)(g)
-	MOV	X2, (g_sched+gobuf_sp)(g)
-	MOV	ZERO, (g_sched+gobuf_lr)(g)
-	MOV	g, (g_sched+gobuf_g)(g)
+	CALL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOV	T1, g
 	CALL	runtime·save_g(SB)
 	MOV	(g_sched+gobuf_sp)(g), T0
-	// make it look like mstart called systemstack on g0, to stop traceback
-	ADD	$-8, T0
-	MOV	$runtime·mstart(SB), T1
-	MOV	T1, 0(T0)
 	MOV	T0, X2
 
 	// call target function
@@ -297,12 +288,19 @@ TEXT runtime·mcall(SB), NOSPLIT|NOFRAME, $0-8
 	JALR	RA, T1
 	JMP	runtime·badmcall2(SB)
 
-// Save state of caller into g->sched. Smashes X31.
-TEXT gosave<>(SB),NOSPLIT|NOFRAME,$0
-	MOV	X1, (g_sched+gobuf_pc)(g)
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes X31.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT|NOFRAME,$0
+	MOV	$runtime·systemstack_switch(SB), X31
+	ADD	$8, X31	// get past prologue
+	MOV	X31, (g_sched+gobuf_pc)(g)
 	MOV	X2, (g_sched+gobuf_sp)(g)
 	MOV	ZERO, (g_sched+gobuf_lr)(g)
 	MOV	ZERO, (g_sched+gobuf_ret)(g)
+	MOV	g, (g_sched+gobuf_g)(g)
 	// Assert ctxt is zero. See func save.
 	MOV	(g_sched+gobuf_ctxt)(g), X31
 	BEQ	ZERO, X31, 2(PC)
@@ -327,7 +325,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-20
 	MOV	m_g0(X6), X7
 	BEQ	X7, g, g0
 
-	CALL	gosave<>(SB)
+	CALL	gosave_systemstack_switch<>(SB)
 	MOV	X7, g
 	CALL	runtime·save_g(SB)
 	MOV	(g_sched+gobuf_sp)(g), X2
diff --git a/src/runtime/asm_s390x.s b/src/runtime/asm_s390x.s
index fbd185c353..43244d961f 100644
--- a/src/runtime/asm_s390x.s
+++ b/src/runtime/asm_s390x.s
@@ -256,22 +256,12 @@ TEXT runtime·systemstack(SB), NOSPLIT, $0-8
 switch:
 	// save our state in g->sched.  Pretend to
 	// be systemstack_switch if the G stack is scanned.
-	MOVD	$runtime·systemstack_switch(SB), R6
-	ADD	$16, R6	// get past prologue
-	MOVD	R6, (g_sched+gobuf_pc)(g)
-	MOVD	R15, (g_sched+gobuf_sp)(g)
-	MOVD	$0, (g_sched+gobuf_lr)(g)
-	MOVD	g, (g_sched+gobuf_g)(g)
+	BL	gosave_systemstack_switch<>(SB)
 
 	// switch to g0
 	MOVD	R5, g
 	BL	runtime·save_g(SB)
-	MOVD	(g_sched+gobuf_sp)(g), R3
-	// make it look like mstart called systemstack on g0, to stop traceback
-	SUB	$8, R3
-	MOVD	$runtime·mstart(SB), R4
-	MOVD	R4, 0(R3)
-	MOVD	R3, R15
+	MOVD	(g_sched+gobuf_sp)(g), R15
 
 	// call target function
 	MOVD	0(R12), R3	// code pointer
@@ -498,12 +488,19 @@ TEXT runtime·jmpdefer(SB),NOSPLIT|NOFRAME,$0-16
 	MOVD	0(R12), R3
 	BR	(R3)
 
-// Save state of caller into g->sched. Smashes R1.
-TEXT gosave<>(SB),NOSPLIT|NOFRAME,$0
-	MOVD	LR, (g_sched+gobuf_pc)(g)
+// Save state of caller into g->sched,
+// but using fake PC from systemstack_switch.
+// Must only be called from functions with no locals ($0)
+// or else unwinding from systemstack_switch is incorrect.
+// Smashes R1.
+TEXT gosave_systemstack_switch<>(SB),NOSPLIT|NOFRAME,$0
+	MOVD	$runtime·systemstack_switch(SB), R1
+	ADD	$16, R1	// get past prologue
+	MOVD	R1, (g_sched+gobuf_pc)(g)
 	MOVD	R15, (g_sched+gobuf_sp)(g)
 	MOVD	$0, (g_sched+gobuf_lr)(g)
 	MOVD	$0, (g_sched+gobuf_ret)(g)
+	MOVD	g, (g_sched+gobuf_g)(g)
 	// Assert ctxt is zero. See func save.
 	MOVD	(g_sched+gobuf_ctxt)(g), R1
 	CMPBEQ	R1, $0, 2(PC)
@@ -529,7 +526,7 @@ TEXT ·asmcgocall(SB),NOSPLIT,$0-20
 	MOVD	g_m(g), R6
 	MOVD	m_g0(R6), R6
 	CMPBEQ	R6, g, g0
-	BL	gosave<>(SB)
+	BL	gosave_systemstack_switch<>(SB)
 	MOVD	R6, g
 	BL	runtime·save_g(SB)
 	MOVD	(g_sched+gobuf_sp)(g), R15
diff --git a/src/runtime/sys_windows_386.s b/src/runtime/sys_windows_386.s
index c556e3a3c2..576dd28771 100644
--- a/src/runtime/sys_windows_386.s
+++ b/src/runtime/sys_windows_386.s
@@ -94,7 +94,7 @@ TEXT sigtramp<>(SB),NOSPLIT,$0-0
 	JNE	2(PC)
 	CALL	runtime·badsignal2(SB)
 
-	// save g and SP in case of stack switch
+	// save g in case of stack switch
 	MOVL	DX, 32(SP)	// g
 	MOVL	SP, 36(SP)
 
@@ -108,13 +108,9 @@ TEXT sigtramp<>(SB),NOSPLIT,$0-0
 	get_tls(BP)
 	MOVL	BX, g(BP)
 	MOVL	(g_sched+gobuf_sp)(BX), DI
-	// make it look like mstart called us on g0, to stop traceback
-	SUBL	$4, DI
-	MOVL	$runtime·mstart(SB), 0(DI)
-	// traceback will think that we've done SUBL
-	// on this stack, so subtract them here to match.
-	// (we need room for sighandler arguments anyway).
+	// make room for sighandler arguments
 	// and re-save old SP for restoring later.
+	// (note that the 36(DI) here must match the 36(SP) above.)
 	SUBL	$40, DI
 	MOVL	SP, 36(DI)
 	MOVL	DI, SP
@@ -132,7 +128,7 @@ g0:
 	// switch back to original stack and g
 	// no-op if we never left.
 	MOVL	36(SP), SP
-	MOVL	32(SP), DX
+	MOVL	32(SP), DX	// note: different SP
 	get_tls(BP)
 	MOVL	DX, g(BP)
 
diff --git a/src/runtime/sys_windows_amd64.s b/src/runtime/sys_windows_amd64.s
index 9cd14016b0..6f2abb5444 100644
--- a/src/runtime/sys_windows_amd64.s
+++ b/src/runtime/sys_windows_amd64.s
@@ -151,16 +151,10 @@ TEXT sigtramp<>(SB),NOSPLIT|NOFRAME,$0-0
 	get_tls(BP)
 	MOVQ	BX, g(BP)
 	MOVQ	(g_sched+gobuf_sp)(BX), DI
-	// make it look like mstart called us on g0, to stop traceback
-	SUBQ	$8, DI
-	MOVQ	$runtime·mstart(SB), SI
-	MOVQ	SI, 0(DI)
-	// traceback will think that we've done PUSHFQ and SUBQ
-	// on this stack, so subtract them here to match.
-	// (we need room for sighandler arguments anyway).
+	// make room for sighandler arguments
 	// and re-save old SP for restoring later.
-	SUBQ	$(112+8), DI
-	// save g, save old stack pointer.
+	// (note that the 104(DI) here must match the 104(SP) above.)
+	SUBQ	$120, DI
 	MOVQ	SP, 104(DI)
 	MOVQ	DI, SP
 
diff --git a/src/runtime/sys_windows_arm.s b/src/runtime/sys_windows_arm.s
index d2bdc50e3b..7d134dd3f4 100644
--- a/src/runtime/sys_windows_arm.s
+++ b/src/runtime/sys_windows_arm.s
@@ -6,6 +6,8 @@
 #include "go_tls.h"
 #include "textflag.h"
 
+// Note: For system ABI, R0-R3 are args, R4-R11 are callee-save.
+
 // void runtime·asmstdcall(void *c);
 TEXT runtime·asmstdcall(SB),NOSPLIT|NOFRAME,$0
 	MOVM.DB.W [R4, R5, R14], (R13)	// push {r4, r5, lr}
@@ -139,11 +141,10 @@ TEXT sigtramp<>(SB),NOSPLIT|NOFRAME,$0
 	MOVW	(g_sched+gobuf_sp)(g), R3	// R3 = g->gobuf.sp
 	BL      runtime·save_g(SB)
 
-	// traceback will think that we've done PUSH and SUB
-	// on this stack, so subtract them here to match.
-	// (we need room for sighandler arguments anyway).
+	// make room for sighandler arguments
 	// and re-save old SP for restoring later.
-	SUB	$(40+8+20), R3
+	// (note that the 24(R3) here must match the 24(R13) above.)
+	SUB	$40, R3
 	MOVW	R13, 24(R3)		// save old stack pointer
 	MOVW	R3, R13			// switch stack
 
@@ -151,22 +152,14 @@ g0:
 	MOVW	0(R6), R2	// R2 = ExceptionPointers->ExceptionRecord
 	MOVW	4(R6), R3	// R3 = ExceptionPointers->ContextRecord
 
-	// make it look like mstart called us on g0, to stop traceback
-	MOVW    $runtime·mstart(SB), R4
-
-	MOVW	R4, 0(R13)	// Save link register for traceback
+	MOVW	$0, R4
+	MOVW	R4, 0(R13)	// No saved link register.
 	MOVW	R2, 4(R13)	// Move arg0 (ExceptionRecord) into position
 	MOVW	R3, 8(R13)	// Move arg1 (ContextRecord) into position
 	MOVW	R5, 12(R13)	// Move arg2 (original g) into position
 	BL	(R7)		// Call the go routine
 	MOVW	16(R13), R4	// Fetch return value from stack
 
-	// Compute the value of the g0 stack pointer after deallocating
-	// this frame, then allocating 8 bytes. We may need to store
-	// the resume SP and PC on the g0 stack to work around
-	// control flow guard when we resume from the exception.
-	ADD	$(40+20), R13, R12
-
 	// switch back to original stack and g
 	MOVW	24(R13), R13
 	MOVW	20(R13), g
@@ -192,33 +185,36 @@ done:
 	// If returntramp has already been set up by a previous exception
 	// handler, don't clobber the stored SP and PC on the stack.
 	MOVW	4(R3), R3			// PEXCEPTION_POINTERS->Context
-	MOVW	0x40(R3), R2			// load PC from context record
+	MOVW	context_pc(R3), R2		// load PC from context record
 	MOVW	$returntramp<>(SB), R1
 	CMP	R1, R2
 	B.EQ	return				// do not clobber saved SP/PC
 
-	// Save resume SP and PC on g0 stack
-	MOVW	0x38(R3), R2			// load SP from context record
-	MOVW	R2, 0(R12)			// Store resume SP on g0 stack
-	MOVW	0x40(R3), R2			// load PC from context record
-	MOVW	R2, 4(R12)			// Store resume PC on g0 stack
+	// Save resume SP and PC into R0, R1.
+	MOVW	context_spr(R3), R2
+	MOVW	R2, context_r0(R3)
+	MOVW	context_pc(R3), R2
+	MOVW	R2, context_r1(R3)
 
 	// Set up context record to return to returntramp on g0 stack
-	MOVW	R12, 0x38(R3)			// save g0 stack pointer
-						// in context record
-	MOVW	$returntramp<>(SB), R2	// save resume address
-	MOVW	R2, 0x40(R3)			// in context record
+	MOVW	R12, context_spr(R3)
+	MOVW	$returntramp<>(SB), R2
+	MOVW	R2, context_pc(R3)
 
 return:
 	B	(R14)				// return
 
-//
 // Trampoline to resume execution from exception handler.
 // This is part of the control flow guard workaround.
 // It switches stacks and jumps to the continuation address.
-//
+// R0 and R1 are set above at the end of sigtramp<>
+// in the context that starts executing at returntramp<>.
 TEXT returntramp<>(SB),NOSPLIT|NOFRAME,$0
-	MOVM.IA	(R13), [R13, R15]		// ldm sp, [sp, pc]
+	// Important: do not smash LR,
+	// which is set to a live value when handling
+	// a signal by pushing a call to sigpanic onto the stack.
+	MOVW	R0, R13
+	B	(R1)
 
 TEXT runtime·exceptiontramp(SB),NOSPLIT|NOFRAME,$0
 	MOVW	$runtime·exceptionhandler(SB), R1