platform/upstream/libnice.git
9 years agoagent: Expand simple example in documentation to mention GMainContext
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

9 years agoagent: Close pseudo-TCP socket earlier in component_close()
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.

9 years agostun: Rename a symbol to avoid a naming clash with OpenSSL
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_’.

10 years agoFix documentation relating to how remote crendentials should be set for new ICE format
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

10 years agoagent: ignore externally set peer-reflexive candidates
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().

10 years agoVersion 0.1.8.1
Olivier Crête [Thu, 9 Oct 2014 21:37:49 +0000 (17:37 -0400)]
Version 0.1.8.1

10 years agoVersion 0.1.8 0.1.8
Olivier Crête [Thu, 9 Oct 2014 20:43:16 +0000 (16:43 -0400)]
Version 0.1.8

10 years agoagent: Remove extra post condition
Olivier Crête [Thu, 9 Oct 2014 21:05:32 +0000 (17:05 -0400)]
agent: Remove extra post condition

It prevent the GError from happening

10 years agoudp-turn: Don't dereference priv before check that it's valid
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

10 years agoagent: Pass all packets that are not acceptable STUN packets to the application
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.

10 years agoagent: Re-emit gathering done after a new relay has been set
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

10 years agopseudotcp: Print unsigned int with %u not %d
Olivier Crête [Sun, 5 Oct 2014 18:06:58 +0000 (14:06 -0400)]
pseudotcp: Print unsigned int with %u not %d

10 years agoagent: Add function to retrieve the current component state
Olivier Crête [Sun, 5 Oct 2014 17:34:22 +0000 (13:34 -0400)]
agent: Add function to retrieve the current component state

10 years agoagent: Make the stun-pacing-timer app configurable
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().

10 years agopseudotcp: Set default RTO to 1 second
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.

10 years agoconncheck: Computer STUN retransmission timeout (RTO) dynamically
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

10 years agodiscovery: Remove extra \n in g_debug()
Olivier Crête [Fri, 3 Oct 2014 21:44:59 +0000 (17:44 -0400)]
discovery: Remove extra \n in g_debug()

10 years agoagent: Only add pairs once on gathering done
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

10 years agoconncheck: Insert the candidates sorted after setting the priority
Olivier Crête [Fri, 3 Oct 2014 20:30:16 +0000 (16:30 -0400)]
conncheck: Insert the candidates sorted after setting the priority

10 years agoconncheck: Insert peer reflexive pair sorted
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.

10 years agoconncheck: Don't remove elements in the conncheck list while iterating it
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.

10 years agoagent: Avoid leaking UPnP mappings between streams
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

10 years agoagent: Avoid restarting the GUPnP client on every gather
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.

10 years agoagent: Ignore UPnP mappings after the UPnP discovery has finished
Olivier Crête [Wed, 1 Oct 2014 02:49:16 +0000 (22:49 -0400)]
agent: Ignore UPnP mappings after the UPnP discovery has finished

10 years agocomponent: On ICE restart reset selected pair priority to 0
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.

10 years agoagent: Add some preconditions to public API
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.

10 years agobuild: Update .gitignore
Philip Withnall [Fri, 5 Sep 2014 09:54:11 +0000 (10:54 +0100)]
build: Update .gitignore

10 years agooutputstream: Fix double unref
Olivier Crête [Mon, 11 Aug 2014 18:27:40 +0000 (14:27 -0400)]
outputstream: Fix double unref

10 years agoagent: Remove socket source on HUP
Olivier Crête [Sat, 9 Aug 2014 20:18:33 +0000 (16:18 -0400)]
agent: Remove socket source on HUP

10 years agoagent: Declare the stream to be CONNECTED only if a pair is selected
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.

10 years agonice.pc: Include ${includedir} directly
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.

10 years agoagent: Document interaction between closing and removing streams
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

10 years agotests: Add more helgrind suppressions
Philip Withnall [Fri, 26 Sep 2014 11:57:18 +0000 (12:57 +0100)]
tests: Add more helgrind suppressions

10 years agoagent: Delay some assignments until the agent is locked
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.

10 years agoagent: Warn if a Component is freed without first being closed
Philip Withnall [Thu, 25 Sep 2014 12:55:42 +0000 (13:55 +0100)]
agent: Warn if a Component is freed without first being closed

10 years agoagent: Close Streams before freeing them
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.

10 years agotests: Add a Valgrind suppression file
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`.

10 years agotests: Fix various memory leaks in the unit tests
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.

10 years agoagent: Check STUN buffer is non-NULL before getting its message ID
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)

10 years agoagent: Fix a leak of a GCancellable and its GSource
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.

10 years agoagent: Fix a minor leak in an error handling path
Philip Withnall [Thu, 25 Sep 2014 08:21:18 +0000 (09:21 +0100)]
agent: Fix a minor leak in an error handling path

10 years agotests: Move a closure from the heap to the stack
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.

10 years agoconncheck: Don't use CandidateRefresh after freeing it
Olivier Crête [Wed, 24 Sep 2014 02:06:52 +0000 (22:06 -0400)]
conncheck: Don't use CandidateRefresh after freeing it

10 years agosocket: Fix a leak on the slow TURN packet handling path
Philip Withnall [Tue, 23 Sep 2014 14:20:31 +0000 (15:20 +0100)]
socket: Fix a leak on the slow TURN packet handling path

10 years agoagent: Eliminate a memset() for local variable initialisation
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.

10 years agoagent: Use g_slist_free_full() to reduce code
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.

10 years agoagent: Add some preconditions to internal API
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.

10 years agoagent: Add some preconditions to public API
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.

10 years agoagent: Clear GSource timeout also for GUPnP timeout
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

10 years agoagent: Drop valid but unmatched STUN packets
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.

10 years agoagent: Remove dangling pointers on NiceSocket destruction
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.

10 years agoagent: Factor out state transition to FAILED for CandidateCheckPairs
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.

10 years agoagent: Factor out free function for IncomingCheck
Philip Withnall [Fri, 19 Sep 2014 12:19:40 +0000 (13:19 +0100)]
agent: Factor out free function for IncomingCheck

10 years agoagent: Improve comments for container element types
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.

10 years agoagent: Clear existing GSource timeouts before adding new ones
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.

10 years agoagent: Add names to timer GSources
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.

10 years agosocket: Return early from socket functions if the socket is closed
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.

10 years agosocket: Fix return values of socket_send_messages in socks5.c
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.

10 years agostun: Make a warning message more prominent
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.

10 years agoagent: Put one StunAgent per Component
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.

10 years agoconncheck: Fetch stream once instead of twice
Olivier Crête [Mon, 1 Sep 2014 13:26:32 +0000 (09:26 -0400)]
conncheck: Fetch stream once instead of twice

10 years agoagent: Agent can reach EOS after having received something
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.

10 years agoagent: Fix locking when calling component_free()
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.

10 years agoagent: Split component_free() into component_[close|free]()
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).

10 years agoagent: Fix initialisation of the agent mutex for old GLib versions
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.

10 years agotests: Fix test-build-io-stream for EOS changes
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.

10 years agoagent: Add missing bracket
Philip Withnall [Mon, 25 Aug 2014 06:55:45 +0000 (07:55 +0100)]
agent: Add missing bracket

Bad merge in commit 67179ffd8fb1a39936c54ce611e8deeec8705ac3.

10 years agoagent: Add missing stream-closed checks for pseudo-TCP
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.

10 years agoagent: Return 0 on EOS from nice_input_stream_read()
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().

10 years agopseudotcp: Ensure shutdown member is not overwritten
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.

10 years agopseudotcp: Fix EOS return from recv() in non-FIN–ACK mode
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.

10 years agopseudotcp: Ensure socket is closed after shutdown in non-FIN–ACK mode
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.

10 years agopseudotcp: Add a PseudoTcpSocket:support-fin-ack property
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.

10 years agotests: Fix race conditions in test-thread
Philip Withnall [Fri, 8 Aug 2014 13:38:56 +0000 (14:38 +0100)]
tests: Fix race conditions in test-thread

As found by tsan.

10 years agoagent: Close pseudo-TCP streams when closing G[IO|Input|Output]Streams
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.

10 years agodocs: Ignore some undocumentable/non-public API
Philip Withnall [Fri, 8 Aug 2014 11:27:20 +0000 (12:27 +0100)]
docs: Ignore some undocumentable/non-public API

10 years agopseudotcp: Add pseudo_tcp_socket_shutdown() support
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.

10 years agopseudotcp: Shorten the TIME-WAIT state timeout dramatically
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.

10 years agotests: Add a new segment-by-segment test framework for pseudo-TCP
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.

10 years agopseudotcp: Add optional FIN–ACK and RST 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

10 years agoagent: Handle EPIPE from pseudo_tcp_socket_send()
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.

10 years agoagent: Don’t clear the PseudoTcpSocket on error
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.

10 years agopseudotcp: Fix a typo in some documentation
Philip Withnall [Fri, 8 Aug 2014 09:00:03 +0000 (10:00 +0100)]
pseudotcp: Fix a typo in some documentation

10 years agopseudotcp: Factor out state change functions for ESTABLISHED and CLOSED
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.

10 years agopseudotcp: Split out some state checks
Philip Withnall [Fri, 8 Aug 2014 08:54:23 +0000 (09:54 +0100)]
pseudotcp: Split out some state checks

This clarifies the code a little, and does not introduce functional
changes.

10 years agopseudotcp: Tidy up buffer size handling in queue_connect_message()
Philip Withnall [Fri, 8 Aug 2014 08:49:09 +0000 (09:49 +0100)]
pseudotcp: Tidy up buffer size handling in queue_connect_message()

This removes some hard-coded offsets. It does not introduce any
functional changes.

10 years agopseudotcp: Store segment flags in SSegment
Philip Withnall [Fri, 8 Aug 2014 08:44:46 +0000 (09:44 +0100)]
pseudotcp: Store segment flags in SSegment

Rather than reconstructing the flag state when sending a segment, just
store the flags explicitly. This does not introduce any behavioural
changes.

10 years agopseudotcp: Allow the ‘current’ time to be set
Philip Withnall [Wed, 6 Aug 2014 14:37:47 +0000 (15:37 +0100)]
pseudotcp: Allow the ‘current’ time to be set

This is needed for the upcoming new test suite for pseudo-TCP. It
shouldn’t be used in normal code — only in tests.

Ideally, the pseudo-TCP code should originally never have called
g_get_monotonic_time() itself, and should have always taken a time
parameter from the caller; then it would be more testable.
Unfortunately, API guarantees prevent this from being changed now.

10 years agopseudotcp: Add socket state to debug messages
Philip Withnall [Tue, 5 Aug 2014 16:39:01 +0000 (17:39 +0100)]
pseudotcp: Add socket state to debug messages

This should make debugging a little clearer.

10 years agopseudotcp: Put TCP flags in an enum
Philip Withnall [Fri, 8 Aug 2014 08:05:19 +0000 (09:05 +0100)]
pseudotcp: Put TCP flags in an enum

The on-the-wire flags (FIN, RST, ACK, etc.) should be in an enum to
clarify the code a little. This introduces no functional changes.

10 years agopseudotcp: Put TCP options in an enum
Philip Withnall [Tue, 5 Aug 2014 09:42:20 +0000 (10:42 +0100)]
pseudotcp: Put TCP options in an enum

This tidies things up a little. No functional changes.

10 years agotests: Improve debug output of pseudotcp tests slightly
Philip Withnall [Fri, 8 Aug 2014 09:12:36 +0000 (10:12 +0100)]
tests: Improve debug output of pseudotcp tests slightly

10 years agotests: Add setlocale() call to test-pseudotcp.c
Philip Withnall [Fri, 8 Aug 2014 09:10:39 +0000 (10:10 +0100)]
tests: Add setlocale() call to test-pseudotcp.c

This ensures UTF-8 output is printed correctly.

10 years agotests: Add a logging domain for the unit tests
Philip Withnall [Fri, 8 Aug 2014 08:26:11 +0000 (09:26 +0100)]
tests: Add a logging domain for the unit tests

This clarifies the log output a little.

10 years agosocket: Gracefully return from send() if the socket is closed
Philip Withnall [Fri, 8 Aug 2014 10:32:59 +0000 (11:32 +0100)]
socket: Gracefully return from send() if the socket is closed

10 years agoagent: Ensure Component.selected_pair is cleaned when freeing candidates
Philip Withnall [Fri, 8 Aug 2014 10:31:06 +0000 (11:31 +0100)]
agent: Ensure Component.selected_pair is cleaned when freeing candidates

When freeing candidates (component_free_socket_sources()), the sockets
which back Component.selected_pair.[local|remote] are closed and their
addresses destroyed. Component.selected_pair should be cleared as well
to allow calling code to condition on (Component.selected_pair == NULL)
to see if it’s invalid.

10 years agoagent: Factor out common clean up code in Component
Philip Withnall [Fri, 8 Aug 2014 10:30:43 +0000 (11:30 +0100)]
agent: Factor out common clean up code in Component

This introduces no functional changes.

10 years agoagent: Tidy up a few variable references
Philip Withnall [Fri, 8 Aug 2014 10:29:41 +0000 (11:29 +0100)]
agent: Tidy up a few variable references

This simplifies the code a little. No functional changes.

10 years agoagent: Rearchitect message handling to use GErrors for EWOULDBLOCK
Philip Withnall [Fri, 8 Aug 2014 07:48:08 +0000 (08:48 +0100)]
agent: Rearchitect message handling to use GErrors for EWOULDBLOCK

Previously, an EWOULDBLOCK return value from the low-level socket calls
(including PseudoTcpSocket) would be represented by a zero number of
bytes (or messages) read by the agent. This conflicts with the use of
zero to represent end of stream (EOS) for pseudo-TCP connections, where
the sender has indicated that they are not going to send any more bytes.

So, now use GError (G_IO_ERROR_WOULD_BLOCK) to represent EWOULDBLOCK,
just like the GSocket functions. Zero is reserved exclusively for if:
 • the number of requested bytes/messages is zero; or
 • reliable mode is enabled and EOS is reached.

This does change the documented behaviour of the NiceAgent send/recv
API, but only by allowing a new behaviour (returning zero) rather than
by changing an existing one, so it should be OK.

10 years agopseudotcp: Fix the type of an argument to apply_option()
Philip Withnall [Tue, 12 Aug 2014 08:07:49 +0000 (09:07 +0100)]
pseudotcp: Fix the type of an argument to apply_option()

We’re dealing with unsigned bytes here, not chars. This will become
important when adding new TcpOptions numbered from 254 downwards.