Mirror of BoringSSL (grpc依赖) https://boringssl.googlesource.com/boringssl
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

2233 lines
81 KiB

/* Copyright (c) 2016, Google Inc.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
* SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
#include <limits.h>
#include <stdio.h>
#include <map>
#include <string>
#include <vector>
#include <gtest/gtest.h>
#include <openssl/asn1.h>
#include <openssl/asn1t.h>
#include <openssl/bytestring.h>
#include <openssl/err.h>
#include <openssl/mem.h>
#include <openssl/obj.h>
#include <openssl/span.h>
#include <openssl/x509v3.h>
#include "../test/test_util.h"
#include "internal.h"
#if defined(OPENSSL_THREADS)
#include <thread>
#endif
// kTag128 is an ASN.1 structure with a universal tag with number 128.
static const uint8_t kTag128[] = {
0x1f, 0x81, 0x00, 0x01, 0x00,
};
// kTag258 is an ASN.1 structure with a universal tag with number 258.
static const uint8_t kTag258[] = {
0x1f, 0x82, 0x02, 0x01, 0x00,
};
static_assert(V_ASN1_NEG_INTEGER == 258,
"V_ASN1_NEG_INTEGER changed. Update kTag258 to collide with it.");
// kTagOverflow is an ASN.1 structure with a universal tag with number 2^35-1,
// which will not fit in an int.
static const uint8_t kTagOverflow[] = {
0x1f, 0xff, 0xff, 0xff, 0xff, 0x7f, 0x01, 0x00,
};
TEST(ASN1Test, LargeTags) {
const uint8_t *p = kTag258;
bssl::UniquePtr<ASN1_TYPE> obj(d2i_ASN1_TYPE(NULL, &p, sizeof(kTag258)));
EXPECT_FALSE(obj) << "Parsed value with illegal tag" << obj->type;
ERR_clear_error();
p = kTagOverflow;
obj.reset(d2i_ASN1_TYPE(NULL, &p, sizeof(kTagOverflow)));
EXPECT_FALSE(obj) << "Parsed value with tag overflow" << obj->type;
ERR_clear_error();
p = kTag128;
obj.reset(d2i_ASN1_TYPE(NULL, &p, sizeof(kTag128)));
ASSERT_TRUE(obj);
EXPECT_EQ(128, obj->type);
const uint8_t kZero = 0;
EXPECT_EQ(Bytes(&kZero, 1), Bytes(obj->value.asn1_string->data,
obj->value.asn1_string->length));
}
// |obj| and |i2d_func| require different template parameters because C++ may
// deduce, say, |ASN1_STRING*| via |obj| and |const ASN1_STRING*| via
// |i2d_func|. Template argument deduction then fails. The language is not able
// to resolve this by observing that |const ASN1_STRING*| works for both.
template <typename T, typename U>
void TestSerialize(T obj, int (*i2d_func)(U a, uint8_t **pp),
bssl::Span<const uint8_t> expected) {
static_assert(std::is_convertible<T, U>::value,
"incompatible parameter to i2d_func");
Compute ASN.1 BIT STRING sizes more consistently. OpenSSL's BIT STRING representation has two modes, one where it implicitly trims trailing zeros and the other where the number of unused bits is explicitly set. This means logic in ASN1_item_verify, or elsewhere in callers, that checks flags and ASN1_STRING_length is inconsistent with i2c_ASN1_BIT_STRING. Add ASN1_BIT_STRING_num_bytes for code that needs to deal with X.509 using BIT STRING for some fields instead of OCTET STRING. Switch ASN1_item_verify to it. Some external code does this too, so export it as public API. This is mostly a theoretical issue. All parsed BIT STRINGS use explicit byte strings, and there are no APIs (apart from not-yet-opaquified structs) to specify the ASN1_STRING in X509, etc., structures. We intentionally made X509_set1_signature_value, etc., internally construct the ASN1_STRING. Still having an API is more consistent and helps nudge callers towards rejecting excess bits when they want bytes. It may also be worth a public API for consistently accessing the bit count. I've left it alone for now because I've not seen callers that need it, and it saves worrying about bytes-to-bits overflows. This also fixes a bug in the original version of the truncating logic when the entire string was all zeros, and const-corrects a few parameters. Change-Id: I9d29842a3d3264b0cde61ca8cfea07d02177dbc2 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48225 Commit-Queue: David Benjamin <davidben@google.com> Commit-Queue: Adam Langley <agl@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
// Test the allocating version first. It is easiest to debug.
uint8_t *ptr = nullptr;
int len = i2d_func(obj, &ptr);
ASSERT_GT(len, 0);
EXPECT_EQ(Bytes(expected), Bytes(ptr, len));
OPENSSL_free(ptr);
len = i2d_func(obj, nullptr);
ASSERT_GT(len, 0);
EXPECT_EQ(len, static_cast<int>(expected.size()));
Compute ASN.1 BIT STRING sizes more consistently. OpenSSL's BIT STRING representation has two modes, one where it implicitly trims trailing zeros and the other where the number of unused bits is explicitly set. This means logic in ASN1_item_verify, or elsewhere in callers, that checks flags and ASN1_STRING_length is inconsistent with i2c_ASN1_BIT_STRING. Add ASN1_BIT_STRING_num_bytes for code that needs to deal with X.509 using BIT STRING for some fields instead of OCTET STRING. Switch ASN1_item_verify to it. Some external code does this too, so export it as public API. This is mostly a theoretical issue. All parsed BIT STRINGS use explicit byte strings, and there are no APIs (apart from not-yet-opaquified structs) to specify the ASN1_STRING in X509, etc., structures. We intentionally made X509_set1_signature_value, etc., internally construct the ASN1_STRING. Still having an API is more consistent and helps nudge callers towards rejecting excess bits when they want bytes. It may also be worth a public API for consistently accessing the bit count. I've left it alone for now because I've not seen callers that need it, and it saves worrying about bytes-to-bits overflows. This also fixes a bug in the original version of the truncating logic when the entire string was all zeros, and const-corrects a few parameters. Change-Id: I9d29842a3d3264b0cde61ca8cfea07d02177dbc2 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48225 Commit-Queue: David Benjamin <davidben@google.com> Commit-Queue: Adam Langley <agl@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
std::vector<uint8_t> buf(len);
ptr = buf.data();
len = i2d_func(obj, &ptr);
ASSERT_EQ(len, static_cast<int>(expected.size()));
EXPECT_EQ(ptr, buf.data() + buf.size());
EXPECT_EQ(Bytes(expected), Bytes(buf));
}
static bssl::UniquePtr<BIGNUM> BIGNUMPow2(unsigned bit) {
bssl::UniquePtr<BIGNUM> bn(BN_new());
if (!bn ||
!BN_set_bit(bn.get(), bit)) {
return nullptr;
}
return bn;
}
TEST(ASN1Test, Integer) {
bssl::UniquePtr<BIGNUM> int64_min = BIGNUMPow2(63);
ASSERT_TRUE(int64_min);
BN_set_negative(int64_min.get(), 1);
bssl::UniquePtr<BIGNUM> int64_max = BIGNUMPow2(63);
ASSERT_TRUE(int64_max);
ASSERT_TRUE(BN_sub_word(int64_max.get(), 1));
bssl::UniquePtr<BIGNUM> int32_min = BIGNUMPow2(31);
ASSERT_TRUE(int32_min);
BN_set_negative(int32_min.get(), 1);
bssl::UniquePtr<BIGNUM> int32_max = BIGNUMPow2(31);
ASSERT_TRUE(int32_max);
ASSERT_TRUE(BN_sub_word(int32_max.get(), 1));
struct {
// der is the DER encoding of the INTEGER, including the tag and length.
std::vector<uint8_t> der;
// type and data are the corresponding fields of the |ASN1_STRING|
// representation.
int type;
std::vector<uint8_t> data;
// bn_asc is the |BIGNUM| representation, as parsed by the |BN_asc2bn|
// function.
const char *bn_asc;
} kTests[] = {
// -2^64 - 1
{{0x02, 0x09, 0xfe, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
V_ASN1_NEG_INTEGER,
{0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
"-0x10000000000000001"},
// -2^64
{{0x02, 0x09, 0xff, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
V_ASN1_NEG_INTEGER,
{0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
"-0x10000000000000000"},
// -2^64 + 1
{{0x02, 0x09, 0xff, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
V_ASN1_NEG_INTEGER,
{0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
"-0xffffffffffffffff"},
// -2^63 - 1
{{0x02, 0x09, 0xff, 0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
V_ASN1_NEG_INTEGER,
{0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
"-0x8000000000000001"},
// -2^63 (INT64_MIN)
{{0x02, 0x08, 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
V_ASN1_NEG_INTEGER,
{0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
"-0x8000000000000000"},
// -2^63 + 1
{{0x02, 0x08, 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
V_ASN1_NEG_INTEGER,
{0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
"-0x7fffffffffffffff"},
// -2^32 - 1
{{0x02, 0x05, 0xfe, 0xff, 0xff, 0xff, 0xff},
V_ASN1_NEG_INTEGER,
{0x01, 0x00, 0x00, 0x00, 0x01},
"-0x100000001"},
// -2^32
{{0x02, 0x05, 0xff, 0x00, 0x00, 0x00, 0x00},
V_ASN1_NEG_INTEGER,
{0x01, 0x00, 0x00, 0x00, 0x00},
"-0x100000000"},
// -2^32 + 1
{{0x02, 0x05, 0xff, 0x00, 0x00, 0x00, 0x01},
V_ASN1_NEG_INTEGER,
{0xff, 0xff, 0xff, 0xff},
"-0xffffffff"},
// -2^31 - 1
{{0x02, 0x05, 0xff, 0x7f, 0xff, 0xff, 0xff},
V_ASN1_NEG_INTEGER,
{0x80, 0x00, 0x00, 0x01},
"-0x80000001"},
// -2^31 (INT32_MIN)
{{0x02, 0x04, 0x80, 0x00, 0x00, 0x00},
V_ASN1_NEG_INTEGER,
{0x80, 0x00, 0x00, 0x00},
"-0x80000000"},
// -2^31 + 1
{{0x02, 0x04, 0x80, 0x00, 0x00, 0x01},
V_ASN1_NEG_INTEGER,
{0x7f, 0xff, 0xff, 0xff},
"-0x7fffffff"},
// -257
{{0x02, 0x02, 0xfe, 0xff}, V_ASN1_NEG_INTEGER, {0x01, 0x01}, "-257"},
// -256
{{0x02, 0x02, 0xff, 0x00}, V_ASN1_NEG_INTEGER, {0x01, 0x00}, "-256"},
// -255
{{0x02, 0x02, 0xff, 0x01}, V_ASN1_NEG_INTEGER, {0xff}, "-255"},
// -129
{{0x02, 0x02, 0xff, 0x7f}, V_ASN1_NEG_INTEGER, {0x81}, "-129"},
// -128
{{0x02, 0x01, 0x80}, V_ASN1_NEG_INTEGER, {0x80}, "-128"},
// -127
{{0x02, 0x01, 0x81}, V_ASN1_NEG_INTEGER, {0x7f}, "-127"},
// -1
{{0x02, 0x01, 0xff}, V_ASN1_NEG_INTEGER, {0x01}, "-1"},
// 0
Rewrite and tighten ASN1_INTEGER encoding and decoding. This fixes several issues around ASN1_INTEGER handling. First, invalid INTEGERs (not allowed in BER or DER) will no longer be accepted by d2i_ASN1_INTEGER. This aligns with upstream OpenSSL, which became strict in 6c5b6cb035666d46495ccbe4a4f3d5e3a659cd40, part of OpenSSL 1.1.0. In addition to matching the standard, this is needed to avoid round-tripping issues: ASN1_INTEGER uses a sign-and-magnitude representation, different from the DER two's complement representation. That means we cannot represent invalid DER INTEGERs. Attempting to do so messes up some invariants and causes values to not round-trip correctly when re-encoded. Thanks to Tavis Ormandy for catching this. Next, this CL tidies the story around invalid ASN1_INTEGERs (non-minimal and negative zero). Although we will never produce them in parsing, it is still possible to manually construct them with ASN1_STRING APIs. Historically (CVE-2016-2108), it was possible to get them out of the parser, due to a different bug, *and* i2d_ASN1_INTEGER had a memory error in doing so. That different bug has since been fixed, but we should still handle them correctly and test this. (To that end, this CL adds a test we ought to have added importing upstream's 3661bb4e7934668bd99ca777ea8b30eedfafa871 back in c4eec0c16b02c97a62a95b6a08656c3a9ddb6baa.) As the two's complement invariants are subtle as it is, I've opted to just fix the invalid values before encoding. However, invalid ASN1_INTEGERs still do not quite work right because ASN1_INTEGER_get, ASN1_INTEGER_cmp, and ASN1_STRING_cmp will all return surprising values with them. I've left those alone. Finally, that leads to the zero value. Almost every function believes the representation of 0 is a "\0" rather than "". However, a default-constructed INTEGER, like any other string type, is "". Those do not compare as equal. crypto/asn1 treats ASN1_INTEGER generically as ASN1_STRING enough that I think changing the other functions to match is cleaner than changing default-constructed ASN1_INTEGERs. Thus this CL removes all the special cases around zero. Update-Note: Invalid INTEGERs will no longer parse, but they already would not have parsed in OpenSSL. Additionally, zero is now internally represented as "" rather than "\0". Bug: 354 Change-Id: Id4d51a18f32afe90fd4df7455b21e0c8bdbc5389 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/51632 Reviewed-by: Adam Langley <agl@google.com> Commit-Queue: David Benjamin <davidben@google.com>
3 years ago
{{0x02, 0x01, 0x00}, V_ASN1_INTEGER, {}, "0"},
// 1
{{0x02, 0x01, 0x01}, V_ASN1_INTEGER, {0x01}, "1"},
// 127
{{0x02, 0x01, 0x7f}, V_ASN1_INTEGER, {0x7f}, "127"},
// 128
{{0x02, 0x02, 0x00, 0x80}, V_ASN1_INTEGER, {0x80}, "128"},
// 129
{{0x02, 0x02, 0x00, 0x81}, V_ASN1_INTEGER, {0x81}, "129"},
// 255
{{0x02, 0x02, 0x00, 0xff}, V_ASN1_INTEGER, {0xff}, "255"},
// 256
{{0x02, 0x02, 0x01, 0x00}, V_ASN1_INTEGER, {0x01, 0x00}, "256"},
// 257
{{0x02, 0x02, 0x01, 0x01}, V_ASN1_INTEGER, {0x01, 0x01}, "257"},
// 2^31 - 2
{{0x02, 0x04, 0x7f, 0xff, 0xff, 0xfe},
V_ASN1_INTEGER,
{0x7f, 0xff, 0xff, 0xfe},
"0x7ffffffe"},
// 2^31 - 1 (INT32_MAX)
{{0x02, 0x04, 0x7f, 0xff, 0xff, 0xff},
V_ASN1_INTEGER,
{0x7f, 0xff, 0xff, 0xff},
"0x7fffffff"},
// 2^31
{{0x02, 0x05, 0x00, 0x80, 0x00, 0x00, 0x00},
V_ASN1_INTEGER,
{0x80, 0x00, 0x00, 0x00},
"0x80000000"},
// 2^32 - 2
{{0x02, 0x05, 0x00, 0xff, 0xff, 0xff, 0xfe},
V_ASN1_INTEGER,
{0xff, 0xff, 0xff, 0xfe},
"0xfffffffe"},
// 2^32 - 1 (UINT32_MAX)
{{0x02, 0x05, 0x00, 0xff, 0xff, 0xff, 0xff},
V_ASN1_INTEGER,
{0xff, 0xff, 0xff, 0xff},
"0xffffffff"},
// 2^32
{{0x02, 0x05, 0x01, 0x00, 0x00, 0x00, 0x00},
V_ASN1_INTEGER,
{0x01, 0x00, 0x00, 0x00, 0x00},
"0x100000000"},
// 2^63 - 2
{{0x02, 0x08, 0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe},
V_ASN1_INTEGER,
{0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe},
"0x7ffffffffffffffe"},
// 2^63 - 1 (INT64_MAX)
{{0x02, 0x08, 0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
V_ASN1_INTEGER,
{0x7f, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
"0x7fffffffffffffff"},
// 2^63
{{0x02, 0x09, 0x00, 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
V_ASN1_INTEGER,
{0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
"0x8000000000000000"},
// 2^64 - 2
{{0x02, 0x09, 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe},
V_ASN1_INTEGER,
{0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe},
"0xfffffffffffffffe"},
// 2^64 - 1 (UINT64_MAX)
{{0x02, 0x09, 0x00, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
V_ASN1_INTEGER,
{0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff},
"0xffffffffffffffff"},
// 2^64
{{0x02, 0x09, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
V_ASN1_INTEGER,
{0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
"0x10000000000000000"},
// 2^64 + 1
{{0x02, 0x09, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
V_ASN1_INTEGER,
{0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
"0x10000000000000001"},
};
for (const auto &t : kTests) {
SCOPED_TRACE(t.bn_asc);
// Collect a map of different ways to construct the integer. The key is the
// method used and is only retained to aid debugging.
std::map<std::string, bssl::UniquePtr<ASN1_INTEGER>> objs;
// Construct |ASN1_INTEGER| by setting the type and data manually.
bssl::UniquePtr<ASN1_INTEGER> by_data(ASN1_STRING_type_new(t.type));
ASSERT_TRUE(by_data);
ASSERT_TRUE(ASN1_STRING_set(by_data.get(), t.data.data(), t.data.size()));
objs["data"] = std::move(by_data);
// Construct |ASN1_INTEGER| from a |BIGNUM|.
BIGNUM *bn_raw = nullptr;
ASSERT_TRUE(BN_asc2bn(&bn_raw, t.bn_asc));
bssl::UniquePtr<BIGNUM> bn(bn_raw);
bssl::UniquePtr<ASN1_INTEGER> by_bn(BN_to_ASN1_INTEGER(bn.get(), nullptr));
ASSERT_TRUE(by_bn);
objs["bn"] = std::move(by_bn);
// Construct |ASN1_INTEGER| from decoding.
const uint8_t *ptr = t.der.data();
bssl::UniquePtr<ASN1_INTEGER> by_der(
d2i_ASN1_INTEGER(nullptr, &ptr, t.der.size()));
ASSERT_TRUE(by_der);
EXPECT_EQ(ptr, t.der.data() + t.der.size());
objs["der"] = std::move(by_der);
// Construct |ASN1_INTEGER| from |long| or |uint64_t|, if it fits.
bool fits_in_long = false, fits_in_u64 = false;
uint64_t u64 = 0;
long l = 0;
uint64_t abs_u64;
if (BN_get_u64(bn.get(), &abs_u64)) {
fits_in_u64 = !BN_is_negative(bn.get());
if (fits_in_u64) {
u64 = abs_u64;
bssl::UniquePtr<ASN1_INTEGER> by_u64(ASN1_INTEGER_new());
ASSERT_TRUE(by_u64);
ASSERT_TRUE(ASN1_INTEGER_set_uint64(by_u64.get(), u64));
objs["u64"] = std::move(by_u64);
}
if (sizeof(long) == 8) {
fits_in_long = BN_cmp(int64_min.get(), bn.get()) <= 0 &&
BN_cmp(bn.get(), int64_max.get()) <= 0;
} else {
ASSERT_EQ(4u, sizeof(long));
fits_in_long = BN_cmp(int32_min.get(), bn.get()) <= 0 &&
BN_cmp(bn.get(), int32_max.get()) <= 0;
}
if (fits_in_long) {
if (BN_is_negative(bn.get())) {
l = static_cast<long>(0u - abs_u64);
} else {
l = static_cast<long>(abs_u64);
}
bssl::UniquePtr<ASN1_INTEGER> by_long(ASN1_INTEGER_new());
ASSERT_TRUE(by_long);
ASSERT_TRUE(ASN1_INTEGER_set(by_long.get(), l));
objs["long"] = std::move(by_long);
}
}
Rewrite and tighten ASN1_INTEGER encoding and decoding. This fixes several issues around ASN1_INTEGER handling. First, invalid INTEGERs (not allowed in BER or DER) will no longer be accepted by d2i_ASN1_INTEGER. This aligns with upstream OpenSSL, which became strict in 6c5b6cb035666d46495ccbe4a4f3d5e3a659cd40, part of OpenSSL 1.1.0. In addition to matching the standard, this is needed to avoid round-tripping issues: ASN1_INTEGER uses a sign-and-magnitude representation, different from the DER two's complement representation. That means we cannot represent invalid DER INTEGERs. Attempting to do so messes up some invariants and causes values to not round-trip correctly when re-encoded. Thanks to Tavis Ormandy for catching this. Next, this CL tidies the story around invalid ASN1_INTEGERs (non-minimal and negative zero). Although we will never produce them in parsing, it is still possible to manually construct them with ASN1_STRING APIs. Historically (CVE-2016-2108), it was possible to get them out of the parser, due to a different bug, *and* i2d_ASN1_INTEGER had a memory error in doing so. That different bug has since been fixed, but we should still handle them correctly and test this. (To that end, this CL adds a test we ought to have added importing upstream's 3661bb4e7934668bd99ca777ea8b30eedfafa871 back in c4eec0c16b02c97a62a95b6a08656c3a9ddb6baa.) As the two's complement invariants are subtle as it is, I've opted to just fix the invalid values before encoding. However, invalid ASN1_INTEGERs still do not quite work right because ASN1_INTEGER_get, ASN1_INTEGER_cmp, and ASN1_STRING_cmp will all return surprising values with them. I've left those alone. Finally, that leads to the zero value. Almost every function believes the representation of 0 is a "\0" rather than "". However, a default-constructed INTEGER, like any other string type, is "". Those do not compare as equal. crypto/asn1 treats ASN1_INTEGER generically as ASN1_STRING enough that I think changing the other functions to match is cleaner than changing default-constructed ASN1_INTEGERs. Thus this CL removes all the special cases around zero. Update-Note: Invalid INTEGERs will no longer parse, but they already would not have parsed in OpenSSL. Additionally, zero is now internally represented as "" rather than "\0". Bug: 354 Change-Id: Id4d51a18f32afe90fd4df7455b21e0c8bdbc5389 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/51632 Reviewed-by: Adam Langley <agl@google.com> Commit-Queue: David Benjamin <davidben@google.com>
3 years ago
// Default construction should return the zero |ASN1_INTEGER|.
if (BN_is_zero(bn.get())) {
bssl::UniquePtr<ASN1_INTEGER> by_default(ASN1_INTEGER_new());
ASSERT_TRUE(by_default);
objs["default"] = std::move(by_default);
}
// Test that every |ASN1_INTEGER| constructed behaves as expected.
for (const auto &pair : objs) {
// The fields should be as expected.
SCOPED_TRACE(pair.first);
const ASN1_INTEGER *obj = pair.second.get();
EXPECT_EQ(t.type, ASN1_STRING_type(obj));
EXPECT_EQ(Bytes(t.data), Bytes(ASN1_STRING_get0_data(obj),
ASN1_STRING_length(obj)));
// The object should encode correctly.
TestSerialize(obj, i2d_ASN1_INTEGER, t.der);
bssl::UniquePtr<BIGNUM> bn2(ASN1_INTEGER_to_BN(obj, nullptr));
ASSERT_TRUE(bn2);
EXPECT_EQ(0, BN_cmp(bn.get(), bn2.get()));
if (fits_in_u64) {
uint64_t v;
ASSERT_TRUE(ASN1_INTEGER_get_uint64(&v, obj));
EXPECT_EQ(v, u64);
} else {
uint64_t v;
EXPECT_FALSE(ASN1_INTEGER_get_uint64(&v, obj));
}
if (fits_in_long) {
EXPECT_EQ(l, ASN1_INTEGER_get(obj));
} else {
EXPECT_EQ(-1, ASN1_INTEGER_get(obj));
}
// All variations of integers should compare as equal to each other, as
// strings or integers. (Functions like |ASN1_TYPE_cmp| rely on
// string-based comparison.)
for (const auto &pair2 : objs) {
SCOPED_TRACE(pair2.first);
EXPECT_EQ(0, ASN1_INTEGER_cmp(obj, pair2.second.get()));
EXPECT_EQ(0, ASN1_STRING_cmp(obj, pair2.second.get()));
}
}
Rewrite and tighten ASN1_INTEGER encoding and decoding. This fixes several issues around ASN1_INTEGER handling. First, invalid INTEGERs (not allowed in BER or DER) will no longer be accepted by d2i_ASN1_INTEGER. This aligns with upstream OpenSSL, which became strict in 6c5b6cb035666d46495ccbe4a4f3d5e3a659cd40, part of OpenSSL 1.1.0. In addition to matching the standard, this is needed to avoid round-tripping issues: ASN1_INTEGER uses a sign-and-magnitude representation, different from the DER two's complement representation. That means we cannot represent invalid DER INTEGERs. Attempting to do so messes up some invariants and causes values to not round-trip correctly when re-encoded. Thanks to Tavis Ormandy for catching this. Next, this CL tidies the story around invalid ASN1_INTEGERs (non-minimal and negative zero). Although we will never produce them in parsing, it is still possible to manually construct them with ASN1_STRING APIs. Historically (CVE-2016-2108), it was possible to get them out of the parser, due to a different bug, *and* i2d_ASN1_INTEGER had a memory error in doing so. That different bug has since been fixed, but we should still handle them correctly and test this. (To that end, this CL adds a test we ought to have added importing upstream's 3661bb4e7934668bd99ca777ea8b30eedfafa871 back in c4eec0c16b02c97a62a95b6a08656c3a9ddb6baa.) As the two's complement invariants are subtle as it is, I've opted to just fix the invalid values before encoding. However, invalid ASN1_INTEGERs still do not quite work right because ASN1_INTEGER_get, ASN1_INTEGER_cmp, and ASN1_STRING_cmp will all return surprising values with them. I've left those alone. Finally, that leads to the zero value. Almost every function believes the representation of 0 is a "\0" rather than "". However, a default-constructed INTEGER, like any other string type, is "". Those do not compare as equal. crypto/asn1 treats ASN1_INTEGER generically as ASN1_STRING enough that I think changing the other functions to match is cleaner than changing default-constructed ASN1_INTEGERs. Thus this CL removes all the special cases around zero. Update-Note: Invalid INTEGERs will no longer parse, but they already would not have parsed in OpenSSL. Additionally, zero is now internally represented as "" rather than "\0". Bug: 354 Change-Id: Id4d51a18f32afe90fd4df7455b21e0c8bdbc5389 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/51632 Reviewed-by: Adam Langley <agl@google.com> Commit-Queue: David Benjamin <davidben@google.com>
3 years ago
// Although our parsers will never output non-minimal |ASN1_INTEGER|s, it is
// possible to construct them manually. They should encode correctly.
std::vector<uint8_t> data = t.data;
const int kMaxExtraBytes = 5;
for (int i = 0; i < kMaxExtraBytes; i++) {
data.insert(data.begin(), 0x00);
SCOPED_TRACE(Bytes(data));
bssl::UniquePtr<ASN1_INTEGER> non_minimal(ASN1_STRING_type_new(t.type));
ASSERT_TRUE(non_minimal);
ASSERT_TRUE(ASN1_STRING_set(non_minimal.get(), data.data(), data.size()));
TestSerialize(non_minimal.get(), i2d_ASN1_INTEGER, t.der);
}
}
for (size_t i = 0; i < OPENSSL_ARRAY_SIZE(kTests); i++) {
SCOPED_TRACE(Bytes(kTests[i].der));
const uint8_t *ptr = kTests[i].der.data();
bssl::UniquePtr<ASN1_INTEGER> a(
d2i_ASN1_INTEGER(nullptr, &ptr, kTests[i].der.size()));
ASSERT_TRUE(a);
for (size_t j = 0; j < OPENSSL_ARRAY_SIZE(kTests); j++) {
SCOPED_TRACE(Bytes(kTests[j].der));
ptr = kTests[j].der.data();
bssl::UniquePtr<ASN1_INTEGER> b(
d2i_ASN1_INTEGER(nullptr, &ptr, kTests[j].der.size()));
ASSERT_TRUE(b);
// |ASN1_INTEGER_cmp| should compare numerically. |ASN1_STRING_cmp| does
// not but should preserve equality.
if (i < j) {
EXPECT_LT(ASN1_INTEGER_cmp(a.get(), b.get()), 0);
EXPECT_NE(ASN1_STRING_cmp(a.get(), b.get()), 0);
} else if (i > j) {
EXPECT_GT(ASN1_INTEGER_cmp(a.get(), b.get()), 0);
EXPECT_NE(ASN1_STRING_cmp(a.get(), b.get()), 0);
} else {
EXPECT_EQ(ASN1_INTEGER_cmp(a.get(), b.get()), 0);
EXPECT_EQ(ASN1_STRING_cmp(a.get(), b.get()), 0);
}
}
}
Rewrite and tighten ASN1_INTEGER encoding and decoding. This fixes several issues around ASN1_INTEGER handling. First, invalid INTEGERs (not allowed in BER or DER) will no longer be accepted by d2i_ASN1_INTEGER. This aligns with upstream OpenSSL, which became strict in 6c5b6cb035666d46495ccbe4a4f3d5e3a659cd40, part of OpenSSL 1.1.0. In addition to matching the standard, this is needed to avoid round-tripping issues: ASN1_INTEGER uses a sign-and-magnitude representation, different from the DER two's complement representation. That means we cannot represent invalid DER INTEGERs. Attempting to do so messes up some invariants and causes values to not round-trip correctly when re-encoded. Thanks to Tavis Ormandy for catching this. Next, this CL tidies the story around invalid ASN1_INTEGERs (non-minimal and negative zero). Although we will never produce them in parsing, it is still possible to manually construct them with ASN1_STRING APIs. Historically (CVE-2016-2108), it was possible to get them out of the parser, due to a different bug, *and* i2d_ASN1_INTEGER had a memory error in doing so. That different bug has since been fixed, but we should still handle them correctly and test this. (To that end, this CL adds a test we ought to have added importing upstream's 3661bb4e7934668bd99ca777ea8b30eedfafa871 back in c4eec0c16b02c97a62a95b6a08656c3a9ddb6baa.) As the two's complement invariants are subtle as it is, I've opted to just fix the invalid values before encoding. However, invalid ASN1_INTEGERs still do not quite work right because ASN1_INTEGER_get, ASN1_INTEGER_cmp, and ASN1_STRING_cmp will all return surprising values with them. I've left those alone. Finally, that leads to the zero value. Almost every function believes the representation of 0 is a "\0" rather than "". However, a default-constructed INTEGER, like any other string type, is "". Those do not compare as equal. crypto/asn1 treats ASN1_INTEGER generically as ASN1_STRING enough that I think changing the other functions to match is cleaner than changing default-constructed ASN1_INTEGERs. Thus this CL removes all the special cases around zero. Update-Note: Invalid INTEGERs will no longer parse, but they already would not have parsed in OpenSSL. Additionally, zero is now internally represented as "" rather than "\0". Bug: 354 Change-Id: Id4d51a18f32afe90fd4df7455b21e0c8bdbc5389 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/51632 Reviewed-by: Adam Langley <agl@google.com> Commit-Queue: David Benjamin <davidben@google.com>
3 years ago
std::vector<uint8_t> kInvalidTests[] = {
// The empty string is not an integer.
{0x02, 0x00},
// Integers must be minimally-encoded.
{0x02, 0x02, 0x00, 0x00},
{0x02, 0x02, 0x00, 0x7f},
{0x02, 0x02, 0xff, 0xff},
{0x02, 0x02, 0xff, 0x80},
};
for (const auto &invalid : kInvalidTests) {
SCOPED_TRACE(Bytes(invalid));
const uint8_t *ptr = invalid.data();
bssl::UniquePtr<ASN1_INTEGER> integer(
d2i_ASN1_INTEGER(nullptr, &ptr, invalid.size()));
EXPECT_FALSE(integer);
}
// Callers expect |ASN1_INTEGER_get| and |ASN1_ENUMERATED_get| to return zero
// given NULL.
EXPECT_EQ(0, ASN1_INTEGER_get(nullptr));
EXPECT_EQ(0, ASN1_ENUMERATED_get(nullptr));
}
Rewrite and tighten ASN1_INTEGER encoding and decoding. This fixes several issues around ASN1_INTEGER handling. First, invalid INTEGERs (not allowed in BER or DER) will no longer be accepted by d2i_ASN1_INTEGER. This aligns with upstream OpenSSL, which became strict in 6c5b6cb035666d46495ccbe4a4f3d5e3a659cd40, part of OpenSSL 1.1.0. In addition to matching the standard, this is needed to avoid round-tripping issues: ASN1_INTEGER uses a sign-and-magnitude representation, different from the DER two's complement representation. That means we cannot represent invalid DER INTEGERs. Attempting to do so messes up some invariants and causes values to not round-trip correctly when re-encoded. Thanks to Tavis Ormandy for catching this. Next, this CL tidies the story around invalid ASN1_INTEGERs (non-minimal and negative zero). Although we will never produce them in parsing, it is still possible to manually construct them with ASN1_STRING APIs. Historically (CVE-2016-2108), it was possible to get them out of the parser, due to a different bug, *and* i2d_ASN1_INTEGER had a memory error in doing so. That different bug has since been fixed, but we should still handle them correctly and test this. (To that end, this CL adds a test we ought to have added importing upstream's 3661bb4e7934668bd99ca777ea8b30eedfafa871 back in c4eec0c16b02c97a62a95b6a08656c3a9ddb6baa.) As the two's complement invariants are subtle as it is, I've opted to just fix the invalid values before encoding. However, invalid ASN1_INTEGERs still do not quite work right because ASN1_INTEGER_get, ASN1_INTEGER_cmp, and ASN1_STRING_cmp will all return surprising values with them. I've left those alone. Finally, that leads to the zero value. Almost every function believes the representation of 0 is a "\0" rather than "". However, a default-constructed INTEGER, like any other string type, is "". Those do not compare as equal. crypto/asn1 treats ASN1_INTEGER generically as ASN1_STRING enough that I think changing the other functions to match is cleaner than changing default-constructed ASN1_INTEGERs. Thus this CL removes all the special cases around zero. Update-Note: Invalid INTEGERs will no longer parse, but they already would not have parsed in OpenSSL. Additionally, zero is now internally represented as "" rather than "\0". Bug: 354 Change-Id: Id4d51a18f32afe90fd4df7455b21e0c8bdbc5389 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/51632 Reviewed-by: Adam Langley <agl@google.com> Commit-Queue: David Benjamin <davidben@google.com>
3 years ago
// Although invalid, a negative zero should encode correctly.
TEST(ASN1Test, NegativeZero) {
bssl::UniquePtr<ASN1_INTEGER> neg_zero(
ASN1_STRING_type_new(V_ASN1_NEG_INTEGER));
ASSERT_TRUE(neg_zero);
EXPECT_EQ(0, ASN1_INTEGER_get(neg_zero.get()));
static const uint8_t kDER[] = {0x02, 0x01, 0x00};
TestSerialize(neg_zero.get(), i2d_ASN1_INTEGER, kDER);
}
TEST(ASN1Test, SerializeObject) {
static const uint8_t kDER[] = {0x06, 0x09, 0x2a, 0x86, 0x48, 0x86,
0xf7, 0x0d, 0x01, 0x01, 0x01};
const ASN1_OBJECT *obj = OBJ_nid2obj(NID_rsaEncryption);
TestSerialize(obj, i2d_ASN1_OBJECT, kDER);
}
TEST(ASN1Test, Boolean) {
static const uint8_t kTrue[] = {0x01, 0x01, 0xff};
TestSerialize(0xff, i2d_ASN1_BOOLEAN, kTrue);
// Other constants are also correctly encoded as TRUE.
TestSerialize(1, i2d_ASN1_BOOLEAN, kTrue);
TestSerialize(0x100, i2d_ASN1_BOOLEAN, kTrue);
const uint8_t *ptr = kTrue;
EXPECT_EQ(0xff, d2i_ASN1_BOOLEAN(nullptr, &ptr, sizeof(kTrue)));
EXPECT_EQ(ptr, kTrue + sizeof(kTrue));
static const uint8_t kFalse[] = {0x01, 0x01, 0x00};
TestSerialize(0x00, i2d_ASN1_BOOLEAN, kFalse);
ptr = kFalse;
EXPECT_EQ(0, d2i_ASN1_BOOLEAN(nullptr, &ptr, sizeof(kFalse)));
EXPECT_EQ(ptr, kFalse + sizeof(kFalse));
const std::vector<uint8_t> kInvalidBooleans[] = {
// No tag header.
{},
// No length.
{0x01},
// Truncated contents.
{0x01, 0x01},
// Contents too short or too long.
{0x01, 0x00},
{0x01, 0x02, 0x00, 0x00},
// Wrong tag number.
{0x02, 0x01, 0x00},
// Wrong tag class.
{0x81, 0x01, 0x00},
// Element is constructed.
{0x21, 0x01, 0x00},
// TODO(https://crbug.com/boringssl/354): Reject non-DER encodings of TRUE
// and test this.
};
for (const auto &invalid : kInvalidBooleans) {
SCOPED_TRACE(Bytes(invalid));
ptr = invalid.data();
EXPECT_EQ(-1, d2i_ASN1_BOOLEAN(nullptr, &ptr, invalid.size()));
ERR_clear_error();
}
}
// The templates go through a different codepath, so test them separately.
TEST(ASN1Test, SerializeEmbeddedBoolean) {
bssl::UniquePtr<BASIC_CONSTRAINTS> val(BASIC_CONSTRAINTS_new());
ASSERT_TRUE(val);
// BasicConstraints defaults to FALSE, so the encoding should be empty.
static const uint8_t kLeaf[] = {0x30, 0x00};
val->ca = 0;
TestSerialize(val.get(), i2d_BASIC_CONSTRAINTS, kLeaf);
// TRUE should always be encoded as 0xff, independent of what value the caller
// placed in the |ASN1_BOOLEAN|.
static const uint8_t kCA[] = {0x30, 0x03, 0x01, 0x01, 0xff};
val->ca = 0xff;
TestSerialize(val.get(), i2d_BASIC_CONSTRAINTS, kCA);
val->ca = 1;
TestSerialize(val.get(), i2d_BASIC_CONSTRAINTS, kCA);
val->ca = 0x100;
TestSerialize(val.get(), i2d_BASIC_CONSTRAINTS, kCA);
}
TEST(ASN1Test, ASN1Type) {
const struct {
int type;
std::vector<uint8_t> der;
} kTests[] = {
// BOOLEAN { TRUE }
{V_ASN1_BOOLEAN, {0x01, 0x01, 0xff}},
// BOOLEAN { FALSE }
{V_ASN1_BOOLEAN, {0x01, 0x01, 0x00}},
// OCTET_STRING { "a" }
{V_ASN1_OCTET_STRING, {0x04, 0x01, 0x61}},
Correctly handle invalid ASN1_OBJECTs when encoding. asn1_ex_i2c actually does have an error condition, it just wasn't being handled. 628b3c7f2fdf68519c27dc087c400ca616616f4e, imported from upstream's f3f8e72f494b36d05e0d04fe418f92b692fbb261, tried to check for OID-less ASN1_OBJECTs and return an error. But it and the upstream change didn't actually work. -1 in this function means to omit the object, so OpenSSL was silently misinterpreting the input structure. This changes the calling convention for asn1_ex_i2c to support this. It is, unfortunately, a little messy because: 1. One cannot check for object presense without walking the ASN1_ITEM/ASN1_TEMPLATE structures. You can *almost* check if *pval is NULL, but ASN1_BOOLEAN is an int with -1 to indicate an omitted optional. There are also FBOOLEAN/TBOOLEAN types that omit FALSE/TRUE for DEFAULT. Thus, without more invasive changes, asn1_ex_i2c must be able to report an omitted element. 2. While the i2d functions report an omitted element by successfully writing zero bytes, i2c only writes the contents. It thus must distinguish between an omitted element and an element with zero-length contents. 3. i2c_ASN1_INTEGER and i2c_ASN1_BIT_STRING return zero on error rather than -1. Those error paths are not actually reachable because they only check for NULL. In fact, OpenSSL has even unexported them. But I found a few callers. Rather than unwind all this and change the calling convention, I've just made it handle 0 and map to -1 for now. It's all a no-op anyway, and hopefully we can redo all this with CBB later. I've just added an output parameter for now. In writing tests, I also noticed that the hand-written i2d_ASN1_OBJECT and i2d_ASN1_BOOLEAN return the wrong value for errors, so I've fixed that. Update-Note: A default-constructed object with a required ASN1_OBJECT field can no longer be encoded without initializing the ASN1_OBJECT. Note this affects X509: the signature algorithm is an ASN1_OBJECT. Tests that try to serialize an X509_new() must fill in all required fields. (Production code is unlikely to be affected because the output was unparsable anyway, while tests sometimes wouldn't notice.) Bug: 429 Change-Id: I04417f5ad6b994cc5ccca540c8a7714b9b3af33d Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49348 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
// OCTET_STRING { }
{V_ASN1_OCTET_STRING, {0x04, 0x00}},
// BIT_STRING { `01` `00` }
{V_ASN1_BIT_STRING, {0x03, 0x02, 0x01, 0x00}},
// INTEGER { -1 }
{V_ASN1_INTEGER, {0x02, 0x01, 0xff}},
// OBJECT_IDENTIFIER { 1.2.840.113554.4.1.72585.2 }
{V_ASN1_OBJECT,
{0x06, 0x0c, 0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12, 0x04, 0x01, 0x84, 0xb7,
0x09, 0x02}},
// NULL {}
{V_ASN1_NULL, {0x05, 0x00}},
// SEQUENCE {}
{V_ASN1_SEQUENCE, {0x30, 0x00}},
// SET {}
{V_ASN1_SET, {0x31, 0x00}},
// [0] { UTF8String { "a" } }
{V_ASN1_OTHER, {0xa0, 0x03, 0x0c, 0x01, 0x61}},
};
for (const auto &t : kTests) {
SCOPED_TRACE(Bytes(t.der));
// The input should successfully parse.
const uint8_t *ptr = t.der.data();
bssl::UniquePtr<ASN1_TYPE> val(d2i_ASN1_TYPE(nullptr, &ptr, t.der.size()));
ASSERT_TRUE(val);
EXPECT_EQ(ASN1_TYPE_get(val.get()), t.type);
EXPECT_EQ(val->type, t.type);
TestSerialize(val.get(), i2d_ASN1_TYPE, t.der);
}
}
// Test that reading |value.ptr| from a FALSE |ASN1_TYPE| behaves correctly. The
// type historically supported this, so maintain the invariant in case external
// code relies on it.
TEST(ASN1Test, UnusedBooleanBits) {
// OCTET_STRING { "a" }
static const uint8_t kDER[] = {0x04, 0x01, 0x61};
const uint8_t *ptr = kDER;
bssl::UniquePtr<ASN1_TYPE> val(d2i_ASN1_TYPE(nullptr, &ptr, sizeof(kDER)));
ASSERT_TRUE(val);
EXPECT_EQ(V_ASN1_OCTET_STRING, val->type);
EXPECT_TRUE(val->value.ptr);
// Set |val| to a BOOLEAN containing FALSE.
ASN1_TYPE_set(val.get(), V_ASN1_BOOLEAN, NULL);
EXPECT_EQ(V_ASN1_BOOLEAN, val->type);
EXPECT_FALSE(val->value.ptr);
}
TEST(ASN1Test, ParseASN1Object) {
// 1.2.840.113554.4.1.72585.2, an arbitrary unknown OID.
static const uint8_t kOID[] = {0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12,
0x04, 0x01, 0x84, 0xb7, 0x09, 0x02};
ASN1_OBJECT *obj = ASN1_OBJECT_create(NID_undef, kOID, sizeof(kOID),
"short name", "long name");
ASSERT_TRUE(obj);
// OBJECT_IDENTIFIER { 1.3.101.112 }
static const uint8_t kDER[] = {0x06, 0x03, 0x2b, 0x65, 0x70};
const uint8_t *ptr = kDER;
// Parse an |ASN1_OBJECT| with object reuse.
EXPECT_TRUE(d2i_ASN1_OBJECT(&obj, &ptr, sizeof(kDER)));
EXPECT_EQ(NID_ED25519, OBJ_obj2nid(obj));
ASN1_OBJECT_free(obj);
// Repeat the test, this time overriding a static |ASN1_OBJECT|. It should
// detect this and construct a new one.
obj = OBJ_nid2obj(NID_rsaEncryption);
ptr = kDER;
EXPECT_TRUE(d2i_ASN1_OBJECT(&obj, &ptr, sizeof(kDER)));
EXPECT_EQ(NID_ED25519, OBJ_obj2nid(obj));
ASN1_OBJECT_free(obj);
const std::vector<uint8_t> kInvalidObjects[] = {
// No tag header.
{},
// No length.
{0x06},
// Truncated contents.
{0x06, 0x01},
// An OID may not be empty.
{0x06, 0x00},
// The last byte may not be a continuation byte (high bit set).
{0x06, 0x03, 0x2b, 0x65, 0xf0},
// Each component must be minimally-encoded.
{0x06, 0x03, 0x2b, 0x65, 0x80, 0x70},
{0x06, 0x03, 0x80, 0x2b, 0x65, 0x70},
// Wrong tag number.
{0x01, 0x03, 0x2b, 0x65, 0x70},
// Wrong tag class.
{0x86, 0x03, 0x2b, 0x65, 0x70},
// Element is constructed.
{0x26, 0x03, 0x2b, 0x65, 0x70},
};
for (const auto &invalid : kInvalidObjects) {
SCOPED_TRACE(Bytes(invalid));
ptr = invalid.data();
obj = d2i_ASN1_OBJECT(nullptr, &ptr, invalid.size());
EXPECT_FALSE(obj);
ASN1_OBJECT_free(obj);
ERR_clear_error();
}
}
Compute ASN.1 BIT STRING sizes more consistently. OpenSSL's BIT STRING representation has two modes, one where it implicitly trims trailing zeros and the other where the number of unused bits is explicitly set. This means logic in ASN1_item_verify, or elsewhere in callers, that checks flags and ASN1_STRING_length is inconsistent with i2c_ASN1_BIT_STRING. Add ASN1_BIT_STRING_num_bytes for code that needs to deal with X.509 using BIT STRING for some fields instead of OCTET STRING. Switch ASN1_item_verify to it. Some external code does this too, so export it as public API. This is mostly a theoretical issue. All parsed BIT STRINGS use explicit byte strings, and there are no APIs (apart from not-yet-opaquified structs) to specify the ASN1_STRING in X509, etc., structures. We intentionally made X509_set1_signature_value, etc., internally construct the ASN1_STRING. Still having an API is more consistent and helps nudge callers towards rejecting excess bits when they want bytes. It may also be worth a public API for consistently accessing the bit count. I've left it alone for now because I've not seen callers that need it, and it saves worrying about bytes-to-bits overflows. This also fixes a bug in the original version of the truncating logic when the entire string was all zeros, and const-corrects a few parameters. Change-Id: I9d29842a3d3264b0cde61ca8cfea07d02177dbc2 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48225 Commit-Queue: David Benjamin <davidben@google.com> Commit-Queue: Adam Langley <agl@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
TEST(ASN1Test, BitString) {
const size_t kNotWholeBytes = static_cast<size_t>(-1);
const struct {
std::vector<uint8_t> in;
size_t num_bytes;
} kValidInputs[] = {
// Empty bit string
{{0x03, 0x01, 0x00}, 0},
// 0b1
{{0x03, 0x02, 0x07, 0x80}, kNotWholeBytes},
// 0b1010
{{0x03, 0x02, 0x04, 0xa0}, kNotWholeBytes},
// 0b1010101
{{0x03, 0x02, 0x01, 0xaa}, kNotWholeBytes},
// 0b10101010
{{0x03, 0x02, 0x00, 0xaa}, 1},
// Bits 0 and 63 are set
{{0x03, 0x09, 0x00, 0x80, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01}, 8},
// 64 zero bits
{{0x03, 0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}, 8},
};
for (const auto &test : kValidInputs) {
SCOPED_TRACE(Bytes(test.in));
// The input should parse and round-trip correctly.
const uint8_t *ptr = test.in.data();
bssl::UniquePtr<ASN1_BIT_STRING> val(
d2i_ASN1_BIT_STRING(nullptr, &ptr, test.in.size()));
ASSERT_TRUE(val);
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, test.in);
// Check the byte count.
size_t num_bytes;
if (test.num_bytes == kNotWholeBytes) {
EXPECT_FALSE(ASN1_BIT_STRING_num_bytes(val.get(), &num_bytes));
} else {
ASSERT_TRUE(ASN1_BIT_STRING_num_bytes(val.get(), &num_bytes));
EXPECT_EQ(num_bytes, test.num_bytes);
}
}
const std::vector<uint8_t> kInvalidInputs[] = {
// Wrong tag
{0x04, 0x01, 0x00},
// Missing leading byte
{0x03, 0x00},
// Leading byte too high
{0x03, 0x02, 0x08, 0x00},
{0x03, 0x02, 0xff, 0x00},
// Empty bit strings must have a zero leading byte.
{0x03, 0x01, 0x01},
Compute ASN.1 BIT STRING sizes more consistently. OpenSSL's BIT STRING representation has two modes, one where it implicitly trims trailing zeros and the other where the number of unused bits is explicitly set. This means logic in ASN1_item_verify, or elsewhere in callers, that checks flags and ASN1_STRING_length is inconsistent with i2c_ASN1_BIT_STRING. Add ASN1_BIT_STRING_num_bytes for code that needs to deal with X.509 using BIT STRING for some fields instead of OCTET STRING. Switch ASN1_item_verify to it. Some external code does this too, so export it as public API. This is mostly a theoretical issue. All parsed BIT STRINGS use explicit byte strings, and there are no APIs (apart from not-yet-opaquified structs) to specify the ASN1_STRING in X509, etc., structures. We intentionally made X509_set1_signature_value, etc., internally construct the ASN1_STRING. Still having an API is more consistent and helps nudge callers towards rejecting excess bits when they want bytes. It may also be worth a public API for consistently accessing the bit count. I've left it alone for now because I've not seen callers that need it, and it saves worrying about bytes-to-bits overflows. This also fixes a bug in the original version of the truncating logic when the entire string was all zeros, and const-corrects a few parameters. Change-Id: I9d29842a3d3264b0cde61ca8cfea07d02177dbc2 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48225 Commit-Queue: David Benjamin <davidben@google.com> Commit-Queue: Adam Langley <agl@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
// Unused bits must all be zero.
{0x03, 0x02, 0x06, 0xc1 /* 0b11000001 */},
Compute ASN.1 BIT STRING sizes more consistently. OpenSSL's BIT STRING representation has two modes, one where it implicitly trims trailing zeros and the other where the number of unused bits is explicitly set. This means logic in ASN1_item_verify, or elsewhere in callers, that checks flags and ASN1_STRING_length is inconsistent with i2c_ASN1_BIT_STRING. Add ASN1_BIT_STRING_num_bytes for code that needs to deal with X.509 using BIT STRING for some fields instead of OCTET STRING. Switch ASN1_item_verify to it. Some external code does this too, so export it as public API. This is mostly a theoretical issue. All parsed BIT STRINGS use explicit byte strings, and there are no APIs (apart from not-yet-opaquified structs) to specify the ASN1_STRING in X509, etc., structures. We intentionally made X509_set1_signature_value, etc., internally construct the ASN1_STRING. Still having an API is more consistent and helps nudge callers towards rejecting excess bits when they want bytes. It may also be worth a public API for consistently accessing the bit count. I've left it alone for now because I've not seen callers that need it, and it saves worrying about bytes-to-bits overflows. This also fixes a bug in the original version of the truncating logic when the entire string was all zeros, and const-corrects a few parameters. Change-Id: I9d29842a3d3264b0cde61ca8cfea07d02177dbc2 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48225 Commit-Queue: David Benjamin <davidben@google.com> Commit-Queue: Adam Langley <agl@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
};
for (const auto &test : kInvalidInputs) {
SCOPED_TRACE(Bytes(test));
const uint8_t *ptr = test.data();
bssl::UniquePtr<ASN1_BIT_STRING> val(
d2i_ASN1_BIT_STRING(nullptr, &ptr, test.size()));
EXPECT_FALSE(val);
}
}
TEST(ASN1Test, SetBit) {
bssl::UniquePtr<ASN1_BIT_STRING> val(ASN1_BIT_STRING_new());
ASSERT_TRUE(val);
static const uint8_t kBitStringEmpty[] = {0x03, 0x01, 0x00};
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitStringEmpty);
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 100));
// Set a few bits via |ASN1_BIT_STRING_set_bit|.
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 0, 1));
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 1, 1));
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 2, 0));
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 3, 1));
static const uint8_t kBitString1101[] = {0x03, 0x02, 0x04, 0xd0};
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitString1101);
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 1));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 2));
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 3));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 4));
// Bits that were set may be cleared.
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 1, 0));
static const uint8_t kBitString1001[] = {0x03, 0x02, 0x04, 0x90};
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitString1001);
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 1));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 2));
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 3));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 4));
// Clearing trailing bits truncates the string.
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 3, 0));
static const uint8_t kBitString1[] = {0x03, 0x02, 0x07, 0x80};
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitString1);
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 1));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 2));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 3));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 4));
// Bits may be set beyond the end of the string.
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 63, 1));
static const uint8_t kBitStringLong[] = {0x03, 0x09, 0x00, 0x80, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x01};
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitStringLong);
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 62));
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 63));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 64));
// The string can be truncated back down again.
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 63, 0));
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitString1);
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 62));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 63));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 64));
// |ASN1_BIT_STRING_set_bit| also truncates when starting from a parsed
// string.
const uint8_t *ptr = kBitStringLong;
val.reset(d2i_ASN1_BIT_STRING(nullptr, &ptr, sizeof(kBitStringLong)));
ASSERT_TRUE(val);
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitStringLong);
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 63, 0));
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitString1);
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 62));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 63));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 64));
// A parsed bit string preserves trailing zero bits.
static const uint8_t kBitString10010[] = {0x03, 0x02, 0x03, 0x90};
ptr = kBitString10010;
val.reset(d2i_ASN1_BIT_STRING(nullptr, &ptr, sizeof(kBitString10010)));
ASSERT_TRUE(val);
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitString10010);
// But |ASN1_BIT_STRING_set_bit| will truncate it even if otherwise a no-op.
ASSERT_TRUE(ASN1_BIT_STRING_set_bit(val.get(), 0, 1));
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitString1001);
EXPECT_EQ(1, ASN1_BIT_STRING_get_bit(val.get(), 0));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 62));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 63));
EXPECT_EQ(0, ASN1_BIT_STRING_get_bit(val.get(), 64));
// By default, a BIT STRING implicitly truncates trailing zeros.
val.reset(ASN1_BIT_STRING_new());
ASSERT_TRUE(val);
static const uint8_t kZeros[64] = {0};
ASSERT_TRUE(ASN1_STRING_set(val.get(), kZeros, sizeof(kZeros)));
TestSerialize(val.get(), i2d_ASN1_BIT_STRING, kBitStringEmpty);
}
TEST(ASN1Test, StringToUTF8) {
static const struct {
std::vector<uint8_t> in;
int type;
const char *expected;
} kTests[] = {
// Non-minimal, two-byte UTF-8.
{{0xc0, 0x81}, V_ASN1_UTF8STRING, nullptr},
// Non-minimal, three-byte UTF-8.
{{0xe0, 0x80, 0x81}, V_ASN1_UTF8STRING, nullptr},
// Non-minimal, four-byte UTF-8.
{{0xf0, 0x80, 0x80, 0x81}, V_ASN1_UTF8STRING, nullptr},
// Truncated, four-byte UTF-8.
{{0xf0, 0x80, 0x80}, V_ASN1_UTF8STRING, nullptr},
// Low-surrogate value.
{{0xed, 0xa0, 0x80}, V_ASN1_UTF8STRING, nullptr},
// High-surrogate value.
{{0xed, 0xb0, 0x81}, V_ASN1_UTF8STRING, nullptr},
// Initial BOMs should be rejected from UCS-2 and UCS-4.
{{0xfe, 0xff, 0, 88}, V_ASN1_BMPSTRING, nullptr},
{{0, 0, 0xfe, 0xff, 0, 0, 0, 88}, V_ASN1_UNIVERSALSTRING, nullptr},
// Otherwise, BOMs should pass through.
{{0, 88, 0xfe, 0xff}, V_ASN1_BMPSTRING, "X\xef\xbb\xbf"},
{{0, 0, 0, 88, 0, 0, 0xfe, 0xff}, V_ASN1_UNIVERSALSTRING,
"X\xef\xbb\xbf"},
// The maximum code-point should pass though.
{{0, 16, 0xff, 0xfd}, V_ASN1_UNIVERSALSTRING, "\xf4\x8f\xbf\xbd"},
// Values outside the Unicode space should not.
{{0, 17, 0, 0}, V_ASN1_UNIVERSALSTRING, nullptr},
// Non-characters should be rejected.
{{0, 1, 0xff, 0xff}, V_ASN1_UNIVERSALSTRING, nullptr},
{{0, 1, 0xff, 0xfe}, V_ASN1_UNIVERSALSTRING, nullptr},
{{0, 0, 0xfd, 0xd5}, V_ASN1_UNIVERSALSTRING, nullptr},
// BMPString is UCS-2, not UTF-16, so surrogate pairs are invalid.
{{0xd8, 0, 0xdc, 1}, V_ASN1_BMPSTRING, nullptr},
};
for (const auto &test : kTests) {
SCOPED_TRACE(Bytes(test.in));
SCOPED_TRACE(test.type);
bssl::UniquePtr<ASN1_STRING> s(ASN1_STRING_type_new(test.type));
ASSERT_TRUE(s);
ASSERT_TRUE(ASN1_STRING_set(s.get(), test.in.data(), test.in.size()));
uint8_t *utf8;
const int utf8_len = ASN1_STRING_to_UTF8(&utf8, s.get());
EXPECT_EQ(utf8_len < 0, test.expected == nullptr);
if (utf8_len >= 0) {
if (test.expected != nullptr) {
EXPECT_EQ(Bytes(test.expected), Bytes(utf8, utf8_len));
}
OPENSSL_free(utf8);
} else {
ERR_clear_error();
}
}
}
static std::string ASN1StringToStdString(const ASN1_STRING *str) {
return std::string(ASN1_STRING_get0_data(str),
ASN1_STRING_get0_data(str) + ASN1_STRING_length(str));
}
TEST(ASN1Test, SetTime) {
static const struct {
time_t time;
const char *generalized;
const char *utc;
} kTests[] = {
{-631152001, "19491231235959Z", nullptr},
{-631152000, "19500101000000Z", "500101000000Z"},
{0, "19700101000000Z", "700101000000Z"},
{981173106, "20010203040506Z", "010203040506Z"},
#if defined(OPENSSL_64_BIT)
// TODO(https://crbug.com/boringssl/416): These cases overflow 32-bit
// |time_t| and do not consistently work on 32-bit platforms. For now,
// disable the tests on 32-bit. Re-enable them once the bug is fixed.
{2524607999, "20491231235959Z", "491231235959Z"},
{2524608000, "20500101000000Z", nullptr},
// Test boundary conditions.
{-62167219200, "00000101000000Z", nullptr},
{-62167219201, nullptr, nullptr},
{253402300799, "99991231235959Z", nullptr},
{253402300800, nullptr, nullptr},
#endif
};
for (const auto &t : kTests) {
SCOPED_TRACE(t.time);
#if defined(OPENSSL_WINDOWS)
// Windows |time_t| functions can only handle 1970 through 3000. See
// https://docs.microsoft.com/en-us/cpp/c-runtime-library/reference/gmtime-s-gmtime32-s-gmtime64-s?view=msvc-160
if (t.time < 0 || int64_t{t.time} > 32535215999) {
continue;
}
#endif
bssl::UniquePtr<ASN1_UTCTIME> utc(ASN1_UTCTIME_set(nullptr, t.time));
if (t.utc) {
ASSERT_TRUE(utc);
EXPECT_EQ(V_ASN1_UTCTIME, ASN1_STRING_type(utc.get()));
EXPECT_EQ(t.utc, ASN1StringToStdString(utc.get()));
} else {
EXPECT_FALSE(utc);
}
bssl::UniquePtr<ASN1_GENERALIZEDTIME> generalized(
ASN1_GENERALIZEDTIME_set(nullptr, t.time));
if (t.generalized) {
ASSERT_TRUE(generalized);
EXPECT_EQ(V_ASN1_GENERALIZEDTIME, ASN1_STRING_type(generalized.get()));
EXPECT_EQ(t.generalized, ASN1StringToStdString(generalized.get()));
} else {
EXPECT_FALSE(generalized);
}
bssl::UniquePtr<ASN1_TIME> choice(ASN1_TIME_set(nullptr, t.time));
if (t.generalized) {
ASSERT_TRUE(choice);
if (t.utc) {
EXPECT_EQ(V_ASN1_UTCTIME, ASN1_STRING_type(choice.get()));
EXPECT_EQ(t.utc, ASN1StringToStdString(choice.get()));
} else {
EXPECT_EQ(V_ASN1_GENERALIZEDTIME, ASN1_STRING_type(choice.get()));
EXPECT_EQ(t.generalized, ASN1StringToStdString(choice.get()));
}
} else {
EXPECT_FALSE(choice);
}
}
}
static std::vector<uint8_t> StringToVector(const std::string &str) {
return std::vector<uint8_t>(str.begin(), str.end());
}
TEST(ASN1Test, StringPrintEx) {
const struct {
int type;
std::vector<uint8_t> data;
int str_flags;
unsigned long flags;
std::string expected;
} kTests[] = {
// A string like "hello" is never escaped or quoted.
// |ASN1_STRFLGS_ESC_QUOTE| only introduces quotes when needed. Note
// OpenSSL interprets T61String as Latin-1.
{V_ASN1_T61STRING, StringToVector("hello"), 0, 0, "hello"},
{V_ASN1_T61STRING, StringToVector("hello"), 0,
ASN1_STRFLGS_ESC_2253 | ASN1_STRFLGS_ESC_CTRL | ASN1_STRFLGS_ESC_MSB,
"hello"},
{V_ASN1_T61STRING, StringToVector("hello"), 0,
ASN1_STRFLGS_ESC_2253 | ASN1_STRFLGS_ESC_CTRL | ASN1_STRFLGS_ESC_MSB |
ASN1_STRFLGS_ESC_QUOTE,
"hello"},
// By default, 8-bit characters are printed without escaping.
{V_ASN1_T61STRING,
{0, '\n', 0x80, 0xff, ',', '+', '"', '\\', '<', '>', ';'},
0,
0,
std::string(1, '\0') + "\n\x80\xff,+\"\\<>;"},
// Flags control different escapes. Note that any escape flag will cause
// blackslashes to be escaped.
{V_ASN1_T61STRING,
{0, '\n', 0x80, 0xff, ',', '+', '"', '\\', '<', '>', ';'},
0,
ASN1_STRFLGS_ESC_2253,
std::string(1, '\0') + "\n\x80\xff\\,\\+\\\"\\\\\\<\\>\\;"},
{V_ASN1_T61STRING,
{0, '\n', 0x80, 0xff, ',', '+', '"', '\\', '<', '>', ';'},
0,
ASN1_STRFLGS_ESC_CTRL,
"\\00\\0A\x80\xff,+\"\\\\<>;"},
{V_ASN1_T61STRING,
{0, '\n', 0x80, 0xff, ',', '+', '"', '\\', '<', '>', ';'},
0,
ASN1_STRFLGS_ESC_MSB,
std::string(1, '\0') + "\n\\80\\FF,+\"\\\\<>;"},
{V_ASN1_T61STRING,
{0, '\n', 0x80, 0xff, ',', '+', '"', '\\', '<', '>', ';'},
0,
ASN1_STRFLGS_ESC_2253 | ASN1_STRFLGS_ESC_CTRL | ASN1_STRFLGS_ESC_MSB,
"\\00\\0A\\80\\FF\\,\\+\\\"\\\\\\<\\>\\;"},
// When quoted, fewer characters need to be escaped in RFC 2253.
{V_ASN1_T61STRING,
{0, '\n', 0x80, 0xff, ',', '+', '"', '\\', '<', '>', ';'},
0,
ASN1_STRFLGS_ESC_2253 | ASN1_STRFLGS_ESC_CTRL | ASN1_STRFLGS_ESC_MSB |
ASN1_STRFLGS_ESC_QUOTE,
"\"\\00\\0A\\80\\FF,+\\\"\\\\<>;\""},
// If no characters benefit from quotes, no quotes are added.
{V_ASN1_T61STRING,
{0, '\n', 0x80, 0xff, '"', '\\'},
0,
ASN1_STRFLGS_ESC_2253 | ASN1_STRFLGS_ESC_CTRL | ASN1_STRFLGS_ESC_MSB |
ASN1_STRFLGS_ESC_QUOTE,
"\\00\\0A\\80\\FF\\\"\\\\"},
// RFC 2253 only escapes spaces at the start and end of a string.
{V_ASN1_T61STRING, StringToVector(" "), 0, ASN1_STRFLGS_ESC_2253,
"\\ \\ "},
{V_ASN1_T61STRING, StringToVector(" "), 0,
ASN1_STRFLGS_ESC_2253 | ASN1_STRFLGS_ESC_QUOTE, "\" \""},
// RFC 2253 only escapes # at the start of a string.
{V_ASN1_T61STRING, StringToVector("###"), 0, ASN1_STRFLGS_ESC_2253,
"\\###"},
{V_ASN1_T61STRING, StringToVector("###"), 0,
ASN1_STRFLGS_ESC_2253 | ASN1_STRFLGS_ESC_QUOTE, "\"###\""},
// By default, strings are decoded and Unicode code points are
// individually escaped.
{V_ASN1_UTF8STRING, StringToVector("a\xc2\x80\xc4\x80\xf0\x90\x80\x80"),
0, ASN1_STRFLGS_ESC_MSB, "a\\80\\U0100\\W00010000"},
{V_ASN1_BMPSTRING,
{0x00, 'a', 0x00, 0x80, 0x01, 0x00},
0,
ASN1_STRFLGS_ESC_MSB,
"a\\80\\U0100"},
{V_ASN1_UNIVERSALSTRING,
{0x00, 0x00, 0x00, 'a', //
0x00, 0x00, 0x00, 0x80, //
0x00, 0x00, 0x01, 0x00, //
0x00, 0x01, 0x00, 0x00},
0,
ASN1_STRFLGS_ESC_MSB,
"a\\80\\U0100\\W00010000"},
// |ASN1_STRFLGS_UTF8_CONVERT| normalizes everything to UTF-8 and then
// escapes individual bytes.
{V_ASN1_IA5STRING, StringToVector("a\x80"), 0,
ASN1_STRFLGS_ESC_MSB | ASN1_STRFLGS_UTF8_CONVERT, "a\\C2\\80"},
{V_ASN1_T61STRING, StringToVector("a\x80"), 0,
ASN1_STRFLGS_ESC_MSB | ASN1_STRFLGS_UTF8_CONVERT, "a\\C2\\80"},
{V_ASN1_UTF8STRING, StringToVector("a\xc2\x80\xc4\x80\xf0\x90\x80\x80"),
0, ASN1_STRFLGS_ESC_MSB | ASN1_STRFLGS_UTF8_CONVERT,
"a\\C2\\80\\C4\\80\\F0\\90\\80\\80"},
{V_ASN1_BMPSTRING,
{0x00, 'a', 0x00, 0x80, 0x01, 0x00},
0,
ASN1_STRFLGS_ESC_MSB | ASN1_STRFLGS_UTF8_CONVERT,
"a\\C2\\80\\C4\\80"},
{V_ASN1_UNIVERSALSTRING,
{0x00, 0x00, 0x00, 'a', //
0x00, 0x00, 0x00, 0x80, //
0x00, 0x00, 0x01, 0x00, //
0x00, 0x01, 0x00, 0x00},
0,
ASN1_STRFLGS_ESC_MSB | ASN1_STRFLGS_UTF8_CONVERT,
"a\\C2\\80\\C4\\80\\F0\\90\\80\\80"},
// The same as above, but without escaping the UTF-8 encoding.
{V_ASN1_IA5STRING, StringToVector("a\x80"), 0, ASN1_STRFLGS_UTF8_CONVERT,
"a\xc2\x80"},
{V_ASN1_T61STRING, StringToVector("a\x80"), 0, ASN1_STRFLGS_UTF8_CONVERT,
"a\xc2\x80"},
{V_ASN1_UTF8STRING, StringToVector("a\xc2\x80\xc4\x80\xf0\x90\x80\x80"),
0, ASN1_STRFLGS_UTF8_CONVERT, "a\xc2\x80\xc4\x80\xf0\x90\x80\x80"},
{V_ASN1_BMPSTRING,
{0x00, 'a', 0x00, 0x80, 0x01, 0x00},
0,
ASN1_STRFLGS_UTF8_CONVERT,
"a\xc2\x80\xc4\x80"},
{V_ASN1_UNIVERSALSTRING,
{0x00, 0x00, 0x00, 'a', //
0x00, 0x00, 0x00, 0x80, //
0x00, 0x00, 0x01, 0x00, //
0x00, 0x01, 0x00, 0x00},
0,
ASN1_STRFLGS_UTF8_CONVERT,
"a\xc2\x80\xc4\x80\xf0\x90\x80\x80"},
// Types that cannot be decoded are, by default, treated as a byte string.
{V_ASN1_OCTET_STRING, {0xff}, 0, 0, "\xff"},
{-1, {0xff}, 0, 0, "\xff"},
{100, {0xff}, 0, 0, "\xff"},
// |ASN1_STRFLGS_UTF8_CONVERT| still converts these bytes to UTF-8.
//
// TODO(davidben): This seems like a bug. Although it's unclear because
// the non-RFC-2253 options aren't especially sound. Can we just remove
// them?
{V_ASN1_OCTET_STRING, {0xff}, 0, ASN1_STRFLGS_UTF8_CONVERT, "\xc3\xbf"},
{-1, {0xff}, 0, ASN1_STRFLGS_UTF8_CONVERT, "\xc3\xbf"},
{100, {0xff}, 0, ASN1_STRFLGS_UTF8_CONVERT, "\xc3\xbf"},
// |ASN1_STRFLGS_IGNORE_TYPE| causes the string type to be ignored, so it
// is always treated as a byte string, even if it is not a valid encoding.
{V_ASN1_UTF8STRING, {0xff}, 0, ASN1_STRFLGS_IGNORE_TYPE, "\xff"},
{V_ASN1_BMPSTRING, {0xff}, 0, ASN1_STRFLGS_IGNORE_TYPE, "\xff"},
{V_ASN1_UNIVERSALSTRING, {0xff}, 0, ASN1_STRFLGS_IGNORE_TYPE, "\xff"},
// |ASN1_STRFLGS_SHOW_TYPE| prepends the type name.
{V_ASN1_UTF8STRING, {'a'}, 0, ASN1_STRFLGS_SHOW_TYPE, "UTF8STRING:a"},
{-1, {'a'}, 0, ASN1_STRFLGS_SHOW_TYPE, "(unknown):a"},
{100, {'a'}, 0, ASN1_STRFLGS_SHOW_TYPE, "(unknown):a"},
// |ASN1_STRFLGS_DUMP_ALL| and |ASN1_STRFLGS_DUMP_UNKNOWN| cause
// non-string types to be printed in hex, though without the DER wrapper
// by default.
{V_ASN1_UTF8STRING, StringToVector("\xe2\x98\x83"), 0,
ASN1_STRFLGS_DUMP_UNKNOWN, "\\U2603"},
{V_ASN1_UTF8STRING, StringToVector("\xe2\x98\x83"), 0,
ASN1_STRFLGS_DUMP_ALL, "#E29883"},
{V_ASN1_OCTET_STRING, StringToVector("\xe2\x98\x83"), 0,
ASN1_STRFLGS_DUMP_UNKNOWN, "#E29883"},
{V_ASN1_OCTET_STRING, StringToVector("\xe2\x98\x83"), 0,
ASN1_STRFLGS_DUMP_ALL, "#E29883"},
// |ASN1_STRFLGS_DUMP_DER| includes the entire element.
{V_ASN1_UTF8STRING, StringToVector("\xe2\x98\x83"), 0,
ASN1_STRFLGS_DUMP_ALL | ASN1_STRFLGS_DUMP_DER, "#0C03E29883"},
{V_ASN1_OCTET_STRING, StringToVector("\xe2\x98\x83"), 0,
ASN1_STRFLGS_DUMP_ALL | ASN1_STRFLGS_DUMP_DER, "#0403E29883"},
{V_ASN1_BIT_STRING,
{0x80},
ASN1_STRING_FLAG_BITS_LEFT | 4,
ASN1_STRFLGS_DUMP_ALL | ASN1_STRFLGS_DUMP_DER,
"#03020480"},
// INTEGER { 1 }
{V_ASN1_INTEGER,
{0x01},
0,
ASN1_STRFLGS_DUMP_ALL | ASN1_STRFLGS_DUMP_DER,
"#020101"},
// INTEGER { -1 }
{V_ASN1_NEG_INTEGER,
{0x01},
0,
ASN1_STRFLGS_DUMP_ALL | ASN1_STRFLGS_DUMP_DER,
"#0201FF"},
// ENUMERATED { 1 }
{V_ASN1_ENUMERATED,
{0x01},
0,
ASN1_STRFLGS_DUMP_ALL | ASN1_STRFLGS_DUMP_DER,
"#0A0101"},
// ENUMERATED { -1 }
{V_ASN1_NEG_ENUMERATED,
{0x01},
0,
ASN1_STRFLGS_DUMP_ALL | ASN1_STRFLGS_DUMP_DER,
"#0A01FF"},
};
for (const auto &t : kTests) {
SCOPED_TRACE(t.type);
SCOPED_TRACE(Bytes(t.data));
SCOPED_TRACE(t.str_flags);
SCOPED_TRACE(t.flags);
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_type_new(t.type));
ASSERT_TRUE(ASN1_STRING_set(str.get(), t.data.data(), t.data.size()));
str->flags = t.str_flags;
// If the |BIO| is null, it should measure the size.
int len = ASN1_STRING_print_ex(nullptr, str.get(), t.flags);
EXPECT_EQ(len, static_cast<int>(t.expected.size()));
// Measuring the size should also work for the |FILE| version
len = ASN1_STRING_print_ex_fp(nullptr, str.get(), t.flags);
EXPECT_EQ(len, static_cast<int>(t.expected.size()));
// Actually print the string.
bssl::UniquePtr<BIO> bio(BIO_new(BIO_s_mem()));
ASSERT_TRUE(bio);
len = ASN1_STRING_print_ex(bio.get(), str.get(), t.flags);
EXPECT_EQ(len, static_cast<int>(t.expected.size()));
const uint8_t *bio_contents;
size_t bio_len;
ASSERT_TRUE(BIO_mem_contents(bio.get(), &bio_contents, &bio_len));
EXPECT_EQ(t.expected, std::string(bio_contents, bio_contents + bio_len));
}
const struct {
int type;
std::vector<uint8_t> data;
int str_flags;
unsigned long flags;
} kUnprintableTests[] = {
// When decoding strings, invalid codepoints are errors.
{V_ASN1_UTF8STRING, {0xff}, 0, ASN1_STRFLGS_ESC_MSB},
{V_ASN1_BMPSTRING, {0xff}, 0, ASN1_STRFLGS_ESC_MSB},
{V_ASN1_BMPSTRING, {0xff}, 0, ASN1_STRFLGS_ESC_MSB},
{V_ASN1_UNIVERSALSTRING, {0xff}, 0, ASN1_STRFLGS_ESC_MSB},
};
for (const auto &t : kUnprintableTests) {
SCOPED_TRACE(t.type);
SCOPED_TRACE(Bytes(t.data));
SCOPED_TRACE(t.str_flags);
SCOPED_TRACE(t.flags);
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_type_new(t.type));
ASSERT_TRUE(ASN1_STRING_set(str.get(), t.data.data(), t.data.size()));
str->flags = t.str_flags;
// If the |BIO| is null, it should measure the size.
int len = ASN1_STRING_print_ex(nullptr, str.get(), t.flags);
EXPECT_EQ(len, -1);
ERR_clear_error();
// Measuring the size should also work for the |FILE| version
len = ASN1_STRING_print_ex_fp(nullptr, str.get(), t.flags);
EXPECT_EQ(len, -1);
ERR_clear_error();
// Actually print the string.
bssl::UniquePtr<BIO> bio(BIO_new(BIO_s_mem()));
ASSERT_TRUE(bio);
len = ASN1_STRING_print_ex(bio.get(), str.get(), t.flags);
EXPECT_EQ(len, -1);
ERR_clear_error();
}
}
TEST(ASN1Test, MBString) {
const unsigned long kAll = B_ASN1_PRINTABLESTRING | B_ASN1_IA5STRING |
B_ASN1_T61STRING | B_ASN1_BMPSTRING |
B_ASN1_UNIVERSALSTRING | B_ASN1_UTF8STRING;
const struct {
int format;
std::vector<uint8_t> in;
unsigned long mask;
int expected_type;
std::vector<uint8_t> expected_data;
int num_codepoints;
} kTests[] = {
// Given a choice of formats, we pick the smallest that fits.
{MBSTRING_UTF8, {}, kAll, V_ASN1_PRINTABLESTRING, {}, 0},
{MBSTRING_UTF8, {'a'}, kAll, V_ASN1_PRINTABLESTRING, {'a'}, 1},
{MBSTRING_UTF8,
{'a', 'A', '0', '\'', '(', ')', '+', ',', '-', '.', '/', ':', '=', '?'},
kAll,
V_ASN1_PRINTABLESTRING,
{'a', 'A', '0', '\'', '(', ')', '+', ',', '-', '.', '/', ':', '=', '?'},
14},
{MBSTRING_UTF8, {'*'}, kAll, V_ASN1_IA5STRING, {'*'}, 1},
{MBSTRING_UTF8, {'\n'}, kAll, V_ASN1_IA5STRING, {'\n'}, 1},
{MBSTRING_UTF8,
{0xc2, 0x80 /* U+0080 */},
kAll,
V_ASN1_T61STRING,
{0x80},
1},
{MBSTRING_UTF8,
{0xc4, 0x80 /* U+0100 */},
kAll,
V_ASN1_BMPSTRING,
{0x01, 0x00},
1},
{MBSTRING_UTF8,
{0xf0, 0x90, 0x80, 0x80 /* U+10000 */},
kAll,
V_ASN1_UNIVERSALSTRING,
{0x00, 0x01, 0x00, 0x00},
1},
{MBSTRING_UTF8,
{0xf0, 0x90, 0x80, 0x80 /* U+10000 */},
kAll & ~B_ASN1_UNIVERSALSTRING,
V_ASN1_UTF8STRING,
{0xf0, 0x90, 0x80, 0x80},
1},
// NUL is not printable. It should also not terminate iteration.
{MBSTRING_UTF8, {0}, kAll, V_ASN1_IA5STRING, {0}, 1},
{MBSTRING_UTF8, {0, 'a'}, kAll, V_ASN1_IA5STRING, {0, 'a'}, 2},
// When a particular format is specified, we use it.
{MBSTRING_UTF8,
{'a'},
B_ASN1_PRINTABLESTRING,
V_ASN1_PRINTABLESTRING,
{'a'},
1},
{MBSTRING_UTF8, {'a'}, B_ASN1_IA5STRING, V_ASN1_IA5STRING, {'a'}, 1},
{MBSTRING_UTF8, {'a'}, B_ASN1_T61STRING, V_ASN1_T61STRING, {'a'}, 1},
{MBSTRING_UTF8, {'a'}, B_ASN1_UTF8STRING, V_ASN1_UTF8STRING, {'a'}, 1},
{MBSTRING_UTF8,
{'a'},
B_ASN1_BMPSTRING,
V_ASN1_BMPSTRING,
{0x00, 'a'},
1},
{MBSTRING_UTF8,
{'a'},
B_ASN1_UNIVERSALSTRING,
V_ASN1_UNIVERSALSTRING,
{0x00, 0x00, 0x00, 'a'},
1},
// A long string with characters of many widths, to test sizes are
// measured in code points.
{MBSTRING_UTF8,
{
'a', //
0xc2, 0x80, // U+0080
0xc4, 0x80, // U+0100
0xf0, 0x90, 0x80, 0x80, // U+10000
},
B_ASN1_UNIVERSALSTRING,
V_ASN1_UNIVERSALSTRING,
{
0x00, 0x00, 0x00, 'a', //
0x00, 0x00, 0x00, 0x80, //
0x00, 0x00, 0x01, 0x00, //
0x00, 0x01, 0x00, 0x00, //
},
4},
};
for (const auto &t : kTests) {
SCOPED_TRACE(t.format);
SCOPED_TRACE(Bytes(t.in));
SCOPED_TRACE(t.mask);
// Passing in nullptr should do a dry run.
EXPECT_EQ(t.expected_type,
ASN1_mbstring_copy(nullptr, t.in.data(), t.in.size(), t.format,
t.mask));
// Test allocating a new object.
ASN1_STRING *str = nullptr;
EXPECT_EQ(
t.expected_type,
ASN1_mbstring_copy(&str, t.in.data(), t.in.size(), t.format, t.mask));
ASSERT_TRUE(str);
EXPECT_EQ(t.expected_type, ASN1_STRING_type(str));
EXPECT_EQ(Bytes(t.expected_data),
Bytes(ASN1_STRING_get0_data(str), ASN1_STRING_length(str)));
// Test writing into an existing object.
ASN1_STRING_free(str);
str = ASN1_STRING_new();
ASSERT_TRUE(str);
ASN1_STRING *old_str = str;
EXPECT_EQ(
t.expected_type,
ASN1_mbstring_copy(&str, t.in.data(), t.in.size(), t.format, t.mask));
ASSERT_EQ(old_str, str);
EXPECT_EQ(t.expected_type, ASN1_STRING_type(str));
EXPECT_EQ(Bytes(t.expected_data),
Bytes(ASN1_STRING_get0_data(str), ASN1_STRING_length(str)));
ASN1_STRING_free(str);
str = nullptr;
// minsize and maxsize should be enforced, even in a dry run.
EXPECT_EQ(t.expected_type,
ASN1_mbstring_ncopy(nullptr, t.in.data(), t.in.size(), t.format,
t.mask, /*minsize=*/t.num_codepoints,
/*maxsize=*/t.num_codepoints));
EXPECT_EQ(t.expected_type,
ASN1_mbstring_ncopy(&str, t.in.data(), t.in.size(), t.format,
t.mask, /*minsize=*/t.num_codepoints,
/*maxsize=*/t.num_codepoints));
ASSERT_TRUE(str);
EXPECT_EQ(t.expected_type, ASN1_STRING_type(str));
EXPECT_EQ(Bytes(t.expected_data),
Bytes(ASN1_STRING_get0_data(str), ASN1_STRING_length(str)));
ASN1_STRING_free(str);
str = nullptr;
EXPECT_EQ(-1, ASN1_mbstring_ncopy(
nullptr, t.in.data(), t.in.size(), t.format, t.mask,
/*minsize=*/t.num_codepoints + 1, /*maxsize=*/0));
ERR_clear_error();
EXPECT_EQ(-1, ASN1_mbstring_ncopy(
&str, t.in.data(), t.in.size(), t.format, t.mask,
/*minsize=*/t.num_codepoints + 1, /*maxsize=*/0));
EXPECT_FALSE(str);
ERR_clear_error();
if (t.num_codepoints > 1) {
EXPECT_EQ(-1, ASN1_mbstring_ncopy(
nullptr, t.in.data(), t.in.size(), t.format, t.mask,
/*minsize=*/0, /*maxsize=*/t.num_codepoints - 1));
ERR_clear_error();
EXPECT_EQ(-1, ASN1_mbstring_ncopy(
&str, t.in.data(), t.in.size(), t.format, t.mask,
/*minsize=*/0, /*maxsize=*/t.num_codepoints - 1));
EXPECT_FALSE(str);
ERR_clear_error();
}
}
const struct {
int format;
std::vector<uint8_t> in;
unsigned long mask;
} kInvalidTests[] = {
// Invalid encodings are rejected.
{MBSTRING_UTF8, {0xff}, B_ASN1_UTF8STRING},
{MBSTRING_BMP, {0xff}, B_ASN1_UTF8STRING},
{MBSTRING_UNIV, {0xff}, B_ASN1_UTF8STRING},
// Lone surrogates are not code points.
{MBSTRING_UTF8, {0xed, 0xa0, 0x80}, B_ASN1_UTF8STRING},
{MBSTRING_BMP, {0xd8, 0x00}, B_ASN1_UTF8STRING},
{MBSTRING_UNIV, {0x00, 0x00, 0xd8, 0x00}, B_ASN1_UTF8STRING},
// The input does not fit in the allowed output types.
{MBSTRING_UTF8, {'\n'}, B_ASN1_PRINTABLESTRING},
{MBSTRING_UTF8,
{0xc2, 0x80 /* U+0080 */},
B_ASN1_PRINTABLESTRING | B_ASN1_IA5STRING},
{MBSTRING_UTF8,
{0xc4, 0x80 /* U+0100 */},
B_ASN1_PRINTABLESTRING | B_ASN1_IA5STRING | B_ASN1_T61STRING},
{MBSTRING_UTF8,
{0xf0, 0x90, 0x80, 0x80 /* U+10000 */},
B_ASN1_PRINTABLESTRING | B_ASN1_IA5STRING | B_ASN1_T61STRING |
B_ASN1_BMPSTRING},
// Unrecognized bits are ignored.
{MBSTRING_UTF8, {'\n'}, B_ASN1_PRINTABLESTRING | B_ASN1_SEQUENCE},
};
for (const auto &t : kInvalidTests) {
SCOPED_TRACE(t.format);
SCOPED_TRACE(Bytes(t.in));
SCOPED_TRACE(t.mask);
EXPECT_EQ(-1, ASN1_mbstring_copy(nullptr, t.in.data(), t.in.size(),
t.format, t.mask));
ERR_clear_error();
ASN1_STRING *str = nullptr;
EXPECT_EQ(-1, ASN1_mbstring_copy(&str, t.in.data(), t.in.size(),
t.format, t.mask));
ERR_clear_error();
EXPECT_EQ(nullptr, str);
}
}
TEST(ASN1Test, StringByNID) {
// |ASN1_mbstring_*| tests above test most of the interactions with |inform|,
// so all tests below use UTF-8.
const struct {
int nid;
std::string in;
int expected_type;
std::string expected;
} kTests[] = {
// Although DirectoryString and PKCS9String allow many types of strings,
// we prefer UTF8String.
{NID_commonName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_commonName, "\xe2\x98\x83", V_ASN1_UTF8STRING, "\xe2\x98\x83"},
{NID_localityName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_stateOrProvinceName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_organizationName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_organizationalUnitName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_pkcs9_unstructuredName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_pkcs9_challengePassword, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_pkcs9_unstructuredAddress, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_givenName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_givenName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_givenName, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_surname, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_initials, "abc", V_ASN1_UTF8STRING, "abc"},
{NID_name, "abc", V_ASN1_UTF8STRING, "abc"},
// Some attribute types use a particular string type.
{NID_countryName, "US", V_ASN1_PRINTABLESTRING, "US"},
{NID_pkcs9_emailAddress, "example@example.com", V_ASN1_IA5STRING,
"example@example.com"},
{NID_serialNumber, "1234", V_ASN1_PRINTABLESTRING, "1234"},
{NID_friendlyName, "abc", V_ASN1_BMPSTRING,
std::string({'\0', 'a', '\0', 'b', '\0', 'c'})},
{NID_dnQualifier, "US", V_ASN1_PRINTABLESTRING, "US"},
{NID_domainComponent, "com", V_ASN1_IA5STRING, "com"},
{NID_ms_csp_name, "abc", V_ASN1_BMPSTRING,
std::string({'\0', 'a', '\0', 'b', '\0', 'c'})},
// Unknown NIDs default to UTF8String.
{NID_rsaEncryption, "abc", V_ASN1_UTF8STRING, "abc"},
};
for (const auto &t : kTests) {
SCOPED_TRACE(t.nid);
SCOPED_TRACE(t.in);
// Test allocating a new object.
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_set_by_NID(
nullptr, reinterpret_cast<const uint8_t *>(t.in.data()), t.in.size(),
MBSTRING_UTF8, t.nid));
ASSERT_TRUE(str);
EXPECT_EQ(t.expected_type, ASN1_STRING_type(str.get()));
EXPECT_EQ(Bytes(t.expected), Bytes(ASN1_STRING_get0_data(str.get()),
ASN1_STRING_length(str.get())));
// Test writing into an existing object.
str.reset(ASN1_STRING_new());
ASSERT_TRUE(str);
ASN1_STRING *old_str = str.get();
ASSERT_TRUE(ASN1_STRING_set_by_NID(
&old_str, reinterpret_cast<const uint8_t *>(t.in.data()), t.in.size(),
MBSTRING_UTF8, t.nid));
ASSERT_EQ(old_str, str.get());
EXPECT_EQ(t.expected_type, ASN1_STRING_type(str.get()));
EXPECT_EQ(Bytes(t.expected), Bytes(ASN1_STRING_get0_data(str.get()),
ASN1_STRING_length(str.get())));
}
const struct {
int nid;
std::string in;
} kInvalidTests[] = {
// DirectoryString forbids empty inputs.
{NID_commonName, ""},
{NID_localityName, ""},
{NID_stateOrProvinceName, ""},
{NID_organizationName, ""},
{NID_organizationalUnitName, ""},
{NID_pkcs9_unstructuredName, ""},
{NID_pkcs9_challengePassword, ""},
{NID_pkcs9_unstructuredAddress, ""},
{NID_givenName, ""},
{NID_givenName, ""},
{NID_givenName, ""},
{NID_surname, ""},
{NID_initials, ""},
{NID_name, ""},
// Test upper bounds from RFC 5280.
{NID_name, std::string(32769, 'a')},
{NID_commonName, std::string(65, 'a')},
{NID_localityName, std::string(129, 'a')},
{NID_stateOrProvinceName, std::string(129, 'a')},
{NID_organizationName, std::string(65, 'a')},
{NID_organizationalUnitName, std::string(65, 'a')},
{NID_pkcs9_emailAddress, std::string(256, 'a')},
{NID_serialNumber, std::string(65, 'a')},
// X520countryName must be exactly two characters.
{NID_countryName, "A"},
{NID_countryName, "AAA"},
// Some string types cannot represent all codepoints.
{NID_countryName, "\xe2\x98\x83"},
{NID_pkcs9_emailAddress, "\xe2\x98\x83"},
{NID_serialNumber, "\xe2\x98\x83"},
{NID_dnQualifier, "\xe2\x98\x83"},
{NID_domainComponent, "\xe2\x98\x83"},
};
for (const auto &t : kInvalidTests) {
SCOPED_TRACE(t.nid);
SCOPED_TRACE(t.in);
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_set_by_NID(
nullptr, reinterpret_cast<const uint8_t *>(t.in.data()), t.in.size(),
MBSTRING_UTF8, t.nid));
EXPECT_FALSE(str);
ERR_clear_error();
}
}
TEST(ASN1Test, StringByCustomNID) {
// This test affects library-global state. We rely on nothing else in the test
// suite using these OIDs.
int nid1 = OBJ_create("1.2.840.113554.4.1.72585.1000", "custom OID 1000",
"custom OID 1000");
ASSERT_NE(NID_undef, nid1);
int nid2 = OBJ_create("1.2.840.113554.4.1.72585.1001", "custom OID 1001",
"custom OID 1001");
ASSERT_NE(NID_undef, nid2);
// Values registered in the string table should be picked up.
ASSERT_TRUE(ASN1_STRING_TABLE_add(nid1, 5, 10, V_ASN1_PRINTABLESTRING,
STABLE_NO_MASK));
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_set_by_NID(
nullptr, reinterpret_cast<const uint8_t *>("12345"), 5, MBSTRING_UTF8,
nid1));
ASSERT_TRUE(str);
EXPECT_EQ(V_ASN1_PRINTABLESTRING, ASN1_STRING_type(str.get()));
EXPECT_EQ(Bytes("12345"), Bytes(ASN1_STRING_get0_data(str.get()),
ASN1_STRING_length(str.get())));
// Minimum and maximum lengths are enforced.
str.reset(ASN1_STRING_set_by_NID(
nullptr, reinterpret_cast<const uint8_t *>("1234"), 4, MBSTRING_UTF8,
nid1));
EXPECT_FALSE(str);
ERR_clear_error();
str.reset(ASN1_STRING_set_by_NID(
nullptr, reinterpret_cast<const uint8_t *>("12345678901"), 11,
MBSTRING_UTF8, nid1));
EXPECT_FALSE(str);
ERR_clear_error();
// Without |STABLE_NO_MASK|, we always pick UTF8String. -1 means there is no
// length limit.
ASSERT_TRUE(ASN1_STRING_TABLE_add(nid2, -1, -1, DIRSTRING_TYPE, 0));
str.reset(ASN1_STRING_set_by_NID(nullptr,
reinterpret_cast<const uint8_t *>("12345"),
5, MBSTRING_UTF8, nid2));
ASSERT_TRUE(str);
EXPECT_EQ(V_ASN1_UTF8STRING, ASN1_STRING_type(str.get()));
EXPECT_EQ(Bytes("12345"), Bytes(ASN1_STRING_get0_data(str.get()),
ASN1_STRING_length(str.get())));
// Overriding existing entries, built-in or custom, is an error.
EXPECT_FALSE(
ASN1_STRING_TABLE_add(NID_countryName, -1, -1, DIRSTRING_TYPE, 0));
EXPECT_FALSE(ASN1_STRING_TABLE_add(nid1, -1, -1, DIRSTRING_TYPE, 0));
}
#if defined(OPENSSL_THREADS)
TEST(ASN1Test, StringByCustomNIDThreads) {
// This test affects library-global state. We rely on nothing else in the test
// suite using these OIDs.
int nid1 = OBJ_create("1.2.840.113554.4.1.72585.1002", "custom OID 1002",
"custom OID 1002");
ASSERT_NE(NID_undef, nid1);
int nid2 = OBJ_create("1.2.840.113554.4.1.72585.1003", "custom OID 1003",
"custom OID 1003");
ASSERT_NE(NID_undef, nid2);
std::vector<std::thread> threads;
threads.emplace_back([&] {
ASSERT_TRUE(ASN1_STRING_TABLE_add(nid1, 5, 10, V_ASN1_PRINTABLESTRING,
STABLE_NO_MASK));
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_set_by_NID(
nullptr, reinterpret_cast<const uint8_t *>("12345"), 5, MBSTRING_UTF8,
nid1));
ASSERT_TRUE(str);
EXPECT_EQ(V_ASN1_PRINTABLESTRING, ASN1_STRING_type(str.get()));
EXPECT_EQ(Bytes("12345"), Bytes(ASN1_STRING_get0_data(str.get()),
ASN1_STRING_length(str.get())));
});
threads.emplace_back([&] {
ASSERT_TRUE(ASN1_STRING_TABLE_add(nid2, 5, 10, V_ASN1_PRINTABLESTRING,
STABLE_NO_MASK));
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_set_by_NID(
nullptr, reinterpret_cast<const uint8_t *>("12345"), 5, MBSTRING_UTF8,
nid2));
ASSERT_TRUE(str);
EXPECT_EQ(V_ASN1_PRINTABLESTRING, ASN1_STRING_type(str.get()));
EXPECT_EQ(Bytes("12345"), Bytes(ASN1_STRING_get0_data(str.get()),
ASN1_STRING_length(str.get())));
});
for (auto &thread : threads) {
thread.join();
}
}
#endif // OPENSSL_THREADS
Fix negative ENUMERATED values in multi-strings. I noticed this while I was reading through the encoder. OpenSSL's ASN.1 library is very sloppy when it comes to reusing enums. It has... - Universal tag numbers. These are just tag numbers from ASN.1 - utype. These are used in the ASN1_TYPE type field, as well as the ASN1_ITEM utype fields They are the same as universal tag numbers, except non-universal types map to V_ASN1_OTHER. I believe ASN1_TYPE types and ASN1_ITEM utypes are the same, but I am not positive. - ASN1_STRING types. These are the same as utypes, except V_ASN1_OTHER appears to only be possible when embedded inside ASN1_TYPE, and negative INTEGER and ENUMERATED values get mapped to V_ASN1_NEG_INTEGER and V_ASN1_NEG_ENUMERATED. Additionally, some values like V_ASN1_OBJECT are possible in a utype but not possible in an ASN1_STRING (and will cause lots of problems if ever placed in one). - Sometimes one of these enums is augmented with V_ASN1_UNDEF and/or V_ASN1_APP_CHOOSE for extra behaviors. - Probably others I'm missing. These get mixed up all the time. asn1_ex_i2c's MSTRING path converts from ASN1_STRING type to utype and forgets to normalize V_ASN1_NEG_*. This means that negative INTEGERs and ENUMERATEDs in MSTRINGs do not get encoded right. The negative INTEGER case is unreachable (unless the caller passes the wrong ASN1_STRING to an MSTRING i2d function, but mismatching i2d functions generally does wrong things), but the negative ENUMERATED case is reachable. Fix this and add a test. Change-Id: I762d482e72ebf03fd64bba291e751ab0b51af2a9 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/48805 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
// Test that multi-string types correctly encode negative ENUMERATED.
// Multi-string types cannot contain INTEGER, so we only test ENUMERATED.
TEST(ASN1Test, NegativeEnumeratedMultistring) {
static const uint8_t kMinusOne[] = {0x0a, 0x01, 0xff}; // ENUMERATED { -1 }
// |ASN1_PRINTABLE| is a multi-string type that allows ENUMERATED.
const uint8_t *p = kMinusOne;
bssl::UniquePtr<ASN1_STRING> str(
d2i_ASN1_PRINTABLE(nullptr, &p, sizeof(kMinusOne)));
ASSERT_TRUE(str);
TestSerialize(str.get(), i2d_ASN1_PRINTABLE, kMinusOne);
}
TEST(ASN1Test, PrintableType) {
const struct {
std::vector<uint8_t> in;
int result;
} kTests[] = {
{{}, V_ASN1_PRINTABLESTRING},
{{'a', 'A', '0', '\'', '(', ')', '+', ',', '-', '.', '/', ':', '=', '?'},
V_ASN1_PRINTABLESTRING},
{{'*'}, V_ASN1_IA5STRING},
{{'\0'}, V_ASN1_IA5STRING},
{{'\0', 'a'}, V_ASN1_IA5STRING},
{{0, 1, 2, 3, 125, 126, 127}, V_ASN1_IA5STRING},
{{0, 1, 2, 3, 125, 126, 127, 128}, V_ASN1_T61STRING},
{{128, 0, 1, 2, 3, 125, 126, 127}, V_ASN1_T61STRING},
};
for (const auto &t : kTests) {
SCOPED_TRACE(Bytes(t.in));
EXPECT_EQ(t.result, ASN1_PRINTABLE_type(t.in.data(), t.in.size()));
}
}
// Encoding a CHOICE type with an invalid selector should fail.
TEST(ASN1Test, InvalidChoice) {
bssl::UniquePtr<GENERAL_NAME> name(GENERAL_NAME_new());
ASSERT_TRUE(name);
// CHOICE types are initialized with an invalid selector.
EXPECT_EQ(-1, name->type);
// |name| should fail to encode.
EXPECT_EQ(-1, i2d_GENERAL_NAME(name.get(), nullptr));
// The error should be propagated through types containing |name|.
bssl::UniquePtr<GENERAL_NAMES> names(GENERAL_NAMES_new());
ASSERT_TRUE(names);
EXPECT_TRUE(bssl::PushToStack(names.get(), std::move(name)));
EXPECT_EQ(-1, i2d_GENERAL_NAMES(names.get(), nullptr));
}
Correctly handle invalid ASN1_OBJECTs when encoding. asn1_ex_i2c actually does have an error condition, it just wasn't being handled. 628b3c7f2fdf68519c27dc087c400ca616616f4e, imported from upstream's f3f8e72f494b36d05e0d04fe418f92b692fbb261, tried to check for OID-less ASN1_OBJECTs and return an error. But it and the upstream change didn't actually work. -1 in this function means to omit the object, so OpenSSL was silently misinterpreting the input structure. This changes the calling convention for asn1_ex_i2c to support this. It is, unfortunately, a little messy because: 1. One cannot check for object presense without walking the ASN1_ITEM/ASN1_TEMPLATE structures. You can *almost* check if *pval is NULL, but ASN1_BOOLEAN is an int with -1 to indicate an omitted optional. There are also FBOOLEAN/TBOOLEAN types that omit FALSE/TRUE for DEFAULT. Thus, without more invasive changes, asn1_ex_i2c must be able to report an omitted element. 2. While the i2d functions report an omitted element by successfully writing zero bytes, i2c only writes the contents. It thus must distinguish between an omitted element and an element with zero-length contents. 3. i2c_ASN1_INTEGER and i2c_ASN1_BIT_STRING return zero on error rather than -1. Those error paths are not actually reachable because they only check for NULL. In fact, OpenSSL has even unexported them. But I found a few callers. Rather than unwind all this and change the calling convention, I've just made it handle 0 and map to -1 for now. It's all a no-op anyway, and hopefully we can redo all this with CBB later. I've just added an output parameter for now. In writing tests, I also noticed that the hand-written i2d_ASN1_OBJECT and i2d_ASN1_BOOLEAN return the wrong value for errors, so I've fixed that. Update-Note: A default-constructed object with a required ASN1_OBJECT field can no longer be encoded without initializing the ASN1_OBJECT. Note this affects X509: the signature algorithm is an ASN1_OBJECT. Tests that try to serialize an X509_new() must fill in all required fields. (Production code is unlikely to be affected because the output was unparsable anyway, while tests sometimes wouldn't notice.) Bug: 429 Change-Id: I04417f5ad6b994cc5ccca540c8a7714b9b3af33d Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49348 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
// Encoding NID-only |ASN1_OBJECT|s should fail.
TEST(ASN1Test, InvalidObject) {
EXPECT_EQ(-1, i2d_ASN1_OBJECT(OBJ_nid2obj(NID_kx_ecdhe), nullptr));
bssl::UniquePtr<X509_ALGOR> alg(X509_ALGOR_new());
ASSERT_TRUE(alg);
ASSERT_TRUE(X509_ALGOR_set0(alg.get(), OBJ_nid2obj(NID_kx_ecdhe),
V_ASN1_UNDEF, nullptr));
EXPECT_EQ(-1, i2d_X509_ALGOR(alg.get(), nullptr));
}
// Encoding invalid |ASN1_TYPE|s should fail. |ASN1_TYPE|s are
// default-initialized to an invalid type.
TEST(ASN1Test, InvalidASN1Type) {
bssl::UniquePtr<ASN1_TYPE> obj(ASN1_TYPE_new());
ASSERT_TRUE(obj);
EXPECT_EQ(-1, obj->type);
EXPECT_EQ(-1, i2d_ASN1_TYPE(obj.get(), nullptr));
}
// Encoding invalid MSTRING types should fail. An MSTRING is a CHOICE of
// string-like types. They are initialized to an invalid type.
TEST(ASN1Test, InvalidMSTRING) {
bssl::UniquePtr<ASN1_STRING> obj(ASN1_TIME_new());
ASSERT_TRUE(obj);
EXPECT_EQ(-1, obj->type);
EXPECT_EQ(-1, i2d_ASN1_TIME(obj.get(), nullptr));
obj.reset(DIRECTORYSTRING_new());
ASSERT_TRUE(obj);
EXPECT_EQ(-1, obj->type);
EXPECT_EQ(-1, i2d_DIRECTORYSTRING(obj.get(), nullptr));
}
TEST(ASN1Test, StringTableSorted) {
const ASN1_STRING_TABLE *table;
size_t table_len;
asn1_get_string_table_for_testing(&table, &table_len);
for (size_t i = 1; i < table_len; i++) {
EXPECT_LT(table[i-1].nid, table[i].nid);
}
}
Make ASN1_NULL an opaque pointer. crypto/asn1 represents an ASN.1 NULL value as a non-null ASN1_NULL* pointer, (ASN1_NULL*)1. It is a non-null pointer because a null pointer represents an omitted OPTIONAL NULL. It is an opaque pointer because there is no sense in allocating anything. This pointer cannot be dereferenced, yet ASN1_NULL is a typedef for int. This is confusing and probably undefined behavior. (N1548, 6.3.2.3, clause 7 requires pointer conversions between two pointer types be correctly aligned, even if the pointer is never dereferenced. Strangely, clause 5 above does not impose the same requirement when converting from integer to pointer, though it mostly punts to the implementation definition.) Of course, all of tasn_*.c is a giant strict aliasing violation anyway, but an opaque struct pointer is a slightly better choice here. (Note that, although ASN1_BOOLEAN is also a typedef for int, that situation is different: the ASN1_BOOLEAN representation is a plain ASN1_BOOLEAN, not ASN1_BOOLEAN*, while the ASN1_NULL representation is a pointer. ASN1_NULL could have had the same treatment and even used a little less memory, but changing that would break the API.) Update-Note: Code that was assuming ASN1_NULL was an int typedef will fail to compile. Given this was never dereferencable, it is hard to imagine anything relying on this. Bug: 438 Change-Id: Ia0c652eed66e76f82a3843af1fc877f06c8d5e8f Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49805 Reviewed-by: Adam Langley <agl@google.com>
3 years ago
TEST(ASN1Test, Null) {
// An |ASN1_NULL| is an opaque, non-null pointer. It is an arbitrary signaling
// value and does not need to be freed. (If the pointer is null, this is an
// omitted OPTIONAL NULL.)
EXPECT_NE(nullptr, ASN1_NULL_new());
// It is safe to free either the non-null pointer or the null one.
ASN1_NULL_free(ASN1_NULL_new());
ASN1_NULL_free(nullptr);
// A NULL may be decoded.
static const uint8_t kNull[] = {0x05, 0x00};
const uint8_t *ptr = kNull;
EXPECT_NE(nullptr, d2i_ASN1_NULL(nullptr, &ptr, sizeof(kNull)));
EXPECT_EQ(ptr, kNull + sizeof(kNull));
// It may also be re-encoded.
uint8_t *enc = nullptr;
int enc_len = i2d_ASN1_NULL(ASN1_NULL_new(), &enc);
ASSERT_GE(enc_len, 0);
EXPECT_EQ(Bytes(kNull), Bytes(enc, enc_len));
OPENSSL_free(enc);
enc = nullptr;
// Although the standalone representation of NULL is a non-null pointer, the
// |ASN1_TYPE| representation is a null pointer.
ptr = kNull;
bssl::UniquePtr<ASN1_TYPE> null_type(
d2i_ASN1_TYPE(nullptr, &ptr, sizeof(kNull)));
ASSERT_TRUE(null_type);
EXPECT_EQ(ptr, kNull + sizeof(kNull));
EXPECT_EQ(V_ASN1_NULL, ASN1_TYPE_get(null_type.get()));
EXPECT_EQ(nullptr, null_type->value.ptr);
}
TEST(ASN1Test, Pack) {
bssl::UniquePtr<BASIC_CONSTRAINTS> val(BASIC_CONSTRAINTS_new());
ASSERT_TRUE(val);
val->ca = 0;
// Test all three calling conventions.
static const uint8_t kExpected[] = {0x30, 0x00};
bssl::UniquePtr<ASN1_STRING> str(
ASN1_item_pack(val.get(), ASN1_ITEM_rptr(BASIC_CONSTRAINTS), nullptr));
ASSERT_TRUE(str);
EXPECT_EQ(
Bytes(ASN1_STRING_get0_data(str.get()), ASN1_STRING_length(str.get())),
Bytes(kExpected));
ASN1_STRING *raw = nullptr;
str.reset(ASN1_item_pack(val.get(), ASN1_ITEM_rptr(BASIC_CONSTRAINTS), &raw));
ASSERT_TRUE(str);
EXPECT_EQ(raw, str.get());
EXPECT_EQ(
Bytes(ASN1_STRING_get0_data(str.get()), ASN1_STRING_length(str.get())),
Bytes(kExpected));
str.reset(ASN1_STRING_new());
ASSERT_TRUE(str);
raw = str.get();
EXPECT_TRUE(
ASN1_item_pack(val.get(), ASN1_ITEM_rptr(BASIC_CONSTRAINTS), &raw));
EXPECT_EQ(raw, str.get());
EXPECT_EQ(
Bytes(ASN1_STRING_get0_data(str.get()), ASN1_STRING_length(str.get())),
Bytes(kExpected));
}
TEST(ASN1Test, Unpack) {
bssl::UniquePtr<ASN1_STRING> str(ASN1_STRING_new());
ASSERT_TRUE(str);
static const uint8_t kValid[] = {0x30, 0x00};
ASSERT_TRUE(
ASN1_STRING_set(str.get(), kValid, sizeof(kValid)));
bssl::UniquePtr<BASIC_CONSTRAINTS> val(static_cast<BASIC_CONSTRAINTS *>(
ASN1_item_unpack(str.get(), ASN1_ITEM_rptr(BASIC_CONSTRAINTS))));
ASSERT_TRUE(val);
EXPECT_EQ(val->ca, 0);
EXPECT_EQ(val->pathlen, nullptr);
static const uint8_t kInvalid[] = {0x31, 0x00};
ASSERT_TRUE(ASN1_STRING_set(str.get(), kInvalid, sizeof(kInvalid)));
val.reset(static_cast<BASIC_CONSTRAINTS *>(
ASN1_item_unpack(str.get(), ASN1_ITEM_rptr(BASIC_CONSTRAINTS))));
EXPECT_FALSE(val);
static const uint8_t kTraiilingData[] = {0x30, 0x00, 0x00};
ASSERT_TRUE(
ASN1_STRING_set(str.get(), kTraiilingData, sizeof(kTraiilingData)));
val.reset(static_cast<BASIC_CONSTRAINTS *>(
ASN1_item_unpack(str.get(), ASN1_ITEM_rptr(BASIC_CONSTRAINTS))));
EXPECT_FALSE(val);
}
TEST(ASN1Test, StringCmp) {
struct Input {
int type;
std::vector<uint8_t> data;
int flags;
bool equals_previous;
};
// kInputs is a list of |ASN1_STRING| parameters, in sorted order. The input
// should be sorted by bit length, then data, then type.
const Input kInputs[] = {
{V_ASN1_BIT_STRING, {}, ASN1_STRING_FLAG_BITS_LEFT | 0, false},
{V_ASN1_BIT_STRING, {}, 0, true},
// When |ASN1_STRING_FLAG_BITS_LEFT| is unset, BIT STRINGs implicitly
// drop trailing zeros.
{V_ASN1_BIT_STRING, {0x00, 0x00, 0x00, 0x00}, 0, true},
{V_ASN1_OCTET_STRING, {}, 0, false},
{V_ASN1_UTF8STRING, {}, 0, false},
// BIT STRINGs with padding bits (i.e. not part of the actual value) are
// shorter and thus sort earlier:
// 1-bit inputs.
{V_ASN1_BIT_STRING, {0x00}, ASN1_STRING_FLAG_BITS_LEFT | 7, false},
{V_ASN1_BIT_STRING, {0x80}, ASN1_STRING_FLAG_BITS_LEFT | 7, false},
// 2-bit inputs.
{V_ASN1_BIT_STRING, {0x00}, ASN1_STRING_FLAG_BITS_LEFT | 6, false},
{V_ASN1_BIT_STRING, {0xc0}, ASN1_STRING_FLAG_BITS_LEFT | 6, false},
// 3-bit inputs.
{V_ASN1_BIT_STRING, {0x00}, ASN1_STRING_FLAG_BITS_LEFT | 5, false},
{V_ASN1_BIT_STRING, {0xe0}, ASN1_STRING_FLAG_BITS_LEFT | 5, false},
// 4-bit inputs.
{V_ASN1_BIT_STRING, {0xf0}, ASN1_STRING_FLAG_BITS_LEFT | 4, false},
{V_ASN1_BIT_STRING, {0xf0}, 0, true}, // 4 trailing zeros dropped.
{V_ASN1_BIT_STRING, {0xf0, 0x00}, 0, true}, // 12 trailing zeros dropped.
// 5-bit inputs.
{V_ASN1_BIT_STRING, {0x00}, ASN1_STRING_FLAG_BITS_LEFT | 3, false},
{V_ASN1_BIT_STRING, {0xf0}, ASN1_STRING_FLAG_BITS_LEFT | 3, false},
{V_ASN1_BIT_STRING, {0xf8}, ASN1_STRING_FLAG_BITS_LEFT | 3, false},
// 6-bit inputs.
{V_ASN1_BIT_STRING, {0x00}, ASN1_STRING_FLAG_BITS_LEFT | 2, false},
{V_ASN1_BIT_STRING, {0xf0}, ASN1_STRING_FLAG_BITS_LEFT | 2, false},
{V_ASN1_BIT_STRING, {0xfc}, ASN1_STRING_FLAG_BITS_LEFT | 2, false},
// 7-bit inputs.
{V_ASN1_BIT_STRING, {0x00}, ASN1_STRING_FLAG_BITS_LEFT | 1, false},
{V_ASN1_BIT_STRING, {0xf0}, ASN1_STRING_FLAG_BITS_LEFT | 1, false},
{V_ASN1_BIT_STRING, {0xfe}, ASN1_STRING_FLAG_BITS_LEFT | 1, false},
// 8-bit inputs.
{V_ASN1_BIT_STRING, {0x00}, ASN1_STRING_FLAG_BITS_LEFT | 0, false},
{V_ASN1_OCTET_STRING, {0x00}, 0, false},
{V_ASN1_UTF8STRING, {0x00}, 0, false},
{V_ASN1_BIT_STRING, {0x80}, ASN1_STRING_FLAG_BITS_LEFT | 0, false},
{V_ASN1_OCTET_STRING, {0x80}, 0, false},
{V_ASN1_UTF8STRING, {0x80}, 0, false},
{V_ASN1_BIT_STRING, {0xff}, ASN1_STRING_FLAG_BITS_LEFT | 0, false},
{V_ASN1_BIT_STRING, {0xff}, 0, true}, // No trailing zeros to drop.
{V_ASN1_OCTET_STRING, {0xff}, 0, false},
{V_ASN1_UTF8STRING, {0xff}, 0, false},
// Bytes are compared lexicographically.
{V_ASN1_BIT_STRING, {0x00, 0x00}, ASN1_STRING_FLAG_BITS_LEFT | 0, false},
{V_ASN1_OCTET_STRING, {0x00, 0x00}, 0, false},
{V_ASN1_UTF8STRING, {0x00, 0x00}, 0, false},
{V_ASN1_BIT_STRING, {0x00, 0xff}, ASN1_STRING_FLAG_BITS_LEFT | 0, false},
{V_ASN1_OCTET_STRING, {0x00, 0xff}, 0, false},
{V_ASN1_UTF8STRING, {0x00, 0xff}, 0, false},
{V_ASN1_BIT_STRING, {0xff, 0x00}, ASN1_STRING_FLAG_BITS_LEFT | 0, false},
{V_ASN1_OCTET_STRING, {0xff, 0x00}, 0, false},
{V_ASN1_UTF8STRING, {0xff, 0x00}, 0, false},
};
std::vector<bssl::UniquePtr<ASN1_STRING>> strs;
strs.reserve(OPENSSL_ARRAY_SIZE(kInputs));
for (const auto &input : kInputs) {
strs.emplace_back(ASN1_STRING_type_new(input.type));
ASSERT_TRUE(strs.back());
ASSERT_TRUE(ASN1_STRING_set(strs.back().get(), input.data.data(),
input.data.size()));
strs.back()->flags = input.flags;
}
for (size_t i = 0; i < strs.size(); i++) {
SCOPED_TRACE(i);
bool expect_equal = true;
for (size_t j = i; j < strs.size(); j++) {
SCOPED_TRACE(j);
if (j > i && !kInputs[j].equals_previous) {
expect_equal = false;
}
const int cmp_i_j = ASN1_STRING_cmp(strs[i].get(), strs[j].get());
const int cmp_j_i = ASN1_STRING_cmp(strs[j].get(), strs[i].get());
if (expect_equal) {
EXPECT_EQ(cmp_i_j, 0);
EXPECT_EQ(cmp_j_i, 0);
} else if (i < j) {
EXPECT_LT(cmp_i_j, 0);
EXPECT_GT(cmp_j_i, 0);
} else {
EXPECT_GT(cmp_i_j, 0);
EXPECT_LT(cmp_j_i, 0);
}
}
}
}
TEST(ASN1Test, PrintASN1Object) {
const struct {
std::vector<uint8_t> in;
const char *expected;
} kDataTests[] = {
// Known OIDs print as the name.
{{0x2a, 0x86, 0x48, 0x86, 0xf7, 0x0d, 0x01, 0x01, 0x01}, "rsaEncryption"},
// Unknown OIDs print in decimal.
{{0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12, 0x04, 0x01, 0x84, 0xb7, 0x09, 0x00},
"1.2.840.113554.4.1.72585.0"},
// Inputs which cannot be parsed as OIDs print as "<INVALID>".
{{0xff}, "<INVALID>"},
// The function has an internal 80-byte buffer. Test inputs at that
// boundary. First, 78 characters.
{{0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12, 0x04, 0x01, 0x84, 0xb7,
0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
"1.2.840.113554.4.1.72585.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0."
"0.0.0.1"},
// 79 characters.
{{0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12, 0x04, 0x01, 0x84, 0xb7,
0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x0a},
"1.2.840.113554.4.1.72585.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0."
"0.0.0.10"},
// 80 characters.
{{0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12, 0x04, 0x01, 0x84, 0xb7,
0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x64},
"1.2.840.113554.4.1.72585.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0."
"0.0.0.100"},
// 81 characters.
{{0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12, 0x04, 0x01, 0x84, 0xb7,
0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x87, 0x68},
"1.2.840.113554.4.1.72585.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0."
"0.0.0.1000"},
// 82 characters.
{{0x2a, 0x86, 0x48, 0x86, 0xf7, 0x12, 0x04, 0x01, 0x84, 0xb7,
0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xce, 0x10},
"1.2.840.113554.4.1.72585.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0."
"0.0.0.10000"},
};
for (const auto &t : kDataTests) {
SCOPED_TRACE(Bytes(t.in));
bssl::UniquePtr<ASN1_OBJECT> obj(ASN1_OBJECT_create(
NID_undef, t.in.data(), t.in.size(), /*sn=*/nullptr, /*ln=*/nullptr));
ASSERT_TRUE(obj);
bssl::UniquePtr<BIO> bio(BIO_new(BIO_s_mem()));
ASSERT_TRUE(bio);
int len = i2a_ASN1_OBJECT(bio.get(), obj.get());
EXPECT_EQ(len, static_cast<int>(strlen(t.expected)));
const uint8_t *bio_data;
size_t bio_len;
BIO_mem_contents(bio.get(), &bio_data, &bio_len);
EXPECT_EQ(t.expected,
std::string(reinterpret_cast<const char *>(bio_data), bio_len));
}
// Test writing NULL.
bssl::UniquePtr<BIO> bio(BIO_new(BIO_s_mem()));
ASSERT_TRUE(bio);
int len = i2a_ASN1_OBJECT(bio.get(), nullptr);
EXPECT_EQ(len, 4);
const uint8_t *bio_data;
size_t bio_len;
BIO_mem_contents(bio.get(), &bio_data, &bio_len);
EXPECT_EQ("NULL",
std::string(reinterpret_cast<const char *>(bio_data), bio_len));
}
TEST(ASN1Test, GetObject) {
// The header is valid, but there are not enough bytes for the length.
static const uint8_t kTruncated[] = {0x30, 0x01};
const uint8_t *ptr = kTruncated;
long length;
int tag;
int tag_class;
EXPECT_EQ(0x80, ASN1_get_object(&ptr, &length, &tag, &tag_class,
sizeof(kTruncated)));
static const uint8_t kIndefinite[] = {0x30, 0x80, 0x00, 0x00};
ptr = kIndefinite;
EXPECT_EQ(0x80, ASN1_get_object(&ptr, &length, &tag, &tag_class,
sizeof(kIndefinite)));
}
// The ASN.1 macros do not work on Windows shared library builds, where usage of
// |OPENSSL_EXPORT| is a bit stricter.
#if !defined(OPENSSL_WINDOWS) || !defined(BORINGSSL_SHARED_LIBRARY)
typedef struct asn1_linked_list_st {
struct asn1_linked_list_st *next;
} ASN1_LINKED_LIST;
DECLARE_ASN1_ITEM(ASN1_LINKED_LIST)
DECLARE_ASN1_FUNCTIONS(ASN1_LINKED_LIST)
ASN1_SEQUENCE(ASN1_LINKED_LIST) = {
Reject missing required fields in i2d functions. See also 006906cddda37e24a66443199444ef4476697477 from OpenSSL, though this CL uses a different strategy from upstream. Upstream makes ASN1_item_ex_i2d continue to allow optionals and checks afterwards at every non-optional call site. This CL pushes down an optional parameter and says functions cannot omit items unless explicitly allowed. I think this is a better default, though it is a larger change. Fields are only optional when they come from an ASN1_TEMPLATE with the OPTIONAL flag. Upstream's strategy misses top-level calls. This CL additionally adds checks for optional ASN1_TEMPLATEs in contexts where it doesn't make sense. Only fields of SEQUENCEs and SETs may be OPTIONAL, but the ASN1_ITEM/ASN1_TEMPLATE split doesn't quite match ASN.1 itself. ASN1_TEMPLATE is additionally responsible for explicit/implicit tagging, and SEQUENCE/SET OF. That means CHOICE arms and the occasional top-level type (ASN1_ITEM_TEMPLATE) use ASN1_TEMPLATE but will get confused if marked optional. As part of this, i2d_FOO(NULL) now returns -1 rather than "successfully" writing 0 bytes. If we want to allow NULL at the top-level, that's not too hard to arrange, but our CBB-based i2d functions do not. Update-Note: Structures with missing mandatory fields can no longer be encoded. Note that, apart from the cases already handled by preceding CLs, tasn_new.c will fill in non-NULL empty objects everywhere. The main downstream impact I've seen of this particular change is in combination with other bugs. Consider a caller that does: GENERAL_NAME *name = GENERAL_NAME_new(); name->type = GEN_DNS; name->d.dNSName = DoSomethingComplicated(...); Suppose DoSomethingComplicated() was actually fallible and returned NULL, but the caller forgot to check. They'd now construct a GENERAL_NAME with a missing field. Previously, this would silently serialize some garbage (omitted field) or empty string. Now we fail to encode, but the true error was the uncaught DoSomethingComplicated() failure. (Which likely was itself a bug.) Bug: 429 Change-Id: I37fe618761be64a619be9fdc8d416f24ecbb8c46 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49350 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
ASN1_OPT(ASN1_LINKED_LIST, next, ASN1_LINKED_LIST),
} ASN1_SEQUENCE_END(ASN1_LINKED_LIST)
IMPLEMENT_ASN1_FUNCTIONS(ASN1_LINKED_LIST)
static bool MakeLinkedList(bssl::UniquePtr<uint8_t> *out, size_t *out_len,
size_t count) {
bssl::ScopedCBB cbb;
std::vector<CBB> cbbs(count);
if (!CBB_init(cbb.get(), 2 * count) ||
!CBB_add_asn1(cbb.get(), &cbbs[0], CBS_ASN1_SEQUENCE)) {
return false;
}
for (size_t i = 1; i < count; i++) {
if (!CBB_add_asn1(&cbbs[i - 1], &cbbs[i], CBS_ASN1_SEQUENCE)) {
return false;
}
}
uint8_t *ptr;
if (!CBB_finish(cbb.get(), &ptr, out_len)) {
return false;
}
out->reset(ptr);
return true;
}
TEST(ASN1Test, Recursive) {
bssl::UniquePtr<uint8_t> data;
size_t len;
// Sanity-check that MakeLinkedList can be parsed.
ASSERT_TRUE(MakeLinkedList(&data, &len, 5));
const uint8_t *ptr = data.get();
ASN1_LINKED_LIST *list = d2i_ASN1_LINKED_LIST(nullptr, &ptr, len);
EXPECT_TRUE(list);
ASN1_LINKED_LIST_free(list);
// Excessively deep structures are rejected.
ASSERT_TRUE(MakeLinkedList(&data, &len, 100));
ptr = data.get();
list = d2i_ASN1_LINKED_LIST(nullptr, &ptr, len);
EXPECT_FALSE(list);
// Note checking the error queue here does not work. The error "stack trace"
// is too deep, so the |ASN1_R_NESTED_TOO_DEEP| entry drops off the queue.
ASN1_LINKED_LIST_free(list);
}
struct IMPLICIT_CHOICE {
ASN1_STRING *string;
};
DECLARE_ASN1_FUNCTIONS(IMPLICIT_CHOICE)
ASN1_SEQUENCE(IMPLICIT_CHOICE) = {
Reject missing required fields in i2d functions. See also 006906cddda37e24a66443199444ef4476697477 from OpenSSL, though this CL uses a different strategy from upstream. Upstream makes ASN1_item_ex_i2d continue to allow optionals and checks afterwards at every non-optional call site. This CL pushes down an optional parameter and says functions cannot omit items unless explicitly allowed. I think this is a better default, though it is a larger change. Fields are only optional when they come from an ASN1_TEMPLATE with the OPTIONAL flag. Upstream's strategy misses top-level calls. This CL additionally adds checks for optional ASN1_TEMPLATEs in contexts where it doesn't make sense. Only fields of SEQUENCEs and SETs may be OPTIONAL, but the ASN1_ITEM/ASN1_TEMPLATE split doesn't quite match ASN.1 itself. ASN1_TEMPLATE is additionally responsible for explicit/implicit tagging, and SEQUENCE/SET OF. That means CHOICE arms and the occasional top-level type (ASN1_ITEM_TEMPLATE) use ASN1_TEMPLATE but will get confused if marked optional. As part of this, i2d_FOO(NULL) now returns -1 rather than "successfully" writing 0 bytes. If we want to allow NULL at the top-level, that's not too hard to arrange, but our CBB-based i2d functions do not. Update-Note: Structures with missing mandatory fields can no longer be encoded. Note that, apart from the cases already handled by preceding CLs, tasn_new.c will fill in non-NULL empty objects everywhere. The main downstream impact I've seen of this particular change is in combination with other bugs. Consider a caller that does: GENERAL_NAME *name = GENERAL_NAME_new(); name->type = GEN_DNS; name->d.dNSName = DoSomethingComplicated(...); Suppose DoSomethingComplicated() was actually fallible and returned NULL, but the caller forgot to check. They'd now construct a GENERAL_NAME with a missing field. Previously, this would silently serialize some garbage (omitted field) or empty string. Now we fail to encode, but the true error was the uncaught DoSomethingComplicated() failure. (Which likely was itself a bug.) Bug: 429 Change-Id: I37fe618761be64a619be9fdc8d416f24ecbb8c46 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49350 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
ASN1_IMP(IMPLICIT_CHOICE, string, DIRECTORYSTRING, 0),
} ASN1_SEQUENCE_END(IMPLICIT_CHOICE)
IMPLEMENT_ASN1_FUNCTIONS(IMPLICIT_CHOICE)
// Test that the ASN.1 templates reject types with implicitly-tagged CHOICE
// types.
TEST(ASN1Test, ImplicitChoice) {
// Serializing a type with an implicitly tagged CHOICE should fail.
std::unique_ptr<IMPLICIT_CHOICE, decltype(&IMPLICIT_CHOICE_free)> obj(
IMPLICIT_CHOICE_new(), IMPLICIT_CHOICE_free);
EXPECT_EQ(-1, i2d_IMPLICIT_CHOICE(obj.get(), nullptr));
// An implicitly-tagged CHOICE is an error. Depending on the implementation,
// it may be misinterpreted as without the tag, or as clobbering the CHOICE
// tag. Test both inputs and ensure they fail.
// SEQUENCE { UTF8String {} }
static const uint8_t kInput1[] = {0x30, 0x02, 0x0c, 0x00};
const uint8_t *ptr = kInput1;
EXPECT_EQ(nullptr, d2i_IMPLICIT_CHOICE(nullptr, &ptr, sizeof(kInput1)));
// SEQUENCE { [0 PRIMITIVE] {} }
static const uint8_t kInput2[] = {0x30, 0x02, 0x80, 0x00};
ptr = kInput2;
EXPECT_EQ(nullptr, d2i_IMPLICIT_CHOICE(nullptr, &ptr, sizeof(kInput2)));
}
Reject missing required fields in i2d functions. See also 006906cddda37e24a66443199444ef4476697477 from OpenSSL, though this CL uses a different strategy from upstream. Upstream makes ASN1_item_ex_i2d continue to allow optionals and checks afterwards at every non-optional call site. This CL pushes down an optional parameter and says functions cannot omit items unless explicitly allowed. I think this is a better default, though it is a larger change. Fields are only optional when they come from an ASN1_TEMPLATE with the OPTIONAL flag. Upstream's strategy misses top-level calls. This CL additionally adds checks for optional ASN1_TEMPLATEs in contexts where it doesn't make sense. Only fields of SEQUENCEs and SETs may be OPTIONAL, but the ASN1_ITEM/ASN1_TEMPLATE split doesn't quite match ASN.1 itself. ASN1_TEMPLATE is additionally responsible for explicit/implicit tagging, and SEQUENCE/SET OF. That means CHOICE arms and the occasional top-level type (ASN1_ITEM_TEMPLATE) use ASN1_TEMPLATE but will get confused if marked optional. As part of this, i2d_FOO(NULL) now returns -1 rather than "successfully" writing 0 bytes. If we want to allow NULL at the top-level, that's not too hard to arrange, but our CBB-based i2d functions do not. Update-Note: Structures with missing mandatory fields can no longer be encoded. Note that, apart from the cases already handled by preceding CLs, tasn_new.c will fill in non-NULL empty objects everywhere. The main downstream impact I've seen of this particular change is in combination with other bugs. Consider a caller that does: GENERAL_NAME *name = GENERAL_NAME_new(); name->type = GEN_DNS; name->d.dNSName = DoSomethingComplicated(...); Suppose DoSomethingComplicated() was actually fallible and returned NULL, but the caller forgot to check. They'd now construct a GENERAL_NAME with a missing field. Previously, this would silently serialize some garbage (omitted field) or empty string. Now we fail to encode, but the true error was the uncaught DoSomethingComplicated() failure. (Which likely was itself a bug.) Bug: 429 Change-Id: I37fe618761be64a619be9fdc8d416f24ecbb8c46 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49350 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
struct REQUIRED_FIELD {
ASN1_INTEGER *value;
ASN1_INTEGER *value_imp;
ASN1_INTEGER *value_exp;
STACK_OF(ASN1_INTEGER) *seq;
STACK_OF(ASN1_INTEGER) *seq_imp;
STACK_OF(ASN1_INTEGER) *seq_exp;
Make ASN1_NULL an opaque pointer. crypto/asn1 represents an ASN.1 NULL value as a non-null ASN1_NULL* pointer, (ASN1_NULL*)1. It is a non-null pointer because a null pointer represents an omitted OPTIONAL NULL. It is an opaque pointer because there is no sense in allocating anything. This pointer cannot be dereferenced, yet ASN1_NULL is a typedef for int. This is confusing and probably undefined behavior. (N1548, 6.3.2.3, clause 7 requires pointer conversions between two pointer types be correctly aligned, even if the pointer is never dereferenced. Strangely, clause 5 above does not impose the same requirement when converting from integer to pointer, though it mostly punts to the implementation definition.) Of course, all of tasn_*.c is a giant strict aliasing violation anyway, but an opaque struct pointer is a slightly better choice here. (Note that, although ASN1_BOOLEAN is also a typedef for int, that situation is different: the ASN1_BOOLEAN representation is a plain ASN1_BOOLEAN, not ASN1_BOOLEAN*, while the ASN1_NULL representation is a pointer. ASN1_NULL could have had the same treatment and even used a little less memory, but changing that would break the API.) Update-Note: Code that was assuming ASN1_NULL was an int typedef will fail to compile. Given this was never dereferencable, it is hard to imagine anything relying on this. Bug: 438 Change-Id: Ia0c652eed66e76f82a3843af1fc877f06c8d5e8f Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49805 Reviewed-by: Adam Langley <agl@google.com>
3 years ago
ASN1_NULL *null;
ASN1_NULL *null_imp;
ASN1_NULL *null_exp;
Reject missing required fields in i2d functions. See also 006906cddda37e24a66443199444ef4476697477 from OpenSSL, though this CL uses a different strategy from upstream. Upstream makes ASN1_item_ex_i2d continue to allow optionals and checks afterwards at every non-optional call site. This CL pushes down an optional parameter and says functions cannot omit items unless explicitly allowed. I think this is a better default, though it is a larger change. Fields are only optional when they come from an ASN1_TEMPLATE with the OPTIONAL flag. Upstream's strategy misses top-level calls. This CL additionally adds checks for optional ASN1_TEMPLATEs in contexts where it doesn't make sense. Only fields of SEQUENCEs and SETs may be OPTIONAL, but the ASN1_ITEM/ASN1_TEMPLATE split doesn't quite match ASN.1 itself. ASN1_TEMPLATE is additionally responsible for explicit/implicit tagging, and SEQUENCE/SET OF. That means CHOICE arms and the occasional top-level type (ASN1_ITEM_TEMPLATE) use ASN1_TEMPLATE but will get confused if marked optional. As part of this, i2d_FOO(NULL) now returns -1 rather than "successfully" writing 0 bytes. If we want to allow NULL at the top-level, that's not too hard to arrange, but our CBB-based i2d functions do not. Update-Note: Structures with missing mandatory fields can no longer be encoded. Note that, apart from the cases already handled by preceding CLs, tasn_new.c will fill in non-NULL empty objects everywhere. The main downstream impact I've seen of this particular change is in combination with other bugs. Consider a caller that does: GENERAL_NAME *name = GENERAL_NAME_new(); name->type = GEN_DNS; name->d.dNSName = DoSomethingComplicated(...); Suppose DoSomethingComplicated() was actually fallible and returned NULL, but the caller forgot to check. They'd now construct a GENERAL_NAME with a missing field. Previously, this would silently serialize some garbage (omitted field) or empty string. Now we fail to encode, but the true error was the uncaught DoSomethingComplicated() failure. (Which likely was itself a bug.) Bug: 429 Change-Id: I37fe618761be64a619be9fdc8d416f24ecbb8c46 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49350 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
};
DECLARE_ASN1_FUNCTIONS(REQUIRED_FIELD)
ASN1_SEQUENCE(REQUIRED_FIELD) = {
ASN1_SIMPLE(REQUIRED_FIELD, value, ASN1_INTEGER),
ASN1_IMP(REQUIRED_FIELD, value_imp, ASN1_INTEGER, 0),
ASN1_EXP(REQUIRED_FIELD, value_exp, ASN1_INTEGER, 1),
ASN1_SEQUENCE_OF(REQUIRED_FIELD, seq, ASN1_INTEGER),
ASN1_IMP_SEQUENCE_OF(REQUIRED_FIELD, seq_imp, ASN1_INTEGER, 2),
ASN1_EXP_SEQUENCE_OF(REQUIRED_FIELD, seq_exp, ASN1_INTEGER, 3),
Make ASN1_NULL an opaque pointer. crypto/asn1 represents an ASN.1 NULL value as a non-null ASN1_NULL* pointer, (ASN1_NULL*)1. It is a non-null pointer because a null pointer represents an omitted OPTIONAL NULL. It is an opaque pointer because there is no sense in allocating anything. This pointer cannot be dereferenced, yet ASN1_NULL is a typedef for int. This is confusing and probably undefined behavior. (N1548, 6.3.2.3, clause 7 requires pointer conversions between two pointer types be correctly aligned, even if the pointer is never dereferenced. Strangely, clause 5 above does not impose the same requirement when converting from integer to pointer, though it mostly punts to the implementation definition.) Of course, all of tasn_*.c is a giant strict aliasing violation anyway, but an opaque struct pointer is a slightly better choice here. (Note that, although ASN1_BOOLEAN is also a typedef for int, that situation is different: the ASN1_BOOLEAN representation is a plain ASN1_BOOLEAN, not ASN1_BOOLEAN*, while the ASN1_NULL representation is a pointer. ASN1_NULL could have had the same treatment and even used a little less memory, but changing that would break the API.) Update-Note: Code that was assuming ASN1_NULL was an int typedef will fail to compile. Given this was never dereferencable, it is hard to imagine anything relying on this. Bug: 438 Change-Id: Ia0c652eed66e76f82a3843af1fc877f06c8d5e8f Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49805 Reviewed-by: Adam Langley <agl@google.com>
3 years ago
ASN1_SIMPLE(REQUIRED_FIELD, null, ASN1_NULL),
ASN1_IMP(REQUIRED_FIELD, null_imp, ASN1_NULL, 4),
ASN1_EXP(REQUIRED_FIELD, null_exp, ASN1_NULL, 5),
Reject missing required fields in i2d functions. See also 006906cddda37e24a66443199444ef4476697477 from OpenSSL, though this CL uses a different strategy from upstream. Upstream makes ASN1_item_ex_i2d continue to allow optionals and checks afterwards at every non-optional call site. This CL pushes down an optional parameter and says functions cannot omit items unless explicitly allowed. I think this is a better default, though it is a larger change. Fields are only optional when they come from an ASN1_TEMPLATE with the OPTIONAL flag. Upstream's strategy misses top-level calls. This CL additionally adds checks for optional ASN1_TEMPLATEs in contexts where it doesn't make sense. Only fields of SEQUENCEs and SETs may be OPTIONAL, but the ASN1_ITEM/ASN1_TEMPLATE split doesn't quite match ASN.1 itself. ASN1_TEMPLATE is additionally responsible for explicit/implicit tagging, and SEQUENCE/SET OF. That means CHOICE arms and the occasional top-level type (ASN1_ITEM_TEMPLATE) use ASN1_TEMPLATE but will get confused if marked optional. As part of this, i2d_FOO(NULL) now returns -1 rather than "successfully" writing 0 bytes. If we want to allow NULL at the top-level, that's not too hard to arrange, but our CBB-based i2d functions do not. Update-Note: Structures with missing mandatory fields can no longer be encoded. Note that, apart from the cases already handled by preceding CLs, tasn_new.c will fill in non-NULL empty objects everywhere. The main downstream impact I've seen of this particular change is in combination with other bugs. Consider a caller that does: GENERAL_NAME *name = GENERAL_NAME_new(); name->type = GEN_DNS; name->d.dNSName = DoSomethingComplicated(...); Suppose DoSomethingComplicated() was actually fallible and returned NULL, but the caller forgot to check. They'd now construct a GENERAL_NAME with a missing field. Previously, this would silently serialize some garbage (omitted field) or empty string. Now we fail to encode, but the true error was the uncaught DoSomethingComplicated() failure. (Which likely was itself a bug.) Bug: 429 Change-Id: I37fe618761be64a619be9fdc8d416f24ecbb8c46 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49350 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
} ASN1_SEQUENCE_END(REQUIRED_FIELD)
IMPLEMENT_ASN1_FUNCTIONS(REQUIRED_FIELD)
// Test that structures with missing required fields cannot be serialized. Test
// the full combination of tagging and SEQUENCE OF.
TEST(ASN1Test, MissingRequiredField) {
EXPECT_EQ(-1, i2d_REQUIRED_FIELD(nullptr, nullptr));
std::unique_ptr<REQUIRED_FIELD, decltype(&REQUIRED_FIELD_free)> obj(
nullptr, REQUIRED_FIELD_free);
for (auto field : {&REQUIRED_FIELD::value, &REQUIRED_FIELD::value_imp,
&REQUIRED_FIELD::value_exp}) {
obj.reset(REQUIRED_FIELD_new());
ASSERT_TRUE(obj);
ASN1_INTEGER_free((*obj).*field);
(*obj).*field = nullptr;
EXPECT_EQ(-1, i2d_REQUIRED_FIELD(obj.get(), nullptr));
}
for (auto field : {&REQUIRED_FIELD::seq, &REQUIRED_FIELD::seq_imp,
&REQUIRED_FIELD::seq_exp}) {
obj.reset(REQUIRED_FIELD_new());
ASSERT_TRUE(obj);
sk_ASN1_INTEGER_pop_free((*obj).*field, ASN1_INTEGER_free);
(*obj).*field = nullptr;
EXPECT_EQ(-1, i2d_REQUIRED_FIELD(obj.get(), nullptr));
}
Make ASN1_NULL an opaque pointer. crypto/asn1 represents an ASN.1 NULL value as a non-null ASN1_NULL* pointer, (ASN1_NULL*)1. It is a non-null pointer because a null pointer represents an omitted OPTIONAL NULL. It is an opaque pointer because there is no sense in allocating anything. This pointer cannot be dereferenced, yet ASN1_NULL is a typedef for int. This is confusing and probably undefined behavior. (N1548, 6.3.2.3, clause 7 requires pointer conversions between two pointer types be correctly aligned, even if the pointer is never dereferenced. Strangely, clause 5 above does not impose the same requirement when converting from integer to pointer, though it mostly punts to the implementation definition.) Of course, all of tasn_*.c is a giant strict aliasing violation anyway, but an opaque struct pointer is a slightly better choice here. (Note that, although ASN1_BOOLEAN is also a typedef for int, that situation is different: the ASN1_BOOLEAN representation is a plain ASN1_BOOLEAN, not ASN1_BOOLEAN*, while the ASN1_NULL representation is a pointer. ASN1_NULL could have had the same treatment and even used a little less memory, but changing that would break the API.) Update-Note: Code that was assuming ASN1_NULL was an int typedef will fail to compile. Given this was never dereferencable, it is hard to imagine anything relying on this. Bug: 438 Change-Id: Ia0c652eed66e76f82a3843af1fc877f06c8d5e8f Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49805 Reviewed-by: Adam Langley <agl@google.com>
3 years ago
for (auto field : {&REQUIRED_FIELD::null, &REQUIRED_FIELD::null_imp,
&REQUIRED_FIELD::null_exp}) {
obj.reset(REQUIRED_FIELD_new());
ASSERT_TRUE(obj);
(*obj).*field = nullptr;
EXPECT_EQ(-1, i2d_REQUIRED_FIELD(obj.get(), nullptr));
}
Reject missing required fields in i2d functions. See also 006906cddda37e24a66443199444ef4476697477 from OpenSSL, though this CL uses a different strategy from upstream. Upstream makes ASN1_item_ex_i2d continue to allow optionals and checks afterwards at every non-optional call site. This CL pushes down an optional parameter and says functions cannot omit items unless explicitly allowed. I think this is a better default, though it is a larger change. Fields are only optional when they come from an ASN1_TEMPLATE with the OPTIONAL flag. Upstream's strategy misses top-level calls. This CL additionally adds checks for optional ASN1_TEMPLATEs in contexts where it doesn't make sense. Only fields of SEQUENCEs and SETs may be OPTIONAL, but the ASN1_ITEM/ASN1_TEMPLATE split doesn't quite match ASN.1 itself. ASN1_TEMPLATE is additionally responsible for explicit/implicit tagging, and SEQUENCE/SET OF. That means CHOICE arms and the occasional top-level type (ASN1_ITEM_TEMPLATE) use ASN1_TEMPLATE but will get confused if marked optional. As part of this, i2d_FOO(NULL) now returns -1 rather than "successfully" writing 0 bytes. If we want to allow NULL at the top-level, that's not too hard to arrange, but our CBB-based i2d functions do not. Update-Note: Structures with missing mandatory fields can no longer be encoded. Note that, apart from the cases already handled by preceding CLs, tasn_new.c will fill in non-NULL empty objects everywhere. The main downstream impact I've seen of this particular change is in combination with other bugs. Consider a caller that does: GENERAL_NAME *name = GENERAL_NAME_new(); name->type = GEN_DNS; name->d.dNSName = DoSomethingComplicated(...); Suppose DoSomethingComplicated() was actually fallible and returned NULL, but the caller forgot to check. They'd now construct a GENERAL_NAME with a missing field. Previously, this would silently serialize some garbage (omitted field) or empty string. Now we fail to encode, but the true error was the uncaught DoSomethingComplicated() failure. (Which likely was itself a bug.) Bug: 429 Change-Id: I37fe618761be64a619be9fdc8d416f24ecbb8c46 Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/49350 Commit-Queue: David Benjamin <davidben@google.com> Reviewed-by: Adam Langley <agl@google.com>
4 years ago
}
#endif // !WINDOWS || !SHARED_LIBRARY