With PR 5170, I added logic that requires a EVP_CTRL_GCM_SET_IV_FIXED command be
issued before a EVP_CTRL_GCM_IV_GEN command. This matches OpenSSL's behavior.
However, OpenSSL also clears the flag enabling EVP_CTRL_GCM_IV_GEN after
EVP_CTRL_GCM_SET_IV_FIXED if EVP_CipherInit is called with a NULL key.
Otherwise, the flag retains its value. We didn't mirror this logic, and that
caused problems in OpenSSH unit testing. This commit aligns our logic with
OpenSSL's and adds a regression test to test_evp_cipher_aes_gcm for this case.
Discovered the AES-GCM flow using this command didn't work in our OpenSSH port.
This commit makes the behavior match OpenSSL and adds testing using known
OpenSSL-generated test vectors to prevent regressions. This was one of those
problems where two ends of a connection would work fine if they were both using
wolfSSL but not if one was using OpenSSL (i.e. OpenSSH interop with AES-GCM
was broken).
CBIORecv/Send are only assigned once.
IOCB_ReadCtx/WriteCtx only assigned once.
BuildMessage checks output and input wiht sizeOnly every call - state
machine doesn't cache these.
Renamed alias_tbl variables to something unique.
Other cppcheck fixes.
Also reworked pem_read_bio_key().
* Fix wolfSSL_RSA_public_decrypt() return value to match Openssl
* Add support for EVP_PKEY_verify_init() and EVP_PKEY_verify()
* wpa_supplicant SAE public key functionality requires this function.
* Add DSA support for EVP_PKEY_sign/verify()
* Add ECDSA support for EVP_PKEY_sign/verify()
* Add tests for EVP_PKEY_sign_verify()
* Fix "siglen = keySz" at error cases
* Fix wolfSSL_DSA_do_sign() usage
1. Check wolfSSL_BN_num_bytes() return value
2. Check siglen size
3. Double the siglen
* Check return code of wolfSSL_i2d_ECDSA_SIG() in wolfSSL_EVP_DigestSignFinal()
* Add size calculations to `wolfSSL_EVP_PKEY_sign`
* Add size checks to wolfSSL_EVP_PKEY_sign before writing out signature
* Use wc_ecc_sig_size() to calculate ECC signature size
Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
Co-authored-by: Juliusz Sosinowicz <juliusz@wolfssl.com>
- Add support for ASN1_TIME_compare, DH_new_by_nid, OBJ_length, OBJ_get0_data,
and ChaCha20-Poly1305 and HKDF in the EVP layer.
- Map EC_POINT_get_affine_coordinates to
wolfSSL_EC_POINT_get_affine_coordinates_GFp and EC_POINT_set_affine_coordinates
to wolfSSL_EC_POINT_set_affine_coordinates_GFp.
- Add kdf.h compatibility header.
- return `1` from `wolfSSL_BIO_set_mem_eof_return` instead of `0` for success
- bind requires ALPN
- `OPENSSL_COMPATIBLE_DEFAULT` defined for bind
- `WOLFSSL_ERROR_CODE_OPENSSL` defined when using compatibility layer
- return `bio->eof` on no pending data to read in memory BIO (defaults to `WOLFSSL_BIO_ERROR`)
- `flags` is no longer an input parameter in `wolfSSL_ERR_get_error_line_data`
- allow lazy parameter loading in `wolfSSL_DH_set0_key`
- implement reference counter in `WOLFSSL_EC_KEY`
- load serial number from `x509->serialNumber` if `x509->serial` is empty
Currently, it only supports ECC, which is all we need it for for the OpenSplice
port we're working on. In the ECC case, all it needs to do is set the group
appropriately. The logic is very similar to `wolfSSL_EVP_PKEY_keygen`, minus
the final step of actually generating the key.
Found with the configuration running the unit test through valgrind.
% ./configure CFLAGS=-DNO_WOLFSSL_CIPHER_SUITE_TEST \
--enable-all --disable-fastmath --enable-debug --disable-shared
1. ssl.c: In wolfSSL_DSA_generate_key(), we initialize (and allocate)
all the parameters in the key (p, q, g, x, y), and then we generate a
key, initializes (and allocates) x and y, again. mp_clear them
first.
2. evp.c: When printing public keys, the temporary mp_int wasn't getting
correctly freed.
3. evp.c: When printing public keys, modified the utility functions to
return once with a do-while-0 loop.
In `wolfSSL_EVP_CipherInit`, `ctx`'s `ivSz` field isn't being accounted for.
A common OpenSSL EVP AES-GCM flow looks like this:
- `EVP_CIPHER_CTX_new`
- `EVP_EncryptInit_ex`
- `EVP_CIPHER_CTX_ctrl` with command `EVP_CTRL_GCM_SET_IVLEN` to set the IV
length to 16 (AES block size) instead of the default 12
- `EVP_EncryptInit_ex` again to set the key and IV
- `EVP_EncryptUpdate` however many times
- `EVP_EncryptFinal`
In fact, we test this flow in our unit test `test_wolfssl_EVP_aes_gcm`. However,
in our implementation, the second call to `EVP_EncryptInit_ex` unconditionally
resets the IV length back to 12. This doesn't cause a test failure because
decryption has the same problem, so both sides of the equation have the same
wrong view of the IV.
The solution is to preserve the IV length in wolfSSL_EVP_CipherInit if ctx->ivSz
is non-zero. Otherwise, use the default of 12 (`GCM_NONCE_MID_SZ`).
This was discovered by a user migrating to the compatibility layer. As I
mentioned, it isn't exposed by our testing. It is exposed if you try to use the
same key and IV with OpenSSL and compare the resulting ciphertext with wolfSSL.
They won't be the same and thus won't interoperate.
* Fix for `EVP_CIPHER_CTX_flags`, which mapped to a missing function (broke openvpn)
* Added stack of name entries for ipsec/racoon support.
* Added `X509_STORE_CTX_set_flags` stub.
* Added PKCS7 NID types.
* Improved FIPS "SHA" logic in `test_wolfSSL_SHA`
* Added some uncommon NID type definitions.
* Expose the DH `DH_set_length` and `DH_set0_pqg` with OPENSSL_ALL
- Define `OPENSSL_COMPATIBLE_DEFAULTS` and `WOLFSSL_NO_OCSP_ISSUER_CHECK` for Apache config
- Fix `SSL_set_timeout` to match OpenSSL signature
- Implement `pkey` in `X509_INFO`
- Detect attempt to connect with plain HTTP
- Implement `wolfSSL_OCSP_request_add1_nonce`
- Set `ssl->cipher.bits` when calling `wolfSSL_get_current_cipher`
- Use custom flush method in `wolfSSL_BIO_flush` when set in BIO method
- Set the TLS version options in the `ssl->options` at the end of ClientHello parsing
- Don't modify the `ssl->version` when in a handshake (`ssl->msgsReceived.got_client_hello` is set)
- `wolfSSL_get_shutdown` returns a full bidirectional return when the SSL object is cleared. `wolfSSL_get_shutdown` calls `wolfSSL_clear` on a successful shutdown so if we detect a cleared SSL object, assume full shutdown was performed.
