libgo: update to Go1.14rc1 release

Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/218017

35 files changed, 664 insertions, 319 deletions

diff --git a/libgo/go/runtime/chan.go b/libgo/go/runtime/chan.go
index 549e566..ec8252b 100644
--- a/libgo/go/runtime/chan.go
+++ b/libgo/go/runtime/chan.go
@@ -133,21 +133,6 @@ func chanbuf(c *hchan, i uint) unsafe.Pointer {
 	return add(c.buf, uintptr(i)*uintptr(c.elemsize))
 }
 
-// full reports whether a send on c would block (that is, the channel is full).
-// It uses a single word-sized read of mutable state, so although
-// the answer is instantaneously true, the correct answer may have changed
-// by the time the calling function receives the return value.
-func full(c *hchan) bool {
-	// c.dataqsiz is immutable (never written after the channel is created)
-	// so it is safe to read at any time during channel operation.
-	if c.dataqsiz == 0 {
-		// Assumes that a pointer read is relaxed-atomic.
-		return c.recvq.first == nil
-	}
-	// Assumes that a uint read is relaxed-atomic.
-	return c.qcount == c.dataqsiz
-}
-
 // entry point for c <- x from compiled code
 //go:nosplit
 func chansend1(c *hchan, elem unsafe.Pointer) {
@@ -192,7 +177,7 @@ func chansend(c *hchan, ep unsafe.Pointer, block bool, callerpc uintptr) bool {
 	//
 	// After observing that the channel is not closed, we observe that the channel is
 	// not ready for sending. Each of these observations is a single word-sized read
-	// (first c.closed and second full()).
+	// (first c.closed and second c.recvq.first or c.qcount depending on kind of channel).
 	// Because a closed channel cannot transition from 'ready for sending' to
 	// 'not ready for sending', even if the channel is closed between the two observations,
 	// they imply a moment between the two when the channel was both not yet closed
@@ -201,10 +186,9 @@ func chansend(c *hchan, ep unsafe.Pointer, block bool, callerpc uintptr) bool {
 	//
 	// It is okay if the reads are reordered here: if we observe that the channel is not
 	// ready for sending and then observe that it is not closed, that implies that the
-	// channel wasn't closed during the first observation. However, nothing here
-	// guarantees forward progress. We rely on the side effects of lock release in
-	// chanrecv() and closechan() to update this thread's view of c.closed and full().
-	if !block && c.closed == 0 && full(c) {
+	// channel wasn't closed during the first observation.
+	if !block && c.closed == 0 && ((c.dataqsiz == 0 && c.recvq.first == nil) ||
+		(c.dataqsiz > 0 && c.qcount == c.dataqsiz)) {
 		return false
 	}
 
@@ -434,16 +418,6 @@ func closechan(c *hchan) {
 	}
 }
 
-// empty reports whether a read from c would block (that is, the channel is
-// empty).  It uses a single atomic read of mutable state.
-func empty(c *hchan) bool {
-	// c.dataqsiz is immutable.
-	if c.dataqsiz == 0 {
-		return atomic.Loadp(unsafe.Pointer(&c.sendq.first)) == nil
-	}
-	return atomic.Loaduint(&c.qcount) == 0
-}
-
 // entry points for <- c from compiled code
 //go:nosplit
 func chanrecv1(c *hchan, elem unsafe.Pointer) {
@@ -484,33 +458,21 @@ func chanrecv(c *hchan, ep unsafe.Pointer, block bool) (selected, received bool)
 	}
 
 	// Fast path: check for failed non-blocking operation without acquiring the lock.
-	if !block && empty(c) {
-		// After observing that the channel is not ready for receiving, we observe whether the
-		// channel is closed.
-		//
-		// Reordering of these checks could lead to incorrect behavior when racing with a close.
-		// For example, if the channel was open and not empty, was closed, and then drained,
-		// reordered reads could incorrectly indicate "open and empty". To prevent reordering,
-		// we use atomic loads for both checks, and rely on emptying and closing to happen in
-		// separate critical sections under the same lock.  This assumption fails when closing
-		// an unbuffered channel with a blocked send, but that is an error condition anyway.
-		if atomic.Load(&c.closed) == 0 {
-			// Because a channel cannot be reopened, the later observation of the channel
-			// being not closed implies that it was also not closed at the moment of the
-			// first observation. We behave as if we observed the channel at that moment
-			// and report that the receive cannot proceed.
-			return
-		}
-		// The channel is irreversibly closed. Re-check whether the channel has any pending data
-		// to receive, which could have arrived between the empty and closed checks above.
-		// Sequential consistency is also required here, when racing with such a send.
-		if empty(c) {
-			// The channel is irreversibly closed and empty.
-			if ep != nil {
-				typedmemclr(c.elemtype, ep)
-			}
-			return true, false
-		}
+	//
+	// After observing that the channel is not ready for receiving, we observe that the
+	// channel is not closed. Each of these observations is a single word-sized read
+	// (first c.sendq.first or c.qcount, and second c.closed).
+	// Because a channel cannot be reopened, the later observation of the channel
+	// being not closed implies that it was also not closed at the moment of the
+	// first observation. We behave as if we observed the channel at that moment
+	// and report that the receive cannot proceed.
+	//
+	// The order of operations is important here: reversing the operations can lead to
+	// incorrect behavior when racing with a close.
+	if !block && (c.dataqsiz == 0 && c.sendq.first == nil ||
+		c.dataqsiz > 0 && atomic.Loaduint(&c.qcount) == 0) &&
+		atomic.Load(&c.closed) == 0 {
+		return
 	}
 
 	var t0 int64
diff --git a/libgo/go/runtime/chan_test.go b/libgo/go/runtime/chan_test.go
index ac81d40..c194781 100644
--- a/libgo/go/runtime/chan_test.go
+++ b/libgo/go/runtime/chan_test.go
@@ -1132,20 +1132,6 @@ func BenchmarkChanPopular(b *testing.B) {
 	wg.Wait()
 }
 
-func BenchmarkChanClosed(b *testing.B) {
-	c := make(chan struct{})
-	close(c)
-	b.RunParallel(func(pb *testing.PB) {
-		for pb.Next() {
-			select {
-			case <-c:
-			default:
-				b.Error("Unreachable")
-			}
-		}
-	})
-}
-
 var (
 	alwaysFalse = false
 	workSink    = 0
diff --git a/libgo/go/runtime/checkptr.go b/libgo/go/runtime/checkptr.go
index f478ddd..974f0a0 100644
--- a/libgo/go/runtime/checkptr.go
+++ b/libgo/go/runtime/checkptr.go
@@ -8,45 +8,22 @@ package runtime
 
 import "unsafe"
 
-type ptrAlignError struct {
-	ptr  unsafe.Pointer
-	elem *_type
-	n    uintptr
-}
-
-func (e ptrAlignError) RuntimeError() {}
-
-func (e ptrAlignError) Error() string {
-	return "runtime error: unsafe pointer conversion"
-}
-
 func checkptrAlignment(p unsafe.Pointer, elem *_type, n uintptr) {
 	// Check that (*[n]elem)(p) is appropriately aligned.
 	// TODO(mdempsky): What about fieldAlign?
 	if uintptr(p)&(uintptr(elem.align)-1) != 0 {
-		panic(ptrAlignError{p, elem, n})
+		throw("checkptr: unsafe pointer conversion")
 	}
 
 	// Check that (*[n]elem)(p) doesn't straddle multiple heap objects.
 	if size := n * elem.size; size > 1 && checkptrBase(p) != checkptrBase(add(p, size-1)) {
-		panic(ptrAlignError{p, elem, n})
+		throw("checkptr: unsafe pointer conversion")
 	}
 }
 
-type ptrArithError struct {
-	ptr       unsafe.Pointer
-	originals []unsafe.Pointer
-}
-
-func (e ptrArithError) RuntimeError() {}
-
-func (e ptrArithError) Error() string {
-	return "runtime error: unsafe pointer arithmetic"
-}
-
 func checkptrArithmetic(p unsafe.Pointer, originals []unsafe.Pointer) {
 	if 0 < uintptr(p) && uintptr(p) < minLegalPointer {
-		panic(ptrArithError{p, originals})
+		throw("checkptr: unsafe pointer arithmetic")
 	}
 
 	// Check that if the computed pointer p points into a heap
@@ -63,7 +40,7 @@ func checkptrArithmetic(p unsafe.Pointer, originals []unsafe.Pointer) {
 		}
 	}
 
-	panic(ptrArithError{p, originals})
+	throw("checkptr: unsafe pointer arithmetic")
 }
 
 // checkptrBase returns the base address for the allocation containing
diff --git a/libgo/go/runtime/checkptr_test.go b/libgo/go/runtime/checkptr_test.go
new file mode 100644
index 0000000..ab3058f
--- /dev/null
+++ b/libgo/go/runtime/checkptr_test.go
@@ -0,0 +1,50 @@
+// 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 runtime_test
+
+import (
+	"internal/testenv"
+	"os/exec"
+	"runtime"
+	"strings"
+	"testing"
+)
+
+func TestCheckPtr(t *testing.T) {
+	if runtime.Compiler == "gccgo" {
+		t.Skip("gccgo does not have -d=checkptr")
+	}
+	t.Parallel()
+	testenv.MustHaveGoRun(t)
+
+	exe, err := buildTestProg(t, "testprog", "-gcflags=all=-d=checkptr=1")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	testCases := []struct {
+		cmd  string
+		want string
+	}{
+		{"CheckPtrAlignment", "fatal error: checkptr: unsafe pointer conversion\n"},
+		{"CheckPtrArithmetic", "fatal error: checkptr: unsafe pointer arithmetic\n"},
+		{"CheckPtrSize", "fatal error: checkptr: unsafe pointer conversion\n"},
+		{"CheckPtrSmall", "fatal error: checkptr: unsafe pointer arithmetic\n"},
+	}
+
+	for _, tc := range testCases {
+		tc := tc
+		t.Run(tc.cmd, func(t *testing.T) {
+			t.Parallel()
+			got, err := testenv.CleanCmdEnv(exec.Command(exe, tc.cmd)).CombinedOutput()
+			if err != nil {
+				t.Log(err)
+			}
+			if !strings.HasPrefix(string(got), tc.want) {
+				t.Errorf("output:\n%s\n\nwant output starting with: %s", got, tc.want)
+			}
+		})
+	}
+}
diff --git a/libgo/go/runtime/debug.go b/libgo/go/runtime/debug.go
index 1202e36..e480466 100644
--- a/libgo/go/runtime/debug.go
+++ b/libgo/go/runtime/debug.go
@@ -26,12 +26,12 @@ func GOMAXPROCS(n int) int {
 		return ret
 	}
 
-	stopTheWorldGC("GOMAXPROCS")
+	stopTheWorld("GOMAXPROCS")
 
 	// newprocs will be processed by startTheWorld
 	newprocs = int32(n)
 
-	startTheWorldGC()
+	startTheWorld()
 	return ret
 }
 
diff --git a/libgo/go/runtime/export_test.go b/libgo/go/runtime/export_test.go
index 9a977d8..b60c19b 100644
--- a/libgo/go/runtime/export_test.go
+++ b/libgo/go/runtime/export_test.go
@@ -45,6 +45,9 @@ var NetpollGenericInit = netpollGenericInit
 
 var ParseRelease = parseRelease
 
+var Memmove = memmove
+var MemclrNoHeapPointers = memclrNoHeapPointers
+
 const PreemptMSupported = preemptMSupported
 
 type LFNode struct {
@@ -573,6 +576,7 @@ const (
 	PageSize         = pageSize
 	PallocChunkPages = pallocChunkPages
 	PageAlloc64Bit   = pageAlloc64Bit
+	PallocSumBytes   = pallocSumBytes
 )
 
 // Expose pallocSum for testing.
diff --git a/libgo/go/runtime/extern.go b/libgo/go/runtime/extern.go
index e2e601f..96af606 100644
--- a/libgo/go/runtime/extern.go
+++ b/libgo/go/runtime/extern.go
@@ -78,21 +78,6 @@ It is a comma-separated list of name=val pairs setting these named variables:
 	If the line ends with "(forced)", this GC was forced by a
 	runtime.GC() call.
 
-	Setting gctrace to any value > 0 also causes the garbage collector
-	to emit a summary when memory is released back to the system.
-	This process of returning memory to the system is called scavenging.
-	The format of this summary is subject to change.
-	Currently it is:
-		scvg#: # MB released  printed only if non-zero
-		scvg#: inuse: # idle: # sys: # released: # consumed: # (MB)
-	where the fields are as follows:
-		scvg#        the scavenge cycle number, incremented at each scavenge
-		inuse: #     MB used or partially used spans
-		idle: #      MB spans pending scavenging
-		sys: #       MB mapped from the system
-		released: #  MB released to the system
-		consumed: #  MB allocated from the system
-
 	madvdontneed: setting madvdontneed=1 will use MADV_DONTNEED
 	instead of MADV_FREE on Linux when returning memory to the
 	kernel. This is less efficient, but causes RSS numbers to drop
@@ -112,6 +97,19 @@ It is a comma-separated list of name=val pairs setting these named variables:
 
 	scavenge: scavenge=1 enables debugging mode of heap scavenger.
 
+	scavtrace: setting scavtrace=1 causes the runtime to emit a single line to standard
+	error, roughly once per GC cycle, summarizing the amount of work done by the
+	scavenger as well as the total amount of memory returned to the operating system
+	and an estimate of physical memory utilization. The format of this line is subject
+	to change, but currently it is:
+		scav # KiB work, # KiB total, #% util
+	where the fields are as follows:
+		# KiB work   the amount of memory returned to the OS since the last scav line
+		# KiB total  how much of the heap at this point in time has been released to the OS
+		#% util      the fraction of all unscavenged memory which is in-use
+	If the line ends with "(forced)", then scavenging was forced by a
+	debug.FreeOSMemory() call.
+
 	scheddetail: setting schedtrace=X and scheddetail=1 causes the scheduler to emit
 	detailed multiline info every X milliseconds, describing state of the scheduler,
 	processors, threads and goroutines.
diff --git a/libgo/go/runtime/gcinfo_test.go b/libgo/go/runtime/gcinfo_test.go
index fc24f04..ddbe5dd 100644
--- a/libgo/go/runtime/gcinfo_test.go
+++ b/libgo/go/runtime/gcinfo_test.go
@@ -165,7 +165,7 @@ func infoBigStruct() []byte {
 			typeScalar, typeScalar, typeScalar, typeScalar, // t int; y uint16; u uint64
 			typePointer, typeScalar, // i string
 		}
-	case "arm64", "amd64", "mips64", "mips64le", "ppc64", "ppc64le", "s390x", "wasm":
+	case "arm64", "amd64", "mips64", "mips64le", "ppc64", "ppc64le", "riscv64", "s390x", "wasm":
 		return []byte{
 			typePointer,                        // q *int
 			typeScalar, typeScalar, typeScalar, // w byte; e [17]byte
diff --git a/libgo/go/runtime/hash64.go b/libgo/go/runtime/hash64.go
index cff663a..704bbe6 100644
--- a/libgo/go/runtime/hash64.go
+++ b/libgo/go/runtime/hash64.go
@@ -6,7 +6,7 @@
 //   xxhash: https://code.google.com/p/xxhash/
 // cityhash: https://code.google.com/p/cityhash/
 
-// +build amd64 amd64p32 arm64 mips64 mips64le ppc64 ppc64le s390x wasm alpha arm64be ia64 mips64p32 mips64p32le sparc64 riscv64
+// +build amd64 arm64 mips64 mips64le ppc64 ppc64le riscv64 s390x wasm alpha amd64p32 arm64be ia64 mips64p32 mips64p32le sparc64
 
 package runtime
 
diff --git a/libgo/go/runtime/lfstack_64bit.go b/libgo/go/runtime/lfstack_64bit.go
index de40a00..af9e7d1 100644
--- a/libgo/go/runtime/lfstack_64bit.go
+++ b/libgo/go/runtime/lfstack_64bit.go
@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64 arm64 mips64 mips64le ppc64 ppc64le s390x wasm arm64be alpha sparc64 ia64 riscv64
+// +build amd64 arm64 mips64 mips64le ppc64 ppc64le riscv64 s390x wasm arm64be alpha sparc64 ia64
 
 package runtime
 
diff --git a/libgo/go/runtime/malloc.go b/libgo/go/runtime/malloc.go
index fda2273..35ace7f 100644
--- a/libgo/go/runtime/malloc.go
+++ b/libgo/go/runtime/malloc.go
@@ -513,6 +513,7 @@ func mallocinit() {
 		// allocation at 0x40 << 32 because when using 4k pages with 3-level
 		// translation buffers, the user address space is limited to 39 bits
 		// On darwin/arm64, the address space is even smaller.
+		//
 		// On AIX, mmaps starts at 0x0A00000000000000 for 64-bit.
 		// processes.
 		for i := 0x7f; i >= 0; i-- {
diff --git a/libgo/go/runtime/malloc_test.go b/libgo/go/runtime/malloc_test.go
index bd30bc1..45555ee 100644
--- a/libgo/go/runtime/malloc_test.go
+++ b/libgo/go/runtime/malloc_test.go
@@ -206,14 +206,6 @@ type acLink struct {
 var arenaCollisionSink []*acLink
 
 func TestArenaCollision(t *testing.T) {
-	if GOOS == "darwin" && race.Enabled {
-		// Skip this test on Darwin in race mode because Darwin 10.10 has
-		// issues following arena hints and runs out of them in race mode, so
-		// MAP_FIXED is used to ensure we keep the heap in the memory region the
-		// race detector expects.
-		// TODO(mknyszek): Delete this when Darwin 10.10 is no longer supported.
-		t.Skip("disabled on Darwin with race mode since MAP_FIXED is used")
-	}
 	testenv.MustHaveExec(t)
 
 	// Test that mheap.sysAlloc handles collisions with other
diff --git a/libgo/go/runtime/memmove_test.go b/libgo/go/runtime/memmove_test.go
index 0b2e191..396c130 100644
--- a/libgo/go/runtime/memmove_test.go
+++ b/libgo/go/runtime/memmove_test.go
@@ -11,7 +11,9 @@ import (
 	"internal/race"
 	"internal/testenv"
 	. "runtime"
+	"sync/atomic"
 	"testing"
+	"unsafe"
 )
 
 func TestMemmove(t *testing.T) {
@@ -206,6 +208,71 @@ func cmpb(a, b []byte) int {
 	return l
 }
 
+// Ensure that memmove writes pointers atomically, so the GC won't
+// observe a partially updated pointer.
+func TestMemmoveAtomicity(t *testing.T) {
+	if race.Enabled {
+		t.Skip("skip under the race detector -- this test is intentionally racy")
+	}
+
+	var x int
+
+	for _, backward := range []bool{true, false} {
+		for _, n := range []int{3, 4, 5, 6, 7, 8, 9, 10, 15, 25, 49} {
+			n := n
+
+			// test copying [N]*int.
+			sz := uintptr(n * PtrSize)
+			name := fmt.Sprint(sz)
+			if backward {
+				name += "-backward"
+			} else {
+				name += "-forward"
+			}
+			t.Run(name, func(t *testing.T) {
+				// Use overlapping src and dst to force forward/backward copy.
+				var s [100]*int
+				src := s[n-1 : 2*n-1]
+				dst := s[:n]
+				if backward {
+					src, dst = dst, src
+				}
+				for i := range src {
+					src[i] = &x
+				}
+				for i := range dst {
+					dst[i] = nil
+				}
+
+				var ready uint32
+				go func() {
+					sp := unsafe.Pointer(&src[0])
+					dp := unsafe.Pointer(&dst[0])
+					atomic.StoreUint32(&ready, 1)
+					for i := 0; i < 10000; i++ {
+						Memmove(dp, sp, sz)
+						MemclrNoHeapPointers(dp, sz)
+					}
+					atomic.StoreUint32(&ready, 2)
+				}()
+
+				for atomic.LoadUint32(&ready) == 0 {
+					Gosched()
+				}
+
+				for atomic.LoadUint32(&ready) != 2 {
+					for i := range dst {
+						p := dst[i]
+						if p != nil && p != &x {
+							t.Fatalf("got partially updated pointer %p at dst[%d], want either nil or %p", p, i, &x)
+						}
+					}
+				}
+			})
+		}
+	}
+}
+
 func benchmarkSizes(b *testing.B, sizes []int, fn func(b *testing.B, n int)) {
 	for _, n := range sizes {
 		b.Run(fmt.Sprint(n), func(b *testing.B) {
diff --git a/libgo/go/runtime/mgc.go b/libgo/go/runtime/mgc.go
index b0040f9..8ded306 100644
--- a/libgo/go/runtime/mgc.go
+++ b/libgo/go/runtime/mgc.go
@@ -1271,7 +1271,6 @@ func gcStart(trigger gcTrigger) {
 	}
 
 	// Ok, we're doing it! Stop everybody else
-	semacquire(&gcsema)
 	semacquire(&worldsema)
 
 	if trace.enabled {
@@ -1370,13 +1369,6 @@ func gcStart(trigger gcTrigger) {
 		work.pauseNS += now - work.pauseStart
 		work.tMark = now
 	})
-
-	// Release the world sema before Gosched() in STW mode
-	// because we will need to reacquire it later but before
-	// this goroutine becomes runnable again, and we could
-	// self-deadlock otherwise.
-	semrelease(&worldsema)
-
 	// In STW mode, we could block the instant systemstack
 	// returns, so don't do anything important here. Make sure we
 	// block rather than returning to user code.
@@ -1446,10 +1438,6 @@ top:
 		return
 	}
 
-	// forEachP needs worldsema to execute, and we'll need it to
-	// stop the world later, so acquire worldsema now.
-	semacquire(&worldsema)
-
 	// Flush all local buffers and collect flushedWork flags.
 	gcMarkDoneFlushed = 0
 	systemstack(func() {
@@ -1510,7 +1498,6 @@ top:
 		// work to do. Keep going. It's possible the
 		// transition condition became true again during the
 		// ragged barrier, so re-check it.
-		semrelease(&worldsema)
 		goto top
 	}
 
@@ -1587,7 +1574,6 @@ top:
 				now := startTheWorldWithSema(true)
 				work.pauseNS += now - work.pauseStart
 			})
-			semrelease(&worldsema)
 			goto top
 		}
 	}
@@ -1802,7 +1788,6 @@ func gcMarkTermination(nextTriggerRatio float64) {
 	}
 
 	semrelease(&worldsema)
-	semrelease(&gcsema)
 	// Careful: another GC cycle may start now.
 
 	releasem(mp)
diff --git a/libgo/go/runtime/mgcscavenge.go b/libgo/go/runtime/mgcscavenge.go
index f3856db..3b60b3d 100644
--- a/libgo/go/runtime/mgcscavenge.go
+++ b/libgo/go/runtime/mgcscavenge.go
@@ -80,6 +80,17 @@ const (
 	// maxPagesPerPhysPage is the maximum number of supported runtime pages per
 	// physical page, based on maxPhysPageSize.
 	maxPagesPerPhysPage = maxPhysPageSize / pageSize
+
+	// scavengeCostRatio is the approximate ratio between the costs of using previously
+	// scavenged memory and scavenging memory.
+	//
+	// For most systems the cost of scavenging greatly outweighs the costs
+	// associated with using scavenged memory, making this constant 0. On other systems
+	// (especially ones where "sysUsed" is not just a no-op) this cost is non-trivial.
+	//
+	// This ratio is used as part of multiplicative factor to help the scavenger account
+	// for the additional costs of using scavenged memory in its pacing.
+	scavengeCostRatio = 0.7 * sys.GoosDarwin
 )
 
 // heapRetained returns an estimate of the current heap RSS.
@@ -248,7 +259,7 @@ func bgscavenge(c chan int) {
 		released := uintptr(0)
 
 		// Time in scavenging critical section.
-		crit := int64(0)
+		crit := float64(0)
 
 		// Run on the system stack since we grab the heap lock,
 		// and a stack growth with the heap lock means a deadlock.
@@ -266,16 +277,10 @@ func bgscavenge(c chan int) {
 			// Scavenge one page, and measure the amount of time spent scavenging.
 			start := nanotime()
 			released = mheap_.pages.scavengeOne(physPageSize, false)
-			crit = nanotime() - start
+			atomic.Xadduintptr(&mheap_.pages.scavReleased, released)
+			crit = float64(nanotime() - start)
 		})
 
-		if debug.gctrace > 0 {
-			if released > 0 {
-				print("scvg: ", released>>10, " KB released\n")
-			}
-			print("scvg: inuse: ", memstats.heap_inuse>>20, ", idle: ", memstats.heap_idle>>20, ", sys: ", memstats.heap_sys>>20, ", released: ", memstats.heap_released>>20, ", consumed: ", (memstats.heap_sys-memstats.heap_released)>>20, " (MB)\n")
-		}
-
 		if released == 0 {
 			lock(&scavenge.lock)
 			scavenge.parked = true
@@ -283,6 +288,14 @@ func bgscavenge(c chan int) {
 			continue
 		}
 
+		// Multiply the critical time by 1 + the ratio of the costs of using
+		// scavenged memory vs. scavenging memory. This forces us to pay down
+		// the cost of reusing this memory eagerly by sleeping for a longer period
+		// of time and scavenging less frequently. More concretely, we avoid situations
+		// where we end up scavenging so often that we hurt allocation performance
+		// because of the additional overheads of using scavenged memory.
+		crit *= 1 + scavengeCostRatio
+
 		// If we spent more than 10 ms (for example, if the OS scheduled us away, or someone
 		// put their machine to sleep) in the critical section, bound the time we use to
 		// calculate at 10 ms to avoid letting the sleep time get arbitrarily high.
@@ -298,13 +311,13 @@ func bgscavenge(c chan int) {
 		// much, then scavengeEMWA < idealFraction, so we'll adjust the sleep time
 		// down.
 		adjust := scavengeEWMA / idealFraction
-		sleepTime := int64(adjust * float64(crit) / (scavengePercent / 100.0))
+		sleepTime := int64(adjust * crit / (scavengePercent / 100.0))
 
 		// Go to sleep.
 		slept := scavengeSleep(sleepTime)
 
 		// Compute the new ratio.
-		fraction := float64(crit) / float64(crit+slept)
+		fraction := crit / (crit + float64(slept))
 
 		// Set a lower bound on the fraction.
 		// Due to OS-related anomalies we may "sleep" for an inordinate amount
@@ -348,12 +361,39 @@ func (s *pageAlloc) scavenge(nbytes uintptr, locked bool) uintptr {
 	return released
 }
 
+// printScavTrace prints a scavenge trace line to standard error.
+//
+// released should be the amount of memory released since the last time this
+// was called, and forced indicates whether the scavenge was forced by the
+// application.
+func printScavTrace(released uintptr, forced bool) {
+	printlock()
+	print("scav ",
+		released>>10, " KiB work, ",
+		atomic.Load64(&memstats.heap_released)>>10, " KiB total, ",
+		(atomic.Load64(&memstats.heap_inuse)*100)/heapRetained(), "% util",
+	)
+	if forced {
+		print(" (forced)")
+	}
+	println()
+	printunlock()
+}
+
 // resetScavengeAddr sets the scavenge start address to the top of the heap's
 // address space. This should be called each time the scavenger's pacing
 // changes.
 //
 // s.mheapLock must be held.
 func (s *pageAlloc) resetScavengeAddr() {
+	released := atomic.Loaduintptr(&s.scavReleased)
+	if debug.scavtrace > 0 {
+		printScavTrace(released, false)
+	}
+	// Subtract from scavReleased instead of just setting it to zero because
+	// the scavenger could have increased scavReleased concurrently with the
+	// load above, and we may miss an update by just blindly zeroing the field.
+	atomic.Xadduintptr(&s.scavReleased, -released)
 	s.scavAddr = chunkBase(s.end) - 1
 }
 
@@ -415,7 +455,10 @@ func (s *pageAlloc) scavengeOne(max uintptr, locked bool) uintptr {
 
 	// Check the chunk containing the scav addr, starting at the addr
 	// and see if there are any free and unscavenged pages.
-	if s.summary[len(s.summary)-1][ci].max() >= uint(minPages) {
+	//
+	// Only check this if s.scavAddr is covered by any address range
+	// in s.inUse, so that we know our check of the summary is safe.
+	if s.inUse.contains(s.scavAddr) && s.summary[len(s.summary)-1][ci].max() >= uint(minPages) {
 		// We only bother looking for a candidate if there at least
 		// minPages free pages at all. It's important that we only
 		// continue if the summary says we can because that's how
diff --git a/libgo/go/runtime/mgcscavenge_test.go b/libgo/go/runtime/mgcscavenge_test.go
index 518d5ab..58f9e3a 100644
--- a/libgo/go/runtime/mgcscavenge_test.go
+++ b/libgo/go/runtime/mgcscavenge_test.go
@@ -272,6 +272,9 @@ func TestPallocDataFindScavengeCandidate(t *testing.T) {
 
 // Tests end-to-end scavenging on a pageAlloc.
 func TestPageAllocScavenge(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	type test struct {
 		request, expect uintptr
 	}
@@ -279,12 +282,13 @@ func TestPageAllocScavenge(t *testing.T) {
 	if minPages < 1 {
 		minPages = 1
 	}
-	tests := map[string]struct {
+	type setup struct {
 		beforeAlloc map[ChunkIdx][]BitRange
 		beforeScav  map[ChunkIdx][]BitRange
 		expect      []test
 		afterScav   map[ChunkIdx][]BitRange
-	}{
+	}
+	tests := map[string]setup{
 		"AllFreeUnscavExhaust": {
 			beforeAlloc: map[ChunkIdx][]BitRange{
 				BaseChunkIdx:     {},
@@ -393,6 +397,26 @@ func TestPageAllocScavenge(t *testing.T) {
 			},
 		},
 	}
+	if PageAlloc64Bit != 0 {
+		tests["ScavAllVeryDiscontiguous"] = setup{
+			beforeAlloc: map[ChunkIdx][]BitRange{
+				BaseChunkIdx:          {},
+				BaseChunkIdx + 0x1000: {},
+			},
+			beforeScav: map[ChunkIdx][]BitRange{
+				BaseChunkIdx:          {},
+				BaseChunkIdx + 0x1000: {},
+			},
+			expect: []test{
+				{^uintptr(0), 2 * PallocChunkPages * PageSize},
+				{^uintptr(0), 0},
+			},
+			afterScav: map[ChunkIdx][]BitRange{
+				BaseChunkIdx:          {{0, PallocChunkPages}},
+				BaseChunkIdx + 0x1000: {{0, PallocChunkPages}},
+			},
+		}
+	}
 	for name, v := range tests {
 		v := v
 		runTest := func(t *testing.T, locked bool) {
diff --git a/libgo/go/runtime/mheap.go b/libgo/go/runtime/mheap.go
index f40589a..c40c9e2 100644
--- a/libgo/go/runtime/mheap.go
+++ b/libgo/go/runtime/mheap.go
@@ -70,7 +70,7 @@ type mheap struct {
 	// on the swept stack.
 	sweepSpans [2]gcSweepBuf
 
-	_ uint32 // align uint64 fields on 32-bit for atomics
+	// _ uint32 // align uint64 fields on 32-bit for atomics
 
 	// Proportional sweep
 	//
@@ -786,7 +786,9 @@ func (h *mheap) reclaim(npage uintptr) {
 // reclaimChunk sweeps unmarked spans that start at page indexes [pageIdx, pageIdx+n).
 // It returns the number of pages returned to the heap.
 //
-// h.lock must be held and the caller must be non-preemptible.
+// h.lock must be held and the caller must be non-preemptible. Note: h.lock may be
+// temporarily unlocked and re-locked in order to do sweeping or if tracing is
+// enabled.
 func (h *mheap) reclaimChunk(arenas []arenaIdx, pageIdx, n uintptr) uintptr {
 	// The heap lock must be held because this accesses the
 	// heapArena.spans arrays using potentially non-live pointers.
@@ -842,8 +844,10 @@ func (h *mheap) reclaimChunk(arenas []arenaIdx, pageIdx, n uintptr) uintptr {
 		n -= uintptr(len(inUse) * 8)
 	}
 	if trace.enabled {
+		unlock(&h.lock)
 		// Account for pages scanned but not reclaimed.
 		traceGCSweepSpan((n0 - nFreed) * pageSize)
+		lock(&h.lock)
 	}
 	return nFreed
 }
@@ -1430,11 +1434,8 @@ func (h *mheap) scavengeAll() {
 	unlock(&h.lock)
 	gp.m.mallocing--
 
-	if debug.gctrace > 0 {
-		if released > 0 {
-			print("forced scvg: ", released>>20, " MB released\n")
-		}
-		print("forced scvg: inuse: ", memstats.heap_inuse>>20, ", idle: ", memstats.heap_idle>>20, ", sys: ", memstats.heap_sys>>20, ", released: ", memstats.heap_released>>20, ", consumed: ", (memstats.heap_sys-memstats.heap_released)>>20, " (MB)\n")
+	if debug.scavtrace > 0 {
+		printScavTrace(released, true)
 	}
 }
 
diff --git a/libgo/go/runtime/mkpreempt.go b/libgo/go/runtime/mkpreempt.go
index 615ec18..64e2207 100644
--- a/libgo/go/runtime/mkpreempt.go
+++ b/libgo/go/runtime/mkpreempt.go
@@ -83,6 +83,7 @@ var arches = map[string]func(){
 	"mips64x": func() { genMIPS(true) },
 	"mipsx":   func() { genMIPS(false) },
 	"ppc64x":  genPPC64,
+	"riscv64": genRISCV64,
 	"s390x":   genS390X,
 	"wasm":    genWasm,
 }
@@ -478,6 +479,11 @@ func genPPC64() {
 	p("JMP (CTR)")
 }
 
+func genRISCV64() {
+	p("// No async preemption on riscv64 - see issue 36711")
+	p("UNDEF")
+}
+
 func genS390X() {
 	// Add integer registers R0-R12
 	// R13 (g), R14 (LR), R15 (SP) are special, and not saved here.
diff --git a/libgo/go/runtime/mpagealloc.go b/libgo/go/runtime/mpagealloc.go
index 572e6a9..bb751f1 100644
--- a/libgo/go/runtime/mpagealloc.go
+++ b/libgo/go/runtime/mpagealloc.go
@@ -225,7 +225,9 @@ type pageAlloc struct {
 	// the bitmaps align better on zero-values.
 	chunks [1 << pallocChunksL1Bits]*[1 << pallocChunksL2Bits]pallocData
 
-	// The address to start an allocation search with.
+	// The address to start an allocation search with. It must never
+	// point to any memory that is not contained in inUse, i.e.
+	// inUse.contains(searchAddr) must always be true.
 	//
 	// When added with arenaBaseOffset, we guarantee that
 	// all valid heap addresses (when also added with
@@ -237,9 +239,15 @@ type pageAlloc struct {
 	// space on architectures with segmented address spaces.
 	searchAddr uintptr
 
-	// The address to start a scavenge candidate search with.
+	// The address to start a scavenge candidate search with. It
+	// need not point to memory contained in inUse.
 	scavAddr uintptr
 
+	// The amount of memory scavenged since the last scavtrace print.
+	//
+	// Read and updated atomically.
+	scavReleased uintptr
+
 	// start and end represent the chunk indices
 	// which pageAlloc knows about. It assumes
 	// chunks in the range [start, end) are
diff --git a/libgo/go/runtime/mpagealloc_64bit.go b/libgo/go/runtime/mpagealloc_64bit.go
index dd44da1..385b7b3 100644
--- a/libgo/go/runtime/mpagealloc_64bit.go
+++ b/libgo/go/runtime/mpagealloc_64bit.go
@@ -2,7 +2,7 @@
 // Use of this source code is governed by a BSD-style
 // license that can be found in the LICENSE file.
 
-// +build amd64 !darwin,arm64 mips64 mips64le ppc64 ppc64le s390x arm64be alpha sparc64 ia64 riscv64
+// +build amd64 !darwin,arm64 mips64 mips64le ppc64 ppc64le riscv64 s390x arm64be alpha sparc64 ia64
 
 // See mpagealloc_32bit.go for why darwin/arm64 is excluded here.
 
diff --git a/libgo/go/runtime/mpagealloc_test.go b/libgo/go/runtime/mpagealloc_test.go
index 6c48296..89a4a25 100644
--- a/libgo/go/runtime/mpagealloc_test.go
+++ b/libgo/go/runtime/mpagealloc_test.go
@@ -41,6 +41,9 @@ func checkPageAlloc(t *testing.T, want, got *PageAlloc) {
 }
 
 func TestPageAllocGrow(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	type test struct {
 		chunks []ChunkIdx
 		inUse  []AddrRange
@@ -216,15 +219,19 @@ func TestPageAllocGrow(t *testing.T) {
 }
 
 func TestPageAllocAlloc(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	type hit struct {
 		npages, base, scav uintptr
 	}
-	tests := map[string]struct {
+	type test struct {
 		scav   map[ChunkIdx][]BitRange
 		before map[ChunkIdx][]BitRange
 		after  map[ChunkIdx][]BitRange
 		hits   []hit
-	}{
+	}
+	tests := map[string]test{
 		"AllFree1": {
 			before: map[ChunkIdx][]BitRange{
 				BaseChunkIdx: {},
@@ -365,7 +372,6 @@ func TestPageAllocAlloc(t *testing.T) {
 				BaseChunkIdx: {{0, 195}},
 			},
 		},
-		// TODO(mknyszek): Add tests close to the chunk size.
 		"ExhaustPallocChunkPages-3": {
 			before: map[ChunkIdx][]BitRange{
 				BaseChunkIdx: {},
@@ -565,6 +571,48 @@ func TestPageAllocAlloc(t *testing.T) {
 			},
 		},
 	}
+	if PageAlloc64Bit != 0 {
+		const chunkIdxBigJump = 0x100000 // chunk index offset which translates to O(TiB)
+
+		// This test attempts to trigger a bug wherein we look at unmapped summary
+		// memory that isn't just in the case where we exhaust the heap.
+		//
+		// It achieves this by placing a chunk such that its summary will be
+		// at the very end of a physical page. It then also places another chunk
+		// much further up in the address space, such that any allocations into the
+		// first chunk do not exhaust the heap and the second chunk's summary is not in the
+		// page immediately adjacent to the first chunk's summary's page.
+		// Allocating into this first chunk to exhaustion and then into the second
+		// chunk may then trigger a check in the allocator which erroneously looks at
+		// unmapped summary memory and crashes.
+
+		// Figure out how many chunks are in a physical page, then align BaseChunkIdx
+		// to a physical page in the chunk summary array. Here we only assume that
+		// each summary array is aligned to some physical page.
+		sumsPerPhysPage := ChunkIdx(PhysPageSize / PallocSumBytes)
+		baseChunkIdx := BaseChunkIdx &^ (sumsPerPhysPage - 1)
+		tests["DiscontiguousMappedSumBoundary"] = test{
+			before: map[ChunkIdx][]BitRange{
+				baseChunkIdx + sumsPerPhysPage - 1: {},
+				baseChunkIdx + chunkIdxBigJump:     {},
+			},
+			scav: map[ChunkIdx][]BitRange{
+				baseChunkIdx + sumsPerPhysPage - 1: {},
+				baseChunkIdx + chunkIdxBigJump:     {},
+			},
+			hits: []hit{
+				{PallocChunkPages - 1, PageBase(baseChunkIdx+sumsPerPhysPage-1, 0), 0},
+				{1, PageBase(baseChunkIdx+sumsPerPhysPage-1, PallocChunkPages-1), 0},
+				{1, PageBase(baseChunkIdx+chunkIdxBigJump, 0), 0},
+				{PallocChunkPages - 1, PageBase(baseChunkIdx+chunkIdxBigJump, 1), 0},
+				{1, 0, 0},
+			},
+			after: map[ChunkIdx][]BitRange{
+				baseChunkIdx + sumsPerPhysPage - 1: {{0, PallocChunkPages}},
+				baseChunkIdx + chunkIdxBigJump:     {{0, PallocChunkPages}},
+			},
+		}
+	}
 	for name, v := range tests {
 		v := v
 		t.Run(name, func(t *testing.T) {
@@ -589,6 +637,9 @@ func TestPageAllocAlloc(t *testing.T) {
 }
 
 func TestPageAllocExhaust(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	for _, npages := range []uintptr{1, 2, 3, 4, 5, 8, 16, 64, 1024, 1025, 2048, 2049} {
 		npages := npages
 		t.Run(fmt.Sprintf("%d", npages), func(t *testing.T) {
@@ -638,6 +689,9 @@ func TestPageAllocExhaust(t *testing.T) {
 }
 
 func TestPageAllocFree(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	tests := map[string]struct {
 		before map[ChunkIdx][]BitRange
 		after  map[ChunkIdx][]BitRange
@@ -867,6 +921,9 @@ func TestPageAllocFree(t *testing.T) {
 }
 
 func TestPageAllocAllocAndFree(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	type hit struct {
 		alloc  bool
 		npages uintptr
diff --git a/libgo/go/runtime/mpagecache_test.go b/libgo/go/runtime/mpagecache_test.go
index 6fdaa04..b8cc0bd 100644
--- a/libgo/go/runtime/mpagecache_test.go
+++ b/libgo/go/runtime/mpagecache_test.go
@@ -180,6 +180,9 @@ func TestPageCacheAlloc(t *testing.T) {
 }
 
 func TestPageCacheFlush(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	bits64ToBitRanges := func(bits uint64, base uint) []BitRange {
 		var ranges []BitRange
 		start, size := uint(0), uint(0)
@@ -254,6 +257,9 @@ func TestPageCacheFlush(t *testing.T) {
 }
 
 func TestPageAllocAllocToCache(t *testing.T) {
+	if GOOS == "openbsd" && testing.Short() {
+		t.Skip("skipping because virtual memory is limited; see #36210")
+	}
 	tests := map[string]struct {
 		before map[ChunkIdx][]BitRange
 		scav   map[ChunkIdx][]BitRange
diff --git a/libgo/go/runtime/mpallocbits.go b/libgo/go/runtime/mpallocbits.go
index 9d01ff8..a801134 100644
--- a/libgo/go/runtime/mpallocbits.go
+++ b/libgo/go/runtime/mpallocbits.go
@@ -202,17 +202,11 @@ func (b *pallocBits) summarize() pallocSum {
 // If find fails to find any free space, it returns an index of ^uint(0) and
 // the new searchIdx should be ignored.
 //
-// The returned searchIdx is always the index of the first free page found
-// in this bitmap during the search, except if npages == 1, in which
-// case it will be the index just after the first free page, because the
-// index returned as the first result is assumed to be allocated and so
-// represents a minor optimization for that case.
+// Note that if npages == 1, the two returned values will always be identical.
 func (b *pallocBits) find(npages uintptr, searchIdx uint) (uint, uint) {
 	if npages == 1 {
 		addr := b.find1(searchIdx)
-		// Return a searchIdx of addr + 1 since we assume addr will be
-		// allocated.
-		return addr, addr + 1
+		return addr, addr
 	} else if npages <= 64 {
 		return b.findSmallN(npages, searchIdx)
 	}
diff --git a/libgo/go/runtime/mranges.go b/libgo/go/runtime/mranges.go
index c14e5c7..b133851 100644
--- a/libgo/go/runtime/mranges.go
+++ b/libgo/go/runtime/mranges.go
@@ -29,6 +29,11 @@ func (a addrRange) size() uintptr {
 	return a.limit - a.base
 }
 
+// contains returns whether or not the range contains a given address.
+func (a addrRange) contains(addr uintptr) bool {
+	return addr >= a.base && addr < a.limit
+}
+
 // subtract takes the addrRange toPrune and cuts out any overlap with
 // from, then returns the new range. subtract assumes that a and b
 // either don't overlap at all, only overlap on one side, or are equal.
@@ -87,6 +92,15 @@ func (a *addrRanges) findSucc(base uintptr) int {
 	return len(a.ranges)
 }
 
+// contains returns true if a covers the address addr.
+func (a *addrRanges) contains(addr uintptr) bool {
+	i := a.findSucc(addr)
+	if i == 0 {
+		return false
+	}
+	return a.ranges[i-1].contains(addr)
+}
+
 // add inserts a new address range to a.
 //
 // r must not overlap with any address range in a.
diff --git a/libgo/go/runtime/preempt_nonwindows.go b/libgo/go/runtime/preempt_nonwindows.go
new file mode 100644
index 0000000..3066a152
--- /dev/null
+++ b/libgo/go/runtime/preempt_nonwindows.go
@@ -0,0 +1,13 @@
+// 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.
+
+// +build !windows
+
+package runtime
+
+//go:nosplit
+func osPreemptExtEnter(mp *m) {}
+
+//go:nosplit
+func osPreemptExtExit(mp *m) {}
diff --git a/libgo/go/runtime/proc.go b/libgo/go/runtime/proc.go
index e3f934a..f75cacf 100644
--- a/libgo/go/runtime/proc.go
+++ b/libgo/go/runtime/proc.go
@@ -841,23 +841,8 @@ func casGFromPreempted(gp *g, old, new uint32) bool {
 // goroutines.
 func stopTheWorld(reason string) {
 	semacquire(&worldsema)
-	gp := getg()
-	gp.m.preemptoff = reason
-	systemstack(func() {
-		// Mark the goroutine which called stopTheWorld preemptible so its
-		// stack may be scanned.
-		// This lets a mark worker scan us while we try to stop the world
-		// since otherwise we could get in a mutual preemption deadlock.
-		// We must not modify anything on the G stack because a stack shrink
-		// may occur. A stack shrink is otherwise OK though because in order
-		// to return from this function (and to leave the system stack) we
-		// must have preempted all goroutines, including any attempting
-		// to scan our stack, in which case, any stack shrinking will
-		// have already completed by the time we exit.
-		casgstatus(gp, _Grunning, _Gwaiting)
-		stopTheWorldWithSema()
-		casgstatus(gp, _Gwaiting, _Grunning)
-	})
+	getg().m.preemptoff = reason
+	systemstack(stopTheWorldWithSema)
 }
 
 // startTheWorld undoes the effects of stopTheWorld.
@@ -869,31 +854,10 @@ func startTheWorld() {
 	getg().m.preemptoff = ""
 }
 
-// stopTheWorldGC has the same effect as stopTheWorld, but blocks
-// until the GC is not running. It also blocks a GC from starting
-// until startTheWorldGC is called.
-func stopTheWorldGC(reason string) {
-	semacquire(&gcsema)
-	stopTheWorld(reason)
-}
-
-// startTheWorldGC undoes the effects of stopTheWorldGC.
-func startTheWorldGC() {
-	startTheWorld()
-	semrelease(&gcsema)
-}
-
-// Holding worldsema grants an M the right to try to stop the world.
+// Holding worldsema grants an M the right to try to stop the world
+// and prevents gomaxprocs from changing concurrently.
 var worldsema uint32 = 1
 
-// Holding gcsema grants the M the right to block a GC, and blocks
-// until the current GC is done. In particular, it prevents gomaxprocs
-// from changing concurrently.
-//
-// TODO(mknyszek): Once gomaxprocs and the execution tracer can handle
-// being changed/enabled during a GC, remove this.
-var gcsema uint32 = 1
-
 // stopTheWorldWithSema is the core implementation of stopTheWorld.
 // The caller is responsible for acquiring worldsema and disabling
 // preemption first and then should stopTheWorldWithSema on the system
@@ -2577,6 +2541,27 @@ func dropg() {
 // We pass now in and out to avoid extra calls of nanotime.
 //go:yeswritebarrierrec
 func checkTimers(pp *p, now int64) (rnow, pollUntil int64, ran bool) {
+	// If there are no timers to adjust, and the first timer on
+	// the heap is not yet ready to run, then there is nothing to do.
+	if atomic.Load(&pp.adjustTimers) == 0 {
+		next := int64(atomic.Load64(&pp.timer0When))
+		if next == 0 {
+			return now, 0, false
+		}
+		if now == 0 {
+			now = nanotime()
+		}
+		if now < next {
+			// Next timer is not ready to run.
+			// But keep going if we would clear deleted timers.
+			// This corresponds to the condition below where
+			// we decide whether to call clearDeletedTimers.
+			if pp != getg().m.p.ptr() || int(atomic.Load(&pp.deletedTimers)) <= int(atomic.Load(&pp.numTimers)/4) {
+				return now, next, false
+			}
+		}
+	}
+
 	lock(&pp.timersLock)
 
 	adjusttimers(pp)
@@ -2599,6 +2584,13 @@ func checkTimers(pp *p, now int64) (rnow, pollUntil int64, ran bool) {
 		}
 	}
 
+	// If this is the local P, and there are a lot of deleted timers,
+	// clear them out. We only do this for the local P to reduce
+	// lock contention on timersLock.
+	if pp == getg().m.p.ptr() && int(atomic.Load(&pp.deletedTimers)) > len(pp.timers)/4 {
+		clearDeletedTimers(pp)
+	}
+
 	unlock(&pp.timersLock)
 
 	return rnow, pollUntil, ran
@@ -2723,7 +2715,7 @@ func preemptPark(gp *g) {
 }
 
 // goyield is like Gosched, but it:
-// - does not emit a GoSched trace event
+// - emits a GoPreempt trace event instead of a GoSched trace event
 // - puts the current G on the runq of the current P instead of the globrunq
 func goyield() {
 	checkTimeouts()
@@ -2731,6 +2723,9 @@ func goyield() {
 }
 
 func goyield_m(gp *g) {
+	if trace.enabled {
+		traceGoPreempt()
+	}
 	pp := gp.m.p.ptr()
 	casgstatus(gp, _Grunning, _Grunnable)
 	dropg()
@@ -3816,7 +3811,10 @@ func (pp *p) destroy() {
 		lock(&pp.timersLock)
 		moveTimers(plocal, pp.timers)
 		pp.timers = nil
+		pp.numTimers = 0
 		pp.adjustTimers = 0
+		pp.deletedTimers = 0
+		atomic.Store64(&pp.timer0When, 0)
 		unlock(&pp.timersLock)
 		unlock(&plocal.timersLock)
 	}
@@ -4122,23 +4120,26 @@ func checkdead() {
 	}
 
 	// Maybe jump time forward for playground.
-	_p_ := timejump()
-	if _p_ != nil {
-		for pp := &sched.pidle; *pp != 0; pp = &(*pp).ptr().link {
-			if (*pp).ptr() == _p_ {
-				*pp = _p_.link
-				break
+	if faketime != 0 {
+		when, _p_ := timeSleepUntil()
+		if _p_ != nil {
+			faketime = when
+			for pp := &sched.pidle; *pp != 0; pp = &(*pp).ptr().link {
+				if (*pp).ptr() == _p_ {
+					*pp = _p_.link
+					break
+				}
 			}
+			mp := mget()
+			if mp == nil {
+				// There should always be a free M since
+				// nothing is running.
+				throw("checkdead: no m for timer")
+			}
+			mp.nextp.set(_p_)
+			notewakeup(&mp.park)
+			return
 		}
-		mp := mget()
-		if mp == nil {
-			// There should always be a free M since
-			// nothing is running.
-			throw("checkdead: no m for timer")
-		}
-		mp.nextp.set(_p_)
-		notewakeup(&mp.park)
-		return
 	}
 
 	// There are no goroutines running, so we can look at the P's.
@@ -4183,7 +4184,7 @@ func sysmon() {
 		}
 		usleep(delay)
 		now := nanotime()
-		next := timeSleepUntil()
+		next, _ := timeSleepUntil()
 		if debug.schedtrace <= 0 && (sched.gcwaiting != 0 || atomic.Load(&sched.npidle) == uint32(gomaxprocs)) {
 			lock(&sched.lock)
 			if atomic.Load(&sched.gcwaiting) != 0 || atomic.Load(&sched.npidle) == uint32(gomaxprocs) {
@@ -4205,7 +4206,7 @@ func sysmon() {
 						osRelax(false)
 					}
 					now = nanotime()
-					next = timeSleepUntil()
+					next, _ = timeSleepUntil()
 					lock(&sched.lock)
 					atomic.Store(&sched.sysmonwait, 0)
 					noteclear(&sched.sysmonnote)
diff --git a/libgo/go/runtime/runtime1.go b/libgo/go/runtime/runtime1.go
index 60aa90f..6edf7a5 100644
--- a/libgo/go/runtime/runtime1.go
+++ b/libgo/go/runtime/runtime1.go
@@ -323,6 +323,7 @@ var debug struct {
 	madvdontneed       int32 // for Linux; issue 28466
 	sbrk               int32
 	scavenge           int32
+	scavtrace          int32
 	scheddetail        int32
 	schedtrace         int32
 	tracebackancestors int32
@@ -343,6 +344,7 @@ var dbgvars = []dbgVar{
 	{"madvdontneed", &debug.madvdontneed},
 	{"sbrk", &debug.sbrk},
 	{"scavenge", &debug.scavenge},
+	{"scavtrace", &debug.scavtrace},
 	{"scheddetail", &debug.scheddetail},
 	{"schedtrace", &debug.schedtrace},
 	{"tracebackancestors", &debug.tracebackancestors},
diff --git a/libgo/go/runtime/runtime2.go b/libgo/go/runtime/runtime2.go
index d50f82a..f5bfc08 100644
--- a/libgo/go/runtime/runtime2.go
+++ b/libgo/go/runtime/runtime2.go
@@ -677,6 +677,11 @@ type p struct {
 
 	_ uint32 // Alignment for atomic fields below
 
+	// The when field of the first entry on the timer heap.
+	// This is updated using atomic functions.
+	// This is 0 if the timer heap is empty.
+	timer0When uint64
+
 	// Per-P GC state
 	gcAssistTime         int64    // Nanoseconds in assistAlloc
 	gcFractionalMarkTime int64    // Nanoseconds in fractional mark worker (atomic)
@@ -708,12 +713,20 @@ type p struct {
 	// Must hold timersLock to access.
 	timers []*timer
 
+	// Number of timers in P's heap.
+	// Modified using atomic instructions.
+	numTimers uint32
+
 	// Number of timerModifiedEarlier timers on P's heap.
 	// This should only be modified while holding timersLock,
 	// or while the timer status is in a transient state
 	// such as timerModifying.
 	adjustTimers uint32
 
+	// Number of timerDeleted timers in P's heap.
+	// Modified using atomic instructions.
+	deletedTimers uint32
+
 	// Race context used while executing timer functions.
 	// Not for gccgo: timerRaceCtx uintptr
 
diff --git a/libgo/go/runtime/sema.go b/libgo/go/runtime/sema.go
index fb16796..b6fab6d 100644
--- a/libgo/go/runtime/sema.go
+++ b/libgo/go/runtime/sema.go
@@ -199,9 +199,9 @@ func semrelease1(addr *uint32, handoff bool, skipframes int) {
 			// the waiter G immediately.
 			// Note that waiter inherits our time slice: this is desirable
 			// to avoid having a highly contended semaphore hog the P
-			// indefinitely. goyield is like Gosched, but it does not emit a
-			// GoSched trace event and, more importantly, puts the current G
-			// on the local runq instead of the global one.
+			// indefinitely. goyield is like Gosched, but it emits a
+			// "preempted" trace event instead and, more importantly, puts
+			// the current G on the local runq instead of the global one.
 			// We only do this in the starving regime (handoff=true), as in
 			// the non-starving case it is possible for a different waiter
 			// to acquire the semaphore while we are yielding/scheduling,
diff --git a/libgo/go/runtime/signal_unix.go b/libgo/go/runtime/signal_unix.go
index 29f9443..150345f 100644
--- a/libgo/go/runtime/signal_unix.go
+++ b/libgo/go/runtime/signal_unix.go
@@ -399,6 +399,16 @@ func sigtrampgo(sig uint32, info *_siginfo_t, ctx unsafe.Pointer) {
 			sigprofNonGo(pc)
 			return
 		}
+		if sig == sigPreempt && preemptMSupported && debug.asyncpreemptoff == 0 {
+			// This is probably a signal from preemptM sent
+			// while executing Go code but received while
+			// executing non-Go code.
+			// We got past sigfwdgo, so we know that there is
+			// no non-Go signal handler for sigPreempt.
+			// The default behavior for sigPreempt is to ignore
+			// the signal, so badsignal will be a no-op anyway.
+			return
+		}
 		badsignal(uintptr(sig), &c)
 		return
 	}
diff --git a/libgo/go/runtime/testdata/testprog/checkptr.go b/libgo/go/runtime/testdata/testprog/checkptr.go
new file mode 100644
index 0000000..177db38
--- /dev/null
+++ b/libgo/go/runtime/testdata/testprog/checkptr.go
@@ -0,0 +1,36 @@
+// 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"
+
+func init() {
+	register("CheckPtrAlignment", CheckPtrAlignment)
+	register("CheckPtrArithmetic", CheckPtrArithmetic)
+	register("CheckPtrSize", CheckPtrSize)
+	register("CheckPtrSmall", CheckPtrSmall)
+}
+
+func CheckPtrAlignment() {
+	var x [2]int64
+	p := unsafe.Pointer(&x[0])
+	sink2 = (*int64)(unsafe.Pointer(uintptr(p) + 1))
+}
+
+func CheckPtrArithmetic() {
+	var x int
+	i := uintptr(unsafe.Pointer(&x))
+	sink2 = (*int)(unsafe.Pointer(i))
+}
+
+func CheckPtrSize() {
+	p := new(int64)
+	sink2 = p
+	sink2 = (*[100]int64)(unsafe.Pointer(p))
+}
+
+func CheckPtrSmall() {
+	sink2 = unsafe.Pointer(uintptr(1))
+}
diff --git a/libgo/go/runtime/time.go b/libgo/go/runtime/time.go
index ded68ed..d0dd3a4 100644
--- a/libgo/go/runtime/time.go
+++ b/libgo/go/runtime/time.go
@@ -73,14 +73,15 @@ type timer struct {
 //   timerNoStatus   -> timerWaiting
 //   anything else   -> panic: invalid value
 // deltimer:
-//   timerWaiting    -> timerDeleted
-//   timerModifiedXX -> timerDeleted
-//   timerNoStatus   -> do nothing
-//   timerDeleted    -> do nothing
-//   timerRemoving   -> do nothing
-//   timerRemoved    -> do nothing
-//   timerRunning    -> wait until status changes
-//   timerMoving     -> wait until status changes
+//   timerWaiting         -> timerDeleted
+//   timerModifiedEarlier -> timerModifying -> timerDeleted
+//   timerModifiedLater   -> timerDeleted
+//   timerNoStatus        -> do nothing
+//   timerDeleted         -> do nothing
+//   timerRemoving        -> do nothing
+//   timerRemoved         -> do nothing
+//   timerRunning         -> wait until status changes
+//   timerMoving          -> wait until status changes
 //   timerModifying  -> panic: concurrent deltimer/modtimer calls
 // modtimer:
 //   timerWaiting    -> timerModifying -> timerModifiedXX
@@ -168,6 +169,10 @@ const (
 // maxWhen is the maximum value for timer's when field.
 const maxWhen = 1<<63 - 1
 
+// verifyTimers can be set to true to add debugging checks that the
+// timer heaps are valid.
+const verifyTimers = false
+
 // Package time APIs.
 // Godoc uses the comments in package time, not these.
 
@@ -283,7 +288,12 @@ func doaddtimer(pp *p, t *timer) bool {
 	t.pp.set(pp)
 	i := len(pp.timers)
 	pp.timers = append(pp.timers, t)
-	return siftupTimer(pp.timers, i)
+	ok := siftupTimer(pp.timers, i)
+	if t == pp.timers[0] {
+		atomic.Store64(&pp.timer0When, uint64(t.when))
+	}
+	atomic.Xadd(&pp.numTimers, 1)
+	return ok
 }
 
 // deltimer deletes the timer t. It may be on some other P, so we can't
@@ -294,7 +304,9 @@ func deltimer(t *timer) bool {
 	for {
 		switch s := atomic.Load(&t.status); s {
 		case timerWaiting, timerModifiedLater:
+			tpp := t.pp.ptr()
 			if atomic.Cas(&t.status, s, timerDeleted) {
+				atomic.Xadd(&tpp.deletedTimers, 1)
 				// Timer was not yet run.
 				return true
 			}
@@ -305,6 +317,7 @@ func deltimer(t *timer) bool {
 				if !atomic.Cas(&t.status, timerModifying, timerDeleted) {
 					badTimer()
 				}
+				atomic.Xadd(&tpp.deletedTimers, 1)
 				// Timer was not yet run.
 				return true
 			}
@@ -355,6 +368,10 @@ func dodeltimer(pp *p, i int) bool {
 			ok = false
 		}
 	}
+	if i == 0 {
+		updateTimer0When(pp)
+	}
+	atomic.Xadd(&pp.numTimers, -1)
 	return ok
 }
 
@@ -378,6 +395,8 @@ func dodeltimer0(pp *p) bool {
 	if last > 0 {
 		ok = siftdownTimer(pp.timers, 0)
 	}
+	updateTimer0When(pp)
+	atomic.Xadd(&pp.numTimers, -1)
 	return ok
 }
 
@@ -485,6 +504,7 @@ func resettimer(t *timer, when int64) {
 				return
 			}
 		case timerDeleted:
+			tpp := t.pp.ptr()
 			if atomic.Cas(&t.status, s, timerModifying) {
 				t.nextwhen = when
 				newStatus := uint32(timerModifiedLater)
@@ -495,6 +515,7 @@ func resettimer(t *timer, when int64) {
 				if !atomic.Cas(&t.status, timerModifying, newStatus) {
 					badTimer()
 				}
+				atomic.Xadd(&tpp.deletedTimers, -1)
 				if newStatus == timerModifiedEarlier {
 					wakeNetPoller(when)
 				}
@@ -542,6 +563,7 @@ func cleantimers(pp *p) bool {
 			if !atomic.Cas(&t.status, timerRemoving, timerRemoved) {
 				return false
 			}
+			atomic.Xadd(&pp.deletedTimers, -1)
 		case timerModifiedEarlier, timerModifiedLater:
 			if !atomic.Cas(&t.status, s, timerMoving) {
 				continue
@@ -630,9 +652,13 @@ func adjusttimers(pp *p) {
 		return
 	}
 	if atomic.Load(&pp.adjustTimers) == 0 {
+		if verifyTimers {
+			verifyTimerHeap(pp)
+		}
 		return
 	}
 	var moved []*timer
+loop:
 	for i := 0; i < len(pp.timers); i++ {
 		t := pp.timers[i]
 		if t.pp.ptr() != pp {
@@ -647,6 +673,7 @@ func adjusttimers(pp *p) {
 				if !atomic.Cas(&t.status, timerRemoving, timerRemoved) {
 					badTimer()
 				}
+				atomic.Xadd(&pp.deletedTimers, -1)
 				// Look at this heap position again.
 				i--
 			}
@@ -664,10 +691,11 @@ func adjusttimers(pp *p) {
 				moved = append(moved, t)
 				if s == timerModifiedEarlier {
 					if n := atomic.Xadd(&pp.adjustTimers, -1); int32(n) <= 0 {
-						addAdjustedTimers(pp, moved)
-						return
+						break loop
 					}
 				}
+				// Look at this heap position again.
+				i--
 			}
 		case timerNoStatus, timerRunning, timerRemoving, timerRemoved, timerMoving:
 			badTimer()
@@ -685,6 +713,10 @@ func adjusttimers(pp *p) {
 	if len(moved) > 0 {
 		addAdjustedTimers(pp, moved)
 	}
+
+	if verifyTimers {
+		verifyTimerHeap(pp)
+	}
 }
 
 // addAdjustedTimers adds any timers we adjusted in adjusttimers
@@ -708,17 +740,11 @@ func addAdjustedTimers(pp *p, moved []*timer) {
 // The netpoller M will wake up and adjust timers before sleeping again.
 //go:nowritebarrierrec
 func nobarrierWakeTime(pp *p) int64 {
-	lock(&pp.timersLock)
-	ret := int64(0)
-	if len(pp.timers) > 0 {
-		if atomic.Load(&pp.adjustTimers) > 0 {
-			ret = nanotime()
-		} else {
-			ret = pp.timers[0].when
-		}
+	if atomic.Load(&pp.adjustTimers) > 0 {
+		return nanotime()
+	} else {
+		return int64(atomic.Load64(&pp.timer0When))
 	}
-	unlock(&pp.timersLock)
-	return ret
 }
 
 // runtimer examines the first timer in timers. If it is ready based on now,
@@ -759,6 +785,7 @@ func runtimer(pp *p, now int64) int64 {
 			if !atomic.Cas(&t.status, timerRemoving, timerRemoved) {
 				badTimer()
 			}
+			atomic.Xadd(&pp.deletedTimers, -1)
 			if len(pp.timers) == 0 {
 				return -1
 			}
@@ -817,6 +844,7 @@ func runOneTimer(pp *p, t *timer, now int64) {
 		if !atomic.Cas(&t.status, timerRunning, timerWaiting) {
 			badTimer()
 		}
+		updateTimer0When(pp)
 	} else {
 		// Remove from heap.
 		if !dodeltimer0(pp) {
@@ -834,69 +862,131 @@ func runOneTimer(pp *p, t *timer, now int64) {
 	lock(&pp.timersLock)
 }
 
-func timejump() *p {
-	if faketime == 0 {
-		return nil
-	}
-
-	// Nothing is running, so we can look at all the P's.
-	// Determine a timer bucket with minimum when.
-	var (
-		minT    *timer
-		minWhen int64
-		minP    *p
-	)
-	for _, pp := range allp {
-		if pp.status != _Pidle && pp.status != _Pdead {
-			throw("non-idle P in timejump")
-		}
-		if len(pp.timers) == 0 {
-			continue
-		}
-		c := pp.adjustTimers
-		for _, t := range pp.timers {
+// clearDeletedTimers removes all deleted timers from the P's timer heap.
+// This is used to avoid clogging up the heap if the program
+// starts a lot of long-running timers and then stops them.
+// For example, this can happen via context.WithTimeout.
+//
+// This is the only function that walks through the entire timer heap,
+// other than moveTimers which only runs when the world is stopped.
+//
+// The caller must have locked the timers for pp.
+func clearDeletedTimers(pp *p) {
+	cdel := int32(0)
+	cearlier := int32(0)
+	to := 0
+	changedHeap := false
+	timers := pp.timers
+nextTimer:
+	for _, t := range timers {
+		for {
 			switch s := atomic.Load(&t.status); s {
 			case timerWaiting:
-				if minT == nil || t.when < minWhen {
-					minT = t
-					minWhen = t.when
-					minP = pp
+				if changedHeap {
+					timers[to] = t
+					siftupTimer(timers, to)
 				}
+				to++
+				continue nextTimer
 			case timerModifiedEarlier, timerModifiedLater:
-				if minT == nil || t.nextwhen < minWhen {
-					minT = t
-					minWhen = t.nextwhen
-					minP = pp
+				if atomic.Cas(&t.status, s, timerMoving) {
+					t.when = t.nextwhen
+					timers[to] = t
+					siftupTimer(timers, to)
+					to++
+					changedHeap = true
+					if !atomic.Cas(&t.status, timerMoving, timerWaiting) {
+						badTimer()
+					}
+					if s == timerModifiedEarlier {
+						cearlier++
+					}
+					continue nextTimer
 				}
-				if s == timerModifiedEarlier {
-					c--
+			case timerDeleted:
+				if atomic.Cas(&t.status, s, timerRemoving) {
+					t.pp = 0
+					cdel++
+					if !atomic.Cas(&t.status, timerRemoving, timerRemoved) {
+						badTimer()
+					}
+					changedHeap = true
+					continue nextTimer
 				}
-			case timerRunning, timerModifying, timerMoving:
+			case timerModifying:
+				// Loop until modification complete.
+				osyield()
+			case timerNoStatus, timerRemoved:
+				// We should not see these status values in a timer heap.
+				badTimer()
+			case timerRunning, timerRemoving, timerMoving:
+				// Some other P thinks it owns this timer,
+				// which should not happen.
+				badTimer()
+			default:
 				badTimer()
-			}
-			// The timers are sorted, so we only have to check
-			// the first timer for each P, unless there are
-			// some timerModifiedEarlier timers. The number
-			// of timerModifiedEarlier timers is in the adjustTimers
-			// field, used to initialize c, above.
-			if c == 0 {
-				break
 			}
 		}
 	}
 
-	if minT == nil || minWhen <= faketime {
-		return nil
+	// Set remaining slots in timers slice to nil,
+	// so that the timer values can be garbage collected.
+	for i := to; i < len(timers); i++ {
+		timers[i] = nil
+	}
+
+	atomic.Xadd(&pp.deletedTimers, -cdel)
+	atomic.Xadd(&pp.numTimers, -cdel)
+	atomic.Xadd(&pp.adjustTimers, -cearlier)
+
+	timers = timers[:to]
+	pp.timers = timers
+	updateTimer0When(pp)
+
+	if verifyTimers {
+		verifyTimerHeap(pp)
+	}
+}
+
+// verifyTimerHeap verifies that the timer heap is in a valid state.
+// This is only for debugging, and is only called if verifyTimers is true.
+// The caller must have locked the timers.
+func verifyTimerHeap(pp *p) {
+	for i, t := range pp.timers {
+		if i == 0 {
+			// First timer has no parent.
+			continue
+		}
+
+		// The heap is 4-ary. See siftupTimer and siftdownTimer.
+		p := (i - 1) / 4
+		if t.when < pp.timers[p].when {
+			print("bad timer heap at ", i, ": ", p, ": ", pp.timers[p].when, ", ", i, ": ", t.when, "\n")
+			throw("bad timer heap")
+		}
+	}
+	if numTimers := int(atomic.Load(&pp.numTimers)); len(pp.timers) != numTimers {
+		println("timer heap len", len(pp.timers), "!= numTimers", numTimers)
+		throw("bad timer heap len")
 	}
+}
 
-	faketime = minWhen
-	return minP
+// updateTimer0When sets the P's timer0When field.
+// The caller must have locked the timers for pp.
+func updateTimer0When(pp *p) {
+	if len(pp.timers) == 0 {
+		atomic.Store64(&pp.timer0When, 0)
+	} else {
+		atomic.Store64(&pp.timer0When, uint64(pp.timers[0].when))
+	}
 }
 
-// timeSleepUntil returns the time when the next timer should fire.
-// This is only called by sysmon.
-func timeSleepUntil() int64 {
+// timeSleepUntil returns the time when the next timer should fire,
+// and the P that holds the timer heap that that timer is on.
+// This is only called by sysmon and checkdead.
+func timeSleepUntil() (int64, *p) {
 	next := int64(maxWhen)
+	var pret *p
 
 	// Prevent allp slice changes. This is like retake.
 	lock(&allpLock)
@@ -907,8 +997,17 @@ func timeSleepUntil() int64 {
 			continue
 		}
 
-		lock(&pp.timersLock)
 		c := atomic.Load(&pp.adjustTimers)
+		if c == 0 {
+			w := int64(atomic.Load64(&pp.timer0When))
+			if w != 0 && w < next {
+				next = w
+				pret = pp
+			}
+			continue
+		}
+
+		lock(&pp.timersLock)
 		for _, t := range pp.timers {
 			switch s := atomic.Load(&t.status); s {
 			case timerWaiting:
@@ -943,7 +1042,7 @@ func timeSleepUntil() int64 {
 	}
 	unlock(&allpLock)
 
-	return next
+	return next, pret
 }
 
 // Heap maintenance algorithms.
diff --git a/libgo/go/runtime/trace.go b/libgo/go/runtime/trace.go
index 81ff0ca..358674b 100644
--- a/libgo/go/runtime/trace.go
+++ b/libgo/go/runtime/trace.go
@@ -181,12 +181,9 @@ func traceBufPtrOf(b *traceBuf) traceBufPtr {
 // Most clients should use the runtime/trace package or the testing package's
 // -test.trace flag instead of calling StartTrace directly.
 func StartTrace() error {
-	// Stop the world so that we can take a consistent snapshot
+	// Stop the world, so that we can take a consistent snapshot
 	// of all goroutines at the beginning of the trace.
-	// Do not stop the world during GC so we ensure we always see
-	// a consistent view of GC-related events (e.g. a start is always
-	// paired with an end).
-	stopTheWorldGC("start tracing")
+	stopTheWorld("start tracing")
 
 	// We are in stop-the-world, but syscalls can finish and write to trace concurrently.
 	// Exitsyscall could check trace.enabled long before and then suddenly wake up
@@ -197,7 +194,7 @@ func StartTrace() error {
 
 	if trace.enabled || trace.shutdown {
 		unlock(&trace.bufLock)
-		startTheWorldGC()
+		startTheWorld()
 		return errorString("tracing is already enabled")
 	}
 
@@ -268,7 +265,7 @@ func StartTrace() error {
 
 	unlock(&trace.bufLock)
 
-	startTheWorldGC()
+	startTheWorld()
 	return nil
 }
 
@@ -277,14 +274,14 @@ func StartTrace() error {
 func StopTrace() {
 	// Stop the world so that we can collect the trace buffers from all p's below,
 	// and also to avoid races with traceEvent.
-	stopTheWorldGC("stop tracing")
+	stopTheWorld("stop tracing")
 
 	// See the comment in StartTrace.
 	lock(&trace.bufLock)
 
 	if !trace.enabled {
 		unlock(&trace.bufLock)
-		startTheWorldGC()
+		startTheWorld()
 		return
 	}
 
@@ -321,7 +318,7 @@ func StopTrace() {
 	trace.shutdown = true
 	unlock(&trace.bufLock)
 
-	startTheWorldGC()
+	startTheWorld()
 
 	// The world is started but we've set trace.shutdown, so new tracing can't start.
 	// Wait for the trace reader to flush pending buffers and stop.
diff --git a/libgo/go/runtime/trace/trace_stack_test.go b/libgo/go/runtime/trace/trace_stack_test.go
index e3608c6..62c06e6 100644
--- a/libgo/go/runtime/trace/trace_stack_test.go
+++ b/libgo/go/runtime/trace/trace_stack_test.go
@@ -233,7 +233,6 @@ func TestTraceSymbolize(t *testing.T) {
 		}},
 		{trace.EvGomaxprocs, []frame{
 			{"runtime.startTheWorld", 0}, // this is when the current gomaxprocs is logged.
-			{"runtime.startTheWorldGC", 0},
 			{"runtime.GOMAXPROCS", 0},
 			{"runtime/trace_test.TestTraceSymbolize", 0},
 			{"testing.tRunner", 0},
diff --git a/libgo/go/runtime/utf8.go b/libgo/go/runtime/utf8.go
index 6590472..a404a33 100644
--- a/libgo/go/runtime/utf8.go
+++ b/libgo/go/runtime/utf8.go
@@ -13,7 +13,7 @@ import _ "unsafe" // For go:linkname.
 // Numbers fundamental to the encoding.
 const (
 	runeError = '\uFFFD'     // the "error" Rune or "Unicode replacement character"
-	runeSelf  = 0x80         // characters below Runeself are represented as themselves in a single byte.
+	runeSelf  = 0x80         // characters below runeSelf are represented as themselves in a single byte.
 	maxRune   = '\U0010FFFF' // Maximum valid Unicode code point.
 )