[libcxx-commits] [libcxx] [ASan][libc++] Turn on ASan annotations for short strings (PR #75882)
via libcxx-commits
libcxx-commits at lists.llvm.org
Tue Jan 9 16:59:27 PST 2024
https://github.com/AdvenamTacet updated https://github.com/llvm/llvm-project/pull/75882
>From 8f71de6fe389d8b9658b7aeaf49bd48d552c2ead Mon Sep 17 00:00:00 2001
From: Advenam Tacet <advenam.tacet at trailofbits.com>
Date: Mon, 18 Dec 2023 23:30:29 +0100
Subject: [PATCH] [ASan][libc++] Turn on ASan annotations for short strings
This commit turns on ASan annotations in `std::basic_string` for short stings (SSO case).
Originally suggested here: https://reviews.llvm.org/D147680
String annotations added here: https://github.com/llvm/llvm-project/pull/72677
Annotating `std::basic_string` with default allocator is implemented in https://github.com/llvm/llvm-project/pull/72677 but annotations for short strings (SSO - Short String Optimization) are turned off there. This commit turns them on. This also removes `_LIBCPP_SHORT_STRING_ANNOTATIONS_ALLOWED`, because we do not plan to support turning on and off short string annotations.
Support in ASan API exists since https://github.com/llvm/llvm-project/commit/dd1b7b797a116eed588fd752fbe61d34deeb24e4.
You can turn off annotations for a specific allocator based on changes from https://github.com/llvm/llvm-project/commit/2fa1bec7a20bb23f2e6620085adb257dafaa3be0.
This PR is a part of a series of patches extending AddressSanitizer C++ container overflow detection capabilities by adding annotations, similar to those existing in `std::vector` and `std::deque` collections. These enhancements empower ASan to effectively detect instances where the instrumented program attempts to access memory within a collection's internal allocation that remains unused. This includes cases where access occurs before or after the stored elements in `std::deque`, or between the `std::basic_string`'s size (including the null terminator) and capacity bounds.
The introduction of these annotations was spurred by a real-world software bug discovered by Trail of Bits, involving an out-of-bounds memory access during the comparison of two strings using the `std::equals` function. This function was taking iterators (`iter1_begin`, `iter1_end`, `iter2_begin`) to perform the comparison, using a custom comparison function. When the `iter1` object exceeded the length of `iter2`, an out-of-bounds read could occur on the `iter2` object. Container sanitization, upon enabling these annotations, would effectively identify and flag this potential vulnerability.
If you have any questions, please email:
advenam.tacet at trailofbits.com
disconnect3d at trailofbits.com
---
libcxx/include/string | 3 +-
.../asan_deque_integration.pass.cpp | 180 ++++++++++++++++++
.../strings/basic.string/asan_short.pass.cpp | 54 ++++++
.../asan_vector_integration.pass.cpp | 180 ++++++++++++++++++
libcxx/test/support/asan_testing.h | 16 +-
5 files changed, 420 insertions(+), 13 deletions(-)
create mode 100644 libcxx/test/libcxx/containers/strings/basic.string/asan_deque_integration.pass.cpp
create mode 100644 libcxx/test/libcxx/containers/strings/basic.string/asan_short.pass.cpp
create mode 100644 libcxx/test/libcxx/containers/strings/basic.string/asan_vector_integration.pass.cpp
diff --git a/libcxx/include/string b/libcxx/include/string
index c676182fba8bac..b7b3bf07f9a08b 100644
--- a/libcxx/include/string
+++ b/libcxx/include/string
@@ -659,7 +659,6 @@ _LIBCPP_PUSH_MACROS
#else
# define _LIBCPP_STRING_INTERNAL_MEMORY_ACCESS
#endif
-#define _LIBCPP_SHORT_STRING_ANNOTATIONS_ALLOWED false
_LIBCPP_BEGIN_NAMESPACE_STD
@@ -1899,7 +1898,7 @@ private:
// ASan: short string is poisoned if and only if this function returns true.
_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 bool __asan_short_string_is_annotated() const _NOEXCEPT {
- return _LIBCPP_SHORT_STRING_ANNOTATIONS_ALLOWED && !__libcpp_is_constant_evaluated();
+ return !__libcpp_is_constant_evaluated();
}
_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 void __annotate_new(size_type __current_size) const _NOEXCEPT {
diff --git a/libcxx/test/libcxx/containers/strings/basic.string/asan_deque_integration.pass.cpp b/libcxx/test/libcxx/containers/strings/basic.string/asan_deque_integration.pass.cpp
new file mode 100644
index 00000000000000..cb5bbd67b14a83
--- /dev/null
+++ b/libcxx/test/libcxx/containers/strings/basic.string/asan_deque_integration.pass.cpp
@@ -0,0 +1,180 @@
+//===----------------------------------------------------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+// REQUIRES: asan
+// UNSUPPORTED: c++03
+
+#include <cassert>
+#include <array>
+#include "test_macros.h"
+#include "asan_testing.h" // includes deque and string - don't do it before
+
+// This tests exists to check if strings work well with deque, as those
+// may be partialy annotated, we cannot simply call
+// is_double_ended_contiguous_container_asan_correct, as it assumes that
+// object memory inside is not annotated, so we check everything in a more careful way.
+
+template <typename D>
+bool verify_inside(D const& d) {
+ for (size_t i = 0; i < d.size(); ++i) {
+ if (!is_string_asan_correct(d[i]))
+ return false;
+ }
+
+ return true;
+}
+
+template <typename S, size_t N>
+S get_s(char c) {
+ S s;
+ for (size_t i = 0; i < N; ++i)
+ s.push_back(c);
+
+ return s;
+}
+
+template <class C, class S>
+void test_string() {
+ size_t const N = sizeof(S) < 256 ? (4096 / sizeof(S)) : 16;
+
+ {
+ C d1a(1), d1b(N), d1c(N + 1), d1d(32 * N);
+ assert(verify_inside(d1a));
+ assert(verify_inside(d1b));
+ assert(verify_inside(d1c));
+ assert(verify_inside(d1d));
+ }
+ {
+ C d2;
+ for (size_t i = 0; i < 16 * N; ++i) {
+ d2.push_back(get_s<S, 1>(i % 10 + 'a'));
+ assert(verify_inside(d2));
+ d2.push_back(get_s<S, 222>(i % 10 + 'b'));
+ assert(verify_inside(d2));
+
+ d2.pop_front();
+ assert(verify_inside(d2));
+ }
+ }
+ {
+ C d3;
+ for (size_t i = 0; i < 16 * N; ++i) {
+ d3.push_front(get_s<S, 1>(i % 10 + 'a'));
+ assert(verify_inside(d3));
+ d3.push_front(get_s<S, 222>(i % 10 + 'b'));
+ assert(verify_inside(d3));
+
+ d3.pop_back();
+ assert(verify_inside(d3));
+ }
+ }
+ {
+ C d4;
+ for (size_t i = 0; i < 16 * N; ++i) {
+ // When there is no SSO, all elements inside should not be poisoned,
+ // so we can verify deque poisoning.
+ d4.push_front(get_s<S, 333>(i % 10 + 'a'));
+ assert(verify_inside(d4));
+ assert(is_double_ended_contiguous_container_asan_correct(d4));
+ d4.push_back(get_s<S, 222>(i % 10 + 'b'));
+ assert(verify_inside(d4));
+ assert(is_double_ended_contiguous_container_asan_correct(d4));
+ }
+ }
+ {
+ C d5;
+ for (size_t i = 0; i < 5 * N; ++i) {
+ // In d4 we never had poisoned memory inside deque.
+ // Here we start with SSO, so part of the inside of the container,
+ // will be poisoned.
+ d5.push_front(S());
+ assert(verify_inside(d5));
+ }
+ for (size_t i = 0; i < d5.size(); ++i) {
+ // We change the size to have long string.
+ // Memory owne by deque should not be poisoned by string.
+ d5[i].resize(1000);
+ assert(verify_inside(d5));
+ }
+
+ assert(is_double_ended_contiguous_container_asan_correct(d5));
+
+ d5.erase(d5.begin() + 2);
+ assert(verify_inside(d5));
+
+ d5.erase(d5.end() - 2);
+ assert(verify_inside(d5));
+
+ assert(is_double_ended_contiguous_container_asan_correct(d5));
+ }
+ {
+ C d6a;
+ assert(is_double_ended_contiguous_container_asan_correct(d6a));
+
+ C d6b(N + 2, get_s<S, 1000>('a'));
+ d6b.push_front(get_s<S, 1001>('b'));
+ while (!d6b.empty()) {
+ d6b.pop_back();
+ assert(is_double_ended_contiguous_container_asan_correct(d6b));
+ }
+
+ C d6c(N + 2, get_s<S, 1002>('c'));
+ while (!d6c.empty()) {
+ d6c.pop_back();
+ assert(is_double_ended_contiguous_container_asan_correct(d6c));
+ }
+ }
+ {
+ C d7(9 * N + 2);
+
+ d7.insert(d7.begin() + 1, S());
+ assert(verify_inside(d7));
+
+ d7.insert(d7.end() - 3, S());
+ assert(verify_inside(d7));
+
+ d7.insert(d7.begin() + 2 * N, get_s<S, 1>('a'));
+ assert(verify_inside(d7));
+
+ d7.insert(d7.end() - 2 * N, get_s<S, 1>('b'));
+ assert(verify_inside(d7));
+
+ d7.insert(d7.begin() + 2 * N, 3 * N, get_s<S, 1>('c'));
+ assert(verify_inside(d7));
+
+ // It may not be short for big element types, but it will be checked correctly:
+ d7.insert(d7.end() - 2 * N, 3 * N, get_s<S, 2>('d'));
+ assert(verify_inside(d7));
+
+ d7.erase(d7.begin() + 2);
+ assert(verify_inside(d7));
+
+ d7.erase(d7.end() - 2);
+ assert(verify_inside(d7));
+ }
+}
+
+template <class S>
+void test_container() {
+ test_string<std::deque<S, std::allocator<S>>, S>();
+ test_string<std::deque<S, min_allocator<S>>, S>();
+ test_string<std::deque<S, safe_allocator<S>>, S>();
+}
+
+int main(int, char**) {
+ // Those tests support only types based on std::basic_string.
+ test_container<std::string>();
+ test_container<std::wstring>();
+#if TEST_STD_VER >= 11
+ test_container<std::u16string>();
+ test_container<std::u32string>();
+#endif
+#if TEST_STD_VER >= 20
+ test_container<std::u8string>();
+#endif
+}
diff --git a/libcxx/test/libcxx/containers/strings/basic.string/asan_short.pass.cpp b/libcxx/test/libcxx/containers/strings/basic.string/asan_short.pass.cpp
new file mode 100644
index 00000000000000..adeb7c4abdc552
--- /dev/null
+++ b/libcxx/test/libcxx/containers/strings/basic.string/asan_short.pass.cpp
@@ -0,0 +1,54 @@
+//===----------------------------------------------------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+// REQUIRES: asan
+// UNSUPPORTED: c++03
+
+// <string>
+
+// Basic test if ASan annotations work for short strings.
+
+#include <string>
+#include <cassert>
+#include <cstdlib>
+
+#include "asan_testing.h"
+#include "min_allocator.h"
+#include "test_iterators.h"
+#include "test_macros.h"
+
+extern "C" void __sanitizer_set_death_callback(void (*callback)(void));
+
+void do_exit() { exit(0); }
+
+int main(int, char**) {
+ {
+ typedef cpp17_input_iterator<char*> MyInputIter;
+ // Should not trigger ASan.
+ std::basic_string<char, std::char_traits<char>, safe_allocator<char>> v;
+ char i[] = {'a', 'b', 'c', 'd'};
+
+ v.insert(v.begin(), MyInputIter(i), MyInputIter(i + 4));
+ assert(v[0] == 'a');
+ assert(is_string_asan_correct(v));
+ }
+
+ __sanitizer_set_death_callback(do_exit);
+ {
+ using T = char;
+ using C = std::basic_string<T, std::char_traits<T>, safe_allocator<T>>;
+ const T t[] = {'a', 'b', 'c', 'd', 'e', 'f', 'g'};
+ C c(std::begin(t), std::end(t));
+ assert(is_string_asan_correct(c));
+ assert(__sanitizer_verify_contiguous_container(c.data(), c.data() + c.size() + 1, c.data() + c.capacity() + 1) !=
+ 0);
+ volatile T foo = c[c.size() + 1]; // should trigger ASAN. Use volatile to prevent being optimized away.
+ assert(false); // if we got here, ASAN didn't trigger
+ ((void)foo);
+ }
+}
diff --git a/libcxx/test/libcxx/containers/strings/basic.string/asan_vector_integration.pass.cpp b/libcxx/test/libcxx/containers/strings/basic.string/asan_vector_integration.pass.cpp
new file mode 100644
index 00000000000000..0ae1b9f65dfd49
--- /dev/null
+++ b/libcxx/test/libcxx/containers/strings/basic.string/asan_vector_integration.pass.cpp
@@ -0,0 +1,180 @@
+//===----------------------------------------------------------------------===//
+//
+// 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
+//
+//===----------------------------------------------------------------------===//
+
+// REQUIRES: asan
+// UNSUPPORTED: c++03
+
+#include <cassert>
+#include <array>
+#include "test_macros.h"
+#include "asan_testing.h" // includes vector and string - don't do it before
+
+// This tests exists to check if strings work well with vector, as those
+// may be partialy annotated, we cannot simply call
+// is_contiguous_container_asan_correct, as it assumes that
+// object memory inside is not annotated, so we check everything in a more careful way.
+
+template <typename D>
+bool verify_inside(D const& d) {
+ for (size_t i = 0; i < d.size(); ++i) {
+ if (!is_string_asan_correct(d[i]))
+ return false;
+ }
+
+ return true;
+}
+
+template <typename S, size_t N>
+S get_s(char c) {
+ S s;
+ for (size_t i = 0; i < N; ++i)
+ s.push_back(c);
+
+ return s;
+}
+
+template <class C, class S>
+void test_string() {
+ size_t const N = sizeof(S) < 256 ? (4096 / sizeof(S)) : 16;
+
+ {
+ C d1a(1), d1b(N), d1c(N + 1), d1d(32 * N);
+ assert(verify_inside(d1a));
+ assert(verify_inside(d1b));
+ assert(verify_inside(d1c));
+ assert(verify_inside(d1d));
+ }
+ {
+ C d2;
+ for (size_t i = 0; i < 16 * N; ++i) {
+ d2.push_back(get_s<S, 1>(i % 10 + 'a'));
+ assert(verify_inside(d2));
+ d2.push_back(get_s<S, 222>(i % 10 + 'b'));
+ assert(verify_inside(d2));
+
+ d2.erase(d2.cbegin());
+ assert(verify_inside(d2));
+ }
+ }
+ {
+ C d3;
+ for (size_t i = 0; i < 16 * N; ++i) {
+ d3.push_back(get_s<S, 1>(i % 10 + 'a'));
+ assert(verify_inside(d3));
+ d3.push_back(get_s<S, 222>(i % 10 + 'b'));
+ assert(verify_inside(d3));
+
+ d3.pop_back();
+ assert(verify_inside(d3));
+ }
+ }
+ {
+ C d4;
+ for (size_t i = 0; i < 16 * N; ++i) {
+ // When there is no SSO, all elements inside should not be poisoned,
+ // so we can verify vector poisoning.
+ d4.push_back(get_s<S, 333>(i % 10 + 'a'));
+ assert(verify_inside(d4));
+ assert(is_contiguous_container_asan_correct(d4));
+ d4.push_back(get_s<S, 222>(i % 10 + 'b'));
+ assert(verify_inside(d4));
+ assert(is_contiguous_container_asan_correct(d4));
+ }
+ }
+ {
+ C d5;
+ for (size_t i = 0; i < 5 * N; ++i) {
+ // In d4 we never had poisoned memory inside vector.
+ // Here we start with SSO, so part of the inside of the container,
+ // will be poisoned.
+ d5.push_back(S());
+ assert(verify_inside(d5));
+ }
+ for (size_t i = 0; i < d5.size(); ++i) {
+ // We change the size to have long string.
+ // Memory owne by vector should not be poisoned by string.
+ d5[i].resize(1000);
+ assert(verify_inside(d5));
+ }
+
+ assert(is_contiguous_container_asan_correct(d5));
+
+ d5.erase(d5.begin() + 2);
+ assert(verify_inside(d5));
+
+ d5.erase(d5.end() - 2);
+ assert(verify_inside(d5));
+
+ assert(is_contiguous_container_asan_correct(d5));
+ }
+ {
+ C d6a;
+ assert(is_contiguous_container_asan_correct(d6a));
+
+ C d6b(N + 2, get_s<S, 1000>('a'));
+ d6b.push_back(get_s<S, 1001>('b'));
+ while (!d6b.empty()) {
+ d6b.pop_back();
+ assert(is_contiguous_container_asan_correct(d6b));
+ }
+
+ C d6c(N + 2, get_s<S, 1002>('c'));
+ while (!d6c.empty()) {
+ d6c.pop_back();
+ assert(is_contiguous_container_asan_correct(d6c));
+ }
+ }
+ {
+ C d7(9 * N + 2);
+
+ d7.insert(d7.begin() + 1, S());
+ assert(verify_inside(d7));
+
+ d7.insert(d7.end() - 3, S());
+ assert(verify_inside(d7));
+
+ d7.insert(d7.begin() + 2 * N, get_s<S, 1>('a'));
+ assert(verify_inside(d7));
+
+ d7.insert(d7.end() - 2 * N, get_s<S, 1>('b'));
+ assert(verify_inside(d7));
+
+ d7.insert(d7.begin() + 2 * N, 3 * N, get_s<S, 1>('c'));
+ assert(verify_inside(d7));
+
+ // It may not be short for big element types, but it will be checked correctly:
+ d7.insert(d7.end() - 2 * N, 3 * N, get_s<S, 2>('d'));
+ assert(verify_inside(d7));
+
+ d7.erase(d7.begin() + 2);
+ assert(verify_inside(d7));
+
+ d7.erase(d7.end() - 2);
+ assert(verify_inside(d7));
+ }
+}
+
+template <class S>
+void test_container() {
+ test_string<std::vector<S, std::allocator<S>>, S>();
+ test_string<std::vector<S, min_allocator<S>>, S>();
+ test_string<std::vector<S, safe_allocator<S>>, S>();
+}
+
+int main(int, char**) {
+ // Those tests support only types based on std::basic_string.
+ test_container<std::string>();
+ test_container<std::wstring>();
+#if TEST_STD_VER >= 11
+ test_container<std::u16string>();
+ test_container<std::u32string>();
+#endif
+#if TEST_STD_VER >= 20
+ test_container<std::u8string>();
+#endif
+}
diff --git a/libcxx/test/support/asan_testing.h b/libcxx/test/support/asan_testing.h
index 2dfec5c42b00b2..5b07b97a2d151c 100644
--- a/libcxx/test/support/asan_testing.h
+++ b/libcxx/test/support/asan_testing.h
@@ -74,17 +74,11 @@ TEST_CONSTEXPR bool is_string_asan_correct(const std::basic_string<ChrT, TraitsT
if (TEST_IS_CONSTANT_EVALUATED)
return true;
- if (!is_string_short(c) || _LIBCPP_SHORT_STRING_ANNOTATIONS_ALLOWED) {
- if (std::is_same<Alloc, std::allocator<ChrT>>::value)
- return __sanitizer_verify_contiguous_container(c.data(), c.data() + c.size() + 1, c.data() + c.capacity() + 1) !=
- 0;
- else
- return __sanitizer_verify_contiguous_container(
- c.data(), c.data() + c.capacity() + 1, c.data() + c.capacity() + 1) != 0;
- } else {
- return __sanitizer_verify_contiguous_container(std::addressof(c), std::addressof(c) + 1, std::addressof(c) + 1) !=
- 0;
- }
+ if (std::is_same<Alloc, std::allocator<ChrT>>::value)
+ return __sanitizer_verify_contiguous_container(c.data(), c.data() + c.size() + 1, c.data() + c.capacity() + 1) != 0;
+ else
+ return __sanitizer_verify_contiguous_container(
+ c.data(), c.data() + c.capacity() + 1, c.data() + c.capacity() + 1) != 0;
}
#else
# include <string>
More information about the libcxx-commits
mailing list