runtime, cmd/compile: rename memclr -> memclrNoHeapPointers

Since barrier-less memclr is only safe in very narrow circumstances,
this commit renames memclr to avoid accidentally calling memclr on
typed memory. This can cause subtle, non-deterministic bugs, so it's
worth some effort to prevent. In the near term, this will also prevent
bugs creeping in from any concurrent CLs that add calls to memclr; if
this happens, whichever patch hits master second will fail to compile.

This also adds the other new memclr variants to the compiler's
builtin.go to minimize the churn on that binary blob. We'll use these
in future commits.

Updates #17503.

Change-Id: I00eead049f5bd35ca107ea525966831f3d1ed9ca
Reviewed-on: https://go-review.googlesource.com/31369
Reviewed-by: Keith Randall <khr@golang.org>
Reviewed-by: Rick Hudson <rlh@golang.org>

src/cmd/compile/internal/gc/builtin.go

@@ -87,51 +87,53 @@ var runtimeDecls = [...]struct {
 	{"writeBarrier", varTag, 76},
 	{"writebarrierptr", funcTag, 77},
 	{"typedmemmove", funcTag, 78},
-	{"typedslicecopy", funcTag, 79},
-	{"selectnbsend", funcTag, 80},
-	{"selectnbrecv", funcTag, 81},
-	{"selectnbrecv2", funcTag, 83},
-	{"newselect", funcTag, 84},
-	{"selectsend", funcTag, 80},
+	{"typedmemclr", funcTag, 79},
+	{"typedslicecopy", funcTag, 80},
+	{"selectnbsend", funcTag, 81},
+	{"selectnbrecv", funcTag, 82},
+	{"selectnbrecv2", funcTag, 84},
+	{"newselect", funcTag, 85},
+	{"selectsend", funcTag, 81},
 	{"selectrecv", funcTag, 73},
-	{"selectrecv2", funcTag, 85},
-	{"selectdefault", funcTag, 86},
-	{"selectgo", funcTag, 87},
+	{"selectrecv2", funcTag, 86},
+	{"selectdefault", funcTag, 87},
+	{"selectgo", funcTag, 88},
 	{"block", funcTag, 5},
-	{"makeslice", funcTag, 89},
-	{"makeslice64", funcTag, 90},
-	{"growslice", funcTag, 91},
-	{"memmove", funcTag, 92},
-	{"memclr", funcTag, 93},
-	{"memequal", funcTag, 94},
-	{"memequal8", funcTag, 95},
-	{"memequal16", funcTag, 95},
-	{"memequal32", funcTag, 95},
-	{"memequal64", funcTag, 95},
-	{"memequal128", funcTag, 95},
-	{"int64div", funcTag, 96},
-	{"uint64div", funcTag, 97},
-	{"int64mod", funcTag, 96},
-	{"uint64mod", funcTag, 97},
-	{"float64toint64", funcTag, 98},
-	{"float64touint64", funcTag, 99},
-	{"float64touint32", funcTag, 101},
-	{"int64tofloat64", funcTag, 102},
-	{"uint64tofloat64", funcTag, 103},
-	{"uint32tofloat64", funcTag, 104},
-	{"complex128div", funcTag, 105},
-	{"racefuncenter", funcTag, 106},
+	{"makeslice", funcTag, 90},
+	{"makeslice64", funcTag, 91},
+	{"growslice", funcTag, 92},
+	{"memmove", funcTag, 93},
+	{"memclrNoHeapPointers", funcTag, 94},
+	{"memclrHasPointers", funcTag, 94},
+	{"memequal", funcTag, 95},
+	{"memequal8", funcTag, 96},
+	{"memequal16", funcTag, 96},
+	{"memequal32", funcTag, 96},
+	{"memequal64", funcTag, 96},
+	{"memequal128", funcTag, 96},
+	{"int64div", funcTag, 97},
+	{"uint64div", funcTag, 98},
+	{"int64mod", funcTag, 97},
+	{"uint64mod", funcTag, 98},
+	{"float64toint64", funcTag, 99},
+	{"float64touint64", funcTag, 100},
+	{"float64touint32", funcTag, 102},
+	{"int64tofloat64", funcTag, 103},
+	{"uint64tofloat64", funcTag, 104},
+	{"uint32tofloat64", funcTag, 105},
+	{"complex128div", funcTag, 106},
+	{"racefuncenter", funcTag, 107},
 	{"racefuncexit", funcTag, 5},
-	{"raceread", funcTag, 106},
-	{"racewrite", funcTag, 106},
-	{"racereadrange", funcTag, 107},
-	{"racewriterange", funcTag, 107},
-	{"msanread", funcTag, 107},
-	{"msanwrite", funcTag, 107},
+	{"raceread", funcTag, 107},
+	{"racewrite", funcTag, 107},
+	{"racereadrange", funcTag, 108},
+	{"racewriterange", funcTag, 108},
+	{"msanread", funcTag, 108},
+	{"msanwrite", funcTag, 108},
 }
 
 func runtimeTypes() []*Type {
-	var typs [108]*Type
+	var typs [109]*Type
 	typs[0] = bytetype
 	typs[1] = typPtr(typs[0])
 	typs[2] = Types[TANY]
@@ -211,34 +213,35 @@ func runtimeTypes() []*Type {
 	typs[76] = tostruct([]*Node{namedfield("enabled", typs[13]), namedfield("needed", typs[13]), namedfield("cgo", typs[13])})
 	typs[77] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[2])}, nil)
 	typs[78] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3]), anonfield(typs[3])}, nil)
-	typs[79] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[2]), anonfield(typs[2])}, []*Node{anonfield(typs[33])})
-	typs[80] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[74]), anonfield(typs[3])}, []*Node{anonfield(typs[13])})
-	typs[81] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3]), anonfield(typs[71])}, []*Node{anonfield(typs[13])})
-	typs[82] = typPtr(typs[13])
-	typs[83] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3]), anonfield(typs[82]), anonfield(typs[71])}, []*Node{anonfield(typs[13])})
-	typs[84] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[17]), anonfield(typs[10])}, nil)
-	typs[85] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[71]), anonfield(typs[3]), anonfield(typs[82])}, []*Node{anonfield(typs[13])})
-	typs[86] = functype(nil, []*Node{anonfield(typs[1])}, []*Node{anonfield(typs[13])})
-	typs[87] = functype(nil, []*Node{anonfield(typs[1])}, nil)
-	typs[88] = typSlice(typs[2])
-	typs[89] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[33]), anonfield(typs[33])}, []*Node{anonfield(typs[88])})
-	typs[90] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[17]), anonfield(typs[17])}, []*Node{anonfield(typs[88])})
-	typs[91] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[88]), anonfield(typs[33])}, []*Node{anonfield(typs[88])})
-	typs[92] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[50])}, nil)
-	typs[93] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[50])}, nil)
-	typs[94] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[50])}, []*Node{anonfield(typs[13])})
-	typs[95] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3])}, []*Node{anonfield(typs[13])})
-	typs[96] = functype(nil, []*Node{anonfield(typs[17]), anonfield(typs[17])}, []*Node{anonfield(typs[17])})
-	typs[97] = functype(nil, []*Node{anonfield(typs[19]), anonfield(typs[19])}, []*Node{anonfield(typs[19])})
-	typs[98] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[17])})
-	typs[99] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[19])})
-	typs[100] = Types[TUINT32]
-	typs[101] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[100])})
-	typs[102] = functype(nil, []*Node{anonfield(typs[17])}, []*Node{anonfield(typs[15])})
-	typs[103] = functype(nil, []*Node{anonfield(typs[19])}, []*Node{anonfield(typs[15])})
-	typs[104] = functype(nil, []*Node{anonfield(typs[100])}, []*Node{anonfield(typs[15])})
-	typs[105] = functype(nil, []*Node{anonfield(typs[21]), anonfield(typs[21])}, []*Node{anonfield(typs[21])})
-	typs[106] = functype(nil, []*Node{anonfield(typs[50])}, nil)
-	typs[107] = functype(nil, []*Node{anonfield(typs[50]), anonfield(typs[50])}, nil)
+	typs[79] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3])}, nil)
+	typs[80] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[2]), anonfield(typs[2])}, []*Node{anonfield(typs[33])})
+	typs[81] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[74]), anonfield(typs[3])}, []*Node{anonfield(typs[13])})
+	typs[82] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3]), anonfield(typs[71])}, []*Node{anonfield(typs[13])})
+	typs[83] = typPtr(typs[13])
+	typs[84] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3]), anonfield(typs[83]), anonfield(typs[71])}, []*Node{anonfield(typs[13])})
+	typs[85] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[17]), anonfield(typs[10])}, nil)
+	typs[86] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[71]), anonfield(typs[3]), anonfield(typs[83])}, []*Node{anonfield(typs[13])})
+	typs[87] = functype(nil, []*Node{anonfield(typs[1])}, []*Node{anonfield(typs[13])})
+	typs[88] = functype(nil, []*Node{anonfield(typs[1])}, nil)
+	typs[89] = typSlice(typs[2])
+	typs[90] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[33]), anonfield(typs[33])}, []*Node{anonfield(typs[89])})
+	typs[91] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[17]), anonfield(typs[17])}, []*Node{anonfield(typs[89])})
+	typs[92] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[89]), anonfield(typs[33])}, []*Node{anonfield(typs[89])})
+	typs[93] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[50])}, nil)
+	typs[94] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[50])}, nil)
+	typs[95] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[50])}, []*Node{anonfield(typs[13])})
+	typs[96] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3])}, []*Node{anonfield(typs[13])})
+	typs[97] = functype(nil, []*Node{anonfield(typs[17]), anonfield(typs[17])}, []*Node{anonfield(typs[17])})
+	typs[98] = functype(nil, []*Node{anonfield(typs[19]), anonfield(typs[19])}, []*Node{anonfield(typs[19])})
+	typs[99] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[17])})
+	typs[100] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[19])})
+	typs[101] = Types[TUINT32]
+	typs[102] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[101])})
+	typs[103] = functype(nil, []*Node{anonfield(typs[17])}, []*Node{anonfield(typs[15])})
+	typs[104] = functype(nil, []*Node{anonfield(typs[19])}, []*Node{anonfield(typs[15])})
+	typs[105] = functype(nil, []*Node{anonfield(typs[101])}, []*Node{anonfield(typs[15])})
+	typs[106] = functype(nil, []*Node{anonfield(typs[21]), anonfield(typs[21])}, []*Node{anonfield(typs[21])})
+	typs[107] = functype(nil, []*Node{anonfield(typs[50])}, nil)
+	typs[108] = functype(nil, []*Node{anonfield(typs[50]), anonfield(typs[50])}, nil)
 	return typs[:]
 }

src/cmd/compile/internal/gc/builtin/runtime.go

@@ -113,6 +113,7 @@ func writebarrierptr(dst *any, src any)
 
 // *byte is really *runtime.Type
 func typedmemmove(typ *byte, dst *any, src *any)
+func typedmemclr(typ *byte, dst *any)
 func typedslicecopy(typ *byte, dst any, src any) int
 
 func selectnbsend(chanType *byte, hchan chan<- any, elem *any) bool
@@ -131,7 +132,8 @@ func makeslice(typ *byte, len int, cap int) (ary []any)
 func makeslice64(typ *byte, len int64, cap int64) (ary []any)
 func growslice(typ *byte, old []any, cap int) (ary []any)
 func memmove(to *any, frm *any, length uintptr)
-func memclr(ptr *byte, length uintptr)
+func memclrNoHeapPointers(ptr *byte, length uintptr)
+func memclrHasPointers(ptr *byte, length uintptr)
 
 func memequal(x, y *any, size uintptr) bool
 func memequal8(x, y *any) bool

src/cmd/compile/internal/gc/range.go

@@ -393,11 +393,13 @@ func memclrrange(n, v1, v2, a *Node) bool {
 		return false
 	}
 
+	// TODO: Use memclrHasPointers if there are pointers.
+
 	// Convert to
 	// if len(a) != 0 {
 	// 	hp = &a[0]
 	// 	hn = len(a)*sizeof(elem(a))
-	// 	memclr(hp, hn)
+	// 	memclrNoHeapPointers(hp, hn)
 	// 	i = len(a) - 1
 	// }
 	n.Op = OIF
@@ -423,8 +425,8 @@ func memclrrange(n, v1, v2, a *Node) bool {
 	tmp = conv(tmp, Types[TUINTPTR])
 	n.Nbody.Append(nod(OAS, hn, tmp))
 
-	// memclr(hp, hn)
-	fn := mkcall("memclr", nil, nil, hp, hn)
+	// memclrNoHeapPointers(hp, hn)
+	fn := mkcall("memclrNoHeapPointers", nil, nil, hp, hn)
 
 	n.Nbody.Append(fn)
 

src/reflect/value.go

@@ -442,13 +442,13 @@ func (v Value) call(op string, in []Value) []Value {
 	if nout == 0 {
 		// This is untyped because the frame is really a
 		// stack, even though it's a heap object.
-		memclr(args, frametype.size)
+		memclrNoHeapPointers(args, frametype.size)
 		framePool.Put(args)
 	} else {
 		// Zero the now unused input area of args,
 		// because the Values returned by this function contain pointers to the args object,
 		// and will thus keep the args object alive indefinitely.
-		memclr(args, retOffset)
+		memclrNoHeapPointers(args, retOffset)
 		// Wrap Values around return values in args.
 		ret = make([]Value, nout)
 		off = retOffset
@@ -648,7 +648,7 @@ func callMethod(ctxt *methodValue, frame unsafe.Pointer) {
 
 	// This is untyped because the frame is really a stack, even
 	// though it's a heap object.
-	memclr(args, frametype.size)
+	memclrNoHeapPointers(args, frametype.size)
 	framePool.Put(args)
 }
 
@@ -2512,7 +2512,7 @@ func typedmemmovepartial(t *rtype, dst, src unsafe.Pointer, off, size uintptr)
 func typedslicecopy(elemType *rtype, dst, src sliceHeader) int
 
 //go:noescape
-func memclr(ptr unsafe.Pointer, n uintptr)
+func memclrNoHeapPointers(ptr unsafe.Pointer, n uintptr)
 
 // Dummy annotation marking that the value x escapes,
 // for use in cases where the reflect code is so clever that

src/runtime/alg.go

@@ -275,12 +275,6 @@ func ifaceHash(i interface {
 	return algarray[alg_INTER].hash(noescape(unsafe.Pointer(&i)), seed)
 }
 
-// Testing adapter for memclr
-func memclrBytes(b []byte) {
-	s := (*slice)(unsafe.Pointer(&b))
-	memclr(s.array, uintptr(s.len))
-}
-
 const hashRandomBytes = sys.PtrSize / 4 * 64
 
 // used in asm_{386,amd64}.s to seed the hash function

src/runtime/asm_amd64p32.s

@@ -484,7 +484,7 @@ TEXT runtime·stackcheck(SB), NOSPLIT, $0-0
 	MOVL	0, AX
 	RET
 
-TEXT runtime·memclr(SB),NOSPLIT,$0-8
+TEXT runtime·memclrNoHeapPointers(SB),NOSPLIT,$0-8
 	MOVL	ptr+0(FP), DI
 	MOVL	n+4(FP), CX
 	MOVQ	CX, BX

src/runtime/export_test.go

@@ -155,7 +155,11 @@ var Int32Hash = int32Hash
 var Int64Hash = int64Hash
 var EfaceHash = efaceHash
 var IfaceHash = ifaceHash
-var MemclrBytes = memclrBytes
+
+func MemclrBytes(b []byte) {
+	s := (*slice)(unsafe.Pointer(&b))
+	memclrNoHeapPointers(s.array, uintptr(s.len))
+}
 
 var HashLoad = &hashLoad
 

src/runtime/hashmap.go

@@ -1090,7 +1090,7 @@ func evacuate(t *maptype, h *hmap, oldbucket uintptr) {
 			if t.bucket.kind&kindNoPointers == 0 {
 				memclrHasPointers(add(unsafe.Pointer(b), dataOffset), uintptr(t.bucketsize)-dataOffset)
 			} else {
-				memclr(add(unsafe.Pointer(b), dataOffset), uintptr(t.bucketsize)-dataOffset)
+				memclrNoHeapPointers(add(unsafe.Pointer(b), dataOffset), uintptr(t.bucketsize)-dataOffset)
 			}
 		}
 	}

src/runtime/heapdump.go

@@ -631,7 +631,7 @@ func mdump() {
 			s.ensureSwept()
 		}
 	}
-	memclr(unsafe.Pointer(&typecache), unsafe.Sizeof(typecache))
+	memclrNoHeapPointers(unsafe.Pointer(&typecache), unsafe.Sizeof(typecache))
 	dwrite(unsafe.Pointer(&dumphdr[0]), uintptr(len(dumphdr)))
 	dumpparams()
 	dumpitabs()

src/runtime/malloc.go

@@ -682,7 +682,7 @@ func mallocgc(size uintptr, typ *_type, needzero bool) unsafe.Pointer {
 			}
 			x = unsafe.Pointer(v)
 			if needzero && span.needzero != 0 {
-				memclr(unsafe.Pointer(v), size)
+				memclrNoHeapPointers(unsafe.Pointer(v), size)
 			}
 		}
 	} else {

src/runtime/mbarrier.go

@@ -333,14 +333,17 @@ func reflect_typedslicecopy(elemType *_type, dst, src slice) int {
 }
 
 // typedmemclr clears the typed memory at ptr with type typ. The
-// memory at ptr must already be type-safe.
+// memory at ptr must already be initialized (and hence in type-safe
+// state). If the memory is being initialized for the first time, see
+// memclrNoHeapPointers.
 //
 // If the caller knows that typ has pointers, it can alternatively
 // call memclrHasPointers.
 //
 //go:nosplit
 func typedmemclr(typ *_type, ptr unsafe.Pointer) {
-	memclr(ptr, typ.size)
+	// TODO(austin): Call the hybrid barrier.
+	memclrNoHeapPointers(ptr, typ.size)
 }
 
 // memclrHasPointers clears n bytes of typed memory starting at ptr.
@@ -350,5 +353,6 @@ func typedmemclr(typ *_type, ptr unsafe.Pointer) {
 //
 //go:nosplit
 func memclrHasPointers(ptr unsafe.Pointer, n uintptr) {
-	memclr(ptr, n)
+	// TODO(austin): Call the hybrid barrier.
+	memclrNoHeapPointers(ptr, n)
 }

src/runtime/mbitmap.go

@@ -742,7 +742,7 @@ func (h heapBits) initSpan(s *mspan) {
 		}
 		return
 	}
-	memclr(unsafe.Pointer(subtractb(h.bitp, nbyte-1)), nbyte)
+	memclrNoHeapPointers(unsafe.Pointer(subtractb(h.bitp, nbyte-1)), nbyte)
 }
 
 // initCheckmarkSpan initializes a span for being checkmarked.
@@ -1433,7 +1433,7 @@ func heapBitsSetTypeGCProg(h heapBits, progSize, elemSize, dataSize, allocSize u
 	}
 	endProg := unsafe.Pointer(subtractb(h.bitp, (totalBits+3)/4))
 	endAlloc := unsafe.Pointer(subtractb(h.bitp, allocSize/heapBitmapScale))
-	memclr(add(endAlloc, 1), uintptr(endProg)-uintptr(endAlloc))
+	memclrNoHeapPointers(add(endAlloc, 1), uintptr(endProg)-uintptr(endAlloc))
 }
 
 // progToPointerMask returns the 1-bit pointer mask output by the GC program prog.

src/runtime/mem_plan9.go

@@ -38,7 +38,7 @@ func memAlloc(n uintptr) unsafe.Pointer {
 				p.size -= n
 				p = (*memHdr)(add(unsafe.Pointer(p), p.size))
 			}
-			memclr(unsafe.Pointer(p), unsafe.Sizeof(memHdr{}))
+			*p = memHdr{}
 			return unsafe.Pointer(p)
 		}
 		prevp = p
@@ -48,7 +48,7 @@ func memAlloc(n uintptr) unsafe.Pointer {
 
 func memFree(ap unsafe.Pointer, n uintptr) {
 	n = memRound(n)
-	memclr(ap, n)
+	memclrNoHeapPointers(ap, n)
 	bp := (*memHdr)(ap)
 	bp.size = n
 	bpn := uintptr(ap)
@@ -63,7 +63,7 @@ func memFree(ap unsafe.Pointer, n uintptr) {
 		if bpn+bp.size == uintptr(unsafe.Pointer(p)) {
 			bp.size += p.size
 			bp.next = p.next
-			memclr(unsafe.Pointer(p), unsafe.Sizeof(memHdr{}))
+			*p = memHdr{}
 		} else {
 			bp.next.set(p)
 		}
@@ -77,14 +77,14 @@ func memFree(ap unsafe.Pointer, n uintptr) {
 	if bpn+bp.size == uintptr(unsafe.Pointer(p.next)) {
 		bp.size += p.next.ptr().size
 		bp.next = p.next.ptr().next
-		memclr(unsafe.Pointer(p.next), unsafe.Sizeof(memHdr{}))
+		*p.next.ptr() = memHdr{}
 	} else {
 		bp.next = p.next
 	}
 	if uintptr(unsafe.Pointer(p))+p.size == bpn {
 		p.size += bp.size
 		p.next = bp.next
-		memclr(unsafe.Pointer(bp), unsafe.Sizeof(memHdr{}))
+		*bp = memHdr{}
 	} else {
 		p.next.set(bp)
 	}

src/runtime/memclr_386.s

@@ -8,8 +8,8 @@
 
 // NOTE: Windows externalthreadhandler expects memclr to preserve DX.
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB), NOSPLIT, $0-8
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB), NOSPLIT, $0-8
 	MOVL	ptr+0(FP), DI
 	MOVL	n+4(FP), BX
 	XORL	AX, AX

src/runtime/memclr_amd64.s

@@ -8,8 +8,8 @@
 
 // NOTE: Windows externalthreadhandler expects memclr to preserve DX.
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB), NOSPLIT, $0-16
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB), NOSPLIT, $0-16
 	MOVQ	ptr+0(FP), DI
 	MOVQ	n+8(FP), BX
 	XORQ	AX, AX

src/runtime/memclr_arm.s

@@ -30,7 +30,7 @@
 #define N	R12
 #define TMP	R12				/* N and TMP don't overlap */
 
-TEXT runtime·memclr(SB),NOSPLIT,$0-8
+TEXT runtime·memclrNoHeapPointers(SB),NOSPLIT,$0-8
 	MOVW	ptr+0(FP), TO
 	MOVW	n+4(FP), N
 	MOVW	$0, R0

src/runtime/memclr_arm64.s

@@ -4,8 +4,8 @@
 
 #include "textflag.h"
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB),NOSPLIT,$0-16
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB),NOSPLIT,$0-16
 	MOVD	ptr+0(FP), R3
 	MOVD	n+8(FP), R4
 	// TODO(mwhudson): this is written this way to avoid tickling

src/runtime/memclr_mips64x.s

@@ -6,8 +6,8 @@
 
 #include "textflag.h"
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB),NOSPLIT,$0-16
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB),NOSPLIT,$0-16
 	MOVV	ptr+0(FP), R1
 	MOVV	n+8(FP), R2
 	ADDV	R1, R2, R4

src/runtime/memclr_plan9_386.s

@@ -4,8 +4,8 @@
 
 #include "textflag.h"
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB), NOSPLIT, $0-8
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB), NOSPLIT, $0-8
 	MOVL	ptr+0(FP), DI
 	MOVL	n+4(FP), BX
 	XORL	AX, AX

src/runtime/memclr_plan9_amd64.s

@@ -4,8 +4,8 @@
 
 #include "textflag.h"
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB),NOSPLIT,$0-16
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB),NOSPLIT,$0-16
 	MOVQ	ptr+0(FP), DI
 	MOVQ	n+8(FP), CX
 	MOVQ	CX, BX

src/runtime/memclr_ppc64x.s

@@ -6,8 +6,8 @@
 
 #include "textflag.h"
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB), NOSPLIT|NOFRAME, $0-16
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB), NOSPLIT|NOFRAME, $0-16
 	MOVD ptr+0(FP), R3
 	MOVD n+8(FP), R4
 

src/runtime/memclr_s390x.s

@@ -4,8 +4,8 @@
 
 #include "textflag.h"
 
-// void runtime·memclr(void*, uintptr)
-TEXT runtime·memclr(SB),NOSPLIT|NOFRAME,$0-16
+// void runtime·memclrNoHeapPointers(void*, uintptr)
+TEXT runtime·memclrNoHeapPointers(SB),NOSPLIT|NOFRAME,$0-16
 	MOVD	ptr+0(FP), R4
 	MOVD	n+8(FP), R5
 

src/runtime/mfixalloc.go

@@ -72,7 +72,7 @@ func (f *fixalloc) alloc() unsafe.Pointer {
 		f.list = f.list.next
 		f.inuse += f.size
 		if f.zero {
-			memclr(v, f.size)
+			memclrNoHeapPointers(v, f.size)
 		}
 		return v
 	}

src/runtime/mheap.go

@@ -629,7 +629,7 @@ func (h *mheap) alloc(npage uintptr, sizeclass int32, large bool, needzero bool)
 
 	if s != nil {
 		if needzero && s.needzero != 0 {
-			memclr(unsafe.Pointer(s.base()), s.npages<<_PageShift)
+			memclrNoHeapPointers(unsafe.Pointer(s.base()), s.npages<<_PageShift)
 		}
 		s.needzero = 0
 	}
@@ -1418,7 +1418,7 @@ func newArena() *gcBits {
 	} else {
 		result = gcBitsArenas.free
 		gcBitsArenas.free = gcBitsArenas.free.next
-		memclr(unsafe.Pointer(result), gcBitsChunkBytes)
+		memclrNoHeapPointers(unsafe.Pointer(result), gcBitsChunkBytes)
 	}
 	result.next = nil
 	// If result.bits is not 8 byte aligned adjust index so

src/runtime/os3_solaris.go

@@ -317,7 +317,7 @@ func semacreate(mp *m) {
 	// here because it could cause a deadlock.
 	_g_.m.libcall.fn = uintptr(unsafe.Pointer(&libc_malloc))
 	_g_.m.libcall.n = 1
-	memclr(unsafe.Pointer(&_g_.m.scratch), uintptr(len(_g_.m.scratch.v)))
+	_g_.m.scratch = mscratch{}
 	_g_.m.scratch.v[0] = unsafe.Sizeof(*sem)
 	_g_.m.libcall.args = uintptr(unsafe.Pointer(&_g_.m.scratch))
 	asmcgocall(unsafe.Pointer(&asmsysvicall6), unsafe.Pointer(&_g_.m.libcall))
@@ -337,7 +337,7 @@ func semasleep(ns int64) int32 {
 
 		_m_.libcall.fn = uintptr(unsafe.Pointer(&libc_sem_reltimedwait_np))
 		_m_.libcall.n = 2
-		memclr(unsafe.Pointer(&_m_.scratch), uintptr(len(_m_.scratch.v)))
+		_m_.scratch = mscratch{}
 		_m_.scratch.v[0] = _m_.waitsema
 		_m_.scratch.v[1] = uintptr(unsafe.Pointer(&_m_.ts))
 		_m_.libcall.args = uintptr(unsafe.Pointer(&_m_.scratch))
@@ -353,7 +353,7 @@ func semasleep(ns int64) int32 {
 	for {
 		_m_.libcall.fn = uintptr(unsafe.Pointer(&libc_sem_wait))
 		_m_.libcall.n = 1
-		memclr(unsafe.Pointer(&_m_.scratch), uintptr(len(_m_.scratch.v)))
+		_m_.scratch = mscratch{}
 		_m_.scratch.v[0] = _m_.waitsema
 		_m_.libcall.args = uintptr(unsafe.Pointer(&_m_.scratch))
 		asmcgocall(unsafe.Pointer(&asmsysvicall6), unsafe.Pointer(&_m_.libcall))

src/runtime/proc.go

@@ -2812,7 +2812,7 @@ func newproc1(fn *funcval, argp *uint8, narg int32, nret int32, callerpc uintptr
 	}
 	memmove(unsafe.Pointer(spArg), unsafe.Pointer(argp), uintptr(narg))
 
-	memclr(unsafe.Pointer(&newg.sched), unsafe.Sizeof(newg.sched))
+	memclrNoHeapPointers(unsafe.Pointer(&newg.sched), unsafe.Sizeof(newg.sched))
 	newg.sched.sp = sp
 	newg.stktopsp = sp
 	newg.sched.pc = funcPC(goexit) + sys.PCQuantum // +PCQuantum so that previous instruction is in same function

src/runtime/slice.go

@@ -141,7 +141,7 @@ func growslice(et *_type, old slice, cap int) slice {
 		memmove(p, old.array, lenmem)
 		// The append() that calls growslice is going to overwrite from old.len to cap (which will be the new length).
 		// Only clear the part that will not be overwritten.
-		memclr(add(p, newlenmem), capmem-newlenmem)
+		memclrNoHeapPointers(add(p, newlenmem), capmem-newlenmem)
 	} else {
 		// Note: can't use rawmem (which avoids zeroing of memory), because then GC can scan uninitialized memory.
 		p = mallocgc(capmem, et, true)

src/runtime/stack.go

@@ -436,7 +436,7 @@ func stackfree(stk stack, n uintptr) {
 	}
 	if stackDebug >= 1 {
 		println("stackfree", v, n)
-		memclr(v, n) // for testing, clobber stack data
+		memclrNoHeapPointers(v, n) // for testing, clobber stack data
 	}
 	if debug.efence != 0 || stackFromSystem != 0 {
 		if debug.efence != 0 || stackFaultOnFree != 0 {

src/runtime/string.go

@@ -249,7 +249,7 @@ func rawbyteslice(size int) (b []byte) {
 	cap := roundupsize(uintptr(size))
 	p := mallocgc(cap, nil, false)
 	if cap != uintptr(size) {
-		memclr(add(p, uintptr(size)), cap-uintptr(size))
+		memclrNoHeapPointers(add(p, uintptr(size)), cap-uintptr(size))
 	}
 
 	*(*slice)(unsafe.Pointer(&b)) = slice{p, size, int(cap)}
@@ -264,7 +264,7 @@ func rawruneslice(size int) (b []rune) {
 	mem := roundupsize(uintptr(size) * 4)
 	p := mallocgc(mem, nil, false)
 	if mem != uintptr(size)*4 {
-		memclr(add(p, uintptr(size)*4), mem-uintptr(size)*4)
+		memclrNoHeapPointers(add(p, uintptr(size)*4), mem-uintptr(size)*4)
 	}
 
 	*(*slice)(unsafe.Pointer(&b)) = slice{p, size, int(mem / 4)}

src/runtime/stubs.go

@@ -60,20 +60,24 @@ func badsystemstack() {
 	throw("systemstack called from unexpected goroutine")
 }
 
-// memclr clears n bytes starting at ptr.
+// memclrNoHeapPointers clears n bytes starting at ptr.
 //
-// Usually you should use typedmemclr. memclr should be used only when
-// the caller knows that *ptr contains no heap pointers or to
-// initialize memory to a type-safe state when allocation reuses dead
-// memory.
+// Usually you should use typedmemclr. memclrNoHeapPointers should be
+// used only when the caller knows that *ptr contains no heap pointers
+// because either:
+//
+// 1. *ptr is initialized memory and its type is pointer-free.
+//
+// 2. *ptr is uninitialized memory (e.g., memory that's being reused
+//    for a new allocation) and hence contains only "junk".
 //
 // in memclr_*.s
 //go:noescape
-func memclr(ptr unsafe.Pointer, n uintptr)
+func memclrNoHeapPointers(ptr unsafe.Pointer, n uintptr)
 
-//go:linkname reflect_memclr reflect.memclr
-func reflect_memclr(ptr unsafe.Pointer, n uintptr) {
-	memclr(ptr, n)
+//go:linkname reflect_memclrNoHeapPointers reflect.memclrNoHeapPointers
+func reflect_memclrNoHeapPointers(ptr unsafe.Pointer, n uintptr) {
+	memclrNoHeapPointers(ptr, n)
 }
 
 // memmove copies n bytes from "from" to "to".

src/runtime/sys_windows_386.s

@@ -192,7 +192,7 @@ TEXT runtime·externalthreadhandler(SB),NOSPLIT,$0
 	SUBL	$m__size, SP		// space for M
 	MOVL	SP, 0(SP)
 	MOVL	$m__size, 4(SP)
-	CALL	runtime·memclr(SB)	// smashes AX,BX,CX
+	CALL	runtime·memclrNoHeapPointers(SB)	// smashes AX,BX,CX
 
 	LEAL	m_tls(SP), CX
 	MOVL	CX, 0x14(FS)
@@ -203,7 +203,7 @@ TEXT runtime·externalthreadhandler(SB),NOSPLIT,$0
 
 	MOVL	SP, 0(SP)
 	MOVL	$g__size, 4(SP)
-	CALL	runtime·memclr(SB)	// smashes AX,BX,CX
+	CALL	runtime·memclrNoHeapPointers(SB)	// smashes AX,BX,CX
 	LEAL	g__size(SP), BX
 	MOVL	BX, g_m(SP)
 

src/runtime/sys_windows_amd64.s

@@ -236,7 +236,7 @@ TEXT runtime·externalthreadhandler(SB),NOSPLIT|NOFRAME,$0
 	SUBQ	$m__size, SP		// space for M
 	MOVQ	SP, 0(SP)
 	MOVQ	$m__size, 8(SP)
-	CALL	runtime·memclr(SB)	// smashes AX,BX,CX, maybe BP
+	CALL	runtime·memclrNoHeapPointers(SB)	// smashes AX,BX,CX, maybe BP
 
 	LEAQ	m_tls(SP), CX
 	MOVQ	CX, 0x28(GS)
@@ -247,7 +247,7 @@ TEXT runtime·externalthreadhandler(SB),NOSPLIT|NOFRAME,$0
 
 	MOVQ	SP, 0(SP)
 	MOVQ	$g__size, 8(SP)
-	CALL	runtime·memclr(SB)	// smashes AX,BX,CX, maybe BP
+	CALL	runtime·memclrNoHeapPointers(SB)	// smashes AX,BX,CX, maybe BP
 	LEAQ	g__size(SP), BX
 	MOVQ	BX, g_m(SP)
 

src/runtime/write_err_android.go

@@ -75,7 +75,9 @@ func writeErr(b []byte) {
 		if v == '\n' || writePos == len(dst)-1 {
 			dst[writePos] = 0
 			write(writeFD, unsafe.Pointer(&writeBuf[0]), int32(hlen+writePos))
-			memclrBytes(dst)
+			for i := range dst {
+				dst[i] = 0
+			}
 			writePos = 0
 		}
 	}