runtime: support windows/arm

Updates #26148

Change-Id: I8f68b2c926c7b11dc86c9664ed7ff2d2f78b64b4
Reviewed-on: https://go-review.googlesource.com/128715
Run-TryBot: Ian Lance Taylor <iant@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Ian Lance Taylor <iant@golang.org>

src/cmd/vet/all/whitelist/windows_386.txt

@@ -3,7 +3,6 @@
 runtime/sys_windows_386.s: [386] profileloop: use of 4(SP) points beyond argument frame
 runtime/sys_windows_386.s: [386] ctrlhandler: 4(SP) should be _type+0(FP)
 runtime/sys_windows_386.s: [386] setldt: function setldt missing Go declaration
-runtime/zcallback_windows.s: [386] callbackasm: function callbackasm missing Go declaration
 runtime/sys_windows_386.s: [386] callbackasm1+0: function callbackasm1+0 missing Go declaration
 runtime/sys_windows_386.s: [386] tstart: function tstart missing Go declaration
 runtime/sys_windows_386.s: [386] tstart_stdcall: RET without writing to 4-byte ret+4(FP)

src/cmd/vet/all/whitelist/windows_amd64.txt

@@ -6,4 +6,3 @@ runtime/sys_windows_amd64.s: [amd64] callbackasm1: function callbackasm1 missing
 runtime/sys_windows_amd64.s: [amd64] tstart_stdcall: RET without writing to 4-byte ret+8(FP)
 runtime/sys_windows_amd64.s: [amd64] settls: function settls missing Go declaration
 runtime/sys_windows_amd64.s: [amd64] cannot check cross-package assembly function: now is in package time
-runtime/zcallback_windows.s: [amd64] callbackasm: function callbackasm missing Go declaration

src/runtime/asm_arm.s

@@ -784,6 +784,9 @@ TEXT setg<>(SB),NOSPLIT|NOFRAME,$0-0
 	MOVW	R0, g
 
 	// Save g to thread-local storage.
+#ifdef GOOS_windows
+	B	runtime·save_g(SB)
+#else
 	MOVB	runtime·iscgo(SB), R0
 	CMP	$0, R0
 	B.EQ	2(PC)
@@ -791,6 +794,7 @@ TEXT setg<>(SB),NOSPLIT|NOFRAME,$0-0
 
 	MOVW	g, R0
 	RET
+#endif
 
 TEXT runtime·emptyfunc(SB),0,$0-0
 	RET

src/runtime/defs_windows_386.go

@@ -104,6 +104,9 @@ type context struct {
 func (c *context) ip() uintptr { return uintptr(c.eip) }
 func (c *context) sp() uintptr { return uintptr(c.esp) }
 
+// 386 does not have link register, so this returns 0.
+func (c *context) lr() uintptr { return 0 }
+
 func (c *context) setip(x uintptr) { c.eip = uint32(x) }
 func (c *context) setsp(x uintptr) { c.esp = uint32(x) }
 

src/runtime/defs_windows_amd64.go

@@ -119,6 +119,9 @@ type context struct {
 func (c *context) ip() uintptr { return uintptr(c.rip) }
 func (c *context) sp() uintptr { return uintptr(c.rsp) }
 
+// Amd64 does not have link register, so this returns 0.
+func (c *context) lr() uintptr { return 0 }
+
 func (c *context) setip(x uintptr) { c.rip = uint64(x) }
 func (c *context) setsp(x uintptr) { c.rsp = uint64(x) }
 

src/runtime/defs_windows_arm.go

+// Copyright 2018 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 runtime
+
+const (
+	_PROT_NONE  = 0
+	_PROT_READ  = 1
+	_PROT_WRITE = 2
+	_PROT_EXEC  = 4
+
+	_MAP_ANON    = 1
+	_MAP_PRIVATE = 2
+
+	_DUPLICATE_SAME_ACCESS   = 0x2
+	_THREAD_PRIORITY_HIGHEST = 0x2
+
+	_SIGINT           = 0x2
+	_CTRL_C_EVENT     = 0x0
+	_CTRL_BREAK_EVENT = 0x1
+
+	_CONTEXT_CONTROL = 0x10001
+	_CONTEXT_FULL    = 0x10007
+
+	_EXCEPTION_ACCESS_VIOLATION     = 0xc0000005
+	_EXCEPTION_BREAKPOINT           = 0x80000003
+	_EXCEPTION_FLT_DENORMAL_OPERAND = 0xc000008d
+	_EXCEPTION_FLT_DIVIDE_BY_ZERO   = 0xc000008e
+	_EXCEPTION_FLT_INEXACT_RESULT   = 0xc000008f
+	_EXCEPTION_FLT_OVERFLOW         = 0xc0000091
+	_EXCEPTION_FLT_UNDERFLOW        = 0xc0000093
+	_EXCEPTION_INT_DIVIDE_BY_ZERO   = 0xc0000094
+	_EXCEPTION_INT_OVERFLOW         = 0xc0000095
+
+	_INFINITE     = 0xffffffff
+	_WAIT_TIMEOUT = 0x102
+
+	_EXCEPTION_CONTINUE_EXECUTION = -0x1
+	_EXCEPTION_CONTINUE_SEARCH    = 0x0
+)
+
+type systeminfo struct {
+	anon0                       [4]byte
+	dwpagesize                  uint32
+	lpminimumapplicationaddress *byte
+	lpmaximumapplicationaddress *byte
+	dwactiveprocessormask       uint32
+	dwnumberofprocessors        uint32
+	dwprocessortype             uint32
+	dwallocationgranularity     uint32
+	wprocessorlevel             uint16
+	wprocessorrevision          uint16
+}
+
+type exceptionrecord struct {
+	exceptioncode        uint32
+	exceptionflags       uint32
+	exceptionrecord      *exceptionrecord
+	exceptionaddress     *byte
+	numberparameters     uint32
+	exceptioninformation [15]uint32
+}
+
+type neon128 struct {
+	low  uint64
+	high int64
+}
+
+type context struct {
+	contextflags uint32
+	r0           uint32
+	r1           uint32
+	r2           uint32
+	r3           uint32
+	r4           uint32
+	r5           uint32
+	r6           uint32
+	r7           uint32
+	r8           uint32
+	r9           uint32
+	r10          uint32
+	r11          uint32
+	r12          uint32
+
+	spr  uint32
+	lrr  uint32
+	pc   uint32
+	cpsr uint32
+
+	fpscr   uint32
+	padding uint32
+
+	floatNeon [16]neon128
+
+	bvr      [8]uint32
+	bcr      [8]uint32
+	wvr      [1]uint32
+	wcr      [1]uint32
+	padding2 [2]uint32
+}
+
+func (c *context) ip() uintptr { return uintptr(c.pc) }
+func (c *context) sp() uintptr { return uintptr(c.spr) }
+func (c *context) lr() uintptr { return uintptr(c.lrr) }
+
+func (c *context) setip(x uintptr) { c.pc = uint32(x) }
+func (c *context) setsp(x uintptr) { c.spr = uint32(x) }
+
+func dumpregs(r *context) {
+	print("r0   ", hex(r.r0), "\n")
+	print("r1   ", hex(r.r1), "\n")
+	print("r2   ", hex(r.r2), "\n")
+	print("r3   ", hex(r.r3), "\n")
+	print("r4   ", hex(r.r4), "\n")
+	print("r5   ", hex(r.r5), "\n")
+	print("r6   ", hex(r.r6), "\n")
+	print("r7   ", hex(r.r7), "\n")
+	print("r8   ", hex(r.r8), "\n")
+	print("r9   ", hex(r.r9), "\n")
+	print("r10  ", hex(r.r10), "\n")
+	print("r11  ", hex(r.r11), "\n")
+	print("r12  ", hex(r.r12), "\n")
+	print("sp   ", hex(r.spr), "\n")
+	print("lr   ", hex(r.lrr), "\n")
+	print("pc   ", hex(r.pc), "\n")
+	print("cpsr ", hex(r.cpsr), "\n")
+}
+
+type overlapped struct {
+	internal     uint32
+	internalhigh uint32
+	anon0        [8]byte
+	hevent       *byte
+}
+
+type memoryBasicInformation struct {
+	baseAddress       uintptr
+	allocationBase    uintptr
+	allocationProtect uint32
+	regionSize        uintptr
+	state             uint32
+	protect           uint32
+	type_             uint32
+}
+
+func stackcheck() {
+	// TODO: not implemented on ARM
+}

src/runtime/os_windows.go

@@ -43,6 +43,7 @@ const (
 //go:cgo_import_dynamic runtime._SetWaitableTimer SetWaitableTimer%6 "kernel32.dll"
 //go:cgo_import_dynamic runtime._SuspendThread SuspendThread%1 "kernel32.dll"
 //go:cgo_import_dynamic runtime._SwitchToThread SwitchToThread%0 "kernel32.dll"
+//go:cgo_import_dynamic runtime._TlsAlloc TlsAlloc%0 "kernel32.dll"
 //go:cgo_import_dynamic runtime._VirtualAlloc VirtualAlloc%4 "kernel32.dll"
 //go:cgo_import_dynamic runtime._VirtualFree VirtualFree%3 "kernel32.dll"
 //go:cgo_import_dynamic runtime._VirtualQuery VirtualQuery%3 "kernel32.dll"
@@ -91,6 +92,7 @@ var (
 	_SetWaitableTimer,
 	_SuspendThread,
 	_SwitchToThread,
+	_TlsAlloc,
 	_VirtualAlloc,
 	_VirtualFree,
 	_VirtualQuery,
@@ -860,14 +862,34 @@ func profilem(mp *m) {
 	var r *context
 	rbuf := make([]byte, unsafe.Sizeof(*r)+15)
 
-	tls := &mp.tls[0]
-	gp := *((**g)(unsafe.Pointer(tls)))
-
 	// align Context to 16 bytes
 	r = (*context)(unsafe.Pointer((uintptr(unsafe.Pointer(&rbuf[15]))) &^ 15))
 	r.contextflags = _CONTEXT_CONTROL
 	stdcall2(_GetThreadContext, mp.thread, uintptr(unsafe.Pointer(r)))
+
+	var gp *g
+	switch GOARCH {
+	default:
+		panic("unsupported architecture")
+	case "arm":
+		// TODO(jordanrh1): this is incorrect when Go is executing
+		// on the system or signal stacks because curg returns
+		// the current user g. The true g is stored in thread
+		// local storage, which we cannot access from another CPU.
+		// We cannot pull R10 from the thread context because
+		// it might be executing C code, in which case R10
+		// would not be g.
+		gp = mp.curg
+	case "386", "amd64":
+		tls := &mp.tls[0]
+		gp = *((**g)(unsafe.Pointer(tls)))
+	}
+
+	if gp == nil {
+		sigprofNonGoPC(r.ip())
+	} else {
 		sigprof(r.ip(), r.sp(), 0, gp, mp)
+	}
 }
 
 func profileloop1(param uintptr) uint32 {

src/runtime/os_windows_arm.go

+// Copyright 2018 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 runtime
+
+//go:nosplit
+func cputicks() int64 {
+	return nanotime()
+}
+
+func checkgoarm() {
+	if goarm < 7 {
+		print("Need atomic synchronization instructions, coprocessor ",
+			"access instructions. Recompile using GOARM=7.\n")
+		exit(1)
+	}
+}

src/runtime/rt0_windows_arm.s

+// Copyright 2018 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 "go_asm.h"
+#include "go_tls.h"
+#include "textflag.h"
+
+// This is the entry point for the program from the
+// kernel for an ordinary -buildmode=exe program.
+TEXT _rt0_arm_windows(SB),NOSPLIT|NOFRAME,$0
+	B	·rt0_go(SB)

src/runtime/signal_windows.go

@@ -27,7 +27,7 @@ func lastcontinuetramp()
 
 func initExceptionHandler() {
 	stdcall2(_AddVectoredExceptionHandler, 1, funcPC(exceptiontramp))
-	if _AddVectoredContinueHandler == nil || unsafe.Sizeof(&_AddVectoredContinueHandler) == 4 {
+	if _AddVectoredContinueHandler == nil || GOARCH == "386" {
 		// use SetUnhandledExceptionFilter for windows-386 or
 		// if VectoredContinueHandler is unavailable.
 		// note: SetUnhandledExceptionFilter handler won't be called, if debugging.
@@ -177,9 +177,15 @@ func lastcontinuehandler(info *exceptionrecord, r *context, gp *g) int32 {
 	}
 	print("\n")
 
+	// TODO(jordanrh1): This may be needed for 386/AMD64 as well.
+	if GOARCH == "arm" {
+		_g_.m.throwing = 1
+		_g_.m.caughtsig.set(gp)
+	}
+
 	level, _, docrash := gotraceback()
 	if level > 0 {
-		tracebacktrap(r.ip(), r.sp(), 0, gp)
+		tracebacktrap(r.ip(), r.sp(), r.lr(), gp)
 		tracebackothers(gp)
 		dumpregs(r)
 	}

src/runtime/syscall_windows.go

@@ -25,18 +25,32 @@ func (c *wincallbackcontext) setCleanstack(cleanstack bool) {
 var (
 	cbs     callbacks
 	cbctxts **wincallbackcontext = &cbs.ctxt[0] // to simplify access to cbs.ctxt in sys_windows_*.s
-
-	callbackasm byte // type isn't really byte, it's code in runtime
 )
 
+func callbackasm()
+
 // callbackasmAddr returns address of runtime.callbackasm
 // function adjusted by i.
-// runtime.callbackasm is just a series of CALL instructions
-// (each is 5 bytes long), and we want callback to arrive at
+// On x86 and amd64, runtime.callbackasm is a series of CALL instructions,
+// and we want callback to arrive at
 // correspondent call instruction instead of start of
 // runtime.callbackasm.
+// On ARM, runtime.callbackasm is a series of mov and branch instructions.
+// R12 is loaded with the callback index. Each entry is two instructions,
+// hence 8 bytes.
 func callbackasmAddr(i int) uintptr {
-	return uintptr(add(unsafe.Pointer(&callbackasm), uintptr(i*5)))
+	var entrySize int
+	switch GOARCH {
+	default:
+		panic("unsupported architecture")
+	case "386", "amd64":
+		entrySize = 5
+	case "arm":
+		// On ARM, each entry is a MOV instruction
+		// followed by a branch instruction
+		entrySize = 8
+	}
+	return funcPC(callbackasm) + uintptr(i*entrySize)
 }
 
 //go:linkname compileCallback syscall.compileCallback

src/runtime/tls_arm.s

@@ -23,6 +23,9 @@
 #ifdef GOOS_darwin
 #define TLSG_IS_VARIABLE
 #endif
+#ifdef GOOS_windows
+#define TLSG_IS_VARIABLE
+#endif
 
 // save_g saves the g register into pthread-provided
 // thread-local memory, so that we can call externally compiled
@@ -35,6 +38,17 @@ TEXT runtime·save_g(SB),NOSPLIT|NOFRAME,$0
 	// nothing to do as nacl/arm does not use TLS at all.
 	MOVW	g, R0 // preserve R0 across call to setg<>
 	RET
+#else
+#ifdef GOOS_windows
+	// Save the value in the _TEB->TlsSlots array.
+	// Effectively implements TlsSetValue().
+	MRC	15, 0, R0, C13, C0, 2
+	ADD	$0xe10, R0
+	MOVW 	$runtime·tls_g(SB), R11
+	MOVW	(R11), R11
+	MOVW	g, R11<<2(R0)
+	MOVW	g, R0	// preserve R0 accross call to setg<>
+	RET
 #else
 	// If the host does not support MRC the linker will replace it with
 	// a call to runtime.read_tls_fallback which jumps to __kuser_get_tls.
@@ -47,6 +61,7 @@ TEXT runtime·save_g(SB),NOSPLIT|NOFRAME,$0
 	MOVW	g, R0 // preserve R0 across call to setg<>
 	RET
 #endif
+#endif
 
 // load_g loads the g register from pthread-provided
 // thread-local memory, for use after calling externally compiled
@@ -55,6 +70,16 @@ TEXT runtime·load_g(SB),NOSPLIT,$0
 #ifdef GOOS_nacl
 	// nothing to do as nacl/arm does not use TLS at all.
 	RET
+#else
+#ifdef GOOS_windows
+	// Get the value from the _TEB->TlsSlots array.
+	// Effectively implements TlsGetValue().
+	MRC	15, 0, R0, C13, C0, 2
+	ADD	$0xe10, R0
+	MOVW 	$runtime·tls_g(SB), g
+	MOVW	(g), g
+	MOVW	g<<2(R0), g
+	RET
 #else
 	// See save_g
 	MRC	15, 0, R0, C13, C0, 3 // fetch TLS base pointer
@@ -64,6 +89,7 @@ TEXT runtime·load_g(SB),NOSPLIT,$0
 	MOVW	0(R0), g
 	RET
 #endif
+#endif
 
 // This is called from rt0_go, which runs on the system stack
 // using the initial stack allocated by the OS.
@@ -76,6 +102,20 @@ TEXT runtime·load_g(SB),NOSPLIT,$0
 // Declare a dummy word ($4, not $0) to make sure the
 // frame is 8 bytes and stays 8-byte-aligned.
 TEXT runtime·_initcgo(SB),NOSPLIT,$4
+#ifdef GOOS_windows
+	MOVW	R13, R4
+	BIC	$0x7, R13
+	MOVW 	$runtime·_TlsAlloc(SB), R0
+	MOVW	(R0), R0
+	BL	(R0)
+	// Assert that slot is less than 64 so we can use _TEB->TlsSlots
+	CMP	$64, R0
+	MOVW	$runtime·abort(SB), R1
+	BL.GE	(R1)
+	MOVW 	$runtime·tls_g(SB), R1
+	MOVW	R0, (R1)
+	MOVW	R4, R13
+#else
 #ifndef GOOS_nacl
 	// if there is an _cgo_init, call it.
 	MOVW	_cgo_init(SB), R4
@@ -91,7 +131,8 @@ TEXT runtime·_initcgo(SB),NOSPLIT,$4
 	MOVW	$setg_gcc<>(SB), R1 	// arg 1: setg
 	MOVW	g, R0 			// arg 0: G
 	BL	(R4) // will clobber R0-R3
-#endif
+#endif // GOOS_nacl
+#endif // GOOS_windows
 nocgo:
 	RET
 

src/runtime/wincallback.go

@@ -17,11 +17,12 @@ import (
 
 const maxCallback = 2000
 
-func genasm() {
+func genasm386Amd64() {
 	var buf bytes.Buffer
 
-	buf.WriteString(`// generated by wincallback.go; run go generate
+	buf.WriteString(`// Code generated by wincallback.go using 'go generate'. DO NOT EDIT.
 
+// +build 386 amd64
 // runtime·callbackasm is called by external code to
 // execute Go implemented callback function. It is not
 // called from the start, instead runtime·compilecallback
@@ -29,13 +30,43 @@ func genasm() {
 // appropriately so different callbacks start with different
 // CALL instruction in runtime·callbackasm. This determines
 // which Go callback function is executed later on.
+
 TEXT runtime·callbackasm(SB),7,$0
 `)
 	for i := 0; i < maxCallback; i++ {
 		buf.WriteString("\tCALL\truntime·callbackasm1(SB)\n")
 	}
 
-	err := ioutil.WriteFile("zcallback_windows.s", buf.Bytes(), 0666)
+	filename := fmt.Sprintf("zcallback_windows.s")
+	err := ioutil.WriteFile(filename, buf.Bytes(), 0666)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "wincallback: %s\n", err)
+		os.Exit(2)
+	}
+}
+
+func genasmArm() {
+	var buf bytes.Buffer
+
+	buf.WriteString(`// Code generated by wincallback.go using 'go generate'. DO NOT EDIT.
+
+// External code calls into callbackasm at an offset corresponding
+// to the callback index. Callbackasm is a table of MOV and B instructions.
+// The MOV instruction loads R12 with the callback index, and the
+// B instruction branches to callbackasm1.
+// callbackasm1 takes the callback index from R12 and
+// indexes into an array that stores information about each callback.
+// It then calls the Go implementation for that callback.
+#include "textflag.h"
+
+TEXT runtime·callbackasm(SB),NOSPLIT|NOFRAME,$0
+`)
+	for i := 0; i < maxCallback; i++ {
+		buf.WriteString(fmt.Sprintf("\tMOVW\t$%d, R12\n", i))
+		buf.WriteString("\tB\truntime·callbackasm1(SB)\n")
+	}
+
+	err := ioutil.WriteFile("zcallback_windows_arm.s", buf.Bytes(), 0666)
 	if err != nil {
 		fmt.Fprintf(os.Stderr, "wincallback: %s\n", err)
 		os.Exit(2)
@@ -45,7 +76,7 @@ TEXT runtime·callbackasm(SB),7,$0
 func gengo() {
 	var buf bytes.Buffer
 
-	buf.WriteString(fmt.Sprintf(`// generated by wincallback.go; run go generate
+	buf.WriteString(fmt.Sprintf(`// Code generated by wincallback.go using 'go generate'. DO NOT EDIT.
 
 package runtime
 
@@ -59,6 +90,7 @@ const cb_max = %d // maximum number of windows callbacks allowed
 }
 
 func main() {
-	genasm()
+	genasm386Amd64()
+	genasmArm()
 	gengo()
 }

src/runtime/zcallback_windows.go

-// generated by wincallback.go; run go generate
+// Code generated by wincallback.go using 'go generate'. DO NOT EDIT.
 
 package runtime
 

src/runtime/zcallback_windows.s

-// generated by wincallback.go; run go generate
+// Code generated by wincallback.go using 'go generate'. DO NOT EDIT.
 
+// +build 386 amd64
 // runtime·callbackasm is called by external code to
 // execute Go implemented callback function. It is not
 // called from the start, instead runtime·compilecallback
@@ -7,6 +8,7 @@
 // appropriately so different callbacks start with different
 // CALL instruction in runtime·callbackasm. This determines
 // which Go callback function is executed later on.
+
 TEXT runtime·callbackasm(SB),7,$0
 	CALL	runtime·callbackasm1(SB)
 	CALL	runtime·callbackasm1(SB)