runtime: remove legacy eager write barrier

Now that the buffered write barrier is implemented for all
architectures, we can remove the old eager write barrier
implementation. This CL removes the implementation from the runtime,
support in the compiler for calling it, and updates some compiler
tests that relied on the old eager barrier support. It also makes sure
that all of the useful comments from the old write barrier
implementation still have a place to live.

Fixes #22460.

Updates #21640 since this fixes the layering concerns of the write
barrier (but not the other things in that issue).

Change-Id: I580f93c152e89607e0a72fe43370237ba97bae74
Reviewed-on: https://go-review.googlesource.com/92705
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

24 files changed, 129 insertions, 251 deletions

diff --git a/src/cmd/compile/internal/gc/asm_test.go b/src/cmd/compile/internal/gc/asm_test.go
index 50857e6533..73c55018f9 100644
--- a/src/cmd/compile/internal/gc/asm_test.go
+++ b/src/cmd/compile/internal/gc/asm_test.go
@@ -468,7 +468,7 @@ var linuxAMD64Tests = []*asmTest{
 			*t = T2{}
 		}
 		`,
-		pos: []string{"\tXORPS\tX., X", "\tMOVUPS\tX., \\(.*\\)", "\tMOVQ\t\\$0, 16\\(.*\\)", "\tCALL\truntime\\.(writebarrierptr|gcWriteBarrier)\\(SB\\)"},
+		pos: []string{"\tXORPS\tX., X", "\tMOVUPS\tX., \\(.*\\)", "\tMOVQ\t\\$0, 16\\(.*\\)", "\tCALL\truntime\\.gcWriteBarrier\\(SB\\)"},
 	},
 	// Rotate tests
 	{
diff --git a/src/cmd/compile/internal/gc/builtin.go b/src/cmd/compile/internal/gc/builtin.go
index 0733a460d5..3bb17673a5 100644
--- a/src/cmd/compile/internal/gc/builtin.go
+++ b/src/cmd/compile/internal/gc/builtin.go
@@ -105,56 +105,55 @@ var runtimeDecls = [...]struct {
 	{"chansend1", funcTag, 81},
 	{"closechan", funcTag, 23},
 	{"writeBarrier", varTag, 83},
-	{"writebarrierptr", funcTag, 84},
-	{"typedmemmove", funcTag, 85},
-	{"typedmemclr", funcTag, 86},
-	{"typedslicecopy", funcTag, 87},
-	{"selectnbsend", funcTag, 88},
-	{"selectnbrecv", funcTag, 89},
-	{"selectnbrecv2", funcTag, 91},
-	{"newselect", funcTag, 92},
-	{"selectsend", funcTag, 93},
-	{"selectrecv", funcTag, 94},
+	{"typedmemmove", funcTag, 84},
+	{"typedmemclr", funcTag, 85},
+	{"typedslicecopy", funcTag, 86},
+	{"selectnbsend", funcTag, 87},
+	{"selectnbrecv", funcTag, 88},
+	{"selectnbrecv2", funcTag, 90},
+	{"newselect", funcTag, 91},
+	{"selectsend", funcTag, 92},
+	{"selectrecv", funcTag, 93},
 	{"selectdefault", funcTag, 55},
-	{"selectgo", funcTag, 95},
+	{"selectgo", funcTag, 94},
 	{"block", funcTag, 5},
-	{"makeslice", funcTag, 97},
-	{"makeslice64", funcTag, 98},
-	{"growslice", funcTag, 99},
-	{"memmove", funcTag, 100},
-	{"memclrNoHeapPointers", funcTag, 101},
-	{"memclrHasPointers", funcTag, 101},
-	{"memequal", funcTag, 102},
-	{"memequal8", funcTag, 103},
-	{"memequal16", funcTag, 103},
-	{"memequal32", funcTag, 103},
-	{"memequal64", funcTag, 103},
-	{"memequal128", funcTag, 103},
-	{"int64div", funcTag, 104},
-	{"uint64div", funcTag, 105},
-	{"int64mod", funcTag, 104},
-	{"uint64mod", funcTag, 105},
-	{"float64toint64", funcTag, 106},
-	{"float64touint64", funcTag, 107},
-	{"float64touint32", funcTag, 108},
-	{"int64tofloat64", funcTag, 109},
-	{"uint64tofloat64", funcTag, 110},
-	{"uint32tofloat64", funcTag, 111},
-	{"complex128div", funcTag, 112},
-	{"racefuncenter", funcTag, 113},
+	{"makeslice", funcTag, 96},
+	{"makeslice64", funcTag, 97},
+	{"growslice", funcTag, 98},
+	{"memmove", funcTag, 99},
+	{"memclrNoHeapPointers", funcTag, 100},
+	{"memclrHasPointers", funcTag, 100},
+	{"memequal", funcTag, 101},
+	{"memequal8", funcTag, 102},
+	{"memequal16", funcTag, 102},
+	{"memequal32", funcTag, 102},
+	{"memequal64", funcTag, 102},
+	{"memequal128", funcTag, 102},
+	{"int64div", funcTag, 103},
+	{"uint64div", funcTag, 104},
+	{"int64mod", funcTag, 103},
+	{"uint64mod", funcTag, 104},
+	{"float64toint64", funcTag, 105},
+	{"float64touint64", funcTag, 106},
+	{"float64touint32", funcTag, 107},
+	{"int64tofloat64", funcTag, 108},
+	{"uint64tofloat64", funcTag, 109},
+	{"uint32tofloat64", funcTag, 110},
+	{"complex128div", funcTag, 111},
+	{"racefuncenter", funcTag, 112},
 	{"racefuncexit", funcTag, 5},
-	{"raceread", funcTag, 113},
-	{"racewrite", funcTag, 113},
-	{"racereadrange", funcTag, 114},
-	{"racewriterange", funcTag, 114},
-	{"msanread", funcTag, 114},
-	{"msanwrite", funcTag, 114},
+	{"raceread", funcTag, 112},
+	{"racewrite", funcTag, 112},
+	{"racereadrange", funcTag, 113},
+	{"racewriterange", funcTag, 113},
+	{"msanread", funcTag, 113},
+	{"msanwrite", funcTag, 113},
 	{"support_popcnt", varTag, 11},
 	{"support_sse41", varTag, 11},
 }
 
 func runtimeTypes() []*types.Type {
-	var typs [115]*types.Type
+	var typs [114]*types.Type
 	typs[0] = types.Bytetype
 	typs[1] = types.NewPtr(typs[0])
 	typs[2] = types.Types[TANY]
@@ -239,36 +238,35 @@ func runtimeTypes() []*types.Type {
 	typs[81] = functype(nil, []*Node{anonfield(typs[80]), anonfield(typs[3])}, nil)
 	typs[82] = types.NewArray(typs[0], 3)
 	typs[83] = tostruct([]*Node{namedfield("enabled", typs[11]), namedfield("pad", typs[82]), namedfield("needed", typs[11]), namedfield("cgo", typs[11]), namedfield("alignme", typs[17])})
-	typs[84] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[2])}, nil)
-	typs[85] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3]), anonfield(typs[3])}, nil)
-	typs[86] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3])}, nil)
-	typs[87] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[2]), anonfield(typs[2])}, []*Node{anonfield(typs[32])})
-	typs[88] = functype(nil, []*Node{anonfield(typs[80]), anonfield(typs[3])}, []*Node{anonfield(typs[11])})
-	typs[89] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[77])}, []*Node{anonfield(typs[11])})
-	typs[90] = types.NewPtr(typs[11])
-	typs[91] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[90]), anonfield(typs[77])}, []*Node{anonfield(typs[11])})
-	typs[92] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[8])}, nil)
-	typs[93] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[80]), anonfield(typs[3])}, nil)
-	typs[94] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[77]), anonfield(typs[3]), anonfield(typs[90])}, nil)
-	typs[95] = functype(nil, []*Node{anonfield(typs[1])}, []*Node{anonfield(typs[32])})
-	typs[96] = types.NewSlice(typs[2])
-	typs[97] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[32]), anonfield(typs[32])}, []*Node{anonfield(typs[96])})
-	typs[98] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[15])}, []*Node{anonfield(typs[96])})
-	typs[99] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[96]), anonfield(typs[32])}, []*Node{anonfield(typs[96])})
-	typs[100] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[48])}, nil)
-	typs[101] = functype(nil, []*Node{anonfield(typs[57]), anonfield(typs[48])}, nil)
-	typs[102] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[48])}, []*Node{anonfield(typs[11])})
-	typs[103] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3])}, []*Node{anonfield(typs[11])})
-	typs[104] = functype(nil, []*Node{anonfield(typs[15]), anonfield(typs[15])}, []*Node{anonfield(typs[15])})
-	typs[105] = functype(nil, []*Node{anonfield(typs[17]), anonfield(typs[17])}, []*Node{anonfield(typs[17])})
-	typs[106] = functype(nil, []*Node{anonfield(typs[13])}, []*Node{anonfield(typs[15])})
-	typs[107] = functype(nil, []*Node{anonfield(typs[13])}, []*Node{anonfield(typs[17])})
-	typs[108] = functype(nil, []*Node{anonfield(typs[13])}, []*Node{anonfield(typs[59])})
-	typs[109] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[13])})
-	typs[110] = functype(nil, []*Node{anonfield(typs[17])}, []*Node{anonfield(typs[13])})
-	typs[111] = functype(nil, []*Node{anonfield(typs[59])}, []*Node{anonfield(typs[13])})
-	typs[112] = functype(nil, []*Node{anonfield(typs[19]), anonfield(typs[19])}, []*Node{anonfield(typs[19])})
-	typs[113] = functype(nil, []*Node{anonfield(typs[48])}, nil)
-	typs[114] = functype(nil, []*Node{anonfield(typs[48]), anonfield(typs[48])}, nil)
+	typs[84] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3]), anonfield(typs[3])}, nil)
+	typs[85] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[3])}, nil)
+	typs[86] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[2]), anonfield(typs[2])}, []*Node{anonfield(typs[32])})
+	typs[87] = functype(nil, []*Node{anonfield(typs[80]), anonfield(typs[3])}, []*Node{anonfield(typs[11])})
+	typs[88] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[77])}, []*Node{anonfield(typs[11])})
+	typs[89] = types.NewPtr(typs[11])
+	typs[90] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[89]), anonfield(typs[77])}, []*Node{anonfield(typs[11])})
+	typs[91] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[8])}, nil)
+	typs[92] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[80]), anonfield(typs[3])}, nil)
+	typs[93] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[77]), anonfield(typs[3]), anonfield(typs[89])}, nil)
+	typs[94] = functype(nil, []*Node{anonfield(typs[1])}, []*Node{anonfield(typs[32])})
+	typs[95] = types.NewSlice(typs[2])
+	typs[96] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[32]), anonfield(typs[32])}, []*Node{anonfield(typs[95])})
+	typs[97] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[15])}, []*Node{anonfield(typs[95])})
+	typs[98] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[95]), anonfield(typs[32])}, []*Node{anonfield(typs[95])})
+	typs[99] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[48])}, nil)
+	typs[100] = functype(nil, []*Node{anonfield(typs[57]), anonfield(typs[48])}, nil)
+	typs[101] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[48])}, []*Node{anonfield(typs[11])})
+	typs[102] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3])}, []*Node{anonfield(typs[11])})
+	typs[103] = functype(nil, []*Node{anonfield(typs[15]), anonfield(typs[15])}, []*Node{anonfield(typs[15])})
+	typs[104] = functype(nil, []*Node{anonfield(typs[17]), anonfield(typs[17])}, []*Node{anonfield(typs[17])})
+	typs[105] = functype(nil, []*Node{anonfield(typs[13])}, []*Node{anonfield(typs[15])})
+	typs[106] = functype(nil, []*Node{anonfield(typs[13])}, []*Node{anonfield(typs[17])})
+	typs[107] = functype(nil, []*Node{anonfield(typs[13])}, []*Node{anonfield(typs[59])})
+	typs[108] = functype(nil, []*Node{anonfield(typs[15])}, []*Node{anonfield(typs[13])})
+	typs[109] = functype(nil, []*Node{anonfield(typs[17])}, []*Node{anonfield(typs[13])})
+	typs[110] = functype(nil, []*Node{anonfield(typs[59])}, []*Node{anonfield(typs[13])})
+	typs[111] = functype(nil, []*Node{anonfield(typs[19]), anonfield(typs[19])}, []*Node{anonfield(typs[19])})
+	typs[112] = functype(nil, []*Node{anonfield(typs[48])}, nil)
+	typs[113] = functype(nil, []*Node{anonfield(typs[48]), anonfield(typs[48])}, nil)
 	return typs[:]
 }
diff --git a/src/cmd/compile/internal/gc/builtin/runtime.go b/src/cmd/compile/internal/gc/builtin/runtime.go
index de17d51d8a..bda9d1d03c 100644
--- a/src/cmd/compile/internal/gc/builtin/runtime.go
+++ b/src/cmd/compile/internal/gc/builtin/runtime.go
@@ -136,8 +136,6 @@ var writeBarrier struct {
 	alignme uint64
 }
 
-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)
diff --git a/src/cmd/compile/internal/gc/go.go b/src/cmd/compile/internal/gc/go.go
index dc94cf4f98..d6db7acc59 100644
--- a/src/cmd/compile/internal/gc/go.go
+++ b/src/cmd/compile/internal/gc/go.go
@@ -292,7 +292,6 @@ var (
 	assertI2I2,
 	goschedguarded,
 	writeBarrier,
-	writebarrierptr,
 	gcWriteBarrier,
 	typedmemmove,
 	typedmemclr,
diff --git a/src/cmd/compile/internal/gc/main.go b/src/cmd/compile/internal/gc/main.go
index 7f947530b1..f1591c16d2 100644
--- a/src/cmd/compile/internal/gc/main.go
+++ b/src/cmd/compile/internal/gc/main.go
@@ -44,7 +44,6 @@ var (
 	Debug_slice        int
 	Debug_vlog         bool
 	Debug_wb           int
-	Debug_eagerwb      int
 	Debug_pctab        string
 	Debug_locationlist int
 	Debug_typecheckinl int
@@ -73,7 +72,6 @@ var debugtab = []struct {
 	{"slice", "print information about slice compilation", &Debug_slice},
 	{"typeassert", "print information about type assertion inlining", &Debug_typeassert},
 	{"wb", "print information about write barriers", &Debug_wb},
-	{"eagerwb", "use unbuffered write barrier", &Debug_eagerwb},
 	{"export", "print export data", &Debug_export},
 	{"pctab", "print named pc-value table", &Debug_pctab},
 	{"locationlists", "print information about DWARF location list creation", &Debug_locationlist},
@@ -407,14 +405,6 @@ func Main(archInit func(*Arch)) {
 		Debug['l'] = 1 - Debug['l']
 	}
 
-	switch objabi.GOARCH {
-	case "amd64", "amd64p32", "386", "arm", "arm64", "ppc64", "ppc64le", "mips64", "mips64le", "mips", "mipsle", "s390x":
-	default:
-		// Other architectures don't support the buffered
-		// write barrier yet.
-		Debug_eagerwb = 1
-	}
-
 	trackScopes = flagDWARF && ((Debug['l'] == 0 && Debug['N'] != 0) || Ctxt.Flag_locationlists)
 
 	Widthptr = thearch.LinkArch.PtrSize
diff --git a/src/cmd/compile/internal/gc/ssa.go b/src/cmd/compile/internal/gc/ssa.go
index fe062da409..5ec01b6a61 100644
--- a/src/cmd/compile/internal/gc/ssa.go
+++ b/src/cmd/compile/internal/gc/ssa.go
@@ -95,7 +95,6 @@ func initssaconfig() {
 	assertI2I2 = sysfunc("assertI2I2")
 	goschedguarded = sysfunc("goschedguarded")
 	writeBarrier = sysfunc("writeBarrier")
-	writebarrierptr = sysfunc("writebarrierptr")
 	gcWriteBarrier = sysfunc("gcWriteBarrier")
 	typedmemmove = sysfunc("typedmemmove")
 	typedmemclr = sysfunc("typedmemclr")
@@ -5380,10 +5379,6 @@ func (e *ssafn) Debug_checknil() bool {
 	return Debug_checknil != 0
 }
 
-func (e *ssafn) Debug_eagerwb() bool {
-	return Debug_eagerwb != 0
-}
-
 func (e *ssafn) UseWriteBarrier() bool {
 	return use_writebarrier
 }
@@ -5394,8 +5389,6 @@ func (e *ssafn) Syslook(name string) *obj.LSym {
 		return goschedguarded
 	case "writeBarrier":
 		return writeBarrier
-	case "writebarrierptr":
-		return writebarrierptr
 	case "gcWriteBarrier":
 		return gcWriteBarrier
 	case "typedmemmove":
diff --git a/src/cmd/compile/internal/ssa/config.go b/src/cmd/compile/internal/ssa/config.go
index 725cdcd9a4..13e5c50da1 100644
--- a/src/cmd/compile/internal/ssa/config.go
+++ b/src/cmd/compile/internal/ssa/config.go
@@ -89,7 +89,6 @@ type Logger interface {
 
 	// Forwards the Debug flags from gc
 	Debug_checknil() bool
-	Debug_eagerwb() bool
 }
 
 type Frontend interface {
diff --git a/src/cmd/compile/internal/ssa/export_test.go b/src/cmd/compile/internal/ssa/export_test.go
index 28ae494505..d1d6831eb3 100644
--- a/src/cmd/compile/internal/ssa/export_test.go
+++ b/src/cmd/compile/internal/ssa/export_test.go
@@ -134,7 +134,6 @@ func (d DummyFrontend) Log() bool                            { return true }
 func (d DummyFrontend) Fatalf(_ src.XPos, msg string, args ...interface{}) { d.t.Fatalf(msg, args...) }
 func (d DummyFrontend) Warnl(_ src.XPos, msg string, args ...interface{})  { d.t.Logf(msg, args...) }
 func (d DummyFrontend) Debug_checknil() bool                               { return false }
-func (d DummyFrontend) Debug_eagerwb() bool                                { return false }
 
 var dummyTypes Types
 
diff --git a/src/cmd/compile/internal/ssa/writebarrier.go b/src/cmd/compile/internal/ssa/writebarrier.go
index b711d8d2bf..c41a677159 100644
--- a/src/cmd/compile/internal/ssa/writebarrier.go
+++ b/src/cmd/compile/internal/ssa/writebarrier.go
@@ -31,7 +31,7 @@ func needwb(v *Value) bool {
 // and runtime calls, like
 //
 // if writeBarrier.enabled {
-//   writebarrierptr(ptr, val)
+//   gcWriteBarrier(ptr, val)	// Not a regular Go call
 // } else {
 //   *ptr = val
 // }
@@ -44,7 +44,7 @@ func writebarrier(f *Func) {
 	}
 
 	var sb, sp, wbaddr, const0 *Value
-	var writebarrierptr, typedmemmove, typedmemclr, gcWriteBarrier *obj.LSym
+	var typedmemmove, typedmemclr, gcWriteBarrier *obj.LSym
 	var stores, after []*Value
 	var sset *sparseSet
 	var storeNumber []int32
@@ -96,10 +96,7 @@ func writebarrier(f *Func) {
 			}
 			wbsym := f.fe.Syslook("writeBarrier")
 			wbaddr = f.Entry.NewValue1A(initpos, OpAddr, f.Config.Types.UInt32Ptr, wbsym, sb)
-			writebarrierptr = f.fe.Syslook("writebarrierptr")
-			if !f.fe.Debug_eagerwb() {
-				gcWriteBarrier = f.fe.Syslook("gcWriteBarrier")
-			}
+			gcWriteBarrier = f.fe.Syslook("gcWriteBarrier")
 			typedmemmove = f.fe.Syslook("typedmemmove")
 			typedmemclr = f.fe.Syslook("typedmemclr")
 			const0 = f.ConstInt32(initpos, f.Config.Types.UInt32, 0)
@@ -198,7 +195,6 @@ func writebarrier(f *Func) {
 			var val *Value
 			switch w.Op {
 			case OpStoreWB:
-				fn = writebarrierptr
 				val = w.Args[1]
 				nWBops--
 			case OpMoveWB:
@@ -217,11 +213,13 @@ func writebarrier(f *Func) {
 			switch w.Op {
 			case OpStoreWB, OpMoveWB, OpZeroWB:
 				volatile := w.Op == OpMoveWB && isVolatile(val)
-				if w.Op == OpStoreWB && !f.fe.Debug_eagerwb() {
+				if w.Op == OpStoreWB {
 					memThen = bThen.NewValue3A(pos, OpWB, types.TypeMem, gcWriteBarrier, ptr, val, memThen)
 				} else {
 					memThen = wbcall(pos, bThen, fn, typ, ptr, val, memThen, sp, sb, volatile)
 				}
+				// Note that we set up a writebarrier function call.
+				f.fe.SetWBPos(pos)
 			case OpVarDef, OpVarLive, OpVarKill:
 				memThen = bThen.NewValue1A(pos, w.Op, types.TypeMem, w.Aux, memThen)
 			}
@@ -239,11 +237,6 @@ func writebarrier(f *Func) {
 			case OpVarDef, OpVarLive, OpVarKill:
 				memElse = bElse.NewValue1A(pos, w.Op, types.TypeMem, w.Aux, memElse)
 			}
-
-			if fn != nil {
-				// Note that we set up a writebarrier function call.
-				f.fe.SetWBPos(pos)
-			}
 		}
 
 		// merge memory
diff --git a/src/cmd/vet/all/whitelist/386.txt b/src/cmd/vet/all/whitelist/386.txt
index 744ac654fd..505856f368 100644
--- a/src/cmd/vet/all/whitelist/386.txt
+++ b/src/cmd/vet/all/whitelist/386.txt
@@ -25,5 +25,3 @@ runtime/asm_386.s: [386] uint32tofloat64: function uint32tofloat64 missing Go de
 runtime/asm_386.s: [386] float64touint32: function float64touint32 missing Go declaration
 
 runtime/asm_386.s: [386] stackcheck: function stackcheck missing Go declaration
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/all.txt b/src/cmd/vet/all/whitelist/all.txt
index 6792d263a5..960ef6b541 100644
--- a/src/cmd/vet/all/whitelist/all.txt
+++ b/src/cmd/vet/all/whitelist/all.txt
@@ -15,6 +15,9 @@ runtime/asm_ARCHSUFF.s: [GOARCH] cannot check cross-package assembly function: E
 runtime/asm_ARCHSUFF.s: [GOARCH] cannot check cross-package assembly function: IndexByte is in package bytes
 runtime/asm_ARCHSUFF.s: [GOARCH] cannot check cross-package assembly function: IndexByte is in package strings
 
+// The write barrier is called directly by the compiler, so no Go def
+runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
+
 // Legitimate vet complaints in which we are testing for correct runtime behavior
 // in bad situations that vet can also detect statically.
 encoding/json/decode_test.go: struct field m has json tag but is not exported
diff --git a/src/cmd/vet/all/whitelist/amd64.txt b/src/cmd/vet/all/whitelist/amd64.txt
index ebde7be58b..56a6e2eb8d 100644
--- a/src/cmd/vet/all/whitelist/amd64.txt
+++ b/src/cmd/vet/all/whitelist/amd64.txt
@@ -31,4 +31,3 @@ runtime/duff_amd64.s: [amd64] duffcopy: function duffcopy missing Go declaration
 runtime/asm_amd64.s: [amd64] stackcheck: function stackcheck missing Go declaration
 runtime/asm_amd64.s: [amd64] indexShortStr: function indexShortStr missing Go declaration
 runtime/asm_amd64.s: [amd64] countByte: function countByte missing Go declaration
-runtime/asm_amd64.s: [amd64] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/arm.txt b/src/cmd/vet/all/whitelist/arm.txt
index 51b3d6bf51..770008c9f0 100644
--- a/src/cmd/vet/all/whitelist/arm.txt
+++ b/src/cmd/vet/all/whitelist/arm.txt
@@ -16,5 +16,3 @@ runtime/tls_arm.s: [arm] load_g: function load_g missing Go declaration
 runtime/tls_arm.s: [arm] _initcgo: function _initcgo missing Go declaration
 
 runtime/internal/atomic/asm_arm.s: [arm] cas: function cas missing Go declaration
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/arm64.txt b/src/cmd/vet/all/whitelist/arm64.txt
index af2d42a62f..24fc6f4223 100644
--- a/src/cmd/vet/all/whitelist/arm64.txt
+++ b/src/cmd/vet/all/whitelist/arm64.txt
@@ -9,5 +9,3 @@ runtime/duff_arm64.s: [arm64] duffzero: function duffzero missing Go declaration
 runtime/duff_arm64.s: [arm64] duffcopy: function duffcopy missing Go declaration
 runtime/tls_arm64.s: [arm64] load_g: function load_g missing Go declaration
 runtime/tls_arm64.s: [arm64] save_g: function save_g missing Go declaration
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/mips64x.txt b/src/cmd/vet/all/whitelist/mips64x.txt
index 45efdc6ae4..5354d21c64 100644
--- a/src/cmd/vet/all/whitelist/mips64x.txt
+++ b/src/cmd/vet/all/whitelist/mips64x.txt
@@ -4,5 +4,3 @@ runtime/asm_mips64x.s: [GOARCH] abort: function abort missing Go declaration
 runtime/duff_mips64x.s: [GOARCH] duffzero: function duffzero missing Go declaration
 runtime/tls_mips64x.s: [GOARCH] save_g: function save_g missing Go declaration
 runtime/tls_mips64x.s: [GOARCH] load_g: function load_g missing Go declaration
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/mipsx.txt b/src/cmd/vet/all/whitelist/mipsx.txt
index fbf4499bdc..ff6c0e613b 100644
--- a/src/cmd/vet/all/whitelist/mipsx.txt
+++ b/src/cmd/vet/all/whitelist/mipsx.txt
@@ -7,5 +7,3 @@ runtime/asm_mipsx.s: [GOARCH] cannot check cross-package assembly function: Comp
 runtime/sys_linux_mipsx.s: [GOARCH] clone: 12(R29) should be mp+8(FP)
 runtime/sys_linux_mipsx.s: [GOARCH] clone: 4(R29) should be flags+0(FP)
 runtime/sys_linux_mipsx.s: [GOARCH] clone: 8(R29) should be stk+4(FP)
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/nacl_amd64p32.txt b/src/cmd/vet/all/whitelist/nacl_amd64p32.txt
index 5c6b3344a0..4b2aad2aac 100644
--- a/src/cmd/vet/all/whitelist/nacl_amd64p32.txt
+++ b/src/cmd/vet/all/whitelist/nacl_amd64p32.txt
@@ -27,5 +27,3 @@ runtime/asm_amd64p32.s: [amd64p32] indexbytebody: function indexbytebody missing
 runtime/asm_amd64p32.s: [amd64p32] asmcgocall: RET without writing to 4-byte ret+8(FP)
 
 runtime/asm_amd64p32.s: [amd64p32] stackcheck: function stackcheck missing Go declaration
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/ppc64x.txt b/src/cmd/vet/all/whitelist/ppc64x.txt
index 84b8f18b53..4f6444e102 100644
--- a/src/cmd/vet/all/whitelist/ppc64x.txt
+++ b/src/cmd/vet/all/whitelist/ppc64x.txt
@@ -10,5 +10,3 @@ runtime/asm_ppc64x.s: [GOARCH] addmoduledata: function addmoduledata missing Go
 runtime/duff_ppc64x.s: [GOARCH] duffzero: function duffzero missing Go declaration
 runtime/tls_ppc64x.s: [GOARCH] save_g: function save_g missing Go declaration
 runtime/tls_ppc64x.s: [GOARCH] load_g: function load_g missing Go declaration
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/cmd/vet/all/whitelist/s390x.txt b/src/cmd/vet/all/whitelist/s390x.txt
index 68e5461a3c..f18236c4f1 100644
--- a/src/cmd/vet/all/whitelist/s390x.txt
+++ b/src/cmd/vet/all/whitelist/s390x.txt
@@ -15,5 +15,3 @@ runtime/memclr_s390x.s: [s390x] memclr_s390x_exrl_xc: function memclr_s390x_exrl
 runtime/memmove_s390x.s: [s390x] memmove_s390x_exrl_mvc: function memmove_s390x_exrl_mvc missing Go declaration
 runtime/tls_s390x.s: [s390x] save_g: function save_g missing Go declaration
 runtime/tls_s390x.s: [s390x] load_g: function load_g missing Go declaration
-
-runtime/asm_ARCHSUFF.s: [GOARCH] gcWriteBarrier: function gcWriteBarrier missing Go declaration
diff --git a/src/runtime/asm_amd64.s b/src/runtime/asm_amd64.s
index 576a61ca6c..82b7832ae3 100644
--- a/src/runtime/asm_amd64.s
+++ b/src/runtime/asm_amd64.s
@@ -2397,7 +2397,13 @@ TEXT runtime·gcWriteBarrier(SB),NOSPLIT,$120
 	CMPQ	R14, (p_wbBuf+wbBuf_end)(R13)
 	// Record the write.
 	MOVQ	AX, -16(R14)	// Record value
-	MOVQ	(DI), R13	// TODO: This turns bad writes into bad reads.
+	// Note: This turns bad pointer writes into bad
+	// pointer reads, which could be confusing. We could avoid
+	// reading from obviously bad pointers, which would
+	// take care of the vast majority of these. We could
+	// patch this up in the signal handler, or use XCHG to
+	// combine the read and the write.
+	MOVQ	(DI), R13
 	MOVQ	R13, -8(R14)	// Record *slot
 	// Is the buffer full? (flags set in CMPQ above)
 	JEQ	flush
diff --git a/src/runtime/mbarrier.go b/src/runtime/mbarrier.go
index e28bdb8b8d..c071728900 100644
--- a/src/runtime/mbarrier.go
+++ b/src/runtime/mbarrier.go
@@ -6,10 +6,10 @@
 //
 // For the concurrent garbage collector, the Go compiler implements
 // updates to pointer-valued fields that may be in heap objects by
-// emitting calls to write barriers. This file contains the actual write barrier
-// implementation, gcmarkwb_m, and the various wrappers called by the
-// compiler to implement pointer assignment, slice assignment,
-// typed memmove, and so on.
+// emitting calls to write barriers. The main write barrier for
+// individual pointer writes is gcWriteBarrier and is implemented in
+// assembly. This file contains write barrier entry points for bulk
+// operations. See also mwbbuf.go.
 
 package runtime
 
@@ -18,10 +18,7 @@ import (
 	"unsafe"
 )
 
-// gcmarkwb_m is the mark-phase write barrier, the only barrier we have.
-// The rest of this file exists only to make calls to this function.
-//
-// This is a hybrid barrier that combines a Yuasa-style deletion
+// Go uses a hybrid barrier that combines a Yuasa-style deletion
 // barrier—which shades the object whose reference is being
 // overwritten—with Dijkstra insertion barrier—which shades the object
 // whose reference is being written. The insertion part of the barrier
@@ -137,105 +134,17 @@ import (
 // reachable by some goroutine that currently cannot reach it.
 //
 //
-//go:nowritebarrierrec
-//go:systemstack
-func gcmarkwb_m(slot *uintptr, ptr uintptr) {
-	if writeBarrier.needed {
-		// Note: This turns bad pointer writes into bad
-		// pointer reads, which could be confusing. We avoid
-		// reading from obviously bad pointers, which should
-		// take care of the vast majority of these. We could
-		// patch this up in the signal handler, or use XCHG to
-		// combine the read and the write. Checking inheap is
-		// insufficient since we need to track changes to
-		// roots outside the heap.
-		//
-		// Note: profbuf.go omits a barrier during signal handler
-		// profile logging; that's safe only because this deletion barrier exists.
-		// If we remove the deletion barrier, we'll have to work out
-		// a new way to handle the profile logging.
-		if slot1 := uintptr(unsafe.Pointer(slot)); slot1 >= minPhysPageSize {
-			if optr := *slot; optr != 0 {
-				shade(optr)
-			}
-		}
-		// TODO: Make this conditional on the caller's stack color.
-		if ptr != 0 && inheap(ptr) {
-			shade(ptr)
-		}
-	}
-}
-
-// writebarrierptr_prewrite1 invokes a write barrier for *dst = src
-// prior to the write happening.
-//
-// Write barrier calls must not happen during critical GC and scheduler
-// related operations. In particular there are times when the GC assumes
-// that the world is stopped but scheduler related code is still being
-// executed, dealing with syscalls, dealing with putting gs on runnable
-// queues and so forth. This code cannot execute write barriers because
-// the GC might drop them on the floor. Stopping the world involves removing
-// the p associated with an m. We use the fact that m.p == nil to indicate
-// that we are in one these critical section and throw if the write is of
-// a pointer to a heap object.
-//go:nosplit
-func writebarrierptr_prewrite1(dst *uintptr, src uintptr) {
-	mp := acquirem()
-	if mp.inwb || mp.dying > 0 {
-		// We explicitly allow write barriers in startpanic_m,
-		// since we're going down anyway. Ignore them here.
-		releasem(mp)
-		return
-	}
-	systemstack(func() {
-		if mp.p == 0 && memstats.enablegc && !mp.inwb && inheap(src) {
-			throw("writebarrierptr_prewrite1 called with mp.p == nil")
-		}
-		mp.inwb = true
-		gcmarkwb_m(dst, src)
-	})
-	mp.inwb = false
-	releasem(mp)
-}
-
-// NOTE: Really dst *unsafe.Pointer, src unsafe.Pointer,
-// but if we do that, Go inserts a write barrier on *dst = src.
-//go:nosplit
-func writebarrierptr(dst *uintptr, src uintptr) {
-	if writeBarrier.cgo {
-		cgoCheckWriteBarrier(dst, src)
-	}
-	if !writeBarrier.needed {
-		*dst = src
-		return
-	}
-	if src != 0 && src < minPhysPageSize {
-		systemstack(func() {
-			print("runtime: writebarrierptr *", dst, " = ", hex(src), "\n")
-			throw("bad pointer in write barrier")
-		})
-	}
-	writebarrierptr_prewrite1(dst, src)
-	*dst = src
-}
-
-// writebarrierptr_prewrite is like writebarrierptr, but the store
-// will be performed by the caller after this call. The caller must
-// not allow preemption between this call and the write.
+// Signal handler pointer writes:
 //
-//go:nosplit
-func writebarrierptr_prewrite(dst *uintptr, src uintptr) {
-	if writeBarrier.cgo {
-		cgoCheckWriteBarrier(dst, src)
-	}
-	if !writeBarrier.needed {
-		return
-	}
-	if src != 0 && src < minPhysPageSize {
-		systemstack(func() { throw("bad pointer in write barrier") })
-	}
-	writebarrierptr_prewrite1(dst, src)
-}
+// In general, the signal handler cannot safely invoke the write
+// barrier because it may run without a P or even during the write
+// barrier.
+//
+// There is exactly one exception: profbuf.go omits a barrier during
+// signal handler profile logging. That's safe only because of the
+// deletion barrier. See profbuf.go for a detailed argument. If we
+// remove the deletion barrier, we'll have to work out a new way to
+// handle the profile logging.
 
 // typedmemmove copies a value of type t to dst from src.
 // Must be nosplit, see #16026.
diff --git a/src/runtime/mwbbuf.go b/src/runtime/mwbbuf.go
index 4a2d1ad988..c5619ed3fb 100644
--- a/src/runtime/mwbbuf.go
+++ b/src/runtime/mwbbuf.go
@@ -5,6 +5,9 @@
 // This implements the write barrier buffer. The write barrier itself
 // is gcWriteBarrier and is implemented in assembly.
 //
+// See mbarrier.go for algorithmic details on the write barrier. This
+// file deals only with the buffer.
+//
 // The write barrier has a fast path and a slow path. The fast path
 // simply enqueues to a per-P write barrier buffer. It's written in
 // assembly and doesn't clobber any general purpose registers, so it
@@ -111,16 +114,21 @@ func (b *wbBuf) discard() {
 //     if !buf.putFast(old, new) {
 //         wbBufFlush(...)
 //     }
+//     ... actual memory write ...
 //
 // The arguments to wbBufFlush depend on whether the caller is doing
 // its own cgo pointer checks. If it is, then this can be
 // wbBufFlush(nil, 0). Otherwise, it must pass the slot address and
 // new.
 //
-// Since buf is a per-P resource, the caller must ensure there are no
-// preemption points while buf is in use.
+// The caller must ensure there are no preemption points during the
+// above sequence. There must be no preemption points while buf is in
+// use because it is a per-P resource. There must be no preemption
+// points between the buffer put and the write to memory because this
+// could allow a GC phase change, which could result in missed write
+// barriers.
 //
-// It must be nowritebarrierrec to because write barriers here would
+// putFast must be nowritebarrierrec to because write barriers here would
 // corrupt the write barrier buffer. It (and everything it calls, if
 // it called anything) has to be nosplit to avoid scheduling on to a
 // different P and a different buffer.
@@ -214,6 +222,14 @@ func wbBufFlush1(_p_ *p) {
 	//
 	// TODO: Should scanobject/scanblock just stuff pointers into
 	// the wbBuf? Then this would become the sole greying path.
+	//
+	// TODO: We could avoid shading any of the "new" pointers in
+	// the buffer if the stack has been shaded, or even avoid
+	// putting them in the buffer at all (which would double its
+	// capacity). This is slightly complicated with the buffer; we
+	// could track whether any un-shaded goroutine has used the
+	// buffer, or just track globally whether there are any
+	// un-shaded stacks and flush after each stack scan.
 	gcw := &_p_.gcw
 	pos := 0
 	arenaStart := mheap_.arena_start
diff --git a/test/fixedbugs/issue15747.go b/test/fixedbugs/issue15747.go
index decabc754e..c7ef96d581 100644
--- a/test/fixedbugs/issue15747.go
+++ b/test/fixedbugs/issue15747.go
@@ -1,4 +1,4 @@
-// errorcheck -0 -live -d=eagerwb
+// errorcheck -0 -live
 
 // Copyright 2016 The Go Authors.  All rights reserved.
 // Use of this source code is governed by a BSD-style
@@ -7,10 +7,6 @@
 // Issue 15747: liveness analysis was marking heap-escaped params live too much,
 // and worse was using the wrong bitmap bits to do so.
 
-// TODO(austin): This expects function calls to the write barrier, so
-// we enable the legacy eager write barrier. Fix this once the
-// buffered write barrier works on all arches.
-
 package p
 
 var global *[]byte
@@ -21,14 +17,14 @@ type T struct{ M string }
 
 var b bool
 
-func f1(q *Q, xx []byte) interface{} { // ERROR "live at call to newobject: xx$" "live at call to writebarrierptr: &xx$" "live at entry to f1: xx$"
+func f1(q *Q, xx []byte) interface{} { // ERROR "live at call to newobject: xx$" "live at entry to f1: xx$"
 	// xx was copied from the stack to the heap on the previous line:
 	// xx was live for the first two prints but then it switched to &xx
 	// being live. We should not see plain xx again.
 	if b {
-		global = &xx // ERROR "live at call to writebarrierptr: &xx$"
+		global = &xx
 	}
-	xx, _, err := f2(xx, 5) // ERROR "live at call to f2: &xx$" "live at call to writebarrierptr: err.data err.type$"
+	xx, _, err := f2(xx, 5) // ERROR "live at call to f2: &xx$"
 	if err != nil {
 		return err
 	}
@@ -38,7 +34,7 @@ func f1(q *Q, xx []byte) interface{} { // ERROR "live at call to newobject: xx$"
 //go:noinline
 func f2(d []byte, n int) (odata, res []byte, e interface{}) { // ERROR "live at entry to f2: d$"
 	if n > len(d) {
-		return d, nil, &T{M: "hello"} // ERROR "live at call to newobject: d" "live at call to writebarrierptr: d"
+		return d, nil, &T{M: "hello"} // ERROR "live at call to newobject: d"
 	}
 	res = d[:n]
 	odata = d[n:]
diff --git a/test/fixedbugs/issue20250.go b/test/fixedbugs/issue20250.go
index 525192a46b..6fc861a8dc 100644
--- a/test/fixedbugs/issue20250.go
+++ b/test/fixedbugs/issue20250.go
@@ -1,4 +1,4 @@
-// errorcheck -0 -live -l -d=compilelater,eagerwb
+// errorcheck -0 -live -l -d=compilelater
 
 // Copyright 2017 The Go Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style
@@ -8,10 +8,6 @@
 // due to propagation of addrtaken to outer variables for
 // closure variables.
 
-// TODO(austin): This expects function calls to the write barrier, so
-// we enable the legacy eager write barrier. Fix this once the
-// buffered write barrier works on all arches.
-
 package p
 
 type T struct {
@@ -21,7 +17,7 @@ type T struct {
 func f(a T) { // ERROR "live at entry to f: a"
 	var e interface{}
 	func() { // ERROR "live at entry to f.func1: a &e"
-		e = a.s // ERROR "live at call to convT2Estring: a &e" "live at call to writebarrierptr: a"
+		e = a.s // ERROR "live at call to convT2Estring: a &e"
 	}() // ERROR "live at call to f.func1: e$"
 	// Before the fix, both a and e were live at the previous line.
 	_ = e