cmd/cgo: special case C ptr types to use uintptr

Some C types are declared as pointers, but C code
stores non-pointers in them.  When the Go garbage
collector sees such a pointer, it gets unhappy.

Instead, for these types represent them on the Go
side with uintptr.

We need this change to handle Apple's CoreFoundation
CF*Ref types. Users of these types might need to
update their code like we do in root_cgo_darwin.go.
The only change that is required under normal
circumstances is converting some nils to 0.
A go fix module is provided to help.

Fixes #21897

RELNOTE=yes

Change-Id: I9716cfb255dc918792625f42952aa171cd31ec1b
Reviewed-on: https://go-review.googlesource.com/66332
Run-TryBot: Keith Randall <[email protected]>
Reviewed-by: Robert Griesemer <[email protected]>
Reviewed-by: Ian Lance Taylor <[email protected]>

Author: randall77

misc/cgo/test/cgo_test.go

@@ -85,5 +85,6 @@ func Test21708(t *testing.T)                 { test21708(t) }
 func Test21809(t *testing.T)                 { test21809(t) }
 func Test6907(t *testing.T)                  { test6907(t) }
 func Test6907Go(t *testing.T)                { test6907Go(t) }
+func Test21897(t *testing.T)                 { test21897(t) }
 
 func BenchmarkCgoCall(b *testing.B) { benchCgoCall(b) }

misc/cgo/test/issue21897.go

@@ -0,0 +1,56 @@
+// Copyright 2017 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.
+
+// +build darwin,cgo,!internal
+
+package cgotest
+
+/*
+#cgo LDFLAGS: -framework CoreFoundation
+#include <CoreFoundation/CoreFoundation.h>
+*/
+import "C"
+import (
+	"runtime/debug"
+	"testing"
+	"unsafe"
+)
+
+func test21897(t *testing.T) {
+	// Please write barrier, kick in soon.
+	defer debug.SetGCPercent(debug.SetGCPercent(1))
+
+	for i := 0; i < 10000; i++ {
+		testCFNumberRef()
+		testCFDateRef()
+		testCFBooleanRef()
+		// Allocate some memory, so eventually the write barrier is enabled
+		// and it will see writes of bad pointers in the test* functions below.
+		byteSliceSink = make([]byte, 1024)
+	}
+}
+
+var byteSliceSink []byte
+
+func testCFNumberRef() {
+	var v int64 = 0
+	xCFNumberRef = C.CFNumberCreate(C.kCFAllocatorSystemDefault, C.kCFNumberSInt64Type, unsafe.Pointer(&v))
+	//fmt.Printf("CFNumberRef: %x\n", uintptr(unsafe.Pointer(xCFNumberRef)))
+}
+
+var xCFNumberRef C.CFNumberRef
+
+func testCFDateRef() {
+	xCFDateRef = C.CFDateCreate(C.kCFAllocatorSystemDefault, 0) // 0 value is 1 Jan 2001 00:00:00 GMT
+	//fmt.Printf("CFDateRef: %x\n", uintptr(unsafe.Pointer(xCFDateRef)))
+}
+
+var xCFDateRef C.CFDateRef
+
+func testCFBooleanRef() {
+	xCFBooleanRef = C.kCFBooleanFalse
+	//fmt.Printf("CFBooleanRef: %x\n", uintptr(unsafe.Pointer(xCFBooleanRef)))
+}
+
+var xCFBooleanRef C.CFBooleanRef

misc/cgo/test/issue21897b.go

@@ -0,0 +1,13 @@
+// Copyright 2017 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.
+
+// +build !darwin !cgo internal
+
+package cgotest
+
+import "testing"
+
+func test21897(t *testing.T) {
+	t.Skip("test runs only on darwin+cgo")
+}

src/cmd/cgo/doc.go

@@ -128,6 +128,10 @@ C.complexfloat (complex float), and C.complexdouble (complex double).
 The C type void* is represented by Go's unsafe.Pointer.
 The C types __int128_t and __uint128_t are represented by [16]byte.
 
+A few special C types which would normally be represented by a pointer
+type in Go are instead represented by a uintptr.  See the Special
+cases section below.
+
 To access a struct, union, or enum type directly, prefix it with
 struct_, union_, or enum_, as in C.struct_stat.
 
@@ -334,6 +338,84 @@ and of course there is nothing stopping the C code from doing anything
 it likes. However, programs that break these rules are likely to fail
 in unexpected and unpredictable ways.
 
+Special cases
+
+A few special C types which would normally be represented by a pointer
+type in Go are instead represented by a uintptr. Those types are
+the CF*Ref types from the CoreFoundation library on Darwin, including:
+
+	CFAllocatorRef
+	CFArrayRef
+	CFAttributedStringRef
+	CFBagRef
+	CFBinaryHeapRef
+	CFBitVectorRef
+	CFBooleanRef
+	CFBundleRef
+	CFCalendarRef
+	CFCharacterSetRef
+	CFDataRef
+	CFDateFormatterRef
+	CFDateRef
+	CFDictionaryRef
+	CFErrorRef
+	CFFileDescriptorRef
+	CFFileSecurityRef
+	CFLocaleRef
+	CFMachPortRef
+	CFMessagePortRef
+	CFMutableArrayRef
+	CFMutableAttributedStringRef
+	CFMutableBagRef
+	CFMutableBitVectorRef
+	CFMutableCharacterSetRef
+	CFMutableDataRef
+	CFMutableDictionaryRef
+	CFMutableSetRef
+	CFMutableStringRef
+	CFNotificationCenterRef
+	CFNullRef
+	CFNumberFormatterRef
+	CFNumberRef
+	CFPlugInInstanceRef
+	CFPlugInRef
+	CFPropertyListRef
+	CFReadStreamRef
+	CFRunLoopObserverRef
+	CFRunLoopRef
+	CFRunLoopSourceRef
+	CFRunLoopTimerRef
+	CFSetRef
+	CFSocketRef
+	CFStringRef
+	CFStringTokenizerRef
+	CFTimeZoneRef
+	CFTreeRef
+	CFTypeRef
+	CFURLCreateFromFSRef
+	CFURLEnumeratorRef
+	CFURLGetFSRef
+	CFURLRef
+	CFUUIDRef
+	CFUserNotificationRef
+	CFWriteStreamRef
+	CFXMLNodeRef
+	CFXMLParserRef
+	CFXMLTreeRef
+
+These types are uintptr on the Go side because they would otherwise
+confuse the Go garbage collector; they are sometimes not really
+pointers but data structures encoded in a pointer type. All operations
+on these types must happen in C. The proper constant to initialize an
+empty such reference is 0, not nil.
+
+This special case was introduced in Go 1.10. For auto-updating code
+from Go 1.9 and earlier, use the cftype rewrite in the Go fix tool:
+
+	go tool fix -r cftype <pkg>
+
+It will replace nil with 0 in the appropriate places.
+
 Using cgo directly
 
 Usage:

src/cmd/cgo/gcc.go

@@ -2057,6 +2057,12 @@ func (c *typeConv) Type(dtype dwarf.Type, pos token.Pos) *Type {
 		name := c.Ident("_Ctype_" + dt.Name)
 		goIdent[name.Name] = name
 		sub := c.Type(dt.Type, pos)
+		if badPointerTypedef(dt.Name) {
+			// Treat this typedef as a uintptr.
+			s := *sub
+			s.Go = c.uintptr
+			sub = &s
+		}
 		t.Go = name
 		if unionWithPointer[sub.Go] {
 			unionWithPointer[t.Go] = true
@@ -2215,6 +2221,11 @@ func (c *typeConv) FuncArg(dtype dwarf.Type, pos token.Pos) *Type {
 			if _, void := base(ptr.Type).(*dwarf.VoidType); void {
 				break
 			}
+			// ...or the typedef is one in which we expect bad pointers.
+			// It will be a uintptr instead of *X.
+			if badPointerTypedef(dt.Name) {
+				break
+			}
 
 			t = c.Type(ptr, pos)
 			if t == nil {
@@ -2547,3 +2558,51 @@ func fieldPrefix(fld []*ast.Field) string {
 	}
 	return prefix
 }
+
+// badPointerTypedef reports whether t is a C typedef that should not be considered a pointer in Go.
+// A typedef is bad if C code sometimes stores non-pointers in this type.
+// TODO: Currently our best solution is to find these manually and list them as
+// they come up. A better solution is desired.
+func badPointerTypedef(t string) bool {
+	// The real bad types are CFNumberRef and CFTypeRef.
+	// Sometimes non-pointers are stored in these types.
+	// CFTypeRef is a supertype of those, so it can have bad pointers in it as well.
+	// We return true for the other CF*Ref types just so casting between them is easier.
+	// See comment below for details about the bad pointers.
+	return goos == "darwin" && strings.HasPrefix(t, "CF") && strings.HasSuffix(t, "Ref")
+}
+
+// Comment from Darwin's CFInternal.h
+/*
+// Tagged pointer support
+// Low-bit set means tagged object, next 3 bits (currently)
+// define the tagged object class, next 4 bits are for type
+// information for the specific tagged object class.  Thus,
+// the low byte is for type info, and the rest of a pointer
+// (32 or 64-bit) is for payload, whatever the tagged class.
+//
+// Note that the specific integers used to identify the
+// specific tagged classes can and will change from release
+// to release (that's why this stuff is in CF*Internal*.h),
+// as can the definition of type info vs payload above.
+//
+#if __LP64__
+#define CF_IS_TAGGED_OBJ(PTR)	((uintptr_t)(PTR) & 0x1)
+#define CF_TAGGED_OBJ_TYPE(PTR)	((uintptr_t)(PTR) & 0xF)
+#else
+#define CF_IS_TAGGED_OBJ(PTR)	0
+#define CF_TAGGED_OBJ_TYPE(PTR)	0
+#endif
+
+enum {
+    kCFTaggedObjectID_Invalid = 0,
+    kCFTaggedObjectID_Atom = (0 << 1) + 1,
+    kCFTaggedObjectID_Undefined3 = (1 << 1) + 1,
+    kCFTaggedObjectID_Undefined2 = (2 << 1) + 1,
+    kCFTaggedObjectID_Integer = (3 << 1) + 1,
+    kCFTaggedObjectID_DateTS = (4 << 1) + 1,
+    kCFTaggedObjectID_ManagedObjectID = (5 << 1) + 1, // Core Data
+    kCFTaggedObjectID_Date = (6 << 1) + 1,
+    kCFTaggedObjectID_Undefined7 = (7 << 1) + 1,
+};
+*/

src/cmd/dist/test.go

@@ -915,7 +915,7 @@ func (t *tester) cgoTest(dt *distTest) error {
 	t.addCmd(dt, "misc/cgo/test", "go", "test", t.tags(), "-ldflags", "-linkmode=auto", t.runFlag(""))
 
 	if t.internalLink() {
-		t.addCmd(dt, "misc/cgo/test", "go", "test", "-ldflags", "-linkmode=internal", t.runFlag(""))
+		t.addCmd(dt, "misc/cgo/test", "go", "test", "-tags", "internal", "-ldflags", "-linkmode=internal", t.runFlag(""))
 	}
 
 	pair := gohostos + "-" + goarch

src/cmd/fix/cftype.go

@@ -0,0 +1,93 @@
+// Copyright 2017 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 main
+
+import (
+	"go/ast"
+	"go/token"
+	"reflect"
+	"strings"
+)
+
+func init() {
+	register(cftypeFix)
+}
+
+var cftypeFix = fix{
+	name:     "cftype",
+	date:     "2017-09-27",
+	f:        cftypefix,
+	desc:     `Fixes initializers of C.CF*Ptr types`,
+	disabled: false,
+}
+
+// Old state:
+//   type CFTypeRef unsafe.Pointer
+// New state:
+//   type CFTypeRef uintptr
+// and similar for other CF*Ref types.
+// This fix finds nils initializing these types and replaces the nils with 0s.
+func cftypefix(f *ast.File) bool {
+	if !imports(f, "C") {
+		return false
+	}
+	typeof, _ := typecheck(&TypeConfig{}, f)
+
+	// step 1: Find all the nils with the offending types.
+	// Compute their replacement.
+	badNils := map[interface{}]ast.Expr{}
+	walk(f, func(n interface{}) {
+		if i, ok := n.(*ast.Ident); ok && i.Name == "nil" && badPointerType(typeof[n]) {
+			badNils[n] = &ast.BasicLit{ValuePos: i.NamePos, Kind: token.INT, Value: "0"}
+		}
+	})
+	if len(badNils) == 0 {
+		return false
+	}
+
+	// step 2: find all uses of the bad nils, replace them with 0.
+	// There's no easy way to map from an ast.Expr to all the places that use them, so
+	// we use reflect to find all such references.
+	exprType := reflect.TypeOf((*ast.Expr)(nil)).Elem()
+	exprSliceType := reflect.TypeOf(([]ast.Expr)(nil))
+	walk(f, func(n interface{}) {
+		if n == nil {
+			return
+		}
+		v := reflect.ValueOf(n)
+		if v.Type().Kind() != reflect.Ptr {
+			return
+		}
+		if v.IsNil() {
+			return
+		}
+		v = v.Elem()
+		if v.Type().Kind() != reflect.Struct {
+			return
+		}
+		for i := 0; i < v.NumField(); i++ {
+			f := v.Field(i)
+			if f.Type() == exprType {
+				if r := badNils[f.Interface()]; r != nil {
+					f.Set(reflect.ValueOf(r))
+				}
+			}
+			if f.Type() == exprSliceType {
+				for j := 0; j < f.Len(); j++ {
+					e := f.Index(j)
+					if r := badNils[e.Interface()]; r != nil {
+						e.Set(reflect.ValueOf(r))
+					}
+				}
+			}
+		}
+	})
+
+	return true
+}
+
+func badPointerType(s string) bool {
+	return strings.HasPrefix(s, "C.CF") && strings.HasSuffix(s, "Ref")
+}