docs: updated the license date

GCC 15 starts warning about non NUL-terminated string literals:

    chacha.c:44:31: error: initializer-string for array of ‘char’ truncates NUL terminator but destination lacks ‘nonstring’ attribute (17 chars into 16 available) [-Werror=unterminated-string-initialization]
       44 | static const char sigma[16] = "expand 32-byte k";
          |                               ^~~~~~~~~~~~~~~~~~

.clang-tidy

@@ -0,0 +1,2 @@
+Checks: 'bugprone-*,-bugprone-easily-swappable-parameters,-bugprone-macro-parentheses,-bugprone-too-small-loop-variable,cert-*,-cert-err33-c,clang-analyzer-*,-clang-analyzer-security.insecureAPI.DeprecatedOrUnsafeBufferHandling,-clang-diagnostic-constant-logical-operand,readability-*,-readability-function-cognitive-complexity,-readability-identifier-length,-readability-inconsistent-declaration-parameter-name,-readability-magic-numbers,-readability-named-parameter,llvm-include-order,misc-*'
+WarningsAsErrors: '*'

.github/dependabot.yml

@@ -0,0 +1,7 @@
+version: 2
+updates:
+  - package-ecosystem: github-actions
+    directory: "/"
+    schedule:
+      interval: daily
+    target-branch: main

.github/workflows/build-and-test.yml

@@ -0,0 +1,55 @@
+name: Build and run tests
+
+on:
+  push:
+  pull_request:
+  schedule:
+    - cron: '0 2 * * *'
+
+jobs:
+  build-ubuntu-gcc:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        version: [12, 13, 14]
+    steps:
+      - uses: actions/checkout@v4
+      - name: Setting up gcc version
+        run: |
+          sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-${{ matrix.version }} 100
+          sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-${{ matrix.version }} 100
+      - name: Build
+        run: make test
+  build-ubuntu-clang:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        version: [14, 15, 16, 17, 18]
+    steps:
+      - uses: actions/checkout@v4
+      - name: Install dependencies
+        run: sudo apt-get update && sudo apt-get install -y --no-install-recommends clang-14 clang-15
+      - name: Setting up clang version
+        run: |
+          sudo update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-${{ matrix.version }} 100
+          sudo update-alternatives --install /usr/bin/clang clang /usr/bin/clang-${{ matrix.version }} 100
+      - name: Build
+        run: CC=clang CXX=clang++ make test
+  build-musl:
+    runs-on: ubuntu-latest
+    container:
+      image: alpine:latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Install dependencies
+        run: apk update && apk add build-base python3
+      - name: Build
+        run: make test
+  build-ubuntu-gcc-aarch64:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Install dependencies
+        run: sudo apt-get update && sudo apt-get install -y --no-install-recommends gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libgcc-s1-arm64-cross cpp-aarch64-linux-gnu
+      - name: Build
+        run: CC=aarch64-linux-gnu-gcc CXX=aarch64-linux-gnu-gcc++ make CONFIG_NATIVE=false

.gitignore

@@ -1,2 +1,2 @@
-*.o
-*.so
+out/
+out-light/

Android.bp

@@ -1,15 +1,14 @@
 common_cflags = [
+    "-pipe",
     "-O3",
     //"-flto",
     "-fPIC",
     "-fvisibility=hidden",
     //"-fno-plt",
-    "-pipe",
-    "-Wall",
-    "-Wextra",
     "-Wcast-align",
     "-Wcast-qual",
     "-Wwrite-strings",
+    "-Werror",
     "-DH_MALLOC_PREFIX",
     "-DZERO_ON_FREE=true",
     "-DWRITE_AFTER_FREE_CHECK=true",
@@ -21,20 +20,21 @@ common_cflags = [
     "-DCONFIG_LARGE_SIZE_CLASSES=true",
     "-DGUARD_SLABS_INTERVAL=1",
     "-DGUARD_SIZE_DIVISOR=2",
-    "-DREGION_QUARANTINE_RANDOM_LENGTH=128",
+    "-DREGION_QUARANTINE_RANDOM_LENGTH=256",
     "-DREGION_QUARANTINE_QUEUE_LENGTH=1024",
     "-DREGION_QUARANTINE_SKIP_THRESHOLD=33554432", // 32MiB
     "-DFREE_SLABS_QUARANTINE_RANDOM_LENGTH=32",
-    "-DCONFIG_CLASS_REGION_SIZE=1073741824", // 1GiB
+    "-DCONFIG_CLASS_REGION_SIZE=34359738368", // 32GiB
     "-DN_ARENA=1",
     "-DCONFIG_STATS=true",
+    "-DCONFIG_SELF_INIT=false",
 ]
 
 cc_defaults {
     name: "hardened_malloc_defaults",
     defaults: ["linux_bionic_supported"],
     cflags: common_cflags,
-    conlyflags: ["-std=c11", "-Wmissing-prototypes"],
+    conlyflags: ["-std=c17", "-Wmissing-prototypes"],
     stl: "none",
 }
 
@@ -47,13 +47,35 @@ lib_src_files = [
     "util.c",
 ]
 
-cc_library_static {
+cc_library {
     name: "libhardened_malloc",
+    ramdisk_available: true,
+    vendor_ramdisk_available: true,
+    recovery_available: true,
     defaults: ["hardened_malloc_defaults"],
     srcs: lib_src_files,
+    export_include_dirs: ["include"],
+    static_libs: ["libasync_safe"],
+    target: {
+        android: {
+            shared: {
+                enabled: false,
+            },
+            system_shared_libs: [],
+        },
+        linux_bionic: {
+            system_shared_libs: [],
+        },
+    },
     product_variables: {
         debuggable: {
             cflags: ["-DLABEL_MEMORY"],
         },
+        device_has_arm_mte: {
+            cflags: ["-DHAS_ARM_MTE", "-march=armv8-a+dotprod+memtag"]
+        },
     },
+    apex_available: [
+        "com.android.runtime",
+    ],
 }

CREDITS

@@ -4,7 +4,7 @@ chacha.c is a simple conversion of chacha-merged.c to a keystream-only implement
     D. J. Bernstein
     Public domain.
 
-malloc.c open-addressed hash table (regions_grow, regions_insert, regions_find, regions_delete):
+h_malloc.c open-addressed hash table (regions_grow, regions_insert, regions_find, regions_delete):
 
     Copyright (c) 2008, 2010, 2011, 2016 Otto Moerbeek <otto@drijf.net>
     Copyright (c) 2012 Matthew Dempsky <matthew@openbsd.org>
@@ -25,7 +25,8 @@ malloc.c open-addressed hash table (regions_grow, regions_insert, regions_find,
 
 libdivide:
 
-    Copyright (C) 2010 ridiculous_fish
+    Copyright (C) 2010 - 2019 ridiculous_fish, <libdivide@ridiculousfish.com>
+    Copyright (C) 2016 - 2019 Kim Walisch, <kim.walisch@gmail.com>
 
     Boost Software License - Version 1.0 - August 17th, 2003
 
@@ -53,3 +54,230 @@ libdivide:
 
 random.c get_random_{type}_uniform functions are based on Fast Random Integer
 Generation in an Interval by Daniel Lemire
+
+arm_mte.h arm_mte_tag_and_clear_mem function contents were copied from storeTags function in scudo:
+
+    ==============================================================================
+    The LLVM Project is under the Apache License v2.0 with LLVM Exceptions:
+    ==============================================================================
+
+                                     Apache License
+                               Version 2.0, January 2004
+                            http://www.apache.org/licenses/
+
+        TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+        1. Definitions.
+
+          "License" shall mean the terms and conditions for use, reproduction,
+          and distribution as defined by Sections 1 through 9 of this document.
+
+          "Licensor" shall mean the copyright owner or entity authorized by
+          the copyright owner that is granting the License.
+
+          "Legal Entity" shall mean the union of the acting entity and all
+          other entities that control, are controlled by, or are under common
+          control with that entity. For the purposes of this definition,
+          "control" means (i) the power, direct or indirect, to cause the
+          direction or management of such entity, whether by contract or
+          otherwise, or (ii) ownership of fifty percent (50%) or more of the
+          outstanding shares, or (iii) beneficial ownership of such entity.
+
+          "You" (or "Your") shall mean an individual or Legal Entity
+          exercising permissions granted by this License.
+
+          "Source" form shall mean the preferred form for making modifications,
+          including but not limited to software source code, documentation
+          source, and configuration files.
+
+          "Object" form shall mean any form resulting from mechanical
+          transformation or translation of a Source form, including but
+          not limited to compiled object code, generated documentation,
+          and conversions to other media types.
+
+          "Work" shall mean the work of authorship, whether in Source or
+          Object form, made available under the License, as indicated by a
+          copyright notice that is included in or attached to the work
+          (an example is provided in the Appendix below).
+
+          "Derivative Works" shall mean any work, whether in Source or Object
+          form, that is based on (or derived from) the Work and for which the
+          editorial revisions, annotations, elaborations, or other modifications
+          represent, as a whole, an original work of authorship. For the purposes
+          of this License, Derivative Works shall not include works that remain
+          separable from, or merely link (or bind by name) to the interfaces of,
+          the Work and Derivative Works thereof.
+
+          "Contribution" shall mean any work of authorship, including
+          the original version of the Work and any modifications or additions
+          to that Work or Derivative Works thereof, that is intentionally
+          submitted to Licensor for inclusion in the Work by the copyright owner
+          or by an individual or Legal Entity authorized to submit on behalf of
+          the copyright owner. For the purposes of this definition, "submitted"
+          means any form of electronic, verbal, or written communication sent
+          to the Licensor or its representatives, including but not limited to
+          communication on electronic mailing lists, source code control systems,
+          and issue tracking systems that are managed by, or on behalf of, the
+          Licensor for the purpose of discussing and improving the Work, but
+          excluding communication that is conspicuously marked or otherwise
+          designated in writing by the copyright owner as "Not a Contribution."
+
+          "Contributor" shall mean Licensor and any individual or Legal Entity
+          on behalf of whom a Contribution has been received by Licensor and
+          subsequently incorporated within the Work.
+
+        2. Grant of Copyright License. Subject to the terms and conditions of
+          this License, each Contributor hereby grants to You a perpetual,
+          worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+          copyright license to reproduce, prepare Derivative Works of,
+          publicly display, publicly perform, sublicense, and distribute the
+          Work and such Derivative Works in Source or Object form.
+
+        3. Grant of Patent License. Subject to the terms and conditions of
+          this License, each Contributor hereby grants to You a perpetual,
+          worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+          (except as stated in this section) patent license to make, have made,
+          use, offer to sell, sell, import, and otherwise transfer the Work,
+          where such license applies only to those patent claims licensable
+          by such Contributor that are necessarily infringed by their
+          Contribution(s) alone or by combination of their Contribution(s)
+          with the Work to which such Contribution(s) was submitted. If You
+          institute patent litigation against any entity (including a
+          cross-claim or counterclaim in a lawsuit) alleging that the Work
+          or a Contribution incorporated within the Work constitutes direct
+          or contributory patent infringement, then any patent licenses
+          granted to You under this License for that Work shall terminate
+          as of the date such litigation is filed.
+
+        4. Redistribution. You may reproduce and distribute copies of the
+          Work or Derivative Works thereof in any medium, with or without
+          modifications, and in Source or Object form, provided that You
+          meet the following conditions:
+
+          (a) You must give any other recipients of the Work or
+              Derivative Works a copy of this License; and
+
+          (b) You must cause any modified files to carry prominent notices
+              stating that You changed the files; and
+
+          (c) You must retain, in the Source form of any Derivative Works
+              that You distribute, all copyright, patent, trademark, and
+              attribution notices from the Source form of the Work,
+              excluding those notices that do not pertain to any part of
+              the Derivative Works; and
+
+          (d) If the Work includes a "NOTICE" text file as part of its
+              distribution, then any Derivative Works that You distribute must
+              include a readable copy of the attribution notices contained
+              within such NOTICE file, excluding those notices that do not
+              pertain to any part of the Derivative Works, in at least one
+              of the following places: within a NOTICE text file distributed
+              as part of the Derivative Works; within the Source form or
+              documentation, if provided along with the Derivative Works; or,
+              within a display generated by the Derivative Works, if and
+              wherever such third-party notices normally appear. The contents
+              of the NOTICE file are for informational purposes only and
+              do not modify the License. You may add Your own attribution
+              notices within Derivative Works that You distribute, alongside
+              or as an addendum to the NOTICE text from the Work, provided
+              that such additional attribution notices cannot be construed
+              as modifying the License.
+
+          You may add Your own copyright statement to Your modifications and
+          may provide additional or different license terms and conditions
+          for use, reproduction, or distribution of Your modifications, or
+          for any such Derivative Works as a whole, provided Your use,
+          reproduction, and distribution of the Work otherwise complies with
+          the conditions stated in this License.
+
+        5. Submission of Contributions. Unless You explicitly state otherwise,
+          any Contribution intentionally submitted for inclusion in the Work
+          by You to the Licensor shall be under the terms and conditions of
+          this License, without any additional terms or conditions.
+          Notwithstanding the above, nothing herein shall supersede or modify
+          the terms of any separate license agreement you may have executed
+          with Licensor regarding such Contributions.
+
+        6. Trademarks. This License does not grant permission to use the trade
+          names, trademarks, service marks, or product names of the Licensor,
+          except as required for reasonable and customary use in describing the
+          origin of the Work and reproducing the content of the NOTICE file.
+
+        7. Disclaimer of Warranty. Unless required by applicable law or
+          agreed to in writing, Licensor provides the Work (and each
+          Contributor provides its Contributions) on an "AS IS" BASIS,
+          WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+          implied, including, without limitation, any warranties or conditions
+          of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+          PARTICULAR PURPOSE. You are solely responsible for determining the
+          appropriateness of using or redistributing the Work and assume any
+          risks associated with Your exercise of permissions under this License.
+
+        8. Limitation of Liability. In no event and under no legal theory,
+          whether in tort (including negligence), contract, or otherwise,
+          unless required by applicable law (such as deliberate and grossly
+          negligent acts) or agreed to in writing, shall any Contributor be
+          liable to You for damages, including any direct, indirect, special,
+          incidental, or consequential damages of any character arising as a
+          result of this License or out of the use or inability to use the
+          Work (including but not limited to damages for loss of goodwill,
+          work stoppage, computer failure or malfunction, or any and all
+          other commercial damages or losses), even if such Contributor
+          has been advised of the possibility of such damages.
+
+        9. Accepting Warranty or Additional Liability. While redistributing
+          the Work or Derivative Works thereof, You may choose to offer,
+          and charge a fee for, acceptance of support, warranty, indemnity,
+          or other liability obligations and/or rights consistent with this
+          License. However, in accepting such obligations, You may act only
+          on Your own behalf and on Your sole responsibility, not on behalf
+          of any other Contributor, and only if You agree to indemnify,
+          defend, and hold each Contributor harmless for any liability
+          incurred by, or claims asserted against, such Contributor by reason
+          of your accepting any such warranty or additional liability.
+
+        END OF TERMS AND CONDITIONS
+
+        APPENDIX: How to apply the Apache License to your work.
+
+          To apply the Apache License to your work, attach the following
+          boilerplate notice, with the fields enclosed by brackets "[]"
+          replaced with your own identifying information. (Don't include
+          the brackets!)  The text should be enclosed in the appropriate
+          comment syntax for the file format. We also recommend that a
+          file or class name and description of purpose be included on the
+          same "printed page" as the copyright notice for easier
+          identification within third-party archives.
+
+        Copyright [yyyy] [name of copyright owner]
+
+        Licensed under the Apache License, Version 2.0 (the "License");
+        you may not use this file except in compliance with the License.
+        You may obtain a copy of the License at
+
+           http://www.apache.org/licenses/LICENSE-2.0
+
+        Unless required by applicable law or agreed to in writing, software
+        distributed under the License is distributed on an "AS IS" BASIS,
+        WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+        See the License for the specific language governing permissions and
+        limitations under the License.
+
+
+    ---- LLVM Exceptions to the Apache 2.0 License ----
+
+    As an exception, if, as a result of your compiling your source code, portions
+    of this Software are embedded into an Object form of such source code, you
+    may redistribute such embedded portions in such Object form without complying
+    with the conditions of Sections 4(a), 4(b) and 4(d) of the License.
+
+    In addition, if you combine or link compiled forms of this Software with
+    software that is licensed under the GPLv2 ("Combined Software") and if a
+    court of competent jurisdiction determines that the patent provision (Section
+    3), the indemnity provision (Section 9) or other Section of the License
+    conflicts with the conditions of the GPLv2, you may retroactively and
+    prospectively choose to deem waived or otherwise exclude such Section(s) of
+    the License, but only in their entirety and only with respect to the Combined
+    Software.
+
+    ==============================================================================

KERNEL_FEATURE_WISHLIST.md

@@ -16,6 +16,8 @@ Somewhat important and an easy sell:
     * also needed by jemalloc for different reasons
     * not needed if the kernel gets first class support for arbitrarily sized
       guard pages and a virtual memory quarantine feature
+    * `MREMAP_DONTUNMAP` is now available but doesn't support expanding the
+      mapping which may be an issue due to VMA merging being unreliable
 
 Fairly infeasible to land but could reduce overhead and extend coverage of
 security features to other code directly using mmap:

LICENSE

@@ -1,4 +1,4 @@
-Copyright (c) 2019 Daniel Micay
+Copyright © 2018-2025 GrapheneOS
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

Makefile

@@ -1,57 +1,60 @@
-CONFIG_NATIVE := true
-CONFIG_CXX_ALLOCATOR := true
-CONFIG_UBSAN := false
-CONFIG_SEAL_METADATA := false
-CONFIG_ZERO_ON_FREE := true
-CONFIG_WRITE_AFTER_FREE_CHECK := true
-CONFIG_SLOT_RANDOMIZE := true
-CONFIG_SLAB_CANARY := true
-CONFIG_SLAB_QUARANTINE_RANDOM_LENGTH := 1
-CONFIG_SLAB_QUARANTINE_QUEUE_LENGTH := 1
-CONFIG_EXTENDED_SIZE_CLASSES := true
-CONFIG_LARGE_SIZE_CLASSES := true
-CONFIG_GUARD_SLABS_INTERVAL := 1
-CONFIG_GUARD_SIZE_DIVISOR := 2
-CONFIG_REGION_QUARANTINE_RANDOM_LENGTH := 128
-CONFIG_REGION_QUARANTINE_QUEUE_LENGTH := 1024
-CONFIG_REGION_QUARANTINE_SKIP_THRESHOLD := 33554432 # 32MiB
-CONFIG_FREE_SLABS_QUARANTINE_RANDOM_LENGTH := 32
-CONFIG_CLASS_REGION_SIZE := 137438953472 # 128GiB
-CONFIG_N_ARENA := 4
-CONFIG_STATS := false
+VARIANT := default
+
+ifneq ($(VARIANT),)
+    CONFIG_FILE := config/$(VARIANT).mk
+    include config/$(VARIANT).mk
+endif
+
+ifeq ($(VARIANT),default)
+    SUFFIX :=
+else
+    SUFFIX := -$(VARIANT)
+endif
+
+OUT := out$(SUFFIX)
 
 define safe_flag
-$(shell $(CC) -E $1 - </dev/null >/dev/null 2>&1 && echo $1 || echo $2)
+$(shell $(CC) $(if $(filter clang%,$(CC)),-Werror=unknown-warning-option) -E $1 - </dev/null >/dev/null 2>&1 && echo $1 || echo $2)
 endef
 
-CPPFLAGS := -D_GNU_SOURCE
-SHARED_FLAGS := -O3 -flto -fPIC -fvisibility=hidden -fno-plt -pipe -Wall -Wextra $(call safe_flag,-Wcast-align=strict) -Wcast-qual -Wwrite-strings
+CPPFLAGS := $(CPPFLAGS) -D_GNU_SOURCE -I include
+SHARED_FLAGS := -pipe -O3 -flto -fPIC -fvisibility=hidden -fno-plt \
+    -fstack-clash-protection $(call safe_flag,-fcf-protection) -fstack-protector-strong \
+    -Wall -Wextra $(call safe_flag,-Wcast-align=strict,-Wcast-align) -Wcast-qual -Wwrite-strings \
+    -Wundef
+
+ifeq ($(CONFIG_WERROR),true)
+    SHARED_FLAGS += -Werror
+endif
 
 ifeq ($(CONFIG_NATIVE),true)
     SHARED_FLAGS += -march=native
 endif
 
-CFLAGS := -std=c11 $(SHARED_FLAGS) -Wmissing-prototypes
-CXXFLAGS := $(call safe_flag,-std=c++17,-std=c++14) $(SHARED_FLAGS)
-LDFLAGS := -Wl,--as-needed,-z,defs,-z,relro,-z,now,-z,nodlopen,-z,text
-TIDY_CHECKS := -checks=bugprone-*,-bugprone-macro-parentheses,cert-*,clang-analyzer-*,readability-*,-readability-inconsistent-declaration-parameter-name,-readability-magic-numbers,-readability-named-parameter,-bugprone-too-small-loop-variable
+ifeq ($(CONFIG_UBSAN),true)
+    SHARED_FLAGS += -fsanitize=undefined -fno-sanitize-recover=undefined
+endif
+
+CFLAGS := $(CFLAGS) -std=c17 $(SHARED_FLAGS) -Wmissing-prototypes -Wstrict-prototypes
+CXXFLAGS := $(CXXFLAGS) -std=c++17 -fsized-deallocation $(SHARED_FLAGS)
+LDFLAGS := $(LDFLAGS) -Wl,-O1,--as-needed,-z,defs,-z,relro,-z,now,-z,nodlopen,-z,text
 
 SOURCES := chacha.c h_malloc.c memory.c pages.c random.c util.c
 OBJECTS := $(SOURCES:.c=.o)
 
 ifeq ($(CONFIG_CXX_ALLOCATOR),true)
+    # make sure LTO is compatible in case CC and CXX don't match (such as clang and g++)
+    CXX := $(CC)
     LDLIBS += -lstdc++
+
     SOURCES += new.cc
     OBJECTS += new.o
 endif
 
-ifeq ($(CONFIG_UBSAN),true)
-    CFLAGS += -fsanitize=undefined
-    CXXFLAGS += -fsanitize=undefined
-endif
+OBJECTS := $(addprefix $(OUT)/,$(OBJECTS))
 
-ifeq ($(CONFIG_SEAL_METADATA),true)
-    CPPFLAGS += -DCONFIG_SEAL_METADATA
+ifeq (,$(filter $(CONFIG_SEAL_METADATA),true false))
+    $(error CONFIG_SEAL_METADATA must be true or false)
 endif
 
 ifeq (,$(filter $(CONFIG_ZERO_ON_FREE),true false))
@@ -82,7 +85,12 @@ ifeq (,$(filter $(CONFIG_STATS),true false))
     $(error CONFIG_STATS must be true or false)
 endif
 
+ifeq (,$(filter $(CONFIG_SELF_INIT),true false))
+    $(error CONFIG_SELF_INIT must be true or false)
+endif
+
 CPPFLAGS += \
+    -DCONFIG_SEAL_METADATA=$(CONFIG_SEAL_METADATA) \
     -DZERO_ON_FREE=$(CONFIG_ZERO_ON_FREE) \
     -DWRITE_AFTER_FREE_CHECK=$(CONFIG_WRITE_AFTER_FREE_CHECK) \
     -DSLOT_RANDOMIZE=$(CONFIG_SLOT_RANDOMIZE) \
@@ -99,23 +107,42 @@ CPPFLAGS += \
     -DFREE_SLABS_QUARANTINE_RANDOM_LENGTH=$(CONFIG_FREE_SLABS_QUARANTINE_RANDOM_LENGTH) \
     -DCONFIG_CLASS_REGION_SIZE=$(CONFIG_CLASS_REGION_SIZE) \
     -DN_ARENA=$(CONFIG_N_ARENA) \
-    -DCONFIG_STATS=$(CONFIG_STATS)
+    -DCONFIG_STATS=$(CONFIG_STATS) \
+    -DCONFIG_SELF_INIT=$(CONFIG_SELF_INIT)
 
-libhardened_malloc.so: $(OBJECTS)
+$(OUT)/libhardened_malloc$(SUFFIX).so: $(OBJECTS) | $(OUT)
 	$(CC) $(CFLAGS) $(LDFLAGS) -shared $^ $(LDLIBS) -o $@
 
-chacha.o: chacha.c chacha.h util.h
-h_malloc.o: h_malloc.c h_malloc.h mutex.h memory.h pages.h random.h util.h
-memory.o: memory.c memory.h util.h
-new.o: new.cc h_malloc.h util.h
-pages.o: pages.c pages.h memory.h util.h
-random.o: random.c random.h chacha.h util.h
-util.o: util.c util.h
+$(OUT):
+	mkdir -p $(OUT)
+
+$(OUT)/chacha.o: chacha.c chacha.h util.h $(CONFIG_FILE) | $(OUT)
+	$(COMPILE.c) $(OUTPUT_OPTION) $<
+$(OUT)/h_malloc.o: h_malloc.c include/h_malloc.h mutex.h memory.h pages.h random.h util.h $(CONFIG_FILE) | $(OUT)
+	$(COMPILE.c) $(OUTPUT_OPTION) $<
+$(OUT)/memory.o: memory.c memory.h util.h $(CONFIG_FILE) | $(OUT)
+	$(COMPILE.c) $(OUTPUT_OPTION) $<
+$(OUT)/new.o: new.cc include/h_malloc.h util.h $(CONFIG_FILE) | $(OUT)
+	$(COMPILE.cc) $(OUTPUT_OPTION) $<
+$(OUT)/pages.o: pages.c pages.h memory.h util.h $(CONFIG_FILE) | $(OUT)
+	$(COMPILE.c) $(OUTPUT_OPTION) $<
+$(OUT)/random.o: random.c random.h chacha.h util.h $(CONFIG_FILE) | $(OUT)
+	$(COMPILE.c) $(OUTPUT_OPTION) $<
+$(OUT)/util.o: util.c util.h $(CONFIG_FILE) | $(OUT)
+	$(COMPILE.c) $(OUTPUT_OPTION) $<
+
+check: tidy
 
 tidy:
-	clang-tidy $(TIDY_CHECKS) $(SOURCES) -- $(CPPFLAGS)
+	clang-tidy --extra-arg=-std=c17 $(filter %.c,$(SOURCES)) -- $(CPPFLAGS)
+	clang-tidy --extra-arg=-std=c++17 $(filter %.cc,$(SOURCES)) -- $(CPPFLAGS)
 
 clean:
-	rm -f libhardened_malloc.so $(OBJECTS)
+	rm -f $(OUT)/libhardened_malloc.so $(OBJECTS)
+	$(MAKE) -C test/ clean
 
-.PHONY: clean tidy
+test: $(OUT)/libhardened_malloc$(SUFFIX).so
+	$(MAKE) -C test/
+	python3 -m unittest discover --start-directory test/
+
+.PHONY: check clean tidy test

README.md

@@ -1,5 +1,30 @@
 # Hardened malloc
 
+* [Introduction](#introduction)
+* [Dependencies](#dependencies)
+* [Testing](#testing)
+    * [Individual Applications](#individual-applications)
+    * [Automated Test Framework](#automated-test-framework)
+* [Compatibility](#compatibility)
+* [OS integration](#os-integration)
+    * [Android-based operating systems](#android-based-operating-systems)
+    * [Traditional Linux-based operating systems](#traditional-linux-based-operating-systems)
+* [Configuration](#configuration)
+* [Core design](#core-design)
+* [Security properties](#security-properties)
+* [Randomness](#randomness)
+* [Size classes](#size-classes)
+* [Scalability](#scalability)
+    * [Small (slab) allocations](#small-slab-allocations)
+        * [Thread caching (or lack thereof)](#thread-caching-or-lack-thereof)
+    * [Large allocations](#large-allocations)
+* [Memory tagging](#memory-tagging)
+* [API extensions](#api-extensions)
+* [Stats](#stats)
+* [System calls](#system-calls)
+
+## Introduction
+
 This is a security-focused general purpose memory allocator providing the
 malloc API along with various extensions. It provides substantial hardening
 against heap corruption vulnerabilities. The security-focused design also leads
@@ -7,7 +32,7 @@ to much less metadata overhead and memory waste from fragmentation than a more
 traditional allocator design. It aims to provide decent overall performance
 with a focus on long-term performance and memory usage rather than allocator
 micro-benchmarks. It offers scalability via a configurable number of entirely
-independently arenas, with the internal locking within arenas further divided
+independent arenas, with the internal locking within arenas further divided
 up per size class.
 
 This project currently supports Bionic (Android), musl and glibc. It may
@@ -20,17 +45,17 @@ and can cover the same use cases.
 This allocator is intended as a successor to a previous implementation based on
 extending OpenBSD malloc with various additional security features. It's still
 heavily based on the OpenBSD malloc design, albeit not on the existing code
-other than reusing the hash table implementation for the time being. The main
-differences in the design are that it is solely focused on hardening rather
-than finding bugs, uses finer-grained size classes along with slab sizes going
-beyond 4k to reduce internal fragmentation, doesn't rely on the kernel having
-fine-grained mmap randomization and only targets 64-bit to make aggressive use
-of the large address space.  There are lots of smaller differences in the
-implementation approach. It incorporates the previous extensions made to
-OpenBSD malloc including adding padding to allocations for canaries (distinct
-from the current OpenBSD malloc canaries), write-after-free detection tied to
-the existing clearing on free, queues alongside the existing randomized arrays
-for quarantining allocations and proper double-free detection for quarantined
+other than reusing the hash table implementation. The main differences in the
+design are that it's solely focused on hardening rather than finding bugs, uses
+finer-grained size classes along with slab sizes going beyond 4k to reduce
+internal fragmentation, doesn't rely on the kernel having fine-grained mmap
+randomization and only targets 64-bit to make aggressive use of the large
+address space. There are lots of smaller differences in the implementation
+approach. It incorporates the previous extensions made to OpenBSD malloc
+including adding padding to allocations for canaries (distinct from the current
+OpenBSD malloc canaries), write-after-free detection tied to the existing
+clearing on free, queues alongside the existing randomized arrays for
+quarantining allocations and proper double-free detection for quarantined
 allocations. The per-size-class memory regions with their own random bases were
 loosely inspired by the size and type-based partitioning in PartitionAlloc. The
 planned changes to OpenBSD malloc ended up being too extensive and invasive so
@@ -40,11 +65,14 @@ used instead as this allocator fundamentally doesn't support that environment.
 
 ## Dependencies
 
-Debian stable determines the most ancient set of supported dependencies:
+Debian stable (currently Debian 12) determines the most ancient set of
+supported dependencies:
 
-* glibc 2.24
-* Linux 4.9
-* Clang 3.8 or GCC 6.3
+* glibc 2.36
+* Linux 6.1
+* Clang 14.0.6 or GCC 12.2.0
+
+For Android, the Linux GKI 5.10, 5.15 and 6.1 branches are supported.
 
 However, using more recent releases is highly recommended. Older versions of
 the dependencies may be compatible at the moment but are not tested and will
@@ -54,18 +82,20 @@ For external malloc replacement with musl, musl 1.1.20 is required. However,
 there will be custom integration offering better performance in the future
 along with other hardening for the C standard library implementation.
 
-For Android, only current generation Android Open Source Project branches will
-be supported, which currently means pie-qpr2-release.
+For Android, only the current generation, actively developed maintenance branch of the Android
+Open Source Project will be supported, which currently means `android15-release`.
 
 ## Testing
 
+### Individual Applications
+
 The `preload.sh` script can be used for testing with dynamically linked
 executables using glibc or musl:
 
     ./preload.sh krita --new-image RGBA,U8,500,500
 
 It can be necessary to substantially increase the `vm.max_map_count` sysctl to
-accomodate the large number of mappings caused by guard slabs and large
+accommodate the large number of mappings caused by guard slabs and large
 allocation guard regions. The number of mappings can also be drastically
 reduced via a significant increase to `CONFIG_GUARD_SLABS_INTERVAL` but the
 feature has a low performance and memory usage cost so that isn't recommended.
@@ -78,6 +108,82 @@ this allocator offers across different size classes. The intention is that this
 will be offered as part of hardened variants of the Bionic and musl C standard
 libraries.
 
+### Automated Test Framework
+
+A collection of simple, automated tests are provided and can be run with the
+make command as follows:
+
+    make test
+
+## Compatibility
+
+OpenSSH 8.1 or higher is required to allow the mprotect `PROT_READ|PROT_WRITE`
+system calls in the seccomp-bpf filter rather than killing the process.
+
+## OS integration
+
+### Android-based operating systems
+
+On GrapheneOS, hardened\_malloc is integrated into the standard C library as
+the standard malloc implementation. Other Android-based operating systems can
+reuse [the integration
+code](https://github.com/GrapheneOS/platform_bionic/commit/20160b81611d6f2acd9ab59241bebeac7cf1d71c)
+to provide it. If desired, jemalloc can be left as a runtime configuration
+option by only conditionally using hardened\_malloc to give users the choice
+between performance and security. However, this reduces security for threat
+models where persistent state is untrusted, i.e. verified boot and attestation
+(see the [attestation sister project](https://attestation.app/about)).
+
+Make sure to raise `vm.max_map_count` substantially too to accommodate the very
+large number of guard pages created by hardened\_malloc. This can be done in
+`init.rc` (`system/core/rootdir/init.rc`) near the other virtual memory
+configuration:
+
+    write /proc/sys/vm/max_map_count 1048576
+
+This is unnecessary if you set `CONFIG_GUARD_SLABS_INTERVAL` to a very large
+value in the build configuration.
+
+### Traditional Linux-based operating systems
+
+On traditional Linux-based operating systems, hardened\_malloc can either be
+integrated into the libc implementation as a replacement for the standard
+malloc implementation or loaded as a dynamic library. Rather than rebuilding
+each executable to be linked against it, it can be added as a preloaded
+library to `/etc/ld.so.preload`. For example, with `libhardened_malloc.so`
+installed to `/usr/local/lib/libhardened_malloc.so`, add that full path as a
+line to the `/etc/ld.so.preload` configuration file:
+
+    /usr/local/lib/libhardened_malloc.so
+
+The format of this configuration file is a whitespace-separated list, so it's
+good practice to put each library on a separate line.
+
+On Debian systems `libhardened_malloc.so` should be installed into `/usr/lib/`
+to avoid preload failures caused by AppArmor profile restrictions.
+
+Using the `LD_PRELOAD` environment variable to load it on a case-by-case basis
+will not work when `AT_SECURE` is set such as with setuid binaries. It's also
+generally not a recommended approach for production usage. The recommendation
+is to enable it globally and make exceptions for performance critical cases by
+running the application in a container / namespace without it enabled.
+
+Make sure to raise `vm.max_map_count` substantially too to accommodate the very
+large number of guard pages created by hardened\_malloc. As an example, in
+`/etc/sysctl.d/hardened_malloc.conf`:
+
+    vm.max_map_count = 1048576
+
+This is unnecessary if you set `CONFIG_GUARD_SLABS_INTERVAL` to a very large
+value in the build configuration.
+
+On arm64, make sure your kernel is configured to use 4k pages since we haven't
+yet added support for 16k and 64k pages. The kernel also has to be configured
+to use 4 level page tables for the full 48 bit address space instead of only
+having a 39 bit address space for the default hardened\_malloc configuration.
+It's possible to reduce the class region size substantially to make a 39 bit
+address space workable but the defaults won't work.
+
 ## Configuration
 
 You can set some configuration options at compile-time via arguments to the
@@ -90,8 +196,45 @@ between portability, performance, memory usage or security. The core design
 choices are not configurable and the allocator remains very security-focused
 even with all the optional features disabled.
 
+The configuration system supports a configuration template system with two
+standard presets: the default configuration (`config/default.mk`) and a light
+configuration (`config/light.mk`). Packagers are strongly encouraged to ship
+both the standard `default` and `light` configuration. You can choose the
+configuration to build using `make VARIANT=light` where `make VARIANT=default`
+is the same as `make`. Non-default configuration templates will build a library
+with the suffix `-variant` such as `libhardened_malloc-light.so` and will use
+an `out-variant` directory instead of `out` for the build.
+
+The `default` configuration template has all normal optional security features
+enabled (just not the niche `CONFIG_SEAL_METADATA`) and is quite aggressive in
+terms of sacrificing performance and memory usage for security. The `light`
+configuration template disables the slab quarantines, write after free check,
+slot randomization and raises the guard slab interval from 1 to 8 but leaves
+zero-on-free and slab canaries enabled. The `light` configuration has solid
+performance and memory usage while still being far more secure than mainstream
+allocators with much better security properties. Disabling zero-on-free would
+gain more performance but doesn't make much difference for small allocations
+without also disabling slab canaries. Slab canaries slightly raise memory use
+and slightly slow down performance but are quite important to mitigate small
+overflows and C string overflows. Disabling slab canaries is not recommended
+in most cases since it would no longer be a strict upgrade over traditional
+allocators with headers on allocations and basic consistency checks for them.
+
+For reduced memory usage at the expense of performance (this will also reduce
+the size of the empty slab caches and quarantines, saving a lot of memory,
+since those are currently based on the size of the largest size class):
+
+    make \
+    N_ARENA=1 \
+    CONFIG_EXTENDED_SIZE_CLASSES=false
+
 The following boolean configuration options are available:
 
+* `CONFIG_WERROR`: `true` (default) or `false` to control whether compiler
+  warnings are treated as errors. This is highly recommended, but it can be
+  disabled to avoid patching the Makefile if a compiler version not tested by
+  the project is being used and has warnings. Investigating these warnings is
+  still recommended and the intention is to always be free of any warnings.
 * `CONFIG_NATIVE`: `true` (default) or `false` to control whether the code is
   optimized for the detected CPU on the host. If this is disabled, setting up a
   custom `-march` higher than the baseline architecture is highly recommended
@@ -104,12 +247,15 @@ The following boolean configuration options are available:
   allocations are zeroed on free, to mitigate use-after-free and uninitialized
   use vulnerabilities along with purging lots of potentially sensitive data
   from the process as soon as possible. This has a performance cost scaling to
-  the size of the allocation, which is usually acceptable.
+  the size of the allocation, which is usually acceptable. This is not relevant
+  to large allocations because the pages are given back to the kernel.
 * `CONFIG_WRITE_AFTER_FREE_CHECK`: `true` (default) or `false` to control
-  sanity checking that new allocations contain zeroed memory. This can detect
-  writes caused by a write-after-free vulnerability and mixes well with the
-  features for making memory reuse randomized / delayed. This has a performance
-  cost scaling to the size of the allocation, which is usually acceptable.
+  sanity checking that new small allocations contain zeroed memory. This can
+  detect writes caused by a write-after-free vulnerability and mixes well with
+  the features for making memory reuse randomized / delayed. This has a
+  performance cost scaling to the size of the allocation, which is usually
+  acceptable. This is not relevant to large allocations because they're always
+  a fresh memory mapping from the kernel.
 * `CONFIG_SLOT_RANDOMIZE`: `true` (default) or `false` to randomize selection
   of free slots within slabs. This has a measurable performance cost and isn't
   one of the important security features, but the cost has been deemed more
@@ -126,84 +272,112 @@ The following boolean configuration options are available:
 * `CONFIG_SEAL_METADATA`: `true` or `false` (default) to control whether Memory
   Protection Keys are used to disable access to all writable allocator state
   outside of the memory allocator code. It's currently disabled by default due
-  to being extremely experimental and a significant performance cost for this
-  use case on current generation hardware, which may become drastically lower
-  in the future. Whether or not this feature is enabled, the metadata is all
-  contained within an isolated memory region with high entropy random guard
-  regions around it.
+  to a significant performance cost for this use case on current generation
+  hardware, which may become drastically lower in the future. Whether or not
+  this feature is enabled, the metadata is all contained within an isolated
+  memory region with high entropy random guard regions around it.
 
-The following integer configuration options are available. Proper sanity checks
-for the chosen values are not written yet, so use them at your own peril:
+The following integer configuration options are available:
 
 * `CONFIG_SLAB_QUARANTINE_RANDOM_LENGTH`: `1` (default) to control the number
   of slots in the random array used to randomize reuse for small memory
-  allocations. This sets the length for the largest size class (currently
-  16384) and the quarantine length for smaller size classes is scaled to match
-  the total memory of the quarantined allocations (1 becomes 1024 for 16 byte
-  allocations).
+  allocations. This sets the length for the largest size class (either 16kiB
+  or 128kiB based on `CONFIG_EXTENDED_SIZE_CLASSES`) and the quarantine length
+  for smaller size classes is scaled to match the total memory of the
+  quarantined allocations (1 becomes 1024 for 16 byte allocations with 16kiB
+  as the largest size class, or 8192 with 128kiB as the largest).
 * `CONFIG_SLAB_QUARANTINE_QUEUE_LENGTH`: `1` (default) to control the number of
   slots in the queue used to delay reuse for small memory allocations. This
-  sets the length for the largest size class (currently 16384) and the
-  quarantine length for smaller size classes is scaled to match the total
-  memory of the quarantined allocations (1 becomes 1024 for 16 byte
-  allocations).
+  sets the length for the largest size class (either 16kiB or 128kiB based on
+  `CONFIG_EXTENDED_SIZE_CLASSES`) and the quarantine length for smaller size
+  classes is scaled to match the total memory of the quarantined allocations (1
+  becomes 1024 for 16 byte allocations with 16kiB as the largest size class, or
+  8192 with 128kiB as the largest).
 * `CONFIG_GUARD_SLABS_INTERVAL`: `1` (default) to control the number of slabs
-  before a slab is skipped and left as an unused memory protected guard slab
+  before a slab is skipped and left as an unused memory protected guard slab.
+  The default of `1` leaves a guard slab between every slab. This feature does
+  not have a *direct* performance cost, but it makes the address space usage
+  sparser which can indirectly hurt performance. The kernel also needs to track
+  a lot more memory mappings, which uses a bit of extra memory and slows down
+  memory mapping and memory protection changes in the process. The kernel uses
+  O(log n) algorithms for this and system calls are already fairly slow anyway,
+  so having many extra mappings doesn't usually add up to a significant cost.
 * `CONFIG_GUARD_SIZE_DIVISOR`: `2` (default) to control the maximum size of the
   guard regions placed on both sides of large memory allocations, relative to
-  the usable size of the memory allocation
-* `CONFIG_REGION_QUARANTINE_RANDOM_LENGTH`: `128` (default) to control the
+  the usable size of the memory allocation.
+* `CONFIG_REGION_QUARANTINE_RANDOM_LENGTH`: `256` (default) to control the
   number of slots in the random array used to randomize region reuse for large
-  memory allocations
+  memory allocations.
 * `CONFIG_REGION_QUARANTINE_QUEUE_LENGTH`: `1024` (default) to control the
   number of slots in the queue used to delay region reuse for large memory
-  allocations
+  allocations.
 * `CONFIG_REGION_QUARANTINE_SKIP_THRESHOLD`: `33554432` (default) to control
-  the size threshold where large allocations will not be quarantined
+  the size threshold where large allocations will not be quarantined.
 * `CONFIG_FREE_SLABS_QUARANTINE_RANDOM_LENGTH`: `32` (default) to control the
-  number of slots in the random array used to randomize free slab reuse
+  number of slots in the random array used to randomize free slab reuse.
 * `CONFIG_CLASS_REGION_SIZE`: `34359738368` (default) to control the size of
-  the size class regions
-* `CONFIG_N_ARENA`: `1` (default) to control the number of arenas
+  the size class regions.
+* `CONFIG_N_ARENA`: `4` (default) to control the number of arenas
 * `CONFIG_STATS`: `false` (default) to control whether stats on allocation /
-  deallocation count and active allocations are tracked. This is currently only
-  exposed via the mallinfo APIs on Android.
+  deallocation count and active allocations are tracked. See the [section on
+  stats](#stats) for more details.
 * `CONFIG_EXTENDED_SIZE_CLASSES`: `true` (default) to control whether small
-  size class go up to 64k instead of the minimum requirement for avoiding
-  memory waste of 16k. The option to extend it even further will be offered in
-  the future when better support for larger slab allocations is added.
+  size class go up to 128kiB instead of the minimum requirement for avoiding
+  memory waste of 16kiB. The option to extend it even further will be offered
+  in the future when better support for larger slab allocations is added. See
+  the [section on size classes](#size-classes) below for details.
 * `CONFIG_LARGE_SIZE_CLASSES`: `true` (default) to control whether large
   allocations use the slab allocation size class scheme instead of page size
-  granularity (see the section on size classes below)
+  granularity. See the [section on size classes](#size-classes) below for
+  details.
 
 There will be more control over enabled features in the future along with
 control over fairly arbitrarily chosen values like the size of empty slab
 caches (making them smaller improves security and reduces memory usage while
 larger caches can substantially improves performance).
 
-## Basic design
+## Core design
 
-The current design is very simple and will become a bit more sophisticated as
-the basic features are completed and the implementation is hardened and
-optimized. The allocator is exclusive to 64-bit platforms in order to take full
-advantage of the abundant address space without being constrained by needing to
-keep the design compatible with 32-bit.
+The core design of the allocator is very simple / minimalist. The allocator is
+exclusive to 64-bit platforms in order to take full advantage of the abundant
+address space without being constrained by needing to keep the design
+compatible with 32-bit.
+
+The mutable allocator state is entirely located within a dedicated metadata
+region, and the allocator is designed around this approach for both small
+(slab) allocations and large allocations. This provides reliable, deterministic
+protections against invalid free including double frees, and protects metadata
+from attackers. Traditional allocator exploitation techniques do not work with
+the hardened\_malloc implementation.
 
 Small allocations are always located in a large memory region reserved for slab
-allocations. It can be determined that an allocation is one of the small size
-classes from the address range. Each small size class has a separate reserved
-region within the larger region, and the size of a small allocation can simply
-be determined from the range. Each small size class has a separate out-of-line
-metadata array outside of the overall allocation region, with the index of the
-metadata struct within the array mapping to the index of the slab within the
-dedicated size class region. Slabs are a multiple of the page size and are
-page aligned. The entire small size class region starts out memory protected
-and becomes readable / writable as it gets allocated, with idle slabs beyond
-the cache limit having their pages dropped and the memory protected again.
+allocations. On free, it can be determined that an allocation is one of the
+small size classes from the address range. If arenas are enabled, the arena is
+also determined from the address range as each arena has a dedicated sub-region
+in the slab allocation region. Arenas provide totally independent slab
+allocators with their own allocator state and no coordination between them.
+Once the base region is determined (simply the slab allocation region as a
+whole without any arenas enabled), the size class is determined from the
+address range too, since it's divided up into a sub-region for each size class.
+There's a top level slab allocation region, divided up into arenas, with each
+of those divided up into size class regions. The size class regions each have a
+random base within a large guard region. Once the size class is determined, the
+slab size is known, and the index of the slab is calculated and used to obtain
+the slab metadata for the slab from the slab metadata array. Finally, the index
+of the slot within the slab provides the index of the bit tracking the slot in
+the bitmap. Every slab allocation slot has a dedicated bit in a bitmap tracking
+whether it's free, along with a separate bitmap for tracking allocations in the
+quarantine. The slab metadata entries in the array have intrusive lists
+threaded through them to track partial slabs (partially filled, and these are
+the first choice for allocation), empty slabs (limited amount of cached free
+memory) and free slabs (purged / memory protected).
 
 Large allocations are tracked via a global hash table mapping their address to
-their size and guard size. They're simply memory mappings and get mapped on
-allocation and then unmapped on free.
+their size and random guard size. They're simply memory mappings and get mapped
+on allocation and then unmapped on free. Large allocations are the only dynamic
+memory mappings made by the allocator, since the address space for allocator
+state (including both small / large allocation metadata) and slab allocations
+is statically reserved.
 
 This allocator is aimed at production usage, not aiding with finding and fixing
 memory corruption bugs for software development. It does find many latent bugs
@@ -268,6 +442,7 @@ was a bit less important and if a core goal was finding latent bugs.
 * Slab allocations are zeroed on free
 * Detection of write-after-free for slab allocations by verifying zero filling
   is intact at allocation time
+* Delayed free via a combination of FIFO and randomization for slab allocations
 * Large allocations are purged and memory protected on free with the memory
   mapping kept reserved in a quarantine to detect use-after-free
     * The quarantine is primarily based on a FIFO ring buffer, with the oldest
@@ -278,7 +453,6 @@ was a bit less important and if a core goal was finding latent bugs.
       of the quarantine
 * Memory in fresh allocations is consistently zeroed due to it either being
   fresh pages or zeroed on free after previous usage
-* Delayed free via a combination of FIFO and randomization for slab allocations
 * Random canaries placed after each slab allocation to *absorb*
   and then later detect overflows/underflows
     * High entropy per-slab random values
@@ -287,8 +461,9 @@ was a bit less important and if a core goal was finding latent bugs.
   size class regions interspersed with guard pages
 * Zero size allocations are a dedicated size class with the entire region
   remaining non-readable and non-writable
-* Extension for retrieving the size of allocations with fallback
-  to a sentinel for pointers not managed by the allocator
+* Extension for retrieving the size of allocations with fallback to a sentinel
+  for pointers not managed by the allocator [in-progress, full implementation
+  needs to be ported from the previous OpenBSD malloc-based allocator]
     * Can also return accurate values for pointers *within* small allocations
     * The same applies to pointers within the first page of large allocations,
       otherwise it currently has to return a sentinel
@@ -298,38 +473,39 @@ was a bit less important and if a core goal was finding latent bugs.
     * Errors other than ENOMEM from mmap, munmap, mprotect and mremap treated
       as fatal, which can help to detect memory management gone wrong elsewhere
       in the process.
-* [future] Memory tagging for slab allocations via MTE on ARMv8.5+
+* Memory tagging for slab allocations via MTE on ARMv8.5+
     * random memory tags as the baseline, providing probabilistic protection
       against various forms of memory corruption
     * dedicated tag for free slots, set on free, for deterministic protection
       against accessing freed memory
-    * store previous random tag within freed slab allocations, and increment it
-      to get the next tag for that slot to provide deterministic use-after-free
-      detection through multiple cycles of memory reuse
     * guarantee distinct tags for adjacent memory allocations by incrementing
       past matching values for deterministic detection of linear overflows
+    * [future] store previous random tag and increment it to get the next tag
+      for that slot to provide deterministic use-after-free detection through
+      multiple cycles of memory reuse
 
 ## Randomness
 
 The current implementation of random number generation for randomization-based
 mitigations is based on generating a keystream from a stream cipher (ChaCha8)
-in small chunks. A separate CSPRNG is used for each small size class, large
-allocations, etc. in order to fit into the existing fine-grained locking model
-without needing to waste memory per thread by having the CSPRNG state in Thread
-Local Storage. Similarly, it's protected via the same approach taken for the
-rest of the metadata. The stream cipher is regularly reseeded from the OS to
-provide backtracking and prediction resistance with a negligible cost. The
-reseed interval simply needs to be adjusted to the point that it stops
-registering as having any significant performance impact. The performance
-impact on recent Linux kernels is primarily from the high cost of system calls
-and locking since the implementation is quite efficient (ChaCha20), especially
-for just generating the key and nonce for another stream cipher (ChaCha8).
+in small chunks. Separate CSPRNGs are used for each small size class in each
+arena, large allocations and initialization in order to fit into the
+fine-grained locking model without needing to waste memory per thread by
+having the CSPRNG state in Thread Local Storage. Similarly, it's protected via
+the same approach taken for the rest of the metadata. The stream cipher is
+regularly reseeded from the OS to provide backtracking and prediction
+resistance with a negligible cost. The reseed interval simply needs to be
+adjusted to the point that it stops registering as having any significant
+performance impact. The performance impact on recent Linux kernels is
+primarily from the high cost of system calls and locking since the
+implementation is quite efficient (ChaCha20), especially for just generating
+the key and nonce for another stream cipher (ChaCha8).
 
 ChaCha8 is a great fit because it's extremely fast across platforms without
 relying on hardware support or complex platform-specific code. The security
 margins of ChaCha20 would be completely overkill for the use case. Using
 ChaCha8 avoids needing to resort to a non-cryptographically secure PRNG or
-something without a lot of scrunity. The current implementation is simply the
+something without a lot of scrutiny. The current implementation is simply the
 reference implementation of ChaCha8 converted into a pure keystream by ripping
 out the XOR of the message into the keystream.
 
@@ -367,41 +543,41 @@ preliminary set of values.
 | size class | worst case internal fragmentation | slab slots | slab size | internal fragmentation for slabs |
 | - | - | - | - | - |
 | 16 | 93.75% | 256 | 4096 | 0.0% |
-| 32 | 46.875% | 128 | 4096 | 0.0% |
+| 32 | 46.88% | 128 | 4096 | 0.0% |
 | 48 | 31.25% | 85 | 4096 | 0.390625% |
-| 64 | 23.4375% | 64 | 4096 | 0.0% |
+| 64 | 23.44% | 64 | 4096 | 0.0% |
 | 80 | 18.75% | 51 | 4096 | 0.390625% |
-| 96 | 15.625% | 42 | 4096 | 1.5625% |
-| 112 | 13.392857142857139% | 36 | 4096 | 1.5625% |
-| 128 | 11.71875% | 64 | 8192 | 0.0% |
-| 160 | 19.375% | 51 | 8192 | 0.390625% |
-| 192 | 16.145833333333343% | 64 | 12288 | 0.0% |
-| 224 | 13.839285714285708% | 54 | 12288 | 1.5625% |
-| 256 | 12.109375% | 64 | 16384 | 0.0% |
-| 320 | 19.6875% | 64 | 20480 | 0.0% |
-| 384 | 16.40625% | 64 | 24576 | 0.0% |
-| 448 | 14.0625% | 64 | 28672 | 0.0% |
-| 512 | 12.3046875% | 64 | 32768 | 0.0% |
-| 640 | 19.84375% | 64 | 40960 | 0.0% |
-| 768 | 16.536458333333343% | 64 | 49152 | 0.0% |
-| 896 | 14.174107142857139% | 64 | 57344 | 0.0% |
-| 1024 | 12.40234375% | 64 | 65536 | 0.0% |
-| 1280 | 19.921875% | 16 | 20480 | 0.0% |
-| 1536 | 16.6015625% | 16 | 24576 | 0.0% |
-| 1792 | 14.229910714285708% | 16 | 28672 | 0.0% |
-| 2048 | 12.451171875% | 16 | 32768 | 0.0% |
-| 2560 | 19.9609375% | 8 | 20480 | 0.0% |
-| 3072 | 16.634114583333343% | 8 | 24576 | 0.0% |
-| 3584 | 14.2578125% | 8 | 28672 | 0.0% |
-| 4096 | 12.4755859375% | 8 | 32768 | 0.0% |
-| 5120 | 19.98046875% | 8 | 40960 | 0.0% |
-| 6144 | 16.650390625% | 8 | 49152 | 0.0% |
-| 7168 | 14.271763392857139% | 8 | 57344 | 0.0% |
-| 8192 | 12.48779296875% | 8 | 65536 | 0.0% |
-| 10240 | 19.990234375% | 6 | 61440 | 0.0% |
-| 12288 | 16.658528645833343% | 5 | 61440 | 0.0% |
-| 14336 | 14.278738839285708% | 4 | 57344 | 0.0% |
-| 16384 | 12.493896484375% | 4 | 65536 | 0.0% |
+| 96 | 15.62% | 42 | 4096 | 1.5625% |
+| 112 | 13.39% | 36 | 4096 | 1.5625% |
+| 128 | 11.72% | 64 | 8192 | 0.0% |
+| 160 | 19.38% | 51 | 8192 | 0.390625% |
+| 192 | 16.15% | 64 | 12288 | 0.0% |
+| 224 | 13.84% | 54 | 12288 | 1.5625% |
+| 256 | 12.11% | 64 | 16384 | 0.0% |
+| 320 | 19.69% | 64 | 20480 | 0.0% |
+| 384 | 16.41% | 64 | 24576 | 0.0% |
+| 448 | 14.06% | 64 | 28672 | 0.0% |
+| 512 | 12.3% | 64 | 32768 | 0.0% |
+| 640 | 19.84% | 64 | 40960 | 0.0% |
+| 768 | 16.54% | 64 | 49152 | 0.0% |
+| 896 | 14.17% | 64 | 57344 | 0.0% |
+| 1024 | 12.4% | 64 | 65536 | 0.0% |
+| 1280 | 19.92% | 16 | 20480 | 0.0% |
+| 1536 | 16.6% | 16 | 24576 | 0.0% |
+| 1792 | 14.23% | 16 | 28672 | 0.0% |
+| 2048 | 12.45% | 16 | 32768 | 0.0% |
+| 2560 | 19.96% | 8 | 20480 | 0.0% |
+| 3072 | 16.63% | 8 | 24576 | 0.0% |
+| 3584 | 14.26% | 8 | 28672 | 0.0% |
+| 4096 | 12.48% | 8 | 32768 | 0.0% |
+| 5120 | 19.98% | 8 | 40960 | 0.0% |
+| 6144 | 16.65% | 8 | 49152 | 0.0% |
+| 7168 | 14.27% | 8 | 57344 | 0.0% |
+| 8192 | 12.49% | 8 | 65536 | 0.0% |
+| 10240 | 19.99% | 6 | 61440 | 0.0% |
+| 12288 | 16.66% | 5 | 61440 | 0.0% |
+| 14336 | 14.28% | 4 | 57344 | 0.0% |
+| 16384 | 12.49% | 4 | 65536 | 0.0% |
 
 The slab allocation size classes end at 16384 since that's the final size for
 2048 byte spacing and the next spacing class matches the page size of 4096
@@ -423,18 +599,18 @@ retaining the isolation.
 
 | size class | worst case internal fragmentation | slab slots | slab size | internal fragmentation for slabs |
 | - | - | - | - | - |
-| 20480 | 19.9951171875% | 2 | 40960 | 0.0% |
-| 24576 | 16.66259765625% | 2 | 49152 | 0.0% |
-| 28672 | 14.2822265625% | 2 | 57344 | 0.0% |
-| 32768 | 12.4969482421875% | 2 | 65536 | 0.0% |
-| 40960 | 19.99755859375% | 1 | 40960 | 0.0% |
-| 49152 | 16.664632161458343% | 1 | 49152 | 0.0% |
-| 57344 | 14.283970424107139% | 1 | 57344 | 0.0% |
-| 65536 | 12.49847412109375% | 1 | 65536 | 0.0% |
-| 81920 | 19.998779296875% | 1 | 81920 | 0.0% |
-| 98304 | 16.6656494140625% | 1 | 98304 | 0.0% |
-| 114688 | 14.284842354910708% | 1 | 114688 | 0.0% |
-| 131072 | 12.499237060546875% | 1 | 131072 | 0.0% |
+| 20480 | 20.0% | 1 | 20480 | 0.0% |
+| 24576 | 16.66% | 1 | 24576 | 0.0% |
+| 28672 | 14.28% | 1 | 28672 | 0.0% |
+| 32768 | 12.5% | 1 | 32768 | 0.0% |
+| 40960 | 20.0% | 1 | 40960 | 0.0% |
+| 49152 | 16.66% | 1 | 49152 | 0.0% |
+| 57344 | 14.28% | 1 | 57344 | 0.0% |
+| 65536 | 12.5% | 1 | 65536 | 0.0% |
+| 81920 | 20.0% | 1 | 81920 | 0.0% |
+| 98304 | 16.67% | 1 | 98304 | 0.0% |
+| 114688 | 14.28% | 1 | 114688 | 0.0% |
+| 131072 | 12.5% | 1 | 131072 | 0.0% |
 
 The `CONFIG_LARGE_SIZE_CLASSES` option controls whether large allocations use
 the same size class scheme providing 4 size classes for every doubling of size.
@@ -461,7 +637,7 @@ to finding the per-size-class metadata. The part that's still open to different
 design choices is how arenas are assigned to threads. One approach is
 statically assigning arenas via round-robin like the standard jemalloc
 implementation, or statically assigning to a random arena which is essentially
-the current implementation.  Another option is dynamic load balancing via a
+the current implementation. Another option is dynamic load balancing via a
 heuristic like `sched_getcpu` for per-CPU arenas, which would offer better
 performance than randomly choosing an arena each time while being more
 predictable for an attacker. There are actually some security benefits from
@@ -472,7 +648,7 @@ varying usage of size classes.
 When there's substantial allocation or deallocation pressure, the allocator
 does end up calling into the kernel to purge / protect unused slabs by
 replacing them with fresh `PROT_NONE` regions along with unprotecting slabs
-when partially filled and cached empty slabs are depleted.  There will be
+when partially filled and cached empty slabs are depleted. There will be
 configuration over the amount of cached empty slabs, but it's not entirely a
 performance vs. memory trade-off since memory protecting unused slabs is a nice
 opportunistic boost to security. However, it's not really part of the core
@@ -548,77 +724,46 @@ freeing as there would be if the kernel supported these features directly.
 
 ## Memory tagging
 
-Integrating extensive support for ARMv8.5 memory tagging is planned and this
-section will be expanded cover the details on the chosen design. The approach
-for slab allocations is currently covered, but it can also be used for the
-allocator metadata region and large allocations.
+Random tags are set for all slab allocations when allocated, with 4 excluded values:
 
-Memory allocations are already always multiples of naturally aligned 16 byte
-units, so memory tags are a natural fit into a malloc implementation due to the
-16 byte alignment requirement. The only extra memory consumption will come from
-the hardware supported storage for the tag values (4 bits per 16 bytes).
+1. the reserved `0` tag
+2. the previous tag used for the slot
+3. the current (or previous) tag used for the slot to the left
+4. the current (or previous) tag used for the slot to the right
 
-The baseline policy will be to generate random tags for each slab allocation
-slot on first use. The highest value will be reserved for marking freed memory
-allocations to detect any accesses to freed memory so it won't be part of the
-generated range. Adjacent slots will be guaranteed to have distinct memory tags
-in order to guarantee that linear overflows are detected. There are a few ways
-of implementing this and it will end up depending on the performance costs of
-different approaches. If there's an efficient way to fetch the adjacent tag
-values without wasting extra memory, it will be possible to check for them and
-skip them either by generating a new random value in a loop or incrementing
-past them since the tiny bit of bias wouldn't matter. Another approach would be
-alternating odd and even tag values but that would substantially reduce the
-overall randomness of the tags and there's very little entropy from the start.
+When a slab allocation is freed, the reserved `0` tag is set for the slot.
+Slab allocation slots are cleared before reuse when memory tagging is enabled.
 
-Once a slab allocation has been freed, the tag will be set to the reserved
-value for free memory and the previous tag value will be stored inside the
-allocation itself. The next time the slot is allocated, the chosen tag value
-will be the previous value incremented by one to provide use-after-free
-detection between generations of allocations. The stored tag will be wiped
-before retagging the memory, to avoid leaking it and as part of preserving the
-security property of newly allocated memory being zeroed due to zero-on-free.
-It will eventually wrap all the way around, but this ends up providing a strong
-guarantee for many allocation cycles due to the combination of 4 bit tags with
-the FIFO quarantine feature providing delayed free. It also benefits from
-random slot allocation and the randomized portion of delayed free, which result
-in a further delay along with preventing a deterministic bypass by forcing a
-reuse after a certain number of allocation cycles. Similarly to the initial tag
-generation, tag values for adjacent allocations will be skipped by incrementing
-past them.
+This ensures the following properties:
 
-For example, consider this slab of allocations that are not yet used with 16
-representing the tag for free memory. For the sake of simplicity, there will be
-no quarantine or other slabs for this example:
+- Linear overflows are deterministically detected.
+- Use-after-free are deterministically detected until the freed slot goes through
+  both the random and FIFO quarantines, gets allocated again, goes through both
+  quarantines again and then finally gets allocated again for a 2nd time.
+- Since the default `0` tag is reserved, untagged pointers can't access slab
+  allocations and vice versa.
 
-    | 16 | 16 | 16 | 16 | 16 | 16 |
+Slab allocations are done in a statically reserved region for each size class
+and all metadata is in a statically reserved region, so interactions between
+different uses of the same address space is not applicable.
 
-Three slots are randomly chosen for allocations, with random tags assigned (2,
-15, 7) since these slots haven't ever been used and don't have saved values:
+Large allocations beyond the largest slab allocation size class (128k by
+default) are guaranteed to have randomly sized guard regions to the left and
+right. Random and FIFO address space quarantines provide use-after-free
+detection. We need to test whether the cost of random tags is acceptable to enabled them by default,
+since they would be useful for:
 
-    | 16 | 2  | 16 | 15 |  7 | 16 |
+- probabilistic detection of overflows
+- probabilistic detection of use-after-free once the address space is
+  out of the quarantine and reused for another allocation
+- deterministic detection of use-after-free for reuse by another allocator.
 
-The 2nd allocation slot is freed, and is set back to the tag for free memory
-(16), but with the previous tag value stored in the freed space:
-
-    | 16 | 16 | 16 | 7  | 15 | 16 |
-
-The first slot is allocated for the first time, receiving the random value 3:
-
-    | 3  | 16 | 16 | 7  | 15 | 16 |
-
-The 2nd slot is randomly chosen again, so the previous tag (2) is retrieved and
-incremented to 3 as part of the use-after-free mitigation. An adjacent
-allocation already uses the tag 3, so the tag is further incremented to 4 (it
-would be incremented to 5 if one of the adjacent tags was 4):
-
-    | 3  | 4  | 16 | 7  | 15 | 16 |
-
-The last slot is randomly chosen for the next alocation, and is assigned the
-random value 15. However, it's placed next to an allocation with the tag 15 so
-the tag is incremented and wraps around to 0:
-
-    | 3  | 4  | 16 | 7  | 15 | 0 |
+When memory tagging is enabled, checking for write-after-free at allocation
+time and checking canaries are both disabled. Canaries will be more thoroughly
+disabled when using memory tagging in the future, but Android currently has
+[very dynamic memory tagging support](https://source.android.com/docs/security/test/memory-safety/arm-mte)
+where it can be disabled at any time which creates a barrier to optimizing
+by disabling redundant features.
 
 ## API extensions
 
@@ -647,6 +792,183 @@ this implementation, it retrieves an upper bound on the size for small memory
 allocations based on calculating the size class region. This function is safe
 to use from signal handlers already.
 
+## Stats
+
+If stats are enabled, hardened\_malloc keeps tracks allocator statistics in
+order to provide implementations of `mallinfo` and `malloc_info`.
+
+On Android, `mallinfo` is used for [mallinfo-based garbage collection
+triggering](https://developer.android.com/preview/features#mallinfo) so
+hardened\_malloc enables `CONFIG_STATS` by default. The `malloc_info`
+implementation on Android is the standard one in Bionic, with the information
+provided to Bionic via Android's internal extended `mallinfo` API with support
+for arenas and size class bins. This means the `malloc_info` output is fully
+compatible, including still having `jemalloc-1` as the version of the data
+format to retain compatibility with existing tooling.
+
+On non-Android Linux, `mallinfo` has zeroed fields even with `CONFIG_STATS`
+enabled because glibc `mallinfo` is inherently broken. It defines the fields as
+`int` instead of `size_t`, resulting in undefined signed overflows. It also
+misuses the fields and provides a strange, idiosyncratic set of values rather
+than following the SVID/XPG `mallinfo` definition. The `malloc_info` function
+is still provided, with a similar format as what Android uses, with tweaks for
+hardened\_malloc and the version set to `hardened_malloc-1`. The data format
+may be changed in the future.
+
+As an example, consider the following program from the hardened\_malloc tests:
+
+```c
+#include <pthread.h>
+
+#include <malloc.h>
+
+__attribute__((optimize(0)))
+void leak_memory(void) {
+    (void)malloc(1024 * 1024 * 1024);
+    (void)malloc(16);
+    (void)malloc(32);
+    (void)malloc(4096);
+}
+
+void *do_work(void *p) {
+    leak_memory();
+    return NULL;
+}
+
+int main(void) {
+    pthread_t thread[4];
+    for (int i = 0; i < 4; i++) {
+        pthread_create(&thread[i], NULL, do_work, NULL);
+    }
+    for (int i = 0; i < 4; i++) {
+        pthread_join(thread[i], NULL);
+    }
+
+    malloc_info(0, stdout);
+}
+```
+
+This produces the following output when piped through `xmllint --format -`:
+
+```xml
+<?xml version="1.0"?>
+<malloc version="hardened_malloc-1">
+  <heap nr="0">
+    <bin nr="2" size="32">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>32</allocated>
+    </bin>
+    <bin nr="3" size="48">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>48</allocated>
+    </bin>
+    <bin nr="13" size="320">
+      <nmalloc>4</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>20480</slab_allocated>
+      <allocated>1280</allocated>
+    </bin>
+    <bin nr="29" size="5120">
+      <nmalloc>2</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>40960</slab_allocated>
+      <allocated>10240</allocated>
+    </bin>
+    <bin nr="45" size="81920">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>81920</slab_allocated>
+      <allocated>81920</allocated>
+    </bin>
+  </heap>
+  <heap nr="1">
+    <bin nr="2" size="32">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>32</allocated>
+    </bin>
+    <bin nr="3" size="48">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>48</allocated>
+    </bin>
+    <bin nr="29" size="5120">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>40960</slab_allocated>
+      <allocated>5120</allocated>
+    </bin>
+  </heap>
+  <heap nr="2">
+    <bin nr="2" size="32">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>32</allocated>
+    </bin>
+    <bin nr="3" size="48">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>48</allocated>
+    </bin>
+    <bin nr="29" size="5120">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>40960</slab_allocated>
+      <allocated>5120</allocated>
+    </bin>
+  </heap>
+  <heap nr="3">
+    <bin nr="2" size="32">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>32</allocated>
+    </bin>
+    <bin nr="3" size="48">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>4096</slab_allocated>
+      <allocated>48</allocated>
+    </bin>
+    <bin nr="29" size="5120">
+      <nmalloc>1</nmalloc>
+      <ndalloc>0</ndalloc>
+      <slab_allocated>40960</slab_allocated>
+      <allocated>5120</allocated>
+    </bin>
+  </heap>
+  <heap nr="4">
+    <allocated_large>4294967296</allocated_large>
+  </heap>
+</malloc>
+```
+
+The heap entries correspond to the arenas. Unlike jemalloc, hardened\_malloc
+doesn't handle large allocations within the arenas, so it presents those in the
+`malloc_info` statistics as a separate arena dedicated to large allocations.
+For example, with 4 arenas enabled, there will be a 5th arena in the statistics
+for the large allocations.
+
+The `nmalloc` / `ndalloc` fields are 64-bit integers tracking allocation and
+deallocation count. These are defined as wrapping on overflow, per the jemalloc
+implementation.
+
+See the [section on size classes](#size-classes) to map the size class bin
+number to the corresponding size class. The bin index begins at 0, mapping to
+the 0 byte size class, followed by 1 for the 16 bytes, 2 for 32 bytes, etc. and
+large allocations are treated as one group.
+
+When stats aren't enabled, the `malloc_info` output will be an empty `malloc`
+element.
+
 ## System calls
 
 This is intended to aid with creating system call whitelists via seccomp-bpf
@@ -665,6 +987,7 @@ System calls used by all build configurations:
 * `mremap(old, old_size, new_size, MREMAP_MAYMOVE|MREMAP_FIXED, new)`
 * `munmap`
 * `write(STDERR_FILENO, buf, len)` (before aborting due to memory corruption)
+* `madvise(ptr, size, MADV_DONTNEED)`
 
 The main distinction from a typical malloc implementation is the use of
 getrandom. A common compatibility issue is that existing system call whitelists
@@ -677,7 +1000,6 @@ Additional system calls when `CONFIG_SEAL_METADATA=true` is set:
 * `pkey_alloc`
 * `pkey_mprotect` instead of `mprotect` with an additional `pkey` parameter,
   but otherwise the same (regular `mprotect` is never called)
-* `uname` (to detect old buggy kernel versions)
 
 Additional system calls for Android builds with `LABEL_MEMORY`:
 

androidtest/Android.bp

@@ -0,0 +1,25 @@
+java_test_host {
+    name: "HMallocTest",
+    srcs: [
+        "src/**/*.java",
+    ],
+
+    libs: [
+        "tradefed",
+        "compatibility-tradefed",
+        "compatibility-host-util",
+    ],
+
+    static_libs: [
+        "cts-host-utils",
+        "frameworks-base-hostutils",
+    ],
+
+    test_suites: [
+        "general-tests",
+    ],
+
+    data_device_bins_64: [
+        "memtag_test",
+    ],
+}

androidtest/AndroidTest.xml

@@ -0,0 +1,13 @@
+<?xml version="1.0" encoding="utf-8"?>
+<configuration description="hardened_malloc test">
+
+    <target_preparer class="com.android.compatibility.common.tradefed.targetprep.FilePusher">
+        <option name="cleanup" value="true" />
+        <option name="push" value="memtag_test->/data/local/tmp/memtag_test" />
+    </target_preparer>
+
+    <test class="com.android.compatibility.common.tradefed.testtype.JarHostTest" >
+        <option name="jar" value="HMallocTest.jar" />
+    </test>
+
+</configuration>

androidtest/memtag/Android.bp

@@ -0,0 +1,17 @@
+cc_test {
+    name: "memtag_test",
+    srcs: ["memtag_test.cc"],
+    cflags: [
+        "-Wall",
+        "-Werror",
+        "-Wextra",
+        "-O0",
+        "-march=armv9-a+memtag",
+    ],
+
+    compile_multilib: "64",
+
+    sanitize: {
+        memtag_heap: true,
+    },
+}

androidtest/memtag/memtag_test.cc

@@ -0,0 +1,351 @@
+// needed to uncondionally enable assertions
+#undef NDEBUG
+#include <assert.h>
+#include <malloc.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <sys/mman.h>
+#include <sys/utsname.h>
+#include <unistd.h>
+
+#include <map>
+#include <set>
+#include <string>
+#include <unordered_map>
+
+#include "../../arm_mte.h"
+
+using namespace std;
+
+using u8 = uint8_t;
+using uptr = uintptr_t;
+using u64 = uint64_t;
+
+const size_t DEFAULT_ALLOC_SIZE = 8;
+const size_t CANARY_SIZE = 8;
+
+void do_context_switch() {
+    utsname s;
+    uname(&s);
+}
+
+u8 get_pointer_tag(void *ptr) {
+    return (((uptr) ptr) >> 56) & 0xf;
+}
+
+void *untag_pointer(void *ptr) {
+    const uintptr_t mask = UINTPTR_MAX >> 8;
+    return (void *) ((uintptr_t) ptr & mask);
+}
+
+void *set_pointer_tag(void *ptr, u8 tag) {
+    return (void *) (((uintptr_t) tag << 56) | (uintptr_t) untag_pointer(ptr));
+}
+
+// This test checks that slab slot allocation uses tag that is distint from tags of its neighbors
+// and from the tag of the previous allocation that used the same slot
+void tag_distinctness() {
+    // tag 0 is reserved
+    const int min_tag = 1;
+    const int max_tag = 0xf;
+
+    struct SizeClass {
+        int size;
+        int slot_cnt;
+    };
+
+    // values from size_classes[] and size_class_slots[] in h_malloc.c
+    SizeClass size_classes[] = {
+        { .size = 16,    .slot_cnt = 256, },
+        { .size = 32,    .slot_cnt = 128, },
+        // this size class is used by allocations that are made by the addr_tag_map, which breaks
+        // tag distinctess checks
+        // { .size = 48,    .slot_cnt = 85,  },
+        { .size = 64,    .slot_cnt = 64,  },
+        { .size = 80,    .slot_cnt = 51,  },
+        { .size = 96,    .slot_cnt = 42,  },
+        { .size = 112,   .slot_cnt = 36,  },
+        { .size = 128,   .slot_cnt = 64,  },
+        { .size = 160,   .slot_cnt = 51,  },
+        { .size = 192,   .slot_cnt = 64,  },
+        { .size = 224,   .slot_cnt = 54,  },
+        { .size = 10240, .slot_cnt = 6,   },
+        { .size = 20480, .slot_cnt = 1,   },
+    };
+
+    int tag_usage[max_tag + 1];
+
+    for (size_t sc_idx = 0; sc_idx < sizeof(size_classes) / sizeof(SizeClass); ++sc_idx) {
+        SizeClass &sc = size_classes[sc_idx];
+
+        const size_t full_alloc_size = sc.size;
+        const size_t alloc_size = full_alloc_size - CANARY_SIZE;
+
+        // "tdc" is short for "tag distinctness check"
+        int left_neighbor_tdc_cnt = 0;
+        int right_neighbor_tdc_cnt = 0;
+        int prev_alloc_tdc_cnt = 0;
+
+        int iter_cnt = 600;
+
+        unordered_map<uptr, u8> addr_tag_map;
+        addr_tag_map.reserve(iter_cnt * sc.slot_cnt);
+
+        u64 seen_tags = 0;
+
+        for (int iter = 0; iter < iter_cnt; ++iter) {
+            uptr allocations[256]; // 256 is max slot count
+
+            for (int i = 0; i < sc.slot_cnt; ++i) {
+                u8 *p = (u8 *) malloc(alloc_size);
+                assert(p);
+                uptr addr = (uptr) untag_pointer(p);
+                u8 tag = get_pointer_tag(p);
+
+                assert(tag >= min_tag && tag <= max_tag);
+                seen_tags |= 1 << tag;
+                ++tag_usage[tag];
+
+                // check most recent tags of left and right neighbors
+
+                auto left = addr_tag_map.find(addr - full_alloc_size);
+                if (left != addr_tag_map.end()) {
+                    assert(left->second != tag);
+                    ++left_neighbor_tdc_cnt;
+                }
+
+                auto right = addr_tag_map.find(addr + full_alloc_size);
+                if (right != addr_tag_map.end()) {
+                    assert(right->second != tag);
+                    ++right_neighbor_tdc_cnt;
+                }
+
+                // check previous tag of this slot
+                auto prev = addr_tag_map.find(addr);
+                if (prev != addr_tag_map.end()) {
+                    assert(prev->second != tag);
+                    ++prev_alloc_tdc_cnt;
+                    addr_tag_map.erase(addr);
+                }
+
+                addr_tag_map.emplace(addr, tag);
+
+                for (size_t j = 0; j < alloc_size; ++j) {
+                    // check that slot is zeroed
+                    assert(p[j] == 0);
+                    // check that slot is readable and writable
+                    p[j]++;
+                }
+
+                allocations[i] = addr;
+            }
+
+            // free some of allocations to allow their slots to be reused
+            for (int i = sc.slot_cnt - 1; i >= 0; i -= 2) {
+                free((void *) allocations[i]);
+            }
+        }
+
+        // check that all of the tags were used, except for the reserved tag 0
+        assert(seen_tags == (0xffff & ~(1 << 0)));
+
+        printf("size_class\t%i\t" "tdc_left %i\t" "tdc_right %i\t" "tdc_prev_alloc %i\n",
+               sc.size, left_neighbor_tdc_cnt, right_neighbor_tdc_cnt, prev_alloc_tdc_cnt);
+
+        // make sure tag distinctess checks were actually performed
+        int min_tdc_cnt = sc.slot_cnt * iter_cnt / 5;
+
+        assert(prev_alloc_tdc_cnt > min_tdc_cnt);
+
+        if (sc.slot_cnt > 1) {
+            assert(left_neighbor_tdc_cnt > min_tdc_cnt);
+            assert(right_neighbor_tdc_cnt > min_tdc_cnt);
+        }
+
+        // async tag check failures are reported on context switch
+        do_context_switch();
+    }
+
+    printf("\nTag use counters:\n");
+
+    int min = INT_MAX;
+    int max = 0;
+    double geomean = 0.0;
+    for (int i = min_tag; i <= max_tag; ++i) {
+        int v = tag_usage[i];
+        geomean += log(v);
+        min = std::min(min, v);
+        max = std::max(max, v);
+        printf("%i\t%i\n", i, tag_usage[i]);
+    }
+    int tag_cnt = 1 + max_tag - min_tag;
+    geomean = exp(geomean / tag_cnt);
+
+    double max_deviation = std::max((double) max - geomean, geomean - min);
+
+    printf("geomean: %.2f, max deviation from geomean: %.2f%%\n", geomean, (100.0 * max_deviation) / geomean);
+}
+
+u8* alloc_default() {
+    const size_t full_alloc_size = DEFAULT_ALLOC_SIZE + CANARY_SIZE;
+    set<uptr> addrs;
+
+    // make sure allocation has both left and right neighbors, otherwise overflow/underflow tests
+    // will fail when allocation is at the end/beginning of slab
+    for (;;) {
+        u8 *p = (u8 *) malloc(DEFAULT_ALLOC_SIZE);
+        assert(p);
+
+        uptr addr = (uptr) untag_pointer(p);
+        uptr left = addr - full_alloc_size;
+        if (addrs.find(left) != addrs.end()) {
+            uptr right = addr + full_alloc_size;
+            if (addrs.find(right) != addrs.end()) {
+                return p;
+            }
+        }
+
+        addrs.emplace(addr);
+    }
+}
+
+int expected_segv_code;
+
+#define expect_segv(exp, segv_code) ({\
+    expected_segv_code = segv_code; \
+    volatile auto val = exp; \
+    (void) val; \
+    do_context_switch(); \
+    fprintf(stderr, "didn't receive SEGV code %i", segv_code); \
+    exit(1); })
+
+// it's expected that the device is configured to use asymm MTE tag checking mode (sync read checks,
+// async write checks)
+#define expect_read_segv(exp) expect_segv(exp, SEGV_MTESERR)
+#define expect_write_segv(exp) expect_segv(exp, SEGV_MTEAERR)
+
+void read_after_free() {
+    u8 *p = alloc_default();
+    free(p);
+    expect_read_segv(p[0]);
+}
+
+void write_after_free() {
+    u8 *p = alloc_default();
+    free(p);
+    expect_write_segv(p[0] = 1);
+}
+
+void underflow_read() {
+    u8 *p = alloc_default();
+    expect_read_segv(p[-1]);
+}
+
+void underflow_write() {
+    u8 *p = alloc_default();
+    expect_write_segv(p[-1] = 1);
+}
+
+void overflow_read() {
+    u8 *p = alloc_default();
+    expect_read_segv(p[DEFAULT_ALLOC_SIZE + CANARY_SIZE]);
+}
+
+void overflow_write() {
+    u8 *p = alloc_default();
+    expect_write_segv(p[DEFAULT_ALLOC_SIZE + CANARY_SIZE] = 1);
+}
+
+void untagged_read() {
+    u8 *p = alloc_default();
+    p = (u8 *) untag_pointer(p);
+    expect_read_segv(p[0]);
+}
+
+void untagged_write() {
+    u8 *p = alloc_default();
+    p = (u8 *) untag_pointer(p);
+    expect_write_segv(p[0] = 1);
+}
+
+// checks that each of memory locations inside the buffer is tagged with expected_tag
+void check_tag(void *buf, size_t len, u8 expected_tag) {
+    for (size_t i = 0; i < len; ++i) {
+        assert(get_pointer_tag(__arm_mte_get_tag((void *) ((uintptr_t) buf + i))) == expected_tag);
+    }
+}
+
+void madvise_dontneed() {
+    const size_t len = 100'000;
+    void *ptr = mmap(NULL, len, PROT_READ | PROT_WRITE | PROT_MTE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
+    assert(ptr != MAP_FAILED);
+
+    // check that 0 is the initial tag
+    check_tag(ptr, len, 0);
+
+    arm_mte_tag_and_clear_mem(set_pointer_tag(ptr, 1), len);
+    check_tag(ptr, len, 1);
+
+    memset(set_pointer_tag(ptr, 1), 1, len);
+
+    assert(madvise(ptr, len, MADV_DONTNEED) == 0);
+    // check that MADV_DONTNEED resets the tag
+    check_tag(ptr, len, 0);
+
+    // check that MADV_DONTNEED clears the memory
+    for (size_t i = 0; i < len; ++i) {
+        assert(((u8 *) ptr)[i] == 0);
+    }
+
+    // check that mistagged read after MADV_DONTNEED fails
+    expect_read_segv(*((u8 *) set_pointer_tag(ptr, 1)));
+}
+
+map<string, function<void()>> tests = {
+#define TEST(s) { #s, s }
+    TEST(tag_distinctness),
+    TEST(read_after_free),
+    TEST(write_after_free),
+    TEST(overflow_read),
+    TEST(overflow_write),
+    TEST(underflow_read),
+    TEST(underflow_write),
+    TEST(untagged_read),
+    TEST(untagged_write),
+    TEST(madvise_dontneed),
+#undef TEST
+};
+
+void segv_handler(int, siginfo_t *si, void *) {
+    if (expected_segv_code == 0 || expected_segv_code != si->si_code) {
+        fprintf(stderr, "received unexpected SEGV_CODE %i", si->si_code);
+        exit(139); // standard exit code for SIGSEGV
+    }
+
+    exit(0);
+}
+
+int main(int argc, char **argv) {
+    setbuf(stdout, NULL);
+    assert(argc == 2);
+
+    auto test_name = string(argv[1]);
+    auto test_fn = tests[test_name];
+    assert(test_fn != nullptr);
+
+    assert(mallopt(M_BIONIC_SET_HEAP_TAGGING_LEVEL, M_HEAP_TAGGING_LEVEL_ASYNC) == 1);
+
+    struct sigaction sa = {
+        .sa_sigaction = segv_handler,
+        .sa_flags = SA_SIGINFO,
+    };
+
+    assert(sigaction(SIGSEGV, &sa, nullptr) == 0);
+
+    test_fn();
+    do_context_switch();
+
+    return 0;
+}

androidtest/src/grapheneos/hmalloc/MemtagTest.java

@@ -0,0 +1,79 @@
+package grapheneos.hmalloc;
+
+import com.android.tradefed.device.DeviceNotAvailableException;
+import com.android.tradefed.testtype.DeviceJUnit4ClassRunner;
+import com.android.tradefed.testtype.junit4.BaseHostJUnit4Test;
+
+import org.junit.Test;
+import org.junit.runner.RunWith;
+
+import java.util.ArrayList;
+
+import static org.junit.Assert.assertEquals;
+
+@RunWith(DeviceJUnit4ClassRunner.class)
+public class MemtagTest extends BaseHostJUnit4Test {
+    private static final String TEST_BINARY = "/data/local/tmp/memtag_test";
+
+    private void runTest(String name) throws DeviceNotAvailableException {
+        var args = new ArrayList<String>();
+        args.add(TEST_BINARY);
+        args.add(name);
+        String cmdLine = String.join(" ", args);
+
+        var result = getDevice().executeShellV2Command(cmdLine);
+
+        assertEquals("stderr", "", result.getStderr());
+        assertEquals("process exit code", 0, result.getExitCode().intValue());
+    }
+
+    @Test
+    public void tag_distinctness() throws DeviceNotAvailableException {
+        runTest("tag_distinctness");
+    }
+
+    @Test
+    public void read_after_free() throws DeviceNotAvailableException {
+        runTest("read_after_free");
+    }
+
+    @Test
+    public void write_after_free() throws DeviceNotAvailableException {
+        runTest("write_after_free");
+    }
+
+    @Test
+    public void underflow_read() throws DeviceNotAvailableException {
+        runTest("underflow_read");
+    }
+
+    @Test
+    public void underflow_write() throws DeviceNotAvailableException {
+        runTest("underflow_write");
+    }
+
+    @Test
+    public void overflow_read() throws DeviceNotAvailableException {
+        runTest("overflow_read");
+    }
+
+    @Test
+    public void overflow_write() throws DeviceNotAvailableException {
+        runTest("overflow_write");
+    }
+
+    @Test
+    public void untagged_read() throws DeviceNotAvailableException {
+        runTest("untagged_read");
+    }
+
+    @Test
+    public void untagged_write() throws DeviceNotAvailableException {
+        runTest("untagged_write");
+    }
+
+    @Test
+    public void madvise_dontneed() throws DeviceNotAvailableException {
+        runTest("madvise_dontneed");
+    }
+}

arm_mte.h

@@ -0,0 +1,91 @@
+#ifndef ARM_MTE_H
+#define ARM_MTE_H
+
+#include <arm_acle.h>
+#include <stdint.h>
+
+// Returns a tagged pointer.
+// See https://developer.arm.com/documentation/ddi0602/2023-09/Base-Instructions/IRG--Insert-Random-Tag-
+static inline void *arm_mte_create_random_tag(void *p, uint64_t exclusion_mask) {
+    return __arm_mte_create_random_tag(p, exclusion_mask);
+}
+
+// Tag the memory region with the tag specified in tag bits of tagged_ptr. Memory region itself is
+// zeroed.
+// tagged_ptr has to be aligned by 16, and len has to be a multiple of 16 (tag granule size).
+//
+// Arm's software optimization guide says:
+// "it is recommended to use STZGM (or DCZGVA) to set tag if data is not a concern." (STZGM and
+// DCGZVA are zeroing variants of tagging instructions).
+//
+// Contents of this function were copied from scudo:
+// https://android.googlesource.com/platform/external/scudo/+/refs/tags/android-14.0.0_r1/standalone/memtag.h#167
+//
+// scudo is licensed under the Apache License v2.0 with LLVM Exceptions, which is compatible with
+// the hardened_malloc's MIT license
+static inline void arm_mte_tag_and_clear_mem(void *tagged_ptr, size_t len) {
+    uintptr_t Begin = (uintptr_t) tagged_ptr;
+    uintptr_t End = Begin + len;
+    uintptr_t LineSize, Next, Tmp;
+    __asm__ __volatile__(
+        ".arch_extension memtag \n\t"
+
+        // Compute the cache line size in bytes (DCZID_EL0 stores it as the log2
+        // of the number of 4-byte words) and bail out to the slow path if DCZID_EL0
+        // indicates that the DC instructions are unavailable.
+        "DCZID .req %[Tmp] \n\t"
+        "mrs DCZID, dczid_el0 \n\t"
+        "tbnz DCZID, #4, 3f \n\t"
+        "and DCZID, DCZID, #15 \n\t"
+        "mov %[LineSize], #4 \n\t"
+        "lsl %[LineSize], %[LineSize], DCZID \n\t"
+        ".unreq DCZID \n\t"
+
+        // Our main loop doesn't handle the case where we don't need to perform any
+        // DC GZVA operations. If the size of our tagged region is less than
+        // twice the cache line size, bail out to the slow path since it's not
+        // guaranteed that we'll be able to do a DC GZVA.
+        "Size .req %[Tmp] \n\t"
+        "sub Size, %[End], %[Cur] \n\t"
+        "cmp Size, %[LineSize], lsl #1 \n\t"
+        "b.lt 3f \n\t"
+        ".unreq Size \n\t"
+
+        "LineMask .req %[Tmp] \n\t"
+        "sub LineMask, %[LineSize], #1 \n\t"
+
+        // STZG until the start of the next cache line.
+        "orr %[Next], %[Cur], LineMask \n\t"
+
+        "1:\n\t"
+        "stzg %[Cur], [%[Cur]], #16 \n\t"
+        "cmp %[Cur], %[Next] \n\t"
+        "b.lt 1b \n\t"
+
+        // DC GZVA cache lines until we have no more full cache lines.
+        "bic %[Next], %[End], LineMask \n\t"
+        ".unreq LineMask \n\t"
+
+        "2: \n\t"
+        "dc gzva, %[Cur] \n\t"
+        "add %[Cur], %[Cur], %[LineSize] \n\t"
+        "cmp %[Cur], %[Next] \n\t"
+        "b.lt 2b \n\t"
+
+        // STZG until the end of the tagged region. This loop is also used to handle
+        // slow path cases.
+
+        "3: \n\t"
+        "cmp %[Cur], %[End] \n\t"
+        "b.ge 4f \n\t"
+        "stzg %[Cur], [%[Cur]], #16 \n\t"
+        "b 3b \n\t"
+
+        "4: \n\t"
+
+        : [Cur] "+&r"(Begin), [LineSize] "=&r"(LineSize), [Next] "=&r"(Next), [Tmp] "=&r"(Tmp)
+        : [End] "r"(End)
+        : "memory"
+    );
+}
+#endif

calculate-waste

@@ -49,7 +49,7 @@ for size, slots, fragmentation in zip(size_classes, size_class_slots, fragmentat
     used = size * slots
     real = page_align(used)
     print("| ", end='')
-    print(size, str(fragmentation) + "%", slots, real, str(100 - used / real * 100) + "%", sep=" | ", end=" |\n")
+    print(size, f"{fragmentation:.4}%", slots, real, str(100 - used / real * 100) + "%", sep=" | ", end=" |\n")
 
 if len(argv) < 2:
     exit()

chacha.c

@@ -41,7 +41,7 @@ static const unsigned rounds = 8;
     a = PLUS(a, b); d = ROTATE(XOR(d, a), 8); \
     c = PLUS(c, d); b = ROTATE(XOR(b, c), 7);
 
-static const char sigma[16] = "expand 32-byte k";
+static const char sigma[16] NONSTRING = "expand 32-byte k";
 
 void chacha_keysetup(chacha_ctx *x, const u8 *k) {
     x->input[0] = U8TO32_LITTLE(sigma + 0);

config/default.mk

@@ -0,0 +1,23 @@
+CONFIG_WERROR := true
+CONFIG_NATIVE := true
+CONFIG_CXX_ALLOCATOR := true
+CONFIG_UBSAN := false
+CONFIG_SEAL_METADATA := false
+CONFIG_ZERO_ON_FREE := true
+CONFIG_WRITE_AFTER_FREE_CHECK := true
+CONFIG_SLOT_RANDOMIZE := true
+CONFIG_SLAB_CANARY := true
+CONFIG_SLAB_QUARANTINE_RANDOM_LENGTH := 1
+CONFIG_SLAB_QUARANTINE_QUEUE_LENGTH := 1
+CONFIG_EXTENDED_SIZE_CLASSES := true
+CONFIG_LARGE_SIZE_CLASSES := true
+CONFIG_GUARD_SLABS_INTERVAL := 1
+CONFIG_GUARD_SIZE_DIVISOR := 2
+CONFIG_REGION_QUARANTINE_RANDOM_LENGTH := 256
+CONFIG_REGION_QUARANTINE_QUEUE_LENGTH := 1024
+CONFIG_REGION_QUARANTINE_SKIP_THRESHOLD := 33554432 # 32MiB
+CONFIG_FREE_SLABS_QUARANTINE_RANDOM_LENGTH := 32
+CONFIG_CLASS_REGION_SIZE := 34359738368 # 32GiB
+CONFIG_N_ARENA := 4
+CONFIG_STATS := false
+CONFIG_SELF_INIT := true

config/light.mk

@@ -0,0 +1,23 @@
+CONFIG_WERROR := true
+CONFIG_NATIVE := true
+CONFIG_CXX_ALLOCATOR := true
+CONFIG_UBSAN := false
+CONFIG_SEAL_METADATA := false
+CONFIG_ZERO_ON_FREE := true
+CONFIG_WRITE_AFTER_FREE_CHECK := false
+CONFIG_SLOT_RANDOMIZE := false
+CONFIG_SLAB_CANARY := true
+CONFIG_SLAB_QUARANTINE_RANDOM_LENGTH := 0
+CONFIG_SLAB_QUARANTINE_QUEUE_LENGTH := 0
+CONFIG_EXTENDED_SIZE_CLASSES := true
+CONFIG_LARGE_SIZE_CLASSES := true
+CONFIG_GUARD_SLABS_INTERVAL := 8
+CONFIG_GUARD_SIZE_DIVISOR := 2
+CONFIG_REGION_QUARANTINE_RANDOM_LENGTH := 256
+CONFIG_REGION_QUARANTINE_QUEUE_LENGTH := 1024
+CONFIG_REGION_QUARANTINE_SKIP_THRESHOLD := 33554432 # 32MiB
+CONFIG_FREE_SLABS_QUARANTINE_RANDOM_LENGTH := 32
+CONFIG_CLASS_REGION_SIZE := 34359738368 # 32GiB
+CONFIG_N_ARENA := 4
+CONFIG_STATS := false
+CONFIG_SELF_INIT := true

include/h_malloc.h

@@ -5,7 +5,9 @@
 
 #include <malloc.h>
 
-__BEGIN_DECLS
+#ifdef __cplusplus
+extern "C" {
+#endif
 
 #ifndef H_MALLOC_PREFIX
 #define h_malloc malloc
@@ -21,6 +23,7 @@ __BEGIN_DECLS
 #define h_malloc_trim malloc_trim
 #define h_malloc_stats malloc_stats
 #define h_mallinfo mallinfo
+#define h_mallinfo2 mallinfo2
 #define h_malloc_info malloc_info
 
 #define h_memalign memalign
@@ -30,7 +33,12 @@ __BEGIN_DECLS
 #define h_malloc_get_state malloc_get_state
 #define h_malloc_set_state malloc_set_state
 
-#define h_iterate iterate
+#define h_mallinfo_narenas mallinfo_narenas
+#define h_mallinfo_nbins mallinfo_nbins
+#define h_mallinfo_arena_info mallinfo_arena_info
+#define h_mallinfo_bin_info mallinfo_bin_info
+
+#define h_malloc_iterate malloc_iterate
 #define h_malloc_disable malloc_disable
 #define h_malloc_enable malloc_enable
 
@@ -40,9 +48,10 @@ __BEGIN_DECLS
 #endif
 
 // C standard
-void *h_malloc(size_t size);
-void *h_calloc(size_t nmemb, size_t size);
-void *h_realloc(void *ptr, size_t size);
+__attribute__((malloc)) __attribute__((alloc_size(1))) void *h_malloc(size_t size);
+__attribute__((malloc)) __attribute__((alloc_size(1, 2))) void *h_calloc(size_t nmemb, size_t size);
+__attribute__((alloc_size(2))) void *h_realloc(void *ptr, size_t size);
+__attribute__((malloc)) __attribute__((alloc_size(2))) __attribute__((alloc_align(1)))
 void *h_aligned_alloc(size_t alignment, size_t size);
 void h_free(void *ptr);
 
@@ -68,36 +77,38 @@ int h_malloc_info(int options, FILE *fp);
 #endif
 
 // obsolete glibc extensions
+__attribute__((malloc)) __attribute__((alloc_size(2))) __attribute__((alloc_align(1)))
 void *h_memalign(size_t alignment, size_t size);
 #ifndef __ANDROID__
-void *h_valloc(size_t size);
-void *h_pvalloc(size_t size);
+__attribute__((malloc)) __attribute__((alloc_size(1))) void *h_valloc(size_t size);
+__attribute__((malloc)) void *h_pvalloc(size_t size);
 #endif
 #ifdef __GLIBC__
-void h_cfree(void *ptr);
+void h_cfree(void *ptr) __THROW;
 void *h_malloc_get_state(void);
 int h_malloc_set_state(void *state);
 #endif
 
 // Android extensions
 #ifdef __ANDROID__
-size_t __mallinfo_narenas(void);
-size_t __mallinfo_nbins(void);
-struct mallinfo __mallinfo_arena_info(size_t arena);
-struct mallinfo __mallinfo_bin_info(size_t arena, size_t bin);
-int h_iterate(uintptr_t base, size_t size, void (*callback)(uintptr_t ptr, size_t size, void *arg),
+size_t h_mallinfo_narenas(void);
+size_t h_mallinfo_nbins(void);
+struct mallinfo h_mallinfo_arena_info(size_t arena);
+struct mallinfo h_mallinfo_bin_info(size_t arena, size_t bin);
+int h_malloc_iterate(uintptr_t base, size_t size, void (*callback)(uintptr_t ptr, size_t size, void *arg),
               void *arg);
 void h_malloc_disable(void);
 void h_malloc_enable(void);
+void h_malloc_disable_memory_tagging(void);
 #endif
 
 // hardened_malloc extensions
 
 // return an upper bound on object size for any pointer based on malloc metadata
-size_t h_malloc_object_size(void *ptr);
+size_t h_malloc_object_size(const void *ptr);
 
 // similar to malloc_object_size, but avoiding locking so the results are much more limited
-size_t h_malloc_object_size_fast(void *ptr);
+size_t h_malloc_object_size_fast(const void *ptr);
 
 // The free function with an extra parameter for passing the size requested at
 // allocation time.
@@ -111,6 +122,8 @@ size_t h_malloc_object_size_fast(void *ptr);
 // passed size matches the allocated size.
 void h_free_sized(void *ptr, size_t expected_size);
 
-__END_DECLS
+#ifdef __cplusplus
+}
+#endif
 
 #endif

memory.c

@@ -1,7 +1,10 @@
 #include <errno.h>
 
 #include <sys/mman.h>
+
+#ifdef LABEL_MEMORY
 #include <sys/prctl.h>
+#endif
 
 #ifndef PR_SET_VMA
 #define PR_SET_VMA 0x53564d41
@@ -14,8 +17,8 @@
 #include "memory.h"
 #include "util.h"
 
-void *memory_map(size_t size) {
-    void *p = mmap(NULL, size, PROT_NONE, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
+static void *memory_map_prot(size_t size, int prot) {
+    void *p = mmap(NULL, size, prot, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
     if (unlikely(p == MAP_FAILED)) {
         if (errno != ENOMEM) {
             fatal_error("non-ENOMEM mmap failure");
@@ -25,30 +28,50 @@ void *memory_map(size_t size) {
     return p;
 }
 
-int memory_map_fixed(void *ptr, size_t size) {
-    void *p = mmap(ptr, size, PROT_NONE, MAP_ANONYMOUS|MAP_PRIVATE|MAP_FIXED, -1, 0);
-    if (unlikely(p == MAP_FAILED)) {
-        if (errno != ENOMEM) {
-            fatal_error("non-ENOMEM MAP_FIXED mmap failure");
-        }
-        return 1;
-    }
-    return 0;
+void *memory_map(size_t size) {
+    return memory_map_prot(size, PROT_NONE);
 }
 
-int memory_unmap(void *ptr, size_t size) {
-    int ret = munmap(ptr, size);
+#ifdef HAS_ARM_MTE
+// Note that PROT_MTE can't be cleared via mprotect
+void *memory_map_mte(size_t size) {
+    return memory_map_prot(size, PROT_MTE);
+}
+#endif
+
+static bool memory_map_fixed_prot(void *ptr, size_t size, int prot) {
+    void *p = mmap(ptr, size, prot, MAP_ANONYMOUS|MAP_PRIVATE|MAP_FIXED, -1, 0);
+    bool ret = p == MAP_FAILED;
+    if (unlikely(ret) && errno != ENOMEM) {
+        fatal_error("non-ENOMEM MAP_FIXED mmap failure");
+    }
+    return ret;
+}
+
+bool memory_map_fixed(void *ptr, size_t size) {
+    return memory_map_fixed_prot(ptr, size, PROT_NONE);
+}
+
+#ifdef HAS_ARM_MTE
+// Note that PROT_MTE can't be cleared via mprotect
+bool memory_map_fixed_mte(void *ptr, size_t size) {
+    return memory_map_fixed_prot(ptr, size, PROT_MTE);
+}
+#endif
+
+bool memory_unmap(void *ptr, size_t size) {
+    bool ret = munmap(ptr, size);
     if (unlikely(ret) && errno != ENOMEM) {
         fatal_error("non-ENOMEM munmap failure");
     }
     return ret;
 }
 
-static int memory_protect_prot(void *ptr, size_t size, int prot, UNUSED int pkey) {
+static bool memory_protect_prot(void *ptr, size_t size, int prot, UNUSED int pkey) {
 #ifdef USE_PKEY
-    int ret = pkey_mprotect(ptr, size, prot, pkey);
+    bool ret = pkey_mprotect(ptr, size, prot, pkey);
 #else
-    int ret = mprotect(ptr, size, prot);
+    bool ret = mprotect(ptr, size, prot);
 #endif
     if (unlikely(ret) && errno != ENOMEM) {
         fatal_error("non-ENOMEM mprotect failure");
@@ -56,42 +79,50 @@ static int memory_protect_prot(void *ptr, size_t size, int prot, UNUSED int pkey
     return ret;
 }
 
-int memory_protect_ro(void *ptr, size_t size) {
+bool memory_protect_ro(void *ptr, size_t size) {
     return memory_protect_prot(ptr, size, PROT_READ, -1);
 }
 
-int memory_protect_rw(void *ptr, size_t size) {
+bool memory_protect_rw(void *ptr, size_t size) {
     return memory_protect_prot(ptr, size, PROT_READ|PROT_WRITE, -1);
 }
 
-int memory_protect_rw_metadata(void *ptr, size_t size) {
+bool memory_protect_rw_metadata(void *ptr, size_t size) {
     return memory_protect_prot(ptr, size, PROT_READ|PROT_WRITE, get_metadata_key());
 }
 
-int memory_remap(void *old, size_t old_size, size_t new_size) {
+#ifdef HAVE_COMPATIBLE_MREMAP
+bool memory_remap(void *old, size_t old_size, size_t new_size) {
     void *ptr = mremap(old, old_size, new_size, 0);
-    if (unlikely(ptr == MAP_FAILED)) {
-        if (errno != ENOMEM) {
-            fatal_error("non-ENOMEM mremap failure");
-        }
-        return 1;
+    bool ret = ptr == MAP_FAILED;
+    if (unlikely(ret) && errno != ENOMEM) {
+        fatal_error("non-ENOMEM mremap failure");
     }
-    return 0;
+    return ret;
 }
 
-int memory_remap_fixed(void *old, size_t old_size, void *new, size_t new_size) {
+bool memory_remap_fixed(void *old, size_t old_size, void *new, size_t new_size) {
     void *ptr = mremap(old, old_size, new_size, MREMAP_MAYMOVE|MREMAP_FIXED, new);
-    if (unlikely(ptr == MAP_FAILED)) {
-        if (errno != ENOMEM) {
-            fatal_error("non-ENOMEM MREMAP_FIXED mremap failure");
-        }
-        return 1;
+    bool ret = ptr == MAP_FAILED;
+    if (unlikely(ret) && errno != ENOMEM) {
+        fatal_error("non-ENOMEM MREMAP_FIXED mremap failure");
     }
-    return 0;
+    return ret;
+}
+#endif
+
+bool memory_purge(void *ptr, size_t size) {
+    int ret = madvise(ptr, size, MADV_DONTNEED);
+    if (unlikely(ret) && errno != ENOMEM) {
+        fatal_error("non-ENOMEM MADV_DONTNEED madvise failure");
+    }
+    return ret;
 }
 
-void memory_set_name(UNUSED void *ptr, UNUSED size_t size, UNUSED const char *name) {
+bool memory_set_name(UNUSED void *ptr, UNUSED size_t size, UNUSED const char *name) {
 #ifdef LABEL_MEMORY
-    prctl(PR_SET_VMA, PR_SET_VMA_ANON_NAME, ptr, size, name);
+    return prctl(PR_SET_VMA, PR_SET_VMA_ANON_NAME, ptr, size, name);
+#else
+    return false;
 #endif
 }

memory.h

@@ -1,18 +1,32 @@
 #ifndef MEMORY_H
 #define MEMORY_H
 
+#include <stdbool.h>
 #include <stddef.h>
 
+#ifdef __linux__
+#define HAVE_COMPATIBLE_MREMAP
+#endif
+
 int get_metadata_key(void);
 
 void *memory_map(size_t size);
-int memory_map_fixed(void *ptr, size_t size);
-int memory_unmap(void *ptr, size_t size);
-int memory_protect_ro(void *ptr, size_t size);
-int memory_protect_rw(void *ptr, size_t size);
-int memory_protect_rw_metadata(void *ptr, size_t size);
-int memory_remap(void *old, size_t old_size, size_t new_size);
-int memory_remap_fixed(void *old, size_t old_size, void *new, size_t new_size);
-void memory_set_name(void *ptr, size_t size, const char *name);
+#ifdef HAS_ARM_MTE
+void *memory_map_mte(size_t size);
+#endif
+bool memory_map_fixed(void *ptr, size_t size);
+#ifdef HAS_ARM_MTE
+bool memory_map_fixed_mte(void *ptr, size_t size);
+#endif
+bool memory_unmap(void *ptr, size_t size);
+bool memory_protect_ro(void *ptr, size_t size);
+bool memory_protect_rw(void *ptr, size_t size);
+bool memory_protect_rw_metadata(void *ptr, size_t size);
+#ifdef HAVE_COMPATIBLE_MREMAP
+bool memory_remap(void *old, size_t old_size, size_t new_size);
+bool memory_remap_fixed(void *old, size_t old_size, void *new, size_t new_size);
+#endif
+bool memory_purge(void *ptr, size_t size);
+bool memory_set_name(void *ptr, size_t size, const char *name);
 
 #endif

memtag.h

@@ -0,0 +1,50 @@
+#ifndef MEMTAG_H
+#define MEMTAG_H
+
+#include "util.h"
+
+#ifdef HAS_ARM_MTE
+#include "arm_mte.h"
+#define MEMTAG 1
+// Note that bionic libc always reserves tag 0 via PR_MTE_TAG_MASK prctl
+#define RESERVED_TAG 0
+#define TAG_WIDTH 4
+#endif
+
+static inline void *untag_pointer(void *ptr) {
+#ifdef HAS_ARM_MTE
+    const uintptr_t mask = UINTPTR_MAX >> 8;
+    return (void *) ((uintptr_t) ptr & mask);
+#else
+    return ptr;
+#endif
+}
+
+static inline const void *untag_const_pointer(const void *ptr) {
+#ifdef HAS_ARM_MTE
+    const uintptr_t mask = UINTPTR_MAX >> 8;
+    return (const void *) ((uintptr_t) ptr & mask);
+#else
+    return ptr;
+#endif
+}
+
+static inline void *set_pointer_tag(void *ptr, u8 tag) {
+#ifdef HAS_ARM_MTE
+    return (void *) (((uintptr_t) tag << 56) | (uintptr_t) untag_pointer(ptr));
+#else
+    (void) tag;
+    return ptr;
+#endif
+}
+
+static inline u8 get_pointer_tag(void *ptr) {
+#ifdef HAS_ARM_MTE
+    return (((uintptr_t) ptr) >> 56) & 0xf;
+#else
+    (void) ptr;
+    return 0;
+#endif
+}
+
+#endif

new.cc

@@ -1,7 +1,9 @@
+// needed with libstdc++ but not libc++
+#if __has_include(<bits/functexcept.h>)
 #include <bits/functexcept.h>
-#include <new>
+#endif
 
-#define noreturn
+#include <new>
 
 #include "h_malloc.h"
 #include "util.h"
@@ -78,7 +80,6 @@ EXPORT void operator delete[](void *ptr, size_t size) noexcept {
     h_free_sized(ptr, size);
 }
 
-#if __cplusplus >= 201703L
 COLD static void *handle_out_of_memory(size_t size, size_t alignment, bool nothrow) {
     void *ptr = nullptr;
 
@@ -150,4 +151,3 @@ EXPORT void operator delete(void *ptr, size_t size, std::align_val_t) noexcept {
 EXPORT void operator delete[](void *ptr, size_t size, std::align_val_t) noexcept {
     h_free_sized(ptr, size);
 }
-#endif

pages.c

@@ -9,10 +9,6 @@ static bool add_guards(size_t size, size_t guard_size, size_t *total_size) {
         __builtin_add_overflow(*total_size, guard_size, total_size);
 }
 
-static uintptr_t alignment_ceiling(uintptr_t s, uintptr_t alignment) {
-    return ((s) + (alignment - 1)) & ((~alignment) + 1);
-}
-
 void *allocate_pages(size_t usable_size, size_t guard_size, bool unprotect, const char *name) {
     size_t real_size;
     if (unlikely(add_guards(usable_size, guard_size, &real_size))) {
@@ -33,7 +29,7 @@ void *allocate_pages(size_t usable_size, size_t guard_size, bool unprotect, cons
 }
 
 void *allocate_pages_aligned(size_t usable_size, size_t alignment, size_t guard_size, const char *name) {
-    usable_size = PAGE_CEILING(usable_size);
+    usable_size = page_align(usable_size);
     if (unlikely(!usable_size)) {
         errno = ENOMEM;
         return NULL;
@@ -59,7 +55,7 @@ void *allocate_pages_aligned(size_t usable_size, size_t alignment, size_t guard_
 
     void *usable = (char *)real + guard_size;
 
-    size_t lead_size = alignment_ceiling((uintptr_t)usable, alignment) - (uintptr_t)usable;
+    size_t lead_size = align((uintptr_t)usable, alignment) - (uintptr_t)usable;
     size_t trail_size = alloc_size - lead_size - usable_size;
     void *base = (char *)usable + lead_size;
 
@@ -86,5 +82,7 @@ void *allocate_pages_aligned(size_t usable_size, size_t alignment, size_t guard_
 }
 
 void deallocate_pages(void *usable, size_t usable_size, size_t guard_size) {
-    memory_unmap((char *)usable - guard_size, usable_size + guard_size * 2);
+    if (unlikely(memory_unmap((char *)usable - guard_size, usable_size + guard_size * 2))) {
+        memory_purge(usable, usable_size);
+    }
 }

pages.h

@@ -5,16 +5,21 @@
 #include <stddef.h>
 #include <stdint.h>
 
+#include "util.h"
+
 #define PAGE_SHIFT 12
 #ifndef PAGE_SIZE
 #define PAGE_SIZE ((size_t)1 << PAGE_SHIFT)
 #endif
-#define PAGE_CEILING(s) (((s) + PAGE_SIZE - 1) & ~(PAGE_SIZE - 1))
 
 void *allocate_pages(size_t usable_size, size_t guard_size, bool unprotect, const char *name);
 void *allocate_pages_aligned(size_t usable_size, size_t alignment, size_t guard_size, const char *name);
 void deallocate_pages(void *usable, size_t usable_size, size_t guard_size);
 
+static inline size_t page_align(size_t size) {
+    return align(size, PAGE_SIZE);
+}
+
 static inline size_t hash_page(const void *p) {
     uintptr_t u = (uintptr_t)p >> PAGE_SHIFT;
     size_t sum = u;

random.c

@@ -5,17 +5,7 @@
 #include "random.h"
 #include "util.h"
 
-#if __has_include(<sys/random.h>)
-// glibc 2.25 and later
 #include <sys/random.h>
-#else
-#include <unistd.h>
-#include <sys/syscall.h>
-
-static ssize_t getrandom(void *buf, size_t buflen, unsigned int flags) {
-    return syscall(SYS_getrandom, buf, buflen, flags);
-}
-#endif
 
 static void get_random_seed(void *buf, size_t size) {
     while (size) {
@@ -102,7 +92,7 @@ u16 get_random_u16_uniform(struct random_state *state, u16 bound) {
     if (leftover < bound) {
         u16 threshold = -bound % bound;
         while (leftover < threshold) {
-            random =  get_random_u16(state);
+            random = get_random_u16(state);
             multiresult = random * bound;
             leftover = (u16)multiresult;
         }
@@ -129,7 +119,7 @@ u64 get_random_u64_uniform(struct random_state *state, u64 bound) {
     if (leftover < bound) {
         u64 threshold = -bound % bound;
         while (leftover < threshold) {
-            random =  get_random_u64(state);
+            random = get_random_u64(state);
             multiresult = random * bound;
             leftover = multiresult;
         }

test/.gitignore

@@ -1,4 +1,44 @@
 large_array_growth
 mallinfo
+mallinfo2
 malloc_info
 offset
+delete_type_size_mismatch
+double_free_large
+double_free_large_delayed
+double_free_small
+double_free_small_delayed
+invalid_free_protected
+invalid_free_small_region
+invalid_free_small_region_far
+invalid_free_unprotected
+read_after_free_large
+read_after_free_small
+read_zero_size
+string_overflow
+unaligned_free_large
+unaligned_free_small
+uninitialized_free
+uninitialized_malloc_usable_size
+uninitialized_realloc
+write_after_free_large
+write_after_free_large_reuse
+write_after_free_small
+write_after_free_small_reuse
+write_zero_size
+unaligned_malloc_usable_size_small
+invalid_malloc_usable_size_small
+invalid_malloc_usable_size_small_quarantine
+malloc_object_size
+malloc_object_size_offset
+invalid_malloc_object_size_small
+invalid_malloc_object_size_small_quarantine
+impossibly_large_malloc
+overflow_large_1_byte
+overflow_large_8_byte
+overflow_small_1_byte
+overflow_small_8_byte
+uninitialized_read_large
+uninitialized_read_small
+realloc_init
+__pycache__/

test/Makefile

@@ -1,21 +1,76 @@
 CONFIG_SLAB_CANARY := true
+CONFIG_EXTENDED_SIZE_CLASSES := true
+
+ifneq ($(VARIANT),)
+    $(error testing non-default variants not yet supported)
+endif
 
 ifeq (,$(filter $(CONFIG_SLAB_CANARY),true false))
     $(error CONFIG_SLAB_CANARY must be true or false)
 endif
 
-LDLIBS := -lpthread
+dir=$(dir $(realpath $(firstword $(MAKEFILE_LIST))))
 
-CPPFLAGS += \
-    -DSLAB_CANARY=$(CONFIG_SLAB_CANARY)
+CPPFLAGS := \
+    -D_GNU_SOURCE \
+    -DSLAB_CANARY=$(CONFIG_SLAB_CANARY) \
+    -DCONFIG_EXTENDED_SIZE_CLASSES=$(CONFIG_EXTENDED_SIZE_CLASSES)
+
+SHARED_FLAGS := -O3
+
+CFLAGS := -std=c17 $(SHARED_FLAGS) -Wmissing-prototypes
+CXXFLAGS := -std=c++17 -fsized-deallocation $(SHARED_FLAGS)
+LDFLAGS := -Wl,-L$(dir)../out,-R,$(dir)../out
+
+LDLIBS := -lpthread -lhardened_malloc
 
 EXECUTABLES := \
     offset \
     mallinfo \
+    mallinfo2 \
     malloc_info \
-    large_array_growth
+    large_array_growth \
+    double_free_large \
+    double_free_large_delayed \
+    double_free_small \
+    double_free_small_delayed \
+    unaligned_free_large \
+    unaligned_free_small \
+    read_after_free_large \
+    read_after_free_small \
+    write_after_free_large \
+    write_after_free_large_reuse \
+    write_after_free_small \
+    write_after_free_small_reuse \
+    read_zero_size \
+    write_zero_size \
+    invalid_free_protected \
+    invalid_free_unprotected \
+    invalid_free_small_region \
+    invalid_free_small_region_far \
+    uninitialized_read_small \
+    uninitialized_read_large \
+    uninitialized_free \
+    uninitialized_realloc \
+    uninitialized_malloc_usable_size \
+    overflow_large_1_byte \
+    overflow_large_8_byte \
+    overflow_small_1_byte \
+    overflow_small_8_byte \
+    string_overflow \
+    delete_type_size_mismatch \
+    unaligned_malloc_usable_size_small \
+    invalid_malloc_usable_size_small \
+    invalid_malloc_usable_size_small_quarantine \
+    malloc_object_size \
+    malloc_object_size_offset \
+    invalid_malloc_object_size_small \
+    invalid_malloc_object_size_small_quarantine \
+    impossibly_large_malloc \
+    realloc_init
 
 all: $(EXECUTABLES)
 
 clean:
 	rm -f $(EXECUTABLES)
+	rm -fr ./__pycache__

test/delete_type_size_mismatch.cc

@@ -1,11 +1,12 @@
 #include <stdint.h>
 
+#include "test_util.h"
+
 struct foo {
     uint64_t a, b, c, d;
 };
 
-__attribute__((optimize(0)))
-int main(void) {
+OPTNONE int main(void) {
     void *p = new char;
     struct foo *c = (struct foo *)p;
     delete c;

test/double_free_large.c

@@ -1,8 +1,9 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
-    void *p = malloc(128 * 1024);
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    void *p = malloc(256 * 1024);
     if (!p) {
         return 1;
     }

test/double_free_large_delayed.c

@@ -1,12 +1,13 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
-    void *p = malloc(128 * 1024);
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    void *p = malloc(256 * 1024);
     if (!p) {
         return 1;
     }
-    void *q = malloc(128 * 1024);
+    void *q = malloc(256 * 1024);
     if (!q) {
         return 1;
     }

test/double_free_small.c

@@ -1,7 +1,8 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     void *p = malloc(16);
     if (!p) {
         return 1;

test/double_free_small_delayed.c

@@ -1,7 +1,8 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     void *p = malloc(16);
     if (!p) {
         return 1;

test/impossibly_large_malloc.c

@@ -0,0 +1,8 @@
+#include <stdlib.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(-8);
+    return !(p == NULL);
+}

test/invalid_free_protected.c

@@ -2,8 +2,9 @@
 
 #include <sys/mman.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     free(malloc(16));
     char *p = mmap(NULL, 4096 * 16, PROT_NONE, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
     if (p == MAP_FAILED) {

test/invalid_free_small_region.c

@@ -1,7 +1,8 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(16);
     if (!p) {
         return 1;

test/invalid_free_small_region_far.c

@@ -1,7 +1,8 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(16);
     if (!p) {
         return 1;

test/invalid_free_unprotected.c

@@ -2,8 +2,9 @@
 
 #include <sys/mman.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     free(malloc(16));
     char *p = mmap(NULL, 4096 * 16, PROT_READ|PROT_WRITE, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
     if (p == MAP_FAILED) {

test/invalid_malloc_object_size_small.c

@@ -0,0 +1,15 @@
+#include <stdlib.h>
+
+#include "test_util.h"
+
+size_t malloc_object_size(void *ptr);
+
+OPTNONE int main(void) {
+    char *p = malloc(16);
+    if (!p) {
+        return 1;
+    }
+    char *q = p + 4096 * 4;
+    malloc_object_size(q);
+    return 0;
+}

test/invalid_malloc_object_size_small_quarantine.c

@@ -0,0 +1,15 @@
+#include <stdlib.h>
+
+#include "test_util.h"
+
+size_t malloc_object_size(void *ptr);
+
+OPTNONE int main(void) {
+    void *p = malloc(16);
+    if (!p) {
+        return 1;
+    }
+    free(p);
+    malloc_object_size(p);
+    return 0;
+}

test/invalid_malloc_usable_size_small.c

@@ -0,0 +1,13 @@
+#include <malloc.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(16);
+    if (!p) {
+        return 1;
+    }
+    char *q = p + 4096 * 4;
+    malloc_usable_size(q);
+    return 0;
+}

test/invalid_malloc_usable_size_small_quarantine.c

@@ -0,0 +1,13 @@
+#include <malloc.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    void *p = malloc(16);
+    if (!p) {
+        return 1;
+    }
+    free(p);
+    malloc_usable_size(p);
+    return 0;
+}

test/large_array_growth.c

@@ -1,8 +1,9 @@
 #include <stdlib.h>
 #include <string.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     void *p = NULL;
     size_t size = 256 * 1024;
 

test/mallinfo.c

@@ -1,21 +1,44 @@
+#include <stdlib.h>
+#include <stdio.h>
+
+#if defined(__GLIBC__) || defined(__ANDROID__)
 #include <malloc.h>
+#endif
 
-__attribute__((optimize(0)))
-int main(void) {
-    malloc(1024 * 1024 * 1024);
-    malloc(16);
-    malloc(32);
-    malloc(64);
+#include "test_util.h"
 
+static void print_mallinfo(void) {
+#if defined(__GLIBC__) || defined(__ANDROID__)
     struct mallinfo info = mallinfo();
-    printf("arena: %zu\n", info.arena);
-    printf("ordblks: %zu\n", info.ordblks);
-    printf("smblks: %zu\n", info.smblks);
-    printf("hblks: %zu\n", info.hblks);
-    printf("hblkhd: %zu\n", info.hblkhd);
-    printf("usmblks: %zu\n", info.usmblks);
-    printf("fsmblks: %zu\n", info.fsmblks);
-    printf("uordblks: %zu\n", info.uordblks);
-    printf("fordblks: %zu\n", info.fordblks);
-    printf("keepcost: %zu\n", info.keepcost);
+    printf("mallinfo:\n");
+    printf("arena: %zu\n", (size_t)info.arena);
+    printf("ordblks: %zu\n", (size_t)info.ordblks);
+    printf("smblks: %zu\n", (size_t)info.smblks);
+    printf("hblks: %zu\n", (size_t)info.hblks);
+    printf("hblkhd: %zu\n", (size_t)info.hblkhd);
+    printf("usmblks: %zu\n", (size_t)info.usmblks);
+    printf("fsmblks: %zu\n", (size_t)info.fsmblks);
+    printf("uordblks: %zu\n", (size_t)info.uordblks);
+    printf("fordblks: %zu\n", (size_t)info.fordblks);
+    printf("keepcost: %zu\n", (size_t)info.keepcost);
+#endif
+}
+
+OPTNONE int main(void) {
+    void *a[4];
+
+    a[0] = malloc(1024 * 1024 * 1024);
+    a[1] = malloc(16);
+    a[2] = malloc(32);
+    a[3] = malloc(64);
+
+    print_mallinfo();
+
+    free(a[0]);
+    free(a[1]);
+    free(a[2]);
+    free(a[3]);
+
+    printf("\n");
+    print_mallinfo();
 }

test/mallinfo2.c

@@ -0,0 +1,44 @@
+#include <stdio.h>
+#include <stdlib.h>
+
+#if defined(__GLIBC__)
+#include <malloc.h>
+#endif
+
+#include "test_util.h"
+
+static void print_mallinfo2(void) {
+#if defined(__GLIBC__)
+    struct mallinfo2 info = mallinfo2();
+    printf("mallinfo2:\n");
+    printf("arena: %zu\n", (size_t)info.arena);
+    printf("ordblks: %zu\n", (size_t)info.ordblks);
+    printf("smblks: %zu\n", (size_t)info.smblks);
+    printf("hblks: %zu\n", (size_t)info.hblks);
+    printf("hblkhd: %zu\n", (size_t)info.hblkhd);
+    printf("usmblks: %zu\n", (size_t)info.usmblks);
+    printf("fsmblks: %zu\n", (size_t)info.fsmblks);
+    printf("uordblks: %zu\n", (size_t)info.uordblks);
+    printf("fordblks: %zu\n", (size_t)info.fordblks);
+    printf("keepcost: %zu\n", (size_t)info.keepcost);
+#endif
+}
+
+OPTNONE int main(void) {
+    void *a[4];
+
+    a[0] = malloc(1024 * 1024 * 1024);
+    a[1] = malloc(16);
+    a[2] = malloc(32);
+    a[3] = malloc(64);
+
+    print_mallinfo2();
+
+    free(a[0]);
+    free(a[1]);
+    free(a[2]);
+    free(a[3]);
+
+    printf("\n");
+    print_mallinfo2();
+}

test/malloc_info.c

@@ -1,16 +1,22 @@
 #include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
 
+#if defined(__GLIBC__) || defined(__ANDROID__)
 #include <malloc.h>
+#endif
 
-__attribute__((optimize(0)))
-void leak_memory(void) {
-    (void)malloc(1024 * 1024 * 1024);
-    (void)malloc(16);
-    (void)malloc(32);
-    (void)malloc(4096);
+#include "test_util.h"
+#include "../util.h"
+
+OPTNONE static void leak_memory(void) {
+    (void)!malloc(1024 * 1024 * 1024);
+    (void)!malloc(16);
+    (void)!malloc(32);
+    (void)!malloc(4096);
 }
 
-void *do_work(void *p) {
+static void *do_work(UNUSED void *p) {
     leak_memory();
     return NULL;
 }
@@ -24,5 +30,7 @@ int main(void) {
         pthread_join(thread[i], NULL);
     }
 
+#if defined(__GLIBC__) || defined(__ANDROID__)
     malloc_info(0, stdout);
+#endif
 }

test/malloc_object_size.c

@@ -0,0 +1,12 @@
+#include <stdbool.h>
+#include <stdlib.h>
+
+#include "test_util.h"
+
+size_t malloc_object_size(void *ptr);
+
+OPTNONE int main(void) {
+    char *p = malloc(16);
+    size_t size = malloc_object_size(p);
+    return size != (SLAB_CANARY ? 24 : 32);
+}

test/malloc_object_size_offset.c

@@ -0,0 +1,12 @@
+#include <stdbool.h>
+#include <stdlib.h>
+
+#include "test_util.h"
+
+size_t malloc_object_size(void *ptr);
+
+OPTNONE int main(void) {
+    char *p = malloc(16);
+    size_t size = malloc_object_size(p + 5);
+    return size != (SLAB_CANARY ? 19 : 27);
+}

test/offset.c

@@ -3,7 +3,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 
-static unsigned size_classes[] = {
+static size_t size_classes[] = {
     /* large */ 4 * 1024 * 1024,
     /* 0 */ 0,
     /* 16 */ 16, 32, 48, 64, 80, 96, 112, 128,
@@ -13,7 +13,12 @@ static unsigned size_classes[] = {
     /* 256 */ 1280, 1536, 1792, 2048,
     /* 512 */ 2560, 3072, 3584, 4096,
     /* 1024 */ 5120, 6144, 7168, 8192,
-    /* 2048 */ 10240, 12288, 14336, 16384
+    /* 2048 */ 10240, 12288, 14336, 16384,
+#if CONFIG_EXTENDED_SIZE_CLASSES
+    /* 4096 */ 20480, 24576, 28672, 32768,
+    /* 8192 */ 40960, 49152, 57344, 65536,
+    /* 16384 */ 81920, 98304, 114688, 131072,
+#endif
 };
 
 #define N_SIZE_CLASSES (sizeof(size_classes) / sizeof(size_classes[0]))
@@ -27,9 +32,9 @@ int main(void) {
 
     void *p[N_SIZE_CLASSES];
     for (unsigned i = 0; i < N_SIZE_CLASSES; i++) {
-        unsigned size = size_classes[i];
+        size_t size = size_classes[i];
         p[i] = malloc(size);
-        if (!p) {
+        if (!p[i]) {
             return 1;
         }
         void *q = malloc(size);

test/overflow_large_1_byte.c

@@ -0,0 +1,15 @@
+#include <malloc.h>
+#include <stdlib.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(256 * 1024);
+    if (!p) {
+        return 1;
+    }
+    size_t size = malloc_usable_size(p);
+    *(p + size) = 0;
+    free(p);
+    return 0;
+}

test/overflow_large_8_byte.c

@@ -0,0 +1,15 @@
+#include <malloc.h>
+#include <stdlib.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(256 * 1024);
+    if (!p) {
+        return 1;
+    }
+    size_t size = malloc_usable_size(p);
+    *(p + size + 7) = 0;
+    free(p);
+    return 0;
+}

test/overflow_small_1_byte.c

@@ -0,0 +1,15 @@
+#include <malloc.h>
+#include <stdlib.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(8);
+    if (!p) {
+        return 1;
+    }
+    size_t size = malloc_usable_size(p);
+    *(p + size) = 1;
+    free(p);
+    return 0;
+}

test/overflow_small_8_byte.c

@@ -0,0 +1,16 @@
+#include <malloc.h>
+#include <stdlib.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(8);
+    if (!p) {
+        return 1;
+    }
+    size_t size = malloc_usable_size(p);
+    // XOR is used to avoid the test having a 1/256 chance to fail
+    *(p + size + 7) ^= 1;
+    free(p);
+    return 0;
+}

test/read_after_free_large.c

@@ -0,0 +1,21 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(256 * 1024);
+    if (!p) {
+        return 1;
+    }
+    memset(p, 'a', 16);
+    free(p);
+    for (size_t i = 0; i < 256 * 1024; i++) {
+        printf("%x\n", p[i]);
+        if (p[i] != '\0') {
+            return 1;
+        }
+    }
+    return 0;
+}

test/read_after_free_small.c

@@ -2,8 +2,9 @@
 #include <stdlib.h>
 #include <string.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(16);
     if (!p) {
         return 1;
@@ -12,6 +13,9 @@ int main(void) {
     free(p);
     for (size_t i = 0; i < 16; i++) {
         printf("%x\n", p[i]);
+        if (p[i] != '\0') {
+            return 1;
+        }
     }
     return 0;
 }

test/read_zero_size.c

@@ -1,8 +1,9 @@
 #include <stdlib.h>
 #include <stdio.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(0);
     if (!p) {
         return 1;

test/realloc_init.c

@@ -0,0 +1,33 @@
+#include <pthread.h>
+#include <stdlib.h>
+
+static void *thread_func(void *arg) {
+    arg = realloc(arg, 1024);
+    if (!arg) {
+        exit(EXIT_FAILURE);
+    }
+
+    free(arg);
+
+    return NULL;
+}
+
+int main(void) {
+    void *mem = realloc(NULL, 12);
+    if (!mem) {
+        return EXIT_FAILURE;
+    }
+
+    pthread_t thread;
+    int r = pthread_create(&thread, NULL, thread_func, mem);
+    if (r != 0) {
+        return EXIT_FAILURE;
+    }
+
+    r = pthread_join(thread, NULL);
+    if (r != 0) {
+        return EXIT_FAILURE;
+    }
+
+    return EXIT_SUCCESS;
+}

test/simple-memory-corruption/.gitignore

@@ -1,25 +0,0 @@
-delete_type_size_mismatch
-double_free_large
-double_free_large_delayed
-double_free_small
-double_free_small_delayed
-eight_byte_overflow_large
-eight_byte_overflow_small
-invalid_free_protected
-invalid_free_small_region
-invalid_free_small_region_far
-invalid_free_unprotected
-read_after_free_large
-read_after_free_small
-read_zero_size
-string_overflow
-unaligned_free_large
-unaligned_free_small
-uninitialized_free
-uninitialized_malloc_usable_size
-uninitialized_realloc
-write_after_free_large
-write_after_free_large_reuse
-write_after_free_small
-write_after_free_small_reuse
-write_zero_size

test/simple-memory-corruption/Makefile

@@ -1,31 +0,0 @@
-EXECUTABLES := \
-    double_free_large \
-    double_free_large_delayed \
-    double_free_small \
-    double_free_small_delayed \
-    unaligned_free_large \
-    unaligned_free_small \
-    read_after_free_large \
-    read_after_free_small \
-    write_after_free_large \
-    write_after_free_large_reuse \
-    write_after_free_small \
-    write_after_free_small_reuse \
-    read_zero_size \
-    write_zero_size \
-    invalid_free_protected \
-    invalid_free_unprotected \
-    invalid_free_small_region \
-    invalid_free_small_region_far \
-    uninitialized_free \
-    uninitialized_realloc \
-    uninitialized_malloc_usable_size \
-    eight_byte_overflow_small \
-    eight_byte_overflow_large \
-    string_overflow \
-    delete_type_size_mismatch
-
-all: $(EXECUTABLES)
-
-clean:
-	rm -f $(EXECUTABLES)

test/simple-memory-corruption/eight_byte_overflow_large.c

@@ -1,12 +0,0 @@
-#include <stdlib.h>
-
-__attribute__((optimize(0)))
-int main(void) {
-    char *p = malloc(128 * 1024);
-    if (!p) {
-        return 1;
-    }
-    *(p + 128 * 1024 + 7) = 0;
-    free(p);
-    return 0;
-}

test/simple-memory-corruption/eight_byte_overflow_small.c

@@ -1,12 +0,0 @@
-#include <stdlib.h>
-
-__attribute__((optimize(0)))
-int main(void) {
-    char *p = malloc(8);
-    if (!p) {
-        return 1;
-    }
-    *(p + 8 + 7) = 0;
-    free(p);
-    return 0;
-}

test/simple-memory-corruption/read_after_free_large.c

@@ -1,17 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-
-__attribute__((optimize(0)))
-int main(void) {
-    char *p = malloc(128 * 1024);
-    if (!p) {
-        return 1;
-    }
-    memset(p, 'a', 16);
-    free(p);
-    for (size_t i = 0; i < 128 * 1024; i++) {
-        printf("%x\n", p[i]);
-    }
-    return 0;
-}

test/simple-memory-corruption/write_after_free_large_reuse.c

@@ -1,14 +0,0 @@
-#include <stdlib.h>
-#include <string.h>
-
-__attribute__((optimize(0)))
-int main(void) {
-    char *p = malloc(128 * 1024);
-    if (!p) {
-        return 1;
-    }
-    free(p);
-    char *q = malloc(128 * 1024);
-    p[64 * 1024 + 1] = 'a';
-    return 0;
-}

test/string_overflow.c

@@ -4,8 +4,9 @@
 
 #include <malloc.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(16);
     if (!p) {
         return 1;

test/test_smc.py

@@ -0,0 +1,242 @@
+import os
+import subprocess
+import unittest
+
+
+class TestSimpleMemoryCorruption(unittest.TestCase):
+
+    @classmethod
+    def setUpClass(self):
+        self.dir = os.path.dirname(os.path.realpath(__file__))
+
+    def run_test(self, test_name):
+        sub = subprocess.Popen(self.dir + "/" + test_name,
+                               stdout=subprocess.PIPE, stderr=subprocess.PIPE)
+        stdout, stderr = sub.communicate()
+        return stdout, stderr, sub.returncode
+
+    def test_delete_type_size_mismatch(self):
+        _stdout, stderr, returncode = self.run_test(
+            "delete_type_size_mismatch")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode(
+            "utf-8"), "fatal allocator error: sized deallocation mismatch (small)\n")
+
+    def test_double_free_large_delayed(self):
+        _stdout, stderr, returncode = self.run_test(
+            "double_free_large_delayed")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid free\n")
+
+    def test_double_free_large(self):
+        _stdout, stderr, returncode = self.run_test("double_free_large")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid free\n")
+
+    def test_double_free_small_delayed(self):
+        _stdout, stderr, returncode = self.run_test(
+            "double_free_small_delayed")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: double free (quarantine)\n")
+
+    def test_double_free_small(self):
+        _stdout, stderr, returncode = self.run_test("double_free_small")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: double free (quarantine)\n")
+
+    def test_overflow_large_1_byte(self):
+        _stdout, _stderr, returncode = self.run_test(
+            "overflow_large_1_byte")
+        self.assertEqual(returncode, -11)
+
+    def test_overflow_large_8_byte(self):
+        _stdout, _stderr, returncode = self.run_test(
+            "overflow_large_8_byte")
+        self.assertEqual(returncode, -11)
+
+    def test_overflow_small_1_byte(self):
+        _stdout, stderr, returncode = self.run_test(
+            "overflow_small_1_byte")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: canary corrupted\n")
+
+    def test_overflow_small_8_byte(self):
+        _stdout, stderr, returncode = self.run_test(
+            "overflow_small_8_byte")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: canary corrupted\n")
+
+    def test_invalid_free_protected(self):
+        _stdout, stderr, returncode = self.run_test("invalid_free_protected")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid free\n")
+
+    def test_invalid_free_small_region_far(self):
+        _stdout, stderr, returncode = self.run_test(
+            "invalid_free_small_region_far")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode(
+            "utf-8"), "fatal allocator error: invalid free within a slab yet to be used\n")
+
+    def test_invalid_free_small_region(self):
+        _stdout, stderr, returncode = self.run_test(
+            "invalid_free_small_region")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: double free\n")
+
+    def test_invalid_free_unprotected(self):
+        _stdout, stderr, returncode = self.run_test("invalid_free_unprotected")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid free\n")
+
+    def test_invalid_malloc_usable_size_small_quarantene(self):
+        _stdout, stderr, returncode = self.run_test(
+            "invalid_malloc_usable_size_small_quarantine")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode(
+            "utf-8"), "fatal allocator error: invalid malloc_usable_size (quarantine)\n")
+
+    def test_invalid_malloc_usable_size_small(self):
+        _stdout, stderr, returncode = self.run_test(
+            "invalid_malloc_usable_size_small")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode(
+            "utf-8"), "fatal allocator error: invalid malloc_usable_size\n")
+
+    def test_read_after_free_large(self):
+        _stdout, _stderr, returncode = self.run_test("read_after_free_large")
+        self.assertEqual(returncode, -11)
+
+    def test_read_after_free_small(self):
+        stdout, _stderr, returncode = self.run_test("read_after_free_small")
+        self.assertEqual(returncode, 0)
+        self.assertEqual(stdout.decode("utf-8"),
+                         "0\n0\n0\n0\n0\n0\n0\n0\n0\n0\n0\n0\n0\n0\n0\n0\n")
+
+    def test_read_zero_size(self):
+        _stdout, _stderr, returncode = self.run_test("read_zero_size")
+        self.assertEqual(returncode, -11)
+
+    def test_string_overflow(self):
+        stdout, _stderr, returncode = self.run_test("string_overflow")
+        self.assertEqual(returncode, 0)
+        self.assertEqual(stdout.decode("utf-8"), "overflow by 0 bytes\n")
+
+    def test_unaligned_free_large(self):
+        _stdout, stderr, returncode = self.run_test("unaligned_free_large")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid free\n")
+
+    def test_unaligned_free_small(self):
+        _stdout, stderr, returncode = self.run_test("unaligned_free_small")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid unaligned free\n")
+
+    def test_unaligned_malloc_usable_size_small(self):
+        _stdout, stderr, returncode = self.run_test(
+            "unaligned_malloc_usable_size_small")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid unaligned malloc_usable_size\n")
+
+    def test_uninitialized_free(self):
+        _stdout, stderr, returncode = self.run_test("uninitialized_free")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid free\n")
+
+    def test_uninitialized_malloc_usable_size(self):
+        _stdout, stderr, returncode = self.run_test(
+            "uninitialized_malloc_usable_size")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid malloc_usable_size\n")
+
+    def test_uninitialized_realloc(self):
+        _stdout, stderr, returncode = self.run_test("uninitialized_realloc")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: invalid realloc\n")
+
+    def test_write_after_free_large_reuse(self):
+        _stdout, _stderr, returncode = self.run_test(
+            "write_after_free_large_reuse")
+        self.assertEqual(returncode, -11)
+
+    def test_write_after_free_large(self):
+        _stdout, _stderr, returncode = self.run_test("write_after_free_large")
+        self.assertEqual(returncode, -11)
+
+    def test_write_after_free_small_reuse(self):
+        _stdout, stderr, returncode = self.run_test(
+            "write_after_free_small_reuse")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: detected write after free\n")
+
+    def test_write_after_free_small(self):
+        _stdout, stderr, returncode = self.run_test("write_after_free_small")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode("utf-8"),
+                         "fatal allocator error: detected write after free\n")
+
+    def test_write_zero_size(self):
+        _stdout, _stderr, returncode = self.run_test("write_zero_size")
+        self.assertEqual(returncode, -11)
+
+    def test_malloc_object_size(self):
+        _stdout, _stderr, returncode = self.run_test("malloc_object_size")
+        self.assertEqual(returncode, 0)
+
+    def test_malloc_object_size_offset(self):
+        _stdout, _stderr, returncode = self.run_test(
+            "malloc_object_size_offset")
+        self.assertEqual(returncode, 0)
+
+    def test_invalid_malloc_object_size_small(self):
+        _stdout, stderr, returncode = self.run_test(
+            "invalid_malloc_object_size_small")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode(
+            "utf-8"), "fatal allocator error: invalid malloc_object_size\n")
+
+    def test_invalid_malloc_object_size_small_quarantine(self):
+        _stdout, stderr, returncode = self.run_test(
+            "invalid_malloc_object_size_small_quarantine")
+        self.assertEqual(returncode, -6)
+        self.assertEqual(stderr.decode(
+            "utf-8"), "fatal allocator error: invalid malloc_object_size (quarantine)\n")
+
+    def test_impossibly_large_malloc(self):
+        _stdout, stderr, returncode = self.run_test(
+            "impossibly_large_malloc")
+        self.assertEqual(returncode, 0)
+
+    def test_uninitialized_read_small(self):
+        _stdout, stderr, returncode = self.run_test(
+            "uninitialized_read_small")
+        self.assertEqual(returncode, 0)
+
+    def test_uninitialized_read_large(self):
+        _stdout, stderr, returncode = self.run_test(
+            "uninitialized_read_large")
+        self.assertEqual(returncode, 0)
+
+    def test_realloc_init(self):
+        _stdout, _stderr, returncode = self.run_test(
+            "realloc_init")
+        self.assertEqual(returncode, 0)
+
+if __name__ == '__main__':
+    unittest.main()

test/test_util.h

@@ -0,0 +1,10 @@
+#ifndef TEST_UTIL_H
+#define TEST_UTIL_H
+
+#ifdef __clang__
+#define OPTNONE __attribute__((optnone))
+#else
+#define OPTNONE __attribute__((optimize(0)))
+#endif
+
+#endif

test/unaligned_free_large.c

@@ -1,8 +1,9 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
-    char *p = malloc(128 * 1024);
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(256 * 1024);
     if (!p) {
         return 1;
     }

test/unaligned_free_small.c

@@ -1,7 +1,8 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(16);
     if (!p) {
         return 1;

test/unaligned_malloc_usable_size_small.c

@@ -0,0 +1,12 @@
+#include <malloc.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(16);
+    if (!p) {
+        return 1;
+    }
+    malloc_usable_size(p + 1);
+    return 0;
+}

test/uninitialized_free.c

@@ -1,7 +1,8 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     free((void *)1);
     return 0;
 }

test/uninitialized_malloc_usable_size.c

@@ -1,7 +1,8 @@
 #include <malloc.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     malloc_usable_size((void *)1);
     return 0;
 }

test/uninitialized_read_large.c

@@ -0,0 +1,14 @@
+#include <stdlib.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(256 * 1024);
+    for (unsigned i = 0; i < 256 * 1024; i++) {
+        if (p[i] != 0) {
+            return 1;
+        }
+    }
+    free(p);
+    return 0;
+}

test/uninitialized_read_small.c

@@ -0,0 +1,14 @@
+#include <stdlib.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(8);
+    for (unsigned i = 0; i < 8; i++) {
+        if (p[i] != 0) {
+            return 1;
+        }
+    }
+    free(p);
+    return 0;
+}

test/uninitialized_realloc.c

@@ -1,7 +1,8 @@
 #include <stdlib.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     void *p = realloc((void *)1, 16);
     if (!p) {
         return 1;

test/write_after_free_large.c

@@ -1,9 +1,9 @@
 #include <stdlib.h>
-#include <string.h>
 
-__attribute__((optimize(0)))
-int main(void) {
-    char *p = malloc(128 * 1024);
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(256 * 1024);
     if (!p) {
         return 1;
     }

test/write_after_free_large_reuse.c

@@ -0,0 +1,16 @@
+#include <stdlib.h>
+#include <string.h>
+
+#include "test_util.h"
+#include "../util.h"
+
+OPTNONE int main(void) {
+    char *p = malloc(256 * 1024);
+    if (!p) {
+        return 1;
+    }
+    free(p);
+    UNUSED char *q = malloc(256 * 1024);
+    p[64 * 1024 + 1] = 'a';
+    return 0;
+}

test/write_after_free_small.c

@@ -1,8 +1,8 @@
 #include <stdlib.h>
-#include <string.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(128);
     if (!p) {
         return 1;

test/write_after_free_small_reuse.c

@@ -1,14 +1,15 @@
 #include <stdlib.h>
-#include <string.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+#include "../util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(128);
     if (!p) {
         return 1;
     }
     free(p);
-    char *q = malloc(128);
+    UNUSED char *q = malloc(128);
 
     p[65] = 'a';
 

test/write_zero_size.c

@@ -1,8 +1,8 @@
 #include <stdlib.h>
-#include <stdio.h>
 
-__attribute__((optimize(0)))
-int main(void) {
+#include "test_util.h"
+
+OPTNONE int main(void) {
     char *p = malloc(0);
     if (!p) {
         return 1;

util.c

@@ -4,8 +4,15 @@
 
 #include <unistd.h>
 
+#ifdef __ANDROID__
+#include <async_safe/log.h>
+int mallopt(int param, int value);
+#define M_BIONIC_RESTORE_DEFAULT_SIGABRT_HANDLER (-1003)
+#endif
+
 #include "util.h"
 
+#ifndef __ANDROID__
 static int write_full(int fd, const char *buf, size_t length) {
     do {
         ssize_t bytes_written = write(fd, buf, length);
@@ -21,11 +28,17 @@ static int write_full(int fd, const char *buf, size_t length) {
 
     return 0;
 }
+#endif
 
 COLD noreturn void fatal_error(const char *s) {
+#ifdef __ANDROID__
+    mallopt(M_BIONIC_RESTORE_DEFAULT_SIGABRT_HANDLER, 0);
+    async_safe_fatal("hardened_malloc: fatal allocator error: %s", s);
+#else
     const char *prefix = "fatal allocator error: ";
     (void)(write_full(STDERR_FILENO, prefix, strlen(prefix)) != -1 &&
         write_full(STDERR_FILENO, s, strlen(s)) != -1 &&
         write_full(STDERR_FILENO, "\n", 1));
     abort();
+#endif
 }

util.h

@@ -1,11 +1,17 @@
 #ifndef UTIL_H
 #define UTIL_H
 
+#include <stdbool.h>
+#include <stddef.h>
 #include <stdint.h>
-#include <stdnoreturn.h>
+
+// C11 noreturn doesn't work in C++
+#define noreturn __attribute__((noreturn))
 
 #define likely(x) __builtin_expect(!!(x), 1)
+#define likely51(x) __builtin_expect_with_probability(!!(x), 1, 0.51)
 #define unlikely(x) __builtin_expect(!!(x), 0)
+#define unlikely51(x) __builtin_expect_with_probability(!!(x), 0, 0.51)
 
 #define min(x, y) ({ \
     __typeof__(x) _x = (x); \
@@ -26,11 +32,12 @@
 #define STRINGIFY(s) #s
 #define ALIAS(f) __attribute__((alias(STRINGIFY(f))))
 
-static inline int ffzl(long x) {
-    return __builtin_ffsl(~x);
-}
-
-COLD noreturn void fatal_error(const char *s);
+// supported since GCC 15
+#if __has_attribute (nonstring)
+#  define NONSTRING __attribute__ ((nonstring))
+#else
+#  define NONSTRING
+#endif
 
 typedef uint8_t u8;
 typedef uint16_t u16;
@@ -38,28 +45,50 @@ typedef uint32_t u32;
 typedef uint64_t u64;
 typedef unsigned __int128 u128;
 
-// use __register_atfork directly to avoid linking with libpthread for glibc < 2.28
-#ifdef __GLIBC__
-#if !__GLIBC_PREREQ(2, 28)
-extern void *__dso_handle;
-extern int __register_atfork(void (*)(void), void (*)(void), void (*)(void), void *);
-#define atfork(prepare, parent, child) __register_atfork(prepare, parent, child, __dso_handle)
-#endif
-#endif
+#define U64_WIDTH 64
 
-#ifndef atfork
-#define atfork pthread_atfork
-#endif
+static inline int ffz64(u64 x) {
+    return __builtin_ffsll(~x);
+}
 
-#ifdef CONFIG_SEAL_METADATA
+// parameter must not be 0
+static inline int clz64(u64 x) {
+    return __builtin_clzll(x);
+}
+
+// parameter must not be 0
+static inline u64 log2u64(u64 x) {
+    return U64_WIDTH - clz64(x) - 1;
+}
+
+static inline size_t align(size_t size, size_t align) {
+    size_t mask = align - 1;
+    return (size + mask) & ~mask;
+}
+
+// u4_arr_{set,get} are helper functions for using u8 array as an array of unsigned 4-bit values.
+
+// val is treated as a 4-bit value
+static inline void u4_arr_set(u8 *arr, size_t idx, u8 val) {
+    size_t off = idx >> 1;
+    size_t shift = (idx & 1) << 2;
+    u8 mask = (u8) (0xf0 >> shift);
+    arr[off] = (arr[off] & mask) | (val << shift);
+}
+
+static inline u8 u4_arr_get(const u8 *arr, size_t idx) {
+    size_t off = idx >> 1;
+    size_t shift = (idx & 1) << 2;
+    return (u8) ((arr[off] >> shift) & 0xf);
+}
+
+COLD noreturn void fatal_error(const char *s);
+
+#if CONFIG_SEAL_METADATA
 
 #ifdef __GLIBC__
-#if __GLIBC_PREREQ(2, 27)
 #define USE_PKEY
-#endif
-#endif
-
-#ifndef USE_PKEY
+#else
 #error "CONFIG_SEAL_METADATA requires Memory Protection Key support"
 #endif