[dev.link] all: merge branch 'master' into dev.link

Change-Id: Ic66b5138f3ecd9e9a48d7ab05782297c06e4a5b5

254 files changed, 4493 insertions, 3426 deletions

diff --git a/src/bufio/bufio.go b/src/bufio/bufio.go
index 7cbd5424ea..6baf9b9e40 100644
--- a/src/bufio/bufio.go
+++ b/src/bufio/bufio.go
@@ -425,7 +425,7 @@ func (b *Reader) ReadLine() (line []byte, isPrefix bool, err error) {
 // of bytes in the combined first two elements, error).
 // The complete result is equal to
 // `bytes.Join(append(fullBuffers, finalFragment), nil)`, which has a
-// length of `totalLen`. The result is strucured in this way to allow callers
+// length of `totalLen`. The result is structured in this way to allow callers
 // to minimize allocations and copies.
 func (b *Reader) collectFragments(delim byte) (fullBuffers [][]byte, finalFragment []byte, totalLen int, err error) {
 	var frag []byte
diff --git a/src/cmd/asm/internal/asm/testdata/arm64.s b/src/cmd/asm/internal/asm/testdata/arm64.s
index 69267bfa63..5a6db05074 100644
--- a/src/cmd/asm/internal/asm/testdata/arm64.s
+++ b/src/cmd/asm/internal/asm/testdata/arm64.s
@@ -77,6 +77,10 @@ TEXT	foo(SB), DUPOK|NOSPLIT, $-8
 	SHA1H	V5, V4                          // a408285e
 	SHA1M	V8.S4, V7, V6                   // e620085e
 	SHA1P	V11.S4, V10, V9                 // 49110b5e
+	SHA512H	V2.D2, V1, V0                   // 208062ce
+	SHA512H2	V4.D2, V3, V2           // 628464ce
+	SHA512SU0	V9.D2, V8.D2            // 2881c0ce
+	SHA512SU1	V7.D2, V6.D2, V5.D2     // c58867ce
 	VADDV	V0.S4, V0                       // 00b8b14e
 	VMOVI	$82, V0.B16                     // 40e6024f
 	VUADDLV	V6.B16, V6                      // c638306e
diff --git a/src/cmd/asm/internal/asm/testdata/ppc64.s b/src/cmd/asm/internal/asm/testdata/ppc64.s
index b3736bf6a4..ba64d84a35 100644
--- a/src/cmd/asm/internal/asm/testdata/ppc64.s
+++ b/src/cmd/asm/internal/asm/testdata/ppc64.s
@@ -1037,6 +1037,7 @@ label1:
 // VSX load with length X-form (also left-justified)
 	LXVL        R3,R4, VS0
 	LXVLL       R3,R4, VS0
+	LXVX        R3,R4, VS0
 // VSX load, DQ-form
 // <MNEMONIC> DQ(RA), XS produces
 // <mnemonic> XS, DQ(RA)
@@ -1060,6 +1061,7 @@ label1:
 // VSX store with length, X-form (also left-justified)
 	STXVL	    VS0, R3,R4
 	STXVLL      VS0, R3,R4
+	STXVX       VS0, R3,R4
 
 //	VSX move from VSR, XX1-form
 //	<MNEMONIC> XS,RA produces
diff --git a/src/cmd/asm/internal/asm/testdata/ppc64enc.s b/src/cmd/asm/internal/asm/testdata/ppc64enc.s
index 07a8a540cd..10a05ec402 100644
--- a/src/cmd/asm/internal/asm/testdata/ppc64enc.s
+++ b/src/cmd/asm/internal/asm/testdata/ppc64enc.s
@@ -595,11 +595,13 @@ TEXT asmtest(SB),DUPOK|NOSPLIT,$0
 	LXV 16(R3), VS1                 // f4230011
 	LXVL R3, R4, VS1                // 7c23221a
 	LXVLL R3, R4, VS1               // 7c23225a
+	LXVX R3, R4, VS1                // 7c232218
 	LXSDX (R3)(R4), VS1             // 7c241c98
 	STXVD2X VS1, (R3)(R4)           // 7c241f98
 	STXV VS1,16(R3)                 // f4230015
 	STXVL VS1, R3, R4               // 7c23231a
 	STXVLL VS1, R3, R4              // 7c23235a
+	STXVX VS1, R3, R4               // 7c232318
 	STXSDX VS1, (R3)(R4)            // 7c241d98
 	LXSIWAX (R3)(R4), VS1           // 7c241898
 	STXSIWX VS1, (R3)(R4)           // 7c241918
diff --git a/src/cmd/cgo/out.go b/src/cmd/cgo/out.go
index 4064f0ae41..50d2811f1b 100644
--- a/src/cmd/cgo/out.go
+++ b/src/cmd/cgo/out.go
@@ -243,6 +243,7 @@ func (p *Package) writeDefs() {
 		if err != nil {
 			fatalf("%s", err)
 		}
+		defer fgcch.Close()
 		_, err = io.Copy(fexp, fgcch)
 		if err != nil {
 			fatalf("%s", err)
diff --git a/src/cmd/compile/internal/gc/builtin.go b/src/cmd/compile/internal/gc/builtin.go
index 2cf2f4687e..861ffaaa5b 100644
--- a/src/cmd/compile/internal/gc/builtin.go
+++ b/src/cmd/compile/internal/gc/builtin.go
@@ -126,74 +126,74 @@ var runtimeDecls = [...]struct {
 	{"selectnbsend", funcTag, 94},
 	{"selectnbrecv", funcTag, 95},
 	{"selectnbrecv2", funcTag, 97},
-	{"selectsetpc", funcTag, 62},
-	{"selectgo", funcTag, 98},
+	{"selectsetpc", funcTag, 98},
+	{"selectgo", funcTag, 99},
 	{"block", funcTag, 9},
-	{"makeslice", funcTag, 99},
-	{"makeslice64", funcTag, 100},
-	{"makeslicecopy", funcTag, 101},
-	{"growslice", funcTag, 103},
-	{"memmove", funcTag, 104},
-	{"memclrNoHeapPointers", funcTag, 105},
-	{"memclrHasPointers", funcTag, 105},
-	{"memequal", funcTag, 106},
-	{"memequal0", funcTag, 107},
-	{"memequal8", funcTag, 107},
-	{"memequal16", funcTag, 107},
-	{"memequal32", funcTag, 107},
-	{"memequal64", funcTag, 107},
-	{"memequal128", funcTag, 107},
-	{"f32equal", funcTag, 108},
-	{"f64equal", funcTag, 108},
-	{"c64equal", funcTag, 108},
-	{"c128equal", funcTag, 108},
-	{"strequal", funcTag, 108},
-	{"interequal", funcTag, 108},
-	{"nilinterequal", funcTag, 108},
-	{"memhash", funcTag, 109},
-	{"memhash0", funcTag, 110},
-	{"memhash8", funcTag, 110},
-	{"memhash16", funcTag, 110},
-	{"memhash32", funcTag, 110},
-	{"memhash64", funcTag, 110},
-	{"memhash128", funcTag, 110},
-	{"f32hash", funcTag, 110},
-	{"f64hash", funcTag, 110},
-	{"c64hash", funcTag, 110},
-	{"c128hash", funcTag, 110},
-	{"strhash", funcTag, 110},
-	{"interhash", funcTag, 110},
-	{"nilinterhash", funcTag, 110},
-	{"int64div", funcTag, 111},
-	{"uint64div", funcTag, 112},
-	{"int64mod", funcTag, 111},
-	{"uint64mod", funcTag, 112},
-	{"float64toint64", funcTag, 113},
-	{"float64touint64", funcTag, 114},
-	{"float64touint32", funcTag, 115},
-	{"int64tofloat64", funcTag, 116},
-	{"uint64tofloat64", funcTag, 117},
-	{"uint32tofloat64", funcTag, 118},
-	{"complex128div", funcTag, 119},
-	{"racefuncenter", funcTag, 120},
+	{"makeslice", funcTag, 100},
+	{"makeslice64", funcTag, 101},
+	{"makeslicecopy", funcTag, 102},
+	{"growslice", funcTag, 104},
+	{"memmove", funcTag, 105},
+	{"memclrNoHeapPointers", funcTag, 106},
+	{"memclrHasPointers", funcTag, 106},
+	{"memequal", funcTag, 107},
+	{"memequal0", funcTag, 108},
+	{"memequal8", funcTag, 108},
+	{"memequal16", funcTag, 108},
+	{"memequal32", funcTag, 108},
+	{"memequal64", funcTag, 108},
+	{"memequal128", funcTag, 108},
+	{"f32equal", funcTag, 109},
+	{"f64equal", funcTag, 109},
+	{"c64equal", funcTag, 109},
+	{"c128equal", funcTag, 109},
+	{"strequal", funcTag, 109},
+	{"interequal", funcTag, 109},
+	{"nilinterequal", funcTag, 109},
+	{"memhash", funcTag, 110},
+	{"memhash0", funcTag, 111},
+	{"memhash8", funcTag, 111},
+	{"memhash16", funcTag, 111},
+	{"memhash32", funcTag, 111},
+	{"memhash64", funcTag, 111},
+	{"memhash128", funcTag, 111},
+	{"f32hash", funcTag, 111},
+	{"f64hash", funcTag, 111},
+	{"c64hash", funcTag, 111},
+	{"c128hash", funcTag, 111},
+	{"strhash", funcTag, 111},
+	{"interhash", funcTag, 111},
+	{"nilinterhash", funcTag, 111},
+	{"int64div", funcTag, 112},
+	{"uint64div", funcTag, 113},
+	{"int64mod", funcTag, 112},
+	{"uint64mod", funcTag, 113},
+	{"float64toint64", funcTag, 114},
+	{"float64touint64", funcTag, 115},
+	{"float64touint32", funcTag, 116},
+	{"int64tofloat64", funcTag, 117},
+	{"uint64tofloat64", funcTag, 118},
+	{"uint32tofloat64", funcTag, 119},
+	{"complex128div", funcTag, 120},
+	{"racefuncenter", funcTag, 121},
 	{"racefuncenterfp", funcTag, 9},
 	{"racefuncexit", funcTag, 9},
-	{"raceread", funcTag, 120},
-	{"racewrite", funcTag, 120},
-	{"racereadrange", funcTag, 121},
-	{"racewriterange", funcTag, 121},
-	{"msanread", funcTag, 121},
-	{"msanwrite", funcTag, 121},
-	{"checkptrAlignment", funcTag, 122},
-	{"checkptrArithmetic", funcTag, 124},
-	{"libfuzzerTraceCmp1", funcTag, 126},
-	{"libfuzzerTraceCmp2", funcTag, 128},
-	{"libfuzzerTraceCmp4", funcTag, 129},
-	{"libfuzzerTraceCmp8", funcTag, 130},
-	{"libfuzzerTraceConstCmp1", funcTag, 126},
-	{"libfuzzerTraceConstCmp2", funcTag, 128},
-	{"libfuzzerTraceConstCmp4", funcTag, 129},
-	{"libfuzzerTraceConstCmp8", funcTag, 130},
+	{"raceread", funcTag, 121},
+	{"racewrite", funcTag, 121},
+	{"racereadrange", funcTag, 122},
+	{"racewriterange", funcTag, 122},
+	{"msanread", funcTag, 122},
+	{"msanwrite", funcTag, 122},
+	{"checkptrAlignment", funcTag, 123},
+	{"checkptrArithmetic", funcTag, 125},
+	{"libfuzzerTraceCmp1", funcTag, 127},
+	{"libfuzzerTraceCmp2", funcTag, 129},
+	{"libfuzzerTraceCmp4", funcTag, 130},
+	{"libfuzzerTraceCmp8", funcTag, 131},
+	{"libfuzzerTraceConstCmp1", funcTag, 127},
+	{"libfuzzerTraceConstCmp2", funcTag, 129},
+	{"libfuzzerTraceConstCmp4", funcTag, 130},
+	{"libfuzzerTraceConstCmp8", funcTag, 131},
 	{"x86HasPOPCNT", varTag, 6},
 	{"x86HasSSE41", varTag, 6},
 	{"x86HasFMA", varTag, 6},
@@ -202,7 +202,7 @@ var runtimeDecls = [...]struct {
 }
 
 func runtimeTypes() []*types.Type {
-	var typs [131]*types.Type
+	var typs [132]*types.Type
 	typs[0] = types.Bytetype
 	typs[1] = types.NewPtr(typs[0])
 	typs[2] = types.Types[TANY]
@@ -301,38 +301,39 @@ func runtimeTypes() []*types.Type {
 	typs[95] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[84])}, []*Node{anonfield(typs[6])})
 	typs[96] = types.NewPtr(typs[6])
 	typs[97] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[96]), anonfield(typs[84])}, []*Node{anonfield(typs[6])})
-	typs[98] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[1]), anonfield(typs[15])}, []*Node{anonfield(typs[15]), anonfield(typs[6])})
-	typs[99] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[15])}, []*Node{anonfield(typs[7])})
-	typs[100] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[22]), anonfield(typs[22])}, []*Node{anonfield(typs[7])})
-	typs[101] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[15]), anonfield(typs[7])}, []*Node{anonfield(typs[7])})
-	typs[102] = types.NewSlice(typs[2])
-	typs[103] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[102]), anonfield(typs[15])}, []*Node{anonfield(typs[102])})
-	typs[104] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[5])}, nil)
-	typs[105] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[5])}, nil)
-	typs[106] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[5])}, []*Node{anonfield(typs[6])})
-	typs[107] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3])}, []*Node{anonfield(typs[6])})
-	typs[108] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[7])}, []*Node{anonfield(typs[6])})
-	typs[109] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[5]), anonfield(typs[5])}, []*Node{anonfield(typs[5])})
-	typs[110] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[5])}, []*Node{anonfield(typs[5])})
-	typs[111] = functype(nil, []*Node{anonfield(typs[22]), anonfield(typs[22])}, []*Node{anonfield(typs[22])})
-	typs[112] = functype(nil, []*Node{anonfield(typs[24]), anonfield(typs[24])}, []*Node{anonfield(typs[24])})
-	typs[113] = functype(nil, []*Node{anonfield(typs[20])}, []*Node{anonfield(typs[22])})
-	typs[114] = functype(nil, []*Node{anonfield(typs[20])}, []*Node{anonfield(typs[24])})
-	typs[115] = functype(nil, []*Node{anonfield(typs[20])}, []*Node{anonfield(typs[65])})
-	typs[116] = functype(nil, []*Node{anonfield(typs[22])}, []*Node{anonfield(typs[20])})
-	typs[117] = functype(nil, []*Node{anonfield(typs[24])}, []*Node{anonfield(typs[20])})
-	typs[118] = functype(nil, []*Node{anonfield(typs[65])}, []*Node{anonfield(typs[20])})
-	typs[119] = functype(nil, []*Node{anonfield(typs[26]), anonfield(typs[26])}, []*Node{anonfield(typs[26])})
-	typs[120] = functype(nil, []*Node{anonfield(typs[5])}, nil)
-	typs[121] = functype(nil, []*Node{anonfield(typs[5]), anonfield(typs[5])}, nil)
-	typs[122] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[1]), anonfield(typs[5])}, nil)
-	typs[123] = types.NewSlice(typs[7])
-	typs[124] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[123])}, nil)
-	typs[125] = types.Types[TUINT8]
-	typs[126] = functype(nil, []*Node{anonfield(typs[125]), anonfield(typs[125])}, nil)
-	typs[127] = types.Types[TUINT16]
-	typs[128] = functype(nil, []*Node{anonfield(typs[127]), anonfield(typs[127])}, nil)
-	typs[129] = functype(nil, []*Node{anonfield(typs[65]), anonfield(typs[65])}, nil)
-	typs[130] = functype(nil, []*Node{anonfield(typs[24]), anonfield(typs[24])}, nil)
+	typs[98] = functype(nil, []*Node{anonfield(typs[63])}, nil)
+	typs[99] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[1]), anonfield(typs[63]), anonfield(typs[15]), anonfield(typs[15]), anonfield(typs[6])}, []*Node{anonfield(typs[15]), anonfield(typs[6])})
+	typs[100] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[15])}, []*Node{anonfield(typs[7])})
+	typs[101] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[22]), anonfield(typs[22])}, []*Node{anonfield(typs[7])})
+	typs[102] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[15]), anonfield(typs[15]), anonfield(typs[7])}, []*Node{anonfield(typs[7])})
+	typs[103] = types.NewSlice(typs[2])
+	typs[104] = functype(nil, []*Node{anonfield(typs[1]), anonfield(typs[103]), anonfield(typs[15])}, []*Node{anonfield(typs[103])})
+	typs[105] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[5])}, nil)
+	typs[106] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[5])}, nil)
+	typs[107] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3]), anonfield(typs[5])}, []*Node{anonfield(typs[6])})
+	typs[108] = functype(nil, []*Node{anonfield(typs[3]), anonfield(typs[3])}, []*Node{anonfield(typs[6])})
+	typs[109] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[7])}, []*Node{anonfield(typs[6])})
+	typs[110] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[5]), anonfield(typs[5])}, []*Node{anonfield(typs[5])})
+	typs[111] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[5])}, []*Node{anonfield(typs[5])})
+	typs[112] = functype(nil, []*Node{anonfield(typs[22]), anonfield(typs[22])}, []*Node{anonfield(typs[22])})
+	typs[113] = functype(nil, []*Node{anonfield(typs[24]), anonfield(typs[24])}, []*Node{anonfield(typs[24])})
+	typs[114] = functype(nil, []*Node{anonfield(typs[20])}, []*Node{anonfield(typs[22])})
+	typs[115] = functype(nil, []*Node{anonfield(typs[20])}, []*Node{anonfield(typs[24])})
+	typs[116] = functype(nil, []*Node{anonfield(typs[20])}, []*Node{anonfield(typs[65])})
+	typs[117] = functype(nil, []*Node{anonfield(typs[22])}, []*Node{anonfield(typs[20])})
+	typs[118] = functype(nil, []*Node{anonfield(typs[24])}, []*Node{anonfield(typs[20])})
+	typs[119] = functype(nil, []*Node{anonfield(typs[65])}, []*Node{anonfield(typs[20])})
+	typs[120] = functype(nil, []*Node{anonfield(typs[26]), anonfield(typs[26])}, []*Node{anonfield(typs[26])})
+	typs[121] = functype(nil, []*Node{anonfield(typs[5])}, nil)
+	typs[122] = functype(nil, []*Node{anonfield(typs[5]), anonfield(typs[5])}, nil)
+	typs[123] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[1]), anonfield(typs[5])}, nil)
+	typs[124] = types.NewSlice(typs[7])
+	typs[125] = functype(nil, []*Node{anonfield(typs[7]), anonfield(typs[124])}, nil)
+	typs[126] = types.Types[TUINT8]
+	typs[127] = functype(nil, []*Node{anonfield(typs[126]), anonfield(typs[126])}, nil)
+	typs[128] = types.Types[TUINT16]
+	typs[129] = functype(nil, []*Node{anonfield(typs[128]), anonfield(typs[128])}, nil)
+	typs[130] = functype(nil, []*Node{anonfield(typs[65]), anonfield(typs[65])}, nil)
+	typs[131] = functype(nil, []*Node{anonfield(typs[24]), anonfield(typs[24])}, nil)
 	return typs[:]
 }
diff --git a/src/cmd/compile/internal/gc/builtin/runtime.go b/src/cmd/compile/internal/gc/builtin/runtime.go
index 00448272c5..635da80f7c 100644
--- a/src/cmd/compile/internal/gc/builtin/runtime.go
+++ b/src/cmd/compile/internal/gc/builtin/runtime.go
@@ -169,8 +169,8 @@ func selectnbsend(hchan chan<- any, elem *any) bool
 func selectnbrecv(elem *any, hchan <-chan any) bool
 func selectnbrecv2(elem *any, received *bool, hchan <-chan any) bool
 
-func selectsetpc(cas *byte)
-func selectgo(cas0 *byte, order0 *byte, ncases int) (int, bool)
+func selectsetpc(pc *uintptr)
+func selectgo(cas0 *byte, order0 *byte, pc0 *uintptr, nsends int, nrecvs int, block bool) (int, bool)
 func block()
 
 func makeslice(typ *byte, len int, cap int) unsafe.Pointer
diff --git a/src/cmd/compile/internal/gc/closure.go b/src/cmd/compile/internal/gc/closure.go
index 3bb7bb9834..23e48939b4 100644
--- a/src/cmd/compile/internal/gc/closure.go
+++ b/src/cmd/compile/internal/gc/closure.go
@@ -108,7 +108,17 @@ func typecheckclosure(clo *Node, top int) {
 
 	xfunc.Func.Nname.Sym = closurename(Curfn)
 	disableExport(xfunc.Func.Nname.Sym)
-	declare(xfunc.Func.Nname, PFUNC)
+	if xfunc.Func.Nname.Sym.Def != nil {
+		// The only case we can reach here is when the outer function was redeclared.
+		// In that case, don't bother to redeclare the closure. Otherwise, we will get
+		// a spurious error message, see #17758. While we are here, double check that
+		// we already reported other error.
+		if nsavederrors+nerrors == 0 {
+			Fatalf("unexpected symbol collision %v", xfunc.Func.Nname.Sym)
+		}
+	} else {
+		declare(xfunc.Func.Nname, PFUNC)
+	}
 	xfunc = typecheck(xfunc, ctxStmt)
 
 	// Type check the body now, but only if we're inside a function.
@@ -526,7 +536,7 @@ func walkpartialcall(n *Node, init *Nodes) *Node {
 	// Create closure in the form of a composite literal.
 	// For x.M with receiver (x) type T, the generated code looks like:
 	//
-	//	clos = &struct{F uintptr; R T}{M.T·f, x}
+	//	clos = &struct{F uintptr; R T}{T.M·f, x}
 	//
 	// Like walkclosure above.
 
diff --git a/src/cmd/compile/internal/gc/esc.go b/src/cmd/compile/internal/gc/esc.go
index f3e9ab78ef..628953741a 100644
--- a/src/cmd/compile/internal/gc/esc.go
+++ b/src/cmd/compile/internal/gc/esc.go
@@ -187,6 +187,13 @@ func mustHeapAlloc(n *Node) bool {
 		return true
 	}
 
+	if n.Op == OCLOSURE && closureType(n).Size() >= maxImplicitStackVarSize {
+		return true
+	}
+	if n.Op == OCALLPART && partialCallType(n).Size() >= maxImplicitStackVarSize {
+		return true
+	}
+
 	if n.Op == OMAKESLICE && !isSmallMakeSlice(n) {
 		return true
 	}
diff --git a/src/cmd/compile/internal/gc/iimport.go b/src/cmd/compile/internal/gc/iimport.go
index 0eeb047c06..4169222c14 100644
--- a/src/cmd/compile/internal/gc/iimport.go
+++ b/src/cmd/compile/internal/gc/iimport.go
@@ -191,9 +191,9 @@ func iimport(pkg *types.Pkg, in *bio.Reader) (fingerprint goobj.FingerprintType)
 		}
 	}
 
-	// Fingerprint
-	n, err := io.ReadFull(in, fingerprint[:])
-	if err != nil || n != len(fingerprint) {
+	// Fingerprint.
+	_, err = io.ReadFull(in, fingerprint[:])
+	if err != nil {
 		yyerror("import %s: error reading fingerprint", pkg.Path)
 		errorexit()
 	}
diff --git a/src/cmd/compile/internal/gc/obj.go b/src/cmd/compile/internal/gc/obj.go
index 0826b04e33..af5037c5a8 100644
--- a/src/cmd/compile/internal/gc/obj.go
+++ b/src/cmd/compile/internal/gc/obj.go
@@ -352,7 +352,7 @@ func stringsym(pos src.XPos, s string) (data *obj.LSym) {
 
 	symdata := Ctxt.Lookup(symdataname)
 
-	if !symdata.SeenGlobl() {
+	if !symdata.OnList() {
 		// string data
 		off := dsname(symdata, 0, s, pos, "string")
 		ggloblsym(symdata, int32(off), obj.DUPOK|obj.RODATA|obj.LOCAL)
diff --git a/src/cmd/compile/internal/gc/plive.go b/src/cmd/compile/internal/gc/plive.go
index b366c8a4a0..0cb2661997 100644
--- a/src/cmd/compile/internal/gc/plive.go
+++ b/src/cmd/compile/internal/gc/plive.go
@@ -140,24 +140,14 @@ type Liveness struct {
 	regMaps     []liveRegMask
 
 	cache progeffectscache
-
-	// These are only populated if open-coded defers are being used.
-	// List of vars/stack slots storing defer args
-	openDeferVars []openDeferVarInfo
-	// Map from defer arg OpVarDef to the block where the OpVarDef occurs.
-	openDeferVardefToBlockMap map[*Node]*ssa.Block
-	// Map of blocks that cannot reach a return or exit (panic)
-	nonReturnBlocks map[*ssa.Block]bool
-}
-
-type openDeferVarInfo struct {
-	n         *Node // Var/stack slot storing a defer arg
-	varsIndex int   // Index of variable in lv.vars
 }
 
 // LivenessMap maps from *ssa.Value to LivenessIndex.
 type LivenessMap struct {
 	vals map[ssa.ID]LivenessIndex
+	// The set of live, pointer-containing variables at the deferreturn
+	// call (only set when open-coded defers are used).
+	deferreturn LivenessIndex
 }
 
 func (m *LivenessMap) reset() {
@@ -168,6 +158,7 @@ func (m *LivenessMap) reset() {
 			delete(m.vals, k)
 		}
 	}
+	m.deferreturn = LivenessInvalid
 }
 
 func (m *LivenessMap) set(v *ssa.Value, i LivenessIndex) {
@@ -542,7 +533,7 @@ func newliveness(fn *Node, f *ssa.Func, vars []*Node, idx map[*Node]int32, stkpt
 		if cap(lc.be) >= f.NumBlocks() {
 			lv.be = lc.be[:f.NumBlocks()]
 		}
-		lv.livenessMap = LivenessMap{lc.livenessMap.vals}
+		lv.livenessMap = LivenessMap{vals: lc.livenessMap.vals, deferreturn: LivenessInvalid}
 		lc.livenessMap.vals = nil
 	}
 	if lv.be == nil {
@@ -893,58 +884,12 @@ func (lv *Liveness) hasStackMap(v *ssa.Value) bool {
 func (lv *Liveness) prologue() {
 	lv.initcache()
 
-	if lv.fn.Func.HasDefer() && !lv.fn.Func.OpenCodedDeferDisallowed() {
-		lv.openDeferVardefToBlockMap = make(map[*Node]*ssa.Block)
-		for i, n := range lv.vars {
-			if n.Name.OpenDeferSlot() {
-				lv.openDeferVars = append(lv.openDeferVars, openDeferVarInfo{n: n, varsIndex: i})
-			}
-		}
-
-		// Find any blocks that cannot reach a return or a BlockExit
-		// (panic) -- these must be because of an infinite loop.
-		reachesRet := make(map[ssa.ID]bool)
-		blockList := make([]*ssa.Block, 0, 256)
-
-		for _, b := range lv.f.Blocks {
-			if b.Kind == ssa.BlockRet || b.Kind == ssa.BlockRetJmp || b.Kind == ssa.BlockExit {
-				blockList = append(blockList, b)
-			}
-		}
-
-		for len(blockList) > 0 {
-			b := blockList[0]
-			blockList = blockList[1:]
-			if reachesRet[b.ID] {
-				continue
-			}
-			reachesRet[b.ID] = true
-			for _, e := range b.Preds {
-				blockList = append(blockList, e.Block())
-			}
-		}
-
-		lv.nonReturnBlocks = make(map[*ssa.Block]bool)
-		for _, b := range lv.f.Blocks {
-			if !reachesRet[b.ID] {
-				lv.nonReturnBlocks[b] = true
-				//fmt.Println("No reach ret", lv.f.Name, b.ID, b.Kind)
-			}
-		}
-	}
-
 	for _, b := range lv.f.Blocks {
 		be := lv.blockEffects(b)
 
 		// Walk the block instructions backward and update the block
 		// effects with the each prog effects.
 		for j := len(b.Values) - 1; j >= 0; j-- {
-			if b.Values[j].Op == ssa.OpVarDef {
-				n := b.Values[j].Aux.(*Node)
-				if n.Name.OpenDeferSlot() {
-					lv.openDeferVardefToBlockMap[n] = b
-				}
-			}
 			pos, e := lv.valueEffects(b.Values[j])
 			regUevar, regKill := lv.regEffects(b.Values[j])
 			if e&varkill != 0 {
@@ -961,20 +906,6 @@ func (lv *Liveness) prologue() {
 	}
 }
 
-// markDeferVarsLive marks each variable storing an open-coded defer arg as
-// specially live in block b if the variable definition dominates block b.
-func (lv *Liveness) markDeferVarsLive(b *ssa.Block, newliveout *varRegVec) {
-	// Only force computation of dominators if we have a block where we need
-	// to specially mark defer args live.
-	sdom := lv.f.Sdom()
-	for _, info := range lv.openDeferVars {
-		defB := lv.openDeferVardefToBlockMap[info.n]
-		if sdom.IsAncestorEq(defB, b) {
-			newliveout.vars.Set(int32(info.varsIndex))
-		}
-	}
-}
-
 // Solve the liveness dataflow equations.
 func (lv *Liveness) solve() {
 	// These temporary bitvectors exist to avoid successive allocations and
@@ -1018,23 +949,6 @@ func (lv *Liveness) solve() {
 				}
 			}
 
-			if lv.fn.Func.HasDefer() && !lv.fn.Func.OpenCodedDeferDisallowed() &&
-				(b.Kind == ssa.BlockExit || lv.nonReturnBlocks[b]) {
-				// Open-coded defer args slots must be live
-				// everywhere in a function, since a panic can
-				// occur (almost) anywhere. Force all appropriate
-				// defer arg slots to be live in BlockExit (panic)
-				// blocks and in blocks that do not reach a return
-				// (because of infinite loop).
-				//
-				// We are assuming that the defer exit code at
-				// BlockReturn/BlockReturnJmp accesses all of the
-				// defer args (with pointers), and so keeps them
-				// live. This analysis may have to be adjusted if
-				// that changes (because of optimizations).
-				lv.markDeferVarsLive(b, &newliveout)
-			}
-
 			if !be.liveout.Eq(newliveout) {
 				change = true
 				be.liveout.Copy(newliveout)
@@ -1087,6 +1001,17 @@ func (lv *Liveness) epilogue() {
 				n.Name.SetNeedzero(true)
 				livedefer.Set(int32(i))
 			}
+			if n.Name.OpenDeferSlot() {
+				// Open-coded defer args slots must be live
+				// everywhere in a function, since a panic can
+				// occur (almost) anywhere. Because it is live
+				// everywhere, it must be zeroed on entry.
+				livedefer.Set(int32(i))
+				// It was already marked as Needzero when created.
+				if !n.Name.Needzero() {
+					Fatalf("all pointer-containing defer arg slots should have Needzero set")
+				}
+			}
 		}
 	}
 
@@ -1188,6 +1113,17 @@ func (lv *Liveness) epilogue() {
 		lv.compact(b)
 	}
 
+	// If we have an open-coded deferreturn call, make a liveness map for it.
+	if lv.fn.Func.OpenCodedDeferDisallowed() {
+		lv.livenessMap.deferreturn = LivenessInvalid
+	} else {
+		lv.livenessMap.deferreturn = LivenessIndex{
+			stackMapIndex: lv.stackMapSet.add(livedefer),
+			regMapIndex:   0, // entry regMap, containing no live registers
+			isUnsafePoint: false,
+		}
+	}
+
 	// Done compacting. Throw out the stack map set.
 	lv.stackMaps = lv.stackMapSet.extractUniqe()
 	lv.stackMapSet = bvecSet{}
diff --git a/src/cmd/compile/internal/gc/racewalk.go b/src/cmd/compile/internal/gc/racewalk.go
index 6f251377c9..3552617401 100644
--- a/src/cmd/compile/internal/gc/racewalk.go
+++ b/src/cmd/compile/internal/gc/racewalk.go
@@ -42,7 +42,7 @@ var omit_pkgs = []string{
 	"internal/cpu",
 }
 
-// Only insert racefuncenterfp/racefuncexit into the following packages.
+// Don't insert racefuncenterfp/racefuncexit into the following packages.
 // Memory accesses in the packages are either uninteresting or will cause false positives.
 var norace_inst_pkgs = []string{"sync", "sync/atomic"}
 
diff --git a/src/cmd/compile/internal/gc/select.go b/src/cmd/compile/internal/gc/select.go
index 49cc23cd3d..3812a0e1fa 100644
--- a/src/cmd/compile/internal/gc/select.go
+++ b/src/cmd/compile/internal/gc/select.go
@@ -106,18 +106,16 @@ func walkselect(sel *Node) {
 }
 
 func walkselectcases(cases *Nodes) []*Node {
-	n := cases.Len()
+	ncas := cases.Len()
 	sellineno := lineno
 
 	// optimization: zero-case select
-	if n == 0 {
+	if ncas == 0 {
 		return []*Node{mkcall("block", nil, nil)}
 	}
 
 	// optimization: one-case select: single op.
-	// TODO(rsc): Reenable optimization once order.go can handle it.
-	// golang.org/issue/7672.
-	if n == 1 {
+	if ncas == 1 {
 		cas := cases.First()
 		setlineno(cas)
 		l := cas.Ninit.Slice()
@@ -125,17 +123,14 @@ func walkselectcases(cases *Nodes) []*Node {
 			n := cas.Left
 			l = append(l, n.Ninit.Slice()...)
 			n.Ninit.Set(nil)
-			var ch *Node
 			switch n.Op {
 			default:
 				Fatalf("select %v", n.Op)
 
-				// ok already
 			case OSEND:
-				ch = n.Left
+				// already ok
 
 			case OSELRECV, OSELRECV2:
-				ch = n.Right.Left
 				if n.Op == OSELRECV || n.List.Len() == 0 {
 					if n.Left == nil {
 						n = n.Right
@@ -159,16 +154,7 @@ func walkselectcases(cases *Nodes) []*Node {
 				n = typecheck(n, ctxStmt)
 			}
 
-			// if ch == nil { block() }; n;
-			a := nod(OIF, nil, nil)
-
-			a.Left = nod(OEQ, ch, nodnil())
-			var ln Nodes
-			ln.Set(l)
-			a.Nbody.Set1(mkcall("block", nil, &ln))
-			l = ln.Slice()
-			a = typecheck(a, ctxStmt)
-			l = append(l, a, n)
+			l = append(l, n)
 		}
 
 		l = append(l, cas.Nbody.Slice()...)
@@ -178,10 +164,12 @@ func walkselectcases(cases *Nodes) []*Node {
 
 	// convert case value arguments to addresses.
 	// this rewrite is used by both the general code and the next optimization.
+	var dflt *Node
 	for _, cas := range cases.Slice() {
 		setlineno(cas)
 		n := cas.Left
 		if n == nil {
+			dflt = cas
 			continue
 		}
 		switch n.Op {
@@ -202,15 +190,10 @@ func walkselectcases(cases *Nodes) []*Node {
 	}
 
 	// optimization: two-case select but one is default: single non-blocking op.
-	if n == 2 && (cases.First().Left == nil || cases.Second().Left == nil) {
-		var cas *Node
-		var dflt *Node
-		if cases.First().Left == nil {
+	if ncas == 2 && dflt != nil {
+		cas := cases.First()
+		if cas == dflt {
 			cas = cases.Second()
-			dflt = cases.First()
-		} else {
-			dflt = cases.Second()
-			cas = cases.First()
 		}
 
 		n := cas.Left
@@ -228,8 +211,6 @@ func walkselectcases(cases *Nodes) []*Node {
 
 		case OSELRECV:
 			// if selectnbrecv(&v, c) { body } else { default body }
-			r = nod(OIF, nil, nil)
-			r.Ninit.Set(cas.Ninit.Slice())
 			ch := n.Right.Left
 			elem := n.Left
 			if elem == nil {
@@ -239,8 +220,6 @@ func walkselectcases(cases *Nodes) []*Node {
 
 		case OSELRECV2:
 			// if selectnbrecv2(&v, &received, c) { body } else { default body }
-			r = nod(OIF, nil, nil)
-			r.Ninit.Set(cas.Ninit.Slice())
 			ch := n.Right.Left
 			elem := n.Left
 			if elem == nil {
@@ -257,66 +236,73 @@ func walkselectcases(cases *Nodes) []*Node {
 		return []*Node{r, nod(OBREAK, nil, nil)}
 	}
 
+	if dflt != nil {
+		ncas--
+	}
+	casorder := make([]*Node, ncas)
+	nsends, nrecvs := 0, 0
+
 	var init []*Node
 
 	// generate sel-struct
 	lineno = sellineno
-	selv := temp(types.NewArray(scasetype(), int64(n)))
+	selv := temp(types.NewArray(scasetype(), int64(ncas)))
 	r := nod(OAS, selv, nil)
 	r = typecheck(r, ctxStmt)
 	init = append(init, r)
 
-	order := temp(types.NewArray(types.Types[TUINT16], 2*int64(n)))
+	order := temp(types.NewArray(types.Types[TUINT16], 2*int64(ncas)))
 	r = nod(OAS, order, nil)
 	r = typecheck(r, ctxStmt)
 	init = append(init, r)
 
+	var pc0, pcs *Node
+	if flag_race {
+		pcs = temp(types.NewArray(types.Types[TUINTPTR], int64(ncas)))
+		pc0 = typecheck(nod(OADDR, nod(OINDEX, pcs, nodintconst(0)), nil), ctxExpr)
+	} else {
+		pc0 = nodnil()
+	}
+
 	// register cases
-	for i, cas := range cases.Slice() {
+	for _, cas := range cases.Slice() {
 		setlineno(cas)
 
 		init = append(init, cas.Ninit.Slice()...)
 		cas.Ninit.Set(nil)
 
-		// Keep in sync with runtime/select.go.
-		const (
-			caseNil = iota
-			caseRecv
-			caseSend
-			caseDefault
-		)
+		n := cas.Left
+		if n == nil { // default:
+			continue
+		}
 
+		var i int
 		var c, elem *Node
-		var kind int64 = caseDefault
-
-		if n := cas.Left; n != nil {
-			init = append(init, n.Ninit.Slice()...)
-
-			switch n.Op {
-			default:
-				Fatalf("select %v", n.Op)
-			case OSEND:
-				kind = caseSend
-				c = n.Left
-				elem = n.Right
-			case OSELRECV, OSELRECV2:
-				kind = caseRecv
-				c = n.Right.Left
-				elem = n.Left
-			}
+		switch n.Op {
+		default:
+			Fatalf("select %v", n.Op)
+		case OSEND:
+			i = nsends
+			nsends++
+			c = n.Left
+			elem = n.Right
+		case OSELRECV, OSELRECV2:
+			nrecvs++
+			i = ncas - nrecvs
+			c = n.Right.Left
+			elem = n.Left
 		}
 
+		casorder[i] = cas
+
 		setField := func(f string, val *Node) {
 			r := nod(OAS, nodSym(ODOT, nod(OINDEX, selv, nodintconst(int64(i))), lookup(f)), val)
 			r = typecheck(r, ctxStmt)
 			init = append(init, r)
 		}
 
-		setField("kind", nodintconst(kind))
-		if c != nil {
-			c = convnop(c, types.Types[TUNSAFEPTR])
-			setField("c", c)
-		}
+		c = convnop(c, types.Types[TUNSAFEPTR])
+		setField("c", c)
 		if elem != nil {
 			elem = convnop(elem, types.Types[TUNSAFEPTR])
 			setField("elem", elem)
@@ -324,11 +310,14 @@ func walkselectcases(cases *Nodes) []*Node {
 
 		// TODO(mdempsky): There should be a cleaner way to
 		// handle this.
-		if instrumenting {
-			r = mkcall("selectsetpc", nil, nil, bytePtrToIndex(selv, int64(i)))
+		if flag_race {
+			r = mkcall("selectsetpc", nil, nil, nod(OADDR, nod(OINDEX, pcs, nodintconst(int64(i))), nil))
 			init = append(init, r)
 		}
 	}
+	if nsends+nrecvs != ncas {
+		Fatalf("walkselectcases: miscount: %v + %v != %v", nsends, nrecvs, ncas)
+	}
 
 	// run the select
 	lineno = sellineno
@@ -337,23 +326,23 @@ func walkselectcases(cases *Nodes) []*Node {
 	r = nod(OAS2, nil, nil)
 	r.List.Set2(chosen, recvOK)
 	fn := syslook("selectgo")
-	r.Rlist.Set1(mkcall1(fn, fn.Type.Results(), nil, bytePtrToIndex(selv, 0), bytePtrToIndex(order, 0), nodintconst(int64(n))))
+	r.Rlist.Set1(mkcall1(fn, fn.Type.Results(), nil, bytePtrToIndex(selv, 0), bytePtrToIndex(order, 0), pc0, nodintconst(int64(nsends)), nodintconst(int64(nrecvs)), nodbool(dflt == nil)))
 	r = typecheck(r, ctxStmt)
 	init = append(init, r)
 
 	// selv and order are no longer alive after selectgo.
 	init = append(init, nod(OVARKILL, selv, nil))
 	init = append(init, nod(OVARKILL, order, nil))
+	if flag_race {
+		init = append(init, nod(OVARKILL, pcs, nil))
+	}
 
 	// dispatch cases
-	for i, cas := range cases.Slice() {
-		setlineno(cas)
-
-		cond := nod(OEQ, chosen, nodintconst(int64(i)))
+	dispatch := func(cond, cas *Node) {
 		cond = typecheck(cond, ctxExpr)
 		cond = defaultlit(cond, nil)
 
-		r = nod(OIF, cond, nil)
+		r := nod(OIF, cond, nil)
 
 		if n := cas.Left; n != nil && n.Op == OSELRECV2 {
 			x := nod(OAS, n.List.First(), recvOK)
@@ -366,6 +355,15 @@ func walkselectcases(cases *Nodes) []*Node {
 		init = append(init, r)
 	}
 
+	if dflt != nil {
+		setlineno(dflt)
+		dispatch(nod(OLT, chosen, nodintconst(0)), dflt)
+	}
+	for i, cas := range casorder {
+		setlineno(cas)
+		dispatch(nod(OEQ, chosen, nodintconst(int64(i))), cas)
+	}
+
 	return init
 }
 
@@ -384,9 +382,6 @@ func scasetype() *types.Type {
 		scase = tostruct([]*Node{
 			namedfield("c", types.Types[TUNSAFEPTR]),
 			namedfield("elem", types.Types[TUNSAFEPTR]),
-			namedfield("kind", types.Types[TUINT16]),
-			namedfield("pc", types.Types[TUINTPTR]),
-			namedfield("releasetime", types.Types[TINT64]),
 		})
 		scase.SetNoalg(true)
 	}
diff --git a/src/cmd/compile/internal/gc/sinit.go b/src/cmd/compile/internal/gc/sinit.go
index 4a2edc7d21..71ed558461 100644
--- a/src/cmd/compile/internal/gc/sinit.go
+++ b/src/cmd/compile/internal/gc/sinit.go
@@ -506,6 +506,7 @@ const (
 // fixedlit handles struct, array, and slice literals.
 // TODO: expand documentation.
 func fixedlit(ctxt initContext, kind initKind, n *Node, var_ *Node, init *Nodes) {
+	isBlank := var_ == nblank
 	var splitnode func(*Node) (a *Node, value *Node)
 	switch n.Op {
 	case OARRAYLIT, OSLICELIT:
@@ -520,6 +521,9 @@ func fixedlit(ctxt initContext, kind initKind, n *Node, var_ *Node, init *Nodes)
 			}
 			a := nod(OINDEX, var_, nodintconst(k))
 			k++
+			if isBlank {
+				a = nblank
+			}
 			return a, r
 		}
 	case OSTRUCTLIT:
@@ -527,7 +531,7 @@ func fixedlit(ctxt initContext, kind initKind, n *Node, var_ *Node, init *Nodes)
 			if r.Op != OSTRUCTKEY {
 				Fatalf("fixedlit: rhs not OSTRUCTKEY: %v", r)
 			}
-			if r.Sym.IsBlank() {
+			if r.Sym.IsBlank() || isBlank {
 				return nblank, r.Left
 			}
 			setlineno(r)
diff --git a/src/cmd/compile/internal/gc/ssa.go b/src/cmd/compile/internal/gc/ssa.go
index d4d23a2956..4124655b79 100644
--- a/src/cmd/compile/internal/gc/ssa.go
+++ b/src/cmd/compile/internal/gc/ssa.go
@@ -338,6 +338,10 @@ func buildssa(fn *Node, worker int) *ssa.Func {
 	s.panics = map[funcLine]*ssa.Block{}
 	s.softFloat = s.config.SoftFloat
 
+	// Allocate starting block
+	s.f.Entry = s.f.NewBlock(ssa.BlockPlain)
+	s.f.Entry.Pos = fn.Pos
+
 	if printssa {
 		s.f.HTMLWriter = ssa.NewHTMLWriter(ssaDumpFile, s.f, ssaDumpCFG)
 		// TODO: generate and print a mapping from nodes to values and blocks
@@ -345,9 +349,6 @@ func buildssa(fn *Node, worker int) *ssa.Func {
 		s.f.HTMLWriter.WriteAST("AST", astBuf)
 	}
 
-	// Allocate starting block
-	s.f.Entry = s.f.NewBlock(ssa.BlockPlain)
-
 	// Allocate starting values
 	s.labels = map[string]*ssaLabel{}
 	s.labeledNodes = map[*Node]*ssaLabel{}
@@ -4318,12 +4319,6 @@ func (s *state) openDeferExit() {
 			}
 		}
 
-		if i == len(s.openDefers)-1 {
-			// Record the call of the first defer. This will be used
-			// to set liveness info for the deferreturn (which is also
-			// used for any location that causes a runtime panic)
-			s.f.LastDeferExit = call
-		}
 		s.endBlock()
 		s.startBlock(bEnd)
 	}
@@ -5807,11 +5802,6 @@ type SSAGenState struct {
 
 	// wasm: The number of values on the WebAssembly stack. This is only used as a safeguard.
 	OnWasmStackSkipped int
-
-	// Liveness index for the first function call in the final defer exit code
-	// path that we generated. All defer functions and args should be live at
-	// this point. This will be used to set the liveness for the deferreturn.
-	lastDeferLiveness LivenessIndex
 }
 
 // Prog appends a new Prog.
@@ -6056,12 +6046,6 @@ func genssa(f *ssa.Func, pp *Progs) {
 				// instruction.
 				s.pp.nextLive = s.livenessMap.Get(v)
 
-				// Remember the liveness index of the first defer call of
-				// the last defer exit
-				if v.Block.Func.LastDeferExit != nil && v == v.Block.Func.LastDeferExit {
-					s.lastDeferLiveness = s.pp.nextLive
-				}
-
 				// Special case for first line in function; move it to the start.
 				if firstPos != src.NoXPos {
 					s.SetPos(firstPos)
@@ -6122,7 +6106,7 @@ func genssa(f *ssa.Func, pp *Progs) {
 		// When doing open-coded defers, generate a disconnected call to
 		// deferreturn and a return. This will be used to during panic
 		// recovery to unwind the stack and return back to the runtime.
-		s.pp.nextLive = s.lastDeferLiveness
+		s.pp.nextLive = s.livenessMap.deferreturn
 		gencallret(pp, Deferreturn)
 	}
 
diff --git a/src/cmd/compile/internal/gc/syntax.go b/src/cmd/compile/internal/gc/syntax.go
index b658410c53..47e5e59156 100644
--- a/src/cmd/compile/internal/gc/syntax.go
+++ b/src/cmd/compile/internal/gc/syntax.go
@@ -141,8 +141,8 @@ const (
 	nodeInitorder, _                   // tracks state during init1; two bits
 	_, _                               // second nodeInitorder bit
 	_, nodeHasBreak
-	_, nodeNoInline // used internally by inliner to indicate that a function call should not be inlined; set for OCALLFUNC and OCALLMETH only
-	_, nodeImplicit
+	_, nodeNoInline  // used internally by inliner to indicate that a function call should not be inlined; set for OCALLFUNC and OCALLMETH only
+	_, nodeImplicit  // implicit OADDR or ODEREF; ++/-- statement represented as OASOP; or ANDNOT lowered to OAND
 	_, nodeIsDDD     // is the argument variadic
 	_, nodeDiag      // already printed error about this
 	_, nodeColas     // OAS resulting from :=
diff --git a/src/cmd/compile/internal/gc/walk.go b/src/cmd/compile/internal/gc/walk.go
index 8ae3d9a5c7..74ed0411bd 100644
--- a/src/cmd/compile/internal/gc/walk.go
+++ b/src/cmd/compile/internal/gc/walk.go
@@ -973,6 +973,7 @@ opswitch:
 	case OANDNOT:
 		n.Left = walkexpr(n.Left, init)
 		n.Op = OAND
+		n.SetImplicit(true) // for walkCheckPtrArithmetic
 		n.Right = nod(OBITNOT, n.Right, nil)
 		n.Right = typecheck(n.Right, ctxExpr)
 		n.Right = walkexpr(n.Right, init)
@@ -4003,8 +4004,12 @@ func walkCheckPtrArithmetic(n *Node, init *Nodes) *Node {
 		case OADD:
 			walk(n.Left)
 			walk(n.Right)
-		case OSUB, OANDNOT:
+		case OSUB:
 			walk(n.Left)
+		case OAND:
+			if n.Implicit() { // was OANDNOT
+				walk(n.Left)
+			}
 		case OCONVNOP:
 			if n.Left.Type.Etype == TUNSAFEPTR {
 				n.Left = cheapexpr(n.Left, init)
diff --git a/src/cmd/compile/internal/ppc64/ssa.go b/src/cmd/compile/internal/ppc64/ssa.go
index 0efdd710fb..4d2ad48135 100644
--- a/src/cmd/compile/internal/ppc64/ssa.go
+++ b/src/cmd/compile/internal/ppc64/ssa.go
@@ -601,6 +601,20 @@ func ssaGenValue(s *gc.SSAGenState, v *ssa.Value) {
 		p.To.Type = obj.TYPE_REG
 		p.To.Reg = v.Reg()
 
+	case ssa.OpPPC64MADDLD:
+		r := v.Reg()
+		r1 := v.Args[0].Reg()
+		r2 := v.Args[1].Reg()
+		r3 := v.Args[2].Reg()
+		// r = r1*r2 ± r3
+		p := s.Prog(v.Op.Asm())
+		p.From.Type = obj.TYPE_REG
+		p.From.Reg = r1
+		p.Reg = r2
+		p.SetFrom3(obj.Addr{Type: obj.TYPE_REG, Reg: r3})
+		p.To.Type = obj.TYPE_REG
+		p.To.Reg = r
+
 	case ssa.OpPPC64FMADD, ssa.OpPPC64FMADDS, ssa.OpPPC64FMSUB, ssa.OpPPC64FMSUBS:
 		r := v.Reg()
 		r1 := v.Args[0].Reg()
diff --git a/src/cmd/compile/internal/ssa/compile.go b/src/cmd/compile/internal/ssa/compile.go
index dbdd027716..444475d67a 100644
--- a/src/cmd/compile/internal/ssa/compile.go
+++ b/src/cmd/compile/internal/ssa/compile.go
@@ -160,15 +160,12 @@ func Compile(f *Func) {
 	phaseName = ""
 }
 
-// TODO: should be a config field
-var dumpFileSeq int
-
 // dumpFile creates a file from the phase name and function name
 // Dumping is done to files to avoid buffering huge strings before
 // output.
 func (f *Func) dumpFile(phaseName string) {
-	dumpFileSeq++
-	fname := fmt.Sprintf("%s_%02d__%s.dump", f.Name, dumpFileSeq, phaseName)
+	f.dumpFileSeq++
+	fname := fmt.Sprintf("%s_%02d__%s.dump", f.Name, int(f.dumpFileSeq), phaseName)
 	fname = strings.Replace(fname, " ", "_", -1)
 	fname = strings.Replace(fname, "/", "_", -1)
 	fname = strings.Replace(fname, ":", "_", -1)
diff --git a/src/cmd/compile/internal/ssa/debug.go b/src/cmd/compile/internal/ssa/debug.go
index 13fe67cbca..6353f72897 100644
--- a/src/cmd/compile/internal/ssa/debug.go
+++ b/src/cmd/compile/internal/ssa/debug.go
@@ -1,6 +1,7 @@
 // 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 ssa
 
 import (
diff --git a/src/cmd/compile/internal/ssa/func.go b/src/cmd/compile/internal/ssa/func.go
index 7cf72a8e37..9e40b6214c 100644
--- a/src/cmd/compile/internal/ssa/func.go
+++ b/src/cmd/compile/internal/ssa/func.go
@@ -33,15 +33,8 @@ type Func struct {
 	Blocks []*Block    // unordered set of all basic blocks (note: not indexable by ID)
 	Entry  *Block      // the entry basic block
 
-	// If we are using open-coded defers, this is the first call to a deferred
-	// function in the final defer exit sequence that we generated. This call
-	// should be after all defer statements, and will have all args, etc. of
-	// all defer calls as live. The liveness info of this call will be used
-	// for the deferreturn/ret segment generated for functions with open-coded
-	// defers.
-	LastDeferExit *Value
-	bid           idAlloc // block ID allocator
-	vid           idAlloc // value ID allocator
+	bid idAlloc // block ID allocator
+	vid idAlloc // value ID allocator
 
 	// Given an environment variable used for debug hash match,
 	// what file (if any) receives the yes/no logging?
@@ -51,9 +44,10 @@ type Func struct {
 	PrintOrHtmlSSA bool           // true if GOSSAFUNC matches, true even if fe.Log() (spew phase results to stdout) is false.
 	ruleMatches    map[string]int // number of times countRule was called during compilation for any given string
 
-	scheduled bool // Values in Blocks are in final order
-	laidout   bool // Blocks are ordered
-	NoSplit   bool // true if function is marked as nosplit.  Used by schedule check pass.
+	scheduled   bool  // Values in Blocks are in final order
+	laidout     bool  // Blocks are ordered
+	NoSplit     bool  // true if function is marked as nosplit.  Used by schedule check pass.
+	dumpFileSeq uint8 // the sequence numbers of dump file. (%s_%02d__%s.dump", funcname, dumpFileSeq, phaseName)
 
 	// when register allocation is done, maps value ids to locations
 	RegAlloc []Location
diff --git a/src/cmd/compile/internal/ssa/gen/AMD64.rules b/src/cmd/compile/internal/ssa/gen/AMD64.rules
index 9967c7b030..5111ef79d3 100644
--- a/src/cmd/compile/internal/ssa/gen/AMD64.rules
+++ b/src/cmd/compile/internal/ssa/gen/AMD64.rules
@@ -1463,14 +1463,14 @@
 (MULQconst [c] (NEGQ x)) && c != -(1<<31) -> (MULQconst [-c] x)
 
 // checking AND against 0.
-(CMPQconst (ANDQ x y) [0]) -> (TESTQ x y)
-(CMPLconst (ANDL x y) [0]) -> (TESTL x y)
-(CMPWconst (ANDL x y) [0]) -> (TESTW x y)
-(CMPBconst (ANDL x y) [0]) -> (TESTB x y)
-(CMPQconst (ANDQconst [c] x) [0]) -> (TESTQconst [c] x)
-(CMPLconst (ANDLconst [c] x) [0]) -> (TESTLconst [c] x)
-(CMPWconst (ANDLconst [c] x) [0]) -> (TESTWconst [int64(int16(c))] x)
-(CMPBconst (ANDLconst [c] x) [0]) -> (TESTBconst [int64(int8(c))] x)
+(CMPQconst a:(ANDQ x y) [0]) && a.Uses == 1 -> (TESTQ x y)
+(CMPLconst a:(ANDL x y) [0]) && a.Uses == 1 -> (TESTL x y)
+(CMPWconst a:(ANDL x y) [0]) && a.Uses == 1 -> (TESTW x y)
+(CMPBconst a:(ANDL x y) [0]) && a.Uses == 1 -> (TESTB x y)
+(CMPQconst a:(ANDQconst [c] x) [0]) && a.Uses == 1 -> (TESTQconst [c] x)
+(CMPLconst a:(ANDLconst [c] x) [0]) && a.Uses == 1 -> (TESTLconst [c] x)
+(CMPWconst a:(ANDLconst [c] x) [0]) && a.Uses == 1 -> (TESTWconst [int64(int16(c))] x)
+(CMPBconst a:(ANDLconst [c] x) [0]) && a.Uses == 1 -> (TESTBconst [int64(int8(c))] x)
 
 // Convert TESTx to TESTxconst if possible.
 (TESTQ (MOVQconst [c]) x) && is32Bit(c) -> (TESTQconst [c] x)
diff --git a/src/cmd/compile/internal/ssa/gen/AMD64Ops.go b/src/cmd/compile/internal/ssa/gen/AMD64Ops.go
index a3b29049df..e6d66957dd 100644
--- a/src/cmd/compile/internal/ssa/gen/AMD64Ops.go
+++ b/src/cmd/compile/internal/ssa/gen/AMD64Ops.go
@@ -748,6 +748,7 @@ func init() {
 				clobbers: buildReg("DI"),
 			},
 			faultOnNilArg0: true,
+			unsafePoint:    true, // FP maintenance around DUFFCOPY can be clobbered by interrupts
 		},
 		{name: "MOVOconst", reg: regInfo{nil, 0, []regMask{fp}}, typ: "Int128", aux: "Int128", rematerializeable: true},
 
@@ -786,6 +787,7 @@ func init() {
 			clobberFlags:   true,
 			faultOnNilArg0: true,
 			faultOnNilArg1: true,
+			unsafePoint:    true, // FP maintenance around DUFFCOPY can be clobbered by interrupts
 		},
 
 		// arg0 = destination pointer
diff --git a/src/cmd/compile/internal/ssa/gen/ARM64.rules b/src/cmd/compile/internal/ssa/gen/ARM64.rules
index 442d769fdd..80e8c7137b 100644
--- a/src/cmd/compile/internal/ssa/gen/ARM64.rules
+++ b/src/cmd/compile/internal/ssa/gen/ARM64.rules
@@ -279,6 +279,16 @@
 (Less32F x y) => (LessThanF (FCMPS x y))
 (Less64F x y) => (LessThanF (FCMPD x y))
 
+// For an unsigned integer x, the following rules are useful when combining branch
+// 0 <  x  =>  x != 0
+// x <= 0  =>  x == 0
+// x <  1  =>  x == 0
+// 1 <= x  =>  x != 0
+(Less(8U|16U|32U|64U) zero:(MOVDconst [0]) x) => (Neq(8|16|32|64) zero x)
+(Leq(8U|16U|32U|64U) x zero:(MOVDconst [0]))  => (Eq(8|16|32|64) x zero)
+(Less(8U|16U|32U|64U) x (MOVDconst [1])) => (Eq(8|16|32|64) x (MOVDconst [0]))
+(Leq(8U|16U|32U|64U) (MOVDconst [1]) x)  => (Neq(8|16|32|64) (MOVDconst [0]) x)
+
 (Less8U x y)  => (LessThanU (CMPW (ZeroExt8to32 x) (ZeroExt8to32 y)))
 (Less16U x y) => (LessThanU (CMPW (ZeroExt16to32 x) (ZeroExt16to32 y)))
 (Less32U x y) => (LessThanU (CMPW x y))
@@ -1313,6 +1323,7 @@
 (AND x (MVN y)) -> (BIC x y)
 (XOR x (MVN y)) -> (EON x y)
 (OR  x (MVN y)) -> (ORN x y)
+(MVN (XOR x y)) -> (EON x y)
 (CSEL {cc} x (MOVDconst [0]) flag) -> (CSEL0 {cc} x flag)
 (CSEL {cc} (MOVDconst [0]) y flag) -> (CSEL0 {arm64Negate(cc.(Op))} y flag)
 (SUB x (SUB y z)) -> (SUB (ADD <v.Type> x z) y)
diff --git a/src/cmd/compile/internal/ssa/gen/ARM64Ops.go b/src/cmd/compile/internal/ssa/gen/ARM64Ops.go
index b402e35ea6..2424e67e20 100644
--- a/src/cmd/compile/internal/ssa/gen/ARM64Ops.go
+++ b/src/cmd/compile/internal/ssa/gen/ARM64Ops.go
@@ -507,6 +507,7 @@ func init() {
 				clobbers: buildReg("R20 R30"),
 			},
 			faultOnNilArg0: true,
+			unsafePoint:    true, // FP maintenance around DUFFZERO can be clobbered by interrupts
 		},
 
 		// large zeroing
@@ -547,6 +548,7 @@ func init() {
 			},
 			faultOnNilArg0: true,
 			faultOnNilArg1: true,
+			unsafePoint:    true, // FP maintenance around DUFFCOPY can be clobbered by interrupts
 		},
 
 		// large move
diff --git a/src/cmd/compile/internal/ssa/gen/PPC64.rules b/src/cmd/compile/internal/ssa/gen/PPC64.rules
index fd28e10098..14942d50f9 100644
--- a/src/cmd/compile/internal/ssa/gen/PPC64.rules
+++ b/src/cmd/compile/internal/ssa/gen/PPC64.rules
@@ -11,6 +11,9 @@
 (Sub32F ...) => (FSUBS ...)
 (Sub64F ...) => (FSUB ...)
 
+// Combine 64 bit integer multiply and adds
+(ADD l:(MULLD x y) z) && objabi.GOPPC64 >= 9 && l.Uses == 1 && clobber(l) => (MADDLD x y z)
+
 (Mod16 x y) => (Mod32 (SignExt16to32 x) (SignExt16to32 y))
 (Mod16u x y) => (Mod32u (ZeroExt16to32 x) (ZeroExt16to32 y))
 (Mod8 x y) => (Mod32 (SignExt8to32 x) (SignExt8to32 y))
diff --git a/src/cmd/compile/internal/ssa/gen/PPC64Ops.go b/src/cmd/compile/internal/ssa/gen/PPC64Ops.go
index f8bc6cb20b..825d0faf34 100644
--- a/src/cmd/compile/internal/ssa/gen/PPC64Ops.go
+++ b/src/cmd/compile/internal/ssa/gen/PPC64Ops.go
@@ -137,6 +137,7 @@ func init() {
 		gp01        = regInfo{inputs: nil, outputs: []regMask{gp}}
 		gp11        = regInfo{inputs: []regMask{gp | sp | sb}, outputs: []regMask{gp}}
 		gp21        = regInfo{inputs: []regMask{gp | sp | sb, gp | sp | sb}, outputs: []regMask{gp}}
+		gp31        = regInfo{inputs: []regMask{gp | sp | sb, gp | sp | sb, gp | sp | sb}, outputs: []regMask{gp}}
 		gp22        = regInfo{inputs: []regMask{gp | sp | sb, gp | sp | sb}, outputs: []regMask{gp, gp}}
 		gp32        = regInfo{inputs: []regMask{gp | sp | sb, gp | sp | sb, gp | sp | sb}, outputs: []regMask{gp, gp}}
 		gp1cr       = regInfo{inputs: []regMask{gp | sp | sb}}
@@ -179,6 +180,7 @@ func init() {
 
 		{name: "MULLD", argLength: 2, reg: gp21, asm: "MULLD", typ: "Int64", commutative: true}, // arg0*arg1 (signed 64-bit)
 		{name: "MULLW", argLength: 2, reg: gp21, asm: "MULLW", typ: "Int32", commutative: true}, // arg0*arg1 (signed 32-bit)
+		{name: "MADDLD", argLength: 3, reg: gp31, asm: "MADDLD", typ: "Int64"},                  // (arg0*arg1)+arg2 (signed 64-bit)
 
 		{name: "MULHD", argLength: 2, reg: gp21, asm: "MULHD", commutative: true},   // (arg0 * arg1) >> 64, signed
 		{name: "MULHW", argLength: 2, reg: gp21, asm: "MULHW", commutative: true},   // (arg0 * arg1) >> 32, signed
@@ -645,9 +647,9 @@ func init() {
 		{name: "LoweredAtomicOr8", argLength: 3, reg: gpstore, asm: "OR", faultOnNilArg0: true, hasSideEffects: true},
 
 		// LoweredWB invokes runtime.gcWriteBarrier. arg0=destptr, arg1=srcptr, arg2=mem, aux=runtime.gcWriteBarrier
-		// It preserves R0 through R15, g, and its arguments R20 and R21,
+		// It preserves R0 through R17 (except special registers R1, R2, R11, R12, R13), g, and its arguments R20 and R21,
 		// but may clobber anything else, including R31 (REGTMP).
-		{name: "LoweredWB", argLength: 3, reg: regInfo{inputs: []regMask{buildReg("R20"), buildReg("R21")}, clobbers: (callerSave &^ buildReg("R0 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R14 R15 R20 R21 g")) | buildReg("R31")}, clobberFlags: true, aux: "Sym", symEffect: "None"},
+		{name: "LoweredWB", argLength: 3, reg: regInfo{inputs: []regMask{buildReg("R20"), buildReg("R21")}, clobbers: (callerSave &^ buildReg("R0 R3 R4 R5 R6 R7 R8 R9 R10 R14 R15 R16 R17 R20 R21 g")) | buildReg("R31")}, clobberFlags: true, aux: "Sym", symEffect: "None"},
 
 		// There are three of these functions so that they can have three different register inputs.
 		// When we check 0 <= c <= cap (A), then 0 <= b <= c (B), then 0 <= a <= b (C), we want the
diff --git a/src/cmd/compile/internal/ssa/gen/S390X.rules b/src/cmd/compile/internal/ssa/gen/S390X.rules
index d3234c1a00..5e4c436ca1 100644
--- a/src/cmd/compile/internal/ssa/gen/S390X.rules
+++ b/src/cmd/compile/internal/ssa/gen/S390X.rules
@@ -716,20 +716,40 @@
 (ANDWconst [0xFF] x) => (MOVBZreg x)
 (ANDWconst [0xFFFF] x) => (MOVHZreg x)
 
-// strength reduction
-(MULLDconst [-1] x) => (NEG x)
-(MULLDconst [0] _) => (MOVDconst [0])
-(MULLDconst [1] x) => x
-(MULLDconst [c] x) && isPowerOfTwo(c) -> (SLDconst [log2(c)] x)
-(MULLDconst [c] x) && isPowerOfTwo(c+1) && c >= 15 -> (SUB (SLDconst <v.Type> [log2(c+1)] x) x)
-(MULLDconst [c] x) && isPowerOfTwo(c-1) && c >= 17 -> (ADD (SLDconst <v.Type> [log2(c-1)] x) x)
+// Strength reduce multiplication to the sum (or difference) of two powers of two.
+//
+// Examples:
+//     5x -> 4x + 1x
+//    10x -> 8x + 2x
+//   120x -> 128x - 8x
+//  -120x -> 8x - 128x
+//
+// We know that the rightmost bit of any positive value, once isolated, must either
+// be a power of 2 (because it is a single bit) or 0 (if the original value is 0).
+// In all of these rules we use a rightmost bit calculation to determine one operand
+// for the addition or subtraction. We then just need to calculate if the other
+// operand is a valid power of 2 before we can match the rule.
+//
+// Notes:
+//   - the generic rules have already matched single powers of two so we ignore them here
+//   - isPowerOfTwo32 asserts that its argument is greater than 0
+//   - c&(c-1) = clear rightmost bit
+//   - c&^(c-1) = isolate rightmost bit
 
-(MULLWconst [-1] x) => (NEGW x)
-(MULLWconst [0] _) => (MOVDconst [0])
-(MULLWconst [1] x) => x
-(MULLWconst [c] x) && isPowerOfTwo(c) -> (SLWconst [log2(c)] x)
-(MULLWconst [c] x) && isPowerOfTwo(c+1) && c >= 15 -> (SUBW (SLWconst <v.Type> [log2(c+1)] x) x)
-(MULLWconst [c] x) && isPowerOfTwo(c-1) && c >= 17 -> (ADDW (SLWconst <v.Type> [log2(c-1)] x) x)
+// c = 2ˣ + 2ʸ => c - 2ˣ = 2ʸ
+(MULL(D|W)const <t> x [c]) && isPowerOfTwo32(c&(c-1))
+  => ((ADD|ADDW) (SL(D|W)const <t> x [int8(log32(c&(c-1)))])
+                 (SL(D|W)const <t> x [int8(log32(c&^(c-1)))]))
+
+// c = 2ʸ - 2ˣ => c + 2ˣ = 2ʸ
+(MULL(D|W)const <t> x [c]) && isPowerOfTwo32(c+(c&^(c-1)))
+  => ((SUB|SUBW) (SL(D|W)const <t> x [int8(log32(c+(c&^(c-1))))])
+                 (SL(D|W)const <t> x [int8(log32(c&^(c-1)))]))
+
+// c = 2ˣ - 2ʸ => -c + 2ˣ = 2ʸ
+(MULL(D|W)const <t> x [c]) && isPowerOfTwo32(-c+(-c&^(-c-1)))
+  => ((SUB|SUBW) (SL(D|W)const <t> x [int8(log32(-c&^(-c-1)))])
+                 (SL(D|W)const <t> x [int8(log32(-c+(-c&^(-c-1))))]))
 
 // Fold ADD into MOVDaddr. Odd offsets from SB shouldn't be folded (LARL can't handle them).
 (ADDconst [c] (MOVDaddr [d] {s} x:(SB))) && ((c+d)&1 == 0) && is32Bit(c+d) -> (MOVDaddr [c+d] {s} x)
@@ -1133,6 +1153,9 @@
 (XORconst [0] x)                  => x
 (XORWconst [c] x) && int32(c)==0   => x
 
+// Shifts by zero (may be inserted during multiplication strength reduction).
+((SLD|SLW|SRD|SRW|SRAD|SRAW)const x [0]) => x
+
 // Convert constant subtracts to constant adds.
 (SUBconst [c] x) && c != -(1<<31) => (ADDconst [-c] x)
 (SUBWconst [c] x) -> (ADDWconst [int64(int32(-c))] x)
diff --git a/src/cmd/compile/internal/ssa/gen/generic.rules b/src/cmd/compile/internal/ssa/gen/generic.rules
index ed5bfc81fd..2d39d27226 100644
--- a/src/cmd/compile/internal/ssa/gen/generic.rules
+++ b/src/cmd/compile/internal/ssa/gen/generic.rules
@@ -545,6 +545,10 @@
 (Or(64|32|16|8) x (Or(64|32|16|8) x y)) => (Or(64|32|16|8) x y)
 (Xor(64|32|16|8) x (Xor(64|32|16|8) x y)) => y
 
+// Unsigned comparisons to zero.
+(Less(64U|32U|16U|8U) _ (Const(64|32|16|8) [0])) => (ConstBool [false])
+(Leq(64U|32U|16U|8U) (Const(64|32|16|8) [0]) _)  => (ConstBool [true])
+
 // Ands clear bits. Ors set bits.
 // If a subsequent Or will set all the bits
 // that an And cleared, we can skip the And.
diff --git a/src/cmd/compile/internal/ssa/opGen.go b/src/cmd/compile/internal/ssa/opGen.go
index 9efa1bfcc4..4cd72799e8 100644
--- a/src/cmd/compile/internal/ssa/opGen.go
+++ b/src/cmd/compile/internal/ssa/opGen.go
@@ -1832,6 +1832,7 @@ const (
 	OpPPC64FSUBS
 	OpPPC64MULLD
 	OpPPC64MULLW
+	OpPPC64MADDLD
 	OpPPC64MULHD
 	OpPPC64MULHW
 	OpPPC64MULHDU
@@ -13119,6 +13120,7 @@ var opcodeTable = [...]opInfo{
 		auxType:        auxInt64,
 		argLen:         3,
 		faultOnNilArg0: true,
+		unsafePoint:    true,
 		reg: regInfo{
 			inputs: []inputInfo{
 				{0, 128},   // DI
@@ -13196,6 +13198,7 @@ var opcodeTable = [...]opInfo{
 		clobberFlags:   true,
 		faultOnNilArg0: true,
 		faultOnNilArg1: true,
+		unsafePoint:    true,
 		reg: regInfo{
 			inputs: []inputInfo{
 				{0, 128}, // DI
@@ -20734,6 +20737,7 @@ var opcodeTable = [...]opInfo{
 		auxType:        auxInt64,
 		argLen:         2,
 		faultOnNilArg0: true,
+		unsafePoint:    true,
 		reg: regInfo{
 			inputs: []inputInfo{
 				{0, 1048576}, // R20
@@ -20760,6 +20764,7 @@ var opcodeTable = [...]opInfo{
 		argLen:         3,
 		faultOnNilArg0: true,
 		faultOnNilArg1: true,
+		unsafePoint:    true,
 		reg: regInfo{
 			inputs: []inputInfo{
 				{0, 2097152}, // R21
@@ -24371,6 +24376,21 @@ var opcodeTable = [...]opInfo{
 		},
 	},
 	{
+		name:   "MADDLD",
+		argLen: 3,
+		asm:    ppc64.AMADDLD,
+		reg: regInfo{
+			inputs: []inputInfo{
+				{0, 1073733630}, // SP SB R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29
+				{1, 1073733630}, // SP SB R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29
+				{2, 1073733630}, // SP SB R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29
+			},
+			outputs: []outputInfo{
+				{0, 1073733624}, // R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R14 R15 R16 R17 R18 R19 R20 R21 R22 R23 R24 R25 R26 R27 R28 R29
+			},
+		},
+	},
+	{
 		name:        "MULHD",
 		argLen:      2,
 		commutative: true,
@@ -26885,7 +26905,7 @@ var opcodeTable = [...]opInfo{
 				{0, 1048576}, // R20
 				{1, 2097152}, // R21
 			},
-			clobbers: 576460746931503104, // R16 R17 R18 R19 R22 R23 R24 R25 R26 R27 R28 R29 R31 F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 F13 F14 F15 F16 F17 F18 F19 F20 F21 F22 F23 F24 F25 F26
+			clobbers: 576460746931312640, // R11 R12 R18 R19 R22 R23 R24 R25 R26 R27 R28 R29 R31 F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12 F13 F14 F15 F16 F17 F18 F19 F20 F21 F22 F23 F24 F25 F26
 		},
 	},
 	{
diff --git a/src/cmd/compile/internal/ssa/passbm_test.go b/src/cmd/compile/internal/ssa/passbm_test.go
index eefdbb8722..3fd3eb579b 100644
--- a/src/cmd/compile/internal/ssa/passbm_test.go
+++ b/src/cmd/compile/internal/ssa/passbm_test.go
@@ -1,6 +1,7 @@
 // Copyright 2015 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 ssa
 
 import (
diff --git a/src/cmd/compile/internal/ssa/phiopt.go b/src/cmd/compile/internal/ssa/phiopt.go
index 8643fa584c..db7b02275c 100644
--- a/src/cmd/compile/internal/ssa/phiopt.go
+++ b/src/cmd/compile/internal/ssa/phiopt.go
@@ -154,7 +154,7 @@ func phioptint(v *Value, b0 *Block, reverse int) {
 	}
 	v.AddArg(a)
 
-	cvt := v.Block.NewValue1(v.Pos, OpCvtBoolToUint8, a.Type, a)
+	cvt := v.Block.NewValue1(v.Pos, OpCvtBoolToUint8, v.Block.Func.Config.Types.UInt8, a)
 	switch v.Type.Size() {
 	case 1:
 		v.reset(OpCopy)
diff --git a/src/cmd/compile/internal/ssa/prove.go b/src/cmd/compile/internal/ssa/prove.go
index 6c6be39d34..ce7d689f93 100644
--- a/src/cmd/compile/internal/ssa/prove.go
+++ b/src/cmd/compile/internal/ssa/prove.go
@@ -1334,7 +1334,7 @@ func removeBranch(b *Block, branch branch) {
 // isNonNegative reports whether v is known to be greater or equal to zero.
 func isNonNegative(v *Value) bool {
 	if !v.Type.IsInteger() {
-		panic("isNonNegative bad type")
+		v.Fatalf("isNonNegative bad type: %v", v.Type)
 	}
 	// TODO: return true if !v.Type.IsSigned()
 	// SSA isn't type-safe enough to do that now (issue 37753).
diff --git a/src/cmd/compile/internal/ssa/rewrite.go b/src/cmd/compile/internal/ssa/rewrite.go
index 2152b1675a..e082bb1dfa 100644
--- a/src/cmd/compile/internal/ssa/rewrite.go
+++ b/src/cmd/compile/internal/ssa/rewrite.go
@@ -1379,6 +1379,15 @@ func needRaceCleanup(sym Sym, v *Value) bool {
 			}
 		}
 	}
+	if symNamed(sym, "runtime.racefuncenter") {
+		// If we're removing racefuncenter, remove its argument as well.
+		if v.Args[0].Op != OpStore {
+			return false
+		}
+		mem := v.Args[0].Args[2]
+		v.Args[0].reset(OpCopy)
+		v.Args[0].AddArg(mem)
+	}
 	return true
 }
 
diff --git a/src/cmd/compile/internal/ssa/rewriteAMD64.go b/src/cmd/compile/internal/ssa/rewriteAMD64.go
index 20eab05e9c..cda9df56f4 100644
--- a/src/cmd/compile/internal/ssa/rewriteAMD64.go
+++ b/src/cmd/compile/internal/ssa/rewriteAMD64.go
@@ -6924,26 +6924,42 @@ func rewriteValueAMD64_OpAMD64CMPBconst(v *Value) bool {
 		v.reset(OpAMD64FlagLT_ULT)
 		return true
 	}
-	// match: (CMPBconst (ANDL x y) [0])
+	// match: (CMPBconst a:(ANDL x y) [0])
+	// cond: a.Uses == 1
 	// result: (TESTB x y)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDL {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDL {
+			break
+		}
+		y := a.Args[1]
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		y := v_0.Args[1]
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTB)
 		v.AddArg2(x, y)
 		return true
 	}
-	// match: (CMPBconst (ANDLconst [c] x) [0])
+	// match: (CMPBconst a:(ANDLconst [c] x) [0])
+	// cond: a.Uses == 1
 	// result: (TESTBconst [int64(int8(c))] x)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDLconst {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDLconst {
+			break
+		}
+		c := a.AuxInt
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		c := v_0.AuxInt
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTBconst)
 		v.AuxInt = int64(int8(c))
 		v.AddArg(x)
@@ -7309,26 +7325,42 @@ func rewriteValueAMD64_OpAMD64CMPLconst(v *Value) bool {
 		v.reset(OpAMD64FlagLT_ULT)
 		return true
 	}
-	// match: (CMPLconst (ANDL x y) [0])
+	// match: (CMPLconst a:(ANDL x y) [0])
+	// cond: a.Uses == 1
 	// result: (TESTL x y)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDL {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDL {
+			break
+		}
+		y := a.Args[1]
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		y := v_0.Args[1]
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTL)
 		v.AddArg2(x, y)
 		return true
 	}
-	// match: (CMPLconst (ANDLconst [c] x) [0])
+	// match: (CMPLconst a:(ANDLconst [c] x) [0])
+	// cond: a.Uses == 1
 	// result: (TESTLconst [c] x)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDLconst {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDLconst {
+			break
+		}
+		c := a.AuxInt
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		c := v_0.AuxInt
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTLconst)
 		v.AuxInt = c
 		v.AddArg(x)
@@ -7874,26 +7906,42 @@ func rewriteValueAMD64_OpAMD64CMPQconst(v *Value) bool {
 		v.reset(OpAMD64FlagLT_ULT)
 		return true
 	}
-	// match: (CMPQconst (ANDQ x y) [0])
+	// match: (CMPQconst a:(ANDQ x y) [0])
+	// cond: a.Uses == 1
 	// result: (TESTQ x y)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDQ {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDQ {
+			break
+		}
+		y := a.Args[1]
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		y := v_0.Args[1]
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTQ)
 		v.AddArg2(x, y)
 		return true
 	}
-	// match: (CMPQconst (ANDQconst [c] x) [0])
+	// match: (CMPQconst a:(ANDQconst [c] x) [0])
+	// cond: a.Uses == 1
 	// result: (TESTQconst [c] x)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDQconst {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDQconst {
+			break
+		}
+		c := a.AuxInt
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		c := v_0.AuxInt
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTQconst)
 		v.AuxInt = c
 		v.AddArg(x)
@@ -8244,26 +8292,42 @@ func rewriteValueAMD64_OpAMD64CMPWconst(v *Value) bool {
 		v.reset(OpAMD64FlagLT_ULT)
 		return true
 	}
-	// match: (CMPWconst (ANDL x y) [0])
+	// match: (CMPWconst a:(ANDL x y) [0])
+	// cond: a.Uses == 1
 	// result: (TESTW x y)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDL {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDL {
+			break
+		}
+		y := a.Args[1]
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		y := v_0.Args[1]
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTW)
 		v.AddArg2(x, y)
 		return true
 	}
-	// match: (CMPWconst (ANDLconst [c] x) [0])
+	// match: (CMPWconst a:(ANDLconst [c] x) [0])
+	// cond: a.Uses == 1
 	// result: (TESTWconst [int64(int16(c))] x)
 	for {
-		if v.AuxInt != 0 || v_0.Op != OpAMD64ANDLconst {
+		if v.AuxInt != 0 {
+			break
+		}
+		a := v_0
+		if a.Op != OpAMD64ANDLconst {
+			break
+		}
+		c := a.AuxInt
+		x := a.Args[0]
+		if !(a.Uses == 1) {
 			break
 		}
-		c := v_0.AuxInt
-		x := v_0.Args[0]
 		v.reset(OpAMD64TESTWconst)
 		v.AuxInt = int64(int16(c))
 		v.AddArg(x)
diff --git a/src/cmd/compile/internal/ssa/rewriteARM64.go b/src/cmd/compile/internal/ssa/rewriteARM64.go
index 8e48b33628..842eddbf4a 100644
--- a/src/cmd/compile/internal/ssa/rewriteARM64.go
+++ b/src/cmd/compile/internal/ssa/rewriteARM64.go
@@ -14593,6 +14593,18 @@ func rewriteValueARM64_OpARM64MULW(v *Value) bool {
 }
 func rewriteValueARM64_OpARM64MVN(v *Value) bool {
 	v_0 := v.Args[0]
+	// match: (MVN (XOR x y))
+	// result: (EON x y)
+	for {
+		if v_0.Op != OpARM64XOR {
+			break
+		}
+		y := v_0.Args[1]
+		x := v_0.Args[0]
+		v.reset(OpARM64EON)
+		v.AddArg2(x, y)
+		return true
+	}
 	// match: (MVN (MOVDconst [c]))
 	// result: (MOVDconst [^c])
 	for {
@@ -21976,6 +21988,31 @@ func rewriteValueARM64_OpLeq16U(v *Value) bool {
 	v_0 := v.Args[0]
 	b := v.Block
 	typ := &b.Func.Config.Types
+	// match: (Leq16U x zero:(MOVDconst [0]))
+	// result: (Eq16 x zero)
+	for {
+		x := v_0
+		zero := v_1
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpEq16)
+		v.AddArg2(x, zero)
+		return true
+	}
+	// match: (Leq16U (MOVDconst [1]) x)
+	// result: (Neq16 (MOVDconst [0]) x)
+	for {
+		if v_0.Op != OpARM64MOVDconst || auxIntToInt64(v_0.AuxInt) != 1 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq16)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(v0, x)
+		return true
+	}
 	// match: (Leq16U x y)
 	// result: (LessEqualU (CMPW (ZeroExt16to32 x) (ZeroExt16to32 y)))
 	for {
@@ -22028,6 +22065,32 @@ func rewriteValueARM64_OpLeq32U(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
 	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (Leq32U x zero:(MOVDconst [0]))
+	// result: (Eq32 x zero)
+	for {
+		x := v_0
+		zero := v_1
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpEq32)
+		v.AddArg2(x, zero)
+		return true
+	}
+	// match: (Leq32U (MOVDconst [1]) x)
+	// result: (Neq32 (MOVDconst [0]) x)
+	for {
+		if v_0.Op != OpARM64MOVDconst || auxIntToInt64(v_0.AuxInt) != 1 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq32)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(v0, x)
+		return true
+	}
 	// match: (Leq32U x y)
 	// result: (LessEqualU (CMPW x y))
 	for {
@@ -22076,6 +22139,32 @@ func rewriteValueARM64_OpLeq64U(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
 	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (Leq64U x zero:(MOVDconst [0]))
+	// result: (Eq64 x zero)
+	for {
+		x := v_0
+		zero := v_1
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpEq64)
+		v.AddArg2(x, zero)
+		return true
+	}
+	// match: (Leq64U (MOVDconst [1]) x)
+	// result: (Neq64 (MOVDconst [0]) x)
+	for {
+		if v_0.Op != OpARM64MOVDconst || auxIntToInt64(v_0.AuxInt) != 1 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq64)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(v0, x)
+		return true
+	}
 	// match: (Leq64U x y)
 	// result: (LessEqualU (CMP x y))
 	for {
@@ -22114,6 +22203,31 @@ func rewriteValueARM64_OpLeq8U(v *Value) bool {
 	v_0 := v.Args[0]
 	b := v.Block
 	typ := &b.Func.Config.Types
+	// match: (Leq8U x zero:(MOVDconst [0]))
+	// result: (Eq8 x zero)
+	for {
+		x := v_0
+		zero := v_1
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpEq8)
+		v.AddArg2(x, zero)
+		return true
+	}
+	// match: (Leq8U (MOVDconst [1]) x)
+	// result: (Neq8 (MOVDconst [0]) x)
+	for {
+		if v_0.Op != OpARM64MOVDconst || auxIntToInt64(v_0.AuxInt) != 1 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq8)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(v0, x)
+		return true
+	}
 	// match: (Leq8U x y)
 	// result: (LessEqualU (CMPW (ZeroExt8to32 x) (ZeroExt8to32 y)))
 	for {
@@ -22156,6 +22270,31 @@ func rewriteValueARM64_OpLess16U(v *Value) bool {
 	v_0 := v.Args[0]
 	b := v.Block
 	typ := &b.Func.Config.Types
+	// match: (Less16U zero:(MOVDconst [0]) x)
+	// result: (Neq16 zero x)
+	for {
+		zero := v_0
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq16)
+		v.AddArg2(zero, x)
+		return true
+	}
+	// match: (Less16U x (MOVDconst [1]))
+	// result: (Eq16 x (MOVDconst [0]))
+	for {
+		x := v_0
+		if v_1.Op != OpARM64MOVDconst || auxIntToInt64(v_1.AuxInt) != 1 {
+			break
+		}
+		v.reset(OpEq16)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Less16U x y)
 	// result: (LessThanU (CMPW (ZeroExt16to32 x) (ZeroExt16to32 y)))
 	for {
@@ -22208,6 +22347,32 @@ func rewriteValueARM64_OpLess32U(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
 	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (Less32U zero:(MOVDconst [0]) x)
+	// result: (Neq32 zero x)
+	for {
+		zero := v_0
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq32)
+		v.AddArg2(zero, x)
+		return true
+	}
+	// match: (Less32U x (MOVDconst [1]))
+	// result: (Eq32 x (MOVDconst [0]))
+	for {
+		x := v_0
+		if v_1.Op != OpARM64MOVDconst || auxIntToInt64(v_1.AuxInt) != 1 {
+			break
+		}
+		v.reset(OpEq32)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Less32U x y)
 	// result: (LessThanU (CMPW x y))
 	for {
@@ -22256,6 +22421,32 @@ func rewriteValueARM64_OpLess64U(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
 	b := v.Block
+	typ := &b.Func.Config.Types
+	// match: (Less64U zero:(MOVDconst [0]) x)
+	// result: (Neq64 zero x)
+	for {
+		zero := v_0
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq64)
+		v.AddArg2(zero, x)
+		return true
+	}
+	// match: (Less64U x (MOVDconst [1]))
+	// result: (Eq64 x (MOVDconst [0]))
+	for {
+		x := v_0
+		if v_1.Op != OpARM64MOVDconst || auxIntToInt64(v_1.AuxInt) != 1 {
+			break
+		}
+		v.reset(OpEq64)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Less64U x y)
 	// result: (LessThanU (CMP x y))
 	for {
@@ -22294,6 +22485,31 @@ func rewriteValueARM64_OpLess8U(v *Value) bool {
 	v_0 := v.Args[0]
 	b := v.Block
 	typ := &b.Func.Config.Types
+	// match: (Less8U zero:(MOVDconst [0]) x)
+	// result: (Neq8 zero x)
+	for {
+		zero := v_0
+		if zero.Op != OpARM64MOVDconst || auxIntToInt64(zero.AuxInt) != 0 {
+			break
+		}
+		x := v_1
+		v.reset(OpNeq8)
+		v.AddArg2(zero, x)
+		return true
+	}
+	// match: (Less8U x (MOVDconst [1]))
+	// result: (Eq8 x (MOVDconst [0]))
+	for {
+		x := v_0
+		if v_1.Op != OpARM64MOVDconst || auxIntToInt64(v_1.AuxInt) != 1 {
+			break
+		}
+		v.reset(OpEq8)
+		v0 := b.NewValue0(v.Pos, OpARM64MOVDconst, typ.UInt64)
+		v0.AuxInt = int64ToAuxInt(0)
+		v.AddArg2(x, v0)
+		return true
+	}
 	// match: (Less8U x y)
 	// result: (LessThanU (CMPW (ZeroExt8to32 x) (ZeroExt8to32 y)))
 	for {
diff --git a/src/cmd/compile/internal/ssa/rewritePPC64.go b/src/cmd/compile/internal/ssa/rewritePPC64.go
index 37b75cc58a..7704b80dc6 100644
--- a/src/cmd/compile/internal/ssa/rewritePPC64.go
+++ b/src/cmd/compile/internal/ssa/rewritePPC64.go
@@ -3852,6 +3852,27 @@ func rewriteValuePPC64_OpPPC64ADD(v *Value) bool {
 	v_0 := v.Args[0]
 	b := v.Block
 	typ := &b.Func.Config.Types
+	// match: (ADD l:(MULLD x y) z)
+	// cond: objabi.GOPPC64 >= 9 && l.Uses == 1 && clobber(l)
+	// result: (MADDLD x y z)
+	for {
+		for _i0 := 0; _i0 <= 1; _i0, v_0, v_1 = _i0+1, v_1, v_0 {
+			l := v_0
+			if l.Op != OpPPC64MULLD {
+				continue
+			}
+			y := l.Args[1]
+			x := l.Args[0]
+			z := v_1
+			if !(objabi.GOPPC64 >= 9 && l.Uses == 1 && clobber(l)) {
+				continue
+			}
+			v.reset(OpPPC64MADDLD)
+			v.AddArg3(x, y, z)
+			return true
+		}
+		break
+	}
 	// match: (ADD (SLDconst x [c]) (SRDconst x [d]))
 	// cond: d == 64-c
 	// result: (ROTLconst [c] x)
diff --git a/src/cmd/compile/internal/ssa/rewriteS390X.go b/src/cmd/compile/internal/ssa/rewriteS390X.go
index dc9b143562..536f8db320 100644
--- a/src/cmd/compile/internal/ssa/rewriteS390X.go
+++ b/src/cmd/compile/internal/ssa/rewriteS390X.go
@@ -732,8 +732,12 @@ func rewriteValueS390X(v *Value) bool {
 		return rewriteValueS390X_OpS390XRLLG(v)
 	case OpS390XSLD:
 		return rewriteValueS390X_OpS390XSLD(v)
+	case OpS390XSLDconst:
+		return rewriteValueS390X_OpS390XSLDconst(v)
 	case OpS390XSLW:
 		return rewriteValueS390X_OpS390XSLW(v)
+	case OpS390XSLWconst:
+		return rewriteValueS390X_OpS390XSLWconst(v)
 	case OpS390XSRAD:
 		return rewriteValueS390X_OpS390XSRAD(v)
 	case OpS390XSRADconst:
@@ -748,6 +752,8 @@ func rewriteValueS390X(v *Value) bool {
 		return rewriteValueS390X_OpS390XSRDconst(v)
 	case OpS390XSRW:
 		return rewriteValueS390X_OpS390XSRW(v)
+	case OpS390XSRWconst:
+		return rewriteValueS390X_OpS390XSRWconst(v)
 	case OpS390XSTM2:
 		return rewriteValueS390X_OpS390XSTM2(v)
 	case OpS390XSTMG2:
@@ -13853,81 +13859,64 @@ func rewriteValueS390X_OpS390XMULLD(v *Value) bool {
 func rewriteValueS390X_OpS390XMULLDconst(v *Value) bool {
 	v_0 := v.Args[0]
 	b := v.Block
-	// match: (MULLDconst [-1] x)
-	// result: (NEG x)
+	// match: (MULLDconst <t> x [c])
+	// cond: isPowerOfTwo32(c&(c-1))
+	// result: (ADD (SLDconst <t> x [int8(log32(c&(c-1)))]) (SLDconst <t> x [int8(log32(c&^(c-1)))]))
 	for {
-		if auxIntToInt32(v.AuxInt) != -1 {
-			break
-		}
-		x := v_0
-		v.reset(OpS390XNEG)
-		v.AddArg(x)
-		return true
-	}
-	// match: (MULLDconst [0] _)
-	// result: (MOVDconst [0])
-	for {
-		if auxIntToInt32(v.AuxInt) != 0 {
-			break
-		}
-		v.reset(OpS390XMOVDconst)
-		v.AuxInt = int64ToAuxInt(0)
-		return true
-	}
-	// match: (MULLDconst [1] x)
-	// result: x
-	for {
-		if auxIntToInt32(v.AuxInt) != 1 {
-			break
-		}
-		x := v_0
-		v.copyOf(x)
-		return true
-	}
-	// match: (MULLDconst [c] x)
-	// cond: isPowerOfTwo(c)
-	// result: (SLDconst [log2(c)] x)
-	for {
-		c := v.AuxInt
+		t := v.Type
+		c := auxIntToInt32(v.AuxInt)
 		x := v_0
-		if !(isPowerOfTwo(c)) {
+		if !(isPowerOfTwo32(c & (c - 1))) {
 			break
 		}
-		v.reset(OpS390XSLDconst)
-		v.AuxInt = log2(c)
-		v.AddArg(x)
+		v.reset(OpS390XADD)
+		v0 := b.NewValue0(v.Pos, OpS390XSLDconst, t)
+		v0.AuxInt = int8ToAuxInt(int8(log32(c & (c - 1))))
+		v0.AddArg(x)
+		v1 := b.NewValue0(v.Pos, OpS390XSLDconst, t)
+		v1.AuxInt = int8ToAuxInt(int8(log32(c &^ (c - 1))))
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
 		return true
 	}
-	// match: (MULLDconst [c] x)
-	// cond: isPowerOfTwo(c+1) && c >= 15
-	// result: (SUB (SLDconst <v.Type> [log2(c+1)] x) x)
+	// match: (MULLDconst <t> x [c])
+	// cond: isPowerOfTwo32(c+(c&^(c-1)))
+	// result: (SUB (SLDconst <t> x [int8(log32(c+(c&^(c-1))))]) (SLDconst <t> x [int8(log32(c&^(c-1)))]))
 	for {
-		c := v.AuxInt
+		t := v.Type
+		c := auxIntToInt32(v.AuxInt)
 		x := v_0
-		if !(isPowerOfTwo(c+1) && c >= 15) {
+		if !(isPowerOfTwo32(c + (c &^ (c - 1)))) {
 			break
 		}
 		v.reset(OpS390XSUB)
-		v0 := b.NewValue0(v.Pos, OpS390XSLDconst, v.Type)
-		v0.AuxInt = log2(c + 1)
+		v0 := b.NewValue0(v.Pos, OpS390XSLDconst, t)
+		v0.AuxInt = int8ToAuxInt(int8(log32(c + (c &^ (c - 1)))))
 		v0.AddArg(x)
-		v.AddArg2(v0, x)
+		v1 := b.NewValue0(v.Pos, OpS390XSLDconst, t)
+		v1.AuxInt = int8ToAuxInt(int8(log32(c &^ (c - 1))))
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
 		return true
 	}
-	// match: (MULLDconst [c] x)
-	// cond: isPowerOfTwo(c-1) && c >= 17
-	// result: (ADD (SLDconst <v.Type> [log2(c-1)] x) x)
+	// match: (MULLDconst <t> x [c])
+	// cond: isPowerOfTwo32(-c+(-c&^(-c-1)))
+	// result: (SUB (SLDconst <t> x [int8(log32(-c&^(-c-1)))]) (SLDconst <t> x [int8(log32(-c+(-c&^(-c-1))))]))
 	for {
-		c := v.AuxInt
+		t := v.Type
+		c := auxIntToInt32(v.AuxInt)
 		x := v_0
-		if !(isPowerOfTwo(c-1) && c >= 17) {
+		if !(isPowerOfTwo32(-c + (-c &^ (-c - 1)))) {
 			break
 		}
-		v.reset(OpS390XADD)
-		v0 := b.NewValue0(v.Pos, OpS390XSLDconst, v.Type)
-		v0.AuxInt = log2(c - 1)
+		v.reset(OpS390XSUB)
+		v0 := b.NewValue0(v.Pos, OpS390XSLDconst, t)
+		v0.AuxInt = int8ToAuxInt(int8(log32(-c &^ (-c - 1))))
 		v0.AddArg(x)
-		v.AddArg2(v0, x)
+		v1 := b.NewValue0(v.Pos, OpS390XSLDconst, t)
+		v1.AuxInt = int8ToAuxInt(int8(log32(-c + (-c &^ (-c - 1)))))
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
 		return true
 	}
 	// match: (MULLDconst [c] (MOVDconst [d]))
@@ -14097,81 +14086,64 @@ func rewriteValueS390X_OpS390XMULLW(v *Value) bool {
 func rewriteValueS390X_OpS390XMULLWconst(v *Value) bool {
 	v_0 := v.Args[0]
 	b := v.Block
-	// match: (MULLWconst [-1] x)
-	// result: (NEGW x)
-	for {
-		if auxIntToInt32(v.AuxInt) != -1 {
-			break
-		}
-		x := v_0
-		v.reset(OpS390XNEGW)
-		v.AddArg(x)
-		return true
-	}
-	// match: (MULLWconst [0] _)
-	// result: (MOVDconst [0])
-	for {
-		if auxIntToInt32(v.AuxInt) != 0 {
-			break
-		}
-		v.reset(OpS390XMOVDconst)
-		v.AuxInt = int64ToAuxInt(0)
-		return true
-	}
-	// match: (MULLWconst [1] x)
-	// result: x
+	// match: (MULLWconst <t> x [c])
+	// cond: isPowerOfTwo32(c&(c-1))
+	// result: (ADDW (SLWconst <t> x [int8(log32(c&(c-1)))]) (SLWconst <t> x [int8(log32(c&^(c-1)))]))
 	for {
-		if auxIntToInt32(v.AuxInt) != 1 {
-			break
-		}
-		x := v_0
-		v.copyOf(x)
-		return true
-	}
-	// match: (MULLWconst [c] x)
-	// cond: isPowerOfTwo(c)
-	// result: (SLWconst [log2(c)] x)
-	for {
-		c := v.AuxInt
+		t := v.Type
+		c := auxIntToInt32(v.AuxInt)
 		x := v_0
-		if !(isPowerOfTwo(c)) {
+		if !(isPowerOfTwo32(c & (c - 1))) {
 			break
 		}
-		v.reset(OpS390XSLWconst)
-		v.AuxInt = log2(c)
-		v.AddArg(x)
+		v.reset(OpS390XADDW)
+		v0 := b.NewValue0(v.Pos, OpS390XSLWconst, t)
+		v0.AuxInt = int8ToAuxInt(int8(log32(c & (c - 1))))
+		v0.AddArg(x)
+		v1 := b.NewValue0(v.Pos, OpS390XSLWconst, t)
+		v1.AuxInt = int8ToAuxInt(int8(log32(c &^ (c - 1))))
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
 		return true
 	}
-	// match: (MULLWconst [c] x)
-	// cond: isPowerOfTwo(c+1) && c >= 15
-	// result: (SUBW (SLWconst <v.Type> [log2(c+1)] x) x)
+	// match: (MULLWconst <t> x [c])
+	// cond: isPowerOfTwo32(c+(c&^(c-1)))
+	// result: (SUBW (SLWconst <t> x [int8(log32(c+(c&^(c-1))))]) (SLWconst <t> x [int8(log32(c&^(c-1)))]))
 	for {
-		c := v.AuxInt
+		t := v.Type
+		c := auxIntToInt32(v.AuxInt)
 		x := v_0
-		if !(isPowerOfTwo(c+1) && c >= 15) {
+		if !(isPowerOfTwo32(c + (c &^ (c - 1)))) {
 			break
 		}
 		v.reset(OpS390XSUBW)
-		v0 := b.NewValue0(v.Pos, OpS390XSLWconst, v.Type)
-		v0.AuxInt = log2(c + 1)
+		v0 := b.NewValue0(v.Pos, OpS390XSLWconst, t)
+		v0.AuxInt = int8ToAuxInt(int8(log32(c + (c &^ (c - 1)))))
 		v0.AddArg(x)
-		v.AddArg2(v0, x)
+		v1 := b.NewValue0(v.Pos, OpS390XSLWconst, t)
+		v1.AuxInt = int8ToAuxInt(int8(log32(c &^ (c - 1))))
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
 		return true
 	}
-	// match: (MULLWconst [c] x)
-	// cond: isPowerOfTwo(c-1) && c >= 17
-	// result: (ADDW (SLWconst <v.Type> [log2(c-1)] x) x)
+	// match: (MULLWconst <t> x [c])
+	// cond: isPowerOfTwo32(-c+(-c&^(-c-1)))
+	// result: (SUBW (SLWconst <t> x [int8(log32(-c&^(-c-1)))]) (SLWconst <t> x [int8(log32(-c+(-c&^(-c-1))))]))
 	for {
-		c := v.AuxInt
+		t := v.Type
+		c := auxIntToInt32(v.AuxInt)
 		x := v_0
-		if !(isPowerOfTwo(c-1) && c >= 17) {
+		if !(isPowerOfTwo32(-c + (-c &^ (-c - 1)))) {
 			break
 		}
-		v.reset(OpS390XADDW)
-		v0 := b.NewValue0(v.Pos, OpS390XSLWconst, v.Type)
-		v0.AuxInt = log2(c - 1)
+		v.reset(OpS390XSUBW)
+		v0 := b.NewValue0(v.Pos, OpS390XSLWconst, t)
+		v0.AuxInt = int8ToAuxInt(int8(log32(-c &^ (-c - 1))))
 		v0.AddArg(x)
-		v.AddArg2(v0, x)
+		v1 := b.NewValue0(v.Pos, OpS390XSLWconst, t)
+		v1.AuxInt = int8ToAuxInt(int8(log32(-c + (-c &^ (-c - 1)))))
+		v1.AddArg(x)
+		v.AddArg2(v0, v1)
 		return true
 	}
 	// match: (MULLWconst [c] (MOVDconst [d]))
@@ -16826,6 +16798,20 @@ func rewriteValueS390X_OpS390XSLD(v *Value) bool {
 	}
 	return false
 }
+func rewriteValueS390X_OpS390XSLDconst(v *Value) bool {
+	v_0 := v.Args[0]
+	// match: (SLDconst x [0])
+	// result: x
+	for {
+		if auxIntToInt8(v.AuxInt) != 0 {
+			break
+		}
+		x := v_0
+		v.copyOf(x)
+		return true
+	}
+	return false
+}
 func rewriteValueS390X_OpS390XSLW(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
@@ -16960,6 +16946,20 @@ func rewriteValueS390X_OpS390XSLW(v *Value) bool {
 	}
 	return false
 }
+func rewriteValueS390X_OpS390XSLWconst(v *Value) bool {
+	v_0 := v.Args[0]
+	// match: (SLWconst x [0])
+	// result: x
+	for {
+		if auxIntToInt8(v.AuxInt) != 0 {
+			break
+		}
+		x := v_0
+		v.copyOf(x)
+		return true
+	}
+	return false
+}
 func rewriteValueS390X_OpS390XSRAD(v *Value) bool {
 	v_1 := v.Args[1]
 	v_0 := v.Args[0]
@@ -17096,6 +17096,16 @@ func rewriteValueS390X_OpS390XSRAD(v *Value) bool {
 }
 func rewriteValueS390X_OpS390XSRADconst(v *Value) bool {
 	v_0 := v.Args[0]
+	// match: (SRADconst x [0])
+	// result: x
+	for {
+		if auxIntToInt8(v.AuxInt) != 0 {
+			break
+		}
+		x := v_0
+		v.copyOf(x)
+		return true
+	}
 	// match: (SRADconst [c] (MOVDconst [d]))
 	// result: (MOVDconst [d>>uint64(c)])
 	for {
@@ -17246,6 +17256,16 @@ func rewriteValueS390X_OpS390XSRAW(v *Value) bool {
 }
 func rewriteValueS390X_OpS390XSRAWconst(v *Value) bool {
 	v_0 := v.Args[0]
+	// match: (SRAWconst x [0])
+	// result: x
+	for {
+		if auxIntToInt8(v.AuxInt) != 0 {
+			break
+		}
+		x := v_0
+		v.copyOf(x)
+		return true
+	}
 	// match: (SRAWconst [c] (MOVDconst [d]))
 	// result: (MOVDconst [int64(int32(d))>>uint64(c)])
 	for {
@@ -17416,6 +17436,16 @@ func rewriteValueS390X_OpS390XSRDconst(v *Value) bool {
 		v.AddArg(v0)
 		return true
 	}
+	// match: (SRDconst x [0])
+	// result: x
+	for {
+		if auxIntToInt8(v.AuxInt) != 0 {
+			break
+		}
+		x := v_0
+		v.copyOf(x)
+		return true
+	}
 	return false
 }
 func rewriteValueS390X_OpS390XSRW(v *Value) bool {
@@ -17552,6 +17582,20 @@ func rewriteValueS390X_OpS390XSRW(v *Value) bool {
 	}
 	return false
 }
+func rewriteValueS390X_OpS390XSRWconst(v *Value) bool {
+	v_0 := v.Args[0]
+	// match: (SRWconst x [0])
+	// result: x
+	for {
+		if auxIntToInt8(v.AuxInt) != 0 {
+			break
+		}
+		x := v_0
+		v.copyOf(x)
+		return true
+	}
+	return false
+}
 func rewriteValueS390X_OpS390XSTM2(v *Value) bool {
 	v_3 := v.Args[3]
 	v_2 := v.Args[2]
diff --git a/src/cmd/compile/internal/ssa/rewritegeneric.go b/src/cmd/compile/internal/ssa/rewritegeneric.go
index 9f4e1b95bd..68e49f46f3 100644
--- a/src/cmd/compile/internal/ssa/rewritegeneric.go
+++ b/src/cmd/compile/internal/ssa/rewritegeneric.go
@@ -9701,6 +9701,16 @@ func rewriteValuegeneric_OpLeq16U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint16(c) <= uint16(d))
 		return true
 	}
+	// match: (Leq16U (Const16 [0]) _)
+	// result: (ConstBool [true])
+	for {
+		if v_0.Op != OpConst16 || auxIntToInt16(v_0.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(true)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLeq32(v *Value) bool {
@@ -9805,6 +9815,16 @@ func rewriteValuegeneric_OpLeq32U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint32(c) <= uint32(d))
 		return true
 	}
+	// match: (Leq32U (Const32 [0]) _)
+	// result: (ConstBool [true])
+	for {
+		if v_0.Op != OpConst32 || auxIntToInt32(v_0.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(true)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLeq64(v *Value) bool {
@@ -9909,6 +9929,16 @@ func rewriteValuegeneric_OpLeq64U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint64(c) <= uint64(d))
 		return true
 	}
+	// match: (Leq64U (Const64 [0]) _)
+	// result: (ConstBool [true])
+	for {
+		if v_0.Op != OpConst64 || auxIntToInt64(v_0.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(true)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLeq8(v *Value) bool {
@@ -9993,6 +10023,16 @@ func rewriteValuegeneric_OpLeq8U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint8(c) <= uint8(d))
 		return true
 	}
+	// match: (Leq8U (Const8 [0]) _)
+	// result: (ConstBool [true])
+	for {
+		if v_0.Op != OpConst8 || auxIntToInt8(v_0.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(true)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLess16(v *Value) bool {
@@ -10033,6 +10073,16 @@ func rewriteValuegeneric_OpLess16U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint16(c) < uint16(d))
 		return true
 	}
+	// match: (Less16U _ (Const16 [0]))
+	// result: (ConstBool [false])
+	for {
+		if v_1.Op != OpConst16 || auxIntToInt16(v_1.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(false)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLess32(v *Value) bool {
@@ -10093,6 +10143,16 @@ func rewriteValuegeneric_OpLess32U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint32(c) < uint32(d))
 		return true
 	}
+	// match: (Less32U _ (Const32 [0]))
+	// result: (ConstBool [false])
+	for {
+		if v_1.Op != OpConst32 || auxIntToInt32(v_1.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(false)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLess64(v *Value) bool {
@@ -10153,6 +10213,16 @@ func rewriteValuegeneric_OpLess64U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint64(c) < uint64(d))
 		return true
 	}
+	// match: (Less64U _ (Const64 [0]))
+	// result: (ConstBool [false])
+	for {
+		if v_1.Op != OpConst64 || auxIntToInt64(v_1.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(false)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLess8(v *Value) bool {
@@ -10193,6 +10263,16 @@ func rewriteValuegeneric_OpLess8U(v *Value) bool {
 		v.AuxInt = boolToAuxInt(uint8(c) < uint8(d))
 		return true
 	}
+	// match: (Less8U _ (Const8 [0]))
+	// result: (ConstBool [false])
+	for {
+		if v_1.Op != OpConst8 || auxIntToInt8(v_1.AuxInt) != 0 {
+			break
+		}
+		v.reset(OpConstBool)
+		v.AuxInt = boolToAuxInt(false)
+		return true
+	}
 	return false
 }
 func rewriteValuegeneric_OpLoad(v *Value) bool {
diff --git a/src/cmd/compile/internal/test/mulconst_test.go b/src/cmd/compile/internal/test/mulconst_test.go
new file mode 100644
index 0000000000..314cab32de
--- /dev/null
+++ b/src/cmd/compile/internal/test/mulconst_test.go
@@ -0,0 +1,242 @@
+// Copyright 2020 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 test
+
+import "testing"
+
+// Benchmark multiplication of an integer by various constants.
+//
+// The comment above each sub-benchmark provides an example of how the
+// target multiplication operation might be implemented using shift
+// (multiplication by a power of 2), addition and subtraction
+// operations. It is platform-dependent whether these transformations
+// are actually applied.
+
+var (
+	mulSinkI32 int32
+	mulSinkI64 int64
+	mulSinkU32 uint32
+	mulSinkU64 uint64
+)
+
+func BenchmarkMulconstI32(b *testing.B) {
+	// 3x = 2x + x
+	b.Run("3", func(b *testing.B) {
+		x := int32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 3
+		}
+		mulSinkI32 = x
+	})
+	// 5x = 4x + x
+	b.Run("5", func(b *testing.B) {
+		x := int32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 5
+		}
+		mulSinkI32 = x
+	})
+	// 12x = 8x + 4x
+	b.Run("12", func(b *testing.B) {
+		x := int32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 12
+		}
+		mulSinkI32 = x
+	})
+	// 120x = 128x - 8x
+	b.Run("120", func(b *testing.B) {
+		x := int32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 120
+		}
+		mulSinkI32 = x
+	})
+	// -120x = 8x - 120x
+	b.Run("-120", func(b *testing.B) {
+		x := int32(1)
+		for i := 0; i < b.N; i++ {
+			x *= -120
+		}
+		mulSinkI32 = x
+	})
+	// 65537x = 65536x + x
+	b.Run("65537", func(b *testing.B) {
+		x := int32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65537
+		}
+		mulSinkI32 = x
+	})
+	// 65538x = 65536x + 2x
+	b.Run("65538", func(b *testing.B) {
+		x := int32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65538
+		}
+		mulSinkI32 = x
+	})
+}
+
+func BenchmarkMulconstI64(b *testing.B) {
+	// 3x = 2x + x
+	b.Run("3", func(b *testing.B) {
+		x := int64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 3
+		}
+		mulSinkI64 = x
+	})
+	// 5x = 4x + x
+	b.Run("5", func(b *testing.B) {
+		x := int64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 5
+		}
+		mulSinkI64 = x
+	})
+	// 12x = 8x + 4x
+	b.Run("12", func(b *testing.B) {
+		x := int64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 12
+		}
+		mulSinkI64 = x
+	})
+	// 120x = 128x - 8x
+	b.Run("120", func(b *testing.B) {
+		x := int64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 120
+		}
+		mulSinkI64 = x
+	})
+	// -120x = 8x - 120x
+	b.Run("-120", func(b *testing.B) {
+		x := int64(1)
+		for i := 0; i < b.N; i++ {
+			x *= -120
+		}
+		mulSinkI64 = x
+	})
+	// 65537x = 65536x + x
+	b.Run("65537", func(b *testing.B) {
+		x := int64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65537
+		}
+		mulSinkI64 = x
+	})
+	// 65538x = 65536x + 2x
+	b.Run("65538", func(b *testing.B) {
+		x := int64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65538
+		}
+		mulSinkI64 = x
+	})
+}
+
+func BenchmarkMulconstU32(b *testing.B) {
+	// 3x = 2x + x
+	b.Run("3", func(b *testing.B) {
+		x := uint32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 3
+		}
+		mulSinkU32 = x
+	})
+	// 5x = 4x + x
+	b.Run("5", func(b *testing.B) {
+		x := uint32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 5
+		}
+		mulSinkU32 = x
+	})
+	// 12x = 8x + 4x
+	b.Run("12", func(b *testing.B) {
+		x := uint32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 12
+		}
+		mulSinkU32 = x
+	})
+	// 120x = 128x - 8x
+	b.Run("120", func(b *testing.B) {
+		x := uint32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 120
+		}
+		mulSinkU32 = x
+	})
+	// 65537x = 65536x + x
+	b.Run("65537", func(b *testing.B) {
+		x := uint32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65537
+		}
+		mulSinkU32 = x
+	})
+	// 65538x = 65536x + 2x
+	b.Run("65538", func(b *testing.B) {
+		x := uint32(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65538
+		}
+		mulSinkU32 = x
+	})
+}
+
+func BenchmarkMulconstU64(b *testing.B) {
+	// 3x = 2x + x
+	b.Run("3", func(b *testing.B) {
+		x := uint64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 3
+		}
+		mulSinkU64 = x
+	})
+	// 5x = 4x + x
+	b.Run("5", func(b *testing.B) {
+		x := uint64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 5
+		}
+		mulSinkU64 = x
+	})
+	// 12x = 8x + 4x
+	b.Run("12", func(b *testing.B) {
+		x := uint64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 12
+		}
+		mulSinkU64 = x
+	})
+	// 120x = 128x - 8x
+	b.Run("120", func(b *testing.B) {
+		x := uint64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 120
+		}
+		mulSinkU64 = x
+	})
+	// 65537x = 65536x + x
+	b.Run("65537", func(b *testing.B) {
+		x := uint64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65537
+		}
+		mulSinkU64 = x
+	})
+	// 65538x = 65536x + 2x
+	b.Run("65538", func(b *testing.B) {
+		x := uint64(1)
+		for i := 0; i < b.N; i++ {
+			x *= 65538
+		}
+		mulSinkU64 = x
+	})
+}
diff --git a/src/cmd/compile/internal/types/type.go b/src/cmd/compile/internal/types/type.go
index 3b7b31c5d6..91b54b43d4 100644
--- a/src/cmd/compile/internal/types/type.go
+++ b/src/cmd/compile/internal/types/type.go
@@ -131,6 +131,7 @@ type Type struct {
 	// TPTR: Ptr
 	// TARRAY: *Array
 	// TSLICE: Slice
+	// TSSA: string
 	Extra interface{}
 
 	// Width is the width of this Type in bytes.
@@ -1026,7 +1027,7 @@ func (t *Type) cmp(x *Type) Cmp {
 
 	case TSSA:
 		tname := t.Extra.(string)
-		xname := t.Extra.(string)
+		xname := x.Extra.(string)
 		// desire fast sorting, not pretty sorting.
 		if len(tname) == len(xname) {
 			if tname == xname {
diff --git a/src/cmd/compile/internal/types/type_test.go b/src/cmd/compile/internal/types/type_test.go
new file mode 100644
index 0000000000..fe3f380b21
--- /dev/null
+++ b/src/cmd/compile/internal/types/type_test.go
@@ -0,0 +1,28 @@
+// Copyright 2020 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 types_test
+
+import (
+	"cmd/compile/internal/types"
+	"testing"
+)
+
+func TestSSACompare(t *testing.T) {
+	a := []*types.Type{
+		types.TypeInvalid,
+		types.TypeMem,
+		types.TypeFlags,
+		types.TypeVoid,
+		types.TypeInt128,
+	}
+	for _, x := range a {
+		for _, y := range a {
+			c := x.Compare(y)
+			if x == y && c != types.CMPeq || x != y && c == types.CMPeq {
+				t.Errorf("%s compare %s == %d\n", x.Extra, y.Extra, c)
+			}
+		}
+	}
+}
diff --git a/src/cmd/cover/cover_test.go b/src/cmd/cover/cover_test.go
index 8a56e39011..1c252e6e45 100644
--- a/src/cmd/cover/cover_test.go
+++ b/src/cmd/cover/cover_test.go
@@ -179,7 +179,7 @@ func TestCover(t *testing.T) {
 	}
 	lines := bytes.Split(file, []byte("\n"))
 	for i, line := range lines {
-		lines[i] = bytes.Replace(line, []byte("LINE"), []byte(fmt.Sprint(i+1)), -1)
+		lines[i] = bytes.ReplaceAll(line, []byte("LINE"), []byte(fmt.Sprint(i+1)))
 	}
 
 	// Add a function that is not gofmt'ed. This used to cause a crash.
diff --git a/src/cmd/dist/build.go b/src/cmd/dist/build.go
index a817e6fcd7..397b3bb88f 100644
--- a/src/cmd/dist/build.go
+++ b/src/cmd/dist/build.go
@@ -1209,7 +1209,7 @@ func timelog(op, name string) {
 		}
 		i := strings.Index(s, " start")
 		if i < 0 {
-			log.Fatalf("time log %s does not begin with start line", os.Getenv("GOBULDTIMELOGFILE"))
+			log.Fatalf("time log %s does not begin with start line", os.Getenv("GOBUILDTIMELOGFILE"))
 		}
 		t, err := time.Parse(time.UnixDate, s[:i])
 		if err != nil {
diff --git a/src/cmd/dist/buildtool.go b/src/cmd/dist/buildtool.go
index 40d28b535b..79eab24d29 100644
--- a/src/cmd/dist/buildtool.go
+++ b/src/cmd/dist/buildtool.go
@@ -113,14 +113,15 @@ var ignorePrefixes = []string{
 
 // File suffixes that use build tags introduced since Go 1.4.
 // These must not be copied into the bootstrap build directory.
+// Also ignore test files.
 var ignoreSuffixes = []string{
 	"_arm64.s",
-	"_arm64_test.s",
 	"_arm64.go",
 	"_riscv64.s",
 	"_riscv64.go",
 	"_wasm.s",
 	"_wasm.go",
+	"_test.s",
 }
 
 func bootstrapBuildTools() {
diff --git a/src/cmd/dist/test.go b/src/cmd/dist/test.go
index 2dc9459215..a83ae35293 100644
--- a/src/cmd/dist/test.go
+++ b/src/cmd/dist/test.go
@@ -241,13 +241,15 @@ func (t *tester) shouldRunTest(name string) bool {
 	return false
 }
 
-// short returns a -short flag to pass to 'go test'.
-// It returns "-short", unless the environment variable
+// short returns a -short flag value to use with 'go test'
+// or a test binary for tests intended to run in short mode.
+// It returns "true", unless the environment variable
 // GO_TEST_SHORT is set to a non-empty, false-ish string.
 //
 // This environment variable is meant to be an internal
-// detail between the Go build system and cmd/dist
-// and is not intended for use by users.
+// detail between the Go build system and cmd/dist for
+// the purpose of longtest builders, and is not intended
+// for use by users. See golang.org/issue/12508.
 func short() string {
 	if v := os.Getenv("GO_TEST_SHORT"); v != "" {
 		short, err := strconv.ParseBool(v)
@@ -255,10 +257,10 @@ func short() string {
 			fatalf("invalid GO_TEST_SHORT %q: %v", v, err)
 		}
 		if !short {
-			return "-short=false"
+			return "false"
 		}
 	}
-	return "-short"
+	return "true"
 }
 
 // goTest returns the beginning of the go test command line.
@@ -266,7 +268,7 @@ func short() string {
 // defaults as later arguments in the command line.
 func (t *tester) goTest() []string {
 	return []string{
-		"go", "test", short(), "-count=1", t.tags(), t.runFlag(""),
+		"go", "test", "-short=" + short(), "-count=1", t.tags(), t.runFlag(""),
 	}
 }
 
@@ -335,7 +337,7 @@ func (t *tester) registerStdTest(pkg string) {
 			}
 			args := []string{
 				"test",
-				short(),
+				"-short=" + short(),
 				t.tags(),
 				t.timeout(timeoutSec),
 				"-gcflags=all=" + gogcflags,
@@ -373,7 +375,7 @@ func (t *tester) registerRaceBenchTest(pkg string) {
 			ranGoBench = true
 			args := []string{
 				"test",
-				short(),
+				"-short=" + short(),
 				"-race",
 				t.timeout(1200), // longer timeout for race with benchmarks
 				"-run=^$",       // nothing. only benchmarks.
@@ -1069,7 +1071,7 @@ func (t *tester) runHostTest(dir, pkg string) error {
 	if err := cmd.Run(); err != nil {
 		return err
 	}
-	return t.dirCmd(dir, f.Name(), "-test.short").Run()
+	return t.dirCmd(dir, f.Name(), "-test.short="+short()).Run()
 }
 
 func (t *tester) cgoTest(dt *distTest) error {
@@ -1570,7 +1572,7 @@ func (t *tester) prebuiltGoPackageTestBinary() string {
 func (t *tester) runPrecompiledStdTest(timeout time.Duration) error {
 	bin := t.prebuiltGoPackageTestBinary()
 	fmt.Fprintf(os.Stderr, "# %s: using pre-built %s...\n", stdMatches[0], bin)
-	cmd := exec.Command(bin, "-test.short", "-test.timeout="+timeout.String())
+	cmd := exec.Command(bin, "-test.short="+short(), "-test.timeout="+timeout.String())
 	cmd.Dir = filepath.Dir(bin)
 	cmd.Stdout = os.Stdout
 	cmd.Stderr = os.Stderr
diff --git a/src/cmd/go.mod b/src/cmd/go.mod
index 6d57ceee79..21670b9996 100644
--- a/src/cmd/go.mod
+++ b/src/cmd/go.mod
@@ -7,7 +7,7 @@ require (
 	github.com/ianlancetaylor/demangle v0.0.0-20200414190113-039b1ae3a340 // indirect
 	golang.org/x/arch v0.0.0-20200511175325-f7c78586839d
 	golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9
-	golang.org/x/mod v0.3.0
+	golang.org/x/mod v0.3.1-0.20200625141748-0b26df4a2231
 	golang.org/x/sys v0.0.0-20200501145240-bc7a7d42d5c3 // indirect
 	golang.org/x/tools v0.0.0-20200616133436-c1934b75d054
 	golang.org/x/xerrors v0.0.0-20200806184451-1a77d5e9f316 // indirect
diff --git a/src/cmd/go.sum b/src/cmd/go.sum
index 3fc693e3bf..1b5ef515c2 100644
--- a/src/cmd/go.sum
+++ b/src/cmd/go.sum
@@ -14,8 +14,8 @@ golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8U
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9 h1:psW17arqaxU48Z5kZ0CQnkZWQJsqcURM6tKiBApRjXI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
-golang.org/x/mod v0.3.0 h1:RM4zey1++hCTbCVQfnWeKs9/IEsaBLA8vTkd0WVtmH4=
-golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
+golang.org/x/mod v0.3.1-0.20200625141748-0b26df4a2231 h1:R11LxkoUvECaAHdM5/ZOevSR7n+016EgTw8nbE1l+XM=
+golang.org/x/mod v0.3.1-0.20200625141748-0b26df4a2231/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/net v0.0.0-20190404232315-eb5bcb51f2a3/go.mod h1:t9HGtf8HONx5eT2rtn7q6eTqICYqUVnKs3thJo3Qplg=
 golang.org/x/net v0.0.0-20190620200207-3b0461eec859/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
 golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLLGU+0bjPO0LkuOLi4/5GtJWs/s=
diff --git a/src/cmd/go/internal/base/base.go b/src/cmd/go/internal/base/base.go
index ab2f1bb4e2..db3ebef933 100644
--- a/src/cmd/go/internal/base/base.go
+++ b/src/cmd/go/internal/base/base.go
@@ -7,6 +7,7 @@
 package base
 
 import (
+	"context"
 	"flag"
 	"fmt"
 	"log"
@@ -24,7 +25,7 @@ import (
 type Command struct {
 	// Run runs the command.
 	// The args are the arguments after the command name.
-	Run func(cmd *Command, args []string)
+	Run func(ctx context.Context, cmd *Command, args []string)
 
 	// UsageLine is the one-line usage message.
 	// The words between "go" and the first flag or argument in the line are taken to be the command name.
diff --git a/src/cmd/go/internal/bug/bug.go b/src/cmd/go/internal/bug/bug.go
index fe71281ef0..52bd40f2fb 100644
--- a/src/cmd/go/internal/bug/bug.go
+++ b/src/cmd/go/internal/bug/bug.go
@@ -7,6 +7,7 @@ package bug
 
 import (
 	"bytes"
+	"context"
 	"fmt"
 	"io"
 	"io/ioutil"
@@ -37,7 +38,7 @@ func init() {
 	CmdBug.Flag.BoolVar(&cfg.BuildV, "v", false, "")
 }
 
-func runBug(cmd *base.Command, args []string) {
+func runBug(ctx context.Context, cmd *base.Command, args []string) {
 	if len(args) > 0 {
 		base.Fatalf("go bug: bug takes no arguments")
 	}
diff --git a/src/cmd/go/internal/cfg/cfg.go b/src/cmd/go/internal/cfg/cfg.go
index 7f8f8e92be..f9bbcd9180 100644
--- a/src/cmd/go/internal/cfg/cfg.go
+++ b/src/cmd/go/internal/cfg/cfg.go
@@ -51,6 +51,7 @@ var (
 	CmdName string // "build", "install", "list", "mod tidy", etc.
 
 	DebugActiongraph string // -debug-actiongraph flag (undocumented, unstable)
+	DebugTrace       string // -debug-trace flag
 )
 
 func defaultContext() build.Context {
diff --git a/src/cmd/go/internal/clean/clean.go b/src/cmd/go/internal/clean/clean.go
index 99704cb2b1..6bfd7ae21e 100644
--- a/src/cmd/go/internal/clean/clean.go
+++ b/src/cmd/go/internal/clean/clean.go
@@ -6,6 +6,7 @@
 package clean
 
 import (
+	"context"
 	"fmt"
 	"io/ioutil"
 	"os"
@@ -105,7 +106,7 @@ func init() {
 	work.AddBuildFlags(CmdClean, work.DefaultBuildFlags)
 }
 
-func runClean(cmd *base.Command, args []string) {
+func runClean(ctx context.Context, cmd *base.Command, args []string) {
 	// golang.org/issue/29925: only load packages before cleaning if
 	// either the flags and arguments explicitly imply a package,
 	// or no other target (such as a cache) was requested to be cleaned.
@@ -116,7 +117,7 @@ func runClean(cmd *base.Command, args []string) {
 	}
 
 	if cleanPkg {
-		for _, pkg := range load.PackagesAndErrors(args) {
+		for _, pkg := range load.PackagesAndErrors(ctx, args) {
 			clean(pkg)
 		}
 	}
diff --git a/src/cmd/go/internal/doc/doc.go b/src/cmd/go/internal/doc/doc.go
index 4ff08bb928..67f76e2256 100644
--- a/src/cmd/go/internal/doc/doc.go
+++ b/src/cmd/go/internal/doc/doc.go
@@ -8,6 +8,7 @@ package doc
 import (
 	"cmd/go/internal/base"
 	"cmd/go/internal/cfg"
+	"context"
 )
 
 var CmdDoc = &base.Command{
@@ -129,6 +130,6 @@ Flags:
 `,
 }
 
-func runDoc(cmd *base.Command, args []string) {
+func runDoc(ctx context.Context, cmd *base.Command, args []string) {
 	base.Run(cfg.BuildToolexec, base.Tool("doc"), args)
 }
diff --git a/src/cmd/go/internal/envcmd/env.go b/src/cmd/go/internal/envcmd/env.go
index 252025dc25..7bd75f7305 100644
--- a/src/cmd/go/internal/envcmd/env.go
+++ b/src/cmd/go/internal/envcmd/env.go
@@ -6,6 +6,7 @@
 package envcmd
 
 import (
+	"context"
 	"encoding/json"
 	"fmt"
 	"go/build"
@@ -62,9 +63,6 @@ var (
 )
 
 func MkEnv() []cfg.EnvVar {
-	var b work.Builder
-	b.Init()
-
 	envFile, _ := cfg.EnvFile()
 	env := []cfg.EnvVar{
 		{Name: "GO111MODULE", Value: cfg.Getenv("GO111MODULE")},
@@ -186,7 +184,7 @@ func argKey(arg string) string {
 	return arg[:i]
 }
 
-func runEnv(cmd *base.Command, args []string) {
+func runEnv(ctx context.Context, cmd *base.Command, args []string) {
 	if *envJson && *envU {
 		base.Fatalf("go env: cannot use -json with -u")
 	}
diff --git a/src/cmd/go/internal/fix/fix.go b/src/cmd/go/internal/fix/fix.go
index 4d741df2b4..825624fcbb 100644
--- a/src/cmd/go/internal/fix/fix.go
+++ b/src/cmd/go/internal/fix/fix.go
@@ -11,6 +11,7 @@ import (
 	"cmd/go/internal/load"
 	"cmd/go/internal/modload"
 	"cmd/go/internal/str"
+	"context"
 	"fmt"
 	"os"
 )
@@ -31,9 +32,9 @@ See also: go fmt, go vet.
 	`,
 }
 
-func runFix(cmd *base.Command, args []string) {
+func runFix(ctx context.Context, cmd *base.Command, args []string) {
 	printed := false
-	for _, pkg := range load.Packages(args) {
+	for _, pkg := range load.Packages(ctx, args) {
 		if modload.Enabled() && pkg.Module != nil && !pkg.Module.Main {
 			if !printed {
 				fmt.Fprintf(os.Stderr, "go: not fixing packages in dependency modules\n")
diff --git a/src/cmd/go/internal/fmtcmd/fmt.go b/src/cmd/go/internal/fmtcmd/fmt.go
index d6894edc9f..f96cff429c 100644
--- a/src/cmd/go/internal/fmtcmd/fmt.go
+++ b/src/cmd/go/internal/fmtcmd/fmt.go
@@ -6,6 +6,7 @@
 package fmtcmd
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"os"
@@ -48,7 +49,7 @@ See also: go fix, go vet.
 	`,
 }
 
-func runFmt(cmd *base.Command, args []string) {
+func runFmt(ctx context.Context, cmd *base.Command, args []string) {
 	printed := false
 	gofmt := gofmtPath()
 	procs := runtime.GOMAXPROCS(0)
@@ -63,7 +64,7 @@ func runFmt(cmd *base.Command, args []string) {
 			}
 		}()
 	}
-	for _, pkg := range load.PackagesAndErrors(args) {
+	for _, pkg := range load.PackagesAndErrors(ctx, args) {
 		if modload.Enabled() && pkg.Module != nil && !pkg.Module.Main {
 			if !printed {
 				fmt.Fprintf(os.Stderr, "go: not formatting packages in dependency modules\n")
diff --git a/src/cmd/go/internal/generate/generate.go b/src/cmd/go/internal/generate/generate.go
index 093b19817b..98c17bba8c 100644
--- a/src/cmd/go/internal/generate/generate.go
+++ b/src/cmd/go/internal/generate/generate.go
@@ -8,6 +8,7 @@ package generate
 import (
 	"bufio"
 	"bytes"
+	"context"
 	"fmt"
 	"go/parser"
 	"go/token"
@@ -160,7 +161,7 @@ func init() {
 	CmdGenerate.Flag.StringVar(&generateRunFlag, "run", "", "")
 }
 
-func runGenerate(cmd *base.Command, args []string) {
+func runGenerate(ctx context.Context, cmd *base.Command, args []string) {
 	load.IgnoreImports = true
 
 	if generateRunFlag != "" {
@@ -175,7 +176,7 @@ func runGenerate(cmd *base.Command, args []string) {
 
 	// Even if the arguments are .go files, this loop suffices.
 	printed := false
-	for _, pkg := range load.PackagesAndErrors(args) {
+	for _, pkg := range load.PackagesAndErrors(ctx, args) {
 		if modload.Enabled() && pkg.Module != nil && !pkg.Module.Main {
 			if !printed {
 				fmt.Fprintf(os.Stderr, "go: not generating in packages in dependency modules\n")
diff --git a/src/cmd/go/internal/get/get.go b/src/cmd/go/internal/get/get.go
index d38350c2a8..e5bacadaa3 100644
--- a/src/cmd/go/internal/get/get.go
+++ b/src/cmd/go/internal/get/get.go
@@ -6,6 +6,7 @@
 package get
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -112,7 +113,7 @@ func init() {
 	CmdGet.Flag.BoolVar(&Insecure, "insecure", Insecure, "")
 }
 
-func runGet(cmd *base.Command, args []string) {
+func runGet(ctx context.Context, cmd *base.Command, args []string) {
 	if cfg.ModulesEnabled {
 		// Should not happen: main.go should install the separate module-enabled get code.
 		base.Fatalf("go get: modules not implemented")
@@ -171,7 +172,7 @@ func runGet(cmd *base.Command, args []string) {
 	// everything.
 	load.ClearPackageCache()
 
-	pkgs := load.PackagesForBuild(args)
+	pkgs := load.PackagesForBuild(ctx, args)
 
 	// Phase 3. Install.
 	if *getD {
@@ -181,7 +182,7 @@ func runGet(cmd *base.Command, args []string) {
 		return
 	}
 
-	work.InstallPackages(args, pkgs)
+	work.InstallPackages(ctx, args, pkgs)
 }
 
 // downloadPaths prepares the list of paths to pass to download.
@@ -245,9 +246,9 @@ func download(arg string, parent *load.Package, stk *load.ImportStack, mode int)
 	load1 := func(path string, mode int) *load.Package {
 		if parent == nil {
 			mode := 0 // don't do module or vendor resolution
-			return load.LoadImport(path, base.Cwd, nil, stk, nil, mode)
+			return load.LoadImport(context.TODO(), path, base.Cwd, nil, stk, nil, mode)
 		}
-		return load.LoadImport(path, parent.Dir, parent, stk, nil, mode|load.ResolveModule)
+		return load.LoadImport(context.TODO(), path, parent.Dir, parent, stk, nil, mode|load.ResolveModule)
 	}
 
 	p := load1(arg, mode)
diff --git a/src/cmd/go/internal/list/list.go b/src/cmd/go/internal/list/list.go
index 6ca1561121..e68c39f392 100644
--- a/src/cmd/go/internal/list/list.go
+++ b/src/cmd/go/internal/list/list.go
@@ -8,6 +8,7 @@ package list
 import (
 	"bufio"
 	"bytes"
+	"context"
 	"encoding/json"
 	"io"
 	"os"
@@ -19,6 +20,7 @@ import (
 	"cmd/go/internal/cache"
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/load"
+	"cmd/go/internal/modinfo"
 	"cmd/go/internal/modload"
 	"cmd/go/internal/str"
 	"cmd/go/internal/work"
@@ -309,7 +311,7 @@ var (
 
 var nl = []byte{'\n'}
 
-func runList(cmd *base.Command, args []string) {
+func runList(ctx context.Context, cmd *base.Command, args []string) {
 	modload.LoadTests = *listTest
 	work.BuildInit()
 	out := newTrackingWriter(os.Stdout)
@@ -348,7 +350,7 @@ func runList(cmd *base.Command, args []string) {
 		fm := template.FuncMap{
 			"join":    strings.Join,
 			"context": context,
-			"module":  modload.ModuleInfo,
+			"module":  func(path string) *modinfo.ModulePublic { return modload.ModuleInfo(ctx, path) },
 		}
 		tmpl, err := template.New("main").Funcs(fm).Parse(*listFmt)
 		if err != nil {
@@ -388,7 +390,7 @@ func runList(cmd *base.Command, args []string) {
 			base.Fatalf("go list -m: not using modules")
 		}
 
-		modload.InitMod() // Parses go.mod and sets cfg.BuildMod.
+		modload.InitMod(ctx) // Parses go.mod and sets cfg.BuildMod.
 		if cfg.BuildMod == "vendor" {
 			const actionDisabledFormat = "go list -m: can't %s using the vendor directory\n\t(Use -mod=mod or -mod=readonly to bypass.)"
 
@@ -412,9 +414,9 @@ func runList(cmd *base.Command, args []string) {
 			}
 		}
 
-		modload.LoadBuildList()
+		modload.LoadBuildList(ctx)
 
-		mods := modload.ListModules(args, *listU, *listVersions)
+		mods := modload.ListModules(ctx, args, *listU, *listVersions)
 		if !*listE {
 			for _, m := range mods {
 				if m.Error != nil {
@@ -448,9 +450,9 @@ func runList(cmd *base.Command, args []string) {
 	load.IgnoreImports = *listFind
 	var pkgs []*load.Package
 	if *listE {
-		pkgs = load.PackagesAndErrors(args)
+		pkgs = load.PackagesAndErrors(ctx, args)
 	} else {
-		pkgs = load.Packages(args)
+		pkgs = load.Packages(ctx, args)
 		base.ExitIfErrors()
 	}
 
@@ -476,9 +478,9 @@ func runList(cmd *base.Command, args []string) {
 				var pmain, ptest, pxtest *load.Package
 				var err error
 				if *listE {
-					pmain, ptest, pxtest = load.TestPackagesAndErrors(p, nil)
+					pmain, ptest, pxtest = load.TestPackagesAndErrors(ctx, p, nil)
 				} else {
-					pmain, ptest, pxtest, err = load.TestPackagesFor(p, nil)
+					pmain, ptest, pxtest, err = load.TestPackagesFor(ctx, p, nil)
 					if err != nil {
 						base.Errorf("can't load test package: %s", err)
 					}
@@ -538,7 +540,7 @@ func runList(cmd *base.Command, args []string) {
 				a.Deps = append(a.Deps, b.AutoAction(work.ModeInstall, work.ModeInstall, p))
 			}
 		}
-		b.Do(a)
+		b.Do(ctx, a)
 	}
 
 	for _, p := range pkgs {
diff --git a/src/cmd/go/internal/load/pkg.go b/src/cmd/go/internal/load/pkg.go
index 2b5fbb1c5b..71fd9b5538 100644
--- a/src/cmd/go/internal/load/pkg.go
+++ b/src/cmd/go/internal/load/pkg.go
@@ -7,6 +7,7 @@ package load
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -30,6 +31,7 @@ import (
 	"cmd/go/internal/par"
 	"cmd/go/internal/search"
 	"cmd/go/internal/str"
+	"cmd/go/internal/trace"
 )
 
 var (
@@ -40,10 +42,10 @@ var (
 	ModBinDir            func() string                                                                            // return effective bin directory
 	ModLookup            func(parentPath string, parentIsStd bool, path string) (dir, realPath string, err error) // lookup effective meaning of import
 	ModPackageModuleInfo func(path string) *modinfo.ModulePublic                                                  // return module info for Package struct
-	ModImportPaths       func(args []string) []*search.Match                                                      // expand import paths
+	ModImportPaths       func(ctx context.Context, args []string) []*search.Match                                 // expand import paths
 	ModPackageBuildInfo  func(main string, deps []string) string                                                  // return module info to embed in binary
 	ModInfoProg          func(info string, isgccgo bool) []byte                                                   // wrap module info in .go code for binary
-	ModImportFromFiles   func([]string)                                                                           // update go.mod to add modules for imports in these files
+	ModImportFromFiles   func(context.Context, []string)                                                          // update go.mod to add modules for imports in these files
 	ModDirImportPath     func(string) string                                                                      // return effective import path for directory
 )
 
@@ -551,7 +553,7 @@ func ReloadPackageNoFlags(arg string, stk *ImportStack) *Package {
 		})
 		packageDataCache.Delete(p.ImportPath)
 	}
-	return LoadImport(arg, base.Cwd, nil, stk, nil, 0)
+	return LoadImport(context.TODO(), arg, base.Cwd, nil, stk, nil, 0)
 }
 
 // dirToImportPath returns the pseudo-import path we use for a package
@@ -603,11 +605,11 @@ const (
 // LoadImport does not set tool flags and should only be used by
 // this package, as part of a bigger load operation, and by GOPATH-based "go get".
 // TODO(rsc): When GOPATH-based "go get" is removed, unexport this function.
-func LoadImport(path, srcDir string, parent *Package, stk *ImportStack, importPos []token.Position, mode int) *Package {
-	return loadImport(nil, path, srcDir, parent, stk, importPos, mode)
+func LoadImport(ctx context.Context, path, srcDir string, parent *Package, stk *ImportStack, importPos []token.Position, mode int) *Package {
+	return loadImport(ctx, nil, path, srcDir, parent, stk, importPos, mode)
 }
 
-func loadImport(pre *preload, path, srcDir string, parent *Package, stk *ImportStack, importPos []token.Position, mode int) *Package {
+func loadImport(ctx context.Context, pre *preload, path, srcDir string, parent *Package, stk *ImportStack, importPos []token.Position, mode int) *Package {
 	if path == "" {
 		panic("LoadImport called with empty package path")
 	}
@@ -655,7 +657,7 @@ func loadImport(pre *preload, path, srcDir string, parent *Package, stk *ImportS
 		// Load package.
 		// loadPackageData may return bp != nil even if an error occurs,
 		// in order to return partial information.
-		p.load(path, stk, importPos, bp, err)
+		p.load(ctx, path, stk, importPos, bp, err)
 
 		if !cfg.ModulesEnabled && path != cleanImport(path) {
 			p.Error = &PackageError{
@@ -1589,7 +1591,7 @@ func (p *Package) DefaultExecName() string {
 // load populates p using information from bp, err, which should
 // be the result of calling build.Context.Import.
 // stk contains the import stack, not including path itself.
-func (p *Package) load(path string, stk *ImportStack, importPos []token.Position, bp *build.Package, err error) {
+func (p *Package) load(ctx context.Context, path string, stk *ImportStack, importPos []token.Position, bp *build.Package, err error) {
 	p.copyBuild(bp)
 
 	// The localPrefix is the path we interpret ./ imports relative to.
@@ -1798,7 +1800,7 @@ func (p *Package) load(path string, stk *ImportStack, importPos []token.Position
 		if path == "C" {
 			continue
 		}
-		p1 := LoadImport(path, p.Dir, p, stk, p.Internal.Build.ImportPos[path], ResolveImport)
+		p1 := LoadImport(ctx, path, p.Dir, p, stk, p.Internal.Build.ImportPos[path], ResolveImport)
 
 		path = p1.ImportPath
 		importPaths[i] = path
@@ -2071,7 +2073,7 @@ func PackageList(roots []*Package) []*Package {
 // TestPackageList returns the list of packages in the dag rooted at roots
 // as visited in a depth-first post-order traversal, including the test
 // imports of the roots. This ignores errors in test packages.
-func TestPackageList(roots []*Package) []*Package {
+func TestPackageList(ctx context.Context, roots []*Package) []*Package {
 	seen := map[*Package]bool{}
 	all := []*Package{}
 	var walk func(*Package)
@@ -2087,7 +2089,7 @@ func TestPackageList(roots []*Package) []*Package {
 	}
 	walkTest := func(root *Package, path string) {
 		var stk ImportStack
-		p1 := LoadImport(path, root.Dir, root, &stk, root.Internal.Build.TestImportPos[path], ResolveImport)
+		p1 := LoadImport(ctx, path, root.Dir, root, &stk, root.Internal.Build.TestImportPos[path], ResolveImport)
 		if p1.Error == nil {
 			walk(p1)
 		}
@@ -2110,7 +2112,7 @@ func TestPackageList(roots []*Package) []*Package {
 // TODO(jayconrod): delete this function and set flags automatically
 // in LoadImport instead.
 func LoadImportWithFlags(path, srcDir string, parent *Package, stk *ImportStack, importPos []token.Position, mode int) *Package {
-	p := LoadImport(path, srcDir, parent, stk, importPos, mode)
+	p := LoadImport(context.TODO(), path, srcDir, parent, stk, importPos, mode)
 	setToolFlags(p)
 	return p
 }
@@ -2123,9 +2125,9 @@ func LoadImportWithFlags(path, srcDir string, parent *Package, stk *ImportStack,
 // to load dependencies of a named package, the named
 // package is still returned, with p.Incomplete = true
 // and details in p.DepsErrors.
-func Packages(args []string) []*Package {
+func Packages(ctx context.Context, args []string) []*Package {
 	var pkgs []*Package
-	for _, pkg := range PackagesAndErrors(args) {
+	for _, pkg := range PackagesAndErrors(ctx, args) {
 		if pkg.Error != nil {
 			base.Errorf("%v", pkg.Error)
 			continue
@@ -2139,7 +2141,10 @@ func Packages(args []string) []*Package {
 // *Package for every argument, even the ones that
 // cannot be loaded at all.
 // The packages that fail to load will have p.Error != nil.
-func PackagesAndErrors(patterns []string) []*Package {
+func PackagesAndErrors(ctx context.Context, patterns []string) []*Package {
+	ctx, span := trace.StartSpan(ctx, "load.PackagesAndErrors")
+	defer span.Done()
+
 	for _, p := range patterns {
 		// Listing is only supported with all patterns referring to either:
 		// - Files that are part of the same directory.
@@ -2148,12 +2153,12 @@ func PackagesAndErrors(patterns []string) []*Package {
 			// We need to test whether the path is an actual Go file and not a
 			// package path or pattern ending in '.go' (see golang.org/issue/34653).
 			if fi, err := os.Stat(p); err == nil && !fi.IsDir() {
-				return []*Package{GoFilesPackage(patterns)}
+				return []*Package{GoFilesPackage(ctx, patterns)}
 			}
 		}
 	}
 
-	matches := ImportPaths(patterns)
+	matches := ImportPaths(ctx, patterns)
 	var (
 		pkgs    []*Package
 		stk     ImportStack
@@ -2169,7 +2174,7 @@ func PackagesAndErrors(patterns []string) []*Package {
 			if pkg == "" {
 				panic(fmt.Sprintf("ImportPaths returned empty package for pattern %s", m.Pattern()))
 			}
-			p := loadImport(pre, pkg, base.Cwd, nil, &stk, nil, 0)
+			p := loadImport(ctx, pre, pkg, base.Cwd, nil, &stk, nil, 0)
 			p.Match = append(p.Match, m.Pattern())
 			p.Internal.CmdlinePkg = true
 			if m.IsLiteral() {
@@ -2223,9 +2228,9 @@ func setToolFlags(pkgs ...*Package) {
 	}
 }
 
-func ImportPaths(args []string) []*search.Match {
+func ImportPaths(ctx context.Context, args []string) []*search.Match {
 	if ModInit(); cfg.ModulesEnabled {
-		return ModImportPaths(args)
+		return ModImportPaths(ctx, args)
 	}
 	return search.ImportPaths(args)
 }
@@ -2233,8 +2238,8 @@ func ImportPaths(args []string) []*search.Match {
 // PackagesForBuild is like Packages but exits
 // if any of the packages or their dependencies have errors
 // (cannot be built).
-func PackagesForBuild(args []string) []*Package {
-	pkgs := PackagesAndErrors(args)
+func PackagesForBuild(ctx context.Context, args []string) []*Package {
+	pkgs := PackagesAndErrors(ctx, args)
 	printed := map[*PackageError]bool{}
 	for _, pkg := range pkgs {
 		if pkg.Error != nil {
@@ -2276,7 +2281,7 @@ func PackagesForBuild(args []string) []*Package {
 // GoFilesPackage creates a package for building a collection of Go files
 // (typically named on the command line). The target is named p.a for
 // package p or named after the first Go file for package main.
-func GoFilesPackage(gofiles []string) *Package {
+func GoFilesPackage(ctx context.Context, gofiles []string) *Package {
 	ModInit()
 
 	for _, f := range gofiles {
@@ -2324,7 +2329,7 @@ func GoFilesPackage(gofiles []string) *Package {
 	ctxt.ReadDir = func(string) ([]os.FileInfo, error) { return dirent, nil }
 
 	if cfg.ModulesEnabled {
-		ModImportFromFiles(gofiles)
+		ModImportFromFiles(ctx, gofiles)
 	}
 
 	var err error
@@ -2340,7 +2345,7 @@ func GoFilesPackage(gofiles []string) *Package {
 	pkg := new(Package)
 	pkg.Internal.Local = true
 	pkg.Internal.CmdlineFiles = true
-	pkg.load("command-line-arguments", &stk, nil, bp, err)
+	pkg.load(ctx, "command-line-arguments", &stk, nil, bp, err)
 	pkg.Internal.LocalPrefix = dirToImportPath(dir)
 	pkg.ImportPath = "command-line-arguments"
 	pkg.Target = ""
diff --git a/src/cmd/go/internal/load/test.go b/src/cmd/go/internal/load/test.go
index 6d251e8358..a0e275095b 100644
--- a/src/cmd/go/internal/load/test.go
+++ b/src/cmd/go/internal/load/test.go
@@ -6,7 +6,7 @@ package load
 
 import (
 	"bytes"
-	"cmd/go/internal/str"
+	"context"
 	"errors"
 	"fmt"
 	"go/ast"
@@ -20,6 +20,9 @@ import (
 	"strings"
 	"unicode"
 	"unicode/utf8"
+
+	"cmd/go/internal/str"
+	"cmd/go/internal/trace"
 )
 
 var TestMainDeps = []string{
@@ -42,8 +45,8 @@ type TestCover struct {
 // TestPackagesFor is like TestPackagesAndErrors but it returns
 // an error if the test packages or their dependencies have errors.
 // Only test packages without errors are returned.
-func TestPackagesFor(p *Package, cover *TestCover) (pmain, ptest, pxtest *Package, err error) {
-	pmain, ptest, pxtest = TestPackagesAndErrors(p, cover)
+func TestPackagesFor(ctx context.Context, p *Package, cover *TestCover) (pmain, ptest, pxtest *Package, err error) {
+	pmain, ptest, pxtest = TestPackagesAndErrors(ctx, p, cover)
 	for _, p1 := range []*Package{ptest, pxtest, pmain} {
 		if p1 == nil {
 			// pxtest may be nil
@@ -89,7 +92,10 @@ func TestPackagesFor(p *Package, cover *TestCover) (pmain, ptest, pxtest *Packag
 //
 // The caller is expected to have checked that len(p.TestGoFiles)+len(p.XTestGoFiles) > 0,
 // or else there's no point in any of this.
-func TestPackagesAndErrors(p *Package, cover *TestCover) (pmain, ptest, pxtest *Package) {
+func TestPackagesAndErrors(ctx context.Context, p *Package, cover *TestCover) (pmain, ptest, pxtest *Package) {
+	ctx, span := trace.StartSpan(ctx, "load.TestPackagesAndErrors")
+	defer span.Done()
+
 	pre := newPreload()
 	defer pre.flush()
 	allImports := append([]string{}, p.TestImports...)
@@ -102,7 +108,7 @@ func TestPackagesAndErrors(p *Package, cover *TestCover) (pmain, ptest, pxtest *
 	stk.Push(p.ImportPath + " (test)")
 	rawTestImports := str.StringList(p.TestImports)
 	for i, path := range p.TestImports {
-		p1 := loadImport(pre, path, p.Dir, p, &stk, p.Internal.Build.TestImportPos[path], ResolveImport)
+		p1 := loadImport(ctx, pre, path, p.Dir, p, &stk, p.Internal.Build.TestImportPos[path], ResolveImport)
 		if str.Contains(p1.Deps, p.ImportPath) || p1.ImportPath == p.ImportPath {
 			// Same error that loadPackage returns (via reusePackage) in pkg.go.
 			// Can't change that code, because that code is only for loading the
@@ -121,7 +127,7 @@ func TestPackagesAndErrors(p *Package, cover *TestCover) (pmain, ptest, pxtest *
 	pxtestNeedsPtest := false
 	rawXTestImports := str.StringList(p.XTestImports)
 	for i, path := range p.XTestImports {
-		p1 := loadImport(pre, path, p.Dir, p, &stk, p.Internal.Build.XTestImportPos[path], ResolveImport)
+		p1 := loadImport(ctx, pre, path, p.Dir, p, &stk, p.Internal.Build.XTestImportPos[path], ResolveImport)
 		if p1.ImportPath == p.ImportPath {
 			pxtestNeedsPtest = true
 		} else {
@@ -238,7 +244,7 @@ func TestPackagesAndErrors(p *Package, cover *TestCover) (pmain, ptest, pxtest *
 		if dep == ptest.ImportPath {
 			pmain.Internal.Imports = append(pmain.Internal.Imports, ptest)
 		} else {
-			p1 := loadImport(pre, dep, "", nil, &stk, nil, 0)
+			p1 := loadImport(ctx, pre, dep, "", nil, &stk, nil, 0)
 			pmain.Internal.Imports = append(pmain.Internal.Imports, p1)
 		}
 	}
diff --git a/src/cmd/go/internal/modcmd/download.go b/src/cmd/go/internal/modcmd/download.go
index 584434935b..d4c161fca1 100644
--- a/src/cmd/go/internal/modcmd/download.go
+++ b/src/cmd/go/internal/modcmd/download.go
@@ -5,14 +5,15 @@
 package modcmd
 
 import (
+	"context"
 	"encoding/json"
 	"os"
+	"runtime"
 
 	"cmd/go/internal/base"
 	"cmd/go/internal/cfg"
-	"cmd/go/internal/modfetch"
 	"cmd/go/internal/modload"
-	"cmd/go/internal/par"
+	"cmd/go/internal/modfetch"
 	"cmd/go/internal/work"
 
 	"golang.org/x/mod/module"
@@ -78,7 +79,7 @@ type moduleJSON struct {
 	GoModSum string `json:",omitempty"`
 }
 
-func runDownload(cmd *base.Command, args []string) {
+func runDownload(ctx context.Context, cmd *base.Command, args []string) {
 	// Check whether modules are enabled and whether we're in a module.
 	if cfg.Getenv("GO111MODULE") == "off" {
 		base.Fatalf("go: modules disabled by GO111MODULE=off; see 'go help modules'")
@@ -89,7 +90,7 @@ func runDownload(cmd *base.Command, args []string) {
 	if len(args) == 0 {
 		args = []string{"all"}
 	} else if modload.HasModRoot() {
-		modload.InitMod() // to fill Target
+		modload.InitMod(ctx) // to fill Target
 		targetAtLatest := modload.Target.Path + "@latest"
 		targetAtUpgrade := modload.Target.Path + "@upgrade"
 		targetAtPatch := modload.Target.Path + "@patch"
@@ -101,33 +102,7 @@ func runDownload(cmd *base.Command, args []string) {
 		}
 	}
 
-	var mods []*moduleJSON
-	var work par.Work
-	listU := false
-	listVersions := false
-	for _, info := range modload.ListModules(args, listU, listVersions) {
-		if info.Replace != nil {
-			info = info.Replace
-		}
-		if info.Version == "" && info.Error == nil {
-			// main module or module replaced with file path.
-			// Nothing to download.
-			continue
-		}
-		m := &moduleJSON{
-			Path:    info.Path,
-			Version: info.Version,
-		}
-		mods = append(mods, m)
-		if info.Error != nil {
-			m.Error = info.Error.Err
-			continue
-		}
-		work.Add(m)
-	}
-
-	work.Do(10, func(item interface{}) {
-		m := item.(*moduleJSON)
+	downloadModule := func(m *moduleJSON) {
 		var err error
 		m.Info, err = modfetch.InfoFile(m.Path, m.Version)
 		if err != nil {
@@ -145,18 +120,53 @@ func runDownload(cmd *base.Command, args []string) {
 			return
 		}
 		mod := module.Version{Path: m.Path, Version: m.Version}
-		m.Zip, err = modfetch.DownloadZip(mod)
+		m.Zip, err = modfetch.DownloadZip(ctx, mod)
 		if err != nil {
 			m.Error = err.Error()
 			return
 		}
 		m.Sum = modfetch.Sum(mod)
-		m.Dir, err = modfetch.Download(mod)
+		m.Dir, err = modfetch.Download(ctx, mod)
 		if err != nil {
 			m.Error = err.Error()
 			return
 		}
-	})
+	}
+
+	var mods []*moduleJSON
+	listU := false
+	listVersions := false
+	type token struct{}
+	sem := make(chan token, runtime.GOMAXPROCS(0))
+	for _, info := range modload.ListModules(ctx, args, listU, listVersions) {
+		if info.Replace != nil {
+			info = info.Replace
+		}
+		if info.Version == "" && info.Error == nil {
+			// main module or module replaced with file path.
+			// Nothing to download.
+			continue
+		}
+		m := &moduleJSON{
+			Path:    info.Path,
+			Version: info.Version,
+		}
+		mods = append(mods, m)
+		if info.Error != nil {
+			m.Error = info.Error.Err
+			continue
+		}
+		sem <- token{}
+		go func() {
+			downloadModule(m)
+			<-sem
+		}()
+	}
+
+	// Fill semaphore channel to wait for goroutines to finish.
+	for n := cap(sem); n > 0; n-- {
+		sem <- token{}
+	}
 
 	if *downloadJSON {
 		for _, m := range mods {
diff --git a/src/cmd/go/internal/modcmd/edit.go b/src/cmd/go/internal/modcmd/edit.go
index dbbfb96e42..a81c25270f 100644
--- a/src/cmd/go/internal/modcmd/edit.go
+++ b/src/cmd/go/internal/modcmd/edit.go
@@ -8,6 +8,7 @@ package modcmd
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -141,7 +142,7 @@ func init() {
 	base.AddBuildFlagsNX(&cmdEdit.Flag)
 }
 
-func runEdit(cmd *base.Command, args []string) {
+func runEdit(ctx context.Context, cmd *base.Command, args []string) {
 	anyFlags :=
 		*editModule != "" ||
 			*editGo != "" ||
diff --git a/src/cmd/go/internal/modcmd/graph.go b/src/cmd/go/internal/modcmd/graph.go
index 27ae9354f3..6da12b9cab 100644
--- a/src/cmd/go/internal/modcmd/graph.go
+++ b/src/cmd/go/internal/modcmd/graph.go
@@ -8,13 +8,13 @@ package modcmd
 
 import (
 	"bufio"
+	"context"
 	"os"
 	"sort"
 
 	"cmd/go/internal/base"
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/modload"
-	"cmd/go/internal/par"
 	"cmd/go/internal/work"
 
 	"golang.org/x/mod/module"
@@ -36,7 +36,7 @@ func init() {
 	work.AddModCommonFlags(cmdGraph)
 }
 
-func runGraph(cmd *base.Command, args []string) {
+func runGraph(ctx context.Context, cmd *base.Command, args []string) {
 	if len(args) > 0 {
 		base.Fatalf("go mod graph: graph takes no arguments")
 	}
@@ -48,7 +48,7 @@ func runGraph(cmd *base.Command, args []string) {
 			base.Fatalf("go: cannot find main module; see 'go help modules'")
 		}
 	}
-	modload.LoadBuildList()
+	modload.LoadBuildList(ctx)
 
 	reqs := modload.MinReqs()
 	format := func(m module.Version) string {
@@ -58,23 +58,25 @@ func runGraph(cmd *base.Command, args []string) {
 		return m.Path + "@" + m.Version
 	}
 
-	// Note: using par.Work only to manage work queue.
-	// No parallelism here, so no locking.
 	var out []string
 	var deps int // index in out where deps start
-	var work par.Work
-	work.Add(modload.Target)
-	work.Do(1, func(item interface{}) {
-		m := item.(module.Version)
+	seen := map[module.Version]bool{modload.Target: true}
+	queue := []module.Version{modload.Target}
+	for len(queue) > 0 {
+		var m module.Version
+		m, queue = queue[0], queue[1:]
 		list, _ := reqs.Required(m)
 		for _, r := range list {
-			work.Add(r)
+			if !seen[r] {
+				queue = append(queue, r)
+				seen[r] = true
+			}
 			out = append(out, format(m)+" "+format(r)+"\n")
 		}
 		if m == modload.Target {
 			deps = len(out)
 		}
-	})
+	}
 
 	sort.Slice(out[deps:], func(i, j int) bool {
 		return out[deps+i][0] < out[deps+j][0]
diff --git a/src/cmd/go/internal/modcmd/init.go b/src/cmd/go/internal/modcmd/init.go
index 714ff2e205..b6cffd332d 100644
--- a/src/cmd/go/internal/modcmd/init.go
+++ b/src/cmd/go/internal/modcmd/init.go
@@ -10,6 +10,7 @@ import (
 	"cmd/go/internal/base"
 	"cmd/go/internal/modload"
 	"cmd/go/internal/work"
+	"context"
 	"os"
 	"strings"
 )
@@ -32,7 +33,7 @@ func init() {
 	work.AddModCommonFlags(cmdInit)
 }
 
-func runInit(cmd *base.Command, args []string) {
+func runInit(ctx context.Context, cmd *base.Command, args []string) {
 	modload.CmdModInit = true
 	if len(args) > 1 {
 		base.Fatalf("go mod init: too many arguments")
@@ -50,5 +51,6 @@ func runInit(cmd *base.Command, args []string) {
 	if strings.Contains(modload.CmdModModule, "@") {
 		base.Fatalf("go mod init: module path must not contain '@'")
 	}
-	modload.InitMod() // does all the hard work
+	modload.InitMod(ctx) // does all the hard work
+	modload.WriteGoMod()
 }
diff --git a/src/cmd/go/internal/modcmd/tidy.go b/src/cmd/go/internal/modcmd/tidy.go
index af2b04c0c2..c7c53d7c0c 100644
--- a/src/cmd/go/internal/modcmd/tidy.go
+++ b/src/cmd/go/internal/modcmd/tidy.go
@@ -9,11 +9,9 @@ package modcmd
 import (
 	"cmd/go/internal/base"
 	"cmd/go/internal/cfg"
-	"cmd/go/internal/modfetch"
 	"cmd/go/internal/modload"
 	"cmd/go/internal/work"
-
-	"golang.org/x/mod/module"
+	"context"
 )
 
 var cmdTidy = &base.Command{
@@ -37,46 +35,13 @@ func init() {
 	work.AddModCommonFlags(cmdTidy)
 }
 
-func runTidy(cmd *base.Command, args []string) {
+func runTidy(ctx context.Context, cmd *base.Command, args []string) {
 	if len(args) > 0 {
 		base.Fatalf("go mod tidy: no arguments allowed")
 	}
 
-	modload.LoadALL()
+	modload.LoadALL(ctx)
 	modload.TidyBuildList()
-	modTidyGoSum() // updates memory copy; WriteGoMod on next line flushes it out
+	modload.TrimGoSum()
 	modload.WriteGoMod()
 }
-
-// modTidyGoSum resets the go.sum file content
-// to be exactly what's needed for the current go.mod.
-func modTidyGoSum() {
-	// Assuming go.sum already has at least enough from the successful load,
-	// we only have to tell modfetch what needs keeping.
-	reqs := modload.Reqs()
-	keep := make(map[module.Version]bool)
-	replaced := make(map[module.Version]bool)
-	var walk func(module.Version)
-	walk = func(m module.Version) {
-		// If we build using a replacement module, keep the sum for the replacement,
-		// since that's the code we'll actually use during a build.
-		//
-		// TODO(golang.org/issue/29182): Perhaps we should keep both sums, and the
-		// sums for both sets of transitive requirements.
-		r := modload.Replacement(m)
-		if r.Path == "" {
-			keep[m] = true
-		} else {
-			keep[r] = true
-			replaced[m] = true
-		}
-		list, _ := reqs.Required(m)
-		for _, r := range list {
-			if !keep[r] && !replaced[r] {
-				walk(r)
-			}
-		}
-	}
-	walk(modload.Target)
-	modfetch.TrimGoSum(keep)
-}
diff --git a/src/cmd/go/internal/modcmd/vendor.go b/src/cmd/go/internal/modcmd/vendor.go
index 8509ceb7a8..e5353b5c7f 100644
--- a/src/cmd/go/internal/modcmd/vendor.go
+++ b/src/cmd/go/internal/modcmd/vendor.go
@@ -6,6 +6,7 @@ package modcmd
 
 import (
 	"bytes"
+	"context"
 	"fmt"
 	"io"
 	"io/ioutil"
@@ -43,11 +44,11 @@ func init() {
 	work.AddModCommonFlags(cmdVendor)
 }
 
-func runVendor(cmd *base.Command, args []string) {
+func runVendor(ctx context.Context, cmd *base.Command, args []string) {
 	if len(args) != 0 {
 		base.Fatalf("go mod vendor: vendor takes no arguments")
 	}
-	pkgs := modload.LoadVendor()
+	pkgs := modload.LoadVendor(ctx)
 
 	vdir := filepath.Join(modload.ModRoot(), "vendor")
 	if err := os.RemoveAll(vdir); err != nil {
diff --git a/src/cmd/go/internal/modcmd/verify.go b/src/cmd/go/internal/modcmd/verify.go
index b7fd7fa8e0..73ab714d10 100644
--- a/src/cmd/go/internal/modcmd/verify.go
+++ b/src/cmd/go/internal/modcmd/verify.go
@@ -6,6 +6,7 @@ package modcmd
 
 import (
 	"bytes"
+	"context"
 	"errors"
 	"fmt"
 	"io/ioutil"
@@ -40,7 +41,7 @@ func init() {
 	work.AddModCommonFlags(cmdVerify)
 }
 
-func runVerify(cmd *base.Command, args []string) {
+func runVerify(ctx context.Context, cmd *base.Command, args []string) {
 	if len(args) != 0 {
 		// NOTE(rsc): Could take a module pattern.
 		base.Fatalf("go mod verify: verify takes no arguments")
@@ -59,7 +60,7 @@ func runVerify(cmd *base.Command, args []string) {
 	sem := make(chan token, runtime.GOMAXPROCS(0))
 
 	// Use a slice of result channels, so that the output is deterministic.
-	mods := modload.LoadBuildList()[1:]
+	mods := modload.LoadBuildList(ctx)[1:]
 	errsChans := make([]<-chan []error, len(mods))
 
 	for i, mod := range mods {
diff --git a/src/cmd/go/internal/modcmd/why.go b/src/cmd/go/internal/modcmd/why.go
index 40d238519b..da33fff89e 100644
--- a/src/cmd/go/internal/modcmd/why.go
+++ b/src/cmd/go/internal/modcmd/why.go
@@ -5,6 +5,7 @@
 package modcmd
 
 import (
+	"context"
 	"fmt"
 	"strings"
 
@@ -60,7 +61,7 @@ func init() {
 	work.AddModCommonFlags(cmdWhy)
 }
 
-func runWhy(cmd *base.Command, args []string) {
+func runWhy(ctx context.Context, cmd *base.Command, args []string) {
 	loadALL := modload.LoadALL
 	if *whyVendor {
 		loadALL = modload.LoadVendor
@@ -73,9 +74,9 @@ func runWhy(cmd *base.Command, args []string) {
 				base.Fatalf("go mod why: module query not allowed")
 			}
 		}
-		mods := modload.ListModules(args, listU, listVersions)
+		mods := modload.ListModules(ctx, args, listU, listVersions)
 		byModule := make(map[module.Version][]string)
-		for _, path := range loadALL() {
+		for _, path := range loadALL(ctx) {
 			m := modload.PackageModule(path)
 			if m.Path != "" {
 				byModule[m] = append(byModule[m], path)
@@ -104,8 +105,8 @@ func runWhy(cmd *base.Command, args []string) {
 			sep = "\n"
 		}
 	} else {
-		matches := modload.ImportPaths(args) // resolve to packages
-		loadALL()                            // rebuild graph, from main module (not from named packages)
+		matches := modload.ImportPaths(ctx, args) // resolve to packages
+		loadALL(ctx)                              // rebuild graph, from main module (not from named packages)
 		sep := ""
 		for _, m := range matches {
 			for _, path := range m.Pkgs {
diff --git a/src/cmd/go/internal/modconv/convert.go b/src/cmd/go/internal/modconv/convert.go
index f465a9f395..5d4165c944 100644
--- a/src/cmd/go/internal/modconv/convert.go
+++ b/src/cmd/go/internal/modconv/convert.go
@@ -7,13 +7,12 @@ package modconv
 import (
 	"fmt"
 	"os"
+	"runtime"
 	"sort"
 	"strings"
-	"sync"
 
 	"cmd/go/internal/base"
 	"cmd/go/internal/modfetch"
-	"cmd/go/internal/par"
 
 	"golang.org/x/mod/modfile"
 	"golang.org/x/mod/module"
@@ -42,46 +41,54 @@ func ConvertLegacyConfig(f *modfile.File, file string, data []byte) error {
 
 	// Convert requirements block, which may use raw SHA1 hashes as versions,
 	// to valid semver requirement list, respecting major versions.
-	var (
-		work    par.Work
-		mu      sync.Mutex
-		need    = make(map[string]string)
-		replace = make(map[string]*modfile.Replace)
-	)
+	versions := make([]module.Version, len(mf.Require))
+	replace := make(map[string]*modfile.Replace)
 
 	for _, r := range mf.Replace {
 		replace[r.New.Path] = r
 		replace[r.Old.Path] = r
 	}
-	for _, r := range mf.Require {
+
+	type token struct{}
+	sem := make(chan token, runtime.GOMAXPROCS(0))
+	for i, r := range mf.Require {
 		m := r.Mod
 		if m.Path == "" {
 			continue
 		}
 		if re, ok := replace[m.Path]; ok {
-			work.Add(re.New)
-			continue
+			m = re.New
 		}
-		work.Add(r.Mod)
+		sem <- token{}
+		go func(i int, m module.Version) {
+			defer func() { <-sem }()
+			repo, info, err := modfetch.ImportRepoRev(m.Path, m.Version)
+			if err != nil {
+				fmt.Fprintf(os.Stderr, "go: converting %s: stat %s@%s: %v\n", base.ShortPath(file), m.Path, m.Version, err)
+				return
+			}
+
+			path := repo.ModulePath()
+			versions[i].Path = path
+			versions[i].Version = info.Version
+		}(i, m)
+	}
+	// Fill semaphore channel to wait for all tasks to finish.
+	for n := cap(sem); n > 0; n-- {
+		sem <- token{}
 	}
 
-	work.Do(10, func(item interface{}) {
-		r := item.(module.Version)
-		repo, info, err := modfetch.ImportRepoRev(r.Path, r.Version)
-		if err != nil {
-			fmt.Fprintf(os.Stderr, "go: converting %s: stat %s@%s: %v\n", base.ShortPath(file), r.Path, r.Version, err)
-			return
+	need := map[string]string{}
+	for _, v := range versions {
+		if v.Path == "" {
+			continue
 		}
-		mu.Lock()
-		path := repo.ModulePath()
 		// Don't use semver.Max here; need to preserve +incompatible suffix.
-		if v, ok := need[path]; !ok || semver.Compare(v, info.Version) < 0 {
-			need[path] = info.Version
+		if needv, ok := need[v.Path]; !ok || semver.Compare(needv, v.Version) < 0 {
+			need[v.Path] = v.Version
 		}
-		mu.Unlock()
-	})
-
-	var paths []string
+	}
+	paths := make([]string, 0, len(need))
 	for path := range need {
 		paths = append(paths, path)
 	}
diff --git a/src/cmd/go/internal/modconv/convert_test.go b/src/cmd/go/internal/modconv/convert_test.go
index a04a13b14f..faa2b4c606 100644
--- a/src/cmd/go/internal/modconv/convert_test.go
+++ b/src/cmd/go/internal/modconv/convert_test.go
@@ -6,6 +6,7 @@ package modconv
 
 import (
 	"bytes"
+	"context"
 	"fmt"
 	"internal/testenv"
 	"io/ioutil"
@@ -146,6 +147,8 @@ func TestConvertLegacyConfig(t *testing.T) {
 		},
 	}
 
+	ctx := context.Background()
+
 	for _, tt := range tests {
 		t.Run(strings.ReplaceAll(tt.path, "/", "_")+"_"+tt.vers, func(t *testing.T) {
 			f, err := modfile.Parse("golden", []byte(tt.gomod), nil)
@@ -157,7 +160,7 @@ func TestConvertLegacyConfig(t *testing.T) {
 				t.Fatal(err)
 			}
 
-			dir, err := modfetch.Download(module.Version{Path: tt.path, Version: tt.vers})
+			dir, err := modfetch.Download(ctx, module.Version{Path: tt.path, Version: tt.vers})
 			if err != nil {
 				t.Fatal(err)
 			}
diff --git a/src/cmd/go/internal/modfetch/fetch.go b/src/cmd/go/internal/modfetch/fetch.go
index fd7a5cef83..e29eb0a942 100644
--- a/src/cmd/go/internal/modfetch/fetch.go
+++ b/src/cmd/go/internal/modfetch/fetch.go
@@ -7,6 +7,7 @@ package modfetch
 import (
 	"archive/zip"
 	"bytes"
+	"context"
 	"errors"
 	"fmt"
 	"io"
@@ -23,6 +24,7 @@ import (
 	"cmd/go/internal/par"
 	"cmd/go/internal/renameio"
 	"cmd/go/internal/robustio"
+	"cmd/go/internal/trace"
 
 	"golang.org/x/mod/module"
 	"golang.org/x/mod/sumdb/dirhash"
@@ -34,7 +36,7 @@ var downloadCache par.Cache
 // Download downloads the specific module version to the
 // local download cache and returns the name of the directory
 // corresponding to the root of the module's file tree.
-func Download(mod module.Version) (dir string, err error) {
+func Download(ctx context.Context, mod module.Version) (dir string, err error) {
 	if cfg.GOMODCACHE == "" {
 		// modload.Init exits if GOPATH[0] is empty, and cfg.GOMODCACHE
 		// is set to GOPATH[0]/pkg/mod if GOMODCACHE is empty, so this should never happen.
@@ -47,7 +49,7 @@ func Download(mod module.Version) (dir string, err error) {
 		err error
 	}
 	c := downloadCache.Do(mod, func() interface{} {
-		dir, err := download(mod)
+		dir, err := download(ctx, mod)
 		if err != nil {
 			return cached{"", err}
 		}
@@ -57,7 +59,10 @@ func Download(mod module.Version) (dir string, err error) {
 	return c.dir, c.err
 }
 
-func download(mod module.Version) (dir string, err error) {
+func download(ctx context.Context, mod module.Version) (dir string, err error) {
+	ctx, span := trace.StartSpan(ctx, "modfetch.download "+mod.String())
+	defer span.Done()
+
 	// If the directory exists, and no .partial file exists, the module has
 	// already been completely extracted. .partial files may be created when a
 	// module zip directory is extracted in place instead of being extracted to a
@@ -72,7 +77,7 @@ func download(mod module.Version) (dir string, err error) {
 	// To avoid cluttering the cache with extraneous files,
 	// DownloadZip uses the same lockfile as Download.
 	// Invoke DownloadZip before locking the file.
-	zipfile, err := DownloadZip(mod)
+	zipfile, err := DownloadZip(ctx, mod)
 	if err != nil {
 		return "", err
 	}
@@ -142,6 +147,7 @@ func download(mod module.Version) (dir string, err error) {
 		return "", err
 	}
 
+	ctx, span = trace.StartSpan(ctx, "unzip "+zipfile)
 	if unzipInPlace {
 		if err := ioutil.WriteFile(partialPath, nil, 0666); err != nil {
 			return "", err
@@ -171,6 +177,7 @@ func download(mod module.Version) (dir string, err error) {
 			return "", err
 		}
 	}
+	defer span.Done()
 
 	if !cfg.ModCacheRW {
 		// Make dir read-only only *after* renaming it.
@@ -195,7 +202,7 @@ var downloadZipCache par.Cache
 
 // DownloadZip downloads the specific module version to the
 // local zip cache and returns the name of the zip file.
-func DownloadZip(mod module.Version) (zipfile string, err error) {
+func DownloadZip(ctx context.Context, mod module.Version) (zipfile string, err error) {
 	// The par.Cache here avoids duplicate work.
 	type cached struct {
 		zipfile string
@@ -230,7 +237,7 @@ func DownloadZip(mod module.Version) (zipfile string, err error) {
 		if err := os.MkdirAll(filepath.Dir(zipfile), 0777); err != nil {
 			return cached{"", err}
 		}
-		if err := downloadZip(mod, zipfile); err != nil {
+		if err := downloadZip(ctx, mod, zipfile); err != nil {
 			return cached{"", err}
 		}
 		return cached{zipfile, nil}
@@ -238,7 +245,10 @@ func DownloadZip(mod module.Version) (zipfile string, err error) {
 	return c.zipfile, c.err
 }
 
-func downloadZip(mod module.Version, zipfile string) (err error) {
+func downloadZip(ctx context.Context, mod module.Version, zipfile string) (err error) {
+	ctx, span := trace.StartSpan(ctx, "modfetch.downloadZip "+zipfile)
+	defer span.Done()
+
 	// Clean up any remaining tempfiles from previous runs.
 	// This is only safe to do because the lock file ensures that their
 	// writers are no longer active.
@@ -374,12 +384,14 @@ type modSum struct {
 
 var goSum struct {
 	mu        sync.Mutex
-	m         map[module.Version][]string // content of go.sum file (+ go.modverify if present)
-	checked   map[modSum]bool             // sums actually checked during execution
-	dirty     bool                        // whether we added any new sums to m
+	m         map[module.Version][]string // content of go.sum file
+	status    map[modSum]modSumStatus     // state of sums in m
 	overwrite bool                        // if true, overwrite go.sum without incorporating its contents
 	enabled   bool                        // whether to use go.sum at all
-	modverify string                      // path to go.modverify, to be deleted
+}
+
+type modSumStatus struct {
+	used, dirty bool
 }
 
 // initGoSum initializes the go.sum data.
@@ -395,7 +407,7 @@ func initGoSum() (bool, error) {
 	}
 
 	goSum.m = make(map[module.Version][]string)
-	goSum.checked = make(map[modSum]bool)
+	goSum.status = make(map[modSum]modSumStatus)
 	data, err := lockedfile.Read(GoSumFile)
 	if err != nil && !os.IsNotExist(err) {
 		return false, err
@@ -403,19 +415,6 @@ func initGoSum() (bool, error) {
 	goSum.enabled = true
 	readGoSum(goSum.m, GoSumFile, data)
 
-	// Add old go.modverify file.
-	// We'll delete go.modverify in WriteGoSum.
-	alt := strings.TrimSuffix(GoSumFile, ".sum") + ".modverify"
-	if data, err := renameio.ReadFile(alt); err == nil {
-		migrate := make(map[module.Version][]string)
-		readGoSum(migrate, alt, data)
-		for mod, sums := range migrate {
-			for _, sum := range sums {
-				addModSumLocked(mod, sum)
-			}
-		}
-		goSum.modverify = alt
-	}
 	return true, nil
 }
 
@@ -518,6 +517,11 @@ func checkModSum(mod module.Version, h string) error {
 		return err
 	}
 	done := inited && haveModSumLocked(mod, h)
+	if inited {
+		st := goSum.status[modSum{mod, h}]
+		st.used = true
+		goSum.status[modSum{mod, h}] = st
+	}
 	goSum.mu.Unlock()
 
 	if done {
@@ -537,6 +541,9 @@ func checkModSum(mod module.Version, h string) error {
 	if inited {
 		goSum.mu.Lock()
 		addModSumLocked(mod, h)
+		st := goSum.status[modSum{mod, h}]
+		st.dirty = true
+		goSum.status[modSum{mod, h}] = st
 		goSum.mu.Unlock()
 	}
 	return nil
@@ -546,7 +553,6 @@ func checkModSum(mod module.Version, h string) error {
 // If it finds a conflicting pair instead, it calls base.Fatalf.
 // goSum.mu must be locked.
 func haveModSumLocked(mod module.Version, h string) bool {
-	goSum.checked[modSum{mod, h}] = true
 	for _, vh := range goSum.m[mod] {
 		if h == vh {
 			return true
@@ -568,7 +574,6 @@ func addModSumLocked(mod module.Version, h string) {
 		fmt.Fprintf(os.Stderr, "warning: verifying %s@%s: unknown hashes in go.sum: %v; adding %v"+hashVersionMismatch, mod.Path, mod.Version, strings.Join(goSum.m[mod], ", "), h)
 	}
 	goSum.m[mod] = append(goSum.m[mod], h)
-	goSum.dirty = true
 }
 
 // checkSumDB checks the mod, h pair against the Go checksum database.
@@ -612,18 +617,35 @@ func Sum(mod module.Version) string {
 }
 
 // WriteGoSum writes the go.sum file if it needs to be updated.
-func WriteGoSum() {
+//
+// keep is used to check whether a newly added sum should be saved in go.sum.
+// It should have entries for both module content sums and go.mod sums
+// (version ends with "/go.mod"). Existing sums will be preserved unless they
+// have been marked for deletion with TrimGoSum.
+func WriteGoSum(keep map[module.Version]bool) {
 	goSum.mu.Lock()
 	defer goSum.mu.Unlock()
 
+	// If we haven't read the go.sum file yet, don't bother writing it.
 	if !goSum.enabled {
-		// If we haven't read the go.sum file yet, don't bother writing it: at best,
-		// we could rename the go.modverify file if it isn't empty, but we haven't
-		// needed to touch it so far — how important could it be?
 		return
 	}
-	if !goSum.dirty {
-		// Don't bother opening the go.sum file if we don't have anything to add.
+
+	// Check whether we need to add sums for which keep[m] is true or remove
+	// unused sums marked with TrimGoSum. If there are no changes to make,
+	// just return without opening go.sum.
+	dirty := false
+Outer:
+	for m, hs := range goSum.m {
+		for _, h := range hs {
+			st := goSum.status[modSum{m, h}]
+			if st.dirty && (!st.used || keep[m]) {
+				dirty = true
+				break Outer
+			}
+		}
+	}
+	if !dirty {
 		return
 	}
 	if cfg.BuildMod == "readonly" {
@@ -644,9 +666,10 @@ func WriteGoSum() {
 			// them without good reason.
 			goSum.m = make(map[module.Version][]string, len(goSum.m))
 			readGoSum(goSum.m, GoSumFile, data)
-			for ms := range goSum.checked {
-				addModSumLocked(ms.mod, ms.sum)
-				goSum.dirty = true
+			for ms, st := range goSum.status {
+				if st.used {
+					addModSumLocked(ms.mod, ms.sum)
+				}
 			}
 		}
 
@@ -661,7 +684,10 @@ func WriteGoSum() {
 			list := goSum.m[m]
 			sort.Strings(list)
 			for _, h := range list {
-				fmt.Fprintf(&buf, "%s %s %s\n", m.Path, m.Version, h)
+				st := goSum.status[modSum{m, h}]
+				if !st.dirty || (st.used && keep[m]) {
+					fmt.Fprintf(&buf, "%s %s %s\n", m.Path, m.Version, h)
+				}
 			}
 		}
 		return buf.Bytes(), nil
@@ -671,16 +697,16 @@ func WriteGoSum() {
 		base.Fatalf("go: updating go.sum: %v", err)
 	}
 
-	goSum.checked = make(map[modSum]bool)
-	goSum.dirty = false
+	goSum.status = make(map[modSum]modSumStatus)
 	goSum.overwrite = false
-
-	if goSum.modverify != "" {
-		os.Remove(goSum.modverify) // best effort
-	}
 }
 
-// TrimGoSum trims go.sum to contain only the modules for which keep[m] is true.
+// TrimGoSum trims go.sum to contain only the modules needed for reproducible
+// builds.
+//
+// keep is used to check whether a sum should be retained in go.mod. It should
+// have entries for both module content sums and go.mod sums (version ends
+// with "/go.mod").
 func TrimGoSum(keep map[module.Version]bool) {
 	goSum.mu.Lock()
 	defer goSum.mu.Unlock()
@@ -692,13 +718,11 @@ func TrimGoSum(keep map[module.Version]bool) {
 		return
 	}
 
-	for m := range goSum.m {
-		// If we're keeping x@v we also keep x@v/go.mod.
-		// Map x@v/go.mod back to x@v for the keep lookup.
-		noGoMod := module.Version{Path: m.Path, Version: strings.TrimSuffix(m.Version, "/go.mod")}
-		if !keep[m] && !keep[noGoMod] {
-			delete(goSum.m, m)
-			goSum.dirty = true
+	for m, hs := range goSum.m {
+		if !keep[m] {
+			for _, h := range hs {
+				goSum.status[modSum{m, h}] = modSumStatus{used: false, dirty: true}
+			}
 			goSum.overwrite = true
 		}
 	}
diff --git a/src/cmd/go/internal/modfetch/insecure.go b/src/cmd/go/internal/modfetch/insecure.go
index 8420432d6c..b692669cba 100644
--- a/src/cmd/go/internal/modfetch/insecure.go
+++ b/src/cmd/go/internal/modfetch/insecure.go
@@ -7,10 +7,11 @@ package modfetch
 import (
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/get"
-	"cmd/go/internal/str"
+
+	"golang.org/x/mod/module"
 )
 
 // allowInsecure reports whether we are allowed to fetch this path in an insecure manner.
 func allowInsecure(path string) bool {
-	return get.Insecure || str.GlobsMatchPath(cfg.GOINSECURE, path)
+	return get.Insecure || module.MatchPrefixPatterns(cfg.GOINSECURE, path)
 }
diff --git a/src/cmd/go/internal/modfetch/proxy.go b/src/cmd/go/internal/modfetch/proxy.go
index 1c35d0b99b..4ac26650a9 100644
--- a/src/cmd/go/internal/modfetch/proxy.go
+++ b/src/cmd/go/internal/modfetch/proxy.go
@@ -242,8 +242,9 @@ func TryProxies(f func(proxy string) error) error {
 }
 
 type proxyRepo struct {
-	url  *url.URL
-	path string
+	url         *url.URL
+	path        string
+	redactedURL string
 }
 
 func newProxyRepo(baseURL, path string) (Repo, error) {
@@ -268,10 +269,10 @@ func newProxyRepo(baseURL, path string) (Repo, error) {
 	if err != nil {
 		return nil, err
 	}
-
+	redactedURL := base.Redacted()
 	base.Path = strings.TrimSuffix(base.Path, "/") + "/" + enc
 	base.RawPath = strings.TrimSuffix(base.RawPath, "/") + "/" + pathEscape(enc)
-	return &proxyRepo{base, path}, nil
+	return &proxyRepo{base, path, redactedURL}, nil
 }
 
 func (p *proxyRepo) ModulePath() string {
@@ -413,7 +414,7 @@ func (p *proxyRepo) Stat(rev string) (*RevInfo, error) {
 	}
 	info := new(RevInfo)
 	if err := json.Unmarshal(data, info); err != nil {
-		return nil, p.versionError(rev, err)
+		return nil, p.versionError(rev, fmt.Errorf("invalid response from proxy %q: %w", p.redactedURL, err))
 	}
 	if info.Version != rev && rev == module.CanonicalVersion(rev) && module.Check(p.path, rev) == nil {
 		// If we request a correct, appropriate version for the module path, the
@@ -434,7 +435,7 @@ func (p *proxyRepo) Latest() (*RevInfo, error) {
 	}
 	info := new(RevInfo)
 	if err := json.Unmarshal(data, info); err != nil {
-		return nil, p.versionError("", err)
+		return nil, p.versionError("", fmt.Errorf("invalid response from proxy %q: %w", p.redactedURL, err))
 	}
 	return info, nil
 }
diff --git a/src/cmd/go/internal/modfetch/repo.go b/src/cmd/go/internal/modfetch/repo.go
index f03bdd8d03..34f805d58a 100644
--- a/src/cmd/go/internal/modfetch/repo.go
+++ b/src/cmd/go/internal/modfetch/repo.go
@@ -16,9 +16,9 @@ import (
 	"cmd/go/internal/get"
 	"cmd/go/internal/modfetch/codehost"
 	"cmd/go/internal/par"
-	"cmd/go/internal/str"
 	web "cmd/go/internal/web"
 
+	"golang.org/x/mod/module"
 	"golang.org/x/mod/semver"
 )
 
@@ -217,7 +217,7 @@ func lookup(proxy, path string) (r Repo, err error) {
 		return nil, errLookupDisabled
 	}
 
-	if str.GlobsMatchPath(cfg.GONOPROXY, path) {
+	if module.MatchPrefixPatterns(cfg.GONOPROXY, path) {
 		switch proxy {
 		case "noproxy", "direct":
 			return lookupDirect(path)
diff --git a/src/cmd/go/internal/modfetch/sumdb.go b/src/cmd/go/internal/modfetch/sumdb.go
index 7973f47426..783c4a433b 100644
--- a/src/cmd/go/internal/modfetch/sumdb.go
+++ b/src/cmd/go/internal/modfetch/sumdb.go
@@ -24,7 +24,6 @@ import (
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/get"
 	"cmd/go/internal/lockedfile"
-	"cmd/go/internal/str"
 	"cmd/go/internal/web"
 
 	"golang.org/x/mod/module"
@@ -34,7 +33,7 @@ import (
 
 // useSumDB reports whether to use the Go checksum database for the given module.
 func useSumDB(mod module.Version) bool {
-	return cfg.GOSUMDB != "off" && !get.Insecure && !str.GlobsMatchPath(cfg.GONOSUMDB, mod.Path)
+	return cfg.GOSUMDB != "off" && !get.Insecure && !module.MatchPrefixPatterns(cfg.GONOSUMDB, mod.Path)
 }
 
 // lookupSumDB returns the Go checksum database's go.sum lines for the given module,
diff --git a/src/cmd/go/internal/modfetch/zip_sum_test/zip_sum_test.go b/src/cmd/go/internal/modfetch/zip_sum_test/zip_sum_test.go
index eac9b32fa8..82398ebfed 100644
--- a/src/cmd/go/internal/modfetch/zip_sum_test/zip_sum_test.go
+++ b/src/cmd/go/internal/modfetch/zip_sum_test/zip_sum_test.go
@@ -16,6 +16,7 @@
 package zip_sum_test
 
 import (
+	"context"
 	"crypto/sha256"
 	"encoding/csv"
 	"encoding/hex"
@@ -119,7 +120,7 @@ func TestZipSums(t *testing.T) {
 		name := fmt.Sprintf("%s@%s", strings.ReplaceAll(test.m.Path, "/", "_"), test.m.Version)
 		t.Run(name, func(t *testing.T) {
 			t.Parallel()
-			zipPath, err := modfetch.DownloadZip(test.m)
+			zipPath, err := modfetch.DownloadZip(context.Background(), test.m)
 			if err != nil {
 				if *updateTestData {
 					t.Logf("%s: could not download module: %s (will remove from testdata)", test.m, err)
diff --git a/src/cmd/go/internal/modget/get.go b/src/cmd/go/internal/modget/get.go
index 4c6982426f..ee9757912b 100644
--- a/src/cmd/go/internal/modget/get.go
+++ b/src/cmd/go/internal/modget/get.go
@@ -6,10 +6,12 @@
 package modget
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
+	"runtime"
 	"sort"
 	"strings"
 	"sync"
@@ -20,7 +22,6 @@ import (
 	"cmd/go/internal/load"
 	"cmd/go/internal/modload"
 	"cmd/go/internal/mvs"
-	"cmd/go/internal/par"
 	"cmd/go/internal/search"
 	"cmd/go/internal/work"
 
@@ -259,7 +260,7 @@ type query struct {
 	m module.Version
 }
 
-func runGet(cmd *base.Command, args []string) {
+func runGet(ctx context.Context, cmd *base.Command, args []string) {
 	switch getU {
 	case "", "upgrade", "patch":
 		// ok
@@ -277,7 +278,7 @@ func runGet(cmd *base.Command, args []string) {
 	}
 	modload.LoadTests = *getT
 
-	buildList := modload.LoadBuildList()
+	buildList := modload.LoadBuildList(ctx)
 	buildList = buildList[:len(buildList):len(buildList)] // copy on append
 	versionByPath := make(map[string]string)
 	for _, m := range buildList {
@@ -352,7 +353,7 @@ func runGet(cmd *base.Command, args []string) {
 			if !strings.Contains(path, "...") {
 				m := search.NewMatch(path)
 				if pkgPath := modload.DirImportPath(path); pkgPath != "." {
-					m = modload.TargetPackages(pkgPath)
+					m = modload.TargetPackages(ctx, pkgPath)
 				}
 				if len(m.Pkgs) == 0 {
 					for _, err := range m.Errs {
@@ -398,7 +399,7 @@ func runGet(cmd *base.Command, args []string) {
 		default:
 			// The argument is a package or module path.
 			if modload.HasModRoot() {
-				if m := modload.TargetPackages(path); len(m.Pkgs) != 0 {
+				if m := modload.TargetPackages(ctx, path); len(m.Pkgs) != 0 {
 					// The path is in the main module. Nothing to query.
 					if vers != "upgrade" && vers != "patch" {
 						base.Errorf("go get %s: can't request explicit version of path in main module", arg)
@@ -443,7 +444,7 @@ func runGet(cmd *base.Command, args []string) {
 	// packages in unknown modules can't be expanded. This also avoids looking
 	// up new modules while loading packages, only to downgrade later.
 	queryCache := make(map[querySpec]*query)
-	byPath := runQueries(queryCache, queries, nil)
+	byPath := runQueries(ctx, queryCache, queries, nil)
 
 	// Add missing modules to the build list.
 	// We call SetBuildList here and elsewhere, since newUpgrader,
@@ -490,7 +491,7 @@ func runGet(cmd *base.Command, args []string) {
 		if q.path == q.m.Path {
 			wg.Add(1)
 			go func(q *query) {
-				if hasPkg, err := modload.ModuleHasRootPackage(q.m); err != nil {
+				if hasPkg, err := modload.ModuleHasRootPackage(ctx, q.m); err != nil {
 					base.Errorf("go get: %v", err)
 				} else if !hasPkg {
 					modOnlyMu.Lock()
@@ -535,7 +536,7 @@ func runGet(cmd *base.Command, args []string) {
 			// Don't load packages if pkgPatterns is empty. Both
 			// modload.ImportPathsQuiet and ModulePackages convert an empty list
 			// of patterns to []string{"."}, which is not what we want.
-			matches = modload.ImportPathsQuiet(pkgPatterns, imports.AnyTags())
+			matches = modload.ImportPathsQuiet(ctx, pkgPatterns, imports.AnyTags())
 			seenPkgs = make(map[string]bool)
 			for i, match := range matches {
 				arg := pkgGets[i]
@@ -585,7 +586,7 @@ func runGet(cmd *base.Command, args []string) {
 
 		// Query target versions for modules providing packages matched by
 		// command line arguments.
-		byPath = runQueries(queryCache, queries, modOnly)
+		byPath = runQueries(ctx, queryCache, queries, modOnly)
 
 		// Handle upgrades. This is needed for arguments that didn't match
 		// modules or matched different modules from a previous iteration. It
@@ -714,8 +715,8 @@ func runGet(cmd *base.Command, args []string) {
 		return
 	}
 	work.BuildInit()
-	pkgs := load.PackagesForBuild(pkgPatterns)
-	work.InstallPackages(pkgPatterns, pkgs)
+	pkgs := load.PackagesForBuild(ctx, pkgPatterns)
+	work.InstallPackages(ctx, pkgPatterns, pkgs)
 }
 
 // runQueries looks up modules at target versions in parallel. Results will be
@@ -723,30 +724,45 @@ func runGet(cmd *base.Command, args []string) {
 // versions (including earlier queries in the modOnly map), an error will be
 // reported. A map from module paths to queries is returned, which includes
 // queries and modOnly.
-func runQueries(cache map[querySpec]*query, queries []*query, modOnly map[string]*query) map[string]*query {
-	var lookup par.Work
-	for _, q := range queries {
-		if cached := cache[q.querySpec]; cached != nil {
-			*q = *cached
-		} else {
-			cache[q.querySpec] = q
-			lookup.Add(q)
-		}
-	}
+func runQueries(ctx context.Context, cache map[querySpec]*query, queries []*query, modOnly map[string]*query) map[string]*query {
 
-	lookup.Do(10, func(item interface{}) {
-		q := item.(*query)
+	runQuery := func(q *query) {
 		if q.vers == "none" {
 			// Wait for downgrade step.
 			q.m = module.Version{Path: q.path, Version: "none"}
 			return
 		}
-		m, err := getQuery(q.path, q.vers, q.prevM, q.forceModulePath)
+		m, err := getQuery(ctx, q.path, q.vers, q.prevM, q.forceModulePath)
 		if err != nil {
 			base.Errorf("go get %s: %v", q.arg, err)
 		}
 		q.m = m
-	})
+	}
+
+	type token struct{}
+	sem := make(chan token, runtime.GOMAXPROCS(0))
+	for _, q := range queries {
+		if cached := cache[q.querySpec]; cached != nil {
+			*q = *cached
+		} else {
+			sem <- token{}
+			go func(q *query) {
+				runQuery(q)
+				<-sem
+			}(q)
+		}
+	}
+
+	// Fill semaphore channel to wait for goroutines to finish.
+	for n := cap(sem); n > 0; n-- {
+		sem <- token{}
+	}
+
+	// Add to cache after concurrent section to avoid races...
+	for _, q := range queries {
+		cache[q.querySpec] = q
+	}
+
 	base.ExitIfErrors()
 
 	byPath := make(map[string]*query)
@@ -774,7 +790,7 @@ func runQueries(cache map[querySpec]*query, queries []*query, modOnly map[string
 // to determine the underlying module version being requested.
 // If forceModulePath is set, getQuery must interpret path
 // as a module path.
-func getQuery(path, vers string, prevM module.Version, forceModulePath bool) (module.Version, error) {
+func getQuery(ctx context.Context, path, vers string, prevM module.Version, forceModulePath bool) (module.Version, error) {
 	if (prevM.Version != "") != forceModulePath {
 		// We resolve package patterns by calling QueryPattern, which does not
 		// accept a previous version and therefore cannot take it into account for
@@ -796,7 +812,7 @@ func getQuery(path, vers string, prevM module.Version, forceModulePath bool) (mo
 			}
 		}
 
-		info, err := modload.Query(path, vers, prevM.Version, modload.Allowed)
+		info, err := modload.Query(ctx, path, vers, prevM.Version, modload.Allowed)
 		if err == nil {
 			if info.Version != vers && info.Version != prevM.Version {
 				logOncef("go: %s %s => %s", path, vers, info.Version)
@@ -822,7 +838,7 @@ func getQuery(path, vers string, prevM module.Version, forceModulePath bool) (mo
 	// If it turns out to only exist as a module, we can detect the resulting
 	// PackageNotInModuleError and avoid a second round-trip through (potentially)
 	// all of the configured proxies.
-	results, err := modload.QueryPattern(path, vers, modload.Allowed)
+	results, err := modload.QueryPattern(ctx, path, vers, modload.Allowed)
 	if err != nil {
 		// If the path doesn't contain a wildcard, check whether it was actually a
 		// module path instead. If so, return that.
@@ -978,7 +994,7 @@ func (u *upgrader) Upgrade(m module.Version) (module.Version, error) {
 	// If we're querying "upgrade" or "patch", Query will compare the current
 	// version against the chosen version and will return the current version
 	// if it is newer.
-	info, err := modload.Query(m.Path, string(getU), m.Version, modload.Allowed)
+	info, err := modload.Query(context.TODO(), m.Path, string(getU), m.Version, modload.Allowed)
 	if err != nil {
 		// Report error but return m, to let version selection continue.
 		// (Reporting the error will fail the command at the next base.ExitIfErrors.)
diff --git a/src/cmd/go/internal/modload/build.go b/src/cmd/go/internal/modload/build.go
index 5f8a2e7e05..a101681a1f 100644
--- a/src/cmd/go/internal/modload/build.go
+++ b/src/cmd/go/internal/modload/build.go
@@ -6,6 +6,7 @@ package modload
 
 import (
 	"bytes"
+	"context"
 	"encoding/hex"
 	"fmt"
 	"internal/goroot"
@@ -57,21 +58,21 @@ func PackageModuleInfo(pkgpath string) *modinfo.ModulePublic {
 	if !ok {
 		return nil
 	}
-	return moduleInfo(m, true)
+	return moduleInfo(context.TODO(), m, true)
 }
 
-func ModuleInfo(path string) *modinfo.ModulePublic {
+func ModuleInfo(ctx context.Context, path string) *modinfo.ModulePublic {
 	if !Enabled() {
 		return nil
 	}
 
 	if i := strings.Index(path, "@"); i >= 0 {
-		return moduleInfo(module.Version{Path: path[:i], Version: path[i+1:]}, false)
+		return moduleInfo(ctx, module.Version{Path: path[:i], Version: path[i+1:]}, false)
 	}
 
 	for _, m := range BuildList() {
 		if m.Path == path {
-			return moduleInfo(m, true)
+			return moduleInfo(ctx, m, true)
 		}
 	}
 
@@ -84,12 +85,12 @@ func ModuleInfo(path string) *modinfo.ModulePublic {
 }
 
 // addUpdate fills in m.Update if an updated version is available.
-func addUpdate(m *modinfo.ModulePublic) {
+func addUpdate(ctx context.Context, m *modinfo.ModulePublic) {
 	if m.Version == "" {
 		return
 	}
 
-	if info, err := Query(m.Path, "upgrade", m.Version, Allowed); err == nil && semver.Compare(info.Version, m.Version) > 0 {
+	if info, err := Query(ctx, m.Path, "upgrade", m.Version, Allowed); err == nil && semver.Compare(info.Version, m.Version) > 0 {
 		m.Update = &modinfo.ModulePublic{
 			Path:    m.Path,
 			Version: info.Version,
@@ -103,7 +104,7 @@ func addVersions(m *modinfo.ModulePublic) {
 	m.Versions, _ = versions(m.Path)
 }
 
-func moduleInfo(m module.Version, fromBuildList bool) *modinfo.ModulePublic {
+func moduleInfo(ctx context.Context, m module.Version, fromBuildList bool) *modinfo.ModulePublic {
 	if m == Target {
 		info := &modinfo.ModulePublic{
 			Path:    m.Path,
@@ -132,7 +133,7 @@ func moduleInfo(m module.Version, fromBuildList bool) *modinfo.ModulePublic {
 	// completeFromModCache fills in the extra fields in m using the module cache.
 	completeFromModCache := func(m *modinfo.ModulePublic) {
 		if m.Version != "" {
-			if q, err := Query(m.Path, m.Version, "", nil); err != nil {
+			if q, err := Query(ctx, m.Path, m.Version, "", nil); err != nil {
 				m.Error = &modinfo.ModuleError{Err: err.Error()}
 			} else {
 				m.Version = q.Version
diff --git a/src/cmd/go/internal/modload/import.go b/src/cmd/go/internal/modload/import.go
index 4d2bc805e2..5c51a79124 100644
--- a/src/cmd/go/internal/modload/import.go
+++ b/src/cmd/go/internal/modload/import.go
@@ -5,6 +5,7 @@
 package modload
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"go/build"
@@ -110,7 +111,7 @@ var _ load.ImportPathError = &AmbiguousImportError{}
 // Import returns an ImportMissingError as the error.
 // If Import can identify a module that could be added to supply the package,
 // the ImportMissingError records that module.
-func Import(path string) (m module.Version, dir string, err error) {
+func Import(ctx context.Context, path string) (m module.Version, dir string, err error) {
 	if strings.Contains(path, "@") {
 		return module.Version{}, "", fmt.Errorf("import path should not have @version")
 	}
@@ -165,7 +166,7 @@ func Import(path string) (m module.Version, dir string, err error) {
 			// Avoid possibly downloading irrelevant modules.
 			continue
 		}
-		root, isLocal, err := fetch(m)
+		root, isLocal, err := fetch(ctx, m)
 		if err != nil {
 			// Report fetch error.
 			// Note that we don't know for sure this module is necessary,
@@ -248,7 +249,7 @@ func Import(path string) (m module.Version, dir string, err error) {
 			return len(mods[i].Path) > len(mods[j].Path)
 		})
 		for _, m := range mods {
-			root, isLocal, err := fetch(m)
+			root, isLocal, err := fetch(ctx, m)
 			if err != nil {
 				// Report fetch error as above.
 				return module.Version{}, "", err
@@ -285,7 +286,7 @@ func Import(path string) (m module.Version, dir string, err error) {
 
 	fmt.Fprintf(os.Stderr, "go: finding module for package %s\n", path)
 
-	candidates, err := QueryPackage(path, "latest", Allowed)
+	candidates, err := QueryPackage(ctx, path, "latest", Allowed)
 	if err != nil {
 		if errors.Is(err, os.ErrNotExist) {
 			// Return "cannot find module providing package […]" instead of whatever
diff --git a/src/cmd/go/internal/modload/import_test.go b/src/cmd/go/internal/modload/import_test.go
index accc60eecd..47ce89a084 100644
--- a/src/cmd/go/internal/modload/import_test.go
+++ b/src/cmd/go/internal/modload/import_test.go
@@ -5,6 +5,7 @@
 package modload
 
 import (
+	"context"
 	"internal/testenv"
 	"regexp"
 	"strings"
@@ -49,10 +50,12 @@ func TestImport(t *testing.T) {
 	}(allowMissingModuleImports)
 	AllowMissingModuleImports()
 
+	ctx := context.Background()
+
 	for _, tt := range importTests {
 		t.Run(strings.ReplaceAll(tt.path, "/", "_"), func(t *testing.T) {
 			// Note that there is no build list, so Import should always fail.
-			m, dir, err := Import(tt.path)
+			m, dir, err := Import(ctx, tt.path)
 			if err == nil {
 				t.Fatalf("Import(%q) = %v, %v, nil; expected error", tt.path, m, dir)
 			}
diff --git a/src/cmd/go/internal/modload/init.go b/src/cmd/go/internal/modload/init.go
index 664a2a1594..93027c44c4 100644
--- a/src/cmd/go/internal/modload/init.go
+++ b/src/cmd/go/internal/modload/init.go
@@ -6,6 +6,7 @@ package modload
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -20,7 +21,6 @@ import (
 	"strings"
 
 	"cmd/go/internal/base"
-	"cmd/go/internal/cache"
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/load"
 	"cmd/go/internal/lockedfile"
@@ -162,12 +162,6 @@ func Init() {
 
 	// We're in module mode. Install the hooks to make it work.
 
-	if c := cache.Default(); c == nil {
-		// With modules, there are no install locations for packages
-		// other than the build cache.
-		base.Fatalf("go: cannot use modules with build cache disabled")
-	}
-
 	list := filepath.SplitList(cfg.BuildContext.GOPATH)
 	if len(list) == 0 || list[0] == "" {
 		base.Fatalf("missing $GOPATH")
@@ -331,11 +325,15 @@ func die() {
 }
 
 // InitMod sets Target and, if there is a main module, parses the initial build
-// list from its go.mod file, creating and populating that file if needed.
+// list from its go.mod file. If InitMod is called by 'go mod init', InitMod
+// will populate go.mod in memory, possibly importing dependencies from a
+// legacy configuration file. For other commands, InitMod may make other
+// adjustments in memory, like adding a go directive. WriteGoMod should be
+// called later to write changes out to disk.
 //
 // As a side-effect, InitMod sets a default for cfg.BuildMod if it does not
 // already have an explicit value.
-func InitMod() {
+func InitMod(ctx context.Context) {
 	if len(buildList) > 0 {
 		return
 	}
@@ -352,7 +350,6 @@ func InitMod() {
 		// Running go mod init: do legacy module conversion
 		legacyModInit()
 		modFileToBuildList()
-		WriteGoMod()
 		return
 	}
 
@@ -363,7 +360,7 @@ func InitMod() {
 	}
 
 	var fixed bool
-	f, err := modfile.Parse(gomod, data, fixVersion(&fixed))
+	f, err := modfile.Parse(gomod, data, fixVersion(ctx, &fixed))
 	if err != nil {
 		// Errors returned by modfile.Parse begin with file:line.
 		base.Fatalf("go: errors parsing go.mod:\n%s\n", err)
@@ -391,9 +388,6 @@ func InitMod() {
 	if cfg.BuildMod == "vendor" {
 		readVendorList()
 		checkVendorConsistency()
-	} else {
-		// TODO(golang.org/issue/33326): if cfg.BuildMod != "readonly"?
-		WriteGoMod()
 	}
 }
 
@@ -404,7 +398,7 @@ func InitMod() {
 // and does nothing for versions that already appear to be canonical.
 //
 // The VersionFixer sets 'fixed' if it ever returns a non-canonical version.
-func fixVersion(fixed *bool) modfile.VersionFixer {
+func fixVersion(ctx context.Context, fixed *bool) modfile.VersionFixer {
 	return func(path, vers string) (resolved string, err error) {
 		defer func() {
 			if err == nil && resolved != vers {
@@ -436,7 +430,7 @@ func fixVersion(fixed *bool) modfile.VersionFixer {
 			}
 		}
 
-		info, err := Query(path, vers, "", nil)
+		info, err := Query(ctx, path, vers, "", nil)
 		if err != nil {
 			return "", err
 		}
@@ -797,9 +791,10 @@ func WriteGoMod() {
 			base.Fatalf("go: updates to go.mod needed, disabled by -mod=readonly")
 		}
 	}
+
 	// Always update go.sum, even if we didn't change go.mod: we may have
 	// downloaded modules that we didn't have before.
-	modfetch.WriteGoSum()
+	modfetch.WriteGoSum(keepSums())
 
 	if !dirty && cfg.CmdName != "mod tidy" {
 		// The go.mod file has the same semantic content that it had before
@@ -849,3 +844,59 @@ func WriteGoMod() {
 		base.Fatalf("go: updating go.mod: %v", err)
 	}
 }
+
+// keepSums returns a set of module sums to preserve in go.sum. The set
+// includes entries for all modules used to load packages (according to
+// the last load function like ImportPaths, LoadALL, etc.). It also contains
+// entries for go.mod files needed for MVS (the version of these entries
+// ends with "/go.mod").
+func keepSums() map[module.Version]bool {
+	// Walk the module graph and keep sums needed by MVS.
+	modkey := func(m module.Version) module.Version {
+		return module.Version{Path: m.Path, Version: m.Version + "/go.mod"}
+	}
+	keep := make(map[module.Version]bool)
+	replaced := make(map[module.Version]bool)
+	reqs := Reqs()
+	var walk func(module.Version)
+	walk = func(m module.Version) {
+		// If we build using a replacement module, keep the sum for the replacement,
+		// since that's the code we'll actually use during a build.
+		//
+		// TODO(golang.org/issue/29182): Perhaps we should keep both sums, and the
+		// sums for both sets of transitive requirements.
+		r := Replacement(m)
+		if r.Path == "" {
+			keep[modkey(m)] = true
+		} else {
+			replaced[m] = true
+			keep[modkey(r)] = true
+		}
+		list, _ := reqs.Required(m)
+		for _, r := range list {
+			if !keep[modkey(r)] && !replaced[r] {
+				walk(r)
+			}
+		}
+	}
+	walk(Target)
+
+	// Add entries for modules that provided packages loaded with ImportPaths,
+	// LoadALL, or similar functions.
+	if loaded != nil {
+		for _, pkg := range loaded.pkgs {
+			m := pkg.mod
+			if r := Replacement(m); r.Path != "" {
+				keep[r] = true
+			} else {
+				keep[m] = true
+			}
+		}
+	}
+
+	return keep
+}
+
+func TrimGoSum() {
+	modfetch.TrimGoSum(keepSums())
+}
diff --git a/src/cmd/go/internal/modload/list.go b/src/cmd/go/internal/modload/list.go
index 9400793bcb..7bf4e86c8d 100644
--- a/src/cmd/go/internal/modload/list.go
+++ b/src/cmd/go/internal/modload/list.go
@@ -5,47 +5,59 @@
 package modload
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"os"
+	"runtime"
 	"strings"
 
 	"cmd/go/internal/base"
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/modinfo"
-	"cmd/go/internal/par"
 	"cmd/go/internal/search"
 
 	"golang.org/x/mod/module"
 )
 
-func ListModules(args []string, listU, listVersions bool) []*modinfo.ModulePublic {
-	mods := listModules(args, listVersions)
+func ListModules(ctx context.Context, args []string, listU, listVersions bool) []*modinfo.ModulePublic {
+	mods := listModules(ctx, args, listVersions)
+
+	type token struct{}
+	sem := make(chan token, runtime.GOMAXPROCS(0))
 	if listU || listVersions {
-		var work par.Work
 		for _, m := range mods {
-			work.Add(m)
+			add := func(m *modinfo.ModulePublic) {
+				sem <- token{}
+				go func() {
+					if listU {
+						addUpdate(ctx, m)
+					}
+					if listVersions {
+						addVersions(m)
+					}
+					<-sem
+				}()
+			}
+
+			add(m)
 			if m.Replace != nil {
-				work.Add(m.Replace)
+				add(m.Replace)
 			}
 		}
-		work.Do(10, func(item interface{}) {
-			m := item.(*modinfo.ModulePublic)
-			if listU {
-				addUpdate(m)
-			}
-			if listVersions {
-				addVersions(m)
-			}
-		})
 	}
+	// Fill semaphore channel to wait for all tasks to finish.
+	for n := cap(sem); n > 0; n-- {
+		sem <- token{}
+	}
+
 	return mods
 }
 
-func listModules(args []string, listVersions bool) []*modinfo.ModulePublic {
-	LoadBuildList()
+func listModules(ctx context.Context, args []string, listVersions bool) []*modinfo.ModulePublic {
+	LoadBuildList(ctx)
 	if len(args) == 0 {
-		return []*modinfo.ModulePublic{moduleInfo(buildList[0], true)}
+		return []*modinfo.ModulePublic{moduleInfo(ctx, buildList[0], true)}
 	}
 
 	var mods []*modinfo.ModulePublic
@@ -71,7 +83,7 @@ func listModules(args []string, listVersions bool) []*modinfo.ModulePublic {
 				}
 			}
 
-			info, err := Query(path, vers, current, nil)
+			info, err := Query(ctx, path, vers, current, nil)
 			if err != nil {
 				mods = append(mods, &modinfo.ModulePublic{
 					Path:    path,
@@ -80,7 +92,7 @@ func listModules(args []string, listVersions bool) []*modinfo.ModulePublic {
 				})
 				continue
 			}
-			mods = append(mods, moduleInfo(module.Version{Path: path, Version: info.Version}, false))
+			mods = append(mods, moduleInfo(ctx, module.Version{Path: path, Version: info.Version}, false))
 			continue
 		}
 
@@ -105,7 +117,7 @@ func listModules(args []string, listVersions bool) []*modinfo.ModulePublic {
 				matched = true
 				if !matchedBuildList[i] {
 					matchedBuildList[i] = true
-					mods = append(mods, moduleInfo(m, true))
+					mods = append(mods, moduleInfo(ctx, m, true))
 				}
 			}
 		}
@@ -115,9 +127,9 @@ func listModules(args []string, listVersions bool) []*modinfo.ModulePublic {
 					// Don't make the user provide an explicit '@latest' when they're
 					// explicitly asking what the available versions are.
 					// Instead, resolve the module, even if it isn't an existing dependency.
-					info, err := Query(arg, "latest", "", nil)
+					info, err := Query(ctx, arg, "latest", "", nil)
 					if err == nil {
-						mods = append(mods, moduleInfo(module.Version{Path: arg, Version: info.Version}, false))
+						mods = append(mods, moduleInfo(ctx, module.Version{Path: arg, Version: info.Version}, false))
 					} else {
 						mods = append(mods, &modinfo.ModulePublic{
 							Path:  arg,
diff --git a/src/cmd/go/internal/modload/load.go b/src/cmd/go/internal/modload/load.go
index 30992e0cc2..686d491219 100644
--- a/src/cmd/go/internal/modload/load.go
+++ b/src/cmd/go/internal/modload/load.go
@@ -6,14 +6,7 @@ package modload
 
 import (
 	"bytes"
-	"cmd/go/internal/base"
-	"cmd/go/internal/cfg"
-	"cmd/go/internal/imports"
-	"cmd/go/internal/modfetch"
-	"cmd/go/internal/mvs"
-	"cmd/go/internal/par"
-	"cmd/go/internal/search"
-	"cmd/go/internal/str"
+	"context"
 	"errors"
 	"fmt"
 	"go/build"
@@ -24,6 +17,16 @@ import (
 	"sort"
 	"strings"
 
+	"cmd/go/internal/base"
+	"cmd/go/internal/cfg"
+	"cmd/go/internal/imports"
+	"cmd/go/internal/modfetch"
+	"cmd/go/internal/mvs"
+	"cmd/go/internal/par"
+	"cmd/go/internal/search"
+	"cmd/go/internal/str"
+	"cmd/go/internal/trace"
+
 	"golang.org/x/mod/module"
 )
 
@@ -49,8 +52,8 @@ var loaded *loader
 // ImportPaths returns the set of packages matching the args (patterns),
 // on the target platform. Modules may be added to the build list
 // to satisfy new imports.
-func ImportPaths(patterns []string) []*search.Match {
-	matches := ImportPathsQuiet(patterns, imports.Tags())
+func ImportPaths(ctx context.Context, patterns []string) []*search.Match {
+	matches := ImportPathsQuiet(ctx, patterns, imports.Tags())
 	search.WarnUnmatched(matches)
 	return matches
 }
@@ -59,7 +62,7 @@ func ImportPaths(patterns []string) []*search.Match {
 // no matches. It also lets the caller specify a set of build tags to match
 // packages. The build tags should typically be imports.Tags() or
 // imports.AnyTags(); a nil map has no special meaning.
-func ImportPathsQuiet(patterns []string, tags map[string]bool) []*search.Match {
+func ImportPathsQuiet(ctx context.Context, patterns []string, tags map[string]bool) []*search.Match {
 	updateMatches := func(matches []*search.Match, iterating bool) {
 		for _, m := range matches {
 			switch {
@@ -100,7 +103,7 @@ func ImportPathsQuiet(patterns []string, tags map[string]bool) []*search.Match {
 
 			case strings.Contains(m.Pattern(), "..."):
 				m.Errs = m.Errs[:0]
-				matchPackages(m, loaded.tags, includeStd, buildList)
+				matchPackages(ctx, m, loaded.tags, includeStd, buildList)
 
 			case m.Pattern() == "all":
 				loaded.testAll = true
@@ -108,7 +111,7 @@ func ImportPathsQuiet(patterns []string, tags map[string]bool) []*search.Match {
 					// Enumerate the packages in the main module.
 					// We'll load the dependencies as we find them.
 					m.Errs = m.Errs[:0]
-					matchPackages(m, loaded.tags, omitStd, []module.Version{Target})
+					matchPackages(ctx, m, loaded.tags, omitStd, []module.Version{Target})
 				} else {
 					// Starting with the packages in the main module,
 					// enumerate the full list of "all".
@@ -126,7 +129,7 @@ func ImportPathsQuiet(patterns []string, tags map[string]bool) []*search.Match {
 		}
 	}
 
-	InitMod()
+	InitMod(ctx)
 
 	var matches []*search.Match
 	for _, pattern := range search.CleanPatterns(patterns) {
@@ -335,8 +338,8 @@ func pathInModuleCache(dir string) string {
 
 // ImportFromFiles adds modules to the build list as needed
 // to satisfy the imports in the named Go source files.
-func ImportFromFiles(gofiles []string) {
-	InitMod()
+func ImportFromFiles(ctx context.Context, gofiles []string) {
+	InitMod(ctx)
 
 	tags := imports.Tags()
 	imports, testImports, err := imports.ScanFiles(gofiles, tags)
@@ -385,8 +388,10 @@ func DirImportPath(dir string) string {
 // LoadBuildList need only be called if ImportPaths is not
 // (typically in commands that care about the module but
 // no particular package).
-func LoadBuildList() []module.Version {
-	InitMod()
+func LoadBuildList(ctx context.Context) []module.Version {
+	ctx, span := trace.StartSpan(ctx, "LoadBuildList")
+	defer span.Done()
+	InitMod(ctx)
 	ReloadBuildList()
 	WriteGoMod()
 	return buildList
@@ -404,20 +409,20 @@ func ReloadBuildList() []module.Version {
 // It adds modules to the build list as needed to satisfy new imports.
 // This set is useful for deciding whether a particular import is needed
 // anywhere in a module.
-func LoadALL() []string {
-	return loadAll(true)
+func LoadALL(ctx context.Context) []string {
+	return loadAll(ctx, true)
 }
 
 // LoadVendor is like LoadALL but only follows test dependencies
 // for tests in the main module. Tests in dependency modules are
 // ignored completely.
 // This set is useful for identifying the which packages to include in a vendor directory.
-func LoadVendor() []string {
-	return loadAll(false)
+func LoadVendor(ctx context.Context) []string {
+	return loadAll(ctx, false)
 }
 
-func loadAll(testAll bool) []string {
-	InitMod()
+func loadAll(ctx context.Context, testAll bool) []string {
+	InitMod(ctx)
 
 	loaded = newLoader(imports.AnyTags())
 	loaded.isALL = true
@@ -425,7 +430,7 @@ func loadAll(testAll bool) []string {
 	if !testAll {
 		loaded.testRoots = true
 	}
-	all := TargetPackages("...")
+	all := TargetPackages(ctx, "...")
 	loaded.load(func() []string { return all.Pkgs })
 	checkMultiplePaths()
 	WriteGoMod()
@@ -448,13 +453,13 @@ func loadAll(testAll bool) []string {
 // TargetPackages returns the list of packages in the target (top-level) module
 // matching pattern, which may be relative to the working directory, under all
 // build tag settings.
-func TargetPackages(pattern string) *search.Match {
+func TargetPackages(ctx context.Context, pattern string) *search.Match {
 	// TargetPackages is relative to the main module, so ensure that the main
 	// module is a thing that can contain packages.
 	ModRoot()
 
 	m := search.NewMatch(pattern)
-	matchPackages(m, imports.AnyTags(), omitStd, []module.Version{Target})
+	matchPackages(ctx, m, imports.AnyTags(), omitStd, []module.Version{Target})
 	return m
 }
 
@@ -812,7 +817,8 @@ func (ld *loader) doPkg(item interface{}) {
 			return
 		}
 
-		pkg.mod, pkg.dir, pkg.err = Import(pkg.path)
+		// TODO(matloob): Handle TODO context. This needs to be threaded through Do.
+		pkg.mod, pkg.dir, pkg.err = Import(context.TODO(), pkg.path)
 		if pkg.dir == "" {
 			return
 		}
diff --git a/src/cmd/go/internal/modload/mvs.go b/src/cmd/go/internal/modload/mvs.go
index 5dd009d31d..67eb2c2e19 100644
--- a/src/cmd/go/internal/modload/mvs.go
+++ b/src/cmd/go/internal/modload/mvs.go
@@ -5,6 +5,7 @@
 package modload
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"os"
@@ -224,7 +225,7 @@ func (*mvsReqs) next(m module.Version) (module.Version, error) {
 //
 // The isLocal return value reports whether the replacement,
 // if any, is local to the filesystem.
-func fetch(mod module.Version) (dir string, isLocal bool, err error) {
+func fetch(ctx context.Context, mod module.Version) (dir string, isLocal bool, err error) {
 	if mod == Target {
 		return ModRoot(), true, nil
 	}
@@ -254,6 +255,6 @@ func fetch(mod module.Version) (dir string, isLocal bool, err error) {
 		mod = r
 	}
 
-	dir, err = modfetch.Download(mod)
+	dir, err = modfetch.Download(ctx, mod)
 	return dir, false, err
 }
diff --git a/src/cmd/go/internal/modload/query.go b/src/cmd/go/internal/modload/query.go
index acc886bf21..e82eb1506f 100644
--- a/src/cmd/go/internal/modload/query.go
+++ b/src/cmd/go/internal/modload/query.go
@@ -5,6 +5,7 @@
 package modload
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"os"
@@ -18,6 +19,7 @@ import (
 	"cmd/go/internal/modfetch"
 	"cmd/go/internal/search"
 	"cmd/go/internal/str"
+	"cmd/go/internal/trace"
 
 	"golang.org/x/mod/module"
 	"golang.org/x/mod/semver"
@@ -55,10 +57,10 @@ import (
 //
 // If path is the path of the main module and the query is "latest",
 // Query returns Target.Version as the version.
-func Query(path, query, current string, allowed func(module.Version) bool) (*modfetch.RevInfo, error) {
+func Query(ctx context.Context, path, query, current string, allowed func(module.Version) bool) (*modfetch.RevInfo, error) {
 	var info *modfetch.RevInfo
 	err := modfetch.TryProxies(func(proxy string) (err error) {
-		info, err = queryProxy(proxy, path, query, current, allowed)
+		info, err = queryProxy(ctx, proxy, path, query, current, allowed)
 		return err
 	})
 	return info, err
@@ -75,7 +77,10 @@ func (queryDisabledError) Error() string {
 	return fmt.Sprintf("cannot query module due to -mod=%s\n\t(%s)", cfg.BuildMod, cfg.BuildModReason)
 }
 
-func queryProxy(proxy, path, query, current string, allowed func(module.Version) bool) (*modfetch.RevInfo, error) {
+func queryProxy(ctx context.Context, proxy, path, query, current string, allowed func(module.Version) bool) (*modfetch.RevInfo, error) {
+	ctx, span := trace.StartSpan(ctx, "modload.queryProxy "+path+" "+query)
+	defer span.Done()
+
 	if current != "" && !semver.IsValid(current) {
 		return nil, fmt.Errorf("invalid previous version %q", current)
 	}
@@ -243,7 +248,7 @@ func queryProxy(proxy, path, query, current string, allowed func(module.Version)
 	if err != nil {
 		return nil, err
 	}
-	releases, prereleases, err := filterVersions(path, versions, ok, preferIncompatible)
+	releases, prereleases, err := filterVersions(ctx, path, versions, ok, preferIncompatible)
 	if err != nil {
 		return nil, err
 	}
@@ -327,7 +332,7 @@ func matchSemverPrefix(p, v string) bool {
 // 	1. versions that do not satisfy the 'ok' predicate, and
 // 	2. "+incompatible" versions, if a compatible one satisfies the predicate
 // 	   and the incompatible version is not preferred.
-func filterVersions(path string, versions []string, ok func(module.Version) bool, preferIncompatible bool) (releases, prereleases []string, err error) {
+func filterVersions(ctx context.Context, path string, versions []string, ok func(module.Version) bool, preferIncompatible bool) (releases, prereleases []string, err error) {
 	var lastCompatible string
 	for _, v := range versions {
 		if !ok(module.Version{Path: path, Version: v}) {
@@ -343,7 +348,7 @@ func filterVersions(path string, versions []string, ok func(module.Version) bool
 				// https://golang.org/issue/34165.) Note that we even prefer a
 				// compatible pre-release over an incompatible release.
 
-				ok, err := versionHasGoMod(module.Version{Path: path, Version: lastCompatible})
+				ok, err := versionHasGoMod(ctx, module.Version{Path: path, Version: lastCompatible})
 				if err != nil {
 					return nil, nil, err
 				}
@@ -380,12 +385,12 @@ type QueryResult struct {
 // If the package is in the main module, QueryPackage considers only the main
 // module and only the version "latest", without checking for other possible
 // modules.
-func QueryPackage(path, query string, allowed func(module.Version) bool) ([]QueryResult, error) {
+func QueryPackage(ctx context.Context, path, query string, allowed func(module.Version) bool) ([]QueryResult, error) {
 	m := search.NewMatch(path)
 	if m.IsLocal() || !m.IsLiteral() {
 		return nil, fmt.Errorf("pattern %s is not an importable package", path)
 	}
-	return QueryPattern(path, query, allowed)
+	return QueryPattern(ctx, path, query, allowed)
 }
 
 // QueryPattern looks up the module(s) containing at least one package matching
@@ -401,7 +406,10 @@ func QueryPackage(path, query string, allowed func(module.Version) bool) ([]Quer
 // If any matching package is in the main module, QueryPattern considers only
 // the main module and only the version "latest", without checking for other
 // possible modules.
-func QueryPattern(pattern, query string, allowed func(module.Version) bool) ([]QueryResult, error) {
+func QueryPattern(ctx context.Context, pattern, query string, allowed func(module.Version) bool) ([]QueryResult, error) {
+	ctx, span := trace.StartSpan(ctx, "modload.QueryPattern "+pattern+" "+query)
+	defer span.Done()
+
 	base := pattern
 
 	firstError := func(m *search.Match) error {
@@ -417,7 +425,7 @@ func QueryPattern(pattern, query string, allowed func(module.Version) bool) ([]Q
 		base = pathpkg.Dir(pattern[:i+3])
 		match = func(mod module.Version, root string, isLocal bool) *search.Match {
 			m := search.NewMatch(pattern)
-			matchPackages(m, imports.AnyTags(), omitStd, []module.Version{mod})
+			matchPackages(ctx, m, imports.AnyTags(), omitStd, []module.Version{mod})
 			return m
 		}
 	} else {
@@ -469,15 +477,18 @@ func QueryPattern(pattern, query string, allowed func(module.Version) bool) ([]Q
 	}
 
 	err := modfetch.TryProxies(func(proxy string) error {
-		queryModule := func(path string) (r QueryResult, err error) {
+		queryModule := func(ctx context.Context, path string) (r QueryResult, err error) {
+			ctx, span := trace.StartSpan(ctx, "modload.QueryPattern.queryModule ["+proxy+"] "+path)
+			defer span.Done()
+
 			current := findCurrentVersion(path)
 			r.Mod.Path = path
-			r.Rev, err = queryProxy(proxy, path, query, current, allowed)
+			r.Rev, err = queryProxy(ctx, proxy, path, query, current, allowed)
 			if err != nil {
 				return r, err
 			}
 			r.Mod.Version = r.Rev.Version
-			root, isLocal, err := fetch(r.Mod)
+			root, isLocal, err := fetch(ctx, r.Mod)
 			if err != nil {
 				return r, err
 			}
@@ -498,7 +509,7 @@ func QueryPattern(pattern, query string, allowed func(module.Version) bool) ([]Q
 		}
 
 		var err error
-		results, err = queryPrefixModules(candidateModules, queryModule)
+		results, err = queryPrefixModules(ctx, candidateModules, queryModule)
 		return err
 	})
 
@@ -542,7 +553,10 @@ type prefixResult struct {
 	err error
 }
 
-func queryPrefixModules(candidateModules []string, queryModule func(path string) (QueryResult, error)) (found []QueryResult, err error) {
+func queryPrefixModules(ctx context.Context, candidateModules []string, queryModule func(ctx context.Context, path string) (QueryResult, error)) (found []QueryResult, err error) {
+	ctx, span := trace.StartSpan(ctx, "modload.queryPrefixModules")
+	defer span.Done()
+
 	// If the path we're attempting is not in the module cache and we don't have a
 	// fetch result cached either, we'll end up making a (potentially slow)
 	// request to the proxy or (often even slower) the origin server.
@@ -555,8 +569,9 @@ func queryPrefixModules(candidateModules []string, queryModule func(path string)
 	var wg sync.WaitGroup
 	wg.Add(len(candidateModules))
 	for i, p := range candidateModules {
+		ctx := trace.StartGoroutine(ctx)
 		go func(p string, r *result) {
-			r.QueryResult, r.err = queryModule(p)
+			r.QueryResult, r.err = queryModule(ctx, p)
 			wg.Done()
 		}(p, &results[i])
 	}
@@ -698,8 +713,8 @@ func (e *PackageNotInModuleError) ImportPath() string {
 }
 
 // ModuleHasRootPackage returns whether module m contains a package m.Path.
-func ModuleHasRootPackage(m module.Version) (bool, error) {
-	root, isLocal, err := fetch(m)
+func ModuleHasRootPackage(ctx context.Context, m module.Version) (bool, error) {
+	root, isLocal, err := fetch(ctx, m)
 	if err != nil {
 		return false, err
 	}
@@ -707,8 +722,8 @@ func ModuleHasRootPackage(m module.Version) (bool, error) {
 	return ok, err
 }
 
-func versionHasGoMod(m module.Version) (bool, error) {
-	root, _, err := fetch(m)
+func versionHasGoMod(ctx context.Context, m module.Version) (bool, error) {
+	root, _, err := fetch(ctx, m)
 	if err != nil {
 		return false, err
 	}
diff --git a/src/cmd/go/internal/modload/query_test.go b/src/cmd/go/internal/modload/query_test.go
index 247e4c40d2..77080e9b5b 100644
--- a/src/cmd/go/internal/modload/query_test.go
+++ b/src/cmd/go/internal/modload/query_test.go
@@ -5,6 +5,7 @@
 package modload
 
 import (
+	"context"
 	"internal/testenv"
 	"io/ioutil"
 	"log"
@@ -179,6 +180,8 @@ func TestQuery(t *testing.T) {
 	testenv.MustHaveExternalNetwork(t)
 	testenv.MustHaveExecPath(t, "git")
 
+	ctx := context.Background()
+
 	for _, tt := range queryTests {
 		allow := tt.allow
 		if allow == "" {
@@ -192,7 +195,7 @@ func TestQuery(t *testing.T) {
 		t.Run(strings.ReplaceAll(tt.path, "/", "_")+"/"+tt.query+"/"+tt.current+"/"+allow, func(t *testing.T) {
 			t.Parallel()
 
-			info, err := Query(tt.path, tt.query, tt.current, allowed)
+			info, err := Query(ctx, tt.path, tt.query, tt.current, allowed)
 			if tt.err != "" {
 				if err == nil {
 					t.Errorf("Query(%q, %q, %v) = %v, want error %q", tt.path, tt.query, allow, info.Version, tt.err)
diff --git a/src/cmd/go/internal/modload/search.go b/src/cmd/go/internal/modload/search.go
index c28e7c0c1e..a9bee0af4e 100644
--- a/src/cmd/go/internal/modload/search.go
+++ b/src/cmd/go/internal/modload/search.go
@@ -5,6 +5,7 @@
 package modload
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -27,7 +28,7 @@ const (
 // matchPackages is like m.MatchPackages, but uses a local variable (rather than
 // a global) for tags, can include or exclude packages in the standard library,
 // and is restricted to the given list of modules.
-func matchPackages(m *search.Match, tags map[string]bool, filter stdFilter, modules []module.Version) {
+func matchPackages(ctx context.Context, m *search.Match, tags map[string]bool, filter stdFilter, modules []module.Version) {
 	m.Pkgs = []string{}
 
 	isMatch := func(string) bool { return true }
@@ -153,7 +154,7 @@ func matchPackages(m *search.Match, tags map[string]bool, filter stdFilter, modu
 			isLocal = true
 		} else {
 			var err error
-			root, isLocal, err = fetch(mod)
+			root, isLocal, err = fetch(ctx, mod)
 			if err != nil {
 				m.AddError(err)
 				continue
diff --git a/src/cmd/go/internal/run/run.go b/src/cmd/go/internal/run/run.go
index 2edae38cca..99578b244c 100644
--- a/src/cmd/go/internal/run/run.go
+++ b/src/cmd/go/internal/run/run.go
@@ -6,6 +6,7 @@
 package run
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"path"
@@ -57,7 +58,7 @@ func printStderr(args ...interface{}) (int, error) {
 	return fmt.Fprint(os.Stderr, args...)
 }
 
-func runRun(cmd *base.Command, args []string) {
+func runRun(ctx context.Context, cmd *base.Command, args []string) {
 	work.BuildInit()
 	var b work.Builder
 	b.Init()
@@ -76,9 +77,9 @@ func runRun(cmd *base.Command, args []string) {
 				base.Fatalf("go run: cannot run *_test.go files (%s)", file)
 			}
 		}
-		p = load.GoFilesPackage(files)
+		p = load.GoFilesPackage(ctx, files)
 	} else if len(args) > 0 && !strings.HasPrefix(args[0], "-") {
-		pkgs := load.PackagesAndErrors(args[:1])
+		pkgs := load.PackagesAndErrors(ctx, args[:1])
 		if len(pkgs) == 0 {
 			base.Fatalf("go run: no packages loaded from %s", args[0])
 		}
@@ -140,12 +141,12 @@ func runRun(cmd *base.Command, args []string) {
 	}
 	a1 := b.LinkAction(work.ModeBuild, work.ModeBuild, p)
 	a := &work.Action{Mode: "go run", Func: buildRunProgram, Args: cmdArgs, Deps: []*work.Action{a1}}
-	b.Do(a)
+	b.Do(ctx, a)
 }
 
 // buildRunProgram is the action for running a binary that has already
 // been compiled. We ignore exit status.
-func buildRunProgram(b *work.Builder, a *work.Action) error {
+func buildRunProgram(b *work.Builder, ctx context.Context, a *work.Action) error {
 	cmdline := str.StringList(work.FindExecCmd(), a.Deps[0].Target, a.Args)
 	if cfg.BuildN || cfg.BuildX {
 		b.Showcmd("", "%s", strings.Join(cmdline, " "))
diff --git a/src/cmd/go/internal/str/path.go b/src/cmd/go/internal/str/path.go
index 95d91a3332..51ab2af82b 100644
--- a/src/cmd/go/internal/str/path.go
+++ b/src/cmd/go/internal/str/path.go
@@ -5,7 +5,6 @@
 package str
 
 import (
-	"path"
 	"path/filepath"
 	"strings"
 )
@@ -50,47 +49,3 @@ func HasFilePathPrefix(s, prefix string) bool {
 		return s[len(prefix)] == filepath.Separator && s[:len(prefix)] == prefix
 	}
 }
-
-// GlobsMatchPath reports whether any path prefix of target
-// matches one of the glob patterns (as defined by path.Match)
-// in the comma-separated globs list.
-// It ignores any empty or malformed patterns in the list.
-func GlobsMatchPath(globs, target string) bool {
-	for globs != "" {
-		// Extract next non-empty glob in comma-separated list.
-		var glob string
-		if i := strings.Index(globs, ","); i >= 0 {
-			glob, globs = globs[:i], globs[i+1:]
-		} else {
-			glob, globs = globs, ""
-		}
-		if glob == "" {
-			continue
-		}
-
-		// A glob with N+1 path elements (N slashes) needs to be matched
-		// against the first N+1 path elements of target,
-		// which end just before the N+1'th slash.
-		n := strings.Count(glob, "/")
-		prefix := target
-		// Walk target, counting slashes, truncating at the N+1'th slash.
-		for i := 0; i < len(target); i++ {
-			if target[i] == '/' {
-				if n == 0 {
-					prefix = target[:i]
-					break
-				}
-				n--
-			}
-		}
-		if n > 0 {
-			// Not enough prefix elements.
-			continue
-		}
-		matched, _ := path.Match(glob, prefix)
-		if matched {
-			return true
-		}
-	}
-	return false
-}
diff --git a/src/cmd/go/internal/test/test.go b/src/cmd/go/internal/test/test.go
index 873a76aa38..3aee6939d2 100644
--- a/src/cmd/go/internal/test/test.go
+++ b/src/cmd/go/internal/test/test.go
@@ -6,6 +6,7 @@ package test
 
 import (
 	"bytes"
+	"context"
 	"crypto/sha256"
 	"errors"
 	"fmt"
@@ -30,6 +31,7 @@ import (
 	"cmd/go/internal/lockedfile"
 	"cmd/go/internal/modload"
 	"cmd/go/internal/str"
+	"cmd/go/internal/trace"
 	"cmd/go/internal/work"
 	"cmd/internal/test2json"
 )
@@ -565,18 +567,35 @@ var defaultVetFlags = []string{
 	// "-unusedresult",
 }
 
-func runTest(cmd *base.Command, args []string) {
+func runTest(ctx context.Context, cmd *base.Command, args []string) {
 	modload.LoadTests = true
 
 	pkgArgs, testArgs = testFlags(args)
 
+	if cfg.DebugTrace != "" {
+		var close func() error
+		var err error
+		ctx, close, err = trace.Start(ctx, cfg.DebugTrace)
+		if err != nil {
+			base.Fatalf("failed to start trace: %v", err)
+		}
+		defer func() {
+			if err := close(); err != nil {
+				base.Fatalf("failed to stop trace: %v", err)
+			}
+		}()
+	}
+
+	ctx, span := trace.StartSpan(ctx, fmt.Sprint("Running ", cmd.Name(), " command"))
+	defer span.Done()
+
 	work.FindExecCmd() // initialize cached result
 
 	work.BuildInit()
 	work.VetFlags = testVet.flags
 	work.VetExplicit = testVet.explicit
 
-	pkgs = load.PackagesForBuild(pkgArgs)
+	pkgs = load.PackagesForBuild(ctx, pkgArgs)
 	if len(pkgs) == 0 {
 		base.Fatalf("no packages to test")
 	}
@@ -658,7 +677,7 @@ func runTest(cmd *base.Command, args []string) {
 		sort.Strings(all)
 
 		a := &work.Action{Mode: "go test -i"}
-		for _, p := range load.PackagesForBuild(all) {
+		for _, p := range load.PackagesForBuild(ctx, all) {
 			if cfg.BuildToolchainName == "gccgo" && p.Standard {
 				// gccgo's standard library packages
 				// can not be reinstalled.
@@ -666,7 +685,7 @@ func runTest(cmd *base.Command, args []string) {
 			}
 			a.Deps = append(a.Deps, b.CompileAction(work.ModeInstall, work.ModeInstall, p))
 		}
-		b.Do(a)
+		b.Do(ctx, a)
 		if !testC || a.Failed {
 			return
 		}
@@ -683,7 +702,7 @@ func runTest(cmd *base.Command, args []string) {
 		}
 
 		// Select for coverage all dependencies matching the testCoverPaths patterns.
-		for _, p := range load.TestPackageList(pkgs) {
+		for _, p := range load.TestPackageList(ctx, pkgs) {
 			haveMatch := false
 			for i := range testCoverPaths {
 				if match[i](p) {
@@ -745,7 +764,7 @@ func runTest(cmd *base.Command, args []string) {
 			ensureImport(p, "sync/atomic")
 		}
 
-		buildTest, runTest, printTest, err := builderTest(&b, p)
+		buildTest, runTest, printTest, err := builderTest(&b, ctx, p)
 		if err != nil {
 			str := err.Error()
 			str = strings.TrimPrefix(str, "\n")
@@ -786,7 +805,7 @@ func runTest(cmd *base.Command, args []string) {
 		}
 	}
 
-	b.Do(root)
+	b.Do(ctx, root)
 }
 
 // ensures that package p imports the named package
@@ -812,7 +831,7 @@ var windowsBadWords = []string{
 	"update",
 }
 
-func builderTest(b *work.Builder, p *load.Package) (buildAction, runAction, printAction *work.Action, err error) {
+func builderTest(b *work.Builder, ctx context.Context, p *load.Package) (buildAction, runAction, printAction *work.Action, err error) {
 	if len(p.TestGoFiles)+len(p.XTestGoFiles) == 0 {
 		build := b.CompileAction(work.ModeBuild, work.ModeBuild, p)
 		run := &work.Action{Mode: "test run", Package: p, Deps: []*work.Action{build}}
@@ -835,7 +854,7 @@ func builderTest(b *work.Builder, p *load.Package) (buildAction, runAction, prin
 			DeclVars: declareCoverVars,
 		}
 	}
-	pmain, ptest, pxtest, err := load.TestPackagesFor(p, cover)
+	pmain, ptest, pxtest, err := load.TestPackagesFor(ctx, p, cover)
 	if err != nil {
 		return nil, nil, nil, err
 	}
@@ -1068,7 +1087,7 @@ func (lockedStdout) Write(b []byte) (int, error) {
 }
 
 // builderRunTest is the action for running a test binary.
-func (c *runCache) builderRunTest(b *work.Builder, a *work.Action) error {
+func (c *runCache) builderRunTest(b *work.Builder, ctx context.Context, a *work.Action) error {
 	if a.Failed {
 		// We were unable to build the binary.
 		a.Failed = false
@@ -1079,9 +1098,13 @@ func (c *runCache) builderRunTest(b *work.Builder, a *work.Action) error {
 	}
 
 	var stdout io.Writer = os.Stdout
+	var err error
 	if testJSON {
 		json := test2json.NewConverter(lockedStdout{}, a.Package.ImportPath, test2json.Timestamp)
-		defer json.Close()
+		defer func() {
+			json.Exited(err)
+			json.Close()
+		}()
 		stdout = json
 	}
 
@@ -1185,7 +1208,7 @@ func (c *runCache) builderRunTest(b *work.Builder, a *work.Action) error {
 	}
 
 	t0 := time.Now()
-	err := cmd.Start()
+	err = cmd.Start()
 
 	// This is a last-ditch deadline to detect and
 	// stop wedged test binaries, to keep the builders
@@ -1641,7 +1664,7 @@ func coveragePercentage(out []byte) string {
 }
 
 // builderCleanTest is the action for cleaning up after a test.
-func builderCleanTest(b *work.Builder, a *work.Action) error {
+func builderCleanTest(b *work.Builder, ctx context.Context, a *work.Action) error {
 	if cfg.BuildWork {
 		return nil
 	}
@@ -1653,7 +1676,7 @@ func builderCleanTest(b *work.Builder, a *work.Action) error {
 }
 
 // builderPrintTest is the action for printing a test result.
-func builderPrintTest(b *work.Builder, a *work.Action) error {
+func builderPrintTest(b *work.Builder, ctx context.Context, a *work.Action) error {
 	clean := a.Deps[0]
 	run := clean.Deps[0]
 	if run.TestOutput != nil {
@@ -1664,7 +1687,7 @@ func builderPrintTest(b *work.Builder, a *work.Action) error {
 }
 
 // builderNoTest is the action for testing a package with no test files.
-func builderNoTest(b *work.Builder, a *work.Action) error {
+func builderNoTest(b *work.Builder, ctx context.Context, a *work.Action) error {
 	var stdout io.Writer = os.Stdout
 	if testJSON {
 		json := test2json.NewConverter(lockedStdout{}, a.Package.ImportPath, test2json.Timestamp)
@@ -1676,7 +1699,7 @@ func builderNoTest(b *work.Builder, a *work.Action) error {
 }
 
 // printExitStatus is the action for printing the exit status
-func printExitStatus(b *work.Builder, a *work.Action) error {
+func printExitStatus(b *work.Builder, ctx context.Context, a *work.Action) error {
 	if !testJSON && len(pkgArgs) != 0 {
 		if base.GetExitStatus() != 0 {
 			fmt.Println("FAIL")
diff --git a/src/cmd/go/internal/test/testflag.go b/src/cmd/go/internal/test/testflag.go
index 1ff34f7445..4f0a8924f1 100644
--- a/src/cmd/go/internal/test/testflag.go
+++ b/src/cmd/go/internal/test/testflag.go
@@ -214,9 +214,13 @@ func testFlags(args []string) (packageNames, passToTest []string) {
 
 	explicitArgs := make([]string, 0, len(args))
 	inPkgList := false
+	afterFlagWithoutValue := false
 	for len(args) > 0 {
 		f, remainingArgs, err := cmdflag.ParseOne(&CmdTest.Flag, args)
 
+		wasAfterFlagWithoutValue := afterFlagWithoutValue
+		afterFlagWithoutValue = false // provisionally
+
 		if errors.Is(err, flag.ErrHelp) {
 			exitWithUsage()
 		}
@@ -233,10 +237,24 @@ func testFlags(args []string) (packageNames, passToTest []string) {
 		if nf := (cmdflag.NonFlagError{}); errors.As(err, &nf) {
 			if !inPkgList && packageNames != nil {
 				// We already saw the package list previously, and this argument is not
-				// a flag, so it — and everything after it — must be a literal argument
-				// to the test binary.
-				explicitArgs = append(explicitArgs, args...)
-				break
+				// a flag, so it — and everything after it — must be either a value for
+				// a preceding flag or a literal argument to the test binary.
+				if wasAfterFlagWithoutValue {
+					// This argument could syntactically be a flag value, so
+					// optimistically assume that it is and keep looking for go command
+					// flags after it.
+					//
+					// (If we're wrong, we'll at least be consistent with historical
+					// behavior; see https://golang.org/issue/40763.)
+					explicitArgs = append(explicitArgs, nf.RawArg)
+					args = remainingArgs
+					continue
+				} else {
+					// This argument syntactically cannot be a flag value, so it must be a
+					// positional argument, and so must everything after it.
+					explicitArgs = append(explicitArgs, args...)
+					break
+				}
 			}
 
 			inPkgList = true
@@ -272,6 +290,9 @@ func testFlags(args []string) (packageNames, passToTest []string) {
 
 			explicitArgs = append(explicitArgs, nd.RawArg)
 			args = remainingArgs
+			if !nd.HasValue {
+				afterFlagWithoutValue = true
+			}
 			continue
 		}
 
diff --git a/src/cmd/go/internal/tool/tool.go b/src/cmd/go/internal/tool/tool.go
index 930eecb63f..7f4dc86802 100644
--- a/src/cmd/go/internal/tool/tool.go
+++ b/src/cmd/go/internal/tool/tool.go
@@ -6,6 +6,7 @@
 package tool
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"os/exec"
@@ -48,7 +49,7 @@ func init() {
 	CmdTool.Flag.BoolVar(&toolN, "n", false, "")
 }
 
-func runTool(cmd *base.Command, args []string) {
+func runTool(ctx context.Context, cmd *base.Command, args []string) {
 	if len(args) == 0 {
 		listTools()
 		return
diff --git a/src/cmd/go/internal/trace/trace.go b/src/cmd/go/internal/trace/trace.go
new file mode 100644
index 0000000000..f108a2b6ca
--- /dev/null
+++ b/src/cmd/go/internal/trace/trace.go
@@ -0,0 +1,206 @@
+// Copyright 2020 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 trace
+
+import (
+	"cmd/internal/traceviewer"
+	"context"
+	"encoding/json"
+	"errors"
+	"os"
+	"strings"
+	"sync/atomic"
+	"time"
+)
+
+// Constants used in event fields.
+// See https://docs.google.com/document/d/1CvAClvFfyA5R-PhYUmn5OOQtYMH4h6I0nSsKchNAySU
+// for more details.
+const (
+	phaseDurationBegin = "B"
+	phaseDurationEnd   = "E"
+	phaseFlowStart     = "s"
+	phaseFlowEnd       = "f"
+
+	bindEnclosingSlice = "e"
+)
+
+var traceStarted int32
+
+func getTraceContext(ctx context.Context) (traceContext, bool) {
+	if atomic.LoadInt32(&traceStarted) == 0 {
+		return traceContext{}, false
+	}
+	v := ctx.Value(traceKey{})
+	if v == nil {
+		return traceContext{}, false
+	}
+	return v.(traceContext), true
+}
+
+// StartSpan starts a trace event with the given name. The Span ends when its Done method is called.
+func StartSpan(ctx context.Context, name string) (context.Context, *Span) {
+	tc, ok := getTraceContext(ctx)
+	if !ok {
+		return ctx, nil
+	}
+	childSpan := &Span{t: tc.t, name: name, tid: tc.tid, start: time.Now()}
+	tc.t.writeEvent(&traceviewer.Event{
+		Name:  childSpan.name,
+		Time:  float64(childSpan.start.UnixNano()) / float64(time.Microsecond),
+		TID:   childSpan.tid,
+		Phase: phaseDurationBegin,
+	})
+	ctx = context.WithValue(ctx, traceKey{}, traceContext{tc.t, tc.tid})
+	return ctx, childSpan
+}
+
+// StartGoroutine associates the context with a new Thread ID. The Chrome trace viewer associates each
+// trace event with a thread, and doesn't expect events with the same thread id to happen at the
+// same time.
+func StartGoroutine(ctx context.Context) context.Context {
+	tc, ok := getTraceContext(ctx)
+	if !ok {
+		return ctx
+	}
+	return context.WithValue(ctx, traceKey{}, traceContext{tc.t, tc.t.getNextTID()})
+}
+
+// Flow marks a flow indicating that the 'to' span depends on the 'from' span.
+// Flow should be called while the 'to' span is in progress.
+func Flow(ctx context.Context, from *Span, to *Span) {
+	tc, ok := getTraceContext(ctx)
+	if !ok || from == nil || to == nil {
+		return
+	}
+
+	id := tc.t.getNextFlowID()
+	tc.t.writeEvent(&traceviewer.Event{
+		Name:     from.name + " -> " + to.name,
+		Category: "flow",
+		ID:       id,
+		Time:     float64(from.end.UnixNano()) / float64(time.Microsecond),
+		Phase:    phaseFlowStart,
+		TID:      from.tid,
+	})
+	tc.t.writeEvent(&traceviewer.Event{
+		Name:      from.name + " -> " + to.name,
+		Category:  "flow", // TODO(matloob): Add Category to Flow?
+		ID:        id,
+		Time:      float64(to.start.UnixNano()) / float64(time.Microsecond),
+		Phase:     phaseFlowEnd,
+		TID:       to.tid,
+		BindPoint: bindEnclosingSlice,
+	})
+}
+
+type Span struct {
+	t *tracer
+
+	name  string
+	tid   uint64
+	start time.Time
+	end   time.Time
+}
+
+func (s *Span) Done() {
+	if s == nil {
+		return
+	}
+	s.end = time.Now()
+	s.t.writeEvent(&traceviewer.Event{
+		Name:  s.name,
+		Time:  float64(s.end.UnixNano()) / float64(time.Microsecond),
+		TID:   s.tid,
+		Phase: phaseDurationEnd,
+	})
+}
+
+type tracer struct {
+	file chan traceFile // 1-buffered
+
+	nextTID    uint64
+	nextFlowID uint64
+}
+
+func (t *tracer) writeEvent(ev *traceviewer.Event) error {
+	f := <-t.file
+	defer func() { t.file <- f }()
+	var err error
+	if f.entries == 0 {
+		_, err = f.sb.WriteString("[\n")
+	} else {
+		_, err = f.sb.WriteString(",")
+	}
+	f.entries++
+	if err != nil {
+		return nil
+	}
+
+	if err := f.enc.Encode(ev); err != nil {
+		return err
+	}
+
+	// Write event string to output file.
+	_, err = f.f.WriteString(f.sb.String())
+	f.sb.Reset()
+	return err
+}
+
+func (t *tracer) Close() error {
+	f := <-t.file
+	defer func() { t.file <- f }()
+
+	_, firstErr := f.f.WriteString("]")
+	if err := f.f.Close(); firstErr == nil {
+		firstErr = err
+	}
+	return firstErr
+}
+
+func (t *tracer) getNextTID() uint64 {
+	return atomic.AddUint64(&t.nextTID, 1)
+}
+
+func (t *tracer) getNextFlowID() uint64 {
+	return atomic.AddUint64(&t.nextFlowID, 1)
+}
+
+// traceKey is the context key for tracing information. It is unexported to prevent collisions with context keys defined in
+// other packages.
+type traceKey struct{}
+
+type traceContext struct {
+	t   *tracer
+	tid uint64
+}
+
+// Start starts a trace which writes to the given file.
+func Start(ctx context.Context, file string) (context.Context, func() error, error) {
+	atomic.StoreInt32(&traceStarted, 1)
+	if file == "" {
+		return nil, nil, errors.New("no trace file supplied")
+	}
+	f, err := os.Create(file)
+	if err != nil {
+		return nil, nil, err
+	}
+	t := &tracer{file: make(chan traceFile, 1)}
+	sb := new(strings.Builder)
+	t.file <- traceFile{
+		f:   f,
+		sb:  sb,
+		enc: json.NewEncoder(sb),
+	}
+	ctx = context.WithValue(ctx, traceKey{}, traceContext{t: t})
+	return ctx, t.Close, nil
+}
+
+type traceFile struct {
+	f       *os.File
+	sb      *strings.Builder
+	enc     *json.Encoder
+	entries int64
+}
diff --git a/src/cmd/go/internal/version/version.go b/src/cmd/go/internal/version/version.go
index ac2ae50155..c2de8d326d 100644
--- a/src/cmd/go/internal/version/version.go
+++ b/src/cmd/go/internal/version/version.go
@@ -7,6 +7,7 @@ package version
 
 import (
 	"bytes"
+	"context"
 	"encoding/binary"
 	"fmt"
 	"os"
@@ -51,7 +52,7 @@ var (
 	versionV = CmdVersion.Flag.Bool("v", false, "")
 )
 
-func runVersion(cmd *base.Command, args []string) {
+func runVersion(ctx context.Context, cmd *base.Command, args []string) {
 	if len(args) == 0 {
 		if *versionM || *versionV {
 			fmt.Fprintf(os.Stderr, "go version: flags can only be used with arguments\n")
@@ -137,7 +138,7 @@ func scanFile(file string, info os.FileInfo, mustPrint bool) {
 
 	fmt.Printf("%s: %s\n", file, vers)
 	if *versionM && mod != "" {
-		fmt.Printf("\t%s\n", strings.Replace(mod[:len(mod)-1], "\n", "\n\t", -1))
+		fmt.Printf("\t%s\n", strings.ReplaceAll(mod[:len(mod)-1], "\n", "\n\t"))
 	}
 }
 
diff --git a/src/cmd/go/internal/vet/vet.go b/src/cmd/go/internal/vet/vet.go
index 4ec58de785..cf2c8d59e8 100644
--- a/src/cmd/go/internal/vet/vet.go
+++ b/src/cmd/go/internal/vet/vet.go
@@ -6,11 +6,16 @@
 package vet
 
 import (
+	"context"
+	"fmt"
+	"path/filepath"
+
 	"cmd/go/internal/base"
+	"cmd/go/internal/cfg"
 	"cmd/go/internal/load"
 	"cmd/go/internal/modload"
+	"cmd/go/internal/trace"
 	"cmd/go/internal/work"
-	"path/filepath"
 )
 
 // Break init loop.
@@ -48,11 +53,28 @@ See also: go fmt, go fix.
 	`,
 }
 
-func runVet(cmd *base.Command, args []string) {
+func runVet(ctx context.Context, cmd *base.Command, args []string) {
 	modload.LoadTests = true
 
 	vetFlags, pkgArgs := vetFlags(args)
 
+	if cfg.DebugTrace != "" {
+		var close func() error
+		var err error
+		ctx, close, err = trace.Start(ctx, cfg.DebugTrace)
+		if err != nil {
+			base.Fatalf("failed to start trace: %v", err)
+		}
+		defer func() {
+			if err := close(); err != nil {
+				base.Fatalf("failed to stop trace: %v", err)
+			}
+		}()
+	}
+
+	ctx, span := trace.StartSpan(ctx, fmt.Sprint("Running ", cmd.Name(), " command"))
+	defer span.Done()
+
 	work.BuildInit()
 	work.VetFlags = vetFlags
 	if len(vetFlags) > 0 {
@@ -66,7 +88,7 @@ func runVet(cmd *base.Command, args []string) {
 		}
 	}
 
-	pkgs := load.PackagesForBuild(pkgArgs)
+	pkgs := load.PackagesForBuild(ctx, pkgArgs)
 	if len(pkgs) == 0 {
 		base.Fatalf("no packages to vet")
 	}
@@ -76,7 +98,7 @@ func runVet(cmd *base.Command, args []string) {
 
 	root := &work.Action{Mode: "go vet"}
 	for _, p := range pkgs {
-		_, ptest, pxtest, err := load.TestPackagesFor(p, nil)
+		_, ptest, pxtest, err := load.TestPackagesFor(ctx, p, nil)
 		if err != nil {
 			base.Errorf("%v", err)
 			continue
@@ -92,5 +114,5 @@ func runVet(cmd *base.Command, args []string) {
 			root.Deps = append(root.Deps, b.VetAction(work.ModeBuild, work.ModeBuild, pxtest))
 		}
 	}
-	b.Do(root)
+	b.Do(ctx, root)
 }
diff --git a/src/cmd/go/internal/work/action.go b/src/cmd/go/internal/work/action.go
index 6b5f9e4807..825e763c03 100644
--- a/src/cmd/go/internal/work/action.go
+++ b/src/cmd/go/internal/work/action.go
@@ -10,6 +10,7 @@ import (
 	"bufio"
 	"bytes"
 	"container/heap"
+	"context"
 	"debug/elf"
 	"encoding/json"
 	"fmt"
@@ -25,6 +26,7 @@ import (
 	"cmd/go/internal/cache"
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/load"
+	"cmd/go/internal/trace"
 	"cmd/internal/buildid"
 )
 
@@ -63,13 +65,13 @@ type Builder struct {
 
 // An Action represents a single action in the action graph.
 type Action struct {
-	Mode       string                        // description of action operation
-	Package    *load.Package                 // the package this action works on
-	Deps       []*Action                     // actions that must happen before this one
-	Func       func(*Builder, *Action) error // the action itself (nil = no-op)
-	IgnoreFail bool                          // whether to run f even if dependencies fail
-	TestOutput *bytes.Buffer                 // test output buffer
-	Args       []string                      // additional args for runProgram
+	Mode       string                                         // description of action operation
+	Package    *load.Package                                  // the package this action works on
+	Deps       []*Action                                      // actions that must happen before this one
+	Func       func(*Builder, context.Context, *Action) error // the action itself (nil = no-op)
+	IgnoreFail bool                                           // whether to run f even if dependencies fail
+	TestOutput *bytes.Buffer                                  // test output buffer
+	Args       []string                                       // additional args for runProgram
 
 	triggers []*Action // inverse of deps
 
@@ -91,10 +93,11 @@ type Action struct {
 	output    []byte     // output redirect buffer (nil means use b.Print)
 
 	// Execution state.
-	pending  int         // number of deps yet to complete
-	priority int         // relative execution priority
-	Failed   bool        // whether the action failed
-	json     *actionJSON // action graph information
+	pending   int         // number of deps yet to complete
+	priority  int         // relative execution priority
+	Failed    bool        // whether the action failed
+	json      *actionJSON // action graph information
+	traceSpan *trace.Span
 }
 
 // BuildActionID returns the action ID section of a's build ID.
diff --git a/src/cmd/go/internal/work/build.go b/src/cmd/go/internal/work/build.go
index 7146c9ce00..d020aa6e9f 100644
--- a/src/cmd/go/internal/work/build.go
+++ b/src/cmd/go/internal/work/build.go
@@ -5,6 +5,7 @@
 package work
 
 import (
+	"context"
 	"errors"
 	"fmt"
 	"go/build"
@@ -18,6 +19,7 @@ import (
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/load"
 	"cmd/go/internal/search"
+	"cmd/go/internal/trace"
 )
 
 var CmdBuild = &base.Command{
@@ -270,6 +272,7 @@ func AddBuildFlags(cmd *base.Command, mask BuildFlagMask) {
 
 	// Undocumented, unstable debugging flags.
 	cmd.Flag.StringVar(&cfg.DebugActiongraph, "debug-actiongraph", "", "")
+	cmd.Flag.StringVar(&cfg.DebugTrace, "debug-trace", "", "")
 }
 
 // AddModCommonFlags adds the module-related flags common to build commands
@@ -343,12 +346,12 @@ var pkgsFilter = func(pkgs []*load.Package) []*load.Package { return pkgs }
 
 var runtimeVersion = runtime.Version()
 
-func runBuild(cmd *base.Command, args []string) {
+func runBuild(ctx context.Context, cmd *base.Command, args []string) {
 	BuildInit()
 	var b Builder
 	b.Init()
 
-	pkgs := load.PackagesForBuild(args)
+	pkgs := load.PackagesForBuild(ctx, args)
 
 	explicitO := len(cfg.BuildO) > 0
 
@@ -377,7 +380,7 @@ func runBuild(cmd *base.Command, args []string) {
 		depMode = ModeInstall
 	}
 
-	pkgs = omitTestOnly(pkgsFilter(load.Packages(args)))
+	pkgs = omitTestOnly(pkgsFilter(load.Packages(ctx, args)))
 
 	// Special case -o /dev/null by not writing at all.
 	if cfg.BuildO == os.DevNull {
@@ -407,7 +410,7 @@ func runBuild(cmd *base.Command, args []string) {
 			if len(a.Deps) == 0 {
 				base.Fatalf("go build: no main packages to build")
 			}
-			b.Do(a)
+			b.Do(ctx, a)
 			return
 		}
 		if len(pkgs) > 1 {
@@ -420,7 +423,7 @@ func runBuild(cmd *base.Command, args []string) {
 		p.Stale = true // must build - not up to date
 		p.StaleReason = "build -o flag in use"
 		a := b.AutoAction(ModeInstall, depMode, p)
-		b.Do(a)
+		b.Do(ctx, a)
 		return
 	}
 
@@ -431,7 +434,7 @@ func runBuild(cmd *base.Command, args []string) {
 	if cfg.BuildBuildmode == "shared" {
 		a = b.buildmodeShared(ModeBuild, depMode, args, pkgs, a)
 	}
-	b.Do(a)
+	b.Do(ctx, a)
 }
 
 var CmdInstall = &base.Command{
@@ -514,9 +517,9 @@ func libname(args []string, pkgs []*load.Package) (string, error) {
 	return "lib" + libname + ".so", nil
 }
 
-func runInstall(cmd *base.Command, args []string) {
+func runInstall(ctx context.Context, cmd *base.Command, args []string) {
 	BuildInit()
-	InstallPackages(args, load.PackagesForBuild(args))
+	InstallPackages(ctx, args, load.PackagesForBuild(ctx, args))
 }
 
 // omitTestOnly returns pkgs with test-only packages removed.
@@ -536,7 +539,10 @@ func omitTestOnly(pkgs []*load.Package) []*load.Package {
 	return list
 }
 
-func InstallPackages(patterns []string, pkgs []*load.Package) {
+func InstallPackages(ctx context.Context, patterns []string, pkgs []*load.Package) {
+	ctx, span := trace.StartSpan(ctx, "InstallPackages "+strings.Join(patterns, " "))
+	defer span.Done()
+
 	if cfg.GOBIN != "" && !filepath.IsAbs(cfg.GOBIN) {
 		base.Fatalf("cannot install, GOBIN must be an absolute path")
 	}
@@ -605,7 +611,7 @@ func InstallPackages(patterns []string, pkgs []*load.Package) {
 		a = b.buildmodeShared(ModeInstall, ModeInstall, patterns, pkgs, a)
 	}
 
-	b.Do(a)
+	b.Do(ctx, a)
 	base.ExitIfErrors()
 
 	// Success. If this command is 'go install' with no arguments
diff --git a/src/cmd/go/internal/work/exec.go b/src/cmd/go/internal/work/exec.go
index 071c9d2db9..d975c36306 100644
--- a/src/cmd/go/internal/work/exec.go
+++ b/src/cmd/go/internal/work/exec.go
@@ -8,6 +8,7 @@ package work
 
 import (
 	"bytes"
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -31,6 +32,7 @@ import (
 	"cmd/go/internal/cfg"
 	"cmd/go/internal/load"
 	"cmd/go/internal/str"
+	"cmd/go/internal/trace"
 )
 
 // actionList returns the list of actions in the dag rooted at root
@@ -54,7 +56,10 @@ func actionList(root *Action) []*Action {
 }
 
 // do runs the action graph rooted at root.
-func (b *Builder) Do(root *Action) {
+func (b *Builder) Do(ctx context.Context, root *Action) {
+	ctx, span := trace.StartSpan(ctx, "exec.Builder.Do ("+root.Mode+" "+root.Target+")")
+	defer span.Done()
+
 	if !b.IsCmdList {
 		// If we're doing real work, take time at the end to trim the cache.
 		c := cache.Default()
@@ -110,13 +115,24 @@ func (b *Builder) Do(root *Action) {
 
 	// Handle runs a single action and takes care of triggering
 	// any actions that are runnable as a result.
-	handle := func(a *Action) {
+	handle := func(ctx context.Context, a *Action) {
 		if a.json != nil {
 			a.json.TimeStart = time.Now()
 		}
 		var err error
 		if a.Func != nil && (!a.Failed || a.IgnoreFail) {
-			err = a.Func(b, a)
+			// TODO(matloob): Better action descriptions
+			desc := "Executing action "
+			if a.Package != nil {
+				desc += "(" + a.Mode + " " + a.Package.Desc() + ")"
+			}
+			ctx, span := trace.StartSpan(ctx, desc)
+			a.traceSpan = span
+			for _, d := range a.Deps {
+				trace.Flow(ctx, d.traceSpan, a.traceSpan)
+			}
+			err = a.Func(b, ctx, a)
+			span.Done()
 		}
 		if a.json != nil {
 			a.json.TimeDone = time.Now()
@@ -164,6 +180,7 @@ func (b *Builder) Do(root *Action) {
 	for i := 0; i < par; i++ {
 		wg.Add(1)
 		go func() {
+			ctx := trace.StartGoroutine(ctx)
 			defer wg.Done()
 			for {
 				select {
@@ -176,7 +193,7 @@ func (b *Builder) Do(root *Action) {
 					b.exec.Lock()
 					a := b.ready.pop()
 					b.exec.Unlock()
-					handle(a)
+					handle(ctx, a)
 				case <-base.Interrupted:
 					base.SetExitStatus(1)
 					return
@@ -386,7 +403,7 @@ const (
 
 // build is the action for building a single package.
 // Note that any new influence on this logic must be reported in b.buildActionID above as well.
-func (b *Builder) build(a *Action) (err error) {
+func (b *Builder) build(ctx context.Context, a *Action) (err error) {
 	p := a.Package
 
 	bit := func(x uint32, b bool) uint32 {
@@ -991,7 +1008,7 @@ var VetFlags []string
 // VetExplicit records whether the vet flags were set explicitly on the command line.
 var VetExplicit bool
 
-func (b *Builder) vet(a *Action) error {
+func (b *Builder) vet(ctx context.Context, a *Action) error {
 	// a.Deps[0] is the build of the package being vetted.
 	// a.Deps[1] is the build of the "fmt" package.
 
@@ -1182,7 +1199,7 @@ func (b *Builder) printLinkerConfig(h io.Writer, p *load.Package) {
 
 // link is the action for linking a single command.
 // Note that any new influence on this logic must be reported in b.linkActionID above as well.
-func (b *Builder) link(a *Action) (err error) {
+func (b *Builder) link(ctx context.Context, a *Action) (err error) {
 	if b.useCache(a, b.linkActionID(a), a.Package.Target) || b.IsCmdList {
 		return nil
 	}
@@ -1374,7 +1391,7 @@ func (b *Builder) getPkgConfigFlags(p *load.Package) (cflags, ldflags []string,
 	return
 }
 
-func (b *Builder) installShlibname(a *Action) error {
+func (b *Builder) installShlibname(ctx context.Context, a *Action) error {
 	if err := allowInstall(a); err != nil {
 		return err
 	}
@@ -1423,7 +1440,7 @@ func (b *Builder) linkSharedActionID(a *Action) cache.ActionID {
 	return h.Sum()
 }
 
-func (b *Builder) linkShared(a *Action) (err error) {
+func (b *Builder) linkShared(ctx context.Context, a *Action) (err error) {
 	if b.useCache(a, b.linkSharedActionID(a), a.Target) || b.IsCmdList {
 		return nil
 	}
@@ -1449,7 +1466,7 @@ func (b *Builder) linkShared(a *Action) (err error) {
 }
 
 // BuildInstallFunc is the action for installing a single package or executable.
-func BuildInstallFunc(b *Builder, a *Action) (err error) {
+func BuildInstallFunc(b *Builder, ctx context.Context, a *Action) (err error) {
 	defer func() {
 		if err != nil && err != errPrintedOutput {
 			// a.Package == nil is possible for the go install -buildmode=shared
@@ -1702,7 +1719,7 @@ func (b *Builder) writeFile(file string, text []byte) error {
 }
 
 // Install the cgo export header file, if there is one.
-func (b *Builder) installHeader(a *Action) error {
+func (b *Builder) installHeader(ctx context.Context, a *Action) error {
 	src := a.Objdir + "_cgo_install.h"
 	if _, err := os.Stat(src); os.IsNotExist(err) {
 		// If the file does not exist, there are no exported
@@ -2883,7 +2900,7 @@ func (b *Builder) swigDoIntSize(objdir string) (intsize string, err error) {
 	}
 	srcs := []string{src}
 
-	p := load.GoFilesPackage(srcs)
+	p := load.GoFilesPackage(context.TODO(), srcs)
 
 	if _, _, e := BuildToolchain.gc(b, &Action{Mode: "swigDoIntSize", Package: p, Objdir: objdir}, "", nil, "", false, srcs); e != nil {
 		return "32", nil
diff --git a/src/cmd/go/main.go b/src/cmd/go/main.go
index fdf49b7380..37bb7d6d27 100644
--- a/src/cmd/go/main.go
+++ b/src/cmd/go/main.go
@@ -7,6 +7,7 @@
 package main
 
 import (
+	"context"
 	"flag"
 	"fmt"
 	"log"
@@ -34,6 +35,7 @@ import (
 	"cmd/go/internal/run"
 	"cmd/go/internal/test"
 	"cmd/go/internal/tool"
+	"cmd/go/internal/trace"
 	"cmd/go/internal/version"
 	"cmd/go/internal/vet"
 	"cmd/go/internal/work"
@@ -187,7 +189,10 @@ BigCmdLoop:
 				cmd.Flag.Parse(args[1:])
 				args = cmd.Flag.Args()
 			}
-			cmd.Run(cmd, args)
+			ctx := maybeStartTrace(context.Background())
+			ctx, span := trace.StartSpan(ctx, fmt.Sprint("Running ", cmd.Name(), " command"))
+			cmd.Run(ctx, cmd, args)
+			span.Done()
 			base.Exit()
 			return
 		}
@@ -209,3 +214,21 @@ func mainUsage() {
 	help.PrintUsage(os.Stderr, base.Go)
 	os.Exit(2)
 }
+
+func maybeStartTrace(pctx context.Context) context.Context {
+	if cfg.DebugTrace == "" {
+		return pctx
+	}
+
+	ctx, close, err := trace.Start(pctx, cfg.DebugTrace)
+	if err != nil {
+		base.Fatalf("failed to start trace: %v", err)
+	}
+	base.AtExit(func() {
+		if err := close(); err != nil {
+			base.Fatalf("failed to stop trace: %v", err)
+		}
+	})
+
+	return ctx
+}
diff --git a/src/cmd/go/proxy_test.go b/src/cmd/go/proxy_test.go
index 2a4d2935b3..7f58fb8ce4 100644
--- a/src/cmd/go/proxy_test.go
+++ b/src/cmd/go/proxy_test.go
@@ -131,6 +131,12 @@ func proxyHandler(w http.ResponseWriter, r *http.Request) {
 	}
 	path := r.URL.Path[len("/mod/"):]
 
+	// /mod/invalid returns faulty responses.
+	if strings.HasPrefix(path, "invalid/") {
+		w.Write([]byte("invalid"))
+		return
+	}
+
 	// /mod/quiet/ does not print errors.
 	quiet := false
 	if strings.HasPrefix(path, "quiet/") {
diff --git a/src/cmd/go/script_test.go b/src/cmd/go/script_test.go
index 2e8f18a897..986646252a 100644
--- a/src/cmd/go/script_test.go
+++ b/src/cmd/go/script_test.go
@@ -22,6 +22,7 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
+	"sync"
 	"testing"
 	"time"
 
@@ -296,6 +297,8 @@ Script:
 				ok = os.Geteuid() == 0
 			case "symlink":
 				ok = testenv.HasSymlink()
+			case "case-sensitive":
+				ok = isCaseSensitive(ts.t)
 			default:
 				if strings.HasPrefix(cond.tag, "exec:") {
 					prog := cond.tag[len("exec:"):]
@@ -364,6 +367,41 @@ Script:
 	}
 }
 
+var (
+	onceCaseSensitive sync.Once
+	caseSensitive     bool
+)
+
+func isCaseSensitive(t *testing.T) bool {
+	onceCaseSensitive.Do(func() {
+		tmpdir, err := ioutil.TempDir("", "case-sensitive")
+		if err != nil {
+			t.Fatal("failed to create directory to determine case-sensitivity:", err)
+		}
+		defer os.RemoveAll(tmpdir)
+
+		fcap := filepath.Join(tmpdir, "FILE")
+		if err := ioutil.WriteFile(fcap, []byte{}, 0644); err != nil {
+			t.Fatal("error writing file to determine case-sensitivity:", err)
+		}
+
+		flow := filepath.Join(tmpdir, "file")
+		_, err = ioutil.ReadFile(flow)
+		switch {
+		case err == nil:
+			caseSensitive = false
+			return
+		case os.IsNotExist(err):
+			caseSensitive = true
+			return
+		default:
+			t.Fatal("unexpected error reading file when determining case-sensitivity:", err)
+		}
+	})
+
+	return caseSensitive
+}
+
 // scriptCmds are the script command implementations.
 // Keep list and the implementations below sorted by name.
 //
diff --git a/src/cmd/go/testdata/script/README b/src/cmd/go/testdata/script/README
index 76d6651718..d658cebfce 100644
--- a/src/cmd/go/testdata/script/README
+++ b/src/cmd/go/testdata/script/README
@@ -85,6 +85,7 @@ should only run when the condition is satisfied. The available conditions are:
    - [link] for testenv.HasLink()
    - [root] for os.Geteuid() == 0
    - [symlink] for testenv.HasSymlink()
+   - [case-sensitive] for whether the file system is case-sensitive
    - [exec:prog] for whether prog is available for execution (found by exec.LookPath)
    - [GODEBUG:value] for whether value is one of the comma-separated entries in the GODEBUG variable
    - [buildmode:value] for whether -buildmode=value is supported
diff --git a/src/cmd/go/testdata/script/build_GOTMPDIR.txt b/src/cmd/go/testdata/script/build_GOTMPDIR.txt
index c93ca932ca..1073517c29 100644
--- a/src/cmd/go/testdata/script/build_GOTMPDIR.txt
+++ b/src/cmd/go/testdata/script/build_GOTMPDIR.txt
@@ -1,15 +1,50 @@
-env GO111MODULE=off
-[short] skip
-
 # Set GOCACHE to a clean directory to ensure that 'go build' has work to report.
-env GOCACHE=$WORK/gocache
+[!windows] env GOCACHE=$WORK/gocache
+[windows] env GOCACHE=$WORK\gocache
 
-# Build should use GOTMPDIR if set.
-env GOTMPDIR=$WORK/my-favorite-tmpdir
+# 'go build' should use GOTMPDIR if set.
+[!windows] env GOTMPDIR=$WORK/my-favorite-tmpdir
+[windows] env GOTMPDIR=$WORK\my-favorite-tmpdir
 mkdir $GOTMPDIR
-go build -work hello.go
+go build -x hello.go
 stderr ^WORK=.*my-favorite-tmpdir
 
+# Make GOTMPDIR a regular file. This prevents the creation of work directories,
+# so we can check that certain commands don't create them.
+# This simulates running on a full disk or a read-only volume.
+rm $GOTMPDIR
+cp hello.go $GOTMPDIR # any file will do
+
+# 'go build' should fail if GOTMPDIR is read-only.
+! go build -x .
+stderr '^go: creating work dir: \w+ '$GOTMPDIR
+
+# 'go list' should only fail if it needs to build something.
+go list -x .
+! stderr 'creating work dir'
+stdout m
+go list -m all
+stdout m
+! go list -x -export .
+stderr '^go: creating work dir: \w+ '$GOTMPDIR
+
+# 'go clean -cache' and 'go clean -modcache' should not fail.
+go clean -x -cache
+! stderr 'creating work dir'
+go clean -x -modcache
+! stderr 'creating work dir'
+
+# 'go env' should not fail for specific variables.
+# Without arguments, it needs to initialize a builder to load cgo flags, and
+# that uses a temporary directory.
+! go env
+stderr '^go: creating work dir: \w+ '$GOTMPDIR
+go env GOROOT
+
+-- go.mod --
+module m
+
+go 1.15
 -- hello.go --
 package main
 func main() { println("hello") }
diff --git a/src/cmd/go/testdata/script/build_cache_disabled.txt b/src/cmd/go/testdata/script/build_cache_disabled.txt
new file mode 100644
index 0000000000..2e1327880b
--- /dev/null
+++ b/src/cmd/go/testdata/script/build_cache_disabled.txt
@@ -0,0 +1,46 @@
+# The build cache is required to build anything. It also may be needed to
+# initialize the build system, which is needed for commands like 'go env'.
+# However, there are lots of commands the cache is not needed for, and we
+# shouldn't require it when it won't be used.
+#
+# TODO(golang.org/issue/39882): commands below should work, too.
+# * go clean -modcache
+# * go env
+# * go fix
+# * go fmt
+# * go generate
+# * go get -d
+# * go list (without -export or -compiled)
+
+env GOCACHE=off
+
+# Commands that don't completely load packages should work.
+go doc fmt
+stdout Printf
+
+go fmt .
+
+! go tool compile -h
+stderr usage:
+
+go version
+stdout '^go version'
+
+
+# Module commands that don't load packages should work.
+go mod init m
+exists go.mod
+
+go mod edit -require rsc.io/quote@v1.5.2
+
+go mod download rsc.io/quote
+
+go mod graph
+stdout rsc.io/quote
+
+go mod verify
+
+-- main.go --
+package main
+
+func main() {}
diff --git a/src/cmd/go/testdata/script/list_case_collision.txt b/src/cmd/go/testdata/script/list_case_collision.txt
index 1b5f305587..73f44b63a0 100644
--- a/src/cmd/go/testdata/script/list_case_collision.txt
+++ b/src/cmd/go/testdata/script/list_case_collision.txt
@@ -6,23 +6,20 @@ stdout 'case-insensitive import collision'
 ! go build example/a
 stderr 'case-insensitive import collision'
 
-# If we're not guaranteed to have a case-sensitive file system, list files explicitly on command line.
-# Otherwise, let directory read find both files.
-[darwin] ! go list example/b/file.go example/b/FILE.go
-[windows] ! go list example/b/file.go example/b/FILE.go
-[!darwin] [!windows] ! go list example/b
+# List files explicitly on command line, to encounter case-checking
+# logic even on case-insensitive filesystems.
+cp example/b/file.go example/b/FILE.go  # no-op on case-insensitive filesystems
+! go list example/b/file.go example/b/FILE.go
 stderr 'case-insensitive file name collision'
 
+mkdir example/a/Pkg  # no-op on case-insensitive filesystems
+cp example/a/pkg/pkg.go example/a/Pkg/pkg.go  # no-op on case-insensitive filesystems
 ! go list example/a/pkg example/a/Pkg
-stderr 'case-insensitive import collision'
-go list -json -e example/a/pkg example/a/Pkg
-stdout 'case-insensitive import collision'
-! go build example/a/pkg example/a/Pkg
-stderr 'case-insensitive import collision'
 
 # Test that the path reported with an indirect import is correct.
-[!darwin] [!windows] ! go build example/c
-[!darwin] [!windows] stderr '^package example/c\n\timports example/b: case-insensitive file name collision: "FILE.go" and "file.go"$'
+cp example/b/file.go example/b/FILE.go
+[case-sensitive] ! go build example/c
+[case-sensitive] stderr '^package example/c\n\timports example/b: case-insensitive file name collision: "FILE.go" and "file.go"$'
 
 -- example/a/a.go --
 package p
@@ -32,12 +29,8 @@ import (
 )
 -- example/a/pkg/pkg.go --
 package pkg
--- example/a/Pkg/pkg.go --
-package pkg
 -- example/b/file.go --
 package b
--- example/b/FILE.go --
-package b
 -- example/c/c.go --
 package c
 
diff --git a/src/cmd/go/testdata/script/mod_proxy_invalid.txt b/src/cmd/go/testdata/script/mod_proxy_invalid.txt
new file mode 100644
index 0000000000..6427cc1527
--- /dev/null
+++ b/src/cmd/go/testdata/script/mod_proxy_invalid.txt
@@ -0,0 +1,8 @@
+env GO111MODULE=on
+env GOPROXY=$GOPROXY/invalid
+
+! go list -m rsc.io/quote@latest
+stderr '^go list -m: module rsc.io/quote: invalid response from proxy "'$GOPROXY'": invalid character ''i'' looking for beginning of value$'
+
+! go list -m rsc.io/quote@1.5.2
+stderr '^go list -m: rsc.io/quote@1.5.2: invalid version: invalid response from proxy "'$GOPROXY'": invalid character ''i'' looking for beginning of value$'
diff --git a/src/cmd/go/testdata/script/mod_query_empty.txt b/src/cmd/go/testdata/script/mod_query_empty.txt
index 4d8259b40f..b3ea3e3de0 100644
--- a/src/cmd/go/testdata/script/mod_query_empty.txt
+++ b/src/cmd/go/testdata/script/mod_query_empty.txt
@@ -40,7 +40,7 @@ env GOPROXY=file:///$WORK/gatekeeper
 chmod 0000 $WORK/gatekeeper/example.com/join/subpkg/@latest
 cp go.mod.orig go.mod
 ! go get -d example.com/join/subpkg
-stderr 'go get example.com/join/subpkg: module example.com/join/subpkg: (invalid character .+|reading file://.*/gatekeeper/example.com/join/subpkg/@latest: .+)'
+stderr 'go get example.com/join/subpkg: module example.com/join/subpkg: (invalid response from proxy ".+": invalid character .+|reading file://.*/gatekeeper/example.com/join/subpkg/@latest: .+)'
 
 -- go.mod.orig --
 module example.com/othermodule
diff --git a/src/cmd/go/testdata/script/mod_sum_lookup.txt b/src/cmd/go/testdata/script/mod_sum_lookup.txt
new file mode 100644
index 0000000000..ed80a44984
--- /dev/null
+++ b/src/cmd/go/testdata/script/mod_sum_lookup.txt
@@ -0,0 +1,33 @@
+# When we attempt to resolve an import that doesn't exist, we should not save
+# hashes for downloaded modules.
+# Verifies golang.org/issue/36260.
+go list -e -tags=ignore ./noexist
+! exists go.sum
+
+# When an import is resolved successfully, we should only save hashes for
+# the module that provides the package, not for other modules looked up.
+# Verifies golang.org/issue/31580.
+go list ./exist
+grep '^example.com/join v1.1.0 h1:' go.sum
+! grep '^example.com/join/subpkg' go.sum
+cp go.sum go.list.sum
+go mod tidy
+cmp go.sum go.list.sum
+
+-- go.mod --
+module m
+
+go 1.15
+
+-- noexist/use.go --
+// ignore tags prevents errors in 'go mod tidy'
+// +build ignore
+
+package use
+
+import _ "example.com/join/subpkg/noexist"
+
+-- exist/use.go --
+package use
+
+import _ "example.com/join/subpkg"
diff --git a/src/cmd/go/testdata/script/mod_tidy_old.txt b/src/cmd/go/testdata/script/mod_tidy_old.txt
new file mode 100644
index 0000000000..7428f0ce8a
--- /dev/null
+++ b/src/cmd/go/testdata/script/mod_tidy_old.txt
@@ -0,0 +1,46 @@
+# 'go mod tidy' should remove content sums for module versions that aren't
+# in the build list. It should preserve go.mod sums for module versions that
+# are in the module graph though.
+# Verifies golang.org/issue/33008.
+go mod tidy
+! grep '^rsc.io/quote v1.5.0 h1:' go.sum
+grep '^rsc.io/quote v1.5.0/go.mod h1:' go.sum
+
+-- go.mod --
+module m
+
+go 1.15
+
+require (
+	rsc.io/quote v1.5.2
+	example.com/r v0.0.0
+)
+
+replace example.com/r => ./r
+
+-- go.sum --
+golang.org/x/text v0.0.0-20170915032832-14c0d48ead0c h1:pvCbr/wm8HzDD3fVywevekufpn6tCGPY3spdHeZJEsw=
+golang.org/x/text v0.0.0-20170915032832-14c0d48ead0c/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
+rsc.io/quote v1.5.0 h1:6fJa6E+wGadANKkUMlZ0DhXFpoKlslOQDCo259XtdIE=
+rsc.io/quote v1.5.0/go.mod h1:LzX7hefJvL54yjefDEDHNONDjII0t9xZLPXsUe+TKr0=
+rsc.io/quote v1.5.2 h1:3fEykkD9k7lYzXqCYrwGAf7iNhbk4yCjHmKBN9td4L0=
+rsc.io/quote v1.5.2/go.mod h1:LzX7hefJvL54yjefDEDHNONDjII0t9xZLPXsUe+TKr0=
+rsc.io/sampler v1.3.0 h1:HLGR/BgEtI3r0uymSP/nl2uPLsUnNJX8toRyhfpBTII=
+rsc.io/sampler v1.3.0/go.mod h1:T1hPZKmBbMNahiBKFy5HrXp6adAjACjK9JXDnKaTXpA=
+rsc.io/testonly v1.0.0 h1:K/VWHdO+Jv7woUXG0GzVNx1czBXUt3Ib1deaMn+xk64=
+rsc.io/testonly v1.0.0/go.mod h1:OqmGbIFOcF+XrFReLOGZ6BhMM7uMBiQwZsyNmh74SzY=
+
+-- r/go.mod --
+module example.com/r
+
+require rsc.io/quote v1.5.0
+
+-- use.go --
+package use
+
+import _ "example.com/r"
+
+-- r/use.go --
+package use
+
+import _ "rsc.io/quote"
diff --git a/src/cmd/go/testdata/script/mod_verify.txt b/src/cmd/go/testdata/script/mod_verify.txt
index 646bc62bb7..3918400435 100644
--- a/src/cmd/go/testdata/script/mod_verify.txt
+++ b/src/cmd/go/testdata/script/mod_verify.txt
@@ -12,20 +12,18 @@ go mod verify
 ! exists $GOPATH/pkg/mod/cache/download/rsc.io/quote/@v/v1.1.0.zip
 
 # With bad go.sum, sync (which must download) fails.
-# Even if the bad sum is in the old legacy go.modverify file.
 rm go.sum
-cp go.sum.bad go.modverify
+cp go.sum.bad go.sum
 ! go mod tidy
 stderr 'checksum mismatch'
 ! exists $GOPATH/pkg/mod/cache/download/rsc.io/quote/@v/v1.1.0.zip
 
-# With good go.sum, sync works (and moves go.modverify to go.sum).
+# With good go.sum, sync works.
 rm go.sum
-cp go.sum.good go.modverify
+cp go.sum.good go.sum
 go mod tidy
 exists $GOPATH/pkg/mod/cache/download/rsc.io/quote/@v/v1.1.0.zip
 exists $GOPATH/pkg/mod/rsc.io/quote@v1.1.0/quote.go
-! exists go.modverify
 
 # go.sum should have the new checksum for go.mod
 grep '^rsc.io/quote v1.1.0/go.mod ' go.sum
diff --git a/src/cmd/go/testdata/script/test_flags.txt b/src/cmd/go/testdata/script/test_flags.txt
index d38e37f238..63385e6997 100644
--- a/src/cmd/go/testdata/script/test_flags.txt
+++ b/src/cmd/go/testdata/script/test_flags.txt
@@ -10,7 +10,7 @@ stdout '\Aok\s+example.com/x\s+[0-9.s]+\n\z'
 ! stderr .
 
 # For backward-compatibility with previous releases of the 'go' command,
-# arguments that appear after unrecognized flags should not be  treated
+# arguments that appear after unrecognized flags should not be treated
 # as packages, even if they are unambiguously not arguments to flags.
 # Even though ./x looks like a package path, the real package should be
 # the implicit '.'.
@@ -18,6 +18,22 @@ stdout '\Aok\s+example.com/x\s+[0-9.s]+\n\z'
 stderr '^no Go files in .+$'
 ! stderr '/x'
 
+# However, *flags* that appear after unrecognized flags should still be
+# interpreted as flags, under the (possibly-erroneous) assumption that
+# unrecognized flags are non-boolean.
+
+go test -v -x ./x -timeout 24h -boolflag=true foo -timeout 25h
+stdout 'args: foo -timeout 25h'
+stdout 'timeout: 24h0m0s$'  # -timeout is unambiguously not a flag, so the real flag wins.
+
+go test -v -x ./x -timeout 24h -boolflag foo -timeout 25h
+stdout 'args: foo -test\.timeout=25h0m0s'  # For legacy reasons, '-timeout ' is erroneously rewritten to -test.timeout; see https://golang.org/issue/40763.
+stdout 'timeout: 24h0m0s$'  # Actual flag wins.
+
+go test -v -x ./x -timeout 24h -stringflag foo -timeout 25h
+stdout 'args: $'
+stdout 'timeout: 25h0m0s$'  # Later flag wins.
+
 # An explicit '-outputdir=' argument should set test.outputdir
 # to the 'go' command's working directory, not zero it out
 # for the test binary.
@@ -30,23 +46,23 @@ exists ./cover.out
 # with the 'test.' prefix in the GOFLAGS entry...
 env GOFLAGS='-test.timeout=24h0m0s -count=1'
 go test -v -x ./x
-stdout '.*: 24h0m0s$'
+stdout 'timeout: 24h0m0s$'
 stderr '-test.count=1'
 
 # ...or without.
 env GOFLAGS='-timeout=24h0m0s -count=1'
 go test -v -x ./x
-stdout '.*: 24h0m0s$'
+stdout 'timeout: 24h0m0s$'
 stderr '-test.count=1'
 
 # Arguments from the command line should override GOFLAGS...
 go test -v -x -timeout=25h0m0s ./x
-stdout '.*: 25h0m0s$'
+stdout 'timeout: 25h0m0s$'
 stderr '-test.count=1'
 
 # ...even if they use a different flag name.
 go test -v -x -test.timeout=26h0m0s ./x
-stdout '.*: 26h0m0s$'
+stdout 'timeout: 26h0m0s$'
 stderr '-test\.timeout=26h0m0s'
 ! stderr 'timeout=24h0m0s'
 stderr '-test.count=1'
@@ -99,11 +115,18 @@ package x
 
 import (
 	"flag"
+	"strings"
 	"testing"
 )
 
 var _ = flag.String("usage_message", "", "dummy flag to check usage message")
+var boolflag = flag.Bool("boolflag", false, "ignored boolean flag")
+var stringflag = flag.String("stringflag", "", "ignored string flag")
 
 func TestLogTimeout(t *testing.T) {
-	t.Log(flag.Lookup("test.timeout").Value)
+	t.Logf("timeout: %v", flag.Lookup("test.timeout").Value)
+}
+
+func TestLogArgs(t *testing.T) {
+	t.Logf("args: %s", strings.Join(flag.Args(), " "))
 }
diff --git a/src/cmd/go/testdata/script/test_json_exit.txt b/src/cmd/go/testdata/script/test_json_exit.txt
new file mode 100644
index 0000000000..dc7ffb06cf
--- /dev/null
+++ b/src/cmd/go/testdata/script/test_json_exit.txt
@@ -0,0 +1,102 @@
+[short] skip
+
+go test -c -o mainpanic.exe ./mainpanic &
+go test -c -o mainexit0.exe ./mainexit0 &
+go test -c -o testpanic.exe ./testpanic &
+go test -c -o testbgpanic.exe ./testbgpanic &
+wait
+
+# Test binaries that panic in TestMain should be marked as failing.
+
+! go test -json ./mainpanic
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+! go tool test2json ./mainpanic.exe
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+# Test binaries that exit with status 0 should be marked as passing.
+
+go test -json ./mainexit0
+stdout '"Action":"pass"'
+! stdout '"Action":"fail"'
+
+go tool test2json ./mainexit0.exe
+stdout '"Action":"pass"'
+! stdout '"Action":"fail"'
+
+# Test functions that panic should never be marked as passing
+# (https://golang.org/issue/40132).
+
+! go test -json ./testpanic
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+! go tool test2json ./testpanic.exe -test.v
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+! go tool test2json ./testpanic.exe
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+# Tests that panic in a background goroutine should be marked as failing.
+
+! go test -json ./testbgpanic
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+! go tool test2json ./testbgpanic.exe -test.v
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+! go tool test2json ./testbgpanic.exe
+stdout '"Action":"fail"'
+! stdout '"Action":"pass"'
+
+-- go.mod --
+module m
+go 1.14
+-- mainpanic/mainpanic_test.go --
+package mainpanic_test
+
+import "testing"
+
+func TestMain(m *testing.M) {
+	panic("haha no")
+}
+-- mainexit0/mainexit0_test.go --
+package mainexit0_test
+
+import (
+	"fmt"
+	"os"
+	"testing"
+)
+
+func TestMain(m *testing.M) {
+	fmt.Println("nothing to do")
+	os.Exit(0)
+}
+-- testpanic/testpanic_test.go --
+package testpanic_test
+
+import "testing"
+
+func TestPanic(*testing.T) {
+	panic("haha no")
+}
+-- testbgpanic/testbgpanic_test.go --
+package testbgpanic_test
+
+import "testing"
+
+func TestPanicInBackground(*testing.T) {
+	c := make(chan struct{})
+	go func() {
+		panic("haha no")
+		close(c)
+	}()
+	<-c
+}
diff --git a/src/cmd/go/testdata/script/test_json_interleaved.txt b/src/cmd/go/testdata/script/test_json_interleaved.txt
new file mode 100644
index 0000000000..e2d349e3fb
--- /dev/null
+++ b/src/cmd/go/testdata/script/test_json_interleaved.txt
@@ -0,0 +1,27 @@
+# Regression test for https://golang.org/issue/40657: output from the main test
+# function should be attributed correctly even if interleaved with the PAUSE
+# line for a new parallel subtest.
+
+[short] skip
+
+go test -json
+stdout '"Test":"TestWeirdTiming","Output":"[^"]* logging to outer again\\n"'
+
+-- go.mod --
+module example.com
+go 1.15
+-- main_test.go --
+package main
+
+import (
+	"testing"
+)
+
+func TestWeirdTiming(outer *testing.T) {
+	outer.Run("pauser", func(pauser *testing.T) {
+		outer.Logf("logging to outer")
+		pauser.Parallel()
+	})
+
+	outer.Logf("logging to outer again")
+}
diff --git a/src/cmd/internal/archive/archive.go b/src/cmd/internal/archive/archive.go
index db67ce424b..c1661d7711 100644
--- a/src/cmd/internal/archive/archive.go
+++ b/src/cmd/internal/archive/archive.go
@@ -17,6 +17,7 @@ import (
 	"log"
 	"os"
 	"strconv"
+	"strings"
 	"time"
 	"unicode/utf8"
 )
@@ -83,6 +84,7 @@ func (e *Entry) String() string {
 
 type GoObj struct {
 	TextHeader []byte
+	Arch       string
 	Data
 }
 
@@ -404,6 +406,10 @@ func (r *objReader) parseObject(o *GoObj, size int64) error {
 		}
 	}
 	o.TextHeader = h
+	hs := strings.Fields(string(h))
+	if len(hs) >= 4 {
+		o.Arch = hs[3]
+	}
 	o.Offset = r.offset
 	o.Size = size - int64(len(h))
 
diff --git a/src/cmd/internal/obj/arm/asm5.go b/src/cmd/internal/obj/arm/asm5.go
index f66f8aaf84..7b7e42ee2e 100644
--- a/src/cmd/internal/obj/arm/asm5.go
+++ b/src/cmd/internal/obj/arm/asm5.go
@@ -327,6 +327,9 @@ var optab = []Optab{
 	{obj.APCDATA, C_LCON, C_NONE, C_LCON, 0, 0, 0, 0, 0, 0},
 	{obj.AFUNCDATA, C_LCON, C_NONE, C_ADDR, 0, 0, 0, 0, 0, 0},
 	{obj.ANOP, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0, 0, 0},
+	{obj.ANOP, C_LCON, C_NONE, C_NONE, 0, 0, 0, 0, 0, 0}, // nop variants, see #40689
+	{obj.ANOP, C_REG, C_NONE, C_NONE, 0, 0, 0, 0, 0, 0},
+	{obj.ANOP, C_FREG, C_NONE, C_NONE, 0, 0, 0, 0, 0, 0},
 	{obj.ADUFFZERO, C_NONE, C_NONE, C_SBRA, 5, 4, 0, 0, 0, 0}, // same as ABL
 	{obj.ADUFFCOPY, C_NONE, C_NONE, C_SBRA, 5, 4, 0, 0, 0, 0}, // same as ABL
 	{obj.AXXX, C_NONE, C_NONE, C_NONE, 0, 4, 0, 0, 0, 0},
diff --git a/src/cmd/internal/obj/arm/obj5.go b/src/cmd/internal/obj/arm/obj5.go
index 008118c47b..86831f2b44 100644
--- a/src/cmd/internal/obj/arm/obj5.go
+++ b/src/cmd/internal/obj/arm/obj5.go
@@ -276,67 +276,21 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 
 	/*
 	 * find leaf subroutines
-	 * strip NOPs
-	 * expand RET
-	 * expand BECOME pseudo
 	 */
-	var q1 *obj.Prog
-	var q *obj.Prog
 	for p := cursym.Func.Text; p != nil; p = p.Link {
 		switch p.As {
 		case obj.ATEXT:
 			p.Mark |= LEAF
 
-		case obj.ARET:
-			break
-
 		case ADIV, ADIVU, AMOD, AMODU:
-			q = p
 			cursym.Func.Text.Mark &^= LEAF
-			continue
-
-		case obj.ANOP:
-			q1 = p.Link
-			q.Link = q1 /* q is non-nop */
-			if q1 != nil {
-				q1.Mark |= p.Mark
-			}
-			continue
 
 		case ABL,
 			ABX,
 			obj.ADUFFZERO,
 			obj.ADUFFCOPY:
 			cursym.Func.Text.Mark &^= LEAF
-			fallthrough
-
-		case AB,
-			ABEQ,
-			ABNE,
-			ABCS,
-			ABHS,
-			ABCC,
-			ABLO,
-			ABMI,
-			ABPL,
-			ABVS,
-			ABVC,
-			ABHI,
-			ABLS,
-			ABGE,
-			ABLT,
-			ABGT,
-			ABLE:
-			q1 = p.Pcond
-			if q1 != nil {
-				for q1.As == obj.ANOP {
-					q1 = q1.Link
-					p.Pcond = q1
-				}
-			}
 		}
-
-		q = p
 	}
 
 	var q2 *obj.Prog
diff --git a/src/cmd/internal/obj/arm64/a.out.go b/src/cmd/internal/obj/arm64/a.out.go
index 152c493a65..03e0278a33 100644
--- a/src/cmd/internal/obj/arm64/a.out.go
+++ b/src/cmd/internal/obj/arm64/a.out.go
@@ -946,6 +946,10 @@ const (
 	ASHA256H2
 	ASHA256SU0
 	ASHA256SU1
+	ASHA512H
+	ASHA512H2
+	ASHA512SU0
+	ASHA512SU1
 	AVADD
 	AVADDP
 	AVAND
diff --git a/src/cmd/internal/obj/arm64/anames.go b/src/cmd/internal/obj/arm64/anames.go
index 565f70aaf9..65ecd007ea 100644
--- a/src/cmd/internal/obj/arm64/anames.go
+++ b/src/cmd/internal/obj/arm64/anames.go
@@ -453,6 +453,10 @@ var Anames = []string{
 	"SHA256H2",
 	"SHA256SU0",
 	"SHA256SU1",
+	"SHA512H",
+	"SHA512H2",
+	"SHA512SU0",
+	"SHA512SU1",
 	"VADD",
 	"VADDP",
 	"VAND",
diff --git a/src/cmd/internal/obj/arm64/asm7.go b/src/cmd/internal/obj/arm64/asm7.go
index df17729a76..7a5a8ff38c 100644
--- a/src/cmd/internal/obj/arm64/asm7.go
+++ b/src/cmd/internal/obj/arm64/asm7.go
@@ -837,6 +837,9 @@ var optab = []Optab{
 	{obj.APCDATA, C_VCON, C_NONE, C_NONE, C_VCON, 0, 0, 0, 0, 0},
 	{obj.AFUNCDATA, C_VCON, C_NONE, C_NONE, C_ADDR, 0, 0, 0, 0, 0},
 	{obj.ANOP, C_NONE, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0, 0},
+	{obj.ANOP, C_LCON, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0, 0}, // nop variants, see #40689
+	{obj.ANOP, C_REG, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0, 0},
+	{obj.ANOP, C_VREG, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0, 0},
 	{obj.ADUFFZERO, C_NONE, C_NONE, C_NONE, C_SBRA, 5, 4, 0, 0, 0}, // same as AB/ABL
 	{obj.ADUFFCOPY, C_NONE, C_NONE, C_NONE, C_SBRA, 5, 4, 0, 0, 0}, // same as AB/ABL
 	{obj.APCALIGN, C_LCON, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0, 0},  // align code
@@ -2747,6 +2750,7 @@ func buildop(ctxt *obj.Link) {
 			oprangeset(AAESIMC, t)
 			oprangeset(ASHA1SU1, t)
 			oprangeset(ASHA256SU0, t)
+			oprangeset(ASHA512SU0, t)
 
 		case ASHA1C:
 			oprangeset(ASHA1P, t)
@@ -2754,9 +2758,12 @@ func buildop(ctxt *obj.Link) {
 
 		case ASHA256H:
 			oprangeset(ASHA256H2, t)
+			oprangeset(ASHA512H, t)
+			oprangeset(ASHA512H2, t)
 
 		case ASHA1SU0:
 			oprangeset(ASHA256SU1, t)
+			oprangeset(ASHA512SU1, t)
 
 		case AVADDV:
 			oprangeset(AVUADDLV, t)
@@ -5391,6 +5398,18 @@ func (c *ctxt7) oprrr(p *obj.Prog, a obj.As) uint32 {
 	case ASHA256SU0:
 		return 0x5E<<24 | 2<<20 | 8<<16 | 2<<12 | 2<<10
 
+	case ASHA512H:
+		return 0xCE<<24 | 3<<21 | 8<<12
+
+	case ASHA512H2:
+		return 0xCE<<24 | 3<<21 | 8<<12 | 4<<8
+
+	case ASHA512SU1:
+		return 0xCE<<24 | 3<<21 | 8<<12 | 8<<8
+
+	case ASHA512SU0:
+		return 0xCE<<24 | 3<<22 | 8<<12
+
 	case AFCVTZSD:
 		return FPCVTI(1, 0, 1, 3, 0)
 
diff --git a/src/cmd/internal/obj/arm64/obj7.go b/src/cmd/internal/obj/arm64/obj7.go
index b046685ada..0d74430053 100644
--- a/src/cmd/internal/obj/arm64/obj7.go
+++ b/src/cmd/internal/obj/arm64/obj7.go
@@ -468,73 +468,21 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 
 	/*
 	 * find leaf subroutines
-	 * strip NOPs
-	 * expand RET
 	 */
-	q := (*obj.Prog)(nil)
-	var q1 *obj.Prog
 	for p := c.cursym.Func.Text; p != nil; p = p.Link {
 		switch p.As {
 		case obj.ATEXT:
 			p.Mark |= LEAF
 
-		case obj.ARET:
-			break
-
-		case obj.ANOP:
-			if p.Link != nil {
-				q1 = p.Link
-				q.Link = q1 /* q is non-nop */
-				q1.Mark |= p.Mark
-			}
-			continue
-
 		case ABL,
 			obj.ADUFFZERO,
 			obj.ADUFFCOPY:
 			c.cursym.Func.Text.Mark &^= LEAF
-			fallthrough
-
-		case ACBNZ,
-			ACBZ,
-			ACBNZW,
-			ACBZW,
-			ATBZ,
-			ATBNZ,
-			AB,
-			ABEQ,
-			ABNE,
-			ABCS,
-			ABHS,
-			ABCC,
-			ABLO,
-			ABMI,
-			ABPL,
-			ABVS,
-			ABVC,
-			ABHI,
-			ABLS,
-			ABGE,
-			ABLT,
-			ABGT,
-			ABLE,
-			AADR, /* strange */
-			AADRP:
-			q1 = p.Pcond
-
-			if q1 != nil {
-				for q1.As == obj.ANOP {
-					q1 = q1.Link
-					p.Pcond = q1
-				}
-			}
-
-			break
 		}
-
-		q = p
 	}
 
+	var q *obj.Prog
+	var q1 *obj.Prog
 	var retjmp *obj.LSym
 	for p := c.cursym.Func.Text; p != nil; p = p.Link {
 		o := p.As
diff --git a/src/cmd/internal/obj/link.go b/src/cmd/internal/obj/link.go
index 11fab63065..2660a564db 100644
--- a/src/cmd/internal/obj/link.go
+++ b/src/cmd/internal/obj/link.go
@@ -480,7 +480,6 @@ const (
 	AttrWrapper
 	AttrNeedCtxt
 	AttrNoFrame
-	AttrSeenGlobl
 	AttrOnList
 	AttrStatic
 
@@ -537,7 +536,6 @@ func (a Attribute) MakeTypelink() bool       { return a&AttrMakeTypelink != 0 }
 func (a Attribute) CFunc() bool              { return a&AttrCFunc != 0 }
 func (a Attribute) NoSplit() bool            { return a&AttrNoSplit != 0 }
 func (a Attribute) Leaf() bool               { return a&AttrLeaf != 0 }
-func (a Attribute) SeenGlobl() bool          { return a&AttrSeenGlobl != 0 }
 func (a Attribute) OnList() bool             { return a&AttrOnList != 0 }
 func (a Attribute) ReflectMethod() bool      { return a&AttrReflectMethod != 0 }
 func (a Attribute) Local() bool              { return a&AttrLocal != 0 }
@@ -574,7 +572,6 @@ var textAttrStrings = [...]struct {
 	{bit: AttrCFunc, s: "CFUNC"},
 	{bit: AttrNoSplit, s: "NOSPLIT"},
 	{bit: AttrLeaf, s: "LEAF"},
-	{bit: AttrSeenGlobl, s: ""},
 	{bit: AttrOnList, s: ""},
 	{bit: AttrReflectMethod, s: "REFLECTMETHOD"},
 	{bit: AttrLocal, s: "LOCAL"},
diff --git a/src/cmd/internal/obj/mips/asm0.go b/src/cmd/internal/obj/mips/asm0.go
index faa12bf133..faa827da9f 100644
--- a/src/cmd/internal/obj/mips/asm0.go
+++ b/src/cmd/internal/obj/mips/asm0.go
@@ -391,6 +391,9 @@ var optab = []Optab{
 	{obj.APCDATA, C_LCON, C_NONE, C_LCON, 0, 0, 0, 0, 0},
 	{obj.AFUNCDATA, C_SCON, C_NONE, C_ADDR, 0, 0, 0, 0, 0},
 	{obj.ANOP, C_NONE, C_NONE, C_NONE, 0, 0, 0, 0, 0},
+	{obj.ANOP, C_LCON, C_NONE, C_NONE, 0, 0, 0, 0, 0}, // nop variants, see #40689
+	{obj.ANOP, C_REG, C_NONE, C_NONE, 0, 0, 0, 0, 0},
+	{obj.ANOP, C_FREG, C_NONE, C_NONE, 0, 0, 0, 0, 0},
 	{obj.ADUFFZERO, C_NONE, C_NONE, C_LBRA, 11, 4, 0, 0, 0}, // same as AJMP
 	{obj.ADUFFCOPY, C_NONE, C_NONE, C_LBRA, 11, 4, 0, 0, 0}, // same as AJMP
 
diff --git a/src/cmd/internal/obj/mips/obj0.go b/src/cmd/internal/obj/mips/obj0.go
index 3106143844..77cad979a6 100644
--- a/src/cmd/internal/obj/mips/obj0.go
+++ b/src/cmd/internal/obj/mips/obj0.go
@@ -158,19 +158,14 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 
 	/*
 	 * find leaf subroutines
-	 * strip NOPs
 	 * expand RET
 	 * expand BECOME pseudo
 	 */
 
-	var q *obj.Prog
-	var q1 *obj.Prog
 	for p := c.cursym.Func.Text; p != nil; p = p.Link {
 		switch p.As {
 		/* too hard, just leave alone */
 		case obj.ATEXT:
-			q = p
-
 			p.Mark |= LABEL | LEAF | SYNC
 			if p.Link != nil {
 				p.Link.Mark |= LABEL
@@ -179,7 +174,6 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 		/* too hard, just leave alone */
 		case AMOVW,
 			AMOVV:
-			q = p
 			if p.To.Type == obj.TYPE_REG && p.To.Reg >= REG_SPECIAL {
 				p.Mark |= LABEL | SYNC
 				break
@@ -195,11 +189,9 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 			ATLBWI,
 			ATLBP,
 			ATLBR:
-			q = p
 			p.Mark |= LABEL | SYNC
 
 		case ANOR:
-			q = p
 			if p.To.Type == obj.TYPE_REG {
 				if p.To.Reg == REGZERO {
 					p.Mark |= LABEL | SYNC
@@ -235,8 +227,7 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 			} else {
 				p.Mark |= BRANCH
 			}
-			q = p
-			q1 = p.Pcond
+			q1 := p.Pcond
 			if q1 != nil {
 				for q1.As == obj.ANOP {
 					q1 = q1.Link
@@ -254,24 +245,11 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 			if q1 != nil {
 				q1.Mark |= LABEL
 			}
-			continue
 
 		case ARET:
-			q = p
 			if p.Link != nil {
 				p.Link.Mark |= LABEL
 			}
-			continue
-
-		case obj.ANOP:
-			q1 = p.Link
-			q.Link = q1 /* q is non-nop */
-			q1.Mark |= p.Mark
-			continue
-
-		default:
-			q = p
-			continue
 		}
 	}
 
@@ -284,6 +262,8 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 		mov = AMOVW
 	}
 
+	var q *obj.Prog
+	var q1 *obj.Prog
 	autosize := int32(0)
 	var p1 *obj.Prog
 	var p2 *obj.Prog
diff --git a/src/cmd/internal/obj/pcln.go b/src/cmd/internal/obj/pcln.go
index 7750637796..ce0d3714c0 100644
--- a/src/cmd/internal/obj/pcln.go
+++ b/src/cmd/internal/obj/pcln.go
@@ -286,6 +286,21 @@ func linkpcln(ctxt *Link, cursym *LSym) {
 	pcln.Pcfile = funcpctab(ctxt, cursym, "pctofile", pctofileline, pcln)
 	pcln.Pcline = funcpctab(ctxt, cursym, "pctoline", pctofileline, nil)
 
+	// Check that all the Progs used as inline markers are still reachable.
+	// See issue #40473.
+	inlMarkProgs := make(map[*Prog]struct{}, len(cursym.Func.InlMarks))
+	for _, inlMark := range cursym.Func.InlMarks {
+		inlMarkProgs[inlMark.p] = struct{}{}
+	}
+	for p := cursym.Func.Text; p != nil; p = p.Link {
+		if _, ok := inlMarkProgs[p]; ok {
+			delete(inlMarkProgs, p)
+		}
+	}
+	if len(inlMarkProgs) > 0 {
+		ctxt.Diag("one or more instructions used as inline markers are no longer reachable")
+	}
+
 	pcinlineState := new(pcinlineState)
 	pcln.Pcinline = funcpctab(ctxt, cursym, "pctoinline", pcinlineState.pctoinline, nil)
 	for _, inlMark := range cursym.Func.InlMarks {
diff --git a/src/cmd/internal/obj/plist.go b/src/cmd/internal/obj/plist.go
index afe0ee4ee0..6e33f29959 100644
--- a/src/cmd/internal/obj/plist.go
+++ b/src/cmd/internal/obj/plist.go
@@ -145,10 +145,6 @@ func (ctxt *Link) InitTextSym(s *LSym, flag int) {
 }
 
 func (ctxt *Link) Globl(s *LSym, size int64, flag int) {
-	if s.SeenGlobl() {
-		fmt.Printf("duplicate %v\n", s)
-	}
-	s.Set(AttrSeenGlobl, true)
 	if s.OnList() {
 		ctxt.Diag("symbol %s listed multiple times", s.Name)
 	}
diff --git a/src/cmd/internal/obj/ppc64/asm9.go b/src/cmd/internal/obj/ppc64/asm9.go
index 0fd0744a42..3c82477fc4 100644
--- a/src/cmd/internal/obj/ppc64/asm9.go
+++ b/src/cmd/internal/obj/ppc64/asm9.go
@@ -613,6 +613,9 @@ var optab = []Optab{
 	{obj.APCDATA, C_LCON, C_NONE, C_NONE, C_LCON, 0, 0, 0},
 	{obj.AFUNCDATA, C_SCON, C_NONE, C_NONE, C_ADDR, 0, 0, 0},
 	{obj.ANOP, C_NONE, C_NONE, C_NONE, C_NONE, 0, 0, 0},
+	{obj.ANOP, C_LCON, C_NONE, C_NONE, C_NONE, 0, 0, 0}, // NOP operand variations added for #40689
+	{obj.ANOP, C_REG, C_NONE, C_NONE, C_NONE, 0, 0, 0},  // to preserve previous behavior
+	{obj.ANOP, C_FREG, C_NONE, C_NONE, C_NONE, 0, 0, 0},
 	{obj.ADUFFZERO, C_NONE, C_NONE, C_NONE, C_LBRA, 11, 4, 0}, // same as ABR/ABL
 	{obj.ADUFFCOPY, C_NONE, C_NONE, C_NONE, C_LBRA, 11, 4, 0}, // same as ABR/ABL
 	{obj.APCALIGN, C_LCON, C_NONE, C_NONE, C_NONE, 0, 0, 0},   // align code
@@ -1581,8 +1584,9 @@ func buildop(ctxt *obj.Link) {
 		case ALXV: /* lxv */
 			opset(ALXV, r0)
 
-		case ALXVL: /* lxvl */
+		case ALXVL: /* lxvl, lxvll, lxvx */
 			opset(ALXVLL, r0)
+			opset(ALXVX, r0)
 
 		case ASTXVD2X: /* stxvd2x, stxvdsx, stxvw4x, stxvh8x, stxvb16x */
 			opset(ASTXVW4X, r0)
@@ -1592,8 +1596,9 @@ func buildop(ctxt *obj.Link) {
 		case ASTXV: /* stxv */
 			opset(ASTXV, r0)
 
-		case ASTXVL: /* stxvl, stxvll */
+		case ASTXVL: /* stxvl, stxvll, stvx */
 			opset(ASTXVLL, r0)
+			opset(ASTXVX, r0)
 
 		case ALXSDX: /* lxsdx  */
 			opset(ALXSDX, r0)
@@ -5017,11 +5022,13 @@ func (c *ctxt9) opload(a obj.As) uint32 {
 	case AMOVW:
 		return OPVCC(58, 0, 0, 0) | 1<<1 /* lwa */
 	case ALXV:
-		return OPDQ(61, 1, 0) /* lxv - ISA v3.00 */
+		return OPDQ(61, 1, 0) /* lxv - ISA v3.0 */
 	case ALXVL:
-		return OPVXX1(31, 269, 0) /* lxvl - ISA v3.00 */
+		return OPVXX1(31, 269, 0) /* lxvl - ISA v3.0 */
 	case ALXVLL:
-		return OPVXX1(31, 301, 0) /* lxvll - ISA v3.00 */
+		return OPVXX1(31, 301, 0) /* lxvll - ISA v3.0 */
+	case ALXVX:
+		return OPVXX1(31, 268, 0) /* lxvx - ISA v3.0 */
 
 		/* no AMOVWU */
 	case AMOVB, AMOVBZ:
@@ -5119,8 +5126,6 @@ func (c *ctxt9) oploadx(a obj.As) uint32 {
 		return OPVCC(31, 309, 0, 0) /* ldmx */
 
 	/* Vector (VMX/Altivec) instructions */
-	/* ISA 2.03 enables these for PPC970. For POWERx processors, these */
-	/* are enabled starting at POWER6 (ISA 2.05). */
 	case ALVEBX:
 		return OPVCC(31, 7, 0, 0) /* lvebx - v2.03 */
 	case ALVEHX:
@@ -5138,7 +5143,8 @@ func (c *ctxt9) oploadx(a obj.As) uint32 {
 		/* End of vector instructions */
 
 	/* Vector scalar (VSX) instructions */
-	/* ISA 2.06 enables these for POWER7. */
+	case ALXVX:
+		return OPVXX1(31, 268, 0) /* lxvx - ISA v3.0 */
 	case ALXVD2X:
 		return OPVXX1(31, 844, 0) /* lxvd2x - v2.06 */
 	case ALXVW4X:
@@ -5205,6 +5211,8 @@ func (c *ctxt9) opstore(a obj.As) uint32 {
 		return OPVXX1(31, 397, 0) /* stxvl ISA 3.0 */
 	case ASTXVLL:
 		return OPVXX1(31, 429, 0) /* stxvll ISA 3.0 */
+	case ASTXVX:
+		return OPVXX1(31, 396, 0) /* stxvx - ISA v3.0 */
 
 	}
 
@@ -5268,8 +5276,6 @@ func (c *ctxt9) opstorex(a obj.As) uint32 {
 		return OPVCC(31, 181, 0, 0) /* stdux */
 
 	/* Vector (VMX/Altivec) instructions */
-	/* ISA 2.03 enables these for PPC970. For POWERx processors, these */
-	/* are enabled starting at POWER6 (ISA 2.05). */
 	case ASTVEBX:
 		return OPVCC(31, 135, 0, 0) /* stvebx - v2.03 */
 	case ASTVEHX:
@@ -5283,15 +5289,16 @@ func (c *ctxt9) opstorex(a obj.As) uint32 {
 		/* End of vector instructions */
 
 	/* Vector scalar (VSX) instructions */
-	/* ISA 2.06 enables these for POWER7. */
+	case ASTXVX:
+		return OPVXX1(31, 396, 0) /* stxvx - v3.0 */
 	case ASTXVD2X:
 		return OPVXX1(31, 972, 0) /* stxvd2x - v2.06 */
 	case ASTXVW4X:
 		return OPVXX1(31, 908, 0) /* stxvw4x - v2.06 */
 	case ASTXVH8X:
-		return OPVXX1(31, 940, 0) /* stxvh8x - v3.00 */
+		return OPVXX1(31, 940, 0) /* stxvh8x - v3.0 */
 	case ASTXVB16X:
-		return OPVXX1(31, 1004, 0) /* stxvb16x - v3.00 */
+		return OPVXX1(31, 1004, 0) /* stxvb16x - v3.0 */
 
 	case ASTXSDX:
 		return OPVXX1(31, 716, 0) /* stxsdx - v2.06 */
diff --git a/src/cmd/internal/obj/ppc64/obj9.go b/src/cmd/internal/obj/ppc64/obj9.go
index 16881c634b..749f7066de 100644
--- a/src/cmd/internal/obj/ppc64/obj9.go
+++ b/src/cmd/internal/obj/ppc64/obj9.go
@@ -429,7 +429,6 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 
 	/*
 	 * find leaf subroutines
-	 * strip NOPs
 	 * expand RET
 	 * expand BECOME pseudo
 	 */
@@ -559,10 +558,7 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 			q = p
 			q1 = p.Pcond
 			if q1 != nil {
-				for q1.As == obj.ANOP {
-					q1 = q1.Link
-					p.Pcond = q1
-				}
+				// NOPs are not removed due to #40689.
 
 				if q1.Mark&LEAF == 0 {
 					q1.Mark |= LABEL
@@ -589,9 +585,8 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 			continue
 
 		case obj.ANOP:
-			q1 = p.Link
-			q.Link = q1 /* q is non-nop */
-			q1.Mark |= p.Mark
+			// NOPs are not removed due to
+			// #40689
 			continue
 
 		default:
diff --git a/src/cmd/internal/obj/s390x/objz.go b/src/cmd/internal/obj/s390x/objz.go
index b14dc810fa..ef6335d849 100644
--- a/src/cmd/internal/obj/s390x/objz.go
+++ b/src/cmd/internal/obj/s390x/objz.go
@@ -283,17 +283,6 @@ func preprocess(ctxt *obj.Link, cursym *obj.LSym, newprog obj.ProgAlloc) {
 			ACMPUBNE:
 			q = p
 			p.Mark |= BRANCH
-			if p.Pcond != nil {
-				q := p.Pcond
-				for q.As == obj.ANOP {
-					q = q.Link
-					p.Pcond = q
-				}
-			}
-
-		case obj.ANOP:
-			q.Link = p.Link /* q is non-nop */
-			p.Link.Mark |= p.Mark
 
 		default:
 			q = p
diff --git a/src/cmd/internal/objfile/goobj.go b/src/cmd/internal/objfile/goobj.go
index 8eecebb1df..7f74a8256c 100644
--- a/src/cmd/internal/objfile/goobj.go
+++ b/src/cmd/internal/objfile/goobj.go
@@ -17,13 +17,13 @@ import (
 	"fmt"
 	"io"
 	"os"
-	"strings"
 )
 
 type goobjFile struct {
 	goobj *archive.GoObj
 	r     *goobj.Reader
 	f     *os.File
+	arch  *sys.Arch
 }
 
 func openGoFile(f *os.File) (*File, error) {
@@ -45,9 +45,16 @@ L:
 				return nil, err
 			}
 			r := goobj.NewReaderFromBytes(b, false)
+			var arch *sys.Arch
+			for _, a := range sys.Archs {
+				if a.Name == e.Obj.Arch {
+					arch = a
+					break
+				}
+			}
 			entries = append(entries, &Entry{
 				name: e.Name,
-				raw:  &goobjFile{e.Obj, r, f},
+				raw:  &goobjFile{e.Obj, r, f, arch},
 			})
 			continue
 		case archive.EntryNativeObj:
@@ -223,17 +230,8 @@ func (f *goobjFile) pcln() (textStart uint64, symtab, pclntab []byte, err error)
 // Returns "",0,nil if unknown.
 // This function implements the Liner interface in preference to pcln() above.
 func (f *goobjFile) PCToLine(pc uint64) (string, int, *gosym.Func) {
-	// TODO: this is really inefficient. Binary search? Memoize last result?
 	r := f.r
-	var arch *sys.Arch
-	archname := f.goarch()
-	for _, a := range sys.Archs {
-		if a.Name == archname {
-			arch = a
-			break
-		}
-	}
-	if arch == nil {
+	if f.arch == nil {
 		return "", 0, nil
 	}
 	getSymData := func(s goobj.SymRef) []byte {
@@ -271,9 +269,9 @@ func (f *goobjFile) PCToLine(pc uint64) (string, int, *gosym.Func) {
 		var info *goobj.FuncInfo
 		lengths := info.ReadFuncInfoLengths(b)
 		pcline := getSymData(info.ReadPcline(b))
-		line := int(pcValue(pcline, pc-addr, arch))
+		line := int(pcValue(pcline, pc-addr, f.arch))
 		pcfile := getSymData(info.ReadPcfile(b))
-		fileID := pcValue(pcfile, pc-addr, arch)
+		fileID := pcValue(pcfile, pc-addr, f.arch)
 		globalFileID := info.ReadFile(b, lengths.FileOff, uint32(fileID))
 		fileName := r.File(int(globalFileID))
 		// Note: we provide only the name in the Func structure.
@@ -338,11 +336,7 @@ func (f *goobjFile) text() (textStart uint64, text []byte, err error) {
 }
 
 func (f *goobjFile) goarch() string {
-	hs := strings.Fields(string(f.goobj.TextHeader))
-	if len(hs) >= 4 {
-		return hs[3]
-	}
-	return ""
+	return f.goobj.Arch
 }
 
 func (f *goobjFile) loadAddress() (uint64, error) {
diff --git a/src/cmd/internal/test2json/test2json.go b/src/cmd/internal/test2json/test2json.go
index a01a8900e8..4eb6dd4838 100644
--- a/src/cmd/internal/test2json/test2json.go
+++ b/src/cmd/internal/test2json/test2json.go
@@ -45,10 +45,10 @@ type textBytes []byte
 
 func (b textBytes) MarshalText() ([]byte, error) { return b, nil }
 
-// A converter holds the state of a test-to-JSON conversion.
+// A Converter holds the state of a test-to-JSON conversion.
 // It implements io.WriteCloser; the caller writes test output in,
 // and the converter writes JSON output to w.
-type converter struct {
+type Converter struct {
 	w        io.Writer  // JSON output stream
 	pkg      string     // package to name in events
 	mode     Mode       // mode bits
@@ -100,9 +100,9 @@ var (
 //
 // The pkg string, if present, specifies the import path to
 // report in the JSON stream.
-func NewConverter(w io.Writer, pkg string, mode Mode) io.WriteCloser {
-	c := new(converter)
-	*c = converter{
+func NewConverter(w io.Writer, pkg string, mode Mode) *Converter {
+	c := new(Converter)
+	*c = Converter{
 		w:     w,
 		pkg:   pkg,
 		mode:  mode,
@@ -122,11 +122,20 @@ func NewConverter(w io.Writer, pkg string, mode Mode) io.WriteCloser {
 }
 
 // Write writes the test input to the converter.
-func (c *converter) Write(b []byte) (int, error) {
+func (c *Converter) Write(b []byte) (int, error) {
 	c.input.write(b)
 	return len(b), nil
 }
 
+// Exited marks the test process as having exited with the given error.
+func (c *Converter) Exited(err error) {
+	if err == nil {
+		c.result = "pass"
+	} else {
+		c.result = "fail"
+	}
+}
+
 var (
 	// printed by test on successful run.
 	bigPass = []byte("PASS\n")
@@ -160,7 +169,7 @@ var (
 // handleInputLine handles a single whole test output line.
 // It must write the line to c.output but may choose to do so
 // before or after emitting other events.
-func (c *converter) handleInputLine(line []byte) {
+func (c *Converter) handleInputLine(line []byte) {
 	// Final PASS or FAIL.
 	if bytes.Equal(line, bigPass) || bytes.Equal(line, bigFail) || bytes.HasPrefix(line, bigFailErrorPrefix) {
 		c.flushReport(0)
@@ -286,7 +295,7 @@ func (c *converter) handleInputLine(line []byte) {
 }
 
 // flushReport flushes all pending PASS/FAIL reports at levels >= depth.
-func (c *converter) flushReport(depth int) {
+func (c *Converter) flushReport(depth int) {
 	c.testName = ""
 	for len(c.report) > depth {
 		e := c.report[len(c.report)-1]
@@ -298,23 +307,22 @@ func (c *converter) flushReport(depth int) {
 // Close marks the end of the go test output.
 // It flushes any pending input and then output (only partial lines at this point)
 // and then emits the final overall package-level pass/fail event.
-func (c *converter) Close() error {
+func (c *Converter) Close() error {
 	c.input.flush()
 	c.output.flush()
-	e := &event{Action: "pass"}
 	if c.result != "" {
-		e.Action = c.result
-	}
-	if c.mode&Timestamp != 0 {
-		dt := time.Since(c.start).Round(1 * time.Millisecond).Seconds()
-		e.Elapsed = &dt
+		e := &event{Action: c.result}
+		if c.mode&Timestamp != 0 {
+			dt := time.Since(c.start).Round(1 * time.Millisecond).Seconds()
+			e.Elapsed = &dt
+		}
+		c.writeEvent(e)
 	}
-	c.writeEvent(e)
 	return nil
 }
 
 // writeOutputEvent writes a single output event with the given bytes.
-func (c *converter) writeOutputEvent(out []byte) {
+func (c *Converter) writeOutputEvent(out []byte) {
 	c.writeEvent(&event{
 		Action: "output",
 		Output: (*textBytes)(&out),
@@ -323,7 +331,7 @@ func (c *converter) writeOutputEvent(out []byte) {
 
 // writeEvent writes a single event.
 // It adds the package, time (if requested), and test name (if needed).
-func (c *converter) writeEvent(e *event) {
+func (c *Converter) writeEvent(e *event) {
 	e.Package = c.pkg
 	if c.mode&Timestamp != 0 {
 		t := time.Now()
diff --git a/src/cmd/internal/test2json/testdata/benchshort.json b/src/cmd/internal/test2json/testdata/benchshort.json
index 28e287c848..34b03b9362 100644
--- a/src/cmd/internal/test2json/testdata/benchshort.json
+++ b/src/cmd/internal/test2json/testdata/benchshort.json
@@ -4,4 +4,3 @@
 {"Action":"output","Output":"# but to avoid questions of timing, we just use a file with no \\n at all.\n"}
 {"Action":"output","Output":"BenchmarkFoo   \t"}
 {"Action":"output","Output":"10000 early EOF"}
-{"Action":"pass"}
diff --git a/src/cmd/internal/test2json/testdata/empty.json b/src/cmd/internal/test2json/testdata/empty.json
index 80b5217501..e69de29bb2 100644
--- a/src/cmd/internal/test2json/testdata/empty.json
+++ b/src/cmd/internal/test2json/testdata/empty.json
@@ -1 +0,0 @@
-{"Action":"pass"}
diff --git a/src/cmd/internal/traceviewer/format.go b/src/cmd/internal/traceviewer/format.go
new file mode 100644
index 0000000000..871477447f
--- /dev/null
+++ b/src/cmd/internal/traceviewer/format.go
@@ -0,0 +1,38 @@
+// Copyright 2020 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 traceviewer provides definitions of the JSON data structures
+// used by the Chrome trace viewer.
+//
+// The official description of the format is in this file:
+// https://docs.google.com/document/d/1CvAClvFfyA5R-PhYUmn5OOQtYMH4h6I0nSsKchNAySU/preview
+package traceviewer
+
+type Data struct {
+	Events   []*Event         `json:"traceEvents"`
+	Frames   map[string]Frame `json:"stackFrames"`
+	TimeUnit string           `json:"displayTimeUnit"`
+}
+
+type Event struct {
+	Name      string      `json:"name,omitempty"`
+	Phase     string      `json:"ph"`
+	Scope     string      `json:"s,omitempty"`
+	Time      float64     `json:"ts"`
+	Dur       float64     `json:"dur,omitempty"`
+	PID       uint64      `json:"pid"`
+	TID       uint64      `json:"tid"`
+	ID        uint64      `json:"id,omitempty"`
+	BindPoint string      `json:"bp,omitempty"`
+	Stack     int         `json:"sf,omitempty"`
+	EndStack  int         `json:"esf,omitempty"`
+	Arg       interface{} `json:"args,omitempty"`
+	Cname     string      `json:"cname,omitempty"`
+	Category  string      `json:"cat,omitempty"`
+}
+
+type Frame struct {
+	Name   string `json:"name"`
+	Parent int    `json:"parent,omitempty"`
+}
diff --git a/src/cmd/link/doc.go b/src/cmd/link/doc.go
index 219499be0a..604675caec 100644
--- a/src/cmd/link/doc.go
+++ b/src/cmd/link/doc.go
@@ -3,7 +3,7 @@
 // license that can be found in the LICENSE file.
 
 /*
-Link, typically invoked as ``go tool link,'' reads the Go archive or object
+Link, typically invoked as ``go tool link'', reads the Go archive or object
 for a package main, along with its dependencies, and combines them
 into an executable binary.
 
diff --git a/src/cmd/link/internal/arm/asm.go b/src/cmd/link/internal/arm/asm.go
index 22bcb518df..611c96ce35 100644
--- a/src/cmd/link/internal/arm/asm.go
+++ b/src/cmd/link/internal/arm/asm.go
@@ -220,7 +220,7 @@ func adddynrel(target *ld.Target, ldr *loader.Loader, syms *ld.ArchSyms, s loade
 		addpltsym(target, ldr, syms, targ)
 		su := ldr.MakeSymbolUpdater(s)
 		su.SetRelocSym(rIdx, syms.PLT)
-		su.SetRelocAdd(rIdx, int64(ldr.SymPlt(targ)))
+		su.SetRelocAdd(rIdx, int64(braddoff(int32(r.Add()), ldr.SymPlt(targ)/4))) // TODO: don't use r.Add for instruction bytes (issue 19811)
 		return true
 
 	case objabi.R_ADDR:
diff --git a/src/cmd/link/internal/benchmark/bench_test.go b/src/cmd/link/internal/benchmark/bench_test.go
index d8ec717c7c..419dc55724 100644
--- a/src/cmd/link/internal/benchmark/bench_test.go
+++ b/src/cmd/link/internal/benchmark/bench_test.go
@@ -1,6 +1,7 @@
 // Copyright 2020 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 benchmark
 
 import (
diff --git a/src/cmd/link/internal/ld/elf.go b/src/cmd/link/internal/ld/elf.go
index 957f5081f6..2862f65f9f 100644
--- a/src/cmd/link/internal/ld/elf.go
+++ b/src/cmd/link/internal/ld/elf.go
@@ -2378,6 +2378,7 @@ func elfadddynsym(ldr *loader.Loader, target *Target, syms *ArchSyms, s loader.S
 		if target.Arch.Family == sys.AMD64 && !cgoeDynamic && dil != "" && !seenlib[dil] {
 			du := ldr.MakeSymbolUpdater(syms.Dynamic)
 			Elfwritedynent(target.Arch, du, DT_NEEDED, uint64(dstru.Addstring(dil)))
+			seenlib[dil] = true
 		}
 	} else {
 
diff --git a/src/cmd/link/internal/ld/elf_test.go b/src/cmd/link/internal/ld/elf_test.go
index 8e86beb1ec..37f0e77336 100644
--- a/src/cmd/link/internal/ld/elf_test.go
+++ b/src/cmd/link/internal/ld/elf_test.go
@@ -13,6 +13,7 @@ import (
 	"os"
 	"os/exec"
 	"path/filepath"
+	"runtime"
 	"testing"
 )
 
@@ -77,3 +78,57 @@ func main() {
 		t.Fatalf("Unexpected sh info, want greater than 0, got: %d", section.Info)
 	}
 }
+
+func TestNoDuplicateNeededEntries(t *testing.T) {
+	testenv.MustHaveGoBuild(t)
+	testenv.MustHaveCGO(t)
+
+	// run this test on just a small set of platforms (no need to test it
+	// across the board given the nature of the test).
+	pair := runtime.GOOS + "-" + runtime.GOARCH
+	switch pair {
+	case "linux-amd64", "freebsd-amd64", "openbsd-amd64":
+	default:
+		t.Skip("no need for test on " + pair)
+	}
+
+	t.Parallel()
+
+	dir, err := ioutil.TempDir("", "no-dup-needed")
+	if err != nil {
+		t.Fatalf("Failed to create temp dir: %v", err)
+	}
+	defer os.RemoveAll(dir)
+
+	wd, err := os.Getwd()
+	if err != nil {
+		t.Fatalf("Failed to get working directory: %v", err)
+	}
+
+	path := filepath.Join(dir, "x")
+	argv := []string{"build", "-o", path, filepath.Join(wd, "testdata", "issue39256")}
+	out, err := exec.Command(testenv.GoToolPath(t), argv...).CombinedOutput()
+	if err != nil {
+		t.Fatalf("Build failure: %s\n%s\n", err, string(out))
+	}
+
+	f, err := elf.Open(path)
+	if err != nil {
+		t.Fatalf("Failed to open ELF file: %v", err)
+	}
+	libs, err := f.ImportedLibraries()
+	if err != nil {
+		t.Fatalf("Failed to read imported libraries: %v", err)
+	}
+
+	var count int
+	for _, lib := range libs {
+		if lib == "libc.so" {
+			count++
+		}
+	}
+
+	if got, want := count, 1; got != want {
+		t.Errorf("Got %d entries for `libc.so`, want %d", got, want)
+	}
+}
diff --git a/src/cmd/link/internal/ld/errors.go b/src/cmd/link/internal/ld/errors.go
index c5ce097fde..d6e8ff236d 100644
--- a/src/cmd/link/internal/ld/errors.go
+++ b/src/cmd/link/internal/ld/errors.go
@@ -1,6 +1,7 @@
 // Copyright 2020 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 ld
 
 import (
diff --git a/src/cmd/link/internal/ld/go.go b/src/cmd/link/internal/ld/go.go
index bf5c9ca1ba..b3541c46c0 100644
--- a/src/cmd/link/internal/ld/go.go
+++ b/src/cmd/link/internal/ld/go.go
@@ -183,6 +183,9 @@ func setCgoAttr(ctxt *Link, lookup func(string, int) loader.Sym, file string, pk
 					hostObjSyms[s] = struct{}{}
 				}
 				havedynamic = 1
+				if lib != "" && ctxt.IsDarwin() {
+					machoadddynlib(lib, ctxt.LinkMode)
+				}
 			}
 
 			continue
diff --git a/src/cmd/link/internal/ld/testdata/issue39256/x.go b/src/cmd/link/internal/ld/testdata/issue39256/x.go
new file mode 100644
index 0000000000..d8562ad172
--- /dev/null
+++ b/src/cmd/link/internal/ld/testdata/issue39256/x.go
@@ -0,0 +1,20 @@
+// Copyright 2020 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 (
+	_ "unsafe"
+)
+
+//go:cgo_import_dynamic libc_getpid getpid "libc.so"
+//go:cgo_import_dynamic libc_kill kill "libc.so"
+//go:cgo_import_dynamic libc_close close "libc.so"
+//go:cgo_import_dynamic libc_open open "libc.so"
+
+func trampoline()
+
+func main() {
+	trampoline()
+}
diff --git a/src/cmd/link/internal/ld/testdata/issue39256/x.s b/src/cmd/link/internal/ld/testdata/issue39256/x.s
new file mode 100644
index 0000000000..41a54b2e04
--- /dev/null
+++ b/src/cmd/link/internal/ld/testdata/issue39256/x.s
@@ -0,0 +1,10 @@
+// Copyright 2020 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.
+
+TEXT ·trampoline(SB),0,$0
+	CALL	libc_getpid(SB)
+	CALL	libc_kill(SB)
+	CALL	libc_open(SB)
+	CALL	libc_close(SB)
+	RET
diff --git a/src/cmd/test2json/main.go b/src/cmd/test2json/main.go
index 0385d8f246..57a874193e 100644
--- a/src/cmd/test2json/main.go
+++ b/src/cmd/test2json/main.go
@@ -118,12 +118,16 @@ func main() {
 		w := &countWriter{0, c}
 		cmd.Stdout = w
 		cmd.Stderr = w
-		if err := cmd.Run(); err != nil {
+		err := cmd.Run()
+		if err != nil {
 			if w.n > 0 {
 				// Assume command printed why it failed.
 			} else {
 				fmt.Fprintf(c, "test2json: %v\n", err)
 			}
+		}
+		c.Exited(err)
+		if err != nil {
 			c.Close()
 			os.Exit(1)
 		}
diff --git a/src/cmd/trace/trace.go b/src/cmd/trace/trace.go
index b452376627..30c80f0e04 100644
--- a/src/cmd/trace/trace.go
+++ b/src/cmd/trace/trace.go
@@ -5,6 +5,7 @@
 package main
 
 import (
+	"cmd/internal/traceviewer"
 	"encoding/json"
 	"fmt"
 	"internal/trace"
@@ -325,7 +326,7 @@ func splittingTraceConsumer(max int) (*splitter, traceConsumer) {
 	}
 
 	var (
-		data = ViewerData{Frames: make(map[string]ViewerFrame)}
+		data = traceviewer.Data{Frames: make(map[string]traceviewer.Frame)}
 
 		sizes []eventSz
 		cw    countingWriter
@@ -337,7 +338,7 @@ func splittingTraceConsumer(max int) (*splitter, traceConsumer) {
 		consumeTimeUnit: func(unit string) {
 			data.TimeUnit = unit
 		},
-		consumeViewerEvent: func(v *ViewerEvent, required bool) {
+		consumeViewerEvent: func(v *traceviewer.Event, required bool) {
 			if required {
 				// Store required events inside data
 				// so flush can include them in the required
@@ -350,7 +351,7 @@ func splittingTraceConsumer(max int) (*splitter, traceConsumer) {
 			sizes = append(sizes, eventSz{v.Time, cw.size + 1}) // +1 for ",".
 			cw.size = 0
 		},
-		consumeViewerFrame: func(k string, v ViewerFrame) {
+		consumeViewerFrame: func(k string, v traceviewer.Frame) {
 			data.Frames[k] = v
 		},
 		flush: func() {
@@ -478,36 +479,6 @@ type gInfo struct {
 	markAssist *trace.Event // if non-nil, the mark assist currently running.
 }
 
-type ViewerData struct {
-	Events   []*ViewerEvent         `json:"traceEvents"`
-	Frames   map[string]ViewerFrame `json:"stackFrames"`
-	TimeUnit string                 `json:"displayTimeUnit"`
-
-	// This is where mandatory part of the trace starts (e.g. thread names)
-	footer int
-}
-
-type ViewerEvent struct {
-	Name     string      `json:"name,omitempty"`
-	Phase    string      `json:"ph"`
-	Scope    string      `json:"s,omitempty"`
-	Time     float64     `json:"ts"`
-	Dur      float64     `json:"dur,omitempty"`
-	Pid      uint64      `json:"pid"`
-	Tid      uint64      `json:"tid"`
-	ID       uint64      `json:"id,omitempty"`
-	Stack    int         `json:"sf,omitempty"`
-	EndStack int         `json:"esf,omitempty"`
-	Arg      interface{} `json:"args,omitempty"`
-	Cname    string      `json:"cname,omitempty"`
-	Category string      `json:"cat,omitempty"`
-}
-
-type ViewerFrame struct {
-	Name   string `json:"name"`
-	Parent int    `json:"parent,omitempty"`
-}
-
 type NameArg struct {
 	Name string `json:"name"`
 }
@@ -528,8 +499,8 @@ type SortIndexArg struct {
 
 type traceConsumer struct {
 	consumeTimeUnit    func(unit string)
-	consumeViewerEvent func(v *ViewerEvent, required bool)
-	consumeViewerFrame func(key string, f ViewerFrame)
+	consumeViewerEvent func(v *traceviewer.Event, required bool)
+	consumeViewerFrame func(key string, f traceviewer.Frame)
 	flush              func()
 }
 
@@ -775,23 +746,23 @@ func generateTrace(params *traceParams, consumer traceConsumer) error {
 		ctx.emitSectionFooter(procsSection, "PROCS", 2)
 	}
 
-	ctx.emitFooter(&ViewerEvent{Name: "thread_name", Phase: "M", Pid: procsSection, Tid: trace.GCP, Arg: &NameArg{"GC"}})
-	ctx.emitFooter(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: procsSection, Tid: trace.GCP, Arg: &SortIndexArg{-6}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_name", Phase: "M", PID: procsSection, TID: trace.GCP, Arg: &NameArg{"GC"}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: procsSection, TID: trace.GCP, Arg: &SortIndexArg{-6}})
 
-	ctx.emitFooter(&ViewerEvent{Name: "thread_name", Phase: "M", Pid: procsSection, Tid: trace.NetpollP, Arg: &NameArg{"Network"}})
-	ctx.emitFooter(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: procsSection, Tid: trace.NetpollP, Arg: &SortIndexArg{-5}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_name", Phase: "M", PID: procsSection, TID: trace.NetpollP, Arg: &NameArg{"Network"}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: procsSection, TID: trace.NetpollP, Arg: &SortIndexArg{-5}})
 
-	ctx.emitFooter(&ViewerEvent{Name: "thread_name", Phase: "M", Pid: procsSection, Tid: trace.TimerP, Arg: &NameArg{"Timers"}})
-	ctx.emitFooter(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: procsSection, Tid: trace.TimerP, Arg: &SortIndexArg{-4}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_name", Phase: "M", PID: procsSection, TID: trace.TimerP, Arg: &NameArg{"Timers"}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: procsSection, TID: trace.TimerP, Arg: &SortIndexArg{-4}})
 
-	ctx.emitFooter(&ViewerEvent{Name: "thread_name", Phase: "M", Pid: procsSection, Tid: trace.SyscallP, Arg: &NameArg{"Syscalls"}})
-	ctx.emitFooter(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: procsSection, Tid: trace.SyscallP, Arg: &SortIndexArg{-3}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_name", Phase: "M", PID: procsSection, TID: trace.SyscallP, Arg: &NameArg{"Syscalls"}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: procsSection, TID: trace.SyscallP, Arg: &SortIndexArg{-3}})
 
 	// Display rows for Ps if we are in the default trace view mode (not goroutine-oriented presentation)
 	if ctx.mode&modeGoroutineOriented == 0 {
 		for i := 0; i <= maxProc; i++ {
-			ctx.emitFooter(&ViewerEvent{Name: "thread_name", Phase: "M", Pid: procsSection, Tid: uint64(i), Arg: &NameArg{fmt.Sprintf("Proc %v", i)}})
-			ctx.emitFooter(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: procsSection, Tid: uint64(i), Arg: &SortIndexArg{i}})
+			ctx.emitFooter(&traceviewer.Event{Name: "thread_name", Phase: "M", PID: procsSection, TID: uint64(i), Arg: &NameArg{fmt.Sprintf("Proc %v", i)}})
+			ctx.emitFooter(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: procsSection, TID: uint64(i), Arg: &SortIndexArg{i}})
 		}
 	}
 
@@ -829,27 +800,27 @@ func generateTrace(params *traceParams, consumer traceConsumer) error {
 			if !ctx.gs[k] {
 				continue
 			}
-			ctx.emitFooter(&ViewerEvent{Name: "thread_name", Phase: "M", Pid: procsSection, Tid: k, Arg: &NameArg{v.name}})
+			ctx.emitFooter(&traceviewer.Event{Name: "thread_name", Phase: "M", PID: procsSection, TID: k, Arg: &NameArg{v.name}})
 		}
 		// Row for the main goroutine (maing)
-		ctx.emitFooter(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: procsSection, Tid: ctx.maing, Arg: &SortIndexArg{-2}})
+		ctx.emitFooter(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: procsSection, TID: ctx.maing, Arg: &SortIndexArg{-2}})
 		// Row for GC or global state (specified with G=0)
-		ctx.emitFooter(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: procsSection, Tid: 0, Arg: &SortIndexArg{-1}})
+		ctx.emitFooter(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: procsSection, TID: 0, Arg: &SortIndexArg{-1}})
 	}
 
 	return nil
 }
 
-func (ctx *traceContext) emit(e *ViewerEvent) {
+func (ctx *traceContext) emit(e *traceviewer.Event) {
 	ctx.consumer.consumeViewerEvent(e, false)
 }
 
-func (ctx *traceContext) emitFooter(e *ViewerEvent) {
+func (ctx *traceContext) emitFooter(e *traceviewer.Event) {
 	ctx.consumer.consumeViewerEvent(e, true)
 }
 func (ctx *traceContext) emitSectionFooter(sectionID uint64, name string, priority int) {
-	ctx.emitFooter(&ViewerEvent{Name: "process_name", Phase: "M", Pid: sectionID, Arg: &NameArg{name}})
-	ctx.emitFooter(&ViewerEvent{Name: "process_sort_index", Phase: "M", Pid: sectionID, Arg: &SortIndexArg{priority}})
+	ctx.emitFooter(&traceviewer.Event{Name: "process_name", Phase: "M", PID: sectionID, Arg: &NameArg{name}})
+	ctx.emitFooter(&traceviewer.Event{Name: "process_sort_index", Phase: "M", PID: sectionID, Arg: &SortIndexArg{priority}})
 }
 
 func (ctx *traceContext) time(ev *trace.Event) float64 {
@@ -880,7 +851,7 @@ func (ctx *traceContext) emitSlice(ev *trace.Event, name string) {
 	ctx.emit(ctx.makeSlice(ev, name))
 }
 
-func (ctx *traceContext) makeSlice(ev *trace.Event, name string) *ViewerEvent {
+func (ctx *traceContext) makeSlice(ev *trace.Event, name string) *traceviewer.Event {
 	// If ViewerEvent.Dur is not a positive value,
 	// trace viewer handles it as a non-terminating time interval.
 	// Avoid it by setting the field with a small value.
@@ -888,12 +859,12 @@ func (ctx *traceContext) makeSlice(ev *trace.Event, name string) *ViewerEvent {
 	if ev.Link.Ts-ev.Ts <= 0 {
 		durationUsec = 0.0001 // 0.1 nanoseconds
 	}
-	sl := &ViewerEvent{
+	sl := &traceviewer.Event{
 		Name:     name,
 		Phase:    "X",
 		Time:     ctx.time(ev),
 		Dur:      durationUsec,
-		Tid:      ctx.proc(ev),
+		TID:      ctx.proc(ev),
 		Stack:    ctx.stack(ev.Stk),
 		EndStack: ctx.stack(ev.Link.Stk),
 	}
@@ -927,16 +898,16 @@ func (ctx *traceContext) emitTask(task *taskDesc, sortIndex int) {
 	taskName := task.name
 	durationUsec := float64(task.lastTimestamp()-task.firstTimestamp()) / 1e3
 
-	ctx.emitFooter(&ViewerEvent{Name: "thread_name", Phase: "M", Pid: tasksSection, Tid: taskRow, Arg: &NameArg{fmt.Sprintf("T%d %s", task.id, taskName)}})
-	ctx.emit(&ViewerEvent{Name: "thread_sort_index", Phase: "M", Pid: tasksSection, Tid: taskRow, Arg: &SortIndexArg{sortIndex}})
+	ctx.emitFooter(&traceviewer.Event{Name: "thread_name", Phase: "M", PID: tasksSection, TID: taskRow, Arg: &NameArg{fmt.Sprintf("T%d %s", task.id, taskName)}})
+	ctx.emit(&traceviewer.Event{Name: "thread_sort_index", Phase: "M", PID: tasksSection, TID: taskRow, Arg: &SortIndexArg{sortIndex}})
 	ts := float64(task.firstTimestamp()) / 1e3
-	sl := &ViewerEvent{
+	sl := &traceviewer.Event{
 		Name:  taskName,
 		Phase: "X",
 		Time:  ts,
 		Dur:   durationUsec,
-		Pid:   tasksSection,
-		Tid:   taskRow,
+		PID:   tasksSection,
+		TID:   taskRow,
 		Cname: pickTaskColor(task.id),
 	}
 	targ := TaskArg{ID: task.id}
@@ -953,8 +924,8 @@ func (ctx *traceContext) emitTask(task *taskDesc, sortIndex int) {
 
 	if task.create != nil && task.create.Type == trace.EvUserTaskCreate && task.create.Args[1] != 0 {
 		ctx.arrowSeq++
-		ctx.emit(&ViewerEvent{Name: "newTask", Phase: "s", Tid: task.create.Args[1], ID: ctx.arrowSeq, Time: ts, Pid: tasksSection})
-		ctx.emit(&ViewerEvent{Name: "newTask", Phase: "t", Tid: taskRow, ID: ctx.arrowSeq, Time: ts, Pid: tasksSection})
+		ctx.emit(&traceviewer.Event{Name: "newTask", Phase: "s", TID: task.create.Args[1], ID: ctx.arrowSeq, Time: ts, PID: tasksSection})
+		ctx.emit(&traceviewer.Event{Name: "newTask", Phase: "t", TID: taskRow, ID: ctx.arrowSeq, Time: ts, PID: tasksSection})
 	}
 }
 
@@ -975,12 +946,12 @@ func (ctx *traceContext) emitRegion(s regionDesc) {
 	scopeID := fmt.Sprintf("%x", id)
 	name := s.Name
 
-	sl0 := &ViewerEvent{
+	sl0 := &traceviewer.Event{
 		Category: "Region",
 		Name:     name,
 		Phase:    "b",
 		Time:     float64(s.firstTimestamp()) / 1e3,
-		Tid:      s.G, // only in goroutine-oriented view
+		TID:      s.G, // only in goroutine-oriented view
 		ID:       uint64(regionID),
 		Scope:    scopeID,
 		Cname:    pickTaskColor(s.TaskID),
@@ -990,12 +961,12 @@ func (ctx *traceContext) emitRegion(s regionDesc) {
 	}
 	ctx.emit(sl0)
 
-	sl1 := &ViewerEvent{
+	sl1 := &traceviewer.Event{
 		Category: "Region",
 		Name:     name,
 		Phase:    "e",
 		Time:     float64(s.lastTimestamp()) / 1e3,
-		Tid:      s.G,
+		TID:      s.G,
 		ID:       uint64(regionID),
 		Scope:    scopeID,
 		Cname:    pickTaskColor(s.TaskID),
@@ -1021,7 +992,7 @@ func (ctx *traceContext) emitHeapCounters(ev *trace.Event) {
 		diff = ctx.heapStats.nextGC - ctx.heapStats.heapAlloc
 	}
 	if tsWithinRange(ev.Ts, ctx.startTime, ctx.endTime) {
-		ctx.emit(&ViewerEvent{Name: "Heap", Phase: "C", Time: ctx.time(ev), Pid: 1, Arg: &heapCountersArg{ctx.heapStats.heapAlloc, diff}})
+		ctx.emit(&traceviewer.Event{Name: "Heap", Phase: "C", Time: ctx.time(ev), PID: 1, Arg: &heapCountersArg{ctx.heapStats.heapAlloc, diff}})
 	}
 	ctx.prevHeapStats = ctx.heapStats
 }
@@ -1037,7 +1008,7 @@ func (ctx *traceContext) emitGoroutineCounters(ev *trace.Event) {
 		return
 	}
 	if tsWithinRange(ev.Ts, ctx.startTime, ctx.endTime) {
-		ctx.emit(&ViewerEvent{Name: "Goroutines", Phase: "C", Time: ctx.time(ev), Pid: 1, Arg: &goroutineCountersArg{uint64(ctx.gstates[gRunning]), uint64(ctx.gstates[gRunnable]), uint64(ctx.gstates[gWaitingGC])}})
+		ctx.emit(&traceviewer.Event{Name: "Goroutines", Phase: "C", Time: ctx.time(ev), PID: 1, Arg: &goroutineCountersArg{uint64(ctx.gstates[gRunning]), uint64(ctx.gstates[gRunnable]), uint64(ctx.gstates[gWaitingGC])}})
 	}
 	ctx.prevGstates = ctx.gstates
 }
@@ -1052,7 +1023,7 @@ func (ctx *traceContext) emitThreadCounters(ev *trace.Event) {
 		return
 	}
 	if tsWithinRange(ev.Ts, ctx.startTime, ctx.endTime) {
-		ctx.emit(&ViewerEvent{Name: "Threads", Phase: "C", Time: ctx.time(ev), Pid: 1, Arg: &threadCountersArg{
+		ctx.emit(&traceviewer.Event{Name: "Threads", Phase: "C", Time: ctx.time(ev), PID: 1, Arg: &threadCountersArg{
 			Running:   ctx.threadStats.prunning,
 			InSyscall: ctx.threadStats.insyscall}})
 	}
@@ -1090,13 +1061,13 @@ func (ctx *traceContext) emitInstant(ev *trace.Event, name, category string) {
 		}
 		arg = &Arg{ev.Args[0]}
 	}
-	ctx.emit(&ViewerEvent{
+	ctx.emit(&traceviewer.Event{
 		Name:     name,
 		Category: category,
 		Phase:    "I",
 		Scope:    "t",
 		Time:     ctx.time(ev),
-		Tid:      ctx.proc(ev),
+		TID:      ctx.proc(ev),
 		Stack:    ctx.stack(ev.Stk),
 		Cname:    cname,
 		Arg:      arg})
@@ -1134,8 +1105,8 @@ func (ctx *traceContext) emitArrow(ev *trace.Event, name string) {
 	}
 
 	ctx.arrowSeq++
-	ctx.emit(&ViewerEvent{Name: name, Phase: "s", Tid: ctx.proc(ev), ID: ctx.arrowSeq, Time: ctx.time(ev), Stack: ctx.stack(ev.Stk), Cname: color})
-	ctx.emit(&ViewerEvent{Name: name, Phase: "t", Tid: ctx.proc(ev.Link), ID: ctx.arrowSeq, Time: ctx.time(ev.Link), Cname: color})
+	ctx.emit(&traceviewer.Event{Name: name, Phase: "s", TID: ctx.proc(ev), ID: ctx.arrowSeq, Time: ctx.time(ev), Stack: ctx.stack(ev.Stk), Cname: color})
+	ctx.emit(&traceviewer.Event{Name: name, Phase: "t", TID: ctx.proc(ev.Link), ID: ctx.arrowSeq, Time: ctx.time(ev.Link), Cname: color})
 }
 
 func (ctx *traceContext) stack(stk []*trace.Frame) int {
@@ -1157,7 +1128,7 @@ func (ctx *traceContext) buildBranch(parent frameNode, stk []*trace.Frame) int {
 		node.id = ctx.frameSeq
 		node.children = make(map[uint64]frameNode)
 		parent.children[frame.PC] = node
-		ctx.consumer.consumeViewerFrame(strconv.Itoa(node.id), ViewerFrame{fmt.Sprintf("%v:%v", frame.Fn, frame.Line), parent.id})
+		ctx.consumer.consumeViewerFrame(strconv.Itoa(node.id), traceviewer.Frame{Name: fmt.Sprintf("%v:%v", frame.Fn, frame.Line), Parent: parent.id})
 	}
 	return ctx.buildBranch(node, stk)
 }
@@ -1192,7 +1163,7 @@ type jsonWriter struct {
 }
 
 func viewerDataTraceConsumer(w io.Writer, start, end int64) traceConsumer {
-	frames := make(map[string]ViewerFrame)
+	frames := make(map[string]traceviewer.Frame)
 	enc := json.NewEncoder(w)
 	written := 0
 	index := int64(-1)
@@ -1204,7 +1175,7 @@ func viewerDataTraceConsumer(w io.Writer, start, end int64) traceConsumer {
 			enc.Encode(unit)
 			io.WriteString(w, ",")
 		},
-		consumeViewerEvent: func(v *ViewerEvent, required bool) {
+		consumeViewerEvent: func(v *traceviewer.Event, required bool) {
 			index++
 			if !required && (index < start || index > end) {
 				// not in the range. Skip!
@@ -1221,7 +1192,7 @@ func viewerDataTraceConsumer(w io.Writer, start, end int64) traceConsumer {
 			// Same should be applied to splittingTraceConsumer.
 			written++
 		},
-		consumeViewerFrame: func(k string, v ViewerFrame) {
+		consumeViewerFrame: func(k string, v traceviewer.Frame) {
 			frames[k] = v
 		},
 		flush: func() {
diff --git a/src/cmd/trace/trace_test.go b/src/cmd/trace/trace_test.go
index ef2d06c961..dd12e8cd20 100644
--- a/src/cmd/trace/trace_test.go
+++ b/src/cmd/trace/trace_test.go
@@ -7,6 +7,7 @@
 package main
 
 import (
+	"cmd/internal/traceviewer"
 	"context"
 	"internal/trace"
 	"io/ioutil"
@@ -78,7 +79,7 @@ func TestGoroutineCount(t *testing.T) {
 	// Use the default viewerDataTraceConsumer but replace
 	// consumeViewerEvent to intercept the ViewerEvents for testing.
 	c := viewerDataTraceConsumer(ioutil.Discard, 0, 1<<63-1)
-	c.consumeViewerEvent = func(ev *ViewerEvent, _ bool) {
+	c.consumeViewerEvent = func(ev *traceviewer.Event, _ bool) {
 		if ev.Name == "Goroutines" {
 			cnt := ev.Arg.(*goroutineCountersArg)
 			if cnt.Runnable+cnt.Running > 2 {
@@ -165,7 +166,7 @@ func TestPreemptedMarkAssist(t *testing.T) {
 	c := viewerDataTraceConsumer(ioutil.Discard, 0, 1<<63-1)
 
 	marks := 0
-	c.consumeViewerEvent = func(ev *ViewerEvent, _ bool) {
+	c.consumeViewerEvent = func(ev *traceviewer.Event, _ bool) {
 		if strings.Contains(ev.Name, "MARK ASSIST") {
 			marks++
 		}
@@ -216,7 +217,7 @@ func TestFoo(t *testing.T) {
 	c := viewerDataTraceConsumer(ioutil.Discard, 0, 1<<63-1)
 
 	var logBeforeTaskEnd, logAfterTaskEnd bool
-	c.consumeViewerEvent = func(ev *ViewerEvent, _ bool) {
+	c.consumeViewerEvent = func(ev *traceviewer.Event, _ bool) {
 		if ev.Name == "log before task ends" {
 			logBeforeTaskEnd = true
 		}
diff --git a/src/cmd/trace/trace_unix_test.go b/src/cmd/trace/trace_unix_test.go
index fec060e121..645978e0f8 100644
--- a/src/cmd/trace/trace_unix_test.go
+++ b/src/cmd/trace/trace_unix_test.go
@@ -8,6 +8,7 @@ package main
 
 import (
 	"bytes"
+	"cmd/internal/traceviewer"
 	traceparser "internal/trace"
 	"io/ioutil"
 	"runtime"
@@ -83,7 +84,7 @@ func TestGoroutineInSyscall(t *testing.T) {
 	// Check only one thread for the pipe read goroutine is
 	// considered in-syscall.
 	c := viewerDataTraceConsumer(ioutil.Discard, 0, 1<<63-1)
-	c.consumeViewerEvent = func(ev *ViewerEvent, _ bool) {
+	c.consumeViewerEvent = func(ev *traceviewer.Event, _ bool) {
 		if ev.Name == "Threads" {
 			arg := ev.Arg.(*threadCountersArg)
 			if arg.InSyscall > 1 {
diff --git a/src/cmd/vendor/golang.org/x/mod/module/module.go b/src/cmd/vendor/golang.org/x/mod/module/module.go
index 6cd37280a8..3a8b080c7b 100644
--- a/src/cmd/vendor/golang.org/x/mod/module/module.go
+++ b/src/cmd/vendor/golang.org/x/mod/module/module.go
@@ -97,6 +97,7 @@ package module
 
 import (
 	"fmt"
+	"path"
 	"sort"
 	"strings"
 	"unicode"
@@ -716,3 +717,49 @@ func unescapeString(escaped string) (string, bool) {
 	}
 	return string(buf), true
 }
+
+// MatchPrefixPatterns reports whether any path prefix of target matches one of
+// the glob patterns (as defined by path.Match) in the comma-separated globs
+// list. This implements the algorithm used when matching a module path to the
+// GOPRIVATE environment variable, as described by 'go help module-private'.
+//
+// It ignores any empty or malformed patterns in the list.
+func MatchPrefixPatterns(globs, target string) bool {
+	for globs != "" {
+		// Extract next non-empty glob in comma-separated list.
+		var glob string
+		if i := strings.Index(globs, ","); i >= 0 {
+			glob, globs = globs[:i], globs[i+1:]
+		} else {
+			glob, globs = globs, ""
+		}
+		if glob == "" {
+			continue
+		}
+
+		// A glob with N+1 path elements (N slashes) needs to be matched
+		// against the first N+1 path elements of target,
+		// which end just before the N+1'th slash.
+		n := strings.Count(glob, "/")
+		prefix := target
+		// Walk target, counting slashes, truncating at the N+1'th slash.
+		for i := 0; i < len(target); i++ {
+			if target[i] == '/' {
+				if n == 0 {
+					prefix = target[:i]
+					break
+				}
+				n--
+			}
+		}
+		if n > 0 {
+			// Not enough prefix elements.
+			continue
+		}
+		matched, _ := path.Match(glob, prefix)
+		if matched {
+			return true
+		}
+	}
+	return false
+}
diff --git a/src/cmd/vendor/modules.txt b/src/cmd/vendor/modules.txt
index 21fc78c237..7272f04ff3 100644
--- a/src/cmd/vendor/modules.txt
+++ b/src/cmd/vendor/modules.txt
@@ -29,7 +29,7 @@ golang.org/x/arch/x86/x86asm
 golang.org/x/crypto/ed25519
 golang.org/x/crypto/ed25519/internal/edwards25519
 golang.org/x/crypto/ssh/terminal
-# golang.org/x/mod v0.3.0
+# golang.org/x/mod v0.3.1-0.20200625141748-0b26df4a2231
 ## explicit
 golang.org/x/mod/internal/lazyregexp
 golang.org/x/mod/modfile
diff --git a/src/crypto/elliptic/p256_asm_amd64.s b/src/crypto/elliptic/p256_asm_amd64.s
index 7afa54a58c..c77b11bcf2 100644
--- a/src/crypto/elliptic/p256_asm_amd64.s
+++ b/src/crypto/elliptic/p256_asm_amd64.s
@@ -1336,7 +1336,7 @@ TEXT p256SubInternal(SB),NOSPLIT,$0
 
 	RET
 /* ---------------------------------------*/
-TEXT p256MulInternal(SB),NOSPLIT,$0
+TEXT p256MulInternal(SB),NOSPLIT,$8
 	MOVQ acc4, mul0
 	MULQ t0
 	MOVQ mul0, acc0
@@ -1519,7 +1519,7 @@ TEXT p256MulInternal(SB),NOSPLIT,$0
 
 	RET
 /* ---------------------------------------*/
-TEXT p256SqrInternal(SB),NOSPLIT,$0
+TEXT p256SqrInternal(SB),NOSPLIT,$8
 
 	MOVQ acc4, mul0
 	MULQ acc5
@@ -2345,4 +2345,3 @@ TEXT ·p256PointDoubleAsm(SB),NOSPLIT,$256-48
 
 	RET
 /* ---------------------------------------*/
-
diff --git a/src/crypto/md5/md5block_amd64.s b/src/crypto/md5/md5block_amd64.s
index 90d932b146..7c7d92d7e8 100644
--- a/src/crypto/md5/md5block_amd64.s
+++ b/src/crypto/md5/md5block_amd64.s
@@ -13,7 +13,7 @@
 // Licence: I hereby disclaim the copyright on this code and place it
 // in the public domain.
 
-TEXT	·block(SB),NOSPLIT,$0-32
+TEXT	·block(SB),NOSPLIT,$8-32
 	MOVQ	dig+0(FP),	BP
 	MOVQ	p+8(FP),	SI
 	MOVQ	p_len+16(FP), DX
diff --git a/src/database/sql/sql.go b/src/database/sql/sql.go
index b3d0653f5c..0b85db66b9 100644
--- a/src/database/sql/sql.go
+++ b/src/database/sql/sql.go
@@ -3110,6 +3110,9 @@ func rowsColumnInfoSetupConnLocked(rowsi driver.Rows) []*ColumnType {
 // "select cursor(select * from my_table) from dual", into a
 // *Rows value that can itself be scanned from. The parent
 // select query will close any cursor *Rows if the parent *Rows is closed.
+//
+// If any of the first arguments implementing Scanner returns an error,
+// that error will be wrapped in the returned error
 func (rs *Rows) Scan(dest ...interface{}) error {
 	rs.closemu.RLock()
 
@@ -3133,7 +3136,7 @@ func (rs *Rows) Scan(dest ...interface{}) error {
 	for i, sv := range rs.lastcols {
 		err := convertAssignRows(dest[i], sv, rs)
 		if err != nil {
-			return fmt.Errorf(`sql: Scan error on column index %d, name %q: %v`, i, rs.rowsi.Columns()[i], err)
+			return fmt.Errorf(`sql: Scan error on column index %d, name %q: %w`, i, rs.rowsi.Columns()[i], err)
 		}
 	}
 	return nil
diff --git a/src/database/sql/sql_test.go b/src/database/sql/sql_test.go
index 5727f0d8aa..762d42f54b 100644
--- a/src/database/sql/sql_test.go
+++ b/src/database/sql/sql_test.go
@@ -4149,6 +4149,41 @@ func TestQueryExecContextOnly(t *testing.T) {
 	}
 }
 
+type alwaysErrScanner struct{}
+
+var errTestScanWrap = errors.New("errTestScanWrap")
+
+func (alwaysErrScanner) Scan(interface{}) error {
+	return errTestScanWrap
+}
+
+// Issue 38099: Ensure that Rows.Scan properly wraps underlying errors.
+func TestRowsScanProperlyWrapsErrors(t *testing.T) {
+	db := newTestDB(t, "people")
+	defer closeDB(t, db)
+
+	rows, err := db.Query("SELECT|people|age|")
+	if err != nil {
+		t.Fatalf("Query: %v", err)
+	}
+
+	var res alwaysErrScanner
+
+	for rows.Next() {
+		err = rows.Scan(&res)
+		if err == nil {
+			t.Fatal("expecting back an error")
+		}
+		if !errors.Is(err, errTestScanWrap) {
+			t.Fatalf("errors.Is mismatch\n%v\nWant: %v", err, errTestScanWrap)
+		}
+		// Ensure that error substring matching still correctly works.
+		if !strings.Contains(err.Error(), errTestScanWrap.Error()) {
+			t.Fatalf("Error %v does not contain %v", err, errTestScanWrap)
+		}
+	}
+}
+
 // badConn implements a bad driver.Conn, for TestBadDriver.
 // The Exec method panics.
 type badConn struct{}
diff --git a/src/encoding/asn1/common.go b/src/encoding/asn1/common.go
index e2aa8bd9c5..1c712e1eff 100644
--- a/src/encoding/asn1/common.go
+++ b/src/encoding/asn1/common.go
@@ -92,7 +92,16 @@ type fieldParameters struct {
 // parseFieldParameters will parse it into a fieldParameters structure,
 // ignoring unknown parts of the string.
 func parseFieldParameters(str string) (ret fieldParameters) {
-	for _, part := range strings.Split(str, ",") {
+	var part string
+	for len(str) > 0 {
+		// This loop uses IndexByte and explicit slicing
+		// instead of strings.Split(str, ",") to reduce allocations.
+		i := strings.IndexByte(str, ',')
+		if i < 0 {
+			part, str = str, ""
+		} else {
+			part, str = str[:i], str[i+1:]
+		}
 		switch {
 		case part == "optional":
 			ret.optional = true
diff --git a/src/fmt/fmt_test.go b/src/fmt/fmt_test.go
index 6004061020..87fb323809 100644
--- a/src/fmt/fmt_test.go
+++ b/src/fmt/fmt_test.go
@@ -290,11 +290,11 @@ var fmtTests = []struct {
 	{"%q", '\U00000e00', `'\u0e00'`},
 	{"%q", '\U0010ffff', `'\U0010ffff'`},
 	// Runes that are not valid.
-	{"%q", int32(-1), "%!q(int32=-1)"},
+	{"%q", int32(-1), `'�'`},
 	{"%q", 0xDC80, `'�'`},
-	{"%q", rune(0x110000), "%!q(int32=1114112)"},
-	{"%q", int64(0xFFFFFFFFF), "%!q(int64=68719476735)"},
-	{"%q", uint64(0xFFFFFFFFF), "%!q(uint64=68719476735)"},
+	{"%q", rune(0x110000), `'�'`},
+	{"%q", int64(0xFFFFFFFFF), `'�'`},
+	{"%q", uint64(0xFFFFFFFFF), `'�'`},
 
 	// width
 	{"%5s", "abc", "  abc"},
diff --git a/src/fmt/print.go b/src/fmt/print.go
index 595869140a..778b5b0938 100644
--- a/src/fmt/print.go
+++ b/src/fmt/print.go
@@ -388,11 +388,7 @@ func (p *pp) fmtInteger(v uint64, isSigned bool, verb rune) {
 	case 'c':
 		p.fmt.fmtC(v)
 	case 'q':
-		if v <= utf8.MaxRune {
-			p.fmt.fmtQc(v)
-		} else {
-			p.badVerb(verb)
-		}
+		p.fmt.fmtQc(v)
 	case 'U':
 		p.fmt.fmtUnicode(v)
 	default:
diff --git a/src/go/types/gotype.go b/src/go/types/gotype.go
index 19dd702c45..eacf68f52f 100644
--- a/src/go/types/gotype.go
+++ b/src/go/types/gotype.go
@@ -48,9 +48,9 @@ The flags are:
 
 Flags controlling additional output:
 	-ast
-		print AST (forces -seq)
+		print AST
 	-trace
-		print parse trace (forces -seq)
+		print parse trace
 	-comments
 		parse comments (ignored unless -ast or -trace is provided)
 	-panic
@@ -104,8 +104,8 @@ var (
 	compiler   = flag.String("c", "source", "compiler used for installed packages (gc, gccgo, or source)")
 
 	// additional output control
-	printAST      = flag.Bool("ast", false, "print AST (forces -seq)")
-	printTrace    = flag.Bool("trace", false, "print parse trace (forces -seq)")
+	printAST      = flag.Bool("ast", false, "print AST")
+	printTrace    = flag.Bool("trace", false, "print parse trace")
 	parseComments = flag.Bool("comments", false, "parse comments (ignored unless -ast or -trace is provided)")
 	panicOnError  = flag.Bool("panic", false, "panic on first error")
 )
diff --git a/src/hash/maphash/maphash_test.go b/src/hash/maphash/maphash_test.go
index caea43a8c8..daf6eb4786 100644
--- a/src/hash/maphash/maphash_test.go
+++ b/src/hash/maphash/maphash_test.go
@@ -165,3 +165,32 @@ func TestSeedFromReset(t *testing.T) {
 // Make sure a Hash implements the hash.Hash and hash.Hash64 interfaces.
 var _ hash.Hash = &Hash{}
 var _ hash.Hash64 = &Hash{}
+
+func benchmarkSize(b *testing.B, size int) {
+	h := &Hash{}
+	buf := make([]byte, size)
+	b.SetBytes(int64(size))
+	b.ResetTimer()
+
+	for i := 0; i < b.N; i++ {
+		h.Reset()
+		h.Write(buf)
+		h.Sum64()
+	}
+}
+
+func BenchmarkHash8Bytes(b *testing.B) {
+	benchmarkSize(b, 8)
+}
+
+func BenchmarkHash320Bytes(b *testing.B) {
+	benchmarkSize(b, 320)
+}
+
+func BenchmarkHash1K(b *testing.B) {
+	benchmarkSize(b, 1024)
+}
+
+func BenchmarkHash8K(b *testing.B) {
+	benchmarkSize(b, 8192)
+}
diff --git a/src/image/gif/writer_test.go b/src/image/gif/writer_test.go
index 9b15c8d99d..5d1b2c439e 100644
--- a/src/image/gif/writer_test.go
+++ b/src/image/gif/writer_test.go
@@ -9,6 +9,7 @@ import (
 	"image"
 	"image/color"
 	"image/color/palette"
+	"image/draw"
 	_ "image/png"
 	"io/ioutil"
 	"math/rand"
@@ -656,25 +657,14 @@ func TestEncodeWrappedImage(t *testing.T) {
 	}
 }
 
-func BenchmarkEncode(b *testing.B) {
+func BenchmarkEncodeRandomPaletted(b *testing.B) {
+	img := image.NewPaletted(image.Rect(0, 0, 640, 480), palette.Plan9)
 	rnd := rand.New(rand.NewSource(123))
-
-	// Restrict to a 256-color paletted image to avoid quantization path.
-	palette := make(color.Palette, 256)
-	for i := range palette {
-		palette[i] = color.RGBA{
-			uint8(rnd.Intn(256)),
-			uint8(rnd.Intn(256)),
-			uint8(rnd.Intn(256)),
-			255,
-		}
-	}
-	img := image.NewPaletted(image.Rect(0, 0, 640, 480), palette)
 	for i := range img.Pix {
 		img.Pix[i] = uint8(rnd.Intn(256))
 	}
 
-	b.SetBytes(640 * 480 * 4)
+	b.SetBytes(640 * 480 * 1)
 	b.ReportAllocs()
 	b.ResetTimer()
 	for i := 0; i < b.N; i++ {
@@ -682,7 +672,7 @@ func BenchmarkEncode(b *testing.B) {
 	}
 }
 
-func BenchmarkQuantizedEncode(b *testing.B) {
+func BenchmarkEncodeRandomRGBA(b *testing.B) {
 	img := image.NewRGBA(image.Rect(0, 0, 640, 480))
 	bo := img.Bounds()
 	rnd := rand.New(rand.NewSource(123))
@@ -696,6 +686,7 @@ func BenchmarkQuantizedEncode(b *testing.B) {
 			})
 		}
 	}
+
 	b.SetBytes(640 * 480 * 4)
 	b.ReportAllocs()
 	b.ResetTimer()
@@ -703,3 +694,35 @@ func BenchmarkQuantizedEncode(b *testing.B) {
 		Encode(ioutil.Discard, img, nil)
 	}
 }
+
+func BenchmarkEncodeRealisticPaletted(b *testing.B) {
+	rgba, err := readImg("../testdata/video-001.png")
+	if err != nil {
+		b.Fatalf("readImg: %v", err)
+	}
+	bo := rgba.Bounds()
+	img := image.NewPaletted(bo, palette.Plan9)
+	draw.Draw(img, bo, rgba, bo.Min, draw.Src)
+
+	b.SetBytes(int64(bo.Dx() * bo.Dy() * 1))
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		Encode(ioutil.Discard, img, nil)
+	}
+}
+
+func BenchmarkEncodeRealisticRGBA(b *testing.B) {
+	img, err := readImg("../testdata/video-001.png")
+	if err != nil {
+		b.Fatalf("readImg: %v", err)
+	}
+	bo := img.Bounds()
+
+	b.SetBytes(int64(bo.Dx() * bo.Dy() * 4))
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		Encode(ioutil.Discard, img, nil)
+	}
+}
diff --git a/src/internal/bytealg/compare_arm64.s b/src/internal/bytealg/compare_arm64.s
index 32e2ba200d..56d56f241e 100644
--- a/src/internal/bytealg/compare_arm64.s
+++ b/src/internal/bytealg/compare_arm64.s
@@ -36,8 +36,7 @@ TEXT cmpbody<>(SB),NOSPLIT|NOFRAME,$0-0
 	CMP	R0, R1
 	CSEL	LT, R1, R0, R6    // R6 is min(R0, R1)
 
-	CMP	$0, R6
-	BEQ	samebytes
+	CBZ	R6, samebytes
 	BIC	$0xf, R6, R10
 	CBZ	R10, small        // length < 16
 	ADD	R2, R10           // end of chunk16
diff --git a/src/internal/bytealg/index_amd64.s b/src/internal/bytealg/index_amd64.s
index 4459820801..6193b57239 100644
--- a/src/internal/bytealg/index_amd64.s
+++ b/src/internal/bytealg/index_amd64.s
@@ -8,7 +8,7 @@
 TEXT ·Index(SB),NOSPLIT,$0-56
 	MOVQ a_base+0(FP), DI
 	MOVQ a_len+8(FP), DX
-	MOVQ b_base+24(FP), BP
+	MOVQ b_base+24(FP), R8
 	MOVQ b_len+32(FP), AX
 	MOVQ DI, R10
 	LEAQ ret+48(FP), R11
@@ -17,7 +17,7 @@ TEXT ·Index(SB),NOSPLIT,$0-56
 TEXT ·IndexString(SB),NOSPLIT,$0-40
 	MOVQ a_base+0(FP), DI
 	MOVQ a_len+8(FP), DX
-	MOVQ b_base+16(FP), BP
+	MOVQ b_base+16(FP), R8
 	MOVQ b_len+24(FP), AX
 	MOVQ DI, R10
 	LEAQ ret+32(FP), R11
@@ -26,7 +26,7 @@ TEXT ·IndexString(SB),NOSPLIT,$0-40
 // AX: length of string, that we are searching for
 // DX: length of string, in which we are searching
 // DI: pointer to string, in which we are searching
-// BP: pointer to string, that we are searching for
+// R8: pointer to string, that we are searching for
 // R11: address, where to put return value
 // Note: We want len in DX and AX, because PCMPESTRI implicitly consumes them
 TEXT indexbody<>(SB),NOSPLIT,$0
@@ -37,11 +37,11 @@ TEXT indexbody<>(SB),NOSPLIT,$0
 no_sse42:
 	CMPQ AX, $2
 	JA   _3_or_more
-	MOVW (BP), BP
+	MOVW (R8), R8
 	LEAQ -1(DI)(DX*1), DX
 loop2:
 	MOVW (DI), SI
-	CMPW SI,BP
+	CMPW SI,R8
 	JZ success
 	ADDQ $1,DI
 	CMPQ DI,DX
@@ -50,12 +50,12 @@ loop2:
 _3_or_more:
 	CMPQ AX, $3
 	JA   _4_or_more
-	MOVW 1(BP), BX
-	MOVW (BP), BP
+	MOVW 1(R8), BX
+	MOVW (R8), R8
 	LEAQ -2(DI)(DX*1), DX
 loop3:
 	MOVW (DI), SI
-	CMPW SI,BP
+	CMPW SI,R8
 	JZ   partial_success3
 	ADDQ $1,DI
 	CMPQ DI,DX
@@ -72,11 +72,11 @@ partial_success3:
 _4_or_more:
 	CMPQ AX, $4
 	JA   _5_or_more
-	MOVL (BP), BP
+	MOVL (R8), R8
 	LEAQ -3(DI)(DX*1), DX
 loop4:
 	MOVL (DI), SI
-	CMPL SI,BP
+	CMPL SI,R8
 	JZ   success
 	ADDQ $1,DI
 	CMPQ DI,DX
@@ -87,11 +87,11 @@ _5_or_more:
 	JA   _8_or_more
 	LEAQ 1(DI)(DX*1), DX
 	SUBQ AX, DX
-	MOVL -4(BP)(AX*1), BX
-	MOVL (BP), BP
+	MOVL -4(R8)(AX*1), BX
+	MOVL (R8), R8
 loop5to7:
 	MOVL (DI), SI
-	CMPL SI,BP
+	CMPL SI,R8
 	JZ   partial_success5to7
 	ADDQ $1,DI
 	CMPQ DI,DX
@@ -108,11 +108,11 @@ partial_success5to7:
 _8_or_more:
 	CMPQ AX, $8
 	JA   _9_or_more
-	MOVQ (BP), BP
+	MOVQ (R8), R8
 	LEAQ -7(DI)(DX*1), DX
 loop8:
 	MOVQ (DI), SI
-	CMPQ SI,BP
+	CMPQ SI,R8
 	JZ   success
 	ADDQ $1,DI
 	CMPQ DI,DX
@@ -123,11 +123,11 @@ _9_or_more:
 	JA   _16_or_more
 	LEAQ 1(DI)(DX*1), DX
 	SUBQ AX, DX
-	MOVQ -8(BP)(AX*1), BX
-	MOVQ (BP), BP
+	MOVQ -8(R8)(AX*1), BX
+	MOVQ (R8), R8
 loop9to15:
 	MOVQ (DI), SI
-	CMPQ SI,BP
+	CMPQ SI,R8
 	JZ   partial_success9to15
 	ADDQ $1,DI
 	CMPQ DI,DX
@@ -144,7 +144,7 @@ partial_success9to15:
 _16_or_more:
 	CMPQ AX, $16
 	JA   _17_or_more
-	MOVOU (BP), X1
+	MOVOU (R8), X1
 	LEAQ -15(DI)(DX*1), DX
 loop16:
 	MOVOU (DI), X2
@@ -161,8 +161,8 @@ _17_or_more:
 	JA   _32_or_more
 	LEAQ 1(DI)(DX*1), DX
 	SUBQ AX, DX
-	MOVOU -16(BP)(AX*1), X0
-	MOVOU (BP), X1
+	MOVOU -16(R8)(AX*1), X0
+	MOVOU (R8), X1
 loop17to31:
 	MOVOU (DI), X2
 	PCMPEQB X1,X2
@@ -188,7 +188,7 @@ partial_success17to31:
 _32_or_more:
 	CMPQ AX, $32
 	JA   _33_to_63
-	VMOVDQU (BP), Y1
+	VMOVDQU (R8), Y1
 	LEAQ -31(DI)(DX*1), DX
 loop32:
 	VMOVDQU (DI), Y2
@@ -203,8 +203,8 @@ loop32:
 _33_to_63:
 	LEAQ 1(DI)(DX*1), DX
 	SUBQ AX, DX
-	VMOVDQU -32(BP)(AX*1), Y0
-	VMOVDQU (BP), Y1
+	VMOVDQU -32(R8)(AX*1), Y0
+	VMOVDQU (R8), Y1
 loop33to63:
 	VMOVDQU (DI), Y2
 	VPCMPEQB Y1, Y2, Y3
@@ -241,10 +241,10 @@ sse42:
 	// This value was determined experimentally and is the ~same
 	// on Nehalem (first with SSE42) and Haswell.
 	JAE _9_or_more
-	LEAQ 16(BP), SI
+	LEAQ 16(R8), SI
 	TESTW $0xff0, SI
 	JEQ no_sse42
-	MOVOU (BP), X1
+	MOVOU (R8), X1
 	LEAQ -15(DI)(DX*1), SI
 	MOVQ $16, R9
 	SUBQ AX, R9 // We advance by 16-len(sep) each iteration, so precalculate it into R9
diff --git a/src/internal/poll/copy_file_range_linux.go b/src/internal/poll/copy_file_range_linux.go
index 604607f774..09de299ff7 100644
--- a/src/internal/poll/copy_file_range_linux.go
+++ b/src/internal/poll/copy_file_range_linux.go
@@ -41,7 +41,7 @@ func CopyFileRange(dst, src *FD, remain int64) (written int64, handled bool, err
 			// use copy_file_range(2) again.
 			atomic.StoreInt32(&copyFileRangeSupported, 0)
 			return 0, false, nil
-		case syscall.EXDEV, syscall.EINVAL:
+		case syscall.EXDEV, syscall.EINVAL, syscall.EOPNOTSUPP, syscall.EPERM:
 			// Prior to Linux 5.3, it was not possible to
 			// copy_file_range across file systems. Similarly to
 			// the ENOSYS case above, if we see EXDEV, we have
@@ -52,6 +52,14 @@ func CopyFileRange(dst, src *FD, remain int64) (written int64, handled bool, err
 			// dst or src refer to a pipe rather than a regular
 			// file. This is another case where no data has been
 			// transfered, so we consider it unhandled.
+			//
+			// If the file is on NFS, we can see EOPNOTSUPP.
+			// See issue #40731.
+			//
+			// If the process is running inside a Docker container,
+			// we might see EPERM instead of ENOSYS. See issue
+			// #40893. Since EPERM might also be a legitimate error,
+			// don't mark copy_file_range(2) as unsupported.
 			return 0, false, nil
 		case nil:
 			if n == 0 {
diff --git a/src/internal/poll/fd_fsync_posix.go b/src/internal/poll/fd_fsync_posix.go
index 69358297f4..dd7956f14d 100644
--- a/src/internal/poll/fd_fsync_posix.go
+++ b/src/internal/poll/fd_fsync_posix.go
@@ -14,5 +14,7 @@ func (fd *FD) Fsync() error {
 		return err
 	}
 	defer fd.decref()
-	return syscall.Fsync(fd.Sysfd)
+	return ignoringEINTR(func() error {
+		return syscall.Fsync(fd.Sysfd)
+	})
 }
diff --git a/src/internal/poll/fd_opendir_darwin.go b/src/internal/poll/fd_opendir_darwin.go
index c7d3318c72..8eb770c358 100644
--- a/src/internal/poll/fd_opendir_darwin.go
+++ b/src/internal/poll/fd_opendir_darwin.go
@@ -19,7 +19,13 @@ func (fd *FD) OpenDir() (uintptr, string, error) {
 	if err != nil {
 		return 0, call, err
 	}
-	dir, err := fdopendir(fd2)
+	var dir uintptr
+	for {
+		dir, err = fdopendir(fd2)
+		if err != syscall.EINTR {
+			break
+		}
+	}
 	if err != nil {
 		syscall.Close(fd2)
 		return 0, "fdopendir", err
diff --git a/src/internal/poll/fd_posix.go b/src/internal/poll/fd_posix.go
index 54747b4c99..e5fb05c9c2 100644
--- a/src/internal/poll/fd_posix.go
+++ b/src/internal/poll/fd_posix.go
@@ -35,7 +35,9 @@ func (fd *FD) Fchmod(mode uint32) error {
 		return err
 	}
 	defer fd.decref()
-	return syscall.Fchmod(fd.Sysfd, mode)
+	return ignoringEINTR(func() error {
+		return syscall.Fchmod(fd.Sysfd, mode)
+	})
 }
 
 // Fchown wraps syscall.Fchown.
@@ -44,7 +46,9 @@ func (fd *FD) Fchown(uid, gid int) error {
 		return err
 	}
 	defer fd.decref()
-	return syscall.Fchown(fd.Sysfd, uid, gid)
+	return ignoringEINTR(func() error {
+		return syscall.Fchown(fd.Sysfd, uid, gid)
+	})
 }
 
 // Ftruncate wraps syscall.Ftruncate.
@@ -53,7 +57,9 @@ func (fd *FD) Ftruncate(size int64) error {
 		return err
 	}
 	defer fd.decref()
-	return syscall.Ftruncate(fd.Sysfd, size)
+	return ignoringEINTR(func() error {
+		return syscall.Ftruncate(fd.Sysfd, size)
+	})
 }
 
 // RawControl invokes the user-defined function f for a non-IO
@@ -66,3 +72,19 @@ func (fd *FD) RawControl(f func(uintptr)) error {
 	f(uintptr(fd.Sysfd))
 	return nil
 }
+
+// ignoringEINTR makes a function call and repeats it if it returns
+// an EINTR error. This appears to be required even though we install all
+// signal handlers with SA_RESTART: see #22838, #38033, #38836, #40846.
+// Also #20400 and #36644 are issues in which a signal handler is
+// installed without setting SA_RESTART. None of these are the common case,
+// but there are enough of them that it seems that we can't avoid
+// an EINTR loop.
+func ignoringEINTR(fn func() error) error {
+	for {
+		err := fn()
+		if err != syscall.EINTR {
+			return err
+		}
+	}
+}
diff --git a/src/internal/poll/fd_unix.go b/src/internal/poll/fd_unix.go
index 4872fa9851..1d5101eac3 100644
--- a/src/internal/poll/fd_unix.go
+++ b/src/internal/poll/fd_unix.go
@@ -152,7 +152,7 @@ func (fd *FD) Read(p []byte) (int, error) {
 		p = p[:maxRW]
 	}
 	for {
-		n, err := ignoringEINTR(syscall.Read, fd.Sysfd, p)
+		n, err := ignoringEINTRIO(syscall.Read, fd.Sysfd, p)
 		if err != nil {
 			n = 0
 			if err == syscall.EAGAIN && fd.pd.pollable() {
@@ -264,7 +264,7 @@ func (fd *FD) Write(p []byte) (int, error) {
 		if fd.IsStream && max-nn > maxRW {
 			max = nn + maxRW
 		}
-		n, err := ignoringEINTR(syscall.Write, fd.Sysfd, p[nn:max])
+		n, err := ignoringEINTRIO(syscall.Write, fd.Sysfd, p[nn:max])
 		if n > 0 {
 			nn += n
 		}
@@ -423,7 +423,7 @@ func (fd *FD) ReadDirent(buf []byte) (int, error) {
 	}
 	defer fd.decref()
 	for {
-		n, err := ignoringEINTR(syscall.ReadDirent, fd.Sysfd, buf)
+		n, err := ignoringEINTRIO(syscall.ReadDirent, fd.Sysfd, buf)
 		if err != nil {
 			n = 0
 			if err == syscall.EAGAIN && fd.pd.pollable() {
@@ -452,7 +452,9 @@ func (fd *FD) Fstat(s *syscall.Stat_t) error {
 		return err
 	}
 	defer fd.decref()
-	return syscall.Fstat(fd.Sysfd, s)
+	return ignoringEINTR(func() error {
+		return syscall.Fstat(fd.Sysfd, s)
+	})
 }
 
 // tryDupCloexec indicates whether F_DUPFD_CLOEXEC should be used.
@@ -514,7 +516,7 @@ func (fd *FD) WriteOnce(p []byte) (int, error) {
 		return 0, err
 	}
 	defer fd.writeUnlock()
-	return ignoringEINTR(syscall.Write, fd.Sysfd, p)
+	return ignoringEINTRIO(syscall.Write, fd.Sysfd, p)
 }
 
 // RawRead invokes the user-defined function f for a read operation.
@@ -555,14 +557,8 @@ func (fd *FD) RawWrite(f func(uintptr) bool) error {
 	}
 }
 
-// ignoringEINTR makes a function call and repeats it if it returns
-// an EINTR error. This appears to be required even though we install
-// all signal handlers with SA_RESTART: see #22838, #38033, #38836.
-// Also #20400 and #36644 are issues in which a signal handler is
-// installed without setting SA_RESTART. None of these are the common case,
-// but there are enough of them that it seems that we can't avoid
-// an EINTR loop.
-func ignoringEINTR(fn func(fd int, p []byte) (int, error), fd int, p []byte) (int, error) {
+// ignoringEINTRIO is like ignoringEINTR, but just for IO calls.
+func ignoringEINTRIO(fn func(fd int, p []byte) (int, error), fd int, p []byte) (int, error) {
 	for {
 		n, err := fn(fd, p)
 		if err != syscall.EINTR {
diff --git a/src/math/big/arith_ppc64x.s b/src/math/big/arith_ppc64x.s
index dbb168a376..409e10ab48 100644
--- a/src/math/big/arith_ppc64x.s
+++ b/src/math/big/arith_ppc64x.s
@@ -394,17 +394,16 @@ loop:
 	ADDZE   R21
 	MULLD   R9, R22, R26
 	MULHDU  R9, R22, R22
-	ADDC    R21, R26
-	ADDZE   R22
 	MULLD   R9, R23, R27
 	MULHDU  R9, R23, R23
-	ADDC    R22, R27
-	ADDZE   R23
+	ADDC    R21, R26
+	ADDZE   R22
 	MOVD    R24, 8(R10)       // z[i]
 	MOVD    R25, 16(R10)      // z[i+1]
+	ADDC    R22, R27
+	ADDZE   R23,R4		  // update carry
 	MOVD    R26, 24(R10)      // z[i+2]
 	MOVDU   R27, 32(R10)      // z[i+3]
-	MOVD    R23, R4           // R4 = c
 	ADD     $-4, R11          // R11 = z_len - 4
 	BC      16, 0, loop       // bdnz
 
diff --git a/src/math/log1p.go b/src/math/log1p.go
index c4ec61b225..e34e1ff4f2 100644
--- a/src/math/log1p.go
+++ b/src/math/log1p.go
@@ -122,10 +122,7 @@ func log1p(x float64) float64 {
 		return Inf(1)
 	}
 
-	absx := x
-	if absx < 0 {
-		absx = -absx
-	}
+	absx := Abs(x)
 
 	var f float64
 	var iu uint64
diff --git a/src/net/http/client.go b/src/net/http/client.go
index 3860d97d8f..6ca0d2e6cf 100644
--- a/src/net/http/client.go
+++ b/src/net/http/client.go
@@ -321,7 +321,7 @@ func knownRoundTripperImpl(rt RoundTripper, req *Request) bool {
 		return true
 	}
 	// There's a very minor chance of a false positive with this.
-	// Insted of detecting our golang.org/x/net/http2.Transport,
+	// Instead of detecting our golang.org/x/net/http2.Transport,
 	// it might detect a Transport type in a different http2
 	// package. But I know of none, and the only problem would be
 	// some temporarily leaked goroutines if the transport didn't
diff --git a/src/net/http/roundtrip_js.go b/src/net/http/roundtrip_js.go
index 509d229aad..b09923c386 100644
--- a/src/net/http/roundtrip_js.go
+++ b/src/net/http/roundtrip_js.go
@@ -98,9 +98,11 @@ func (t *Transport) RoundTrip(req *Request) (*Response, error) {
 			return nil, err
 		}
 		req.Body.Close()
-		buf := uint8Array.New(len(body))
-		js.CopyBytesToJS(buf, body)
-		opt.Set("body", buf)
+		if len(body) != 0 {
+			buf := uint8Array.New(len(body))
+			js.CopyBytesToJS(buf, body)
+			opt.Set("body", buf)
+		}
 	}
 
 	fetchPromise := js.Global().Call("fetch", req.URL.String(), opt)
diff --git a/src/net/rawconn_unix_test.go b/src/net/rawconn_unix_test.go
index a71b6f3041..0194ba67c8 100644
--- a/src/net/rawconn_unix_test.go
+++ b/src/net/rawconn_unix_test.go
@@ -24,10 +24,7 @@ func readRawConn(c syscall.RawConn, b []byte) (int, error) {
 	if err != nil {
 		return n, err
 	}
-	if operr != nil {
-		return n, operr
-	}
-	return n, nil
+	return n, operr
 }
 
 func writeRawConn(c syscall.RawConn, b []byte) error {
@@ -42,10 +39,7 @@ func writeRawConn(c syscall.RawConn, b []byte) error {
 	if err != nil {
 		return err
 	}
-	if operr != nil {
-		return operr
-	}
-	return nil
+	return operr
 }
 
 func controlRawConn(c syscall.RawConn, addr Addr) error {
@@ -87,10 +81,7 @@ func controlRawConn(c syscall.RawConn, addr Addr) error {
 	if err := c.Control(fn); err != nil {
 		return err
 	}
-	if operr != nil {
-		return operr
-	}
-	return nil
+	return operr
 }
 
 func controlOnConnSetup(network string, address string, c syscall.RawConn) error {
@@ -120,8 +111,5 @@ func controlOnConnSetup(network string, address string, c syscall.RawConn) error
 	if err := c.Control(fn); err != nil {
 		return err
 	}
-	if operr != nil {
-		return operr
-	}
-	return nil
+	return operr
 }
diff --git a/src/net/rawconn_windows_test.go b/src/net/rawconn_windows_test.go
index 2774c97e5c..5febf08f77 100644
--- a/src/net/rawconn_windows_test.go
+++ b/src/net/rawconn_windows_test.go
@@ -26,10 +26,7 @@ func readRawConn(c syscall.RawConn, b []byte) (int, error) {
 	if err != nil {
 		return n, err
 	}
-	if operr != nil {
-		return n, operr
-	}
-	return n, nil
+	return n, operr
 }
 
 func writeRawConn(c syscall.RawConn, b []byte) error {
@@ -45,10 +42,7 @@ func writeRawConn(c syscall.RawConn, b []byte) error {
 	if err != nil {
 		return err
 	}
-	if operr != nil {
-		return operr
-	}
-	return nil
+	return operr
 }
 
 func controlRawConn(c syscall.RawConn, addr Addr) error {
@@ -92,10 +86,7 @@ func controlRawConn(c syscall.RawConn, addr Addr) error {
 	if err := c.Control(fn); err != nil {
 		return err
 	}
-	if operr != nil {
-		return operr
-	}
-	return nil
+	return operr
 }
 
 func controlOnConnSetup(network string, address string, c syscall.RawConn) error {
@@ -121,8 +112,5 @@ func controlOnConnSetup(network string, address string, c syscall.RawConn) error
 	if err := c.Control(fn); err != nil {
 		return err
 	}
-	if operr != nil {
-		return operr
-	}
-	return nil
+	return operr
 }
diff --git a/src/os/dir_darwin.go b/src/os/dir_darwin.go
index 2f9ba78d68..87797e2dda 100644
--- a/src/os/dir_darwin.go
+++ b/src/os/dir_darwin.go
@@ -47,6 +47,9 @@ func (f *File) readdirnames(n int) (names []string, err error) {
 	var entptr *syscall.Dirent
 	for len(names) < size || n == -1 {
 		if res := readdir_r(d.dir, &dirent, &entptr); res != 0 {
+			if syscall.Errno(res) == syscall.EINTR {
+				continue
+			}
 			return names, wrapSyscallError("readdir", syscall.Errno(res))
 		}
 		if entptr == nil { // EOF
diff --git a/src/os/file.go b/src/os/file.go
index a2b71cb61a..05d2f83283 100644
--- a/src/os/file.go
+++ b/src/os/file.go
@@ -255,7 +255,10 @@ func Mkdir(name string, perm FileMode) error {
 	if runtime.GOOS == "windows" && isWindowsNulName(name) {
 		return &PathError{"mkdir", name, syscall.ENOTDIR}
 	}
-	e := syscall.Mkdir(fixLongPath(name), syscallMode(perm))
+	longName := fixLongPath(name)
+	e := ignoringEINTR(func() error {
+		return syscall.Mkdir(longName, syscallMode(perm))
+	})
 
 	if e != nil {
 		return &PathError{"mkdir", name, e}
diff --git a/src/os/file_plan9.go b/src/os/file_plan9.go
index eb158905ab..043500744b 100644
--- a/src/os/file_plan9.go
+++ b/src/os/file_plan9.go
@@ -558,3 +558,7 @@ func (c *rawConn) Write(f func(uintptr) bool) error {
 func newRawConn(file *File) (*rawConn, error) {
 	return nil, syscall.EPLAN9
 }
+
+func ignoringEINTR(fn func() error) error {
+	return fn()
+}
diff --git a/src/os/file_posix.go b/src/os/file_posix.go
index 24ea554b62..ae23d22d0a 100644
--- a/src/os/file_posix.go
+++ b/src/os/file_posix.go
@@ -76,7 +76,11 @@ func syscallMode(i FileMode) (o uint32) {
 
 // See docs in file.go:Chmod.
 func chmod(name string, mode FileMode) error {
-	if e := syscall.Chmod(fixLongPath(name), syscallMode(mode)); e != nil {
+	longName := fixLongPath(name)
+	e := ignoringEINTR(func() error {
+		return syscall.Chmod(longName, syscallMode(mode))
+	})
+	if e != nil {
 		return &PathError{"chmod", name, e}
 	}
 	return nil
@@ -101,7 +105,10 @@ func (f *File) chmod(mode FileMode) error {
 // On Windows or Plan 9, Chown always returns the syscall.EWINDOWS or
 // EPLAN9 error, wrapped in *PathError.
 func Chown(name string, uid, gid int) error {
-	if e := syscall.Chown(name, uid, gid); e != nil {
+	e := ignoringEINTR(func() error {
+		return syscall.Chown(name, uid, gid)
+	})
+	if e != nil {
 		return &PathError{"chown", name, e}
 	}
 	return nil
@@ -114,7 +121,10 @@ func Chown(name string, uid, gid int) error {
 // On Windows, it always returns the syscall.EWINDOWS error, wrapped
 // in *PathError.
 func Lchown(name string, uid, gid int) error {
-	if e := syscall.Lchown(name, uid, gid); e != nil {
+	e := ignoringEINTR(func() error {
+		return syscall.Lchown(name, uid, gid)
+	})
+	if e != nil {
 		return &PathError{"lchown", name, e}
 	}
 	return nil
@@ -222,3 +232,19 @@ func (f *File) checkValid(op string) error {
 	}
 	return nil
 }
+
+// ignoringEINTR makes a function call and repeats it if it returns an
+// EINTR error. This appears to be required even though we install all
+// signal handlers with SA_RESTART: see #22838, #38033, #38836, #40846.
+// Also #20400 and #36644 are issues in which a signal handler is
+// installed without setting SA_RESTART. None of these are the common case,
+// but there are enough of them that it seems that we can't avoid
+// an EINTR loop.
+func ignoringEINTR(fn func() error) error {
+	for {
+		err := fn()
+		if err != syscall.EINTR {
+			return err
+		}
+	}
+}
diff --git a/src/os/file_unix.go b/src/os/file_unix.go
index f2c00ae0cb..5446dd5003 100644
--- a/src/os/file_unix.go
+++ b/src/os/file_unix.go
@@ -39,7 +39,9 @@ func rename(oldname, newname string) error {
 			return &LinkError{"rename", oldname, newname, syscall.EEXIST}
 		}
 	}
-	err = syscall.Rename(oldname, newname)
+	err = ignoringEINTR(func() error {
+		return syscall.Rename(oldname, newname)
+	})
 	if err != nil {
 		return &LinkError{"rename", oldname, newname, err}
 	}
@@ -129,7 +131,9 @@ func newFile(fd uintptr, name string, kind newFileKind) *File {
 		switch runtime.GOOS {
 		case "darwin", "dragonfly", "freebsd", "netbsd", "openbsd":
 			var st syscall.Stat_t
-			err := syscall.Fstat(fdi, &st)
+			err := ignoringEINTR(func() error {
+				return syscall.Fstat(fdi, &st)
+			})
 			typ := st.Mode & syscall.S_IFMT
 			// Don't try to use kqueue with regular files on *BSDs.
 			// On FreeBSD a regular file is always
@@ -264,7 +268,10 @@ func (f *File) seek(offset int64, whence int) (ret int64, err error) {
 // If the file is a symbolic link, it changes the size of the link's target.
 // If there is an error, it will be of type *PathError.
 func Truncate(name string, size int64) error {
-	if e := syscall.Truncate(name, size); e != nil {
+	e := ignoringEINTR(func() error {
+		return syscall.Truncate(name, size)
+	})
+	if e != nil {
 		return &PathError{"truncate", name, e}
 	}
 	return nil
@@ -277,11 +284,15 @@ func Remove(name string) error {
 	// whether name is a file or directory.
 	// Try both: it is cheaper on average than
 	// doing a Stat plus the right one.
-	e := syscall.Unlink(name)
+	e := ignoringEINTR(func() error {
+		return syscall.Unlink(name)
+	})
 	if e == nil {
 		return nil
 	}
-	e1 := syscall.Rmdir(name)
+	e1 := ignoringEINTR(func() error {
+		return syscall.Rmdir(name)
+	})
 	if e1 == nil {
 		return nil
 	}
@@ -316,7 +327,9 @@ func tempDir() string {
 // Link creates newname as a hard link to the oldname file.
 // If there is an error, it will be of type *LinkError.
 func Link(oldname, newname string) error {
-	e := syscall.Link(oldname, newname)
+	e := ignoringEINTR(func() error {
+		return syscall.Link(oldname, newname)
+	})
 	if e != nil {
 		return &LinkError{"link", oldname, newname, e}
 	}
@@ -326,7 +339,9 @@ func Link(oldname, newname string) error {
 // Symlink creates newname as a symbolic link to oldname.
 // If there is an error, it will be of type *LinkError.
 func Symlink(oldname, newname string) error {
-	e := syscall.Symlink(oldname, newname)
+	e := ignoringEINTR(func() error {
+		return syscall.Symlink(oldname, newname)
+	})
 	if e != nil {
 		return &LinkError{"symlink", oldname, newname, e}
 	}
@@ -365,7 +380,16 @@ func (f *File) readdir(n int) (fi []FileInfo, err error) {
 func Readlink(name string) (string, error) {
 	for len := 128; ; len *= 2 {
 		b := make([]byte, len)
-		n, e := fixCount(syscall.Readlink(name, b))
+		var (
+			n int
+			e error
+		)
+		for {
+			n, e = fixCount(syscall.Readlink(name, b))
+			if e != syscall.EINTR {
+				break
+			}
+		}
 		// buffer too small
 		if runtime.GOOS == "aix" && e == syscall.ERANGE {
 			continue
diff --git a/src/os/getwd.go b/src/os/getwd.go
index 6d25466bb4..f3afd8c06c 100644
--- a/src/os/getwd.go
+++ b/src/os/getwd.go
@@ -45,7 +45,16 @@ func Getwd() (dir string, err error) {
 	// If the operating system provides a Getwd call, use it.
 	// Otherwise, we're trying to find our way back to ".".
 	if syscall.ImplementsGetwd {
-		s, e := syscall.Getwd()
+		var (
+			s string
+			e error
+		)
+		for {
+			s, e = syscall.Getwd()
+			if e != syscall.EINTR {
+				break
+			}
+		}
 		if useSyscallwd(e) {
 			return s, NewSyscallError("getwd", e)
 		}
diff --git a/src/os/stat_unix.go b/src/os/stat_unix.go
index 0a7e6029ac..ef74a43758 100644
--- a/src/os/stat_unix.go
+++ b/src/os/stat_unix.go
@@ -28,7 +28,9 @@ func (f *File) Stat() (FileInfo, error) {
 // statNolog stats a file with no test logging.
 func statNolog(name string) (FileInfo, error) {
 	var fs fileStat
-	err := syscall.Stat(name, &fs.sys)
+	err := ignoringEINTR(func() error {
+		return syscall.Stat(name, &fs.sys)
+	})
 	if err != nil {
 		return nil, &PathError{"stat", name, err}
 	}
@@ -39,7 +41,9 @@ func statNolog(name string) (FileInfo, error) {
 // lstatNolog lstats a file with no test logging.
 func lstatNolog(name string) (FileInfo, error) {
 	var fs fileStat
-	err := syscall.Lstat(name, &fs.sys)
+	err := ignoringEINTR(func() error {
+		return syscall.Lstat(name, &fs.sys)
+	})
 	if err != nil {
 		return nil, &PathError{"lstat", name, err}
 	}
diff --git a/src/path/example_test.go b/src/path/example_test.go
index 67b9718664..e30ebd13dc 100644
--- a/src/path/example_test.go
+++ b/src/path/example_test.go
@@ -79,13 +79,18 @@ func ExampleJoin() {
 	fmt.Println(path.Join("a", "b", "c"))
 	fmt.Println(path.Join("a", "b/c"))
 	fmt.Println(path.Join("a/b", "c"))
+
+	fmt.Println(path.Join("a/b", "../../../xyz"))
+
 	fmt.Println(path.Join("", ""))
 	fmt.Println(path.Join("a", ""))
 	fmt.Println(path.Join("", "a"))
+
 	// Output:
 	// a/b/c
 	// a/b/c
 	// a/b/c
+	// ../xyz
 	//
 	// a
 	// a
diff --git a/src/path/filepath/example_unix_test.go b/src/path/filepath/example_unix_test.go
index 23f21380d0..c9d6944518 100644
--- a/src/path/filepath/example_unix_test.go
+++ b/src/path/filepath/example_unix_test.go
@@ -72,12 +72,16 @@ func ExampleJoin() {
 	fmt.Println(filepath.Join("a", "b/c"))
 	fmt.Println(filepath.Join("a/b", "c"))
 	fmt.Println(filepath.Join("a/b", "/c"))
+
+	fmt.Println(filepath.Join("a/b", "../../../xyz"))
+
 	// Output:
 	// On Unix:
 	// a/b/c
 	// a/b/c
 	// a/b/c
 	// a/b/c
+	// ../xyz
 }
 
 func ExampleMatch() {
diff --git a/src/reflect/all_test.go b/src/reflect/all_test.go
index 63f6a92157..5a12699472 100644
--- a/src/reflect/all_test.go
+++ b/src/reflect/all_test.go
@@ -74,6 +74,10 @@ var typeTests = []pair{
 	{struct{ x ([]int8) }{}, "[]int8"},
 	{struct{ x (map[string]int32) }{}, "map[string]int32"},
 	{struct{ x (chan<- string) }{}, "chan<- string"},
+	{struct{ x (chan<- chan string) }{}, "chan<- chan string"},
+	{struct{ x (chan<- <-chan string) }{}, "chan<- <-chan string"},
+	{struct{ x (<-chan <-chan string) }{}, "<-chan <-chan string"},
+	{struct{ x (chan (<-chan string)) }{}, "chan (<-chan string)"},
 	{struct {
 		x struct {
 			c chan *int32
@@ -1721,6 +1725,14 @@ func TestSelectMaxCases(t *testing.T) {
 	_, _, _ = Select(sCases)
 }
 
+func TestSelectNop(t *testing.T) {
+	// "select { default: }" should always return the default case.
+	chosen, _, _ := Select([]SelectCase{{Dir: SelectDefault}})
+	if chosen != 0 {
+		t.Fatalf("expected Select to return 0, but got %#v", chosen)
+	}
+}
+
 func BenchmarkSelect(b *testing.B) {
 	channel := make(chan int)
 	close(channel)
@@ -5491,6 +5503,18 @@ func TestChanOf(t *testing.T) {
 	// check that type already in binary is found
 	type T1 int
 	checkSameType(t, ChanOf(BothDir, TypeOf(T1(1))), (chan T1)(nil))
+
+	// Check arrow token association in undefined chan types.
+	var left chan<- chan T
+	var right chan (<-chan T)
+	tLeft := ChanOf(SendDir, ChanOf(BothDir, TypeOf(T(""))))
+	tRight := ChanOf(BothDir, ChanOf(RecvDir, TypeOf(T(""))))
+	if tLeft != TypeOf(left) {
+		t.Errorf("chan<-chan: have %s, want %T", tLeft, left)
+	}
+	if tRight != TypeOf(right) {
+		t.Errorf("chan<-chan: have %s, want %T", tRight, right)
+	}
 }
 
 func TestChanOfDir(t *testing.T) {
@@ -6451,12 +6475,9 @@ func verifyGCBitsSlice(t *testing.T, typ Type, cap int, bits []byte) {
 	// Repeat the bitmap for the slice size, trimming scalars in
 	// the last element.
 	bits = rep(cap, bits)
-	for len(bits) > 2 && bits[len(bits)-1] == 0 {
+	for len(bits) > 0 && bits[len(bits)-1] == 0 {
 		bits = bits[:len(bits)-1]
 	}
-	if len(bits) == 2 && bits[0] == 0 && bits[1] == 0 {
-		bits = bits[:0]
-	}
 	if !bytes.Equal(heapBits, bits) {
 		t.Errorf("heapBits incorrect for make(%v, 0, %v)\nhave %v\nwant %v", typ, cap, heapBits, bits)
 	}
diff --git a/src/reflect/deepequal.go b/src/reflect/deepequal.go
index 8a2bf8b09e..be66464129 100644
--- a/src/reflect/deepequal.go
+++ b/src/reflect/deepequal.go
@@ -21,7 +21,7 @@ type visit struct {
 // Tests for deep equality using reflected types. The map argument tracks
 // comparisons that have already been seen, which allows short circuiting on
 // recursive types.
-func deepValueEqual(v1, v2 Value, visited map[visit]bool, depth int) bool {
+func deepValueEqual(v1, v2 Value, visited map[visit]bool) bool {
 	if !v1.IsValid() || !v2.IsValid() {
 		return v1.IsValid() == v2.IsValid()
 	}
@@ -29,8 +29,6 @@ func deepValueEqual(v1, v2 Value, visited map[visit]bool, depth int) bool {
 		return false
 	}
 
-	// if depth > 10 { panic("deepValueEqual") }	// for debugging
-
 	// We want to avoid putting more in the visited map than we need to.
 	// For any possible reference cycle that might be encountered,
 	// hard(v1, v2) needs to return true for at least one of the types in the cycle,
@@ -79,7 +77,7 @@ func deepValueEqual(v1, v2 Value, visited map[visit]bool, depth int) bool {
 	switch v1.Kind() {
 	case Array:
 		for i := 0; i < v1.Len(); i++ {
-			if !deepValueEqual(v1.Index(i), v2.Index(i), visited, depth+1) {
+			if !deepValueEqual(v1.Index(i), v2.Index(i), visited) {
 				return false
 			}
 		}
@@ -95,7 +93,7 @@ func deepValueEqual(v1, v2 Value, visited map[visit]bool, depth int) bool {
 			return true
 		}
 		for i := 0; i < v1.Len(); i++ {
-			if !deepValueEqual(v1.Index(i), v2.Index(i), visited, depth+1) {
+			if !deepValueEqual(v1.Index(i), v2.Index(i), visited) {
 				return false
 			}
 		}
@@ -104,15 +102,15 @@ func deepValueEqual(v1, v2 Value, visited map[visit]bool, depth int) bool {
 		if v1.IsNil() || v2.IsNil() {
 			return v1.IsNil() == v2.IsNil()
 		}
-		return deepValueEqual(v1.Elem(), v2.Elem(), visited, depth+1)
+		return deepValueEqual(v1.Elem(), v2.Elem(), visited)
 	case Ptr:
 		if v1.Pointer() == v2.Pointer() {
 			return true
 		}
-		return deepValueEqual(v1.Elem(), v2.Elem(), visited, depth+1)
+		return deepValueEqual(v1.Elem(), v2.Elem(), visited)
 	case Struct:
 		for i, n := 0, v1.NumField(); i < n; i++ {
-			if !deepValueEqual(v1.Field(i), v2.Field(i), visited, depth+1) {
+			if !deepValueEqual(v1.Field(i), v2.Field(i), visited) {
 				return false
 			}
 		}
@@ -130,7 +128,7 @@ func deepValueEqual(v1, v2 Value, visited map[visit]bool, depth int) bool {
 		for _, k := range v1.MapKeys() {
 			val1 := v1.MapIndex(k)
 			val2 := v2.MapIndex(k)
-			if !val1.IsValid() || !val2.IsValid() || !deepValueEqual(val1, val2, visited, depth+1) {
+			if !val1.IsValid() || !val2.IsValid() || !deepValueEqual(val1, val2, visited) {
 				return false
 			}
 		}
@@ -207,5 +205,5 @@ func DeepEqual(x, y interface{}) bool {
 	if v1.Type() != v2.Type() {
 		return false
 	}
-	return deepValueEqual(v1, v2, make(map[visit]bool), 0)
+	return deepValueEqual(v1, v2, make(map[visit]bool))
 }
diff --git a/src/reflect/type.go b/src/reflect/type.go
index 38b1283d42..44c96fea82 100644
--- a/src/reflect/type.go
+++ b/src/reflect/type.go
@@ -1789,7 +1789,6 @@ func ChanOf(dir ChanDir, t Type) Type {
 	}
 
 	// Look in known types.
-	// TODO: Precedence when constructing string.
 	var s string
 	switch dir {
 	default:
@@ -1799,7 +1798,16 @@ func ChanOf(dir ChanDir, t Type) Type {
 	case RecvDir:
 		s = "<-chan " + typ.String()
 	case BothDir:
-		s = "chan " + typ.String()
+		typeStr := typ.String()
+		if typeStr[0] == '<' {
+			// typ is recv chan, need parentheses as "<-" associates with leftmost
+			// chan possible, see:
+			// * https://golang.org/ref/spec#Channel_types
+			// * https://github.com/golang/go/issues/39897
+			s = "chan (" + typeStr + ")"
+		} else {
+			s = "chan " + typeStr
+		}
 	}
 	for _, tt := range typesByString(s) {
 		ch := (*chanType)(unsafe.Pointer(tt))
diff --git a/src/runtime/asm.s b/src/runtime/asm.s
index 95a3424de2..27d8df9e06 100644
--- a/src/runtime/asm.s
+++ b/src/runtime/asm.s
@@ -11,24 +11,3 @@
 DATA runtime·no_pointers_stackmap+0x00(SB)/4, $2
 DATA runtime·no_pointers_stackmap+0x04(SB)/4, $0
 GLOBL runtime·no_pointers_stackmap(SB),RODATA, $8
-
-// NaCl requires that these skips be verifiable machine code.
-#ifdef GOARCH_amd64
-#define SKIP4 BYTE $0x90; BYTE $0x90; BYTE $0x90; BYTE $0x90
-#endif
-#ifdef GOARCH_386
-#define SKIP4 BYTE $0x90; BYTE $0x90; BYTE $0x90; BYTE $0x90
-#endif
-#ifdef GOARCH_wasm
-#define SKIP4 UNDEF; UNDEF; UNDEF; UNDEF
-#endif
-#ifndef SKIP4
-#define SKIP4 WORD $0
-#endif
-
-#define SKIP16 SKIP4; SKIP4; SKIP4; SKIP4
-#define SKIP64 SKIP16; SKIP16; SKIP16; SKIP16
-
-// This function must be sizeofSkipFunction bytes.
-TEXT runtime·skipPleaseUseCallersFrames(SB),NOSPLIT,$0-0
-	SKIP64; SKIP64; SKIP64; SKIP64
diff --git a/src/runtime/asm_ppc64x.s b/src/runtime/asm_ppc64x.s
index 11d2f2f51a..23387a2165 100644
--- a/src/runtime/asm_ppc64x.s
+++ b/src/runtime/asm_ppc64x.s
@@ -916,23 +916,23 @@ TEXT ·checkASM(SB),NOSPLIT,$0-1
 // - R20 is the destination of the write
 // - R21 is the value being written at R20.
 // It clobbers condition codes.
-// It does not clobber R0 through R15,
+// It does not clobber R0 through R17 (except special registers),
 // but may clobber any other register, *including* R31.
 TEXT runtime·gcWriteBarrier(SB),NOSPLIT,$112
 	// The standard prologue clobbers R31.
-	// We use R16 and R17 as scratch registers.
-	MOVD	g_m(g), R16
-	MOVD	m_p(R16), R16
-	MOVD	(p_wbBuf+wbBuf_next)(R16), R17
+	// We use R18 and R19 as scratch registers.
+	MOVD	g_m(g), R18
+	MOVD	m_p(R18), R18
+	MOVD	(p_wbBuf+wbBuf_next)(R18), R19
 	// Increment wbBuf.next position.
-	ADD	$16, R17
-	MOVD	R17, (p_wbBuf+wbBuf_next)(R16)
-	MOVD	(p_wbBuf+wbBuf_end)(R16), R16
-	CMP	R16, R17
+	ADD	$16, R19
+	MOVD	R19, (p_wbBuf+wbBuf_next)(R18)
+	MOVD	(p_wbBuf+wbBuf_end)(R18), R18
+	CMP	R18, R19
 	// Record the write.
-	MOVD	R21, -16(R17)	// Record value
-	MOVD	(R20), R16	// TODO: This turns bad writes into bad reads.
-	MOVD	R16, -8(R17)	// Record *slot
+	MOVD	R21, -16(R19)	// Record value
+	MOVD	(R20), R18	// TODO: This turns bad writes into bad reads.
+	MOVD	R18, -8(R19)	// Record *slot
 	// Is the buffer full? (flags set in CMP above)
 	BEQ	flush
 ret:
@@ -956,11 +956,12 @@ flush:
 	MOVD	R8, (FIXED_FRAME+56)(R1)
 	MOVD	R9, (FIXED_FRAME+64)(R1)
 	MOVD	R10, (FIXED_FRAME+72)(R1)
-	MOVD	R11, (FIXED_FRAME+80)(R1)
-	MOVD	R12, (FIXED_FRAME+88)(R1)
+	// R11, R12 may be clobbered by external-linker-inserted trampoline
 	// R13 is REGTLS
-	MOVD	R14, (FIXED_FRAME+96)(R1)
-	MOVD	R15, (FIXED_FRAME+104)(R1)
+	MOVD	R14, (FIXED_FRAME+80)(R1)
+	MOVD	R15, (FIXED_FRAME+88)(R1)
+	MOVD	R16, (FIXED_FRAME+96)(R1)
+	MOVD	R17, (FIXED_FRAME+104)(R1)
 
 	// This takes arguments R20 and R21.
 	CALL	runtime·wbBufFlush(SB)
@@ -975,10 +976,10 @@ flush:
 	MOVD	(FIXED_FRAME+56)(R1), R8
 	MOVD	(FIXED_FRAME+64)(R1), R9
 	MOVD	(FIXED_FRAME+72)(R1), R10
-	MOVD	(FIXED_FRAME+80)(R1), R11
-	MOVD	(FIXED_FRAME+88)(R1), R12
-	MOVD	(FIXED_FRAME+96)(R1), R14
-	MOVD	(FIXED_FRAME+104)(R1), R15
+	MOVD	(FIXED_FRAME+80)(R1), R14
+	MOVD	(FIXED_FRAME+88)(R1), R15
+	MOVD	(FIXED_FRAME+96)(R1), R16
+	MOVD	(FIXED_FRAME+104)(R1), R17
 	JMP	ret
 
 // Note: these functions use a special calling convention to save generated code space.
diff --git a/src/runtime/asm_riscv64.s b/src/runtime/asm_riscv64.s
index d7c45a183d..8f6c8773eb 100644
--- a/src/runtime/asm_riscv64.s
+++ b/src/runtime/asm_riscv64.s
@@ -79,7 +79,7 @@ TEXT setg_gcc<>(SB),NOSPLIT,$0-0
 
 // func cputicks() int64
 TEXT runtime·cputicks(SB),NOSPLIT,$0-8
-	WORD	$0xc0102573	// rdtime a0
+	RDTIME	A0
 	MOV	A0, ret+0(FP)
 	RET
 
diff --git a/src/runtime/cgocall.go b/src/runtime/cgocall.go
index a4e64b00cc..099aa540e0 100644
--- a/src/runtime/cgocall.go
+++ b/src/runtime/cgocall.go
@@ -605,7 +605,7 @@ func cgoCheckUnknownPointer(p unsafe.Pointer, msg string) (base, i uintptr) {
 		hbits := heapBitsForAddr(base)
 		n := span.elemsize
 		for i = uintptr(0); i < n; i += sys.PtrSize {
-			if i != 1*sys.PtrSize && !hbits.morePointers() {
+			if !hbits.morePointers() {
 				// No more possible pointers.
 				break
 			}
diff --git a/src/runtime/chan.go b/src/runtime/chan.go
index f6f4ffd02e..0afe5d962b 100644
--- a/src/runtime/chan.go
+++ b/src/runtime/chan.go
@@ -263,18 +263,19 @@ func chansend(c *hchan, ep unsafe.Pointer, block bool, callerpc uintptr) bool {
 	}
 	gp.waiting = nil
 	gp.activeStackChans = false
-	if gp.param == nil {
-		if c.closed == 0 {
-			throw("chansend: spurious wakeup")
-		}
-		panic(plainError("send on closed channel"))
-	}
+	closed := !mysg.success
 	gp.param = nil
 	if mysg.releasetime > 0 {
 		blockevent(mysg.releasetime-t0, 2)
 	}
 	mysg.c = nil
 	releaseSudog(mysg)
+	if closed {
+		if c.closed == 0 {
+			throw("chansend: spurious wakeup")
+		}
+		panic(plainError("send on closed channel"))
+	}
 	return true
 }
 
@@ -311,6 +312,7 @@ func send(c *hchan, sg *sudog, ep unsafe.Pointer, unlockf func(), skip int) {
 	gp := sg.g
 	unlockf()
 	gp.param = unsafe.Pointer(sg)
+	sg.success = true
 	if sg.releasetime != 0 {
 		sg.releasetime = cputicks()
 	}
@@ -384,7 +386,8 @@ func closechan(c *hchan) {
 			sg.releasetime = cputicks()
 		}
 		gp := sg.g
-		gp.param = nil
+		gp.param = unsafe.Pointer(sg)
+		sg.success = false
 		if raceenabled {
 			raceacquireg(gp, c.raceaddr())
 		}
@@ -402,7 +405,8 @@ func closechan(c *hchan) {
 			sg.releasetime = cputicks()
 		}
 		gp := sg.g
-		gp.param = nil
+		gp.param = unsafe.Pointer(sg)
+		sg.success = false
 		if raceenabled {
 			raceacquireg(gp, c.raceaddr())
 		}
@@ -575,11 +579,11 @@ func chanrecv(c *hchan, ep unsafe.Pointer, block bool) (selected, received bool)
 	if mysg.releasetime > 0 {
 		blockevent(mysg.releasetime-t0, 2)
 	}
-	closed := gp.param == nil
+	success := mysg.success
 	gp.param = nil
 	mysg.c = nil
 	releaseSudog(mysg)
-	return true, !closed
+	return true, success
 }
 
 // recv processes a receive operation on a full channel c.
@@ -632,6 +636,7 @@ func recv(c *hchan, sg *sudog, ep unsafe.Pointer, unlockf func(), skip int) {
 	gp := sg.g
 	unlockf()
 	gp.param = unsafe.Pointer(sg)
+	sg.success = true
 	if sg.releasetime != 0 {
 		sg.releasetime = cputicks()
 	}
diff --git a/src/runtime/checkptr_test.go b/src/runtime/checkptr_test.go
index 8ab8a4937c..194cc1243a 100644
--- a/src/runtime/checkptr_test.go
+++ b/src/runtime/checkptr_test.go
@@ -27,6 +27,7 @@ func TestCheckPtr(t *testing.T) {
 		{"CheckPtrAlignmentPtr", "fatal error: checkptr: misaligned pointer conversion\n"},
 		{"CheckPtrAlignmentNoPtr", ""},
 		{"CheckPtrArithmetic", "fatal error: checkptr: pointer arithmetic result points to invalid allocation\n"},
+		{"CheckPtrArithmetic2", "fatal error: checkptr: pointer arithmetic result points to invalid allocation\n"},
 		{"CheckPtrSize", "fatal error: checkptr: converted pointer straddles multiple allocations\n"},
 		{"CheckPtrSmall", "fatal error: checkptr: pointer arithmetic computed bad pointer value\n"},
 	}
diff --git a/src/runtime/closure_test.go b/src/runtime/closure_test.go
index ea65fbd5f5..741c932eab 100644
--- a/src/runtime/closure_test.go
+++ b/src/runtime/closure_test.go
@@ -1,6 +1,7 @@
 // Copyright 2011 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_test
 
 import "testing"
diff --git a/src/runtime/defs_linux_386.go b/src/runtime/defs_linux_386.go
index f4db8cf927..64a0fbcaaa 100644
--- a/src/runtime/defs_linux_386.go
+++ b/src/runtime/defs_linux_386.go
@@ -226,14 +226,3 @@ type sockaddr_un struct {
 	family uint16
 	path   [108]byte
 }
-
-const __NEW_UTS_LEN = 64
-
-type new_utsname struct {
-	sysname    [__NEW_UTS_LEN + 1]byte
-	nodename   [__NEW_UTS_LEN + 1]byte
-	release    [__NEW_UTS_LEN + 1]byte
-	version    [__NEW_UTS_LEN + 1]byte
-	machine    [__NEW_UTS_LEN + 1]byte
-	domainname [__NEW_UTS_LEN + 1]byte
-}
diff --git a/src/runtime/defs_linux_amd64.go b/src/runtime/defs_linux_amd64.go
index 8480d85219..1ae18a309b 100644
--- a/src/runtime/defs_linux_amd64.go
+++ b/src/runtime/defs_linux_amd64.go
@@ -262,14 +262,3 @@ type sockaddr_un struct {
 	family uint16
 	path   [108]byte
 }
-
-const __NEW_UTS_LEN = 64
-
-type new_utsname struct {
-	sysname    [__NEW_UTS_LEN + 1]byte
-	nodename   [__NEW_UTS_LEN + 1]byte
-	release    [__NEW_UTS_LEN + 1]byte
-	version    [__NEW_UTS_LEN + 1]byte
-	machine    [__NEW_UTS_LEN + 1]byte
-	domainname [__NEW_UTS_LEN + 1]byte
-}
diff --git a/src/runtime/export_test.go b/src/runtime/export_test.go
index 5ab03f3f99..d591fdc4e9 100644
--- a/src/runtime/export_test.go
+++ b/src/runtime/export_test.go
@@ -43,8 +43,6 @@ var PhysHugePageSize = physHugePageSize
 
 var NetpollGenericInit = netpollGenericInit
 
-var ParseRelease = parseRelease
-
 var Memmove = memmove
 var MemclrNoHeapPointers = memclrNoHeapPointers
 
diff --git a/src/runtime/gcinfo_test.go b/src/runtime/gcinfo_test.go
index ec1ba90c2e..0808b416f0 100644
--- a/src/runtime/gcinfo_test.go
+++ b/src/runtime/gcinfo_test.go
@@ -77,7 +77,7 @@ func TestGCInfo(t *testing.T) {
 	}
 
 	for i := 0; i < 10; i++ {
-		verifyGCInfo(t, "heap Ptr", escape(new(Ptr)), trimDead(padDead(infoPtr)))
+		verifyGCInfo(t, "heap Ptr", escape(new(Ptr)), trimDead(infoPtr))
 		verifyGCInfo(t, "heap PtrSlice", escape(&make([]*byte, 10)[0]), trimDead(infoPtr10))
 		verifyGCInfo(t, "heap ScalarPtr", escape(new(ScalarPtr)), trimDead(infoScalarPtr))
 		verifyGCInfo(t, "heap ScalarPtrSlice", escape(&make([]ScalarPtr, 4)[0]), trimDead(infoScalarPtr4))
@@ -97,25 +97,10 @@ func verifyGCInfo(t *testing.T, name string, p interface{}, mask0 []byte) {
 	}
 }
 
-func padDead(mask []byte) []byte {
-	// Because the dead bit isn't encoded in the second word,
-	// and because on 32-bit systems a one-word allocation
-	// uses a two-word block, the pointer info for a one-word
-	// object needs to be expanded to include an extra scalar
-	// on 32-bit systems to match the heap bitmap.
-	if runtime.PtrSize == 4 && len(mask) == 1 {
-		return []byte{mask[0], 0}
-	}
-	return mask
-}
-
 func trimDead(mask []byte) []byte {
-	for len(mask) > 2 && mask[len(mask)-1] == typeScalar {
+	for len(mask) > 0 && mask[len(mask)-1] == typeScalar {
 		mask = mask[:len(mask)-1]
 	}
-	if len(mask) == 2 && mask[0] == typeScalar && mask[1] == typeScalar {
-		mask = mask[:0]
-	}
 	return mask
 }
 
diff --git a/src/runtime/heapdump.go b/src/runtime/heapdump.go
index cfd5c251b4..4c35309211 100644
--- a/src/runtime/heapdump.go
+++ b/src/runtime/heapdump.go
@@ -713,7 +713,7 @@ func makeheapobjbv(p uintptr, size uintptr) bitvector {
 	i := uintptr(0)
 	hbits := heapBitsForAddr(p)
 	for ; i < nptr; i++ {
-		if i != 1 && !hbits.morePointers() {
+		if !hbits.morePointers() {
 			break // end of object
 		}
 		if hbits.isPointer() {
diff --git a/src/runtime/lockrank.go b/src/runtime/lockrank.go
index 000193585d..b23cf767be 100644
--- a/src/runtime/lockrank.go
+++ b/src/runtime/lockrank.go
@@ -67,8 +67,6 @@ const (
 	lockRankRwmutexW
 	lockRankRwmutexR
 
-	lockRankMcentral // For !go115NewMCentralImpl
-	lockRankSpine    // For !go115NewMCentralImpl
 	lockRankSpanSetSpine
 	lockRankGscan
 	lockRankStackpool
@@ -149,8 +147,6 @@ var lockNames = []string{
 	lockRankRwmutexW: "rwmutexW",
 	lockRankRwmutexR: "rwmutexR",
 
-	lockRankMcentral:     "mcentral",
-	lockRankSpine:        "spine",
 	lockRankSpanSetSpine: "spanSetSpine",
 	lockRankGscan:        "gscan",
 	lockRankStackpool:    "stackpool",
@@ -228,18 +224,16 @@ var lockPartialOrder [][]lockRank = [][]lockRank{
 	lockRankRwmutexW: {},
 	lockRankRwmutexR: {lockRankRwmutexW},
 
-	lockRankMcentral:     {lockRankSysmon, lockRankScavenge, lockRankForcegc, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankAllg, lockRankAllp, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankHchan},
-	lockRankSpine:        {lockRankSysmon, lockRankScavenge, lockRankAssistQueue, lockRankCpuprof, lockRankSched, lockRankAllg, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankHchan},
 	lockRankSpanSetSpine: {lockRankSysmon, lockRankScavenge, lockRankForcegc, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankAllg, lockRankAllp, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankHchan},
-	lockRankGscan:        {lockRankSysmon, lockRankScavenge, lockRankForcegc, lockRankSweepWaiters, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankHchan, lockRankFin, lockRankTraceBuf, lockRankTraceStrings, lockRankRoot, lockRankNotifyList, lockRankProf, lockRankGcBitsArenas, lockRankTrace, lockRankTraceStackTab, lockRankNetpollInit, lockRankMcentral, lockRankSpine, lockRankSpanSetSpine},
-	lockRankStackpool:    {lockRankSysmon, lockRankScavenge, lockRankSweepWaiters, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankPollDesc, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankHchan, lockRankFin, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankProf, lockRankGcBitsArenas, lockRankRoot, lockRankTrace, lockRankTraceStackTab, lockRankNetpollInit, lockRankRwmutexR, lockRankMcentral, lockRankSpine, lockRankSpanSetSpine, lockRankGscan},
-	lockRankStackLarge:   {lockRankSysmon, lockRankAssistQueue, lockRankSched, lockRankItab, lockRankHchan, lockRankProf, lockRankGcBitsArenas, lockRankRoot, lockRankMcentral, lockRankSpanSetSpine, lockRankGscan},
+	lockRankGscan:        {lockRankSysmon, lockRankScavenge, lockRankForcegc, lockRankSweepWaiters, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankHchan, lockRankFin, lockRankTraceBuf, lockRankTraceStrings, lockRankRoot, lockRankNotifyList, lockRankProf, lockRankGcBitsArenas, lockRankTrace, lockRankTraceStackTab, lockRankNetpollInit, lockRankSpanSetSpine},
+	lockRankStackpool:    {lockRankSysmon, lockRankScavenge, lockRankSweepWaiters, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankPollDesc, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankHchan, lockRankFin, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankProf, lockRankGcBitsArenas, lockRankRoot, lockRankTrace, lockRankTraceStackTab, lockRankNetpollInit, lockRankRwmutexR, lockRankSpanSetSpine, lockRankGscan},
+	lockRankStackLarge:   {lockRankSysmon, lockRankAssistQueue, lockRankSched, lockRankItab, lockRankHchan, lockRankProf, lockRankGcBitsArenas, lockRankRoot, lockRankSpanSetSpine, lockRankGscan},
 	lockRankDefer:        {},
 	lockRankSudog:        {lockRankNotifyList, lockRankHchan},
 	lockRankWbufSpans:    {lockRankSysmon, lockRankScavenge, lockRankSweepWaiters, lockRankAssistQueue, lockRankSweep, lockRankSched, lockRankAllg, lockRankPollDesc, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankHchan, lockRankNotifyList, lockRankTraceStrings, lockRankMspanSpecial, lockRankProf, lockRankRoot, lockRankGscan, lockRankDefer, lockRankSudog},
-	lockRankMheap:        {lockRankSysmon, lockRankScavenge, lockRankSweepWaiters, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankAllg, lockRankAllp, lockRankPollDesc, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankHchan, lockRankMspanSpecial, lockRankProf, lockRankGcBitsArenas, lockRankRoot, lockRankMcentral, lockRankGscan, lockRankStackpool, lockRankStackLarge, lockRankDefer, lockRankSudog, lockRankWbufSpans, lockRankSpanSetSpine},
+	lockRankMheap:        {lockRankSysmon, lockRankScavenge, lockRankSweepWaiters, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankAllg, lockRankAllp, lockRankPollDesc, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankHchan, lockRankMspanSpecial, lockRankProf, lockRankGcBitsArenas, lockRankRoot, lockRankGscan, lockRankStackpool, lockRankStackLarge, lockRankDefer, lockRankSudog, lockRankWbufSpans, lockRankSpanSetSpine},
 	lockRankMheapSpecial: {lockRankSysmon, lockRankScavenge, lockRankAssistQueue, lockRankCpuprof, lockRankSweep, lockRankSched, lockRankAllg, lockRankAllp, lockRankTimers, lockRankItab, lockRankReflectOffs, lockRankNotifyList, lockRankTraceBuf, lockRankTraceStrings, lockRankHchan},
-	lockRankGlobalAlloc:  {lockRankProf, lockRankSpine, lockRankSpanSetSpine, lockRankMheap, lockRankMheapSpecial},
+	lockRankGlobalAlloc:  {lockRankProf, lockRankSpanSetSpine, lockRankMheap, lockRankMheapSpecial},
 
 	lockRankGFree:     {lockRankSched},
 	lockRankHchanLeaf: {lockRankGscan, lockRankHchanLeaf},
diff --git a/src/runtime/malloc.go b/src/runtime/malloc.go
index b3fac3de24..e46327f9ce 100644
--- a/src/runtime/malloc.go
+++ b/src/runtime/malloc.go
@@ -1178,11 +1178,9 @@ func largeAlloc(size uintptr, needzero bool, noscan bool) *mspan {
 	if s == nil {
 		throw("out of memory")
 	}
-	if go115NewMCentralImpl {
-		// Put the large span in the mcentral swept list so that it's
-		// visible to the background sweeper.
-		mheap_.central[spc].mcentral.fullSwept(mheap_.sweepgen).push(s)
-	}
+	// Put the large span in the mcentral swept list so that it's
+	// visible to the background sweeper.
+	mheap_.central[spc].mcentral.fullSwept(mheap_.sweepgen).push(s)
 	s.limit = s.base() + size
 	heapBitsForAddr(s.base()).initSpan(s)
 	return s
diff --git a/src/runtime/map_benchmark_test.go b/src/runtime/map_benchmark_test.go
index 893cb6c5b6..d0becc9ddb 100644
--- a/src/runtime/map_benchmark_test.go
+++ b/src/runtime/map_benchmark_test.go
@@ -1,6 +1,7 @@
 // Copyright 2013 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_test
 
 import (
diff --git a/src/runtime/map_fast32.go b/src/runtime/map_fast32.go
index 534454f3ad..d035ed0386 100644
--- a/src/runtime/map_fast32.go
+++ b/src/runtime/map_fast32.go
@@ -299,8 +299,12 @@ search:
 				continue
 			}
 			// Only clear key if there are pointers in it.
-			if t.key.ptrdata != 0 {
-				memclrHasPointers(k, t.key.size)
+			// This can only happen if pointers are 32 bit
+			// wide as 64 bit pointers do not fit into a 32 bit key.
+			if sys.PtrSize == 4 && t.key.ptrdata != 0 {
+				// The key must be a pointer as we checked pointers are
+				// 32 bits wide and the key is 32 bits wide also.
+				*(*unsafe.Pointer)(k) = nil
 			}
 			e := add(unsafe.Pointer(b), dataOffset+bucketCnt*4+i*uintptr(t.elemsize))
 			if t.elem.ptrdata != 0 {
diff --git a/src/runtime/map_fast64.go b/src/runtime/map_fast64.go
index 1669c7cfe9..f1f3927598 100644
--- a/src/runtime/map_fast64.go
+++ b/src/runtime/map_fast64.go
@@ -300,7 +300,13 @@ search:
 			}
 			// Only clear key if there are pointers in it.
 			if t.key.ptrdata != 0 {
-				memclrHasPointers(k, t.key.size)
+				if sys.PtrSize == 8 {
+					*(*unsafe.Pointer)(k) = nil
+				} else {
+					// There are three ways to squeeze at one ore more 32 bit pointers into 64 bits.
+					// Just call memclrHasPointers instead of trying to handle all cases here.
+					memclrHasPointers(k, 8)
+				}
 			}
 			e := add(unsafe.Pointer(b), dataOffset+bucketCnt*8+i*uintptr(t.elemsize))
 			if t.elem.ptrdata != 0 {
diff --git a/src/runtime/map_test.go b/src/runtime/map_test.go
index 1b7ccad6ed..302b3c23c1 100644
--- a/src/runtime/map_test.go
+++ b/src/runtime/map_test.go
@@ -993,6 +993,27 @@ func benchmarkMapDeleteStr(b *testing.B, n int) {
 	}
 }
 
+func benchmarkMapDeletePointer(b *testing.B, n int) {
+	i2p := make([]*int, n)
+	for i := 0; i < n; i++ {
+		i2p[i] = new(int)
+	}
+	a := make(map[*int]int, n)
+	b.ResetTimer()
+	k := 0
+	for i := 0; i < b.N; i++ {
+		if len(a) == 0 {
+			b.StopTimer()
+			for j := 0; j < n; j++ {
+				a[i2p[j]] = j
+			}
+			k = i
+			b.StartTimer()
+		}
+		delete(a, i2p[i-k])
+	}
+}
+
 func runWith(f func(*testing.B, int), v ...int) func(*testing.B) {
 	return func(b *testing.B) {
 		for _, n := range v {
@@ -1023,6 +1044,7 @@ func BenchmarkMapDelete(b *testing.B) {
 	b.Run("Int32", runWith(benchmarkMapDeleteInt32, 100, 1000, 10000))
 	b.Run("Int64", runWith(benchmarkMapDeleteInt64, 100, 1000, 10000))
 	b.Run("Str", runWith(benchmarkMapDeleteStr, 100, 1000, 10000))
+	b.Run("Pointer", runWith(benchmarkMapDeletePointer, 100, 1000, 10000))
 }
 
 func TestDeferDeleteSlow(t *testing.T) {
diff --git a/src/runtime/mbitmap.go b/src/runtime/mbitmap.go
index 35332c91c4..8de44c14b9 100644
--- a/src/runtime/mbitmap.go
+++ b/src/runtime/mbitmap.go
@@ -6,10 +6,11 @@
 //
 // Stack, data, and bss bitmaps
 //
-// Stack frames and global variables in the data and bss sections are described
-// by 1-bit bitmaps in which 0 means uninteresting and 1 means live pointer
-// to be visited during GC. The bits in each byte are consumed starting with
-// the low bit: 1<<0, 1<<1, and so on.
+// Stack frames and global variables in the data and bss sections are
+// described by bitmaps with 1 bit per pointer-sized word. A "1" bit
+// means the word is a live pointer to be visited by the GC (referred to
+// as "pointer"). A "0" bit means the word should be ignored by GC
+// (referred to as "scalar", though it could be a dead pointer value).
 //
 // Heap bitmap
 //
@@ -20,18 +21,13 @@
 // through start+3*ptrSize, ha.bitmap[1] holds the entries for
 // start+4*ptrSize through start+7*ptrSize, and so on.
 //
-// In each 2-bit entry, the lower bit holds the same information as in the 1-bit
-// bitmaps: 0 means uninteresting and 1 means live pointer to be visited during GC.
-// The meaning of the high bit depends on the position of the word being described
-// in its allocated object. In all words *except* the second word, the
-// high bit indicates that the object is still being described. In
-// these words, if a bit pair with a high bit 0 is encountered, the
-// low bit can also be assumed to be 0, and the object description is
-// over. This 00 is called the ``dead'' encoding: it signals that the
-// rest of the words in the object are uninteresting to the garbage
-// collector.
-//
-// In the second word, the high bit is the GC ``checkmarked'' bit (see below).
+// In each 2-bit entry, the lower bit is a pointer/scalar bit, just
+// like in the stack/data bitmaps described above. The upper bit
+// indicates scan/dead: a "1" value ("scan") indicates that there may
+// be pointers in later words of the allocation, and a "0" value
+// ("dead") indicates there are no more pointers in the allocation. If
+// the upper bit is 0, the lower bit must also be 0, and this
+// indicates scanning can ignore the rest of the allocation.
 //
 // The 2-bit entries are split when written into the byte, so that the top half
 // of the byte contains 4 high bits and the bottom half contains 4 low (pointer)
@@ -39,38 +35,14 @@
 // This form allows a copy from the 1-bit to the 4-bit form to keep the
 // pointer bits contiguous, instead of having to space them out.
 //
-// The code makes use of the fact that the zero value for a heap bitmap
-// has no live pointer bit set and is (depending on position), not used,
-// not checkmarked, and is the dead encoding.
-// These properties must be preserved when modifying the encoding.
+// The code makes use of the fact that the zero value for a heap
+// bitmap means scalar/dead. This property must be preserved when
+// modifying the encoding.
 //
 // The bitmap for noscan spans is not maintained. Code must ensure
 // that an object is scannable before consulting its bitmap by
 // checking either the noscan bit in the span or by consulting its
 // type's information.
-//
-// Checkmarks
-//
-// In a concurrent garbage collector, one worries about failing to mark
-// a live object due to mutations without write barriers or bugs in the
-// collector implementation. As a sanity check, the GC has a 'checkmark'
-// mode that retraverses the object graph with the world stopped, to make
-// sure that everything that should be marked is marked.
-// In checkmark mode, in the heap bitmap, the high bit of the 2-bit entry
-// for the second word of the object holds the checkmark bit.
-// When not in checkmark mode, this bit is set to 1.
-//
-// The smallest possible allocation is 8 bytes. On a 32-bit machine, that
-// means every allocated object has two words, so there is room for the
-// checkmark bit. On a 64-bit machine, however, the 8-byte allocation is
-// just one word, so the second bit pair is not available for encoding the
-// checkmark. However, because non-pointer allocations are combined
-// into larger 16-byte (maxTinySize) allocations, a plain 8-byte allocation
-// must be a pointer, so the type bit in the first word is not actually needed.
-// It is still used in general, except in checkmark the type bit is repurposed
-// as the checkmark bit and then reinitialized (to 1) as the type bit when
-// finished.
-//
 
 package runtime
 
@@ -551,33 +523,6 @@ func (h heapBits) isPointer() bool {
 	return h.bits()&bitPointer != 0
 }
 
-// isCheckmarked reports whether the heap bits have the checkmarked bit set.
-// It must be told how large the object at h is, because the encoding of the
-// checkmark bit varies by size.
-// h must describe the initial word of the object.
-func (h heapBits) isCheckmarked(size uintptr) bool {
-	if size == sys.PtrSize {
-		return (*h.bitp>>h.shift)&bitPointer != 0
-	}
-	// All multiword objects are 2-word aligned,
-	// so we know that the initial word's 2-bit pair
-	// and the second word's 2-bit pair are in the
-	// same heap bitmap byte, *h.bitp.
-	return (*h.bitp>>(heapBitsShift+h.shift))&bitScan != 0
-}
-
-// setCheckmarked sets the checkmarked bit.
-// It must be told how large the object at h is, because the encoding of the
-// checkmark bit varies by size.
-// h must describe the initial word of the object.
-func (h heapBits) setCheckmarked(size uintptr) {
-	if size == sys.PtrSize {
-		atomic.Or8(h.bitp, bitPointer<<h.shift)
-		return
-	}
-	atomic.Or8(h.bitp, bitScan<<(heapBitsShift+h.shift))
-}
-
 // bulkBarrierPreWrite executes a write barrier
 // for every pointer slot in the memory range [src, src+size),
 // using pointer/scalar information from [dst, dst+size).
@@ -795,7 +740,6 @@ func typeBitsBulkBarrier(typ *_type, dst, src, size uintptr) {
 // TODO(rsc): Perhaps introduce a different heapBitsSpan type.
 
 // initSpan initializes the heap bitmap for a span.
-// It clears all checkmark bits.
 // If this is a span of pointer-sized objects, it initializes all
 // words to pointer/scan.
 // Otherwise, it initializes all words to scalar/dead.
@@ -826,45 +770,6 @@ func (h heapBits) initSpan(s *mspan) {
 	}
 }
 
-// initCheckmarkSpan initializes a span for being checkmarked.
-// It clears the checkmark bits, which are set to 1 in normal operation.
-func (h heapBits) initCheckmarkSpan(size, n, total uintptr) {
-	// The ptrSize == 8 is a compile-time constant false on 32-bit and eliminates this code entirely.
-	if sys.PtrSize == 8 && size == sys.PtrSize {
-		// Checkmark bit is type bit, bottom bit of every 2-bit entry.
-		// Only possible on 64-bit system, since minimum size is 8.
-		// Must clear type bit (checkmark bit) of every word.
-		// The type bit is the lower of every two-bit pair.
-		for i := uintptr(0); i < n; i += wordsPerBitmapByte {
-			*h.bitp &^= bitPointerAll
-			h = h.forward(wordsPerBitmapByte)
-		}
-		return
-	}
-	for i := uintptr(0); i < n; i++ {
-		*h.bitp &^= bitScan << (heapBitsShift + h.shift)
-		h = h.forward(size / sys.PtrSize)
-	}
-}
-
-// clearCheckmarkSpan undoes all the checkmarking in a span.
-// The actual checkmark bits are ignored, so the only work to do
-// is to fix the pointer bits. (Pointer bits are ignored by scanobject
-// but consulted by typedmemmove.)
-func (h heapBits) clearCheckmarkSpan(size, n, total uintptr) {
-	// The ptrSize == 8 is a compile-time constant false on 32-bit and eliminates this code entirely.
-	if sys.PtrSize == 8 && size == sys.PtrSize {
-		// Checkmark bit is type bit, bottom bit of every 2-bit entry.
-		// Only possible on 64-bit system, since minimum size is 8.
-		// Must clear type bit (checkmark bit) of every word.
-		// The type bit is the lower of every two-bit pair.
-		for i := uintptr(0); i < n; i += wordsPerBitmapByte {
-			*h.bitp |= bitPointerAll
-			h = h.forward(wordsPerBitmapByte)
-		}
-	}
-}
-
 // countAlloc returns the number of objects allocated in span s by
 // scanning the allocation bitmap.
 func (s *mspan) countAlloc() int {
@@ -957,11 +862,11 @@ func heapBitsSetType(x, size, dataSize uintptr, typ *_type) {
 			if sys.PtrSize == 4 && dataSize == sys.PtrSize {
 				// 1 pointer object. On 32-bit machines clear the bit for the
 				// unused second word.
-				*h.bitp &^= (bitPointer | bitScan | ((bitPointer | bitScan) << heapBitsShift)) << h.shift
+				*h.bitp &^= (bitPointer | bitScan | (bitPointer|bitScan)<<heapBitsShift) << h.shift
 				*h.bitp |= (bitPointer | bitScan) << h.shift
 			} else {
 				// 2-element slice of pointer.
-				*h.bitp |= (bitPointer | bitScan | bitPointer<<heapBitsShift) << h.shift
+				*h.bitp |= (bitPointer | bitScan | (bitPointer|bitScan)<<heapBitsShift) << h.shift
 			}
 			return
 		}
@@ -974,11 +879,10 @@ func heapBitsSetType(x, size, dataSize uintptr, typ *_type) {
 			}
 		}
 		b := uint32(*ptrmask)
-		hb := (b & 3) | bitScan
-		// bitPointer == 1, bitScan is 1 << 4, heapBitsShift is 1.
-		// 110011 is shifted h.shift and complemented.
-		// This clears out the bits that are about to be
-		// ored into *h.hbitp in the next instructions.
+		hb := b & 3
+		hb |= bitScanAll & ((bitScan << (typ.ptrdata / sys.PtrSize)) - 1)
+		// Clear the bits for this object so we can set the
+		// appropriate ones.
 		*h.bitp &^= (bitPointer | bitScan | ((bitPointer | bitScan) << heapBitsShift)) << h.shift
 		*h.bitp |= uint8(hb << h.shift)
 		return
@@ -1155,11 +1059,6 @@ func heapBitsSetType(x, size, dataSize uintptr, typ *_type) {
 		throw("heapBitsSetType: called with non-pointer type")
 		return
 	}
-	if nw < 2 {
-		// Must write at least 2 words, because the "no scan"
-		// encoding doesn't take effect until the third word.
-		nw = 2
-	}
 
 	// Phase 1: Special case for leading byte (shift==0) or half-byte (shift==2).
 	// The leading byte is special because it contains the bits for word 1,
@@ -1172,21 +1071,22 @@ func heapBitsSetType(x, size, dataSize uintptr, typ *_type) {
 
 	case h.shift == 0:
 		// Ptrmask and heap bitmap are aligned.
-		// Handle first byte of bitmap specially.
+		//
+		// This is a fast path for small objects.
 		//
 		// The first byte we write out covers the first four
 		// words of the object. The scan/dead bit on the first
 		// word must be set to scan since there are pointers
-		// somewhere in the object. The scan/dead bit on the
-		// second word is the checkmark, so we don't set it.
+		// somewhere in the object.
 		// In all following words, we set the scan/dead
 		// appropriately to indicate that the object contains
 		// to the next 2-bit entry in the bitmap.
 		//
-		// TODO: It doesn't matter if we set the checkmark, so
-		// maybe this case isn't needed any more.
+		// We set four bits at a time here, but if the object
+		// is fewer than four words, phase 3 will clear
+		// unnecessary bits.
 		hb = b & bitPointerAll
-		hb |= bitScan | bitScan<<(2*heapBitsShift) | bitScan<<(3*heapBitsShift)
+		hb |= bitScanAll
 		if w += 4; w >= nw {
 			goto Phase3
 		}
@@ -1203,14 +1103,13 @@ func heapBitsSetType(x, size, dataSize uintptr, typ *_type) {
 		// We took care of 1-word and 2-word objects above,
 		// so this is at least a 6-word object.
 		hb = (b & (bitPointer | bitPointer<<heapBitsShift)) << (2 * heapBitsShift)
-		// This is not noscan, so set the scan bit in the
-		// first word.
 		hb |= bitScan << (2 * heapBitsShift)
+		if nw > 1 {
+			hb |= bitScan << (3 * heapBitsShift)
+		}
 		b >>= 2
 		nb -= 2
-		// Note: no bitScan for second word because that's
-		// the checkmark.
-		*hbitp &^= uint8((bitPointer | bitScan | (bitPointer << heapBitsShift)) << (2 * heapBitsShift))
+		*hbitp &^= uint8((bitPointer | bitScan | ((bitPointer | bitScan) << heapBitsShift)) << (2 * heapBitsShift))
 		*hbitp |= uint8(hb)
 		hbitp = add1(hbitp)
 		if w += 2; w >= nw {
@@ -1403,17 +1302,20 @@ Phase4:
 	// Double check the whole bitmap.
 	if doubleCheck {
 		// x+size may not point to the heap, so back up one
-		// word and then call next().
-		end := heapBitsForAddr(x + size - sys.PtrSize).next()
-		endAI := arenaIdx(end.arena)
-		if !outOfPlace && (end.bitp == nil || (end.shift == 0 && end.bitp == &mheap_.arenas[endAI.l1()][endAI.l2()].bitmap[0])) {
-			// The unrolling code above walks hbitp just
-			// past the bitmap without moving to the next
-			// arena. Synthesize this for end.bitp.
-			end.arena--
-			endAI = arenaIdx(end.arena)
-			end.bitp = addb(&mheap_.arenas[endAI.l1()][endAI.l2()].bitmap[0], heapArenaBitmapBytes)
-			end.last = nil
+		// word and then advance it the way we do above.
+		end := heapBitsForAddr(x + size - sys.PtrSize)
+		if outOfPlace {
+			// In out-of-place copying, we just advance
+			// using next.
+			end = end.next()
+		} else {
+			// Don't use next because that may advance to
+			// the next arena and the in-place logic
+			// doesn't do that.
+			end.shift += heapBitsShift
+			if end.shift == 4*heapBitsShift {
+				end.bitp, end.shift = add1(end.bitp), 0
+			}
 		}
 		if typ.kind&kindGCProg == 0 && (hbitp != end.bitp || (w == nw+2) != (end.shift == 2)) {
 			println("ended at wrong bitmap byte for", typ.string(), "x", dataSize/typ.size)
@@ -1437,19 +1339,16 @@ Phase4:
 			var have, want uint8
 			have = (*h.bitp >> h.shift) & (bitPointer | bitScan)
 			if i >= totalptr {
-				want = 0 // deadmarker
 				if typ.kind&kindGCProg != 0 && i < (totalptr+3)/4*4 {
+					// heapBitsSetTypeGCProg always fills
+					// in full nibbles of bitScan.
 					want = bitScan
 				}
 			} else {
 				if j < nptr && (*addb(ptrmask, j/8)>>(j%8))&1 != 0 {
 					want |= bitPointer
 				}
-				if i != 1 {
-					want |= bitScan
-				} else {
-					have &^= bitScan
-				}
+				want |= bitScan
 			}
 			if have != want {
 				println("mismatch writing bits for", typ.string(), "x", dataSize/typ.size)
@@ -2009,7 +1908,7 @@ func getgcmask(ep interface{}) (mask []byte) {
 			if hbits.isPointer() {
 				mask[i/sys.PtrSize] = 1
 			}
-			if i != 1*sys.PtrSize && !hbits.morePointers() {
+			if !hbits.morePointers() {
 				mask = mask[:i/sys.PtrSize]
 				break
 			}
diff --git a/src/runtime/mcache.go b/src/runtime/mcache.go
index 5bceb51ac9..7a7d33ccae 100644
--- a/src/runtime/mcache.go
+++ b/src/runtime/mcache.go
@@ -131,11 +131,7 @@ func (c *mcache) refill(spc spanClass) {
 		if s.sweepgen != mheap_.sweepgen+3 {
 			throw("bad sweepgen in refill")
 		}
-		if go115NewMCentralImpl {
-			mheap_.central[spc].mcentral.uncacheSpan(s)
-		} else {
-			atomic.Store(&s.sweepgen, mheap_.sweepgen)
-		}
+		mheap_.central[spc].mcentral.uncacheSpan(s)
 	}
 
 	// Get a new cached span from the central lists.
diff --git a/src/runtime/mcentral.go b/src/runtime/mcentral.go
index ed49d86d0c..ed49e01677 100644
--- a/src/runtime/mcentral.go
+++ b/src/runtime/mcentral.go
@@ -18,7 +18,6 @@ import "runtime/internal/atomic"
 //
 //go:notinheap
 type mcentral struct {
-	lock      mutex
 	spanclass spanClass
 
 	// For !go115NewMCentralImpl.
@@ -55,16 +54,10 @@ type mcentral struct {
 // Initialize a single central free list.
 func (c *mcentral) init(spc spanClass) {
 	c.spanclass = spc
-	if go115NewMCentralImpl {
-		lockInit(&c.partial[0].spineLock, lockRankSpanSetSpine)
-		lockInit(&c.partial[1].spineLock, lockRankSpanSetSpine)
-		lockInit(&c.full[0].spineLock, lockRankSpanSetSpine)
-		lockInit(&c.full[1].spineLock, lockRankSpanSetSpine)
-	} else {
-		c.nonempty.init()
-		c.empty.init()
-		lockInit(&c.lock, lockRankMcentral)
-	}
+	lockInit(&c.partial[0].spineLock, lockRankSpanSetSpine)
+	lockInit(&c.partial[1].spineLock, lockRankSpanSetSpine)
+	lockInit(&c.full[0].spineLock, lockRankSpanSetSpine)
+	lockInit(&c.full[1].spineLock, lockRankSpanSetSpine)
 }
 
 // partialUnswept returns the spanSet which holds partially-filled
@@ -93,9 +86,6 @@ func (c *mcentral) fullSwept(sweepgen uint32) *spanSet {
 
 // Allocate a span to use in an mcache.
 func (c *mcentral) cacheSpan() *mspan {
-	if !go115NewMCentralImpl {
-		return c.oldCacheSpan()
-	}
 	// Deduct credit for this span allocation and sweep if necessary.
 	spanBytes := uintptr(class_to_allocnpages[c.spanclass.sizeclass()]) * _PageSize
 	deductSweepCredit(spanBytes, 0)
@@ -213,127 +203,11 @@ havespan:
 	return s
 }
 
-// Allocate a span to use in an mcache.
-//
-// For !go115NewMCentralImpl.
-func (c *mcentral) oldCacheSpan() *mspan {
-	// Deduct credit for this span allocation and sweep if necessary.
-	spanBytes := uintptr(class_to_allocnpages[c.spanclass.sizeclass()]) * _PageSize
-	deductSweepCredit(spanBytes, 0)
-
-	lock(&c.lock)
-	traceDone := false
-	if trace.enabled {
-		traceGCSweepStart()
-	}
-	sg := mheap_.sweepgen
-retry:
-	var s *mspan
-	for s = c.nonempty.first; s != nil; s = s.next {
-		if s.sweepgen == sg-2 && atomic.Cas(&s.sweepgen, sg-2, sg-1) {
-			c.nonempty.remove(s)
-			c.empty.insertBack(s)
-			unlock(&c.lock)
-			s.sweep(true)
-			goto havespan
-		}
-		if s.sweepgen == sg-1 {
-			// the span is being swept by background sweeper, skip
-			continue
-		}
-		// we have a nonempty span that does not require sweeping, allocate from it
-		c.nonempty.remove(s)
-		c.empty.insertBack(s)
-		unlock(&c.lock)
-		goto havespan
-	}
-
-	for s = c.empty.first; s != nil; s = s.next {
-		if s.sweepgen == sg-2 && atomic.Cas(&s.sweepgen, sg-2, sg-1) {
-			// we have an empty span that requires sweeping,
-			// sweep it and see if we can free some space in it
-			c.empty.remove(s)
-			// swept spans are at the end of the list
-			c.empty.insertBack(s)
-			unlock(&c.lock)
-			s.sweep(true)
-			freeIndex := s.nextFreeIndex()
-			if freeIndex != s.nelems {
-				s.freeindex = freeIndex
-				goto havespan
-			}
-			lock(&c.lock)
-			// the span is still empty after sweep
-			// it is already in the empty list, so just retry
-			goto retry
-		}
-		if s.sweepgen == sg-1 {
-			// the span is being swept by background sweeper, skip
-			continue
-		}
-		// already swept empty span,
-		// all subsequent ones must also be either swept or in process of sweeping
-		break
-	}
-	if trace.enabled {
-		traceGCSweepDone()
-		traceDone = true
-	}
-	unlock(&c.lock)
-
-	// Replenish central list if empty.
-	s = c.grow()
-	if s == nil {
-		return nil
-	}
-	lock(&c.lock)
-	c.empty.insertBack(s)
-	unlock(&c.lock)
-
-	// At this point s is a non-empty span, queued at the end of the empty list,
-	// c is unlocked.
-havespan:
-	if trace.enabled && !traceDone {
-		traceGCSweepDone()
-	}
-	n := int(s.nelems) - int(s.allocCount)
-	if n == 0 || s.freeindex == s.nelems || uintptr(s.allocCount) == s.nelems {
-		throw("span has no free objects")
-	}
-	// Assume all objects from this span will be allocated in the
-	// mcache. If it gets uncached, we'll adjust this.
-	atomic.Xadd64(&c.nmalloc, int64(n))
-	usedBytes := uintptr(s.allocCount) * s.elemsize
-	atomic.Xadd64(&memstats.heap_live, int64(spanBytes)-int64(usedBytes))
-	if trace.enabled {
-		// heap_live changed.
-		traceHeapAlloc()
-	}
-	if gcBlackenEnabled != 0 {
-		// heap_live changed.
-		gcController.revise()
-	}
-	freeByteBase := s.freeindex &^ (64 - 1)
-	whichByte := freeByteBase / 8
-	// Init alloc bits cache.
-	s.refillAllocCache(whichByte)
-
-	// Adjust the allocCache so that s.freeindex corresponds to the low bit in
-	// s.allocCache.
-	s.allocCache >>= s.freeindex % 64
-
-	return s
-}
-
 // Return span from an mcache.
 //
 // s must have a span class corresponding to this
 // mcentral and it must not be empty.
 func (c *mcentral) uncacheSpan(s *mspan) {
-	if !go115NewMCentralImpl {
-		c.oldUncacheSpan(s)
-		return
-	}
 	if s.allocCount == 0 {
 		throw("uncaching span but s.allocCount == 0")
 	}
@@ -393,111 +267,6 @@ func (c *mcentral) uncacheSpan(s *mspan) {
 	}
 }
 
-// Return span from an mcache.
-//
-// For !go115NewMCentralImpl.
-func (c *mcentral) oldUncacheSpan(s *mspan) {
-	if s.allocCount == 0 {
-		throw("uncaching span but s.allocCount == 0")
-	}
-
-	sg := mheap_.sweepgen
-	stale := s.sweepgen == sg+1
-	if stale {
-		// Span was cached before sweep began. It's our
-		// responsibility to sweep it.
-		//
-		// Set sweepgen to indicate it's not cached but needs
-		// sweeping and can't be allocated from. sweep will
-		// set s.sweepgen to indicate s is swept.
-		atomic.Store(&s.sweepgen, sg-1)
-	} else {
-		// Indicate that s is no longer cached.
-		atomic.Store(&s.sweepgen, sg)
-	}
-
-	n := int(s.nelems) - int(s.allocCount)
-	if n > 0 {
-		// cacheSpan updated alloc assuming all objects on s
-		// were going to be allocated. Adjust for any that
-		// weren't. We must do this before potentially
-		// sweeping the span.
-		atomic.Xadd64(&c.nmalloc, -int64(n))
-
-		lock(&c.lock)
-		c.empty.remove(s)
-		c.nonempty.insert(s)
-		if !stale {
-			// mCentral_CacheSpan conservatively counted
-			// unallocated slots in heap_live. Undo this.
-			//
-			// If this span was cached before sweep, then
-			// heap_live was totally recomputed since
-			// caching this span, so we don't do this for
-			// stale spans.
-			atomic.Xadd64(&memstats.heap_live, -int64(n)*int64(s.elemsize))
-		}
-		unlock(&c.lock)
-	}
-
-	if stale {
-		// Now that s is in the right mcentral list, we can
-		// sweep it.
-		s.sweep(false)
-	}
-}
-
-// freeSpan updates c and s after sweeping s.
-// It sets s's sweepgen to the latest generation,
-// and, based on the number of free objects in s,
-// moves s to the appropriate list of c or returns it
-// to the heap.
-// freeSpan reports whether s was returned to the heap.
-// If preserve=true, it does not move s (the caller
-// must take care of it).
-//
-// For !go115NewMCentralImpl.
-func (c *mcentral) freeSpan(s *mspan, preserve bool, wasempty bool) bool {
-	if sg := mheap_.sweepgen; s.sweepgen == sg+1 || s.sweepgen == sg+3 {
-		throw("freeSpan given cached span")
-	}
-	s.needzero = 1
-
-	if preserve {
-		// preserve is set only when called from (un)cacheSpan above,
-		// the span must be in the empty list.
-		if !s.inList() {
-			throw("can't preserve unlinked span")
-		}
-		atomic.Store(&s.sweepgen, mheap_.sweepgen)
-		return false
-	}
-
-	lock(&c.lock)
-
-	// Move to nonempty if necessary.
-	if wasempty {
-		c.empty.remove(s)
-		c.nonempty.insert(s)
-	}
-
-	// delay updating sweepgen until here. This is the signal that
-	// the span may be used in an mcache, so it must come after the
-	// linked list operations above (actually, just after the
-	// lock of c above.)
-	atomic.Store(&s.sweepgen, mheap_.sweepgen)
-
-	if s.allocCount != 0 {
-		unlock(&c.lock)
-		return false
-	}
-
-	c.nonempty.remove(s)
-	unlock(&c.lock)
-	mheap_.freeSpan(s)
-	return true
-}
-
 // grow allocates a new empty span from the heap and initializes it for c's size class.
 func (c *mcentral) grow() *mspan {
 	npages := uintptr(class_to_allocnpages[c.spanclass.sizeclass()])
diff --git a/src/runtime/mcheckmark.go b/src/runtime/mcheckmark.go
new file mode 100644
index 0000000000..1fd8e4e78f
--- /dev/null
+++ b/src/runtime/mcheckmark.go
@@ -0,0 +1,100 @@
+// Copyright 2020 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.
+
+// GC checkmarks
+//
+// In a concurrent garbage collector, one worries about failing to mark
+// a live object due to mutations without write barriers or bugs in the
+// collector implementation. As a sanity check, the GC has a 'checkmark'
+// mode that retraverses the object graph with the world stopped, to make
+// sure that everything that should be marked is marked.
+
+package runtime
+
+import (
+	"runtime/internal/atomic"
+	"runtime/internal/sys"
+	"unsafe"
+)
+
+// A checkmarksMap stores the GC marks in "checkmarks" mode. It is a
+// per-arena bitmap with a bit for every word in the arena. The mark
+// is stored on the bit corresponding to the first word of the marked
+// allocation.
+//
+//go:notinheap
+type checkmarksMap [heapArenaBytes / sys.PtrSize / 8]uint8
+
+// If useCheckmark is true, marking of an object uses the checkmark
+// bits instead of the standard mark bits.
+var useCheckmark = false
+
+// startCheckmarks prepares for the checkmarks phase.
+//
+// The world must be stopped.
+func startCheckmarks() {
+	// Clear all checkmarks.
+	for _, ai := range mheap_.allArenas {
+		arena := mheap_.arenas[ai.l1()][ai.l2()]
+		bitmap := arena.checkmarks
+
+		if bitmap == nil {
+			// Allocate bitmap on first use.
+			bitmap = (*checkmarksMap)(persistentalloc(unsafe.Sizeof(*bitmap), 0, &memstats.gc_sys))
+			if bitmap == nil {
+				throw("out of memory allocating checkmarks bitmap")
+			}
+			arena.checkmarks = bitmap
+		} else {
+			// Otherwise clear the existing bitmap.
+			for i := range bitmap {
+				bitmap[i] = 0
+			}
+		}
+	}
+	// Enable checkmarking.
+	useCheckmark = true
+}
+
+// endCheckmarks ends the checkmarks phase.
+func endCheckmarks() {
+	if gcMarkWorkAvailable(nil) {
+		throw("GC work not flushed")
+	}
+	useCheckmark = false
+}
+
+// setCheckmark throws if marking object is a checkmarks violation,
+// and otherwise sets obj's checkmark. It returns true if obj was
+// already checkmarked.
+func setCheckmark(obj, base, off uintptr, mbits markBits) bool {
+	if !mbits.isMarked() {
+		printlock()
+		print("runtime: checkmarks found unexpected unmarked object obj=", hex(obj), "\n")
+		print("runtime: found obj at *(", hex(base), "+", hex(off), ")\n")
+
+		// Dump the source (base) object
+		gcDumpObject("base", base, off)
+
+		// Dump the object
+		gcDumpObject("obj", obj, ^uintptr(0))
+
+		getg().m.traceback = 2
+		throw("checkmark found unmarked object")
+	}
+
+	ai := arenaIndex(obj)
+	arena := mheap_.arenas[ai.l1()][ai.l2()]
+	arenaWord := (obj / heapArenaBytes / 8) % uintptr(len(arena.checkmarks))
+	mask := byte(1 << ((obj / heapArenaBytes) % 8))
+	bytep := &arena.checkmarks[arenaWord]
+
+	if atomic.Load8(bytep)&mask != 0 {
+		// Already checkmarked.
+		return true
+	}
+
+	atomic.Or8(bytep, mask)
+	return false
+}
diff --git a/src/runtime/mgc.go b/src/runtime/mgc.go
index b3499516f6..bd87144355 100644
--- a/src/runtime/mgc.go
+++ b/src/runtime/mgc.go
@@ -1670,13 +1670,13 @@ func gcMarkTermination(nextTriggerRatio float64) {
 			// mark using checkmark bits, to check that we
 			// didn't forget to mark anything during the
 			// concurrent mark process.
+			startCheckmarks()
 			gcResetMarkState()
-			initCheckmarks()
 			gcw := &getg().m.p.ptr().gcw
 			gcDrain(gcw, 0)
 			wbBufFlush1(getg().m.p.ptr())
 			gcw.dispose()
-			clearCheckmarks()
+			endCheckmarks()
 		}
 
 		// marking is complete so we can turn the write barrier off
@@ -2149,21 +2149,13 @@ func gcSweep(mode gcMode) {
 	lock(&mheap_.lock)
 	mheap_.sweepgen += 2
 	mheap_.sweepdone = 0
-	if !go115NewMCentralImpl && mheap_.sweepSpans[mheap_.sweepgen/2%2].index != 0 {
-		// We should have drained this list during the last
-		// sweep phase. We certainly need to start this phase
-		// with an empty swept list.
-		throw("non-empty swept list")
-	}
 	mheap_.pagesSwept = 0
 	mheap_.sweepArenas = mheap_.allArenas
 	mheap_.reclaimIndex = 0
 	mheap_.reclaimCredit = 0
 	unlock(&mheap_.lock)
 
-	if go115NewMCentralImpl {
-		sweep.centralIndex.clear()
-	}
+	sweep.centralIndex.clear()
 
 	if !_ConcurrentSweep || mode == gcForceBlockMode {
 		// Special case synchronous sweep.
diff --git a/src/runtime/mgcmark.go b/src/runtime/mgcmark.go
index fe988c46d9..2b84945471 100644
--- a/src/runtime/mgcmark.go
+++ b/src/runtime/mgcmark.go
@@ -47,10 +47,6 @@ const (
 	// Must be a multiple of the pageInUse bitmap element size and
 	// must also evenly divide pagesPerArena.
 	pagesPerSpanRoot = 512
-
-	// go115NewMarkrootSpans is a feature flag that indicates whether
-	// to use the new bitmap-based markrootSpans implementation.
-	go115NewMarkrootSpans = true
 )
 
 // gcMarkRootPrepare queues root scanning jobs (stacks, globals, and
@@ -87,24 +83,16 @@ func gcMarkRootPrepare() {
 	//
 	// We depend on addfinalizer to mark objects that get
 	// finalizers after root marking.
-	if go115NewMarkrootSpans {
-		// We're going to scan the whole heap (that was available at the time the
-		// mark phase started, i.e. markArenas) for in-use spans which have specials.
-		//
-		// Break up the work into arenas, and further into chunks.
-		//
-		// Snapshot allArenas as markArenas. This snapshot is safe because allArenas
-		// is append-only.
-		mheap_.markArenas = mheap_.allArenas[:len(mheap_.allArenas):len(mheap_.allArenas)]
-		work.nSpanRoots = len(mheap_.markArenas) * (pagesPerArena / pagesPerSpanRoot)
-	} else {
-		// We're only interested in scanning the in-use spans,
-		// which will all be swept at this point. More spans
-		// may be added to this list during concurrent GC, but
-		// we only care about spans that were allocated before
-		// this mark phase.
-		work.nSpanRoots = mheap_.sweepSpans[mheap_.sweepgen/2%2].numBlocks()
-	}
+	//
+	// We're going to scan the whole heap (that was available at the time the
+	// mark phase started, i.e. markArenas) for in-use spans which have specials.
+	//
+	// Break up the work into arenas, and further into chunks.
+	//
+	// Snapshot allArenas as markArenas. This snapshot is safe because allArenas
+	// is append-only.
+	mheap_.markArenas = mheap_.allArenas[:len(mheap_.allArenas):len(mheap_.allArenas)]
+	work.nSpanRoots = len(mheap_.markArenas) * (pagesPerArena / pagesPerSpanRoot)
 
 	// Scan stacks.
 	//
@@ -316,10 +304,6 @@ func markrootFreeGStacks() {
 //
 //go:nowritebarrier
 func markrootSpans(gcw *gcWork, shard int) {
-	if !go115NewMarkrootSpans {
-		oldMarkrootSpans(gcw, shard)
-		return
-	}
 	// Objects with finalizers have two GC-related invariants:
 	//
 	// 1) Everything reachable from the object must be marked.
@@ -396,90 +380,6 @@ func markrootSpans(gcw *gcWork, shard int) {
 	}
 }
 
-// oldMarkrootSpans marks roots for one shard of work.spans.
-//
-// For go115NewMarkrootSpans = false.
-//
-//go:nowritebarrier
-func oldMarkrootSpans(gcw *gcWork, shard int) {
-	// Objects with finalizers have two GC-related invariants:
-	//
-	// 1) Everything reachable from the object must be marked.
-	// This ensures that when we pass the object to its finalizer,
-	// everything the finalizer can reach will be retained.
-	//
-	// 2) Finalizer specials (which are not in the garbage
-	// collected heap) are roots. In practice, this means the fn
-	// field must be scanned.
-	//
-	// TODO(austin): There are several ideas for making this more
-	// efficient in issue #11485.
-
-	sg := mheap_.sweepgen
-	spans := mheap_.sweepSpans[mheap_.sweepgen/2%2].block(shard)
-	// Note that work.spans may not include spans that were
-	// allocated between entering the scan phase and now. We may
-	// also race with spans being added into sweepSpans when they're
-	// just created, and as a result we may see nil pointers in the
-	// spans slice. This is okay because any objects with finalizers
-	// in those spans must have been allocated and given finalizers
-	// after we entered the scan phase, so addfinalizer will have
-	// ensured the above invariants for them.
-	for i := 0; i < len(spans); i++ {
-		// sweepBuf.block requires that we read pointers from the block atomically.
-		// It also requires that we ignore nil pointers.
-		s := (*mspan)(atomic.Loadp(unsafe.Pointer(&spans[i])))
-
-		// This is racing with spans being initialized, so
-		// check the state carefully.
-		if s == nil || s.state.get() != mSpanInUse {
-			continue
-		}
-		// Check that this span was swept (it may be cached or uncached).
-		if !useCheckmark && !(s.sweepgen == sg || s.sweepgen == sg+3) {
-			// sweepgen was updated (+2) during non-checkmark GC pass
-			print("sweep ", s.sweepgen, " ", sg, "\n")
-			throw("gc: unswept span")
-		}
-
-		// Speculatively check if there are any specials
-		// without acquiring the span lock. This may race with
-		// adding the first special to a span, but in that
-		// case addfinalizer will observe that the GC is
-		// active (which is globally synchronized) and ensure
-		// the above invariants. We may also ensure the
-		// invariants, but it's okay to scan an object twice.
-		if s.specials == nil {
-			continue
-		}
-
-		// Lock the specials to prevent a special from being
-		// removed from the list while we're traversing it.
-		lock(&s.speciallock)
-
-		for sp := s.specials; sp != nil; sp = sp.next {
-			if sp.kind != _KindSpecialFinalizer {
-				continue
-			}
-			// don't mark finalized object, but scan it so we
-			// retain everything it points to.
-			spf := (*specialfinalizer)(unsafe.Pointer(sp))
-			// A finalizer can be set for an inner byte of an object, find object beginning.
-			p := s.base() + uintptr(spf.special.offset)/s.elemsize*s.elemsize
-
-			// Mark everything that can be reached from
-			// the object (but *not* the object itself or
-			// we'll never collect it).
-			scanobject(p, gcw)
-
-			// The special itself is a root.
-			scanblock(uintptr(unsafe.Pointer(&spf.fn)), sys.PtrSize, &oneptrmask[0], gcw, nil)
-		}
-
-		unlock(&s.speciallock)
-	}
-}
-
 // gcAssistAlloc performs GC work to make gp's assist debt positive.
 // gp must be the calling user gorountine.
 //
@@ -1354,11 +1254,7 @@ func scanobject(b uintptr, gcw *gcWork) {
 		}
 		// Load bits once. See CL 22712 and issue 16973 for discussion.
 		bits := hbits.bits()
-		// During checkmarking, 1-word objects store the checkmark
-		// in the type bit for the one word. The only one-word objects
-		// are pointers, or else they'd be merged with other non-pointer
-		// data into larger allocations.
-		if i != 1*sys.PtrSize && bits&bitScan == 0 {
+		if bits&bitScan == 0 {
 			break // no more pointers in this object
 		}
 		if bits&bitPointer == 0 {
@@ -1511,28 +1407,10 @@ func greyobject(obj, base, off uintptr, span *mspan, gcw *gcWork, objIndex uintp
 	mbits := span.markBitsForIndex(objIndex)
 
 	if useCheckmark {
-		if !mbits.isMarked() {
-			printlock()
-			print("runtime:greyobject: checkmarks finds unexpected unmarked object obj=", hex(obj), "\n")
-			print("runtime: found obj at *(", hex(base), "+", hex(off), ")\n")
-
-			// Dump the source (base) object
-			gcDumpObject("base", base, off)
-
-			// Dump the object
-			gcDumpObject("obj", obj, ^uintptr(0))
-
-			getg().m.traceback = 2
-			throw("checkmark found unmarked object")
-		}
-		hbits := heapBitsForAddr(obj)
-		if hbits.isCheckmarked(span.elemsize) {
+		if setCheckmark(obj, base, off, mbits) {
+			// Already marked.
 			return
 		}
-		hbits.setCheckmarked(span.elemsize)
-		if !hbits.isCheckmarked(span.elemsize) {
-			throw("setCheckmarked and isCheckmarked disagree")
-		}
 	} else {
 		if debug.gccheckmark > 0 && span.isFree(objIndex) {
 			print("runtime: marking free object ", hex(obj), " found at *(", hex(base), "+", hex(off), ")\n")
@@ -1661,45 +1539,3 @@ func gcMarkTinyAllocs() {
 		greyobject(c.tiny, 0, 0, span, gcw, objIndex)
 	}
 }
-
-// Checkmarking
-
-// To help debug the concurrent GC we remark with the world
-// stopped ensuring that any object encountered has their normal
-// mark bit set. To do this we use an orthogonal bit
-// pattern to indicate the object is marked. The following pattern
-// uses the upper two bits in the object's boundary nibble.
-// 01: scalar  not marked
-// 10: pointer not marked
-// 11: pointer     marked
-// 00: scalar      marked
-// Xoring with 01 will flip the pattern from marked to unmarked and vica versa.
-// The higher bit is 1 for pointers and 0 for scalars, whether the object
-// is marked or not.
-// The first nibble no longer holds the typeDead pattern indicating that the
-// there are no more pointers in the object. This information is held
-// in the second nibble.
-
-// If useCheckmark is true, marking of an object uses the
-// checkmark bits (encoding above) instead of the standard
-// mark bits.
-var useCheckmark = false
-
-//go:nowritebarrier
-func initCheckmarks() {
-	useCheckmark = true
-	for _, s := range mheap_.allspans {
-		if s.state.get() == mSpanInUse {
-			heapBitsForAddr(s.base()).initCheckmarkSpan(s.layout())
-		}
-	}
-}
-
-func clearCheckmarks() {
-	useCheckmark = false
-	for _, s := range mheap_.allspans {
-		if s.state.get() == mSpanInUse {
-			heapBitsForAddr(s.base()).clearCheckmarkSpan(s.layout())
-		}
-	}
-}
diff --git a/src/runtime/mgcsweep.go b/src/runtime/mgcsweep.go
index 3aa3afc028..6b8c56ce35 100644
--- a/src/runtime/mgcsweep.go
+++ b/src/runtime/mgcsweep.go
@@ -132,17 +132,15 @@ func finishsweep_m() {
 		sweep.npausesweep++
 	}
 
-	if go115NewMCentralImpl {
-		// Reset all the unswept buffers, which should be empty.
-		// Do this in sweep termination as opposed to mark termination
-		// so that we can catch unswept spans and reclaim blocks as
-		// soon as possible.
-		sg := mheap_.sweepgen
-		for i := range mheap_.central {
-			c := &mheap_.central[i].mcentral
-			c.partialUnswept(sg).reset()
-			c.fullUnswept(sg).reset()
-		}
+	// Reset all the unswept buffers, which should be empty.
+	// Do this in sweep termination as opposed to mark termination
+	// so that we can catch unswept spans and reclaim blocks as
+	// soon as possible.
+	sg := mheap_.sweepgen
+	for i := range mheap_.central {
+		c := &mheap_.central[i].mcentral
+		c.partialUnswept(sg).reset()
+		c.fullUnswept(sg).reset()
 	}
 
 	// Sweeping is done, so if the scavenger isn't already awake,
@@ -202,11 +200,7 @@ func sweepone() uintptr {
 	var s *mspan
 	sg := mheap_.sweepgen
 	for {
-		if go115NewMCentralImpl {
-			s = mheap_.nextSpanForSweep()
-		} else {
-			s = mheap_.sweepSpans[1-sg/2%2].pop()
-		}
+		s = mheap_.nextSpanForSweep()
 		if s == nil {
 			atomic.Store(&mheap_.sweepdone, 1)
 			break
@@ -322,9 +316,6 @@ func (s *mspan) ensureSwept() {
 // If preserve=true, don't return it to heap nor relink in mcentral lists;
 // caller takes care of it.
 func (s *mspan) sweep(preserve bool) bool {
-	if !go115NewMCentralImpl {
-		return s.oldSweep(preserve)
-	}
 	// It's critical that we enter this function with preemption disabled,
 	// GC must not start while we are in the middle of this function.
 	_g_ := getg()
@@ -568,214 +559,6 @@ func (s *mspan) sweep(preserve bool) bool {
 	return false
 }
 
-// Sweep frees or collects finalizers for blocks not marked in the mark phase.
-// It clears the mark bits in preparation for the next GC round.
-// Returns true if the span was returned to heap.
-// If preserve=true, don't return it to heap nor relink in mcentral lists;
-// caller takes care of it.
-//
-// For !go115NewMCentralImpl.
-func (s *mspan) oldSweep(preserve bool) bool {
-	// It's critical that we enter this function with preemption disabled,
-	// GC must not start while we are in the middle of this function.
-	_g_ := getg()
-	if _g_.m.locks == 0 && _g_.m.mallocing == 0 && _g_ != _g_.m.g0 {
-		throw("mspan.sweep: m is not locked")
-	}
-	sweepgen := mheap_.sweepgen
-	if state := s.state.get(); state != mSpanInUse || s.sweepgen != sweepgen-1 {
-		print("mspan.sweep: state=", state, " sweepgen=", s.sweepgen, " mheap.sweepgen=", sweepgen, "\n")
-		throw("mspan.sweep: bad span state")
-	}
-
-	if trace.enabled {
-		traceGCSweepSpan(s.npages * _PageSize)
-	}
-
-	atomic.Xadd64(&mheap_.pagesSwept, int64(s.npages))
-
-	spc := s.spanclass
-	size := s.elemsize
-	res := false
-
-	c := _g_.m.p.ptr().mcache
-	freeToHeap := false
-
-	// The allocBits indicate which unmarked objects don't need to be
-	// processed since they were free at the end of the last GC cycle
-	// and were not allocated since then.
-	// If the allocBits index is >= s.freeindex and the bit
-	// is not marked then the object remains unallocated
-	// since the last GC.
-	// This situation is analogous to being on a freelist.
-
-	// Unlink & free special records for any objects we're about to free.
-	// Two complications here:
-	// 1. An object can have both finalizer and profile special records.
-	//    In such case we need to queue finalizer for execution,
-	//    mark the object as live and preserve the profile special.
-	// 2. A tiny object can have several finalizers setup for different offsets.
-	//    If such object is not marked, we need to queue all finalizers at once.
-	// Both 1 and 2 are possible at the same time.
-	hadSpecials := s.specials != nil
-	specialp := &s.specials
-	special := *specialp
-	for special != nil {
-		// A finalizer can be set for an inner byte of an object, find object beginning.
-		objIndex := uintptr(special.offset) / size
-		p := s.base() + objIndex*size
-		mbits := s.markBitsForIndex(objIndex)
-		if !mbits.isMarked() {
-			// This object is not marked and has at least one special record.
-			// Pass 1: see if it has at least one finalizer.
-			hasFin := false
-			endOffset := p - s.base() + size
-			for tmp := special; tmp != nil && uintptr(tmp.offset) < endOffset; tmp = tmp.next {
-				if tmp.kind == _KindSpecialFinalizer {
-					// Stop freeing of object if it has a finalizer.
-					mbits.setMarkedNonAtomic()
-					hasFin = true
-					break
-				}
-			}
-			// Pass 2: queue all finalizers _or_ handle profile record.
-			for special != nil && uintptr(special.offset) < endOffset {
-				// Find the exact byte for which the special was setup
-				// (as opposed to object beginning).
-				p := s.base() + uintptr(special.offset)
-				if special.kind == _KindSpecialFinalizer || !hasFin {
-					// Splice out special record.
-					y := special
-					special = special.next
-					*specialp = special
-					freespecial(y, unsafe.Pointer(p), size)
-				} else {
-					// This is profile record, but the object has finalizers (so kept alive).
-					// Keep special record.
-					specialp = &special.next
-					special = *specialp
-				}
-			}
-		} else {
-			// object is still live: keep special record
-			specialp = &special.next
-			special = *specialp
-		}
-	}
-	if go115NewMarkrootSpans && hadSpecials && s.specials == nil {
-		spanHasNoSpecials(s)
-	}
-
-	if debug.allocfreetrace != 0 || debug.clobberfree != 0 || raceenabled || msanenabled {
-		// Find all newly freed objects. This doesn't have to
-		// efficient; allocfreetrace has massive overhead.
-		mbits := s.markBitsForBase()
-		abits := s.allocBitsForIndex(0)
-		for i := uintptr(0); i < s.nelems; i++ {
-			if !mbits.isMarked() && (abits.index < s.freeindex || abits.isMarked()) {
-				x := s.base() + i*s.elemsize
-				if debug.allocfreetrace != 0 {
-					tracefree(unsafe.Pointer(x), size)
-				}
-				if debug.clobberfree != 0 {
-					clobberfree(unsafe.Pointer(x), size)
-				}
-				if raceenabled {
-					racefree(unsafe.Pointer(x), size)
-				}
-				if msanenabled {
-					msanfree(unsafe.Pointer(x), size)
-				}
-			}
-			mbits.advance()
-			abits.advance()
-		}
-	}
-
-	// Count the number of free objects in this span.
-	nalloc := uint16(s.countAlloc())
-	if spc.sizeclass() == 0 && nalloc == 0 {
-		s.needzero = 1
-		freeToHeap = true
-	}
-	nfreed := s.allocCount - nalloc
-	if nalloc > s.allocCount {
-		print("runtime: nelems=", s.nelems, " nalloc=", nalloc, " previous allocCount=", s.allocCount, " nfreed=", nfreed, "\n")
-		throw("sweep increased allocation count")
-	}
-
-	s.allocCount = nalloc
-	wasempty := s.nextFreeIndex() == s.nelems
-	s.freeindex = 0 // reset allocation index to start of span.
-	if trace.enabled {
-		getg().m.p.ptr().traceReclaimed += uintptr(nfreed) * s.elemsize
-	}
-
-	// gcmarkBits becomes the allocBits.
-	// get a fresh cleared gcmarkBits in preparation for next GC
-	s.allocBits = s.gcmarkBits
-	s.gcmarkBits = newMarkBits(s.nelems)
-
-	// Initialize alloc bits cache.
-	s.refillAllocCache(0)
-
-	// We need to set s.sweepgen = h.sweepgen only when all blocks are swept,
-	// because of the potential for a concurrent free/SetFinalizer.
-	// But we need to set it before we make the span available for allocation
-	// (return it to heap or mcentral), because allocation code assumes that a
-	// span is already swept if available for allocation.
-	if freeToHeap || nfreed == 0 {
-		// The span must be in our exclusive ownership until we update sweepgen,
-		// check for potential races.
-		if state := s.state.get(); state != mSpanInUse || s.sweepgen != sweepgen-1 {
-			print("mspan.sweep: state=", state, " sweepgen=", s.sweepgen, " mheap.sweepgen=", sweepgen, "\n")
-			throw("mspan.sweep: bad span state after sweep")
-		}
-		// Serialization point.
-		// At this point the mark bits are cleared and allocation ready
-		// to go so release the span.
-		atomic.Store(&s.sweepgen, sweepgen)
-	}
-
-	if nfreed > 0 && spc.sizeclass() != 0 {
-		c.local_nsmallfree[spc.sizeclass()] += uintptr(nfreed)
-		res = mheap_.central[spc].mcentral.freeSpan(s, preserve, wasempty)
-		// mcentral.freeSpan updates sweepgen
-	} else if freeToHeap {
-		// Free large span to heap
-
-		// NOTE(rsc,dvyukov): The original implementation of efence
-		// in CL 22060046 used sysFree instead of sysFault, so that
-		// the operating system would eventually give the memory
-		// back to us again, so that an efence program could run
-		// longer without running out of memory. Unfortunately,
-		// calling sysFree here without any kind of adjustment of the
-		// heap data structures means that when the memory does
-		// come back to us, we have the wrong metadata for it, either in
-		// the mspan structures or in the garbage collection bitmap.
-		// Using sysFault here means that the program will run out of
-		// memory fairly quickly in efence mode, but at least it won't
-		// have mysterious crashes due to confused memory reuse.
-		// It should be possible to switch back to sysFree if we also
-		// implement and then call some kind of mheap.deleteSpan.
-		if debug.efence > 0 {
-			s.limit = 0 // prevent mlookup from finding this span
-			sysFault(unsafe.Pointer(s.base()), size)
-		} else {
-			mheap_.freeSpan(s)
-		}
-		c.local_nlargefree++
-		c.local_largefree += size
-		res = true
-	}
-	if !res {
-		// The span has been swept and is still in-use, so put
-		// it on the swept in-use list.
-		mheap_.sweepSpans[sweepgen/2%2].push(s)
-	}
-	return res
-}
-
 // reportZombies reports any marked but free objects in s and throws.
 //
 // This generally means one of the following:
diff --git a/src/runtime/mgcsweepbuf.go b/src/runtime/mgcsweepbuf.go
deleted file mode 100644
index 1f722c3d58..0000000000
--- a/src/runtime/mgcsweepbuf.go
+++ /dev/null
@@ -1,176 +0,0 @@
-// Copyright 2016 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
-
-import (
-	"internal/cpu"
-	"runtime/internal/atomic"
-	"runtime/internal/sys"
-	"unsafe"
-)
-
-// A gcSweepBuf is a set of *mspans.
-//
-// gcSweepBuf is safe for concurrent push operations *or* concurrent
-// pop operations, but not both simultaneously.
-type gcSweepBuf struct {
-	// A gcSweepBuf is a two-level data structure consisting of a
-	// growable spine that points to fixed-sized blocks. The spine
-	// can be accessed without locks, but adding a block or
-	// growing it requires taking the spine lock.
-	//
-	// Because each mspan covers at least 8K of heap and takes at
-	// most 8 bytes in the gcSweepBuf, the growth of the spine is
-	// quite limited.
-	//
-	// The spine and all blocks are allocated off-heap, which
-	// allows this to be used in the memory manager and avoids the
-	// need for write barriers on all of these. We never release
-	// this memory because there could be concurrent lock-free
-	// access and we're likely to reuse it anyway. (In principle,
-	// we could do this during STW.)
-
-	spineLock mutex
-	spine     unsafe.Pointer // *[N]*gcSweepBlock, accessed atomically
-	spineLen  uintptr        // Spine array length, accessed atomically
-	spineCap  uintptr        // Spine array cap, accessed under lock
-
-	// index is the first unused slot in the logical concatenation
-	// of all blocks. It is accessed atomically.
-	index uint32
-}
-
-const (
-	gcSweepBlockEntries    = 512 // 4KB on 64-bit
-	gcSweepBufInitSpineCap = 256 // Enough for 1GB heap on 64-bit
-)
-
-type gcSweepBlock struct {
-	spans [gcSweepBlockEntries]*mspan
-}
-
-// push adds span s to buffer b. push is safe to call concurrently
-// with other push operations, but NOT to call concurrently with pop.
-func (b *gcSweepBuf) push(s *mspan) {
-	// Obtain our slot.
-	cursor := uintptr(atomic.Xadd(&b.index, +1) - 1)
-	top, bottom := cursor/gcSweepBlockEntries, cursor%gcSweepBlockEntries
-
-	// Do we need to add a block?
-	spineLen := atomic.Loaduintptr(&b.spineLen)
-	var block *gcSweepBlock
-retry:
-	if top < spineLen {
-		spine := atomic.Loadp(unsafe.Pointer(&b.spine))
-		blockp := add(spine, sys.PtrSize*top)
-		block = (*gcSweepBlock)(atomic.Loadp(blockp))
-	} else {
-		// Add a new block to the spine, potentially growing
-		// the spine.
-		lock(&b.spineLock)
-		// spineLen cannot change until we release the lock,
-		// but may have changed while we were waiting.
-		spineLen = atomic.Loaduintptr(&b.spineLen)
-		if top < spineLen {
-			unlock(&b.spineLock)
-			goto retry
-		}
-
-		if spineLen == b.spineCap {
-			// Grow the spine.
-			newCap := b.spineCap * 2
-			if newCap == 0 {
-				newCap = gcSweepBufInitSpineCap
-			}
-			newSpine := persistentalloc(newCap*sys.PtrSize, cpu.CacheLineSize, &memstats.gc_sys)
-			if b.spineCap != 0 {
-				// Blocks are allocated off-heap, so
-				// no write barriers.
-				memmove(newSpine, b.spine, b.spineCap*sys.PtrSize)
-			}
-			// Spine is allocated off-heap, so no write barrier.
-			atomic.StorepNoWB(unsafe.Pointer(&b.spine), newSpine)
-			b.spineCap = newCap
-			// We can't immediately free the old spine
-			// since a concurrent push with a lower index
-			// could still be reading from it. We let it
-			// leak because even a 1TB heap would waste
-			// less than 2MB of memory on old spines. If
-			// this is a problem, we could free old spines
-			// during STW.
-		}
-
-		// Allocate a new block and add it to the spine.
-		block = (*gcSweepBlock)(persistentalloc(unsafe.Sizeof(gcSweepBlock{}), cpu.CacheLineSize, &memstats.gc_sys))
-		blockp := add(b.spine, sys.PtrSize*top)
-		// Blocks are allocated off-heap, so no write barrier.
-		atomic.StorepNoWB(blockp, unsafe.Pointer(block))
-		atomic.Storeuintptr(&b.spineLen, spineLen+1)
-		unlock(&b.spineLock)
-	}
-
-	// We have a block. Insert the span atomically, since there may be
-	// concurrent readers via the block API.
-	atomic.StorepNoWB(unsafe.Pointer(&block.spans[bottom]), unsafe.Pointer(s))
-}
-
-// pop removes and returns a span from buffer b, or nil if b is empty.
-// pop is safe to call concurrently with other pop operations, but NOT
-// to call concurrently with push.
-func (b *gcSweepBuf) pop() *mspan {
-	cursor := atomic.Xadd(&b.index, -1)
-	if int32(cursor) < 0 {
-		atomic.Xadd(&b.index, +1)
-		return nil
-	}
-
-	// There are no concurrent spine or block modifications during
-	// pop, so we can omit the atomics.
-	top, bottom := cursor/gcSweepBlockEntries, cursor%gcSweepBlockEntries
-	blockp := (**gcSweepBlock)(add(b.spine, sys.PtrSize*uintptr(top)))
-	block := *blockp
-	s := block.spans[bottom]
-	// Clear the pointer for block(i).
-	block.spans[bottom] = nil
-	return s
-}
-
-// numBlocks returns the number of blocks in buffer b. numBlocks is
-// safe to call concurrently with any other operation. Spans that have
-// been pushed prior to the call to numBlocks are guaranteed to appear
-// in some block in the range [0, numBlocks()), assuming there are no
-// intervening pops. Spans that are pushed after the call may also
-// appear in these blocks.
-func (b *gcSweepBuf) numBlocks() int {
-	return int(divRoundUp(uintptr(atomic.Load(&b.index)), gcSweepBlockEntries))
-}
-
-// block returns the spans in the i'th block of buffer b. block is
-// safe to call concurrently with push. The block may contain nil
-// pointers that must be ignored, and each entry in the block must be
-// loaded atomically.
-func (b *gcSweepBuf) block(i int) []*mspan {
-	// Perform bounds check before loading spine address since
-	// push ensures the allocated length is at least spineLen.
-	if i < 0 || uintptr(i) >= atomic.Loaduintptr(&b.spineLen) {
-		throw("block index out of range")
-	}
-
-	// Get block i.
-	spine := atomic.Loadp(unsafe.Pointer(&b.spine))
-	blockp := add(spine, sys.PtrSize*uintptr(i))
-	block := (*gcSweepBlock)(atomic.Loadp(blockp))
-
-	// Slice the block if necessary.
-	cursor := uintptr(atomic.Load(&b.index))
-	top, bottom := cursor/gcSweepBlockEntries, cursor%gcSweepBlockEntries
-	var spans []*mspan
-	if uintptr(i) < top {
-		spans = block.spans[:]
-	} else {
-		spans = block.spans[:bottom]
-	}
-	return spans
-}
diff --git a/src/runtime/mheap.go b/src/runtime/mheap.go
index 2c7bfd8a59..1a57bcd66e 100644
--- a/src/runtime/mheap.go
+++ b/src/runtime/mheap.go
@@ -42,17 +42,8 @@ const (
 	// roughly 100µs.
 	//
 	// Must be a multiple of the pageInUse bitmap element size and
-	// must also evenly divid pagesPerArena.
+	// must also evenly divide pagesPerArena.
 	pagesPerReclaimerChunk = 512
-
-	// go115NewMCentralImpl is a feature flag for the new mcentral implementation.
-	//
-	// This flag depends on go115NewMarkrootSpans because the new mcentral
-	// implementation requires that markroot spans no longer rely on mgcsweepbufs.
-	// The definition of this flag helps ensure that if there's a problem with
-	// the new markroot spans implementation and it gets turned off, that the new
-	// mcentral implementation also gets turned off so the runtime isn't broken.
-	go115NewMCentralImpl = true && go115NewMarkrootSpans
 )
 
 // Main malloc heap.
@@ -85,19 +76,6 @@ type mheap struct {
 	// access (since that may free the backing store).
 	allspans []*mspan // all spans out there
 
-	// sweepSpans contains two mspan stacks: one of swept in-use
-	// spans, and one of unswept in-use spans. These two trade
-	// roles on each GC cycle. Since the sweepgen increases by 2
-	// on each cycle, this means the swept spans are in
-	// sweepSpans[sweepgen/2%2] and the unswept spans are in
-	// sweepSpans[1-sweepgen/2%2]. Sweeping pops spans from the
-	// unswept stack and pushes spans that are still in-use on the
-	// swept stack. Likewise, allocating an in-use span pushes it
-	// on the swept stack.
-	//
-	// For !go115NewMCentralImpl.
-	sweepSpans [2]gcSweepBuf
-
 	_ uint32 // align uint64 fields on 32-bit for atomics
 
 	// Proportional sweep
@@ -220,7 +198,7 @@ type mheap struct {
 		base, end uintptr
 	}
 
-	// _ uint32 // ensure 64-bit alignment of central
+	_ uint32 // ensure 64-bit alignment of central
 
 	// central free lists for small size classes.
 	// the padding makes sure that the mcentrals are
@@ -300,6 +278,10 @@ type heapArena struct {
 	// during marking.
 	pageSpecials [pagesPerArena / 8]uint8
 
+	// checkmarks stores the debug.gccheckmark state. It is only
+	// used if debug.gccheckmark > 0.
+	checkmarks *checkmarksMap
+
 	// zeroedBase marks the first byte of the first page in this
 	// arena which hasn't been used yet and is therefore already
 	// zero. zeroedBase is relative to the arena base.
@@ -715,8 +697,6 @@ func pageIndexOf(p uintptr) (arena *heapArena, pageIdx uintptr, pageMask uint8)
 // Initialize the heap.
 func (h *mheap) init() {
 	lockInit(&h.lock, lockRankMheap)
-	lockInit(&h.sweepSpans[0].spineLock, lockRankSpine)
-	lockInit(&h.sweepSpans[1].spineLock, lockRankSpine)
 	lockInit(&h.speciallock, lockRankMheapSpecial)
 
 	h.spanalloc.init(unsafe.Sizeof(mspan{}), recordspan, unsafe.Pointer(h), &memstats.mspan_sys)
@@ -1290,16 +1270,6 @@ HaveSpan:
 	h.setSpans(s.base(), npages, s)
 
 	if !manual {
-		if !go115NewMCentralImpl {
-			// Add to swept in-use list.
-			//
-			// This publishes the span to root marking.
-			//
-			// h.sweepgen is guaranteed to only change during STW,
-			// and preemption is disabled in the page allocator.
-			h.sweepSpans[h.sweepgen/2%2].push(s)
-		}
-
 		// Mark in-use span in arena page bitmap.
 		//
 		// This publishes the span to the page sweeper, so
@@ -1701,9 +1671,7 @@ func addspecial(p unsafe.Pointer, s *special) bool {
 	s.offset = uint16(offset)
 	s.next = *t
 	*t = s
-	if go115NewMarkrootSpans {
-		spanHasSpecials(span)
-	}
+	spanHasSpecials(span)
 	unlock(&span.speciallock)
 	releasem(mp)
 
@@ -1744,7 +1712,7 @@ func removespecial(p unsafe.Pointer, kind uint8) *special {
 		}
 		t = &s.next
 	}
-	if go115NewMarkrootSpans && span.specials == nil {
+	if span.specials == nil {
 		spanHasNoSpecials(span)
 	}
 	unlock(&span.speciallock)
diff --git a/src/runtime/mkpreempt.go b/src/runtime/mkpreempt.go
index 1fe77663b9..44dea22ef3 100644
--- a/src/runtime/mkpreempt.go
+++ b/src/runtime/mkpreempt.go
@@ -131,7 +131,7 @@ func header(arch string) {
 
 func p(f string, args ...interface{}) {
 	fmted := fmt.Sprintf(f, args...)
-	fmt.Fprintf(out, "\t%s\n", strings.Replace(fmted, "\n", "\n\t", -1))
+	fmt.Fprintf(out, "\t%s\n", strings.ReplaceAll(fmted, "\n", "\n\t"))
 }
 
 func label(l string) {
diff --git a/src/runtime/mpallocbits.go b/src/runtime/mpallocbits.go
index a8011341bc..ff112300c3 100644
--- a/src/runtime/mpallocbits.go
+++ b/src/runtime/mpallocbits.go
@@ -120,84 +120,105 @@ func (b *pageBits) popcntRange(i, n uint) (s uint) {
 // sake of documentation, 0s are free pages and 1s are allocated pages.
 type pallocBits pageBits
 
-// consec8tab is a table containing the number of consecutive
-// zero bits for any uint8 value.
-//
-// The table is generated by calling consec8(i) for each
-// possible uint8 value, which is defined as:
-//
-// // consec8 counts the maximum number of consecutive 0 bits
-// // in a uint8.
-// func consec8(n uint8) int {
-// 	n = ^n
-// 	i := 0
-// 	for n != 0 {
-// 		n &= (n << 1)
-// 		i++
-// 	}
-// 	return i
-// }
-var consec8tab = [256]uint{
-	8, 7, 6, 6, 5, 5, 5, 5, 4, 4, 4, 4, 4, 4, 4, 4,
-	4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
-	5, 4, 3, 3, 2, 2, 2, 2, 3, 2, 2, 2, 2, 2, 2, 2,
-	4, 3, 2, 2, 2, 2, 2, 2, 3, 2, 2, 2, 2, 2, 2, 2,
-	6, 5, 4, 4, 3, 3, 3, 3, 3, 2, 2, 2, 2, 2, 2, 2,
-	4, 3, 2, 2, 2, 1, 1, 1, 3, 2, 1, 1, 2, 1, 1, 1,
-	5, 4, 3, 3, 2, 2, 2, 2, 3, 2, 1, 1, 2, 1, 1, 1,
-	4, 3, 2, 2, 2, 1, 1, 1, 3, 2, 1, 1, 2, 1, 1, 1,
-	7, 6, 5, 5, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3,
-	4, 3, 2, 2, 2, 2, 2, 2, 3, 2, 2, 2, 2, 2, 2, 2,
-	5, 4, 3, 3, 2, 2, 2, 2, 3, 2, 1, 1, 2, 1, 1, 1,
-	4, 3, 2, 2, 2, 1, 1, 1, 3, 2, 1, 1, 2, 1, 1, 1,
-	6, 5, 4, 4, 3, 3, 3, 3, 3, 2, 2, 2, 2, 2, 2, 2,
-	4, 3, 2, 2, 2, 1, 1, 1, 3, 2, 1, 1, 2, 1, 1, 1,
-	5, 4, 3, 3, 2, 2, 2, 2, 3, 2, 1, 1, 2, 1, 1, 1,
-	4, 3, 2, 2, 2, 1, 1, 1, 3, 2, 1, 1, 2, 1, 1, 0,
-}
-
 // summarize returns a packed summary of the bitmap in pallocBits.
 func (b *pallocBits) summarize() pallocSum {
-	// TODO(mknyszek): There may be something more clever to be done
-	// here to make the summarize operation more efficient. For example,
-	// we can compute start and end with 64-bit wide operations easily,
-	// but max is a bit more complex. Perhaps there exists some way to
-	// leverage the 64-bit start and end to our advantage?
-	var start, max, end uint
+	var start, max, cur uint
+	const notSetYet = ^uint(0) // sentinel for start value
+	start = notSetYet
 	for i := 0; i < len(b); i++ {
-		a := b[i]
-		for j := 0; j < 64; j += 8 {
-			k := uint8(a >> j)
+		x := b[i]
+		if x == 0 {
+			cur += 64
+			continue
+		}
+		t := uint(sys.TrailingZeros64(x))
+		l := uint(sys.LeadingZeros64(x))
 
-			// Compute start.
-			si := uint(sys.TrailingZeros8(k))
-			if start == uint(i*64+j) {
-				start += si
-			}
+		// Finish any region spanning the uint64s
+		cur += t
+		if start == notSetYet {
+			start = cur
+		}
+		if cur > max {
+			max = cur
+		}
+		// Final region that might span to next uint64
+		cur = l
+	}
+	if start == notSetYet {
+		// Made it all the way through without finding a single 1 bit.
+		const n = uint(64 * len(b))
+		return packPallocSum(n, n, n)
+	}
+	if cur > max {
+		max = cur
+	}
+	if max >= 64-2 {
+		// There is no way an internal run of zeros could beat max.
+		return packPallocSum(start, max, cur)
+	}
+	// Now look inside each uint64 for runs of zeros.
+	// All uint64s must be nonzero, or we would have aborted above.
+outer:
+	for i := 0; i < len(b); i++ {
+		x := b[i]
 
-			// Compute max.
-			if end+si > max {
-				max = end + si
-			}
-			if mi := consec8tab[k]; mi > max {
-				max = mi
+		// Look inside this uint64. We have a pattern like
+		// 000000 1xxxxx1 000000
+		// We need to look inside the 1xxxxx1 for any contiguous
+		// region of zeros.
+
+		// We already know the trailing zeros are no larger than max. Remove them.
+		x >>= sys.TrailingZeros64(x) & 63
+		if x&(x+1) == 0 { // no more zeros (except at the top).
+			continue
+		}
+
+		// Strategy: shrink all runs of zeros by max. If any runs of zero
+		// remain, then we've identified a larger maxiumum zero run.
+		p := max     // number of zeros we still need to shrink by.
+		k := uint(1) // current minimum length of runs of ones in x.
+		for {
+			// Shrink all runs of zeros by p places (except the top zeros).
+			for p > 0 {
+				if p <= k {
+					// Shift p ones down into the top of each run of zeros.
+					x |= x >> (p & 63)
+					if x&(x+1) == 0 { // no more zeros (except at the top).
+						continue outer
+					}
+					break
+				}
+				// Shift k ones down into the top of each run of zeros.
+				x |= x >> (k & 63)
+				if x&(x+1) == 0 { // no more zeros (except at the top).
+					continue outer
+				}
+				p -= k
+				// We've just doubled the minimum length of 1-runs.
+				// This allows us to shift farther in the next iteration.
+				k *= 2
 			}
 
-			// Compute end.
-			if k == 0 {
-				end += 8
-			} else {
-				end = uint(sys.LeadingZeros8(k))
+			// The length of the lowest-order zero run is an increment to our maximum.
+			j := uint(sys.TrailingZeros64(^x)) // count contiguous trailing ones
+			x >>= j & 63                       // remove trailing ones
+			j = uint(sys.TrailingZeros64(x))   // count contiguous trailing zeros
+			x >>= j & 63                       // remove zeros
+			max += j                           // we have a new maximum!
+			if x&(x+1) == 0 {                  // no more zeros (except at the top).
+				continue outer
 			}
+			p = j // remove j more zeros from each zero run.
 		}
 	}
-	return packPallocSum(start, max, end)
+	return packPallocSum(start, max, cur)
 }
 
 // find searches for npages contiguous free pages in pallocBits and returns
 // the index where that run starts, as well as the index of the first free page
 // it found in the search. searchIdx represents the first known free page and
-// where to begin the search from.
+// where to begin the next search from.
 //
 // If find fails to find any free space, it returns an index of ^uint(0) and
 // the new searchIdx should be ignored.
@@ -218,9 +239,10 @@ func (b *pallocBits) find(npages uintptr, searchIdx uint) (uint, uint) {
 //
 // See find for an explanation of the searchIdx parameter.
 func (b *pallocBits) find1(searchIdx uint) uint {
+	_ = b[0] // lift nil check out of loop
 	for i := searchIdx / 64; i < uint(len(b)); i++ {
 		x := b[i]
-		if x == ^uint64(0) {
+		if ^x == 0 {
 			continue
 		}
 		return i*64 + uint(sys.TrailingZeros64(^x))
@@ -242,18 +264,18 @@ func (b *pallocBits) findSmallN(npages uintptr, searchIdx uint) (uint, uint) {
 	end, newSearchIdx := uint(0), ^uint(0)
 	for i := searchIdx / 64; i < uint(len(b)); i++ {
 		bi := b[i]
-		if bi == ^uint64(0) {
+		if ^bi == 0 {
 			end = 0
 			continue
 		}
 		// First see if we can pack our allocation in the trailing
 		// zeros plus the end of the last 64 bits.
-		start := uint(sys.TrailingZeros64(bi))
 		if newSearchIdx == ^uint(0) {
 			// The new searchIdx is going to be at these 64 bits after any
 			// 1s we file, so count trailing 1s.
 			newSearchIdx = i*64 + uint(sys.TrailingZeros64(^bi))
 		}
+		start := uint(sys.TrailingZeros64(bi))
 		if end+start >= uint(npages) {
 			return i*64 - end, newSearchIdx
 		}
@@ -348,15 +370,33 @@ func (b *pallocBits) pages64(i uint) uint64 {
 // findBitRange64 returns the bit index of the first set of
 // n consecutive 1 bits. If no consecutive set of 1 bits of
 // size n may be found in c, then it returns an integer >= 64.
+// n must be > 0.
 func findBitRange64(c uint64, n uint) uint {
-	i := uint(0)
-	cont := uint(sys.TrailingZeros64(^c))
-	for cont < n && i < 64 {
-		i += cont
-		i += uint(sys.TrailingZeros64(c >> i))
-		cont = uint(sys.TrailingZeros64(^(c >> i)))
+	// This implementation is based on shrinking the length of
+	// runs of contiguous 1 bits. We remove the top n-1 1 bits
+	// from each run of 1s, then look for the first remaining 1 bit.
+	p := n - 1   // number of 1s we want to remove.
+	k := uint(1) // current minimum width of runs of 0 in c.
+	for p > 0 {
+		if p <= k {
+			// Shift p 0s down into the top of each run of 1s.
+			c &= c >> (p & 63)
+			break
+		}
+		// Shift k 0s down into the top of each run of 1s.
+		c &= c >> (k & 63)
+		if c == 0 {
+			return 64
+		}
+		p -= k
+		// We've just doubled the minimum length of 0-runs.
+		// This allows us to shift farther in the next iteration.
+		k *= 2
 	}
-	return i
+	// Find first remaining 1.
+	// Since we shrunk from the top down, the first 1 is in
+	// its correct original position.
+	return uint(sys.TrailingZeros64(c))
 }
 
 // pallocData encapsulates pallocBits and a bitmap for
diff --git a/src/runtime/mpallocbits_test.go b/src/runtime/mpallocbits_test.go
index 71a29f3b3a..5095e24220 100644
--- a/src/runtime/mpallocbits_test.go
+++ b/src/runtime/mpallocbits_test.go
@@ -101,7 +101,7 @@ func invertPallocBits(b *PallocBits) {
 
 // Ensures two packed summaries are identical, and reports a detailed description
 // of the difference if they're not.
-func checkPallocSum(t *testing.T, got, want PallocSum) {
+func checkPallocSum(t testing.TB, got, want PallocSum) {
 	if got.Start() != want.Start() {
 		t.Errorf("inconsistent start: got %d, want %d", got.Start(), want.Start())
 	}
@@ -297,17 +297,29 @@ func TestPallocBitsSummarize(t *testing.T) {
 
 // Benchmarks how quickly we can summarize a PallocBits.
 func BenchmarkPallocBitsSummarize(b *testing.B) {
-	buf0 := new(PallocBits)
-	buf1 := new(PallocBits)
-	for i := 0; i < len(buf1); i++ {
-		buf1[i] = ^uint64(0)
+	patterns := []uint64{
+		0,
+		^uint64(0),
+		0xaa,
+		0xaaaaaaaaaaaaaaaa,
+		0x80000000aaaaaaaa,
+		0xaaaaaaaa00000001,
+		0xbbbbbbbbbbbbbbbb,
+		0x80000000bbbbbbbb,
+		0xbbbbbbbb00000001,
+		0xcccccccccccccccc,
+		0x4444444444444444,
+		0x4040404040404040,
+		0x4000400040004000,
+		0x1000404044ccaaff,
 	}
-	bufa := new(PallocBits)
-	for i := 0; i < len(bufa); i++ {
-		bufa[i] = 0xaa
-	}
-	for _, buf := range []*PallocBits{buf0, buf1, bufa} {
-		b.Run(fmt.Sprintf("Unpacked%02X", buf[0]), func(b *testing.B) {
+	for _, p := range patterns {
+		buf := new(PallocBits)
+		for i := 0; i < len(buf); i++ {
+			buf[i] = p
+		}
+		b.Run(fmt.Sprintf("Unpacked%02X", p), func(b *testing.B) {
+			checkPallocSum(b, buf.Summarize(), SummarizeSlow(buf))
 			for i := 0; i < b.N; i++ {
 				buf.Summarize()
 			}
@@ -492,10 +504,9 @@ func TestFindBitRange64(t *testing.T) {
 			t.Errorf("case (%016x, %d): got %d, want %d", x, n, i, result)
 		}
 	}
-	for i := uint(0); i <= 64; i++ {
+	for i := uint(1); i <= 64; i++ {
 		check(^uint64(0), i, 0)
 	}
-	check(0, 0, 0)
 	for i := uint(1); i <= 64; i++ {
 		check(0, i, ^uint(0))
 	}
@@ -508,3 +519,33 @@ func TestFindBitRange64(t *testing.T) {
 	check(0xffff03ff0107ffff, 16, 0)
 	check(0x0fff03ff01079fff, 16, ^uint(0))
 }
+
+func BenchmarkFindBitRange64(b *testing.B) {
+	patterns := []uint64{
+		0,
+		^uint64(0),
+		0xaa,
+		0xaaaaaaaaaaaaaaaa,
+		0x80000000aaaaaaaa,
+		0xaaaaaaaa00000001,
+		0xbbbbbbbbbbbbbbbb,
+		0x80000000bbbbbbbb,
+		0xbbbbbbbb00000001,
+		0xcccccccccccccccc,
+		0x4444444444444444,
+		0x4040404040404040,
+		0x4000400040004000,
+	}
+	sizes := []uint{
+		2, 8, 32,
+	}
+	for _, pattern := range patterns {
+		for _, size := range sizes {
+			b.Run(fmt.Sprintf("Pattern%02XSize%d", pattern, size), func(b *testing.B) {
+				for i := 0; i < b.N; i++ {
+					FindBitRange64(pattern, size)
+				}
+			})
+		}
+	}
+}
diff --git a/src/runtime/os_linux.go b/src/runtime/os_linux.go
index 7b95ff2428..9702920bcf 100644
--- a/src/runtime/os_linux.go
+++ b/src/runtime/os_linux.go
@@ -249,6 +249,10 @@ func sysargs(argc int32, argv **byte) {
 	sysauxv(buf[:])
 }
 
+// startupRandomData holds random bytes initialized at startup. These come from
+// the ELF AT_RANDOM auxiliary vector.
+var startupRandomData []byte
+
 func sysauxv(auxv []uintptr) int {
 	var i int
 	for ; auxv[i] != _AT_NULL; i += 2 {
@@ -328,20 +332,11 @@ func libpreinit() {
 	initsig(true)
 }
 
-// gsignalInitQuirk, if non-nil, is called for every allocated gsignal G.
-//
-// TODO(austin): Remove this after Go 1.15 when we remove the
-// mlockGsignal workaround.
-var gsignalInitQuirk func(gsignal *g)
-
 // Called to initialize a new m (including the bootstrap m).
 // Called on the parent thread (main thread in case of bootstrap), can allocate memory.
 func mpreinit(mp *m) {
 	mp.gsignal = malg(32 * 1024) // Linux wants >= 2K
 	mp.gsignal.m = mp
-	if gsignalInitQuirk != nil {
-		gsignalInitQuirk(mp.gsignal)
-	}
 }
 
 func gettid() uint32
diff --git a/src/runtime/os_linux_x86.go b/src/runtime/os_linux_x86.go
index 97f870707d..d91fa1a0d1 100644
--- a/src/runtime/os_linux_x86.go
+++ b/src/runtime/os_linux_x86.go
@@ -7,120 +7,4 @@
 
 package runtime
 
-import (
-	"runtime/internal/atomic"
-	"unsafe"
-)
-
-//go:noescape
-func uname(utsname *new_utsname) int
-
-func mlock(addr, len uintptr) int
-
-func osArchInit() {
-	// Linux 5.2 introduced a bug that can corrupt vector
-	// registers on return from a signal if the signal stack isn't
-	// faulted in:
-	// https://bugzilla.kernel.org/show_bug.cgi?id=205663
-	//
-	// It was fixed in 5.3.15, 5.4.2, and all 5.5 and later
-	// kernels.
-	//
-	// If we're on an affected kernel, work around this issue by
-	// mlocking the top page of every signal stack. This doesn't
-	// help for signal stacks created in C, but there's not much
-	// we can do about that.
-	//
-	// TODO(austin): Remove this in Go 1.15, at which point it
-	// will be unlikely to encounter any of the affected kernels
-	// in the wild.
-
-	var uts new_utsname
-	if uname(&uts) < 0 {
-		throw("uname failed")
-	}
-	// Check for null terminator to ensure gostringnocopy doesn't
-	// walk off the end of the release string.
-	found := false
-	for _, b := range uts.release {
-		if b == 0 {
-			found = true
-			break
-		}
-	}
-	if !found {
-		return
-	}
-	rel := gostringnocopy(&uts.release[0])
-
-	major, minor, patch, ok := parseRelease(rel)
-	if !ok {
-		return
-	}
-
-	if major == 5 && minor == 4 && patch < 2 {
-		// All 5.4 versions of Ubuntu are patched.
-		procVersion := []byte("/proc/version\000")
-		f := open(&procVersion[0], _O_RDONLY, 0)
-		if f >= 0 {
-			var buf [512]byte
-			p := noescape(unsafe.Pointer(&buf[0]))
-			n := read(f, p, int32(len(buf)))
-			closefd(f)
-
-			needle := []byte("Ubuntu")
-		contains:
-			for i, c := range buf[:n] {
-				if c != needle[0] {
-					continue
-				}
-				if int(n)-i < len(needle) {
-					break
-				}
-				for j, c2 := range needle {
-					if c2 != buf[i+j] {
-						continue contains
-					}
-				}
-				// This is an Ubuntu system.
-				return
-			}
-		}
-	}
-
-	if major == 5 && (minor == 2 || minor == 3 && patch < 15 || minor == 4 && patch < 2) {
-		gsignalInitQuirk = mlockGsignal
-		if m0.gsignal != nil {
-			throw("gsignal quirk too late")
-		}
-		throwReportQuirk = throwBadKernel
-	}
-}
-
-func mlockGsignal(gsignal *g) {
-	if atomic.Load(&touchStackBeforeSignal) != 0 {
-		// mlock has already failed, don't try again.
-		return
-	}
-
-	// This mlock call may fail, but we don't report the failure.
-	// Instead, if something goes badly wrong, we rely on prepareSignalM
-	// and throwBadKernel to do further mitigation and to report a problem
-	// to the user if mitigation fails. This is because many
-	// systems have a limit on the total mlock size, and many kernels
-	// that appear to have bad versions are actually patched to avoid the
-	// bug described above. We want Go 1.14 to run on those systems.
-	// See #37436.
-	if errno := mlock(gsignal.stack.hi-physPageSize, physPageSize); errno < 0 {
-		atomic.Store(&touchStackBeforeSignal, uint32(-errno))
-	}
-}
-
-// throwBadKernel is called, via throwReportQuirk, by throw.
-func throwBadKernel() {
-	if errno := atomic.Load(&touchStackBeforeSignal); errno != 0 {
-		println("runtime: note: your Linux kernel may be buggy")
-		println("runtime: note: see https://golang.org/wiki/LinuxKernelSignalVectorBug")
-		println("runtime: note: mlock workaround for kernel bug failed with errno", errno)
-	}
-}
+func osArchInit() {}
diff --git a/src/runtime/os_netbsd.go b/src/runtime/os_netbsd.go
index 97106c7b9d..f7f90cedc1 100644
--- a/src/runtime/os_netbsd.go
+++ b/src/runtime/os_netbsd.go
@@ -95,18 +95,28 @@ var sigset_all = sigset{[4]uint32{^uint32(0), ^uint32(0), ^uint32(0), ^uint32(0)
 
 // From NetBSD's <sys/sysctl.h>
 const (
-	_CTL_HW      = 6
-	_HW_NCPU     = 3
-	_HW_PAGESIZE = 7
+	_CTL_HW        = 6
+	_HW_NCPU       = 3
+	_HW_PAGESIZE   = 7
+	_HW_NCPUONLINE = 16
 )
 
-func getncpu() int32 {
-	mib := [2]uint32{_CTL_HW, _HW_NCPU}
-	out := uint32(0)
+func sysctlInt(mib []uint32) (int32, bool) {
+	var out int32
 	nout := unsafe.Sizeof(out)
-	ret := sysctl(&mib[0], 2, (*byte)(unsafe.Pointer(&out)), &nout, nil, 0)
-	if ret >= 0 {
-		return int32(out)
+	ret := sysctl(&mib[0], uint32(len(mib)), (*byte)(unsafe.Pointer(&out)), &nout, nil, 0)
+	if ret < 0 {
+		return 0, false
+	}
+	return out, true
+}
+
+func getncpu() int32 {
+	if n, ok := sysctlInt([]uint32{_CTL_HW, _HW_NCPUONLINE}); ok {
+		return int32(n)
+	}
+	if n, ok := sysctlInt([]uint32{_CTL_HW, _HW_NCPU}); ok {
+		return int32(n)
 	}
 	return 1
 }
diff --git a/src/runtime/os_plan9.go b/src/runtime/os_plan9.go
index 9e187d2220..128c30adeb 100644
--- a/src/runtime/os_plan9.go
+++ b/src/runtime/os_plan9.go
@@ -82,10 +82,10 @@ func sigpanic() {
 	note := gostringnocopy((*byte)(unsafe.Pointer(g.m.notesig)))
 	switch g.sig {
 	case _SIGRFAULT, _SIGWFAULT:
-		i := index(note, "addr=")
+		i := indexNoFloat(note, "addr=")
 		if i >= 0 {
 			i += 5
-		} else if i = index(note, "va="); i >= 0 {
+		} else if i = indexNoFloat(note, "va="); i >= 0 {
 			i += 3
 		} else {
 			panicmem()
@@ -111,6 +111,20 @@ func sigpanic() {
 	}
 }
 
+// indexNoFloat is bytealg.IndexString but safe to use in a note
+// handler.
+func indexNoFloat(s, t string) int {
+	if len(t) == 0 {
+		return 0
+	}
+	for i := 0; i < len(s); i++ {
+		if s[i] == t[0] && hasPrefix(s[i:], t) {
+			return i
+		}
+	}
+	return -1
+}
+
 func atolwhex(p string) int64 {
 	for hasPrefix(p, " ") || hasPrefix(p, "\t") {
 		p = p[1:]
diff --git a/src/runtime/os_windows.go b/src/runtime/os_windows.go
index a584ada702..a62e941229 100644
--- a/src/runtime/os_windows.go
+++ b/src/runtime/os_windows.go
@@ -1010,11 +1010,6 @@ func ctrlhandler1(_type uint32) uint32 {
 	if sigsend(s) {
 		return 1
 	}
-	if !islibrary && !isarchive {
-		// Only exit the program if we don't have a DLL.
-		// See https://golang.org/issues/35965.
-		exit(2) // SIGINT, SIGTERM, etc
-	}
 	return 0
 }
 
diff --git a/src/runtime/panic.go b/src/runtime/panic.go
index 615249f33c..127843b081 100644
--- a/src/runtime/panic.go
+++ b/src/runtime/panic.go
@@ -1283,12 +1283,6 @@ func startpanic_m() bool {
 	}
 }
 
-// throwReportQuirk, if non-nil, is called by throw after dumping the stacks.
-//
-// TODO(austin): Remove this after Go 1.15 when we remove the
-// mlockGsignal workaround.
-var throwReportQuirk func()
-
 var didothers bool
 var deadlock mutex
 
@@ -1335,10 +1329,6 @@ func dopanic_m(gp *g, pc, sp uintptr) bool {
 
 	printDebugLog()
 
-	if throwReportQuirk != nil {
-		throwReportQuirk()
-	}
-
 	return docrash
 }
 
diff --git a/src/runtime/proc.go b/src/runtime/proc.go
index 035822216d..5e38b3194c 100644
--- a/src/runtime/proc.go
+++ b/src/runtime/proc.go
@@ -5,6 +5,7 @@
 package runtime
 
 import (
+	"internal/bytealg"
 	"internal/cpu"
 	"runtime/internal/atomic"
 	"runtime/internal/sys"
@@ -557,7 +558,6 @@ func schedinit() {
 
 	sched.maxmcount = 10000
 
-	tracebackinit()
 	moduledataverify()
 	stackinit()
 	mallocinit()
@@ -2575,15 +2575,20 @@ func injectglist(glist *gList) {
 		return
 	}
 
-	lock(&sched.lock)
-	npidle := int(sched.npidle)
+	npidle := int(atomic.Load(&sched.npidle))
+	var globq gQueue
 	var n int
 	for n = 0; n < npidle && !q.empty(); n++ {
-		globrunqput(q.pop())
+		g := q.pop()
+		globq.pushBack(g)
+	}
+	if n > 0 {
+		lock(&sched.lock)
+		globrunqputbatch(&globq, int32(n))
+		unlock(&sched.lock)
+		startIdle(n)
+		qsize -= n
 	}
-	unlock(&sched.lock)
-	startIdle(n)
-	qsize -= n
 
 	if !q.empty() {
 		runqputbatch(pp, &q, qsize)
@@ -5460,7 +5465,7 @@ func haveexperiment(name string) bool {
 	x := sys.Goexperiment
 	for x != "" {
 		xname := ""
-		i := index(x, ",")
+		i := bytealg.IndexByteString(x, ',')
 		if i < 0 {
 			xname, x = x, ""
 		} else {
diff --git a/src/runtime/rt0_freebsd_arm64.s b/src/runtime/rt0_freebsd_arm64.s
index 3a348c33e2..a938d98262 100644
--- a/src/runtime/rt0_freebsd_arm64.s
+++ b/src/runtime/rt0_freebsd_arm64.s
@@ -45,8 +45,7 @@ TEXT _rt0_arm64_freebsd_lib(SB),NOSPLIT,$184
 
 	// Create a new thread to do the runtime initialization and return.
 	MOVD	_cgo_sys_thread_create(SB), R4
-	CMP	$0, R4
-	BEQ	nocgo
+	CBZ	R4, nocgo
 	MOVD	$_rt0_arm64_freebsd_lib_go(SB), R0
 	MOVD	$0, R1
 	SUB	$16, RSP	// reserve 16 bytes for sp-8 where fp may be saved.
diff --git a/src/runtime/rt0_netbsd_arm64.s b/src/runtime/rt0_netbsd_arm64.s
index 75ecbe5176..2f3b5a5a87 100644
--- a/src/runtime/rt0_netbsd_arm64.s
+++ b/src/runtime/rt0_netbsd_arm64.s
@@ -44,8 +44,7 @@ TEXT _rt0_arm64_netbsd_lib(SB),NOSPLIT,$184
 
 	// Create a new thread to do the runtime initialization and return.
 	MOVD	_cgo_sys_thread_create(SB), R4
-	CMP	$0, R4
-	BEQ	nocgo
+	CBZ	R4, nocgo
 	MOVD	$_rt0_arm64_netbsd_lib_go(SB), R0
 	MOVD	$0, R1
 	SUB	$16, RSP		// reserve 16 bytes for sp-8 where fp may be saved.
diff --git a/src/runtime/rt0_openbsd_arm64.s b/src/runtime/rt0_openbsd_arm64.s
index 12408f2eec..722fab6129 100644
--- a/src/runtime/rt0_openbsd_arm64.s
+++ b/src/runtime/rt0_openbsd_arm64.s
@@ -50,8 +50,7 @@ TEXT _rt0_arm64_openbsd_lib(SB),NOSPLIT,$184
 
 	// Create a new thread to do the runtime initialization and return.
 	MOVD	_cgo_sys_thread_create(SB), R4
-	CMP	$0, R4
-	BEQ	nocgo
+	CBZ	R4, nocgo
 	MOVD	$_rt0_arm64_openbsd_lib_go(SB), R0
 	MOVD	$0, R1
 	SUB	$16, RSP		// reserve 16 bytes for sp-8 where fp may be saved.
diff --git a/src/runtime/runtime1.go b/src/runtime/runtime1.go
index c65a534ef6..7c893aa25c 100644
--- a/src/runtime/runtime1.go
+++ b/src/runtime/runtime1.go
@@ -5,6 +5,7 @@
 package runtime
 
 import (
+	"internal/bytealg"
 	"runtime/internal/atomic"
 	"runtime/internal/sys"
 	"unsafe"
@@ -347,13 +348,13 @@ func parsedebugvars() {
 
 	for p := gogetenv("GODEBUG"); p != ""; {
 		field := ""
-		i := index(p, ",")
+		i := bytealg.IndexByteString(p, ',')
 		if i < 0 {
 			field, p = p, ""
 		} else {
 			field, p = p[:i], p[i+1:]
 		}
-		i = index(field, "=")
+		i = bytealg.IndexByteString(field, '=')
 		if i < 0 {
 			continue
 		}
diff --git a/src/runtime/runtime2.go b/src/runtime/runtime2.go
index 755c409078..eba68da624 100644
--- a/src/runtime/runtime2.go
+++ b/src/runtime/runtime2.go
@@ -366,6 +366,12 @@ type sudog struct {
 	// g.selectDone must be CAS'd to win the wake-up race.
 	isSelect bool
 
+	// success indicates whether communication over channel c
+	// succeeded. It is true if the goroutine was awoken because a
+	// value was delivered over channel c, and false if awoken
+	// because c was closed.
+	success bool
+
 	parent   *sudog // semaRoot binary tree
 	waitlink *sudog // g.waiting list or semaRoot
 	waittail *sudog // semaRoot
@@ -846,10 +852,6 @@ type forcegcstate struct {
 	idle uint32
 }
 
-// startup_random_data holds random bytes initialized at startup. These come from
-// the ELF AT_RANDOM auxiliary vector (vdso_linux_amd64.go or os_linux_386.go).
-var startupRandomData []byte
-
 // extendRandom extends the random numbers in r[:n] to the whole slice r.
 // Treats n<0 as n==0.
 func extendRandom(r []byte, n int) {
diff --git a/src/runtime/select.go b/src/runtime/select.go
index a069e3e050..80768b285b 100644
--- a/src/runtime/select.go
+++ b/src/runtime/select.go
@@ -12,25 +12,12 @@ import (
 
 const debugSelect = false
 
-// scase.kind values.
-// Known to compiler.
-// Changes here must also be made in src/cmd/compile/internal/gc/select.go's walkselectcases.
-const (
-	caseNil = iota
-	caseRecv
-	caseSend
-	caseDefault
-)
-
 // Select case descriptor.
 // Known to compiler.
 // Changes here must also be made in src/cmd/internal/gc/select.go's scasetype.
 type scase struct {
-	c           *hchan         // chan
-	elem        unsafe.Pointer // data element
-	kind        uint16
-	pc          uintptr // race pc (for race detector / msan)
-	releasetime int64
+	c    *hchan         // chan
+	elem unsafe.Pointer // data element
 }
 
 var (
@@ -38,15 +25,15 @@ var (
 	chanrecvpc = funcPC(chanrecv)
 )
 
-func selectsetpc(cas *scase) {
-	cas.pc = getcallerpc()
+func selectsetpc(pc *uintptr) {
+	*pc = getcallerpc()
 }
 
 func sellock(scases []scase, lockorder []uint16) {
 	var c *hchan
 	for _, o := range lockorder {
 		c0 := scases[o].c
-		if c0 != nil && c0 != c {
+		if c0 != c {
 			c = c0
 			lock(&c.lock)
 		}
@@ -62,11 +49,8 @@ func selunlock(scases []scase, lockorder []uint16) {
 	// the G that calls select runnable again and schedules it for execution.
 	// When the G runs on another M, it locks all the locks and frees sel.
 	// Now if the first M touches sel, it will access freed memory.
-	for i := len(scases) - 1; i >= 0; i-- {
+	for i := len(lockorder) - 1; i >= 0; i-- {
 		c := scases[lockorder[i]].c
-		if c == nil {
-			break
-		}
 		if i > 0 && c == scases[lockorder[i-1]].c {
 			continue // will unlock it on the next iteration
 		}
@@ -112,11 +96,15 @@ func block() {
 // Both reside on the goroutine's stack (regardless of any escaping in
 // selectgo).
 //
+// For race detector builds, pc0 points to an array of type
+// [ncases]uintptr (also on the stack); for other builds, it's set to
+// nil.
+//
 // selectgo returns the index of the chosen scase, which matches the
 // ordinal position of its respective select{recv,send,default} call.
 // Also, if the chosen scase was a receive operation, it reports whether
 // a value was received.
-func selectgo(cas0 *scase, order0 *uint16, ncases int) (int, bool) {
+func selectgo(cas0 *scase, order0 *uint16, pc0 *uintptr, nsends, nrecvs int, block bool) (int, bool) {
 	if debugSelect {
 		print("select: cas0=", cas0, "\n")
 	}
@@ -126,25 +114,29 @@ func selectgo(cas0 *scase, order0 *uint16, ncases int) (int, bool) {
 	cas1 := (*[1 << 16]scase)(unsafe.Pointer(cas0))
 	order1 := (*[1 << 17]uint16)(unsafe.Pointer(order0))
 
+	ncases := nsends + nrecvs
 	scases := cas1[:ncases:ncases]
 	pollorder := order1[:ncases:ncases]
 	lockorder := order1[ncases:][:ncases:ncases]
 
-	// Replace send/receive cases involving nil channels with
-	// caseNil so logic below can assume non-nil channel.
-	for i := range scases {
-		cas := &scases[i]
-		if cas.c == nil && cas.kind != caseDefault {
-			*cas = scase{}
+	// Even when raceenabled is true, there might be select
+	// statements in packages compiled without -race (e.g.,
+	// ensureSigM in runtime/signal_unix.go).
+	var pcs []uintptr
+	if raceenabled && pc0 != nil {
+		pc1 := (*[1 << 16]uintptr)(unsafe.Pointer(pc0))
+		pcs = pc1[:ncases:ncases]
+	}
+	casePC := func(casi int) uintptr {
+		if pcs == nil {
+			return 0
 		}
+		return pcs[casi]
 	}
 
 	var t0 int64
 	if blockprofilerate > 0 {
 		t0 = cputicks()
-		for i := 0; i < ncases; i++ {
-			scases[i].releasetime = -1
-		}
 	}
 
 	// The compiler rewrites selects that statically have
@@ -156,15 +148,27 @@ func selectgo(cas0 *scase, order0 *uint16, ncases int) (int, bool) {
 	// optimizing (and needing to test).
 
 	// generate permuted order
-	for i := 1; i < ncases; i++ {
-		j := fastrandn(uint32(i + 1))
-		pollorder[i] = pollorder[j]
+	norder := 0
+	for i := range scases {
+		cas := &scases[i]
+
+		// Omit cases without channels from the poll and lock orders.
+		if cas.c == nil {
+			cas.elem = nil // allow GC
+			continue
+		}
+
+		j := fastrandn(uint32(norder + 1))
+		pollorder[norder] = pollorder[j]
 		pollorder[j] = uint16(i)
+		norder++
 	}
+	pollorder = pollorder[:norder]
+	lockorder = lockorder[:norder]
 
 	// sort the cases by Hchan address to get the locking order.
 	// simple heap sort, to guarantee n log n time and constant stack footprint.
-	for i := 0; i < ncases; i++ {
+	for i := range lockorder {
 		j := i
 		// Start with the pollorder to permute cases on the same channel.
 		c := scases[pollorder[i]].c
@@ -175,7 +179,7 @@ func selectgo(cas0 *scase, order0 *uint16, ncases int) (int, bool) {
 		}
 		lockorder[j] = pollorder[i]
 	}
-	for i := ncases - 1; i >= 0; i-- {
+	for i := len(lockorder) - 1; i >= 0; i-- {
 		o := lockorder[i]
 		c := scases[o].c
 		lockorder[i] = lockorder[0]
@@ -199,7 +203,7 @@ func selectgo(cas0 *scase, order0 *uint16, ncases int) (int, bool) {
 	}
 
 	if debugSelect {
-		for i := 0; i+1 < ncases; i++ {
+		for i := 0; i+1 < len(lockorder); i++ {
 			if scases[lockorder[i]].c.sortkey() > scases[lockorder[i+1]].c.sortkey() {
 				print("i=", i, " x=", lockorder[i], " y=", lockorder[i+1], "\n")
 				throw("select: broken sort")
@@ -221,23 +225,18 @@ func selectgo(cas0 *scase, order0 *uint16, ncases int) (int, bool) {
 		nextp  **sudog
 	)
 
-loop:
 	// pass 1 - look for something already waiting
-	var dfli int
-	var dfl *scase
 	var casi int
 	var cas *scase
+	var caseSuccess bool
+	var caseReleaseTime int64 = -1
 	var recvOK bool
-	for i := 0; i < ncases; i++ {
-		casi = int(pollorder[i])
+	for _, casei := range pollorder {
+		casi = int(casei)
 		cas = &scases[casi]
 		c = cas.c
 
-		switch cas.kind {
-		case caseNil:
-			continue
-
-		case caseRecv:
+		if casi >= nsends {
 			sg = c.sendq.dequeue()
 			if sg != nil {
 				goto recv
@@ -248,10 +247,9 @@ loop:
 			if c.closed != 0 {
 				goto rclose
 			}
-
-		case caseSend:
+		} else {
 			if raceenabled {
-				racereadpc(c.raceaddr(), cas.pc, chansendpc)
+				racereadpc(c.raceaddr(), casePC(casi), chansendpc)
 			}
 			if c.closed != 0 {
 				goto sclose
@@ -263,17 +261,12 @@ loop:
 			if c.qcount < c.dataqsiz {
 				goto bufsend
 			}
-
-		case caseDefault:
-			dfli = casi
-			dfl = cas
 		}
 	}
 
-	if dfl != nil {
+	if !block {
 		selunlock(scases, lockorder)
-		casi = dfli
-		cas = dfl
+		casi = -1
 		goto retc
 	}
 
@@ -286,9 +279,6 @@ loop:
 	for _, casei := range lockorder {
 		casi = int(casei)
 		cas = &scases[casi]
-		if cas.kind == caseNil {
-			continue
-		}
 		c = cas.c
 		sg := acquireSudog()
 		sg.g = gp
@@ -305,12 +295,10 @@ loop:
 		*nextp = sg
 		nextp = &sg.waitlink
 
-		switch cas.kind {
-		case caseRecv:
-			c.recvq.enqueue(sg)
-
-		case caseSend:
+		if casi < nsends {
 			c.sendq.enqueue(sg)
+		} else {
+			c.recvq.enqueue(sg)
 		}
 	}
 
@@ -331,6 +319,7 @@ loop:
 	// We singly-linked up the SudoGs in lock order.
 	casi = -1
 	cas = nil
+	caseSuccess = false
 	sglist = gp.waiting
 	// Clear all elem before unlinking from gp.waiting.
 	for sg1 := gp.waiting; sg1 != nil; sg1 = sg1.waitlink {
@@ -342,19 +331,17 @@ loop:
 
 	for _, casei := range lockorder {
 		k = &scases[casei]
-		if k.kind == caseNil {
-			continue
-		}
-		if sglist.releasetime > 0 {
-			k.releasetime = sglist.releasetime
-		}
 		if sg == sglist {
 			// sg has already been dequeued by the G that woke us up.
 			casi = int(casei)
 			cas = k
+			caseSuccess = sglist.success
+			if sglist.releasetime > 0 {
+				caseReleaseTime = sglist.releasetime
+			}
 		} else {
 			c = k.c
-			if k.kind == caseSend {
+			if int(casei) < nsends {
 				c.sendq.dequeueSudoG(sglist)
 			} else {
 				c.recvq.dequeueSudoG(sglist)
@@ -367,40 +354,35 @@ loop:
 	}
 
 	if cas == nil {
-		// We can wake up with gp.param == nil (so cas == nil)
-		// when a channel involved in the select has been closed.
-		// It is easiest to loop and re-run the operation;
-		// we'll see that it's now closed.
-		// Maybe some day we can signal the close explicitly,
-		// but we'd have to distinguish close-on-reader from close-on-writer.
-		// It's easiest not to duplicate the code and just recheck above.
-		// We know that something closed, and things never un-close,
-		// so we won't block again.
-		goto loop
+		throw("selectgo: bad wakeup")
 	}
 
 	c = cas.c
 
 	if debugSelect {
-		print("wait-return: cas0=", cas0, " c=", c, " cas=", cas, " kind=", cas.kind, "\n")
+		print("wait-return: cas0=", cas0, " c=", c, " cas=", cas, " send=", casi < nsends, "\n")
 	}
 
-	if cas.kind == caseRecv {
-		recvOK = true
+	if casi < nsends {
+		if !caseSuccess {
+			goto sclose
+		}
+	} else {
+		recvOK = caseSuccess
 	}
 
 	if raceenabled {
-		if cas.kind == caseRecv && cas.elem != nil {
-			raceWriteObjectPC(c.elemtype, cas.elem, cas.pc, chanrecvpc)
-		} else if cas.kind == caseSend {
-			raceReadObjectPC(c.elemtype, cas.elem, cas.pc, chansendpc)
+		if casi < nsends {
+			raceReadObjectPC(c.elemtype, cas.elem, casePC(casi), chansendpc)
+		} else if cas.elem != nil {
+			raceWriteObjectPC(c.elemtype, cas.elem, casePC(casi), chanrecvpc)
 		}
 	}
 	if msanenabled {
-		if cas.kind == caseRecv && cas.elem != nil {
-			msanwrite(cas.elem, c.elemtype.size)
-		} else if cas.kind == caseSend {
+		if casi < nsends {
 			msanread(cas.elem, c.elemtype.size)
+		} else if cas.elem != nil {
+			msanwrite(cas.elem, c.elemtype.size)
 		}
 	}
 
@@ -411,7 +393,7 @@ bufrecv:
 	// can receive from buffer
 	if raceenabled {
 		if cas.elem != nil {
-			raceWriteObjectPC(c.elemtype, cas.elem, cas.pc, chanrecvpc)
+			raceWriteObjectPC(c.elemtype, cas.elem, casePC(casi), chanrecvpc)
 		}
 		raceacquire(chanbuf(c, c.recvx))
 		racerelease(chanbuf(c, c.recvx))
@@ -438,7 +420,7 @@ bufsend:
 	if raceenabled {
 		raceacquire(chanbuf(c, c.sendx))
 		racerelease(chanbuf(c, c.sendx))
-		raceReadObjectPC(c.elemtype, cas.elem, cas.pc, chansendpc)
+		raceReadObjectPC(c.elemtype, cas.elem, casePC(casi), chansendpc)
 	}
 	if msanenabled {
 		msanread(cas.elem, c.elemtype.size)
@@ -476,7 +458,7 @@ rclose:
 send:
 	// can send to a sleeping receiver (sg)
 	if raceenabled {
-		raceReadObjectPC(c.elemtype, cas.elem, cas.pc, chansendpc)
+		raceReadObjectPC(c.elemtype, cas.elem, casePC(casi), chansendpc)
 	}
 	if msanenabled {
 		msanread(cas.elem, c.elemtype.size)
@@ -488,8 +470,8 @@ send:
 	goto retc
 
 retc:
-	if cas.releasetime > 0 {
-		blockevent(cas.releasetime-t0, 1)
+	if caseReleaseTime > 0 {
+		blockevent(caseReleaseTime-t0, 1)
 	}
 	return casi, recvOK
 
@@ -528,23 +510,57 @@ func reflect_rselect(cases []runtimeSelect) (int, bool) {
 		block()
 	}
 	sel := make([]scase, len(cases))
-	order := make([]uint16, 2*len(cases))
-	for i := range cases {
-		rc := &cases[i]
+	orig := make([]int, len(cases))
+	nsends, nrecvs := 0, 0
+	dflt := -1
+	for i, rc := range cases {
+		var j int
 		switch rc.dir {
 		case selectDefault:
-			sel[i] = scase{kind: caseDefault}
+			dflt = i
+			continue
 		case selectSend:
-			sel[i] = scase{kind: caseSend, c: rc.ch, elem: rc.val}
+			j = nsends
+			nsends++
 		case selectRecv:
-			sel[i] = scase{kind: caseRecv, c: rc.ch, elem: rc.val}
+			nrecvs++
+			j = len(cases) - nrecvs
 		}
-		if raceenabled || msanenabled {
-			selectsetpc(&sel[i])
+
+		sel[j] = scase{c: rc.ch, elem: rc.val}
+		orig[j] = i
+	}
+
+	// Only a default case.
+	if nsends+nrecvs == 0 {
+		return dflt, false
+	}
+
+	// Compact sel and orig if necessary.
+	if nsends+nrecvs < len(cases) {
+		copy(sel[nsends:], sel[len(cases)-nrecvs:])
+		copy(orig[nsends:], orig[len(cases)-nrecvs:])
+	}
+
+	order := make([]uint16, 2*(nsends+nrecvs))
+	var pc0 *uintptr
+	if raceenabled {
+		pcs := make([]uintptr, nsends+nrecvs)
+		for i := range pcs {
+			selectsetpc(&pcs[i])
 		}
+		pc0 = &pcs[0]
 	}
 
-	return selectgo(&sel[0], &order[0], len(cases))
+	chosen, recvOK := selectgo(&sel[0], &order[0], pc0, nsends, nrecvs, dflt == -1)
+
+	// Translate chosen back to caller's ordering.
+	if chosen < 0 {
+		chosen = dflt
+	} else {
+		chosen = orig[chosen]
+	}
+	return chosen, recvOK
 }
 
 func (q *waitq) dequeueSudoG(sgp *sudog) {
diff --git a/src/runtime/signal_unix.go b/src/runtime/signal_unix.go
index dd6d79f8ec..064a0ea100 100644
--- a/src/runtime/signal_unix.go
+++ b/src/runtime/signal_unix.go
@@ -272,6 +272,12 @@ func setProcessCPUProfiler(hz int32) {
 			atomic.Storeuintptr(&fwdSig[_SIGPROF], getsig(_SIGPROF))
 			setsig(_SIGPROF, funcPC(sighandler))
 		}
+
+		var it itimerval
+		it.it_interval.tv_sec = 0
+		it.it_interval.set_usec(1000000 / hz)
+		it.it_value = it.it_interval
+		setitimer(_ITIMER_PROF, &it, nil)
 	} else {
 		// If the Go signal handler should be disabled by default,
 		// switch back to the signal handler that was installed
@@ -296,23 +302,16 @@ func setProcessCPUProfiler(hz int32) {
 				setsig(_SIGPROF, h)
 			}
 		}
+
+		setitimer(_ITIMER_PROF, &itimerval{}, nil)
 	}
 }
 
 // setThreadCPUProfiler makes any thread-specific changes required to
 // implement profiling at a rate of hz.
+// No changes required on Unix systems.
 func setThreadCPUProfiler(hz int32) {
-	var it itimerval
-	if hz == 0 {
-		setitimer(_ITIMER_PROF, &it, nil)
-	} else {
-		it.it_interval.tv_sec = 0
-		it.it_interval.set_usec(1000000 / hz)
-		it.it_value = it.it_interval
-		setitimer(_ITIMER_PROF, &it, nil)
-	}
-	_g_ := getg()
-	_g_.m.profilehz = hz
+	getg().m.profilehz = hz
 }
 
 func sigpipe() {
@@ -616,7 +615,7 @@ func sighandler(sig uint32, info *siginfo, ctxt unsafe.Pointer, gp *g) {
 		print("signal arrived during cgo execution\n")
 		gp = _g_.m.lockedg.ptr()
 	}
-	if sig == _SIGILL {
+	if sig == _SIGILL || sig == _SIGFPE {
 		// It would be nice to know how long the instruction is.
 		// Unfortunately, that's complicated to do in general (mostly for x86
 		// and s930x, but other archs have non-standard instruction lengths also).
diff --git a/src/runtime/slice_test.go b/src/runtime/slice_test.go
index e963a43dd3..cd2bc26d1e 100644
--- a/src/runtime/slice_test.go
+++ b/src/runtime/slice_test.go
@@ -1,6 +1,7 @@
 // Copyright 2011 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_test
 
 import (
diff --git a/src/runtime/string.go b/src/runtime/string.go
index 0515b56573..9a601f0094 100644
--- a/src/runtime/string.go
+++ b/src/runtime/string.go
@@ -335,22 +335,6 @@ func gostringn(p *byte, l int) string {
 	return s
 }
 
-func index(s, t string) int {
-	if len(t) == 0 {
-		return 0
-	}
-	for i := 0; i < len(s); i++ {
-		if s[i] == t[0] && hasPrefix(s[i:], t) {
-			return i
-		}
-	}
-	return -1
-}
-
-func contains(s, t string) bool {
-	return index(s, t) >= 0
-}
-
 func hasPrefix(s, prefix string) bool {
 	return len(s) >= len(prefix) && s[:len(prefix)] == prefix
 }
@@ -499,37 +483,3 @@ func gostringw(strw *uint16) string {
 	b[n2] = 0 // for luck
 	return s[:n2]
 }
-
-// parseRelease parses a dot-separated version number. It follows the
-// semver syntax, but allows the minor and patch versions to be
-// elided.
-func parseRelease(rel string) (major, minor, patch int, ok bool) {
-	// Strip anything after a dash or plus.
-	for i := 0; i < len(rel); i++ {
-		if rel[i] == '-' || rel[i] == '+' {
-			rel = rel[:i]
-			break
-		}
-	}
-
-	next := func() (int, bool) {
-		for i := 0; i < len(rel); i++ {
-			if rel[i] == '.' {
-				ver, ok := atoi(rel[:i])
-				rel = rel[i+1:]
-				return ver, ok
-			}
-		}
-		ver, ok := atoi(rel)
-		rel = ""
-		return ver, ok
-	}
-	if major, ok = next(); !ok || rel == "" {
-		return
-	}
-	if minor, ok = next(); !ok || rel == "" {
-		return
-	}
-	patch, ok = next()
-	return
-}
diff --git a/src/runtime/string_test.go b/src/runtime/string_test.go
index b9ac667533..4eda12c35d 100644
--- a/src/runtime/string_test.go
+++ b/src/runtime/string_test.go
@@ -454,34 +454,3 @@ func TestAtoi32(t *testing.T) {
 		}
 	}
 }
-
-type parseReleaseTest struct {
-	in                  string
-	major, minor, patch int
-}
-
-var parseReleaseTests = []parseReleaseTest{
-	{"", -1, -1, -1},
-	{"x", -1, -1, -1},
-	{"5", 5, 0, 0},
-	{"5.12", 5, 12, 0},
-	{"5.12-x", 5, 12, 0},
-	{"5.12.1", 5, 12, 1},
-	{"5.12.1-x", 5, 12, 1},
-	{"5.12.1.0", 5, 12, 1},
-	{"5.20496382327982653440", -1, -1, -1},
-}
-
-func TestParseRelease(t *testing.T) {
-	for _, test := range parseReleaseTests {
-		major, minor, patch, ok := runtime.ParseRelease(test.in)
-		if !ok {
-			major, minor, patch = -1, -1, -1
-		}
-		if test.major != major || test.minor != minor || test.patch != patch {
-			t.Errorf("parseRelease(%q) = (%v, %v, %v) want (%v, %v, %v)",
-				test.in, major, minor, patch,
-				test.major, test.minor, test.patch)
-		}
-	}
-}
diff --git a/src/runtime/sys_darwin.go b/src/runtime/sys_darwin.go
index 06474434c9..e4f19bbf41 100644
--- a/src/runtime/sys_darwin.go
+++ b/src/runtime/sys_darwin.go
@@ -489,9 +489,3 @@ func setNonblock(fd int32) {
 //go:cgo_import_dynamic libc_pthread_cond_wait pthread_cond_wait "/usr/lib/libSystem.B.dylib"
 //go:cgo_import_dynamic libc_pthread_cond_timedwait_relative_np pthread_cond_timedwait_relative_np "/usr/lib/libSystem.B.dylib"
 //go:cgo_import_dynamic libc_pthread_cond_signal pthread_cond_signal "/usr/lib/libSystem.B.dylib"
-
-// Magic incantation to get libSystem and friends actually dynamically linked.
-// TODO: Why does the code require this?  See cmd/link/internal/ld/go.go
-//go:cgo_import_dynamic _ _ "/usr/lib/libSystem.B.dylib"
-//go:cgo_import_dynamic _ _ "/System/Library/Frameworks/Security.framework/Versions/A/Security"
-//go:cgo_import_dynamic _ _ "/System/Library/Frameworks/CoreFoundation.framework/Versions/A/CoreFoundation"
diff --git a/src/runtime/sys_linux_386.s b/src/runtime/sys_linux_386.s
index 5b9b638ad7..1e3a834812 100644
--- a/src/runtime/sys_linux_386.s
+++ b/src/runtime/sys_linux_386.s
@@ -39,8 +39,6 @@
 #define SYS_socketcall		102
 #define SYS_setittimer		104
 #define SYS_clone		120
-#define SYS_uname		122
-#define SYS_mlock		150
 #define SYS_sched_yield 	158
 #define SYS_nanosleep		162
 #define SYS_rt_sigreturn	173
@@ -808,20 +806,3 @@ TEXT runtime·sbrk0(SB),NOSPLIT,$0-4
 	INVOKE_SYSCALL
 	MOVL	AX, ret+0(FP)
 	RET
-
-// func uname(utsname *new_utsname) int
-TEXT ·uname(SB),NOSPLIT,$0-8
-	MOVL    $SYS_uname, AX
-	MOVL    utsname+0(FP), BX
-	INVOKE_SYSCALL
-	MOVL	AX, ret+4(FP)
-	RET
-
-// func mlock(addr, len uintptr) int
-TEXT ·mlock(SB),NOSPLIT,$0-12
-	MOVL    $SYS_mlock, AX
-	MOVL    addr+0(FP), BX
-	MOVL    len+4(FP), CX
-	INVOKE_SYSCALL
-	MOVL	AX, ret+8(FP)
-	RET
diff --git a/src/runtime/sys_linux_amd64.s b/src/runtime/sys_linux_amd64.s
index fe9c6bce85..8d90813589 100644
--- a/src/runtime/sys_linux_amd64.s
+++ b/src/runtime/sys_linux_amd64.s
@@ -33,10 +33,8 @@
 #define SYS_clone		56
 #define SYS_exit		60
 #define SYS_kill		62
-#define SYS_uname		63
 #define SYS_fcntl		72
 #define SYS_sigaltstack 	131
-#define SYS_mlock		149
 #define SYS_arch_prctl		158
 #define SYS_gettid		186
 #define SYS_futex		202
@@ -214,7 +212,7 @@ TEXT runtime·walltime1(SB),NOSPLIT,$16-12
 	// due to stack probes inserted to avoid stack/heap collisions.
 	// See issue #20427.
 
-	MOVQ	SP, BP	// Save old SP; BP unchanged by C code.
+	MOVQ	SP, R12	// Save old SP; R12 unchanged by C code.
 
 	get_tls(CX)
 	MOVQ	g(CX), AX
@@ -252,7 +250,7 @@ noswitch:
 	MOVQ	0(SP), AX	// sec
 	MOVQ	8(SP), DX	// nsec
 ret:
-	MOVQ	BP, SP		// Restore real SP
+	MOVQ	R12, SP		// Restore real SP
 	// Restore vdsoPC, vdsoSP
 	// We don't worry about being signaled between the two stores.
 	// If we are not in a signal handler, we'll restore vdsoSP to 0,
@@ -279,7 +277,7 @@ fallback:
 TEXT runtime·nanotime1(SB),NOSPLIT,$16-8
 	// Switch to g0 stack. See comment above in runtime·walltime.
 
-	MOVQ	SP, BP	// Save old SP; BP unchanged by C code.
+	MOVQ	SP, R12	// Save old SP; R12 unchanged by C code.
 
 	get_tls(CX)
 	MOVQ	g(CX), AX
@@ -317,7 +315,7 @@ noswitch:
 	MOVQ	0(SP), AX	// sec
 	MOVQ	8(SP), DX	// nsec
 ret:
-	MOVQ	BP, SP		// Restore real SP
+	MOVQ	R12, SP		// Restore real SP
 	// Restore vdsoPC, vdsoSP
 	// We don't worry about being signaled between the two stores.
 	// If we are not in a signal handler, we'll restore vdsoSP to 0,
@@ -594,13 +592,25 @@ TEXT runtime·clone(SB),NOSPLIT,$0
 	MOVQ	stk+8(FP), SI
 	MOVQ	$0, DX
 	MOVQ	$0, R10
-
+	MOVQ    $0, R8
 	// Copy mp, gp, fn off parent stack for use by child.
 	// Careful: Linux system call clobbers CX and R11.
-	MOVQ	mp+16(FP), R8
+	MOVQ	mp+16(FP), R13
 	MOVQ	gp+24(FP), R9
 	MOVQ	fn+32(FP), R12
-
+	CMPQ	R13, $0    // m
+	JEQ	nog1
+	CMPQ	R9, $0    // g
+	JEQ	nog1
+	LEAQ	m_tls(R13), R8
+#ifdef GOOS_android
+	// Android stores the TLS offset in runtime·tls_g.
+	SUBQ	runtime·tls_g(SB), R8
+#else
+	ADDQ	$8, R8	// ELF wants to use -8(FS)
+#endif
+	ORQ 	$0x00080000, DI //add flag CLONE_SETTLS(0x00080000) to call clone
+nog1:
 	MOVL	$SYS_clone, AX
 	SYSCALL
 
@@ -614,27 +624,23 @@ TEXT runtime·clone(SB),NOSPLIT,$0
 	MOVQ	SI, SP
 
 	// If g or m are nil, skip Go-related setup.
-	CMPQ	R8, $0    // m
-	JEQ	nog
+	CMPQ	R13, $0    // m
+	JEQ	nog2
 	CMPQ	R9, $0    // g
-	JEQ	nog
+	JEQ	nog2
 
 	// Initialize m->procid to Linux tid
 	MOVL	$SYS_gettid, AX
 	SYSCALL
-	MOVQ	AX, m_procid(R8)
-
-	// Set FS to point at m->tls.
-	LEAQ	m_tls(R8), DI
-	CALL	runtime·settls(SB)
+	MOVQ	AX, m_procid(R13)
 
 	// In child, set up new stack
 	get_tls(CX)
-	MOVQ	R8, g_m(R9)
+	MOVQ	R13, g_m(R9)
 	MOVQ	R9, g(CX)
 	CALL	runtime·stackcheck(SB)
 
-nog:
+nog2:
 	// Call fn
 	CALL	R12
 
@@ -789,20 +795,3 @@ TEXT runtime·sbrk0(SB),NOSPLIT,$0-8
 	SYSCALL
 	MOVQ	AX, ret+0(FP)
 	RET
-
-// func uname(utsname *new_utsname) int
-TEXT ·uname(SB),NOSPLIT,$0-16
-	MOVQ    utsname+0(FP), DI
-	MOVL    $SYS_uname, AX
-	SYSCALL
-	MOVQ	AX, ret+8(FP)
-	RET
-
-// func mlock(addr, len uintptr) int
-TEXT ·mlock(SB),NOSPLIT,$0-24
-	MOVQ    addr+0(FP), DI
-	MOVQ    len+8(FP), SI
-	MOVL    $SYS_mlock, AX
-	SYSCALL
-	MOVQ	AX, ret+16(FP)
-	RET
diff --git a/src/runtime/testdata/testprog/checkptr.go b/src/runtime/testdata/testprog/checkptr.go
index 45e6fb1aa5..e0a2794f4c 100644
--- a/src/runtime/testdata/testprog/checkptr.go
+++ b/src/runtime/testdata/testprog/checkptr.go
@@ -10,6 +10,7 @@ func init() {
 	register("CheckPtrAlignmentNoPtr", CheckPtrAlignmentNoPtr)
 	register("CheckPtrAlignmentPtr", CheckPtrAlignmentPtr)
 	register("CheckPtrArithmetic", CheckPtrArithmetic)
+	register("CheckPtrArithmetic2", CheckPtrArithmetic2)
 	register("CheckPtrSize", CheckPtrSize)
 	register("CheckPtrSmall", CheckPtrSmall)
 }
@@ -32,6 +33,13 @@ func CheckPtrArithmetic() {
 	sink2 = (*int)(unsafe.Pointer(i))
 }
 
+func CheckPtrArithmetic2() {
+	var x [2]int64
+	p := unsafe.Pointer(&x[1])
+	var one uintptr = 1
+	sink2 = unsafe.Pointer(uintptr(p) & ^one)
+}
+
 func CheckPtrSize() {
 	p := new(int64)
 	sink2 = p
diff --git a/src/runtime/time.go b/src/runtime/time.go
index fdb5066b24..f895bf8443 100644
--- a/src/runtime/time.go
+++ b/src/runtime/time.go
@@ -403,7 +403,7 @@ func dodeltimer0(pp *p) {
 }
 
 // modtimer modifies an existing timer.
-// This is called by the netpoll code or time.Ticker.Reset.
+// This is called by the netpoll code or time.Ticker.Reset or time.Timer.Reset.
 // Reports whether the timer was modified before it was run.
 func modtimer(t *timer, when, period int64, f func(interface{}, uintptr), arg interface{}, seq uintptr) bool {
 	if when < 0 {
diff --git a/src/runtime/trace/trace_stack_test.go b/src/runtime/trace/trace_stack_test.go
index cfc0419b72..be3adc9801 100644
--- a/src/runtime/trace/trace_stack_test.go
+++ b/src/runtime/trace/trace_stack_test.go
@@ -252,7 +252,7 @@ func TestTraceSymbolize(t *testing.T) {
 			{trace.EvGoSysCall, []frame{
 				{"syscall.read", 0},
 				{"syscall.Read", 0},
-				{"internal/poll.ignoringEINTR", 0},
+				{"internal/poll.ignoringEINTRIO", 0},
 				{"internal/poll.(*FD).Read", 0},
 				{"os.(*File).read", 0},
 				{"os.(*File).Read", 0},
diff --git a/src/runtime/traceback.go b/src/runtime/traceback.go
index 944c8473d2..7850eceafa 100644
--- a/src/runtime/traceback.go
+++ b/src/runtime/traceback.go
@@ -5,6 +5,7 @@
 package runtime
 
 import (
+	"internal/bytealg"
 	"runtime/internal/atomic"
 	"runtime/internal/sys"
 	"unsafe"
@@ -35,16 +36,6 @@ import (
 
 const usesLR = sys.MinFrameSize > 0
 
-var skipPC uintptr
-
-func tracebackinit() {
-	// Go variable initialization happens late during runtime startup.
-	// Instead of initializing the variables above in the declarations,
-	// schedinit calls this function so that the variables are
-	// initialized and available earlier in the startup sequence.
-	skipPC = funcPC(skipPleaseUseCallersFrames)
-}
-
 // Traceback over the deferred function calls.
 // Report them like calls that have been invoked but not started executing yet.
 func tracebackdefers(gp *g, callback func(*stkframe, unsafe.Pointer) bool, v unsafe.Pointer) {
@@ -82,9 +73,6 @@ func tracebackdefers(gp *g, callback func(*stkframe, unsafe.Pointer) bool, v uns
 
 const sizeofSkipFunction = 256
 
-// This function is defined in asm.s to be sizeofSkipFunction bytes long.
-func skipPleaseUseCallersFrames()
-
 // Generic traceback. Handles runtime stack prints (pcbuf == nil),
 // the runtime.Callers function (pcbuf != nil), as well as the garbage
 // collector (callback != nil).  A little clunky to merge these, but avoids
@@ -848,7 +836,7 @@ func showfuncinfo(f funcInfo, firstFrame bool, funcID, childID funcID) bool {
 		return true
 	}
 
-	return contains(name, ".") && (!hasPrefix(name, "runtime.") || isExportedRuntime(name))
+	return bytealg.IndexByteString(name, '.') >= 0 && (!hasPrefix(name, "runtime.") || isExportedRuntime(name))
 }
 
 // isExportedRuntime reports whether name is an exported runtime function.
diff --git a/src/strconv/isprint.go b/src/strconv/isprint.go
index 7ada2d1b04..994a8e423c 100644
--- a/src/strconv/isprint.go
+++ b/src/strconv/isprint.go
@@ -6,7 +6,7 @@
 
 package strconv
 
-// (442+132+90)*2 + (450)*4 = 3128 bytes
+// (434+132+95)*2 + (468)*4 = 3194 bytes
 
 var isPrint16 = []uint16{
 	0x0020, 0x007e,
@@ -25,7 +25,7 @@ var isPrint16 = []uint16{
 	0x07fd, 0x082d,
 	0x0830, 0x085b,
 	0x085e, 0x086a,
-	0x08a0, 0x08bd,
+	0x08a0, 0x08c7,
 	0x08d3, 0x098c,
 	0x098f, 0x0990,
 	0x0993, 0x09b2,
@@ -56,7 +56,7 @@ var isPrint16 = []uint16{
 	0x0b3c, 0x0b44,
 	0x0b47, 0x0b48,
 	0x0b4b, 0x0b4d,
-	0x0b56, 0x0b57,
+	0x0b55, 0x0b57,
 	0x0b5c, 0x0b63,
 	0x0b66, 0x0b77,
 	0x0b82, 0x0b8a,
@@ -82,8 +82,7 @@ var isPrint16 = []uint16{
 	0x0ce6, 0x0cf2,
 	0x0d00, 0x0d4f,
 	0x0d54, 0x0d63,
-	0x0d66, 0x0d7f,
-	0x0d82, 0x0d96,
+	0x0d66, 0x0d96,
 	0x0d9a, 0x0dbd,
 	0x0dc0, 0x0dc6,
 	0x0dca, 0x0dca,
@@ -138,7 +137,7 @@ var isPrint16 = []uint16{
 	0x1a7f, 0x1a89,
 	0x1a90, 0x1a99,
 	0x1aa0, 0x1aad,
-	0x1ab0, 0x1abe,
+	0x1ab0, 0x1ac0,
 	0x1b00, 0x1b4b,
 	0x1b50, 0x1b7c,
 	0x1b80, 0x1bf3,
@@ -166,30 +165,27 @@ var isPrint16 = []uint16{
 	0x2190, 0x2426,
 	0x2440, 0x244a,
 	0x2460, 0x2b73,
-	0x2b76, 0x2b95,
-	0x2b98, 0x2cf3,
+	0x2b76, 0x2cf3,
 	0x2cf9, 0x2d27,
 	0x2d2d, 0x2d2d,
 	0x2d30, 0x2d67,
 	0x2d6f, 0x2d70,
 	0x2d7f, 0x2d96,
-	0x2da0, 0x2e4f,
+	0x2da0, 0x2e52,
 	0x2e80, 0x2ef3,
 	0x2f00, 0x2fd5,
 	0x2ff0, 0x2ffb,
 	0x3001, 0x3096,
 	0x3099, 0x30ff,
-	0x3105, 0x31ba,
-	0x31c0, 0x31e3,
-	0x31f0, 0x4db5,
-	0x4dc0, 0x9fef,
+	0x3105, 0x31e3,
+	0x31f0, 0x9ffc,
 	0xa000, 0xa48c,
 	0xa490, 0xa4c6,
 	0xa4d0, 0xa62b,
 	0xa640, 0xa6f7,
 	0xa700, 0xa7bf,
-	0xa7c2, 0xa7c6,
-	0xa7f7, 0xa82b,
+	0xa7c2, 0xa7ca,
+	0xa7f5, 0xa82c,
 	0xa830, 0xa839,
 	0xa840, 0xa877,
 	0xa880, 0xa8c5,
@@ -205,7 +201,7 @@ var isPrint16 = []uint16{
 	0xab01, 0xab06,
 	0xab09, 0xab0e,
 	0xab11, 0xab16,
-	0xab20, 0xab67,
+	0xab20, 0xab6b,
 	0xab70, 0xabed,
 	0xabf0, 0xabf9,
 	0xac00, 0xd7a3,
@@ -288,11 +284,11 @@ var isNotPrint16 = []uint16{
 	0x0cc9,
 	0x0cdf,
 	0x0cf0,
-	0x0d04,
 	0x0d0d,
 	0x0d11,
 	0x0d45,
 	0x0d49,
+	0x0d80,
 	0x0d84,
 	0x0db2,
 	0x0dbc,
@@ -335,6 +331,7 @@ var isNotPrint16 = []uint16{
 	0x1fdc,
 	0x1ff5,
 	0x208f,
+	0x2b96,
 	0x2c2f,
 	0x2c5f,
 	0x2d26,
@@ -351,7 +348,6 @@ var isNotPrint16 = []uint16{
 	0x3130,
 	0x318f,
 	0x321f,
-	0x32ff,
 	0xa9ce,
 	0xa9ff,
 	0xab27,
@@ -373,7 +369,7 @@ var isPrint32 = []uint32{
 	0x010080, 0x0100fa,
 	0x010100, 0x010102,
 	0x010107, 0x010133,
-	0x010137, 0x01019b,
+	0x010137, 0x01019c,
 	0x0101a0, 0x0101a0,
 	0x0101d0, 0x0101fd,
 	0x010280, 0x01029c,
@@ -424,19 +420,20 @@ var isPrint32 = []uint32{
 	0x010cc0, 0x010cf2,
 	0x010cfa, 0x010d27,
 	0x010d30, 0x010d39,
-	0x010e60, 0x010e7e,
+	0x010e60, 0x010ead,
+	0x010eb0, 0x010eb1,
 	0x010f00, 0x010f27,
 	0x010f30, 0x010f59,
+	0x010fb0, 0x010fcb,
 	0x010fe0, 0x010ff6,
 	0x011000, 0x01104d,
 	0x011052, 0x01106f,
 	0x01107f, 0x0110c1,
 	0x0110d0, 0x0110e8,
 	0x0110f0, 0x0110f9,
-	0x011100, 0x011146,
+	0x011100, 0x011147,
 	0x011150, 0x011176,
-	0x011180, 0x0111cd,
-	0x0111d0, 0x0111f4,
+	0x011180, 0x0111f4,
 	0x011200, 0x01123e,
 	0x011280, 0x0112a9,
 	0x0112b0, 0x0112ea,
@@ -451,7 +448,7 @@ var isPrint32 = []uint32{
 	0x01135d, 0x011363,
 	0x011366, 0x01136c,
 	0x011370, 0x011374,
-	0x011400, 0x01145f,
+	0x011400, 0x011461,
 	0x011480, 0x0114c7,
 	0x0114d0, 0x0114d9,
 	0x011580, 0x0115b5,
@@ -466,7 +463,11 @@ var isPrint32 = []uint32{
 	0x011730, 0x01173f,
 	0x011800, 0x01183b,
 	0x0118a0, 0x0118f2,
-	0x0118ff, 0x0118ff,
+	0x0118ff, 0x011906,
+	0x011909, 0x011909,
+	0x01190c, 0x011938,
+	0x01193b, 0x011946,
+	0x011950, 0x011959,
 	0x0119a0, 0x0119a7,
 	0x0119aa, 0x0119d7,
 	0x0119da, 0x0119e4,
@@ -483,6 +484,7 @@ var isPrint32 = []uint32{
 	0x011d60, 0x011d98,
 	0x011da0, 0x011da9,
 	0x011ee0, 0x011ef8,
+	0x011fb0, 0x011fb0,
 	0x011fc0, 0x011ff1,
 	0x011fff, 0x012399,
 	0x012400, 0x012474,
@@ -501,9 +503,11 @@ var isPrint32 = []uint32{
 	0x016f00, 0x016f4a,
 	0x016f4f, 0x016f87,
 	0x016f8f, 0x016f9f,
-	0x016fe0, 0x016fe3,
+	0x016fe0, 0x016fe4,
+	0x016ff0, 0x016ff1,
 	0x017000, 0x0187f7,
-	0x018800, 0x018af2,
+	0x018800, 0x018cd5,
+	0x018d00, 0x018d08,
 	0x01b000, 0x01b11e,
 	0x01b150, 0x01b152,
 	0x01b164, 0x01b167,
@@ -557,17 +561,15 @@ var isPrint32 = []uint32{
 	0x01f030, 0x01f093,
 	0x01f0a0, 0x01f0ae,
 	0x01f0b1, 0x01f0f5,
-	0x01f100, 0x01f10c,
-	0x01f110, 0x01f16c,
-	0x01f170, 0x01f1ac,
+	0x01f100, 0x01f1ad,
 	0x01f1e6, 0x01f202,
 	0x01f210, 0x01f23b,
 	0x01f240, 0x01f248,
 	0x01f250, 0x01f251,
 	0x01f260, 0x01f265,
-	0x01f300, 0x01f6d5,
+	0x01f300, 0x01f6d7,
 	0x01f6e0, 0x01f6ec,
-	0x01f6f0, 0x01f6fa,
+	0x01f6f0, 0x01f6fc,
 	0x01f700, 0x01f773,
 	0x01f780, 0x01f7d8,
 	0x01f7e0, 0x01f7eb,
@@ -576,22 +578,25 @@ var isPrint32 = []uint32{
 	0x01f850, 0x01f859,
 	0x01f860, 0x01f887,
 	0x01f890, 0x01f8ad,
-	0x01f900, 0x01f976,
-	0x01f97a, 0x01f9a2,
-	0x01f9a5, 0x01f9aa,
-	0x01f9ae, 0x01f9ca,
-	0x01f9cd, 0x01fa53,
+	0x01f8b0, 0x01f8b1,
+	0x01f900, 0x01fa53,
 	0x01fa60, 0x01fa6d,
-	0x01fa70, 0x01fa73,
+	0x01fa70, 0x01fa74,
 	0x01fa78, 0x01fa7a,
-	0x01fa80, 0x01fa82,
-	0x01fa90, 0x01fa95,
-	0x020000, 0x02a6d6,
+	0x01fa80, 0x01fa86,
+	0x01fa90, 0x01faa8,
+	0x01fab0, 0x01fab6,
+	0x01fac0, 0x01fac2,
+	0x01fad0, 0x01fad6,
+	0x01fb00, 0x01fbca,
+	0x01fbf0, 0x01fbf9,
+	0x020000, 0x02a6dd,
 	0x02a700, 0x02b734,
 	0x02b740, 0x02b81d,
 	0x02b820, 0x02cea1,
 	0x02ceb0, 0x02ebe0,
 	0x02f800, 0x02fa1d,
+	0x030000, 0x03134a,
 	0x0e0100, 0x0e01ef,
 }
 
@@ -609,6 +614,8 @@ var isNotPrint32 = []uint16{ // add 0x10000 to each entry
 	0x0a04,
 	0x0a14,
 	0x0a18,
+	0x0e7f,
+	0x0eaa,
 	0x10bd,
 	0x1135,
 	0x11e0,
@@ -622,8 +629,10 @@ var isNotPrint32 = []uint16{ // add 0x10000 to each entry
 	0x1331,
 	0x1334,
 	0x133a,
-	0x145a,
 	0x145c,
+	0x1914,
+	0x1917,
+	0x1936,
 	0x1c09,
 	0x1c37,
 	0x1ca8,
@@ -684,8 +693,9 @@ var isNotPrint32 = []uint16{ // add 0x10000 to each entry
 	0xeeaa,
 	0xf0c0,
 	0xf0d0,
-	0xf90c,
-	0xf972,
+	0xf979,
+	0xf9cc,
+	0xfb93,
 }
 
 // isGraphic lists the graphic runes not matched by IsPrint.
diff --git a/src/strings/strings.go b/src/strings/strings.go
index d6f5cea6e6..b429735fea 100644
--- a/src/strings/strings.go
+++ b/src/strings/strings.go
@@ -934,8 +934,8 @@ func Replace(s, old, new string, n int) string {
 	}
 
 	// Apply replacements to buffer.
-	t := make([]byte, len(s)+n*(len(new)-len(old)))
-	w := 0
+	var b Builder
+	b.Grow(len(s) + n*(len(new)-len(old)))
 	start := 0
 	for i := 0; i < n; i++ {
 		j := start
@@ -947,12 +947,12 @@ func Replace(s, old, new string, n int) string {
 		} else {
 			j += Index(s[start:], old)
 		}
-		w += copy(t[w:], s[start:j])
-		w += copy(t[w:], new)
+		b.WriteString(s[start:j])
+		b.WriteString(new)
 		start = j + len(old)
 	}
-	w += copy(t[w:], s[start:])
-	return string(t[0:w])
+	b.WriteString(s[start:])
+	return b.String()
 }
 
 // ReplaceAll returns a copy of the string s with all
diff --git a/src/strings/strings_test.go b/src/strings/strings_test.go
index c01c4dabc5..09e5b27cc3 100644
--- a/src/strings/strings_test.go
+++ b/src/strings/strings_test.go
@@ -1900,3 +1900,12 @@ func BenchmarkTrimSpace(b *testing.B) {
 		})
 	}
 }
+
+var stringSink string
+
+func BenchmarkReplaceAll(b *testing.B) {
+	b.ReportAllocs()
+	for i := 0; i < b.N; i++ {
+		stringSink = ReplaceAll("banana", "a", "<>")
+	}
+}
diff --git a/src/sync/cond_test.go b/src/sync/cond_test.go
index 9d0d9adc74..859cae59bc 100644
--- a/src/sync/cond_test.go
+++ b/src/sync/cond_test.go
@@ -1,6 +1,7 @@
 // Copyright 2011 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 sync_test
 
 import (
diff --git a/src/sync/mutex_test.go b/src/sync/mutex_test.go
index e61a853642..98c1bf2a5f 100644
--- a/src/sync/mutex_test.go
+++ b/src/sync/mutex_test.go
@@ -194,7 +194,7 @@ func TestMutexFairness(t *testing.T) {
 			}
 		}
 	}()
-	done := make(chan bool)
+	done := make(chan bool, 1)
 	go func() {
 		for i := 0; i < 10; i++ {
 			time.Sleep(100 * time.Microsecond)
diff --git a/src/syscall/asm_linux_riscv64.s b/src/syscall/asm_linux_riscv64.s
index ad0b6b17d9..f172dd3d9b 100644
--- a/src/syscall/asm_linux_riscv64.s
+++ b/src/syscall/asm_linux_riscv64.s
@@ -104,6 +104,28 @@ err:
 	MOV	A0, err+72(FP)	// errno
 	RET
 
+// func rawVforkSyscall(trap, a1 uintptr) (r1, err uintptr)
+TEXT ·rawVforkSyscall(SB),NOSPLIT|NOFRAME,$0-32
+	MOV	a1+8(FP), A0
+	MOV	ZERO, A1
+	MOV	ZERO, A2
+	MOV	ZERO, A3
+	MOV	ZERO, A4
+	MOV	ZERO, A5
+	MOV	trap+0(FP), A7	// syscall entry
+	ECALL
+	MOV	$-4096, T0
+	BLTU	T0, A0, err
+	MOV	A0, r1+16(FP)	// r1
+	MOV	ZERO, err+24(FP)	// errno
+	RET
+err:
+	MOV	$-1, T0
+	MOV	T0, r1+16(FP)	// r1
+	SUB	A0, ZERO, A0
+	MOV	A0, err+24(FP)	// errno
+	RET
+
 TEXT ·rawSyscallNoError(SB),NOSPLIT,$0-48
 	MOV	a1+8(FP), A0
 	MOV	a2+16(FP), A1
diff --git a/src/syscall/exec_linux.go b/src/syscall/exec_linux.go
index b7351cda82..b6acad96ea 100644
--- a/src/syscall/exec_linux.go
+++ b/src/syscall/exec_linux.go
@@ -207,7 +207,11 @@ func forkAndExecInChild1(argv0 *byte, argv, envv []*byte, chroot, dir *byte, att
 		}
 	}
 
-	hasRawVforkSyscall := runtime.GOARCH == "amd64" || runtime.GOARCH == "ppc64" || runtime.GOARCH == "s390x" || runtime.GOARCH == "arm64"
+	var hasRawVforkSyscall bool
+	switch runtime.GOARCH {
+	case "amd64", "arm64", "ppc64", "riscv64", "s390x":
+		hasRawVforkSyscall = true
+	}
 
 	// About to call fork.
 	// No more allocation or calls of non-assembly functions.
diff --git a/src/syscall/syscall_linux_riscv64.go b/src/syscall/syscall_linux_riscv64.go
index d54bd38510..088e23439f 100644
--- a/src/syscall/syscall_linux_riscv64.go
+++ b/src/syscall/syscall_linux_riscv64.go
@@ -199,6 +199,4 @@ func Pause() error {
 	return err
 }
 
-func rawVforkSyscall(trap, a1 uintptr) (r1 uintptr, err Errno) {
-	panic("not implemented")
-}
+func rawVforkSyscall(trap, a1 uintptr) (r1 uintptr, err Errno)
diff --git a/src/syscall/syscall_unix_test.go b/src/syscall/syscall_unix_test.go
index 13b79ca8d8..3c7982eefe 100644
--- a/src/syscall/syscall_unix_test.go
+++ b/src/syscall/syscall_unix_test.go
@@ -336,11 +336,11 @@ func TestRlimit(t *testing.T) {
 	}
 	set := rlimit
 	set.Cur = set.Max - 1
-	if runtime.GOOS == "darwin" && set.Cur > 10240 {
-		// The max file limit is 10240, even though
-		// the max returned by Getrlimit is 1<<63-1.
-		// This is OPEN_MAX in sys/syslimits.h.
-		set.Cur = 10240
+	if runtime.GOOS == "darwin" && set.Cur > 4096 {
+		// rlim_min for RLIMIT_NOFILE should be equal to
+		// or lower than kern.maxfilesperproc, which on
+		// some machines are 4096. See #40564.
+		set.Cur = 4096
 	}
 	err = syscall.Setrlimit(syscall.RLIMIT_NOFILE, &set)
 	if err != nil {
@@ -353,8 +353,8 @@ func TestRlimit(t *testing.T) {
 	}
 	set = rlimit
 	set.Cur = set.Max - 1
-	if runtime.GOOS == "darwin" && set.Cur > 10240 {
-		set.Cur = 10240
+	if runtime.GOOS == "darwin" && set.Cur > 4096 {
+		set.Cur = 4096
 	}
 	if set != get {
 		t.Fatalf("Rlimit: change failed: wanted %#v got %#v", set, get)
diff --git a/src/testing/iotest/example_test.go b/src/testing/iotest/example_test.go
new file mode 100644
index 0000000000..10f6bd38f7
--- /dev/null
+++ b/src/testing/iotest/example_test.go
@@ -0,0 +1,22 @@
+// Copyright 2020 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 iotest_test
+
+import (
+	"errors"
+	"fmt"
+	"testing/iotest"
+)
+
+func ExampleErrReader() {
+	// A reader that always returns a custom error.
+	r := iotest.ErrReader(errors.New("custom error"))
+	n, err := r.Read(nil)
+	fmt.Printf("n:   %d\nerr: %q\n", n, err)
+
+	// Output:
+	// n:   0
+	// err: "custom error"
+}
diff --git a/src/testing/iotest/logger_test.go b/src/testing/iotest/logger_test.go
index c121bf48f7..fec4467cc6 100644
--- a/src/testing/iotest/logger_test.go
+++ b/src/testing/iotest/logger_test.go
@@ -81,14 +81,6 @@ func TestWriteLogger_errorOnWrite(t *testing.T) {
 	}
 }
 
-type errReader struct {
-	err error
-}
-
-func (r errReader) Read([]byte) (int, error) {
-	return 0, r.err
-}
-
 func TestReadLogger(t *testing.T) {
 	olw := log.Writer()
 	olf := log.Flags()
@@ -146,14 +138,14 @@ func TestReadLogger_errorOnRead(t *testing.T) {
 	data := []byte("Hello, World!")
 	p := make([]byte, len(data))
 
-	lr := errReader{err: errors.New("Read Error!")}
+	lr := ErrReader(errors.New("io failure"))
 	rl := NewReadLogger("read", lr)
 	n, err := rl.Read(p)
 	if err == nil {
 		t.Fatalf("Unexpectedly succeeded to read: %v", err)
 	}
 
-	wantLogWithHex := fmt.Sprintf("lr: read %x: %v\n", p[:n], "Read Error!")
+	wantLogWithHex := fmt.Sprintf("lr: read %x: io failure\n", p[:n])
 	if g, w := lOut.String(), wantLogWithHex; g != w {
 		t.Errorf("ReadLogger mismatch\n\tgot:  %q\n\twant: %q", g, w)
 	}
diff --git a/src/testing/iotest/reader.go b/src/testing/iotest/reader.go
index 8d82018fd6..bc2f72a911 100644
--- a/src/testing/iotest/reader.go
+++ b/src/testing/iotest/reader.go
@@ -68,6 +68,7 @@ func (r *dataErrReader) Read(p []byte) (n int, err error) {
 	return
 }
 
+// ErrTimeout is a fake timeout error.
 var ErrTimeout = errors.New("timeout")
 
 // TimeoutReader returns ErrTimeout on the second read
@@ -86,3 +87,16 @@ func (r *timeoutReader) Read(p []byte) (int, error) {
 	}
 	return r.r.Read(p)
 }
+
+// ErrReader returns an io.Reader that returns 0, err from all Read calls.
+func ErrReader(err error) io.Reader {
+	return &alwaysErrReader{err: err}
+}
+
+type alwaysErrReader struct {
+	err error
+}
+
+func (aer *alwaysErrReader) Read(p []byte) (int, error) {
+	return 0, aer.err
+}
diff --git a/src/testing/iotest/reader_test.go b/src/testing/iotest/reader_test.go
index 9397837e08..6004e841e5 100644
--- a/src/testing/iotest/reader_test.go
+++ b/src/testing/iotest/reader_test.go
@@ -6,6 +6,7 @@ package iotest
 
 import (
 	"bytes"
+	"errors"
 	"io"
 	"testing"
 )
@@ -224,3 +225,27 @@ func TestDataErrReader_emptyReader(t *testing.T) {
 		t.Errorf("Unexpectedly read %d bytes, wanted %d", g, w)
 	}
 }
+
+func TestErrReader(t *testing.T) {
+	cases := []struct {
+		name string
+		err  error
+	}{
+		{"nil error", nil},
+		{"non-nil error", errors.New("io failure")},
+		{"io.EOF", io.EOF},
+	}
+
+	for _, tt := range cases {
+		tt := tt
+		t.Run(tt.name, func(t *testing.T) {
+			n, err := ErrReader(tt.err).Read(nil)
+			if err != tt.err {
+				t.Fatalf("Error mismatch\nGot:  %v\nWant: %v", err, tt.err)
+			}
+			if n != 0 {
+				t.Fatalf("Byte count mismatch: got %d want 0", n)
+			}
+		})
+	}
+}
diff --git a/src/testing/testing.go b/src/testing/testing.go
index 061142b9ab..6fc8c4fa9f 100644
--- a/src/testing/testing.go
+++ b/src/testing/testing.go
@@ -357,10 +357,19 @@ func (p *testPrinter) Fprint(w io.Writer, testName, out string) {
 	defer p.lastNameMu.Unlock()
 
 	if !p.chatty ||
-		strings.HasPrefix(out, "--- PASS") ||
-		strings.HasPrefix(out, "--- FAIL") ||
-		strings.HasPrefix(out, "=== CONT") ||
-		strings.HasPrefix(out, "=== RUN") {
+		strings.HasPrefix(out, "--- PASS: ") ||
+		strings.HasPrefix(out, "--- FAIL: ") ||
+		strings.HasPrefix(out, "--- SKIP: ") ||
+		strings.HasPrefix(out, "=== RUN   ") ||
+		strings.HasPrefix(out, "=== CONT  ") ||
+		strings.HasPrefix(out, "=== PAUSE ") {
+		// If we're buffering test output (!p.chatty), we don't really care which
+		// test is emitting which line so long as they are serialized.
+		//
+		// If the message already implies an association with a specific new test,
+		// we don't need to check what the old test name was or log an extra CONT
+		// line for it. (We're updating it anyway, and the current message already
+		// includes the test name.)
 		p.lastName = testName
 		fmt.Fprint(w, out)
 		return
@@ -976,7 +985,13 @@ func (t *T) Parallel() {
 		for ; root.parent != nil; root = root.parent {
 		}
 		root.mu.Lock()
-		fmt.Fprintf(root.w, "=== PAUSE %s\n", t.name)
+		// Unfortunately, even though PAUSE indicates that the named test is *no
+		// longer* running, cmd/test2json interprets it as changing the active test
+		// for the purpose of log parsing. We could fix cmd/test2json, but that
+		// won't fix existing deployments of third-party tools that already shell
+		// out to older builds of cmd/test2json — so merely fixing cmd/test2json
+		// isn't enough for now.
+		printer.Fprint(root.w, t.name, fmt.Sprintf("=== PAUSE %s\n", t.name))
 		root.mu.Unlock()
 	}
 
diff --git a/src/unicode/tables.go b/src/unicode/tables.go
index 12441682d0..a9b23bfacd 100644
--- a/src/unicode/tables.go
+++ b/src/unicode/tables.go
@@ -3,7 +3,7 @@
 package unicode
 
 // Version is the Unicode edition from which the tables are derived.
-const Version = "12.0.0"
+const Version = "13.0.0"
 
 // Categories is the set of Unicode category tables.
 var Categories = map[string]*RangeTable{
@@ -170,7 +170,7 @@ var _L = &RangeTable{
 		{0x0841, 0x0858, 1},
 		{0x0860, 0x086a, 1},
 		{0x08a0, 0x08b4, 1},
-		{0x08b6, 0x08bd, 1},
+		{0x08b6, 0x08c7, 1},
 		{0x0904, 0x0939, 1},
 		{0x093d, 0x0950, 19},
 		{0x0958, 0x0961, 1},
@@ -241,7 +241,7 @@ var _L = &RangeTable{
 		{0x0cbd, 0x0cde, 33},
 		{0x0ce0, 0x0ce1, 1},
 		{0x0cf1, 0x0cf2, 1},
-		{0x0d05, 0x0d0c, 1},
+		{0x0d04, 0x0d0c, 1},
 		{0x0d0e, 0x0d10, 1},
 		{0x0d12, 0x0d3a, 1},
 		{0x0d3d, 0x0d4e, 17},
@@ -402,10 +402,10 @@ var _L = &RangeTable{
 		{0x30fc, 0x30ff, 1},
 		{0x3105, 0x312f, 1},
 		{0x3131, 0x318e, 1},
-		{0x31a0, 0x31ba, 1},
+		{0x31a0, 0x31bf, 1},
 		{0x31f0, 0x31ff, 1},
-		{0x3400, 0x4db5, 1},
-		{0x4e00, 0x9fef, 1},
+		{0x3400, 0x4dbf, 1},
+		{0x4e00, 0x9ffc, 1},
 		{0xa000, 0xa48c, 1},
 		{0xa4d0, 0xa4fd, 1},
 		{0xa500, 0xa60c, 1},
@@ -417,8 +417,8 @@ var _L = &RangeTable{
 		{0xa717, 0xa71f, 1},
 		{0xa722, 0xa788, 1},
 		{0xa78b, 0xa7bf, 1},
-		{0xa7c2, 0xa7c6, 1},
-		{0xa7f7, 0xa801, 1},
+		{0xa7c2, 0xa7ca, 1},
+		{0xa7f5, 0xa801, 1},
 		{0xa803, 0xa805, 1},
 		{0xa807, 0xa80a, 1},
 		{0xa80c, 0xa822, 1},
@@ -454,7 +454,7 @@ var _L = &RangeTable{
 		{0xab20, 0xab26, 1},
 		{0xab28, 0xab2e, 1},
 		{0xab30, 0xab5a, 1},
-		{0xab5c, 0xab67, 1},
+		{0xab5c, 0xab69, 1},
 		{0xab70, 0xabe2, 1},
 		{0xac00, 0xd7a3, 1},
 		{0xd7b0, 0xd7c6, 1},
@@ -540,16 +540,19 @@ var _L = &RangeTable{
 		{0x10c80, 0x10cb2, 1},
 		{0x10cc0, 0x10cf2, 1},
 		{0x10d00, 0x10d23, 1},
+		{0x10e80, 0x10ea9, 1},
+		{0x10eb0, 0x10eb1, 1},
 		{0x10f00, 0x10f1c, 1},
 		{0x10f27, 0x10f30, 9},
 		{0x10f31, 0x10f45, 1},
+		{0x10fb0, 0x10fc4, 1},
 		{0x10fe0, 0x10ff6, 1},
 		{0x11003, 0x11037, 1},
 		{0x11083, 0x110af, 1},
 		{0x110d0, 0x110e8, 1},
 		{0x11103, 0x11126, 1},
-		{0x11144, 0x11150, 12},
-		{0x11151, 0x11172, 1},
+		{0x11144, 0x11147, 3},
+		{0x11150, 0x11172, 1},
 		{0x11176, 0x11183, 13},
 		{0x11184, 0x111b2, 1},
 		{0x111c1, 0x111c4, 1},
@@ -572,8 +575,8 @@ var _L = &RangeTable{
 		{0x1135d, 0x11361, 1},
 		{0x11400, 0x11434, 1},
 		{0x11447, 0x1144a, 1},
-		{0x1145f, 0x11480, 33},
-		{0x11481, 0x114af, 1},
+		{0x1145f, 0x11461, 1},
+		{0x11480, 0x114af, 1},
 		{0x114c4, 0x114c5, 1},
 		{0x114c7, 0x11580, 185},
 		{0x11581, 0x115ae, 1},
@@ -585,8 +588,13 @@ var _L = &RangeTable{
 		{0x11701, 0x1171a, 1},
 		{0x11800, 0x1182b, 1},
 		{0x118a0, 0x118df, 1},
-		{0x118ff, 0x119a0, 161},
-		{0x119a1, 0x119a7, 1},
+		{0x118ff, 0x11906, 1},
+		{0x11909, 0x1190c, 3},
+		{0x1190d, 0x11913, 1},
+		{0x11915, 0x11916, 1},
+		{0x11918, 0x1192f, 1},
+		{0x1193f, 0x11941, 2},
+		{0x119a0, 0x119a7, 1},
 		{0x119aa, 0x119d0, 1},
 		{0x119e1, 0x119e3, 2},
 		{0x11a00, 0x11a0b, 11},
@@ -608,7 +616,8 @@ var _L = &RangeTable{
 		{0x11d6a, 0x11d89, 1},
 		{0x11d98, 0x11ee0, 328},
 		{0x11ee1, 0x11ef2, 1},
-		{0x12000, 0x12399, 1},
+		{0x11fb0, 0x12000, 80},
+		{0x12001, 0x12399, 1},
 		{0x12480, 0x12543, 1},
 		{0x13000, 0x1342e, 1},
 		{0x14400, 0x14646, 1},
@@ -626,7 +635,8 @@ var _L = &RangeTable{
 		{0x16fe0, 0x16fe1, 1},
 		{0x16fe3, 0x17000, 29},
 		{0x17001, 0x187f7, 1},
-		{0x18800, 0x18af2, 1},
+		{0x18800, 0x18cd5, 1},
+		{0x18d00, 0x18d08, 1},
 		{0x1b000, 0x1b11e, 1},
 		{0x1b150, 0x1b152, 1},
 		{0x1b164, 0x1b167, 1},
@@ -696,12 +706,13 @@ var _L = &RangeTable{
 		{0x1eea1, 0x1eea3, 1},
 		{0x1eea5, 0x1eea9, 1},
 		{0x1eeab, 0x1eebb, 1},
-		{0x20000, 0x2a6d6, 1},
+		{0x20000, 0x2a6dd, 1},
 		{0x2a700, 0x2b734, 1},
 		{0x2b740, 0x2b81d, 1},
 		{0x2b820, 0x2cea1, 1},
 		{0x2ceb0, 0x2ebe0, 1},
 		{0x2f800, 0x2fa1d, 1},
+		{0x30000, 0x3134a, 1},
 	},
 	LatinOffset: 6,
 }
@@ -821,9 +832,11 @@ var _Ll = &RangeTable{
 		{0xa797, 0xa7a9, 2},
 		{0xa7af, 0xa7b5, 6},
 		{0xa7b7, 0xa7bf, 2},
-		{0xa7c3, 0xa7fa, 55},
-		{0xab30, 0xab5a, 1},
-		{0xab60, 0xab67, 1},
+		{0xa7c3, 0xa7c8, 5},
+		{0xa7ca, 0xa7f6, 44},
+		{0xa7fa, 0xab30, 822},
+		{0xab31, 0xab5a, 1},
+		{0xab60, 0xab68, 1},
 		{0xab70, 0xabbf, 1},
 		{0xfb00, 0xfb06, 1},
 		{0xfb13, 0xfb17, 1},
@@ -908,8 +921,8 @@ var _Lm = &RangeTable{
 		{0xaa70, 0xaadd, 109},
 		{0xaaf3, 0xaaf4, 1},
 		{0xab5c, 0xab5f, 1},
-		{0xff70, 0xff9e, 46},
-		{0xff9f, 0xff9f, 1},
+		{0xab69, 0xff70, 21511},
+		{0xff9e, 0xff9f, 1},
 	},
 	R32: []Range32{
 		{0x16b40, 0x16b43, 1},
@@ -945,7 +958,7 @@ var _Lo = &RangeTable{
 		{0x0840, 0x0858, 1},
 		{0x0860, 0x086a, 1},
 		{0x08a0, 0x08b4, 1},
-		{0x08b6, 0x08bd, 1},
+		{0x08b6, 0x08c7, 1},
 		{0x0904, 0x0939, 1},
 		{0x093d, 0x0950, 19},
 		{0x0958, 0x0961, 1},
@@ -1016,7 +1029,7 @@ var _Lo = &RangeTable{
 		{0x0cbd, 0x0cde, 33},
 		{0x0ce0, 0x0ce1, 1},
 		{0x0cf1, 0x0cf2, 1},
-		{0x0d05, 0x0d0c, 1},
+		{0x0d04, 0x0d0c, 1},
 		{0x0d0e, 0x0d10, 1},
 		{0x0d12, 0x0d3a, 1},
 		{0x0d3d, 0x0d4e, 17},
@@ -1127,10 +1140,10 @@ var _Lo = &RangeTable{
 		{0x30ff, 0x3105, 6},
 		{0x3106, 0x312f, 1},
 		{0x3131, 0x318e, 1},
-		{0x31a0, 0x31ba, 1},
+		{0x31a0, 0x31bf, 1},
 		{0x31f0, 0x31ff, 1},
-		{0x3400, 0x4db5, 1},
-		{0x4e00, 0x9fef, 1},
+		{0x3400, 0x4dbf, 1},
+		{0x4e00, 0x9ffc, 1},
 		{0xa000, 0xa014, 1},
 		{0xa016, 0xa48c, 1},
 		{0xa4d0, 0xa4f7, 1},
@@ -1254,16 +1267,19 @@ var _Lo = &RangeTable{
 		{0x10b80, 0x10b91, 1},
 		{0x10c00, 0x10c48, 1},
 		{0x10d00, 0x10d23, 1},
+		{0x10e80, 0x10ea9, 1},
+		{0x10eb0, 0x10eb1, 1},
 		{0x10f00, 0x10f1c, 1},
 		{0x10f27, 0x10f30, 9},
 		{0x10f31, 0x10f45, 1},
+		{0x10fb0, 0x10fc4, 1},
 		{0x10fe0, 0x10ff6, 1},
 		{0x11003, 0x11037, 1},
 		{0x11083, 0x110af, 1},
 		{0x110d0, 0x110e8, 1},
 		{0x11103, 0x11126, 1},
-		{0x11144, 0x11150, 12},
-		{0x11151, 0x11172, 1},
+		{0x11144, 0x11147, 3},
+		{0x11150, 0x11172, 1},
 		{0x11176, 0x11183, 13},
 		{0x11184, 0x111b2, 1},
 		{0x111c1, 0x111c4, 1},
@@ -1286,8 +1302,8 @@ var _Lo = &RangeTable{
 		{0x1135d, 0x11361, 1},
 		{0x11400, 0x11434, 1},
 		{0x11447, 0x1144a, 1},
-		{0x1145f, 0x11480, 33},
-		{0x11481, 0x114af, 1},
+		{0x1145f, 0x11461, 1},
+		{0x11480, 0x114af, 1},
 		{0x114c4, 0x114c5, 1},
 		{0x114c7, 0x11580, 185},
 		{0x11581, 0x115ae, 1},
@@ -1298,8 +1314,13 @@ var _Lo = &RangeTable{
 		{0x116b8, 0x11700, 72},
 		{0x11701, 0x1171a, 1},
 		{0x11800, 0x1182b, 1},
-		{0x118ff, 0x119a0, 161},
-		{0x119a1, 0x119a7, 1},
+		{0x118ff, 0x11906, 1},
+		{0x11909, 0x1190c, 3},
+		{0x1190d, 0x11913, 1},
+		{0x11915, 0x11916, 1},
+		{0x11918, 0x1192f, 1},
+		{0x1193f, 0x11941, 2},
+		{0x119a0, 0x119a7, 1},
 		{0x119aa, 0x119d0, 1},
 		{0x119e1, 0x119e3, 2},
 		{0x11a00, 0x11a0b, 11},
@@ -1321,7 +1342,8 @@ var _Lo = &RangeTable{
 		{0x11d6a, 0x11d89, 1},
 		{0x11d98, 0x11ee0, 328},
 		{0x11ee1, 0x11ef2, 1},
-		{0x12000, 0x12399, 1},
+		{0x11fb0, 0x12000, 80},
+		{0x12001, 0x12399, 1},
 		{0x12480, 0x12543, 1},
 		{0x13000, 0x1342e, 1},
 		{0x14400, 0x14646, 1},
@@ -1334,7 +1356,8 @@ var _Lo = &RangeTable{
 		{0x16f00, 0x16f4a, 1},
 		{0x16f50, 0x17000, 176},
 		{0x17001, 0x187f7, 1},
-		{0x18800, 0x18af2, 1},
+		{0x18800, 0x18cd5, 1},
+		{0x18d00, 0x18d08, 1},
 		{0x1b000, 0x1b11e, 1},
 		{0x1b150, 0x1b152, 1},
 		{0x1b164, 0x1b167, 1},
@@ -1371,12 +1394,13 @@ var _Lo = &RangeTable{
 		{0x1eea1, 0x1eea3, 1},
 		{0x1eea5, 0x1eea9, 1},
 		{0x1eeab, 0x1eebb, 1},
-		{0x20000, 0x2a6d6, 1},
+		{0x20000, 0x2a6dd, 1},
 		{0x2a700, 0x2b734, 1},
 		{0x2b740, 0x2b81d, 1},
 		{0x2b820, 0x2cea1, 1},
 		{0x2ceb0, 0x2ebe0, 1},
 		{0x2f800, 0x2fa1d, 1},
+		{0x30000, 0x3134a, 1},
 	},
 	LatinOffset: 1,
 }
@@ -1500,7 +1524,8 @@ var _Lu = &RangeTable{
 		{0xa7b0, 0xa7b4, 1},
 		{0xa7b6, 0xa7be, 2},
 		{0xa7c2, 0xa7c4, 2},
-		{0xa7c5, 0xa7c6, 1},
+		{0xa7c5, 0xa7c7, 1},
+		{0xa7c9, 0xa7f5, 44},
 		{0xff21, 0xff3a, 1},
 	},
 	R32: []Range32{
@@ -1601,7 +1626,7 @@ var _M = &RangeTable{
 		{0x0b3f, 0x0b44, 1},
 		{0x0b47, 0x0b48, 1},
 		{0x0b4b, 0x0b4d, 1},
-		{0x0b56, 0x0b57, 1},
+		{0x0b55, 0x0b57, 1},
 		{0x0b62, 0x0b63, 1},
 		{0x0b82, 0x0bbe, 60},
 		{0x0bbf, 0x0bc2, 1},
@@ -1627,9 +1652,10 @@ var _M = &RangeTable{
 		{0x0d46, 0x0d48, 1},
 		{0x0d4a, 0x0d4d, 1},
 		{0x0d57, 0x0d62, 11},
-		{0x0d63, 0x0d82, 31},
-		{0x0d83, 0x0dca, 71},
-		{0x0dcf, 0x0dd4, 1},
+		{0x0d63, 0x0d81, 30},
+		{0x0d82, 0x0d83, 1},
+		{0x0dca, 0x0dcf, 5},
+		{0x0dd0, 0x0dd4, 1},
 		{0x0dd6, 0x0dd8, 2},
 		{0x0dd9, 0x0ddf, 1},
 		{0x0df2, 0x0df3, 1},
@@ -1672,7 +1698,7 @@ var _M = &RangeTable{
 		{0x1a55, 0x1a5e, 1},
 		{0x1a60, 0x1a7c, 1},
 		{0x1a7f, 0x1ab0, 49},
-		{0x1ab1, 0x1abe, 1},
+		{0x1ab1, 0x1ac0, 1},
 		{0x1b00, 0x1b04, 1},
 		{0x1b34, 0x1b44, 1},
 		{0x1b6b, 0x1b73, 1},
@@ -1699,8 +1725,9 @@ var _M = &RangeTable{
 		{0xa802, 0xa806, 4},
 		{0xa80b, 0xa823, 24},
 		{0xa824, 0xa827, 1},
-		{0xa880, 0xa881, 1},
-		{0xa8b4, 0xa8c5, 1},
+		{0xa82c, 0xa880, 84},
+		{0xa881, 0xa8b4, 51},
+		{0xa8b5, 0xa8c5, 1},
 		{0xa8e0, 0xa8f1, 1},
 		{0xa8ff, 0xa926, 39},
 		{0xa927, 0xa92d, 1},
@@ -1735,6 +1762,7 @@ var _M = &RangeTable{
 		{0x10a3f, 0x10ae5, 166},
 		{0x10ae6, 0x10d24, 574},
 		{0x10d25, 0x10d27, 1},
+		{0x10eab, 0x10eac, 1},
 		{0x10f46, 0x10f50, 1},
 		{0x11000, 0x11002, 1},
 		{0x11038, 0x11046, 1},
@@ -1747,6 +1775,7 @@ var _M = &RangeTable{
 		{0x11181, 0x11182, 1},
 		{0x111b3, 0x111c0, 1},
 		{0x111c9, 0x111cc, 1},
+		{0x111ce, 0x111cf, 1},
 		{0x1122c, 0x11237, 1},
 		{0x1123e, 0x112df, 161},
 		{0x112e0, 0x112ea, 1},
@@ -1769,7 +1798,12 @@ var _M = &RangeTable{
 		{0x116ab, 0x116b7, 1},
 		{0x1171d, 0x1172b, 1},
 		{0x1182c, 0x1183a, 1},
-		{0x119d1, 0x119d7, 1},
+		{0x11930, 0x11935, 1},
+		{0x11937, 0x11938, 1},
+		{0x1193b, 0x1193e, 1},
+		{0x11940, 0x11942, 2},
+		{0x11943, 0x119d1, 142},
+		{0x119d2, 0x119d7, 1},
 		{0x119da, 0x119e0, 1},
 		{0x119e4, 0x11a01, 29},
 		{0x11a02, 0x11a0a, 1},
@@ -1796,8 +1830,10 @@ var _M = &RangeTable{
 		{0x16f4f, 0x16f51, 2},
 		{0x16f52, 0x16f87, 1},
 		{0x16f8f, 0x16f92, 1},
-		{0x1bc9d, 0x1bc9e, 1},
-		{0x1d165, 0x1d169, 1},
+		{0x16fe4, 0x16ff0, 12},
+		{0x16ff1, 0x1bc9d, 19628},
+		{0x1bc9e, 0x1d165, 5319},
+		{0x1d166, 0x1d169, 1},
 		{0x1d16d, 0x1d172, 1},
 		{0x1d17b, 0x1d182, 1},
 		{0x1d185, 0x1d18b, 1},
@@ -1929,7 +1965,8 @@ var _Mc = &RangeTable{
 		{0x11146, 0x11182, 60},
 		{0x111b3, 0x111b5, 1},
 		{0x111bf, 0x111c0, 1},
-		{0x1122c, 0x1122e, 1},
+		{0x111ce, 0x1122c, 94},
+		{0x1122d, 0x1122e, 1},
 		{0x11232, 0x11233, 1},
 		{0x11235, 0x112e0, 171},
 		{0x112e1, 0x112e2, 1},
@@ -1957,7 +1994,11 @@ var _Mc = &RangeTable{
 		{0x116b6, 0x11720, 106},
 		{0x11721, 0x11726, 5},
 		{0x1182c, 0x1182e, 1},
-		{0x11838, 0x119d1, 409},
+		{0x11838, 0x11930, 248},
+		{0x11931, 0x11935, 1},
+		{0x11937, 0x11938, 1},
+		{0x1193d, 0x11940, 3},
+		{0x11942, 0x119d1, 143},
 		{0x119d2, 0x119d3, 1},
 		{0x119dc, 0x119df, 1},
 		{0x119e4, 0x11a39, 85},
@@ -1970,6 +2011,7 @@ var _Mc = &RangeTable{
 		{0x11d96, 0x11ef5, 351},
 		{0x11ef6, 0x16f51, 20571},
 		{0x16f52, 0x16f87, 1},
+		{0x16ff0, 0x16ff1, 1},
 		{0x1d165, 0x1d166, 1},
 		{0x1d16d, 0x1d172, 1},
 	},
@@ -2038,12 +2080,12 @@ var _Mn = &RangeTable{
 		{0x0b01, 0x0b3c, 59},
 		{0x0b3f, 0x0b41, 2},
 		{0x0b42, 0x0b44, 1},
-		{0x0b4d, 0x0b56, 9},
-		{0x0b62, 0x0b63, 1},
-		{0x0b82, 0x0bc0, 62},
-		{0x0bcd, 0x0c00, 51},
-		{0x0c04, 0x0c3e, 58},
-		{0x0c3f, 0x0c40, 1},
+		{0x0b4d, 0x0b55, 8},
+		{0x0b56, 0x0b62, 12},
+		{0x0b63, 0x0b82, 31},
+		{0x0bc0, 0x0bcd, 13},
+		{0x0c00, 0x0c04, 4},
+		{0x0c3e, 0x0c40, 1},
 		{0x0c46, 0x0c48, 1},
 		{0x0c4a, 0x0c4d, 1},
 		{0x0c55, 0x0c56, 1},
@@ -2056,8 +2098,9 @@ var _Mn = &RangeTable{
 		{0x0d3b, 0x0d3c, 1},
 		{0x0d41, 0x0d44, 1},
 		{0x0d4d, 0x0d62, 21},
-		{0x0d63, 0x0dca, 103},
-		{0x0dd2, 0x0dd4, 1},
+		{0x0d63, 0x0d81, 30},
+		{0x0dca, 0x0dd2, 8},
+		{0x0dd3, 0x0dd4, 1},
 		{0x0dd6, 0x0e31, 91},
 		{0x0e34, 0x0e3a, 1},
 		{0x0e47, 0x0e4e, 1},
@@ -2107,6 +2150,7 @@ var _Mn = &RangeTable{
 		{0x1a73, 0x1a7c, 1},
 		{0x1a7f, 0x1ab0, 49},
 		{0x1ab1, 0x1abd, 1},
+		{0x1abf, 0x1ac0, 1},
 		{0x1b00, 0x1b03, 1},
 		{0x1b34, 0x1b36, 2},
 		{0x1b37, 0x1b3a, 1},
@@ -2142,9 +2186,9 @@ var _Mn = &RangeTable{
 		{0xa6f0, 0xa6f1, 1},
 		{0xa802, 0xa806, 4},
 		{0xa80b, 0xa825, 26},
-		{0xa826, 0xa8c4, 158},
-		{0xa8c5, 0xa8e0, 27},
-		{0xa8e1, 0xa8f1, 1},
+		{0xa826, 0xa82c, 6},
+		{0xa8c4, 0xa8c5, 1},
+		{0xa8e0, 0xa8f1, 1},
 		{0xa8ff, 0xa926, 39},
 		{0xa927, 0xa92d, 1},
 		{0xa947, 0xa951, 1},
@@ -2178,6 +2222,7 @@ var _Mn = &RangeTable{
 		{0x10a3f, 0x10ae5, 166},
 		{0x10ae6, 0x10d24, 574},
 		{0x10d25, 0x10d27, 1},
+		{0x10eab, 0x10eac, 1},
 		{0x10f46, 0x10f50, 1},
 		{0x11001, 0x11038, 55},
 		{0x11039, 0x11046, 1},
@@ -2191,7 +2236,8 @@ var _Mn = &RangeTable{
 		{0x11181, 0x111b6, 53},
 		{0x111b7, 0x111be, 1},
 		{0x111c9, 0x111cc, 1},
-		{0x1122f, 0x11231, 1},
+		{0x111cf, 0x1122f, 96},
+		{0x11230, 0x11231, 1},
 		{0x11234, 0x11236, 2},
 		{0x11237, 0x1123e, 7},
 		{0x112df, 0x112e3, 4},
@@ -2223,6 +2269,8 @@ var _Mn = &RangeTable{
 		{0x11727, 0x1172b, 1},
 		{0x1182f, 0x11837, 1},
 		{0x11839, 0x1183a, 1},
+		{0x1193b, 0x1193c, 1},
+		{0x1193e, 0x11943, 5},
 		{0x119d4, 0x119d7, 1},
 		{0x119da, 0x119db, 1},
 		{0x119e0, 0x11a01, 33},
@@ -2253,8 +2301,9 @@ var _Mn = &RangeTable{
 		{0x16b30, 0x16b36, 1},
 		{0x16f4f, 0x16f8f, 64},
 		{0x16f90, 0x16f92, 1},
-		{0x1bc9d, 0x1bc9e, 1},
-		{0x1d167, 0x1d169, 1},
+		{0x16fe4, 0x1bc9d, 19641},
+		{0x1bc9e, 0x1d167, 5321},
+		{0x1d168, 0x1d169, 1},
 		{0x1d17b, 0x1d182, 1},
 		{0x1d185, 0x1d18b, 1},
 		{0x1d1aa, 0x1d1ad, 1},
@@ -2375,6 +2424,7 @@ var _N = &RangeTable{
 		{0x10e60, 0x10e7e, 1},
 		{0x10f1d, 0x10f26, 1},
 		{0x10f51, 0x10f54, 1},
+		{0x10fc5, 0x10fcb, 1},
 		{0x11052, 0x1106f, 1},
 		{0x110f0, 0x110f9, 1},
 		{0x11136, 0x1113f, 1},
@@ -2387,6 +2437,7 @@ var _N = &RangeTable{
 		{0x116c0, 0x116c9, 1},
 		{0x11730, 0x1173b, 1},
 		{0x118e0, 0x118f2, 1},
+		{0x11950, 0x11959, 1},
 		{0x11c50, 0x11c6c, 1},
 		{0x11d50, 0x11d59, 1},
 		{0x11da0, 0x11da9, 1},
@@ -2409,6 +2460,7 @@ var _N = &RangeTable{
 		{0x1ed01, 0x1ed2d, 1},
 		{0x1ed2f, 0x1ed3d, 1},
 		{0x1f100, 0x1f10c, 1},
+		{0x1fbf0, 0x1fbf9, 1},
 	},
 	LatinOffset: 4,
 }
@@ -2467,6 +2519,7 @@ var _Nd = &RangeTable{
 		{0x116c0, 0x116c9, 1},
 		{0x11730, 0x11739, 1},
 		{0x118e0, 0x118e9, 1},
+		{0x11950, 0x11959, 1},
 		{0x11c50, 0x11c59, 1},
 		{0x11d50, 0x11d59, 1},
 		{0x11da0, 0x11da9, 1},
@@ -2476,6 +2529,7 @@ var _Nd = &RangeTable{
 		{0x1e140, 0x1e149, 1},
 		{0x1e2f0, 0x1e2f9, 1},
 		{0x1e950, 0x1e959, 1},
+		{0x1fbf0, 0x1fbf9, 1},
 	},
 	LatinOffset: 1,
 }
@@ -2554,6 +2608,7 @@ var _No = &RangeTable{
 		{0x10e60, 0x10e7e, 1},
 		{0x10f1d, 0x10f26, 1},
 		{0x10f51, 0x10f54, 1},
+		{0x10fc5, 0x10fcb, 1},
 		{0x11052, 0x11065, 1},
 		{0x111e1, 0x111f4, 1},
 		{0x1173a, 0x1173b, 1},
@@ -2655,7 +2710,8 @@ var _P = &RangeTable{
 		{0x2d70, 0x2e00, 144},
 		{0x2e01, 0x2e2e, 1},
 		{0x2e30, 0x2e4f, 1},
-		{0x3001, 0x3003, 1},
+		{0x2e52, 0x3001, 431},
+		{0x3002, 0x3003, 1},
 		{0x3008, 0x3011, 1},
 		{0x3014, 0x301f, 1},
 		{0x3030, 0x303d, 13},
@@ -2701,7 +2757,8 @@ var _P = &RangeTable{
 		{0x10af1, 0x10af6, 1},
 		{0x10b39, 0x10b3f, 1},
 		{0x10b99, 0x10b9c, 1},
-		{0x10f55, 0x10f59, 1},
+		{0x10ead, 0x10f55, 168},
+		{0x10f56, 0x10f59, 1},
 		{0x11047, 0x1104d, 1},
 		{0x110bb, 0x110bc, 1},
 		{0x110be, 0x110c1, 1},
@@ -2713,14 +2770,16 @@ var _P = &RangeTable{
 		{0x11238, 0x1123d, 1},
 		{0x112a9, 0x1144b, 418},
 		{0x1144c, 0x1144f, 1},
-		{0x1145b, 0x1145d, 2},
-		{0x114c6, 0x115c1, 251},
-		{0x115c2, 0x115d7, 1},
+		{0x1145a, 0x1145b, 1},
+		{0x1145d, 0x114c6, 105},
+		{0x115c1, 0x115d7, 1},
 		{0x11641, 0x11643, 1},
 		{0x11660, 0x1166c, 1},
 		{0x1173c, 0x1173e, 1},
-		{0x1183b, 0x119e2, 423},
-		{0x11a3f, 0x11a46, 1},
+		{0x1183b, 0x11944, 265},
+		{0x11945, 0x11946, 1},
+		{0x119e2, 0x11a3f, 93},
+		{0x11a40, 0x11a46, 1},
 		{0x11a9a, 0x11a9c, 1},
 		{0x11a9e, 0x11aa2, 1},
 		{0x11c41, 0x11c45, 1},
@@ -2764,6 +2823,9 @@ var _Pd = &RangeTable{
 		{0xfe58, 0xfe63, 11},
 		{0xff0d, 0xff0d, 1},
 	},
+	R32: []Range32{
+		{0x10ead, 0x10ead, 1},
+	},
 }
 
 var _Pe = &RangeTable{
@@ -2894,7 +2956,8 @@ var _Po = &RangeTable{
 		{0x2e3c, 0x2e3f, 1},
 		{0x2e41, 0x2e43, 2},
 		{0x2e44, 0x2e4f, 1},
-		{0x3001, 0x3003, 1},
+		{0x2e52, 0x3001, 431},
+		{0x3002, 0x3003, 1},
 		{0x303d, 0x30fb, 190},
 		{0xa4fe, 0xa4ff, 1},
 		{0xa60d, 0xa60f, 1},
@@ -2951,14 +3014,16 @@ var _Po = &RangeTable{
 		{0x11238, 0x1123d, 1},
 		{0x112a9, 0x1144b, 418},
 		{0x1144c, 0x1144f, 1},
-		{0x1145b, 0x1145d, 2},
-		{0x114c6, 0x115c1, 251},
-		{0x115c2, 0x115d7, 1},
+		{0x1145a, 0x1145b, 1},
+		{0x1145d, 0x114c6, 105},
+		{0x115c1, 0x115d7, 1},
 		{0x11641, 0x11643, 1},
 		{0x11660, 0x1166c, 1},
 		{0x1173c, 0x1173e, 1},
-		{0x1183b, 0x119e2, 423},
-		{0x11a3f, 0x11a46, 1},
+		{0x1183b, 0x11944, 265},
+		{0x11945, 0x11946, 1},
+		{0x119e2, 0x11a3f, 93},
+		{0x11a40, 0x11a46, 1},
 		{0x11a9a, 0x11a9c, 1},
 		{0x11a9e, 0x11aa2, 1},
 		{0x11c41, 0x11c45, 1},
@@ -3094,8 +3159,9 @@ var _S = &RangeTable{
 		{0x29dc, 0x29fb, 1},
 		{0x29fe, 0x2b73, 1},
 		{0x2b76, 0x2b95, 1},
-		{0x2b98, 0x2bff, 1},
+		{0x2b97, 0x2bff, 1},
 		{0x2ce5, 0x2cea, 1},
+		{0x2e50, 0x2e51, 1},
 		{0x2e80, 0x2e99, 1},
 		{0x2e9b, 0x2ef3, 1},
 		{0x2f00, 0x2fd5, 1},
@@ -3113,8 +3179,7 @@ var _S = &RangeTable{
 		{0x3250, 0x3260, 16},
 		{0x3261, 0x327f, 1},
 		{0x328a, 0x32b0, 1},
-		{0x32c0, 0x32fe, 1},
-		{0x3300, 0x33ff, 1},
+		{0x32c0, 0x33ff, 1},
 		{0x4dc0, 0x4dff, 1},
 		{0xa490, 0xa4c6, 1},
 		{0xa700, 0xa716, 1},
@@ -3123,7 +3188,8 @@ var _S = &RangeTable{
 		{0xa828, 0xa82b, 1},
 		{0xa836, 0xa839, 1},
 		{0xaa77, 0xaa79, 1},
-		{0xab5b, 0xfb29, 20430},
+		{0xab5b, 0xab6a, 15},
+		{0xab6b, 0xfb29, 20414},
 		{0xfbb2, 0xfbc1, 1},
 		{0xfdfc, 0xfdfd, 1},
 		{0xfe62, 0xfe64, 2},
@@ -3141,7 +3207,7 @@ var _S = &RangeTable{
 		{0x10137, 0x1013f, 1},
 		{0x10179, 0x10189, 1},
 		{0x1018c, 0x1018e, 1},
-		{0x10190, 0x1019b, 1},
+		{0x10190, 0x1019c, 1},
 		{0x101a0, 0x101d0, 48},
 		{0x101d1, 0x101fc, 1},
 		{0x10877, 0x10878, 1},
@@ -3179,16 +3245,15 @@ var _S = &RangeTable{
 		{0x1f0b1, 0x1f0bf, 1},
 		{0x1f0c1, 0x1f0cf, 1},
 		{0x1f0d1, 0x1f0f5, 1},
-		{0x1f110, 0x1f16c, 1},
-		{0x1f170, 0x1f1ac, 1},
+		{0x1f10d, 0x1f1ad, 1},
 		{0x1f1e6, 0x1f202, 1},
 		{0x1f210, 0x1f23b, 1},
 		{0x1f240, 0x1f248, 1},
 		{0x1f250, 0x1f251, 1},
 		{0x1f260, 0x1f265, 1},
-		{0x1f300, 0x1f6d5, 1},
+		{0x1f300, 0x1f6d7, 1},
 		{0x1f6e0, 0x1f6ec, 1},
-		{0x1f6f0, 0x1f6fa, 1},
+		{0x1f6f0, 0x1f6fc, 1},
 		{0x1f700, 0x1f773, 1},
 		{0x1f780, 0x1f7d8, 1},
 		{0x1f7e0, 0x1f7eb, 1},
@@ -3197,18 +3262,20 @@ var _S = &RangeTable{
 		{0x1f850, 0x1f859, 1},
 		{0x1f860, 0x1f887, 1},
 		{0x1f890, 0x1f8ad, 1},
-		{0x1f900, 0x1f90b, 1},
-		{0x1f90d, 0x1f971, 1},
-		{0x1f973, 0x1f976, 1},
-		{0x1f97a, 0x1f9a2, 1},
-		{0x1f9a5, 0x1f9aa, 1},
-		{0x1f9ae, 0x1f9ca, 1},
+		{0x1f8b0, 0x1f8b1, 1},
+		{0x1f900, 0x1f978, 1},
+		{0x1f97a, 0x1f9cb, 1},
 		{0x1f9cd, 0x1fa53, 1},
 		{0x1fa60, 0x1fa6d, 1},
-		{0x1fa70, 0x1fa73, 1},
+		{0x1fa70, 0x1fa74, 1},
 		{0x1fa78, 0x1fa7a, 1},
-		{0x1fa80, 0x1fa82, 1},
-		{0x1fa90, 0x1fa95, 1},
+		{0x1fa80, 0x1fa86, 1},
+		{0x1fa90, 0x1faa8, 1},
+		{0x1fab0, 0x1fab6, 1},
+		{0x1fac0, 0x1fac2, 1},
+		{0x1fad0, 0x1fad6, 1},
+		{0x1fb00, 0x1fb92, 1},
+		{0x1fb94, 0x1fbca, 1},
 	},
 	LatinOffset: 10,
 }
@@ -3257,7 +3324,8 @@ var _Sk = &RangeTable{
 		{0xa700, 0xa716, 1},
 		{0xa720, 0xa721, 1},
 		{0xa789, 0xa78a, 1},
-		{0xab5b, 0xfbb2, 20567},
+		{0xab5b, 0xab6a, 15},
+		{0xab6b, 0xfbb2, 20551},
 		{0xfbb3, 0xfbc1, 1},
 		{0xff3e, 0xff40, 2},
 		{0xffe3, 0xffe3, 1},
@@ -3394,8 +3462,9 @@ var _So = &RangeTable{
 		{0x2b45, 0x2b46, 1},
 		{0x2b4d, 0x2b73, 1},
 		{0x2b76, 0x2b95, 1},
-		{0x2b98, 0x2bff, 1},
+		{0x2b97, 0x2bff, 1},
 		{0x2ce5, 0x2cea, 1},
+		{0x2e50, 0x2e51, 1},
 		{0x2e80, 0x2e99, 1},
 		{0x2e9b, 0x2ef3, 1},
 		{0x2f00, 0x2fd5, 1},
@@ -3412,8 +3481,7 @@ var _So = &RangeTable{
 		{0x3250, 0x3260, 16},
 		{0x3261, 0x327f, 1},
 		{0x328a, 0x32b0, 1},
-		{0x32c0, 0x32fe, 1},
-		{0x3300, 0x33ff, 1},
+		{0x32c0, 0x33ff, 1},
 		{0x4dc0, 0x4dff, 1},
 		{0xa490, 0xa4c6, 1},
 		{0xa828, 0xa82b, 1},
@@ -3429,7 +3497,7 @@ var _So = &RangeTable{
 		{0x10137, 0x1013f, 1},
 		{0x10179, 0x10189, 1},
 		{0x1018c, 0x1018e, 1},
-		{0x10190, 0x1019b, 1},
+		{0x10190, 0x1019c, 1},
 		{0x101a0, 0x101d0, 48},
 		{0x101d1, 0x101fc, 1},
 		{0x10877, 0x10878, 1},
@@ -3461,17 +3529,16 @@ var _So = &RangeTable{
 		{0x1f0b1, 0x1f0bf, 1},
 		{0x1f0c1, 0x1f0cf, 1},
 		{0x1f0d1, 0x1f0f5, 1},
-		{0x1f110, 0x1f16c, 1},
-		{0x1f170, 0x1f1ac, 1},
+		{0x1f10d, 0x1f1ad, 1},
 		{0x1f1e6, 0x1f202, 1},
 		{0x1f210, 0x1f23b, 1},
 		{0x1f240, 0x1f248, 1},
 		{0x1f250, 0x1f251, 1},
 		{0x1f260, 0x1f265, 1},
 		{0x1f300, 0x1f3fa, 1},
-		{0x1f400, 0x1f6d5, 1},
+		{0x1f400, 0x1f6d7, 1},
 		{0x1f6e0, 0x1f6ec, 1},
-		{0x1f6f0, 0x1f6fa, 1},
+		{0x1f6f0, 0x1f6fc, 1},
 		{0x1f700, 0x1f773, 1},
 		{0x1f780, 0x1f7d8, 1},
 		{0x1f7e0, 0x1f7eb, 1},
@@ -3480,18 +3547,20 @@ var _So = &RangeTable{
 		{0x1f850, 0x1f859, 1},
 		{0x1f860, 0x1f887, 1},
 		{0x1f890, 0x1f8ad, 1},
-		{0x1f900, 0x1f90b, 1},
-		{0x1f90d, 0x1f971, 1},
-		{0x1f973, 0x1f976, 1},
-		{0x1f97a, 0x1f9a2, 1},
-		{0x1f9a5, 0x1f9aa, 1},
-		{0x1f9ae, 0x1f9ca, 1},
+		{0x1f8b0, 0x1f8b1, 1},
+		{0x1f900, 0x1f978, 1},
+		{0x1f97a, 0x1f9cb, 1},
 		{0x1f9cd, 0x1fa53, 1},
 		{0x1fa60, 0x1fa6d, 1},
-		{0x1fa70, 0x1fa73, 1},
+		{0x1fa70, 0x1fa74, 1},
 		{0x1fa78, 0x1fa7a, 1},
-		{0x1fa80, 0x1fa82, 1},
-		{0x1fa90, 0x1fa95, 1},
+		{0x1fa80, 0x1fa86, 1},
+		{0x1fa90, 0x1faa8, 1},
+		{0x1fab0, 0x1fab6, 1},
+		{0x1fac0, 0x1fac2, 1},
+		{0x1fad0, 0x1fad6, 1},
+		{0x1fb00, 0x1fb92, 1},
+		{0x1fb94, 0x1fbca, 1},
 	},
 	LatinOffset: 2,
 }
@@ -3607,6 +3676,7 @@ var Scripts = map[string]*RangeTable{
 	"Chakma":                 Chakma,
 	"Cham":                   Cham,
 	"Cherokee":               Cherokee,
+	"Chorasmian":             Chorasmian,
 	"Common":                 Common,
 	"Coptic":                 Coptic,
 	"Cuneiform":              Cuneiform,
@@ -3614,6 +3684,7 @@ var Scripts = map[string]*RangeTable{
 	"Cyrillic":               Cyrillic,
 	"Deseret":                Deseret,
 	"Devanagari":             Devanagari,
+	"Dives_Akuru":            Dives_Akuru,
 	"Dogra":                  Dogra,
 	"Duployan":               Duployan,
 	"Egyptian_Hieroglyphs":   Egyptian_Hieroglyphs,
@@ -3645,6 +3716,7 @@ var Scripts = map[string]*RangeTable{
 	"Katakana":               Katakana,
 	"Kayah_Li":               Kayah_Li,
 	"Kharoshthi":             Kharoshthi,
+	"Khitan_Small_Script":    Khitan_Small_Script,
 	"Khmer":                  Khmer,
 	"Khojki":                 Khojki,
 	"Khudawadi":              Khudawadi,
@@ -3734,6 +3806,7 @@ var Scripts = map[string]*RangeTable{
 	"Vai":                    Vai,
 	"Wancho":                 Wancho,
 	"Warang_Citi":            Warang_Citi,
+	"Yezidi":                 Yezidi,
 	"Yi":                     Yi,
 	"Zanabazar_Square":       Zanabazar_Square,
 }
@@ -3776,7 +3849,7 @@ var _Arabic = &RangeTable{
 		{0x06de, 0x06ff, 1},
 		{0x0750, 0x077f, 1},
 		{0x08a0, 0x08b4, 1},
-		{0x08b6, 0x08bd, 1},
+		{0x08b6, 0x08c7, 1},
 		{0x08d3, 0x08e1, 1},
 		{0x08e3, 0x08ff, 1},
 		{0xfb50, 0xfbc1, 1},
@@ -3820,9 +3893,8 @@ var _Arabic = &RangeTable{
 var _Armenian = &RangeTable{
 	R16: []Range16{
 		{0x0531, 0x0556, 1},
-		{0x0559, 0x0588, 1},
-		{0x058a, 0x058d, 3},
-		{0x058e, 0x058f, 1},
+		{0x0559, 0x058a, 1},
+		{0x058d, 0x058f, 1},
 		{0xfb13, 0xfb17, 1},
 	},
 }
@@ -3899,7 +3971,7 @@ var _Bopomofo = &RangeTable{
 	R16: []Range16{
 		{0x02ea, 0x02eb, 1},
 		{0x3105, 0x312f, 1},
-		{0x31a0, 0x31ba, 1},
+		{0x31a0, 0x31bf, 1},
 	},
 }
 
@@ -3957,7 +4029,7 @@ var _Chakma = &RangeTable{
 	R16: []Range16{},
 	R32: []Range32{
 		{0x11100, 0x11134, 1},
-		{0x11136, 0x11146, 1},
+		{0x11136, 0x11147, 1},
 	},
 }
 
@@ -3978,6 +4050,13 @@ var _Cherokee = &RangeTable{
 	},
 }
 
+var _Chorasmian = &RangeTable{
+	R16: []Range16{},
+	R32: []Range32{
+		{0x10fb0, 0x10fcb, 1},
+	},
+}
+
 var _Common = &RangeTable{
 	R16: []Range16{
 		{0x0000, 0x0040, 1},
@@ -3991,13 +4070,12 @@ var _Common = &RangeTable{
 		{0x02ec, 0x02ff, 1},
 		{0x0374, 0x037e, 10},
 		{0x0385, 0x0387, 2},
-		{0x0589, 0x0605, 124},
-		{0x060c, 0x061b, 15},
-		{0x061f, 0x0640, 33},
-		{0x06dd, 0x08e2, 517},
-		{0x0964, 0x0965, 1},
-		{0x0e3f, 0x0fd5, 406},
-		{0x0fd6, 0x0fd8, 1},
+		{0x0605, 0x060c, 7},
+		{0x061b, 0x061f, 4},
+		{0x0640, 0x06dd, 157},
+		{0x08e2, 0x0964, 130},
+		{0x0965, 0x0e3f, 1242},
+		{0x0fd5, 0x0fd8, 1},
 		{0x10fb, 0x16eb, 1520},
 		{0x16ec, 0x16ed, 1},
 		{0x1735, 0x1736, 1},
@@ -4025,8 +4103,8 @@ var _Common = &RangeTable{
 		{0x2460, 0x27ff, 1},
 		{0x2900, 0x2b73, 1},
 		{0x2b76, 0x2b95, 1},
-		{0x2b98, 0x2bff, 1},
-		{0x2e00, 0x2e4f, 1},
+		{0x2b97, 0x2bff, 1},
+		{0x2e00, 0x2e52, 1},
 		{0x2ff0, 0x2ffb, 1},
 		{0x3000, 0x3004, 1},
 		{0x3006, 0x3008, 2},
@@ -4040,13 +4118,15 @@ var _Common = &RangeTable{
 		{0x31c0, 0x31e3, 1},
 		{0x3220, 0x325f, 1},
 		{0x327f, 0x32cf, 1},
-		{0x3358, 0x33ff, 1},
+		{0x32ff, 0x3358, 89},
+		{0x3359, 0x33ff, 1},
 		{0x4dc0, 0x4dff, 1},
 		{0xa700, 0xa721, 1},
 		{0xa788, 0xa78a, 1},
 		{0xa830, 0xa839, 1},
 		{0xa92e, 0xa9cf, 161},
-		{0xab5b, 0xfd3e, 20963},
+		{0xab5b, 0xab6a, 15},
+		{0xab6b, 0xfd3e, 20947},
 		{0xfd3f, 0xfe10, 209},
 		{0xfe11, 0xfe19, 1},
 		{0xfe30, 0xfe52, 1},
@@ -4066,7 +4146,7 @@ var _Common = &RangeTable{
 		{0x10100, 0x10102, 1},
 		{0x10107, 0x10133, 1},
 		{0x10137, 0x1013f, 1},
-		{0x10190, 0x1019b, 1},
+		{0x10190, 0x1019c, 1},
 		{0x101d0, 0x101fc, 1},
 		{0x102e1, 0x102fb, 1},
 		{0x16fe2, 0x16fe3, 1},
@@ -4110,18 +4190,16 @@ var _Common = &RangeTable{
 		{0x1f0b1, 0x1f0bf, 1},
 		{0x1f0c1, 0x1f0cf, 1},
 		{0x1f0d1, 0x1f0f5, 1},
-		{0x1f100, 0x1f10c, 1},
-		{0x1f110, 0x1f16c, 1},
-		{0x1f170, 0x1f1ac, 1},
+		{0x1f100, 0x1f1ad, 1},
 		{0x1f1e6, 0x1f1ff, 1},
 		{0x1f201, 0x1f202, 1},
 		{0x1f210, 0x1f23b, 1},
 		{0x1f240, 0x1f248, 1},
 		{0x1f250, 0x1f251, 1},
 		{0x1f260, 0x1f265, 1},
-		{0x1f300, 0x1f6d5, 1},
+		{0x1f300, 0x1f6d7, 1},
 		{0x1f6e0, 0x1f6ec, 1},
-		{0x1f6f0, 0x1f6fa, 1},
+		{0x1f6f0, 0x1f6fc, 1},
 		{0x1f700, 0x1f773, 1},
 		{0x1f780, 0x1f7d8, 1},
 		{0x1f7e0, 0x1f7eb, 1},
@@ -4130,18 +4208,21 @@ var _Common = &RangeTable{
 		{0x1f850, 0x1f859, 1},
 		{0x1f860, 0x1f887, 1},
 		{0x1f890, 0x1f8ad, 1},
-		{0x1f900, 0x1f90b, 1},
-		{0x1f90d, 0x1f971, 1},
-		{0x1f973, 0x1f976, 1},
-		{0x1f97a, 0x1f9a2, 1},
-		{0x1f9a5, 0x1f9aa, 1},
-		{0x1f9ae, 0x1f9ca, 1},
+		{0x1f8b0, 0x1f8b1, 1},
+		{0x1f900, 0x1f978, 1},
+		{0x1f97a, 0x1f9cb, 1},
 		{0x1f9cd, 0x1fa53, 1},
 		{0x1fa60, 0x1fa6d, 1},
-		{0x1fa70, 0x1fa73, 1},
+		{0x1fa70, 0x1fa74, 1},
 		{0x1fa78, 0x1fa7a, 1},
-		{0x1fa80, 0x1fa82, 1},
-		{0x1fa90, 0x1fa95, 1},
+		{0x1fa80, 0x1fa86, 1},
+		{0x1fa90, 0x1faa8, 1},
+		{0x1fab0, 0x1fab6, 1},
+		{0x1fac0, 0x1fac2, 1},
+		{0x1fad0, 0x1fad6, 1},
+		{0x1fb00, 0x1fb92, 1},
+		{0x1fb94, 0x1fbca, 1},
+		{0x1fbf0, 0x1fbf9, 1},
 		{0xe0001, 0xe0020, 31},
 		{0xe0021, 0xe007f, 1},
 	},
@@ -4205,6 +4286,20 @@ var _Devanagari = &RangeTable{
 	},
 }
 
+var _Dives_Akuru = &RangeTable{
+	R16: []Range16{},
+	R32: []Range32{
+		{0x11900, 0x11906, 1},
+		{0x11909, 0x1190c, 3},
+		{0x1190d, 0x11913, 1},
+		{0x11915, 0x11916, 1},
+		{0x11918, 0x11935, 1},
+		{0x11937, 0x11938, 1},
+		{0x1193b, 0x11946, 1},
+		{0x11950, 0x11959, 1},
+	},
+}
+
 var _Dogra = &RangeTable{
 	R16: []Range16{},
 	R32: []Range32{
@@ -4435,18 +4530,20 @@ var _Han = &RangeTable{
 		{0x3005, 0x3007, 2},
 		{0x3021, 0x3029, 1},
 		{0x3038, 0x303b, 1},
-		{0x3400, 0x4db5, 1},
-		{0x4e00, 0x9fef, 1},
+		{0x3400, 0x4dbf, 1},
+		{0x4e00, 0x9ffc, 1},
 		{0xf900, 0xfa6d, 1},
 		{0xfa70, 0xfad9, 1},
 	},
 	R32: []Range32{
-		{0x20000, 0x2a6d6, 1},
+		{0x16ff0, 0x16ff1, 1},
+		{0x20000, 0x2a6dd, 1},
 		{0x2a700, 0x2b734, 1},
 		{0x2b740, 0x2b81d, 1},
 		{0x2b820, 0x2cea1, 1},
 		{0x2ceb0, 0x2ebe0, 1},
 		{0x2f800, 0x2fa1d, 1},
+		{0x30000, 0x3134a, 1},
 	},
 }
 
@@ -4533,7 +4630,7 @@ var _Inherited = &RangeTable{
 		{0x064b, 0x0655, 1},
 		{0x0670, 0x0951, 737},
 		{0x0952, 0x0954, 1},
-		{0x1ab0, 0x1abe, 1},
+		{0x1ab0, 0x1ac0, 1},
 		{0x1cd0, 0x1cd2, 1},
 		{0x1cd4, 0x1ce0, 1},
 		{0x1ce2, 0x1ce8, 1},
@@ -4646,6 +4743,14 @@ var _Kharoshthi = &RangeTable{
 	},
 }
 
+var _Khitan_Small_Script = &RangeTable{
+	R16: []Range16{},
+	R32: []Range32{
+		{0x16fe4, 0x18b00, 6940},
+		{0x18b01, 0x18cd5, 1},
+	},
+}
+
 var _Khmer = &RangeTable{
 	R16: []Range16{
 		{0x1780, 0x17dd, 1},
@@ -4710,11 +4815,11 @@ var _Latin = &RangeTable{
 		{0x2c60, 0x2c7f, 1},
 		{0xa722, 0xa787, 1},
 		{0xa78b, 0xa7bf, 1},
-		{0xa7c2, 0xa7c6, 1},
-		{0xa7f7, 0xa7ff, 1},
+		{0xa7c2, 0xa7ca, 1},
+		{0xa7f5, 0xa7ff, 1},
 		{0xab30, 0xab5a, 1},
 		{0xab5c, 0xab64, 1},
-		{0xab66, 0xab67, 1},
+		{0xab66, 0xab69, 1},
 		{0xfb00, 0xfb06, 1},
 		{0xff21, 0xff3a, 1},
 		{0xff41, 0xff5a, 1},
@@ -4766,6 +4871,9 @@ var _Lisu = &RangeTable{
 	R16: []Range16{
 		{0xa4d0, 0xa4ff, 1},
 	},
+	R32: []Range32{
+		{0x11fb0, 0x11fb0, 1},
+	},
 }
 
 var _Lycian = &RangeTable{
@@ -4799,8 +4907,7 @@ var _Makasar = &RangeTable{
 
 var _Malayalam = &RangeTable{
 	R16: []Range16{
-		{0x0d00, 0x0d03, 1},
-		{0x0d05, 0x0d0c, 1},
+		{0x0d00, 0x0d0c, 1},
 		{0x0d0e, 0x0d10, 1},
 		{0x0d12, 0x0d44, 1},
 		{0x0d46, 0x0d48, 1},
@@ -4974,9 +5081,8 @@ var _New_Tai_Lue = &RangeTable{
 var _Newa = &RangeTable{
 	R16: []Range16{},
 	R32: []Range32{
-		{0x11400, 0x11459, 1},
-		{0x1145b, 0x1145d, 2},
-		{0x1145e, 0x1145f, 1},
+		{0x11400, 0x1145b, 1},
+		{0x1145d, 0x11461, 1},
 	},
 }
 
@@ -5089,7 +5195,7 @@ var _Oriya = &RangeTable{
 		{0x0b3c, 0x0b44, 1},
 		{0x0b47, 0x0b48, 1},
 		{0x0b4b, 0x0b4d, 1},
-		{0x0b56, 0x0b57, 1},
+		{0x0b55, 0x0b57, 1},
 		{0x0b5c, 0x0b5d, 1},
 		{0x0b5f, 0x0b63, 1},
 		{0x0b66, 0x0b77, 1},
@@ -5191,8 +5297,7 @@ var _Saurashtra = &RangeTable{
 var _Sharada = &RangeTable{
 	R16: []Range16{},
 	R32: []Range32{
-		{0x11180, 0x111cd, 1},
-		{0x111d0, 0x111df, 1},
+		{0x11180, 0x111df, 1},
 	},
 }
 
@@ -5222,7 +5327,7 @@ var _SignWriting = &RangeTable{
 
 var _Sinhala = &RangeTable{
 	R16: []Range16{
-		{0x0d82, 0x0d83, 1},
+		{0x0d81, 0x0d83, 1},
 		{0x0d85, 0x0d96, 1},
 		{0x0d9a, 0x0db1, 1},
 		{0x0db3, 0x0dbb, 1},
@@ -5271,7 +5376,7 @@ var _Sundanese = &RangeTable{
 
 var _Syloti_Nagri = &RangeTable{
 	R16: []Range16{
-		{0xa800, 0xa82b, 1},
+		{0xa800, 0xa82c, 1},
 	},
 }
 
@@ -5360,7 +5465,8 @@ var _Tangut = &RangeTable{
 	R32: []Range32{
 		{0x16fe0, 0x17000, 32},
 		{0x17001, 0x187f7, 1},
-		{0x18800, 0x18af2, 1},
+		{0x18800, 0x18aff, 1},
+		{0x18d00, 0x18d08, 1},
 	},
 }
 
@@ -5452,6 +5558,15 @@ var _Warang_Citi = &RangeTable{
 	},
 }
 
+var _Yezidi = &RangeTable{
+	R16: []Range16{},
+	R32: []Range32{
+		{0x10e80, 0x10ea9, 1},
+		{0x10eab, 0x10ead, 1},
+		{0x10eb0, 0x10eb1, 1},
+	},
+}
+
 var _Yi = &RangeTable{
 	R16: []Range16{
 		{0xa000, 0xa48c, 1},
@@ -5491,6 +5606,7 @@ var (
 	Chakma                 = _Chakma                 // Chakma is the set of Unicode characters in script Chakma.
 	Cham                   = _Cham                   // Cham is the set of Unicode characters in script Cham.
 	Cherokee               = _Cherokee               // Cherokee is the set of Unicode characters in script Cherokee.
+	Chorasmian             = _Chorasmian             // Chorasmian is the set of Unicode characters in script Chorasmian.
 	Common                 = _Common                 // Common is the set of Unicode characters in script Common.
 	Coptic                 = _Coptic                 // Coptic is the set of Unicode characters in script Coptic.
 	Cuneiform              = _Cuneiform              // Cuneiform is the set of Unicode characters in script Cuneiform.
@@ -5498,6 +5614,7 @@ var (
 	Cyrillic               = _Cyrillic               // Cyrillic is the set of Unicode characters in script Cyrillic.
 	Deseret                = _Deseret                // Deseret is the set of Unicode characters in script Deseret.
 	Devanagari             = _Devanagari             // Devanagari is the set of Unicode characters in script Devanagari.
+	Dives_Akuru            = _Dives_Akuru            // Dives_Akuru is the set of Unicode characters in script Dives_Akuru.
 	Dogra                  = _Dogra                  // Dogra is the set of Unicode characters in script Dogra.
 	Duployan               = _Duployan               // Duployan is the set of Unicode characters in script Duployan.
 	Egyptian_Hieroglyphs   = _Egyptian_Hieroglyphs   // Egyptian_Hieroglyphs is the set of Unicode characters in script Egyptian_Hieroglyphs.
@@ -5529,6 +5646,7 @@ var (
 	Katakana               = _Katakana               // Katakana is the set of Unicode characters in script Katakana.
 	Kayah_Li               = _Kayah_Li               // Kayah_Li is the set of Unicode characters in script Kayah_Li.
 	Kharoshthi             = _Kharoshthi             // Kharoshthi is the set of Unicode characters in script Kharoshthi.
+	Khitan_Small_Script    = _Khitan_Small_Script    // Khitan_Small_Script is the set of Unicode characters in script Khitan_Small_Script.
 	Khmer                  = _Khmer                  // Khmer is the set of Unicode characters in script Khmer.
 	Khojki                 = _Khojki                 // Khojki is the set of Unicode characters in script Khojki.
 	Khudawadi              = _Khudawadi              // Khudawadi is the set of Unicode characters in script Khudawadi.
@@ -5618,6 +5736,7 @@ var (
 	Vai                    = _Vai                    // Vai is the set of Unicode characters in script Vai.
 	Wancho                 = _Wancho                 // Wancho is the set of Unicode characters in script Wancho.
 	Warang_Citi            = _Warang_Citi            // Warang_Citi is the set of Unicode characters in script Warang_Citi.
+	Yezidi                 = _Yezidi                 // Yezidi is the set of Unicode characters in script Yezidi.
 	Yi                     = _Yi                     // Yi is the set of Unicode characters in script Yi.
 	Zanabazar_Square       = _Zanabazar_Square       // Zanabazar_Square is the set of Unicode characters in script Zanabazar_Square.
 )
@@ -5695,6 +5814,9 @@ var _Dash = &RangeTable{
 		{0xfe58, 0xfe63, 11},
 		{0xff0d, 0xff0d, 1},
 	},
+	R32: []Range32{
+		{0x10ead, 0x10ead, 1},
+	},
 }
 
 var _Deprecated = &RangeTable{
@@ -5746,10 +5868,11 @@ var _Diacritic = &RangeTable{
 		{0x0acd, 0x0afd, 48},
 		{0x0afe, 0x0aff, 1},
 		{0x0b3c, 0x0b4d, 17},
-		{0x0bcd, 0x0c4d, 128},
-		{0x0cbc, 0x0ccd, 17},
-		{0x0d3b, 0x0d3c, 1},
-		{0x0d4d, 0x0e47, 125},
+		{0x0b55, 0x0bcd, 120},
+		{0x0c4d, 0x0cbc, 111},
+		{0x0ccd, 0x0d3b, 110},
+		{0x0d3c, 0x0d4d, 17},
+		{0x0dca, 0x0e47, 125},
 		{0x0e48, 0x0e4c, 1},
 		{0x0e4e, 0x0eba, 108},
 		{0x0ec8, 0x0ecc, 1},
@@ -5811,6 +5934,7 @@ var _Diacritic = &RangeTable{
 		{0xaabf, 0xaac2, 1},
 		{0xaaf6, 0xab5b, 101},
 		{0xab5c, 0xab5f, 1},
+		{0xab69, 0xab6b, 1},
 		{0xabec, 0xabed, 1},
 		{0xfb1e, 0xfe20, 770},
 		{0xfe21, 0xfe2f, 1},
@@ -5838,14 +5962,16 @@ var _Diacritic = &RangeTable{
 		{0x1163f, 0x116b6, 119},
 		{0x116b7, 0x1172b, 116},
 		{0x11839, 0x1183a, 1},
-		{0x119e0, 0x11a34, 84},
-		{0x11a47, 0x11a99, 82},
-		{0x11c3f, 0x11d42, 259},
-		{0x11d44, 0x11d45, 1},
-		{0x11d97, 0x16af0, 19801},
-		{0x16af1, 0x16af4, 1},
+		{0x1193d, 0x1193e, 1},
+		{0x11943, 0x119e0, 157},
+		{0x11a34, 0x11a47, 19},
+		{0x11a99, 0x11c3f, 422},
+		{0x11d42, 0x11d44, 2},
+		{0x11d45, 0x11d97, 82},
+		{0x16af0, 0x16af4, 1},
 		{0x16b30, 0x16b36, 1},
 		{0x16f8f, 0x16f9f, 1},
+		{0x16ff0, 0x16ff1, 1},
 		{0x1d167, 0x1d169, 1},
 		{0x1d16d, 0x1d172, 1},
 		{0x1d17b, 0x1d182, 1},
@@ -5864,12 +5990,12 @@ var _Extender = &RangeTable{
 	R16: []Range16{
 		{0x00b7, 0x02d0, 537},
 		{0x02d1, 0x0640, 879},
-		{0x07fa, 0x0e46, 1612},
-		{0x0ec6, 0x180a, 2372},
-		{0x1843, 0x1aa7, 612},
-		{0x1c36, 0x1c7b, 69},
-		{0x3005, 0x3031, 44},
-		{0x3032, 0x3035, 1},
+		{0x07fa, 0x0b55, 859},
+		{0x0e46, 0x0ec6, 128},
+		{0x180a, 0x1843, 57},
+		{0x1aa7, 0x1c36, 399},
+		{0x1c7b, 0x3005, 5002},
+		{0x3031, 0x3035, 1},
 		{0x309d, 0x309e, 1},
 		{0x30fc, 0x30fe, 1},
 		{0xa015, 0xa60c, 1527},
@@ -5931,21 +6057,24 @@ var _Ideographic = &RangeTable{
 		{0x3006, 0x3007, 1},
 		{0x3021, 0x3029, 1},
 		{0x3038, 0x303a, 1},
-		{0x3400, 0x4db5, 1},
-		{0x4e00, 0x9fef, 1},
+		{0x3400, 0x4dbf, 1},
+		{0x4e00, 0x9ffc, 1},
 		{0xf900, 0xfa6d, 1},
 		{0xfa70, 0xfad9, 1},
 	},
 	R32: []Range32{
-		{0x17000, 0x187f7, 1},
-		{0x18800, 0x18af2, 1},
+		{0x16fe4, 0x17000, 28},
+		{0x17001, 0x187f7, 1},
+		{0x18800, 0x18cd5, 1},
+		{0x18d00, 0x18d08, 1},
 		{0x1b170, 0x1b2fb, 1},
-		{0x20000, 0x2a6d6, 1},
+		{0x20000, 0x2a6dd, 1},
 		{0x2a700, 0x2b734, 1},
 		{0x2b740, 0x2b81d, 1},
 		{0x2b820, 0x2cea1, 1},
 		{0x2ceb0, 0x2ebe0, 1},
 		{0x2f800, 0x2fa1d, 1},
+		{0x30000, 0x3134a, 1},
 	},
 }
 
@@ -6066,9 +6195,9 @@ var _Other_Alphabetic = &RangeTable{
 		{0x0d46, 0x0d48, 1},
 		{0x0d4a, 0x0d4c, 1},
 		{0x0d57, 0x0d62, 11},
-		{0x0d63, 0x0d82, 31},
-		{0x0d83, 0x0dcf, 76},
-		{0x0dd0, 0x0dd4, 1},
+		{0x0d63, 0x0d81, 30},
+		{0x0d82, 0x0d83, 1},
+		{0x0dcf, 0x0dd4, 1},
 		{0x0dd6, 0x0dd8, 2},
 		{0x0dd9, 0x0ddf, 1},
 		{0x0df2, 0x0df3, 1},
@@ -6104,6 +6233,7 @@ var _Other_Alphabetic = &RangeTable{
 		{0x1a17, 0x1a1b, 1},
 		{0x1a55, 0x1a5e, 1},
 		{0x1a61, 0x1a74, 1},
+		{0x1abf, 0x1ac0, 1},
 		{0x1b00, 0x1b04, 1},
 		{0x1b35, 0x1b43, 1},
 		{0x1b80, 0x1b82, 1},
@@ -6145,6 +6275,7 @@ var _Other_Alphabetic = &RangeTable{
 		{0x10a05, 0x10a06, 1},
 		{0x10a0c, 0x10a0f, 1},
 		{0x10d24, 0x10d27, 1},
+		{0x10eab, 0x10eac, 1},
 		{0x11000, 0x11002, 1},
 		{0x11038, 0x11045, 1},
 		{0x11082, 0x110b0, 46},
@@ -6154,6 +6285,7 @@ var _Other_Alphabetic = &RangeTable{
 		{0x11145, 0x11146, 1},
 		{0x11180, 0x11182, 1},
 		{0x111b3, 0x111bf, 1},
+		{0x111ce, 0x111cf, 1},
 		{0x1122c, 0x11234, 1},
 		{0x11237, 0x1123e, 7},
 		{0x112df, 0x112e8, 1},
@@ -6174,6 +6306,10 @@ var _Other_Alphabetic = &RangeTable{
 		{0x116ac, 0x116b5, 1},
 		{0x1171d, 0x1172a, 1},
 		{0x1182c, 0x11838, 1},
+		{0x11930, 0x11935, 1},
+		{0x11937, 0x11938, 1},
+		{0x1193b, 0x1193c, 1},
+		{0x11940, 0x11942, 2},
 		{0x119d1, 0x119d7, 1},
 		{0x119da, 0x119df, 1},
 		{0x119e4, 0x11a01, 29},
@@ -6198,6 +6334,7 @@ var _Other_Alphabetic = &RangeTable{
 		{0x16f4f, 0x16f51, 2},
 		{0x16f52, 0x16f87, 1},
 		{0x16f8f, 0x16f92, 1},
+		{0x16ff0, 0x16ff1, 1},
 		{0x1bc9e, 0x1e000, 9058},
 		{0x1e001, 0x1e006, 1},
 		{0x1e008, 0x1e018, 1},
@@ -6243,8 +6380,9 @@ var _Other_Grapheme_Extend = &RangeTable{
 	R32: []Range32{
 		{0x1133e, 0x11357, 25},
 		{0x114b0, 0x114bd, 13},
-		{0x115af, 0x1d165, 48054},
-		{0x1d16e, 0x1d172, 1},
+		{0x115af, 0x11930, 897},
+		{0x1d165, 0x1d16e, 9},
+		{0x1d16f, 0x1d172, 1},
 		{0xe0020, 0xe007f, 1},
 	},
 }
@@ -6562,6 +6700,7 @@ var _Sentence_Terminal = &RangeTable{
 		{0x115c9, 0x115d7, 1},
 		{0x11641, 0x11642, 1},
 		{0x1173c, 0x1173e, 1},
+		{0x11944, 0x11946, 2},
 		{0x11a42, 0x11a43, 1},
 		{0x11a9b, 0x11a9c, 1},
 		{0x11c41, 0x11c42, 1},
@@ -6679,11 +6818,12 @@ var _Terminal_Punctuation = &RangeTable{
 		{0x11239, 0x1123c, 1},
 		{0x112a9, 0x1144b, 418},
 		{0x1144c, 0x1144d, 1},
-		{0x1145b, 0x115c2, 359},
-		{0x115c3, 0x115c5, 1},
+		{0x1145a, 0x1145b, 1},
+		{0x115c2, 0x115c5, 1},
 		{0x115c9, 0x115d7, 1},
 		{0x11641, 0x11642, 1},
 		{0x1173c, 0x1173e, 1},
+		{0x11944, 0x11946, 2},
 		{0x11a42, 0x11a43, 1},
 		{0x11a9b, 0x11a9c, 1},
 		{0x11aa1, 0x11aa2, 1},
@@ -6703,8 +6843,8 @@ var _Terminal_Punctuation = &RangeTable{
 
 var _Unified_Ideograph = &RangeTable{
 	R16: []Range16{
-		{0x3400, 0x4db5, 1},
-		{0x4e00, 0x9fef, 1},
+		{0x3400, 0x4dbf, 1},
+		{0x4e00, 0x9ffc, 1},
 		{0xfa0e, 0xfa0f, 1},
 		{0xfa11, 0xfa13, 2},
 		{0xfa14, 0xfa1f, 11},
@@ -6713,11 +6853,12 @@ var _Unified_Ideograph = &RangeTable{
 		{0xfa28, 0xfa29, 1},
 	},
 	R32: []Range32{
-		{0x20000, 0x2a6d6, 1},
+		{0x20000, 0x2a6dd, 1},
 		{0x2a700, 0x2b734, 1},
 		{0x2b740, 0x2b81d, 1},
 		{0x2b820, 0x2cea1, 1},
 		{0x2ceb0, 0x2ebe0, 1},
+		{0x30000, 0x3134a, 1},
 	},
 }
 
@@ -7088,6 +7229,8 @@ var _CaseRanges = []CaseRange{
 	{0xA7C4, 0xA7C4, d{0, -48, 0}},
 	{0xA7C5, 0xA7C5, d{0, -42307, 0}},
 	{0xA7C6, 0xA7C6, d{0, -35384, 0}},
+	{0xA7C7, 0xA7CA, d{UpperLower, UpperLower, UpperLower}},
+	{0xA7F5, 0xA7F6, d{UpperLower, UpperLower, UpperLower}},
 	{0xAB53, 0xAB53, d{-928, 0, -928}},
 	{0xAB70, 0xABBF, d{-38864, 0, -38864}},
 	{0xFF21, 0xFF3A, d{0, 32, 0}},
@@ -7711,7 +7854,8 @@ var foldLl = &RangeTable{
 		{0xa7b0, 0xa7b4, 1},
 		{0xa7b6, 0xa7be, 2},
 		{0xa7c2, 0xa7c4, 2},
-		{0xa7c5, 0xa7c6, 1},
+		{0xa7c5, 0xa7c7, 1},
+		{0xa7c9, 0xa7f5, 44},
 		{0xff21, 0xff3a, 1},
 	},
 	R32: []Range32{
@@ -7844,8 +7988,10 @@ var foldLu = &RangeTable{
 		{0xa793, 0xa794, 1},
 		{0xa797, 0xa7a9, 2},
 		{0xa7b5, 0xa7bf, 2},
-		{0xa7c3, 0xab53, 912},
-		{0xab70, 0xabbf, 1},
+		{0xa7c3, 0xa7c8, 5},
+		{0xa7ca, 0xa7f6, 44},
+		{0xab53, 0xab70, 29},
+		{0xab71, 0xabbf, 1},
 		{0xff41, 0xff5a, 1},
 	},
 	R32: []Range32{
@@ -7902,7 +8048,7 @@ var foldInherited = &RangeTable{
 	},
 }
 
-// Range entries: 3483 16-bit, 1730 32-bit, 5213 total.
-// Range bytes: 20898 16-bit, 20760 32-bit, 41658 total.
+// Range entries: 3499 16-bit, 1820 32-bit, 5319 total.
+// Range bytes: 20994 16-bit, 21840 32-bit, 42834 total.
 
 // Fold orbit bytes: 88 pairs, 352 bytes