runtime: fix cgo to handle the case where the G stack is copied.

Tests will come in a separate CL after the funcdata stuff is resolved.

Update #8696

LGTM=iant, rsc
R=rsc, iant
CC=golang-codereviews
https://golang.org/cl/138330045

src/runtime/asm_386.s

@@ -684,15 +684,19 @@ TEXT asmcgocall<>(SB),NOSPLIT,$0-12
 	SUBL	$32, SP
 	ANDL	$~15, SP	// alignment, perhaps unnecessary
 	MOVL	DI, 8(SP)	// save g
-	MOVL	DX, 4(SP)	// save SP
+	MOVL	(g_stack+stack_hi)(DI), DI
+	SUBL	DX, DI
+	MOVL	DI, 4(SP)	// save depth in stack (can't just save SP, as stack might be copied during a callback)
 	MOVL	BX, 0(SP)	// first argument in x86-32 ABI
 	CALL	AX
 
 	// Restore registers, g, stack pointer.
 	get_tls(CX)
 	MOVL	8(SP), DI
+	MOVL	(g_stack+stack_hi)(DI), SI
+	SUBL	4(SP), SI
 	MOVL	DI, g(CX)
-	MOVL	4(SP), SP
+	MOVL	SI, SP
 	RET
 
 // cgocallback(void (*fn)(void*), void *frame, uintptr framesize)

src/runtime/asm_amd64.s

@@ -670,7 +670,9 @@ nosave:
 	SUBQ	$64, SP
 	ANDQ	$~15, SP	// alignment for gcc ABI
 	MOVQ	DI, 48(SP)	// save g
-	MOVQ	DX, 40(SP)	// save SP
+	MOVQ	(g_stack+stack_hi)(DI), DI
+	SUBQ	DX, DI
+	MOVQ	DI, 40(SP)	// save depth in stack (can't just save SP, as stack might be copied during a callback)
 	MOVQ	BX, DI		// DI = first argument in AMD64 ABI
 	MOVQ	BX, CX		// CX = first argument in Win64
 	CALL	AX
@@ -678,8 +680,10 @@ nosave:
 	// Restore registers, g, stack pointer.
 	get_tls(CX)
 	MOVQ	48(SP), DI
+	MOVQ	(g_stack+stack_hi)(DI), SI
+	SUBQ	40(SP), SI
 	MOVQ	DI, g(CX)
-	MOVQ	40(SP), SP
+	MOVQ	SI, SP
 	RET
 
 // cgocallback(void (*fn)(void*), void *frame, uintptr framesize)

src/runtime/asm_arm.s

@@ -520,15 +520,20 @@ asmcgocall_g0:
 	SUB	$24, R13
 	BIC	$0x7, R13	// alignment for gcc ABI
 	MOVW	R4, 20(R13) // save old g
-	MOVW	R2, 16(R13)	// save old SP
+	MOVW	(g_stack+stack_hi)(R4), R4
+	SUB	R2, R4
+	MOVW	R4, 16(R13)	// save depth in stack (can't just save SP, as stack might be copied during a callback)
 	BL	(R1)
 
 	// Restore registers, g, stack pointer.
 	MOVW	R0, R5
 	MOVW	20(R13), R0
 	BL	setg<>(SB)
+	MOVW	(g_stack+stack_hi)(g), R1
+	MOVW	16(R13), R2
+	SUB	R2, R1
 	MOVW	R5, R0
-	MOVW	16(R13), R13
+	MOVW	R1, R13
 	RET
 
 // cgocallback(void (*fn)(void*), void *frame, uintptr framesize)