Import chromium-118.0.5993.65

src/.clang-format

@@ -0,0 +1,41 @@
+# Defines the Chromium style for automatic reformatting.
+# http://clang.llvm.org/docs/ClangFormatStyleOptions.html
+BasedOnStyle: Chromium
+# This defaults to 'Auto'. Explicitly set it for a while, so that
+# 'vector<vector<int> >' in existing files gets formatted to
+# 'vector<vector<int>>'. ('Auto' means that clang-format will only use
+# 'int>>' if the file already contains at least one such instance.)
+Standard: Cpp11
+
+# TODO(crbug.com/1392808): Remove when InsertBraces has been upstreamed into
+# the Chromium style (is implied by BasedOnStyle: Chromium).
+InsertBraces: true
+InsertNewlineAtEOF: true
+
+# Make sure code like:
+# IPC_BEGIN_MESSAGE_MAP()
+#   IPC_MESSAGE_HANDLER(WidgetHostViewHost_Update, OnUpdate)
+# IPC_END_MESSAGE_MAP()
+# gets correctly indented.
+MacroBlockBegin: "^\
+BEGIN_MSG_MAP|\
+BEGIN_MSG_MAP_EX|\
+BEGIN_SAFE_MSG_MAP_EX|\
+CR_BEGIN_MSG_MAP_EX|\
+IPC_BEGIN_MESSAGE_MAP|\
+IPC_BEGIN_MESSAGE_MAP_WITH_PARAM|\
+IPC_PROTOBUF_MESSAGE_TRAITS_BEGIN|\
+IPC_STRUCT_BEGIN|\
+IPC_STRUCT_BEGIN_WITH_PARENT|\
+IPC_STRUCT_TRAITS_BEGIN|\
+POLPARAMS_BEGIN|\
+PPAPI_BEGIN_MESSAGE_MAP$"
+MacroBlockEnd: "^\
+CR_END_MSG_MAP|\
+END_MSG_MAP|\
+IPC_END_MESSAGE_MAP|\
+IPC_PROTOBUF_MESSAGE_TRAITS_END|\
+IPC_STRUCT_END|\
+IPC_STRUCT_TRAITS_END|\
+POLPARAMS_END|\
+PPAPI_END_MESSAGE_MAP$"

src/.gitattributes

@@ -0,0 +1,58 @@
+# Stop Windows python license check presubmit errors by forcing LF checkout.
+*.py  text eol=lf
+
+# Force LF checkout of the pins files to avoid transport_security_state_generator errors.
+/net/http/*.pins  text eol=lf
+
+# Force LF checkout for all source files
+*.bin binary
+*.c text eol=lf
+*.cc text eol=lf
+*.cpp text eol=lf
+*.csv text eol=lf
+*.grd   text eol=lf
+*.grdp   text eol=lf
+*.gn text eol=lf
+*.gni text eol=lf
+*.h text eol=lf
+*.html   text eol=lf
+*.idl text eol=lf
+*.in   text eol=lf
+*.inc   text eol=lf
+*.java   text eol=lf
+*.js   text eol=lf
+*.json   text eol=lf
+*.json5 text eol=lf
+*.md text eol=lf
+*.mm text eol=lf
+*.mojom text eol=lf
+*.pdf -diff
+*.proto text eol=lf
+*.rs text eol=lf
+*.sh text eol=lf
+*.sql   text eol=lf
+*.toml  text eol=lf
+*.txt   text eol=lf
+*.xml   text eol=lf
+*.xslt   text eol=lf
+.clang-format   text eol=lf
+.eslintrc.js   text eol=lf
+.git-blame-ignore-revs   text eol=lf
+.gitattributes   text eol=lf
+.gitignore   text eol=lf
+.vpython   text eol=lf
+codereview.settings   text eol=lf
+DEPS   text eol=lf
+ATL_OWNERS   text eol=lf
+LICENSE text eol=lf
+LICENSE.* text eol=lf
+MAJOR_BRANCH_DATE   text eol=lf
+OWNERS   text eol=lf
+README   text eol=lf
+README.*   text eol=lf
+WATCHLISTS   text eol=lf
+VERSION   text eol=lf
+DIR_METADATA   text eol=lf
+
+# Skip Tricium by default on files in third_party.
+third_party/** -tricium

src/.gn

@@ -0,0 +1,161 @@
+# This file is used by the GN meta build system to find the root of the source
+# tree and to set startup options. For documentation on the values set in this
+# file, run "gn help dotfile" at the command line.
+
+import("//build/dotfile_settings.gni")
+import("//third_party/angle/dotfile_settings.gni")
+
+# The location of the build configuration file.
+buildconfig = "//build/config/BUILDCONFIG.gn"
+
+# The python interpreter to use by default. On Windows, this will look
+# for python3.exe and python3.bat.
+script_executable = "python3"
+
+# These arguments override the default values for items in a declare_args
+# block. "gn args" in turn can override these.
+#
+# In general the value for a build arg in the declare_args block should be the
+# default. In some cases, a DEPS-ed in project will want different defaults for
+# being built as part of Chrome vs. being built standalone. In this case, the
+# Chrome defaults should go here. There should be no overrides here for
+# values declared in the main Chrome repository.
+#
+# Important note for defining defaults: This file is executed before the
+# BUILDCONFIG.gn file. That file sets up the global variables like "is_ios".
+# This means that the default_args can not depend on the platform,
+# architecture, or other build parameters. If you really need that, the other
+# repo should define a flag that toggles on a behavior that implements the
+# additional logic required by Chrome to set the variables.
+default_args = {
+  # TODO(brettw) bug 684096: Chrome on iOS does not build v8, so "gn gen" prints
+  # a warning that "Build argument has no effect". When adding a v8 variable, it
+  # also needs to be defined to src/ios/BUILD.gn (respectively removed from both
+  # location when it is removed).
+
+  v8_enable_gdbjit = false
+  v8_imminent_deprecation_warnings = false
+
+  # Don't include webrtc's builtin task queue implementation.
+  rtc_link_task_queue_impl = false
+
+  # Don't include the iLBC audio codec.
+  # TODO(bugs.webrtc.org/8396): Once WebRTC gets rid of its internal
+  # deps on codecs, we can remove this.
+  rtc_include_ilbc = false
+
+  # Changes some setup for the Crashpad build to set them to build against
+  # Chromium's zlib, base, etc.
+  crashpad_dependencies = "chromium"
+
+  # Override ANGLE's Vulkan dependencies.
+  angle_vulkan_headers_dir = "//third_party/vulkan-deps/vulkan-headers/src"
+  angle_vulkan_loader_dir = "//third_party/vulkan-deps/vulkan-loader/src"
+  angle_vulkan_tools_dir = "//third_party/vulkan-deps/vulkan-tools/src"
+  angle_vulkan_validation_layers_dir =
+      "//third_party/vulkan-deps/vulkan-validation-layers/src"
+
+  # Overwrite default args declared in the Fuchsia sdk
+  fuchsia_sdk_readelf_exec =
+      "//third_party/llvm-build/Release+Asserts/bin/llvm-readelf"
+
+  devtools_visibility = [ "*" ]
+}
+
+# These are the targets to skip header checking by default. The files in targets
+# matching these patterns (see "gn help label_pattern" for format) will not have
+# their includes checked for proper dependencies when you run either
+# "gn check" or "gn gen --check".
+no_check_targets = [
+  # //v8, https://crbug.com/v8/7330
+  "//v8/src/inspector:inspector",  # 20 errors
+  "//v8/test/cctest:cctest_sources",  # 15 errors
+  "//v8/test/unittests:inspector_unittests_sources",  # 2 errors
+  "//v8:cppgc_base",  # 1 error
+  "//v8:v8_internal_headers",  # 11 errors
+  "//v8:v8_libplatform",  # 2 errors
+]
+
+# These are the list of GN files that run exec_script. This whitelist exists
+# to force additional review for new uses of exec_script, which is strongly
+# discouraged.
+#
+# PLEASE READ
+#
+# You should almost never need to add new exec_script calls. exec_script is
+# slow, especially on Windows, and can cause confusing effects. Although
+# individually each call isn't slow or necessarily very confusing, at the scale
+# of our repo things get out of hand quickly. By strongly pushing back on all
+# additions, we keep the build fast and clean. If you think you need to add a
+# new call, please consider:
+#
+# - Do not use a script to check for the existence of a file or directory to
+#   enable a different mode. Instead, use GN build args to enable or disable
+#   functionality and set options. An example is checking for a file in the
+#   src-internal repo to see if the corresponding src-internal feature should
+#   be enabled. There are several things that can go wrong with this:
+#
+#    - It's mysterious what causes some things to happen. Although in many cases
+#      such behavior can be conveniently automatic, GN optimizes for explicit
+#      and obvious behavior so people can more easily diagnose problems.
+#
+#    - The user can't enable a mode for one build and not another. With GN build
+#      args, the user can choose the exact configuration of multiple builds
+#      using one checkout. But implicitly basing flags on the state of the
+#      checkout, this functionality is broken.
+#
+#    - It's easy to get stale files. If for example the user edits the gclient
+#      to stop checking out src-internal (or any other optional thing), it's
+#      easy to end up with stale files still mysteriously triggering build
+#      conditions that are no longer appropriate (yes, this happens in real
+#      life).
+#
+# - Do not use a script to iterate files in a directory (glob):
+#
+#    - This has the same "stale file" problem as the above discussion. Various
+#      operations can leave untracked files in the source tree which can cause
+#      surprising effects.
+#
+#    - It becomes impossible to use "git grep" to find where a certain file is
+#      referenced. This operation is very common and people really do get
+#      confused when things aren't listed.
+#
+#    - It's easy to screw up. One common case is a build-time script that packs
+#      up a directory. The author notices that the script isn't re-run when the
+#      directory is updated, so adds a glob so all the files are listed as
+#      inputs. This seems to work great... until a file is deleted. When a
+#      file is deleted, all the inputs the glob lists will still be up to date
+#      and no command-lines will have been changed. The action will not be
+#      re-run and the build will be broken. It is possible to get this correct
+#      using glob, and it's possible to mess it up without glob, but globs make
+#      this situation much easier to create. if the build always lists the
+#      files and passes them to a script, it will always be correct.
+
+exec_script_whitelist =
+    build_dotfile_settings.exec_script_whitelist +
+    angle_dotfile_settings.exec_script_whitelist +
+    [
+      # Whitelist entries for //build should go into
+      # //build/dotfile_settings.gni instead, so that they can be shared
+      # with other repos. The entries in this list should be only for files
+      # in the Chromium repo outside of //build.
+      "//build_overrides/build.gni",
+
+      "//chrome/android/webapk/shell_apk/prepare_upload_dir/BUILD.gn",
+      "//chrome/version.gni",
+
+      # TODO(dgn): Layer violation but breaks the build otherwise, see
+      # https://crbug.com/474506.
+      "//clank/java/BUILD.gn",
+      "//clank/native/BUILD.gn",
+
+      "//google_apis/BUILD.gn",
+      "//printing/BUILD.gn",
+
+      "//remoting/host/installer/linux/BUILD.gn",
+      "//remoting/remoting_version.gni",
+      "//remoting/host/installer/win/generate_clsids.gni",
+
+      "//tools/grit/grit_rule.gni",
+      "//tools/gritsettings/BUILD.gn",
+    ]

src/LICENSE

@@ -0,0 +1,27 @@
+// Copyright 2015 The Chromium Authors
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//    * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//    * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//    * Neither the name of Google LLC nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

src/base/DEPS

@@ -0,0 +1,41 @@
+include_rules = [
+  "+third_party/ashmem",
+  "+third_party/apple_apsl",
+  "+third_party/boringssl/src/include",
+  "+third_party/ced",
+  "+third_party/libevent",
+  "+third_party/libunwindstack/src/libunwindstack/include",
+  "+third_party/lss",
+  "+third_party/modp_b64",
+  "+third_party/perfetto/include",
+  "+third_party/perfetto/protos/perfetto",
+  # Conversions between base and Rust types (e.g. base::span <-> rust::Slice)
+  # require the cxx.h header from cxx. This is only used if Rust is enabled
+  # in the gn build; see //base/BUILD.gn's conditional dependency on
+  # //build/rust:cxx_cppdeps.
+  "+third_party/rust/cxx",
+  "+third_party/test_fonts",
+  # JSON Deserialization.
+  "+third_party/rust/serde_json_lenient/v0_1/wrapper",
+
+  # These are implicitly brought in from the root, and we don't want them.
+  "-ipc",
+  "-url",
+
+  # ICU dependendencies must be separate from the rest of base.
+  "-i18n",
+
+  # //base/util can use //base but not vice versa.
+  "-util",
+]
+
+specific_include_rules = {
+  # Special case
+  "process/current_process(|_test)\.h": [
+    "+third_party/perfetto/protos/perfetto/trace/track_event/chrome_process_descriptor.pbzero.h",
+  ],
+  # To evaluate the performance effects of using absl's flat_hash_map.
+  "supports_user_data\.h": [
+    "+third_party/abseil-cpp/absl/container/flat_hash_map.h",
+  ]
+}

src/base/DIR_METADATA

@@ -0,0 +1,3 @@
+monorail {
+  component: "Internals>Core"
+}

src/base/OWNERS

@@ -0,0 +1,51 @@
+# See //base/README.md to find qualification for being an owner.
+
+set noparent
+# NOTE: keep this in sync with lsc-owners-override@chromium.org owners
+# by emailing lsc-policy@chromium.org when this list changes.
+altimin@chromium.org
+danakj@chromium.org
+dcheng@chromium.org
+fdoray@chromium.org
+gab@chromium.org
+kylechar@chromium.org
+mark@chromium.org
+thakis@chromium.org
+thestig@chromium.org
+wez@chromium.org
+# NOTE: keep this in sync with lsc-owners-override@chromium.org owners
+# by emailing lsc-policy@chromium.org when this list changes.
+
+# per-file rules:
+# These are for the common case of adding or renaming files. If you're doing
+# structural changes, please get a review from a reviewer in this file.
+per-file BUILD.gn=*
+
+# For Android-specific changes:
+per-file ..._android*=file://base/android/OWNERS
+
+# For Fuchsia-specific changes:
+per-file ..._fuchsia*=file://build/fuchsia/OWNERS
+
+# For Windows-specific changes:
+per-file ..._win*=file://base/win/OWNERS
+
+per-file callback_list*=pkasting@chromium.org
+per-file feature_list*=asvitkine@chromium.org
+per-file feature_list*=isherman@chromium.org
+
+# Logging-related changes:
+per-file check*=olivierli@chromium.org
+per-file check*=pbos@chromium.org
+per-file dcheck*=olivierli@chromium.org
+per-file dcheck*=pbos@chromium.org
+per-file logging*=olivierli@chromium.org
+per-file logging*=pbos@chromium.org
+per-file notreached.h=olivierli@chromium.org
+per-file notreached.h=pbos@chromium.org
+
+# Restricted since rand_util.h also backs the cryptographically secure RNG.
+per-file rand_util*=set noparent
+per-file rand_util*=file://ipc/SECURITY_OWNERS
+
+per-file safe_numerics_unittest.cc=file://base/numerics/OWNERS

src/base/PRESUBMIT.py

@@ -0,0 +1,159 @@
+# Copyright 2012 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+"""Chromium presubmit script for src/base.
+
+See http://dev.chromium.org/developers/how-tos/depottools/presubmit-scripts
+for more details on the presubmit API built into depot_tools.
+"""
+
+def CheckChangeLintsClean(input_api, output_api):
+  """Makes sure that the code is cpplint clean."""
+  # lint_filters=[] stops the OFF_BY_DEFAULT_LINT_FILTERS from being disabled,
+  # finding many more issues. verbose_level=1 finds a small number of additional
+  # issues.
+  # The only valid extensions for cpplint are .cc, .h, .cpp, .cu, and .ch.
+  # Only process those extensions which are used in Chromium, in directories
+  # that currently lint clean.
+  CLEAN_CPP_FILES_ONLY = (r'base/win/.*\.(cc|h)$', )
+  source_file_filter = lambda x: input_api.FilterSourceFile(
+      x,
+      files_to_check=CLEAN_CPP_FILES_ONLY,
+      files_to_skip=input_api.DEFAULT_FILES_TO_SKIP)
+  return input_api.canned_checks.CheckChangeLintsClean(
+      input_api, output_api, source_file_filter=source_file_filter,
+      lint_filters=[], verbose_level=1)
+
+
+def _CheckNoInterfacesInBase(input_api, output_api):
+  """Checks to make sure no files in libbase.a have |@interface|."""
+  pattern = input_api.re.compile(r'^\s*@interface', input_api.re.MULTILINE)
+  files = []
+  for f in input_api.AffectedSourceFiles(input_api.FilterSourceFile):
+    if (f.LocalPath().startswith('base/') and
+        not "/ios/" in f.LocalPath() and
+        not "/test/" in f.LocalPath() and
+        not f.LocalPath().endswith('.java') and
+        not f.LocalPath().endswith('_unittest.mm') and
+        not f.LocalPath().endswith('_spi.h')):
+      contents = input_api.ReadFile(f)
+      if pattern.search(contents):
+        files.append(f)
+
+  if len(files):
+    return [ output_api.PresubmitError(
+        'Objective-C interfaces or categories are forbidden in libbase. ' +
+        'See http://groups.google.com/a/chromium.org/group/chromium-dev/' +
+        'browse_thread/thread/efb28c10435987fd',
+        files) ]
+  return []
+
+
+def _FindLocations(input_api, search_regexes, files_to_check, files_to_skip):
+  """Returns locations matching one of the search_regexes."""
+  def FilterFile(affected_file):
+    return input_api.FilterSourceFile(
+      affected_file,
+      files_to_check=files_to_check,
+      files_to_skip=files_to_skip)
+
+  no_presubmit = r"// no-presubmit-check"
+  locations = []
+  for f in input_api.AffectedSourceFiles(FilterFile):
+    for line_num, line in f.ChangedContents():
+      for search_regex in search_regexes:
+        if (input_api.re.search(search_regex, line) and
+            not input_api.re.search(no_presubmit, line)):
+          locations.append("    %s:%d" % (f.LocalPath(), line_num))
+          break
+  return locations
+
+
+def _CheckNoTraceEventInclude(input_api, output_api):
+  """Verify that //base includes base_tracing.h instead of trace event headers.
+
+  Checks that files outside trace event implementation include the
+  base_tracing.h header instead of specific trace event implementation headers
+  to maintain compatibility with the gn flag "enable_base_tracing = false".
+  """
+  discouraged_includes = [
+    r'^#include "base/trace_event/(?!base_tracing\.h|base_tracing_forward\.h)',
+    r'^#include "third_party/perfetto/include/',
+  ]
+
+  files_to_check = [
+    r".*\.(h|cc|mm)$",
+  ]
+  files_to_skip = [
+    r".*/test/.*",
+    r".*/trace_event/.*",
+    r".*/tracing/.*",
+  ]
+
+  locations = _FindLocations(input_api, discouraged_includes, files_to_check,
+                             files_to_skip)
+  if locations:
+    return [ output_api.PresubmitError(
+        'Base code should include "base/trace_event/base_tracing.h" instead\n' +
+        'of trace_event implementation headers. If you need to include an\n' +
+        'implementation header, verify that "gn check" and base_unittests\n' +
+        'still pass with gn arg "enable_base_tracing = false" and add\n' +
+        '"// no-presubmit-check" after the include. \n' +
+        '\n'.join(locations)) ]
+  return []
+
+
+def _WarnPbzeroIncludes(input_api, output_api):
+  """Warn to check enable_base_tracing=false when including a pbzero header.
+
+  Emits a warning when including a perfetto pbzero header, encouraging the
+  user to verify that //base still builds with enable_base_tracing=false.
+  """
+  warn_includes = [
+    r'^#include "third_party/perfetto/protos/',
+    r'^#include "base/tracing/protos/',
+  ]
+
+  files_to_check = [
+    r".*\.(h|cc|mm)$",
+  ]
+  files_to_skip = [
+    r".*/test/.*",
+    r".*/trace_event/.*",
+    r".*/tracing/.*",
+  ]
+
+  locations = _FindLocations(input_api, warn_includes, files_to_check,
+                             files_to_skip)
+  if locations:
+    return [ output_api.PresubmitPromptWarning(
+        'Please verify that "gn check" and base_unittests still pass with\n' +
+        'gn arg "enable_base_tracing = false" when adding typed trace\n' +
+        'events to //base. You can use "#if BUILDFLAG(ENABLE_BASE_TRACING)"\n' +
+        'to exclude pbzero headers and anything not supported by\n' +
+        '//base/trace_event/trace_event_stub.h.\n' +
+        '\n'.join(locations)) ]
+  return []
+
+
+def _CommonChecks(input_api, output_api):
+  """Checks common to both upload and commit."""
+  results = []
+  results.extend(_CheckNoInterfacesInBase(input_api, output_api))
+  results.extend(_CheckNoTraceEventInclude(input_api, output_api))
+  results.extend(_WarnPbzeroIncludes(input_api, output_api))
+  results.extend(CheckChangeLintsClean(input_api, output_api))
+  return results
+
+
+def CheckChangeOnUpload(input_api, output_api):
+  results = []
+  results.extend(_CommonChecks(input_api, output_api))
+  return results
+
+
+def CheckChangeOnCommit(input_api, output_api):
+  results = []
+  results.extend(_CommonChecks(input_api, output_api))
+  return results

src/base/README.md

@@ -0,0 +1,87 @@
+# What is this
+Contains a written down set of principles and other information on //base.
+Please add to it!
+
+## About //base:
+
+Chromium is a very mature project. Most things that are generally useful are
+already here and things not here aren't generally useful.
+
+The bar for adding stuff to base is that it must have demonstrated wide
+applicability. Prefer to add things closer to where they're used (i.e. "not
+base"), and pull into base only when needed. In a project our size,
+sometimes even duplication is OK and inevitable.
+
+Adding a new logging macro `DPVELOG_NE` is not more clear than just
+writing the stuff you want to log in a regular logging statement, even
+if it makes your calling code longer. Just add it to your own code.
+
+If the code in question does not need to be used inside base, but will have
+multiple consumers across the codebase, consider placing it in a new directory
+under components/ instead.
+
+base is written for the Chromium project and is not intended to be used
+outside it.  Using base outside of src.git is explicitly not supported,
+and base makes no guarantees about API (or even ABI) stability (like all
+other code in Chromium).  New code that depends on base/ must be in
+src.git. Code that's not in src.git but pulled in through DEPS (for
+example, v8) cannot use base.
+
+## Qualifications for being in //base OWNERS
+  * interest and ability to learn low level/high detail/complex c++ stuff
+  * inclination to always ask why and understand everything (including external
+    interactions like win32) rather than just hoping the author did it right
+  * mentorship/experience
+  * demonstrated good judgement (esp with regards to public APIs) over a length
+    of time
+
+Owners are added when a contributor has shown the above qualifications and
+when they express interest. There isn't an upper bound on the number of OWNERS.
+
+## Design and naming
+  * Be sure to use the base namespace.
+  * STL-like constructs should adhere as closely to STL as possible. Functions
+    and behaviors not present in STL should only be added when they are related
+    to the specific data structure implemented by the container.
+  * For STL-like constructs our policy is that they should use STL-like naming
+    even when it may conflict with the style guide. So functions and class names
+    should be lower case with underscores. Non-STL-like classes and functions
+    should use Google naming.
+
+## Performance testing
+
+Since the primitives provided by //base are used very widely, it is important to
+ensure they scale to the necessary workloads and perform well under all
+supported platforms. The `base_perftests` target is a suite of
+synthetic microbenchmarks that measure performance in various scenarios:
+
+  * BasicPostTaskPerfTest: Exercises MessageLoopTaskRunner's multi-threaded
+    queue in isolation.
+  * ConditionVariablePerfTest: Measures thread switching cost of condition
+    variables.
+  * IntegratedPostTaskPerfTest: Exercises the full MessageLoop/RunLoop
+    machinery.
+  * JSONPerfTest: Tests JSONWriter and JSONReader performance.
+  * MessageLoopPerfTest: Measures the speed of task posting in various
+    configurations.
+  * ObserverListPerfTest: Exercises adding, removing and signalling observers.
+  * PartitionLockPerfTest: Tests the implementation of Lock used in
+    PartitionAlloc
+  * PthreadEventPerfTest: Establishes the baseline thread switching cost using
+    pthreads.
+  * RandUtilPerfTest: Measures the time it takes to generate random numbers.
+  * ScheduleWorkTest: Measures the overhead of MessagePump::ScheduleWork.
+  * SequenceManagerPerfTest: Benchmarks SequenceManager scheduling with various
+    underlying task runners.
+  * TaskObserverPerfTest: Measures the incremental cost of adding task
+    observers.
+  * TaskPerfTest: Checks the cost of posting tasks between threads.
+  * ThreadLocalStoragePerfTest: Exercises different mechanisms for accessing
+    data associated with the current thread (C++ `thread_local`, the
+    implementation in //base, the POSIX/WinAPI directly)
+  * WaitableEvent{Thread,}PerfTest: Measures waitable events in single and
+    multithreaded scenarios.
+
+Regressions in these benchmarks can generally by caused by 1) operating system
+changes, 2) compiler version or flag changes or 3) changes in //base code
+itself.

src/base/SECURITY_OWNERS

@@ -0,0 +1,13 @@
+# Changes to code that runs at high privilege and which has a high risk of
+# memory corruption, such as parsers for complex inputs, require a security
+# review to avoid introducing sandbox escapes.
+#
+# Although this file is in base/, it may apply to more than just base, OWNERS
+# files outside of base may also include this file.
+#
+# Security team: If you are uncomfortable reviewing a particular bit of code
+# yourself, don't hesitate to seek help from another security team member!
+# Nobody knows everything, and the only way to learn is from experience.
+dcheng@chromium.org
+rsesek@chromium.org
+tsepez@chromium.org

src/base/allocator/BUILD.gn

@@ -0,0 +1,31 @@
+# Copyright 2013 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import("//base/allocator/allocator.gni")
+import("//base/allocator/partition_allocator/partition_alloc.gni")
+import("//build/buildflag_header.gni")
+import("//build/config/compiler/compiler.gni")
+import("//build/config/dcheck_always_on.gni")
+
+buildflag_header("buildflags") {
+  header = "buildflags.h"
+
+  flags =
+      [ "USE_PARTITION_ALLOC_AS_GWP_ASAN_STORE=$enable_backup_ref_ptr_support" ]
+}
+
+if (is_apple) {
+  source_set("early_zone_registration_apple") {
+    sources = [
+      "early_zone_registration_apple.cc",
+      "early_zone_registration_apple.h",
+      "partition_allocator/shim/early_zone_registration_constants.h",
+    ]
+
+    deps = [
+      ":buildflags",
+      "//base/allocator/partition_allocator:buildflags",
+    ]
+  }
+}

src/base/allocator/DIR_METADATA

@@ -0,0 +1,3 @@
+monorail {
+  component: "Internals"
+}

src/base/allocator/OWNERS

@@ -0,0 +1,7 @@
+lizeb@chromium.org
+primiano@chromium.org
+wfh@chromium.org
+
+per-file allocator.gni=file://base/allocator/partition_allocator/OWNERS
+per-file partition_alloc*=file://base/allocator/partition_allocator/OWNERS
+per-file BUILD.gn=file://base/allocator/partition_allocator/OWNERS

src/base/allocator/README.md

@@ -0,0 +1,156 @@
+This document describes how malloc / new calls are routed in the various Chrome
+platforms.
+
+Bear in mind that the chromium codebase does not always just use `malloc()`.
+Some examples:
+ - Large parts of the renderer (Blink) use two home-brewed allocators,
+   PartitionAlloc and BlinkGC (Oilpan).
+ - Some subsystems, such as the V8 JavaScript engine, handle memory management
+   autonomously.
+ - Various parts of the codebase use abstractions such as `SharedMemory` or
+   `DiscardableMemory` which, similarly to the above, have their own page-level
+   memory management.
+
+Background
+----------
+The `allocator` target defines at compile-time the platform-specific choice of
+the allocator and extra-hooks which services calls to malloc/new. The relevant
+build-time flags involved are `use_allocator_shim` and
+`use_partition_alloc_as_malloc`.
+
+By default, these are true on all platforms except iOS (not yet supported) and
+NaCl (no plan to support).
+Furthermore, when building with a sanitizer (e.g. `asan`, `msan`, ...) both the
+allocator and the shim layer are disabled.
+
+
+Layering and build deps
+-----------------------
+The `allocator` target provides the linker flags required for the Windows shim
+layer. The `base` target is (almost) the only one depending on `allocator`. No
+other targets should depend on it, with the exception of the very few
+executables / dynamic libraries that don't depend, either directly or
+indirectly, on `base` within the scope of a linker unit.
+
+More importantly, **no other place outside of `/base` should depend on the
+specific allocator**.
+If such a functional dependency is required that should be achieved using
+abstractions in `base` (see `/base/allocator/allocator_extension.h` and
+`/base/memory/`)
+
+**Why `base` depends on `allocator`?**
+Because it needs to provide services that depend on the actual allocator
+implementation. In the past `base` used to pretend to be allocator-agnostic
+and get the dependencies injected by other layers. This ended up being an
+inconsistent mess.
+See the [allocator cleanup doc][url-allocator-cleanup] for more context.
+
+Linker unit targets (executables and shared libraries) that depend in some way
+on `base` (most of the targets in the codebase) automatically get the correct
+set of linker flags to pull in the Windows shim-layer (if needed).
+
+
+Source code
+-----------
+This directory contains just the allocator (i.e. shim) layer that switches
+between the different underlying memory allocation implementations.
+
+
+Unified allocator shim
+----------------------
+On most platforms, Chrome overrides the malloc / operator new symbols (and
+corresponding free / delete and other variants). This is to enforce security
+checks and lately to enable the
+[memory-infra heap profiler][url-memory-infra-heap-profiler].
+Historically each platform had its special logic for defining the allocator
+symbols in different places of the codebase. The unified allocator shim is
+a project aimed to unify the symbol definition and allocator routing logic in
+a central place.
+
+ - Full documentation: [Allocator shim design doc][url-allocator-shim].
+ - Current state: Available and enabled by default on Android, CrOS, Linux,
+   Mac OS and Windows.
+ - Tracking bug: [crbug.com/550886](https://crbug.com/550886).
+ - Build-time flag: `use_allocator_shim`.
+
+**Overview of the unified allocator shim**
+The allocator shim consists of three stages:
+```
++-------------------------+    +-----------------------+    +----------------+
+|     malloc & friends    | -> |       shim layer      | -> |   Routing to   |
+|    symbols definition   |    |     implementation    |    |    allocator   |
++-------------------------+    +-----------------------+    +----------------+
+| - libc symbols (malloc, |    | - Security checks     |    | - glibc        |
+|   calloc, free, ...)    |    | - Chain of dispatchers|    | - Android      |
+| - C++ symbols (operator |    |   that can intercept  |    |   bionic       |
+|   new, delete, ...)     |    |   and override        |    | - WinHeap      |
+| - glibc weak symbols    |    |   allocations         |    | - Partition    |
+|   (__libc_malloc, ...)  |    +-----------------------+    |   Alloc        |
++-------------------------+                                 +----------------+
+```
+
+**1. malloc symbols definition**
+This stage takes care of overriding the symbols `malloc`, `free`,
+`operator new`, `operator delete` and friends and routing those calls inside the
+allocator shim (next point).
+This is taken care of by the headers in `allocator_shim_override_*`.
+
+*On Windows*: Windows' UCRT (Universal C Runtime) exports weak symbols, that we
+can override in `allocator_shim_override_ucrt_symbols_win.h`.
+
+*On Linux/CrOS*: the allocator symbols are defined as exported global symbols
+in `allocator_shim_override_libc_symbols.h` (for `malloc`, `free` and friends)
+and in `allocator_shim_override_cpp_symbols.h` (for `operator new`,
+`operator delete` and friends).
+This enables proper interposition of malloc symbols referenced by the main
+executable and any third party libraries. Symbol resolution on Linux is a breadth first search that starts from the root link unit, that is the executable
+(see EXECUTABLE AND LINKABLE FORMAT (ELF) - Portable Formats Specification).
+The Linux/CrOS shim was introduced by
+[crrev.com/1675143004](https://crrev.com/1675143004).
+
+*On Android*: load-time symbol interposition (unlike the Linux/CrOS case) is not
+possible. This is because Android processes are `fork()`-ed from the Android
+zygote, which pre-loads libc.so and only later native code gets loaded via
+`dlopen()` (symbols from `dlopen()`-ed libraries get a different resolution
+scope).
+In this case, the approach instead of wrapping symbol resolution at link time
+(i.e. during the build), via the `--Wl,-wrap,malloc` linker flag.
+The use of this wrapping flag causes:
+ - All references to allocator symbols in the Chrome codebase to be rewritten as
+   references to `__wrap_malloc` and friends. The `__wrap_malloc` symbols are
+   defined in the `allocator_shim_override_linker_wrapped_symbols.h` and
+   route allocator calls inside the shim layer.
+ - The reference to the original `malloc` symbols (which typically is defined by
+   the system's libc.so) are accessible via the special `__real_malloc` and
+   friends symbols (which will be relocated, at load time, against `malloc`).
+
+In summary, this approach is transparent to the dynamic loader, which still sees
+undefined symbol references to malloc symbols.
+These symbols will be resolved against libc.so as usual.
+More details in [crrev.com/1719433002](https://crrev.com/1719433002).
+
+**2. Shim layer implementation**
+This stage contains the actual shim implementation. This consists of:
+- A singly linked list of dispatchers (structs with function pointers to `malloc`-like functions). Dispatchers can be dynamically inserted at runtime
+(using the `InsertAllocatorDispatch` API). They can intercept and override
+allocator calls.
+- The security checks (suicide on malloc-failure via `std::new_handler`, etc).
+This happens inside `allocator_shim.cc`
+
+**3. Final allocator routing**
+The final element of the aforementioned dispatcher chain is statically defined
+at build time and ultimately routes the allocator calls to the actual allocator
+(as described in the *Background* section above). This is taken care of by the
+headers in `allocator_shim_default_dispatch_to_*` files.
+
+
+Related links
+-------------
+- [Unified allocator shim doc - Feb 2016][url-allocator-shim]
+- [Allocator cleanup doc - Jan 2016][url-allocator-cleanup]
+- [Proposal to use PartitionAlloc as default allocator](https://crbug.com/339604)
+- [Memory-Infra: Tools to profile memory usage in Chrome](/docs/memory-infra/README.md)
+
+[url-allocator-cleanup]: https://docs.google.com/document/d/1V77Kgp_4tfaaWPEZVxNevoD02wXiatnAv7Ssgr0hmjg/edit?usp=sharing
+[url-memory-infra-heap-profiler]: /docs/memory-infra/heap_profiler.md
+[url-allocator-shim]: https://docs.google.com/document/d/1yKlO1AO4XjpDad9rjcBOI15EKdAGsuGO_IeZy0g0kxo/edit?usp=sharing

src/base/allocator/allocator.gni

@@ -0,0 +1,27 @@
+# Copyright 2019 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import("//base/allocator/partition_allocator/partition_alloc.gni")
+
+# Chromium-specific asserts. External embedders _may_ elect to use these
+# features even without PA-E.
+if (!use_partition_alloc_as_malloc) {
+  # In theory, BackupRefPtr will work just fine without
+  # PartitionAlloc-Everywhere, but its scope would be limited to partitions
+  # that are invoked explicitly (not via malloc). These are only Blink
+  # partition, where we currently don't even use raw_ptr<T>.
+  assert(!enable_backup_ref_ptr_support,
+         "Chromium does not use BRP without PA-E")
+
+  # Pointer compression works only if all pointers are guaranteed to be
+  # allocated by PA (in one of its core pools, to be precise). In theory,
+  # this could be useful with partitions that are invoked explicitly. In
+  # practice, the pointers we have in mind for compression (scoped_refptr<>,
+  # unique_ptr<>) require PA-E.
+  assert(!enable_pointer_compression_support,
+         "Pointer compressions likely doesn't make sense without PA-E")
+}
+
+assert(use_allocator_shim || !use_partition_alloc_as_malloc,
+       "PartitionAlloc-Everywhere requires the allocator shim")

src/base/allocator/allocator_check.cc

@@ -0,0 +1,38 @@
+// Copyright 2016 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/allocator_check.h"
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_WIN)
+#include "base/allocator/partition_allocator/shim/winheap_stubs_win.h"
+#endif
+
+#if BUILDFLAG(IS_LINUX) || BUILDFLAG(IS_CHROMEOS)
+#include <malloc.h>
+#endif
+
+#if BUILDFLAG(IS_APPLE)
+#include "base/allocator/partition_allocator/shim/allocator_interception_apple.h"
+#endif
+
+namespace base::allocator {
+
+bool IsAllocatorInitialized() {
+#if BUILDFLAG(IS_WIN) && BUILDFLAG(USE_ALLOCATOR_SHIM)
+  // Set by allocator_shim_override_ucrt_symbols_win.h when the
+  // shimmed _set_new_mode() is called.
+  return allocator_shim::g_is_win_shim_layer_initialized;
+#elif BUILDFLAG(IS_APPLE) && !defined(MEMORY_TOOL_REPLACES_ALLOCATOR) && \
+    !BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+  // From allocator_interception_mac.mm.
+  return allocator_shim::g_replaced_default_zone;
+#else
+  return true;
+#endif
+}
+
+}  // namespace base::allocator

src/base/allocator/allocator_check.h

@@ -0,0 +1,18 @@
+// Copyright 2016 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_ALLOCATOR_CHECK_H_
+#define BASE_ALLOCATOR_ALLOCATOR_CHECK_H_
+
+#include "base/base_export.h"
+
+namespace base {
+namespace allocator {
+
+BASE_EXPORT bool IsAllocatorInitialized();
+
+}  // namespace allocator
+}  // namespace base
+
+#endif  // BASE_ALLOCATOR_ALLOCATOR_CHECK_H_

src/base/allocator/allocator_extension.cc

@@ -0,0 +1,15 @@
+// Copyright 2012 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/allocator_extension.h"
+#include "base/allocator/buildflags.h"
+#include "base/check.h"
+
+namespace base {
+namespace allocator {
+
+void ReleaseFreeMemory() {}
+
+}  // namespace allocator
+}  // namespace base

src/base/allocator/allocator_extension.h

@@ -0,0 +1,23 @@
+// Copyright 2012 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_ALLOCATOR_EXTENSION_H_
+#define BASE_ALLOCATOR_ALLOCATOR_EXTENSION_H_
+
+#include <stddef.h>  // for size_t
+
+#include "base/base_export.h"
+#include "build/build_config.h"
+
+namespace base {
+namespace allocator {
+
+// Request that the allocator release any free memory it knows about to the
+// system.
+BASE_EXPORT void ReleaseFreeMemory();
+
+}  // namespace allocator
+}  // namespace base
+
+#endif  // BASE_ALLOCATOR_ALLOCATOR_EXTENSION_H_

src/base/allocator/dispatcher/configuration.h

@@ -0,0 +1,24 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_CONFIGURATION_H_
+#define BASE_ALLOCATOR_DISPATCHER_CONFIGURATION_H_
+
+#include <cstddef>
+
+namespace base::allocator::dispatcher::configuration {
+
+// The maximum number of optional observers that may be present depending on
+// command line parameters.
+constexpr size_t kMaximumNumberOfOptionalObservers = 4;
+
+// The total number of observers including mandatory and optional observers.
+// Primarily the number of observers affects the performance at allocation time.
+// The current value of 4 doesn't have hard evidence. Keep in mind that
+// also a single observer can severely impact performance.
+constexpr size_t kMaximumNumberOfObservers = 4;
+
+}  // namespace base::allocator::dispatcher::configuration
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_CONFIGURATION_H_

src/base/allocator/dispatcher/dispatcher.cc

@@ -0,0 +1,114 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/dispatcher/dispatcher.h"
+
+#include "base/allocator/dispatcher/internal/dispatch_data.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/shim/allocator_shim.h"
+#include "base/check.h"
+#include "base/dcheck_is_on.h"
+#include "base/no_destructor.h"
+
+#if DCHECK_IS_ON()
+#include <atomic>
+#endif
+
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+#include "base/allocator/partition_allocator/partition_alloc_hooks.h"
+#endif
+
+namespace base::allocator::dispatcher {
+
+// The private implementation of Dispatcher.
+struct Dispatcher::Impl {
+  void Initialize(const internal::DispatchData& dispatch_data) {
+#if DCHECK_IS_ON()
+    DCHECK(!is_initialized_check_flag_.test_and_set());
+#endif
+
+    dispatch_data_ = dispatch_data;
+    ConnectToEmitters(dispatch_data_);
+  }
+
+  void Reset() {
+#if DCHECK_IS_ON()
+    DCHECK([&]() {
+      auto const was_set = is_initialized_check_flag_.test_and_set();
+      is_initialized_check_flag_.clear();
+      return was_set;
+    }());
+#endif
+
+    DisconnectFromEmitters(dispatch_data_);
+    dispatch_data_ = {};
+  }
+
+ private:
+  // Connect the hooks to the memory subsystem. In some cases, most notably when
+  // we have no observers at all, the hooks will be invalid and must NOT be
+  // connected. This way we prevent notifications although no observers are
+  // present.
+  static void ConnectToEmitters(const internal::DispatchData& dispatch_data) {
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+    if (auto* const allocator_dispatch = dispatch_data.GetAllocatorDispatch()) {
+      allocator_shim::InsertAllocatorDispatch(allocator_dispatch);
+    }
+#endif
+
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+    {
+      auto* const allocation_hook = dispatch_data.GetAllocationObserverHook();
+      auto* const free_hook = dispatch_data.GetFreeObserverHook();
+      if (allocation_hook && free_hook) {
+        partition_alloc::PartitionAllocHooks::SetObserverHooks(allocation_hook,
+                                                               free_hook);
+      }
+    }
+#endif
+  }
+
+  static void DisconnectFromEmitters(internal::DispatchData& dispatch_data) {
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+    if (auto* const allocator_dispatch = dispatch_data.GetAllocatorDispatch()) {
+      allocator_shim::RemoveAllocatorDispatchForTesting(
+          allocator_dispatch);  // IN-TEST
+    }
+#endif
+
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+    partition_alloc::PartitionAllocHooks::SetObserverHooks(nullptr, nullptr);
+#endif
+  }
+
+  // Information on the hooks.
+  internal::DispatchData dispatch_data_;
+#if DCHECK_IS_ON()
+  // Indicator if the dispatcher has been initialized before.
+#if !defined(__cpp_lib_atomic_value_initialization) || \
+    __cpp_lib_atomic_value_initialization < 201911L
+  std::atomic_flag is_initialized_check_flag_ = ATOMIC_FLAG_INIT;
+#else
+  std::atomic_flag is_initialized_check_flag_;
+#endif
+#endif
+};
+
+Dispatcher::Dispatcher() : impl_(std::make_unique<Impl>()) {}
+
+Dispatcher::~Dispatcher() = default;
+
+Dispatcher& Dispatcher::GetInstance() {
+  static base::NoDestructor<Dispatcher> instance;
+  return *instance;
+}
+
+void Dispatcher::Initialize(const internal::DispatchData& dispatch_data) {
+  impl_->Initialize(dispatch_data);
+}
+
+void Dispatcher::ResetForTesting() {
+  impl_->Reset();
+}
+}  // namespace base::allocator::dispatcher

src/base/allocator/dispatcher/dispatcher.h

@@ -0,0 +1,75 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_DISPATCHER_H_
+#define BASE_ALLOCATOR_DISPATCHER_DISPATCHER_H_
+
+#include "base/allocator/dispatcher/internal/dispatcher_internal.h"
+#include "base/base_export.h"
+
+#include <memory>
+
+namespace base::allocator::dispatcher {
+
+namespace internal {
+struct DispatchData;
+}
+
+// Dispatcher serves as the top level instance for managing the dispatch
+// mechanism. The class instance manages connections to the various memory
+// subsystems such as PartitionAlloc. To keep the public interface as lean as
+// possible it uses a pimpl pattern.
+class BASE_EXPORT Dispatcher {
+ public:
+  static Dispatcher& GetInstance();
+
+  Dispatcher();
+
+  // Initialize the dispatch mechanism with the given tuple of observers. The
+  // observers must be valid (it is only DCHECKed internally at initialization,
+  // but not verified further)
+  // If Initialize is called multiple times, the first one wins. All later
+  // invocations are silently ignored. Initialization is protected from
+  // concurrent invocations. In case of concurrent accesses, the first one to
+  // get the lock wins.
+  // The dispatcher invokes following functions on the observers:
+  // void OnAllocation(void* address,
+  //                   size_t size,
+  //                   AllocationSubsystem sub_system,
+  //                   const char* type_name);
+  // void OnFree(void* address);
+  //
+  // Note: The dispatcher mechanism does NOT bring systematic protection against
+  // recursive invocations. That is, observers which allocate memory on the
+  // heap, i.e. through dynamically allocated containers or by using the
+  // CHECK-macro, are responsible to break these recursions!
+  template <typename... ObserverTypes>
+  void Initialize(const std::tuple<ObserverTypes...>& observers) {
+    // Get the hooks for running these observers and pass them to further
+    // initialization
+    Initialize(internal::GetNotificationHooks(observers));
+  }
+
+  // The following functions provide an interface to setup and tear down the
+  // dispatcher when testing. This must NOT be used from production code since
+  // the hooks cannot be removed reliably under all circumstances.
+  template <typename ObserverType>
+  void InitializeForTesting(ObserverType* observer) {
+    Initialize(std::make_tuple(observer));
+  }
+
+  void ResetForTesting();
+
+ private:
+  // structure and pointer to the private implementation.
+  struct Impl;
+  std::unique_ptr<Impl> const impl_;
+
+  ~Dispatcher();
+
+  void Initialize(const internal::DispatchData& dispatch_data);
+};
+}  // namespace base::allocator::dispatcher
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_DISPATCHER_H_

src/base/allocator/dispatcher/initializer.h

@@ -0,0 +1,206 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_INITIALIZER_H_
+#define BASE_ALLOCATOR_DISPATCHER_INITIALIZER_H_
+
+#include "base/allocator/dispatcher/configuration.h"
+#include "base/allocator/dispatcher/dispatcher.h"
+#include "base/allocator/dispatcher/internal/tools.h"
+
+#include <tuple>
+#include <utility>
+
+namespace base::allocator::dispatcher {
+namespace internal {
+
+// Filter the passed observers and perform initialization of the passed
+// dispatcher.
+template <size_t CurrentIndex,
+          typename DispatcherType,
+          typename CheckObserverPredicate,
+          typename VerifiedObservers,
+          typename UnverifiedObservers,
+          size_t... IndicesToSelect>
+inline void DoInitialize(DispatcherType& dispatcher,
+                         CheckObserverPredicate check_observer,
+                         const VerifiedObservers& verified_observers,
+                         const UnverifiedObservers& unverified_observers,
+                         std::index_sequence<IndicesToSelect...> indices) {
+  if constexpr (CurrentIndex < std::tuple_size<UnverifiedObservers>::value) {
+    // We still have some items left to handle.
+    if (check_observer(std::get<CurrentIndex>(unverified_observers))) {
+      // The current observer is valid. Hence, append the index of the current
+      // item to the set of indices and head on to the next item.
+      DoInitialize<CurrentIndex + 1>(
+          dispatcher, check_observer, verified_observers, unverified_observers,
+          std::index_sequence<IndicesToSelect..., CurrentIndex>{});
+    } else {
+      // The current observer is not valid. Hence, head on to the next item with
+      // an unaltered list of indices.
+      DoInitialize<CurrentIndex + 1>(dispatcher, check_observer,
+                                     verified_observers, unverified_observers,
+                                     indices);
+    }
+  } else if constexpr (CurrentIndex ==
+                       std::tuple_size<UnverifiedObservers>::value) {
+    // So we have met the end of the tuple of observers to verify.
+    // Hence, we extract the additional valid observers, append to the tuple of
+    // already verified observers and hand over to the dispatcher.
+    auto observers = std::tuple_cat(
+        verified_observers,
+        std::make_tuple(std::get<IndicesToSelect>(unverified_observers)...));
+
+    // Do a final check that neither the maximum total number of observers nor
+    // the maximum number of optional observers is exceeded.
+    static_assert(std::tuple_size<decltype(observers)>::value <=
+                  configuration::kMaximumNumberOfObservers);
+    static_assert(sizeof...(IndicesToSelect) <=
+                  configuration::kMaximumNumberOfOptionalObservers);
+
+    dispatcher.Initialize(std::move(observers));
+  }
+}
+
+}  // namespace internal
+
+// The result of concatenating two tuple-types.
+template <typename... tuples>
+using TupleCat = decltype(std::tuple_cat(std::declval<tuples>()...));
+
+// Initializer collects mandatory and optional observers and initializes the
+// passed Dispatcher with only the enabled observers.
+//
+// In some situations, presence of observers depends on runtime. i.e. command
+// line parameters or CPU features. With 3 optional observers we already have 8
+// different combinations. Initializer takes the job of dealing with all
+// combinations from the user. It allows users to pass all observers (including
+// nullptr for disabled optional observers) and initializes the Dispatcher with
+// only the enabled observers.
+//
+// Since this process results in a combinatoric explosion, Initializer
+// distinguishes between optional and mandatory observers. Mandatory observers
+// are not included in the filtering process and must always be enabled (not
+// nullptr).
+//
+// To allow the Initializer to track the number and exact type of observers, it
+// is implemented as a templated class which holds information on the types in
+// the std::tuples passed as template parameters. Therefore, whenever any type
+// observer it set, the initializer changes its type to reflect this.
+template <typename MandatoryObservers = std::tuple<>,
+          typename OptionalObservers = std::tuple<>>
+struct BASE_EXPORT Initializer {
+  Initializer() = default;
+  Initializer(MandatoryObservers mandatory_observers,
+              OptionalObservers optional_observers)
+      : mandatory_observers_(std::move(mandatory_observers)),
+        optional_observers_(std::move(optional_observers)) {}
+
+  // Set the mandatory observers. The number of observers that can be set is
+  // limited by configuration::maximum_number_of_observers.
+  template <typename... NewMandatoryObservers,
+            std::enable_if_t<
+                internal::LessEqual((sizeof...(NewMandatoryObservers) +
+                                     std::tuple_size<OptionalObservers>::value),
+                                    configuration::kMaximumNumberOfObservers),
+                bool> = true>
+  Initializer<std::tuple<NewMandatoryObservers*...>, OptionalObservers>
+  SetMandatoryObservers(NewMandatoryObservers*... mandatory_observers) const {
+    return {std::make_tuple(mandatory_observers...), GetOptionalObservers()};
+  }
+
+  // Add mandatory observers. The number of observers that can be added is
+  // limited by the current number of observers, see
+  // configuration::maximum_number_of_observers.
+  template <typename... AdditionalMandatoryObservers,
+            std::enable_if_t<internal::LessEqual(
+                                 std::tuple_size<MandatoryObservers>::value +
+                                     sizeof...(AdditionalMandatoryObservers) +
+                                     std::tuple_size<OptionalObservers>::value,
+                                 configuration::kMaximumNumberOfObservers),
+                             bool> = true>
+  Initializer<TupleCat<MandatoryObservers,
+                       std::tuple<AdditionalMandatoryObservers*...>>,
+              OptionalObservers>
+  AddMandatoryObservers(
+      AdditionalMandatoryObservers*... additional_mandatory_observers) const {
+    return {std::tuple_cat(GetMandatoryObservers(),
+                           std::make_tuple(additional_mandatory_observers...)),
+            GetOptionalObservers()};
+  }
+
+  // Set the optional observers. The number of observers that can be set is
+  // limited by configuration::maximum_number_of_optional_observers as well as
+  // configuration::maximum_number_of_observers.
+  template <
+      typename... NewOptionalObservers,
+      std::enable_if_t<
+          internal::LessEqual(
+              sizeof...(NewOptionalObservers),
+              configuration::kMaximumNumberOfOptionalObservers) &&
+              internal::LessEqual((sizeof...(NewOptionalObservers) +
+                                   std::tuple_size<MandatoryObservers>::value),
+                                  configuration::kMaximumNumberOfObservers),
+          bool> = true>
+  Initializer<MandatoryObservers, std::tuple<NewOptionalObservers*...>>
+  SetOptionalObservers(NewOptionalObservers*... optional_observers) const {
+    return {GetMandatoryObservers(), std::make_tuple(optional_observers...)};
+  }
+
+  // Add optional observers. The number of observers that can be added is
+  // limited by the current number of optional observers,
+  // configuration::maximum_number_of_optional_observers as well as
+  // configuration::maximum_number_of_observers.
+  template <
+      typename... AdditionalOptionalObservers,
+      std::enable_if_t<
+          internal::LessEqual(
+              std::tuple_size<OptionalObservers>::value +
+                  sizeof...(AdditionalOptionalObservers),
+              configuration::kMaximumNumberOfOptionalObservers) &&
+              internal::LessEqual((std::tuple_size<OptionalObservers>::value +
+                                   sizeof...(AdditionalOptionalObservers) +
+                                   std::tuple_size<MandatoryObservers>::value),
+                                  configuration::kMaximumNumberOfObservers),
+          bool> = true>
+  Initializer<
+      MandatoryObservers,
+      TupleCat<OptionalObservers, std::tuple<AdditionalOptionalObservers*...>>>
+  AddOptionalObservers(
+      AdditionalOptionalObservers*... additional_optional_observers) const {
+    return {GetMandatoryObservers(),
+            std::tuple_cat(GetOptionalObservers(),
+                           std::make_tuple(additional_optional_observers...))};
+  }
+
+  // Perform the actual initialization on the passed dispatcher.
+  // The dispatcher is passed as a template only to provide better testability.
+  template <typename DispatcherType>
+  void DoInitialize(DispatcherType& dispatcher) const {
+    internal::DoInitialize<0>(dispatcher, internal::IsValidObserver{},
+                              GetMandatoryObservers(), GetOptionalObservers(),
+                              {});
+  }
+
+  const MandatoryObservers& GetMandatoryObservers() const {
+    return mandatory_observers_;
+  }
+
+  const OptionalObservers& GetOptionalObservers() const {
+    return optional_observers_;
+  }
+
+ private:
+  MandatoryObservers mandatory_observers_;
+  OptionalObservers optional_observers_;
+};
+
+// Convenience function for creating an empty Initializer.
+inline Initializer<> CreateInitializer() {
+  return {};
+}
+
+}  // namespace base::allocator::dispatcher
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_INITIALIZER_H_

src/base/allocator/dispatcher/internal/dispatch_data.cc

@@ -0,0 +1,42 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/dispatcher/internal/dispatch_data.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+
+namespace base::allocator::dispatcher::internal {
+
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+
+DispatchData& DispatchData::SetAllocationObserverHooks(
+    AllocationObserverHook* allocation_observer_hook,
+    FreeObserverHook* free_observer_hook) {
+  allocation_observer_hook_ = allocation_observer_hook;
+  free_observer_hook_ = free_observer_hook;
+
+  return *this;
+}
+
+DispatchData::AllocationObserverHook* DispatchData::GetAllocationObserverHook()
+    const {
+  return allocation_observer_hook_;
+}
+
+DispatchData::FreeObserverHook* DispatchData::GetFreeObserverHook() const {
+  return free_observer_hook_;
+}
+#endif
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+DispatchData& DispatchData::SetAllocatorDispatch(
+    AllocatorDispatch* allocator_dispatch) {
+  allocator_dispatch_ = allocator_dispatch;
+  return *this;
+}
+
+AllocatorDispatch* DispatchData::GetAllocatorDispatch() const {
+  return allocator_dispatch_;
+}
+#endif
+}  // namespace base::allocator::dispatcher::internal

src/base/allocator/dispatcher/internal/dispatch_data.h

@@ -0,0 +1,58 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_INTERNAL_DISPATCH_DATA_H_
+#define BASE_ALLOCATOR_DISPATCHER_INTERNAL_DISPATCH_DATA_H_
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/base_export.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+#include "base/allocator/partition_allocator/partition_alloc_hooks.h"
+#endif
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+#include "base/allocator/partition_allocator/shim/allocator_shim.h"
+#endif
+
+namespace base::allocator::dispatcher::internal {
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+using allocator_shim::AllocatorDispatch;
+#endif
+
+// A simple utility class to pass all the information required to properly hook
+// into the memory allocation subsystems from DispatcherImpl to the Dispatcher.
+struct BASE_EXPORT DispatchData {
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+  using AllocationObserverHook =
+      partition_alloc::PartitionAllocHooks::AllocationObserverHook;
+  using FreeObserverHook =
+      partition_alloc::PartitionAllocHooks::FreeObserverHook;
+
+  DispatchData& SetAllocationObserverHooks(AllocationObserverHook*,
+                                           FreeObserverHook*);
+  AllocationObserverHook* GetAllocationObserverHook() const;
+  FreeObserverHook* GetFreeObserverHook() const;
+
+ private:
+  AllocationObserverHook* allocation_observer_hook_ = nullptr;
+  FreeObserverHook* free_observer_hook_ = nullptr;
+
+ public:
+#endif
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+  DispatchData& SetAllocatorDispatch(AllocatorDispatch* allocator_dispatch);
+  AllocatorDispatch* GetAllocatorDispatch() const;
+
+ private:
+  AllocatorDispatch* allocator_dispatch_ = nullptr;
+#endif
+};
+
+}  // namespace base::allocator::dispatcher::internal
+
+#endif

src/base/allocator/dispatcher/internal/dispatcher_internal.h

@@ -0,0 +1,353 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_INTERNAL_DISPATCHER_INTERNAL_H_
+#define BASE_ALLOCATOR_DISPATCHER_INTERNAL_DISPATCHER_INTERNAL_H_
+
+#include "base/allocator/dispatcher/configuration.h"
+#include "base/allocator/dispatcher/internal/dispatch_data.h"
+#include "base/allocator/dispatcher/internal/tools.h"
+#include "base/allocator/dispatcher/subsystem.h"
+#include "base/allocator/partition_allocator/partition_alloc_allocation_data.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/check.h"
+#include "base/compiler_specific.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+#include "base/allocator/partition_allocator/shim/allocator_shim.h"
+#endif
+
+#include <tuple>
+
+namespace base::allocator::dispatcher::internal {
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+using allocator_shim::AllocatorDispatch;
+#endif
+
+template <typename CheckObserverPredicate,
+          typename... ObserverTypes,
+          size_t... Indices>
+void inline PerformObserverCheck(const std::tuple<ObserverTypes...>& observers,
+                                 std::index_sequence<Indices...>,
+                                 CheckObserverPredicate check_observer) {
+  ([](bool b) { DCHECK(b); }(check_observer(std::get<Indices>(observers))),
+   ...);
+}
+
+template <typename... ObserverTypes, size_t... Indices>
+ALWAYS_INLINE void PerformAllocationNotification(
+    const std::tuple<ObserverTypes...>& observers,
+    std::index_sequence<Indices...>,
+    const partition_alloc::AllocationNotificationData& notification_data,
+    AllocationSubsystem sub_system) {
+  ((std::get<Indices>(observers)->OnAllocation(
+       notification_data.address(), notification_data.size(), sub_system,
+       notification_data.type_name())),
+   ...);
+}
+
+template <typename... ObserverTypes, size_t... Indices>
+ALWAYS_INLINE void PerformFreeNotification(
+    const std::tuple<ObserverTypes...>& observers,
+    std::index_sequence<Indices...>,
+    const partition_alloc::FreeNotificationData& notification_data) {
+  ((std::get<Indices>(observers)->OnFree(notification_data.address())), ...);
+}
+
+// DispatcherImpl provides hooks into the various memory subsystems. These hooks
+// are responsible for dispatching any notification to the observers.
+// In order to provide as many information on the exact type of the observer and
+// prevent any conditional jumps in the hot allocation path, observers are
+// stored in a std::tuple. DispatcherImpl performs a CHECK at initialization
+// time to ensure they are valid.
+template <typename... ObserverTypes>
+struct DispatcherImpl {
+  using AllObservers = std::index_sequence_for<ObserverTypes...>;
+
+  template <std::enable_if_t<
+                internal::LessEqual(sizeof...(ObserverTypes),
+                                    configuration::kMaximumNumberOfObservers),
+                bool> = true>
+  static DispatchData GetNotificationHooks(
+      std::tuple<ObserverTypes*...> observers) {
+    s_observers = std::move(observers);
+
+    PerformObserverCheck(s_observers, AllObservers{}, IsValidObserver{});
+
+    return CreateDispatchData();
+  }
+
+ private:
+  static DispatchData CreateDispatchData() {
+    return DispatchData()
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+        .SetAllocationObserverHooks(&PartitionAllocatorAllocationHook,
+                                    &PartitionAllocatorFreeHook)
+#endif
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+        .SetAllocatorDispatch(&allocator_dispatch_)
+#endif
+        ;
+  }
+
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+  static void PartitionAllocatorAllocationHook(
+      const partition_alloc::AllocationNotificationData& notification_data) {
+    DoNotifyAllocation(notification_data,
+                       AllocationSubsystem::kPartitionAllocator);
+  }
+
+  static void PartitionAllocatorFreeHook(
+      const partition_alloc::FreeNotificationData& notification_data) {
+    DoNotifyFree(notification_data);
+  }
+#endif
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+  static void* AllocFn(const AllocatorDispatch* self,
+                       size_t size,
+                       void* context) {
+    void* const address = self->next->alloc_function(self->next, size, context);
+
+    DoNotifyAllocation(
+        partition_alloc::AllocationNotificationData(address, size, nullptr),
+        AllocationSubsystem::kAllocatorShim);
+
+    return address;
+  }
+
+  static void* AllocUncheckedFn(const AllocatorDispatch* self,
+                                size_t size,
+                                void* context) {
+    void* const address =
+        self->next->alloc_unchecked_function(self->next, size, context);
+
+    DoNotifyAllocation(
+        partition_alloc::AllocationNotificationData(address, size, nullptr),
+        AllocationSubsystem::kAllocatorShim);
+
+    return address;
+  }
+
+  static void* AllocZeroInitializedFn(const AllocatorDispatch* self,
+                                      size_t n,
+                                      size_t size,
+                                      void* context) {
+    void* const address = self->next->alloc_zero_initialized_function(
+        self->next, n, size, context);
+
+    DoNotifyAllocation(
+        partition_alloc::AllocationNotificationData(address, n * size, nullptr),
+        AllocationSubsystem::kAllocatorShim);
+
+    return address;
+  }
+
+  static void* AllocAlignedFn(const AllocatorDispatch* self,
+                              size_t alignment,
+                              size_t size,
+                              void* context) {
+    void* const address = self->next->alloc_aligned_function(
+        self->next, alignment, size, context);
+
+    DoNotifyAllocation(
+        partition_alloc::AllocationNotificationData(address, size, nullptr),
+        AllocationSubsystem::kAllocatorShim);
+
+    return address;
+  }
+
+  static void* ReallocFn(const AllocatorDispatch* self,
+                         void* address,
+                         size_t size,
+                         void* context) {
+    // Note: size == 0 actually performs free.
+    DoNotifyFree(partition_alloc::FreeNotificationData(address));
+    void* const reallocated_address =
+        self->next->realloc_function(self->next, address, size, context);
+
+    DoNotifyAllocation(partition_alloc::AllocationNotificationData(
+                           reallocated_address, size, nullptr),
+                       AllocationSubsystem::kAllocatorShim);
+
+    return reallocated_address;
+  }
+
+  static void FreeFn(const AllocatorDispatch* self,
+                     void* address,
+                     void* context) {
+    // Note: The RecordFree should be called before free_function (here and in
+    // other places). That is because observers need to handle the allocation
+    // being freed before calling free_function, as once the latter is executed
+    // the address becomes available and can be allocated by another thread.
+    // That would be racy otherwise.
+    DoNotifyFree(partition_alloc::FreeNotificationData(address));
+    self->next->free_function(self->next, address, context);
+  }
+
+  static size_t GetSizeEstimateFn(const AllocatorDispatch* self,
+                                  void* address,
+                                  void* context) {
+    return self->next->get_size_estimate_function(self->next, address, context);
+  }
+
+  static bool ClaimedAddressFn(const AllocatorDispatch* self,
+                               void* address,
+                               void* context) {
+    return self->next->claimed_address_function(self->next, address, context);
+  }
+
+  static unsigned BatchMallocFn(const AllocatorDispatch* self,
+                                size_t size,
+                                void** results,
+                                unsigned num_requested,
+                                void* context) {
+    unsigned const num_allocated = self->next->batch_malloc_function(
+        self->next, size, results, num_requested, context);
+    for (unsigned i = 0; i < num_allocated; ++i) {
+      DoNotifyAllocation(partition_alloc::AllocationNotificationData(
+                             results[i], size, nullptr),
+                         AllocationSubsystem::kAllocatorShim);
+    }
+    return num_allocated;
+  }
+
+  static void BatchFreeFn(const AllocatorDispatch* self,
+                          void** to_be_freed,
+                          unsigned num_to_be_freed,
+                          void* context) {
+    for (unsigned i = 0; i < num_to_be_freed; ++i) {
+      DoNotifyFree(partition_alloc::FreeNotificationData(to_be_freed[i]));
+    }
+
+    self->next->batch_free_function(self->next, to_be_freed, num_to_be_freed,
+                                    context);
+  }
+
+  static void FreeDefiniteSizeFn(const AllocatorDispatch* self,
+                                 void* address,
+                                 size_t size,
+                                 void* context) {
+    DoNotifyFree(partition_alloc::FreeNotificationData(address));
+    self->next->free_definite_size_function(self->next, address, size, context);
+  }
+
+  static void TryFreeDefaultFn(const AllocatorDispatch* self,
+                               void* address,
+                               void* context) {
+    DoNotifyFree(partition_alloc::FreeNotificationData(address));
+    self->next->try_free_default_function(self->next, address, context);
+  }
+
+  static void* AlignedMallocFn(const AllocatorDispatch* self,
+                               size_t size,
+                               size_t alignment,
+                               void* context) {
+    void* const address = self->next->aligned_malloc_function(
+        self->next, size, alignment, context);
+
+    DoNotifyAllocation(
+        partition_alloc::AllocationNotificationData(address, size, nullptr),
+        AllocationSubsystem::kAllocatorShim);
+
+    return address;
+  }
+
+  static void* AlignedReallocFn(const AllocatorDispatch* self,
+                                void* address,
+                                size_t size,
+                                size_t alignment,
+                                void* context) {
+    // Note: size == 0 actually performs free.
+    DoNotifyFree(partition_alloc::FreeNotificationData(address));
+    address = self->next->aligned_realloc_function(self->next, address, size,
+                                                   alignment, context);
+
+    DoNotifyAllocation(
+        partition_alloc::AllocationNotificationData(address, size, nullptr),
+        AllocationSubsystem::kAllocatorShim);
+
+    return address;
+  }
+
+  static void AlignedFreeFn(const AllocatorDispatch* self,
+                            void* address,
+                            void* context) {
+    DoNotifyFree(partition_alloc::FreeNotificationData(address));
+    self->next->aligned_free_function(self->next, address, context);
+  }
+
+  static AllocatorDispatch allocator_dispatch_;
+#endif
+
+  ALWAYS_INLINE static void DoNotifyAllocation(
+      const partition_alloc::AllocationNotificationData& notification_data,
+      AllocationSubsystem sub_system) {
+    PerformAllocationNotification(s_observers, AllObservers{},
+                                  notification_data, sub_system);
+  }
+
+  ALWAYS_INLINE static void DoNotifyFree(
+      const partition_alloc::FreeNotificationData& notification_data) {
+    PerformFreeNotification(s_observers, AllObservers{}, notification_data);
+  }
+
+  static std::tuple<ObserverTypes*...> s_observers;
+};
+
+template <typename... ObserverTypes>
+std::tuple<ObserverTypes*...> DispatcherImpl<ObserverTypes...>::s_observers;
+
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+template <typename... ObserverTypes>
+AllocatorDispatch DispatcherImpl<ObserverTypes...>::allocator_dispatch_ = {
+    &AllocFn,
+    &AllocUncheckedFn,
+    &AllocZeroInitializedFn,
+    &AllocAlignedFn,
+    &ReallocFn,
+    &FreeFn,
+    &GetSizeEstimateFn,
+    &ClaimedAddressFn,
+    &BatchMallocFn,
+    &BatchFreeFn,
+    &FreeDefiniteSizeFn,
+    &TryFreeDefaultFn,
+    &AlignedMallocFn,
+    &AlignedReallocFn,
+    &AlignedFreeFn,
+    nullptr};
+#endif
+
+// Specialization of DispatcherImpl in case we have no observers to notify. In
+// this special case we return a set of null pointers as the Dispatcher must not
+// install any hooks at all.
+template <>
+struct DispatcherImpl<> {
+  static DispatchData GetNotificationHooks(std::tuple<> /*observers*/) {
+    return DispatchData()
+#if BUILDFLAG(USE_PARTITION_ALLOC)
+        .SetAllocationObserverHooks(nullptr, nullptr)
+#endif
+#if BUILDFLAG(USE_ALLOCATOR_SHIM)
+        .SetAllocatorDispatch(nullptr)
+#endif
+        ;
+  }
+};
+
+// A little utility function that helps using DispatcherImpl by providing
+// automated type deduction for templates.
+template <typename... ObserverTypes>
+inline DispatchData GetNotificationHooks(
+    std::tuple<ObserverTypes*...> observers) {
+  return DispatcherImpl<ObserverTypes...>::GetNotificationHooks(
+      std::move(observers));
+}
+
+}  // namespace base::allocator::dispatcher::internal
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_INTERNAL_DISPATCHER_INTERNAL_H_

src/base/allocator/dispatcher/internal/tools.h

@@ -0,0 +1,29 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_INTERNAL_TOOLS_H_
+#define BASE_ALLOCATOR_DISPATCHER_INTERNAL_TOOLS_H_
+
+#include <cstddef>
+
+namespace base::allocator::dispatcher::internal {
+
+constexpr bool LessEqual(size_t lhs, size_t rhs) {
+  return lhs <= rhs;
+}
+
+constexpr bool Equal(size_t lhs, size_t rhs) {
+  return lhs == rhs;
+}
+
+struct IsValidObserver {
+  template <typename T>
+  constexpr bool operator()(T const* ptr) const noexcept {
+    return ptr != nullptr;
+  }
+};
+
+}  // namespace base::allocator::dispatcher::internal
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_INTERNAL_DISPATCHER_H_

src/base/allocator/dispatcher/reentry_guard.cc

@@ -0,0 +1,61 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/dispatcher/reentry_guard.h"
+
+#include "base/check.h"
+#include "base/compiler_specific.h"
+#include "base/debug/crash_logging.h"
+#include "base/strings/string_number_conversions.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_APPLE) || BUILDFLAG(IS_ANDROID)
+#include <pthread.h>
+#endif
+
+namespace base::allocator::dispatcher {
+
+#if BUILDFLAG(IS_APPLE) || BUILDFLAG(IS_ANDROID)
+// pthread_key_t has different signedness on Mac and Android. Store the null
+// value in a strongly-typed constant to avoid "comparison of integers of
+// different signs" warnings when comparing with 0.
+constexpr pthread_key_t kNullKey = 0;
+
+pthread_key_t ReentryGuard::entered_key_ = kNullKey;
+
+void ReentryGuard::InitTLSSlot() {
+  if (entered_key_ == kNullKey) {
+    int error = pthread_key_create(&entered_key_, nullptr);
+    CHECK(!error);
+    // Touch the TLS slot immediately to force any allocations.
+    // TODO(https://crbug.com/1411454): Use this technique to avoid allocations
+    // in PoissonAllocationSampler::ScopedMuteThreadSamples, which will make
+    // ReentryGuard redundant.
+    pthread_setspecific(entered_key_, nullptr);
+  }
+
+  DCHECK_NE(entered_key_, kNullKey);
+}
+
+#else
+
+void ReentryGuard::InitTLSSlot() {}
+
+#endif
+
+void ReentryGuard::RecordTLSSlotToCrashKey() {
+  // Record the key in crash dumps to detect when it's higher than 32
+  // (PTHREAD_KEY_2NDLEVEL_SIZE).
+  // TODO(crbug.com/1411454): Remove this after diagnosing reentry crashes.
+  static auto* const crash_key = base::debug::AllocateCrashKeyString(
+      "reentry_guard_tls_slot", base::debug::CrashKeySize::Size32);
+
+#if BUILDFLAG(IS_APPLE) || BUILDFLAG(IS_ANDROID)
+  base::debug::SetCrashKeyString(crash_key, base::NumberToString(entered_key_));
+#else
+  base::debug::SetCrashKeyString(crash_key, "unused");
+#endif
+}
+
+}  // namespace base::allocator::dispatcher

src/base/allocator/dispatcher/reentry_guard.h

@@ -0,0 +1,75 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_REENTRY_GUARD_H_
+#define BASE_ALLOCATOR_DISPATCHER_REENTRY_GUARD_H_
+
+#include "base/base_export.h"
+#include "base/check.h"
+#include "base/compiler_specific.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_APPLE) || BUILDFLAG(IS_ANDROID)
+#include <pthread.h>
+#endif
+
+namespace base::allocator::dispatcher {
+
+#if BUILDFLAG(IS_APPLE) || BUILDFLAG(IS_ANDROID)
+
+// The macOS implementation of libmalloc sometimes calls malloc recursively,
+// delegating allocations between zones. That causes our hooks being called
+// twice. The scoped guard allows us to detect that.
+//
+// Besides that the implementations of thread_local on macOS and Android
+// seem to allocate memory lazily on the first access to thread_local variables
+// (and on Android at least thread_local is implemented on top of pthread so is
+// strictly worse for performance). Make use of pthread TLS instead of C++
+// thread_local there.
+struct BASE_EXPORT ReentryGuard {
+  ALWAYS_INLINE ReentryGuard() : allowed_(!pthread_getspecific(entered_key_)) {
+    pthread_setspecific(entered_key_, reinterpret_cast<void*>(true));
+  }
+
+  ALWAYS_INLINE ~ReentryGuard() {
+    if (LIKELY(allowed_))
+      pthread_setspecific(entered_key_, nullptr);
+  }
+
+  explicit operator bool() const noexcept { return allowed_; }
+
+  // This function must be called before installing any allocator hooks because
+  // some TLS implementations may allocate (eg. glibc will require a malloc call
+  // to allocate storage for a higher slot number (>= PTHREAD_KEY_2NDLEVEL_SIZE
+  // == 32). This touches the thread-local storage so that any malloc happens
+  // before installing the hooks.
+  static void InitTLSSlot();
+
+  // InitTLSSlot() is called before crash keys are available. At some point
+  // after SetCrashKeyImplementation() is called, this function should be
+  // called to record `entered_key_` to a crash key for debugging. This may
+  // allocate so it must not be called from inside an allocator hook.
+  static void RecordTLSSlotToCrashKey();
+
+ private:
+  static pthread_key_t entered_key_;
+  const bool allowed_;
+};
+
+#else
+
+// Use [[maybe_unused]] as this lightweight stand-in for the more heavyweight
+// ReentryGuard above will otherwise trigger the "unused code" warnings.
+struct [[maybe_unused]] BASE_EXPORT ReentryGuard {
+  constexpr explicit operator bool() const noexcept { return true; }
+
+  static void InitTLSSlot();
+  static void RecordTLSSlotToCrashKey();
+};
+
+#endif
+
+}  // namespace base::allocator::dispatcher
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_REENTRY_GUARD_H_

src/base/allocator/dispatcher/subsystem.h

@@ -0,0 +1,27 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_SUBSYSTEM_H_
+#define BASE_ALLOCATOR_DISPATCHER_SUBSYSTEM_H_
+
+namespace base::allocator::dispatcher {
+
+// Identifiers for the memory subsystem handling the allocation. Some observers
+// require more detailed information on who is performing the allocation, i.e.
+// SamplingHeapProfiler.
+enum class AllocationSubsystem {
+  // Allocation is handled by PartitionAllocator.
+  kPartitionAllocator = 1,
+  // Allocation is handled by AllocatorShims.
+  kAllocatorShim = 2,
+  // Represents a simulated allocation event during testing and is used to
+  // filter out these allocations from real ones.
+  //
+  // Included for backward compatibility, this value becomes obsolete once the
+  // old allocation hooks are removed from PoissonAllocationSampler.
+  kManualForTesting = 3,
+};
+}  // namespace base::allocator::dispatcher
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_SUBSYSTEM_H_

src/base/allocator/dispatcher/testing/dispatcher_test.h

@@ -0,0 +1,27 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_TESTING_DISPATCHER_TEST_H_
+#define BASE_ALLOCATOR_DISPATCHER_TESTING_DISPATCHER_TEST_H_
+
+#include "testing/gtest/include/gtest/gtest.h"
+
+namespace base::allocator::dispatcher::testing {
+
+// DispatcherTest provides some common initialization which most of the
+// unittests of the dispatcher require. DispatcherTest should not be used
+// directly. Instead, derive your test fixture from it.
+struct DispatcherTest : public ::testing::Test {
+  // Perform some commonly required initialization, at them moment
+  // - Initialize the TLS slot for the ReentryGuard
+  DispatcherTest();
+
+ protected:
+  // Protected d'tor only to prevent direct usage of this class.
+  ~DispatcherTest() override;
+};
+
+}  // namespace base::allocator::dispatcher::testing
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_TESTING_DISPATCHER_TEST_H_

src/base/allocator/dispatcher/testing/observer_mock.h

@@ -0,0 +1,32 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_TESTING_OBSERVER_MOCK_H_
+#define BASE_ALLOCATOR_DISPATCHER_TESTING_OBSERVER_MOCK_H_
+
+#include "base/allocator/dispatcher/subsystem.h"
+#include "testing/gmock/include/gmock/gmock.h"
+
+#include <cstddef>
+
+namespace base::allocator::dispatcher::testing {
+
+// ObserverMock is a small mock class based on GoogleMock.
+// It complies to the interface enforced by the dispatcher. The template
+// parameter serves only to create distinct types of observers if required.
+template <typename T = void>
+struct ObserverMock {
+  MOCK_METHOD(void,
+              OnAllocation,
+              (void* address,
+               size_t size,
+               AllocationSubsystem sub_system,
+               const char* type_name),
+              ());
+  MOCK_METHOD(void, OnFree, (void* address), ());
+};
+
+}  // namespace base::allocator::dispatcher::testing
+
+#endif  // BASE_ALLOCATOR_DISPATCHER_TESTING_OBSERVER_MOCK_H_

src/base/allocator/dispatcher/testing/tools.h

@@ -0,0 +1,50 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_TESTING_TOOLS_H_
+#define BASE_ALLOCATOR_DISPATCHER_TESTING_TOOLS_H_
+
+#include <array>
+#include <tuple>
+#include <utility>
+
+namespace base::allocator::dispatcher::testing {
+
+namespace internal {
+template <size_t Size, typename Type, typename... AppendedTypes>
+struct DefineTupleFromSingleType {
+  using type = typename DefineTupleFromSingleType<Size - 1,
+                                                  Type,
+                                                  AppendedTypes...,
+                                                  Type>::type;
+};
+
+template <typename Type, typename... AppendedTypes>
+struct DefineTupleFromSingleType<0, Type, AppendedTypes...> {
+  using type = std::tuple<AppendedTypes...>;
+};
+
+}  // namespace internal
+
+template <size_t Size, typename Type>
+struct DefineTupleFromSingleType {
+  using type = typename internal::DefineTupleFromSingleType<Size, Type>::type;
+};
+
+template <typename Type, size_t Size, size_t... Indices>
+typename internal::DefineTupleFromSingleType<Size, Type*>::type
+CreateTupleOfPointers(std::array<Type, Size>& items,
+                      std::index_sequence<Indices...>) {
+  return std::make_tuple((&items[Indices])...);
+}
+
+template <typename Type, size_t Size>
+typename internal::DefineTupleFromSingleType<Size, Type*>::type
+CreateTupleOfPointers(std::array<Type, Size>& items) {
+  return CreateTupleOfPointers(items, std::make_index_sequence<Size>{});
+}
+
+}  // namespace base::allocator::dispatcher::testing
+
+#endif

src/base/allocator/dispatcher/tls.cc

@@ -0,0 +1,100 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/dispatcher/tls.h"
+
+#if USE_LOCAL_TLS_EMULATION()
+
+#include "base/check.h"
+#include "base/dcheck_is_on.h"
+#include "base/immediate_crash.h"
+#include "build/build_config.h"
+
+#include <sys/mman.h>
+
+#if BUILDFLAG(IS_ANDROID) || BUILDFLAG(IS_LINUX)
+#include <sys/prctl.h>
+#endif
+
+namespace base::allocator::dispatcher::internal {
+
+void* MMapAllocator::AllocateMemory(size_t size_in_bytes) {
+  void* const mmap_res = mmap(nullptr, size_in_bytes, PROT_READ | PROT_WRITE,
+                              MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+#if BUILDFLAG(IS_ANDROID) || BUILDFLAG(IS_LINUX)
+#if defined(PR_SET_VMA) && defined(PR_SET_VMA_ANON_NAME)
+  if (mmap_res != MAP_FAILED) {
+    // Allow the anonymous memory region allocated by mmap(MAP_ANONYMOUS) to
+    // be identified in /proc/$PID/smaps.  This helps improve visibility into
+    // Chromium's memory usage on Android.
+    prctl(PR_SET_VMA, PR_SET_VMA_ANON_NAME, mmap_res, size_in_bytes,
+          "tls-mmap-allocator");
+  }
+#endif
+#endif
+
+  return (mmap_res != MAP_FAILED) ? mmap_res : nullptr;
+}
+
+bool MMapAllocator::FreeMemoryForTesting(void* pointer_to_allocated,
+                                         size_t size_in_bytes) {
+  auto const munmap_res = munmap(pointer_to_allocated, size_in_bytes);
+  return (munmap_res == 0);
+}
+
+bool PThreadTLSSystem::Setup(
+    OnThreadTerminationFunction thread_termination_function) {
+#if DCHECK_IS_ON()
+  // Initialize must happen outside of the allocation path. Therefore, it is
+  // secure to verify with DCHECK.
+  DCHECK(!initialized_.exchange(true, std::memory_order_acq_rel));
+#endif
+
+  auto const key_create_res =
+      pthread_key_create(&data_access_key_, thread_termination_function);
+
+  // On some platforms creating a new pthread-key requires an allocation when a
+  // given number of keys has been created. I.e. in glibc this limit is denoted
+  // by PTHREAD_KEY_2NDLEVEL_SIZE. However, this value is neither present on all
+  // systems nor accessible from here. Hence, we do not do any checks here.
+  // However, we strongly recommend to setup the TLS system as early as possible
+  // to avoid exceeding this limit.
+
+  return (0 == key_create_res);
+}
+
+bool PThreadTLSSystem::TearDownForTesting() {
+#if DCHECK_IS_ON()
+  // TearDownForTesting must happen outside of the allocation path. Therefore,
+  // it is secure to verify with DCHECK.
+  DCHECK(initialized_.exchange(false, std::memory_order_acq_rel));
+#endif
+
+  auto const key_delete_res = pthread_key_delete(data_access_key_);
+  return (0 == key_delete_res);
+}
+
+void* PThreadTLSSystem::GetThreadSpecificData() {
+#if DCHECK_IS_ON()
+  if (!initialized_.load(std::memory_order_acquire)) {
+    return nullptr;
+  }
+#endif
+
+  return pthread_getspecific(data_access_key_);
+}
+
+bool PThreadTLSSystem::SetThreadSpecificData(void* data) {
+#if DCHECK_IS_ON()
+  if (!initialized_.load(std::memory_order_acquire)) {
+    return false;
+  }
+#endif
+
+  return (0 == pthread_setspecific(data_access_key_, data));
+}
+
+}  // namespace base::allocator::dispatcher::internal
+
+#endif  // USE_LOCAL_TLS_EMULATION()

src/base/allocator/dispatcher/tls.h

@@ -0,0 +1,432 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_DISPATCHER_TLS_H_
+#define BASE_ALLOCATOR_DISPATCHER_TLS_H_
+
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_POSIX)  // the current allocation mechanism (mmap) and TLS
+                         // support (pthread) are both defined by POSIX
+#define USE_LOCAL_TLS_EMULATION() true
+#else
+#define USE_LOCAL_TLS_EMULATION() false
+#endif
+
+#if USE_LOCAL_TLS_EMULATION()
+#include <algorithm>
+#include <atomic>
+#include <memory>
+#include <mutex>
+
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+#include "base/base_export.h"
+#include "base/check.h"
+#include "base/compiler_specific.h"
+
+#include <pthread.h>
+
+#if HAS_FEATURE(thread_sanitizer)
+#define DISABLE_TSAN_INSTRUMENTATION __attribute__((no_sanitize("thread")))
+#else
+#define DISABLE_TSAN_INSTRUMENTATION
+#endif
+
+namespace base::allocator::dispatcher {
+namespace internal {
+
+// Allocate memory using POSIX' mmap and unmap functionality. The allocator
+// implements the allocator interface required by ThreadLocalStorage.
+struct BASE_EXPORT MMapAllocator {
+// The minimum size of a memory chunk when allocating. Even for chunks with
+// fewer bytes, at least AllocationChunkSize bytes are allocated. For mmap, this
+// is usually the page size of the system.
+// For various OS-CPU combinations, partition_alloc::PartitionPageSize() is not
+// constexpr. Hence, we can not use this value but define it locally.
+#if defined(PAGE_ALLOCATOR_CONSTANTS_ARE_CONSTEXPR) && \
+    PAGE_ALLOCATOR_CONSTANTS_ARE_CONSTEXPR
+  constexpr static size_t AllocationChunkSize =
+      partition_alloc::PartitionPageSize();
+#elif BUILDFLAG(IS_APPLE)
+  constexpr static size_t AllocationChunkSize = 16384;
+#elif BUILDFLAG(IS_LINUX) && defined(ARCH_CPU_ARM64)
+  constexpr static size_t AllocationChunkSize = 16384;
+#else
+  constexpr static size_t AllocationChunkSize = 4096;
+#endif
+
+  // Allocate size_in_bytes bytes of raw memory. Return nullptr if allocation
+  // fails.
+  void* AllocateMemory(size_t size_in_bytes);
+  // Free the raw memory pointed to by pointer_to_allocated. Returns a boolean
+  // value indicating if the free was successful.
+  bool FreeMemoryForTesting(void* pointer_to_allocated, size_t size_in_bytes);
+};
+
+// The allocator used by default for the thread local storage.
+using DefaultAllocator = MMapAllocator;
+
+using OnThreadTerminationFunction = void (*)(void*);
+
+// The TLS system used by default for the thread local storage. It stores and
+// retrieves thread specific data pointers.
+struct BASE_EXPORT PThreadTLSSystem {
+  // Initialize the TLS system to store a data set for different threads.
+  // @param thread_termination_function An optional function which will be
+  // invoked upon termination of a thread.
+  bool Setup(OnThreadTerminationFunction thread_termination_function);
+  // Tear down the TLS system. After completing tear down, the thread
+  // termination function passed to Setup will not be invoked anymore.
+  bool TearDownForTesting();
+
+  // Get the pointer to the data associated to the current thread. Returns
+  // nullptr if the TLS system is not initialized or no data was set before.
+  void* GetThreadSpecificData();
+  // Set the pointer to the data associated to the current thread. Return true
+  // if stored successfully, false otherwise.
+  bool SetThreadSpecificData(void* data);
+
+ private:
+  pthread_key_t data_access_key_ = 0;
+#if DCHECK_IS_ON()
+  // From POSIX standard at https://www.open-std.org/jtc1/sc22/open/n4217.pdf:
+  // The effect of calling pthread_getspecific() or pthread_setspecific() with a
+  // key value not obtained from pthread_key_create() or after key has been
+  // deleted with pthread_key_delete() is undefined.
+  //
+  // Unfortunately, POSIX doesn't define a special value of pthread_key_t
+  // indicating an invalid key which would allow us to detect accesses outside
+  // of initialized state. Hence, to prevent us from drifting into the evil
+  // realm of undefined behaviour we store whether we're somewhere between Setup
+  // and Teardown.
+  std::atomic_bool initialized_{false};
+#endif
+};
+
+using DefaultTLSSystem = PThreadTLSSystem;
+
+// In some scenarios, most notably when testing, the allocator and TLS system
+// passed to |ThreadLocalStorage| are not copyable and have to be wrapped, i.e.
+// using std::reference_wrapper. |dereference| is a small helper to retrieve the
+// underlying value.
+template <typename T>
+T& dereference(T& ref) {
+  return ref;
+}
+
+template <typename T>
+T& dereference(std::reference_wrapper<T>& ref) {
+  // std::reference_wrapper requires a valid reference for construction,
+  // therefore, no need in checking here.
+  return ref.get();
+}
+
+// Store thread local data. The data is organized in chunks, where each chunk
+// holds |ItemsPerChunk|. Each item may be free or used.
+//
+// When a thread requests data, the chunks are searched for a free data item,
+// which is registered for this thread and marked as |used|. Further requests by
+// this thread will then always return the same item. When a thread terminates,
+// the item will be reset and return to the pool of free items.
+//
+// Upon construction, the first chunk is created. If a thread requests data and
+// there is no free item available, another chunk is created. Upon destruction,
+// all memory is freed. Pointers to data items become invalid!
+//
+// Constructor and destructor are not thread safe.
+//
+// @tparam PayloadType The item type to be stored.
+// @tparam AllocatorType The allocator being used. An allocator must provide
+// the following interface:
+//  void* AllocateMemory(size_t size_in_bytes); // Allocate size_in_bytes bytes
+//  of raw memory.
+//  void FreeMemory(void* pointer_to_allocated, size_t size_in_bytes); // Free
+//  the raw memory pointed to by pointer_to_allocated.
+// Any failure in allocation or free must terminate the process.
+// @tparam TLSSystemType The TLS system being used. A TLS system must provide
+// the following interface:
+//  bool Setup(OnThreadTerminationFunction thread_termination_function);
+//  bool Destroy();
+//  void* GetThreadSpecificData();
+//  bool SetThreadSpecificData(void* data);
+// @tparam AllocationChunkSize The minimum size of a memory chunk that the
+// allocator can handle. We try to size the chunks so that each chunk uses this
+// size to the maximum.
+// @tparam IsDestructibleForTesting For testing purposes we allow the destructor
+// to perform clean up upon destruction. Otherwise, using the destructor will
+// result in a compilation failure.
+template <typename PayloadType,
+          typename AllocatorType,
+          typename TLSSystemType,
+          size_t AllocationChunkSize,
+          bool IsDestructibleForTesting>
+struct ThreadLocalStorage {
+  ThreadLocalStorage() : root_(AllocateAndInitializeChunk()) { Initialize(); }
+
+  // Create a new instance of |ThreadLocalStorage| using the passed allocator
+  // and TLS system. This initializes the underlying TLS system and creates the
+  // first chunk of data.
+  ThreadLocalStorage(AllocatorType allocator, TLSSystemType tlsSystem)
+      : allocator_(std::move(allocator)),
+        tls_system_(std::move(tlsSystem)),
+        root_(AllocateAndInitializeChunk()) {
+    Initialize();
+  }
+
+  // Deletes an instance of |ThreadLocalStorage| and delete all the data chunks
+  // created.
+  ~ThreadLocalStorage() {
+    if constexpr (IsDestructibleForTesting) {
+      TearDownForTesting();
+    } else if constexpr (!IsDestructibleForTesting) {
+      static_assert(
+          IsDestructibleForTesting,
+          "ThreadLocalStorage cannot be destructed outside of test code.");
+    }
+  }
+
+  // Explicitly prevent all forms of Copy/Move construction/assignment. For an
+  // exact copy of ThreadLocalStorage we would need to copy the mapping of
+  // thread to item, which we can't do at the moment. On the other side, our
+  // atomic members do not support moving out of the box.
+  ThreadLocalStorage(const ThreadLocalStorage&) = delete;
+  ThreadLocalStorage(ThreadLocalStorage&& other) = delete;
+  ThreadLocalStorage& operator=(const ThreadLocalStorage&) = delete;
+  ThreadLocalStorage& operator=(ThreadLocalStorage&&) = delete;
+
+  // Get the data item for the current thread. If no data is registered so far,
+  // find a free item in the chunks and register it for the current thread.
+  PayloadType* GetThreadLocalData() {
+    auto& tls_system = dereference(tls_system_);
+
+    auto* slot = static_cast<SingleSlot*>(tls_system.GetThreadSpecificData());
+
+    if (UNLIKELY(slot == nullptr)) {
+      slot = FindAndAllocateFreeSlot(root_.load(std::memory_order_relaxed));
+
+      // We might be called in the course of handling a memory allocation. We do
+      // not use CHECK since they might allocate and cause a recursion.
+      RAW_CHECK(tls_system.SetThreadSpecificData(slot));
+
+      // Reset the content to wipe out any previous data.
+      Reset(slot->item);
+    }
+
+    return &(slot->item);
+  }
+
+ private:
+  // Encapsulate the payload item and some administrative data.
+  struct SingleSlot {
+    PayloadType item;
+#if !defined(__cpp_lib_atomic_value_initialization) || \
+    __cpp_lib_atomic_value_initialization < 201911L
+    std::atomic_flag is_used = ATOMIC_FLAG_INIT;
+#else
+    std::atomic_flag is_used;
+#endif
+  };
+
+  template <size_t NumberOfItems>
+  struct ChunkT {
+    SingleSlot slots[NumberOfItems];
+    // Pointer to the next chunk.
+    std::atomic<ChunkT*> next_chunk = nullptr;
+    // Helper flag to ensure we create the next chunk only once in a multi
+    // threaded environment.
+    std::once_flag create_next_chunk_flag;
+  };
+
+  template <size_t LowerNumberOfItems,
+            size_t UpperNumberOfItems,
+            size_t NumberOfBytes>
+  static constexpr size_t CalculateEffectiveNumberOfItemsBinSearch() {
+    if constexpr (LowerNumberOfItems == UpperNumberOfItems) {
+      return LowerNumberOfItems;
+    }
+
+    constexpr size_t CurrentNumberOfItems =
+        (UpperNumberOfItems - LowerNumberOfItems) / 2 + LowerNumberOfItems;
+
+    if constexpr (sizeof(ChunkT<CurrentNumberOfItems>) > NumberOfBytes) {
+      return CalculateEffectiveNumberOfItemsBinSearch<
+          LowerNumberOfItems, CurrentNumberOfItems, NumberOfBytes>();
+    }
+
+    if constexpr (sizeof(ChunkT<CurrentNumberOfItems + 1>) < NumberOfBytes) {
+      return CalculateEffectiveNumberOfItemsBinSearch<
+          CurrentNumberOfItems + 1, UpperNumberOfItems, NumberOfBytes>();
+    }
+
+    return CurrentNumberOfItems;
+  }
+
+  // Calculate the maximum number of items we can store in one chunk without the
+  // size of the chunk exceeding NumberOfBytes. To avoid things like alignment
+  // and packing tampering with the calculation, instead of calculating the
+  // correct number of items we use sizeof-operator against ChunkT to search for
+  // the correct size. Unfortunately, the number of recursions is limited by the
+  // compiler. Therefore, we use a binary search instead of a simple linear
+  // search.
+  template <size_t MinimumNumberOfItems, size_t NumberOfBytes>
+  static constexpr size_t CalculateEffectiveNumberOfItems() {
+    if constexpr (sizeof(ChunkT<MinimumNumberOfItems>) < NumberOfBytes) {
+      constexpr size_t LowerNumberOfItems = MinimumNumberOfItems;
+      constexpr size_t UpperNumberOfItems =
+          NumberOfBytes / sizeof(PayloadType) + 1;
+      return CalculateEffectiveNumberOfItemsBinSearch<
+          LowerNumberOfItems, UpperNumberOfItems, NumberOfBytes>();
+    }
+
+    return MinimumNumberOfItems;
+  }
+
+ public:
+  // The minimum number of items per chunk. It should be high enough to
+  // accommodate most items in the root chunk whilst not wasting to much space
+  // on unnecessary items.
+  static constexpr size_t MinimumNumberOfItemsPerChunk = 75;
+  // The effective number of items per chunk. We use the AllocationChunkSize as
+  // a hint to calculate to effective number of items so we occupy one of these
+  // memory chunks to the maximum extent possible.
+  static constexpr size_t ItemsPerChunk =
+      CalculateEffectiveNumberOfItems<MinimumNumberOfItemsPerChunk,
+                                      AllocationChunkSize>();
+
+ private:
+  using Chunk = ChunkT<ItemsPerChunk>;
+
+  static_assert(ItemsPerChunk >= MinimumNumberOfItemsPerChunk);
+
+  // Mark an item's slot ready for reuse. This function is used as thread
+  // termination function in the TLS system. We do not destroy anything at this
+  // point but simply mark the slot as unused.
+  static void MarkSlotAsFree(void* data) {
+    // We always store SingleSlots in the TLS system. Therefore, we cast to
+    // SingleSlot and reset the is_used flag.
+    auto* const slot = static_cast<SingleSlot*>(data);
+
+    // We might be called in the course of handling a memory allocation. We do
+    // not use CHECK since they might allocate and cause a recursion.
+    RAW_CHECK(slot && slot->is_used.test_and_set());
+
+    slot->is_used.clear(std::memory_order_relaxed);
+  }
+
+  // Perform common initialization during construction of an instance.
+  void Initialize() {
+    // The constructor must be called outside of the allocation path. Therefore,
+    // it is secure to verify with CHECK.
+
+    // Passing MarkSlotAsFree as thread_termination_function we ensure the
+    // slot/item assigned to the finished thread will be returned to the pool of
+    // unused items.
+    CHECK(dereference(tls_system_).Setup(&MarkSlotAsFree));
+  }
+
+  Chunk* AllocateAndInitializeChunk() {
+    void* const uninitialized_memory =
+        dereference(allocator_).AllocateMemory(sizeof(Chunk));
+
+    // We might be called in the course of handling a memory allocation. We do
+    // not use CHECK since they might allocate and cause a recursion.
+    RAW_CHECK(uninitialized_memory != nullptr);
+
+    return new (uninitialized_memory) Chunk{};
+  }
+
+  void FreeAndDeallocateChunkForTesting(Chunk* chunk_to_erase) {
+    chunk_to_erase->~Chunk();
+
+    // FreeAndDeallocateChunkForTesting must be called outside of the allocation
+    // path. Therefore, it is secure to verify with CHECK.
+    CHECK(dereference(allocator_)
+              .FreeMemoryForTesting(chunk_to_erase, sizeof(Chunk)));
+  }
+
+  // Find a free slot in the passed chunk, reserve it and return it to the
+  // caller. If no free slot can be found, head on to the next chunk. If the
+  // next chunk doesn't exist, create it.
+  SingleSlot* FindAndAllocateFreeSlot(Chunk* const chunk) {
+    SingleSlot* const slot = std::find_if_not(
+        std::begin(chunk->slots), std::end(chunk->slots),
+        [](SingleSlot& candidate_slot) {
+          return candidate_slot.is_used.test_and_set(std::memory_order_relaxed);
+        });
+
+    // So we found a slot. Happily return it to the caller.
+    if (slot != std::end(chunk->slots)) {
+      return slot;
+    }
+
+    // Ok, there are no more free slots in this chunk. First, ensure the next
+    // chunk is valid and create one if necessary.
+    std::call_once(chunk->create_next_chunk_flag, [&] {
+      // From https://eel.is/c++draft/thread.once.callonce#3
+      //
+      // Synchronization: For any given once_­flag: all active executions occur
+      // in a total order; completion of an active execution synchronizes with
+      // the start of the next one in this total order; and the returning
+      // execution synchronizes with the return from all passive executions.
+      //
+      // Therefore, we do only a relaxed store here, call_once synchronizes with
+      // other threads.
+      chunk->next_chunk.store(AllocateAndInitializeChunk(),
+                              std::memory_order_relaxed);
+    });
+
+    return FindAndAllocateFreeSlot(chunk->next_chunk);
+  }
+
+  template <bool IsDestructibleForTestingP = IsDestructibleForTesting>
+  typename std::enable_if<IsDestructibleForTestingP>::type
+  TearDownForTesting() {
+    // The destructor must be called outside of the allocation path. Therefore,
+    // it is secure to verify with CHECK.
+
+    // All accessing threads must be terminated by now. For additional security
+    // we tear down the TLS system first. This way we ensure that
+    // MarkSlotAsFree is not called anymore and we have no accesses from the
+    // TLS system's side.
+    CHECK(dereference(tls_system_).TearDownForTesting());
+
+    // Delete all data chunks.
+    for (auto* chunk = root_.load(); chunk != nullptr;) {
+      auto* next_chunk = chunk->next_chunk.load();
+      FreeAndDeallocateChunkForTesting(chunk);
+      chunk = next_chunk;
+    }
+  }
+
+  // Reset a single item to its default value.
+  // Since items are re-used, they may be accessed from different threads,
+  // causing TSan to trigger. Therefore, the reset is exempt from TSan
+  // instrumentation.
+  DISABLE_TSAN_INSTRUMENTATION void Reset(PayloadType& item) { item = {}; }
+
+  AllocatorType allocator_;
+  TLSSystemType tls_system_;
+  std::atomic<Chunk*> const root_;
+};
+
+}  // namespace internal
+
+// The ThreadLocalStorage visible to the user. This uses the internal default
+// allocator and TLS system.
+template <typename StorageType,
+          typename AllocatorType = internal::DefaultAllocator,
+          typename TLSSystemType = internal::DefaultTLSSystem,
+          size_t AllocationChunkSize = AllocatorType::AllocationChunkSize,
+          bool IsDestructibleForTesting = false>
+using ThreadLocalStorage =
+    internal::ThreadLocalStorage<StorageType,
+                                 AllocatorType,
+                                 TLSSystemType,
+                                 AllocationChunkSize,
+                                 IsDestructibleForTesting>;
+
+}  // namespace base::allocator::dispatcher
+
+#endif  // USE_LOCAL_TLS_EMULATION()
+#endif  // BASE_ALLOCATOR_DISPATCHER_TLS_H_

src/base/allocator/early_zone_registration_apple.cc

@@ -0,0 +1,266 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/early_zone_registration_apple.h"
+
+#include <mach/mach.h>
+#include <malloc/malloc.h>
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/shim/early_zone_registration_constants.h"
+
+// BASE_EXPORT tends to be defined as soon as anything from //base is included.
+#if defined(BASE_EXPORT)
+#error "This file cannot depend on //base"
+#endif
+
+namespace partition_alloc {
+
+#if !BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+
+void EarlyMallocZoneRegistration() {}
+void AllowDoublePartitionAllocZoneRegistration() {}
+
+#else
+
+extern "C" {
+// abort_report_np() records the message in a special section that both the
+// system CrashReporter and Crashpad collect in crash reports. See also in
+// chrome_exe_main_mac.cc.
+void abort_report_np(const char* fmt, ...);
+}
+
+namespace {
+
+malloc_zone_t* GetDefaultMallocZone() {
+  // malloc_default_zone() does not return... the default zone, but the
+  // initial one. The default one is the first element of the default zone
+  // array.
+  unsigned int zone_count = 0;
+  vm_address_t* zones = nullptr;
+  kern_return_t result = malloc_get_all_zones(
+      mach_task_self(), /*reader=*/nullptr, &zones, &zone_count);
+  if (result != KERN_SUCCESS) {
+    abort_report_np("Cannot enumerate malloc() zones");
+  }
+  return reinterpret_cast<malloc_zone_t*>(zones[0]);
+}
+
+}  // namespace
+
+void EarlyMallocZoneRegistration() {
+  // Must have static storage duration, as raw pointers are passed to
+  // libsystem_malloc.
+  static malloc_zone_t g_delegating_zone;
+  static malloc_introspection_t g_delegating_zone_introspect;
+  static malloc_zone_t* g_default_zone;
+
+  // Make sure that the default zone is instantiated.
+  malloc_zone_t* purgeable_zone = malloc_default_purgeable_zone();
+
+  g_default_zone = GetDefaultMallocZone();
+
+  // The delegating zone:
+  // - Forwards all allocations to the existing default zone
+  // - Does *not* claim to own any memory, meaning that it will always be
+  //   skipped in free() in libsystem_malloc.dylib.
+  //
+  // This is a temporary zone, until it gets replaced by PartitionAlloc, inside
+  // the main library. Since the main library depends on many external
+  // libraries, we cannot install PartitionAlloc as the default zone without
+  // concurrency issues.
+  //
+  // Instead, what we do is here, while the process is single-threaded:
+  // - Register the delegating zone as the default one.
+  // - Set the original (libsystem_malloc's) one as the second zone
+  //
+  // Later, when PartitionAlloc initializes, we replace the default (delegating)
+  // zone with ours. The end state is:
+  // 1. PartitionAlloc zone
+  // 2. libsystem_malloc zone
+
+  // Set up of the delegating zone. Note that it doesn't just forward calls to
+  // the default zone. This is because the system zone's malloc_zone_t pointer
+  // actually points to a larger struct, containing allocator metadata. So if we
+  // pass as the first parameter the "simple" delegating zone pointer, then we
+  // immediately crash inside the system zone functions. So we need to replace
+  // the zone pointer as well.
+  //
+  // Calls fall into 4 categories:
+  // - Allocation calls: forwarded to the real system zone
+  // - "Is this pointer yours" calls: always answer no
+  // - free(): Should never be called, but is in practice, see comments below.
+  // - Diagnostics and debugging: these are typically called for every
+  //   zone. They are no-ops for us, as we don't want to double-count, or lock
+  //   the data structures of the real zone twice.
+
+  // Allocation: Forward to the real zone.
+  g_delegating_zone.malloc = [](malloc_zone_t* zone, size_t size) {
+    return g_default_zone->malloc(g_default_zone, size);
+  };
+  g_delegating_zone.calloc = [](malloc_zone_t* zone, size_t num_items,
+                                size_t size) {
+    return g_default_zone->calloc(g_default_zone, num_items, size);
+  };
+  g_delegating_zone.valloc = [](malloc_zone_t* zone, size_t size) {
+    return g_default_zone->valloc(g_default_zone, size);
+  };
+  g_delegating_zone.realloc = [](malloc_zone_t* zone, void* ptr, size_t size) {
+    return g_default_zone->realloc(g_default_zone, ptr, size);
+  };
+  g_delegating_zone.batch_malloc = [](malloc_zone_t* zone, size_t size,
+                                      void** results, unsigned num_requested) {
+    return g_default_zone->batch_malloc(g_default_zone, size, results,
+                                        num_requested);
+  };
+  g_delegating_zone.memalign = [](malloc_zone_t* zone, size_t alignment,
+                                  size_t size) {
+    return g_default_zone->memalign(g_default_zone, alignment, size);
+  };
+
+  // Does ptr belong to this zone? Return value is != 0 if so.
+  g_delegating_zone.size = [](malloc_zone_t* zone, const void* ptr) -> size_t {
+    return 0;
+  };
+
+  // Free functions.
+  // The normal path for freeing memory is:
+  // 1. Try all zones in order, call zone->size(ptr)
+  // 2. If zone->size(ptr) != 0, call zone->free(ptr) (or free_definite_size)
+  // 3. If no zone matches, crash.
+  //
+  // Since this zone always returns 0 in size() (see above), then zone->free()
+  // should never be called. Unfortunately, this is not the case, as some places
+  // in CoreFoundation call malloc_zone_free(zone, ptr) directly. So rather than
+  // crashing, forward the call. It's the caller's responsibility to use the
+  // same zone for free() as for the allocation (this is in the contract of
+  // malloc_zone_free()).
+  //
+  // However, note that the sequence of calls size() -> free() is not possible
+  // for this zone, as size() always returns 0.
+  g_delegating_zone.free = [](malloc_zone_t* zone, void* ptr) {
+    return g_default_zone->free(g_default_zone, ptr);
+  };
+  g_delegating_zone.free_definite_size = [](malloc_zone_t* zone, void* ptr,
+                                            size_t size) {
+    return g_default_zone->free_definite_size(g_default_zone, ptr, size);
+  };
+  g_delegating_zone.batch_free = [](malloc_zone_t* zone, void** to_be_freed,
+                                    unsigned num_to_be_freed) {
+    return g_default_zone->batch_free(g_default_zone, to_be_freed,
+                                      num_to_be_freed);
+  };
+#if PA_TRY_FREE_DEFAULT_IS_AVAILABLE
+  g_delegating_zone.try_free_default = [](malloc_zone_t* zone, void* ptr) {
+    return g_default_zone->try_free_default(g_default_zone, ptr);
+  };
+#endif
+
+  // Diagnostics and debugging.
+  //
+  // Do nothing to reduce memory footprint, the real
+  // zone will do it.
+  g_delegating_zone.pressure_relief = [](malloc_zone_t* zone,
+                                         size_t goal) -> size_t { return 0; };
+
+  // Introspection calls are not all optional, for instance locking and
+  // unlocking before/after fork() is not optional.
+  //
+  // Nothing to enumerate.
+  g_delegating_zone_introspect.enumerator =
+      [](task_t task, void*, unsigned type_mask, vm_address_t zone_address,
+         memory_reader_t reader,
+         vm_range_recorder_t recorder) -> kern_return_t {
+    return KERN_SUCCESS;
+  };
+  // Need to provide a real implementation, it is used for e.g. array sizing.
+  g_delegating_zone_introspect.good_size = [](malloc_zone_t* zone,
+                                              size_t size) {
+    return g_default_zone->introspect->good_size(g_default_zone, size);
+  };
+  // Nothing to do.
+  g_delegating_zone_introspect.check = [](malloc_zone_t* zone) -> boolean_t {
+    return true;
+  };
+  g_delegating_zone_introspect.print = [](malloc_zone_t* zone,
+                                          boolean_t verbose) {};
+  g_delegating_zone_introspect.log = [](malloc_zone_t*, void*) {};
+  // Do not forward the lock / unlock calls. Since the default zone is still
+  // there, we should not lock here, as it would lock the zone twice (all
+  // zones are locked before fork().). Rather, do nothing, since this fake
+  // zone does not need any locking.
+  g_delegating_zone_introspect.force_lock = [](malloc_zone_t* zone) {};
+  g_delegating_zone_introspect.force_unlock = [](malloc_zone_t* zone) {};
+  g_delegating_zone_introspect.reinit_lock = [](malloc_zone_t* zone) {};
+  // No stats.
+  g_delegating_zone_introspect.statistics = [](malloc_zone_t* zone,
+                                               malloc_statistics_t* stats) {};
+  // We are not locked.
+  g_delegating_zone_introspect.zone_locked =
+      [](malloc_zone_t* zone) -> boolean_t { return false; };
+  // Don't support discharge checking.
+  g_delegating_zone_introspect.enable_discharge_checking =
+      [](malloc_zone_t* zone) -> boolean_t { return false; };
+  g_delegating_zone_introspect.disable_discharge_checking =
+      [](malloc_zone_t* zone) {};
+  g_delegating_zone_introspect.discharge = [](malloc_zone_t* zone,
+                                              void* memory) {};
+
+  // Could use something lower to support fewer functions, but this is
+  // consistent with the real zone installed by PartitionAlloc.
+  g_delegating_zone.version = allocator_shim::kZoneVersion;
+  g_delegating_zone.introspect = &g_delegating_zone_introspect;
+  // This name is used in PartitionAlloc's initialization to determine whether
+  // it should replace the delegating zone.
+  g_delegating_zone.zone_name = allocator_shim::kDelegatingZoneName;
+
+  // Register puts the new zone at the end, unregister swaps the new zone with
+  // the last one.
+  // The zone array is, after these lines, in order:
+  // 1. |g_default_zone|...|g_delegating_zone|
+  // 2. |g_delegating_zone|...|  (no more default)
+  // 3. |g_delegating_zone|...|g_default_zone|
+  malloc_zone_register(&g_delegating_zone);
+  malloc_zone_unregister(g_default_zone);
+  malloc_zone_register(g_default_zone);
+
+  // Make sure that the purgeable zone is after the default one.
+  // Will make g_default_zone take the purgeable zone spot
+  malloc_zone_unregister(purgeable_zone);
+  // Add back the purgeable zone as the last one.
+  malloc_zone_register(purgeable_zone);
+
+  // Final configuration:
+  // |g_delegating_zone|...|g_default_zone|purgeable_zone|
+
+  // Sanity check.
+  if (GetDefaultMallocZone() != &g_delegating_zone) {
+    abort_report_np("Failed to install the delegating zone as default.");
+  }
+}
+
+void AllowDoublePartitionAllocZoneRegistration() {
+  unsigned int zone_count = 0;
+  vm_address_t* zones = nullptr;
+  kern_return_t result = malloc_get_all_zones(
+      mach_task_self(), /*reader=*/nullptr, &zones, &zone_count);
+  if (result != KERN_SUCCESS) {
+    abort_report_np("Cannot enumerate malloc() zones");
+  }
+
+  // If PartitionAlloc is one of the zones, *change* its name so that
+  // registration can happen multiple times. This works because zone
+  // registration only keeps a pointer to the struct, it does not copy the data.
+  for (unsigned int i = 0; i < zone_count; i++) {
+    malloc_zone_t* zone = reinterpret_cast<malloc_zone_t*>(zones[i]);
+    if (zone->zone_name &&
+        strcmp(zone->zone_name, allocator_shim::kPartitionAllocZoneName) == 0) {
+      zone->zone_name = "RenamedPartitionAlloc";
+      break;
+    }
+  }
+}
+
+#endif  // BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+}  // namespace partition_alloc

src/base/allocator/early_zone_registration_apple.h

@@ -0,0 +1,29 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_EARLY_ZONE_REGISTRATION_APPLE_H_
+#define BASE_ALLOCATOR_EARLY_ZONE_REGISTRATION_APPLE_H_
+
+// This is an Apple-only file, used to register PartitionAlloc's zone *before*
+// the process becomes multi-threaded.
+
+namespace partition_alloc {
+
+// Must be called *once*, *before* the process becomes multi-threaded.
+void EarlyMallocZoneRegistration();
+
+// Tricks the registration code to believe that PartitionAlloc was not already
+// registered. This allows a future library load to register PartitionAlloc's
+// zone as well, rather than bailing out.
+//
+// This is mutually exclusive with EarlyMallocZoneRegistration(), and should
+// ideally be removed. Indeed, by allowing two zones to be registered, we still
+// end up with a split heap, and more memory usage.
+//
+// This is a hack for https://crbug.com/1274236.
+void AllowDoublePartitionAllocZoneRegistration();
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_EARLY_ZONE_REGISTRATION_APPLE_H_

src/base/allocator/partition_alloc_features.cc

@@ -0,0 +1,408 @@
+// Copyright 2020 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_alloc_features.h"
+
+#include "base/allocator/partition_allocator/partition_alloc_base/time/time.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_root.h"
+#include "base/allocator/partition_allocator/thread_cache.h"
+#include "base/base_export.h"
+#include "base/feature_list.h"
+#include "base/features.h"
+#include "base/metrics/field_trial_params.h"
+#include "base/time/time.h"
+#include "build/build_config.h"
+#include "build/chromecast_buildflags.h"
+#include "build/chromeos_buildflags.h"
+
+namespace base {
+namespace features {
+
+BASE_FEATURE(kPartitionAllocUnretainedDanglingPtr,
+             "PartitionAllocUnretainedDanglingPtr",
+             FEATURE_ENABLED_BY_DEFAULT);
+
+constexpr FeatureParam<UnretainedDanglingPtrMode>::Option
+    kUnretainedDanglingPtrModeOption[] = {
+        {UnretainedDanglingPtrMode::kCrash, "crash"},
+        {UnretainedDanglingPtrMode::kDumpWithoutCrashing,
+         "dump_without_crashing"},
+};
+const base::FeatureParam<UnretainedDanglingPtrMode>
+    kUnretainedDanglingPtrModeParam = {
+        &kPartitionAllocUnretainedDanglingPtr,
+        "mode",
+        UnretainedDanglingPtrMode::kDumpWithoutCrashing,
+        &kUnretainedDanglingPtrModeOption,
+};
+
+BASE_FEATURE(kPartitionAllocDanglingPtr,
+             "PartitionAllocDanglingPtr",
+#if BUILDFLAG(ENABLE_DANGLING_RAW_PTR_FEATURE_FLAG) ||                   \
+    (BUILDFLAG(ENABLE_DANGLING_RAW_PTR_CHECKS) && BUILDFLAG(IS_LINUX) && \
+     !defined(OFFICIAL_BUILD) && (!defined(NDEBUG) || DCHECK_IS_ON()))
+             FEATURE_ENABLED_BY_DEFAULT
+#else
+             FEATURE_DISABLED_BY_DEFAULT
+#endif
+);
+
+constexpr FeatureParam<DanglingPtrMode>::Option kDanglingPtrModeOption[] = {
+    {DanglingPtrMode::kCrash, "crash"},
+    {DanglingPtrMode::kLogOnly, "log_only"},
+};
+const base::FeatureParam<DanglingPtrMode> kDanglingPtrModeParam{
+    &kPartitionAllocDanglingPtr,
+    "mode",
+    DanglingPtrMode::kCrash,
+    &kDanglingPtrModeOption,
+};
+constexpr FeatureParam<DanglingPtrType>::Option kDanglingPtrTypeOption[] = {
+    {DanglingPtrType::kAll, "all"},
+    {DanglingPtrType::kCrossTask, "cross_task"},
+};
+const base::FeatureParam<DanglingPtrType> kDanglingPtrTypeParam{
+    &kPartitionAllocDanglingPtr,
+    "type",
+    DanglingPtrType::kAll,
+    &kDanglingPtrTypeOption,
+};
+
+#if BUILDFLAG(USE_STARSCAN)
+// If enabled, PCScan is turned on by default for all partitions that don't
+// disable it explicitly.
+BASE_FEATURE(kPartitionAllocPCScan,
+             "PartitionAllocPCScan",
+             FEATURE_DISABLED_BY_DEFAULT);
+#endif  // BUILDFLAG(USE_STARSCAN)
+
+#if BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+// If enabled, PCScan is turned on only for the browser's malloc partition.
+BASE_FEATURE(kPartitionAllocPCScanBrowserOnly,
+             "PartitionAllocPCScanBrowserOnly",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+// If enabled, PCScan is turned on only for the renderer's malloc partition.
+BASE_FEATURE(kPartitionAllocPCScanRendererOnly,
+             "PartitionAllocPCScanRendererOnly",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+// Use a larger maximum thread cache cacheable bucket size.
+BASE_FEATURE(kPartitionAllocLargeThreadCacheSize,
+             "PartitionAllocLargeThreadCacheSize",
+             FEATURE_ENABLED_BY_DEFAULT);
+
+const base::FeatureParam<int> kPartitionAllocLargeThreadCacheSizeValue{
+    &kPartitionAllocLargeThreadCacheSize,
+    "PartitionAllocLargeThreadCacheSizeValue",
+    ::partition_alloc::ThreadCacheLimits::kLargeSizeThreshold};
+
+const base::FeatureParam<int>
+    kPartitionAllocLargeThreadCacheSizeValueForLowRAMAndroid{
+        &kPartitionAllocLargeThreadCacheSize,
+        "PartitionAllocLargeThreadCacheSizeValueForLowRAMAndroid",
+        ::partition_alloc::ThreadCacheLimits::kDefaultSizeThreshold};
+
+BASE_FEATURE(kPartitionAllocLargeEmptySlotSpanRing,
+             "PartitionAllocLargeEmptySlotSpanRing",
+             FEATURE_DISABLED_BY_DEFAULT);
+#endif  // BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+
+BASE_FEATURE(kPartitionAllocBackupRefPtr,
+             "PartitionAllocBackupRefPtr",
+#if BUILDFLAG(IS_ANDROID) || BUILDFLAG(IS_WIN) || BUILDFLAG(IS_MAC) || \
+    BUILDFLAG(IS_CHROMEOS_ASH) || BUILDFLAG(IS_CHROMEOS_LACROS) ||     \
+    (BUILDFLAG(IS_LINUX) && !BUILDFLAG(IS_CASTOS)) ||                  \
+    BUILDFLAG(ENABLE_BACKUP_REF_PTR_FEATURE_FLAG)
+             FEATURE_ENABLED_BY_DEFAULT
+#else
+             FEATURE_DISABLED_BY_DEFAULT
+#endif
+);
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocBackupRefPtrForAsh);
+BASE_FEATURE(kPartitionAllocBackupRefPtrForAsh,
+             "PartitionAllocBackupRefPtrForAsh",
+             FEATURE_ENABLED_BY_DEFAULT);
+
+constexpr FeatureParam<BackupRefPtrEnabledProcesses>::Option
+    kBackupRefPtrEnabledProcessesOptions[] = {
+        {BackupRefPtrEnabledProcesses::kBrowserOnly, "browser-only"},
+        {BackupRefPtrEnabledProcesses::kBrowserAndRenderer,
+         "browser-and-renderer"},
+        {BackupRefPtrEnabledProcesses::kNonRenderer, "non-renderer"},
+        {BackupRefPtrEnabledProcesses::kAllProcesses, "all-processes"}};
+
+const base::FeatureParam<BackupRefPtrEnabledProcesses>
+    kBackupRefPtrEnabledProcessesParam{
+        &kPartitionAllocBackupRefPtr, "enabled-processes",
+        BackupRefPtrEnabledProcesses::kNonRenderer,
+        &kBackupRefPtrEnabledProcessesOptions};
+
+constexpr FeatureParam<BackupRefPtrRefCountSize>::Option
+    kBackupRefPtrRefCountSizeOptions[] = {
+        {BackupRefPtrRefCountSize::kNatural, "natural"},
+        {BackupRefPtrRefCountSize::k4B, "4B"},
+        {BackupRefPtrRefCountSize::k8B, "8B"},
+        {BackupRefPtrRefCountSize::k16B, "16B"}};
+
+const base::FeatureParam<BackupRefPtrRefCountSize>
+    kBackupRefPtrRefCountSizeParam{
+        &kPartitionAllocBackupRefPtr, "ref-count-size",
+        BackupRefPtrRefCountSize::kNatural, &kBackupRefPtrRefCountSizeOptions};
+
+// Map -with-memory-reclaimer modes onto their counterpars without the suffix.
+// They are the same, as memory reclaimer is now controlled independently.
+// However, we need to keep both option strings, as there is a long tail of
+// clients that may have an old field trial config, which used these modes.
+//
+// DO NOT USE -with-memory-reclaimer modes in new configs!
+constexpr FeatureParam<BackupRefPtrMode>::Option kBackupRefPtrModeOptions[] = {
+    {BackupRefPtrMode::kDisabled, "disabled"},
+    {BackupRefPtrMode::kEnabled, "enabled"},
+    {BackupRefPtrMode::kEnabled, "enabled-with-memory-reclaimer"},
+    {BackupRefPtrMode::kDisabledButSplitPartitions2Way,
+     "disabled-but-2-way-split"},
+    {BackupRefPtrMode::kDisabledButSplitPartitions2Way,
+     "disabled-but-2-way-split-with-memory-reclaimer"},
+    {BackupRefPtrMode::kDisabledButSplitPartitions3Way,
+     "disabled-but-3-way-split"},
+};
+
+const base::FeatureParam<BackupRefPtrMode> kBackupRefPtrModeParam{
+    &kPartitionAllocBackupRefPtr, "brp-mode", BackupRefPtrMode::kEnabled,
+    &kBackupRefPtrModeOptions};
+
+BASE_FEATURE(kPartitionAllocMemoryTagging,
+             "PartitionAllocMemoryTagging",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+constexpr FeatureParam<MemtagMode>::Option kMemtagModeOptions[] = {
+    {MemtagMode::kSync, "sync"},
+    {MemtagMode::kAsync, "async"}};
+
+const base::FeatureParam<MemtagMode> kMemtagModeParam{
+    &kPartitionAllocMemoryTagging, "memtag-mode", MemtagMode::kAsync,
+    &kMemtagModeOptions};
+
+constexpr FeatureParam<MemoryTaggingEnabledProcesses>::Option
+    kMemoryTaggingEnabledProcessesOptions[] = {
+        {MemoryTaggingEnabledProcesses::kBrowserOnly, "browser-only"},
+        {MemoryTaggingEnabledProcesses::kNonRenderer, "non-renderer"},
+        {MemoryTaggingEnabledProcesses::kAllProcesses, "all-processes"}};
+
+const base::FeatureParam<MemoryTaggingEnabledProcesses>
+    kMemoryTaggingEnabledProcessesParam{
+        &kPartitionAllocMemoryTagging, "enabled-processes",
+        MemoryTaggingEnabledProcesses::kBrowserOnly,
+        &kMemoryTaggingEnabledProcessesOptions};
+
+BASE_FEATURE(kKillPartitionAllocMemoryTagging,
+             "KillPartitionAllocMemoryTagging",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPermissiveMte);
+BASE_FEATURE(kPartitionAllocPermissiveMte,
+             "PartitionAllocPermissiveMte",
+             FEATURE_ENABLED_BY_DEFAULT);
+
+const base::FeatureParam<bool> kBackupRefPtrAsanEnableDereferenceCheckParam{
+    &kPartitionAllocBackupRefPtr, "asan-enable-dereference-check", true};
+const base::FeatureParam<bool> kBackupRefPtrAsanEnableExtractionCheckParam{
+    &kPartitionAllocBackupRefPtr, "asan-enable-extraction-check",
+    false};  // Not much noise at the moment to enable by default.
+const base::FeatureParam<bool> kBackupRefPtrAsanEnableInstantiationCheckParam{
+    &kPartitionAllocBackupRefPtr, "asan-enable-instantiation-check", true};
+
+// If enabled, switches the bucket distribution to a denser one.
+//
+// We enable this by default everywhere except for 32-bit Android, since we saw
+// regressions there.
+BASE_FEATURE(kPartitionAllocUseDenserDistribution,
+             "PartitionAllocUseDenserDistribution",
+#if BUILDFLAG(IS_ANDROID) && defined(ARCH_CPU_32_BITS)
+             FEATURE_DISABLED_BY_DEFAULT
+#else
+             FEATURE_ENABLED_BY_DEFAULT
+#endif  // BUILDFLAG(IS_ANDROID) && defined(ARCH_CPU_32_BITS)
+);
+const base::FeatureParam<BucketDistributionMode>::Option
+    kPartitionAllocBucketDistributionOption[] = {
+        {BucketDistributionMode::kDefault, "default"},
+        {BucketDistributionMode::kDenser, "denser"},
+};
+const base::FeatureParam<BucketDistributionMode>
+    kPartitionAllocBucketDistributionParam {
+  &kPartitionAllocUseDenserDistribution, "mode",
+#if BUILDFLAG(IS_ANDROID) && defined(ARCH_CPU_32_BITS)
+      BucketDistributionMode::kDefault,
+#else
+      BucketDistributionMode::kDenser,
+#endif  // BUILDFLAG(IS_ANDROID) && defined(ARCH_CPU_32_BITS)
+      &kPartitionAllocBucketDistributionOption
+};
+
+BASE_FEATURE(kPartitionAllocMemoryReclaimer,
+             "PartitionAllocMemoryReclaimer",
+             FEATURE_ENABLED_BY_DEFAULT);
+const base::FeatureParam<TimeDelta> kPartitionAllocMemoryReclaimerInterval = {
+    &kPartitionAllocMemoryReclaimer, "interval",
+    TimeDelta(),  // Defaults to zero.
+};
+
+// Configures whether we set a lower limit for renderers that do not have a main
+// frame, similar to the limit that is already done for backgrounded renderers.
+BASE_FEATURE(kLowerPAMemoryLimitForNonMainRenderers,
+             "LowerPAMemoryLimitForNonMainRenderers",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+// If enabled, switches PCScan scheduling to a mutator-aware scheduler. Does not
+// affect whether PCScan is enabled itself.
+BASE_FEATURE(kPartitionAllocPCScanMUAwareScheduler,
+             "PartitionAllocPCScanMUAwareScheduler",
+             FEATURE_ENABLED_BY_DEFAULT);
+
+// If enabled, PCScan frees unconditionally all quarantined objects.
+// This is a performance testing feature.
+BASE_FEATURE(kPartitionAllocPCScanImmediateFreeing,
+             "PartitionAllocPCScanImmediateFreeing",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+// If enabled, PCScan clears eagerly (synchronously) on free().
+BASE_FEATURE(kPartitionAllocPCScanEagerClearing,
+             "PartitionAllocPCScanEagerClearing",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+// In addition to heap, scan also the stack of the current mutator.
+BASE_FEATURE(kPartitionAllocPCScanStackScanning,
+             "PartitionAllocPCScanStackScanning",
+#if BUILDFLAG(PCSCAN_STACK_SUPPORTED)
+             FEATURE_ENABLED_BY_DEFAULT
+#else
+             FEATURE_DISABLED_BY_DEFAULT
+#endif  // BUILDFLAG(PCSCAN_STACK_SUPPORTED)
+);
+
+BASE_FEATURE(kPartitionAllocDCScan,
+             "PartitionAllocDCScan",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+// Whether to straighten free lists for larger slot spans in PurgeMemory() ->
+// ... -> PartitionPurgeSlotSpan().
+BASE_FEATURE(kPartitionAllocStraightenLargerSlotSpanFreeLists,
+             "PartitionAllocStraightenLargerSlotSpanFreeLists",
+             FEATURE_ENABLED_BY_DEFAULT);
+const base::FeatureParam<
+    partition_alloc::StraightenLargerSlotSpanFreeListsMode>::Option
+    kPartitionAllocStraightenLargerSlotSpanFreeListsModeOption[] = {
+        {partition_alloc::StraightenLargerSlotSpanFreeListsMode::
+             kOnlyWhenUnprovisioning,
+         "only-when-unprovisioning"},
+        {partition_alloc::StraightenLargerSlotSpanFreeListsMode::kAlways,
+         "always"},
+};
+const base::FeatureParam<partition_alloc::StraightenLargerSlotSpanFreeListsMode>
+    kPartitionAllocStraightenLargerSlotSpanFreeListsMode = {
+        &kPartitionAllocStraightenLargerSlotSpanFreeLists,
+        "mode",
+        partition_alloc::StraightenLargerSlotSpanFreeListsMode::
+            kOnlyWhenUnprovisioning,
+        &kPartitionAllocStraightenLargerSlotSpanFreeListsModeOption,
+};
+
+// Whether to sort free lists for smaller slot spans in PurgeMemory().
+BASE_FEATURE(kPartitionAllocSortSmallerSlotSpanFreeLists,
+             "PartitionAllocSortSmallerSlotSpanFreeLists",
+             FEATURE_ENABLED_BY_DEFAULT);
+
+// Whether to sort the active slot spans in PurgeMemory().
+BASE_FEATURE(kPartitionAllocSortActiveSlotSpans,
+             "PartitionAllocSortActiveSlotSpans",
+             FEATURE_DISABLED_BY_DEFAULT);
+
+#if BUILDFLAG(IS_WIN)
+// Whether to retry allocations when commit fails.
+BASE_FEATURE(kPageAllocatorRetryOnCommitFailure,
+             "PageAllocatorRetryOnCommitFailure",
+             FEATURE_DISABLED_BY_DEFAULT);
+#endif
+
+#if BUILDFLAG(IS_ANDROID)
+// A parameter to exclude or not exclude PartitionAllocSupport from
+// PartialLowModeOnMidRangeDevices. This is used to see how it affects
+// renderer performances, e.g. blink_perf.parser benchmark.
+// The feature: kPartialLowEndModeOnMidRangeDevices is defined in
+// //base/features.cc. Since the following feature param is related to
+// PartitionAlloc, define the param here.
+const FeatureParam<bool> kPartialLowEndModeExcludePartitionAllocSupport{
+    &kPartialLowEndModeOnMidRangeDevices, "exclude-partition-alloc-support",
+    false};
+#endif
+
+BASE_FEATURE(kEnableConfigurableThreadCacheMultiplier,
+             "EnableConfigurableThreadCacheMultiplier",
+             base::FEATURE_DISABLED_BY_DEFAULT);
+
+const base::FeatureParam<double> kThreadCacheMultiplier{
+    &kEnableConfigurableThreadCacheMultiplier, "ThreadCacheMultiplier", 2.};
+
+const base::FeatureParam<double> kThreadCacheMultiplierForAndroid{
+    &kEnableConfigurableThreadCacheMultiplier,
+    "ThreadCacheMultiplierForAndroid", 1.};
+
+constexpr partition_alloc::internal::base::TimeDelta ToPartitionAllocTimeDelta(
+    base::TimeDelta time_delta) {
+  return partition_alloc::internal::base::Microseconds(
+      time_delta.InMicroseconds());
+}
+
+constexpr base::TimeDelta FromPartitionAllocTimeDelta(
+    partition_alloc::internal::base::TimeDelta time_delta) {
+  return base::Microseconds(time_delta.InMicroseconds());
+}
+
+BASE_FEATURE(kEnableConfigurableThreadCachePurgeInterval,
+             "EnableConfigurableThreadCachePurgeInterval",
+             base::FEATURE_DISABLED_BY_DEFAULT);
+
+const base::FeatureParam<base::TimeDelta> kThreadCacheMinPurgeInterval{
+    &kEnableConfigurableThreadCachePurgeInterval, "ThreadCacheMinPurgeInterval",
+    FromPartitionAllocTimeDelta(partition_alloc::kMinPurgeInterval)};
+
+const base::FeatureParam<base::TimeDelta> kThreadCacheMaxPurgeInterval{
+    &kEnableConfigurableThreadCachePurgeInterval, "ThreadCacheMaxPurgeInterval",
+    FromPartitionAllocTimeDelta(partition_alloc::kMaxPurgeInterval)};
+
+const base::FeatureParam<base::TimeDelta> kThreadCacheDefaultPurgeInterval{
+    &kEnableConfigurableThreadCachePurgeInterval,
+    "ThreadCacheDefaultPurgeInterval",
+    FromPartitionAllocTimeDelta(partition_alloc::kDefaultPurgeInterval)};
+
+const partition_alloc::internal::base::TimeDelta
+GetThreadCacheMinPurgeInterval() {
+  return ToPartitionAllocTimeDelta(kThreadCacheMinPurgeInterval.Get());
+}
+
+const partition_alloc::internal::base::TimeDelta
+GetThreadCacheMaxPurgeInterval() {
+  return ToPartitionAllocTimeDelta(kThreadCacheMaxPurgeInterval.Get());
+}
+
+const partition_alloc::internal::base::TimeDelta
+GetThreadCacheDefaultPurgeInterval() {
+  return ToPartitionAllocTimeDelta(kThreadCacheDefaultPurgeInterval.Get());
+}
+
+BASE_FEATURE(kEnableConfigurableThreadCacheMinCachedMemoryForPurging,
+             "EnableConfigurableThreadCacheMinCachedMemoryForPurging",
+             base::FEATURE_DISABLED_BY_DEFAULT);
+
+const base::FeatureParam<int> kThreadCacheMinCachedMemoryForPurgingBytes{
+    &kEnableConfigurableThreadCacheMinCachedMemoryForPurging,
+    "ThreadCacheMinCachedMemoryForPurgingBytes",
+    partition_alloc::kMinCachedMemoryForPurgingBytes};
+
+}  // namespace features
+}  // namespace base

src/base/allocator/partition_alloc_features.h

@@ -0,0 +1,225 @@
+// Copyright 2020 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOC_FEATURES_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOC_FEATURES_H_
+
+#include "base/allocator/partition_allocator/partition_alloc_base/time/time.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_root.h"
+#include "base/base_export.h"
+#include "base/compiler_specific.h"
+#include "base/feature_list.h"
+#include "base/metrics/field_trial_params.h"
+#include "base/strings/string_piece.h"
+#include "base/time/time.h"
+#include "build/build_config.h"
+
+namespace base {
+namespace features {
+
+extern const BASE_EXPORT Feature kPartitionAllocUnretainedDanglingPtr;
+enum class UnretainedDanglingPtrMode {
+  kCrash,
+  kDumpWithoutCrashing,
+};
+extern const BASE_EXPORT base::FeatureParam<UnretainedDanglingPtrMode>
+    kUnretainedDanglingPtrModeParam;
+
+// See /docs/dangling_ptr.md
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocDanglingPtr);
+enum class DanglingPtrMode {
+  // Crash immediately after detecting a dangling raw_ptr.
+  kCrash,  // (default)
+
+  // Log the signature of every occurrences without crashing. It is used by
+  // bots.
+  // Format "[DanglingSignature]\t<1>\t<2>\t<3>\t<4>"
+  // 1. The function which freed the memory while it was still referenced.
+  // 2. The task in which the memory was freed.
+  // 3. The function which released the raw_ptr reference.
+  // 4. The task in which the raw_ptr was released.
+  kLogOnly,
+
+  // Note: This will be extended with a single shot DumpWithoutCrashing.
+};
+extern const BASE_EXPORT base::FeatureParam<DanglingPtrMode>
+    kDanglingPtrModeParam;
+enum class DanglingPtrType {
+  // Act on any dangling raw_ptr released after being freed.
+  kAll,  // (default)
+
+  // Detect when freeing memory and releasing the dangling raw_ptr happens in
+  // a different task. Those are more likely to cause use after free.
+  kCrossTask,
+
+  // Note: This will be extended with LongLived
+};
+extern const BASE_EXPORT base::FeatureParam<DanglingPtrType>
+    kDanglingPtrTypeParam;
+
+#if BUILDFLAG(USE_STARSCAN)
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPCScan);
+#endif
+#if BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPCScanBrowserOnly);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPCScanRendererOnly);
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocLargeThreadCacheSize);
+extern const BASE_EXPORT base::FeatureParam<int>
+    kPartitionAllocLargeThreadCacheSizeValue;
+extern const BASE_EXPORT base::FeatureParam<int>
+    kPartitionAllocLargeThreadCacheSizeValueForLowRAMAndroid;
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocLargeEmptySlotSpanRing);
+#endif  // BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+
+enum class BackupRefPtrEnabledProcesses {
+  // BRP enabled only in the browser process.
+  kBrowserOnly,
+  // BRP enabled only in the browser and renderer processes.
+  kBrowserAndRenderer,
+  // BRP enabled in all processes, except renderer.
+  kNonRenderer,
+  // BRP enabled in all processes.
+  kAllProcesses,
+};
+
+enum class BackupRefPtrMode {
+  // BRP is disabled across all partitions. Equivalent to the Finch flag being
+  // disabled.
+  kDisabled,
+
+  // BRP is enabled in the main partition, as well as certain Renderer-only
+  // partitions (if enabled in Renderer at all).
+  // This entails splitting the main partition.
+  kEnabled,
+
+  // BRP is disabled, but the main partition is split out, as if BRP was enabled
+  // in the "previous slot" mode.
+  kDisabledButSplitPartitions2Way,
+
+  // BRP is disabled, but the main partition *and* aligned partition are split
+  // out, as if BRP was enabled in the "before allocation" mode.
+  kDisabledButSplitPartitions3Way,
+};
+
+// Decides the amount of memory uses for BRP ref-count. The actual ref-count may
+// be smaller, in which case extra padding is added.
+enum class BackupRefPtrRefCountSize {
+  // Whatever sizeof(PartitionRefCount) happens to be, which is influence by
+  // buildflags.
+  // The remaining options require sizeof(PartitionRefCount) not to exceed the
+  // desired size, which will be asserted.
+  kNatural,
+  // 4 bytes.
+  k4B,
+  // 8 bytes
+  k8B,
+  // 16 bytes.
+  k16B,
+};
+
+enum class MemtagMode {
+  // memtagMode will be SYNC.
+  kSync,
+  // memtagMode will be ASYNC.
+  kAsync,
+};
+
+enum class MemoryTaggingEnabledProcesses {
+  // Memory tagging enabled only in the browser process.
+  kBrowserOnly,
+  // Memory tagging enabled in all processes, except renderer.
+  kNonRenderer,
+  // Memory tagging enabled in all processes.
+  kAllProcesses,
+};
+
+enum class BucketDistributionMode : uint8_t {
+  kDefault,
+  kDenser,
+};
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocBackupRefPtr);
+extern const BASE_EXPORT base::FeatureParam<BackupRefPtrEnabledProcesses>
+    kBackupRefPtrEnabledProcessesParam;
+extern const BASE_EXPORT base::FeatureParam<BackupRefPtrMode>
+    kBackupRefPtrModeParam;
+extern const BASE_EXPORT base::FeatureParam<BackupRefPtrRefCountSize>
+    kBackupRefPtrRefCountSizeParam;
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocMemoryTagging);
+extern const BASE_EXPORT base::FeatureParam<MemtagMode> kMemtagModeParam;
+extern const BASE_EXPORT base::FeatureParam<MemoryTaggingEnabledProcesses>
+    kMemoryTaggingEnabledProcessesParam;
+// Kill switch for memory tagging. Skips any code related to memory tagging when
+// enabled.
+BASE_EXPORT BASE_DECLARE_FEATURE(kKillPartitionAllocMemoryTagging);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPermissiveMte);
+extern const BASE_EXPORT base::FeatureParam<bool>
+    kBackupRefPtrAsanEnableDereferenceCheckParam;
+extern const BASE_EXPORT base::FeatureParam<bool>
+    kBackupRefPtrAsanEnableExtractionCheckParam;
+extern const BASE_EXPORT base::FeatureParam<bool>
+    kBackupRefPtrAsanEnableInstantiationCheckParam;
+extern const BASE_EXPORT base::FeatureParam<BucketDistributionMode>
+    kPartitionAllocBucketDistributionParam;
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocBackupRefPtrForAsh);
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kLowerPAMemoryLimitForNonMainRenderers);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPCScanMUAwareScheduler);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPCScanStackScanning);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocDCScan);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPCScanImmediateFreeing);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocPCScanEagerClearing);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocUseDenserDistribution);
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocMemoryReclaimer);
+extern const BASE_EXPORT base::FeatureParam<TimeDelta>
+    kPartitionAllocMemoryReclaimerInterval;
+BASE_EXPORT BASE_DECLARE_FEATURE(
+    kPartitionAllocStraightenLargerSlotSpanFreeLists);
+extern const BASE_EXPORT
+    base::FeatureParam<partition_alloc::StraightenLargerSlotSpanFreeListsMode>
+        kPartitionAllocStraightenLargerSlotSpanFreeListsMode;
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocSortSmallerSlotSpanFreeLists);
+BASE_EXPORT BASE_DECLARE_FEATURE(kPartitionAllocSortActiveSlotSpans);
+
+#if BUILDFLAG(IS_WIN)
+BASE_EXPORT BASE_DECLARE_FEATURE(kPageAllocatorRetryOnCommitFailure);
+#endif
+
+#if BUILDFLAG(IS_ANDROID)
+extern const base::FeatureParam<bool>
+    kPartialLowEndModeExcludePartitionAllocSupport;
+#endif
+
+// Name of the synthetic trial associated with forcibly enabling BRP in
+// all processes.
+inline constexpr base::StringPiece kRendererLiveBRPSyntheticTrialName =
+    "BackupRefPtrRendererLive";
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kEnableConfigurableThreadCacheMultiplier);
+extern const BASE_EXPORT base::FeatureParam<double> kThreadCacheMultiplier;
+extern const BASE_EXPORT base::FeatureParam<double>
+    kThreadCacheMultiplierForAndroid;
+
+BASE_EXPORT BASE_DECLARE_FEATURE(kEnableConfigurableThreadCachePurgeInterval);
+extern const partition_alloc::internal::base::TimeDelta
+GetThreadCacheMinPurgeInterval();
+extern const partition_alloc::internal::base::TimeDelta
+GetThreadCacheMaxPurgeInterval();
+extern const partition_alloc::internal::base::TimeDelta
+GetThreadCacheDefaultPurgeInterval();
+
+BASE_EXPORT BASE_DECLARE_FEATURE(
+    kEnableConfigurableThreadCacheMinCachedMemoryForPurging);
+extern const BASE_EXPORT base::FeatureParam<int>
+    kThreadCacheMinCachedMemoryForPurgingBytes;
+
+}  // namespace features
+}  // namespace base
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOC_FEATURES_H_

src/base/allocator/partition_alloc_support.h

@@ -0,0 +1,126 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOC_SUPPORT_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOC_SUPPORT_H_
+
+#include <map>
+#include <string>
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_config.h"
+#include "base/allocator/partition_allocator/thread_cache.h"
+#include "base/base_export.h"
+#include "base/memory/scoped_refptr.h"
+#include "base/synchronization/lock.h"
+#include "base/task/sequenced_task_runner.h"
+#include "base/thread_annotations.h"
+
+namespace base::allocator {
+
+#if BUILDFLAG(USE_STARSCAN)
+BASE_EXPORT void RegisterPCScanStatsReporter();
+#endif
+
+// Starts a periodic timer on the current thread to purge all thread caches.
+BASE_EXPORT void StartThreadCachePeriodicPurge();
+
+BASE_EXPORT void StartMemoryReclaimer(
+    scoped_refptr<SequencedTaskRunner> task_runner);
+
+BASE_EXPORT std::map<std::string, std::string> ProposeSyntheticFinchTrials();
+
+// Install handlers for when dangling raw_ptr(s) have been detected. This prints
+// two StackTraces. One where the memory is freed, one where the last dangling
+// raw_ptr stopped referencing it.
+//
+// This is currently effective, only when compiled with
+// `enable_dangling_raw_ptr_checks` build flag.
+BASE_EXPORT void InstallDanglingRawPtrChecks();
+BASE_EXPORT void InstallUnretainedDanglingRawPtrChecks();
+
+// Allows to re-configure PartitionAlloc at run-time.
+class BASE_EXPORT PartitionAllocSupport {
+ public:
+  struct BrpConfiguration {
+    bool enable_brp = false;
+    bool enable_brp_for_ash = false;
+    bool split_main_partition = false;
+    bool use_dedicated_aligned_partition = false;
+    bool process_affected_by_brp_flag = false;
+    size_t ref_count_size = 0;
+  };
+
+  // Reconfigure* functions re-configure PartitionAlloc. It is impossible to
+  // configure PartitionAlloc before/at its initialization using information not
+  // known at compile-time (e.g. process type, Finch), because by the time this
+  // information is available memory allocations would have surely happened,
+  // that requiring a functioning allocator.
+  //
+  // *Earlyish() is called as early as it is reasonably possible.
+  // *AfterZygoteFork() is its complement to finish configuring process-specific
+  // stuff that had to be postponed due to *Earlyish() being called with
+  // |process_type==kZygoteProcess|.
+  // *AfterFeatureListInit() is called in addition to the above, once
+  // FeatureList has been initialized and ready to use. It is guaranteed to be
+  // called on non-zygote processes or after the zygote has been forked.
+  // *AfterTaskRunnerInit() is called once it is possible to post tasks, and
+  // after the previous steps.
+  //
+  // *Earlyish() must be called exactly once. *AfterZygoteFork() must be called
+  // once iff *Earlyish() was called before with |process_type==kZygoteProcess|.
+  //
+  // *AfterFeatureListInit() may be called more than once, but will perform its
+  // re-configuration steps exactly once.
+  //
+  // *AfterTaskRunnerInit() may be called more than once.
+  void ReconfigureForTests();
+  void ReconfigureEarlyish(const std::string& process_type);
+  void ReconfigureAfterZygoteFork(const std::string& process_type);
+  void ReconfigureAfterFeatureListInit(
+      const std::string& process_type,
+      bool configure_dangling_pointer_detector = true);
+  void ReconfigureAfterTaskRunnerInit(const std::string& process_type);
+
+  // |has_main_frame| tells us if the renderer contains a main frame.
+  void OnForegrounded(bool has_main_frame);
+  void OnBackgrounded();
+
+#if BUILDFLAG(ENABLE_DANGLING_RAW_PTR_CHECKS)
+  static std::string ExtractDanglingPtrSignatureForTests(
+      std::string stacktrace);
+#endif
+
+  static PartitionAllocSupport* Get();
+
+  static BrpConfiguration GetBrpConfiguration(const std::string& process_type);
+
+  // Returns true if memory tagging should be enabled if available for the given
+  // process type. May be called multiple times per process.
+  static bool ShouldEnableMemoryTagging(const std::string& process_type);
+
+  // For calling from within third_party/blink/.
+  static bool ShouldEnableMemoryTaggingInRendererProcess();
+
+ private:
+  PartitionAllocSupport();
+
+  base::Lock lock_;
+  bool called_for_tests_ GUARDED_BY(lock_) = false;
+  bool called_earlyish_ GUARDED_BY(lock_) = false;
+  bool called_after_zygote_fork_ GUARDED_BY(lock_) = false;
+  bool called_after_feature_list_init_ GUARDED_BY(lock_) = false;
+  bool called_after_thread_pool_init_ GUARDED_BY(lock_) = false;
+  std::string established_process_type_ GUARDED_BY(lock_) = "INVALID";
+
+#if PA_CONFIG(THREAD_CACHE_SUPPORTED) && \
+    BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+  size_t largest_cached_size_ =
+      ::partition_alloc::ThreadCacheLimits::kDefaultSizeThreshold;
+#endif
+};
+
+}  // namespace base::allocator
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOC_SUPPORT_H_

src/base/allocator/partition_allocator/BUILD.gn

@@ -0,0 +1,782 @@
+# Copyright 2022 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import("//base/allocator/partition_allocator/partition_alloc.gni")
+import("//build/buildflag_header.gni")
+import("//build/config/android/config.gni")
+import("//build/config/chromecast_build.gni")
+import("//build/config/chromeos/ui_mode.gni")
+import("//build/config/compiler/compiler.gni")
+import("//build/config/dcheck_always_on.gni")
+import("//build/config/logging.gni")
+
+# Add partition_alloc.gni and import it for partition_alloc configs.
+
+config("partition_alloc_implementation") {
+  # See also: `partition_alloc_base/component_export.h`
+  defines = [ "IS_PARTITION_ALLOC_IMPL" ]
+}
+
+config("memory_tagging") {
+  if (current_cpu == "arm64" && is_clang &&
+      (is_linux || is_chromeos || is_android || is_fuchsia)) {
+    # base/ has access to the MTE intrinsics because it needs to use them,
+    # but they're not backwards compatible. Use base::CPU::has_mte()
+    # beforehand to confirm or use indirect functions (ifuncs) to select
+    # an MTE-specific implementation at dynamic link-time.
+    cflags = [
+      "-Xclang",
+      "-target-feature",
+      "-Xclang",
+      "+mte",
+    ]
+  }
+}
+
+# Used to shim malloc symbols on Android. see //base/allocator/README.md.
+config("wrap_malloc_symbols") {
+  ldflags = [
+    "-Wl,-wrap,calloc",
+    "-Wl,-wrap,free",
+    "-Wl,-wrap,malloc",
+    "-Wl,-wrap,memalign",
+    "-Wl,-wrap,posix_memalign",
+    "-Wl,-wrap,pvalloc",
+    "-Wl,-wrap,realloc",
+    "-Wl,-wrap,valloc",
+
+    # Not allocating memory, but part of the API
+    "-Wl,-wrap,malloc_usable_size",
+
+    # <stdlib.h> functions
+    "-Wl,-wrap,realpath",
+
+    # <string.h> functions
+    "-Wl,-wrap,strdup",
+    "-Wl,-wrap,strndup",
+
+    # <unistd.h> functions
+    "-Wl,-wrap,getcwd",
+
+    # <stdio.h> functions
+    "-Wl,-wrap,asprintf",
+    "-Wl,-wrap,vasprintf",
+  ]
+}
+
+config("mac_no_default_new_delete_symbols") {
+  if (!is_component_build) {
+    # This is already set when we compile libc++, see
+    # buildtools/third_party/libc++/BUILD.gn. But it needs to be set here as
+    # well, since the shim defines the symbols, to prevent them being exported.
+    cflags = [ "-fvisibility-global-new-delete-hidden" ]
+  }
+}
+
+if (is_fuchsia) {
+  config("fuchsia_sync_lib") {
+    libs = [
+      "sync",  # Used by spinning_mutex.h.
+    ]
+  }
+}
+
+if (enable_pkeys && is_debug) {
+  config("no_stack_protector") {
+    cflags = [ "-fno-stack-protector" ]
+  }
+}
+
+_remove_configs = []
+_add_configs = []
+if (!is_debug || partition_alloc_optimized_debug) {
+  _remove_configs += [ "//build/config/compiler:default_optimization" ]
+
+  # Partition alloc is relatively hot (>1% of cycles for users of CrOS).
+  # Use speed-focused optimizations for it.
+  _add_configs += [ "//build/config/compiler:optimize_speed" ]
+} else {
+  _remove_configs += [ "//build/config/compiler:default_optimization" ]
+  _add_configs += [ "//build/config/compiler:no_optimize" ]
+}
+
+component("partition_alloc") {
+  public_deps = [
+    ":allocator_base",
+    ":allocator_core",
+    ":allocator_shim",
+  ]
+}
+
+# Changes the freelist implementation to use pointer offsets in lieu
+# of full-on pointers. Defaults to false, which implies the use of
+# "encoded next" freelist entry.
+#
+# Only usable when pointers are 64-bit.
+use_freelist_pool_offsets = has_64_bit_pointers && false
+
+source_set("allocator_core") {
+  visibility = [ ":*" ]
+
+  sources = [
+    "address_pool_manager.cc",
+    "address_pool_manager.h",
+    "address_pool_manager_bitmap.cc",
+    "address_pool_manager_bitmap.h",
+    "address_pool_manager_types.h",
+    "address_space_randomization.cc",
+    "address_space_randomization.h",
+    "address_space_stats.h",
+    "allocation_guard.cc",
+    "allocation_guard.h",
+    "compressed_pointer.cc",
+    "compressed_pointer.h",
+    "dangling_raw_ptr_checks.cc",
+    "dangling_raw_ptr_checks.h",
+    "freeslot_bitmap.h",
+    "freeslot_bitmap_constants.h",
+    "gwp_asan_support.cc",
+    "gwp_asan_support.h",
+    "memory_reclaimer.cc",
+    "memory_reclaimer.h",
+    "oom.cc",
+    "oom.h",
+    "oom_callback.cc",
+    "oom_callback.h",
+    "page_allocator.cc",
+    "page_allocator.h",
+    "page_allocator_constants.h",
+    "page_allocator_internal.h",
+    "partition_address_space.cc",
+    "partition_address_space.h",
+    "partition_alloc-inl.h",
+    "partition_alloc.cc",
+    "partition_alloc.h",
+    "partition_alloc_allocation_data.h",
+    "partition_alloc_check.h",
+    "partition_alloc_config.h",
+    "partition_alloc_constants.h",
+    "partition_alloc_forward.h",
+    "partition_alloc_hooks.cc",
+    "partition_alloc_hooks.h",
+    "partition_bucket.cc",
+    "partition_bucket.h",
+    "partition_bucket_lookup.h",
+    "partition_cookie.h",
+    "partition_dcheck_helper.cc",
+    "partition_dcheck_helper.h",
+    "partition_direct_map_extent.h",
+    "partition_freelist_entry.cc",
+    "partition_freelist_entry.h",
+    "partition_lock.h",
+    "partition_oom.cc",
+    "partition_oom.h",
+    "partition_page.cc",
+    "partition_page.h",
+    "partition_page_constants.h",
+    "partition_ref_count.h",
+    "partition_root.cc",
+    "partition_root.h",
+    "partition_stats.cc",
+    "partition_stats.h",
+    "partition_superpage_extent_entry.h",
+    "partition_tls.h",
+    "random.cc",
+    "random.h",
+    "reservation_offset_table.cc",
+    "reservation_offset_table.h",
+    "reverse_bytes.h",
+    "spinning_mutex.cc",
+    "spinning_mutex.h",
+    "tagging.cc",
+    "tagging.h",
+    "thread_cache.cc",
+    "thread_cache.h",
+    "thread_isolation/alignment.h",
+    "thread_isolation/pkey.cc",
+    "thread_isolation/pkey.h",
+    "thread_isolation/thread_isolation.cc",
+    "thread_isolation/thread_isolation.h",
+    "yield_processor.h",
+  ]
+
+  if (use_starscan) {
+    sources += [
+      "starscan/logging.h",
+      "starscan/metadata_allocator.cc",
+      "starscan/metadata_allocator.h",
+      "starscan/pcscan.cc",
+      "starscan/pcscan.h",
+      "starscan/pcscan_internal.cc",
+      "starscan/pcscan_internal.h",
+      "starscan/pcscan_scheduling.cc",
+      "starscan/pcscan_scheduling.h",
+      "starscan/raceful_worklist.h",
+      "starscan/scan_loop.h",
+      "starscan/snapshot.cc",
+      "starscan/snapshot.h",
+      "starscan/stack/stack.cc",
+      "starscan/stack/stack.h",
+      "starscan/starscan_fwd.h",
+      "starscan/state_bitmap.h",
+      "starscan/stats_collector.cc",
+      "starscan/stats_collector.h",
+      "starscan/stats_reporter.h",
+      "starscan/write_protector.cc",
+      "starscan/write_protector.h",
+    ]
+  }
+
+  defines = []
+  if (is_win) {
+    sources += [
+      "page_allocator_internals_win.h",
+      "partition_tls_win.cc",
+    ]
+  } else if (is_posix) {
+    sources += [
+      "page_allocator_internals_posix.cc",
+      "page_allocator_internals_posix.h",
+    ]
+  } else if (is_fuchsia) {
+    sources += [ "page_allocator_internals_fuchsia.h" ]
+  }
+  if (is_android) {
+    # The Android NDK supports PR_MTE_* macros as of NDK r23.
+    if (android_ndk_major_version >= 23) {
+      defines += [ "HAS_PR_MTE_MACROS" ]
+    }
+  }
+  if (use_starscan) {
+    if (current_cpu == "x64") {
+      assert(pcscan_stack_supported)
+      sources += [ "starscan/stack/asm/x64/push_registers_asm.cc" ]
+    } else if (current_cpu == "x86") {
+      assert(pcscan_stack_supported)
+      sources += [ "starscan/stack/asm/x86/push_registers_asm.cc" ]
+    } else if (current_cpu == "arm") {
+      assert(pcscan_stack_supported)
+      sources += [ "starscan/stack/asm/arm/push_registers_asm.cc" ]
+    } else if (current_cpu == "arm64") {
+      assert(pcscan_stack_supported)
+      sources += [ "starscan/stack/asm/arm64/push_registers_asm.cc" ]
+    } else if (current_cpu == "riscv64") {
+      assert(pcscan_stack_supported)
+      sources += [ "starscan/stack/asm/riscv64/push_registers_asm.cc" ]
+    } else {
+      # To support a trampoline for another arch, please refer to v8/src/heap/base.
+      assert(!pcscan_stack_supported)
+    }
+  }
+  if (use_freelist_pool_offsets) {
+    sources += [ "pool_offset_freelist.h" ]
+  } else {
+    sources += [ "encoded_next_freelist.h" ]
+  }
+
+  public_deps = [
+    ":chromecast_buildflags",
+    ":chromeos_buildflags",
+    ":debugging_buildflags",
+    ":partition_alloc_buildflags",
+  ]
+
+  configs += [
+    ":partition_alloc_implementation",
+    ":memory_tagging",
+  ]
+  deps = [ ":allocator_base" ]
+  public_configs = []
+  if (is_android) {
+    # tagging.cc requires __arm_mte_set_* functions.
+    deps += [ "//third_party/cpu_features:ndk_compat" ]
+  }
+  if (is_fuchsia) {
+    deps += [
+      "//third_party/fuchsia-sdk/sdk/fidl/fuchsia.kernel:fuchsia.kernel_cpp",
+      "//third_party/fuchsia-sdk/sdk/pkg/component_incoming_cpp",
+    ]
+    public_deps += [
+      "//third_party/fuchsia-sdk/sdk/pkg/sync",
+      "//third_party/fuchsia-sdk/sdk/pkg/zx",
+    ]
+
+    # Needed for users of spinning_mutex.h, which for performance reasons,
+    # contains inlined calls to `libsync` inside the header file.
+    # It appends an entry to the "libs" section of the dependent target.
+    public_configs += [ ":fuchsia_sync_lib" ]
+  }
+
+  frameworks = []
+  if (is_mac) {
+    # SecTaskGetCodeSignStatus needs:
+    frameworks += [ "Security.framework" ]
+  }
+
+  if (is_apple) {
+    frameworks += [
+      "CoreFoundation.framework",
+      "Foundation.framework",
+    ]
+  }
+
+  configs += [ "//build/config/compiler:wexit_time_destructors" ]
+  configs -= _remove_configs
+  configs += _add_configs
+
+  # We want to be able to test pkey mode without access to the default pkey.
+  # This is incompatible with stack protectors since the TLS won't be pkey-tagged.
+  if (enable_pkeys && is_debug) {
+    configs += [ ":no_stack_protector" ]
+  }
+}
+
+source_set("allocator_base") {
+  visibility = [ ":*" ]
+
+  sources = [
+    "partition_alloc_base/atomic_ref_count.h",
+    "partition_alloc_base/augmentations/compiler_specific.h",
+    "partition_alloc_base/bit_cast.h",
+    "partition_alloc_base/bits.h",
+    "partition_alloc_base/check.cc",
+    "partition_alloc_base/check.h",
+    "partition_alloc_base/compiler_specific.h",
+    "partition_alloc_base/component_export.h",
+    "partition_alloc_base/cpu.cc",
+    "partition_alloc_base/cpu.h",
+    "partition_alloc_base/cxx20_is_constant_evaluated.h",
+    "partition_alloc_base/debug/alias.cc",
+    "partition_alloc_base/debug/alias.h",
+    "partition_alloc_base/debug/stack_trace.cc",
+    "partition_alloc_base/debug/stack_trace.h",
+    "partition_alloc_base/export_template.h",
+    "partition_alloc_base/gtest_prod_util.h",
+    "partition_alloc_base/immediate_crash.h",
+    "partition_alloc_base/log_message.cc",
+    "partition_alloc_base/log_message.h",
+    "partition_alloc_base/logging.cc",
+    "partition_alloc_base/logging.h",
+    "partition_alloc_base/memory/page_size.h",
+    "partition_alloc_base/memory/ref_counted.cc",
+    "partition_alloc_base/memory/ref_counted.h",
+    "partition_alloc_base/memory/scoped_policy.h",
+    "partition_alloc_base/memory/scoped_refptr.h",
+    "partition_alloc_base/no_destructor.h",
+    "partition_alloc_base/notreached.h",
+    "partition_alloc_base/numerics/checked_math.h",
+    "partition_alloc_base/numerics/checked_math_impl.h",
+    "partition_alloc_base/numerics/clamped_math.h",
+    "partition_alloc_base/numerics/clamped_math_impl.h",
+    "partition_alloc_base/numerics/safe_conversions.h",
+    "partition_alloc_base/numerics/safe_conversions_arm_impl.h",
+    "partition_alloc_base/numerics/safe_conversions_impl.h",
+    "partition_alloc_base/numerics/safe_math.h",
+    "partition_alloc_base/numerics/safe_math_arm_impl.h",
+    "partition_alloc_base/numerics/safe_math_clang_gcc_impl.h",
+    "partition_alloc_base/numerics/safe_math_shared_impl.h",
+    "partition_alloc_base/posix/eintr_wrapper.h",
+    "partition_alloc_base/process/process_handle.h",
+    "partition_alloc_base/rand_util.cc",
+    "partition_alloc_base/rand_util.h",
+    "partition_alloc_base/scoped_clear_last_error.h",
+    "partition_alloc_base/strings/cstring_builder.cc",
+    "partition_alloc_base/strings/cstring_builder.h",
+    "partition_alloc_base/strings/safe_sprintf.cc",
+    "partition_alloc_base/strings/safe_sprintf.h",
+    "partition_alloc_base/strings/string_util.cc",
+    "partition_alloc_base/strings/string_util.h",
+    "partition_alloc_base/strings/stringprintf.cc",
+    "partition_alloc_base/strings/stringprintf.h",
+    "partition_alloc_base/system/sys_info.h",
+    "partition_alloc_base/thread_annotations.h",
+    "partition_alloc_base/threading/platform_thread.cc",
+    "partition_alloc_base/threading/platform_thread.h",
+    "partition_alloc_base/threading/platform_thread_ref.h",
+    "partition_alloc_base/time/time.cc",
+    "partition_alloc_base/time/time.h",
+    "partition_alloc_base/time/time_override.cc",
+    "partition_alloc_base/time/time_override.h",
+    "partition_alloc_base/types/strong_alias.h",
+    "partition_alloc_base/win/win_handle_types.h",
+    "partition_alloc_base/win/win_handle_types_list.inc",
+    "partition_alloc_base/win/windows_types.h",
+  ]
+
+  if (is_win) {
+    sources += [
+      "partition_alloc_base/debug/stack_trace_win.cc",
+      "partition_alloc_base/memory/page_size_win.cc",
+      "partition_alloc_base/process/process_handle_win.cc",
+      "partition_alloc_base/rand_util_win.cc",
+      "partition_alloc_base/scoped_clear_last_error_win.cc",
+      "partition_alloc_base/threading/platform_thread_win.cc",
+      "partition_alloc_base/time/time_win.cc",
+    ]
+  } else if (is_posix) {
+    sources += [
+      "partition_alloc_base/debug/stack_trace_posix.cc",
+      "partition_alloc_base/files/file_util.h",
+      "partition_alloc_base/files/file_util_posix.cc",
+      "partition_alloc_base/memory/page_size_posix.cc",
+      "partition_alloc_base/posix/safe_strerror.cc",
+      "partition_alloc_base/posix/safe_strerror.h",
+      "partition_alloc_base/process/process_handle_posix.cc",
+      "partition_alloc_base/rand_util_posix.cc",
+      "partition_alloc_base/threading/platform_thread_internal_posix.h",
+      "partition_alloc_base/threading/platform_thread_posix.cc",
+      "partition_alloc_base/time/time_conversion_posix.cc",
+    ]
+
+    if (is_linux || is_chromeos) {
+      sources += [ "partition_alloc_base/debug/stack_trace_linux.cc" ]
+    }
+
+    if (is_android || is_chromeos_ash) {
+      sources += [ "partition_alloc_base/time/time_android.cc" ]
+    }
+    if (is_apple) {
+      sources += [
+        "partition_alloc_base/debug/stack_trace_mac.cc",
+        "partition_alloc_base/time/time_apple.mm",
+      ]
+    } else {
+      sources += [ "partition_alloc_base/time/time_now_posix.cc" ]
+    }
+  } else if (is_fuchsia) {
+    sources += [
+      "partition_alloc_base/fuchsia/fuchsia_logging.cc",
+      "partition_alloc_base/fuchsia/fuchsia_logging.h",
+      "partition_alloc_base/memory/page_size_posix.cc",
+      "partition_alloc_base/posix/safe_strerror.cc",
+      "partition_alloc_base/posix/safe_strerror.h",
+      "partition_alloc_base/rand_util_fuchsia.cc",
+      "partition_alloc_base/threading/platform_thread_internal_posix.h",
+      "partition_alloc_base/threading/platform_thread_posix.cc",
+      "partition_alloc_base/time/time_conversion_posix.cc",
+      "partition_alloc_base/time/time_fuchsia.cc",
+    ]
+  }
+  if (is_android) {
+    # Only android build requires native_library, and native_library depends
+    # on file_path. So file_path is added if is_android = true.
+    sources += [
+      "partition_alloc_base/debug/stack_trace_android.cc",
+      "partition_alloc_base/files/file_path.cc",
+      "partition_alloc_base/files/file_path.h",
+      "partition_alloc_base/native_library.cc",
+      "partition_alloc_base/native_library.h",
+      "partition_alloc_base/native_library_posix.cc",
+    ]
+  }
+  if (is_apple) {
+    # Apple-specific utilities
+    sources += [
+      "partition_alloc_base/apple/foundation_util.h",
+      "partition_alloc_base/apple/foundation_util.mm",
+      "partition_alloc_base/apple/mach_logging.cc",
+      "partition_alloc_base/apple/mach_logging.h",
+      "partition_alloc_base/apple/scoped_cftyperef.h",
+      "partition_alloc_base/apple/scoped_typeref.h",
+    ]
+    if (is_ios) {
+      sources += [
+        "partition_alloc_base/ios/ios_util.h",
+        "partition_alloc_base/ios/ios_util.mm",
+        "partition_alloc_base/system/sys_info_ios.mm",
+      ]
+    }
+    if (is_mac) {
+      sources += [
+        "partition_alloc_base/mac/mac_util.h",
+        "partition_alloc_base/mac/mac_util.mm",
+        "partition_alloc_base/system/sys_info_mac.mm",
+      ]
+    }
+  }
+
+  public_deps = [
+    ":chromecast_buildflags",
+    ":chromeos_buildflags",
+    ":debugging_buildflags",
+    ":partition_alloc_buildflags",
+  ]
+
+  configs += [ ":partition_alloc_implementation" ]
+
+  deps = []
+  if (is_fuchsia) {
+    public_deps += [ "//third_party/fuchsia-sdk/sdk/pkg/fit" ]
+  }
+
+  frameworks = []
+  if (is_apple) {
+    frameworks += [
+      "CoreFoundation.framework",
+      "Foundation.framework",
+    ]
+  }
+
+  configs -= _remove_configs
+  configs += _add_configs
+}
+
+source_set("allocator_shim") {
+  visibility = [ ":*" ]
+
+  sources = []
+  deps = [ ":allocator_base" ]
+  all_dependent_configs = []
+  configs += [ ":partition_alloc_implementation" ]
+
+  configs -= _remove_configs
+  configs += _add_configs
+
+  if (use_allocator_shim) {
+    sources += [
+      "shim/allocator_shim.cc",
+      "shim/allocator_shim.h",
+      "shim/allocator_shim_internals.h",
+    ]
+    if (use_partition_alloc) {
+      sources += [
+        "shim/allocator_shim_default_dispatch_to_partition_alloc.cc",
+        "shim/allocator_shim_default_dispatch_to_partition_alloc.h",
+        "shim/nonscannable_allocator.cc",
+        "shim/nonscannable_allocator.h",
+      ]
+    }
+    if (is_android) {
+      sources += [
+        "shim/allocator_shim_override_cpp_symbols.h",
+        "shim/allocator_shim_override_linker_wrapped_symbols.h",
+      ]
+      all_dependent_configs += [ ":wrap_malloc_symbols" ]
+    }
+    if (is_apple) {
+      sources += [
+        "shim/allocator_shim_override_apple_default_zone.h",
+        "shim/allocator_shim_override_apple_symbols.h",
+        "shim/early_zone_registration_constants.h",
+      ]
+      configs += [ ":mac_no_default_new_delete_symbols" ]
+    }
+    if (is_chromeos || is_linux) {
+      sources += [
+        "shim/allocator_shim_override_cpp_symbols.h",
+        "shim/allocator_shim_override_glibc_weak_symbols.h",
+        "shim/allocator_shim_override_libc_symbols.h",
+      ]
+    }
+    if (is_win) {
+      sources += [
+        "shim/allocator_shim_override_ucrt_symbols_win.h",
+        "shim/winheap_stubs_win.cc",
+        "shim/winheap_stubs_win.h",
+      ]
+    }
+
+    if (!use_partition_alloc_as_malloc) {
+      if (is_android) {
+        sources += [
+          "shim/allocator_shim_default_dispatch_to_linker_wrapped_symbols.cc",
+        ]
+      }
+      if (is_apple) {
+        sources +=
+            [ "shim/allocator_shim_default_dispatch_to_apple_zoned_malloc.cc" ]
+      }
+      if (is_chromeos || is_linux) {
+        sources += [ "shim/allocator_shim_default_dispatch_to_glibc.cc" ]
+      }
+      if (is_win) {
+        sources += [ "shim/allocator_shim_default_dispatch_to_winheap.cc" ]
+      }
+    }
+
+    deps += [
+      ":allocator_base",
+      ":allocator_core",
+      ":buildflags",
+    ]
+  }
+
+  if (is_apple) {
+    sources += [
+      "shim/allocator_interception_apple.h",
+      "shim/allocator_interception_apple.mm",
+      "shim/malloc_zone_functions_apple.cc",
+      "shim/malloc_zone_functions_apple.h",
+    ]
+
+    # Do not compile with ARC because this target has to interface with
+    # low-level Objective-C and having ARC would interfere.
+    configs -= [ "//build/config/compiler:enable_arc" ]
+    deps += [
+      ":allocator_base",
+      ":allocator_core",
+      ":buildflags",
+    ]
+  }
+}
+
+source_set("raw_ptr") {
+  # `gn check` is unhappy with most `#includes` when PA isn't
+  # actually built.
+  check_includes = use_partition_alloc
+  public = [
+    "pointers/raw_ptr.h",
+    "pointers/raw_ptr_cast.h",
+    "pointers/raw_ptr_exclusion.h",
+    "pointers/raw_ref.h",
+  ]
+  sources = []
+  if (enable_backup_ref_ptr_support) {
+    sources += [
+      "pointers/raw_ptr_backup_ref_impl.cc",
+      "pointers/raw_ptr_backup_ref_impl.h",
+    ]
+  } else if (use_hookable_raw_ptr) {
+    sources += [
+      "pointers/raw_ptr_hookable_impl.cc",
+      "pointers/raw_ptr_hookable_impl.h",
+    ]
+  } else if (use_asan_unowned_ptr) {
+    sources += [
+      "pointers/raw_ptr_asan_unowned_impl.cc",
+      "pointers/raw_ptr_asan_unowned_impl.h",
+    ]
+  } else {
+    sources += [ "pointers/raw_ptr_noop_impl.h" ]
+  }
+  if (use_partition_alloc) {
+    public_deps = [ ":partition_alloc" ]
+  }
+  deps = [ ":buildflags" ]
+
+  # See also: `partition_alloc_base/component_export.h`
+  defines = [ "IS_RAW_PTR_IMPL" ]
+
+  configs -= _remove_configs
+  configs += _add_configs
+}
+
+buildflag_header("partition_alloc_buildflags") {
+  header = "partition_alloc_buildflags.h"
+
+  _record_alloc_info = false
+
+  # GWP-ASan is tied to BRP's "refcount in previous slot" mode, whose
+  # enablement is already gated on BRP enablement.
+  _enable_gwp_asan_support = put_ref_count_in_previous_slot
+
+  # Pools are a logical concept when address space is 32-bit.
+  _glue_core_pools = glue_core_pools && has_64_bit_pointers
+
+  # Pointer compression requires 64-bit pointers.
+  _enable_pointer_compression =
+      enable_pointer_compression_support && has_64_bit_pointers
+
+  # Force-enable live BRP in all processes, ignoring the canonical
+  # experiment state of `PartitionAllocBackupRefPtr`.
+  #
+  # This is not exposed as a GN arg as it is not meant to be used by
+  # developers - it is simply a compile-time hinge that should be
+  # set in the experimental build and then reverted immediately.
+  _force_all_process_brp = false
+
+  # TODO(crbug.com/1151236): Need to refactor the following buildflags.
+  # The buildflags (except RECORD_ALLOC_INFO) are used by both chrome and
+  # partition alloc. For partition alloc,
+  # gen/base/allocator/partition_allocator/partition_alloc_buildflags.h
+  # defines and partition alloc includes the header file. For chrome,
+  # gen/base/allocator/buildflags.h defines and chrome includes.
+  flags = [
+    "HAS_64_BIT_POINTERS=$has_64_bit_pointers",
+
+    "USE_ALLOCATOR_SHIM=$use_allocator_shim",
+    "USE_PARTITION_ALLOC=$use_partition_alloc",
+    "USE_PARTITION_ALLOC_AS_MALLOC=$use_partition_alloc_as_malloc",
+
+    "ENABLE_BACKUP_REF_PTR_SUPPORT=$enable_backup_ref_ptr_support",
+    "ENABLE_BACKUP_REF_PTR_SLOW_CHECKS=$enable_backup_ref_ptr_slow_checks",
+    "ENABLE_BACKUP_REF_PTR_FEATURE_FLAG=$enable_backup_ref_ptr_feature_flag",
+    "ENABLE_DANGLING_RAW_PTR_CHECKS=$enable_dangling_raw_ptr_checks",
+    "ENABLE_DANGLING_RAW_PTR_FEATURE_FLAG=$enable_dangling_raw_ptr_feature_flag",
+    "ENABLE_DANGLING_RAW_PTR_PERF_EXPERIMENT=$enable_dangling_raw_ptr_perf_experiment",
+    "ENABLE_POINTER_SUBTRACTION_CHECK=$enable_pointer_subtraction_check",
+    "BACKUP_REF_PTR_POISON_OOB_PTR=$backup_ref_ptr_poison_oob_ptr",
+    "PUT_REF_COUNT_IN_PREVIOUS_SLOT=$put_ref_count_in_previous_slot",
+    "USE_ASAN_BACKUP_REF_PTR=$use_asan_backup_ref_ptr",
+    "USE_ASAN_UNOWNED_PTR=$use_asan_unowned_ptr",
+    "USE_HOOKABLE_RAW_PTR=$use_hookable_raw_ptr",
+    "ENABLE_GWP_ASAN_SUPPORT=$_enable_gwp_asan_support",
+    "FORCIBLY_ENABLE_BACKUP_REF_PTR_IN_ALL_PROCESSES=$_force_all_process_brp",
+
+    "FORCE_ENABLE_RAW_PTR_EXCLUSION=$force_enable_raw_ptr_exclusion",
+
+    "RECORD_ALLOC_INFO=$_record_alloc_info",
+    "USE_FREESLOT_BITMAP=$use_freeslot_bitmap",
+    "GLUE_CORE_POOLS=$_glue_core_pools",
+    "ENABLE_POINTER_COMPRESSION=$_enable_pointer_compression",
+    "ENABLE_SHADOW_METADATA_FOR_64_BITS_POINTERS=$enable_shadow_metadata",
+    "USE_FREELIST_POOL_OFFSETS=$use_freelist_pool_offsets",
+
+    "USE_STARSCAN=$use_starscan",
+    "PCSCAN_STACK_SUPPORTED=$pcscan_stack_supported",
+
+    "ENABLE_PKEYS=$enable_pkeys",
+    "ENABLE_THREAD_ISOLATION=$enable_pkeys",
+  ]
+
+  if (is_apple) {
+    # TODO(crbug.com/1414153): once TimeTicks::Now behavior is unified on iOS,
+    # this should be removed.
+    flags += [ "PARTITION_ALLOC_ENABLE_MACH_ABSOLUTE_TIME_TICKS=" +
+               "$partition_alloc_enable_mach_absolute_time_ticks" ]
+  }
+}
+
+buildflag_header("chromecast_buildflags") {
+  header = "chromecast_buildflags.h"
+
+  flags = [
+    "PA_IS_CAST_ANDROID=$is_cast_android",
+    "PA_IS_CASTOS=$is_castos",
+  ]
+}
+
+buildflag_header("chromeos_buildflags") {
+  header = "chromeos_buildflags.h"
+
+  flags = [ "PA_IS_CHROMEOS_ASH=$is_chromeos_ash" ]
+}
+
+buildflag_header("debugging_buildflags") {
+  header = "debugging_buildflags.h"
+  header_dir = rebase_path(".", "//") + "/partition_alloc_base/debug"
+
+  # Duplicates the setup Chromium uses to define `DCHECK_IS_ON()`,
+  # but avails it as a buildflag.
+  _dcheck_is_on = is_debug || dcheck_always_on
+
+  flags = [
+    "PA_DCHECK_IS_ON=$_dcheck_is_on",
+    "PA_EXPENSIVE_DCHECKS_ARE_ON=$enable_expensive_dchecks",
+    "PA_DCHECK_IS_CONFIGURABLE=$dcheck_is_configurable",
+    "PA_CAN_UNWIND_WITH_FRAME_POINTERS=$can_unwind_with_frame_pointers",
+  ]
+}
+
+group("buildflags") {
+  public_deps = [
+    ":chromecast_buildflags",
+    ":chromeos_buildflags",
+    ":debugging_buildflags",
+    ":partition_alloc_buildflags",
+  ]
+}
+# TODO(crbug.com/1151236): After making partition_alloc a standalone library,
+# move test code here. i.e. test("partition_alloc_tests") { ... } and
+# test("partition_alloc_perftests").

src/base/allocator/partition_allocator/DEPS

@@ -0,0 +1,49 @@
+# Copyright 2021 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+include_rules = [
+    # The basic rule is that PartitionAlloc library must not depend on Chromium
+    # project in order to be a standalone library.
+    "-base",
+    "-build",
+    "-build_overrides",
+    "-testing",
+    "-third_party",
+
+    "+base/allocator/partition_allocator",
+    "+build/build_config.h",
+    "+build/buildflag.h",
+]
+
+specific_include_rules = {
+  ".*_(perf|unit)test\.cc$": [
+    "+base/allocator/allocator_shim_default_dispatch_to_partition_alloc.h",
+    "+base/allocator/dispatcher/dispatcher.h",
+    "+base/debug/allocation_trace.h",
+    "+base/debug/debugging_buildflags.h",
+    "+base/debug/proc_maps_linux.h",
+    "+base/system/sys_info.h",
+    "+base/test/gtest_util.h",
+    "+base/timer/lap_timer.h",
+    "+base/win/windows_version.h",
+    "+testing/gmock/include/gmock/gmock.h",
+    "+testing/gtest/include/gtest/gtest.h",
+    "+testing/perf/perf_result_reporter.h",
+  ],
+  "extended_api\.cc$": [
+    "+base/allocator/allocator_shim_default_dispatch_to_partition_alloc.h",
+  ],
+  "gtest_prod_util\.h$": [
+    "+testing/gtest/include/gtest/gtest_prod.h",
+  ],
+  "raw_(ptr|ref)_unittest\.cc$": [
+    "+base",
+    "+third_party/abseil-cpp/absl/types/optional.h",
+    "+third_party/abseil-cpp/absl/types/variant.h",
+  ],
+  "raw_ptr_test_support\.h$": [
+    "+testing/gmock/include/gmock/gmock.h",
+    "+third_party/abseil-cpp/absl/types/optional.h",
+  ]
+}

src/base/allocator/partition_allocator/DIR_METADATA

@@ -0,0 +1,6 @@
+monorail {
+  component: "Blink>MemoryAllocator>Partition"
+}
+
+# Also security-dev@chromium.org
+team_email: "platform-architecture-dev@chromium.org"

src/base/allocator/partition_allocator/OWNERS

@@ -0,0 +1,8 @@
+bartekn@chromium.org
+haraken@chromium.org
+keishi@chromium.org
+lizeb@chromium.org
+tasak@google.com
+
+per-file pointers/raw_ptr*=file://base/memory/MIRACLE_PTR_OWNERS
+per-file pointers/raw_ref*=file://base/memory/MIRACLE_PTR_OWNERS

src/base/allocator/partition_allocator/PartitionAlloc.md

@@ -0,0 +1,203 @@
+# PartitionAlloc Design
+
+This document describes PartitionAlloc at a high level, with some architectural
+details. For implementation details, see the comments in
+`partition_alloc_constants.h`.
+
+## Quick Links
+
+* [Glossary](./glossary.md): Definitions of terms commonly used in
+  PartitionAlloc. The present document largely avoids defining terms.
+
+* [Build Config](./build_config.md): Pertinent GN args, buildflags, and
+  macros.
+
+* [Chrome-External Builds](./external_builds.md): Further considerations
+  for standalone PartitionAlloc, plus an embedder's guide for some extra
+  GN args.
+
+## Overview
+
+PartitionAlloc is a memory allocator optimized for space efficiency,
+allocation latency, and security.
+
+### Performance
+
+PartitionAlloc is designed to be extremely fast in its fast paths. The fast
+paths of allocation and deallocation require very few (reasonably predictable)
+branches. The number of operations in the fast paths is minimal, leading to the
+possibility of inlining.
+
+![The central allocator manages slots and spans. It is locked on a
+  per-partition basis. Separately, the thread cache consumes slots
+  from the central allocator, allowing it to hand out memory
+  quickly to individual threads.](./dot/layers.png)
+
+However, even the fast path isn't the fastest, because it requires taking
+a per-partition lock. Although we optimized the lock, there was still room for
+improvement; to this end, we introduced the thread cache.
+The thread cache has been tailored to satisfy a vast majority of requests by
+allocating from and releasing memory to the main allocator in batches,
+amortizing lock acquisition and further improving locality while not trapping
+excess memory.
+
+### Security
+
+Security is one of the important goals of PartitionAlloc.
+
+PartitionAlloc guarantees that different partitions exist in different regions
+of the process's address space. When the caller has freed all objects contained
+in a page in a partition, PartitionAlloc returns the physical memory to the
+operating system, but continues to reserve the region of address space.
+PartitionAlloc will only reuse an address space region for the same partition.
+
+Similarly, one page can contain only objects from the same bucket.
+When freed, PartitionAlloc returns the physical memory, but continues to reserve
+the region for this very bucket.
+
+The above techniques help avoid type confusion attacks. Note, however, these
+apply only to normal buckets and not to direct map, as it'd waste too much
+address space.
+
+PartitionAlloc also guarantees that:
+
+* Linear overflows/underflows cannot corrupt into, out of, or between
+  partitions. There are guard pages at the beginning and the end of each memory
+  region owned by a partition.
+
+* Linear overflows/underflows cannot corrupt the allocation metadata.
+  PartitionAlloc records metadata in a dedicated, out-of-line region (not
+  adjacent to objects), surrounded by guard pages. (Freelist pointers are an
+  exception.)
+
+* Partial pointer overwrite of freelist pointer should fault.
+
+* Direct map allocations have guard pages at the beginning and the end.
+
+### Alignment
+
+PartitionAlloc guarantees that returned pointers are aligned on
+`partition_alloc::internal::kAlignment` boundary (typically 16B on
+64-bit systems, and 8B on 32-bit).
+
+PartitionAlloc also supports higher levels of alignment, that can be requested
+via `PartitionAlloc::AlignedAlloc()` or platform-specific APIs (such as
+`posix_memalign()`). The requested
+alignment has to be a power of two. PartitionAlloc reserves the right to round
+up the requested size to the nearest power of two, greater than or equal to the
+requested alignment. This may be wasteful, but allows taking advantage of
+natural PartitionAlloc alignment guarantees. Allocations with an alignment
+requirement greater than `partition_alloc::internal::kAlignment` are expected
+to be very rare.
+
+## Architecture
+
+### Layout in Memory
+
+PartitionAlloc handles normal buckets by reserving (not committing) 2MiB super
+pages. Each super page is split into partition pages.
+The first and the last partition page are permanently inaccessible and serve
+as guard pages, with the exception of one system page in the middle of the first
+partition page that holds metadata (32B struct per partition page).
+
+![A super page is shown full of slot spans. The slot spans are logically
+  strung together to form buckets. At both extremes of the super page
+  are guard pages. PartitionAlloc metadata is hidden inside the
+  guard pages at the "front."](./dot/super-page.png)
+
+* The slot span numbers provide a visual hint of their size (in partition
+  pages).
+* Colors provide a visual hint of the bucket to which the slot span belongs.
+    * Although only five colors are shown, in reality, a super page holds
+      tens of slot spans, some of which belong to the same bucket.
+* The system page that holds metadata tracks each partition page with one 32B
+  [`PartitionPage` struct][PartitionPage], which is either
+    * a [`SlotSpanMetadata`][SlotSpanMetadata] ("v"s in the diagram) or
+    * a [`SubsequentPageMetadata`][SubsequentPageMetadata] ("+"s in the
+      diagram).
+* Gray fill denotes guard pages (one partition page each at the head and tail
+  of each super page).
+* In some configurations, PartitionAlloc stores more metadata than can
+  fit in the one system page at the front. These are the bitmaps for
+  StarScan and `MTECheckedPtr<T>`, and they are relegated to the head of
+  what would otherwise be usable space for slot spans. One, both, or
+  none of these bitmaps may be present, depending on build
+  configuration, runtime configuration, and type of allocation.
+  See [`SuperPagePayloadBegin()`][payload-start] for details.
+
+As allocation requests arrive, there is eventually a need to allocate a new slot
+span.
+Address space for such a slot span is carved out from the last super page. If
+not enough space, a new super page is allocated. Due to varying sizes of slot
+span, this may lead to leaving space unused (we never go back to fill previous
+super pages), which is fine because this memory is merely reserved, which is far
+less precious than committed memory. Note also that address space reserved for a
+slot span is never released, even if the slot span isn't used for a long time.
+
+All slots in a newly allocated slot span are *free*, i.e. available for
+allocation.
+
+### Freelist Pointers
+
+All free slots within a slot span are chained into a singly-linked free-list,
+by writing the *next* pointer at the beginning of each slot, and the head of the
+list is written in the metadata struct.
+
+However, writing a pointer in each free slot of a newly allocated span would
+require committing and faulting in physical pages upfront, which would be
+unacceptable. Therefore, PartitionAlloc has a concept of *provisioning slots*.
+Only provisioned slots are chained into the freelist.
+Once provisioned slots in a span are depleted, then another page worth of slots
+is provisioned (note, a slot that crosses a page boundary only gets
+provisioned with slots of the next page). See
+`PartitionBucket::ProvisionMoreSlotsAndAllocOne()` for more details.
+
+Freelist pointers are stored at the beginning of each free slot. As such, they
+are the only metadata that is inline, i.e. stored among the
+objects. This makes them prone to overruns. On little-endian systems, the
+pointers are encoded by reversing byte order, so that partial overruns will very
+likely result in destroying the pointer, as opposed to forming a valid pointer
+to a nearby location.
+
+Furthermore, a shadow of a freelist pointer is stored next to it, encoded in a
+different manner. This helps PartitionAlloc detect corruptions.
+
+### Slot Span States
+
+A slot span can be in any of 4 states:
+* *Full*. A full span has no free slots.
+* *Empty*. An empty span has no allocated slots, only free slots.
+* *Active*. An active span is anything in between the above two.
+* *Decommitted*. A decommitted span is a special case of an empty span, where
+  all pages are decommitted from memory.
+
+PartitionAlloc prioritizes getting an available slot from an active span, over
+an empty one, in hope that the latter can be soon transitioned into a
+decommitted state, thus releasing memory. There is no mechanism, however, to
+prioritize selection of a slot span based on the number of already allocated
+slots.
+
+An empty span becomes decommitted either when there are too many empty spans
+(FIFO), or when `PartitionRoot::PurgeMemory()` gets invoked periodically (or in
+low memory pressure conditions). An allocation can be satisfied from
+a decommitted span if there are no active or empty spans available. The slot
+provisioning mechanism kicks back in, committing the pages gradually as needed,
+and the span becomes active. (There is currently no other way
+to unprovision slots than decommitting the entire span).
+
+As mentioned above, a bucket is a collection of slot spans containing slots of
+the same size. In fact, each bucket has 3 linked-lists, chaining active, empty
+and decommitted spans (see `PartitionBucket::*_slot_spans_head`).
+There is no need for a full span list. The lists are updated lazily. An empty,
+decommitted or full span may stay on the active list for some time, until
+`PartitionBucket::SetNewActiveSlotSpan()` encounters it.
+A decommitted span may stay on the empty list for some time,
+until `PartitionBucket::SlowPathAlloc()` encounters it. However,
+the inaccuracy can't happen in the other direction, i.e. an active span can only
+be on the active list, and an empty span can only be on the active or empty
+list.
+
+[PartitionPage]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/partition_allocator/partition_page.h;l=314;drc=e5b03e85ea180d1d1ab0dec471c7fd5d1706a9e4
+[SlotSpanMetadata]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/partition_allocator/partition_page.h;l=120;drc=e5b03e85ea180d1d1ab0dec471c7fd5d1706a9e4
+[SubsequentPageMetadata]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/partition_allocator/partition_page.h;l=295;drc=e5b03e85ea180d1d1ab0dec471c7fd5d1706a9e4
+[payload-start]: https://source.chromium.org/chromium/chromium/src/+/35b2deed603dedd4abb37f204d516ed62aa2b85c:base/allocator/partition_allocator/partition_page.h;l=454

src/base/allocator/partition_allocator/address_pool_manager.cc

@@ -0,0 +1,571 @@
+// Copyright 2020 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/address_pool_manager.h"
+
+#include <algorithm>
+#include <atomic>
+#include <cstdint>
+#include <limits>
+
+#include "base/allocator/partition_allocator/address_space_stats.h"
+#include "base/allocator/partition_allocator/page_allocator.h"
+#include "base/allocator/partition_allocator/page_allocator_constants.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/debug/debugging_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/notreached.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+#include "base/allocator/partition_allocator/reservation_offset_table.h"
+#include "base/allocator/partition_allocator/thread_isolation/alignment.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_APPLE) || BUILDFLAG(ENABLE_THREAD_ISOLATION)
+#include <sys/mman.h>
+#endif
+
+namespace partition_alloc::internal {
+
+AddressPoolManager AddressPoolManager::singleton_;
+
+// static
+AddressPoolManager& AddressPoolManager::GetInstance() {
+  return singleton_;
+}
+
+namespace {
+// Allocations are all performed on behalf of PartitionAlloc.
+constexpr PageTag kPageTag = PageTag::kPartitionAlloc;
+
+}  // namespace
+
+#if BUILDFLAG(HAS_64_BIT_POINTERS)
+
+namespace {
+
+// This will crash if the range cannot be decommitted.
+void DecommitPages(uintptr_t address, size_t size) {
+  // Callers rely on the pages being zero-initialized when recommitting them.
+  // |DecommitSystemPages| doesn't guarantee this on all operating systems, in
+  // particular on macOS, but |DecommitAndZeroSystemPages| does.
+  DecommitAndZeroSystemPages(address, size, kPageTag);
+}
+
+}  // namespace
+
+void AddressPoolManager::Add(pool_handle handle, uintptr_t ptr, size_t length) {
+  PA_DCHECK(!(ptr & kSuperPageOffsetMask));
+  PA_DCHECK(!((ptr + length) & kSuperPageOffsetMask));
+  PA_CHECK(handle > 0 && handle <= std::size(pools_));
+
+  Pool* pool = GetPool(handle);
+  PA_CHECK(!pool->IsInitialized());
+  pool->Initialize(ptr, length);
+}
+
+void AddressPoolManager::GetPoolUsedSuperPages(
+    pool_handle handle,
+    std::bitset<kMaxSuperPagesInPool>& used) {
+  Pool* pool = GetPool(handle);
+  if (!pool) {
+    return;
+  }
+
+  pool->GetUsedSuperPages(used);
+}
+
+uintptr_t AddressPoolManager::GetPoolBaseAddress(pool_handle handle) {
+  Pool* pool = GetPool(handle);
+  if (!pool) {
+    return 0;
+  }
+
+  return pool->GetBaseAddress();
+}
+
+void AddressPoolManager::ResetForTesting() {
+  for (size_t i = 0; i < std::size(pools_); ++i) {
+    pools_[i].Reset();
+  }
+}
+
+void AddressPoolManager::Remove(pool_handle handle) {
+  Pool* pool = GetPool(handle);
+  PA_DCHECK(pool->IsInitialized());
+  pool->Reset();
+}
+
+uintptr_t AddressPoolManager::Reserve(pool_handle handle,
+                                      uintptr_t requested_address,
+                                      size_t length) {
+  Pool* pool = GetPool(handle);
+  if (!requested_address) {
+    return pool->FindChunk(length);
+  }
+  const bool is_available = pool->TryReserveChunk(requested_address, length);
+  if (is_available) {
+    return requested_address;
+  }
+  return pool->FindChunk(length);
+}
+
+void AddressPoolManager::UnreserveAndDecommit(pool_handle handle,
+                                              uintptr_t address,
+                                              size_t length) {
+  PA_DCHECK(kNullPoolHandle < handle && handle <= kNumPools);
+  Pool* pool = GetPool(handle);
+  PA_DCHECK(pool->IsInitialized());
+  DecommitPages(address, length);
+  pool->FreeChunk(address, length);
+}
+
+void AddressPoolManager::Pool::Initialize(uintptr_t ptr, size_t length) {
+  PA_CHECK(ptr != 0);
+  PA_CHECK(!(ptr & kSuperPageOffsetMask));
+  PA_CHECK(!(length & kSuperPageOffsetMask));
+  address_begin_ = ptr;
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+  address_end_ = ptr + length;
+  PA_DCHECK(address_begin_ < address_end_);
+#endif
+
+  total_bits_ = length / kSuperPageSize;
+  PA_CHECK(total_bits_ <= kMaxSuperPagesInPool);
+
+  ScopedGuard scoped_lock(lock_);
+  alloc_bitset_.reset();
+  bit_hint_ = 0;
+}
+
+bool AddressPoolManager::Pool::IsInitialized() {
+  return address_begin_ != 0;
+}
+
+void AddressPoolManager::Pool::Reset() {
+  address_begin_ = 0;
+}
+
+void AddressPoolManager::Pool::GetUsedSuperPages(
+    std::bitset<kMaxSuperPagesInPool>& used) {
+  ScopedGuard scoped_lock(lock_);
+
+  PA_DCHECK(IsInitialized());
+  used = alloc_bitset_;
+}
+
+uintptr_t AddressPoolManager::Pool::GetBaseAddress() {
+  PA_DCHECK(IsInitialized());
+  return address_begin_;
+}
+
+uintptr_t AddressPoolManager::Pool::FindChunk(size_t requested_size) {
+  ScopedGuard scoped_lock(lock_);
+
+  PA_DCHECK(!(requested_size & kSuperPageOffsetMask));
+  const size_t need_bits = requested_size >> kSuperPageShift;
+
+  // Use first-fit policy to find an available chunk from free chunks. Start
+  // from |bit_hint_|, because we know there are no free chunks before.
+  size_t beg_bit = bit_hint_;
+  size_t curr_bit = bit_hint_;
+  while (true) {
+    // |end_bit| points 1 past the last bit that needs to be 0. If it goes past
+    // |total_bits_|, return |nullptr| to signal no free chunk was found.
+    size_t end_bit = beg_bit + need_bits;
+    if (end_bit > total_bits_) {
+      return 0;
+    }
+
+    bool found = true;
+    for (; curr_bit < end_bit; ++curr_bit) {
+      if (alloc_bitset_.test(curr_bit)) {
+        // The bit was set, so this chunk isn't entirely free. Set |found=false|
+        // to ensure the outer loop continues. However, continue the inner loop
+        // to set |beg_bit| just past the last set bit in the investigated
+        // chunk. |curr_bit| is advanced all the way to |end_bit| to prevent the
+        // next outer loop pass from checking the same bits.
+        beg_bit = curr_bit + 1;
+        found = false;
+        if (bit_hint_ == curr_bit) {
+          ++bit_hint_;
+        }
+      }
+    }
+
+    // An entire [beg_bit;end_bit) region of 0s was found. Fill them with 1s (to
+    // mark as allocated) and return the allocated address.
+    if (found) {
+      for (size_t i = beg_bit; i < end_bit; ++i) {
+        PA_DCHECK(!alloc_bitset_.test(i));
+        alloc_bitset_.set(i);
+      }
+      if (bit_hint_ == beg_bit) {
+        bit_hint_ = end_bit;
+      }
+      uintptr_t address = address_begin_ + beg_bit * kSuperPageSize;
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+      PA_DCHECK(address + requested_size <= address_end_);
+#endif
+      return address;
+    }
+  }
+
+  PA_NOTREACHED();
+}
+
+bool AddressPoolManager::Pool::TryReserveChunk(uintptr_t address,
+                                               size_t requested_size) {
+  ScopedGuard scoped_lock(lock_);
+  PA_DCHECK(!(address & kSuperPageOffsetMask));
+  PA_DCHECK(!(requested_size & kSuperPageOffsetMask));
+  const size_t begin_bit = (address - address_begin_) / kSuperPageSize;
+  const size_t need_bits = requested_size / kSuperPageSize;
+  const size_t end_bit = begin_bit + need_bits;
+  // Check that requested address is not too high.
+  if (end_bit > total_bits_) {
+    return false;
+  }
+  // Check if any bit of the requested region is set already.
+  for (size_t i = begin_bit; i < end_bit; ++i) {
+    if (alloc_bitset_.test(i)) {
+      return false;
+    }
+  }
+  // Otherwise, set the bits.
+  for (size_t i = begin_bit; i < end_bit; ++i) {
+    alloc_bitset_.set(i);
+  }
+  return true;
+}
+
+void AddressPoolManager::Pool::FreeChunk(uintptr_t address, size_t free_size) {
+  ScopedGuard scoped_lock(lock_);
+
+  PA_DCHECK(!(address & kSuperPageOffsetMask));
+  PA_DCHECK(!(free_size & kSuperPageOffsetMask));
+
+  PA_DCHECK(address_begin_ <= address);
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+  PA_DCHECK(address + free_size <= address_end_);
+#endif
+
+  const size_t beg_bit = (address - address_begin_) / kSuperPageSize;
+  const size_t end_bit = beg_bit + free_size / kSuperPageSize;
+  for (size_t i = beg_bit; i < end_bit; ++i) {
+    PA_DCHECK(alloc_bitset_.test(i));
+    alloc_bitset_.reset(i);
+  }
+  bit_hint_ = std::min(bit_hint_, beg_bit);
+}
+
+void AddressPoolManager::Pool::GetStats(PoolStats* stats) {
+  std::bitset<kMaxSuperPagesInPool> pages;
+  size_t i;
+  {
+    ScopedGuard scoped_lock(lock_);
+    pages = alloc_bitset_;
+    i = bit_hint_;
+  }
+
+  stats->usage = pages.count();
+
+  size_t largest_run = 0;
+  size_t current_run = 0;
+  for (; i < total_bits_; ++i) {
+    if (!pages[i]) {
+      current_run += 1;
+      continue;
+    } else if (current_run > largest_run) {
+      largest_run = current_run;
+    }
+    current_run = 0;
+  }
+
+  // Fell out of the loop with last bit being zero. Check once more.
+  if (current_run > largest_run) {
+    largest_run = current_run;
+  }
+  stats->largest_available_reservation = largest_run;
+}
+
+void AddressPoolManager::GetPoolStats(const pool_handle handle,
+                                      PoolStats* stats) {
+  Pool* pool = GetPool(handle);
+  if (!pool->IsInitialized()) {
+    return;
+  }
+  pool->GetStats(stats);
+}
+
+bool AddressPoolManager::GetStats(AddressSpaceStats* stats) {
+  // Get 64-bit pool stats.
+  GetPoolStats(kRegularPoolHandle, &stats->regular_pool_stats);
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  GetPoolStats(kBRPPoolHandle, &stats->brp_pool_stats);
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  if (IsConfigurablePoolAvailable()) {
+    GetPoolStats(kConfigurablePoolHandle, &stats->configurable_pool_stats);
+  }
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  GetPoolStats(kThreadIsolatedPoolHandle, &stats->thread_isolated_pool_stats);
+#endif
+  return true;
+}
+
+#else  // BUILDFLAG(HAS_64_BIT_POINTERS)
+
+static_assert(
+    kSuperPageSize % AddressPoolManagerBitmap::kBytesPer1BitOfBRPPoolBitmap ==
+        0,
+    "kSuperPageSize must be a multiple of kBytesPer1BitOfBRPPoolBitmap.");
+static_assert(
+    kSuperPageSize / AddressPoolManagerBitmap::kBytesPer1BitOfBRPPoolBitmap > 0,
+    "kSuperPageSize must be larger than kBytesPer1BitOfBRPPoolBitmap.");
+static_assert(AddressPoolManagerBitmap::kGuardBitsOfBRPPoolBitmap >=
+                  AddressPoolManagerBitmap::kGuardOffsetOfBRPPoolBitmap,
+              "kGuardBitsOfBRPPoolBitmap must be larger than or equal to "
+              "kGuardOffsetOfBRPPoolBitmap.");
+
+template <size_t bitsize>
+void SetBitmap(std::bitset<bitsize>& bitmap,
+               size_t start_bit,
+               size_t bit_length) {
+  const size_t end_bit = start_bit + bit_length;
+  PA_DCHECK(start_bit <= bitsize);
+  PA_DCHECK(end_bit <= bitsize);
+
+  for (size_t i = start_bit; i < end_bit; ++i) {
+    PA_DCHECK(!bitmap.test(i));
+    bitmap.set(i);
+  }
+}
+
+template <size_t bitsize>
+void ResetBitmap(std::bitset<bitsize>& bitmap,
+                 size_t start_bit,
+                 size_t bit_length) {
+  const size_t end_bit = start_bit + bit_length;
+  PA_DCHECK(start_bit <= bitsize);
+  PA_DCHECK(end_bit <= bitsize);
+
+  for (size_t i = start_bit; i < end_bit; ++i) {
+    PA_DCHECK(bitmap.test(i));
+    bitmap.reset(i);
+  }
+}
+
+uintptr_t AddressPoolManager::Reserve(pool_handle handle,
+                                      uintptr_t requested_address,
+                                      size_t length) {
+  PA_DCHECK(!(length & DirectMapAllocationGranularityOffsetMask()));
+  uintptr_t address =
+      AllocPages(requested_address, length, kSuperPageSize,
+                 PageAccessibilityConfiguration(
+                     PageAccessibilityConfiguration::kInaccessible),
+                 kPageTag);
+  return address;
+}
+
+void AddressPoolManager::UnreserveAndDecommit(pool_handle handle,
+                                              uintptr_t address,
+                                              size_t length) {
+  PA_DCHECK(!(address & kSuperPageOffsetMask));
+  PA_DCHECK(!(length & DirectMapAllocationGranularityOffsetMask()));
+  FreePages(address, length);
+}
+
+void AddressPoolManager::MarkUsed(pool_handle handle,
+                                  uintptr_t address,
+                                  size_t length) {
+  ScopedGuard scoped_lock(AddressPoolManagerBitmap::GetLock());
+  // When ENABLE_BACKUP_REF_PTR_SUPPORT is off, BRP pool isn't used.
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  if (handle == kBRPPoolHandle) {
+    PA_DCHECK(
+        (length % AddressPoolManagerBitmap::kBytesPer1BitOfBRPPoolBitmap) == 0);
+
+    // Make IsManagedByBRPPoolPool() return false when an address inside the
+    // first or the last PartitionPageSize()-bytes block is given:
+    //
+    //          ------+---+---------------+---+----
+    // memory   ..... | B | managed by PA | B | ...
+    // regions  ------+---+---------------+---+----
+    //
+    // B: PartitionPageSize()-bytes block. This is used internally by the
+    // allocator and is not available for callers.
+    //
+    // This is required to avoid crash caused by the following code:
+    //   {
+    //     // Assume this allocation happens outside of PartitionAlloc.
+    //     raw_ptr<T> ptr = new T[20];
+    //     for (size_t i = 0; i < 20; i ++) { ptr++; }
+    //     // |ptr| may point to an address inside 'B'.
+    //   }
+    //
+    // Suppose that |ptr| points to an address inside B after the loop. If
+    // IsManagedByBRPPoolPool(ptr) were to return true, ~raw_ptr<T>() would
+    // crash, since the memory is not allocated by PartitionAlloc.
+    SetBitmap(AddressPoolManagerBitmap::brp_pool_bits_,
+              (address >> AddressPoolManagerBitmap::kBitShiftOfBRPPoolBitmap) +
+                  AddressPoolManagerBitmap::kGuardOffsetOfBRPPoolBitmap,
+              (length >> AddressPoolManagerBitmap::kBitShiftOfBRPPoolBitmap) -
+                  AddressPoolManagerBitmap::kGuardBitsOfBRPPoolBitmap);
+  } else
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  {
+    PA_DCHECK(handle == kRegularPoolHandle);
+    PA_DCHECK(
+        (length % AddressPoolManagerBitmap::kBytesPer1BitOfRegularPoolBitmap) ==
+        0);
+    SetBitmap(AddressPoolManagerBitmap::regular_pool_bits_,
+              address >> AddressPoolManagerBitmap::kBitShiftOfRegularPoolBitmap,
+              length >> AddressPoolManagerBitmap::kBitShiftOfRegularPoolBitmap);
+  }
+}
+
+void AddressPoolManager::MarkUnused(pool_handle handle,
+                                    uintptr_t address,
+                                    size_t length) {
+  // Address regions allocated for normal buckets are never released, so this
+  // function can only be called for direct map. However, do not DCHECK on
+  // IsManagedByDirectMap(address), because many tests test this function using
+  // small allocations.
+
+  ScopedGuard scoped_lock(AddressPoolManagerBitmap::GetLock());
+  // When ENABLE_BACKUP_REF_PTR_SUPPORT is off, BRP pool isn't used.
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  if (handle == kBRPPoolHandle) {
+    PA_DCHECK(
+        (length % AddressPoolManagerBitmap::kBytesPer1BitOfBRPPoolBitmap) == 0);
+
+    // Make IsManagedByBRPPoolPool() return false when an address inside the
+    // first or the last PartitionPageSize()-bytes block is given.
+    // (See MarkUsed comment)
+    ResetBitmap(
+        AddressPoolManagerBitmap::brp_pool_bits_,
+        (address >> AddressPoolManagerBitmap::kBitShiftOfBRPPoolBitmap) +
+            AddressPoolManagerBitmap::kGuardOffsetOfBRPPoolBitmap,
+        (length >> AddressPoolManagerBitmap::kBitShiftOfBRPPoolBitmap) -
+            AddressPoolManagerBitmap::kGuardBitsOfBRPPoolBitmap);
+  } else
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  {
+    PA_DCHECK(handle == kRegularPoolHandle);
+    PA_DCHECK(
+        (length % AddressPoolManagerBitmap::kBytesPer1BitOfRegularPoolBitmap) ==
+        0);
+    ResetBitmap(
+        AddressPoolManagerBitmap::regular_pool_bits_,
+        address >> AddressPoolManagerBitmap::kBitShiftOfRegularPoolBitmap,
+        length >> AddressPoolManagerBitmap::kBitShiftOfRegularPoolBitmap);
+  }
+}
+
+void AddressPoolManager::ResetForTesting() {
+  ScopedGuard guard(AddressPoolManagerBitmap::GetLock());
+  AddressPoolManagerBitmap::regular_pool_bits_.reset();
+  AddressPoolManagerBitmap::brp_pool_bits_.reset();
+}
+
+namespace {
+
+// Counts super pages in use represented by `bitmap`.
+template <size_t bitsize>
+size_t CountUsedSuperPages(const std::bitset<bitsize>& bitmap,
+                           const size_t bits_per_super_page) {
+  size_t count = 0;
+  size_t bit_index = 0;
+
+  // Stride over super pages.
+  for (size_t super_page_index = 0; bit_index < bitsize; ++super_page_index) {
+    // Stride over the bits comprising the super page.
+    for (bit_index = super_page_index * bits_per_super_page;
+         bit_index < (super_page_index + 1) * bits_per_super_page &&
+         bit_index < bitsize;
+         ++bit_index) {
+      if (bitmap[bit_index]) {
+        count += 1;
+        // Move on to the next super page.
+        break;
+      }
+    }
+  }
+  return count;
+}
+
+}  // namespace
+
+bool AddressPoolManager::GetStats(AddressSpaceStats* stats) {
+  std::bitset<AddressPoolManagerBitmap::kRegularPoolBits> regular_pool_bits;
+  std::bitset<AddressPoolManagerBitmap::kBRPPoolBits> brp_pool_bits;
+  {
+    ScopedGuard scoped_lock(AddressPoolManagerBitmap::GetLock());
+    regular_pool_bits = AddressPoolManagerBitmap::regular_pool_bits_;
+    brp_pool_bits = AddressPoolManagerBitmap::brp_pool_bits_;
+  }  // scoped_lock
+
+  // Pool usage is read out from the address pool bitmaps.
+  // The output stats are sized in super pages, so we interpret
+  // the bitmaps into super page usage.
+  static_assert(
+      kSuperPageSize %
+              AddressPoolManagerBitmap::kBytesPer1BitOfRegularPoolBitmap ==
+          0,
+      "information loss when calculating metrics");
+  constexpr size_t kRegularPoolBitsPerSuperPage =
+      kSuperPageSize /
+      AddressPoolManagerBitmap::kBytesPer1BitOfRegularPoolBitmap;
+
+  // Get 32-bit pool usage.
+  stats->regular_pool_stats.usage =
+      CountUsedSuperPages(regular_pool_bits, kRegularPoolBitsPerSuperPage);
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  static_assert(
+      kSuperPageSize % AddressPoolManagerBitmap::kBytesPer1BitOfBRPPoolBitmap ==
+          0,
+      "information loss when calculating metrics");
+  constexpr size_t kBRPPoolBitsPerSuperPage =
+      kSuperPageSize / AddressPoolManagerBitmap::kBytesPer1BitOfBRPPoolBitmap;
+  stats->brp_pool_stats.usage =
+      CountUsedSuperPages(brp_pool_bits, kBRPPoolBitsPerSuperPage);
+
+  // Get blocklist size.
+  for (const auto& blocked :
+       AddressPoolManagerBitmap::brp_forbidden_super_page_map_) {
+    if (blocked.load(std::memory_order_relaxed)) {
+      stats->blocklist_size += 1;
+    }
+  }
+
+  // Count failures in finding non-blocklisted addresses.
+  stats->blocklist_hit_count =
+      AddressPoolManagerBitmap::blocklist_hit_count_.load(
+          std::memory_order_relaxed);
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  return true;
+}
+
+#endif  // BUILDFLAG(HAS_64_BIT_POINTERS)
+
+void AddressPoolManager::DumpStats(AddressSpaceStatsDumper* dumper) {
+  AddressSpaceStats stats{};
+  if (GetStats(&stats)) {
+    dumper->DumpStats(&stats);
+  }
+}
+
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+// This function just exists to static_assert the layout of the private fields
+// in Pool.
+void AddressPoolManager::AssertThreadIsolatedLayout() {
+  constexpr size_t last_pool_offset =
+      offsetof(AddressPoolManager, pools_) + sizeof(Pool) * (kNumPools - 1);
+  constexpr size_t alloc_bitset_offset =
+      last_pool_offset + offsetof(Pool, alloc_bitset_);
+  static_assert(alloc_bitset_offset % PA_THREAD_ISOLATED_ALIGN_SZ == 0);
+  static_assert(sizeof(AddressPoolManager) % PA_THREAD_ISOLATED_ALIGN_SZ == 0);
+}
+#endif  // BUILDFLAG(ENABLE_THREAD_ISOLATION)
+
+}  // namespace partition_alloc::internal

src/base/allocator/partition_allocator/address_pool_manager.h

@@ -0,0 +1,202 @@
+// Copyright 2020 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_H_
+
+#include <bitset>
+#include <limits>
+
+#include "base/allocator/partition_allocator/address_pool_manager_types.h"
+#include "base/allocator/partition_allocator/partition_address_space.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/debug/debugging_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/thread_annotations.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+#include "base/allocator/partition_allocator/partition_lock.h"
+#include "base/allocator/partition_allocator/thread_isolation/alignment.h"
+#include "base/allocator/partition_allocator/thread_isolation/thread_isolation.h"
+#include "build/build_config.h"
+
+#if !BUILDFLAG(HAS_64_BIT_POINTERS)
+#include "base/allocator/partition_allocator/address_pool_manager_bitmap.h"
+#endif
+
+namespace partition_alloc {
+
+class AddressSpaceStatsDumper;
+struct AddressSpaceStats;
+struct PoolStats;
+
+}  // namespace partition_alloc
+
+namespace partition_alloc::internal {
+
+// (64bit version)
+// AddressPoolManager takes a reserved virtual address space and manages address
+// space allocation.
+//
+// AddressPoolManager (currently) supports up to 4 pools. Each pool manages a
+// contiguous reserved address space. Alloc() takes a pool_handle and returns
+// address regions from the specified pool. Free() also takes a pool_handle and
+// returns the address region back to the manager.
+//
+// (32bit version)
+// AddressPoolManager wraps AllocPages and FreePages and remembers allocated
+// address regions using bitmaps. IsManagedByPartitionAlloc*Pool use the bitmaps
+// to judge whether a given address is in a pool that supports BackupRefPtr or
+// in a pool that doesn't. All PartitionAlloc allocations must be in either of
+// the pools.
+class PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+    PA_THREAD_ISOLATED_ALIGN AddressPoolManager {
+ public:
+  static AddressPoolManager& GetInstance();
+
+  AddressPoolManager(const AddressPoolManager&) = delete;
+  AddressPoolManager& operator=(const AddressPoolManager&) = delete;
+
+#if BUILDFLAG(HAS_64_BIT_POINTERS)
+  void Add(pool_handle handle, uintptr_t address, size_t length);
+  void Remove(pool_handle handle);
+
+  // Populate a |used| bitset of superpages currently in use.
+  void GetPoolUsedSuperPages(pool_handle handle,
+                             std::bitset<kMaxSuperPagesInPool>& used);
+
+  // Return the base address of a pool.
+  uintptr_t GetPoolBaseAddress(pool_handle handle);
+#endif  // BUILDFLAG(HAS_64_BIT_POINTERS)
+
+  // Reserves address space from the pool.
+  uintptr_t Reserve(pool_handle handle,
+                    uintptr_t requested_address,
+                    size_t length);
+
+  // Frees address space back to the pool and decommits underlying system pages.
+  void UnreserveAndDecommit(pool_handle handle,
+                            uintptr_t address,
+                            size_t length);
+  void ResetForTesting();
+
+#if !BUILDFLAG(HAS_64_BIT_POINTERS)
+  void MarkUsed(pool_handle handle, uintptr_t address, size_t size);
+  void MarkUnused(pool_handle handle, uintptr_t address, size_t size);
+
+  static bool IsManagedByRegularPool(uintptr_t address) {
+    return AddressPoolManagerBitmap::IsManagedByRegularPool(address);
+  }
+
+  static bool IsManagedByBRPPool(uintptr_t address) {
+    return AddressPoolManagerBitmap::IsManagedByBRPPool(address);
+  }
+#endif  // !BUILDFLAG(HAS_64_BIT_POINTERS)
+
+  void DumpStats(AddressSpaceStatsDumper* dumper);
+
+ private:
+  friend class AddressPoolManagerForTesting;
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  // If we use a thread isolated pool, we need to write-protect its metadata.
+  // Allow the function to get access to the pool pointer.
+  friend void WriteProtectThreadIsolatedGlobals(ThreadIsolationOption);
+#endif
+
+  constexpr AddressPoolManager() = default;
+  ~AddressPoolManager() = default;
+
+  // Populates `stats` if applicable.
+  // Returns whether `stats` was populated. (They might not be, e.g.
+  // if PartitionAlloc is wholly unused in this process.)
+  bool GetStats(AddressSpaceStats* stats);
+
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  static void AssertThreadIsolatedLayout();
+#endif  // BUILDFLAG(ENABLE_THREAD_ISOLATION)
+
+#if BUILDFLAG(HAS_64_BIT_POINTERS)
+
+  class Pool {
+   public:
+    constexpr Pool() = default;
+    ~Pool() = default;
+
+    Pool(const Pool&) = delete;
+    Pool& operator=(const Pool&) = delete;
+
+    void Initialize(uintptr_t ptr, size_t length);
+    bool IsInitialized();
+    void Reset();
+
+    uintptr_t FindChunk(size_t size);
+    void FreeChunk(uintptr_t address, size_t size);
+
+    bool TryReserveChunk(uintptr_t address, size_t size);
+
+    void GetUsedSuperPages(std::bitset<kMaxSuperPagesInPool>& used);
+    uintptr_t GetBaseAddress();
+
+    void GetStats(PoolStats* stats);
+
+   private:
+    // The lock needs to be the first field in this class.
+    // We write-protect the pool in the ThreadIsolated case, except that the
+    // lock can be used without acquiring write-permission first (via
+    // DumpStats()). So instead of protecting the whole variable, we only
+    // protect the memory after the lock.
+    // See the alignment of ` below.
+    Lock lock_;
+
+    // The bitset stores the allocation state of the address pool. 1 bit per
+    // super-page: 1 = allocated, 0 = free.
+    std::bitset<kMaxSuperPagesInPool> alloc_bitset_ PA_GUARDED_BY(lock_);
+
+    // An index of a bit in the bitset before which we know for sure there all
+    // 1s. This is a best-effort hint in the sense that there still may be lots
+    // of 1s after this index, but at least we know there is no point in
+    // starting the search before it.
+    size_t bit_hint_ PA_GUARDED_BY(lock_) = 0;
+
+    size_t total_bits_ = 0;
+    uintptr_t address_begin_ = 0;
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+    uintptr_t address_end_ = 0;
+#endif
+
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+    friend class AddressPoolManager;
+    friend void WriteProtectThreadIsolatedGlobals(ThreadIsolationOption);
+#endif  // BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  };
+
+  PA_ALWAYS_INLINE Pool* GetPool(pool_handle handle) {
+    PA_DCHECK(kNullPoolHandle < handle && handle <= kNumPools);
+    return &pools_[handle - 1];
+  }
+
+  // Gets the stats for the pool identified by `handle`, if
+  // initialized.
+  void GetPoolStats(pool_handle handle, PoolStats* stats);
+
+  // If thread isolation support is enabled, we need to write-protect the
+  // isolated pool (which needs to be last). For this, we need to add padding in
+  // front of the pools so that the isolated one starts on a page boundary.
+  // We also skip the Lock at the beginning of the pool since it needs to be
+  // used in contexts where we didn't enable write access to the pool memory.
+  char pad_[PA_THREAD_ISOLATED_ARRAY_PAD_SZ_WITH_OFFSET(
+      Pool,
+      kNumPools,
+      offsetof(Pool, alloc_bitset_))] = {};
+  Pool pools_[kNumPools];
+
+#endif  // BUILDFLAG(HAS_64_BIT_POINTERS)
+
+  static PA_CONSTINIT AddressPoolManager singleton_;
+};
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_H_

src/base/allocator/partition_allocator/address_pool_manager_bitmap.cc

@@ -0,0 +1,37 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/address_pool_manager_bitmap.h"
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+
+#if !BUILDFLAG(HAS_64_BIT_POINTERS)
+
+namespace partition_alloc::internal {
+
+namespace {
+
+Lock g_lock;
+
+}  // namespace
+
+Lock& AddressPoolManagerBitmap::GetLock() {
+  return g_lock;
+}
+
+std::bitset<AddressPoolManagerBitmap::kRegularPoolBits>
+    AddressPoolManagerBitmap::regular_pool_bits_;  // GUARDED_BY(GetLock())
+std::bitset<AddressPoolManagerBitmap::kBRPPoolBits>
+    AddressPoolManagerBitmap::brp_pool_bits_;  // GUARDED_BY(GetLock())
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+std::array<std::atomic_bool,
+           AddressPoolManagerBitmap::kAddressSpaceSize / kSuperPageSize>
+    AddressPoolManagerBitmap::brp_forbidden_super_page_map_;
+std::atomic_size_t AddressPoolManagerBitmap::blocklist_hit_count_;
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+
+}  // namespace partition_alloc::internal
+
+#endif  // !BUILDFLAG(HAS_64_BIT_POINTERS)

src/base/allocator/partition_allocator/address_pool_manager_bitmap.h

@@ -0,0 +1,189 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_BITMAP_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_BITMAP_H_
+
+#include <array>
+#include <atomic>
+#include <bitset>
+#include <limits>
+
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+#include "base/allocator/partition_allocator/partition_lock.h"
+#include "build/build_config.h"
+
+#if !BUILDFLAG(HAS_64_BIT_POINTERS)
+
+namespace partition_alloc {
+
+namespace internal {
+
+// AddressPoolManagerBitmap is a set of bitmaps that track whether a given
+// address is in a pool that supports BackupRefPtr, or in a pool that doesn't
+// support it. All PartitionAlloc allocations must be in either of the pools.
+//
+// This code is specific to 32-bit systems.
+class PA_COMPONENT_EXPORT(PARTITION_ALLOC) AddressPoolManagerBitmap {
+ public:
+  static constexpr uint64_t kGiB = 1024 * 1024 * 1024ull;
+  static constexpr uint64_t kAddressSpaceSize = 4ull * kGiB;
+
+  // For BRP pool, we use partition page granularity to eliminate the guard
+  // pages from the bitmap at the ends:
+  // - Eliminating the guard page at the beginning is needed so that pointers
+  //   to the end of an allocation that immediately precede a super page in BRP
+  //   pool don't accidentally fall into that pool.
+  // - Eliminating the guard page at the end is to ensure that the last page
+  //   of the address space isn't in the BRP pool. This allows using sentinels
+  //   like reinterpret_cast<void*>(-1) without a risk of triggering BRP logic
+  //   on an invalid address. (Note, 64-bit systems don't have this problem as
+  //   the upper half of the address space always belongs to the OS.)
+  //
+  // Note, direct map allocations also belong to this pool. The same logic as
+  // above applies. It is important to note, however, that the granularity used
+  // here has to be a minimum of partition page size and direct map allocation
+  // granularity. Since DirectMapAllocationGranularity() is no smaller than
+  // PageAllocationGranularity(), we don't need to decrease the bitmap
+  // granularity any further.
+  static constexpr size_t kBitShiftOfBRPPoolBitmap = PartitionPageShift();
+  static constexpr size_t kBytesPer1BitOfBRPPoolBitmap = PartitionPageSize();
+  static_assert(kBytesPer1BitOfBRPPoolBitmap == 1 << kBitShiftOfBRPPoolBitmap,
+                "");
+  static constexpr size_t kGuardOffsetOfBRPPoolBitmap = 1;
+  static constexpr size_t kGuardBitsOfBRPPoolBitmap = 2;
+  static constexpr size_t kBRPPoolBits =
+      kAddressSpaceSize / kBytesPer1BitOfBRPPoolBitmap;
+
+  // Regular pool may include both normal bucket and direct map allocations, so
+  // the bitmap granularity has to be at least as small as
+  // DirectMapAllocationGranularity(). No need to eliminate guard pages at the
+  // ends, as this is a BackupRefPtr-specific concern, hence no need to lower
+  // the granularity to partition page size.
+  static constexpr size_t kBitShiftOfRegularPoolBitmap =
+      DirectMapAllocationGranularityShift();
+  static constexpr size_t kBytesPer1BitOfRegularPoolBitmap =
+      DirectMapAllocationGranularity();
+  static_assert(kBytesPer1BitOfRegularPoolBitmap ==
+                    1 << kBitShiftOfRegularPoolBitmap,
+                "");
+  static constexpr size_t kRegularPoolBits =
+      kAddressSpaceSize / kBytesPer1BitOfRegularPoolBitmap;
+
+  // Returns false for nullptr.
+  static bool IsManagedByRegularPool(uintptr_t address) {
+    static_assert(
+        std::numeric_limits<uintptr_t>::max() >> kBitShiftOfRegularPoolBitmap <
+            regular_pool_bits_.size(),
+        "The bitmap is too small, will result in unchecked out of bounds "
+        "accesses.");
+    // It is safe to read |regular_pool_bits_| without a lock since the caller
+    // is responsible for guaranteeing that the address is inside a valid
+    // allocation and the deallocation call won't race with this call.
+    return PA_TS_UNCHECKED_READ(
+        regular_pool_bits_)[address >> kBitShiftOfRegularPoolBitmap];
+  }
+
+  // Returns false for nullptr.
+  static bool IsManagedByBRPPool(uintptr_t address) {
+    static_assert(std::numeric_limits<uintptr_t>::max() >>
+                      kBitShiftOfBRPPoolBitmap < brp_pool_bits_.size(),
+                  "The bitmap is too small, will result in unchecked out of "
+                  "bounds accesses.");
+    // It is safe to read |brp_pool_bits_| without a lock since the caller
+    // is responsible for guaranteeing that the address is inside a valid
+    // allocation and the deallocation call won't race with this call.
+    return PA_TS_UNCHECKED_READ(
+        brp_pool_bits_)[address >> kBitShiftOfBRPPoolBitmap];
+  }
+
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  static void BanSuperPageFromBRPPool(uintptr_t address) {
+    brp_forbidden_super_page_map_[address >> kSuperPageShift].store(
+        true, std::memory_order_relaxed);
+  }
+
+  static bool IsAllowedSuperPageForBRPPool(uintptr_t address) {
+    // The only potentially dangerous scenario, in which this check is used, is
+    // when the assignment of the first raw_ptr<T> object for an address
+    // allocated outside the BRP pool is racing with the allocation of a new
+    // super page at the same address. We assume that if raw_ptr<T> is being
+    // initialized with a raw pointer, the associated allocation is "alive";
+    // otherwise, the issue should be fixed by rewriting the raw pointer
+    // variable as raw_ptr<T>. In the worst case, when such a fix is
+    // impossible, we should just undo the raw pointer -> raw_ptr<T> rewrite of
+    // the problematic field. If the above assumption holds, the existing
+    // allocation will prevent us from reserving the super-page region and,
+    // thus, having the race condition.  Since we rely on that external
+    // synchronization, the relaxed memory ordering should be sufficient.
+    return !brp_forbidden_super_page_map_[address >> kSuperPageShift].load(
+        std::memory_order_relaxed);
+  }
+
+  static void IncrementBlocklistHitCount() { ++blocklist_hit_count_; }
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+
+ private:
+  friend class AddressPoolManager;
+
+  static Lock& GetLock();
+
+  static std::bitset<kRegularPoolBits> regular_pool_bits_
+      PA_GUARDED_BY(GetLock());
+  static std::bitset<kBRPPoolBits> brp_pool_bits_ PA_GUARDED_BY(GetLock());
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  static std::array<std::atomic_bool, kAddressSpaceSize / kSuperPageSize>
+      brp_forbidden_super_page_map_;
+  static std::atomic_size_t blocklist_hit_count_;
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+};
+
+}  // namespace internal
+
+// Returns false for nullptr.
+PA_ALWAYS_INLINE bool IsManagedByPartitionAlloc(uintptr_t address) {
+  // When ENABLE_BACKUP_REF_PTR_SUPPORT is off, BRP pool isn't used.
+  // No need to add IsManagedByConfigurablePool, because Configurable Pool
+  // doesn't exist on 32-bit.
+#if !BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  PA_DCHECK(!internal::AddressPoolManagerBitmap::IsManagedByBRPPool(address));
+#endif
+  return internal::AddressPoolManagerBitmap::IsManagedByRegularPool(address)
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+         || internal::AddressPoolManagerBitmap::IsManagedByBRPPool(address)
+#endif
+      ;
+}
+
+// Returns false for nullptr.
+PA_ALWAYS_INLINE bool IsManagedByPartitionAllocRegularPool(uintptr_t address) {
+  return internal::AddressPoolManagerBitmap::IsManagedByRegularPool(address);
+}
+
+// Returns false for nullptr.
+PA_ALWAYS_INLINE bool IsManagedByPartitionAllocBRPPool(uintptr_t address) {
+  return internal::AddressPoolManagerBitmap::IsManagedByBRPPool(address);
+}
+
+// Returns false for nullptr.
+PA_ALWAYS_INLINE bool IsManagedByPartitionAllocConfigurablePool(
+    uintptr_t address) {
+  // The Configurable Pool is only available on 64-bit builds.
+  return false;
+}
+
+PA_ALWAYS_INLINE bool IsConfigurablePoolAvailable() {
+  // The Configurable Pool is only available on 64-bit builds.
+  return false;
+}
+
+}  // namespace partition_alloc
+
+#endif  // !BUILDFLAG(HAS_64_BIT_POINTERS)
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_BITMAP_H_

src/base/allocator/partition_allocator/address_pool_manager_types.h

@@ -0,0 +1,14 @@
+// Copyright 2020 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_TYPES_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_TYPES_H_
+
+namespace partition_alloc::internal {
+
+enum pool_handle : unsigned;
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_POOL_MANAGER_TYPES_H_

src/base/allocator/partition_allocator/address_space_randomization.cc

@@ -0,0 +1,51 @@
+// Copyright 2014 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/address_space_randomization.h"
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/random.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_WIN)
+#include <windows.h>
+#endif
+
+namespace partition_alloc {
+
+uintptr_t GetRandomPageBase() {
+  uintptr_t random = static_cast<uintptr_t>(internal::RandomValue());
+
+#if BUILDFLAG(HAS_64_BIT_POINTERS)
+  random <<= 32ULL;
+  random |= static_cast<uintptr_t>(internal::RandomValue());
+
+  // The ASLRMask() and ASLROffset() constants will be suitable for the
+  // OS and build configuration.
+  random &= internal::ASLRMask();
+  random += internal::ASLROffset();
+#else  // BUILDFLAG(HAS_64_BIT_POINTERS)
+#if BUILDFLAG(IS_WIN)
+  // On win32 host systems the randomization plus huge alignment causes
+  // excessive fragmentation. Plus most of these systems lack ASLR, so the
+  // randomization isn't buying anything. In that case we just skip it.
+  // TODO(palmer): Just dump the randomization when HE-ASLR is present.
+  static BOOL is_wow64 = -1;
+  if (is_wow64 == -1 && !IsWow64Process(GetCurrentProcess(), &is_wow64)) {
+    is_wow64 = FALSE;
+  }
+  if (!is_wow64) {
+    return 0;
+  }
+#endif  // BUILDFLAG(IS_WIN)
+  random &= internal::ASLRMask();
+  random += internal::ASLROffset();
+#endif  // BUILDFLAG(HAS_64_BIT_POINTERS)
+
+  PA_DCHECK(!(random & internal::PageAllocationGranularityOffsetMask()));
+  return random;
+}
+
+}  // namespace partition_alloc

src/base/allocator/partition_allocator/address_space_randomization.h

@@ -0,0 +1,286 @@
+// Copyright 2014 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_SPACE_RANDOMIZATION_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_SPACE_RANDOMIZATION_H_
+
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/page_allocator_constants.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "build/build_config.h"
+
+namespace partition_alloc {
+
+// Calculates a random preferred mapping address. In calculating an address, we
+// balance good ASLR against not fragmenting the address space too badly.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) uintptr_t GetRandomPageBase();
+
+namespace internal {
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+AslrAddress(uintptr_t mask) {
+  return mask & PageAllocationGranularityBaseMask();
+}
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+AslrMask(uintptr_t bits) {
+  return AslrAddress((1ULL << bits) - 1ULL);
+}
+
+// Turn off formatting, because the thicket of nested ifdefs below is
+// incomprehensible without indentation. It is also incomprehensible with
+// indentation, but the only other option is a combinatorial explosion of
+// *_{win,linux,mac,foo}_{32,64}.h files.
+//
+// clang-format off
+
+#if defined(ARCH_CPU_64_BITS)
+
+  #if defined(MEMORY_TOOL_REPLACES_ALLOCATOR)
+
+    // We shouldn't allocate system pages at all for sanitizer builds. However,
+    // we do, and if random hint addresses interfere with address ranges
+    // hard-coded in those tools, bad things happen. This address range is
+    // copied from TSAN source but works with all tools. See
+    // https://crbug.com/539863.
+    PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+    ASLRMask() {
+      return AslrAddress(0x007fffffffffULL);
+    }
+    PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+    ASLROffset() {
+      return AslrAddress(0x7e8000000000ULL);
+    }
+
+  #elif BUILDFLAG(IS_WIN)
+
+    // Windows 8.10 and newer support the full 48 bit address range. Since
+    // ASLROffset() is non-zero and may cause a carry, use 47 bit masks. See
+    // http://www.alex-ionescu.com/?p=246
+    PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+      return AslrMask(47);
+    }
+    // Try not to map pages into the range where Windows loads DLLs by default.
+    PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+      return 0x80000000ULL;
+    }
+
+  #elif BUILDFLAG(IS_APPLE)
+
+    // macOS as of 10.12.5 does not clean up entries in page map levels 3/4
+    // [PDP/PML4] created from mmap or mach_vm_allocate, even after the region
+    // is destroyed. Using a virtual address space that is too large causes a
+    // leak of about 1 wired [can never be paged out] page per call to mmap. The
+    // page is only reclaimed when the process is killed. Confine the hint to a
+    // 39-bit section of the virtual address space.
+    //
+    // This implementation adapted from
+    // https://chromium-review.googlesource.com/c/v8/v8/+/557958. The difference
+    // is that here we clamp to 39 bits, not 32.
+    //
+    // TODO(crbug.com/738925): Remove this limitation if/when the macOS behavior
+    // changes.
+    PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+    ASLRMask() {
+      return AslrMask(38);
+    }
+    PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+    ASLROffset() {
+      // Be careful, there is a zone where macOS will not map memory, at least
+      // on ARM64. From an ARM64 machine running 12.3, the range seems to be
+      // [0x1000000000, 0x7000000000). Make sure that the range we use is
+      // outside these bounds. In 12.3, there is a reserved area between
+      // MACH_VM_MIN_GPU_CARVEOUT_ADDRESS and MACH_VM_MAX_GPU_CARVEOUT_ADDRESS,
+      // which is reserved on ARM64. See these constants in XNU's source code
+      // for details (xnu-8019.80.24/osfmk/mach/arm/vm_param.h).
+      return AslrAddress(0x10000000000ULL);
+    }
+
+  #elif BUILDFLAG(IS_POSIX) || BUILDFLAG(IS_FUCHSIA)
+
+    #if defined(ARCH_CPU_X86_64)
+
+      // Linux (and macOS) support the full 47-bit user space of x64 processors.
+      // Use only 46 to allow the kernel a chance to fulfill the request.
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+        return AslrMask(46);
+      }
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+        return AslrAddress(0);
+      }
+
+    #elif defined(ARCH_CPU_ARM64)
+
+      #if BUILDFLAG(IS_ANDROID)
+
+      // Restrict the address range on Android to avoid a large performance
+      // regression in single-process WebViews. See https://crbug.com/837640.
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+        return AslrMask(30);
+      }
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+        return AslrAddress(0x20000000ULL);
+      }
+
+      #elif BUILDFLAG(IS_LINUX)
+
+      // Linux on arm64 can use 39, 42, 48, or 52-bit user space, depending on
+      // page size and number of levels of translation pages used. We use
+      // 39-bit as base as all setups should support this, lowered to 38-bit
+      // as ASLROffset() could cause a carry.
+      PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+      ASLRMask() {
+        return AslrMask(38);
+      }
+      PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+      ASLROffset() {
+        return AslrAddress(0x1000000000ULL);
+      }
+
+      #else
+
+      // ARM64 on Linux has 39-bit user space. Use 38 bits since ASLROffset()
+      // could cause a carry.
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+        return AslrMask(38);
+      }
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+        return AslrAddress(0x1000000000ULL);
+      }
+
+      #endif
+
+    #elif defined(ARCH_CPU_PPC64)
+
+      #if BUILDFLAG(IS_AIX)
+
+        // AIX has 64 bits of virtual addressing, but we limit the address range
+        // to (a) minimize segment lookaside buffer (SLB) misses; and (b) use
+        // extra address space to isolate the mmap regions.
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+          return AslrMask(30);
+        }
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+          return AslrAddress(0x400000000000ULL);
+        }
+
+      #elif defined(ARCH_CPU_BIG_ENDIAN)
+
+        // Big-endian Linux PPC has 44 bits of virtual addressing. Use 42.
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+          return AslrMask(42);
+        }
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+          return AslrAddress(0);
+        }
+
+      #else  // !BUILDFLAG(IS_AIX) && !defined(ARCH_CPU_BIG_ENDIAN)
+
+        // Little-endian Linux PPC has 48 bits of virtual addressing. Use 46.
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+          return AslrMask(46);
+        }
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+          return AslrAddress(0);
+        }
+
+      #endif  // !BUILDFLAG(IS_AIX) && !defined(ARCH_CPU_BIG_ENDIAN)
+
+    #elif defined(ARCH_CPU_S390X)
+
+      // Linux on Z uses bits 22 - 32 for Region Indexing, which translates to
+      // 42 bits of virtual addressing. Truncate to 40 bits to allow kernel a
+      // chance to fulfill the request.
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+        return AslrMask(40);
+      }
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+        return AslrAddress(0);
+      }
+
+    #elif defined(ARCH_CPU_S390)
+
+      // 31 bits of virtual addressing. Truncate to 29 bits to allow the kernel
+      // a chance to fulfill the request.
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+        return AslrMask(29);
+      }
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+        return AslrAddress(0);
+      }
+
+    #else  // !defined(ARCH_CPU_X86_64) && !defined(ARCH_CPU_PPC64) &&
+           // !defined(ARCH_CPU_S390X) && !defined(ARCH_CPU_S390)
+
+      // For all other POSIX variants, use 30 bits.
+      PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+        return AslrMask(30);
+      }
+
+      #if BUILDFLAG(IS_SOLARIS)
+
+        // For our Solaris/illumos mmap hint, we pick a random address in the
+        // bottom half of the top half of the address space (that is, the third
+        // quarter). Because we do not MAP_FIXED, this will be treated only as a
+        // hint -- the system will not fail to mmap because something else
+        // happens to already be mapped at our random address. We deliberately
+        // set the hint high enough to get well above the system's break (that
+        // is, the heap); Solaris and illumos will try the hint and if that
+        // fails allocate as if there were no hint at all. The high hint
+        // prevents the break from getting hemmed in at low values, ceding half
+        // of the address space to the system heap.
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+          return AslrAddress(0x80000000ULL);
+        }
+
+      #elif BUILDFLAG(IS_AIX)
+
+        // The range 0x30000000 - 0xD0000000 is available on AIX; choose the
+        // upper range.
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+          return AslrAddress(0x90000000ULL);
+        }
+
+      #else  // !BUILDFLAG(IS_SOLARIS) && !BUILDFLAG(IS_AIX)
+
+        // The range 0x20000000 - 0x60000000 is relatively unpopulated across a
+        // variety of ASLR modes (PAE kernel, NX compat mode, etc) and on macOS
+        // 10.6 and 10.7.
+        PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+          return AslrAddress(0x20000000ULL);
+        }
+
+      #endif  // !BUILDFLAG(IS_SOLARIS) && !BUILDFLAG(IS_AIX)
+
+    #endif  // !defined(ARCH_CPU_X86_64) && !defined(ARCH_CPU_PPC64) &&
+            // !defined(ARCH_CPU_S390X) && !defined(ARCH_CPU_S390)
+
+  #endif  // BUILDFLAG(IS_POSIX)
+
+#elif defined(ARCH_CPU_32_BITS)
+
+  // This is a good range on 32-bit Windows and Android (the only platforms on
+  // which we support 32-bitness). Allocates in the 0.5 - 1.5 GiB region. There
+  // is no issue with carries here.
+  PA_ALWAYS_INLINE constexpr uintptr_t ASLRMask() {
+    return AslrMask(30);
+  }
+  PA_ALWAYS_INLINE constexpr uintptr_t ASLROffset() {
+    return AslrAddress(0x20000000ULL);
+  }
+
+#else
+
+  #error Please tell us about your exotic hardware! Sounds interesting.
+
+#endif  // defined(ARCH_CPU_32_BITS)
+
+// clang-format on
+
+}  // namespace internal
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_SPACE_RANDOMIZATION_H_

src/base/allocator/partition_allocator/address_space_stats.h

@@ -0,0 +1,55 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_SPACE_STATS_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_SPACE_STATS_H_
+
+#include <cstddef>
+
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+
+namespace partition_alloc {
+
+// All members are measured in super pages.
+struct PoolStats {
+  size_t usage = 0;
+
+  // On 32-bit, pools are mainly logical entities, intermingled with
+  // allocations not managed by PartitionAlloc. The "largest available
+  // reservation" is not possible to measure in that case.
+#if BUILDFLAG(HAS_64_BIT_POINTERS)
+  size_t largest_available_reservation = 0;
+#endif
+};
+
+struct AddressSpaceStats {
+  PoolStats regular_pool_stats;
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  PoolStats brp_pool_stats;
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+#if BUILDFLAG(HAS_64_BIT_POINTERS)
+  PoolStats configurable_pool_stats;
+#else
+#if BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+  size_t blocklist_size;  // measured in super pages
+  size_t blocklist_hit_count;
+#endif  // BUILDFLAG(ENABLE_BACKUP_REF_PTR_SUPPORT)
+#endif  // BUILDFLAG(HAS_64_BIT_POINTERS)
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  PoolStats thread_isolated_pool_stats;
+#endif
+};
+
+// Interface passed to `AddressPoolManager::DumpStats()` to mediate
+// for `AddressSpaceDumpProvider`.
+class PA_COMPONENT_EXPORT(PARTITION_ALLOC) AddressSpaceStatsDumper {
+ public:
+  virtual void DumpStats(const AddressSpaceStats* address_space_stats) = 0;
+  virtual ~AddressSpaceStatsDumper() = default;
+};
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ADDRESS_SPACE_STATS_H_

src/base/allocator/partition_allocator/allocation_guard.cc

@@ -0,0 +1,42 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/allocation_guard.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/immediate_crash.h"
+#include "base/allocator/partition_allocator/partition_alloc_config.h"
+
+#if PA_CONFIG(HAS_ALLOCATION_GUARD)
+
+namespace partition_alloc {
+
+namespace {
+thread_local bool g_disallow_allocations;
+}  // namespace
+
+ScopedDisallowAllocations::ScopedDisallowAllocations() {
+  if (g_disallow_allocations) {
+    PA_IMMEDIATE_CRASH();
+  }
+
+  g_disallow_allocations = true;
+}
+
+ScopedDisallowAllocations::~ScopedDisallowAllocations() {
+  g_disallow_allocations = false;
+}
+
+ScopedAllowAllocations::ScopedAllowAllocations() {
+  // Save the previous value, as ScopedAllowAllocations is used in all
+  // partitions, not just the malloc() ones(s).
+  saved_value_ = g_disallow_allocations;
+  g_disallow_allocations = false;
+}
+
+ScopedAllowAllocations::~ScopedAllowAllocations() {
+  g_disallow_allocations = saved_value_;
+}
+
+}  // namespace partition_alloc
+
+#endif  // PA_CONFIG(HAS_ALLOCATION_GUARD)

src/base/allocator/partition_allocator/allocation_guard.h

@@ -0,0 +1,49 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ALLOCATION_GUARD_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ALLOCATION_GUARD_H_
+
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_config.h"
+#include "build/build_config.h"
+
+namespace partition_alloc {
+
+#if PA_CONFIG(HAS_ALLOCATION_GUARD)
+
+// Disallow allocations in the scope. Does not nest.
+class PA_COMPONENT_EXPORT(PARTITION_ALLOC) ScopedDisallowAllocations {
+ public:
+  ScopedDisallowAllocations();
+  ~ScopedDisallowAllocations();
+};
+
+// Disallow allocations in the scope. Does not nest.
+class PA_COMPONENT_EXPORT(PARTITION_ALLOC) ScopedAllowAllocations {
+ public:
+  ScopedAllowAllocations();
+  ~ScopedAllowAllocations();
+
+ private:
+  bool saved_value_;
+};
+
+#else
+
+struct [[maybe_unused]] ScopedDisallowAllocations {};
+struct [[maybe_unused]] ScopedAllowAllocations {};
+
+#endif  // PA_CONFIG(HAS_ALLOCATION_GUARD)
+
+}  // namespace partition_alloc
+
+namespace base::internal {
+
+using ::partition_alloc::ScopedAllowAllocations;
+using ::partition_alloc::ScopedDisallowAllocations;
+
+}  // namespace base::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ALLOCATION_GUARD_H_

src/base/allocator/partition_allocator/arm_bti_test_functions.S

@@ -0,0 +1,50 @@
+# Copyright 2021 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+# This file contains a test function for checking Arm's branch target
+# identification (BTI) feature, which helps mitigate jump-oriented
+# programming. To get it working, BTI instructions must be executed
+# on a compatible core, and the executable pages must be mapped with
+# PROT_BTI. To validate that pages mapped with PROT_BTI are working
+# correctly:
+# 1) Allocate a read-write page.
+# 2) Copy between the start and end symbols into that page.
+# 3) Set the page to read-execute with PROT_BTI.
+# 4) Call the first offset of the page, verify the result.
+# 5) Call the second offset of the page (skipping the landing pad).
+#    Verify that it crashes as expected.
+# This test works irrespective of whether BTI is enabled for C/C++
+# objects via -mbranch-protection=standard.
+
+.text
+.global arm_bti_test_function
+.global arm_bti_test_function_invalid_offset
+.global arm_bti_test_function_end
+arm_bti_test_function:
+    # Mark the start of this function as a valid call target.
+    bti jc
+    add x0, x0, #1
+arm_bti_test_function_invalid_offset:
+    # This label simulates calling an incomplete function.
+    # Jumping here should crash systems which support BTI.
+    add x0, x0, #2
+    ret
+arm_bti_test_function_end:
+    nop
+
+// For details see section "6.2 Program Property" in
+// "ELF for the Arm 64-bit Architecture (AArch64)"
+// https://github.com/ARM-software/abi-aa/blob/main/aaelf64/aaelf64.rst#62program-property
+.pushsection .note.gnu.property, "a";
+.balign 8;
+.long 4;
+.long 0x10;
+.long 0x5;
+.asciz "GNU";
+.long 0xc0000000; /* GNU_PROPERTY_AARCH64_FEATURE_1_AND */
+.long 4;
+.long 1;           /* GNU_PROPERTY_AARCH64_BTI */;
+.long 0;
+.popsection
+

src/base/allocator/partition_allocator/arm_bti_test_functions.h

@@ -0,0 +1,31 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ARM_BTI_TEST_FUNCTIONS_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ARM_BTI_TEST_FUNCTIONS_H_
+
+#include "build/build_config.h"
+
+#if defined(ARCH_CPU_ARM64)
+extern "C" {
+/**
+ * A valid BTI function. Jumping to this funtion should not cause any problem in
+ * a BTI enabled environment.
+ **/
+int64_t arm_bti_test_function(int64_t);
+
+/**
+ * A function without proper BTI landing pad. Jumping here should crash the
+ * program on systems which support BTI.
+ **/
+int64_t arm_bti_test_function_invalid_offset(int64_t);
+
+/**
+ * A simple function which immediately returns to sender.
+ **/
+void arm_bti_test_function_end(void);
+}
+#endif  // defined(ARCH_CPU_ARM64)
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ARM_BTI_TEST_FUNCTIONS_H_

src/base/allocator/partition_allocator/build_config.md

@@ -0,0 +1,103 @@
+# Build Config
+
+PartitionAlloc's behavior and operation can be influenced by many
+different settings. Broadly, these are controlled at the top-level by
+[GN args][gn-declare-args], which propagate via
+[buildflags][buildflag-header] and `#defined` clauses.
+
+*** promo
+Most of what you'll want to know exists between
+
+* [`//base/allocator/partition_allocator/BUILD.gn`][pa-build-gn],
+* Everything else ending in `.gn` or `.gni` in
+  `//base/allocator/partition_allocator/`,
+* [`allocator.gni`][allocator-gni],
+* [`//base/allocator/BUILD.gn`][base-allocator-build-gn], and
+* [`//base/BUILD.gn`][base-build-gn].
+***
+
+*** aside
+While Chromium promotes the `#if BUILDFLAG(FOO)` construct, some of
+PartitionAlloc's behavior is governed by compound conditions `#defined`
+in [`partition_alloc_config.h`][partition-alloc-config].
+***
+
+*** promo
+PartitionAlloc targets C++17. As the team develops standalone
+PartitionAlloc, this may diverge from what the rest of Chrome browser
+does, as we will be obligated to support external clients that
+may not yet support newer C++ standards.
+
+See [Chrome-External Builds](./external_builds.md) for more.
+***
+
+## Select GN Args
+
+### `use_partition_alloc`
+
+Defines whether PartitionAlloc is at all available.
+
+Setting this `false` will entirely remove PartitionAlloc from the
+Chromium build. _You probably do not want this._
+
+*** note
+Back when PartitionAlloc was the dedicated allocator in Blink, disabling
+it was logically identical to wholly disabling it in Chromium. This GN
+arg organically grew in scope with the advent of
+PartitionAlloc-Everywhere and must be `true` as a prerequisite for
+enabling PA-E.
+***
+
+### `use_partition_alloc_as_malloc`
+
+Does nothing special when value is `false`. Enables
+[PartitionAlloc-Everywhere (PA-E)][pae-public-doc] when value is `true`.
+
+*** note
+* While "everywhere" (in "PartitionAlloc-Everywhere") tautologically
+  includes Blink where PartitionAlloc originated, setting
+  `use_partition_alloc_as_malloc = false` does not disable PA usage in Blink,
+  which invokes PA explicitly (not via malloc).
+* `use_partition_alloc_as_malloc = true` must not be confused
+  with `use_partition_alloc` (see above).
+***
+
+## Notable Macros
+
+There is an ongoing effort
+[to break out PartitionAlloc into a standalone library][pa-ee-crbug].
+Once PartitionAlloc stands alone from the larger Chrome build apparatus,
+the code loses access to some macros. This is not an immediate concern,
+but the team needs to decide either
+
+* how to propagate these macros in place, or
+* how to remove them, replacing them with PA-specific build config.
+
+A non-exhaustive list of work items:
+
+* `OFFICIAL_BUILD` - influences crash macros and
+  `PA_THREAD_CACHE_ALLOC_STATS`. These are conceptually distinct enough
+  to be worth separating into dedicated build controls.
+* `IS_PARTITION_ALLOC_IMPL` - must be defined when PartitionAlloc is
+  built as a shared library. This is required to export symbols.
+* `COMPONENT_BUILD` - component builds (as per
+  `//docs/component_build.md`) must `#define COMPONENT_BUILD`.
+  Additionally, to build Win32, invoker must `#define WIN32`.
+* `MEMORY_TOOL_REPLACES_ALLOCATOR`
+* `*_SANITIZER` - mainly influences unit tests.
+
+*** note
+Over time, the above list should evolve into a list of macros / GN args
+that influence PartitionAlloc's behavior.
+***
+
+[gn-declare-args]: https://gn.googlesource.com/gn/+/refs/heads/main/docs/reference.md#func_declare_args
+[buildflag-header]: https://source.chromium.org/chromium/chromium/src/+/main:build/buildflag_header.gni
+[pa-build-gn]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/partition_allocator/BUILD.gn
+[allocator-gni]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/allocator.gni
+[base-allocator-build-gn]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/BUILD.gn
+[base-build-gn]: https://source.chromium.org/chromium/chromium/src/+/main:base/BUILD.gn
+[partition-alloc-config]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/partition_allocator/partition_alloc_config.h
+[pae-public-doc]: https://docs.google.com/document/d/1R1H9z5IVUAnXJgDjnts3nTJVcRbufWWT9ByXLgecSUM/preview
+[miracleptr-doc]: https://docs.google.com/document/d/1pnnOAIz_DMWDI4oIOFoMAqLnf_MZ2GsrJNb_dbQ3ZBg/preview
+[pa-ee-crbug]: https://crbug.com/1151236

src/base/allocator/partition_allocator/build_overrides/build.gni

@@ -0,0 +1,9 @@
+# Copyright 2022 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+# This file will be used to check out PartitionAlloc and to build it as
+# standalone library. In this case, PartitionAlloc needs to define
+# build_with_chromium. If building PartitionAlloc as a part of chromium,
+# chromium will provide build_with_chromium=true.
+build_with_chromium = false

src/base/allocator/partition_allocator/build_overrides/partition_alloc.gni

@@ -0,0 +1,18 @@
+# Copyright 2022 The Chromium Authors
+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+
+import("//build_overrides/build.gni")
+
+# This is the default build configuration when building PartitionAlloc
+# as a standalone library.
+# If embedders want to use PartitionAlloc, they need to create their own
+# //build_overrides/partition_alloc.gni and define their own PartitionAlloc
+# configuration.
+
+use_partition_alloc_as_malloc_default = false
+use_allocator_shim_default = false
+enable_backup_ref_ptr_support_default = false
+put_ref_count_in_previous_slot_default = true
+enable_backup_ref_ptr_slow_checks_default = false
+enable_dangling_raw_ptr_checks_default = false

src/base/allocator/partition_allocator/compressed_pointer.cc

@@ -0,0 +1,29 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/compressed_pointer.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+
+#if BUILDFLAG(ENABLE_POINTER_COMPRESSION)
+
+namespace partition_alloc::internal {
+
+// We keep the useful part in |g_base_| as 1s to speed up decompression.
+alignas(kPartitionCachelineSize)
+    PA_COMPONENT_EXPORT(PARTITION_ALLOC) CompressedPointerBaseGlobal::Base
+    CompressedPointerBaseGlobal::g_base_ = {.base = kUsefulBitsMask};
+
+void CompressedPointerBaseGlobal::SetBase(uintptr_t base) {
+  PA_DCHECK(!IsSet());
+  PA_DCHECK((base & kUsefulBitsMask) == 0);
+  g_base_.base = base | kUsefulBitsMask;
+}
+
+void CompressedPointerBaseGlobal::ResetBaseForTesting() {
+  g_base_.base = kUsefulBitsMask;
+}
+
+}  // namespace partition_alloc::internal
+
+#endif  // BUILDFLAG(ENABLE_POINTER_COMPRESSION)

src/base/allocator/partition_allocator/compressed_pointer.h

@@ -0,0 +1,668 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_COMPRESSED_POINTER_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_COMPRESSED_POINTER_H_
+
+#include <climits>
+#include <type_traits>
+
+#include "base/allocator/partition_allocator/partition_address_space.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/bits.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+
+#if BUILDFLAG(ENABLE_POINTER_COMPRESSION)
+
+#if !BUILDFLAG(GLUE_CORE_POOLS)
+#error "Pointer compression only works with glued pools"
+#endif
+#if PA_CONFIG(DYNAMICALLY_SELECT_POOL_SIZE)
+#error "Pointer compression currently supports constant pool size"
+#endif
+
+#endif  // BUILDFLAG(ENABLE_POINTER_COMPRESSION)
+
+namespace partition_alloc {
+
+namespace internal {
+
+template <typename T1, typename T2>
+constexpr bool IsDecayedSame =
+    std::is_same_v<std::decay_t<T1>, std::decay_t<T2>>;
+
+#if BUILDFLAG(ENABLE_POINTER_COMPRESSION)
+
+// Pointer compression works by storing only the 'useful' 32-bit part of the
+// pointer. The other half (the base) is stored in a global variable
+// (CompressedPointerBaseGlobal::g_base_), which is used on decompression. To
+// support fast branchless decompression of nullptr, we use the most significant
+// bit in the compressed pointer to leverage sign-extension (for non-nullptr
+// pointers, the most significant bit is set, whereas for nullptr it's not).
+// Using this bit and supporting heaps larger than 4GB relies on having
+// alignment bits in pointers. Assuming that all pointers point to at least
+// 8-byte alignment objects, pointer compression can support heaps of size <=
+// 16GB.
+// ((3 alignment bits) = (1 bit for sign-extension) + (2 bits for 16GB heap)).
+//
+// Example: heap base: 0x4b0'ffffffff
+//  - g_base: 0x4b3'ffffffff (lower 34 bits set)
+//  - normal pointer: 0x4b2'a08b6480
+//    - compression:
+//      - shift right by 3:        0x96'54116c90
+//      - truncate:                   0x54116c90
+//      - mark MSB:                   0xd4116c90
+//    - decompression:
+//      - sign-extend:       0xffffffff'd4116c90
+//      - shift left by 3:   0xfffffffe'a08b6480
+//      - 'and' with g_base: 0x000004b2'a08b6480
+//
+//  - nullptr: 0x00000000'00000000
+//    - compression:
+//      - shift right by 3:  0x00000000'00000000
+//      - truncate:                   0x00000000
+//      - (don't mark MSB for nullptr)
+//    - decompression:
+//      - sign-extend:       0x00000000'00000000
+//      - shift left by 3:   0x00000000'00000000
+//      - 'and' with g_base: 0x00000000'00000000
+//
+// Pointer compression relies on having both the regular and the BRP pool (core
+// pools) 'glued', so that the same base could be used for both. For simplicity,
+// the configurations with dynamically selected pool size are not supported.
+// However, they can be at the cost of performing an extra load for
+// core-pools-shift-size on both compression and decompression.
+
+class CompressedPointerBaseGlobal final {
+ public:
+  static constexpr size_t kUsefulBits =
+      base::bits::CountTrailingZeroBits(PartitionAddressSpace::CorePoolsSize());
+  static_assert(kUsefulBits >= sizeof(uint32_t) * CHAR_BIT);
+  static constexpr size_t kBitsToShift =
+      kUsefulBits - sizeof(uint32_t) * CHAR_BIT;
+
+  CompressedPointerBaseGlobal() = delete;
+
+  // Attribute const allows the compiler to assume that
+  // CompressedPointerBaseGlobal::g_base_ doesn't change (e.g. across calls) and
+  // thereby avoid redundant loads.
+  PA_ALWAYS_INLINE __attribute__((const)) static uintptr_t Get() {
+    PA_DCHECK(IsBaseConsistent());
+    return g_base_.base;
+  }
+
+  PA_ALWAYS_INLINE static bool IsSet() {
+    PA_DCHECK(IsBaseConsistent());
+    return (g_base_.base & ~kUsefulBitsMask) != 0;
+  }
+
+ private:
+  static constexpr uintptr_t kUsefulBitsMask =
+      PartitionAddressSpace::CorePoolsSize() - 1;
+
+  static union alignas(kPartitionCachelineSize)
+      PA_COMPONENT_EXPORT(PARTITION_ALLOC) Base {
+    uintptr_t base;
+    char cache_line[kPartitionCachelineSize];
+  } g_base_ PA_CONSTINIT;
+
+  PA_ALWAYS_INLINE static bool IsBaseConsistent() {
+    return kUsefulBitsMask == (g_base_.base & kUsefulBitsMask);
+  }
+
+  static void SetBase(uintptr_t base);
+  static void ResetBaseForTesting();
+
+  friend class PartitionAddressSpace;
+};
+
+#endif  // BUILDFLAG(ENABLE_POINTER_COMPRESSION)
+
+}  // namespace internal
+
+#if BUILDFLAG(ENABLE_POINTER_COMPRESSION)
+
+template <typename T>
+class PA_TRIVIAL_ABI CompressedPointer final {
+ public:
+  using UnderlyingType = uint32_t;
+
+  PA_ALWAYS_INLINE constexpr CompressedPointer() = default;
+  PA_ALWAYS_INLINE explicit CompressedPointer(T* ptr) : value_(Compress(ptr)) {}
+  PA_ALWAYS_INLINE constexpr explicit CompressedPointer(std::nullptr_t)
+      : value_(0u) {}
+
+  PA_ALWAYS_INLINE constexpr CompressedPointer(const CompressedPointer&) =
+      default;
+  PA_ALWAYS_INLINE constexpr CompressedPointer(
+      CompressedPointer&& other) noexcept = default;
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr CompressedPointer(
+      const CompressedPointer<U>& other) {
+    if constexpr (internal::IsDecayedSame<T, U>) {
+      // When pointers have the same type modulo constness, avoid the
+      // compress-decompress round.
+      value_ = other.value_;
+    } else {
+      // When the types are different, perform the round, because the pointer
+      // may need to be adjusted.
+      // TODO(1376980): Avoid the cycle here.
+      value_ = Compress(other.get());
+    }
+  }
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr CompressedPointer(
+      CompressedPointer<U>&& other) noexcept
+      : CompressedPointer(other) {}
+
+  ~CompressedPointer() = default;
+
+  PA_ALWAYS_INLINE constexpr CompressedPointer& operator=(
+      const CompressedPointer&) = default;
+  PA_ALWAYS_INLINE constexpr CompressedPointer& operator=(
+      CompressedPointer&& other) noexcept = default;
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr CompressedPointer& operator=(
+      const CompressedPointer<U>& other) {
+    CompressedPointer copy(other);
+    value_ = copy.value_;
+    return *this;
+  }
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr CompressedPointer& operator=(
+      CompressedPointer<U>&& other) noexcept {
+    *this = other;
+    return *this;
+  }
+
+  // Don't perform compression when assigning to nullptr.
+  PA_ALWAYS_INLINE constexpr CompressedPointer& operator=(std::nullptr_t) {
+    value_ = 0u;
+    return *this;
+  }
+
+  PA_ALWAYS_INLINE T* get() const { return Decompress(value_); }
+
+  PA_ALWAYS_INLINE constexpr bool is_nonnull() const { return value_; }
+
+  PA_ALWAYS_INLINE constexpr UnderlyingType GetAsIntegral() const {
+    return value_;
+  }
+
+  PA_ALWAYS_INLINE constexpr explicit operator bool() const {
+    return is_nonnull();
+  }
+
+  template <typename U = T,
+            std::enable_if_t<!std::is_void_v<std::remove_cv_t<U>>>* = nullptr>
+  PA_ALWAYS_INLINE U& operator*() const {
+    PA_DCHECK(is_nonnull());
+    return *get();
+  }
+
+  PA_ALWAYS_INLINE T* operator->() const {
+    PA_DCHECK(is_nonnull());
+    return get();
+  }
+
+  PA_ALWAYS_INLINE constexpr void swap(CompressedPointer& other) {
+    std::swap(value_, other.value_);
+  }
+
+ private:
+  template <typename>
+  friend class CompressedPointer;
+
+  static constexpr size_t kBitsForSignExtension = 1;
+  static constexpr size_t kOverallBitsToShift =
+      internal::CompressedPointerBaseGlobal::kBitsToShift +
+      kBitsForSignExtension;
+
+  PA_ALWAYS_INLINE static UnderlyingType Compress(T* ptr) {
+    static constexpr size_t kMinimalRequiredAlignment = 8;
+    static_assert((1 << kOverallBitsToShift) == kMinimalRequiredAlignment);
+
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+    PA_DCHECK(reinterpret_cast<uintptr_t>(ptr) % kMinimalRequiredAlignment ==
+              0);
+    PA_DCHECK(internal::CompressedPointerBaseGlobal::IsSet());
+
+    const uintptr_t base = internal::CompressedPointerBaseGlobal::Get();
+    static constexpr size_t kCorePoolsBaseMask =
+        ~(internal::PartitionAddressSpace::CorePoolsSize() - 1);
+    PA_DCHECK(!ptr ||
+              (base & kCorePoolsBaseMask) ==
+                  (reinterpret_cast<uintptr_t>(ptr) & kCorePoolsBaseMask));
+#endif  // BUILDFLAG(PA_DCHECK_IS_ON)
+
+    const auto uptr = reinterpret_cast<uintptr_t>(ptr);
+    // Shift the pointer and truncate.
+    auto compressed = static_cast<UnderlyingType>(uptr >> kOverallBitsToShift);
+    // If the pointer is non-null, mark the most-significant-bit to sign-extend
+    // it on decompression. Assuming compression is a significantly less
+    // frequent operation, we let more work here in favor of faster
+    // decompression.
+    // TODO(1376980): Avoid this by overreserving the heap.
+    if (compressed) {
+      compressed |= (1u << (sizeof(uint32_t) * CHAR_BIT - 1));
+    }
+
+    return compressed;
+  }
+
+  PA_ALWAYS_INLINE static T* Decompress(UnderlyingType ptr) {
+    PA_DCHECK(internal::CompressedPointerBaseGlobal::IsSet());
+    const uintptr_t base = internal::CompressedPointerBaseGlobal::Get();
+    // Treat compressed pointer as signed and cast it to uint64_t, which will
+    // sign-extend it. Then, shift the result by one. It's important to shift
+    // the already unsigned value, as otherwise it would result in undefined
+    // behavior.
+    const uint64_t mask = static_cast<uint64_t>(static_cast<int32_t>(ptr))
+                          << (kOverallBitsToShift);
+    return reinterpret_cast<T*>(mask & base);
+  }
+
+  UnderlyingType value_;
+};
+
+template <typename T>
+PA_ALWAYS_INLINE constexpr void swap(CompressedPointer<T>& a,
+                                     CompressedPointer<T>& b) {
+  a.swap(b);
+}
+
+// operators==.
+template <typename T, typename U>
+PA_ALWAYS_INLINE bool operator==(CompressedPointer<T> a,
+                                 CompressedPointer<U> b) {
+  if constexpr (internal::IsDecayedSame<T, U>) {
+    // When pointers have the same type modulo constness, simply compare
+    // compressed values.
+    return a.GetAsIntegral() == b.GetAsIntegral();
+  } else {
+    // When the types are different, compare decompressed pointers, because the
+    // pointers may need to be adjusted.
+    // TODO(1376980): Avoid decompression here.
+    return a.get() == b.get();
+  }
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator==(CompressedPointer<T> a, U* b) {
+  // Do compression, since it is less expensive.
+  return a == static_cast<CompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator==(T* a, CompressedPointer<U> b) {
+  return b == a;
+}
+
+template <typename T>
+PA_ALWAYS_INLINE constexpr bool operator==(CompressedPointer<T> a,
+                                           std::nullptr_t) {
+  return !a.is_nonnull();
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator==(std::nullptr_t,
+                                           CompressedPointer<U> b) {
+  return b == nullptr;
+}
+
+// operators!=.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(CompressedPointer<T> a,
+                                           CompressedPointer<U> b) {
+  return !(a == b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(CompressedPointer<T> a, U* b) {
+  // Do compression, since it is less expensive.
+  return a != static_cast<CompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(T* a, CompressedPointer<U> b) {
+  return b != a;
+}
+
+template <typename T>
+PA_ALWAYS_INLINE constexpr bool operator!=(CompressedPointer<T> a,
+                                           std::nullptr_t) {
+  return a.is_nonnull();
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(std::nullptr_t,
+                                           CompressedPointer<U> b) {
+  return b != nullptr;
+}
+
+// operators<.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<(CompressedPointer<T> a,
+                                          CompressedPointer<U> b) {
+  if constexpr (internal::IsDecayedSame<T, U>) {
+    // When pointers have the same type modulo constness, simply compare
+    // compressed values.
+    return a.GetAsIntegral() < b.GetAsIntegral();
+  } else {
+    // When the types are different, compare decompressed pointers, because the
+    // pointers may need to be adjusted.
+    // TODO(1376980): Avoid decompression here.
+    return a.get() < b.get();
+  }
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<(CompressedPointer<T> a, U* b) {
+  // Do compression, since it is less expensive.
+  return a < static_cast<CompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<(T* a, CompressedPointer<U> b) {
+  // Do compression, since it is less expensive.
+  return static_cast<CompressedPointer<T>>(a) < b;
+}
+
+// operators<=.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<=(CompressedPointer<T> a,
+                                           CompressedPointer<U> b) {
+  if constexpr (internal::IsDecayedSame<T, U>) {
+    // When pointers have the same type modulo constness, simply compare
+    // compressed values.
+    return a.GetAsIntegral() <= b.GetAsIntegral();
+  } else {
+    // When the types are different, compare decompressed pointers, because the
+    // pointers may need to be adjusted.
+    // TODO(1376980): Avoid decompression here.
+    return a.get() <= b.get();
+  }
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<=(CompressedPointer<T> a, U* b) {
+  // Do compression, since it is less expensive.
+  return a <= static_cast<CompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<=(T* a, CompressedPointer<U> b) {
+  // Do compression, since it is less expensive.
+  return static_cast<CompressedPointer<T>>(a) <= b;
+}
+
+// operators>.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>(CompressedPointer<T> a,
+                                          CompressedPointer<U> b) {
+  return !(a <= b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>(CompressedPointer<T> a, U* b) {
+  // Do compression, since it is less expensive.
+  return a > static_cast<CompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>(T* a, CompressedPointer<U> b) {
+  // Do compression, since it is less expensive.
+  return static_cast<CompressedPointer<T>>(a) > b;
+}
+
+// operators>=.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>=(CompressedPointer<T> a,
+                                           CompressedPointer<U> b) {
+  return !(a < b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>=(CompressedPointer<T> a, U* b) {
+  // Do compression, since it is less expensive.
+  return a >= static_cast<CompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>=(T* a, CompressedPointer<U> b) {
+  // Do compression, since it is less expensive.
+  return static_cast<CompressedPointer<T>>(a) >= b;
+}
+
+#endif  // BUILDFLAG(ENABLE_POINTER_COMPRESSION)
+
+// Simple wrapper over the raw pointer.
+template <typename T>
+class PA_TRIVIAL_ABI UncompressedPointer final {
+ public:
+  PA_ALWAYS_INLINE constexpr UncompressedPointer() = default;
+  PA_ALWAYS_INLINE constexpr explicit UncompressedPointer(T* ptr) : ptr_(ptr) {}
+  PA_ALWAYS_INLINE constexpr explicit UncompressedPointer(std::nullptr_t)
+      : ptr_(nullptr) {}
+
+  PA_ALWAYS_INLINE constexpr UncompressedPointer(const UncompressedPointer&) =
+      default;
+  PA_ALWAYS_INLINE constexpr UncompressedPointer(
+      UncompressedPointer&& other) noexcept = default;
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr explicit UncompressedPointer(
+      const UncompressedPointer<U>& other)
+      : ptr_(other.ptr_) {}
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr explicit UncompressedPointer(
+      UncompressedPointer<U>&& other) noexcept
+      : ptr_(std::move(other.ptr_)) {}
+
+  ~UncompressedPointer() = default;
+
+  PA_ALWAYS_INLINE constexpr UncompressedPointer& operator=(
+      const UncompressedPointer&) = default;
+  PA_ALWAYS_INLINE constexpr UncompressedPointer& operator=(
+      UncompressedPointer&& other) noexcept = default;
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr UncompressedPointer& operator=(
+      const UncompressedPointer<U>& other) {
+    ptr_ = other.ptr_;
+    return *this;
+  }
+
+  template <typename U,
+            std::enable_if_t<std::is_convertible_v<U*, T*>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr UncompressedPointer& operator=(
+      UncompressedPointer<U>&& other) noexcept {
+    ptr_ = std::move(other.ptr_);
+    return *this;
+  }
+
+  PA_ALWAYS_INLINE constexpr UncompressedPointer& operator=(std::nullptr_t) {
+    ptr_ = nullptr;
+    return *this;
+  }
+
+  PA_ALWAYS_INLINE constexpr T* get() const { return ptr_; }
+
+  PA_ALWAYS_INLINE constexpr bool is_nonnull() const { return ptr_; }
+
+  PA_ALWAYS_INLINE constexpr explicit operator bool() const {
+    return is_nonnull();
+  }
+
+  template <typename U = T,
+            std::enable_if_t<!std::is_void_v<std::remove_cv_t<U>>>* = nullptr>
+  PA_ALWAYS_INLINE constexpr U& operator*() const {
+    PA_DCHECK(is_nonnull());
+    return *get();
+  }
+
+  PA_ALWAYS_INLINE constexpr T* operator->() const {
+    PA_DCHECK(is_nonnull());
+    return get();
+  }
+
+  PA_ALWAYS_INLINE constexpr void swap(UncompressedPointer& other) {
+    std::swap(ptr_, other.ptr_);
+  }
+
+ private:
+  template <typename>
+  friend class UncompressedPointer;
+
+  T* ptr_;
+};
+
+template <typename T>
+PA_ALWAYS_INLINE constexpr void swap(UncompressedPointer<T>& a,
+                                     UncompressedPointer<T>& b) {
+  a.swap(b);
+}
+
+// operators==.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator==(UncompressedPointer<T> a,
+                                           UncompressedPointer<U> b) {
+  return a.get() == b.get();
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator==(UncompressedPointer<T> a, U* b) {
+  return a == static_cast<UncompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator==(T* a, UncompressedPointer<U> b) {
+  return b == a;
+}
+
+template <typename T>
+PA_ALWAYS_INLINE constexpr bool operator==(UncompressedPointer<T> a,
+                                           std::nullptr_t) {
+  return !a.is_nonnull();
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator==(std::nullptr_t,
+                                           UncompressedPointer<U> b) {
+  return b == nullptr;
+}
+
+// operators!=.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(UncompressedPointer<T> a,
+                                           UncompressedPointer<U> b) {
+  return !(a == b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(UncompressedPointer<T> a, U* b) {
+  return a != static_cast<UncompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(T* a, UncompressedPointer<U> b) {
+  return b != a;
+}
+
+template <typename T>
+PA_ALWAYS_INLINE constexpr bool operator!=(UncompressedPointer<T> a,
+                                           std::nullptr_t) {
+  return a.is_nonnull();
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator!=(std::nullptr_t,
+                                           UncompressedPointer<U> b) {
+  return b != nullptr;
+}
+
+// operators<.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<(UncompressedPointer<T> a,
+                                          UncompressedPointer<U> b) {
+  return a.get() < b.get();
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<(UncompressedPointer<T> a, U* b) {
+  return a < static_cast<UncompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<(T* a, UncompressedPointer<U> b) {
+  return static_cast<UncompressedPointer<T>>(a) < b;
+}
+
+// operators<=.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<=(UncompressedPointer<T> a,
+                                           UncompressedPointer<U> b) {
+  return a.get() <= b.get();
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<=(UncompressedPointer<T> a, U* b) {
+  return a <= static_cast<UncompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator<=(T* a, UncompressedPointer<U> b) {
+  return static_cast<UncompressedPointer<T>>(a) <= b;
+}
+
+// operators>.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>(UncompressedPointer<T> a,
+                                          UncompressedPointer<U> b) {
+  return !(a <= b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>(UncompressedPointer<T> a, U* b) {
+  return a > static_cast<UncompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>(T* a, UncompressedPointer<U> b) {
+  return static_cast<UncompressedPointer<T>>(a) > b;
+}
+
+// operators>=.
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>=(UncompressedPointer<T> a,
+                                           UncompressedPointer<U> b) {
+  return !(a < b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>=(UncompressedPointer<T> a, U* b) {
+  return a >= static_cast<UncompressedPointer<U>>(b);
+}
+
+template <typename T, typename U>
+PA_ALWAYS_INLINE constexpr bool operator>=(T* a, UncompressedPointer<U> b) {
+  return static_cast<UncompressedPointer<T>>(a) >= b;
+}
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_COMPRESSED_POINTER_H_

src/base/allocator/partition_allocator/dangling_raw_ptr_checks.cc

@@ -0,0 +1,75 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/dangling_raw_ptr_checks.h"
+
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+
+namespace partition_alloc {
+
+namespace {
+DanglingRawPtrDetectedFn* g_dangling_raw_ptr_detected_fn = [](uintptr_t) {};
+DanglingRawPtrReleasedFn* g_dangling_raw_ptr_released_fn = [](uintptr_t) {};
+DanglingRawPtrDetectedFn* g_unretained_dangling_raw_ptr_detected_fn =
+    [](uintptr_t) {};
+bool g_unretained_dangling_raw_ptr_check_enabled = false;
+}  // namespace
+
+DanglingRawPtrDetectedFn* GetDanglingRawPtrDetectedFn() {
+  PA_DCHECK(g_dangling_raw_ptr_detected_fn);
+  return g_dangling_raw_ptr_detected_fn;
+}
+
+DanglingRawPtrDetectedFn* GetDanglingRawPtrReleasedFn() {
+  PA_DCHECK(g_dangling_raw_ptr_released_fn);
+  return g_dangling_raw_ptr_released_fn;
+}
+
+void SetDanglingRawPtrDetectedFn(DanglingRawPtrDetectedFn fn) {
+  PA_DCHECK(fn);
+  g_dangling_raw_ptr_detected_fn = fn;
+}
+
+void SetDanglingRawPtrReleasedFn(DanglingRawPtrReleasedFn fn) {
+  PA_DCHECK(fn);
+  g_dangling_raw_ptr_released_fn = fn;
+}
+
+DanglingRawPtrDetectedFn* GetUnretainedDanglingRawPtrDetectedFn() {
+  return g_unretained_dangling_raw_ptr_detected_fn;
+}
+
+void SetUnretainedDanglingRawPtrDetectedFn(DanglingRawPtrDetectedFn* fn) {
+  PA_DCHECK(fn);
+  g_unretained_dangling_raw_ptr_detected_fn = fn;
+}
+
+bool SetUnretainedDanglingRawPtrCheckEnabled(bool enabled) {
+  bool old = g_unretained_dangling_raw_ptr_check_enabled;
+  g_unretained_dangling_raw_ptr_check_enabled = enabled;
+  return old;
+}
+
+namespace internal {
+
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) void DanglingRawPtrDetected(uintptr_t id) {
+  g_dangling_raw_ptr_detected_fn(id);
+}
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) void DanglingRawPtrReleased(uintptr_t id) {
+  g_dangling_raw_ptr_released_fn(id);
+}
+
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void UnretainedDanglingRawPtrDetected(uintptr_t id) {
+  g_unretained_dangling_raw_ptr_detected_fn(id);
+}
+
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+bool IsUnretainedDanglingRawPtrCheckEnabled() {
+  return g_unretained_dangling_raw_ptr_check_enabled;
+}
+
+}  // namespace internal
+}  // namespace partition_alloc

src/base/allocator/partition_allocator/dangling_raw_ptr_checks.h

@@ -0,0 +1,67 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_DANGLING_RAW_PTR_CHECKS_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_DANGLING_RAW_PTR_CHECKS_H_
+
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+
+// When compiled with build flags `enable_dangling_raw_ptr_checks`, dangling
+// raw_ptr are reported. Its behavior can be configured here.
+//
+// Purpose of this level of indirection:
+// - Ease testing.
+// - Keep partition_alloc/ independent from base/. In most cases, when a
+//   dangling raw_ptr is detected/released, this involves recording a
+//   base::debug::StackTrace, which isn't desirable inside partition_alloc/.
+// - Be able (potentially) to turn this feature on/off at runtime based on
+//   dependant's flags.
+namespace partition_alloc {
+
+// DanglingRawPtrDetected is called when there exists a `raw_ptr` referencing a
+// memory region and the allocator is asked to release it.
+//
+// It won't be called again with the same `id`, up until (potentially) a call to
+// DanglingRawPtrReleased(`id`) is made.
+//
+// This function is called from within the allocator, and is not allowed to
+// allocate memory.
+using DanglingRawPtrDetectedFn = void(uintptr_t /*id*/);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+DanglingRawPtrDetectedFn* GetDanglingRawPtrDetectedFn();
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void SetDanglingRawPtrDetectedFn(DanglingRawPtrDetectedFn);
+
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+DanglingRawPtrDetectedFn* GetUnretainedDanglingRawPtrDetectedFn();
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void SetUnretainedDanglingRawPtrDetectedFn(DanglingRawPtrDetectedFn*);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+bool SetUnretainedDanglingRawPtrCheckEnabled(bool enabled);
+
+// DanglingRawPtrReleased: Called after DanglingRawPtrDetected(id), once the
+// last dangling raw_ptr stops referencing the memory region.
+//
+// This function is allowed to allocate memory.
+using DanglingRawPtrReleasedFn = void(uintptr_t /*id*/);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+DanglingRawPtrReleasedFn* GetDanglingRawPtrReleasedFn();
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void SetDanglingRawPtrReleasedFn(DanglingRawPtrReleasedFn);
+
+namespace internal {
+
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) void DanglingRawPtrDetected(uintptr_t id);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) void DanglingRawPtrReleased(uintptr_t id);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void UnretainedDanglingRawPtrDetected(uintptr_t id);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+bool IsUnretainedDanglingRawPtrCheckEnabled();
+
+}  // namespace internal
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_DANGLING_RAW_PTR_CHECKS_H_

src/base/allocator/partition_allocator/dot/address-space.dot

@@ -0,0 +1,34 @@
+digraph {
+  node[shape=box]
+  edge[dir=both]
+  compound = true
+  bgcolor = transparent
+  dpi = 192
+  nodesep = 0.91
+  // Allows aligning nodes in different subgraphs.
+  newrank = true
+
+  subgraph cluster_0 {
+    label = "Address Space"
+    reg[label="Regular Pool"]
+    brp[label="BRP Pool"]
+    add[label="Additional Pools"]
+    reg->brp->add[style=invis]
+  }
+
+  manager[label="AddressPoolManager"]
+  manager->reg[constraint=false]
+  manager->brp
+  manager->add[constraint=false]
+
+  subgraph cluster_1 {
+    label = "PartitionRoots"
+    pae[label="PA-E Root"]
+    blink[label="Blink Roots"]
+    etc[style=dotted, label="Other Roots"]
+    pae->blink->etc[style=invis]
+  }
+
+  manager->blink[lhead=cluster_1]
+  {rank=same manager brp blink}
+}

src/base/allocator/partition_allocator/dot/bucket.dot

@@ -0,0 +1,60 @@
+digraph {
+  graph[bgcolor=transparent]
+  node[shape=plaintext]
+  edge[style=dashed, color=crimson]
+
+  page1[label=<
+    <table border="0" cellborder="1" cellspacing="0"><tr>
+      <!-- head partition page -->
+      <td port="head" bgcolor="darkgrey" width="40" height="52"></td>
+      <!-- bucket-external memory - not depicted -->
+      <td width="160"></td>
+      <!-- a slot span in this bucket -->
+   <td port="slotspan" bgcolor="crimson" width="80"></td>
+      <!-- bucket-external memory - not depicted -->
+      <td width="320"></td>
+      <!-- tail partition page -->
+      <td bgcolor="darkgrey" width="40"></td>
+    </tr></table>
+  >]
+  page2[label=<
+    <table border="0" cellborder="1" cellspacing="0"><tr>
+      <!-- head partition page -->
+      <td port="head" bgcolor="darkgrey" width="40" height="52"></td>
+      <!-- bucket-external memory - not depicted -->
+      <td width="280"></td>
+      <!-- a slot span in this bucket -->
+      <td port="slotspan" bgcolor="crimson" width="80"></td>
+      <!-- bucket-external memory - not depicted -->
+      <td width="200"></td>
+      <!-- tail partition page -->
+      <td bgcolor="darkgrey" width="40"></td>
+    </tr></table>
+  >]
+  page3[label=<
+    <table border="0" cellborder="1" cellspacing="0"><tr>
+      <!-- head partition page -->
+      <td port="head" bgcolor="darkgrey" width="40" height="52"></td>
+      <!-- bucket-external memory - not depicted -->
+      <td width="40"></td>
+      <!-- a slot span in this bucket -->
+      <td port="slotspan1" bgcolor="crimson" width="80"></td>
+      <!-- bucket-external memory - not depicted -->
+      <td width="120"></td>
+      <!-- a slot span in this bucket -->
+      <td port="slotspan2" bgcolor="crimson" width="80"></td>
+      <!-- bucket-external memory - not depicted -->
+      <td width="240"></td>
+      <!-- tail partition page -->
+      <td bgcolor="darkgrey" width="40"></td>
+    </tr></table>
+  >]
+
+  // Invisibly link the head partition pages to force alignment.
+  page1:head->page2:head->page3:head[style=invis]
+
+  // Inter-super-page links disable constraints so to let the above
+  // fully control alignment.
+  page1:slotspan->page2:slotspan->page3:slotspan1[constraint=false]
+  page3:slotspan1:s->page3:slotspan2:sw
+}

src/base/allocator/partition_allocator/dot/layers.dot

@@ -0,0 +1,23 @@
+digraph G {
+  graph[bgcolor=transparent]
+  node[shape=box,style="filled,rounded",color=deepskyblue]
+
+  subgraph cluster_tc {
+    label = "Thread Cache"
+    rankdir = LR
+    {rank=same;TLS1,TLS2,TLSn}
+    TLS1->TLS2[style=invisible,dir=none]
+    TLS2->TLSn[style=dotted,dir=none]
+  }
+
+  subgraph cluster_central {
+    label = "Central Allocator (per-partition lock)"
+    fast[label="slot span freelists (fast path)"]
+    slow[label="slot span management (slow path)"]
+    # Forces slow path node beneath fast path node.
+    fast->slow[style=invisible,dir=none]
+  }
+
+  # Forces thread-external subgraph beneath thread cache subgraph.
+  TLS2->fast[style=invisible,dir=none]
+}

src/base/allocator/partition_allocator/dot/super-page.dot

@@ -0,0 +1,95 @@
+digraph G {
+  graph[bgcolor=transparent]
+  node[shape=plaintext]
+  edge[style=dashed]
+
+  invisible_a[label=<
+    <TABLE BORDER="0" CELLBORDER="0" CELLSPACING="0">
+      <TR>
+        <TD PORT="red" WIDTH="100"></TD>
+        <TD PORT="green" WIDTH="20"></TD>
+        <TD PORT="blue" WIDTH="40"></TD>
+        <TD PORT="gold" WIDTH="300"></TD>
+        <TD PORT="pink" WIDTH="60"></TD>
+      </TR>
+    </TABLE>
+  >]
+  superpage[xlabel="Super Page",label=<
+    <TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0" WIDTH="10">
+      <TR>
+        <!-- Head Partition Page -->
+        <TD BGCOLOR="darkgrey" HEIGHT="52"></TD>
+        <TD PORT="metadata"></TD>
+        <TD BGCOLOR="darkgrey" WIDTH="18"></TD>
+        <!-- Bitmaps -->
+        <TD WIDTH="100">Bitmaps(?)</TD>
+        <!-- Several Slot Spans -->
+        <TD PORT="red" BGCOLOR="crimson" WIDTH="119">3</TD>
+        <TD PORT="green" BGCOLOR="palegreen" WIDTH="39">1</TD>
+        <TD PORT="blue" BGCOLOR="cornflowerblue" WIDTH="79">2</TD>
+        <TD PORT="gold" BGCOLOR="gold" WIDTH="239">6</TD>
+        <TD PORT="red2" BGCOLOR="crimson" WIDTH="119">3</TD>
+        <TD PORT="pink" BGCOLOR="deeppink" WIDTH="39">1</TD>
+        <TD WIDTH="79">...</TD>
+        <!-- Tail Partition Page -->
+        <TD BGCOLOR="darkgrey" WIDTH="39"></TD>
+      </TR>
+    </TABLE>
+  >]
+  invisible_b[label=<
+    <TABLE BORDER="0" CELLBORDER="0" CELLSPACING="0">
+      <TR>
+        <TD PORT="green" WIDTH="30"></TD>
+        <TD PORT="blue" WIDTH="60"></TD>
+        <TD PORT="gold" WIDTH="180"></TD>
+        <TD PORT="red" WIDTH="90"></TD>
+        <TD PORT="pink" WIDTH="90"></TD>
+      </TR>
+    </TABLE>
+  >]
+  metadata_page[xlabel="Metadata",label=<
+    <TABLE BORDER="0" CELLBORDER="1" CELLSPACING="0">
+      <TR>
+        <!-- Guard Page Metadata -->
+        <TD BGCOLOR="darkgrey"> </TD>
+        <!-- Bitmaps Offset -->
+        <TD> B? </TD>
+        <!-- Red Slot Span Metadata -->
+        <TD BGCOLOR="crimson">v</TD>
+        <TD BGCOLOR="crimson">+</TD>
+        <TD BGCOLOR="crimson">+</TD>
+        <!-- Green Slot Span Metadata -->
+        <TD BGCOLOR="palegreen">v</TD>
+        <!-- Blue Slot Span Metadata -->
+        <TD BGCOLOR="cornflowerblue">v</TD>
+        <TD BGCOLOR="cornflowerblue">+</TD>
+        <!-- Gold Slot Span Metadata -->
+        <TD BGCOLOR="gold">v</TD>
+        <TD BGCOLOR="gold">+</TD>
+        <TD BGCOLOR="gold">+</TD>
+        <TD BGCOLOR="gold">+</TD>
+        <TD BGCOLOR="gold">+</TD>
+        <TD BGCOLOR="gold">+</TD>
+        <!-- Red Slot Span Metadata -->
+        <TD BGCOLOR="crimson">v</TD>
+        <TD BGCOLOR="crimson">+</TD>
+        <TD BGCOLOR="crimson">+</TD>
+        <!-- Pink Slot Span Metadata -->
+        <TD BGCOLOR="deeppink">v</TD>
+        <!-- etc. -->
+        <TD WIDTH="64">...</TD>
+        <!-- Guard Page Metadata -->
+        <TD BGCOLOR="darkgrey"> </TD>
+      </TR>
+    </TABLE>
+  >]
+
+  invisible_a:red->superpage:red->superpage:red2[color=crimson]
+  superpage:red2->invisible_b:red[color=crimson]
+  invisible_a:green->superpage:green->invisible_b:green[color=palegreen]
+  invisible_a:blue->superpage:blue->invisible_b:blue[color=cornflowerblue]
+  invisible_a:gold->superpage:gold->invisible_b:gold[color=gold]
+  invisible_a:pink->superpage:pink->invisible_b:pink[color=deeppink]
+
+  superpage:metadata->metadata_page[style="",arrowhead=odot]
+}

src/base/allocator/partition_allocator/encoded_next_freelist.h

@@ -0,0 +1,320 @@
+// Copyright 2018 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_ENCODED_NEXT_FREELIST_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_ENCODED_NEXT_FREELIST_H_
+
+#include <cstddef>
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/freeslot_bitmap.h"
+#include "base/allocator/partition_allocator/partition_alloc-inl.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/debug/debugging_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/immediate_crash.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/partition_alloc_config.h"
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+#include "base/allocator/partition_allocator/partition_ref_count.h"
+#include "build/build_config.h"
+
+#if !defined(ARCH_CPU_BIG_ENDIAN)
+#include "base/allocator/partition_allocator/reverse_bytes.h"
+#endif  // !defined(ARCH_CPU_BIG_ENDIAN)
+
+namespace partition_alloc::internal {
+
+class EncodedNextFreelistEntry;
+
+class EncodedFreelistPtr {
+ private:
+  PA_ALWAYS_INLINE constexpr explicit EncodedFreelistPtr(std::nullptr_t)
+      : encoded_(Transform(0)) {}
+  PA_ALWAYS_INLINE explicit EncodedFreelistPtr(void* ptr)
+      // The encoded pointer stays MTE-tagged.
+      : encoded_(Transform(reinterpret_cast<uintptr_t>(ptr))) {}
+
+  PA_ALWAYS_INLINE EncodedNextFreelistEntry* Decode() const {
+    return reinterpret_cast<EncodedNextFreelistEntry*>(Transform(encoded_));
+  }
+
+  PA_ALWAYS_INLINE constexpr uintptr_t Inverted() const { return ~encoded_; }
+
+  PA_ALWAYS_INLINE constexpr void Override(uintptr_t encoded) {
+    encoded_ = encoded;
+  }
+
+  PA_ALWAYS_INLINE constexpr explicit operator bool() const { return encoded_; }
+
+  // Transform() works the same in both directions, so can be used for
+  // encoding and decoding.
+  PA_ALWAYS_INLINE static constexpr uintptr_t Transform(uintptr_t address) {
+    // We use bswap on little endian as a fast transformation for two reasons:
+    // 1) On 64 bit architectures, the pointer is very unlikely to be a
+    //    canonical address. Therefore, if an object is freed and its vtable is
+    //    used where the attacker doesn't get the chance to run allocations
+    //    between the free and use, the vtable dereference is likely to fault.
+    // 2) If the attacker has a linear buffer overflow and elects to try and
+    //    corrupt a freelist pointer, partial pointer overwrite attacks are
+    //    thwarted.
+    // For big endian, similar guarantees are arrived at with a negation.
+#if defined(ARCH_CPU_BIG_ENDIAN)
+    uintptr_t transformed = ~address;
+#else
+    uintptr_t transformed = ReverseBytes(address);
+#endif
+    return transformed;
+  }
+
+  uintptr_t encoded_;
+
+  friend EncodedNextFreelistEntry;
+};
+
+// Freelist entries are encoded for security reasons. See
+// //base/allocator/partition_allocator/PartitionAlloc.md and |Transform()| for
+// the rationale and mechanism, respectively.
+class EncodedNextFreelistEntry {
+ private:
+  constexpr explicit EncodedNextFreelistEntry(std::nullptr_t)
+      : encoded_next_(EncodedFreelistPtr(nullptr))
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+        ,
+        shadow_(encoded_next_.Inverted())
+#endif
+  {
+  }
+  explicit EncodedNextFreelistEntry(EncodedNextFreelistEntry* next)
+      : encoded_next_(EncodedFreelistPtr(next))
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+        ,
+        shadow_(encoded_next_.Inverted())
+#endif
+  {
+  }
+  // For testing only.
+  EncodedNextFreelistEntry(void* next, bool make_shadow_match)
+      : encoded_next_(EncodedFreelistPtr(next))
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+        ,
+        shadow_(make_shadow_match ? encoded_next_.Inverted() : 12345)
+#endif
+  {
+  }
+
+ public:
+  ~EncodedNextFreelistEntry() = delete;
+
+  // Emplaces the freelist entry at the beginning of the given slot span, and
+  // initializes it as null-terminated.
+  PA_ALWAYS_INLINE static EncodedNextFreelistEntry* EmplaceAndInitNull(
+      void* slot_start_tagged) {
+    // |slot_start_tagged| is MTE-tagged.
+    auto* entry = new (slot_start_tagged) EncodedNextFreelistEntry(nullptr);
+    return entry;
+  }
+  PA_ALWAYS_INLINE static EncodedNextFreelistEntry* EmplaceAndInitNull(
+      uintptr_t slot_start) {
+    return EmplaceAndInitNull(SlotStartAddr2Ptr(slot_start));
+  }
+
+  // Emplaces the freelist entry at the beginning of the given slot span, and
+  // initializes it with the given |next| pointer, but encoded.
+  //
+  // This freelist is built for the purpose of thread-cache. This means that we
+  // can't perform a check that this and the next pointer belong to the same
+  // super page, as thread-cache spans may chain slots across super pages.
+  PA_ALWAYS_INLINE static EncodedNextFreelistEntry*
+  EmplaceAndInitForThreadCache(uintptr_t slot_start,
+                               EncodedNextFreelistEntry* next) {
+    auto* entry =
+        new (SlotStartAddr2Ptr(slot_start)) EncodedNextFreelistEntry(next);
+    return entry;
+  }
+
+  // Emplaces the freelist entry at the beginning of the given slot span, and
+  // initializes it with the given |next| pointer.
+  //
+  // This is for testing purposes only! |make_shadow_match| allows you to choose
+  // if the shadow matches the next pointer properly or is trash.
+  PA_ALWAYS_INLINE static void EmplaceAndInitForTest(uintptr_t slot_start,
+                                                     void* next,
+                                                     bool make_shadow_match) {
+    new (SlotStartAddr2Ptr(slot_start))
+        EncodedNextFreelistEntry(next, make_shadow_match);
+  }
+
+  void CorruptNextForTesting(uintptr_t v) {
+    // We just need a value that can never be a valid pointer here.
+    encoded_next_.Override(EncodedFreelistPtr::Transform(v));
+  }
+
+  // Puts `slot_size` on the stack before crashing in case of memory
+  // corruption. Meant to be used to report the failed allocation size.
+  template <bool crash_on_corruption>
+  PA_ALWAYS_INLINE EncodedNextFreelistEntry* GetNextForThreadCache(
+      size_t slot_size) const;
+  PA_ALWAYS_INLINE EncodedNextFreelistEntry* GetNext(size_t slot_size) const;
+
+  PA_NOINLINE void CheckFreeList(size_t slot_size) const {
+    for (auto* entry = this; entry; entry = entry->GetNext(slot_size)) {
+      // |GetNext()| checks freelist integrity.
+    }
+  }
+
+  PA_NOINLINE void CheckFreeListForThreadCache(size_t slot_size) const {
+    for (auto* entry = this; entry;
+         entry = entry->GetNextForThreadCache<true>(slot_size)) {
+      // |GetNextForThreadCache()| checks freelist integrity.
+    }
+  }
+
+  PA_ALWAYS_INLINE void SetNext(EncodedNextFreelistEntry* entry) {
+    // SetNext() is either called on the freelist head, when provisioning new
+    // slots, or when GetNext() has been called before, no need to pass the
+    // size.
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+    // Regular freelists always point to an entry within the same super page.
+    //
+    // This is most likely a PartitionAlloc bug if this triggers.
+    if (PA_UNLIKELY(entry &&
+                    (SlotStartPtr2Addr(this) & kSuperPageBaseMask) !=
+                        (SlotStartPtr2Addr(entry) & kSuperPageBaseMask))) {
+      FreelistCorruptionDetected(0);
+    }
+#endif  // BUILDFLAG(PA_DCHECK_IS_ON)
+
+    encoded_next_ = EncodedFreelistPtr(entry);
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+    shadow_ = encoded_next_.Inverted();
+#endif
+  }
+
+  // Zeroes out |this| before returning the slot. The pointer to this memory
+  // will be returned to the user (caller of Alloc()), thus can't have internal
+  // data.
+  PA_ALWAYS_INLINE uintptr_t ClearForAllocation() {
+    encoded_next_.Override(0);
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+    shadow_ = 0;
+#endif
+    return SlotStartPtr2Addr(this);
+  }
+
+  PA_ALWAYS_INLINE constexpr bool IsEncodedNextPtrZero() const {
+    return !encoded_next_;
+  }
+
+ private:
+  template <bool crash_on_corruption>
+  PA_ALWAYS_INLINE EncodedNextFreelistEntry* GetNextInternal(
+      size_t slot_size,
+      bool for_thread_cache) const;
+
+  PA_ALWAYS_INLINE static bool IsSane(const EncodedNextFreelistEntry* here,
+                                      const EncodedNextFreelistEntry* next,
+                                      bool for_thread_cache) {
+    // Don't allow the freelist to be blindly followed to any location.
+    // Checks two constraints:
+    // - here and next must belong to the same superpage, unless this is in the
+    //   thread cache (they even always belong to the same slot span).
+    // - next cannot point inside the metadata area.
+    //
+    // Also, the lightweight UaF detection (pointer shadow) is checked.
+
+    uintptr_t here_address = SlotStartPtr2Addr(here);
+    uintptr_t next_address = SlotStartPtr2Addr(next);
+
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+    bool shadow_ptr_ok = here->encoded_next_.Inverted() == here->shadow_;
+#else
+    bool shadow_ptr_ok = true;
+#endif
+
+    bool same_superpage = (here_address & kSuperPageBaseMask) ==
+                          (next_address & kSuperPageBaseMask);
+#if BUILDFLAG(USE_FREESLOT_BITMAP)
+    bool marked_as_free_in_bitmap =
+        for_thread_cache ? true : !FreeSlotBitmapSlotIsUsed(next_address);
+#else
+    bool marked_as_free_in_bitmap = true;
+#endif
+
+    // This is necessary but not sufficient when quarantine is enabled, see
+    // SuperPagePayloadBegin() in partition_page.h. However we don't want to
+    // fetch anything from the root in this function.
+    bool not_in_metadata =
+        (next_address & kSuperPageOffsetMask) >= PartitionPageSize();
+
+    if (for_thread_cache) {
+      return shadow_ptr_ok & not_in_metadata;
+    } else {
+      return shadow_ptr_ok & same_superpage & marked_as_free_in_bitmap &
+             not_in_metadata;
+    }
+  }
+
+  EncodedFreelistPtr encoded_next_;
+  // This is intended to detect unintentional corruptions of the freelist.
+  // These can happen due to a Use-after-Free, or overflow of the previous
+  // allocation in the slot span.
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+  uintptr_t shadow_;
+#endif
+};
+
+template <bool crash_on_corruption>
+PA_ALWAYS_INLINE EncodedNextFreelistEntry*
+EncodedNextFreelistEntry::GetNextInternal(size_t slot_size,
+                                          bool for_thread_cache) const {
+  // GetNext() can be called on discarded memory, in which case |encoded_next_|
+  // is 0, and none of the checks apply. Don't prefetch nullptr either.
+  if (IsEncodedNextPtrZero()) {
+    return nullptr;
+  }
+
+  auto* ret = encoded_next_.Decode();
+  // We rely on constant propagation to remove the branches coming from
+  // |for_thread_cache|, since the argument is always a compile-time constant.
+  if (PA_UNLIKELY(!IsSane(this, ret, for_thread_cache))) {
+    if constexpr (crash_on_corruption) {
+      // Put the corrupted data on the stack, it may give us more information
+      // about what kind of corruption that was.
+      PA_DEBUG_DATA_ON_STACK("first",
+                             static_cast<size_t>(encoded_next_.encoded_));
+#if PA_CONFIG(HAS_FREELIST_SHADOW_ENTRY)
+      PA_DEBUG_DATA_ON_STACK("second", static_cast<size_t>(shadow_));
+#endif
+      FreelistCorruptionDetected(slot_size);
+    } else {
+      return nullptr;
+    }
+  }
+
+  // In real-world profiles, the load of |encoded_next_| above is responsible
+  // for a large fraction of the allocation cost. However, we cannot anticipate
+  // it enough since it is accessed right after we know its address.
+  //
+  // In the case of repeated allocations, we can prefetch the access that will
+  // be done at the *next* allocation, which will touch *ret, prefetch it.
+  PA_PREFETCH(ret);
+
+  return ret;
+}
+
+template <bool crash_on_corruption>
+PA_ALWAYS_INLINE EncodedNextFreelistEntry*
+EncodedNextFreelistEntry::GetNextForThreadCache(size_t slot_size) const {
+  return GetNextInternal<crash_on_corruption>(slot_size, true);
+}
+
+PA_ALWAYS_INLINE EncodedNextFreelistEntry* EncodedNextFreelistEntry::GetNext(
+    size_t slot_size) const {
+  return GetNextInternal<true>(slot_size, false);
+}
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_ENCODED_NEXT_FREELIST_H_

src/base/allocator/partition_allocator/extended_api.cc

@@ -0,0 +1,135 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/extended_api.h"
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_config.h"
+#include "base/allocator/partition_allocator/shim/allocator_shim_default_dispatch_to_partition_alloc.h"
+#include "base/allocator/partition_allocator/thread_cache.h"
+
+namespace partition_alloc::internal {
+
+#if PA_CONFIG(THREAD_CACHE_SUPPORTED)
+
+namespace {
+
+void DisableThreadCacheForRootIfEnabled(PartitionRoot* root) {
+  // Some platforms don't have a thread cache, or it could already have been
+  // disabled.
+  if (!root || !root->settings.with_thread_cache) {
+    return;
+  }
+
+  ThreadCacheRegistry::Instance().PurgeAll();
+  root->settings.with_thread_cache = false;
+  // Doesn't destroy the thread cache object(s). For background threads, they
+  // will be collected (and free cached memory) at thread destruction
+  // time. For the main thread, we leak it.
+}
+
+void EnablePartitionAllocThreadCacheForRootIfDisabled(PartitionRoot* root) {
+  if (!root) {
+    return;
+  }
+  root->settings.with_thread_cache = true;
+}
+
+#if BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+void DisablePartitionAllocThreadCacheForProcess() {
+  PA_CHECK(allocator_shim::internal::PartitionAllocMalloc::
+               AllocatorConfigurationFinalized());
+  auto* regular_allocator =
+      allocator_shim::internal::PartitionAllocMalloc::Allocator();
+  auto* aligned_allocator =
+      allocator_shim::internal::PartitionAllocMalloc::AlignedAllocator();
+  DisableThreadCacheForRootIfEnabled(regular_allocator);
+  if (aligned_allocator != regular_allocator) {
+    DisableThreadCacheForRootIfEnabled(aligned_allocator);
+  }
+  DisableThreadCacheForRootIfEnabled(
+      allocator_shim::internal::PartitionAllocMalloc::OriginalAllocator());
+}
+#endif  // BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+
+}  // namespace
+
+#endif  // PA_CONFIG(THREAD_CACHE_SUPPORTED)
+
+ThreadAllocStats GetAllocStatsForCurrentThread() {
+  ThreadCache* thread_cache = ThreadCache::Get();
+  if (ThreadCache::IsValid(thread_cache)) {
+    return thread_cache->thread_alloc_stats();
+  }
+  return {};
+}
+
+#if PA_CONFIG(THREAD_CACHE_SUPPORTED)
+ThreadCacheProcessScopeForTesting::ThreadCacheProcessScopeForTesting(
+    PartitionRoot* root)
+    : root_(root) {
+#if BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+  auto* regular_allocator =
+      allocator_shim::internal::PartitionAllocMalloc::Allocator();
+  regular_was_enabled_ =
+      regular_allocator && regular_allocator->settings.with_thread_cache;
+
+  if (root_ != regular_allocator) {
+    // Another |root| is ThreadCache's PartitionRoot. Need to disable
+    // thread cache for the process.
+    DisablePartitionAllocThreadCacheForProcess();
+    EnablePartitionAllocThreadCacheForRootIfDisabled(root_);
+    // Replace ThreadCache's PartitionRoot.
+    ThreadCache::SwapForTesting(root_);
+  } else {
+    if (!regular_was_enabled_) {
+      EnablePartitionAllocThreadCacheForRootIfDisabled(root_);
+      ThreadCache::SwapForTesting(root_);
+    }
+  }
+#else
+  PA_CHECK(!ThreadCache::IsValid(ThreadCache::Get()));
+  EnablePartitionAllocThreadCacheForRootIfDisabled(root_);
+  ThreadCache::SwapForTesting(root_);
+#endif  // BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+
+  PA_CHECK(ThreadCache::Get());
+}
+
+ThreadCacheProcessScopeForTesting::~ThreadCacheProcessScopeForTesting() {
+#if BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+  auto* regular_allocator =
+      allocator_shim::internal::PartitionAllocMalloc::Allocator();
+  bool regular_enabled =
+      regular_allocator && regular_allocator->settings.with_thread_cache;
+
+  if (regular_was_enabled_) {
+    if (!regular_enabled) {
+      // Need to re-enable ThreadCache for the process.
+      EnablePartitionAllocThreadCacheForRootIfDisabled(regular_allocator);
+      // In the case, |regular_allocator| must be ThreadCache's root.
+      ThreadCache::SwapForTesting(regular_allocator);
+    } else {
+      // ThreadCache is enabled for the process, but we need to be
+      // careful about ThreadCache's PartitionRoot. If it is different from
+      // |regular_allocator|, we need to invoke SwapForTesting().
+      if (regular_allocator != root_) {
+        ThreadCache::SwapForTesting(regular_allocator);
+      }
+    }
+  } else {
+    // ThreadCache for all processes was disabled.
+    DisableThreadCacheForRootIfEnabled(regular_allocator);
+    ThreadCache::SwapForTesting(nullptr);
+  }
+#else
+  // First, disable the test thread cache we have.
+  DisableThreadCacheForRootIfEnabled(root_);
+
+  ThreadCache::SwapForTesting(nullptr);
+#endif  // BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+}
+#endif  // PA_CONFIG(THREAD_CACHE_SUPPORTED)
+
+}  // namespace partition_alloc::internal

src/base/allocator/partition_allocator/extended_api.h

@@ -0,0 +1,42 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_EXTENDED_API_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_EXTENDED_API_H_
+
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_root.h"
+#include "base/allocator/partition_allocator/partition_stats.h"
+#include "base/allocator/partition_allocator/thread_cache.h"
+
+namespace partition_alloc::internal {
+// Get allocation stats for the thread cache partition on the current
+// thread. See the documentation of ThreadAllocStats for details.
+ThreadAllocStats GetAllocStatsForCurrentThread();
+
+// Creates a scope for testing which:
+// - if the given |root| is a default malloc root for the entire process,
+//   enables the thread cache for the entire process.
+//   (This may happen if UsePartitionAllocAsMalloc is enabled.)
+// - otherwise, disables the thread cache for the entire process, and
+//   replaces it with a thread cache for |root|.
+// This class is unsafe to run if there are multiple threads running
+// in the process.
+class ThreadCacheProcessScopeForTesting {
+ public:
+  explicit ThreadCacheProcessScopeForTesting(PartitionRoot* root);
+  ~ThreadCacheProcessScopeForTesting();
+
+  ThreadCacheProcessScopeForTesting() = delete;
+
+ private:
+  PartitionRoot* root_ = nullptr;
+#if BUILDFLAG(USE_PARTITION_ALLOC_AS_MALLOC)
+  bool regular_was_enabled_ = false;
+#endif
+};
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_EXTENDED_API_H_

src/base/allocator/partition_allocator/external_builds.md

@@ -0,0 +1,76 @@
+# Chrome-External Builds
+
+Work is ongoing to make PartitionAlloc a standalone library. The
+standalone repository for PartitionAlloc is hosted
+[here][standalone-PA-repo].
+
+## GN Args
+
+External clients should examine the args described in
+`build_overrides/partition_alloc.gni` and add them in their own source
+tree. PartitionAlloc's build will expect them at
+`//build_overrides/partition_alloc.gni`.
+
+In addition, something must provide `build_with_chromium = false` to
+the PA build system.
+
+## `use_partition_alloc`
+
+The `use_partition_alloc` GN arg, described in
+[`build_config.md`](./build_config.md), provides a GN-level seam that
+embedders
+
+1.  can set in their GN args and
+2.  should  observe in their GN recipes to conditionally pull in
+    PartitionAlloc.
+
+I.E. if you have any reason to disable PartitionAlloc, you should do so
+with this GN arg. Avoid pulling in PartitionAlloc headers when the
+corresponding buildflag is false.
+
+Setting `use_partition_alloc` false will also implicitly disable other
+features, e.g. nixing the compilation of BackupRefPtr as the
+implementation of `raw_ptr<T>`.
+
+## Periodic Memory Reduction Routines
+
+PartitionAlloc provides APIs to
+
+* reclaim memory (see [memory\_reclaimer.h](./memory_reclaimer.h)) and
+
+* purge thread caches (see [thread\_cache.h](./thread_cache.h)).
+
+Both of these must be called by the embedder external to PartitionAlloc.
+PA provides neither an event loop nor timers of its own, delegating this
+to its clients.
+
+## Build Considerations
+
+External clients create constraints on PartitionAlloc's implementation.
+
+### C++17
+
+PartitionAlloc targets C++17. This is aligned with our first external
+client, PDFium, and may be further constrained by other clients. These
+impositions prevent us from moving in lockstep with Chrome's target
+C++ version.
+
+We do not even have guarantees of backported future features, e.g.
+C++20's designated initializers. Therefore, these cannot ship with
+PartitionAlloc.
+
+### MSVC Support
+
+PDFium supports MSVC. PartitionAlloc will have to match it.
+
+### MSVC Constraint: No Inline Assembly
+
+MSVC's syntax for `asm` blocks differs from the one widely adopted in
+parts of Chrome. But more generally,
+[MSVC doesn't support inline assembly on ARM and x64 processors][msvc-inline-assembly].
+Assembly blocks should be gated behind compiler-specific flags and
+replaced with intrinsics in the presence of `COMPILER_MSVC` (absent
+`__clang__`).
+
+[standalone-PA-repo]: https://chromium.googlesource.com/chromium/src/base/allocator/partition_allocator.git
+[msvc-inline-assembly]: https://docs.microsoft.com/en-us/cpp/assembler/inline/inline-assembler?view=msvc-170

src/base/allocator/partition_allocator/freeslot_bitmap.h

@@ -0,0 +1,140 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_FREESLOT_BITMAP_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_FREESLOT_BITMAP_H_
+
+#include <climits>
+#include <cstdint>
+#include <utility>
+
+#include "base/allocator/partition_allocator/freeslot_bitmap_constants.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/bits.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+
+#if BUILDFLAG(USE_FREESLOT_BITMAP)
+
+namespace partition_alloc::internal {
+
+PA_ALWAYS_INLINE uintptr_t GetFreeSlotBitmapAddressForPointer(uintptr_t ptr) {
+  uintptr_t super_page = ptr & kSuperPageBaseMask;
+  return SuperPageFreeSlotBitmapAddr(super_page);
+}
+
+// Calculates the cell address and the offset inside the cell corresponding to
+// the |slot_start|.
+PA_ALWAYS_INLINE std::pair<FreeSlotBitmapCellType*, size_t>
+GetFreeSlotBitmapCellPtrAndBitIndex(uintptr_t slot_start) {
+  uintptr_t slot_superpage_offset = slot_start & kSuperPageOffsetMask;
+  uintptr_t superpage_bitmap_start =
+      GetFreeSlotBitmapAddressForPointer(slot_start);
+  uintptr_t cell_addr = base::bits::AlignDown(
+      superpage_bitmap_start +
+          (slot_superpage_offset / kSmallestBucket) / CHAR_BIT,
+      sizeof(FreeSlotBitmapCellType));
+  PA_DCHECK(cell_addr < superpage_bitmap_start + kFreeSlotBitmapSize);
+  size_t bit_index =
+      (slot_superpage_offset / kSmallestBucket) & kFreeSlotBitmapOffsetMask;
+  PA_DCHECK(bit_index < kFreeSlotBitmapBitsPerCell);
+  return {reinterpret_cast<FreeSlotBitmapCellType*>(cell_addr), bit_index};
+}
+
+// This bitmap marks the used slot as 0 and free one as 1. This is because we
+// would like to set all the slots as "used" by default to prevent allocating a
+// used slot when the freelist entry is overwritten. The state of the bitmap is
+// expected to be synced with freelist (i.e. the bitmap is set to 1 if and only
+// if the slot is in the freelist).
+
+PA_ALWAYS_INLINE FreeSlotBitmapCellType CellWithAOne(size_t n) {
+  return static_cast<FreeSlotBitmapCellType>(1) << n;
+}
+
+PA_ALWAYS_INLINE FreeSlotBitmapCellType CellWithTrailingOnes(size_t n) {
+  return (static_cast<FreeSlotBitmapCellType>(1) << n) -
+         static_cast<FreeSlotBitmapCellType>(1);
+}
+
+// Returns true if the bit corresponding to |slot_start| is used( = 0)
+PA_ALWAYS_INLINE bool FreeSlotBitmapSlotIsUsed(uintptr_t slot_start) {
+  auto [cell, bit_index] = GetFreeSlotBitmapCellPtrAndBitIndex(slot_start);
+  return (*cell & CellWithAOne(bit_index)) == 0;
+}
+
+// Mark the bit corresponding to |slot_start| as used( = 0).
+PA_ALWAYS_INLINE void FreeSlotBitmapMarkSlotAsUsed(uintptr_t slot_start) {
+  PA_CHECK(!FreeSlotBitmapSlotIsUsed(slot_start));
+  auto [cell, bit_index] = GetFreeSlotBitmapCellPtrAndBitIndex(slot_start);
+  *cell &= ~CellWithAOne(bit_index);
+}
+
+// Mark the bit corresponding to |slot_start| as free( = 1).
+PA_ALWAYS_INLINE void FreeSlotBitmapMarkSlotAsFree(uintptr_t slot_start) {
+  PA_CHECK(FreeSlotBitmapSlotIsUsed(slot_start));
+  auto [cell, bit_index] = GetFreeSlotBitmapCellPtrAndBitIndex(slot_start);
+  *cell |= CellWithAOne(bit_index);
+}
+
+// Resets (= set to 0) all the bits corresponding to the slot-start addresses
+// within [begin_addr, end_addr). |begin_addr| has to be the beginning of a
+// slot, but |end_addr| does not.
+PA_ALWAYS_INLINE void FreeSlotBitmapReset(uintptr_t begin_addr,
+                                          uintptr_t end_addr,
+                                          uintptr_t slot_size) {
+  PA_DCHECK(begin_addr <= end_addr);
+  // |end_addr| has to be kSmallestBucket-aligned.
+  PA_DCHECK((end_addr & (kSmallestBucket - 1)) == 0u);
+  for (uintptr_t slot_start = begin_addr; slot_start < end_addr;
+       slot_start += slot_size) {
+    auto [cell, bit_index] = GetFreeSlotBitmapCellPtrAndBitIndex(slot_start);
+    *cell &= ~CellWithAOne(bit_index);
+  }
+
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+  // Checks if the cells that are meant to contain only unset bits are really 0.
+  auto [begin_cell, begin_bit_index] =
+      GetFreeSlotBitmapCellPtrAndBitIndex(begin_addr);
+  auto [end_cell, end_bit_index] =
+      GetFreeSlotBitmapCellPtrAndBitIndex(end_addr);
+
+  // The bits that should be marked to 0 are |begin_bit_index|th bit of
+  // |begin_cell| to |end_bit_index - 1|th bit of |end_cell|. We verify all the
+  // bits are set to 0 for the cells between [begin_cell + 1, end_cell). For the
+  // |begin_cell| and |end_cell|, we have to handle them separately to only
+  // check the partial bits.
+  // | begin_cell |     |...|     | end_cell |
+  // |11...100...0|0...0|...|0...0|0...01...1|
+  //        ^                           ^
+  //        |                           |
+  //    begin_addr                   end_addr
+
+  if (begin_cell == end_cell) {
+    PA_DCHECK((*begin_cell & (~CellWithTrailingOnes(begin_bit_index) &
+                              CellWithTrailingOnes(end_bit_index))) == 0u);
+  }
+
+  if (begin_bit_index != 0) {
+    // Checks the bits between [begin_bit_index, kFreeSlotBitmapBitsPerCell) in
+    // the begin_cell are 0
+    PA_DCHECK((*begin_cell & ~CellWithTrailingOnes(begin_bit_index)) == 0u);
+    ++begin_cell;
+  }
+
+  if (end_bit_index != 0) {
+    // Checks the bits between [0, end_bit_index) in the end_cell are 0
+    PA_DCHECK((*end_cell & CellWithTrailingOnes(end_bit_index)) == 0u);
+  }
+
+  for (FreeSlotBitmapCellType* cell = begin_cell; cell != end_cell; ++cell) {
+    PA_DCHECK(*cell == 0u);
+  }
+#endif  // BUILDFLAG(PA_DCHECK_IS_ON)
+}
+
+}  // namespace partition_alloc::internal
+
+#endif  // BUILDFLAG(USE_FREESLOT_BITMAP)
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_FREESLOT_BITMAP_H_

src/base/allocator/partition_allocator/freeslot_bitmap_constants.h

@@ -0,0 +1,61 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_FREESLOT_BITMAP_CONSTANTS_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_FREESLOT_BITMAP_CONSTANTS_H_
+
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/partition_alloc_base/bits.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_constants.h"
+#include "base/allocator/partition_allocator/partition_alloc_forward.h"
+#include "base/allocator/partition_allocator/reservation_offset_table.h"
+
+namespace partition_alloc::internal {
+
+using FreeSlotBitmapCellType = uint64_t;
+constexpr size_t kFreeSlotBitmapBitsPerCell =
+    sizeof(FreeSlotBitmapCellType) * CHAR_BIT;
+constexpr size_t kFreeSlotBitmapOffsetMask = kFreeSlotBitmapBitsPerCell - 1;
+
+// The number of bits necessary for the bitmap is equal to the maximum number of
+// slots in a super page.
+constexpr size_t kFreeSlotBitmapSize =
+    (kSuperPageSize / kSmallestBucket) / CHAR_BIT;
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+ReservedFreeSlotBitmapSize() {
+#if BUILDFLAG(USE_FREESLOT_BITMAP)
+  return base::bits::AlignUp(kFreeSlotBitmapSize, PartitionPageSize());
+#else
+  return 0;
+#endif
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+CommittedFreeSlotBitmapSize() {
+#if BUILDFLAG(USE_FREESLOT_BITMAP)
+  return base::bits::AlignUp(kFreeSlotBitmapSize, SystemPageSize());
+#else
+  return 0;
+#endif
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+NumPartitionPagesPerFreeSlotBitmap() {
+  return ReservedFreeSlotBitmapSize() / PartitionPageSize();
+}
+
+#if BUILDFLAG(USE_FREESLOT_BITMAP)
+PA_ALWAYS_INLINE uintptr_t SuperPageFreeSlotBitmapAddr(uintptr_t super_page) {
+  PA_DCHECK(!(super_page % kSuperPageAlignment));
+  return super_page + PartitionPageSize();
+}
+#endif
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_FREESLOT_BITMAP_CONSTANTS_H_

src/base/allocator/partition_allocator/glossary.md

@@ -0,0 +1,186 @@
+# Glossary
+
+This page describes some core terminology used in PartitionAlloc.
+A weak attempt is made to present terms "in conceptual order" s.t.
+each term depends mainly upon previously defined ones.
+
+* **Partition**: A heap that is separated and protected both from other
+  partitions and from non-PartitionAlloc memory. Each partition holds
+  multiple buckets.
+
+*** promo
+**NOTE**: In code (and comments), "partition," "root," and even
+"allocator" are all conceptually the same thing.
+***
+
+## Pages
+
+* **System Page**: A memory page defined by the CPU/OS. Commonly
+  referred to as a "virtual page" in other contexts. This is typically
+  4KiB, but it can be larger. PartitionAlloc supports up to 64KiB,
+  though this constant isn't always known at compile time (depending
+  on the OS).
+* **Partition Page**: The most common granularity used by
+  PartitionAlloc. Consists of exactly 4 system pages.
+* **Super Page**: A 2MiB region, aligned on a 2MiB boundary. Not to
+  be confused with OS-level terms like "large page" or "huge page",
+  which are also commonly 2MiB. These have to be fully committed /
+  uncommitted in memory, whereas super pages can be partially committed
+  with system page granularity.
+* **Extent**: An extent is a run of consecutive super pages (belonging
+  to a single partition). Extents are to super pages what slot spans are
+  to slots (see below).
+
+## Slots and Spans
+
+* **Slot**: An indivisible allocation unit. Slot sizes are tied to
+  buckets. For example, each allocation that falls into the bucket
+  (224, 256] would be satisfied with a slot of size 256. This
+  applies only to normal buckets, not to direct map.
+* **Slot Span**: A run of same-sized slots that are contiguous in
+  memory. Slot span size is a multiple of partition page size, but it
+  isn't always a multiple of slot size, although we try hard for this
+  to be the case.
+  * **Small Bucket**: Allocations up to 4 partition pages. In these
+    cases, slot spans are always between 1 and 4 partition pages in
+    size. For each slot span size, the slot span is chosen to minimize
+    number of pages used while keeping the rounding waste under a
+    reasonable limit.
+    * For example, for a slot size 96, 64B waste is deemed acceptable
+      when using a single partition page, but for slot size
+      384, the potential waste of 256B wouldn't be, so 3 partition pages
+      are used to achieve 0B waste.
+    * PartitionAlloc may avoid waste by lowering the number of committed
+      system pages compared to the number of reserved pages. For
+      example, for the slot size of 896B we'd use a slot span of 2
+      partition pages of 16KiB, i.e. 8 system pages of 4KiB, but commit
+      only up to 7, thus resulting in perfect packing.
+  * **Single-Slot Span**: Allocations above 4 partition pages (but
+    ≤`kMaxBucketed`). This is because each slot span is guaranteed to
+    hold exactly one slot.
+    * Fun fact: there are sizes ≤4 partition pages that result in a
+      slot span having exactly 1 slot, but nonetheless they're still
+      classified as small buckets. The reason is that single-slot spans
+      are often handled by a different code path, and that distinction
+      is made purely based on slot size, for simplicity and efficiency.
+
+## Buckets
+
+* **Bucket**: A collection of regions in a partition that contains
+  similar-sized objects. For example, one bucket may hold objects of
+  size (224, 256], another (256, 320], etc. Bucket size
+  brackets are geometrically spaced,
+  [going up to `kMaxBucketed`][max-bucket-comment].
+  * Plainly put, all slots (ergo the resulting spans) of a given size
+    class are logically chained into one bucket.
+
+![A bucket, spanning multiple super pages, collects spans whose
+  slots are of a particular size class.](./dot/bucket.png)
+
+* **Normal Bucket**: Any bucket whose size ceiling does not exceed
+  `kMaxBucketed`. This is the common case in PartitionAlloc, and
+  the "normal" modifier is often dropped in casual reference.
+* **Direct Map (Bucket)**: Any allocation whose size exceeds `kMaxBucketed`.
+
+## Other Terms
+
+* **Object**: A chunk of memory returned to the allocating invoker
+  of the size requested. It doesn't have to span the entire slot,
+  nor does it have to begin at the slot start. This term is commonly
+  used as a parameter name in PartitionAlloc code, as opposed to
+  `slot_start`.
+* **Thread Cache**: A [thread-local structure][pa-thread-cache] that
+  holds some not-too-large memory chunks, ready to be allocated. This
+  speeds up in-thread allocation by reducing a lock hold to a
+  thread-local storage lookup, improving cache locality.
+* **Pool**: A large (and contiguous on 64-bit) virtual address region, housing
+  super pages, etc. from which PartitionAlloc services allocations. The
+  primary purpose of the pools is to provide a fast answer to the
+  question, "Did PartitionAlloc allocate the memory for this pointer
+  from this pool?" with a single bit-masking operation.
+  * The regular pool is a general purpose pool that contains allocations that
+    aren't protected by BackupRefPtr.
+  * The BRP pool contains all allocations protected by BackupRefPtr.
+  * [64-bit only] The configurable pool is named generically, because its
+    primary user (the [V8 Sandbox][v8-sandbox]) can configure it at runtime,
+    providing a pre-existing mapping. Its allocations aren't protected by
+    BackupRefPtr.
+  * [64-bit only] The thread isolated pool is returning memory protected with
+    per-thread permissions. At the moment, this is implemented for pkeys on x64.
+    It's primary user is [V8 CFI][v8-cfi].
+
+![The singular AddressPoolManager mediates access to the separate pools
+  for each PartitionRoot.](./dot/address-space.png)
+
+*** promo
+Pools are downgraded into a logical concept in 32-bit environments,
+tracking a non-contiguous set of allocations using a bitmap.
+***
+
+* **Payload**: The usable area of a super page in which slot spans
+  reside. While generally this means "everything between the first
+  and last guard partition pages in a super page," the presence of
+  other metadata (e.g. StarScan bitmaps) can bump the starting offset
+  forward. While this term is entrenched in the code, the team
+  considers it suboptimal and is actively looking for a replacement.
+* **Allocation Fast Path**: A path taken during an allocation that is
+  considered fast.  Usually means that an allocation request can be
+  immediately satisfied by grabbing a slot from the freelist of the
+  first active slot span in the bucket.
+* **Allocation Slow Path**: Anything which is not fast (see above).
+  Can involve
+  * finding another active slot span in the list,
+  * provisioning more slots in a slot span,
+  * bringing back a free (or decommitted) slot span,
+  * allocating a new slot span, or even
+  * allocating a new super page.
+
+*** aside
+By "slow" we may mean something as simple as extra logic (`if`
+statements etc.), or something as costly as system calls.
+***
+
+## Legacy Terms
+
+These terms are (mostly) deprecated and should not be used. They are
+surfaced here to provide a ready reference for readers coming from
+older design documents or documentation.
+
+* **GigaCage**: A memory region several gigabytes wide, reserved by
+  PartitionAlloc upon initialization, from which nearly all allocations
+  are taken. _Pools_ have overtaken GigaCage in conceptual importance,
+  and so and so there is less need today to refer to "GigaCage" or the
+  "cage." This is especially true given the V8 Sandbox and the
+  configurable pool (see above).
+
+## PartitionAlloc-Everywhere
+
+Originally, PartitionAlloc was used only in Blink (Chromium's rendering engine).
+It was invoked explicitly, by calling PartitionAlloc APIs directly.
+
+PartitionAlloc-Everywhere is the name of the project that brought PartitionAlloc
+to the entire-ish codebase (exclusions apply). This was done by intercepting
+`malloc()`, `free()`, `realloc()`, aforementioned `posix_memalign()`, etc. and
+routing them into PartitionAlloc. The shim located in
+`base/allocator/allocator_shim_default_dispatch_to_partition_alloc.h` is
+responsible for intercepting. For more details, see
+[base/allocator/README.md](../../../base/allocator/README.md).
+
+A special, catch-it-all *Malloc* partition has been created for the intercepted
+`malloc()` et al. This is to isolate from already existing Blink partitions.
+The only exception from that is Blink's *FastMalloc* partition, which was also
+catch-it-all in nature, so it's perfectly fine to merge these together, to
+minimize fragmentation.
+
+As of 2022, PartitionAlloc-Everywhere is supported on
+
+* Windows 32- and 64-bit
+* Linux
+* Android 32- and 64-bit
+* macOS
+* Fuchsia
+
+[max-bucket-comment]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/partition_allocator/partition_alloc_constants.h;l=345;drc=667e6b001f438521e1c1a1bc3eabeead7aaa1f37
+[pa-thread-cache]: https://source.chromium.org/chromium/chromium/src/+/main:base/allocator/partition_allocator/thread_cache.h
+[v8-sandbox]: https://docs.google.com/document/d/1FM4fQmIhEqPG8uGp5o9A-mnPB5BOeScZYpkHjo0KKA8/preview#
+[v8-cfi]: https://docs.google.com/document/d/1O2jwK4dxI3nRcOJuPYkonhTkNQfbmwdvxQMyXgeaRHo/preview#

src/base/allocator/partition_allocator/gwp_asan_support.cc

@@ -0,0 +1,125 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/gwp_asan_support.h"
+
+#if BUILDFLAG(ENABLE_GWP_ASAN_SUPPORT)
+
+#include "base/allocator/partition_allocator/freeslot_bitmap_constants.h"
+#include "base/allocator/partition_allocator/page_allocator_constants.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/no_destructor.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/partition_bucket.h"
+#include "base/allocator/partition_allocator/partition_lock.h"
+#include "base/allocator/partition_allocator/partition_page.h"
+#include "base/allocator/partition_allocator/partition_ref_count.h"
+#include "base/allocator/partition_allocator/partition_root.h"
+#include "build/build_config.h"
+
+namespace partition_alloc {
+
+// static
+void* GwpAsanSupport::MapRegion(size_t slot_count,
+                                std::vector<uint16_t>& free_list) {
+  PA_CHECK(slot_count > 0);
+
+  constexpr PartitionOptions kConfig{
+      .backup_ref_ptr = PartitionOptions::kEnabled,
+  };
+  static internal::base::NoDestructor<PartitionRoot> root(kConfig);
+
+  const size_t kSlotSize = 2 * internal::SystemPageSize();
+  uint16_t bucket_index = PartitionRoot::SizeToBucketIndex(
+      kSlotSize, root->GetBucketDistribution());
+  auto* bucket = root->buckets + bucket_index;
+
+  const size_t kSuperPagePayloadStartOffset =
+      internal::SuperPagePayloadStartOffset(
+          /* is_managed_by_normal_buckets = */ true,
+          /* with_quarantine = */ false);
+  PA_CHECK(kSuperPagePayloadStartOffset % kSlotSize == 0);
+  const size_t kSuperPageGwpAsanSlotAreaBeginOffset =
+      kSuperPagePayloadStartOffset;
+  const size_t kSuperPageGwpAsanSlotAreaEndOffset =
+      internal::SuperPagePayloadEndOffset();
+  const size_t kSuperPageGwpAsanSlotAreaSize =
+      kSuperPageGwpAsanSlotAreaEndOffset - kSuperPageGwpAsanSlotAreaBeginOffset;
+  const size_t kSlotsPerSlotSpan = bucket->get_bytes_per_span() / kSlotSize;
+  const size_t kSlotsPerSuperPage =
+      kSuperPageGwpAsanSlotAreaSize / (kSlotsPerSlotSpan * kSlotSize);
+
+  size_t super_page_count = 1 + ((slot_count - 1) / kSlotsPerSuperPage);
+  PA_CHECK(super_page_count <=
+           std::numeric_limits<size_t>::max() / kSuperPageSize);
+  uintptr_t super_page_span_start;
+  {
+    internal::ScopedGuard locker{internal::PartitionRootLock(root.get())};
+    super_page_span_start = bucket->AllocNewSuperPageSpanForGwpAsan(
+        root.get(), super_page_count, 0);
+
+    if (!super_page_span_start) {
+      return nullptr;
+    }
+
+#if defined(ARCH_CPU_64_BITS)
+    // Mapping the GWP-ASan region in to the lower 32-bits of address space
+    // makes it much more likely that a bad pointer dereference points into
+    // our region and triggers a false positive report. We rely on the fact
+    // that PA address pools are never allocated in the first 4GB due to
+    // their alignment requirements.
+    PA_CHECK(super_page_span_start >= (1ULL << 32));
+#endif  // defined(ARCH_CPU_64_BITS)
+
+    uintptr_t super_page_span_end =
+        super_page_span_start + super_page_count * kSuperPageSize;
+    PA_CHECK(super_page_span_start < super_page_span_end);
+
+    for (uintptr_t super_page = super_page_span_start;
+         super_page < super_page_span_end; super_page += kSuperPageSize) {
+      auto* page_metadata =
+          internal::PartitionSuperPageToMetadataArea(super_page);
+
+      // Index 0 is invalid because it is the super page extent metadata.
+      for (size_t partition_page_idx =
+               1 + internal::NumPartitionPagesPerFreeSlotBitmap();
+           partition_page_idx + bucket->get_pages_per_slot_span() <
+           internal::NumPartitionPagesPerSuperPage();
+           partition_page_idx += bucket->get_pages_per_slot_span()) {
+        auto* slot_span_metadata =
+            &page_metadata[partition_page_idx].slot_span_metadata;
+        bucket->InitializeSlotSpanForGwpAsan(slot_span_metadata);
+        auto slot_span_start =
+            internal::SlotSpanMetadata::ToSlotSpanStart(slot_span_metadata);
+
+        for (uintptr_t slot_idx = 0; slot_idx < kSlotsPerSlotSpan; ++slot_idx) {
+          auto slot_start = slot_span_start + slot_idx * kSlotSize;
+          internal::PartitionRefCountPointer(slot_start)->InitalizeForGwpAsan();
+          size_t global_slot_idx = (slot_start - super_page_span_start -
+                                    kSuperPageGwpAsanSlotAreaBeginOffset) /
+                                   kSlotSize;
+          PA_DCHECK(global_slot_idx < std::numeric_limits<uint16_t>::max());
+          free_list.push_back(global_slot_idx);
+          if (free_list.size() == slot_count) {
+            return reinterpret_cast<void*>(
+                super_page_span_start + kSuperPageGwpAsanSlotAreaBeginOffset -
+                internal::SystemPageSize());  // Depends on the PA guard region
+                                              // in front of the super page
+                                              // payload area.
+          }
+        }
+      }
+    }
+  }
+
+  PA_NOTREACHED();
+}
+
+// static
+bool GwpAsanSupport::CanReuse(uintptr_t slot_start) {
+  return internal::PartitionRefCountPointer(slot_start)->CanBeReusedByGwpAsan();
+}
+
+}  // namespace partition_alloc
+
+#endif  // BUILDFLAG(ENABLE_GWP_ASAN_SUPPORT)

src/base/allocator/partition_allocator/gwp_asan_support.h

@@ -0,0 +1,120 @@
+// Copyright 2022 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_GWP_ASAN_SUPPORT_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_GWP_ASAN_SUPPORT_H_
+
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+
+#if BUILDFLAG(ENABLE_GWP_ASAN_SUPPORT)
+
+#include <cstddef>
+#include <cstdint>
+#include <vector>
+
+namespace partition_alloc {
+
+// This class allows GWP-ASan allocations to be backed by PartitionAlloc and,
+// consequently, protected by MiraclePtr.
+//
+// GWP-ASan mainly operates at the system memory page granularity. During
+// process startup, it reserves a certain number of consecutive system pages.
+//
+// The standard layout is as follows:
+//
+//   +-------------------+--------
+//   |                   | ▲   ▲
+//   |   system page 0   |(a) (c)
+//   |                   | ▼   ▼
+//   +-------------------+--------
+//   |                   | ▲   ▲
+//   |   system page 1   |(b)  |
+//   |                   | ▼   |
+//   +-------------------+--- (d)    (a) inaccessible
+//   |                   | ▲   |     (b) accessible
+//   |   system page 2   |(a)  |     (c) initial guard page
+//   |                   | ▼   ▼     (d) allocation slot
+//   +-------------------+--------
+//   |                   | ▲   ▲
+//   |   system page 3   |(b)  |
+//   |                   | ▼   |
+//   +-------------------+--- (d)
+//   |                   | ▲   |
+//   |   system page 4   |(a)  |
+//   |                   | ▼   ▼
+//   |-------------------|--------
+//   |                   | ▲   ▲
+//   |        ...        |(a) (d)
+//
+// Unfortunately, PartitionAlloc can't provide GWP-ASan an arbitrary number of
+// consecutive allocation slots. Allocations need to be grouped into 2MB super
+// pages so that the allocation metadata can be easily located.
+//
+// Below is the new layout:
+//
+//   +-----------------------------------
+//   |                   |         ▲   ▲
+//   |   system page 0   |         |   |
+//   |                   |         |   |
+//   +-------------------+         |   |
+//   |                   |         |   |
+//   |        ...        |        (e)  |
+//   |                   |         |   |
+//   +-------------------+-------  |   |
+//   |                   | ▲   ▲   |   |
+//   |  system page k-1  |(a) (c)  |   |
+//   |                   | ▼   ▼   ▼   |
+//   +-------------------+----------- (f)
+//   |                   | ▲   ▲       |
+//   |   system page k   |(b)  |       |
+//   |                   | ▼   |       |
+//   +-------------------+--- (d)      |
+//   |                   | ▲   |       |
+//   |  system page k+1  |(a)  |       |
+//   |                   | ▼   ▼       |
+//   +-------------------+-----------  |
+//   |                   |             |    (a) inaccessible
+//   |        ...        |             |    (b) accessible
+//   |                   |             ▼    (c) initial guard page
+//   +-----------------------------------   (d) allocation slot
+//   |                   |         ▲   ▲    (e) super page metadata
+//   |   system page m   |         |   |    (f) super page
+//   |                   |         |   |    (g) pseudo allocation slot
+//   +-------------------+-------  |   |
+//   |                   |     ▲   |   |
+//   |        ...        |     |  (e)  |
+//   |                   |     |   |   |
+//   +-------------------+--- (g)  |   |
+//   |                   | ▲   |   |   |
+//   | system page m+k-1 |(a)  |   |   |
+//   |                   | ▼   ▼   ▼   |
+//   +-------------------+----------- (f)
+//   |                   | ▲   ▲       |
+//   |  system page m+k  |(b)  |       |
+//   |                   | ▼   |       |
+//   +-------------------+--- (d)      |
+//   |                   | ▲   |       |
+//   | system page m+k+1 |(a)  |       |
+//   |                   | ▼   ▼       |
+//   +-------------------+-----------  |
+//   |                   |             |
+//   |        ...        |             |
+//   |                   |             ▼
+//   +-------------------+---------------
+//
+// This means some allocation slots will be reserved to hold PA
+// metadata. We exclude these pseudo slots from the GWP-ASan free list so that
+// they are never used for anything other that storing the metadata.
+class PA_COMPONENT_EXPORT(PARTITION_ALLOC) GwpAsanSupport {
+ public:
+  static void* MapRegion(size_t slot_count, std::vector<uint16_t>& free_list);
+  static bool CanReuse(uintptr_t slot_start);
+};
+
+}  // namespace partition_alloc
+
+#endif  // BUILDFLAG(ENABLE_GWP_ASAN_SUPPORT)
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_GWP_ASAN_SUPPORT_H_

src/base/allocator/partition_allocator/memory_reclaimer.cc

@@ -0,0 +1,98 @@
+// Copyright 2019 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/memory_reclaimer.h"
+
+#include "base/allocator/partition_allocator/partition_alloc.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/no_destructor.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/partition_alloc_config.h"
+
+#if BUILDFLAG(USE_STARSCAN)
+#include "base/allocator/partition_allocator/starscan/pcscan.h"
+#endif
+
+namespace partition_alloc {
+
+// static
+MemoryReclaimer* MemoryReclaimer::Instance() {
+  static internal::base::NoDestructor<MemoryReclaimer> instance;
+  return instance.get();
+}
+
+void MemoryReclaimer::RegisterPartition(PartitionRoot* partition) {
+  internal::ScopedGuard lock(lock_);
+  PA_DCHECK(partition);
+  auto it_and_whether_inserted = partitions_.insert(partition);
+  PA_DCHECK(it_and_whether_inserted.second);
+}
+
+void MemoryReclaimer::UnregisterPartition(PartitionRoot* partition) {
+  internal::ScopedGuard lock(lock_);
+  PA_DCHECK(partition);
+  size_t erased_count = partitions_.erase(partition);
+  PA_DCHECK(erased_count == 1u);
+}
+
+MemoryReclaimer::MemoryReclaimer() = default;
+MemoryReclaimer::~MemoryReclaimer() = default;
+
+void MemoryReclaimer::ReclaimAll() {
+  constexpr int kFlags = PurgeFlags::kDecommitEmptySlotSpans |
+                         PurgeFlags::kDiscardUnusedSystemPages |
+                         PurgeFlags::kAggressiveReclaim;
+  Reclaim(kFlags);
+}
+
+void MemoryReclaimer::ReclaimNormal() {
+  constexpr int kFlags = PurgeFlags::kDecommitEmptySlotSpans |
+                         PurgeFlags::kDiscardUnusedSystemPages;
+  Reclaim(kFlags);
+}
+
+void MemoryReclaimer::Reclaim(int flags) {
+  internal::ScopedGuard lock(
+      lock_);  // Has to protect from concurrent (Un)Register calls.
+
+  // PCScan quarantines freed slots. Trigger the scan first to let it call
+  // FreeNoHooksImmediate on slots that pass the quarantine.
+  //
+  // In turn, FreeNoHooksImmediate may add slots to thread cache. Purge it next
+  // so that the slots are actually freed. (This is done synchronously only for
+  // the current thread.)
+  //
+  // Lastly decommit empty slot spans and lastly try to discard unused pages at
+  // the end of the remaining active slots.
+#if PA_CONFIG(STARSCAN_ENABLE_STARSCAN_ON_RECLAIM) && BUILDFLAG(USE_STARSCAN)
+  {
+    using PCScan = internal::PCScan;
+    const auto invocation_mode = flags & PurgeFlags::kAggressiveReclaim
+                                     ? PCScan::InvocationMode::kForcedBlocking
+                                     : PCScan::InvocationMode::kBlocking;
+    PCScan::PerformScanIfNeeded(invocation_mode);
+  }
+#endif  // PA_CONFIG(STARSCAN_ENABLE_STARSCAN_ON_RECLAIM) &&
+        // BUILDFLAG(USE_STARSCAN)
+
+#if PA_CONFIG(THREAD_CACHE_SUPPORTED)
+  // Don't completely empty the thread cache outside of low memory situations,
+  // as there is periodic purge which makes sure that it doesn't take too much
+  // space.
+  if (flags & PurgeFlags::kAggressiveReclaim) {
+    ThreadCacheRegistry::Instance().PurgeAll();
+  }
+#endif  // PA_CONFIG(THREAD_CACHE_SUPPORTED)
+
+  for (auto* partition : partitions_) {
+    partition->PurgeMemory(flags);
+  }
+}
+
+void MemoryReclaimer::ResetForTesting() {
+  internal::ScopedGuard lock(lock_);
+  partitions_.clear();
+}
+
+}  // namespace partition_alloc

src/base/allocator/partition_allocator/memory_reclaimer.h

@@ -0,0 +1,71 @@
+// Copyright 2019 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_MEMORY_RECLAIMER_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_MEMORY_RECLAIMER_H_
+
+#include <memory>
+#include <set>
+
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/no_destructor.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/thread_annotations.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/time/time.h"
+#include "base/allocator/partition_allocator/partition_alloc_forward.h"
+#include "base/allocator/partition_allocator/partition_lock.h"
+
+namespace partition_alloc {
+
+// Posts and handles memory reclaim tasks for PartitionAlloc.
+//
+// PartitionAlloc users are responsible for scheduling and calling the
+// reclamation methods with their own timers / event loops.
+//
+// Singleton as this runs as long as the process is alive, and
+// having multiple instances would be wasteful.
+class PA_COMPONENT_EXPORT(PARTITION_ALLOC) MemoryReclaimer {
+ public:
+  static MemoryReclaimer* Instance();
+
+  MemoryReclaimer(const MemoryReclaimer&) = delete;
+  MemoryReclaimer& operator=(const MemoryReclaimer&) = delete;
+
+  // Internal. Do not use.
+  // Registers a partition to be tracked by the reclaimer.
+  void RegisterPartition(PartitionRoot* partition);
+  // Internal. Do not use.
+  // Unregisters a partition to be tracked by the reclaimer.
+  void UnregisterPartition(PartitionRoot* partition);
+
+  // Triggers an explicit reclaim now to reclaim as much free memory as
+  // possible. The API callers need to invoke this method periodically
+  // if they want to use memory reclaimer.
+  // See also GetRecommendedReclaimIntervalInMicroseconds()'s comment.
+  void ReclaimNormal();
+
+  // Returns a recommended interval to invoke ReclaimNormal.
+  int64_t GetRecommendedReclaimIntervalInMicroseconds() {
+    return internal::base::Seconds(4).InMicroseconds();
+  }
+
+  // Triggers an explicit reclaim now reclaiming all free memory
+  void ReclaimAll();
+
+ private:
+  MemoryReclaimer();
+  ~MemoryReclaimer();
+  // |flags| is an OR of base::PartitionPurgeFlags
+  void Reclaim(int flags);
+  void ResetForTesting();
+
+  internal::Lock lock_;
+  std::set<PartitionRoot*> partitions_ PA_GUARDED_BY(lock_);
+
+  friend class internal::base::NoDestructor<MemoryReclaimer>;
+  friend class MemoryReclaimerTest;
+};
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_MEMORY_RECLAIMER_H_

src/base/allocator/partition_allocator/oom.cc

@@ -0,0 +1,81 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/oom.h"
+
+#include "base/allocator/partition_allocator/oom_callback.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/debug/alias.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/immediate_crash.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_WIN)
+#include <windows.h>
+
+#include <stdlib.h>
+
+#include <array>
+#endif  // BUILDFLAG(IS_WIN)
+
+namespace partition_alloc {
+
+size_t g_oom_size = 0U;
+
+namespace internal {
+
+// Crash server classifies base::internal::OnNoMemoryInternal as OOM.
+// TODO(crbug.com/1151236): Update to
+// partition_alloc::internal::base::internal::OnNoMemoryInternal
+PA_NOINLINE void OnNoMemoryInternal(size_t size) {
+  g_oom_size = size;
+#if BUILDFLAG(IS_WIN)
+  // Kill the process. This is important for security since most of code
+  // does not check the result of memory allocation.
+  // https://msdn.microsoft.com/en-us/library/het71c37.aspx
+  // Pass the size of the failed request in an exception argument.
+  ULONG_PTR exception_args[] = {size};
+  ::RaiseException(win::kOomExceptionCode, EXCEPTION_NONCONTINUABLE,
+                   std::size(exception_args), exception_args);
+
+  // Safety check, make sure process exits here.
+  _exit(win::kOomExceptionCode);
+#else
+  size_t tmp_size = size;
+  internal::base::debug::Alias(&tmp_size);
+
+  // Note: Don't add anything that may allocate here. Depending on the
+  // allocator, this may be called from within the allocator (e.g. with
+  // PartitionAlloc), and would deadlock as our locks are not recursive.
+  //
+  // Additionally, this is unlikely to work, since allocating from an OOM
+  // handler is likely to fail.
+  //
+  // Use PA_IMMEDIATE_CRASH() so that the top frame in the crash is our code,
+  // rather than using abort() or similar; this avoids the crash server needing
+  // to be able to successfully unwind through libc to get to the correct
+  // address, which is particularly an issue on Android.
+  PA_IMMEDIATE_CRASH();
+#endif  // BUILDFLAG(IS_WIN)
+}
+
+}  // namespace internal
+
+void TerminateBecauseOutOfMemory(size_t size) {
+  internal::OnNoMemoryInternal(size);
+}
+
+namespace internal {
+
+// The crash is generated in a PA_NOINLINE function so that we can classify the
+// crash as an OOM solely by analyzing the stack trace. It is tagged as
+// PA_NOT_TAIL_CALLED to ensure that its parent function stays on the stack.
+[[noreturn]] PA_NOINLINE PA_NOT_TAIL_CALLED void OnNoMemory(size_t size) {
+  RunPartitionAllocOomCallback();
+  TerminateBecauseOutOfMemory(size);
+  PA_IMMEDIATE_CRASH();
+}
+
+}  // namespace internal
+
+}  // namespace partition_alloc

src/base/allocator/partition_allocator/oom.h

@@ -0,0 +1,70 @@
+// Copyright 2016 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_OOM_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_OOM_H_
+
+#include <cstddef>
+
+#include "base/allocator/partition_allocator/allocation_guard.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_WIN)
+#include "base/allocator/partition_allocator/partition_alloc_base/win/windows_types.h"
+#endif
+
+namespace partition_alloc {
+
+// Terminates process. Should be called only for out of memory errors.
+// |size| is the size of the failed allocation, or 0 if not known.
+// Crash reporting classifies such crashes as OOM.
+// Must be allocation-safe.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void TerminateBecauseOutOfMemory(size_t size);
+
+// Records the size of the allocation that caused the current OOM crash, for
+// consumption by Breakpad.
+// TODO: this can be removed when Breakpad is no longer supported.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) extern size_t g_oom_size;
+
+#if BUILDFLAG(IS_WIN)
+namespace win {
+
+// Custom Windows exception code chosen to indicate an out of memory error.
+// See https://msdn.microsoft.com/en-us/library/het71c37.aspx.
+// "To make sure that you do not define a code that conflicts with an existing
+// exception code" ... "The resulting error code should therefore have the
+// highest four bits set to hexadecimal E."
+// 0xe0000008 was chosen arbitrarily, as 0x00000008 is ERROR_NOT_ENOUGH_MEMORY.
+const DWORD kOomExceptionCode = 0xe0000008;
+
+}  // namespace win
+#endif
+
+namespace internal {
+
+// The crash is generated in a PA_NOINLINE function so that we can classify the
+// crash as an OOM solely by analyzing the stack trace. It is tagged as
+// PA_NOT_TAIL_CALLED to ensure that its parent function stays on the stack.
+[[noreturn]] PA_NOT_TAIL_CALLED PA_COMPONENT_EXPORT(
+    PARTITION_ALLOC) void OnNoMemory(size_t size);
+
+// OOM_CRASH(size) - Specialization of IMMEDIATE_CRASH which will raise a custom
+// exception on Windows to signal this is OOM and not a normal assert.
+// OOM_CRASH(size) is called by users of PageAllocator (including
+// PartitionAlloc) to signify an allocation failure from the platform.
+#define OOM_CRASH(size)                                     \
+  do {                                                      \
+    /* Raising an exception might allocate, allow that.  */ \
+    ::partition_alloc::ScopedAllowAllocations guard{};      \
+    ::partition_alloc::internal::OnNoMemory(size);          \
+  } while (0)
+
+}  // namespace internal
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_OOM_H_

src/base/allocator/partition_allocator/oom_callback.cc

@@ -0,0 +1,28 @@
+// Copyright 2018 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/oom_callback.h"
+
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+
+namespace partition_alloc {
+
+namespace {
+PartitionAllocOomCallback g_oom_callback;
+}  // namespace
+
+void SetPartitionAllocOomCallback(PartitionAllocOomCallback callback) {
+  PA_DCHECK(!g_oom_callback);
+  g_oom_callback = callback;
+}
+
+namespace internal {
+void RunPartitionAllocOomCallback() {
+  if (g_oom_callback) {
+    g_oom_callback();
+  }
+}
+}  // namespace internal
+
+}  // namespace partition_alloc

src/base/allocator/partition_allocator/oom_callback.h

@@ -0,0 +1,26 @@
+// Copyright 2018 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_OOM_CALLBACK_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_OOM_CALLBACK_H_
+
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+
+namespace partition_alloc {
+
+using PartitionAllocOomCallback = void (*)();
+
+// Registers a callback to be invoked during an OOM_CRASH(). OOM_CRASH is
+// invoked by users of PageAllocator (including PartitionAlloc) to signify an
+// allocation failure from the platform.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void SetPartitionAllocOomCallback(PartitionAllocOomCallback callback);
+
+namespace internal {
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) void RunPartitionAllocOomCallback();
+}  // namespace internal
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_OOM_CALLBACK_H_

src/base/allocator/partition_allocator/page_allocator.cc

@@ -0,0 +1,418 @@
+// Copyright 2013 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/page_allocator.h"
+
+#include <atomic>
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/address_space_randomization.h"
+#include "base/allocator/partition_allocator/page_allocator_internal.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/bits.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/thread_annotations.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/partition_lock.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_WIN)
+#include <windows.h>
+#endif
+
+#if BUILDFLAG(IS_WIN)
+#include "base/allocator/partition_allocator/page_allocator_internals_win.h"
+#elif BUILDFLAG(IS_POSIX)
+#include "base/allocator/partition_allocator/page_allocator_internals_posix.h"
+#elif BUILDFLAG(IS_FUCHSIA)
+#include "base/allocator/partition_allocator/page_allocator_internals_fuchsia.h"
+#else
+#error Platform not supported.
+#endif
+
+namespace partition_alloc {
+
+namespace {
+
+internal::Lock g_reserve_lock;
+
+// We may reserve/release address space on different threads.
+internal::Lock& GetReserveLock() {
+  return g_reserve_lock;
+}
+
+std::atomic<size_t> g_total_mapped_address_space;
+
+// We only support a single block of reserved address space.
+uintptr_t s_reservation_address PA_GUARDED_BY(GetReserveLock()) = 0;
+size_t s_reservation_size PA_GUARDED_BY(GetReserveLock()) = 0;
+
+uintptr_t AllocPagesIncludingReserved(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageTag page_tag,
+    int file_descriptor_for_shared_alloc = -1) {
+  uintptr_t ret =
+      internal::SystemAllocPages(address, length, accessibility, page_tag,
+                                 file_descriptor_for_shared_alloc);
+  if (!ret) {
+    const bool cant_alloc_length = internal::kHintIsAdvisory || !address;
+    if (cant_alloc_length) {
+      // The system cannot allocate |length| bytes. Release any reserved address
+      // space and try once more.
+      ReleaseReservation();
+      ret = internal::SystemAllocPages(address, length, accessibility, page_tag,
+                                       file_descriptor_for_shared_alloc);
+    }
+  }
+  return ret;
+}
+
+// Trims memory at |base_address| to given |trim_length| and |alignment|.
+//
+// On failure, on Windows, this function returns 0 and frees memory at
+// |base_address|.
+uintptr_t TrimMapping(uintptr_t base_address,
+                      size_t base_length,
+                      size_t trim_length,
+                      uintptr_t alignment,
+                      uintptr_t alignment_offset,
+                      PageAccessibilityConfiguration accessibility) {
+  PA_DCHECK(base_length >= trim_length);
+  PA_DCHECK(internal::base::bits::IsPowerOfTwo(alignment));
+  PA_DCHECK(alignment_offset < alignment);
+  uintptr_t new_base =
+      NextAlignedWithOffset(base_address, alignment, alignment_offset);
+  PA_DCHECK(new_base >= base_address);
+  size_t pre_slack = new_base - base_address;
+  size_t post_slack = base_length - pre_slack - trim_length;
+  PA_DCHECK(base_length == trim_length || pre_slack || post_slack);
+  PA_DCHECK(pre_slack < base_length);
+  PA_DCHECK(post_slack < base_length);
+  return internal::TrimMappingInternal(base_address, base_length, trim_length,
+                                       accessibility, pre_slack, post_slack);
+}
+
+}  // namespace
+
+// Align |address| up to the closest, non-smaller address, that gives
+// |requested_offset| remainder modulo |alignment|.
+//
+// Examples for alignment=1024 and requested_offset=64:
+//   64 -> 64
+//   65 -> 1088
+//   1024 -> 1088
+//   1088 -> 1088
+//   1089 -> 2112
+//   2048 -> 2112
+uintptr_t NextAlignedWithOffset(uintptr_t address,
+                                uintptr_t alignment,
+                                uintptr_t requested_offset) {
+  PA_DCHECK(internal::base::bits::IsPowerOfTwo(alignment));
+  PA_DCHECK(requested_offset < alignment);
+
+  uintptr_t actual_offset = address & (alignment - 1);
+  uintptr_t new_address;
+  if (actual_offset <= requested_offset) {
+    new_address = address + requested_offset - actual_offset;
+  } else {
+    new_address = address + alignment + requested_offset - actual_offset;
+  }
+  PA_DCHECK(new_address >= address);
+  PA_DCHECK(new_address - address < alignment);
+  PA_DCHECK(new_address % alignment == requested_offset);
+
+  return new_address;
+}
+
+namespace internal {
+
+uintptr_t SystemAllocPages(uintptr_t hint,
+                           size_t length,
+                           PageAccessibilityConfiguration accessibility,
+                           PageTag page_tag,
+                           int file_descriptor_for_shared_alloc) {
+  PA_DCHECK(!(length & internal::PageAllocationGranularityOffsetMask()));
+  PA_DCHECK(!(hint & internal::PageAllocationGranularityOffsetMask()));
+  uintptr_t ret = internal::SystemAllocPagesInternal(
+      hint, length, accessibility, page_tag, file_descriptor_for_shared_alloc);
+  if (ret) {
+    g_total_mapped_address_space.fetch_add(length, std::memory_order_relaxed);
+  }
+
+  return ret;
+}
+
+}  // namespace internal
+
+uintptr_t AllocPages(size_t length,
+                     size_t align,
+                     PageAccessibilityConfiguration accessibility,
+                     PageTag page_tag,
+                     int file_descriptor_for_shared_alloc) {
+  return AllocPagesWithAlignOffset(0, length, align, 0, accessibility, page_tag,
+                                   file_descriptor_for_shared_alloc);
+}
+uintptr_t AllocPages(uintptr_t address,
+                     size_t length,
+                     size_t align,
+                     PageAccessibilityConfiguration accessibility,
+                     PageTag page_tag) {
+  return AllocPagesWithAlignOffset(address, length, align, 0, accessibility,
+                                   page_tag);
+}
+void* AllocPages(void* address,
+                 size_t length,
+                 size_t align,
+                 PageAccessibilityConfiguration accessibility,
+                 PageTag page_tag) {
+  return reinterpret_cast<void*>(
+      AllocPages(reinterpret_cast<uintptr_t>(address), length, align,
+                 accessibility, page_tag));
+}
+
+uintptr_t AllocPagesWithAlignOffset(
+    uintptr_t address,
+    size_t length,
+    size_t align,
+    size_t align_offset,
+    PageAccessibilityConfiguration accessibility,
+    PageTag page_tag,
+    int file_descriptor_for_shared_alloc) {
+  PA_DCHECK(length >= internal::PageAllocationGranularity());
+  PA_DCHECK(!(length & internal::PageAllocationGranularityOffsetMask()));
+  PA_DCHECK(align >= internal::PageAllocationGranularity());
+  // Alignment must be power of 2 for masking math to work.
+  PA_DCHECK(internal::base::bits::IsPowerOfTwo(align));
+  PA_DCHECK(align_offset < align);
+  PA_DCHECK(!(align_offset & internal::PageAllocationGranularityOffsetMask()));
+  PA_DCHECK(!(address & internal::PageAllocationGranularityOffsetMask()));
+  uintptr_t align_offset_mask = align - 1;
+  uintptr_t align_base_mask = ~align_offset_mask;
+  PA_DCHECK(!address || (address & align_offset_mask) == align_offset);
+
+  // If the client passed null as the address, choose a good one.
+  if (!address) {
+    address = (GetRandomPageBase() & align_base_mask) + align_offset;
+  }
+
+  // First try to force an exact-size, aligned allocation from our random base.
+#if defined(ARCH_CPU_32_BITS)
+  // On 32 bit systems, first try one random aligned address, and then try an
+  // aligned address derived from the value of |ret|.
+  constexpr int kExactSizeTries = 2;
+#else
+  // On 64 bit systems, try 3 random aligned addresses.
+  constexpr int kExactSizeTries = 3;
+#endif
+
+  for (int i = 0; i < kExactSizeTries; ++i) {
+    uintptr_t ret =
+        AllocPagesIncludingReserved(address, length, accessibility, page_tag,
+                                    file_descriptor_for_shared_alloc);
+    if (ret) {
+      // If the alignment is to our liking, we're done.
+      if ((ret & align_offset_mask) == align_offset) {
+        return ret;
+      }
+      // Free the memory and try again.
+      FreePages(ret, length);
+    } else {
+      // |ret| is null; if this try was unhinted, we're OOM.
+      if (internal::kHintIsAdvisory || !address) {
+        return 0;
+      }
+    }
+
+#if defined(ARCH_CPU_32_BITS)
+    // For small address spaces, try the first aligned address >= |ret|. Note
+    // |ret| may be null, in which case |address| becomes null. If
+    // |align_offset| is non-zero, this calculation may get us not the first,
+    // but the next matching address.
+    address = ((ret + align_offset_mask) & align_base_mask) + align_offset;
+#else  // defined(ARCH_CPU_64_BITS)
+    // Keep trying random addresses on systems that have a large address space.
+    address = NextAlignedWithOffset(GetRandomPageBase(), align, align_offset);
+#endif
+  }
+
+  // Make a larger allocation so we can force alignment.
+  size_t try_length = length + (align - internal::PageAllocationGranularity());
+  PA_CHECK(try_length >= length);
+  uintptr_t ret;
+
+  do {
+    // Continue randomizing only on POSIX.
+    address = internal::kHintIsAdvisory ? GetRandomPageBase() : 0;
+    ret =
+        AllocPagesIncludingReserved(address, try_length, accessibility,
+                                    page_tag, file_descriptor_for_shared_alloc);
+    // The retries are for Windows, where a race can steal our mapping on
+    // resize.
+  } while (ret && (ret = TrimMapping(ret, try_length, length, align,
+                                     align_offset, accessibility)) == 0);
+
+  return ret;
+}
+
+void FreePages(uintptr_t address, size_t length) {
+  PA_DCHECK(!(address & internal::PageAllocationGranularityOffsetMask()));
+  PA_DCHECK(!(length & internal::PageAllocationGranularityOffsetMask()));
+  internal::FreePagesInternal(address, length);
+  PA_DCHECK(g_total_mapped_address_space.load(std::memory_order_relaxed) > 0);
+  g_total_mapped_address_space.fetch_sub(length, std::memory_order_relaxed);
+}
+void FreePages(void* address, size_t length) {
+  FreePages(reinterpret_cast<uintptr_t>(address), length);
+}
+
+bool TrySetSystemPagesAccess(uintptr_t address,
+                             size_t length,
+                             PageAccessibilityConfiguration accessibility) {
+  PA_DCHECK(!(length & internal::SystemPageOffsetMask()));
+  return internal::TrySetSystemPagesAccessInternal(address, length,
+                                                   accessibility);
+}
+bool TrySetSystemPagesAccess(void* address,
+                             size_t length,
+                             PageAccessibilityConfiguration accessibility) {
+  return TrySetSystemPagesAccess(reinterpret_cast<uintptr_t>(address), length,
+                                 accessibility);
+}
+
+void SetSystemPagesAccess(uintptr_t address,
+                          size_t length,
+                          PageAccessibilityConfiguration accessibility) {
+  PA_DCHECK(!(length & internal::SystemPageOffsetMask()));
+  internal::SetSystemPagesAccessInternal(address, length, accessibility);
+}
+
+void DecommitSystemPages(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityDisposition accessibility_disposition) {
+  PA_DCHECK(!(address & internal::SystemPageOffsetMask()));
+  PA_DCHECK(!(length & internal::SystemPageOffsetMask()));
+  internal::DecommitSystemPagesInternal(address, length,
+                                        accessibility_disposition);
+}
+void DecommitSystemPages(
+    void* address,
+    size_t length,
+    PageAccessibilityDisposition accessibility_disposition) {
+  DecommitSystemPages(reinterpret_cast<uintptr_t>(address), length,
+                      accessibility_disposition);
+}
+
+void DecommitAndZeroSystemPages(uintptr_t address,
+                                size_t length,
+                                PageTag page_tag) {
+  PA_DCHECK(!(address & internal::SystemPageOffsetMask()));
+  PA_DCHECK(!(length & internal::SystemPageOffsetMask()));
+  internal::DecommitAndZeroSystemPagesInternal(address, length, page_tag);
+}
+
+void DecommitAndZeroSystemPages(void* address,
+                                size_t length,
+                                PageTag page_tag) {
+  DecommitAndZeroSystemPages(reinterpret_cast<uintptr_t>(address), length,
+                             page_tag);
+}
+
+void RecommitSystemPages(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  PA_DCHECK(!(address & internal::SystemPageOffsetMask()));
+  PA_DCHECK(!(length & internal::SystemPageOffsetMask()));
+  PA_DCHECK(accessibility.permissions !=
+            PageAccessibilityConfiguration::kInaccessible);
+  internal::RecommitSystemPagesInternal(address, length, accessibility,
+                                        accessibility_disposition);
+}
+
+bool TryRecommitSystemPages(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // Duplicated because we want errors to be reported at a lower level in the
+  // crashing case.
+  PA_DCHECK(!(address & internal::SystemPageOffsetMask()));
+  PA_DCHECK(!(length & internal::SystemPageOffsetMask()));
+  PA_DCHECK(accessibility.permissions !=
+            PageAccessibilityConfiguration::kInaccessible);
+  return internal::TryRecommitSystemPagesInternal(
+      address, length, accessibility, accessibility_disposition);
+}
+
+void DiscardSystemPages(uintptr_t address, size_t length) {
+  PA_DCHECK(!(length & internal::SystemPageOffsetMask()));
+  internal::DiscardSystemPagesInternal(address, length);
+}
+void DiscardSystemPages(void* address, size_t length) {
+  DiscardSystemPages(reinterpret_cast<uintptr_t>(address), length);
+}
+
+bool ReserveAddressSpace(size_t size) {
+  // To avoid deadlock, call only SystemAllocPages.
+  internal::ScopedGuard guard(GetReserveLock());
+  if (!s_reservation_address) {
+    uintptr_t mem = internal::SystemAllocPages(
+        0, size,
+        PageAccessibilityConfiguration(
+            PageAccessibilityConfiguration::kInaccessible),
+        PageTag::kChromium);
+    if (mem) {
+      // We guarantee this alignment when reserving address space.
+      PA_DCHECK(!(mem & internal::PageAllocationGranularityOffsetMask()));
+      s_reservation_address = mem;
+      s_reservation_size = size;
+      return true;
+    }
+  }
+  return false;
+}
+
+bool ReleaseReservation() {
+  // To avoid deadlock, call only FreePages.
+  internal::ScopedGuard guard(GetReserveLock());
+  if (!s_reservation_address) {
+    return false;
+  }
+
+  FreePages(s_reservation_address, s_reservation_size);
+  s_reservation_address = 0;
+  s_reservation_size = 0;
+  return true;
+}
+
+bool HasReservationForTesting() {
+  internal::ScopedGuard guard(GetReserveLock());
+  return s_reservation_address;
+}
+
+uint32_t GetAllocPageErrorCode() {
+  return internal::s_allocPageErrorCode;
+}
+
+size_t GetTotalMappedSize() {
+  return g_total_mapped_address_space;
+}
+
+#if BUILDFLAG(IS_WIN)
+namespace {
+bool g_retry_on_commit_failure = false;
+}
+
+void SetRetryOnCommitFailure(bool retry_on_commit_failure) {
+  g_retry_on_commit_failure = retry_on_commit_failure;
+}
+
+bool GetRetryOnCommitFailure() {
+  return g_retry_on_commit_failure;
+}
+#endif
+
+}  // namespace partition_alloc

src/base/allocator/partition_allocator/page_allocator.h

@@ -0,0 +1,393 @@
+// Copyright 2013 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_H_
+
+#include <cstddef>
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/page_allocator_constants.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/thread_isolation/thread_isolation.h"
+#include "build/build_config.h"
+
+namespace partition_alloc {
+
+struct PageAccessibilityConfiguration {
+  enum Permissions {
+    kInaccessible,
+    // This flag is valid only with AllocPages(), where in creates kInaccessible
+    // pages that may later be re-mapped as executable, on platforms which
+    // distinguish never-executable and maybe-executable pages.
+    kInaccessibleWillJitLater,
+    kRead,
+    kReadWrite,
+    // This flag is mapped to kReadWrite on systems that
+    // don't support MTE.
+    kReadWriteTagged,
+    // This flag is mapped to kReadExecute on systems
+    // that don't support Arm's BTI.
+    kReadExecuteProtected,
+    kReadExecute,
+    // This flag is deprecated and will go away soon.
+    // TODO(bbudge) Remove this as soon as V8 doesn't need RWX pages.
+    kReadWriteExecute,
+  };
+
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  constexpr explicit PageAccessibilityConfiguration(Permissions permissions)
+      : permissions(permissions) {}
+  constexpr PageAccessibilityConfiguration(
+      Permissions permissions,
+      ThreadIsolationOption thread_isolation)
+      : permissions(permissions), thread_isolation(thread_isolation) {}
+#else
+  constexpr explicit PageAccessibilityConfiguration(Permissions permissions)
+      : permissions(permissions) {}
+#endif  // BUILDFLAG(ENABLE_THREAD_ISOLATION)
+
+  Permissions permissions;
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  // Tag the page with a Memory Protection Key. Use 0 for none.
+  ThreadIsolationOption thread_isolation;
+#endif  // BUILDFLAG(ENABLE_THREAD_ISOLATION)
+};
+
+// Use for De/RecommitSystemPages API.
+enum class PageAccessibilityDisposition {
+  // Enforces permission update (Decommit will set to
+  // PageAccessibilityConfiguration::kInaccessible;
+  // Recommit will set to whatever was requested, other than
+  // PageAccessibilityConfiguration::kInaccessible).
+  kRequireUpdate,
+  // Will not update permissions, if the platform supports that (POSIX & Fuchsia
+  // only).
+  kAllowKeepForPerf,
+};
+
+// Some platforms (including macOS and some Linux-based ones) support tagged
+// memory regions, to help in debugging. On Android, these tags are used to name
+// anonymous mappings.
+//
+// kChromium is the default value, used to distinguish general
+// Chromium-originated allocations from other ones (e.g. from platform
+// libraries).
+enum class PageTag {
+  kSimulation = 251,      // Memory simulator tool.
+  kBlinkGC = 252,         // Blink GC pages.
+  kPartitionAlloc = 253,  // PartitionAlloc, no matter the partition.
+  kChromium = 254,        // Chromium page.
+  kV8 = 255,              // V8 heap pages.
+
+  kFirst = kSimulation,  // Minimum tag value.
+  kLast = kV8            // Maximum tag value.
+};
+
+// See
+// https://github.com/apple-oss-distributions/xnu/blob/5c2921b07a2480ab43ec66f5b9e41cb872bc554f/osfmk/mach/vm_statistics.h#L687
+static_assert(static_cast<int>(PageTag::kLast) >= 240,
+              "The first application-reserved tag on macOS is 240, see "
+              "vm_statistics.h in XNU.");
+static_assert(
+    static_cast<int>(PageTag::kLast) < 256,
+    "Tags are only 1 byte long on macOS, see vm_statistics.h in XNU.");
+
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+uintptr_t NextAlignedWithOffset(uintptr_t ptr,
+                                uintptr_t alignment,
+                                uintptr_t requested_offset);
+
+// Allocates one or more pages.
+//
+// The requested |address| is just a hint; the actual address returned may
+// differ. The returned address will be aligned to |align_offset| modulo |align|
+// bytes.
+//
+// |length|, |align| and |align_offset| are in bytes, and must be a multiple of
+// |PageAllocationGranularity()|. |length| and |align| must be non-zero.
+// |align_offset| must be less than |align|. |align| must be a power of two.
+//
+// If |address| is 0/nullptr, then a suitable and randomized address will be
+// chosen automatically.
+//
+// |accessibility| controls the permission of the allocated pages.
+// PageAccessibilityConfiguration::kInaccessible means uncommitted.
+//
+// |page_tag| is used on some platforms to identify the source of the
+// allocation.
+//
+// |file_descriptor_for_shared_alloc| is only used in mapping the shadow
+// pools to the same physical address as the real one in
+// PartitionAddressSpace::Init(). It should be ignored in other cases.
+//
+// This call will return 0/nullptr if the allocation cannot be satisfied.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+uintptr_t AllocPages(size_t length,
+                     size_t align,
+                     PageAccessibilityConfiguration accessibility,
+                     PageTag page_tag = PageTag::kChromium,
+                     int file_descriptor_for_shared_alloc = -1);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+uintptr_t AllocPages(uintptr_t address,
+                     size_t length,
+                     size_t align,
+                     PageAccessibilityConfiguration accessibility,
+                     PageTag page_tag = PageTag::kChromium);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void* AllocPages(void* address,
+                 size_t length,
+                 size_t align,
+                 PageAccessibilityConfiguration accessibility,
+                 PageTag page_tag = PageTag::kChromium);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+uintptr_t AllocPagesWithAlignOffset(
+    uintptr_t address,
+    size_t length,
+    size_t align,
+    size_t align_offset,
+    PageAccessibilityConfiguration page_accessibility,
+    PageTag page_tag = PageTag::kChromium,
+    int file_descriptor_for_shared_alloc = -1);
+
+// Frees one or more pages starting at |address| and continuing for |length|
+// bytes.
+//
+// |address| and |length| must match a previous call to |AllocPages|. Therefore,
+// |address| must be aligned to |PageAllocationGranularity()| bytes, and
+// |length| must be a multiple of |PageAllocationGranularity()|.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void FreePages(uintptr_t address, size_t length);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void FreePages(void* address, size_t length);
+
+// Marks one or more system pages, starting at |address| with the given
+// |page_accessibility|. |length| must be a multiple of |SystemPageSize()|
+// bytes.
+//
+// Returns true if the permission change succeeded. In most cases you must
+// |CHECK| the result.
+[[nodiscard]] PA_COMPONENT_EXPORT(PARTITION_ALLOC) bool TrySetSystemPagesAccess(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration page_accessibility);
+[[nodiscard]] PA_COMPONENT_EXPORT(PARTITION_ALLOC) bool TrySetSystemPagesAccess(
+    void* address,
+    size_t length,
+    PageAccessibilityConfiguration page_accessibility);
+
+// Marks one or more system pages, starting at |address| with the given
+// |page_accessibility|. |length| must be a multiple of |SystemPageSize()|
+// bytes.
+//
+// Performs a CHECK that the operation succeeds.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void SetSystemPagesAccess(uintptr_t address,
+                          size_t length,
+                          PageAccessibilityConfiguration page_accessibility);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void SetSystemPagesAccess(void* address,
+                          size_t length,
+                          PageAccessibilityConfiguration page_accessibility);
+
+// Decommits one or more system pages starting at |address| and continuing for
+// |length| bytes. |address| and |length| must be aligned to a system page
+// boundary.
+//
+// This API will crash if the operation cannot be performed!
+//
+// If disposition is PageAccessibilityDisposition::kRequireUpdate (recommended),
+// the decommitted pages will be made inaccessible before the call returns.
+// While it is always a programming error to access decommitted pages without
+// first recommitting them, callers may use
+// PageAccessibilityDisposition::kAllowKeepForPerf to allow the implementation
+// to skip changing permissions (use with care), for performance reasons (see
+// crrev.com/c/2567282 and crrev.com/c/2563038 for perf regressions encountered
+// in the past). Implementations may choose to always modify permissions, hence
+// accessing those pages may or may not trigger a fault.
+//
+// Decommitting means that physical resources (RAM or swap/pagefile) backing the
+// allocated virtual address range may be released back to the system, but the
+// address space is still allocated to the process (possibly using up page table
+// entries or other accounting resources). There is no guarantee that the pages
+// are zeroed, unless |DecommittedMemoryIsAlwaysZeroed()| is true.
+//
+// This operation may not be atomic on some platforms.
+//
+// Note: "Committed memory" is a Windows Memory Subsystem concept that ensures
+// processes will not fault when touching a committed memory region. There is
+// no analogue in the POSIX & Fuchsia memory API where virtual memory pages are
+// best-effort allocated resources on the first touch. If
+// PageAccessibilityDisposition::kRequireUpdate disposition is used, this API
+// behaves in a platform-agnostic way by simulating the Windows "decommit" state
+// by both discarding the region (allowing the OS to avoid swap operations)
+// *and* changing the page protections so accesses fault.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void DecommitSystemPages(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityDisposition accessibility_disposition);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void DecommitSystemPages(
+    void* address,
+    size_t length,
+    PageAccessibilityDisposition accessibility_disposition);
+
+// Decommits one or more system pages starting at |address| and continuing for
+// |length| bytes. |address| and |length| must be aligned to a system page
+// boundary.
+//
+// In contrast to |DecommitSystemPages|, this API guarantees that the pages are
+// zeroed and will always mark the region as inaccessible (the equivalent of
+// setting them to PageAccessibilityConfiguration::kInaccessible).
+//
+// This API will crash if the operation cannot be performed.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void DecommitAndZeroSystemPages(uintptr_t address,
+                                size_t length,
+                                PageTag page_tag = PageTag::kChromium);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void DecommitAndZeroSystemPages(void* address,
+                                size_t length,
+                                PageTag page_tag = PageTag::kChromium);
+
+// Whether decommitted memory is guaranteed to be zeroed when it is
+// recommitted. Do not assume that this will not change over time.
+constexpr PA_COMPONENT_EXPORT(
+    PARTITION_ALLOC) bool DecommittedMemoryIsAlwaysZeroed() {
+#if BUILDFLAG(IS_APPLE)
+  return false;
+#else
+  return true;
+#endif
+}
+
+// (Re)Commits one or more system pages, starting at |address| and continuing
+// for |length| bytes with the given |page_accessibility| (must not be
+// PageAccessibilityConfiguration::kInaccessible). |address| and |length|
+// must be aligned to a system page boundary.
+//
+// This API will crash if the operation cannot be performed!
+//
+// If disposition is PageAccessibilityConfiguration::kRequireUpdate, the calls
+// updates the pages to |page_accessibility|. This can be used regardless of
+// what disposition was used to decommit the pages.
+// PageAccessibilityConfiguration::kAllowKeepForPerf allows the implementation
+// to leave the page permissions, if that improves performance. This option can
+// only be used if the pages were previously accessible and decommitted with
+// that same option.
+//
+// The memory will be zeroed when it is committed for the first time. However,
+// there is no such guarantee when memory is recommitted, unless
+// |DecommittedMemoryIsAlwaysZeroed()| is true.
+//
+// This operation may not be atomic on some platforms.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void RecommitSystemPages(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration page_accessibility,
+    PageAccessibilityDisposition accessibility_disposition);
+
+// Like RecommitSystemPages(), but returns false instead of crashing.
+[[nodiscard]] PA_COMPONENT_EXPORT(PARTITION_ALLOC) bool TryRecommitSystemPages(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration page_accessibility,
+    PageAccessibilityDisposition accessibility_disposition);
+
+// Discard one or more system pages starting at |address| and continuing for
+// |length| bytes. |length| must be a multiple of |SystemPageSize()|.
+//
+// Discarding is a hint to the system that the page is no longer required. The
+// hint may:
+//   - Do nothing.
+//   - Discard the page immediately, freeing up physical pages.
+//   - Discard the page at some time in the future in response to memory
+//   pressure.
+//
+// Only committed pages should be discarded. Discarding a page does not decommit
+// it, and it is valid to discard an already-discarded page. A read or write to
+// a discarded page will not fault.
+//
+// Reading from a discarded page may return the original page content, or a page
+// full of zeroes.
+//
+// Writing to a discarded page is the only guaranteed way to tell the system
+// that the page is required again. Once written to, the content of the page is
+// guaranteed stable once more. After being written to, the page content may be
+// based on the original page content, or a page of zeroes.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void DiscardSystemPages(uintptr_t address, size_t length);
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void DiscardSystemPages(void* address, size_t length);
+
+// Rounds up |address| to the next multiple of |SystemPageSize()|. Returns
+// 0 for an |address| of 0.
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+RoundUpToSystemPage(uintptr_t address) {
+  return (address + internal::SystemPageOffsetMask()) &
+         internal::SystemPageBaseMask();
+}
+
+// Rounds down |address| to the previous multiple of |SystemPageSize()|. Returns
+// 0 for an |address| of 0.
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+RoundDownToSystemPage(uintptr_t address) {
+  return address & internal::SystemPageBaseMask();
+}
+
+// Rounds up |address| to the next multiple of |PageAllocationGranularity()|.
+// Returns 0 for an |address| of 0.
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+RoundUpToPageAllocationGranularity(uintptr_t address) {
+  return (address + internal::PageAllocationGranularityOffsetMask()) &
+         internal::PageAllocationGranularityBaseMask();
+}
+
+// Rounds down |address| to the previous multiple of
+// |PageAllocationGranularity()|. Returns 0 for an |address| of 0.
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR uintptr_t
+RoundDownToPageAllocationGranularity(uintptr_t address) {
+  return address & internal::PageAllocationGranularityBaseMask();
+}
+
+// Reserves (at least) |size| bytes of address space, aligned to
+// |PageAllocationGranularity()|. This can be called early on to make it more
+// likely that large allocations will succeed. Returns true if the reservation
+// succeeded, false if the reservation failed or a reservation was already made.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) bool ReserveAddressSpace(size_t size);
+
+// Releases any reserved address space. |AllocPages| calls this automatically on
+// an allocation failure. External allocators may also call this on failure.
+//
+// Returns true when an existing reservation was released.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) bool ReleaseReservation();
+
+// Returns true if there is currently an address space reservation.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) bool HasReservationForTesting();
+
+// Returns |errno| (POSIX) or the result of |GetLastError| (Windows) when |mmap|
+// (POSIX) or |VirtualAlloc| (Windows) fails.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) uint32_t GetAllocPageErrorCode();
+
+// Returns the total amount of mapped pages from all clients of
+// PageAllocator. These pages may or may not be committed. This is mostly useful
+// to assess address space pressure.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC) size_t GetTotalMappedSize();
+
+#if BUILDFLAG(IS_WIN)
+// Sets whether to retry the allocation of pages when a commit failure
+// happens. This doesn't cover cases where the system is out of address space,
+// or reaches another limit.
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+void SetRetryOnCommitFailure(bool retry_on_commit_failure);
+bool GetRetryOnCommitFailure();
+#endif  // BUILDFLAG(IS_WIN)
+
+}  // namespace partition_alloc
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_H_

src/base/allocator/partition_allocator/page_allocator_constants.h

@@ -0,0 +1,180 @@
+// Copyright 2018 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_CONSTANTS_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_CONSTANTS_H_
+
+#include <stddef.h>
+
+#include "base/allocator/partition_allocator/partition_alloc_base/compiler_specific.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/component_export.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_APPLE) && defined(ARCH_CPU_64_BITS)
+
+#include <mach/vm_page_size.h>
+
+// Although page allocator constants are not constexpr, they are run-time
+// constant. Because the underlying variables they access, such as vm_page_size,
+// are not marked const, the compiler normally has no way to know that they
+// don’t change and must obtain their values whenever it can't prove that they
+// haven't been modified, even if they had already been obtained previously.
+// Attaching __attribute__((const)) to these declarations allows these redundant
+// accesses to be omitted under optimization such as common subexpression
+// elimination.
+#define PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR __attribute__((const))
+
+#elif BUILDFLAG(IS_LINUX) && defined(ARCH_CPU_ARM64)
+// This should work for all POSIX (if needed), but currently all other
+// supported OS/architecture combinations use either hard-coded values
+// (such as x86) or have means to determine these values without needing
+// atomics (such as macOS on arm64).
+
+// Page allocator constants are run-time constant
+#define PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR __attribute__((const))
+
+#include <unistd.h>
+#include <atomic>
+
+namespace partition_alloc::internal {
+
+// Holds the current page size and shift, where size = 1 << shift
+// Use PageAllocationGranularity(), PageAllocationGranularityShift()
+// to initialize and retrieve these values safely.
+struct PageCharacteristics {
+  std::atomic<size_t> size;
+  std::atomic<size_t> shift;
+};
+PA_COMPONENT_EXPORT(PARTITION_ALLOC)
+extern PageCharacteristics page_characteristics;
+
+}  // namespace partition_alloc::internal
+
+#else
+
+// When defined, page size constants are fixed at compile time. When not
+// defined, they may vary at run time.
+#define PAGE_ALLOCATOR_CONSTANTS_ARE_CONSTEXPR 1
+
+// Use this macro to declare a function as constexpr or not based on whether
+// PAGE_ALLOCATOR_CONSTANTS_ARE_CONSTEXPR is defined.
+#define PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR constexpr
+
+#endif
+
+// Ability to name anonymous VMAs is available on some, but not all Linux-based
+// systems.
+#if BUILDFLAG(IS_ANDROID) || BUILDFLAG(IS_LINUX)
+#include <sys/prctl.h>
+
+#if defined(PR_SET_VMA) && defined(PR_SET_VMA_ANON_NAME)
+#define LINUX_NAME_REGION 1
+#endif
+
+#endif  // BUILDFLAG(IS_ANDROID) || BUILDFLAG(IS_LINUX)
+
+namespace partition_alloc::internal {
+
+// Forward declaration, implementation below
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+PageAllocationGranularity();
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+PageAllocationGranularityShift() {
+#if BUILDFLAG(IS_WIN) || defined(ARCH_CPU_PPC64)
+  // Modern ppc64 systems support 4kB (shift = 12) and 64kB (shift = 16) page
+  // sizes.  Since 64kB is the de facto standard on the platform and binaries
+  // compiled for 64kB are likely to work on 4kB systems, 64kB is a good choice
+  // here.
+  return 16;  // 64kB
+#elif defined(_MIPS_ARCH_LOONGSON) || defined(ARCH_CPU_LOONGARCH64)
+  return 14;  // 16kB
+#elif BUILDFLAG(IS_APPLE) && defined(ARCH_CPU_64_BITS)
+  return static_cast<size_t>(vm_page_shift);
+#elif BUILDFLAG(IS_LINUX) && defined(ARCH_CPU_ARM64)
+  // arm64 supports 4kb (shift = 12), 16kb (shift = 14), and 64kb (shift = 16)
+  // page sizes. Retrieve from or initialize cache.
+  size_t shift = page_characteristics.shift.load(std::memory_order_relaxed);
+  if (PA_UNLIKELY(shift == 0)) {
+    shift = static_cast<size_t>(
+        __builtin_ctz((unsigned int)PageAllocationGranularity()));
+    page_characteristics.shift.store(shift, std::memory_order_relaxed);
+  }
+  return shift;
+#else
+  return 12;  // 4kB
+#endif
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+PageAllocationGranularity() {
+#if BUILDFLAG(IS_APPLE) && defined(ARCH_CPU_64_BITS)
+  // This is literally equivalent to |1 << PageAllocationGranularityShift()|
+  // below, but was separated out for IS_APPLE to avoid << on a non-constexpr.
+  return vm_page_size;
+#elif BUILDFLAG(IS_LINUX) && defined(ARCH_CPU_ARM64)
+  // arm64 supports 4kb, 16kb, and 64kb page sizes. Retrieve from or
+  // initialize cache.
+  size_t size = page_characteristics.size.load(std::memory_order_relaxed);
+  if (PA_UNLIKELY(size == 0)) {
+    size = static_cast<size_t>(getpagesize());
+    page_characteristics.size.store(size, std::memory_order_relaxed);
+  }
+  return size;
+#else
+  return 1 << PageAllocationGranularityShift();
+#endif
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+PageAllocationGranularityOffsetMask() {
+  return PageAllocationGranularity() - 1;
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+PageAllocationGranularityBaseMask() {
+  return ~PageAllocationGranularityOffsetMask();
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+SystemPageShift() {
+  // On Windows allocation granularity is higher than the page size. This comes
+  // into play when reserving address space range (allocation granularity),
+  // compared to committing pages into memory (system page granularity).
+#if BUILDFLAG(IS_WIN)
+  return 12;  // 4096=1<<12
+#else
+  return PageAllocationGranularityShift();
+#endif
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+SystemPageSize() {
+#if (BUILDFLAG(IS_APPLE) && defined(ARCH_CPU_64_BITS)) || \
+    (BUILDFLAG(IS_LINUX) && defined(ARCH_CPU_ARM64))
+  // This is literally equivalent to |1 << SystemPageShift()| below, but was
+  // separated out for 64-bit IS_APPLE and arm64 on Linux to avoid << on a
+  // non-constexpr.
+  return PageAllocationGranularity();
+#else
+  return 1 << SystemPageShift();
+#endif
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+SystemPageOffsetMask() {
+  return SystemPageSize() - 1;
+}
+
+PA_ALWAYS_INLINE PAGE_ALLOCATOR_CONSTANTS_DECLARE_CONSTEXPR size_t
+SystemPageBaseMask() {
+  return ~SystemPageOffsetMask();
+}
+
+constexpr size_t kPageMetadataShift = 5;  // 32 bytes per partition page.
+constexpr size_t kPageMetadataSize = 1 << kPageMetadataShift;
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_CONSTANTS_H_

src/base/allocator/partition_allocator/page_allocator_internal.h

@@ -0,0 +1,23 @@
+// Copyright 2018 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNAL_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNAL_H_
+
+#include <cstddef>
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/page_allocator.h"
+
+namespace partition_alloc::internal {
+
+uintptr_t SystemAllocPages(uintptr_t hint,
+                           size_t length,
+                           PageAccessibilityConfiguration accessibility,
+                           PageTag page_tag,
+                           int file_descriptor_for_shared_alloc = -1);
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNAL_H_

src/base/allocator/partition_allocator/page_allocator_internals_fuchsia.h

@@ -0,0 +1,267 @@
+// Copyright 2019 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+//
+// This file implements memory allocation primitives for PageAllocator using
+// Fuchsia's VMOs (Virtual Memory Objects). VMO API is documented in
+// https://fuchsia.dev/fuchsia-src/zircon/objects/vm_object . A VMO is a kernel
+// object that corresponds to a set of memory pages. VMO pages may be mapped
+// to an address space. The code below creates VMOs for each memory allocations
+// and maps them to the default address space of the current process.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_FUCHSIA_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_FUCHSIA_H_
+
+#include <fidl/fuchsia.kernel/cpp/fidl.h>
+#include <lib/component/incoming/cpp/protocol.h>
+#include <lib/zx/resource.h>
+#include <lib/zx/vmar.h>
+#include <lib/zx/vmo.h>
+
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/page_allocator.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/fuchsia/fuchsia_logging.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/no_destructor.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/notreached.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+
+namespace partition_alloc::internal {
+
+namespace {
+
+zx::resource GetVmexResource() {
+  auto vmex_resource_client =
+      component::Connect<fuchsia_kernel::VmexResource>();
+  if (vmex_resource_client.is_error()) {
+    PA_LOG(ERROR) << "Connect(VmexResource):"
+                  << vmex_resource_client.status_string();
+    return {};
+  }
+
+  fidl::SyncClient sync_vmex_resource_client(
+      std::move(vmex_resource_client.value()));
+  auto result = sync_vmex_resource_client->Get();
+  if (result.is_error()) {
+    PA_LOG(ERROR) << "VmexResource.Get():"
+                  << result.error_value().FormatDescription().c_str();
+    return {};
+  }
+
+  return std::move(result->resource());
+}
+
+const zx::resource& VmexResource() {
+  static base::NoDestructor<zx::resource> vmex_resource(GetVmexResource());
+  return *vmex_resource;
+}
+
+// Returns VMO name for a PageTag.
+const char* PageTagToName(PageTag tag) {
+  switch (tag) {
+    case PageTag::kBlinkGC:
+      return "cr_blink_gc";
+    case PageTag::kPartitionAlloc:
+      return "cr_partition_alloc";
+    case PageTag::kChromium:
+      return "cr_chromium";
+    case PageTag::kV8:
+      return "cr_v8";
+    case PageTag::kSimulation:
+      PA_NOTREACHED();
+  }
+  PA_NOTREACHED();
+}
+
+zx_vm_option_t PageAccessibilityToZxVmOptions(
+    PageAccessibilityConfiguration accessibility) {
+  switch (accessibility.permissions) {
+    case PageAccessibilityConfiguration::kRead:
+      return ZX_VM_PERM_READ;
+    case PageAccessibilityConfiguration::kReadWrite:
+    case PageAccessibilityConfiguration::kReadWriteTagged:
+      return ZX_VM_PERM_READ | ZX_VM_PERM_WRITE;
+    case PageAccessibilityConfiguration::kReadExecuteProtected:
+    case PageAccessibilityConfiguration::kReadExecute:
+      return ZX_VM_PERM_READ | ZX_VM_PERM_EXECUTE;
+    case PageAccessibilityConfiguration::kReadWriteExecute:
+      return ZX_VM_PERM_READ | ZX_VM_PERM_WRITE | ZX_VM_PERM_EXECUTE;
+    case PageAccessibilityConfiguration::kInaccessible:
+    case PageAccessibilityConfiguration::kInaccessibleWillJitLater:
+      return 0;
+  };
+  PA_NOTREACHED();
+}
+
+}  // namespace
+
+// zx_vmar_map() will fail if the VMO cannot be mapped at |vmar_offset|, i.e.
+// |hint| is not advisory.
+constexpr bool kHintIsAdvisory = false;
+
+std::atomic<int32_t> s_allocPageErrorCode{0};
+
+uintptr_t SystemAllocPagesInternal(
+    uintptr_t hint,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageTag page_tag,
+    [[maybe_unused]] int file_descriptor_for_shared_alloc) {
+  zx::vmo vmo;
+  zx_status_t status = zx::vmo::create(length, 0, &vmo);
+  if (status != ZX_OK) {
+    PA_ZX_DLOG(INFO, status) << "zx_vmo_create";
+    return 0;
+  }
+
+  const char* vmo_name = PageTagToName(page_tag);
+  status = vmo.set_property(ZX_PROP_NAME, vmo_name, strlen(vmo_name));
+
+  // VMO names are used only for debugging, so failure to set a name is not
+  // fatal.
+  PA_ZX_DCHECK(status == ZX_OK, status);
+
+  if (accessibility.permissions ==
+          PageAccessibilityConfiguration::kInaccessibleWillJitLater ||
+      accessibility.permissions ==
+          PageAccessibilityConfiguration::kReadWriteExecute) {
+    // V8 uses JIT. Call zx_vmo_replace_as_executable() to allow code execution
+    // in the new VMO.
+    status = vmo.replace_as_executable(VmexResource(), &vmo);
+    if (status != ZX_OK) {
+      PA_ZX_DLOG(INFO, status) << "zx_vmo_replace_as_executable";
+      return 0;
+    }
+  }
+
+  zx_vm_option_t options = PageAccessibilityToZxVmOptions(accessibility);
+
+  uint64_t vmar_offset = 0;
+  if (hint) {
+    vmar_offset = hint;
+    options |= ZX_VM_SPECIFIC;
+  }
+
+  uint64_t address;
+  status = zx::vmar::root_self()->map(options, vmar_offset, vmo,
+                                      /*vmo_offset=*/0, length, &address);
+  if (status != ZX_OK) {
+    // map() is expected to fail if |hint| is set to an already-in-use location.
+    if (!hint) {
+      PA_ZX_DLOG(ERROR, status) << "zx_vmar_map";
+    }
+    return 0;
+  }
+
+  return address;
+}
+
+uintptr_t TrimMappingInternal(uintptr_t base_address,
+                              size_t base_length,
+                              size_t trim_length,
+                              PageAccessibilityConfiguration accessibility,
+                              size_t pre_slack,
+                              size_t post_slack) {
+  PA_DCHECK(base_length == trim_length + pre_slack + post_slack);
+
+  // Unmap head if necessary.
+  if (pre_slack) {
+    zx_status_t status = zx::vmar::root_self()->unmap(base_address, pre_slack);
+    PA_ZX_CHECK(status == ZX_OK, status);
+  }
+
+  // Unmap tail if necessary.
+  if (post_slack) {
+    zx_status_t status = zx::vmar::root_self()->unmap(
+        base_address + pre_slack + trim_length, post_slack);
+    PA_ZX_CHECK(status == ZX_OK, status);
+  }
+
+  return base_address + pre_slack;
+}
+
+bool TrySetSystemPagesAccessInternal(
+    uint64_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility) {
+  zx_status_t status = zx::vmar::root_self()->protect(
+      PageAccessibilityToZxVmOptions(accessibility), address, length);
+  return status == ZX_OK;
+}
+
+void SetSystemPagesAccessInternal(
+    uint64_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility) {
+  zx_status_t status = zx::vmar::root_self()->protect(
+      PageAccessibilityToZxVmOptions(accessibility), address, length);
+  PA_ZX_CHECK(status == ZX_OK, status);
+}
+
+void FreePagesInternal(uint64_t address, size_t length) {
+  zx_status_t status = zx::vmar::root_self()->unmap(address, length);
+  PA_ZX_CHECK(status == ZX_OK, status);
+}
+
+void DiscardSystemPagesInternal(uint64_t address, size_t length) {
+  zx_status_t status = zx::vmar::root_self()->op_range(
+      ZX_VMO_OP_DECOMMIT, address, length, nullptr, 0);
+  PA_ZX_CHECK(status == ZX_OK, status);
+}
+
+void DecommitSystemPagesInternal(
+    uint64_t address,
+    size_t length,
+    PageAccessibilityDisposition accessibility_disposition) {
+  if (accessibility_disposition ==
+      PageAccessibilityDisposition::kRequireUpdate) {
+    SetSystemPagesAccess(address, length,
+                         PageAccessibilityConfiguration(
+                             PageAccessibilityConfiguration::kInaccessible));
+  }
+
+  DiscardSystemPagesInternal(address, length);
+}
+
+void DecommitAndZeroSystemPagesInternal(uintptr_t address,
+                                        size_t length,
+                                        PageTag page_tag) {
+  SetSystemPagesAccess(address, length,
+                       PageAccessibilityConfiguration(
+                           PageAccessibilityConfiguration::kInaccessible));
+
+  DiscardSystemPagesInternal(address, length);
+}
+
+void RecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // On Fuchsia systems, the caller needs to simply read the memory to recommit
+  // it. However, if decommit changed the permissions, recommit has to change
+  // them back.
+  if (accessibility_disposition ==
+      PageAccessibilityDisposition::kRequireUpdate) {
+    SetSystemPagesAccess(address, length, accessibility);
+  }
+}
+
+bool TryRecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // On Fuchsia systems, the caller needs to simply read the memory to recommit
+  // it. However, if decommit changed the permissions, recommit has to change
+  // them back.
+  if (accessibility_disposition ==
+      PageAccessibilityDisposition::kRequireUpdate) {
+    return TrySetSystemPagesAccess(address, length, accessibility);
+  }
+  return true;
+}
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_FUCHSIA_H_

src/base/allocator/partition_allocator/page_allocator_internals_posix.cc

@@ -0,0 +1,44 @@
+// Copyright 2021 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/allocator/partition_allocator/page_allocator.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/cpu.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/notreached.h"
+
+#include <sys/mman.h>
+
+// PA_PROT_BTI requests a page that supports BTI landing pads.
+#define PA_PROT_BTI 0x10
+// PA_PROT_MTE requests a page that's suitable for memory tagging.
+#define PA_PROT_MTE 0x20
+
+namespace partition_alloc::internal {
+
+int GetAccessFlags(PageAccessibilityConfiguration accessibility) {
+  switch (accessibility.permissions) {
+    case PageAccessibilityConfiguration::kRead:
+      return PROT_READ;
+    case PageAccessibilityConfiguration::kReadWriteTagged:
+#if defined(ARCH_CPU_ARM64)
+      return PROT_READ | PROT_WRITE |
+             (base::CPU::GetInstanceNoAllocation().has_mte() ? PA_PROT_MTE : 0);
+#else
+      [[fallthrough]];
+#endif
+    case PageAccessibilityConfiguration::kReadWrite:
+      return PROT_READ | PROT_WRITE;
+    case PageAccessibilityConfiguration::kReadExecuteProtected:
+      return PROT_READ | PROT_EXEC |
+             (base::CPU::GetInstanceNoAllocation().has_bti() ? PA_PROT_BTI : 0);
+    case PageAccessibilityConfiguration::kReadExecute:
+      return PROT_READ | PROT_EXEC;
+    case PageAccessibilityConfiguration::kReadWriteExecute:
+      return PROT_READ | PROT_WRITE | PROT_EXEC;
+    case PageAccessibilityConfiguration::kInaccessible:
+    case PageAccessibilityConfiguration::kInaccessibleWillJitLater:
+      return PROT_NONE;
+  }
+}
+
+}  // namespace partition_alloc::internal

src/base/allocator/partition_allocator/page_allocator_internals_posix.h

@@ -0,0 +1,426 @@
+// Copyright 2018 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_POSIX_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_POSIX_H_
+
+#include <algorithm>
+#include <atomic>
+#include <cerrno>
+#include <cstdint>
+#include <cstring>
+
+#include <sys/mman.h>
+
+#include "base/allocator/partition_allocator/oom.h"
+#include "base/allocator/partition_allocator/page_allocator.h"
+#include "base/allocator/partition_allocator/page_allocator_constants.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/debug/debugging_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/notreached.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/posix/eintr_wrapper.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+#include "base/allocator/partition_allocator/thread_isolation/thread_isolation.h"
+#include "build/build_config.h"
+
+#if BUILDFLAG(IS_APPLE)
+#include "base/allocator/partition_allocator/partition_alloc_base/apple/foundation_util.h"
+#if BUILDFLAG(IS_IOS)
+#include "base/allocator/partition_allocator/partition_alloc_base/ios/ios_util.h"
+#elif BUILDFLAG(IS_MAC)
+#include "base/allocator/partition_allocator/partition_alloc_base/mac/mac_util.h"
+#else
+#error "Unknown platform"
+#endif
+#include "base/allocator/partition_allocator/partition_alloc_base/apple/scoped_cftyperef.h"
+
+#include <Availability.h>
+#include <Security/Security.h>
+#include <mach/mach.h>
+#endif
+#if BUILDFLAG(IS_ANDROID) || BUILDFLAG(IS_LINUX)
+#include <sys/prctl.h>
+#endif
+#if BUILDFLAG(IS_LINUX) || BUILDFLAG(IS_CHROMEOS)
+#include <sys/resource.h>
+#endif
+
+#ifndef MAP_ANONYMOUS
+#define MAP_ANONYMOUS MAP_ANON
+#endif
+
+#if BUILDFLAG(IS_MAC)
+
+// SecTaskGetCodeSignStatus is marked as unavailable on macOS, although it’s
+// available on iOS and other Apple operating systems. It is, in fact, present
+// on the system since macOS 10.12.
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wavailability"
+uint32_t SecTaskGetCodeSignStatus(SecTaskRef task) API_AVAILABLE(macos(10.12));
+#pragma clang diagnostic pop
+
+#endif  // BUILDFLAG(IS_MAC)
+
+namespace partition_alloc::internal {
+
+namespace {
+
+#if defined(LINUX_NAME_REGION)
+
+void NameRegion(void* start, size_t length, PageTag page_tag) {
+  // Important: All the names should be string literals. As per prctl.h in
+  // //third_party/android_toolchain/ndk the kernel keeps a pointer to the name
+  // instead of copying it.
+  //
+  // Having the name in .rodata ensures that the pointer remains valid as
+  // long as the mapping is alive.
+  const char* name = nullptr;
+  switch (page_tag) {
+    case PageTag::kSimulation:
+      name = "simulation";
+      break;
+    case PageTag::kBlinkGC:
+      name = "blink_gc";
+      break;
+    case PageTag::kPartitionAlloc:
+      name = "partition_alloc";
+      break;
+    case PageTag::kChromium:
+      name = "chromium";
+      break;
+    case PageTag::kV8:
+      name = "v8";
+      break;
+    default:
+      PA_NOTREACHED();
+      break;
+  }
+
+  // No error checking on purpose, testing only.
+  prctl(PR_SET_VMA, PR_SET_VMA_ANON_NAME, start, length, name);
+}
+
+#endif  // defined(LINUX_NAME_REGION)
+
+#if BUILDFLAG(IS_MAC)
+// Tests whether the version of macOS supports the MAP_JIT flag and if the
+// current process is signed with the hardened runtime and the allow-jit
+// entitlement, returning whether MAP_JIT should be used to allocate regions
+// that will contain JIT-compiled executable code.
+bool UseMapJit() {
+  // Until determining that the hardened runtime is enabled, early returns will
+  // return true, so that MAP_JIT will be used. This is important on arm64,
+  // which only allows pages to be simultaneously writable and executable when
+  // in a region allocated with MAP_JIT, regardless of code signing options. On
+  // arm64, an attempt to set a non-MAP_JIT page as simultaneously writable and
+  // executable fails with EPERM. Although this is not enforced on x86_64,
+  // MAP_JIT is harmless in that case.
+
+  base::apple::ScopedCFTypeRef<SecTaskRef> task(
+      SecTaskCreateFromSelf(kCFAllocatorDefault));
+  if (!task) {
+    return true;
+  }
+
+  uint32_t flags = SecTaskGetCodeSignStatus(task);
+  if (!(flags & kSecCodeSignatureRuntime)) {
+    // The hardened runtime is not enabled. Note that kSecCodeSignatureRuntime
+    // == CS_RUNTIME.
+    return true;
+  }
+
+  // The hardened runtime is enabled. From this point on, early returns must
+  // return false, indicating that MAP_JIT is not to be used. It’s an error
+  // (EINVAL) to use MAP_JIT with the hardened runtime unless the JIT
+  // entitlement is specified.
+
+  base::apple::ScopedCFTypeRef<CFTypeRef> jit_entitlement(
+      SecTaskCopyValueForEntitlement(
+          task.get(), CFSTR("com.apple.security.cs.allow-jit"), nullptr));
+  if (!jit_entitlement) {
+    return false;
+  }
+
+  return base::apple::CFCast<CFBooleanRef>(jit_entitlement.get()) ==
+         kCFBooleanTrue;
+}
+#elif BUILDFLAG(IS_IOS)
+bool UseMapJit() {
+// Always enable MAP_JIT in simulator as it is supported unconditionally.
+#if TARGET_IPHONE_SIMULATOR
+  return true;
+#else
+  // TODO(https://crbug.com/1413818): Fill this out when the API it is
+  // available.
+  return false;
+#endif  // TARGET_IPHONE_SIMULATOR
+}
+#endif  // BUILDFLAG(IS_IOS)
+}  // namespace
+
+// |mmap| uses a nearby address if the hint address is blocked.
+constexpr bool kHintIsAdvisory = true;
+std::atomic<int32_t> s_allocPageErrorCode{0};
+
+int GetAccessFlags(PageAccessibilityConfiguration accessibility);
+
+uintptr_t SystemAllocPagesInternal(uintptr_t hint,
+                                   size_t length,
+                                   PageAccessibilityConfiguration accessibility,
+                                   PageTag page_tag,
+                                   int file_descriptor_for_shared_alloc) {
+#if BUILDFLAG(IS_APPLE)
+  // Use a custom tag to make it easier to distinguish Partition Alloc regions
+  // in vmmap(1). Tags between 240-255 are supported.
+  int fd = file_descriptor_for_shared_alloc == -1
+               ? VM_MAKE_TAG(static_cast<int>(page_tag))
+               : file_descriptor_for_shared_alloc;
+#else
+  int fd = file_descriptor_for_shared_alloc;
+#endif
+
+  int access_flag = GetAccessFlags(accessibility);
+  int map_flags = MAP_ANONYMOUS | MAP_PRIVATE;
+
+#if BUILDFLAG(IS_APPLE)
+  // On macOS, executables that are code signed with the "runtime" option cannot
+  // execute writable memory by default. They can opt into this capability by
+  // specifying the "com.apple.security.cs.allow-jit" code signing entitlement
+  // and allocating the region with the MAP_JIT flag.
+  static const bool kUseMapJit = UseMapJit();
+  if (accessibility.permissions ==
+          PageAccessibilityConfiguration::kInaccessibleWillJitLater &&
+      kUseMapJit) {
+    map_flags |= MAP_JIT;
+  }
+#endif
+
+  void* ret = mmap(reinterpret_cast<void*>(hint), length, access_flag,
+                   map_flags, fd, 0);
+  if (ret == MAP_FAILED) {
+    s_allocPageErrorCode = errno;
+    ret = nullptr;
+  }
+
+#if defined(LINUX_NAME_REGION)
+  if (ret) {
+    NameRegion(ret, length, page_tag);
+  }
+#endif
+
+  return reinterpret_cast<uintptr_t>(ret);
+}
+
+bool TrySetSystemPagesAccessInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility) {
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  if (accessibility.thread_isolation.enabled) {
+    return 0 == MprotectWithThreadIsolation(reinterpret_cast<void*>(address),
+                                            length,
+                                            GetAccessFlags(accessibility),
+                                            accessibility.thread_isolation);
+  }
+#endif  // BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  return 0 == PA_HANDLE_EINTR(mprotect(reinterpret_cast<void*>(address), length,
+                                       GetAccessFlags(accessibility)));
+}
+
+void SetSystemPagesAccessInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility) {
+  int access_flags = GetAccessFlags(accessibility);
+  int ret;
+#if BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  if (accessibility.thread_isolation.enabled) {
+    ret = MprotectWithThreadIsolation(reinterpret_cast<void*>(address), length,
+                                      GetAccessFlags(accessibility),
+                                      accessibility.thread_isolation);
+  } else
+#endif  // BUILDFLAG(ENABLE_THREAD_ISOLATION)
+  {
+    ret = PA_HANDLE_EINTR(mprotect(reinterpret_cast<void*>(address), length,
+                                   GetAccessFlags(accessibility)));
+  }
+
+  // On Linux, man mprotect(2) states that ENOMEM is returned when (1) internal
+  // kernel data structures cannot be allocated, (2) the address range is
+  // invalid, or (3) this would split an existing mapping in a way that would
+  // exceed the maximum number of allowed mappings.
+  //
+  // Neither are very likely, but we still get a lot of crashes here. This is
+  // because setrlimit(RLIMIT_DATA)'s limit is checked and enforced here, if the
+  // access flags match a "data" mapping, which in our case would be MAP_PRIVATE
+  // | MAP_ANONYMOUS, and PROT_WRITE. see the call to may_expand_vm() in
+  // mm/mprotect.c in the kernel for details.
+  //
+  // In this case, we are almost certainly bumping into the sandbox limit, mark
+  // the crash as OOM. See SandboxLinux::LimitAddressSpace() for details.
+  if (ret == -1 && errno == ENOMEM && (access_flags & PROT_WRITE)) {
+    OOM_CRASH(length);
+  }
+
+  PA_PCHECK(0 == ret);
+}
+
+void FreePagesInternal(uintptr_t address, size_t length) {
+  PA_PCHECK(0 == munmap(reinterpret_cast<void*>(address), length));
+}
+
+uintptr_t TrimMappingInternal(uintptr_t base_address,
+                              size_t base_length,
+                              size_t trim_length,
+                              PageAccessibilityConfiguration accessibility,
+                              size_t pre_slack,
+                              size_t post_slack) {
+  uintptr_t ret = base_address;
+  // We can resize the allocation run. Release unneeded memory before and after
+  // the aligned range.
+  if (pre_slack) {
+    FreePages(base_address, pre_slack);
+    ret = base_address + pre_slack;
+  }
+  if (post_slack) {
+    FreePages(ret + trim_length, post_slack);
+  }
+  return ret;
+}
+
+void DecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // In POSIX, there is no decommit concept. Discarding is an effective way of
+  // implementing the Windows semantics where the OS is allowed to not swap the
+  // pages in the region.
+  DiscardSystemPages(address, length);
+
+  bool change_permissions =
+      accessibility_disposition == PageAccessibilityDisposition::kRequireUpdate;
+#if BUILDFLAG(PA_DCHECK_IS_ON)
+  // This is not guaranteed, show that we're serious.
+  //
+  // More specifically, several callers have had issues with assuming that
+  // memory is zeroed, this would hopefully make these bugs more visible.  We
+  // don't memset() everything, because ranges can be very large, and doing it
+  // over the entire range could make Chrome unusable with
+  // BUILDFLAG(PA_DCHECK_IS_ON).
+  //
+  // Only do it when we are about to change the permissions, since we don't know
+  // the previous permissions, and cannot restore them.
+  if (!DecommittedMemoryIsAlwaysZeroed() && change_permissions) {
+    // Memory may not be writable.
+    size_t size = std::min(length, 2 * SystemPageSize());
+    void* ptr = reinterpret_cast<void*>(address);
+    PA_CHECK(mprotect(ptr, size, PROT_WRITE) == 0);
+    memset(ptr, 0xcc, size);
+  }
+#endif
+
+  // Make pages inaccessible, unless the caller requested to keep permissions.
+  //
+  // Note, there is a small window between these calls when the pages can be
+  // incorrectly touched and brought back to memory. Not ideal, but doing those
+  // operations in the opposite order resulted in PMF regression on Mac (see
+  // crbug.com/1153021).
+  if (change_permissions) {
+    SetSystemPagesAccess(address, length,
+                         PageAccessibilityConfiguration(
+                             PageAccessibilityConfiguration::kInaccessible));
+  }
+}
+
+void DecommitAndZeroSystemPagesInternal(uintptr_t address,
+                                        size_t length,
+                                        PageTag page_tag) {
+  int fd = -1;
+#if BUILDFLAG(IS_APPLE)
+  fd = VM_MAKE_TAG(static_cast<int>(page_tag));
+#endif
+
+  // https://pubs.opengroup.org/onlinepubs/9699919799/functions/mmap.html: "If
+  // a MAP_FIXED request is successful, then any previous mappings [...] for
+  // those whole pages containing any part of the address range [pa,pa+len)
+  // shall be removed, as if by an appropriate call to munmap(), before the
+  // new mapping is established." As a consequence, the memory will be
+  // zero-initialized on next access.
+  void* ptr = reinterpret_cast<void*>(address);
+  void* ret = mmap(ptr, length, PROT_NONE,
+                   MAP_FIXED | MAP_ANONYMOUS | MAP_PRIVATE, fd, 0);
+  PA_CHECK(ptr == ret);
+  // Since we just remapped the region, need to set is name again.
+#if defined(LINUX_NAME_REGION)
+  NameRegion(ret, length, page_tag);
+#endif
+}
+
+void RecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // On POSIX systems, the caller needs to simply read the memory to recommit
+  // it. However, if decommit changed the permissions, recommit has to change
+  // them back.
+  if (accessibility_disposition ==
+      PageAccessibilityDisposition::kRequireUpdate) {
+    SetSystemPagesAccess(address, length, accessibility);
+  }
+
+#if BUILDFLAG(IS_APPLE)
+  // On macOS, to update accounting, we need to make another syscall. For more
+  // details, see https://crbug.com/823915.
+  madvise(reinterpret_cast<void*>(address), length, MADV_FREE_REUSE);
+#endif
+}
+
+bool TryRecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // On POSIX systems, the caller needs to simply read the memory to recommit
+  // it. However, if decommit changed the permissions, recommit has to change
+  // them back.
+  if (accessibility_disposition ==
+      PageAccessibilityDisposition::kRequireUpdate) {
+    bool ok = TrySetSystemPagesAccess(address, length, accessibility);
+    if (!ok) {
+      return false;
+    }
+  }
+
+#if BUILDFLAG(IS_APPLE)
+  // On macOS, to update accounting, we need to make another syscall. For more
+  // details, see https://crbug.com/823915.
+  madvise(reinterpret_cast<void*>(address), length, MADV_FREE_REUSE);
+#endif
+
+  return true;
+}
+
+void DiscardSystemPagesInternal(uintptr_t address, size_t length) {
+  void* ptr = reinterpret_cast<void*>(address);
+#if BUILDFLAG(IS_APPLE)
+  int ret = madvise(ptr, length, MADV_FREE_REUSABLE);
+  if (ret) {
+    // MADV_FREE_REUSABLE sometimes fails, so fall back to MADV_DONTNEED.
+    ret = madvise(ptr, length, MADV_DONTNEED);
+  }
+  PA_PCHECK(ret == 0);
+#else   // BUILDFLAG(IS_APPLE)
+  // We have experimented with other flags, but with suboptimal results.
+  //
+  // MADV_FREE (Linux): Makes our memory measurements less predictable;
+  // performance benefits unclear.
+  //
+  // Therefore, we just do the simple thing: MADV_DONTNEED.
+  PA_PCHECK(0 == madvise(ptr, length, MADV_DONTNEED));
+#endif  // BUILDFLAG(IS_APPLE)
+}
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_POSIX_H_

src/base/allocator/partition_allocator/page_allocator_internals_win.h

@@ -0,0 +1,240 @@
+// Copyright 2018 The Chromium Authors
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_WIN_H_
+#define BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_WIN_H_
+
+#include <cstdint>
+
+#include "base/allocator/partition_allocator/oom.h"
+#include "base/allocator/partition_allocator/page_allocator.h"
+#include "base/allocator/partition_allocator/page_allocator_internal.h"
+#include "base/allocator/partition_allocator/partition_alloc_base/notreached.h"
+#include "base/allocator/partition_allocator/partition_alloc_buildflags.h"
+#include "base/allocator/partition_allocator/partition_alloc_check.h"
+
+namespace partition_alloc::internal {
+
+// |VirtualAlloc| will fail if allocation at the hint address is blocked.
+constexpr bool kHintIsAdvisory = false;
+std::atomic<int32_t> s_allocPageErrorCode{ERROR_SUCCESS};
+
+bool IsOutOfMemory(DWORD error) {
+  // From
+  // https://learn.microsoft.com/en-us/windows/win32/debug/system-error-codes--0-499-
+  switch (error) {
+    // Page file is being extended.
+    case ERROR_COMMITMENT_MINIMUM:
+      // Page file is too small.
+    case ERROR_COMMITMENT_LIMIT:
+#if BUILDFLAG(HAS_64_BIT_POINTERS)
+    // Not enough memory resources are available to process this command.
+    //
+    // It is not entirely clear whether this error pertains to out of address
+    // space errors, or the kernel being out of memory. Only include it for 64
+    // bit architectures, since address space issues are unlikely there.
+    case ERROR_NOT_ENOUGH_MEMORY:
+#endif
+    case ERROR_PAGEFILE_QUOTA:
+      // Insufficient quota to complete the requested service.
+      return true;
+    default:
+      return false;
+  }
+}
+
+void* VirtualAllocWithRetry(void* address,
+                            size_t size,
+                            DWORD type_flags,
+                            DWORD access_flags) {
+  void* ret = nullptr;
+  // Failure to commit memory can be temporary, in at least two cases:
+  // - The page file is getting extended.
+  // - Another process terminates (most likely because of OOM)
+  //
+  // Wait and retry, since the alternative is crashing. Note that if we
+  // selectively apply this... hum... beautiful hack to some process types only,
+  // "some process crashing" may very well be one of ours, which may be
+  // desirable (e.g. some processes like the browser are more important than
+  // others).
+  //
+  // This approach has been shown to be effective for Firefox, see
+  // crbug.com/1392738 for context. Constants below are accordingly taken from
+  // Firefox as well.
+  constexpr int kMaxTries = 10;
+  constexpr int kDelayMs = 50;
+
+  bool should_retry = GetRetryOnCommitFailure() && (type_flags & MEM_COMMIT) &&
+                      (access_flags != PAGE_NOACCESS);
+  for (int tries = 0; tries < kMaxTries; tries++) {
+    ret = VirtualAlloc(address, size, type_flags, access_flags);
+    // Only retry for commit failures. If this is an address space problem
+    // (e.g. caller asked for an address which is not available), this is
+    // unlikely to be resolved by waiting.
+    if (ret || !should_retry || !IsOutOfMemory(GetLastError())) {
+      break;
+    }
+
+    Sleep(kDelayMs);
+  }
+  return ret;
+}
+
+int GetAccessFlags(PageAccessibilityConfiguration accessibility) {
+  switch (accessibility.permissions) {
+    case PageAccessibilityConfiguration::kRead:
+      return PAGE_READONLY;
+    case PageAccessibilityConfiguration::kReadWrite:
+    case PageAccessibilityConfiguration::kReadWriteTagged:
+      return PAGE_READWRITE;
+    case PageAccessibilityConfiguration::kReadExecute:
+    case PageAccessibilityConfiguration::kReadExecuteProtected:
+      return PAGE_EXECUTE_READ;
+    case PageAccessibilityConfiguration::kReadWriteExecute:
+      return PAGE_EXECUTE_READWRITE;
+    case PageAccessibilityConfiguration::kInaccessible:
+    case PageAccessibilityConfiguration::kInaccessibleWillJitLater:
+      return PAGE_NOACCESS;
+  }
+  PA_NOTREACHED();
+}
+
+uintptr_t SystemAllocPagesInternal(
+    uintptr_t hint,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageTag page_tag,
+    [[maybe_unused]] int file_descriptor_for_shared_alloc) {
+  const DWORD access_flag = GetAccessFlags(accessibility);
+  const DWORD type_flags =
+      (access_flag == PAGE_NOACCESS) ? MEM_RESERVE : (MEM_RESERVE | MEM_COMMIT);
+  void* ret = VirtualAllocWithRetry(reinterpret_cast<void*>(hint), length,
+                                    type_flags, access_flag);
+  if (ret == nullptr) {
+    s_allocPageErrorCode = GetLastError();
+  }
+  return reinterpret_cast<uintptr_t>(ret);
+}
+
+uintptr_t TrimMappingInternal(uintptr_t base_address,
+                              size_t base_length,
+                              size_t trim_length,
+                              PageAccessibilityConfiguration accessibility,
+                              size_t pre_slack,
+                              size_t post_slack) {
+  uintptr_t ret = base_address;
+  if (pre_slack || post_slack) {
+    // We cannot resize the allocation run. Free it and retry at the aligned
+    // address within the freed range.
+    ret = base_address + pre_slack;
+    FreePages(base_address, base_length);
+    ret = SystemAllocPages(ret, trim_length, accessibility, PageTag::kChromium);
+  }
+  return ret;
+}
+
+bool TrySetSystemPagesAccessInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility) {
+  void* ptr = reinterpret_cast<void*>(address);
+  if (GetAccessFlags(accessibility) == PAGE_NOACCESS) {
+    return VirtualFree(ptr, length, MEM_DECOMMIT) != 0;
+  }
+  // Call the retry path even though this function can fail, because callers of
+  // this are likely to crash the process when this function fails, and we don't
+  // want that for transient failures.
+  return nullptr != VirtualAllocWithRetry(ptr, length, MEM_COMMIT,
+                                          GetAccessFlags(accessibility));
+}
+
+void SetSystemPagesAccessInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility) {
+  void* ptr = reinterpret_cast<void*>(address);
+  const DWORD access_flag = GetAccessFlags(accessibility);
+  if (access_flag == PAGE_NOACCESS) {
+    if (!VirtualFree(ptr, length, MEM_DECOMMIT)) {
+      // We check `GetLastError` for `ERROR_SUCCESS` here so that in a crash
+      // report we get the error number.
+      PA_CHECK(static_cast<uint32_t>(ERROR_SUCCESS) == GetLastError());
+    }
+  } else {
+    if (!VirtualAllocWithRetry(ptr, length, MEM_COMMIT, access_flag)) {
+      int32_t error = GetLastError();
+      if (error == ERROR_COMMITMENT_LIMIT) {
+        OOM_CRASH(length);
+      }
+      // We check `GetLastError` for `ERROR_SUCCESS` here so that in a crash
+      // report we get the error number.
+      PA_CHECK(ERROR_SUCCESS == error);
+    }
+  }
+}
+
+void FreePagesInternal(uintptr_t address, size_t length) {
+  PA_CHECK(VirtualFree(reinterpret_cast<void*>(address), 0, MEM_RELEASE));
+}
+
+void DecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // Ignore accessibility_disposition, because decommitting is equivalent to
+  // making pages inaccessible.
+  SetSystemPagesAccess(address, length,
+                       PageAccessibilityConfiguration(
+                           PageAccessibilityConfiguration::kInaccessible));
+}
+
+void DecommitAndZeroSystemPagesInternal(uintptr_t address,
+                                        size_t length,
+                                        PageTag page_tag) {
+  // https://docs.microsoft.com/en-us/windows/win32/api/memoryapi/nf-memoryapi-virtualfree:
+  // "If a page is decommitted but not released, its state changes to reserved.
+  // Subsequently, you can call VirtualAlloc to commit it, or VirtualFree to
+  // release it. Attempts to read from or write to a reserved page results in an
+  // access violation exception."
+  // https://docs.microsoft.com/en-us/windows/win32/api/memoryapi/nf-memoryapi-virtualalloc
+  // for MEM_COMMIT: "The function also guarantees that when the caller later
+  // initially accesses the memory, the contents will be zero."
+  PA_CHECK(VirtualFree(reinterpret_cast<void*>(address), length, MEM_DECOMMIT));
+}
+
+void RecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // Ignore accessibility_disposition, because decommitting is equivalent to
+  // making pages inaccessible.
+  SetSystemPagesAccess(address, length, accessibility);
+}
+
+bool TryRecommitSystemPagesInternal(
+    uintptr_t address,
+    size_t length,
+    PageAccessibilityConfiguration accessibility,
+    PageAccessibilityDisposition accessibility_disposition) {
+  // Ignore accessibility_disposition, because decommitting is equivalent to
+  // making pages inaccessible.
+  return TrySetSystemPagesAccess(address, length, accessibility);
+}
+
+void DiscardSystemPagesInternal(uintptr_t address, size_t length) {
+  void* ptr = reinterpret_cast<void*>(address);
+  // Use DiscardVirtualMemory when available because it releases faster than
+  // MEM_RESET.
+  DWORD ret = DiscardVirtualMemory(ptr, length);
+  // DiscardVirtualMemory is buggy in Win10 SP0, so fall back to MEM_RESET on
+  // failure.
+  if (ret) {
+    PA_CHECK(VirtualAllocWithRetry(ptr, length, MEM_RESET, PAGE_READWRITE));
+  }
+}
+
+}  // namespace partition_alloc::internal
+
+#endif  // BASE_ALLOCATOR_PARTITION_ALLOCATOR_PAGE_ALLOCATOR_INTERNALS_WIN_H_