- 23 Mar, 2018 9 commits
-
-
Fabrice Bellet authored
With this patch we prevent the username and the password of a candidate to be modified during a session, as required by the RFC, sect 9.1.2. This is also needed from a memory management point of view, because the password string pointer may be recorded in the components stun agent sent_ids[] struct key member, and freeing these values there may cause an use-after-free condition, when an inbound stun is received from this candidate. This behavior has been observed with pidgin, xmpp, and farstream when a same remote candidates are "updated" several times, even if the credentials don't change in this case. Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1917
-
Fabrice Bellet authored
The tcp server reflexive discovered local candidates must be ignored when force_relay is set. Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1899
-
Fabrice Bellet authored
Since commit 17f30e44, we may have a stream with an empty conncheck list, and such a stream obviously should not be tested for failed components. Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1895
-
Fabrice Bellet authored
Verify the compatibility of the socket domain with the stun server IP address, before sending a request. Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1894
-
Fabrice Bellet authored
Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1893
-
Fabrice Bellet authored
Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1892
-
Fabrice Bellet authored
With this patch, we put the pair in state failed if we cannot send the connection check, for example due to missing local credentials. Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1891
-
Fabrice Bellet authored
the first case of test-new-dribble (standard-test) is updated, by making the credentials swap between the left and right agent asymmetric. Previously, ragent started to receive stun requests without initially knowing lagent candidates. Now, ragent also ignores lagent credentials. This modification allows to test changes introduced by the previous commit. Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1890
-
Fabrice Bellet authored
With this patch we simplify the code used to handle the incoming stun request when remote candidates or remote credentials have not been received yet. When the remote credentials is unknown, the stun request is stored in a list of incoming_checks for later processing, and no further processing is done, except responding to the request. When the remote credentials are received, the triggered checks for these incoming checks can now be queued, and the related pairs are created. If the remote candidates have not been received when the stun request on a valid local port arrives, a peer-reflexive remote candidate will be created. This candidate may need to be updated later when remote candidates are finally received, including candidate priority and foundation, and also related pairs. Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1889
-
- 22 Mar, 2018 1 commit
-
-
Jozsef Vass authored
The variable tie is actually never read.
-
- 28 Nov, 2017 4 commits
-
-
Youness Alaoui authored
If a relay gives us an alternate-server, we need to cancel and reset every candidate discovery attempt that uses the same server, to avoid ending up with one component on one server and the other component on another server (causing relay candidates with mismatched foundations).
-
Youness Alaoui authored
The discovery_unsched_items is decremented every time a DiscoveryCandidate goes from non-pending to pending. So if we restart a check by setting pending to FALSE, we should re-increase the discovery_unsched_items.
-
Youness Alaoui authored
The MS Office TURN servers will always return the MS_ALTERNATE_SERVER in allocation responses, and if they are not handled, we end up using the main turn server to send allocation requests that then get sent to the alternate server which will return the XOR_MAPPED_ADDRESS containing the IP address of the turn server that proxied the message instead of our own actual external IP.
-
Miguel París authored
One or more streams might not have any connection check list if the number of streams differs from the peer agent. Differential Revision: https://phabricator.freedesktop.org/D1880
-
- 27 Nov, 2017 2 commits
-
-
Fabrice Bellet authored
Differential Revision: https://phabricator.freedesktop.org/D1888
-
Fabrice Bellet authored
With this patch, we stash the controlling mode property change, and apply it safely, when it won't interfere with an ongoing conncheck running. According to RFC5245, sect 5.2. "Determining Role", the role is determined for a session, and persists unless an ICE is restarted. Differential Revision: https://phabricator.freedesktop.org/D1887
-
- 12 Sep, 2017 3 commits
-
-
Philip Withnall authored
https://phabricator.freedesktop.org/T7798Signed-off-by:
Philip Withnall <withnall@endlessm.com> Reviewed-by:
Olivier Crête <olivier.crete@collabora.com> Differential Revision: https://phabricator.freedesktop.org/D1819
-
Philip Withnall authored
Spotted by Lukas Gradl on the mailing list. Signed-off-by:
Philip Withnall <withnall@endlessm.com>
-
Philip Withnall authored
This code is not 1000 years old. Signed-off-by:
Philip Withnall <withnall@endlessm.com>
-
- 05 Sep, 2017 8 commits
-
-
Fabrice Bellet authored
When a pair is nominated while in state failed, we first move back to state connecting, then we update the selected pair, and finally we move to state connected.
-
Fabrice Bellet authored
When a new pair is created from an unknown remote candidate, it should be enqueue for a triggered check, to allow it to be marked as nominated on response arrival in priv_mark_pair_nominated(). Creating it in waiting state is not sufficient since the update in priv_mark_pair_nominated() from previous commits. Differential Revision: https://phabricator.freedesktop.org/D1763
-
Fabrice Bellet authored
With this patch, we fix an ambiguity of some parts of the spec, when the document refers to in-progress pairs, that also concern pairs in the triggered checks list. The first cast is in section 7.1.2.5, "Updating the Nominated Flag", when the in-progress pair will be nominated on response arrival. This is handled in function priv_mark_pair_nominated(), when a pair is put to the triggered check list in reaction to a matching inbound stun request. Such a pair in priv_mark_pair_nominated() will _always_ be in the triggered check list, from the previously called function priv_schedule_triggered_check(). The second case is in section 8.1.2, "Updating State" when an in-progress pair stops its retransmission when another pair of higher priority is already nominated. This is handled by function priv_prune_pending_checks(). Until now, pairs enqueued in the triggered check list move transiently to state waiting, according to 7.2.1.4. But this state causes wrong decisions in the two previous cases, because such pairs should in fact rather be considered "like in-progress", to avoid discarding them inadvertantly. This patch update the state of the triggered check list pairs to in-progress. It allows to remove exception handling cited above: the code is a bit more simple, and allows some refactoring in priv_mark_pair_nominated() between RFC and compatibility modes. Differential Revision: https://phabricator.freedesktop.org/D1762
-
Fabrice Bellet authored
This patch should improve the reliabily of the connection check by keeping the record of several simultaneous ongoing stun requests per pair. A new stun request on an in-progress pair typically is caused by in inbound stun request from the peer on this same pair. This is named "Triggered Checks" in the spec. When this situation arises, it is fair to handle these two stun requests simultaneously, the triggered check, and the initial ordinary check, since both can potentially succeed. Differential Revision: https://phabricator.freedesktop.org/D1761
-
Fabrice Bellet authored
We try to use stun_timer_remainder() less frequently, particularily in the debug messages, and favour of the next_tick value associated to the pair. Differential Revision: https://phabricator.freedesktop.org/D1760
-
Fabrice Bellet authored
We add a helper function to print the pair state in-extenso. Differential Revision: https://phabricator.freedesktop.org/D1759
-
Fabrice Bellet authored
With this patch we simplify the levels of code indentation. Differential Revision: https://phabricator.freedesktop.org/D1758
-
Olivier Crête authored
-
- 22 Jun, 2017 2 commits
-
-
Fabrice Bellet authored
This patch displays explicitely the controlling or controlled role of the agent. Differential Revision: https://phabricator.freedesktop.org/D874
-
Olivier Crête authored
Creates useless warnings when other libraries change. https://phabricator.freedesktop.org/T7770
-
- 21 Jun, 2017 11 commits
-
-
Olivier Crête authored
GClosures are not that cheap to setup
-
Olivier Crête authored
-
Olivier Crête authored
-
Olivier Crête authored
This makes it easier to read and more extensible.
-
Fabrice Bellet authored
Differential Revision: https://phabricator.freedesktop.org/D1756
-
Fabrice Bellet authored
Differential Revision: https://phabricator.freedesktop.org/D1754
-
Fabrice Bellet authored
At the end of the local candidate gathering process, we only create new pairs for streams that are in gathering state. Other stream that may be in ready state for example, due to a previously succeeded conncheck process, may have accumulated some couples (local,remote) candidates that have not resulted in the creation a new pair during this previous conncheck process, and we don't want these new pairs to be added now, because it would generate unneeded transition changes for a stream unconcerned by this gathering. Differential Revision: https://phabricator.freedesktop.org/D1755
-
Fabrice Bellet authored
This patch fixes the transition of a component from connecting to failed, that previously occured due to the propagation of the keep_timer_going variable, and to the final call to function priv_update_check_list_failed_components(), after the global agent timer was stopped. Previously, the code almost never entered to failed state, because the timer was going one, as long as the number of nominated pair was not enough, and as long as there were valid pairs not yet nominated. Even if all pair timers were over. The definition of the Failed state of a conncheck list is somewhat contradictory in the spec, depending on weather you read : * sect 5.7.4. "Computing States", "Failed: In this state, the ICE checks have not completed successfully for this media stream." or * sect 7.1.3.3. "Check List and Timer State Updates", "If all of the pairs in the check list are now either in the Failed or Succeeded state: If there is not a pair in the valid list for each component of the media stream, the state of the check list is set to Failed." Our understanding of the ICE spec is that, the proper way to enter failed state instead in when all connchecks have no longer in-progress pairs. All pairs are either in state succeeded, discovered, or failed. No timer is still running, and we have no hope that the conncheck list changes again, except if an unexpected STUN packet arrives later. All pairs in frozen state is a special case, that is handled separately (sect 7.1.3.3). A special grace delay is added before declaring a component in state Failed. This delay is not part of the RFC, and it is aimed to limit the cases when a conncheck list is reactivated just after it's been declared failed, causing a user visible transition from connecting to failed, and back from failed to connecting again. This is also required by the test suite, that counts exactly the number of time each state is entered, and doesn't expect these transcient failed states to happen (frequent due to the nature of the testsuite, less frequent in real life). Differential Revision: https://phabricator.freedesktop.org/D1111
-
Fabrice Bellet authored
Pairs with the state NICE_CHECK_CANCELLED are the pairs targeted for removal after the nomination of a pair with an higher priority, described in Section 8.1.2 "Updating States", item 2 of RFC 5245. They include also pairs that overflow the conncheck list size, but this is a somewhat more marginal situation. So we are mainly interested in the first use case of this state. This state mixes two different situations, that deserve a distinct handling : on one side, there are waiting or frozen pairs that must be removed, this is an immediate action that doesn't need a dedicated state for that. And on the other side, there are in-progress pairs that should no longer be retransmitted, because another pair with a higher priority has already been nominated. This patch removes the cancelled state, and adds a flag retransmit_on_timeout to deal with this last situation. Note that this case should not generate a triggered check, as per described in section 7.2.1.4, when the state of the pair is In-Progress or Failed, since this pair of lower priority has no hope to replace the nominated one. Differential Revision: https://phabricator.freedesktop.org/D1114
-
Fabrice Bellet authored
The pair recheck on timeout can easily cause repetitive rechecks in a ping-pong effect, if both peers with the same behaviour try to check the same pair almost simultaneously, and if the network rtt is greater than the initial timer rto. The reply to the initial stun request may arrive after the in-progress conncheck cancellation (described in RFC 5245, sect 7.2.1.4). Cancellation creates a new stun request, and forgets the initial one. The conncheck timer is restarted with the same initial value, so the same situation happens again later. We choose to avoid resetting the timer in such situation. After enough retransmissions, the timeout delay, that doubles after each timeout, becomes longer than the rtt, and the stun reply can be handled. Differential Revision: https://phabricator.freedesktop.org/D1115
-
Fabrice Bellet authored
We cancel the potential in-progress transaction cancellation, caused by sect 7.2.1.4 "Triggered Checks", when we receive a valid reply before transmission timeout, or just after timeout, when the pair is temporarily put on the triggered check list on the way to be rechecked. This situation is not covered by the RFC 5245. Differential Revision: https://phabricator.freedesktop.org/D1119
-