runtime: fix uses of ABIInternal PCs in assembly

The covers three kinds of uses:

1. Calls of closures from assembly. These are always ABIInternal calls
without wrappers. I went through every indirect call in the runtime
and I think mcall is the only case of assembly calling a Go closure in
a way that's affected by ABIInternal. systemstack also calls a
closure, but it takes no arguments.

2. Calls of Go functions that expect raw ABIInternal pointers. I also
only found one of these: callbackasm1 -> cgocallback on Windows. These
are trickier to find, though.

3. Finally, I found one case on NetBSD where new OS threads were
directly calling the Go runtime entry-point from assembly via a PC,
rather than going through a wrapper. This meant new threads may not
have special registers set up. In this case, a change on all other
OSes had already forced new thread entry to go through an ABI wrapper,
so I just caught NetBSD up with that change.

With this change, I'm able to run a "hello world" with
GOEXPERIMENT=regabi,regabiargs.

For #40724.

Change-Id: I2a6d0e530c4fd4edf13484d923891c6160d683aa
Reviewed-on: https://go-review.googlesource.com/c/go/+/305669
Trust: Austin Clements <austin@google.com>
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Go Bot <gobot@golang.org>

src/cmd/internal/objabi/util.go

@@ -166,8 +166,8 @@ func init() {
 	if Experiment.RegabiG && !Experiment.RegabiWrappers {
 		panic("GOEXPERIMENT regabig requires regabiwrappers")
 	}
-	if Experiment.RegabiArgs && !(Experiment.RegabiWrappers && Experiment.RegabiReflect && Experiment.RegabiDefer) {
-		panic("GOEXPERIMENT regabiargs requires regabiwrappers,regabireflect,regabidefer")
+	if Experiment.RegabiArgs && !(Experiment.RegabiWrappers && Experiment.RegabiG && Experiment.RegabiReflect && Experiment.RegabiDefer) {
+		panic("GOEXPERIMENT regabiargs requires regabiwrappers,regabig,regabireflect,regabidefer")
 	}
 
 	// Set GOEXPERIMENT to the parsed and canonicalized set of experiments.
@@ -242,7 +242,11 @@ type ExpFlags struct {
 	// RegabiArgs enables register arguments/results in all
 	// compiled Go functions.
 	//
-	// Requires wrappers, reflect, defer.
+	// Requires wrappers (to do ABI translation), g (because
+	// runtime assembly that's been ported to ABIInternal uses the
+	// G register), reflect (so reflection calls use registers),
+	// and defer (because the runtime doesn't support passing
+	// register arguments to defer/go).
 	RegabiArgs bool
 }
 

src/runtime/asm_amd64.s

@@ -285,6 +285,34 @@ TEXT gogo<>(SB), NOSPLIT, $0
 // Switch to m->g0's stack, call fn(g).
 // Fn must never return. It should gogo(&g->sched)
 // to keep running g.
+#ifdef GOEXPERIMENT_REGABI_ARGS
+TEXT runtime·mcall<ABIInternal>(SB), NOSPLIT, $0-8
+	MOVQ	AX, DX	// DX = fn
+
+	// save state in g->sched
+	MOVQ	0(SP), BX	// caller's PC
+	MOVQ	BX, (g_sched+gobuf_pc)(R14)
+	LEAQ	fn+0(FP), BX	// caller's SP
+	MOVQ	BX, (g_sched+gobuf_sp)(R14)
+	MOVQ	BP, (g_sched+gobuf_bp)(R14)
+
+	// switch to m->g0 & its stack, call fn
+	MOVQ	g_m(R14), BX
+	MOVQ	m_g0(BX), SI	// SI = g.m.g0
+	CMPQ	SI, R14	// if g == m->g0 call badmcall
+	JNE	goodm
+	JMP	runtime·badmcall(SB)
+goodm:
+	MOVQ	R14, AX		// AX (and arg 0) = g
+	MOVQ	SI, R14		// g = g.m.g0
+	get_tls(CX)		// Set G in TLS
+	MOVQ	R14, g(CX)
+	MOVQ	(g_sched+gobuf_sp)(R14), SP	// sp = g0.sched.sp
+	MOVQ	0(DX), R12
+	CALL	R12		// fn(g)
+	JMP	runtime·badmcall2(SB)
+	RET
+#else
 TEXT runtime·mcall(SB), NOSPLIT, $0-8
 	MOVQ	fn+0(FP), DI
 
@@ -315,6 +343,7 @@ TEXT runtime·mcall(SB), NOSPLIT, $0-8
 	MOVQ	$runtime·badmcall2(SB), AX
 	JMP	AX
 	RET
+#endif
 
 // systemstack_switch is a dummy routine that systemstack leaves at the bottom
 // of the G stack. We need to distinguish the routine that

src/runtime/cgocall.go

@@ -195,7 +195,7 @@ func cgocall(fn, arg unsafe.Pointer) int32 {
 	return errno
 }
 
-// Call from C back to Go.
+// Call from C back to Go. fn must point to an ABIInternal Go entry-point.
 //go:nosplit
 func cgocallbackg(fn, frame unsafe.Pointer, ctxt uintptr) {
 	gp := getg()

src/runtime/os_netbsd.go

@@ -228,7 +228,11 @@ func newosproc(mp *m) {
 	}
 }
 
-// netbsdMStart is the function call that starts executing a newly
+// mstart is the entry-point for new Ms.
+// It is written in assembly, uses ABI0, is marked TOPFRAME, and calls netbsdMstart0.
+func netbsdMstart()
+
+// netbsdMStart0 is the function call that starts executing a newly
 // created thread. On NetBSD, a new thread inherits the signal stack
 // of the creating thread. That confuses minit, so we remove that
 // signal stack here before calling the regular mstart. It's a bit
@@ -236,10 +240,10 @@ func newosproc(mp *m) {
 // it's a simple change that keeps NetBSD working like other OS's.
 // At this point all signals are blocked, so there is no race.
 //go:nosplit
-func netbsdMstart() {
+func netbsdMstart0() {
 	st := stackt{ss_flags: _SS_DISABLE}
 	sigaltstack(&st, nil)
-	mstart()
+	mstart0()
 }
 
 func osinit() {

src/runtime/proc.go

@@ -1258,7 +1258,7 @@ func mStackIsSystemAllocated() bool {
 }
 
 // mstart is the entry-point for new Ms.
-// It is written in assembly, marked TOPFRAME, and calls mstart0.
+// It is written in assembly, uses ABI0, is marked TOPFRAME, and calls mstart0.
 func mstart()
 
 // mstart0 is the Go entry-point for new Ms.

src/runtime/sys_linux_amd64.s

@@ -652,7 +652,7 @@ nog1:
 	CALL	runtime·stackcheck(SB)
 
 nog2:
-	// Call fn
+	// Call fn. This is the PC of an ABI0 function.
 	CALL	R12
 
 	// It shouldn't return. If it does, exit that thread.

src/runtime/sys_netbsd_386.s

@@ -376,6 +376,10 @@ TEXT runtime·lwp_tramp(SB),NOSPLIT,$0
 	MOVL	$0x1234, 0x1005
 	RET
 
+TEXT ·netbsdMstart(SB),NOSPLIT|TOPFRAME,$0
+	CALL	·netbsdMstart0(SB)
+	RET // not reached
+
 TEXT runtime·sigaltstack(SB),NOSPLIT,$-8
 	MOVL	$SYS___sigaltstack14, AX
 	MOVL	new+0(FP), BX

src/runtime/sys_netbsd_amd64.s

@@ -70,7 +70,7 @@ TEXT runtime·lwp_tramp(SB),NOSPLIT,$0
 	MOVQ	R9, g(CX)
 	CALL	runtime·stackcheck(SB)
 
-	// Call fn
+	// Call fn. This is an ABI0 PC.
 	CALL	R12
 
 	// It shouldn't return. If it does, exit.
@@ -78,6 +78,10 @@ TEXT runtime·lwp_tramp(SB),NOSPLIT,$0
 	SYSCALL
 	JMP	-3(PC)			// keep exiting
 
+TEXT ·netbsdMstart(SB),NOSPLIT|TOPFRAME,$0
+	CALL	·netbsdMstart0(SB)
+	RET // not reached
+
 TEXT runtime·osyield(SB),NOSPLIT,$0
 	MOVL	$SYS_sched_yield, AX
 	SYSCALL

src/runtime/sys_netbsd_arm.s

@@ -177,6 +177,10 @@ TEXT runtime·lwp_tramp(SB),NOSPLIT,$0
 	MOVW R8, (R8)
 	RET
 
+TEXT ·netbsdMstart(SB),NOSPLIT|TOPFRAME,$0
+	BL ·netbsdMstart0(SB)
+	RET // not reached
+
 TEXT runtime·usleep(SB),NOSPLIT,$16
 	MOVW usec+0(FP), R0
 	CALL runtime·usplitR0(SB)

src/runtime/sys_netbsd_arm64.s

@@ -76,6 +76,10 @@ nog:
 	MOVD	$0, R0  // crash (not reached)
 	MOVD	R0, (R8)
 
+TEXT ·netbsdMstart(SB),NOSPLIT|TOPFRAME,$0
+	CALL	·netbsdMstart0(SB)
+	RET // not reached
+
 TEXT runtime·osyield(SB),NOSPLIT,$0
 	SVC	$SYS_sched_yield
 	RET

src/runtime/sys_windows_amd64.s

@@ -320,7 +320,7 @@ TEXT runtime·callbackasm1(SB),NOSPLIT,$0
 	// Call cgocallback, which will call callbackWrap(frame).
 	MOVQ	$0, 16(SP)	// context
 	MOVQ	AX, 8(SP)	// frame (address of callbackArgs)
-	LEAQ	·callbackWrap(SB), BX
+	LEAQ	·callbackWrap<ABIInternal>(SB), BX	// cgocallback takes an ABIInternal entry-point
 	MOVQ	BX, 0(SP)	// PC of function value to call (callbackWrap)
 	CALL	·cgocallback(SB)
 	// Get callback result.