summaryrefslogtreecommitdiff
path: root/src/lib/libssl (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* Cleanup more of tls_decrypt_ticket().jsing2019-04-211-5/+9
| | | | | | | | | | | Separate the malloc() check and EVP_DecryptUpdate() - the malloc() failure is fatal while a EVP_DecryptUpdate() is a decryption failure. Also ensure that we clear the error stack in all cases where we are indicating a failure to decrypt or decode the ticket - otherwise SSL_error() while later return failure when it should not. ok tb@
* Start cleaning up tls_decrypt_ticket().jsing2019-04-211-58/+63
| | | | | | | | | Rather than returning from multiple places and trying to clean up as we go, move to a single exit point and clean/free in one place. Also invert the logic that handles NULL sessions - fail early, rather than having an indented if test for success. ok tb@
* Avoid leak in SSL_dup_CA_list()tb2019-04-131-8/+14
| | | | | | | | In the case that X509_NAME_dup() succeeds, but sk_X509_NAME_push() fails, name is leaked. The entire function is trying to be clever and therefore hard to follow. Let's do it the stupid but safe way. ok jsing
* Recommend SSL_CTX_add1_chain_cert(3) rather thanschwarze2019-04-091-12/+11
| | | | | | | SSL_CTX_add_extra_chain_cert(3). From Dr. Stephen Henson <steve at openssl dot org> via OpenSSL commit a4339ea3 Jan 3 22:38:03 2014 +0000 which is still under a free license.
* Document SSL_CTX_clear_mode(3) and SSL_clear_mode(3).schwarze2019-04-091-22/+48
| | | | | | | From Kurt Roeckx <kurt at roeckx dot be> via OpenSSL commit 57fd5170 May 13 11:24:11 2018 +0200 which is still under a free license. While here, polish awkward wording and reduce duplication.
* By design, our state machine is a DAG contrary to the state machine intb2019-04-054-7/+29
| | | | | | | | | | | | the spec. To avoid the obvious loop in the RFC's state machine, we added a CLIENT_HELLO_RETRY state which is a second ClientHello with special rules. There is, however, no state to react to this second client hello. This adds a matching SERVER_HELLO_RETRY state to the handshakes table. This means in particular that the WITH_HRR state cannot be set in tls13_server_hello_recv(), so remove this now dead check. ok jsing
* Import SSL_CTX_add1_chain_cert(3) from OpenSSL branch 1.1.1, which is stillschwarze2019-04-055-17/+264
| | | | | | | | | under a free license, omitting functions we don't have and tweaked by me; the functions were provided by jsing@ in ssl.h rev. 1.166. While here, also document SSL_CTX_get_extra_chain_certs(3) because it is closely related to companion functions are already documented and the API is kind of incomplete without it.
* Zap two dead #defines that were unused since jsing deleted thetb2019-04-051-5/+1
| | | | | | record_type member of the tls13_handshake_action struct. ok jsing
* Implement legacy fallback for the TLS 1.3 client.jsing2019-04-043-8/+87
| | | | | | | If the Server Hello received indicates that the server did not negotiate TLS 1.3, fallback to the original TLS client implementation. ok bcook@, tb@
* Clean up the cipher/digest table mess.jsing2019-04-043-179/+45
| | | | | | | | | | The original implementation allows for libcrypto to be compiled without a given algorithm and libssl then detects that ciphers or digests are unavailable so that it can disable the associated cipher suites. This is unnecessary since we do not compile out algorithms. ok beck@, tb@ (a while back)
* Bump libssl/libtls minors due to symbol addition.jsing2019-04-041-1/+1
|
* Provide SSL chain/cert chain APIs.jsing2019-04-043-2/+161
| | | | | | | | | | | These allow for chains to be managed on a per-certificate basis rather than as a single "extra certificates" list. Note that "chain" in this context does not actually include the leaf certificate however, unlike SSL_CTX_use_certificate_chain_{file,mem}(). Thanks to sthen@ for running this through a bulk ports build. ok beck@ tb@
* This case also needs to be fatal.jsing2019-04-041-1/+2
|
* mark ERR_R_INTERNAL_ERROR instead of assert or _exit, sighderaadt2019-04-011-3/+3
|
* Correct subtle bug in sigalgs, only care about curve_nid if we arebeck2019-04-011-4/+4
| | | | | checking the curve. ok jsing@ tb@
* Clean up and simplify the client verify code:jsing2019-03-311-50/+49
| | | | | | | | | | - Be consistent with _len naming. - Use size_t where possible/appropriate. - Group the CBB code. - Use EVP_MAX_MD_SIZE consistently, instead of "magic" values. - Switch GOST to EVP_DigestSign*, making it similar to sigalgs. ok tb@ a while back.
* Strip out all of the pkey to sigalg and sigalg to pkey linkages.jsing2019-03-255-59/+8
| | | | | | These are no longer used now that we defer signature algorithm selection. ok beck@
* tls1_process_sigalgs() is no longer needed.jsing2019-03-252-57/+2
| | | | ok beck@
* Defer sigalgs selection until the certificate is known.jsing2019-03-259-48/+124
| | | | | | | | | | | | | Previously the signature algorithm was selected when the TLS extension was parsed (or the client received a certificate request), however the actual certificate to be used is not known at this stage. This leads to various problems, including the selection of a signature algorithm that cannot be used with the certificate key size (as found by jeremy@ via ruby regress). Instead, store the signature algorithms list and only select a signature algorithm when we're ready to do signature generation. Joint work with beck@.
* Rework ssl_ctx_use_certificate_chain_bio() to use the CERT_PKEY chain.jsing2019-03-251-46/+26
| | | | | | | | | | This means that any additional CA certificates end up on the per certificate chain, rather than the single/shared extra_certs. Also simplify this code and in particular, avoid setting the return value to indicate success until we've actually succeeded. ok beck@ tb@
* Remove ssl_get_server_send_cert() which is now unused.jsing2019-03-252-14/+2
| | | | ok beck@ tb@
* Rework ssl3_output_cert_chain() to take a CERT_PKEY and consider chains.jsing2019-03-254-40/+36
| | | | | | | | | | We will now include the certificates in the chain in the certificate list, or use the existing extra_certs if present. Failing that we fall back to the automatic chain building if not disabled. This also simplifies the code significantly. ok beck@ tb@
* Add a chain member to CERT_PKEY and provide functions for manipulating it.jsing2019-03-252-3/+74
| | | | | | | | | Note that this is not the full chain, as the leaf certificate currently remains in the x509 member of CERT_PKEY. Unfortunately we've got to contend with the fact that some OpenSSL *_chain_* APIs exclude the leaf certificate while others include it... ok beck@ tb@
* If ssl_cipher_apply_rule() is given a specific cipher suite, match on it.jsing2019-03-241-3/+4
| | | | | | | | | | Otherwise matching a specific cipher is performed by matching against its characteristics, which can result in multiple rather than a single match. Found by bluhm@'s regress tests. ok bluhm@ tb@
* do not call assert(), which has a tendency to leave traces of stuff inderaadt2019-03-241-2/+3
| | | | | | corefiles. Instead call OPENSSL_assert(), which has recently been trained to do this in a safer (if more awkward to debug) way. discussed with jsing and beck a while back
* Avoid an internal 2 byte overread in ssl_sigalgs().jsing2019-03-191-7/+2
| | | | | | Found by oss-fuzz, fixes issue #13797. ok beck@ tb@
* Revert TLS1_get{,_client}_version simplification because DTLS.jsing2019-03-194-14/+15
|
* * note that the handshake must be completed firstschwarze2019-03-181-4/+15
| | | | | | | * correct the description of "unknown" (the previous are both from OpenSSL 1.1.1, still under a free license) * add a comment saying that TLS1_get_version() and TLS1_get_client_version() are intentionally undocumented (reasons provided by jsing@)
* In ssl.h rev. 1.165, tb@ provided version agnostic DTLS methods.schwarze2019-03-181-3/+26
| | | | Document them.
* bump minors after symbol additiontb2019-03-171-1/+1
|
* Provide version agnostic DTLS methods.jsing2019-03-173-2/+27
| | | | ok tb@
* Correct return value handling in tls13_handshake_recv_action().jsing2019-03-171-2/+4
| | | | | | | | The recv action handler returns success/failure, rather than a TLS13_IO_* value, which is what tls13_handshake_recv_action() needs to return. Failure previously mapped to TLS13_IO_EOF, which is not ideal. ok tb@
* Partially clean up the TLS1_get_{,client}_version macros.jsing2019-03-174-15/+14
| | | | | | | | | LibreSSL only supports TLSv1.0 and above, hence the checks the macros are performing are useless. Simplify them to their effective code. Also place both under #ifndef LIBRESSL_INTERNAL and use the variables directly in our code, which improves readability. ok tb@
* Remove the alert level from the TLSv1.3 alert callback.jsing2019-03-173-7/+6
| | | | | | | | In TLSv1.3 the alert level is effectively meaningless and the record layer has already checked that it is appropriate. As such, drop it from the alert callback. ok tb@
* Don't index a void pointer, fixes compilation with visual studio.millert2019-03-041-2/+3
| | | | | Gcc/clang will treat void * as char * but this is non-standard. OK deraadt@ jsing@ inoguchi@
* Implement non-SSL_MODE_ENABLE_PARTIAL_WRITE in tls13_legacy_write_bytes().jsing2019-02-281-3/+34
| | | | | | | In non-SSL_MODE_ENABLE_PARTIAL_WRITE mode we have to write out all the things and only return success once all of the data has been sent. ok inoguchi@ tb@
* Automatically complete the handshake from tls13_legacy_{read,write}_bytes()jsing2019-02-283-3/+17
| | | | | | | | | | | | | | If the TLS handshake has not been completed, automatically complete the handshake as part of the read/write call, implementing the current SSL_read()/SSL_write() behaviour. Once the TLS handshake is completed we push a WANT_POLLIN or WANT_POLLOUT back up to the caller, since some applications appear to incorrectly call SSL_read() or SSL_write(), rather than repeating the previous call. This can lead to attempts to read data that does not exist, since the WANT_POLLIN was actually triggered as part of the handshake. ok inoguchi@ tb@
* Set the SSL state as part of tls13_legacy_connect().jsing2019-02-281-2/+5
| | | | | | | | Set the SSL state to SSL_ST_CONNECT during the TLSv1.3 handshake and to SSL_ST_OK once the handshake completes, since some applications currently rely on this information to function correctly. ok inoguchi@ tb@
* Add appropriate length checks to tls13_legacy_{read,write}_bytes()jsing2019-02-281-3/+9
| | | | ok inoguchi@ tb@
* Remove unused record_type from TLSv1.3 handshake actions.jsing2019-02-281-16/+1
| | | | ok tb@
* Correct return values for non-fatal alerts.jsing2019-02-261-3/+3
| | | | | | | | | In the close notify case we need to signal EOF and in the user cancelled case we need to return WANT_POLLIN. Returning success results in tls13_record_layer_read_record() thinking that we have record data when we do not, which then results in the content type check later failing. ok tb@
* Correctly propagate EOF from BIO in tls13_legacy_wire_read().jsing2019-02-261-1/+3
| | | | ok tb@
* Fix gcc3 compilation error by using a forward declaration instead of atb2019-02-251-3/+3
| | | | | | repeated typedef. Found the hard way by aoyama who also tested the fix. ok jsing
* revert "Move struct ssl_handshake_tls13_st definition", i.e., movetb2019-02-252-24/+26
| | | | | | SSL_HANDSHAKE_TLS13 back to ssl_locl.h. discussed with jsing and inoguchi
* Correctly handle oversize writes.jsing2019-02-251-2/+4
| | | | | | | | | | | | If the record layer is asked to write more than fits in a plaintext record, cap the amount at that limit. This means that we will effectively write out a single record and return a short-write. This behaviour matches SSL_write() with SSL_MODE_ENABLE_PARTIAL_WRITE enabled and the non-SSL_MODE_ENABLE_PARTIAL_WRITE case will be handled at a higher layer. ok inoguchi@ tb@
* Add a handshake action sent handler and use it for client finished.jsing2019-02-253-11/+22
| | | | | | | | | | The write traffic key needs to be changed to the client application traffic key after the client finished message has been sent. The send handler generates the client finished message, however we cannot switch keys at this stage since the client finished message has not yet been protected by the record layer. ok tb@
* Correct a few error returns.jsing2019-02-251-4/+4
|
* Move struct ssl_handshake_tls13_st definitioninoguchi2019-02-252-26/+24
| | | | | | This solves build error on luna88k with gcc3. ok aoyama@ jca@ jsing@ tb@
* Remove temporary gotos.jsing2019-02-231-7/+3
| | | | | | | | | In the case of a dummy CCS or post-handshake handshake message, return TLS13_WANT_POLLIN rather than using a goto internally. This allows the caller to retry at an appropriate time and reduces the complexity within the record layer. ok beck@ tb@
* Set BIO retry on TLS13_IO_WANT_POLLIN/TLS13_IO_WANT_POLLOUT.jsing2019-02-231-1/+3
| | | | | | | | | | | In most cases a TLS13_IO_WANT_POLLIN or TLS13_IO_WANT_POLLOUT will have bubbled up from the wire callbacks, in which case the BIO retry flag will already be set. However, if we return TLS13_IO_WANT_POLLIN or TLS13_IO_WANT_POLLOUT from a higher layer the BIO retry flag will not be set and that will cause SSL_get_error() to return SSL_ERROR_SYSCALL rather than the intended SSL_ERROR_WANT_READ/SSL_ERROR_WANT_WRITE. ok beck@ tb@
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-03-22 00:22:08 +0000