Matt Jordan [Fri, 4 Nov 2016 20:36:42 +0000 (15:36 -0500)]
main/bridge: Add some verbose logging for video source changes
It's actually quite useful to see the source of a video stream change.
This doesn't happen terribly often, even with talk detection - but when
it does, it's nice to know which channel is now providing your video
stream.
As a verbose 5 level message, it shouldn't be terribly spammy or costly
to have, and is 'lower level' then most other verbose messages that the
bridge system emits.
Matt Jordan [Fri, 4 Nov 2016 20:33:35 +0000 (15:33 -0500)]
bridges/bridge_softmix: Remove SSRC changes on join/leave; update video source
WebRTC clients really, really want to know the SSRC of the media they're
getting. Changing the SSRC is generally not a good thing.
bridge_softmix, starting in Asterisk 12, started changing the SSRC of
parties as they joined or left the bridge. With most phones, this isn't
a problem: phones just play back the stream they're getting. With WebRTC
clients, however, the SSRC is tied to a media stream that may be
negotiated. When a new SSRC just shows up, the media can be dropped.
As it turns out, the SSRC change shouldn't even be necessary. From the
perspective of the client, it's still talking to Asterisk with the same
media stream: why indicate that the far party has suddenly changed to a
different source of media?
This patch opts to just remove the SSRC changes. With this patch, video
clients that join/leave a softmix bridge actually get the video stream
instead of freaking out.
Corey Farrell [Sun, 30 Oct 2016 18:33:12 +0000 (14:33 -0400)]
vector: Prevent NULL argument to memcpy.
Headers declare that memcpy does not accept NULL argument for the first
two parameters. Add a conditional block to prevent memcpy and ast_free
from running on vectors with NULL element array.
Corey Farrell [Sat, 29 Oct 2016 15:19:53 +0000 (11:19 -0400)]
astobj2: Declare private variable data_size for AO2_DEBUG only.
Every ao2 object contains storage for a private variable data_size,
though the value is never read if AO2_DEBUG is disabled. This change
makes the variable conditional, reducing memory usage.
The patch below removes a cached DNS response from the hash table when
another thread is referencing the old entry. The table still contained
the entry when it was destroyed which can result in inexplicable crashes.
George Joseph [Fri, 28 Oct 2016 21:59:19 +0000 (15:59 -0600)]
pjproject_bundled: Fix issue where "/version.mak" wasn't found
main/Makefile includes third-party/pjproject/build.mak but
doesn't set PJDIR beforehand so "include $(PJDIR)/version.mak"
evaluates to "/version.mak". Fix is to set PJDIR in main/Makefile
before the include.
Corey Farrell [Fri, 28 Oct 2016 02:49:43 +0000 (22:49 -0400)]
Fix shutdown crash caused by modules being left open.
It is only safe to run ast_register_cleanup callbacks when all modules
have been unloaded. Previously these callbacks were run during graceful
shutdown, making it possible to crash during shutdown.
George Joseph [Wed, 26 Oct 2016 23:48:24 +0000 (17:48 -0600)]
pjproject_bundled: Remove usage of tar's --strip-components option
Older versions of tar don't support the --strip-components option so
instead of doing 'tar --strip-components=1 -C source', we now just
untar to the tarball's root directory (pjproject-<version>) and
rename that directory to 'source'.
Also fixed an issue where the pjproject source directory is a hard
coded absolute pathname.
Joshua Colp [Thu, 27 Oct 2016 13:07:02 +0000 (13:07 +0000)]
res_pjsip_caller_id: Fix crash on session timers UPDATE on inbound calls.
The res_pjsip_caller_id module wrongly assumed that a
saved From header would always exist on sessions. This
is true until an inbound call is received and a session
timer causes an UPDATE to be sent. In this case there will
be no saved From header and a crash will occur. This change
makes it fall back to the From header of the outgoing request
if no saved From header is present.
Joshua Colp [Wed, 26 Oct 2016 12:51:50 +0000 (12:51 +0000)]
app_voicemail: Clear voice mailbox in MailboxExists and MAILBOX_EXISTS.
When executing the MailboxExists dialplan application and
MAILBOX_EXISTS dialplan function the passed in temporary voice
mailbox was not cleared, causing it to try to free garbage.
Joshua Colp [Sun, 23 Oct 2016 12:38:59 +0000 (12:38 +0000)]
pjsip: Fix a few media bugs with reinvites and asymmetric payloads.
When channel format changes occurred as a result of an RTP
re-negotiation the bridge was not informed this had happened.
As a result the bridge technology was not re-evaluated and the
channel may have been in a bridge technology that was incompatible
with its formats. The bridge is now unbridged and the technology
re-evaluated when this occurs.
The chan_pjsip module also allowed asymmetric codecs for sending
and receiving. This did not work with all devices and caused one
way audio problems. The default has been changed to NOT do this
but to match the sending codec to the receiving codec. For users
who want asymmetric codecs an option has been added, asymmetric_rtp_codec,
which will return chan_pjsip to the previous behavior.
The codecs returned by the chan_pjsip module when queried by
the bridge_native_rtp module were also not reflective of the
actual negotiated codecs. The nativeformats are now returned as
they reflect the actual negotiated codecs.
George Joseph [Tue, 25 Oct 2016 16:20:16 +0000 (10:20 -0600)]
test_astobj2_thrash: Fix multithreaded issues
The test uses 4 threads to grow, count, lookup and shrink 15K objects
in a container. If there's only 1 execution engine available, the test
will complete in <50ms. If each threads gets its own execution engine,
the test may timeout after 60 seconds because the count thread does a
locked ao2_callback on the whole container in a tight loop with only
a sched_yield to give up time. The lock contention makes the test
execution times wildly variable and mostly timeout. 2 execution
engines are OK, 3 results in about 33% failure rate and >=4 causes
a 80% failure rate.
To fix, the sched_yield was changed to a usleep(500).
Also, the number of buckets specified for the container was an even
number so that was changed to the next prime number greater than
(MAX_HASH_ENTRIES / 100). That's 151 currently.
George Joseph [Mon, 24 Oct 2016 15:55:23 +0000 (09:55 -0600)]
pjproject_bundled: Fixed various build issues
* CFLAGS is now properly set when using older gcc.
* All third-party pjproject targets have been removed. This fixes
an issue with older libsrtp in some distros.
* Manually removing the source directory now causes a rebuild.
* EXTERNALS_CACHE_DIR is now properly checked.
* Whitespace fixes.
This change adds support for dual stack automatically. No
configuration is required and the IP address and version
in the SIP messages and SDP will be automatically changed
based on the transport over which the message is being
sent. RTP usage has also been changed to listen on both
IPv4 and IPv6 simultaneously to allow media to flow, and
to allow ICE support on both simultaneously. This also
allows failover between IPv6 and IPv4 to work as expected.
Mark Michelson [Wed, 19 Oct 2016 17:05:28 +0000 (12:05 -0500)]
ARI: Add duplicate channel ID checking for channel creation.
This is similar to what is done for origination, but for the 14 and up
channel creation method. When attempting to create a channel, if a
channel ID is specified and a channel already exists with that ID, then
a 409 is returned.
Mark Michelson [Mon, 17 Oct 2016 19:18:57 +0000 (14:18 -0500)]
ARI: Detect duplicate channel IDs
ARI and AMI allow for an explicit channel ID to be specified
when originating channels. Unfortunately, there is nothing in
place to prevent someone from using the same ID for multiple
channels. Further complicating things, adding ID validation to channel
allocation makes it impossible for ARI to discern why channel allocation
failed, resulting in a vague error code being returned.
The fix for this is to institute a new method for channel errors to be
discerned. The method mirrors errno, in that when an error occurs, the
caller can consult the channel errno value to determine what the error
was. This initial iteration of the feature only introduces "unknown" and
"channel ID exists" errors. However, it's possible to add more errors as
needed.
ARI uses this feature to determine why channel allocation failed and can
return a 409 error during origination to show that a channel with the
given ID already exists.
Michael Walton [Tue, 4 Oct 2016 23:24:54 +0000 (12:24 +1300)]
res_rtp_asterisk: Add ice_blacklist option
Introduces ice_blacklist configuration in rtp.conf. Subnets listed in the
form ice_blacklist = <subnet spec>, e.g. ice_blacklist =
192.168.1.0/255.255.255.0, are excluded from ICE host, srflx and relay
discovery. This is useful for optimizing the ICE process where a system
has multiple host address ranges and/or physical interfaces and certain
of them are not expected to be used for RTP. Multiple ice_blacklist
configuration lines may be used. If left unconfigured, all discovered
host addresses are used, as per previous behavior.
Mark Michelson [Tue, 18 Oct 2016 21:30:17 +0000 (16:30 -0500)]
CDR: Alter destruction pattern for CDR chains.
CDRs form chains. When the root of the chain is destroyed, it then
unreferences the next CDR in the chain. That CDR is destroyed, and it
then unreferences the next CDR in the chain. This repeats until the end
of the chain is reached. While this typically does not cause any sort of
problems, it is possible in strange scenarios for the CDR chain to grow
way longer than expected. In such a scenario, the destruction pattern
can result in a stack overflow.
This patch fixes the problem by switching from a recursive pattern to an
iterative pattern for destruction. When the root CDR is destroyed, it is
responsible for iterating over the rest of the CDRs and unreferencing
each one. Other CDRs in the chain, since they are not the root, will
simply destroy themselves and be done. This causes the stack depth not
to increase.
Alexander Traud [Tue, 18 Oct 2016 08:01:47 +0000 (10:01 +0200)]
cli: Auto-complete File not Module for core set debug.
Since Asterisk 1.8, the command "core set debug" on the command-line interface
asks not for a file (.c) but a module name. This change shows modules (.so) on
the auto-completion via a tabulator or the question mark. Now, when you
partially type a module name, TAB or ?, you get the correct candidiates.
Leandro Dardini [Thu, 13 Oct 2016 19:09:18 +0000 (21:09 +0200)]
app_queue: Added initialization for "context" parameter
When using Asterisk Realtime Architecture, empty fields are skipped and the
default values are used. If the "context" parameter in queue was set and then
cleared from the database, the old value remains in memory and it continues
to be used. This change initialize the "context" parameter with an empty value,
allowing clearing the parameter.
Matt Jordan [Sun, 16 Oct 2016 01:05:05 +0000 (20:05 -0500)]
res/ari: Add the Asterisk EID field to outgoing events
This patch adds the Asterisk EID field to all outgoing ARI events.
Because this field should be added to all events as they are
transmitted, it is appended to the JSON message just prior to it being
handed off to the application message handler. This makes it somewhat
resilient to both new events being added to ARI, as well as other
potential event transport mechanisms.
George Joseph [Sun, 16 Oct 2016 22:25:35 +0000 (16:25 -0600)]
utils.c: Fix ast_set_default_eid for multiple platforms
ast_set_default_eid was searching for ethX, emX, enoX, ensX and even
pciD#U interface names. While this was a good attempt, it wasn't
inclusive enough to capture interfaces like enp6s0 or ens6d1, etc.
Rather than relying on interface names, we now simply find the first
interface returned by the OS that has a hardware address and that
address isn't all 0x00 or all 0xff. The code IS different for BSD,
Solaris and Linux based on what method is available for enumerating
interfaces.
Tested on:
FreeBSD9
CentOS6
Ubuntu14
Fedora24
I was unable to test on Solaris at this time but the code for Solaris
is used elsewhere at Digium.
Michael Kuron [Sat, 15 Oct 2016 09:58:05 +0000 (11:58 +0200)]
chan_sip: Only send video on outgoing channel if incoming channel supports it
Previously, the settings videosupport=always and videosupport=yes behaved
identically and unconditionally caused a video offer to be sent in the SDP on
an outgoing call. This was a regression introduced with commit 5a1d90e1fbfc4b48927aad55311f3b38efbf1f54 in Asterisk 1.6.1.
This commit restores correct behavior: videosupport=always causes a video offer
to be sent unconditionally, while videosupport=yes will only offer video on an
outbound channel if the incoming channel it is bridged to also supports video.
That way, the device receiving the outgoing call can display the correct user
interface elements for audio or video and will not unnecessarily show a blank
video window on an audio-only call.
Corey Farrell [Fri, 14 Oct 2016 05:18:50 +0000 (01:18 -0400)]
Fix issues with bundled pjproject cached download.
Previously when testing I had a preexisting makeopts in ASTTOPDIR. The
ordering of configure.ac causes --with-externals-cache to be processed
after third-party configure. In cases where the Asterisk clone is
cleaned it would cause pjproject to be downloaded to /tmp. This
moves processing of the externals cache and sounds cache to happen
before third-party configure.
This also addresses a possible issue with the third-party Makefile. If
TMPDIR is set by the environment it would override the path given to
--with-externals-cache.
Richard Mudgett [Wed, 12 Oct 2016 21:24:14 +0000 (16:24 -0500)]
Audit ast_json_pack() calls for needed UTF-8 checks.
Added needed UTF-8 checks before constructing json objects in various
files for strings obtained outside the system. In this case string values
from a channel driver's peer and not from the user setting channel
variables.
* aoc.c: Fixed type mismatch in s_to_json() for time and granularity json
object construction.
George Joseph [Wed, 12 Oct 2016 21:30:40 +0000 (15:30 -0600)]
res_config_mysql: Fix several issues related to recent table changes
Unlike any of the other database drivers, res_config_mysql checks that
the table definition matches the requirements for every insert and
update statement. Since all requirements are forced to 'char', any
column that isn't a char, like ps_contacts' expiration_time,
qualify_timeout, etc., will throw a warning. It's kinda harmless but
very misleading. Since no other driver does those checks on insert
or update, they've been removed from res_config_mysql. Also, all
the logic that actually attempted to ALTER the table to fix the issue
has been removed. With the move to alembic, the auto-alter
functionality is not only unnecessary, it's also dangerous.
The other issue is that res_config_mysql calls the mysql_insert_id
function inside store_mysql. Presumably the intention was to return
the number of rows inserted DESPITE A NOTE IN THE CODE THAT THE VALUE
IS NON_PORTABLE AND MAY CHANGE. That value is then returned to
config realtime as the number of rows inserted. Guess what? The value
changed. It now only returns the number of rows inserted if there's an
auto increment column on the table, which ps_contacts doesn't have.
Otherwise it returns 0. So now, the insert worked but we tell config
realtime and sorcery that no rows were inserted. That call to
mysql_insert_id was removed and we now always return 1 if the insert
succeeded. We're only inserting 1 row at a time anyway. If the insert
fails, we still return -1.
ASTERISK-26362 #close Reported-by: Carlos Chavez
Change-Id: I83ce633efdb477b03c8399946994ee16fefceaf4
res_fax: Fix a tight race condition causing fax to crash in audio fallback
When T.38 gets rejected and G711 failback occurs there is a period of
time where neither AST_FAX_TECH_T38 nor AST_FAX_TECH_AUDIO is set,
leading to a crash.
Alexander Traud [Tue, 11 Oct 2016 11:55:13 +0000 (13:55 +0200)]
chan_sip: Support nat=auto_comedia or nat=force_rport,auto_comedia.
In the SIP channel driver chan_sip, auto_comedia was expected to be used in
tandem with auto_force_rport. Or stated differently: Only when auto_force_rport
was chosen (the default), auto_comedia worked. This change allows auto_comedia
to be set independently of the state of (auto_)force_rport. For example,
nat=force_rport,auto_comedia is useful for IPv4/IPv6 Dual Stack deployments
when IPv6 clients are behind a Firewall.
Michael Walton [Wed, 5 Oct 2016 01:46:17 +0000 (14:46 +1300)]
audiohooks: Remove redundant codec translations when using audiohooks
The main frame read and write handlers in main/channel.c don't use the
optimum placement in the processing flow for calling audiohooks
callbacks, as far as codec translation is concerned. This change places
the audiohooks callback code:
* After the channel read translation if the frame is not linear before
the translation, thereby increasing the chance that the frame is linear
as required by audiohooks
* Before the channel write translation if the frame is linear at this
point
This prevents the audiohooks code from instantiating additional
translation paths to/from linear where a linear frame format is already
available, saving valuable CPU cycles
res_pjsip_config_wizard: Memory leak in module_unload
Fixed a memory leak. It removes only the first element.
Added a useful feature in vector.h to remove all items
under the CMP through a callback function / macro.
George Joseph [Mon, 3 Oct 2016 16:30:43 +0000 (10:30 -0600)]
bundled_pjproject: Add tests for programs used by the Makefile, et al.
Added tests for bzip2, tar, patch, sed and nm to configure.ac.
Set DOWNLOAD_TO_STDOUT to a working command line regardless of
whether the download program is wget, curl or fetch.
Added a 'configure.m4' file to the third-party directory which takes
care of calling any third-party project setup. Had to move some
pjproject_bundled stuff up in configure.ac so it was called before
the third-party configure macro.
The pjproject tarball is now downloaded to the externals_cache_dir if
it was specified on the ./configure command line
Removed regeneration of the pjproject aconfigure file. It was only
needed for an old patch that no longer applies.
Converted the tests for symbols to explicit tests since we know that
they're now available in the bundled version. Saves a little time
during configure.
Corey Farrell [Mon, 10 Oct 2016 02:28:52 +0000 (22:28 -0400)]
logger: Prevent output of verbose messages initiated from rasterisk.
Remote asterisk consoles should only display verbose log messages
created by the daemon. The first patch for ASTERISK-26410 caused
a couple verbose messages to be printed when the rasterisk process
ended.
George Joseph [Wed, 5 Oct 2016 19:53:10 +0000 (13:53 -0600)]
pjproject_bundled: Add MALLOC_DEBUG capability
pjproject_bundled will now use the asterisk memory debugging APIs
if MALLOC_DEBUG is turned on in menuselect.
Because this required stubs for the executable programs and the python
bindings, some Makefile reorganization was needed to properly handle
the dependencies. As a result, the makefile now individually makes
each of the pjproject libraries separately instead of making them all
in 1 shot. The only visible change is that there are separate status
lines printed for each library instead oif 1 for all libs. Also, the
making of the pjproject dependency files was eliminated. They're not
needed for building unless you're actively modifying pjproject source
files and it makes the build process faster. Finally, any issues with
parallel builds should be resolved again making the build faster.
George Joseph [Tue, 4 Oct 2016 21:59:54 +0000 (15:59 -0600)]
alembic: Allow cdr, config and voicemail to exist in the same schema
cdr, config and voicemail are all separate alembic trees. Because
alembic's default is to use a table named 'alembic_version' to store
the current tree revision, the 3 trees can't exist in the same schema
without stepping on each other.
Now each tree uses 'alembic_version_<tree_name>' as the version table.
Each tree's env.py script now first checks for 'alembic_version'. If
it finds it AND its revision is in the tree's history, the script
renames it to 'alembic_version_<tree_name>'. Regardless, the script
then continues with the migration using 'alembic_version_<tree_name>'
and creates that table if it's not found. The result is that if an
existing 'alembic_version' table was found but it didn't belong to this
tree, it's left alone and 'alembic_version_<tree_name>' is used or
created.
WARNING: If multiple trees are using the same schema, they MUST NOT
CRU or D any objects with names that might exist in the other trees.
An example would be 'yesno_values' type. If two trees perform
operations on it, one tree could pull it out from under the other.
Thankfully we currently don't share any names among cdr, config and
voicemail.
NOTE: Since the env.py scripts in each tree were identical, a common
env.py has been placed in the ast-db-manage directory and a symlink
to it has been placed in each tree directory.
ASTERISK-24311 #close Reported-by: Dafi Ni
Change-Id: I4d593f000350deb5d21a14fa1e9bc3896844d898
* Compile __ast_assert_failed unconditionally.
* Use __ast_assert_failed to log messages from log_bad_ao2
* Remove calls to ast_assert(0) that happen after log_bad_ao2 was run.
Alexander Traud [Wed, 5 Oct 2016 09:25:11 +0000 (11:25 +0200)]
chan_sip: Honor support of Symmetric Response (rport) for SIP requests.
In the SIP channel driver chan_sip, the default is "auto_force_rport". When no
NAT was detected, for example in case of IPv6, Asterisk uses the IP address
from the headers within the SIP-REGISTER for subsequent SIP signaling. When
the remote party specifies support for Symmetric Response (RFC 3581) via the
parameter "rport", Asterisk should not extract the port from the SIP headers
but reuse the port of the transport. This did not happen because of a typo.
Verbose messages should be printed to the console if the sublevel is
less than option_verbose. This fix ensures the welcome message with
copyright and license are printed at daemon and interactive rasterisk
startup.
projects / thirdparty / asterisk.git / log
