summaryrefslogtreecommitdiff
path: root/src/lib/libssl (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* Only reset TLS extension state when parsing client hello or server hello.jsing2020-05-101-5/+7
| | | | | | | | With TLSv1.3 we end up parsing extensions from more than just these two messages. This can result in variables (like the selected alpn) being freed when things still need them. ok tb@
* Correct tlsext_ocsp_resplen check.jsing2020-05-101-2/+2
| | | | | | | This variable is currently overloaded - a value of -1 means that it is "unset" and any other value is a length. ok tb@
* Back out server side CCS sending. It breaks TLSv1.3 client communicationtb2020-05-093-34/+3
| | | | | | | | with TLSv1.2 servers, since it makes clients send their dummy CCS too early... There's an obvious but dirty bandaid which I can't bring myself to applying - this business is already disgusting enough. Issue found the hard way by sthen
* Forcibly ensure that only PSS may be used with RSA in TLS 1.3.beck2020-05-091-2/+8
| | | | | | | This prevents us from incorrectly choosing a PKCS1 based signature if the client advertises support for them but also prefers them to PSS such as appears to be the case with gnuTLS. ok jsing@
* Send dummy ChangeCipherSpec messages from the TLSv1.3 servertb2020-05-093-3/+34
| | | | | | | | | If the client has requested middle box compatibility mode by sending a non-empty legacy_session_id, the server must send a dummy CCS right after its first handshake message. This means right after ServerHello or HelloRetryRequest. ok jsing
* Send dummy ChangeCipherSpec messages from the TLSv1.3 client.jsing2020-05-094-6/+45
| | | | | | | | When operating in middlebox compatibility mode, the TLSv1.3 client needs to send a dummy ChangeCipherSpec message immediately before its second flight of handshake messages (when early data is not offered). ok tb@
* Correct return value check to handle TLS13_IO_EOF case.jsing2020-05-091-2/+2
|
* Add a middlebox_compat flag and condition session ID randomisation on it.jsing2020-05-093-4/+7
| | | | ok tb@
* Add support for certificate status requests in TLS 1.3 clientbeck2020-05-094-12/+81
| | | | ok jsing@, tb@, inoguchi@
* Make the test for the legacy_compression_method vector in the ClientHellotb2020-05-091-12/+7
| | | | | | | | stricter. Previously, we would accept any vector if it advertised the "null" compression method. RFC 8446 4.1.2 specifies that the only legal vector has length one and contains a zero byte for the null method. ok jsing
* Drop a redundant test. It's effectively doing the same test twicetb2020-05-091-3/+2
| | | | | | | and if the two lengths differed, the later CBS_write_bytes() would correctly fail anyway. Discussed with jsing
* On receiving an overlong session ID terminate with an illegal_parametertb2020-05-091-1/+6
| | | | | | alert. Found with tlsfuzzer. ok jsing
* Add support for HelloRetryRequests in the TLSv1.3 server.jsing2020-05-092-10/+73
| | | | ok inoguchi@ tb@
* crazy whitespace on one linetb2020-05-091-2/+2
|
* Pull the sending of alerts up into tls13_handshake_perform().jsing2020-05-091-14/+11
| | | | | | | | | | This fixes the case where a send function signals that an alert should be sent, then returns failure. Previously the failure would be propagated up, without the alert being sent. Issued noted by tb@ ok tb@
* Refactor tls13_server_hello_sent().jsing2020-05-091-30/+36
| | | | | | | | | Split the record protection engagement code into a separate tls13_server_engage_record_protection() function and call this from tls13_server_hello_sent(). Also move some functions around to keep the logical ordering/grouping. ok inoguchi@ tb@ (as part of a larger diff)
* On receiving a handshake or alert record with empty inner plaintext,tb2020-05-071-1/+11
| | | | | | | | | terminate the connection with an unexpected_message alert. See RFC 8446 section 5.4. Found with tlsfuzzer hint/ok jsing
* Accept two ChangeCipherSpec messages during a TLSv1.3 handshake.jsing2020-05-031-3/+3
| | | | | | | | | | | | | | | | | | | | | In compatibility mode, a TLSv1.3 server MUST send a dummy CCS message immediately after its first handshake message. This is normally after the ServerHello message, but it can be after the HelloRetryRequest message. As such we accept one CCS message from the server during the handshake. However, it turns out that in the HelloRetryRequest case, Facebook's fizz TLSv1.3 stack sends CCS messages after both the HelloRetryRequest message and the ServerHello message. This is unexpected and as far as I'm aware, no other TLSv1.3 implementation does this. Unfortunately the RFC is rather ambiguous here, which probably means it is not strictly an RFC violation. Relax the CCS message handling to allow two dummy CCS messages during a TLSv1.3. This makes our TLSv1.3 client work with Facebook Fizz when HRR is triggered. Issue discovered by inoguchi@ and investigated by tb@. ok deraadt@ tb@
* Add const to TLS1.3 internal vectorsinoguchi2020-05-022-14/+14
| | | | ok tb@
* tls13_record_layer internal functions to static in libsslinoguchi2020-04-291-4/+4
| | | | | | | We might remove static again for further regress around record layer in the future. ok jsing@ tb@
* tls13_handshake internal functions to static in libsslinoguchi2020-04-291-11/+12
| | | | ok jsing@ tb@
* Move legacy stack interfacing functions into tls13_legacy.c.jsing2020-04-284-199/+206
| | | | | | No functional change. ok inoguchi@ tb@
* Rename tls13_client_synthetic_handshake_message() and move to tls13_lib.c.jsing2020-04-283-47/+48
| | | | | | | | The server-side will need to use the same function. No functional change. ok inoguchi@ tb@
* Shuffle some functions around.jsing2020-04-272-329/+328
| | | | | | | | | | Move functions so that they are in the order that the TLSv1.3 messages are processed. While here, also move tls13_client_end_of_early_data_send() from tls13_client.c to tls13_server.c. No functional change. ok beck@ tb@
* Switch to NEGOTIATED when using WITHOUT_HRR.jsing2020-04-251-4/+9
| | | | | | | | | | This ensures that we remain in a valid handshake state in the TLSv1.3 server. Ideally we would not switch to NEGOTIATED until after record protection has been enabled, but we'll revisit this later. Issue noted by inoguchi@ ok tb@
* Move unsupported, obsolete ciphers and deprecated aliases out ofschwarze2020-04-251-31/+29
| | | | | | | | | | | the main list of words to make it more readable, even though it remains long. Avoid using deprecated aliases in explanations what other words mean. Stop documenting aDSS because it is *both* a deprecated alias *and* no longer matches anything at all. General direction discussed with jsing@ some time ago.
* tweak the wording to make it clearer under which conditions exactlyschwarze2020-04-251-4/+4
| | | | | the TLSv1.3 cipher suites are made available, too; related to ssl_ciph.c rev. 1.115
* Improve TLSv1.3 state machine for HelloRetryRequest handling.jsing2020-04-225-66/+104
| | | | | | | | | | | | | | | | | | | The state machine currently handles the HelloRetryRequest case by using WITH_HRR - in other words, we're explicitly indicating when we transition to the alternate path. The problem here is that we do not know if we're going to receive a ServerHello or a HelloRetryRequest until we process the message. This means that the ServerHello processing code has to handle both types of messages. The state machine and associated processing code becomes cleaner if we flip this around so that we assume we are going to receive a HelloRetryRequest and upon discovering that it is not, trigger WITHOUT_HRR and hand off to the ServerHello processing function. In particular, this makes the logic much more straight forward on the server side, when adding support for HRR. With feedback from tb@ ok tb@
* Handle TLSv1.3 key shares other than X25519 on the server side.jsing2020-04-212-16/+34
| | | | | | | | Previously we would only select an X25519 key share from the client, ignoring any others. Change this so that we will select the first of the key shares that matches one of our supported groups. ok beck@ inoguchi@ tb@
* Consolidate TLSv1.3 constants.jsing2020-04-213-40/+47
| | | | | | | Move all of the TLSv1.3 constants to the top of tls13_lib.c. Also mark these all as const so that they end up in .rodata rather than .data. ok tb@
* Provide TLSv1.3 cipher suite aliases to match the names used in RFC 8446.jsing2020-04-191-2/+25
| | | | ok beck@ inoguchi@ tb@
* Fix wrapping/indentation.jsing2020-04-181-4/+3
|
* Expose the peer ephemeral public key used for TLSv1.3 key exchange.jsing2020-04-185-36/+79
| | | | | | | | | SSL_get_server_tmp_key() provides the peer ephemeral public key used for key exchange. In the case of TLSv1.3 this is essentially the peer public key from the key share used for TLSv1.3 key exchange, hence make it availaable via SSL_get_server_tmp_key(). ok inoguchi@ tb@
* Tweak previous active cipher suite code.jsing2020-04-181-6/+5
| | | | | | | Use a boolean value rather than using a counter, as suggested by tb@ during the previous review. ok tb@
* Allow more key share groups for TLSv1.3.jsing2020-04-181-21/+12
| | | | | | | | The key share code previously only allowed for key shares to be generated using one of the groups in our default list (X25519, secp256r1, secp384r1). Relax this and allow key shares using any of the groups in our NID list. ok inoguchi@ tb@
* Only include TLSv1.3 cipher suites if there are active cipher suites.jsing2020-04-171-2/+10
| | | | | | | | | | Revise the previous so that we only include TLSv1.3 cipher suites if the cipher rule string resulted in at least one active cipher suite. This more closely matches OpenSSL behaviour. Noted and fix tested by schwarze@ ok beck@ tb@
* Generate client key share using our preferred group.jsing2020-04-174-25/+37
| | | | | | | | Generate a client key share using our preferred group, rather than always using X25519. This means that the key share group can be controlled via SSL{_CTX,}_set1_groups() and SSL{_CTX,}_set1_groups_list(). ok beck@
* Update in several respects:schwarze2020-04-141-13/+11
| | | | | | | | * mention TLSv1.3 * remove DSS, DES(56), RC4(64), and IDEA(128), which are no longer supported * remove ChaCha20-Poly1305-Old and STREEBOG512 which don't exist in LibreSSL * correct the instruction for printing the complete list OK jsing@
* add the missing sentence "LibreSSL no longer provides any suchschwarze2020-04-141-2/+3
| | | | | cipher suites" to the DES entry and use the same wording for DSS; OK jsing@
* Delete the three sentences listing the ciphers currently includedschwarze2020-04-141-15/+2
| | | | | | | | | in LOW, MEDIUM, and HIGH. That's going to change repeatedly and the extra maintenance effort for keeping it up to date is a waste because people can trivially run "openssl ciphers -v LOW" to look it up. Besides, updating it will usually be forgotten; the LOW entry was already wrong. Suggested by jsing@.
* Document the TLSv1.3 control word, update the description of theschwarze2020-04-111-4/+30
| | | | | | | TLSv1 control word, and explain how TLSv1.3 cipher suites can be configured in LibreSSL and in OpenSSL. While here, also mention how users can inspect the DEFAULT list of cipher suites. Stimulus, feedback and OK from jsing@.
* Include TLSv1.3 cipher suites unless cipher string references TLSv1.3.jsing2020-04-091-6/+19
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | OpenSSL has always taken the approach of enabling almost everything by default. As a result, if you wanted to run a secure TLS client/server you had to specify your own "secure" cipher string, rather than being able to trust the defaults as being sensible and secure. The problem is that with the introduction of TLSv1.3, most of these "secure" cipher strings result in the new TLSv1.3 cipher suites being excluded. The "work around" for this issue in OpenSSL was to add a new TLSv1.3 API (SSL_CTX_set_ciphersuites(), SSL_set_ciphersuites()) and have separate knobs for the pre-TLSv1.3 and TLSv1.3 cipher suites. This of course means that every application now needs to call two APIs, but it does mean that applications that only call SSL_CTX_set_cipher_list()/SSL_set_cipher_list() cannot remove TLSv1.3 cipher suites and prevent TLSv1.3 from working. We've taken a different approach and have allowed TLSv1.3 cipher suites to be manipulated via the existing SSL_set_cipher_list() API. However, in order to avoid problems with hardcoded cipher strings, change this behaviour so that we always include TLSv1.3 cipher suites unless the cipher string has a specific reference to the TLSv1.3 protocol or a TLSv1.3 cipher suite. This means that: $ openssl ciphers -v TLSv1.2:!TLSv1.3 still gives TLSv1.2 only cipher suites and: $ openssl ciphers -v AEAD-CHACHA20-POLY1305-SHA256 only lists a single TLSv1.3 cipher, however: $ openssl ciphers -v ECDHE-RSA-AES256-GCM-SHA384 now includes both TLSv1.2 ECDHE-RSA-AES256-GCM-SHA384 and all TLSv1.3 cipher suites (which also matches OpenSSL's openssl(1) behaviour). Issue encountered by kn@ with mumble. ok tb@
* Tidy line wrapping and remove an extra blank line.jsing2020-04-091-4/+3
|
* ssl_aes_is_accelerated() returns a boolean - treat it as such, rather thanjsing2020-04-091-2/+2
| | | | explicitly comparing against a value.
* Ensure legacy session ID is persistent during client TLS session.jsing2020-04-081-9/+14
| | | | | | | | | | | | Generate an unpredictable 32-byte legacy session ID during client initialisation, rather than when the ClientHello message is being created. Otherwise in the case of a HelloRetryRequest the legacy session ID values will differ between the first and second ClientHello messages, which is not permitted by the RFC. Fixes an issue talking TLSv1.3 to smtp.mail.yahoo.com. ok beck@
* Send a zero-length session identifier if TLSv1.3 is not enabled.jsing2020-04-061-4/+7
| | | | | | | | | If the maximum version is less than TLSv1.3, send a zero-length session identifier (matching the behaviour of the legacy TLS stack), rather than a 32 byte random identifier. The 32 byte random identifier is only needed for "compatibility" mode in TLSv1.3. ok beck@
* Void functions obviously do not return values; no need to elaborate.schwarze2020-03-305-31/+10
| | | | Patch from Martin Vahlensieck <academicsolutions dot ch>.
* Void functions obviously do not return values; no need to elaborate.schwarze2020-03-291-5/+2
| | | | Patch from Martin Vahlensieck <academicsolutions dot ch>.
* Consistently spell 'unsigned' as 'unsigned int', as style(9) seemstb2020-03-167-44/+45
| | | | | | | | | to prefer that. No binary change except in d1_srtp.c where the generated assembly differs only in line numbers (due to a wrapped long line) and in s3_cbc.c where there is no change in the generated assembly. ok inoguchi jsing
* The RFC is clear (section 5.3) that sequence number should never wrap.tb2020-03-161-5/+12
| | | | | | | | | We currently throw an error on overflow, but still wrap. Check up front if we would need to wrap and only increment if that case is excluded. This simplifies the increment loop and makes the returns in this function less magic. ok jsing
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-03-26 12:03:58 +0000