| Age | Commit message (Collapse) | Author |
|---|
|
TLS 1.2 introduces changes on how signatures
are calculate and encoded. This makes the
signature handling version aware
|
|
|
|
TLS 1.2 allows to negotiate the used PRF,
additional the default PRF uses a different
hash. This change make the PRF selectable
and hardwires the PRF for TLS < 1.2
|
|
dh parameter verification is done differently with TLS 1.2.
Prepare for that by passing the verion to verify_dh_params.
|
|
TLS 1.2 changes the layout of several handshake
records. This adds the TLS version to dec_hs/2
so it can decode those.
|
|
TLS 1.2 changed the way digital signatures are
done. key_exchange/3 needs to pass the version
to it.
|
|
TLS/SSL version before 1.2 always used a MD5/SHA combination
for the handshake hashes. With TLS 1.2 the default hash is
SHA256 and it is possible to negotiate a different hash.
This change delays the calculation of the handshake
hashes until they are really needed. At that point the hash
to use should be known.
For now MD5/SHA is still hard coded.
|
|
ssl sockets.
|
|
Do proc_lib:spawn_link instead of proc_lib:start_link as synchronized
init is not used/needed anyway.
|
|
Avoid cach validation with file:file_info/2 as this i too expensive and
causes a bottleneck in the file server. Instead we expose a new API function
ssl:clear_pem_cache/0 to deal with the problem. As we think it will be
of occasional use and the normal case is that the cache will be valid we think
it is the right thing to do.
Convert file paths to binary representation in the ssl API module to
avoid uncessarry calls in file later on.
Also add sanity checks for openssl versions in testsuite due to new
openssl bugs.
|
|
|
|
Instance of state variable that are "updated" in a function is called
for example State0 and the last instance, that should be returned, is
called State possible intermidiat versions are suffixed by increasing
numbers. State0 may be rturned in error cases.
Avoid nesting case statments.
|
|
Aviods storing a lot of data
|
|
The session id keept in the connection processes state must be updated to be
the id selected by ssl_handshake:client_hello, failing to do so will
cause a crash if the session is not reused.
|
|
Do not use ssl_manager process for selecting an id. It's unnecessary
to involve the manager process at all on the client side.
|
|
This is done by using proc_lib and gen_fsm:enter_loop
so that supervisor will not have to wait for the relative long initialization
of an ssl connection process before starting another connection process.
|
|
|
|
transport layer need to generate additional application specific
key material. One way to generate such material is to use the TLS
PRF and key material from the TLS session itself.
This change makes it possible to use a TLS sessions PRF either with
the session internal or caller supplied key material to generate
additional key material.
|
|
|
|
evaluating ssl:recv could be left hanging for ever.
|
|
|
|
The code is refactored and improved to make it easier to insert the
1/n-1 splitting countermeasure Rizzo/Duong-Beast that is really done
in one function clause in ssl:record_split_bin/3
|
|
|
|
Added session status "new" to mark sessions that are
in the session database to reserve the session id
but not resumable yet and that we want to separate from
sessions that has been invalidated for further reuse.
|
|
The time_stamp filed is now initated in the connection process
init function, so that invalidations of sessions due to handshake failiures,
will not cause sessions in the session table to have
an uninitiated time_stamp field.
|
|
|
|
* ia/ssl/remove-old-ssl/OTP-7048:
Remove old ssl implementation and deprecated function ssl:peercert/1
Conflicts:
lib/ssl/test/Makefile
|
|
|
|
* upstream/dev:
Support explicit use of packet option httph and httph_bin
|
|
|
|
* ia/ssl/dist/OTP-7053:
First fully working version
Use ssl instead of being a proxy command
Connect from both sides works now
|
|
|
|
|
|
|
|
|
|
We want the certificate table to be handled the same way as the
session table and not have a global name, so that we may easier
create a separate ssl-manager to handle erlang distribution over ssl.
|
|
|
|
that a session not be resumed. This is a change from TLS 1.0 to
conform with widespread implementation practice. Erlang ssl will now
in TLS 1.0 conform to the widespread implementation practice instead
of the specification to avoid performance issues.
|
|
ssl:send/2 takes iodata() as a second
argument. erlang:iolist_to_binary should really be called
erlang:iodata_to_binary which caused the mismatch in the first place.
|
|
When making an SSL connection (either as client or as server), the
process implementing the connection may use as much as hundreds of
kilobytes of memory, even when idle. This is problematic for any
application maintaining more than just a few SSL connections.
This patch introduces the option { hibernate_after, int() } to the
ssl:connect and ssl:listen functions, making the SSL connection
process go into hibernation after the specified number of milliseconds
of inactivity. This will reduce the memory used by the process to
just a few hundred bytes, making applications with thousands or
more SSL connections feasible, as long as most of the connections
are idle for most of the time (which is typically the case).
The approach of making the process go into hibernation only after
some time of inactivity was chosen because hibernation incurs some
CPU usage, and it is therefore not desirable for a process to
hibernate after each call.
|
|
|
|
Handshake hashes, premaster secret and "public_key_info" does not need
to be saved when the connection has been established. The own certificate is
no longer duplicated in the state.
|
|
-------------------------------------------------------------------------
ssl's published documentation states:
ssl_accept(ListenSocket) ->
ssl_accept(ListenSocket, Timeout) -> ok | {error, Reason}
(see http://www.erlang.org/doc/man/ssl.html#ssl_accept-1)
while its code has the specs:
-spec ssl_accept(#sslsocket{}) -> {ok, #sslsocket{}} | {error, reason()}.
-spec ssl_accept(#sslsocket{}, list() | timeout()) -> {ok, #sslsocket{}} | {error, reason()}.
One of the two cannot be right. This should be fixed.
Moreover, I do not see why the spec just mentions list() for the options when the documentation explicitly mentions the options of ssl.
Kostis
---------------------------------------------------------------------
|
|
Added cache invalidation control of ssl certificates so that
sessions will not be reused if file content is changed.
There was a glitch in ssl:close that made it possible to
to get eaddrinuse even though reuseadder-option was used.
Also improved tests for better user-close handling.
|
|
|
|
Data to sign and verify should be inputed as binaries.
Also cleaned up and moved some dialyzer specs.
|
|
Also changed alert to BAD_RECORD_MAC as:
"differentiating between bad_record_mac and decryption_failed alerts
may permit certain attacks against CBC mode as used in TLS
[CBCATT]. It is preferable to uniformly use the bad_record_mac
alert to hide the specific type of the error."
Also cleaned up the code and changed a few other alert reasons in
according to alert descriptions in the TLS RFC 4346. And added function
terminate_alert/3 so that we can differentiate between a crash
in ssl (a bug in our code) and a crash in the application using ssl.
|
|
* ia/ssl/certificate-verify/wrong-key-method/OTP-8897:
Correct handling of client certificate verify message
Conflicts:
lib/ssl/src/ssl_handshake.erl
|
|
When checking the client certificate verify message the server used
the wrong algorithm identifier to determine the signing algorithm,
causing a function clause error in the public_key application when the
key-exchange algorithm and the public key algorithm of the client
certificate happen to differ.
|
|
For testing purposes ssl now also support some anonymous cipher suites
when explicitly configured to do so.
Also moved session cache tests to its own suite, so that timeout
of end_per_testcase when the mnesia is used as session cache will
not affect other test cases.
|
