Compare commits

...

8 Commits

Author SHA1 Message Date
yhirose ec5ce17929 Release v0.43.3 2026-05-04 16:19:49 +09:00
yhirose f6524c0802 Drop Str2tagTest unit test that broke split / no-exceptions builds
The test referenced detail::can_compress_content_type, which lives below
the split BORDER in httplib.h and is therefore not visible to test.cc in
test_split / Windows-CMake builds. EXPECT_NO_THROW also expanded to a
try/catch that would not compile under -fno-exceptions. The OSS-Fuzz
reproducer in test/fuzzing/corpus already serves as the regression test
for #508087118 and is exercised by make fuzz_test.
2026-05-01 22:20:41 +09:00
yhirose 35c4026c7f Make fuzz_test robust to missing corpus files
When a glob like clusterfuzz-testcase-minimized-foo_fuzzer-* did not
match anything, bash passed the literal pattern through. The standalone
runner then tried to open it, tellg() returned -1, and the resulting
size_t cast (SIZE_MAX) crashed std::vector with length_error. This made
fuzz_test fail loudly during bisects to commits before a corpus file
landed. Filter each glob through a -f test so unmatched patterns are
silently skipped with a "(no XXX corpus)" notice, mirroring what was
already done for url_parser_fuzzer.
2026-05-01 21:50:26 +09:00
yhirose 40e18460bc Document str2tag_core's compile-time-only role 2026-05-01 21:46:13 +09:00
yhirose 92aecf85d8 Fix OSS-Fuzz #508087118: avoid stack overflow in str2tag
str2tag_core is recursive (one frame per character), so a long runtime
input such as a fuzzer-supplied Content-Type would overflow the stack.
Rewrite the runtime entry point str2tag() iteratively while keeping the
recursive constexpr str2tag_core for compile-time UDL evaluation. The
hash output is unchanged for all inputs.
2026-05-01 21:39:46 +09:00
yhirose b223e29778 Add OSS-Fuzz #508370122 reproducer to client_fuzzer corpus
Same root cause as #508342856 (fixed in 2d2efe4): an oversized
Content-Length value (here 4467440718547775) caused res.body.reserve()
to attempt a multi-petabyte allocation. The UBSAN fuzzer job surfaced
it as a std::bad_alloc-driven abort, while the ASAN job for #508342856
reported it as allocation-size-too-big. The payload_max_length_ cap
introduced in 2d2efe4 already addresses both.
2026-05-01 21:34:03 +09:00
yhirose 2d2efe46da Fix OSS-Fuzz #508342856: cap Content-Length reservation by payload_max_length_
A malicious or malformed server response with an enormous Content-Length
header (e.g. 20000000000) caused the client to call res.body.reserve(len)
with the untrusted value, triggering OOM before read_content's
payload_max_length_ check could take effect. Cap the pre-reservation
at payload_max_length_, since reading more than that is never useful.
2026-05-01 21:28:57 +09:00
yhirose cae753425e Run all fuzzers via make fuzz_test 2026-05-01 21:28:45 +09:00
8 changed files with 137 additions and 10 deletions
+3
View File
@@ -43,6 +43,9 @@ test/test_mbedtls
test/test_wolfssl
test/test_no_tls
test/server_fuzzer
test/client_fuzzer
test/header_parser_fuzzer
test/url_parser_fuzzer
test/test_proxy
test/test_proxy_mbedtls
test/test_proxy_wolfssl
+1 -1
View File
@@ -4,7 +4,7 @@ langs = ["en", "ja"]
[site]
title = "cpp-httplib"
version = "0.43.2"
version = "0.43.3"
hostname = "https://yhirose.github.io"
base_path = "/cpp-httplib"
footer_message = "© 2026 Yuji Hirose. All rights reserved."
+25 -4
View File
@@ -8,8 +8,8 @@
#ifndef CPPHTTPLIB_HTTPLIB_H
#define CPPHTTPLIB_HTTPLIB_H
#define CPPHTTPLIB_VERSION "0.43.2"
#define CPPHTTPLIB_VERSION_NUM "0x002b02"
#define CPPHTTPLIB_VERSION "0.43.3"
#define CPPHTTPLIB_VERSION_NUM "0x002b03"
#ifdef _WIN32
#if defined(_WIN32_WINNT) && _WIN32_WINNT < 0x0A00
@@ -6361,6 +6361,10 @@ inline void get_remote_ip_and_port(socket_t sock, std::string &ip, int &port) {
}
}
// Recursive form retained so operator""_t below can compute hashes for
// switch-case labels at compile time (C++11 constexpr forbids loops). Do not
// call from runtime paths with arbitrary-length inputs — use str2tag()
// instead, which is iterative and stack-safe.
inline constexpr unsigned int str2tag_core(const char *s, size_t l,
unsigned int h) {
return (l == 0)
@@ -6374,7 +6378,16 @@ inline constexpr unsigned int str2tag_core(const char *s, size_t l,
}
inline unsigned int str2tag(const std::string &s) {
return str2tag_core(s.data(), s.size(), 0);
// Iterative form of str2tag_core: the recursive constexpr version is kept
// for compile-time UDL evaluation of short string literals, but at runtime
// we may receive arbitrarily long inputs (e.g. fuzzed Content-Type) that
// would blow the stack with one frame per character.
unsigned int h = 0;
for (auto c : s) {
h = (((std::numeric_limits<unsigned int>::max)() >> 6) & h * 33) ^
static_cast<unsigned char>(c);
}
return h;
}
namespace udl {
@@ -13632,7 +13645,15 @@ inline bool ClientImpl::process_request(Stream &strm, Request &req,
output_error_log(error, &req);
return false;
}
res.body.reserve(static_cast<size_t>(len));
// Cap the reservation by payload_max_length_ to avoid OOM when a
// hostile or malformed server sends an enormous Content-Length.
// The actual body read below is bounded by payload_max_length_,
// so reserving more than that is never useful.
auto reserve_len = static_cast<size_t>(len);
if (payload_max_length_ > 0 && reserve_len > payload_max_length_) {
reserve_len = payload_max_length_;
}
res.body.reserve(reserve_len);
}
}
+27 -5
View File
@@ -251,16 +251,38 @@ test_proxy_mbedtls : test_proxy.cc ../httplib.h Makefile cert.pem
test_proxy_wolfssl : test_proxy.cc ../httplib.h Makefile cert.pem
$(CXX) -o $@ -I.. $(CXXFLAGS) test_proxy.cc $(TEST_ARGS_WOLFSSL)
# Runs server_fuzzer.cc based on value of $(LIB_FUZZING_ENGINE).
# Usage: make fuzz_test LIB_FUZZING_ENGINE=/path/to/libFuzzer
fuzz_test: server_fuzzer
./server_fuzzer fuzzing/corpus/*
# Runs all fuzz harnesses based on the value of $(LIB_FUZZING_ENGINE).
# By default LIB_FUZZING_ENGINE is standalone_fuzz_target_runner.o, so each
# fuzzer is replayed over its regression corpus.
# Override for actual fuzzing:
# make fuzz_test LIB_FUZZING_ENGINE=/path/to/libFuzzer
fuzz_test: server_fuzzer client_fuzzer header_parser_fuzzer url_parser_fuzzer
@m=""; for f in fuzzing/corpus/[0-9]* fuzzing/corpus/issue1264 fuzzing/corpus/clusterfuzz-testcase-minimized-server_fuzzer-*; do if [ -f "$$f" ]; then m="$$m $$f"; fi; done; \
if [ -n "$$m" ]; then echo "./server_fuzzer$$m"; ./server_fuzzer $$m; else echo "(no server_fuzzer corpus)"; fi
@m=""; for f in fuzzing/corpus/clusterfuzz-testcase-minimized-client_fuzzer-*; do if [ -f "$$f" ]; then m="$$m $$f"; fi; done; \
if [ -n "$$m" ]; then echo "./client_fuzzer$$m"; ./client_fuzzer $$m; else echo "(no client_fuzzer corpus)"; fi
@m=""; for f in fuzzing/corpus/clusterfuzz-testcase-minimized-header_parser_fuzzer-*; do if [ -f "$$f" ]; then m="$$m $$f"; fi; done; \
if [ -n "$$m" ]; then echo "./header_parser_fuzzer$$m"; ./header_parser_fuzzer $$m; else echo "(no header_parser_fuzzer corpus)"; fi
@m=""; for f in fuzzing/corpus/clusterfuzz-testcase-minimized-url_parser_fuzzer-*; do if [ -f "$$f" ]; then m="$$m $$f"; fi; done; \
if [ -n "$$m" ]; then echo "./url_parser_fuzzer$$m"; ./url_parser_fuzzer $$m; else echo "(no url_parser_fuzzer corpus)"; fi
# Fuzz target, so that you can choose which $(LIB_FUZZING_ENGINE) to use.
server_fuzzer : fuzzing/server_fuzzer.cc ../httplib.h standalone_fuzz_target_runner.o
$(CXX) -o $@ -I.. $(CXXFLAGS) $< $(OPENSSL_SUPPORT) $(ZLIB_SUPPORT) $(BROTLI_SUPPORT) $(LIB_FUZZING_ENGINE) $(ZSTD_SUPPORT) $(LIBS)
@file $@
client_fuzzer : fuzzing/client_fuzzer.cc ../httplib.h standalone_fuzz_target_runner.o
$(CXX) -o $@ -I.. $(CXXFLAGS) $< $(OPENSSL_SUPPORT) $(ZLIB_SUPPORT) $(BROTLI_SUPPORT) $(LIB_FUZZING_ENGINE) $(ZSTD_SUPPORT) $(LIBS)
@file $@
header_parser_fuzzer : fuzzing/header_parser_fuzzer.cc ../httplib.h standalone_fuzz_target_runner.o
$(CXX) -o $@ -I.. $(CXXFLAGS) $< $(OPENSSL_SUPPORT) $(ZLIB_SUPPORT) $(BROTLI_SUPPORT) $(LIB_FUZZING_ENGINE) $(ZSTD_SUPPORT) $(LIBS)
@file $@
url_parser_fuzzer : fuzzing/url_parser_fuzzer.cc ../httplib.h standalone_fuzz_target_runner.o
$(CXX) -o $@ -I.. $(CXXFLAGS) $< $(OPENSSL_SUPPORT) $(ZLIB_SUPPORT) $(BROTLI_SUPPORT) $(LIB_FUZZING_ENGINE) $(ZSTD_SUPPORT) $(LIBS)
@file $@
# Standalone fuzz runner, which just reads inputs from fuzzing/corpus/ dir and
# feeds it to server_fuzzer.
standalone_fuzz_target_runner.o : fuzzing/standalone_fuzz_target_runner.cpp
@@ -273,5 +295,5 @@ cert.pem:
./gen-certs.sh
clean:
rm -rf test test_split test_mbedtls test_split_mbedtls test_wolfssl test_split_wolfssl test_no_tls, test_split_no_tls test_proxy test_proxy_mbedtls test_proxy_wolfssl test_benchmark server_fuzzer *.pem *.0 *.o *.1 *.srl httplib.h httplib.cc _build* *.dSYM *_shard_*.log cpp-httplib
rm -rf test test_split test_mbedtls test_split_mbedtls test_wolfssl test_split_wolfssl test_no_tls, test_split_no_tls test_proxy test_proxy_mbedtls test_proxy_wolfssl test_benchmark server_fuzzer client_fuzzer header_parser_fuzzer url_parser_fuzzer *.pem *.0 *.o *.1 *.srl httplib.h httplib.cc _build* *.dSYM *_shard_*.log cpp-httplib
@@ -0,0 +1,3 @@
HTTP/1.1 777
Content-Length:20000000000
@@ -0,0 +1,3 @@
HTTP/1.1 777
Content-Length:446744071854775
+75
View File
@@ -9288,6 +9288,81 @@ TEST(ClientVulnerabilityTest, PayloadMaxLengthZeroMeansNoLimit) {
<< " bytes without truncation, but only read " << total_read << " bytes.";
}
// Regression test for OSS-Fuzz issue 508342856: a malicious server sending an
// enormous Content-Length must not cause the client to pre-allocate a huge
// response body buffer. The reservation is capped at payload_max_length_, and
// the read itself fails when the body exceeds the limit.
TEST(ClientVulnerabilityTest, HugeContentLengthDoesNotPreallocate) {
#ifndef _WIN32
signal(SIGPIPE, SIG_IGN);
#endif
auto server_thread = std::thread([] {
auto srv = ::socket(AF_INET, SOCK_STREAM, 0);
default_socket_options(srv);
detail::set_socket_opt_time(srv, SOL_SOCKET, SO_RCVTIMEO, 5, 0);
detail::set_socket_opt_time(srv, SOL_SOCKET, SO_SNDTIMEO, 5, 0);
sockaddr_in addr{};
addr.sin_family = AF_INET;
addr.sin_port = htons(static_cast<uint16_t>(PORT + 2));
::inet_pton(AF_INET, "127.0.0.1", &addr.sin_addr);
int opt = 1;
::setsockopt(srv, SOL_SOCKET, SO_REUSEADDR,
#ifdef _WIN32
reinterpret_cast<const char *>(&opt),
#else
&opt,
#endif
sizeof(opt));
::bind(srv, reinterpret_cast<sockaddr *>(&addr), sizeof(addr));
::listen(srv, 1);
sockaddr_in cli_addr{};
socklen_t cli_len = sizeof(cli_addr);
auto cli = ::accept(srv, reinterpret_cast<sockaddr *>(&cli_addr), &cli_len);
if (cli != INVALID_SOCKET) {
char buf[4096];
::recv(cli, buf, sizeof(buf), 0);
// Malicious response: claim a 20GB body but send only a tiny payload.
std::string response = "HTTP/1.1 200 OK\r\n"
"Content-Length: 20000000000\r\n"
"\r\n"
"abc";
::send(cli,
#ifdef _WIN32
static_cast<const char *>(response.c_str()),
static_cast<int>(response.size()),
#else
response.c_str(), response.size(),
#endif
0);
detail::close_socket(cli);
}
detail::close_socket(srv);
});
std::this_thread::sleep_for(std::chrono::milliseconds(200));
{
Client cli("127.0.0.1", PORT + 2);
cli.set_read_timeout(5, 0);
// Default payload_max_length_ is 100MB; a 20GB Content-Length must not
// result in a 20GB pre-allocation. The Get() call is expected to fail
// (server claims more bytes than payload_max_length permits), but it must
// not exhaust memory before getting there.
auto res = cli.Get("/malicious");
EXPECT_FALSE(res); // Read fails because body exceeds payload_max_length_
}
server_thread.join();
}
// Verify that content_receiver bypasses the default payload_max_length,
// allowing streaming downloads larger than 100MB without requiring an explicit
// set_payload_max_length call.