The recent SSL error overhaul left a case where an error occurs but
SSLfatal() is not called.
Credit to OSSfuzz for finding this issue.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4847)
The most likely explanation for us ending up at this point in the code
is that we were called by the user application incorrectly - so use an
appropriate error code.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4778)
Follow up from the conversion to use SSLfatal() in the state machine to
clean things up a bit more.
[extended tests]
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4778)
We shouldn't call SSLfatal() multiple times for the same error condition.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4778)
Sometimes at the top level of the state machine code we know we are
supposed to be in a fatal error condition. This commit adds some sanity
checks to ensure that SSLfatal() has been called.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4778)
This is an initial step towards using SSLfatal() everywhere. Initially in
this commit and in subsequent commits we focus on the state machine code.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4778)
Typically if a fatal error occurs three things need to happen:
- Put an error on the error queue
- Send an alert
- Put the state machine into the error state
Although all 3 of these things need to be done every time we hit a fatal
error the responsibilities for doing this are distributed throughout the
code. The place where the error goes on the queue, where the alert gets
sent and where the state machine goes into the error state are almost
invariably different. It has been a common pattern to pass alert codes up
and down the stack to get the alert information from the point in the code
where the error is detected to the point in the code where the alert gets
sent.
This commit provides an SSLfatal() macro (backed by an ossl_statem_fatal
function) that does all 3 of the above error tasks. This is largely a drop
in replacement for SSLerr, but takes a couple of extra parameters (the SSL
object, and an alert code).
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4778)
This is a specific 1.1.1 change; do not squash if the chacha
prioritization code is to be backported
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4436)
IFF the client has ChaCha first, and server cipher priority is used,
and the new SSL_OP_PRIORITIZE_CHACHA_FOR_MOBILE option is used,
then reprioritize ChaCha above everything else. This way, A matching
ChaCha cipher will be selected if there is a match. If no ChaCha ciphers
match, then the other ciphers are used.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4436)
These functions needed updates for the various state machine states that
have been added for TLSv1.3.
Fixes#4795
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4801)
SNI needs to be consistent before we accept early_data. However a
server may choose to not acknowledge SNI. In that case we have to
expect that a client may send it anyway. We change the consistency
checks so that not acknowledging is treated more a like a "wild card",
accepting any SNI as being consistent.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4738)
We can only send early_data if the SNI is consistent. However it is valid
for the client to set SNI and the server to not use it. This would still be
counted as consistent. OpenSSL client was being overzealous in this check
and disallowing this scenario.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4738)
It's argued that /WX allows to keep better focus on new code, which
motivates its comeback...
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4721)
It's argued that /WX allows to keep better focus on new code, which
motivates its comeback...
[Keep this commit separate as reminder for time overhaul.]
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4721)
Around 138 distinct errors found and fixed; thanks!
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3459)
If SSL_read() is called with a zero length buffer, and we read a zero length
record then we should mark that record as read.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4685)
We currently increment the SSL_CTX stats.sess_accept field in
tls_setup_handshake(), which is invoked from the state machine well
before ClientHello processing would have had a chance to switch
the SSL_CTX attached to the SSL object due to a provided SNI value.
However, stats.sess_accept_good is incremented in tls_finish_handshake(),
and uses the s->ctx.stats field (i.e., the new SSL_CTX that was switched
to as a result of SNI processing). This leads to the confusing
(nonsensical) situation where stats.sess_accept_good is larger than
stats.sess_accept, as the "sess_accept" value was counted on the
s->session_ctx.
In order to provide some more useful numbers, increment
s->ctx.stats.sess_accept after SNI processing if the SNI processing
changed s->ctx to differ from s->session_ctx. To preserve the
property that any given accept is counted only once, make the
corresponding decrement to s->session_ctx.stats.sess_accept when
doing so.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/4549)
For client SSL objects and before any callbacks have had a chance
to be called, we can write the stats accesses using the session_ctx,
which makes sense given that these values are all prefixed with
"sess_".
For servers after a client_hello or servername callback has been
called, retain the existing behavior of modifying the statistics
for the current (non-session) context. This has some value,
in that it allows the statistics to be viewed on a per-vhost level.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/4549)
It is expected that SSL_CTX objects are shared across threads,
and as such we are responsible for ensuring coherent data accesses.
Aligned integer accesses ought to be atomic already on all supported
architectures, but we can be formally correct.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/4549)
Use the newly introduced sk_TYPE_new_reserve API to simplify the
reservation of stack as creating it.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4592)
Add a check for NULL return in t1_lib.c.
Since return type of ssl_cert_lookup_by_idx is pointer and unify coding
style, I changed from zero to NULL in ssl_cert.c.
Remove unnecessary space for ++.
Fix incorrect condition
Expression is always false because 'else if' condition matches previous
condition. SInce the next line of 'else if' condition has substituted
TLSEXT_ECPOINTFORMAT_ansiX962_compressed_char2, the 'else if'
condition should compare with NID_X9_62_characteristic_two_field.
Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4562)
Since return is inconsistent, I removed unnecessary parentheses and
unified them.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4541)
The drbg's lock must be held across calls to RAND_DRBG_generate()
to prevent simultaneous modification of internal state.
This was observed in practice with simultaneous SSL_new() calls attempting
to seed the (separate) per-SSL RAND_DRBG instances from the global
rand_drbg instance; this eventually led to simultaneous calls to
ctr_BCC_update() attempting to increment drbg->bltmp_pos for their
respective partial final block, violating the invariant that bltmp_pos < 16.
The AES operations performed in ctr_BCC_blocks() makes the race window
quite easy to trigger. A value of bltmp_pos greater than 16 induces
catastrophic failure in ctr_BCC_final(), with subtraction overflowing
and leading to an attempt to memset() to zero a very large range,
which eventually reaches an unmapped page and segfaults.
Provide the needed locking in get_entropy_from_parent(), as well as
fixing a similar issue in RAND_priv_bytes(). There is also an
unlocked call to RAND_DRBG_generate() in ssl_randbytes(), but the
requisite serialization is already guaranteed by the requirements on
the application's usage of SSL objects, and no further locking is
needed for correct behavior. In that case, leave a comment noting
the apparent discrepancy and the reason for its safety (at present).
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4328)
Reseeding is handled very differently by the classic RAND_METHOD API
and the new RAND_DRBG api. These differences led to some problems when
the new RAND_DRBG was made the default OpenSSL RNG. In particular,
RAND_add() did not work as expected anymore. These issues are discussed
on the thread '[openssl-dev] Plea for a new public OpenSSL RNG API'
and in Pull Request #4328. This commit fixes the mentioned issues,
introducing the following changes:
- Replace the fixed size RAND_BYTES_BUFFER by a new RAND_POOL API which
facilitates collecting entropy by the get_entropy() callback.
- Don't use RAND_poll()/RAND_add() for collecting entropy from the
get_entropy() callback anymore. Instead, replace RAND_poll() by
RAND_POOL_acquire_entropy().
- Add a new function rand_drbg_restart() which tries to get the DRBG
in an instantiated state by all means, regardless of the current
state (uninstantiated, error, ...) the DRBG is in. If the caller
provides entropy or additional input, it will be used for reseeding.
- Restore the original documented behaviour of RAND_add() and RAND_poll()
(namely to reseed the DRBG immediately) by a new implementation based
on rand_drbg_restart().
- Add automatic error recovery from temporary failures of the entropy
source to RAND_DRBG_generate() using the rand_drbg_restart() function.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4328)
The previous commit removed version negotiation on an HRR. However we should
still sanity check the contents of the version field.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4527)
Previously if a client received an HRR then we would do version negotiation
immediately - because we know we are going to get TLSv1.3. However this
causes a problem when we emit the 2nd ClientHello because we start changing
a whole load of stuff to ommit things that aren't relevant for < TLSv1.3.
The spec requires that the 2nd ClientHello is the same except for changes
required from the HRR. Therefore the simplest thing to do is to defer the
version negotiation until we receive the ServerHello.
Fixes#4292
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4527)
Now that we are moving to support named FFDH groups, these fields are not
ec-specific, so we need them to always be available.
This fixes the no-ec --strict-warnings build, since gcc
5.4.0-6ubuntu1~16.04.4 appears to always try to compile the static inline
functions from ssl_locl.h, even when they are not used in the current
compilation unit.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4518)
The function tls_check_curve is only called on clients and contains
almost identical functionaity to tls1_check_group_id when called from
a client. Merge the two.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4475)
When an SSL's context is swtiched from a ticket-enabled context to
a ticket-disabled context in the servername callback, no session-id
is generated, so the session can't be resumed.
If a servername callback changes the SSL_OP_NO_TICKET option, check
to see if it's changed to disable, and whether a session ticket is
expected (i.e. the client indicated ticket support and the SSL had
tickets enabled at the time), and whether we already have a previous
session (i.e. s->hit is set).
In this case, clear the ticket-expected flag, remove any ticket data
and generate a session-id in the session.
If the SSL hit (resumed) and switched to a ticket-disabled context,
assume that the resumption was via session-id, and don't bother to
update the session.
Before this fix, the updated unit-tests in 06-sni-ticket.conf would
fail test #4 (server1 = SNI, server2 = no SNI).
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/1529)
Remove all stack headers from some includes that don't use them.
Avoid a genearic untyped stack use.
Update stack POD file to include the OPENSSL_sk_ API functions in the notes
section. They were mentioned in the name section but not defined anywhere.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4430)
Rename tls1_get_curvelist to tls1_get_grouplist, change to void as
it can never fail and remove unnecessary return value checks. Clean
up the code.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/=4412)
Replace existing compression and groups check with two functions.
tls1_check_pkey_comp() checks a keys compression algorithms is consistent
with extensions.
tls1_check_group_id() checks is a group is consistent with extensions
and preferences.
Rename tls1_ec_nid2curve_id() to tls1_nid2group_id() and make it static.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/=4412)
Setup EVP_PKEY structure from a group ID in ssl_generate_param_group,
replace duplicate code with this function.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/=4412)
Replace tls1_ec_curve_id2nid() with tls_group_id_lookup() which returns
the TLS_GROUP_INFO for the group.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/=4412)
Instead of storing supported groups in on-the-wire format store
them as parsed uint16_t values. This simplifies handling of groups
as the values can be directly used instead of being converted.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4406)
Compilation failed due to -Werror=misleading-indentation.
Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4395)
Allo RSA certificate to be used for RSA-PSS signatures: this needs
to be explicit because RSA and RSA-PSS certificates are now distinct
types.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4368)
OpenSSL 1.1.0 made SSL_CTX and SSL structs opaque and introduced a new
API to set the minimum and maximum protocol version for SSL_CTX with
TLS_method(). Add getters to introspect the configured versions:
int SSL_CTX_get_min_proto_version(SSL_CTX *ctx);
int SSL_CTX_get_max_proto_version(SSL_CTX *ctx);
int SSL_get_min_proto_version(SSL *ssl);
int SSL_get_max_proto_version(SSL *ssl);
NOTE: The getters do not resolv the version in case when the minimum or
maxium version are configured as '0' (meaning auto-select lowest and
highst version number).
Signed-off-by: Christian Heimes <christian@python.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4364)
It is otherwise unclear what all the magic numbers mean.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4349)
"Early callback" is a little ambiguous now that early data exists.
Perhaps "ClientHello callback"?
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4349)
In OpenSSL 1.1.0, when there were no extensions added to the ServerHello,
we did not write the extension data length bytes to the end of the
ServerHello; this is needed for compatibility with old client implementations
that do not support TLS extensions (such as the default configuration of
OpenSSL 0.9.8). When ServerHello extension construction was converted
to the new extensions framework in commit
7da160b0f4, this behavior was inadvertently
limited to cases when SSLv3 was negotiated (and similarly for ClientHellos),
presumably since extensions are not defined at all for SSLv3. However,
extensions for TLS prior to TLS 1.3 have been defined in separate
RFCs (6066, 4366, and 3546) from the TLS protocol specifications, and as such
should be considered an optional protocol feature in those cases.
Accordingly, be conservative in what we send, and skip the extensions block
when there are no extensions to be sent, regardless of the TLS/SSL version.
(TLS 1.3 requires extensions and can safely be treated differently.)
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/4296)
This is actually not all warnings, only return values.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4297)
Move struct timeval includes into e_os.h (where the Windows ones were).
Enaure that the include is guarded canonically.
Refer #4271
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4312)
SNI and ALPN must be set to be consistent with the PSK. Otherwise this is
an error.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/3926)
The returned ID matches with what IANA specifies (or goes on the
wire anyway, IANA notwithstanding).
Doc is added.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4107)
cryptilib.h is the second.
Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4188)
Removed e_os.h from all bar three headers (apps/apps.h crypto/bio/bio_lcl.h and
ssl/ssl_locl.h).
Added e_os.h into the files that need it now.
Directly reference internal/nelem.h when required.
Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4188)
The one creating the DRBG should instantiate it, it's there that we
know which parameters we should use to instantiate it.
This splits the rand init in two parts to avoid a deadlock
because when the global drbg is created it wands to call
rand_add on the global rand method.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
GH: #4268
A condition was removed by commit 1053a6e2281d; presumably it was an
unintended change. Restore the previous behavior so the get_session_cb
won't be called with zero-length session ID.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/4236)
Remove GETPID_IS_MEANINGLESS and osslargused.
Move socket-related things to new file internal/sockets.h; this is now
only needed by four(!!!) files. Compiles should be a bit faster.
Remove USE_SOCKETS ifdef's
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4209)
Cast arguments to the various ctype functions to unsigned char to match their
documentation.
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4203)
Force non-empty padding extension.
When enabled, force the padding extension to be at least 1 byte long.
WebSphere application server cannot handle having an empty
extension (e.g. EMS/EtM) as the last extension in a client hello.
This moves the SigAlgs extension last for TLSv1.2 to avoid this
issue.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3921)
The existing function SSL_get_current_cipher() queries the
current session for the ciphersuite in use, but there is no way
for application code to determine what ciphersuite has been
negotiated and will be used in the future, prior to ChangeCipherState
(or the TLS 1.3 equivalent) causing the new cipher to take effect and
become visible in the session information. Expose this information
to appropriate application callbacks to use during the handshake.
The name SSL_get_pending_cipher() was chosen for compatibility with
BoringSSL's routine of that name.
Improve the note on macro implementations in SSL_get_current_cipher.pod
while here.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4070)
Commit 02f0274e8c moved ALPN processing
into an extension finalization function, as the only documented ordering
requirement from previous commits was that ALPN processing occur after
SNI processing, and SNI processing is performed before the extension
finalization step. However, it is useful for applications'
alpn_select callbacks to run after ciphersuite selection as well -- at
least one application protocol specification (HTTP/2) imposes restrictions
on which ciphersuites are usable with that protocol. Since it is generally
more preferrable to have a successful TLS connection with a default application
protocol than to fail the TLS connection and not be able to have the preferred
application protocol, it is good to give the alpn_select callback information
about the ciphersuite to be used, so that appropriate restrctions can be
enforced in application code.
Accordingly, split the ALPN handling out into a separate tls_handl_alpn()
function akin to tls_handle_status_request(), called from
tls_post_process_client_hello(). This is an alternative to resuscitating
ssl_check_clienthello_tlsext_late(), something of an awkwward name itself.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4070)
If s->hit is set, s->session corresponds to a session created on
a previous connection, and is a data structure that is potentially
shared across other SSL objects. As such, there are thread-safety
issues with modifying the structure without taking its lock (and
of course all corresponding read accesses would also need to take
the lock as well), which have been observed to cause double-frees.
Regardless of thread-safety, the resumed session object is intended
to reflect parameters of the connection that created the session,
and modifying it to reflect the parameters from the current connection
is confusing. So, modifications to the session object during
ClientHello processing should only be performed on new connections,
i.e., those where s->hit is not set.
The code mostly got this right, providing such checks when processing
SNI and EC point formats, but the supported groups (formerly
supported curves) extension was missing it, which is fixed by this commit.
However, TLS 1.3 makes the suppported_groups extension mandatory
(when using (EC)DHE, which is the normal case), checking for the group
list in the key_share extension processing. But, TLS 1.3 only [0] supports
session tickets for session resumption, so the session object in question
is the output of d2i_SSL_SESSION(), and will not be shared across SSL
objects. Thus, it is safe to modify s->session for TLS 1.3 connections.
[0] A psk_find_session callback can also be used, but the restriction that
each callback execution must produce a distinct SSL_SESSION structure
can be documented when the psk_find_session callback documentation is
completed.
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4123)
Give each SSL object it's own DRBG, chained to the parent global
DRBG which is used only as a source of randomness into the per-SSL
DRBG. This is used for all session, ticket, and pre-master secret keys.
It is NOT used for ECDH key generation which use only the global
DRBG. (Doing that without changing the API is tricky, if not impossible.)
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/4050)
If RAND_add wraps around, XOR with existing. Add test to drbgtest that
does the wrap-around.
Re-order seeding and stop after first success.
Add RAND_poll_ex()
Use the DF and therefore lower RANDOMNESS_NEEDED. Also, for child DRBG's,
mix in the address as the personalization bits.
Centralize the entropy callbacks, from drbg_lib to rand_lib.
(Conceptually, entropy is part of the enclosing application.)
Thanks to Dr. Matthias St Pierre for the suggestion.
Various code cleanups:
-Make state an enum; inline RANDerr calls.
-Add RAND_POLL_RETRIES (thanks Pauli for the idea)
-Remove most RAND_seed calls from rest of library
-Rename DRBG_CTX to RAND_DRBG, etc.
-Move some code from drbg_lib to drbg_rand; drbg_lib is now only the
implementation of NIST DRBG.
-Remove blocklength
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/4019)
Move the definition of ossl_assert() out of e_os.h which is intended for OS
specific things. Instead it is moved into internal/cryptlib.h.
This also changes the definition to remove the (int) cast.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4073)
The comment "The following should not return 1, otherwise, things
are very strange" is from the very first commit of OpenSSL. The
really meaning of the comment is if the identical session can be
found from internal cache after calling get_session_cb but not
found before calling get_session_cb, it is just strange.
The value 1 was originated from the old doc of SSLeay, reversed
from the actual return value of SSL_CTX_add_session().
Anyway either return value of SSL_CTX_add_session() should not
interrupt the session resumption process. So the checking of
return value of SSL_CTX_add_session() is not necessary.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4014)
If a new_session_cb is set then it was only ever getting invoked if !s->hit
is true. This is sensible for <=TLSv1.2 but does not work for TLSv1.3.
Fixes#4045
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4068)
Some extensions were being displayed twice, before they were parsed, and
again after they were parsed.
The supported_versions extension was not being fully displayed, as it
was processed differently than other extensions.
Move the debug callback to where the extensions are first collected, to
catch all the extensions as they come in, so they are ordered correctly.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3911)
TLS_ST_SR_NEXT_PROTO means "SSLv3/TLS read next proto"
Fix typo in the message for TLS_ST_SW_CERT_STATUS
Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4054)
Remove the function prototypes for ssl_cert_get0_next_certificate, ssl_set_default_md, tls1_shared_list,
dtls1_send_newsession_ticket, tls1_ctrl, and tls1_callback_ctrl, all of which are not defined.
It also changed the signature of the function pqueue_next to `pitem *pqueue_next(piterator *item)` in
pqueue.c, making it match the prototype in ssl_locl.h. (`piterator *` is equivalent to `pitem **`.)
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4049)
This patch removes the prototype of function RECORD_LAYER_set_write_sequence from record_locl.h, since this function is not defined.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4051)
Documentation and test cases are also updated
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3783)
OpenSSL already has the feature of SSL_MODE_RELEASE_BUFFERS that can
be set to release the read or write buffers when data has finished
reading or writing. OpenSSL will automatically re-allocate the buffers
as needed. This can be quite aggressive in terms of memory allocation.
This provides a manual mechanism. SSL_free_buffers() will free
the data buffers if there's no pending data. SSL_alloc_buffers()
will realloc them; but this function is not strictly necessary, as it's
still done automatically in the state machine.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2240)
Fixes: issue #3747
make SSL_CIPHER_standard_name globally available and introduce a new
function OPENSSL_cipher_name.
A new option '-convert' is also added to 'openssl ciphers' app.
Documentation and test cases are added.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/3859)
We now allow a different protocol version when reusing a session so we can
unconditionally reset the SSL_METHOD if it has changed.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/3954)
SSL_clear() does not reset the SSL_METHOD if a session already exists in
the SSL object. However, TLSv1.3 does not have an externally visible
version fixed method (only an internal one). The state machine assumes
that we are always starting from a version flexible method for TLSv1.3.
The simplest solution is to just fix SSL_clear() to always reset the method
if it is using the internal TLSv1.3 version fixed method.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/3954)
TLSv1.3 draft-21 requires the ticket nonce to be at least 1 byte in length.
However NSS sends a zero length nonce. This is actually ok because the next
draft will allow zero length nonces anyway, so we should tolerate this.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3957)
The functiontls12_get_pkey_idx is only used to see if a certificate index is
enabled: call ssl_cert_is_disabled instead.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3858)
Add certificate table giving properties of each certificate index:
specifically the NID associated with the index and the the auth mask
value for any cipher the certificate can be used with.
This will be used to generalise certificate handling instead of hard coding
algorithm specific cases.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3858)
In most scenarios the length of the input data is the hashsize, or 0 if
the data is NULL. However with the new ticket_nonce changes the length can
be different.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/3852)
This just adds the processing for sending and receiving the newly added
ticket_nonce field. It doesn't actually use it yet.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/3852)
If the result of a SSL_{CTX_,}set_{min,max}_proto_version() call
leaves the min and max version identical, and support for that version
is compiled out of the library, return an error. Such an object has
no hope of successfully completing a handshake, and this error may
be easier to decipher than the resulting handshake failure.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3422)
We need to use the hashsize in generating the exportsecret not 0! Otherwise
we end up with random garbage for the secret.
It was pure chance that this passed the tests previously. It so happens
that, because we call SSL_export_keying_material() repeatedly for different
scenarios in the test, we end up in the tls13_export_keying_material() at
exactly the same position in the stack each time and therefore end up using
the same random garbage secret each time!
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3810)
There are no public TLSv1_3_*method() functions so
OPENSSL_NO_TLS1_3_METHOD doesn't make any sense and should be removed.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3800)
Properly copy ext.alpn_session in ssl_session_dup()
Use OPENSSL_strndup() as that's used in ssl_asn1.c
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/3770)
In draft-ietf-tls-tls13-20 Appendix B we find that:
This section describes protocol types and constants. Values listed
as _RESERVED were used in previous versions of TLS and are listed
here for completeness. TLS 1.3 implementations MUST NOT send them
but might receive them from older TLS implementations.
Similarly, in section 4.2.3 we see:
Legacy algorithms Indicates algorithms which are being deprecated
because they use algorithms with known weaknesses, specifically
SHA-1 which is used in this context with either with RSA using
RSASSA-PKCS1-v1_5 or ECDSA. These values refer solely to
signatures which appear in certificates (see Section 4.4.2.2) and
are not defined for use in signed TLS handshake messages.
Endpoints SHOULD NOT negotiate these algorithms but are permitted
to do so solely for backward compatibility. Clients offering
these values MUST list them as the lowest priority (listed after
all other algorithms in SignatureSchemeList). TLS 1.3 servers
MUST NOT offer a SHA-1 signed certificate unless no valid
certificate chain can be produced without it (see
Section 4.4.2.2).
However, we are currently sending the SHA2-based DSA signature schemes
and many SHA1-based schemes, which is in contradiction with the specification.
Because TLS 1.3 support will appear in OpenSSL 1.1, we are bound by
stability requirements to continue to offer the DSA signature schemes
and the deprecated hash algorithms. at least until OpenSSL 1.2.
However, for pure TLS 1.3 clients that do not offer lower TLS versions,
we can be compliant. Do so, and leave a note to revisit the issue when
we are permitted to break with sacred historical tradition.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3326)
Following on from the previous commit this fixes another instance where
we need to treat a -ve return from EVP_DigestVerify() as a bad signature.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3756)
Prior to 72ceb6a we treated all failures from the call to
EVP_DigestVerifyFinal() as if it were a bad signature, and failures in
EVP_DigestUpdate() as an internal error. After that commit we replaced
this with the one-shot function EVP_DigestVerify() and treated a 0 return
as a bad signature and a negative return as an internal error. However,
some signature errors can be negative (e.g. according to the docs if the
form of the signature is wrong). Therefore we should treat all <=0
returns as a bad signature.
This fixes a boringssl test failure.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3756)
Per RFC 7905, the cipher suite names end in "_SHA256". The original
implementation targeted the -03 draft, but there was a -04 draft right
before the RFC was published to make the names consistent.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3748)
initialize some local variables
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3741)
The value of BIO_CTRL_DGRAM_SET_PEEK_MODE was clashing with the value for
BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE. In an SCTP enabled build
BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE was used unconditionally with
the reasoning that it would be ignored if SCTP wasn't in use. Unfortunately
due to this clash, this wasn't the case. The BIO ended up going into peek
mode and was continually reading the same data over and over - throwing it
away as a replay.
Fixes#3723
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3724)
This does things as per the recommendation in the TLSv1.3 spec. It also
means that the server will always choose its preferred ciphersuite.
Previously the server would only select ciphersuites compatible with the
session.
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/3623)
Also remove nested OPENSSL_NO_EC conditional; it was properly indented,
but a no-op.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Benjamin Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/3693)
Add "*" as indicator meaning the function/reason is removed, so put an
empty string in the function/reason string table; this preserves backward
compatibility by keeping the #define's.
In state files, trailing backslash means text is on the next line.
Add copyright to state files
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3640)
It is an API to be used from the early callback that indicates what
extensions were present in the ClientHello, and in what order.
This can be used to eliminate unneeded calls to SSL_early_get0_ext()
(which itself scales linearly in the number of extensions supported
by the library).
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2976)
We prevent compression both when the server is parsing the ClientHello
and when the client is constructing the ClientHello. A 1.3 ServerHello
has no way to hand us back a compression method, and we already check
that the server does not try to give us back a compression method that
we did not request, so these checks seem sufficient.
Weaken the INSTALL note slightly, as we do now expect to interoperate
with other implementations.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3131)
Signed-off-by: Paul Yang <paulyang.inf@gmail.com>
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3622)
This uses memset() to clear all of the SRP_CTX when free'ing or
initializing it as well as in error paths instead of having a series
of NULL and zero assignments as it is safer.
It also changes SSL_SRP_CTX_init() to reset all the SRP_CTX to zero
in case or error, previously it could retain pointers to freed
memory, potentially leading to a double free.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3467)
Ownership and lifetime rules of SRP_CTX.info are confusing and different
from those of SRP_CTX.login, making it difficult to use correctly.
This makes the ownership and lifetime be the same as those of SRP_CTX.login,
thet is a copy is made when setting it and is freed when SRP_CTX is freed.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3467)
This used to work but was inadvertently removed as part of the TLSv1.3
work. This adds it back.
Fixes#3633
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3639)
Run perltidy on util/mkerr
Change some mkerr flags, write some doc comments
Make generated tables "const" when genearting lib-internal ones.
Add "state" file for mkerr
Renerate error tables and headers
Rationalize declaration of ERR_load_XXX_strings
Fix out-of-tree build
Add -static; sort flags/vars for options.
Also tweak code output
Moved engines/afalg to engines (from master)
Use -static flag
Standard engine #include's of errors
Don't linewrap err string tables unless necessary
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3392)
At the moment we flush the write BIO if we send a fatal alert, but not a
warning one. This can mean the warning is never sent if we never do another
write and subsequently flush the BIO. Instead we should just always flush
after writing an alert.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3432)
Code was added in commit b3c31a65 that overwrote the last ex_data value
using CRYPTO_dup_ex_data() causing a memory leak, and potentially
confusing the ex_data dup() callback.
In ssl_session_dup(), fix error handling (properly reference and up-ref
shared data) and new-up the ex_data before calling CRYPTO_dup_ex_data();
all other structures that dup ex_data have the destination ex_data new'd
before the dup.
Fix up some of the ex_data documentation.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3323)
The check for SSL3_FLAGS_ALLOW_UNSAFE_LEGACY_RENEGOTIATION is
inconsistent. Most places check SSL->options, one place is checking
SSL_CTX->options; fix that.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
GH: #3523
SSLv3 (specifically with client auth) cannot use one shot APIs: the digested
data and the master secret are handled in separate update operations. So
in the special case of SSLv3 use the streaming API.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3527)
The return code from tls1_mac is supposed to be a boolean 0 for fail, 1 for
success. In one place we returned -1 on error. This would cause code calling
the mac function to erroneously see this as a success (because a non-zero
value is being treated as success in all call sites).
Fortunately, AFAICT, the place that returns -1 can only happen on an
internal error so is not under attacker control. Additionally this code only
appears in master. In 1.1.0 the return codes are treated differently.
Therefore there are no security implications.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3495)
We are quite inconsistent about which alerts get sent. Specifically, these
alerts should be used (normally) in the following circumstances:
SSL_AD_DECODE_ERROR = The peer sent a syntactically incorrect message
SSL_AD_ILLEGAL_PARAMETER = The peer sent a message which was syntactically
correct, but a parameter given is invalid for the context
SSL_AD_HANDSHAKE_FAILURE = The peer's messages were syntactically and
semantically correct, but the parameters provided were unacceptable to us
(e.g. because we do not support the requested parameters)
SSL_AD_INTERNAL_ERROR = We messed up (e.g. malloc failure)
The standards themselves aren't always consistent but I think the above
represents the best interpretation.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3480)
add_key_share() is a helper function used during key_share extension
construction. It is expected to be a simple boolean success/fail return.
It shouldn't be using the new EXT_RETURN type but it was partially converted
anyway. This changes it back.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3488)
The current TLSv1.3 spec says:
'If a server is authenticating via a certificate and the client has not
sent a "signature_algorithms" extension, then the server MUST abort the
handshake with a "missing_extension" alert (see Section 8.2).'
If we are resuming then we are not "authenticating via a certificate" but
we were still aborting with the missing_extension alert if sig algs was
missing.
This commit ensures that we only send the alert if we are not resuming.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3478)
We already did this on an ad-hoc per extension basis (for some extensions).
This centralises it and makes sure we do it for all extensions.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3418)
An alert message is 2 bytes long. In theory it is permissible in SSLv3 -
TLSv1.2 to fragment such alerts across multiple records (some of which
could be empty). In practice it make no sense to send an empty alert
record, or to fragment one. TLSv1.3 prohibts this altogether and other
libraries (BoringSSL, NSS) do not support this at all. Supporting it adds
significant complexity to the record layer, and its removal is unlikely
to cause inter-operability issues.
The DTLS code for this never worked anyway and it is not supported at a
protocol level for DTLS. Similarly fragmented DTLS handshake records only
work at a protocol level where at least the handshake message header
exists within the record. DTLS code existed for trying to handle fragmented
handshake records smaller than this size. This code didn't work either so
has also been removed.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3476)
- Mostly missing fall thru comments
- And uninitialized value used in sslapitest.c
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3440)
Add "single part" digest sign and verify functions. These sign and verify
a message in one function. This simplifies some operations and it will later
be used as the API for algorithms which do not support the update/final
mechanism (e.g. PureEdDSA).
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3409)
The function SSL_set_SSL_CTX() can be used to swap the SSL_CTX used for
a connection as part of an SNI callback. One result of this is that the
s->cert structure is replaced. However this structure contains information
about any custom extensions that have been loaded. In particular flags are
set indicating whether a particular extension has been received in the
ClientHello. By replacing the s->cert structure we lose the custom
extension flag values, and it appears as if a client has not sent those
extensions.
SSL_set_SSL_CTX() should copy any flags for custom extensions that appear
in both the old and the new cert structure.
Fixes#2180
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3425)
It is invalid if we receive an HRR but no change will result in
ClientHello2.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3414)
If an HRR gets sent without a key_share (e.g. cookie only) then the code
fails when it should not.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3414)
It is illegal in a TLSv1.3 ClientHello to send anything other than the
NULL compression method. We should send an alert if we find anything else
there. Previously we were ignoring this error.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3410)
When using the -trace option with TLSv1.3 all records appear as "application
data". This adds the ability to see the inner content type too.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3408)
This trace option does not appear in Configure as a separate option and is
undocumented. It can be switched on using "-DOPENSSL_SSL_TRACE_CRYPTO",
however this does not compile in master or in any 1.1.0 released version.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3408)
The TLSv1.3 spec says that a server SHOULD send supported_groups in the
EE message if there is a group that it prefers to the one used in the
key_share. Clients MAY act on that. At the moment we don't do anything
with it on the client side, but that may change in the future.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3395)
SSL_CTX_use_serverinfo_ex() et al were always processing data as if it was
V2 format, even if it was V1. This bug was masked because, although we had
a test which loaded V1 serverinfo data from a file, the function
SSL_CTX_use_serverinfo_file() transparently converts V1 data to V2 before
calling SSL_CTX_use_serverinfo_ex().
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3382)
|version| "could" be used uninitialized here, not really, but the
compiler doesn't understand the flow
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3373)
Ensure that serverinfo only gets added for the first Certificate in a list.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3298)
We already did this for ServerHello and EncryptedExtensions. We should be
doing it for Certificate and HelloRetryRequest as well.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3298)
This enables us to know what messages the extensions are relevant for in
TLSv1.3. The new file format is not compatible with the previous one so
we call it SERVERINFOV2.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3298)
Add padding callback for application control
Standard block_size callback
Documentation and tests included
Configuration file/s_client/s_srver option
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3130)
This fixes a segfault if a NULL parse_cb is passed to
SSL_CTX_add_{client,server}_custom_ext, which was supported in the
pre-1.1.1 implementation.
This behaviour is consistent with the other custom_ext_*_old_cb_wrap
functions, and with the new SSL_CTX_add_custom_ext function.
CLA: trivial
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3310)
Previously, init and finalization function for extensions are called
per extension block, rather than per message. This commit changes
that behaviour, and now they are called per message. The parse
function is still called per extension block.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3244)
Because NST messages arrive post-handshake, the session may have already
gone into the cache. Once in the cache a session must be immutable -
otherwise you could get multi-thread issues.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3008)
Provide a way to test whether the SSL_SESSION object can be used to resume a
sesion or not.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3008)
TLSv1.3 will do the same thing as TLSv1.2 with tickets with regards to session
ids, i.e. it will create a synthetic session id when the session is established,
so it is reasonable to check the session id length, even in TLSv1.3.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3008)
Ensure that there are ciphersuites enabled for the maximum supported
version we will accept in a ClientHello.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3316)
Ensure that there are ciphersuites enabled for the maximum supported
version we are claiming in the ClientHello.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3316)
The function tls_early_post_process_client_hello() was overwriting the
passed "al" parameter even if it was successful. The caller of that
function, tls_post_process_client_hello(), sets "al" to a sensible default
(HANDSHAKE_FAILURE), but this was being overwritten to be INTERNAL_ERROR.
The result is a "no shared cipher" error (and probably other similar errors)
were being reported back to the client with an incorrect INTERNAL_ERROR
alert.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3314)
We were allocating the write buffer based on the size of max_send_fragment,
but ignoring it when writing data. We should fragment handshake messages
if they exceed max_send_fragment and reject application data writes that
are too large.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3286)
There was code existing which attempted to handle the case where application
data is received after a reneg handshake has started in SCTP. In normal DTLS
we just fail the connection if this occurs, so there doesn't seem any reason
to try and work around it for SCTP. In practice it didn't work properly
anyway and is probably a bad idea to start with.
Fixes#3251
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3286)
ECDHE is not properly defined for SSLv3. Commit fe55c4a2 prevented ECDHE
from being selected in that protocol. However, historically, servers do
still select ECDHE anyway so that commit causes interoperability problems.
Clients that previously worked when talking to an SSLv3 server could now
fail.
This commit introduces an exception which enables a client to continue in
SSLv3 if the server selected ECDHE.
[extended tests]
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3204)
doing the pms assignment after log is successful
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3278)
X509_STORE_add_cert and X509_STORE_add_crl are changed to return
success if the object to be added was already found in the store, rather
than returning an error.
Raise errors if empty or malformed files are read when loading certificates
and CRLs.
Remove NULL checks and allow a segv to occur.
Add error handing for all calls to X509_STORE_add_c{ert|tl}
Refactor these two routines into one.
Bring the unit test for duplicate certificates up to date using the test
framework.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2830)
This resulted in the SCT timestamp check always failing, because the
timestamp appeared to be in the future.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3138)
SSLv3 does not support TLS extensions, and thus, cannot provide any
curves for ECDH(E). With the removal of the default (all) list of curves
being used for connections that didn't provide any curves, ECDHE is no
longer possible.
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3181)
RFC 7301 mandates that the server SHALL respond with a fatal
"no_application_protocol" alert when there is no overlap between
the client's supplied list and the server's list of supported protocols.
In commit 062178678f we changed from
ignoring non-success returns from the supplied alpn_select_cb() to
treating such non-success returns as indicative of non-overlap and
sending the fatal alert.
In effect, this is using the presence of an alpn_select_cb() as a proxy
to attempt to determine whether the application has configured a list
of supported protocols. However, there may be cases in which an
application's architecture leads it to supply an alpn_select_cb() but
have that callback be configured to take no action on connections that
do not have ALPN configured; returning SSL_TLSEXT_ERR_NOACK from
the callback would be the natural way to do so. Unfortunately, the
aforementioned behavior change also treated SSL_TLSEXT_ERR_NOACK as
indicative of no overlap and terminated the connection; this change
supplies special handling for SSL_TLSEXT_ERR_NOACK returns from the
callback. In effect, it provides a way for a callback to obtain the
behavior that would have occurred if no callback was registered at
all, which was not possible prior to this change.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2570)
The old custom extensions API was not TLSv1.3 aware. Extensions are used
extensively in TLSv1.3 and they can appear in many different types of
messages. Therefore we need a new API to be able to cope with that.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3139)
This move prepares for the later addition of the new custom extensions
API. The context codes have an additional "SSL_" added to their name to
ensure we don't have name clashes with other applications.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3139)
This increases portability of SSL_SESSION files between architectures
where the size of |long| may vary. Before this, SSL_SESSION files
produced on a 64-bit long architecture may break on a 32-bit long
architecture.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3088)
Add functions to add/retrieve the certificate_authorities. The older
client_CA functions mainly just call the new versions now.
Rename fields sice new extension can be generated by client and server.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3015)
The macro SSL_get_server_tmp_key() returns information about the temp key
used by the server during a handshake. This was returning NULL for TLSv1.3
and causing s_client to omit this information in its connection summary.
Fixes#3081
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3114)
If we have received the EoED message but not yet had the CF then we are
"in init". Despite that we still want to write application data, so suppress
the "in init" check in ssl3_write_bytes() in that scenario.
Fixes#3041
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3091)
SSL_get_max_early_data() recently added by 3fc8d85610 ("Construct the
ticket_early_data_info extension", 2017-02-17) is supposed to take an
SSL, but it doesn't.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3113)
Numerous changes have been made to the supported built-in extensions and
SSL_extension_supported() has not kept up.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3097)
If the server received EoED then SSL_read_early_data() will return
SSL_READ_EARLY_DATA_FINISH. However if the CF has not yet been processed
then SSL_is_init_finished() will still return 0. Therefore we should still
be able to write early data.
Fixes#3041
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3089)
If read_ahead is set, or SSL_MODE_AUTO_RETRY is used then if
SSL_read_early_data() hits an EndOfEarlyData message then it will
immediately retry automatically, but this time read normal data instead
of early data!
Fixes#3041
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3077)
Variable 'pktype' was set but not used under OPENSSL_NO_GOST. This change
will fix the build warning under [-Werror=unused-but-set-variable].
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2961)
A similar change that probably should have been wrapped into
commit e0926ef49d.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3010)
Fix some comments too
[skip ci]
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3069)
Commit 6b1bb98fa moved the processing of ClientHello extensions into the
state machine post-processing stage. After processing s->init_num is reset
to 0, so by post-processing we cannot rely on its value. Unfortunately we
were using it to handle the PSK extension. This causes the handshake to
fail.
We were using init_num to figure out the length of ClientHello2 so we can
remove it from the handshake_buffer. The handshake_buffer holds the
transcript of all the messages sent so far. For PSK processing though we
only want to add in a partial ClientHello2. This commit changes things so
we just work out where ClientHello2 starts, working forward from the
beginning of handshake_buffer.
Fixes#2983
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2996)
This label for this derivation was incorrectly "derived" or "der" depending
on the pointer size of the build(!). The correct string is "derived secret".
(cherry picked from commit 936dcf272033c1bf59a5e859ec63e2557194f191)
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2989)
In OpenSSL 1.1.0 the padding extension MUST be last because it calculates
the length of everything that has been written into the ClientHello to
determine whether it needs to be padded or not. With TLSv1.3 that isn't
possible because the specification requires that the PSK extension is last.
Therefore we need to fix the padding extension to take account of any PSK
extension that will be later added.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2968)
Choose a new ciphersuite for the HRR. Don't just use the one from the
session.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2895)
Don't include a PSK that does not have the right hash for the selected
ciphersuite following an HRR.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2895)
Draft-19 changes the HRR transcript hash so that the initial ClientHello
is replaced in the transcript with a special synthetic message_hash message
that just contains a hash of ClientHello1 as its message body.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2895)
The end of early data is now indicated by a new handshake message rather
than an alert.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2895)
tls1_get_curvelist() does not read from its third parameter, so
the assignments prior to function call were dead code and can be removed.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2952)
Instead of making a positive comparison against the invalid value
that our server would send, make a negative check against the only
value that is not an error.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2953)
Zero out the length alongside the NULLing of the pointer, to
bring parity between the selected and proposed fields..
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2954)
No need to break out of the loop and repeat the loop termination
condition when we can just return.
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2949)
Found using various (old-ish) versions of gcc.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2940)
... in functions dealing with the SSL object rather than the context.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2870)
that refers to space
deallocated by a call to the free function in tls_decrypt_ticket.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2897)
(cherry picked from commit 13ed1afa92)
The value of SSL3_RT_MAX_ENCRYPTED_LENGTH normally includes the compression
overhead (even if no compression is negotiated for a connection). Except in
a build where no-comp is used the value of SSL3_RT_MAX_ENCRYPTED_LENGTH does
not include the compression overhead.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2872)
In TLSv1.3 the above messages signal a key change. The spec requires that
the end of these messages must align with a record boundary. We can detect
this by checking for decrypted but as yet unread record data sitting in
OpenSSL buffers at the point where we process the messages.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2875)
Also updates SSL_has_pending() to use it. This actually fixes a bug in
SSL_has_pending() which is supposed to return 1 if we have any processed
or unprocessed data sitting in OpenSSL buffers. However it failed to return
1 if we had processed non-application data pending.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2875)
Also, restore 1.0.2 behavior of looping over all BIO's in the chain.
Thanks to Joseph Bester for finding this and suggesting a fix to the
crash.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2651)
If early data is sent to a server, but ALPN is not used then memcmp is
called with a NULL pointer which is undefined behaviour.
Fixes#2841
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2845)
We do not allow the generation of TLSv1.3 cookies. But if we receive one
in an HRR we will echo it back in the ClientHello.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2839)
A leak of an SSL_SESSION object can occur when decoding a psk extension on
an error path when using TLSv1.3
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2843)
Change tls12_sigalg_allowed() so it is passed a SIGALG_LOOKUP parameter,
this avoids multiple lookups.
When we copy signature algorithms return an error if no valid TLS message
signing algorithm is present. For TLS 1.3 this means we need at least one
signature algorithm other than RSA PKCS#1 or SHA1 both of which can only be
used to sign certificates and not TLS messages.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2840)
Change the early data API so that the server must use
SSL_write_early_data() to write to an unauthenticated client.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2737)
This is for consistency with the rest of the API where all the functions
are called *early_data*.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2737)
This is for consistency with the rest of the API where all the functions
are called *early_data*.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/2737)
