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",
@@ -17,21 +16,25 @@ common_cflags = [
     "-DSLAB_CANARY=true",
     "-DSLAB_QUARANTINE_RANDOM_LENGTH=1",
     "-DSLAB_QUARANTINE_QUEUE_LENGTH=1",
+    "-DCONFIG_EXTENDED_SIZE_CLASSES=true",
+    "-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",
 }
 
@@ -44,13 +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

@@ -1,54 +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_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 = 1
+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) $(SHARED_FLAGS)
-LDFLAGS := -Wl,--as-needed,-z,defs,-z,relro,-z,now,-z,nodlopen,-z,text
-TIDY_CHECKS := -checks=bugprone-*,-bugprone-macro-parentheses,cert-*,clang-analyzer-*,readability-*,-readability-inconsistent-declaration-parameter-name,-readability-named-parameter
+ifeq ($(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 $(CONFIG_ZERO_ON_FREE),true false))
@@ -67,13 +73,32 @@ ifeq (,$(filter $(CONFIG_SLAB_CANARY),true false))
     $(error CONFIG_SLAB_CANARY must be true or false)
 endif
 
+ifeq (,$(filter $(CONFIG_EXTENDED_SIZE_CLASSES),true false))
+    $(error CONFIG_EXTENDED_SIZE_CLASSES must be true or false)
+endif
+
+ifeq (,$(filter $(CONFIG_LARGE_SIZE_CLASSES),true false))
+    $(error CONFIG_LARGE_SIZE_CLASSES must be true or false)
+endif
+
+ifeq (,$(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=$(CONFIG_SLOT_RANDOMIZE) \
     -DSLAB_CANARY=$(CONFIG_SLAB_CANARY) \
     -DSLAB_QUARANTINE_RANDOM_LENGTH=$(CONFIG_SLAB_QUARANTINE_RANDOM_LENGTH) \
     -DSLAB_QUARANTINE_QUEUE_LENGTH=$(CONFIG_SLAB_QUARANTINE_QUEUE_LENGTH) \
+    -DCONFIG_EXTENDED_SIZE_CLASSES=$(CONFIG_EXTENDED_SIZE_CLASSES) \
+    -DCONFIG_LARGE_SIZE_CLASSES=$(CONFIG_LARGE_SIZE_CLASSES) \
     -DGUARD_SLABS_INTERVAL=$(CONFIG_GUARD_SLABS_INTERVAL) \
     -DGUARD_SIZE_DIVISOR=$(CONFIG_GUARD_SIZE_DIVISOR) \
     -DREGION_QUARANTINE_RANDOM_LENGTH=$(CONFIG_REGION_QUARANTINE_RANDOM_LENGTH) \
@@ -81,23 +106,43 @@ CPPFLAGS += \
     -DREGION_QUARANTINE_SKIP_THRESHOLD=$(CONFIG_REGION_QUARANTINE_SKIP_THRESHOLD) \
     -DFREE_SLABS_QUARANTINE_RANDOM_LENGTH=$(CONFIG_FREE_SLABS_QUARANTINE_RANDOM_LENGTH) \
     -DCONFIG_CLASS_REGION_SIZE=$(CONFIG_CLASS_REGION_SIZE) \
-    -DN_ARENA=$(CONFIG_N_ARENA)
+    -DN_ARENA=$(CONFIG_N_ARENA) \
+    -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

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

@@ -10,7 +10,10 @@ size_classes = [
     1280, 1536, 1792, 2048,
     2560, 3072, 3584, 4096,
     5120, 6144, 7168, 8192,
-    10240, 12288, 14336, 16384
+    10240, 12288, 14336, 16384,
+    20480, 24576, 28672, 32768,
+    40960, 49152, 57344, 65536,
+    81920, 98304, 114688, 131072,
 ]
 
 size_class_slots = [
@@ -21,7 +24,10 @@ size_class_slots = [
     16, 16, 16, 16,
     8, 8, 8, 8,
     8, 8, 8, 8,
-    6, 5, 4, 4
+    6, 5, 4, 4,
+    2, 2, 2, 2,
+    1, 1, 1, 1,
+    1, 1, 1, 1,
 ]
 
 fragmentation = [100 - 1 / 16 * 100]
@@ -43,7 +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

@@ -4,6 +4,9 @@
 
 #include "chacha.h"
 
+// ChaCha8
+static const unsigned rounds = 8;
+
 #define U8C(v) (v##U)
 #define U32C(v) (v##U)
 
@@ -38,7 +41,7 @@
     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);
@@ -63,55 +66,52 @@ void chacha_ivsetup(chacha_ctx *x, const u8 *iv) {
 }
 
 void chacha_keystream_bytes(chacha_ctx *x, u8 *c, u32 bytes) {
-    u32 x0, x1, x2, x3, x4, x5, x6, x7, x8, x9, x10, x11, x12, x13, x14, x15;
-    u32 j0, j1, j2, j3, j4, j5, j6, j7, j8, j9, j10, j11, j12, j13, j14, j15;
-    u8 *ctarget;
-    u8 tmp[64];
-    unsigned i;
-
     if (!bytes) {
         return;
     }
 
-    j0 = x->input[0];
-    j1 = x->input[1];
-    j2 = x->input[2];
-    j3 = x->input[3];
-    j4 = x->input[4];
-    j5 = x->input[5];
-    j6 = x->input[6];
-    j7 = x->input[7];
-    j8 = x->input[8];
-    j9 = x->input[9];
-    j10 = x->input[10];
-    j11 = x->input[11];
-    j12 = x->input[12];
-    j13 = x->input[13];
-    j14 = x->input[14];
-    j15 = x->input[15];
+    u8 *ctarget;
+    u8 tmp[64];
+
+    u32 j0 = x->input[0];
+    u32 j1 = x->input[1];
+    u32 j2 = x->input[2];
+    u32 j3 = x->input[3];
+    u32 j4 = x->input[4];
+    u32 j5 = x->input[5];
+    u32 j6 = x->input[6];
+    u32 j7 = x->input[7];
+    u32 j8 = x->input[8];
+    u32 j9 = x->input[9];
+    u32 j10 = x->input[10];
+    u32 j11 = x->input[11];
+    u32 j12 = x->input[12];
+    u32 j13 = x->input[13];
+    u32 j14 = x->input[14];
+    u32 j15 = x->input[15];
 
     for (;;) {
         if (bytes < 64) {
             ctarget = c;
             c = tmp;
         }
-        x0 = j0;
-        x1 = j1;
-        x2 = j2;
-        x3 = j3;
-        x4 = j4;
-        x5 = j5;
-        x6 = j6;
-        x7 = j7;
-        x8 = j8;
-        x9 = j9;
-        x10 = j10;
-        x11 = j11;
-        x12 = j12;
-        x13 = j13;
-        x14 = j14;
-        x15 = j15;
-        for (i = 8; i > 0; i -= 2) {
+        u32 x0 = j0;
+        u32 x1 = j1;
+        u32 x2 = j2;
+        u32 x3 = j3;
+        u32 x4 = j4;
+        u32 x5 = j5;
+        u32 x6 = j6;
+        u32 x7 = j7;
+        u32 x8 = j8;
+        u32 x9 = j9;
+        u32 x10 = j10;
+        u32 x11 = j11;
+        u32 x12 = j12;
+        u32 x13 = j13;
+        u32 x14 = j14;
+        u32 x15 = j15;
+        for (unsigned i = rounds; i > 0; i -= 2) {
             QUARTERROUND(x0, x4, x8, x12)
             QUARTERROUND(x1, x5, x9, x13)
             QUARTERROUND(x2, x6, x10, x14)
@@ -163,7 +163,7 @@ void chacha_keystream_bytes(chacha_ctx *x, u8 *c, u32 bytes) {
 
         if (bytes <= 64) {
             if (bytes < 64) {
-                for (i = 0; i < bytes; ++i) {
+                for (unsigned i = 0; i < bytes; ++i) {
                     ctarget[i] = c[i];
                 }
             }

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);
 
@@ -63,35 +72,43 @@ void h_malloc_stats(void);
 #if defined(__GLIBC__) || defined(__ANDROID__)
 struct mallinfo h_mallinfo(void);
 #endif
+#ifndef __ANDROID__
 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);
-void *h_valloc(size_t size);
-void *h_pvalloc(size_t size);
-void h_cfree(void *ptr);
+#ifndef __ANDROID__
+__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) __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
 
-// custom extensions
+// 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.
@@ -105,6 +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

@@ -4,13 +4,14 @@
 #include "pages.h"
 #include "util.h"
 
-static uintptr_t alignment_ceiling(uintptr_t s, uintptr_t alignment) {
-    return ((s) + (alignment - 1)) & ((~alignment) + 1);
+static bool add_guards(size_t size, size_t guard_size, size_t *total_size) {
+    return __builtin_add_overflow(size, guard_size, total_size) ||
+        __builtin_add_overflow(*total_size, guard_size, total_size);
 }
 
 void *allocate_pages(size_t usable_size, size_t guard_size, bool unprotect, const char *name) {
     size_t real_size;
-    if (unlikely(__builtin_add_overflow(usable_size, guard_size * 2, &real_size))) {
+    if (unlikely(add_guards(usable_size, guard_size, &real_size))) {
         errno = ENOMEM;
         return NULL;
     }
@@ -28,7 +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;
@@ -41,7 +42,7 @@ void *allocate_pages_aligned(size_t usable_size, size_t alignment, size_t guard_
     }
 
     size_t real_alloc_size;
-    if (unlikely(__builtin_add_overflow(alloc_size, guard_size * 2, &real_alloc_size))) {
+    if (unlikely(add_guards(alloc_size, guard_size, &real_alloc_size))) {
         errno = ENOMEM;
         return NULL;
     }
@@ -54,7 +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;
 
@@ -81,5 +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

@@ -5,20 +5,10 @@
 #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 > 0) {
+    while (size) {
         ssize_t r;
 
         do {
@@ -39,19 +29,48 @@ void random_state_init(struct random_state *state) {
     get_random_seed(rnd, sizeof(rnd));
     chacha_keysetup(&state->ctx, rnd);
     chacha_ivsetup(&state->ctx, rnd + CHACHA_KEY_SIZE);
-    chacha_keystream_bytes(&state->ctx, state->cache, RANDOM_CACHE_SIZE);
-    state->index = 0;
+    state->index = RANDOM_CACHE_SIZE;
+    state->reseed = 0;
+}
+
+void random_state_init_from_random_state(struct random_state *state, struct random_state *source) {
+    u8 rnd[CHACHA_KEY_SIZE + CHACHA_IV_SIZE];
+    get_random_bytes(source, rnd, sizeof(rnd));
+    chacha_keysetup(&state->ctx, rnd);
+    chacha_ivsetup(&state->ctx, rnd + CHACHA_KEY_SIZE);
+    state->index = RANDOM_CACHE_SIZE;
     state->reseed = 0;
 }
 
 static void refill(struct random_state *state) {
-    if (state->reseed < RANDOM_RESEED_SIZE) {
-        chacha_keystream_bytes(&state->ctx, state->cache, RANDOM_CACHE_SIZE);
-        state->index = 0;
-        state->reseed += RANDOM_CACHE_SIZE;
-    } else {
+    if (state->reseed >= RANDOM_RESEED_SIZE) {
         random_state_init(state);
     }
+    chacha_keystream_bytes(&state->ctx, state->cache, RANDOM_CACHE_SIZE);
+    state->index = 0;
+    state->reseed += RANDOM_CACHE_SIZE;
+}
+
+void get_random_bytes(struct random_state *state, void *buf, size_t size) {
+    // avoid needless copying to and from the cache as an optimization
+    if (size > RANDOM_CACHE_SIZE / 2) {
+        chacha_keystream_bytes(&state->ctx, buf, size);
+        return;
+    }
+
+    while (size) {
+        if (state->index == RANDOM_CACHE_SIZE) {
+            refill(state);
+        }
+
+        size_t remaining = RANDOM_CACHE_SIZE - state->index;
+        size_t copy_size = min(size, remaining);
+        memcpy(buf, state->cache + state->index, copy_size);
+        state->index += copy_size;
+
+        buf = (char *)buf + copy_size;
+        size -= copy_size;
+    }
 }
 
 u16 get_random_u16(struct random_state *state) {
@@ -73,7 +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;
         }
@@ -100,7 +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;
         }

random.h

@@ -15,6 +15,8 @@ struct random_state {
 };
 
 void random_state_init(struct random_state *state);
+void random_state_init_from_random_state(struct random_state *state, struct random_state *source);
+void get_random_bytes(struct random_state *state, void *buf, size_t size);
 u16 get_random_u16(struct random_state *state);
 u16 get_random_u16_uniform(struct random_state *state, u16 bound);
 u64 get_random_u64(struct random_state *state);

test/.gitignore

@@ -1 +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

@@ -1,16 +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
 
-CPPFLAGS += \
-    -DSLAB_CANARY=$(CONFIG_SLAB_CANARY)
+dir=$(dir $(realpath $(firstword $(MAKEFILE_LIST))))
+
+CPPFLAGS := \
+    -D_GNU_SOURCE \
+    -DSLAB_CANARY=$(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
+    offset \
+    mallinfo \
+    mallinfo2 \
+    malloc_info \
+    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

@@ -0,0 +1,18 @@
+#include <stdlib.h>
+#include <string.h>
+
+#include "test_util.h"
+
+OPTNONE int main(void) {
+    void *p = NULL;
+    size_t size = 256 * 1024;
+
+    for (unsigned i = 0; i < 20; i++) {
+        p = realloc(p, size);
+        if (!p) {
+            return 1;
+        }
+        memset(p, 'a', size);
+        size = size * 3 / 2;
+    }
+}

test/mallinfo.c

@@ -0,0 +1,44 @@
+#include <stdlib.h>
+#include <stdio.h>
+
+#if defined(__GLIBC__) || defined(__ANDROID__)
+#include <malloc.h>
+#endif
+
+#include "test_util.h"
+
+static void print_mallinfo(void) {
+#if defined(__GLIBC__) || defined(__ANDROID__)
+    struct mallinfo info = mallinfo();
+    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

@@ -0,0 +1,36 @@
+#include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+#if defined(__GLIBC__) || defined(__ANDROID__)
+#include <malloc.h>
+#endif
+
+#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);
+}
+
+static void *do_work(UNUSED 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);
+    }
+
+#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