Olivier Crête [Thu, 29 Jan 2015 02:46:33 +0000 (21:46 -0500)]
Version 0.1.9.1
Olivier Crête [Thu, 29 Jan 2015 02:44:04 +0000 (21:44 -0500)]
Version 0.1.9
Youness Alaoui [Tue, 20 Jan 2015 17:31:11 +0000 (12:31 -0500)]
Only change the receiving nicesock for udp-turn for ice-tcp
The code was in the wrong spot, for OC2007, we need to fix the
nicesock used only in the case of udp-turn-over-tcp for OC2007
compatibility mode. With the current code, it was thinking that
the data came from TURN for every udp packet.
This should hopefully fix : https://github.com/EricssonResearch/openwebrtc/issues/85
Philip Withnall [Fri, 2 Jan 2015 19:25:32 +0000 (19:25 +0000)]
docs: Enable documentation checks during `make check` by default
Now that they pass, we should endeavour to ensure they continue to pass.
Philip Withnall [Fri, 2 Jan 2015 19:25:07 +0000 (19:25 +0000)]
docs: Add 0.1.8 API to the documentation index
This should have been included before. This fixes `make check` in docs/.
Philip Withnall [Fri, 2 Jan 2015 19:24:31 +0000 (19:24 +0000)]
docs: Add various missing documentation comments and update sections.txt
This straightens out the last few bits of the documentation and almost
allows `make check` to pass in docs/.
Philip Withnall [Fri, 2 Jan 2015 19:23:39 +0000 (19:23 +0000)]
docs: Inform gtk-doc of the libnice namespace
This allows it to order the index a little better.
Philip Withnall [Fri, 2 Jan 2015 19:22:23 +0000 (19:22 +0000)]
docs: Use gtk-doc no-tmpl mode
Templated mode has been deprecated for a long time and will be removed
in future versions of gtk-doc. Switch to no-tmpl mode for faster
documentation builds and future-proofness.
This does not bump our gtk-doc dependency, since gtk-doc 1.9 supported
no-tmpl mode.
Philip Withnall [Fri, 2 Jan 2015 19:21:48 +0000 (19:21 +0000)]
docs: Switch from gtk-doc SGML mode to XML mode
We already used XML everywhere; just formally switch to XML mode and
eliminate the last few traces of SGML.
Philip Withnall [Sat, 27 Dec 2014 09:49:37 +0000 (09:49 +0000)]
agent: Document G_MESSAGES_DEBUG and NICE_DEBUG
As requested on the mailing list, expand the documentation for
nice_debug_enable() to mention them.
Philip Withnall [Mon, 15 Dec 2014 23:35:13 +0000 (23:35 +0000)]
agent: Handle EWOULDBLOCK when transmitting pseudo-TCP segments
The pseudo-TCP code previously didn’t handle EAGAIN or EWOULDBLOCK
errors from the low-level NiceSocket code. This caused pseudo-TCP
connections to be dropped if the transmitting socket ever filled up,
which could cause problems on high bandwidth connections.
Fix the issue by effectively dropping the packet on EWOULDBLOCK. This
will eventually get picked up by the pseudo-TCP recovery mechanism,
retransmitting the packet and throttling the sender. This should
hopefully reduce the system resource usage which caused EWOULDBLOCK in
the first place.
Spotted and debugged by Radosław Kołodziejczyk
<radek.kolodziejczyk@gmail.com>.
https://bugs.freedesktop.org/show_bug.cgi?id=87344
Olivier Crête [Mon, 8 Dec 2014 15:22:21 +0000 (10:22 -0500)]
pseudotcp: Do roll-over safe assertion
In case of a roll-over, only the difference matters.
Philip Withnall [Fri, 14 Nov 2014 10:06:21 +0000 (10:06 +0000)]
agent: Add debug output for lifetime of Components and Streams
Philip Withnall [Thu, 6 Nov 2014 10:41:57 +0000 (10:41 +0000)]
tests: Use G_SOURCE_[CONTINUE|REMOVE] instead of booleans
Makes the test code a little clearer. Introduces no functional changes.
Philip Withnall [Thu, 6 Nov 2014 10:40:53 +0000 (10:40 +0000)]
tests: Fix a memory leak in test-send-recv
If n_messages ≠ n_valid_messages.
Nicolas Dufresne [Thu, 30 Oct 2014 21:09:41 +0000 (17:09 -0400)]
Add ability to build static GST plugins
Note that this only works for 1.0 plugin.
https://bugs.freedesktop.org/show_bug.cgi?id=85929
Philip Withnall [Thu, 30 Oct 2014 17:20:43 +0000 (17:20 +0000)]
agent: Expand simple example in documentation to mention GMainContext
wrt the ML thread:
http://lists.freedesktop.org/archives/nice/2014-October/000981.html
Philip Withnall [Thu, 30 Oct 2014 09:10:51 +0000 (09:10 +0000)]
agent: Close pseudo-TCP socket earlier in component_close()
This tries to mitigate the race condition between finishing the TCP FIN
handshake and closing the underlying sockets, but it’s impossible to
mitigate properly without API changes. See the comment.
Curieux Tres [Mon, 27 Oct 2014 11:32:17 +0000 (11:32 +0000)]
stun: Rename a symbol to avoid a naming clash with OpenSSL
The symbol is not exported, but nevertheless leaks in the static version
of libnice, due to limitations on symbol visibilities with static
libraries. OpenSSL has a symbol named RAND_bytes(), which ours clashes
with if the two are linked together statically.
Avoid this by prefixing ours with ‘nice_’.
Youness Alaoui [Thu, 18 Sep 2014 23:42:53 +0000 (19:42 -0400)]
Fix documentation relating to how remote crendentials should be set for new ICE format
Youness Alaoui [Thu, 18 Sep 2014 23:30:08 +0000 (19:30 -0400)]
agent: ignore externally set peer-reflexive candidates
All discovered peer-reflexive candidates should be added internally in
conncheck.c and should have the sockptr set. We ignore any prflx
candidates added by an external application because their NULL sockptr
could cause a crash in conn_check_send().
Olivier Crête [Thu, 9 Oct 2014 21:37:49 +0000 (17:37 -0400)]
Version 0.1.8.1
Olivier Crête [Thu, 9 Oct 2014 20:43:16 +0000 (16:43 -0400)]
Version 0.1.8
Olivier Crête [Thu, 9 Oct 2014 21:05:32 +0000 (17:05 -0400)]
agent: Remove extra post condition
It prevent the GError from happening
Olivier Crête [Thu, 9 Oct 2014 21:02:56 +0000 (17:02 -0400)]
udp-turn: Don't dereference priv before check that it's valid
Olivier Crête [Thu, 9 Oct 2014 20:21:18 +0000 (16:21 -0400)]
agent: Pass all packets that are not acceptable STUN packets to the application
We attempted to not pass through some packets that looked like
valid STUN but were ot acceptable to us, but that dropped some application
packets.
Olivier Crête [Wed, 8 Oct 2014 23:36:01 +0000 (19:36 -0400)]
agent: Re-emit gathering done after a new relay has been set
If the stream was ever gathering and a new relay has been set,
then re-emit the candidate=gathering-done signal
Olivier Crête [Sun, 5 Oct 2014 18:06:58 +0000 (14:06 -0400)]
pseudotcp: Print unsigned int with %u not %d
Olivier Crête [Sun, 5 Oct 2014 17:34:22 +0000 (13:34 -0400)]
agent: Add function to retrieve the current component state
Olivier Crête [Sun, 5 Oct 2014 17:35:12 +0000 (13:35 -0400)]
agent: Make the stun-pacing-timer app configurable
There is no reason to only allow modifying it in the new().
Olivier Crête [Sun, 5 Oct 2014 17:33:00 +0000 (13:33 -0400)]
pseudotcp: Set default RTO to 1 second
The newer RFC 6298 recommends 1 second instead of 3.
Olivier Crête [Sun, 5 Oct 2014 17:32:24 +0000 (13:32 -0400)]
conncheck: Computer STUN retransmission timeout (RTO) dynamically
This is how it is specified in RFC 5245 section 16
Olivier Crête [Fri, 3 Oct 2014 21:44:59 +0000 (17:44 -0400)]
discovery: Remove extra \n in g_debug()
Olivier Crête [Fri, 3 Oct 2014 21:10:45 +0000 (17:10 -0400)]
agent: Only add pairs once on gathering done
Also don't try to re-add pairs that already have been added
Olivier Crête [Fri, 3 Oct 2014 20:30:16 +0000 (16:30 -0400)]
conncheck: Insert the candidates sorted after setting the priority
Olivier Crête [Thu, 2 Oct 2014 22:41:14 +0000 (18:41 -0400)]
conncheck: Insert peer reflexive pair sorted
The rest of the list is sorted, but this was just appended, making it likely
to be dropped if the list grows too long.
Olivier Crête [Thu, 2 Oct 2014 22:37:34 +0000 (18:37 -0400)]
conncheck: Don't remove elements in the conncheck list while iterating it
priv_limit_conn_check_list_size() would remove elemtns from the conncheck_list
while the calling functions were iterating it. Now instead just mark them as
cancelled. Then later, at the outer function, free all cancelled elements to
prevent the list from growing out of bounds.
Olivier Crête [Wed, 1 Oct 2014 04:00:30 +0000 (00:00 -0400)]
agent: Avoid leaking UPnP mappings between streams
Clean up the UPnP mappings of each stream when closing it
Olivier Crête [Wed, 1 Oct 2014 03:59:59 +0000 (23:59 -0400)]
agent: Avoid restarting the GUPnP client on every gather
This would cause mappings to be dropped on every new gather, which is bad!
Instead, keep the same one with the mappings, and just drop the timer to ignore
new discovered mappings afterwards.
Olivier Crête [Wed, 1 Oct 2014 02:49:16 +0000 (22:49 -0400)]
agent: Ignore UPnP mappings after the UPnP discovery has finished
Olivier Crête [Tue, 30 Sep 2014 22:28:24 +0000 (18:28 -0400)]
component: On ICE restart reset selected pair priority to 0
This is to ensure that the result of the new negotiation will always
have priority.
Philip Withnall [Tue, 23 Sep 2014 13:54:59 +0000 (14:54 +0100)]
agent: Add some preconditions to public API
To try and avoid NULL NiceCandidates entering our internal state.
Philip Withnall [Fri, 5 Sep 2014 09:54:11 +0000 (10:54 +0100)]
build: Update .gitignore
Olivier Crête [Mon, 11 Aug 2014 18:27:40 +0000 (14:27 -0400)]
outputstream: Fix double unref
Olivier Crête [Sat, 9 Aug 2014 20:18:33 +0000 (16:18 -0400)]
agent: Remove socket source on HUP
Olivier Crête [Thu, 31 Jul 2014 19:16:38 +0000 (15:16 -0400)]
agent: Declare the stream to be CONNECTED only if a pair is selected
Don't change the state if no pair is selected. Otherwise we get
a component that claims to be CONNECTED but has no selected pair.
Also, set the selected pair before announcing the state change.
Olivier Crête [Wed, 8 Oct 2014 23:03:12 +0000 (19:03 -0400)]
nice.pc: Include ${includedir} directly
It doesn't make sense to tell applications to include <agent.h>,
including <nice/agent.h> is much safer. So tell the applications
to use the includedir directly.
Philip Withnall [Fri, 26 Sep 2014 15:11:59 +0000 (16:11 +0100)]
agent: Document interaction between closing and removing streams
GIOStream and nice_agent_remove_stream() interact slightly subtly when
closing down the stream.
See: https://bugzilla.gnome.org/show_bug.cgi?id=735754
Philip Withnall [Fri, 26 Sep 2014 11:57:18 +0000 (12:57 +0100)]
tests: Add more helgrind suppressions
Philip Withnall [Fri, 26 Sep 2014 11:56:52 +0000 (12:56 +0100)]
agent: Delay some assignments until the agent is locked
This makes helgrind slightly happier.
Philip Withnall [Thu, 25 Sep 2014 12:55:42 +0000 (13:55 +0100)]
agent: Warn if a Component is freed without first being closed
Philip Withnall [Thu, 25 Sep 2014 12:54:41 +0000 (13:54 +0100)]
agent: Close Streams before freeing them
This could result in leaking the stream’s un-closed state otherwise, in
any case where a NiceAgent is finalised without
nice_agent_remove_stream() being called on all its streams.
Philip Withnall [Thu, 25 Sep 2014 09:33:12 +0000 (10:33 +0100)]
tests: Add a Valgrind suppression file
To be used with `make check-valgrind`.
Philip Withnall [Thu, 25 Sep 2014 08:24:04 +0000 (09:24 +0100)]
tests: Fix various memory leaks in the unit tests
The theory being that if we can get check-valgrind to pass, we could
acutally use it to find more memory leaks and other errors in the agent.
Philip Withnall [Thu, 25 Sep 2014 08:22:26 +0000 (09:22 +0100)]
agent: Check STUN buffer is non-NULL before getting its message ID
This should fix a crash, as detected by Valgrind:
==28354== Invalid read of size 2
==28354== at 0x4C2B5B0: memcpy@@GLIBC_2.14 (in
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==28354== by 0x50C17E2: stun_message_id (stunmessage.c:658)
==28354== by 0x509E4E7: candidate_check_pair_fail (conncheck.c:254)
==28354== by 0x50A4EDB: conn_check_prune_socket (conncheck.c:3145)
==28354== by 0x509B6F8: component_io_cb (agent.c:3951)
Philip Withnall [Thu, 25 Sep 2014 08:21:44 +0000 (09:21 +0100)]
agent: Fix a leak of a GCancellable and its GSource
The GSource holds a reference to the GCancellable, so needs to be
explicitly removed from the GMainContext when it’s finised with.
Philip Withnall [Thu, 25 Sep 2014 08:21:18 +0000 (09:21 +0100)]
agent: Fix a minor leak in an error handling path
Philip Withnall [Wed, 24 Sep 2014 13:42:19 +0000 (14:42 +0100)]
tests: Move a closure from the heap to the stack
There is no need for this to be heap-allocated.
Olivier Crête [Wed, 24 Sep 2014 02:06:52 +0000 (22:06 -0400)]
conncheck: Don't use CandidateRefresh after freeing it
Philip Withnall [Tue, 23 Sep 2014 14:20:31 +0000 (15:20 +0100)]
socket: Fix a leak on the slow TURN packet handling path
Philip Withnall [Tue, 23 Sep 2014 13:57:12 +0000 (14:57 +0100)]
agent: Eliminate a memset() for local variable initialisation
Stack variables can be initialised to zero by explicit assignment.
This introduces no functional changes.
Philip Withnall [Tue, 23 Sep 2014 13:56:52 +0000 (14:56 +0100)]
agent: Use g_slist_free_full() to reduce code
This introduces no functional changes.
Philip Withnall [Tue, 23 Sep 2014 13:55:54 +0000 (14:55 +0100)]
agent: Add some preconditions to internal API
To try and avoid NULL NiceCandidates entering our internal state.
Philip Withnall [Tue, 23 Sep 2014 13:54:59 +0000 (14:54 +0100)]
agent: Add some preconditions to public API
To try and avoid NULL NiceCandidates entering our internal state.
Olivier Crête [Mon, 22 Sep 2014 17:04:26 +0000 (13:04 -0400)]
agent: Clear GSource timeout also for GUPnP timeout
Fix regressions introduced by the following patch:
Clear existing GSource timeouts before adding new ones
Philip Withnall [Fri, 5 Sep 2014 09:50:23 +0000 (10:50 +0100)]
agent: Drop valid but unmatched STUN packets
Rather than passing them through to the application. Invalid STUN
packets are still passed through, but it causes unnecessary noise and
corruption for higher-level applications to receive STUN packets they
weren’t expecting.
This is permitted by RFC 5389, §7.3.0:
If any errors are detected, the message is silently discarded.
In the case when STUN is being multiplexed with another protocol,
an error may indicate that this is not really a STUN message; in
this case, the agent should try to parse the message as a different
protocol.
where I interpret ‘error’ to mean ‘validation error’ rather than (e.g.)
‘unrecognised attribute’ or ‘unmatched response’ where the STUN packet
is otherwise perfectly formed.
Philip Withnall [Fri, 19 Sep 2014 12:22:09 +0000 (13:22 +0100)]
agent: Remove dangling pointers on NiceSocket destruction
If a NiceSocket is destroyed, various pointers are currently left
dangling to it in the conncheck state. These can cause crashes if (for
example) a CandidateCheckPair with such a dangling pointer is then used;
the GSocket methods will fail.
Fix this by explicitly removing the socket and all NiceCandidates which
wrap it from various areas of the state.
Philip Withnall [Fri, 19 Sep 2014 12:20:25 +0000 (13:20 +0100)]
agent: Factor out state transition to FAILED for CandidateCheckPairs
This introduces no functional changes.
Philip Withnall [Fri, 19 Sep 2014 12:19:40 +0000 (13:19 +0100)]
agent: Factor out free function for IncomingCheck
Philip Withnall [Fri, 19 Sep 2014 10:03:27 +0000 (11:03 +0100)]
agent: Improve comments for container element types
To allow enhanced grepping for what structs point to other structs.
Philip Withnall [Fri, 19 Sep 2014 08:56:21 +0000 (09:56 +0100)]
agent: Clear existing GSource timeouts before adding new ones
Modify agent_timeout_add_with_context() to force destroying and freeing
of an existing GSource before overwriting it with a new one (probably
with an updated timeout period).
This fixes a case in priv_map_reply_to_relay_refresh() where the TURN
candidate refresh timer was being overwritten with a new one, without
the old one being destroyed. This lead to two timeouts existing, only
one of which would be destroyed when the CandidateRefresh struct was
freed, leaking the other one (in the main context) and allowing it to be
later dispatched with a dangling CandidateRefresh pointer.
The modification to agent_timeout_add_with_context() should prevent this
happening in new code in future.
Philip Withnall [Fri, 19 Sep 2014 07:57:53 +0000 (08:57 +0100)]
agent: Add names to timer GSources
Modify the agent_timeout_add_with_context() utility function to
automatically add names to the timer GSources it creates. This makes
them a little easier to identify when debugging.
Philip Withnall [Tue, 2 Sep 2014 10:56:37 +0000 (11:56 +0100)]
socket: Return early from socket functions if the socket is closed
Explicitly check whether the socket is closed (universally represented
as sock->priv == NULL) before doing anything else in the socket methods.
This should safely return from unusual situations where the socket has
been closed and part-destroyed but still ends up having send() or recv()
called on it.
Philip Withnall [Tue, 2 Sep 2014 10:55:19 +0000 (11:55 +0100)]
socket: Fix return values of socket_send_messages in socks5.c
The return value is the number of messages sent, or -1 on error — not a
boolean.
Philip Withnall [Mon, 1 Sep 2014 14:56:34 +0000 (15:56 +0100)]
stun: Make a warning message more prominent
Since dropping a STUN message due to having insufficient buffer space in
libstun can cause ICE negotiation to fail. So we want it to be more
obvious in the logs.
Olivier Crête [Sat, 30 Aug 2014 21:13:29 +0000 (17:13 -0400)]
agent: Put one StunAgent per Component
Otherwise we risk running out of space in the space limited saved ids
list. So the easiest way to do that is to put one StunAgent in each
component. It may be advisable to just give up on not allocating
memory inside libstun and just use a regular hash table in there
instead.
Olivier Crête [Mon, 1 Sep 2014 13:26:32 +0000 (09:26 -0400)]
conncheck: Fetch stream once instead of twice
Philip Withnall [Mon, 1 Sep 2014 15:01:02 +0000 (16:01 +0100)]
agent: Agent can reach EOS after having received something
The agent can reach EOS after performing one or more non-empty reads in
its read loop, and then hitting EOS on the underlying stream. That means
reached_eos is TRUE, but n_valid_messages is non-zero.
Weaken the postcondition to reflect this.
Philip Withnall [Tue, 26 Aug 2014 08:31:37 +0000 (09:31 +0100)]
agent: Fix locking when calling component_free()
As mentioned in the previous commit message, component_free() requires
the agent lock to be released, but component_close() requires it to be
held. Fix the locking in nice_agent_remove_stream() to do that.
Philip Withnall [Tue, 26 Aug 2014 08:30:26 +0000 (09:30 +0100)]
agent: Split component_free() into component_[close|free]()
Closing a component and freeing it are slightly conceptually different.
As it happens, freeing it can result in the disposal of the component's
NiceIOStream, which requires the agent lock to be released — but closing
the rest of the Component requires the agent lock to be held. Splitting
the function up simplifies locking handling in other parts of the code
(see next commit).
Philip Withnall [Mon, 25 Aug 2014 10:31:52 +0000 (11:31 +0100)]
agent: Fix initialisation of the agent mutex for old GLib versions
The mutex is now non-recursive. This is left over from commit
1deee693.
Philip Withnall [Mon, 25 Aug 2014 06:58:39 +0000 (07:58 +0100)]
tests: Fix test-build-io-stream for EOS changes
From commit
7b6935c66738d855c84fba291d47ece6ce1c43e2.
Philip Withnall [Mon, 25 Aug 2014 06:55:45 +0000 (07:55 +0100)]
agent: Add missing bracket
Bad merge in commit
67179ffd8fb1a39936c54ce611e8deeec8705ac3.
Philip Withnall [Sat, 23 Aug 2014 07:45:30 +0000 (08:45 +0100)]
agent: Add missing stream-closed checks for pseudo-TCP
This got lost in rebasing the FIN–ACK work. component->tcp is now only
NULL before the pseudo-TCP stream is initially created — afterwards, it
is always non-NULL, but pseudo_tcp_stream_is_closed() returns TRUE if
the stream has closed (gracefully or on errors). This allows
differentiation between the states: TCP support was never initialised;
and TCP support was initialised but is now closed.
Philip Withnall [Fri, 22 Aug 2014 07:24:57 +0000 (08:24 +0100)]
agent: Return 0 on EOS from nice_input_stream_read()
We should return 0 instead of G_IO_ERROR_CLOSED for consistency with
recv() and read().
Philip Withnall [Tue, 12 Aug 2014 08:40:09 +0000 (09:40 +0100)]
pseudotcp: Ensure shutdown member is not overwritten
Add a few safeguards to ensure that once priv->shutdown is set, it is
not overwritten with a different value.
Philip Withnall [Tue, 12 Aug 2014 08:37:34 +0000 (09:37 +0100)]
pseudotcp: Fix EOS return from recv() in non-FIN–ACK mode
Change pseudo_tcp_socket_recv() to return 0 if in non-FIN–ACK mode and
the socket has been shut down. This makes the behaviour of FIN–ACK and
non-FIN–ACK modes consistent.
This is a behaviour change from before the introduction of FIN–ACK mode
— previously, pseudo_tcp_socket_recv() would return -1 and set the
ENOTCONN error if called after the socket had been shut down (i.e. after
pseudo_tcp_socket_get_next_clock() had returned FALSE). The new
behaviour will hopefully not break anything.
Philip Withnall [Tue, 12 Aug 2014 08:35:20 +0000 (09:35 +0100)]
pseudotcp: Ensure socket is closed after shutdown in non-FIN–ACK mode
If the socket is closed (with pseudo_tcp_socket_close()) with FIN–ACK
unsupported, it should transition to the CLOSED state when
pseudo_tcp_socket_get_next_clock() returns FALSE (as that’s the
old-style indication that the socket has closed).
This is a behaviour change from before the introduction of FIN–ACK:
previously, the socket would stay in the ESTABLISHED state. However,
this should not have been easily detectable by consumers of the API,
as pseudo_tcp_socket_is_closed() did not exist then.
Philip Withnall [Tue, 12 Aug 2014 08:05:41 +0000 (09:05 +0100)]
pseudotcp: Add a PseudoTcpSocket:support-fin-ack property
This allows FIN–ACK support to be disabled entirely. This is mostly for
testing purposes, since TCP_OPT_FIN_ACK is negotiated when establishing
the connection, and is disabled if the other side doesn’t support it.
This includes an interoperability test.
Philip Withnall [Fri, 8 Aug 2014 13:38:56 +0000 (14:38 +0100)]
tests: Fix race conditions in test-thread
As found by tsan.
Philip Withnall [Fri, 8 Aug 2014 12:35:17 +0000 (13:35 +0100)]
agent: Close pseudo-TCP streams when closing G[IO|Input|Output]Streams
This means that users of the Nice[Input|Output|IO]Stream API can easily
close TCP connections without having to hack around with libnice
internals.
Philip Withnall [Fri, 8 Aug 2014 11:27:20 +0000 (12:27 +0100)]
docs: Ignore some undocumentable/non-public API
Philip Withnall [Fri, 8 Aug 2014 10:07:17 +0000 (11:07 +0100)]
pseudotcp: Add pseudo_tcp_socket_shutdown() support
This is analogous to the UNIX shutdown() function, allowing either or
both sides of a pseudo-TCP connection to be shut down.
Philip Withnall [Fri, 8 Aug 2014 08:26:39 +0000 (09:26 +0100)]
pseudotcp: Shorten the TIME-WAIT state timeout dramatically
The TIME-WAIT timeout is typically 2×MSL (on the order of 60 seconds),
which is needed to be able to reject delayed segments from closed
conversations. However, the underlying socket layer for pseudo-TCP takes
care of channel numbering so that segments don’t end up being sent to
closed conversations. Therefore, the TIME-WAIT state can be eliminated
(by shortening the timeout). The code for the state is kept around so
that the pseudo-TCP implementation matches the TCP specification
correctly, which will simplify maintenance.
Philip Withnall [Wed, 6 Aug 2014 14:35:20 +0000 (15:35 +0100)]
tests: Add a new segment-by-segment test framework for pseudo-TCP
This explicitly removes all timers and main loop considerations from the
pseudo-TCP testing, which considerably simplifies arranging packet swaps
and mistimings for testing purposes.
This test suite includes a few tests for the FIN–ACK support.
Philip Withnall [Thu, 17 Jul 2014 15:44:45 +0000 (16:44 +0100)]
pseudotcp: Add optional FIN–ACK and RST support
In order to detect cases where the peer closes its connection without an
explicit in-band close message (e.g. in protocols such as Telnet where
there is none), pseudo-TCP needs to grow support for a shutdown
handshake, following the TCP FIN–ACK specification. Arguably it should
have had this all along, but Jingle apparently doesn’t need it.
This adds support for FIN–ACK to the pseudo-TCP implementation. It is
backwards-compatible, only being used if the TCP_OPT_FIN_ACK option is
specified in the SYN segment.
If enabled, full-duplex closes are supported, and the standard method
for notifying a peer of the other end closing its connection (returning
0 from recv()) is used.
Also allow rapidly tearing down a connection, discarding unsent and
unreceived data, by sending an RST segment. This preserves the ability to
do a forced socket closure with pseudo_tcp_socket_close(sock, TRUE).
It also permits graceful socket shutdown in the case where the final ACK
is lost, and one peer gets stuck in the LAST-ACK state: that peer will
eventually re-transmit its FIN segment. The other peer, in the CLOSED
state, will respond with a RST segment, and the first peer will then
reach CLOSED.
References (most useful first):
• http://tools.ietf.org/html/rfc793#section-3.5
• http://tools.ietf.org/html/rfc1122#page-87
• http://vincent.bernat.im/en/blog/2014-tcp-time-wait-state-linux.html
• http://tools.ietf.org/html/rfc675
Diagram:
•
http://en.wikipedia.org/wiki/Transmission_Control_Protocol#mediaviewer/File:TCP_CLOSE.svg
Philip Withnall [Fri, 8 Aug 2014 07:46:25 +0000 (08:46 +0100)]
agent: Handle EPIPE from pseudo_tcp_socket_send()
As with send(), pseudo_tcp_socket_send() may return EPIPE if the local
side of the connection has been closed using close() or shutdown(). It
currently doesn’t, but will do once pseudo-TCP FIN–ACK support has been
implemented.
Philip Withnall [Fri, 8 Aug 2014 07:44:49 +0000 (08:44 +0100)]
agent: Don’t clear the PseudoTcpSocket on error
Instead, keep the closed object around. This allows differentiation
between the states: TCP support was never initialised; and TCP support
was initialised but is now closed.
Philip Withnall [Fri, 8 Aug 2014 09:00:03 +0000 (10:00 +0100)]
pseudotcp: Fix a typo in some documentation
Philip Withnall [Fri, 8 Aug 2014 08:57:40 +0000 (09:57 +0100)]
pseudotcp: Factor out state change functions for ESTABLISHED and CLOSED
Both state changes need some follow-up code executed immediately after,
which was duplicated in a number of places. Factor that out. This
introduces no behavioural changes.