[libc-commits] [libc] [libc] add hashtable fuzzing (PR #87949)

Schrodinger ZHU Yifan via libc-commits libc-commits at lists.llvm.org
Fri Apr 19 14:57:06 PDT 2024


https://github.com/SchrodingerZhu updated https://github.com/llvm/llvm-project/pull/87949

>From b4d7cbc29a448ff490c43440f0ef4b351657782c Mon Sep 17 00:00:00 2001
From: Schrodinger ZHU Yifan <yifanzhu at rochester.edu>
Date: Sun, 7 Apr 2024 18:26:16 -0400
Subject: [PATCH 1/5] [libc] add hashtable fuzzing

---
 libc/fuzzing/__support/CMakeLists.txt     |   9 ++
 libc/fuzzing/__support/hashtable_fuzz.cpp | 157 ++++++++++++++++++++++
 2 files changed, 166 insertions(+)
 create mode 100644 libc/fuzzing/__support/hashtable_fuzz.cpp

diff --git a/libc/fuzzing/__support/CMakeLists.txt b/libc/fuzzing/__support/CMakeLists.txt
index d4f6db71fdd849..b5d2b488447fc5 100644
--- a/libc/fuzzing/__support/CMakeLists.txt
+++ b/libc/fuzzing/__support/CMakeLists.txt
@@ -5,3 +5,12 @@ add_libc_fuzzer(
   DEPENDS
     libc.src.__support.big_int
 )
+
+add_libc_fuzzer(
+  hashtable_fuzz
+  SRCS
+    hashtable_fuzz.cpp
+  DEPENDS
+    libc.src.__support.HashTable.table
+    libc.src.string.memcpy
+)
diff --git a/libc/fuzzing/__support/hashtable_fuzz.cpp b/libc/fuzzing/__support/hashtable_fuzz.cpp
new file mode 100644
index 00000000000000..4b862b03b9d309
--- /dev/null
+++ b/libc/fuzzing/__support/hashtable_fuzz.cpp
@@ -0,0 +1,157 @@
+#include "src/__support/CPP/new.h"
+#include "src/__support/CPP/optional.h"
+#include "src/__support/HashTable/table.h"
+#include "src/string/memcpy.h"
+#include <search.h>
+#include <stdint.h>
+namespace LIBC_NAMESPACE {
+
+enum class Action { Find, Insert, CrossCheck };
+static uint8_t *global_buffer = nullptr;
+static size_t remaining = 0;
+
+static cpp::optional<uint8_t> next_u8() {
+  if (remaining == 0)
+    return cpp::nullopt;
+  uint8_t result = *global_buffer;
+  global_buffer++;
+  remaining--;
+  return result;
+}
+
+static cpp::optional<uint64_t> next_uint64() {
+  uint64_t result;
+  if (remaining < sizeof(result))
+    return cpp::nullopt;
+  memcpy(&result, global_buffer, sizeof(result));
+  global_buffer += sizeof(result);
+  remaining -= sizeof(result);
+  return result;
+}
+
+static cpp::optional<Action> next_action() {
+  if (cpp::optional<uint8_t> action = next_u8()) {
+    switch (*action % 3) {
+    case 0:
+      return Action::Find;
+    case 1:
+      return Action::Insert;
+    case 2:
+      return Action::CrossCheck;
+    }
+  }
+  return cpp::nullopt;
+}
+
+static cpp::optional<char *> next_cstr() {
+  char *result = reinterpret_cast<char *>(global_buffer);
+  if (cpp::optional<uint64_t> len = next_uint64()) {
+    uint64_t length;
+    for (length = 0; length < *len % 128; length++) {
+      if (length >= remaining)
+        return cpp::nullopt;
+      if (*global_buffer == '\0')
+        break;
+    }
+    if (length >= remaining)
+      return cpp::nullopt;
+    global_buffer[length] = '\0';
+    global_buffer += length + 1;
+    remaining -= length + 1;
+    return result;
+  }
+  return cpp::nullopt;
+}
+
+#define GET_VAL(op)                                                            \
+  __extension__({                                                              \
+    auto val = op();                                                           \
+    if (!val)                                                                  \
+      return 0;                                                                \
+    *val;                                                                      \
+  })
+
+template <typename Fn> struct CleanUpHook {
+  cpp::optional<Fn> fn;
+  ~CleanUpHook() {
+    if (fn)
+      (*fn)();
+  }
+  CleanUpHook(Fn fn) : fn(cpp::move(fn)) {}
+  CleanUpHook(const CleanUpHook &) = delete;
+  CleanUpHook(CleanUpHook &&other) : fn(cpp::move(other.fn)) {
+    other.fn = cpp::nullopt;
+  }
+};
+
+#define register_cleanup(ID, ...)                                              \
+  auto cleanup_hook##ID = __extension__({                                      \
+    auto a = __VA_ARGS__;                                                      \
+    CleanUpHook<decltype(a)>{a};                                               \
+  });
+
+static void trap_with_message(const char *msg) { __builtin_trap(); }
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
+  AllocChecker ac;
+  global_buffer = static_cast<uint8_t *>(::operator new(size, ac));
+  register_cleanup(0, [global_buffer = global_buffer, size] {
+    ::operator delete(global_buffer, size);
+  });
+  if (!ac)
+    return 0;
+  memcpy(global_buffer, data, size);
+
+  remaining = size;
+  uint64_t size_a = GET_VAL(next_uint64) % 256;
+  uint64_t size_b = GET_VAL(next_uint64) % 256;
+  uint64_t rand_a = GET_VAL(next_uint64);
+  uint64_t rand_b = GET_VAL(next_uint64);
+  internal::HashTable *table_a = internal::HashTable::allocate(size_a, rand_a);
+  register_cleanup(1, [&table_a] { internal::HashTable::deallocate(table_a); });
+  internal::HashTable *table_b = internal::HashTable::allocate(size_b, rand_b);
+  register_cleanup(2, [&table_b] { internal::HashTable::deallocate(table_b); });
+  if (!table_a || !table_b)
+    return 0;
+  for (;;) {
+    Action action = GET_VAL(next_action);
+    switch (action) {
+    case Action::Find: {
+      const char *key = GET_VAL(next_cstr);
+      if (!key)
+        return 0;
+      if (static_cast<bool>(table_a->find(key)) !=
+          static_cast<bool>(table_b->find(key)))
+        trap_with_message(key);
+      break;
+    }
+    case Action::Insert: {
+      char *key = GET_VAL(next_cstr);
+      if (!key)
+        return 0;
+      ENTRY *a = internal::HashTable::insert(table_a, ENTRY{key, key});
+      ENTRY *b = internal::HashTable::insert(table_b, ENTRY{key, key});
+      if (a->data != b->data)
+        __builtin_trap();
+      break;
+    }
+    case Action::CrossCheck: {
+      for (ENTRY a : *table_a) {
+        if (const ENTRY *b = table_b->find(a.key)) {
+          if (a.data != b->data)
+            __builtin_trap();
+        }
+      }
+      for (ENTRY b : *table_b) {
+        if (const ENTRY *a = table_a->find(b.key)) {
+          if (a->data != b.data)
+            __builtin_trap();
+        }
+      }
+      break;
+    }
+    }
+  }
+}
+
+} // namespace LIBC_NAMESPACE

>From 9d681e18e17317d52de705ebf9a332e894971c09 Mon Sep 17 00:00:00 2001
From: Schrodinger ZHU Yifan <yifanzhu at rochester.edu>
Date: Sun, 7 Apr 2024 18:35:17 -0400
Subject: [PATCH 2/5] remove extra code

---
 libc/fuzzing/__support/hashtable_fuzz.cpp | 14 ++++++++------
 1 file changed, 8 insertions(+), 6 deletions(-)

diff --git a/libc/fuzzing/__support/hashtable_fuzz.cpp b/libc/fuzzing/__support/hashtable_fuzz.cpp
index 4b862b03b9d309..758c8d1aae01bf 100644
--- a/libc/fuzzing/__support/hashtable_fuzz.cpp
+++ b/libc/fuzzing/__support/hashtable_fuzz.cpp
@@ -108,9 +108,15 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
   uint64_t rand_a = GET_VAL(next_uint64);
   uint64_t rand_b = GET_VAL(next_uint64);
   internal::HashTable *table_a = internal::HashTable::allocate(size_a, rand_a);
-  register_cleanup(1, [&table_a] { internal::HashTable::deallocate(table_a); });
+  register_cleanup(1, [&table_a] {
+    if (table_a)
+      internal::HashTable::deallocate(table_a);
+  });
   internal::HashTable *table_b = internal::HashTable::allocate(size_b, rand_b);
-  register_cleanup(2, [&table_b] { internal::HashTable::deallocate(table_b); });
+  register_cleanup(2, [&table_b] {
+    if (table_b)
+      internal::HashTable::deallocate(table_b);
+  });
   if (!table_a || !table_b)
     return 0;
   for (;;) {
@@ -118,8 +124,6 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
     switch (action) {
     case Action::Find: {
       const char *key = GET_VAL(next_cstr);
-      if (!key)
-        return 0;
       if (static_cast<bool>(table_a->find(key)) !=
           static_cast<bool>(table_b->find(key)))
         trap_with_message(key);
@@ -127,8 +131,6 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
     }
     case Action::Insert: {
       char *key = GET_VAL(next_cstr);
-      if (!key)
-        return 0;
       ENTRY *a = internal::HashTable::insert(table_a, ENTRY{key, key});
       ENTRY *b = internal::HashTable::insert(table_b, ENTRY{key, key});
       if (a->data != b->data)

>From 5a72c4d1d63dde2e26f71afa566bc77759188150 Mon Sep 17 00:00:00 2001
From: Schrodinger ZHU Yifan <yifanzhu at rochester.edu>
Date: Mon, 8 Apr 2024 09:18:48 -0400
Subject: [PATCH 3/5] make style consistent

---
 libc/fuzzing/__support/hashtable_fuzz.cpp | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/libc/fuzzing/__support/hashtable_fuzz.cpp b/libc/fuzzing/__support/hashtable_fuzz.cpp
index 758c8d1aae01bf..d5c64970b53234 100644
--- a/libc/fuzzing/__support/hashtable_fuzz.cpp
+++ b/libc/fuzzing/__support/hashtable_fuzz.cpp
@@ -63,7 +63,7 @@ static cpp::optional<char *> next_cstr() {
   return cpp::nullopt;
 }
 
-#define GET_VAL(op)                                                            \
+#define get_value(op)                                                          \
   __extension__({                                                              \
     auto val = op();                                                           \
     if (!val)                                                                  \
@@ -103,10 +103,10 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
   memcpy(global_buffer, data, size);
 
   remaining = size;
-  uint64_t size_a = GET_VAL(next_uint64) % 256;
-  uint64_t size_b = GET_VAL(next_uint64) % 256;
-  uint64_t rand_a = GET_VAL(next_uint64);
-  uint64_t rand_b = GET_VAL(next_uint64);
+  uint64_t size_a = get_value(next_uint64) % 256;
+  uint64_t size_b = get_value(next_uint64) % 256;
+  uint64_t rand_a = get_value(next_uint64);
+  uint64_t rand_b = get_value(next_uint64);
   internal::HashTable *table_a = internal::HashTable::allocate(size_a, rand_a);
   register_cleanup(1, [&table_a] {
     if (table_a)
@@ -120,17 +120,17 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
   if (!table_a || !table_b)
     return 0;
   for (;;) {
-    Action action = GET_VAL(next_action);
+    Action action = get_value(next_action);
     switch (action) {
     case Action::Find: {
-      const char *key = GET_VAL(next_cstr);
+      const char *key = get_value(next_cstr);
       if (static_cast<bool>(table_a->find(key)) !=
           static_cast<bool>(table_b->find(key)))
         trap_with_message(key);
       break;
     }
     case Action::Insert: {
-      char *key = GET_VAL(next_cstr);
+      char *key = get_value(next_cstr);
       ENTRY *a = internal::HashTable::insert(table_a, ENTRY{key, key});
       ENTRY *b = internal::HashTable::insert(table_b, ENTRY{key, key});
       if (a->data != b->data)

>From 4c87431cb4b53eb7528c406150126a5da585a3ad Mon Sep 17 00:00:00 2001
From: Schrodinger ZHU Yifan <yifanzhu at rochester.edu>
Date: Mon, 8 Apr 2024 12:50:59 -0400
Subject: [PATCH 4/5] make information more concentrated

---
 libc/fuzzing/__support/hashtable_fuzz.cpp | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/libc/fuzzing/__support/hashtable_fuzz.cpp b/libc/fuzzing/__support/hashtable_fuzz.cpp
index d5c64970b53234..f346d2b79c6db4 100644
--- a/libc/fuzzing/__support/hashtable_fuzz.cpp
+++ b/libc/fuzzing/__support/hashtable_fuzz.cpp
@@ -45,9 +45,9 @@ static cpp::optional<Action> next_action() {
 
 static cpp::optional<char *> next_cstr() {
   char *result = reinterpret_cast<char *>(global_buffer);
-  if (cpp::optional<uint64_t> len = next_uint64()) {
+  if (cpp::optional<uint8_t> len = next_u8()) {
     uint64_t length;
-    for (length = 0; length < *len % 128; length++) {
+    for (length = 0; length < *len; length++) {
       if (length >= remaining)
         return cpp::nullopt;
       if (*global_buffer == '\0')
@@ -87,7 +87,7 @@ template <typename Fn> struct CleanUpHook {
 #define register_cleanup(ID, ...)                                              \
   auto cleanup_hook##ID = __extension__({                                      \
     auto a = __VA_ARGS__;                                                      \
-    CleanUpHook<decltype(a)>{a};                                               \
+    CleanUpHook<decltype(a)>(cpp::move(a));                                    \
   });
 
 static void trap_with_message(const char *msg) { __builtin_trap(); }
@@ -103,8 +103,8 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
   memcpy(global_buffer, data, size);
 
   remaining = size;
-  uint64_t size_a = get_value(next_uint64) % 256;
-  uint64_t size_b = get_value(next_uint64) % 256;
+  uint64_t size_a = get_value(next_u8);
+  uint64_t size_b = get_value(next_u8);
   uint64_t rand_a = get_value(next_uint64);
   uint64_t rand_b = get_value(next_uint64);
   internal::HashTable *table_a = internal::HashTable::allocate(size_a, rand_a);

>From 43486bd306ac6723393a73e36a55ec1141d504db Mon Sep 17 00:00:00 2001
From: Schrodinger ZHU Yifan <yifanzhu at rochester.edu>
Date: Fri, 19 Apr 2024 17:56:50 -0400
Subject: [PATCH 5/5] add missing license headers

---
 libc/fuzzing/__support/hashtable_fuzz.cpp | 11 +++++++++++
 libc/fuzzing/__support/uint_fuzz.cpp      | 11 +++++++++++
 2 files changed, 22 insertions(+)

diff --git a/libc/fuzzing/__support/hashtable_fuzz.cpp b/libc/fuzzing/__support/hashtable_fuzz.cpp
index f346d2b79c6db4..4a726b229b7818 100644
--- a/libc/fuzzing/__support/hashtable_fuzz.cpp
+++ b/libc/fuzzing/__support/hashtable_fuzz.cpp
@@ -1,3 +1,14 @@
+//===-- hashtable_fuzz.cpp ------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+///
+/// Fuzzing test for llvm-libc hashtable implementations.
+///
+//===----------------------------------------------------------------------===//
 #include "src/__support/CPP/new.h"
 #include "src/__support/CPP/optional.h"
 #include "src/__support/HashTable/table.h"
diff --git a/libc/fuzzing/__support/uint_fuzz.cpp b/libc/fuzzing/__support/uint_fuzz.cpp
index 07149f511b8386..109375f84da780 100644
--- a/libc/fuzzing/__support/uint_fuzz.cpp
+++ b/libc/fuzzing/__support/uint_fuzz.cpp
@@ -1,3 +1,14 @@
+//===-- uint_fuzz.cpp -----------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+///
+/// Fuzzing test for llvm-libc unsigned integer utilities.
+///
+//===----------------------------------------------------------------------===//
 #include "src/__support/CPP/bit.h"
 #include "src/__support/big_int.h"
 #include "src/string/memory_utils/inline_memcpy.h"



More information about the libc-commits mailing list