Joshua Colp [Thu, 24 Aug 2017 16:45:08 +0000 (13:45 -0300)]
core: Reduce video update queueing.
A video update frame is used to indicate that a channel
with video negotiated should provide a full frame so the
decoder decoding the stream is able to do so. In situations
where a queue is used to store frames it makes no sense
for the queue to contain multiple video update frames. One
is sufficient to have a full frame be sent.
Sean Bright [Fri, 25 Aug 2017 18:44:35 +0000 (14:44 -0400)]
voicemail: Fix various abuses of mkstemp
mkstemp() returns a unique filename, but appending an extension to that
filename does not guarantee uniqueness. Instead, use mkdtemp() and we
can put whatever extension we want on the files that we create inside
the directory.
In the case of app_minivm, we also now properly clean up any temporary
files that we create.
Sean Bright [Fri, 25 Aug 2017 17:20:16 +0000 (13:20 -0400)]
app_record: Resolve some absolute vs. relative filename bugs
If the Record() application is called with a relative filename that
includes directories, we were not properly creating the intermediate
directories and Record() would fail.
Secondarily, updated the documentation for RECORDED_FILE to mention
that it does not include a filename extension.
Finally, rewrote the '%d' functionality to be a bit more straight
forward and less noisy.
Torrey Searle [Mon, 21 Aug 2017 09:28:52 +0000 (11:28 +0200)]
res/res_pjsip_session: allow SDP answer to be regenerated
If an SDP answer hasn't been sent yet, it's legal to change it.
This is required for PJSIP_DTMF_MODE to work correctly, and can
also have use in the future for updating codecs too.
Richard Mudgett [Fri, 18 Aug 2017 22:37:12 +0000 (17:37 -0500)]
res_pjsip_session.c: Fix crash when declining an active stream.
If a previously active stream is declined we could crash because the
channel's thread is still using the stream while we are updating the
topology in the serializer thread.
* Defer removing any declined stream's handler until we have blocked the
channel's thread with the channel lock.
Richard Mudgett [Wed, 16 Aug 2017 22:50:18 +0000 (17:50 -0500)]
bridge_channel.c: Fix FRACK when mapping frames to the bridge.
* Add protection checks when mapping streams to the bridge. The channel
and bridge may be in the process of updating the stream mapping when a
media frame comes in so we may not be able to map the frame at the time.
* We need to map the streams to the bridge's stream numbers right before
they are written into the bridge. That way we don't have to keep
locking/unlocking the bridge and we won't have any synchronization
problems before the frames actually go into the bridge.
* Protect the deferred queue with the bridge_channel lock.
Richard Mudgett [Wed, 16 Aug 2017 20:22:04 +0000 (15:22 -0500)]
bridge: Fix softmix bridge deadlock.
* Fix deadlock in
bridge_softmix.c:softmix_bridge_stream_topology_changed() between
bridge_channel and channel locks.
* The new bridge technology topology change callbacks must be called with
the bridge locked. The callback references the bridge channel list, the
bridge technology could change, and the bridge stream mapping is updated.
Michael Kuron [Sun, 20 Aug 2017 13:15:37 +0000 (15:15 +0200)]
res_xmpp: fix inverted return code check in OAuth
fetch_access_token calls func_curl via ast_func_read. The latter returns 0 upon
success and -1 if the function is not available.
This commit inverts the return code check so that an error is printed if the
module is not loaded and not if it is loaded.
George Joseph [Wed, 16 Aug 2017 20:43:10 +0000 (14:43 -0600)]
Fix downloader not working with curl
The codec/dpma downloader wasn't handling curl correctly. The logic
that transforms makeopts into a bash-sourceable file wasn't
handling the make 'or' command in DOWNLOAD_TIMEOUT so bash was
looking for an 'or' command.
That logic has been eliminated. Instead of trying to transform
and source makeopts, the downloader now calls a make scriptlet
to print the value of a specific variable. This way, make handles
the ors (or any other make construct that happens to creep into
that file).
Kevin Harwell [Tue, 15 Aug 2017 18:12:10 +0000 (13:12 -0500)]
manager: hook event is not being raised
When the iostream code went in it introduced a conditional that made it so the
hook event was not being raised even if a hook is present. This patch adds a
check to see if a hook is present in astman_append. If so then call into the
send_string function, which in turn raises the even for specified hook.
Also updated the ami hooks unit test, so the test could be automated.
Richard Mudgett [Tue, 15 Aug 2017 20:15:58 +0000 (15:15 -0500)]
configure: Check cache for valid pjproject tarball before downloading.
On a fresh Asterisk source directory, the bundled pjproject tarball is
unconditionally downloaded even if the tarball is already in a specified
cache directory.
* Made check if the pjproject tarball is valid in the cache directory
before downloading the tarball on a fresh source directory.
Richard Mudgett [Tue, 15 Aug 2017 16:14:20 +0000 (11:14 -0500)]
res_pjsip: Fix prune_on_boot to remove only contacts for the host.
* Check that the contact's reg_server matches the host's name before
deleting any prune_on_boot contacts. We don't want to delete reliable
transport contacts made with other servers if the ps_contacts database
table is shared with other servers.
Thanks to Ross Beer for pointing out that the original prune logic would
delete reliable transport contacts from other servers.
Richard Mudgett [Thu, 10 Aug 2017 19:18:01 +0000 (14:18 -0500)]
STUN/netsock2: Fix some valgrind uninitialized memory findings.
* netsock2.c: Test the addr->len member first as it may be the only member
initialized in the struct.
* stun.c:ast_stun_handle_packet(): The combinded[] local array could get
used uninitialized by ast_stun_request(). The uninitialized string gets
copied to another location and could overflow the destination memory
buffer.
These valgrind findings were found for ASTERISK_27150 but are not
necessarily a fix for the issue.
Richard Mudgett [Wed, 2 Aug 2017 23:44:12 +0000 (18:44 -0500)]
res_pjsip_outbound_registration.c: Re-REGISTER on transport shutdown.
The fix for the issue is broken up into three parts.
This is part three which handles the client side of REGISTER requests.
The registered contact may no longer be valid on the server when the
transport used is reliable and the connection is broken.
* Re-REGISTER our contact if the reliable transport is broken after
registration completes. We attempt to re-REGISTER immediately to minimize
the time we are unreachable. Time may have already passed between the
connection being broken and the loss being detected.
* Reorder sip_outbound_registration_state_alloc() so the STATSD_GUAGE's
are still correct if an allocation failure happens.
Richard Mudgett [Mon, 31 Jul 2017 19:21:06 +0000 (14:21 -0500)]
res_pjsip: Remove ephemeral registered contacts on transport shutdown.
The fix for the issue is broken up into three parts.
This is part two which handles the server side of REGISTER requests when
rewrite_contact is enabled. Any registered reliable transport contact
becomes invalid when the transport connection becomes disconnected.
* Monitor the rewrite_contact's reliable transport REGISTER contact for
shutdown. If it is shutdown then the contact must be removed because it
is no longer valid. Otherwise, when the client attempts to re-REGISTER it
may be blocked because the invalid contact is there. Also if we try to
send a call to the endpoint using the invalid contact then the endpoint is
not likely to see the request. The endpoint either won't be listening on
that port for new connections or a NAT/firewall will block it.
* Prune any rewrite_contact's registered reliable transport contacts on
boot. The reliable transport no longer exists so the contact is invalid.
* Websockets always rewrite the REGISTER contact address and the transport
needs to be monitored for shutdown.
* Made the websocket transport set a unique name since that is what we use
as the ao2 container key. Otherwise, we would not know which transport we
find when one of them shuts down. The names are also used for PJPROJECT
debug logging.
* Made the websocket transport post the PJSIP_TP_STATE_CONNECTED state
event. Now the global keep_alive_interval option, initially idle shutdown
timer, and the server REGISTER contact monitor can work on wetsocket
transports.
* Made the websocket transport set the PJSIP_TP_DIR_INCOMING direction.
Now initially idle websockets will automatically shutdown.
Richard Mudgett [Fri, 28 Jul 2017 23:26:17 +0000 (18:26 -0500)]
res_pjsip: PJSIP Transport state monitor refactor.
The fix for the issue is broken up into three parts.
This is part one which refactors the transport state monitor code to allow
more modules to be able to monitor transports.
* Pull the management of PJPROJECT's transport state callback code from
res_pjsip_transport_management.c into res_pjsip. Now other modules can
dynamically add and remove themselves from transport monitoring without
worrying about breaking PJPROJECT's callback chain.
* Add the ability for other modules to get a callback whenever a specific
transport is shutdown.
When handling an incoming SIP MESSAGE, PJSIP
attaches the IP address that the message was
received from to the message in the variable
PJSIP_RECVADDR. When the IP address is IPv6
the :PORT appended results in an unparseable
mess. By using an additional bit flag on the
pj_sockaddr_print call, the conventional use
of brackets around the address is achieved.
res_rtp_asterisk: enable rtcp & QOS stats on native bridge
Asterisk wasn't generating or forwarding RTCP packets when native
bridge was activated. Also the stats weren't available via
CHANNEL(qos). Now the RTCP stats are always calculated.
res_rtp_asterisk: Make P2P bridge Asymmetric codec aware
Introduce a new property to rtp-engine to make it aware of
the desire for assymetric codecs or not. If asymmetric codecs
is not allowed, the bridge will compare read/write formats
and shut down the p2p bridge if needed
George Joseph [Tue, 8 Aug 2017 18:33:50 +0000 (12:33 -0600)]
Make --with-pjproject-bundled the default for Asterisk 15
'--with-pjproject-bundled' is now the default when running
./configure. It can be disabled with '--without-pjproject-bundled'.
To make building without an internet connection easier, a new
./configure option '--with-download-cache' was added that sets
the cache for externals (like pjproject, the codecs and the DPMA),
AND the sounds files. It can also be specified as an environment
variable named "AST_DOWNLOAD_CACHE". The existing
'--with-sounds-cache' option / SOUNDS_CACHE_DIR env variable and
'--with-externals-cache' option / EXTERNALS_CACHE_DIR env variable
remain and if specified, will override '--with-downloads-cache'.
Joshua Colp [Sat, 5 Aug 2017 11:36:49 +0000 (11:36 +0000)]
res_pjsip_session: Release media resources on session end quicker.
A change was made long ago where the session was kept around
until the underlying INVITE session had been destroyed. This
had the side effect of also keeping the underlying media resources
around for this time as well.
This change ensures that when we are told to terminate the
session we immediately release any media sessions associated
with it.
Joshua Colp [Sun, 6 Aug 2017 16:15:34 +0000 (16:15 +0000)]
bridge: Fix stream topology/participant locking and video misrouting.
This change fixes a few locking issues and some video misrouting.
1. When accessing the stream topology of a channel the channel lock
must be held to guarantee the topology remains valid.
2. When a channel was joined to a bridge the bridge specific
implementation for stream mapping was not invoked, causing video
to be misrouted for a brief period of time.
Kevin Harwell [Fri, 4 Aug 2017 21:47:30 +0000 (16:47 -0500)]
res_pjsip_session/_sdp_rtp: Handling of 'msid' is incorrect
Currently, the handling of the msid attribute is not quite right. According to
the spec the msid's between the offer/answer are not dependent upon one another.
Meaning the same msid's given in an offer do not have to be returned in the
answer for a given stream. And they probably shouldn't be (copied/reused) since
this can potentially cause some browser side confusion.
This patch generates new msids when both an offer and answer are sent from
Asterisk. However, Asterisk does reuse the original msid it sent out for a
reinvite. Also audio+video streams are paired together by sharing the same
stream id, but a different track id.
Corey Farrell [Fri, 4 Aug 2017 01:58:25 +0000 (21:58 -0400)]
Correct some leaks in unit tests.
* chan_sip: channel in test_sip_rtpqos_1.
* test_config: config hook, config info and global config holder.
* test_core_format: format in format_attribute_set_without_interface.
* test_stream: unneeded frame duplication.
* test_taskprocessor: task_data.
When the "webrtc" option was added in res_pjsip it was not added to the alembic
scripts. This patch adds the option for alembic.
Also, changed the sorcery configuration type to an OPT_YESNO_T value instead of
an OPT_BOOL_T so if this field is ever written to a database it will write out
the correct value.
res_pjsip_session: Release media resources on session end quicker.
A change was made long ago where the session was kept around
until the underlying INVITE session had been destroyed. This
had the side effect of also keeping the underlying media resources
around for this time as well.
This change ensures that when we know the session is ending we
release the media resources immediately.
Sean Bright [Wed, 26 Jul 2017 13:48:29 +0000 (09:48 -0400)]
res_pjsip_pidf_eyebeam_body_supplement: Correct status presentation
This change fixes PIDF content generation when the underlying device
state is considered in use. Previously it was incorrectly marked
as closed meaning they were offline/unavailable. The code now
correctly marks them as open.
Additionally:
* Generate an XML element for our activity instead of a using a text
node.
* Consider every extension state other than "unavailable" to be 'open'
status.
* Update the XML namespaces and structure to reflect those
documented in RFC 4480
* Use 'on-the-phone' (defined in RFC 4880) instead of 'busy' as the
"in use" activity. This change results in eyeBeam using the
appropriate icon for the watched user.
This was tested on eyeBeam 1.5.20.2 build 59030 on Windows.
ASTERISK-26659 #close
Reported by: Abraham Liebsch
patches:
ASTERISK-26659.diff submitted by snuffy (license 5024)
res_pjsip: Add support for dnsmgr to external_media_address.
The "external_media_address" option on transports is now
resolved using dnsmgr. This allows it to be automatically
refreshed regularly if refreshes are enabled in dnsmgr.
If the system is using a dynamic IP address a dynamic DNS
hostname can be provided to keep the IP address up to
date.
GCC 7 has added capability to produce warnings, this fixes most of those
warnings. The specific warnings are disabled in a few places:
* app_voicemail.c: truncation of paths more than 4096 chars in many places.
* chan_mgcp.c: callid truncated to 80 chars.
* cdr.c: two userfields are combined to cdr copy, fix would break ABI.
* tcptls.c: ignore use of deprecated method SSLv3_client_method().
Sean Bright [Wed, 26 Jul 2017 14:27:00 +0000 (10:27 -0400)]
app_queue: Add announce-position-only-up option
Setting this option will cause the Queue application to only announce
the caller's position if it has improved since the last time that we
announced it.
George Joseph [Thu, 27 Jul 2017 11:35:51 +0000 (05:35 -0600)]
bundled_pjproject: Improve SSL/TLS error handling
OpenSSL has 2 levels or error processing. It's possible for the
top layer to return SSL_ERROR_SYSCALL but the lower layer return
no error, in which case processing should continue. Only the top
layer was being examined though so connections were being torn
down when they didn't need to be. This patch adds the examination
of the lower level codes, and if they return no errors, allows
processing to continue.
ASTERISK-27001 Reported-by: Ian Gilmour
patches:
pjproject-2.6.patch submitted by Ian Gilmour (license 6889)
Updated-by: George Joseph and Sauw Ming (Teluu)
Merged to upstream pjproject on 7/27/2017 (commit 5631)
bridge_softmix / res_rtp_asterisk: Fix packet loss and renegotiation issues.
This change does a few things to improve packet loss and renegotiation:
1. On outgoing RTP streams we will now properly reflect out of order
packets and packet loss in the sequence number. This allows the
remote jitterbuffer to better reorder things.
2. Video updates can now be discarded for a period of time
after one has been sent to prevent flooding of clients.
3. For declined and removed streams we will now release any
media session resources associated with them. This was not
previously done and caused an issue where old state was being
used for a new stream.
4. RTP bundling was not actually removing bundled RTP instances
from the parent. This has been resolved by removing based on
the RTP instance itself and not the SSRC.
5. The code did not properly handle explicitly unbundling an
RTP instance from its parent. This now works as expected.
projects / thirdparty / asterisk.git / log
