Fabrice Bellet [Sun, 23 Feb 2020 18:38:08 +0000 (19:38 +0100)]
conncheck: more debug
Fabrice Bellet [Sun, 5 Apr 2020 20:27:38 +0000 (22:27 +0200)]
conncheck: add debug about agent mode
Fabrice Bellet [Wed, 8 Apr 2020 14:59:40 +0000 (16:59 +0200)]
discovery: add the agent pointer to turn debug messages
Fabrice Bellet [Fri, 21 Feb 2020 14:40:56 +0000 (15:40 +0100)]
conncheck: drop some excessively verbose comment
Fabrice Bellet [Sun, 5 Apr 2020 19:11:19 +0000 (21:11 +0200)]
conncheck: wait for a pair until all stun requests are completed
Only the newest stun request may need to be retransmitted, according to
the pair retransmit flag. This is the first element of the
stun_transactions list. Older stun requests are just kept around until
their timeout expires, without retransmission.
The newest stun request is usually the last one that will timeout.
Current code was based on that assumption, causing the pair to fail when
the newest stun request timeout expires. This is not always true, and some
older stun requests may have a greater timeout delay.
So, we should wait until *all* stun requests of a given pair have
reached their timeout.
We also refactor this part of the code, to handle the first stun and the
other stun requests in the same loop.
Fabrice Bellet [Sun, 5 Apr 2020 19:18:55 +0000 (21:18 +0200)]
discovery: use different port numbers for every local host candidates
This constraint is added to handle the situation where the agent runs on
a box doing SNAT on one of its outgoing network interface. The NAT does
usually its best to ensure that source port number is preserved on the
external NAT address and port. This is called "port preservation" in RFC
4787.
When two local host candidates are allowed to have the same source port
number, we increase the risk that a first local host candidate *is* the
NAT mapping address and port of a second local host candidate, because
of the "port preservation" effect. When it happens, a server reflexive
candidate and a host candidate will have the same address and port.
For that situation to happen, a stun request must be emitted from the
internal address first, the NAT mapping doing the port preservation will
be created for the internal address, and when a stun request is sent
from the external address thereafter, a new NAT mapping will be created,
but without port preservation, because the previous mapping already took
that reservation.
The problem will occur on the remote agent, when receiving a stun request
from this address and port, that has no way to know wheather it comes from
the host or the server reflexive candidate, if both have been advertised
remotely, resulting in pair type mislabelling.
This case may happen more easily when a source port range is reduced.
Fabrice Bellet [Sun, 5 Apr 2020 18:56:38 +0000 (20:56 +0200)]
agent: stay in aggressive mode after conncheck has started
When remote tcp candidates are received late after the conncheck has
started, RFC 6554 suggests that we switch the nomination mode back from
aggressive to regular. The problem is that some stun requests may
already be inflight with the use-candidate stun flag set, and reverting
to regular mode in that case is too late, because these inflight
requests may nominate a pair on the remote agent, and not on the local
agent. We prefer to just ignore remote tcp candidates that are received
after the component state has reached state CONNECTING.
Fabrice Bellet [Tue, 3 Mar 2020 10:17:04 +0000 (11:17 +0100)]
agent: warn when reaching max number of turn and local addresses
Fabrice Bellet [Tue, 3 Mar 2020 10:13:37 +0000 (11:13 +0100)]
agent: fix boundary test of max turn servers and local addresses
We can accept up to 8 turn servers, with turn preference value
starting at zero. Also fix the error message.
Fabrice Bellet [Thu, 13 Feb 2020 17:50:07 +0000 (18:50 +0100)]
discovery: add a unique local preference value per turn server
This value is built from the position in the component turn servers
list, and from the base address network interface position in the list
of network interfaces. This value is used to ensure a unique candidate
priority for each one. Also ensure that the fields that compose the
local preference don't overlap, by checking their maximum value. See
RFC-8445, section 5.1.2.2 "Guidelines for Choosing Type and Local
Preferences".
Fabrice Bellet [Wed, 12 Feb 2020 15:07:33 +0000 (16:07 +0100)]
conncheck: rename peer-reflexive priority to stun-priority
Fabrice Bellet [Wed, 12 Feb 2020 14:48:12 +0000 (15:48 +0100)]
candidate: ensuring stun priority uniqueness no more needed
The uniqueness of candidate priorities is achieved by the iteration on
the ip local addresses for local host candidates, and also on their base
address for reflexive and relay candidates. Helper function checking
its uniqueness at allocation time is not required anyore.
The priority of the stun request (prflx_priority) is built from the
priority of the local candidate of the pair, according the RFC 5245,
section 7.1.2.1. This priority must be identical to a virtual "local
candidate of type peer-reflexive that would be learned as a consequence
of a check from this local candidate."
Outgoing stun requests will come from local candidates of type host or
type relayed. The priority uniqueness of local candidates of type host
implies the uniqueness of the computed peer-reflexive priority. And
relay local candidates cannot produce a peer-reflexive local candidate
by design, so we can safely use their unique local priority too in
the stun request.
Tim-Philipp Müller [Mon, 4 May 2020 12:01:02 +0000 (13:01 +0100)]
subprojects: bump glib wrap to latest stable branch for msys build
Should fix build failures with latest mingw compiler in msys.
Tim-Philipp Müller [Mon, 4 May 2020 11:52:35 +0000 (12:52 +0100)]
ci: update windows image
The old one (v8) was removed from the gstreamer registry it seems.
Fabrice Bellet [Fri, 21 Feb 2020 14:40:56 +0000 (15:40 +0100)]
conncheck: refactor local and remote candidates validation
The same code to get and validate local and remote candidates from an
incoming stun is shared between regular inbound stun, early checks
replay, and partially in the local peer-reflexive discovery function.
The selection of the matching local and remote candidate from an
incoming stun sometimes requires more information than just the local
socket, and the sender address and port. It happens more frequently when
the port range is reduced, and when the conncheck handles both tcp and
udp candidates.
To help to disambiguate such situations, we add supplementary checks
when two candidates in the list have the same address and and port
number:
* the type of the socket must compatible with the candidate transport.
A socket for a tcp candidate may be active of passive, but also
of type "tcp-bsd" when the parent active or passive socket is
replaced after a bind() or accept(). It gives several cases.
* the remote candidate transport and the local candidate transport must
be compatible
Fabrice Bellet [Thu, 20 Feb 2020 16:05:51 +0000 (17:05 +0100)]
conncheck: match remote candidates on the socket reliability
When the couple (address, port) is identical between two remote
candidates, we may have to match a remote candidate based on its socket
reliability.
Fabrice Bellet [Mon, 17 Feb 2020 15:29:41 +0000 (16:29 +0100)]
conncheck: match transport of peer-reflexive local candidates
Another some rare case, but we may have two local candidates with
the same couple (address, port) and a different transport.
Fabrice Bellet [Mon, 2 Mar 2020 12:23:35 +0000 (13:23 +0100)]
tests: add a unit test for private addresses ranges
This is a unix-only test
Fabrice Bellet [Thu, 13 Feb 2020 15:09:31 +0000 (16:09 +0100)]
interfaces: fix network to host conversion for private addresses
Fabrice Bellet [Wed, 12 Feb 2020 15:29:24 +0000 (16:29 +0100)]
tests: fix local preference based on interface rank in list
The local preference of UDP candidates is (now) determined by the
position of the IP address in the list returned by
nice_interfaces_get_local_ips().
Fabrice Bellet [Wed, 12 Feb 2020 14:40:52 +0000 (15:40 +0100)]
candidate: fix local preferences of udp candidates
udp candidates failed to call nice_candidate_ip_local_preferences()
and were all given the same local preference priority.
Fabrice Bellet [Wed, 19 Feb 2020 20:11:59 +0000 (21:11 +0100)]
agent: enforce turn deallocate stun pacing
In a way similar to other stun packets, we add a delay of Timer TA
(default is 20 ms) between each refresh candidate processing.
Fabrice Bellet [Tue, 18 Feb 2020 13:01:37 +0000 (14:01 +0100)]
conncheck: enforce keepalives stun requests pacing
Keepalives STUN requests should not be sent for each local host
candidate or each selected candidate in the single loop, but with a pacing
of at least Timer TA (default is 20 ms) between each request. It remains
compatible with a pause of Timer TR (default is 20 seconds) between each
keepalives batch.
Fabrice Bellet [Tue, 18 Feb 2020 12:58:09 +0000 (13:58 +0100)]
discovery: enforce stun requests pacing
Only a single STUN request should be sent per discovery tick to enforce
an overall pacing of 20ms per default between two STUN requests.
Stefan Becker [Fri, 28 Feb 2020 18:49:21 +0000 (20:49 +0200)]
interfaces: ignore only interfaces we really want to (II)
This improves commit
bd4b47813844aa0e68e9074b3ac516c383ca8bab
There is a second place where this fix is needed.
Olivier Crête [Thu, 13 Feb 2020 21:19:03 +0000 (16:19 -0500)]
gitlab-ci: Disable trigger for now
Olivier Crête [Fri, 10 May 2019 05:23:39 +0000 (01:23 -0400)]
gitlab-ci: Make the documentation acceptable for the website
Jakub Adam [Thu, 13 Feb 2020 19:33:18 +0000 (20:33 +0100)]
conncheck: Connection check reply must have a remote candidate
Ensure that a conncheck reply is coming from an address and port of a
known remote candidate and that the type of incoming socket matches that
candidate's transport.
Attemps to fix a Coverity issue in which no matching remote_candidate
gets found for a connectivity reply in conn_check_handle_inbound_stun()
(apparently due to transport mismatch), yet
priv_map_reply_to_conn_check_request() still successfully matches it
with a previous request.
Jakub Adam [Thu, 13 Feb 2020 18:59:39 +0000 (19:59 +0100)]
conncheck: Typo fix
Fabrice Bellet [Sat, 8 Feb 2020 21:00:57 +0000 (22:00 +0100)]
conncheck: display priorities in debug by spliting their values
We display 32-bit candidate priorities in hexadecimal to identify each
8-bit-aligned field more easily: type preference, local preference, and
component id. We display 64-bit pair priority by splitting their local
and remote part. See RFC-8445, section 5.1.2.1. "Recommended Formula",
and section 6.1.2.3. "Computing Pair Priority and Ordering Pairs".
Fabrice Bellet [Thu, 13 Feb 2020 15:11:00 +0000 (16:11 +0100)]
candidate: strip link-local ipv6 scope in local preference
This patch makes IPv6 link-local addresses obtain a unique ice
local preference when computing their priority, by stripping the
"%<interfacename>" added to them by getnameinfo(). Previously, all
these addresses obtained the same preference value, since the whole
local ips list was checked without finding a match.
Olivier Crête [Thu, 13 Feb 2020 00:54:49 +0000 (19:54 -0500)]
gitlab: Also push to coverity on manual builds
Olivier Crête [Thu, 13 Feb 2020 00:47:16 +0000 (19:47 -0500)]
stun usage turn: Check return value as is done elsewhere
Olivier Crête [Thu, 13 Feb 2020 00:45:32 +0000 (19:45 -0500)]
udp-bsd: Check all return value
Errors found by coverity
Olivier Crête [Thu, 13 Feb 2020 00:41:45 +0000 (19:41 -0500)]
stun test: Check return value as is done elsewhere
Olivier Crête [Thu, 13 Feb 2020 00:19:45 +0000 (19:19 -0500)]
conncheck: Pass arguments in the right order
The order of the passed arguments doesn't match the prototype of the
function. We exchange the local and remote candidates.
Olivier Crête [Thu, 13 Feb 2020 00:16:00 +0000 (19:16 -0500)]
udp-udp: Make coverity happy by giving it the whole union
Olivier Crête [Wed, 12 Feb 2020 23:41:27 +0000 (18:41 -0500)]
gitlab-ci: Build coverity using meson
Fabrice Bellet [Tue, 11 Feb 2020 12:20:10 +0000 (13:20 +0100)]
agent: refactor conncheck update_selected_pair
Fabrice Bellet [Sat, 8 Feb 2020 18:33:30 +0000 (19:33 +0100)]
agent: update the selected pair priority when foundation changes
The component selected pair priority may have its priority modified
if the foundation of its remote candidate changes. This update needs
to be recorded, even if its not signalled, since this value is used
in the conncheck in priv_update_selected_pair().
Fabrice Bellet [Fri, 7 Feb 2020 09:50:04 +0000 (10:50 +0100)]
agent: make code indentation a bit more homogeneous
Fabrice Bellet [Fri, 7 Feb 2020 08:17:46 +0000 (09:17 +0100)]
agent: verify the best nominated pair after priorities recalculation
When we get a set_remote_candidate late after the start of the
conncheck, some remote candidates type may change from peer-reflexive
to server-reflexive. This required a recalculation of the foundations
and of the priorities of the related pairs.
This recalculation may change the pairs ordering in the conncheck list:
a srflx:prflx pair becoming a srflx:srflx pair will have a lower
priority (RFC 8445, section 5.1.2.1 "Recommended Formula", "the type
preference for peer-reflexive candidates MUST be higher than that of
server-reflexive candidates.")
This recalculation is required to maintain the same pairs ordering in
both agent.
But in some cases, in aggressive nomination mode, this may also change
the selected pair, if this recalculation changed the highest priority
pair with the nominated flag.
Fabrice Bellet [Sun, 9 Feb 2020 20:49:23 +0000 (21:49 +0100)]
conncheck: send keepalive for candidates until connected state
We can start sending keepalive stun request for the selected pair, and
stop sending keepalive for candidates as soon as the conncheck reaches
state connected and not ready.
Fabrice Bellet [Sun, 9 Feb 2020 20:45:55 +0000 (21:45 +0100)]
conncheck: fix peer-reflexive priority in discovered pair
The peer-reflexive priority used in the STUN request is copied from the
parent succeeded pair to the discovered pair. This value is not required
for discovered pair, that won't emit stun requests themselves, but may
be used when such pair becomes the selected pair, and when keepalive
stun are emitted, using the sockptr and prflx_priority values from the
succeeded pair.
Fabrice Bellet [Sat, 8 Feb 2020 21:02:18 +0000 (22:02 +0100)]
conncheck: update peer-reflexive priority of the selected pair
Fabrice Bellet [Fri, 7 Feb 2020 08:09:08 +0000 (09:09 +0100)]
conncheck: do not create a pair from peer-reflexive local candidate
The usual case when forming the initial conncheck list from local and
remote candidates is to discard server-reflexive local candidates, since
the agent can not send requests directly from these candidates. But it
also may happen, when the remote candidates are received late after the
beginning of the conncheck, that we also have discovered peer-reflexive
local candidates, that must be discarded as well. See RFC 8445, section
6.1.2.4 "Pruning the Pairs".
Fabrice Bellet [Mon, 10 Feb 2020 20:33:03 +0000 (20:33 +0000)]
conncheck: prune oversized conncheck list based on pair state
Removing lower-priority pairs to keep the conncheck list under a fixed
size (RFC 8445, sect 6.1.2.5) may accidently remove pairs that we want
to preserve, like succeeded and discovered pairs. We choose to only
remove pairs that are not engaged yet in the connection check
processing, those in frozen state.
Fixes #95
sergey radionov [Sun, 2 Feb 2020 09:06:31 +0000 (16:06 +0700)]
interfaces: use fallback if getifaddrs failed
fix #99
Olivier Crête [Thu, 6 Feb 2020 21:43:12 +0000 (16:43 -0500)]
test-thread: Increase timeout
Olivier Crête [Thu, 6 Feb 2020 21:41:23 +0000 (16:41 -0500)]
test-io-stream-common: Remove io_stream pointer when removing stream
Olivier Crête [Thu, 6 Feb 2020 21:37:50 +0000 (16:37 -0500)]
gitlab-ci: No longer allow meson test failures
Olivier Crête [Thu, 6 Feb 2020 21:30:34 +0000 (16:30 -0500)]
test-send-recv: Change timeout to 150s
Olivier Crête [Thu, 6 Feb 2020 20:21:50 +0000 (15:21 -0500)]
tests: Make iostream test termination more reliable
Olivier Crête [Thu, 6 Feb 2020 02:12:11 +0000 (21:12 -0500)]
test-io-common: Abort on error
Olivier Crête [Thu, 6 Feb 2020 01:38:32 +0000 (20:38 -0500)]
gitlab-ci: Run meson tests in debug setup
This setup stores more debug info
Olivier Crête [Thu, 6 Feb 2020 01:15:28 +0000 (20:15 -0500)]
test-send-recv: Write in a loop until blocked
We're not sure that our callback will be called again if
we haven't filled the pipe.
Fabrice Bellet [Thu, 30 Jan 2020 19:32:44 +0000 (20:32 +0100)]
conncheck: print a supplementary pair flag in the debug dump
Fabrice Bellet [Wed, 5 Feb 2020 21:04:16 +0000 (21:04 +0000)]
conncheck: fix timer remainder return value
We show the timer remainder in milliseconds.
Xavier Claessens [Tue, 17 Dec 2019 15:54:37 +0000 (10:54 -0500)]
Fix building doc when libnice is a subproject
When passing strings to 'src_dir', they are assumed to be relative to
the root source dir of main project. That's possibly a Meson bug, it
should be relative to current source dir. In any case, it's better to
use include_directories() to avoid any confusion.
Olivier Crête [Wed, 11 Dec 2019 00:11:44 +0000 (19:11 -0500)]
Remove GTimeVal as it is deprecated in GLib
Olivier Crête [Fri, 6 Dec 2019 18:11:10 +0000 (13:11 -0500)]
gitlab-ci: Replace dependencies with needs for parallel running
Olivier Crête [Fri, 6 Dec 2019 19:37:09 +0000 (14:37 -0500)]
gitlab-ci: Add workaround for missing msys in Docker image
Olivier Crête [Fri, 6 Dec 2019 19:20:11 +0000 (14:20 -0500)]
gitlab-ci: Move CI to Windows 1809 builder
Also upgrade to Windows image v8 as the image needs to match the host
Jakub Adam [Tue, 26 Nov 2019 15:54:08 +0000 (16:54 +0100)]
conncheck: fix long time to connect if conncheck list gets filled late
When a stream's conncheck list is empty during
priv_conn_check_unfreeze_related() and the check pairs get added only
afterwards, the whole list may remain in frozen state for tens of
seconds until the lists of all other streams are completed.
If a connection check list missed the opportunity to be unfrozen
according to the process in RFC 5245 Section 7.1.3.2.3 because it was
empty at the time, unfreeze it immediately once some check pair gets
added.
Tim-Philipp Müller [Tue, 17 Sep 2019 10:58:16 +0000 (11:58 +0100)]
meson: fix build with Meson 0.49.2
The parser didn't seem to handle line continuations properly back then.
Olivier Crête [Fri, 2 Aug 2019 00:10:45 +0000 (20:10 -0400)]
gitlab-ci: Add MSVC builds
Uses a wrap file to download a prebuilt OpenSSL from our git as a
subproject.
Builds for both amd64 & x86
Olivier Crete [Mon, 5 Aug 2019 19:07:36 +0000 (12:07 -0700)]
tests: Don't use GSubprocessLauncher
It crashes on Windows
Olivier Crête [Fri, 2 Aug 2019 16:57:18 +0000 (12:57 -0400)]
test-fullmode-with-stun: Use portable g_usleep() for MSVC
Olivier Crête [Thu, 1 Aug 2019 23:49:35 +0000 (19:49 -0400)]
meson.build: Simplify shell based test checks
Olivier Crête [Thu, 1 Aug 2019 22:57:43 +0000 (18:57 -0400)]
meson: Add non-pkgconfig test for OpenSSL
This makes it work with Windows MSVC builds of OpenSSL which
don't include a pkg-config file.
Jordan Petridis [Thu, 18 Jul 2019 09:28:00 +0000 (12:28 +0300)]
CI: add basic msys2 build
For now it uses the gstreamer docker image and shared runner, based
on windows server 1607. The runner will migrate to windows 1807
soonish hopefully.
This commit also adds a glib wrap file needed to build
libnice.
Olivier Crête [Wed, 31 Jul 2019 23:20:27 +0000 (19:20 -0400)]
tests: Port test-fullmode-with-stun wrapper to C
Shell scripts don't work well on Windows and Python doesn't work with valgrind.
Olivier Crête [Thu, 1 Aug 2019 23:30:07 +0000 (16:30 -0700)]
stund: Add win32 support
Fabrice Bellet [Fri, 2 Aug 2019 13:19:26 +0000 (13:19 +0000)]
conncheck: fix a wrong transport value in debug
Olivier Crête [Thu, 1 Aug 2019 15:38:48 +0000 (11:38 -0400)]
dockerfiles: Add buildah format export to make gitlab happy
Olivier Crête [Wed, 31 Jul 2019 22:30:31 +0000 (18:30 -0400)]
tests: Pass GST env with valgrind too
Olivier Crête [Wed, 31 Jul 2019 20:09:25 +0000 (16:09 -0400)]
libnice.supp: Add RHEL 7 suppressions
Olivier Crête [Wed, 31 Jul 2019 20:00:27 +0000 (16:00 -0400)]
gitlab-ci: Add valgrind autotools test too
Olivier Crête [Wed, 31 Jul 2019 19:54:32 +0000 (15:54 -0400)]
dockerfiles: Add valgrind to docker files
Olivier Crête [Wed, 31 Jul 2019 19:45:16 +0000 (15:45 -0400)]
gitlab-ci: Use extends to reduce size of script
Olivier Crête [Wed, 31 Jul 2019 19:39:35 +0000 (15:39 -0400)]
gitlab-ci: Add valgrind meson test
Olivier Crête [Wed, 31 Jul 2019 19:38:28 +0000 (15:38 -0400)]
meson: Add valgrind setup
Olivier Crête [Wed, 31 Jul 2019 19:33:51 +0000 (15:33 -0400)]
libnice.supp: Add GStreamer suppressions
Olivier Crête [Wed, 31 Jul 2019 19:33:39 +0000 (15:33 -0400)]
test-gstreamer: fix buffer leak
Olivier Crête [Wed, 31 Jul 2019 19:04:21 +0000 (15:04 -0400)]
conncheck: Drop valid STUN for which we can't find a request
It's most likely caused by a retransmission received after the initial
request already had a reply.
Olivier Crête [Wed, 31 Jul 2019 18:51:16 +0000 (14:51 -0400)]
test-tcp: Only free passive socket after child
The child now calls into the parent..
Olivier Crête [Wed, 31 Jul 2019 18:51:04 +0000 (14:51 -0400)]
test-turn: Initialize variable earlier to avoid race
Olivier Crête [Wed, 31 Jul 2019 18:39:51 +0000 (14:39 -0400)]
test-udp-turn-fragmentation: Avoid leaking test socket
Olivier Crête [Wed, 31 Jul 2019 18:39:35 +0000 (14:39 -0400)]
test-different-number-streams: Avoid leaking NiceAddress
Fabrice Bellet [Mon, 15 Jul 2019 12:47:11 +0000 (14:47 +0200)]
conncheck: ignore selected pairs for nomination that failed
When evaluating the stopping criterion, failed pairs from other streams
having the "use_candidate_on_next_check" flag set should be
ignored.
This should normally not happen, because a pair selected for nomination
has no reason to fail when being rechecked, since it previously
worked... but it may happen with Skype for Business, when libnice
selects a tcp pair for component 1, the peer seems to have no interest
in the second component and lets it fail in the middle of the conncheck.
Fabrice Bellet [Fri, 12 Jul 2019 10:22:28 +0000 (12:22 +0200)]
conncheck: make the stopping criterion a bit more clear
This patch doesn't change the logic of the selection of the pair for
nomination, it makes the code a bit more simple to read.
Fabrice Bellet [Mon, 29 Jul 2019 21:35:04 +0000 (21:35 +0000)]
agent: fix a regression when updating foundations
A previous commit c1fb6f2 introduced a regression in the way the
foundation of a selected pair is updated and signaled, when the
foundation of its remote candidate changes. The previous comparison was
made on *always* identical strings, so the update of the selected pair
was *never* signaled.
Fabrice Bellet [Mon, 29 Jul 2019 21:11:55 +0000 (21:11 +0000)]
conncheck: update a misleading debug statement
We may not have received remote candidates yet, but we may have
discovered remote candidates from the early incoming checks. Only
having stream credentials is required to react to these checks.
Olivier Crête [Sun, 28 Jul 2019 15:33:16 +0000 (11:33 -0400)]
udp-bsd: Use sockaddr_storage union to ensure big enough socket
The IPv6 struct sockaddr variant is bigger than the IPv4 one,
so use the storage struct to ensure that the size is big enough.
Jakub Adam [Tue, 16 Jul 2019 09:55:32 +0000 (11:55 +0200)]
discovery: Don't start STUN/TURN disco on errorneous socket
If the initial attempt at sending discovery message returns a socket
error, don't start the retransmit timer and immediately mark such
discovery item as done. This is to quickly eliminate clearly
non-functioning items from the discovery process.
Particularly improves times to finish discovery on Windows, where
sending data from a link-local (169.254.0.0/16) IP to a destination not
on the same subnet leads to "A socket operation was attempted to an
unreachable network" error. Pointless retransmissions on those sockets
prolonged discovery in the order of seconds.
Jakub Adam [Tue, 16 Jul 2019 08:53:54 +0000 (10:53 +0200)]
udp-bsd: log source and destination addresses on GSocket send error
Fabrice Bellet [Fri, 19 Jul 2019 07:58:34 +0000 (07:58 +0000)]
agent: fix server-reflexive candidates with oc2007r2
The nomination of a pair having such a local candidate breaks SfB when
the libnice agent is behind a nat that does not do port mapping
randomization. In that case a server reflexive local candidate usually
lead to a nominated pair.
The guess made here from observing this behavior is that, it is valid to
discover and signal these local server reflexive candidates to our peer,
but they should be removed from our local candidates list thereafter, so
they do not contribute to build a valid and *even worse* a nominated
pair with the type server-reflexive. They do not appear in the conncheck
list per design anyway.
Instead, the same candidate is discovered again later during the
conncheck, with a peer-reflexive type this time, and with that type, it
just works.
Closes #90
Fabrice Bellet [Fri, 12 Jul 2019 14:06:13 +0000 (14:06 +0000)]
conncheck: avoid transport association mismatch
In some rare cases, the same address and port number may match two
remotes candidates, a tcp and an udp one, and lead to buggy pair
construction with incompatible transport. This supplementary check
prevents this problem. The matching test is not aimed to be exhaustive
but just a way to discard obviously broken associations, and fallback to
accept everything else (because socket type has a great diversity, with
socket types based on other sockets types).
It should fix #81, where such bogus transport association has been
reported (tcp-pass:udp).
Jakub Adam [Fri, 12 Jul 2019 10:15:14 +0000 (12:15 +0200)]
interfaces: ignore only interfaces we really want to
Once an interface got ignored, ALL interfaces coming after it were
dropped too.
Jakub Adam [Fri, 12 Jul 2019 08:47:50 +0000 (10:47 +0200)]
candidate: replace uint8_t -> guint8
Fixes MSVC build.