Age | Commit message (Collapse) | Author |
|
|
|
maint-21
* ingela/ssl/recv-timeout-bug/ERL-884/ERL-883/OTP-14701:
ssl: Cancel recv timer in all places
|
|
Would cause connection to crash although all data was delivered to ssl:recv
|
|
|
|
|
|
Improve the abstraction between the ssl_connection module
and dtls_connection, tls_connection and tls_sender, as well
as towards the lower level tls_record and ssl_record modules.
Remove some dead code.
|
|
To avoid degenerate case with quadratic complexity that
shows up when sending large messages since the the fragment
concatenation was done by binary append. An Okasaki queue
is much more efficient.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Remove duplicate handsake history handling form
ssl_config function. Later we might refactor to avoid
duplication in TLS and DTLS code, but handshake history
does not belong in ssl_config function.
|
|
|
|
|
|
|
|
Could cause connection processes not terminate when they should
|
|
* ingela/ssl/continue-optimize/OTP-15445:
ssl: If possible assemble several received application data records
|
|
|
|
We want to decrease the size of the outer state tuple, and gain
ease of understanding by better grouping. This is the first step
of creating a hs_env (handshake environment) part of the state.
This change will be performed gradually to reduce merge conflicts
complexity and risk of introducing errors.
|
|
New internal active N changed timing, and
new check is needed.
|
|
Wtite connection state was not synchronized when peer initiated renegotiation
|
|
|
|
As the stop wrapper functions are no longer needed after tls_sender
that altered the behaviour of the TLS distribution code.
|
|
Both test case and code needed updates to work as intended. Code needed update due to
new tls_sender process and the test case gave false positive reusult erarlier probably
due to beeing to sloopy in order to avoid timeouts.
|
|
|
|
|
|
Rename Connection:handle_common_event Connection:handle_protocol_record
removing use of unnecessary argument and making code easier to understand.
|
|
State values created at init
|
|
* raimo/ssl/tls_dist-optimization:
Handle socket close in state downgrade
Handle dead sender at terminate
Handle tls_sender exit properly
Optimize split_bin
Improve dist send throughput
|
|
Also avoid code duplication
Conflicts:
lib/ssl/src/dtls_connection.erl
lib/ssl/src/tls_connection.erl
|
|
Make next_record an internal help function to next_event and avoid
duplicate calls to tls_socket:setopts for setting the active option.
|
|
|
|
|
|
* ingela/ssl/deliver-all-data-at-close/ERL-731/OTP-15412:
ssl: Extend check for undelivered data at closing
|
|
This is a timing related bug that alas is hard to test
|
|
With the new TLS sender process, solving ERL-622, TLS ALERTs sent in
the connection state must be encrypted and sent by the TLS sender
process. This to make sure that the correct encryption state is used
to encode the ALERTS. Care must also be taken to ensure a graceful
close down behavior both for normal shutdown and downgrading from TLS
to TCP.
The original TR ERL-738 is verified by cowboy tests, and close down
behavior by our tests. However we alas have not been able to yet
create a minimal test case for the originating problem.
Also it seems it has become less likely that we run in to the TCP
delivery problem, that is the guarantee is only on transport level,
not application level. Keep work around function in ssl_test_lib but
we can have better test as long as we do not get to much wobbling
tests.
|
|
The link should be between the connection process and the tls_sender
process. But the start of the tls_sender process needs to be done
by the process that also starts the connection process in order to
correctly create the opaque #ssl_socket{}.
|
|
We want to make sure that the sender process that may get stuck in
prim_inet:send will die if the tls_connection process is
terminated. And we also like to make sure that it terminates as
gracefully as possible. So when the tls_connection process dies it
spawns a killer process that will brutaly kill the sender if it is
unresponsive and does not terminate due to its monitor of the
tls_connetion process triggering.
When the sender process also acts as distribution controller it
may also have other processess that it is linked with that it
should bring down or that could bring the connection down.
|
|
|
|
Separate sending and receiving when using TCP as transport
as prim_inet:send may block which in turn may result
in a deadlock between two Erlang processes communicating over
TLS, this is especially likely to happen when running Erlang distribution
over TLS.
|
|
When doing ssl:controlling_process on a ssl socket that has not
performed the TLS/DTLS handshake that call will succeed even though
the documentation stated otherwise. However if some other ssl option
was incorrect the call would hang. Now {error, closed} will be
returned in the latter case, which is logical independent on if it
should succeed or not in the former case. The former case will continue
to succeed, as it is not dependent of the TLS/DTLS connection being
established, and the documentation is altered slightly to not
explicitly disallow it. If the TLS/DTLS connection later fails and
the socket mode is active, the new controlling process will be
notified as expected.
|
|
|
|
|
|
|
|
|