runtime: increase profiling stack depth to 128

The current stack depth limit for alloc, mutex, block, threadcreate and
goroutine profiles of 32 frequently leads to truncated stack traces in
production applications. Increase the limit to 128 which is the same
size used by the execution tracer.

Create internal/profilerecord to define variants of the runtime's
StackRecord, MemProfileRecord and BlockProfileRecord types that can hold
arbitrarily big stack traces. Implement internal profiling APIs based on
these new types and use them for creating protobuf profiles and to act
as shims for the public profiling APIs using the old types.

This will lead to an increase in memory usage for applications that
use the impacted profile types and have stack traces exceeding the
current limit of 32. Those applications will also experience a slight
increase in CPU usage, but this will hopefully soon be mitigated via CL
540476 and 533258 which introduce frame pointer unwinding for the
relevant profile types.

For #43669.

Change-Id: Ie53762e65d0f6295f5d4c7d3c87172d5a052164e
Reviewed-on: https://go-review.googlesource.com/c/go/+/572396
LUCI-TryBot-Result: Go LUCI <[email protected]>
Reviewed-by: Austin Clements <[email protected]>
Reviewed-by: Cherry Mui <[email protected]>

doc/next/6-stdlib/99-minor/runtime/pprof/43669.md

@@ -0,0 +1,2 @@
+The maximum stack depth for alloc, mutex, block, threadcreate and goroutine
+profiles has been raised from 32 to 128 frames.

src/cmd/internal/objabi/pkgspecial.go

@@ -58,6 +58,7 @@ var runtimePkgs = []string{
 	"internal/godebugs",
 	"internal/goexperiment",
 	"internal/goos",
+	"internal/profilerecord",
 	"internal/stringslite",
 }
 

src/go/build/deps_test.go

@@ -45,7 +45,7 @@ var depsRules = `
 	  internal/goarch, internal/godebugs,
 	  internal/goexperiment, internal/goos, internal/byteorder,
 	  internal/goversion, internal/nettrace, internal/platform,
-	  internal/trace/traceviewer/format,
+	  internal/profilerecord, internal/trace/traceviewer/format,
 	  log/internal,
 	  unicode/utf8, unicode/utf16, unicode,
 	  unsafe;
@@ -65,7 +65,8 @@ var depsRules = `
 	internal/goarch,
 	internal/godebugs,
 	internal/goexperiment,
-	internal/goos
+	internal/goos,
+	internal/profilerecord
 	< internal/bytealg
 	< internal/stringslite
 	< internal/itoa

src/internal/profilerecord/profilerecord.go

@@ -0,0 +1,28 @@
+// Copyright 2024 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 profilerecord holds internal types used to represent profiling
+// records with deep stack traces.
+//
+// TODO: Consider moving this to internal/runtime, see golang.org/issue/65355.
+package profilerecord
+
+type StackRecord struct {
+	Stack []uintptr
+}
+
+type MemProfileRecord struct {
+	AllocBytes, FreeBytes     int64
+	AllocObjects, FreeObjects int64
+	Stack                     []uintptr
+}
+
+func (r *MemProfileRecord) InUseBytes() int64   { return r.AllocBytes - r.FreeBytes }
+func (r *MemProfileRecord) InUseObjects() int64 { return r.AllocObjects - r.FreeObjects }
+
+type BlockProfileRecord struct {
+	Count  int64
+	Cycles int64
+	Stack  []uintptr
+}

src/runtime/cpuprof.go

@@ -209,8 +209,8 @@ func CPUProfile() []byte {
 	panic("CPUProfile no longer available")
 }
 
-//go:linkname runtime_pprof_runtime_cyclesPerSecond runtime/pprof.runtime_cyclesPerSecond
-func runtime_pprof_runtime_cyclesPerSecond() int64 {
+//go:linkname pprof_cyclesPerSecond
+func pprof_cyclesPerSecond() int64 {
 	return ticksPerSecond()
 }