4 years agores_config_mysql: Fix several issues related to recent table changes
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

4 years agoMerge "Binaural synthesis (confbridge): interleaved two-channel audio."
zuul [Wed, 12 Oct 2016 16:36:06 +0000 (11:36 -0500)]
Merge "Binaural synthesis (confbridge): interleaved two-channel audio."

4 years agoMerge "bundled_pjproject: Add tests for programs used by the Makefile, et al."
zuul [Wed, 12 Oct 2016 16:04:53 +0000 (11:04 -0500)]
Merge "bundled_pjproject:  Add tests for programs used by the Makefile, et al."

4 years agores_fax: Fix a tight race condition causing fax to crash in audio fallback
Torrey Searle [Thu, 29 Sep 2016 18:08:07 +0000 (20:08 +0200)]
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.

Change-Id: Icc3f457b2292d48a9d7843dac0028347420cc982

4 years agoMerge "Add text of cdr directory into for ast-db-manage"
zuul [Wed, 12 Oct 2016 00:45:14 +0000 (19:45 -0500)]
Merge "Add text of cdr directory into for ast-db-manage"

4 years agoMerge "res_calendar: Add support for fetching calendars when reloading"
zuul [Wed, 12 Oct 2016 00:22:24 +0000 (19:22 -0500)]
Merge "res_calendar: Add support for fetching calendars when reloading"

4 years agoMerge "audiohooks: Remove redundant codec translations when using audiohooks"
zuul [Tue, 11 Oct 2016 22:45:56 +0000 (17:45 -0500)]
Merge "audiohooks: Remove redundant codec translations when using audiohooks"

4 years agoMerge "vector: After remove element recheck index"
zuul [Tue, 11 Oct 2016 21:41:33 +0000 (16:41 -0500)]
Merge "vector: After remove element recheck index"

4 years agoMerge "app_dial: Add the "Q" option to set the cause on unanswered channels"
zuul [Tue, 11 Oct 2016 20:15:56 +0000 (15:15 -0500)]
Merge "app_dial:  Add the "Q" option to set the cause on unanswered channels"

4 years agoMerge "logger: Prevent output of verbose messages initiated from rasterisk."
zuul [Tue, 11 Oct 2016 18:57:56 +0000 (13:57 -0500)]
Merge "logger: Prevent output of verbose messages initiated from rasterisk."

4 years agoapp_dial: Add the "Q" option to set the cause on unanswered channels
George Joseph [Thu, 6 Oct 2016 14:58:26 +0000 (08:58 -0600)]
app_dial:  Add the "Q" option to set the cause on unanswered channels

The "Q" option will set the cause on the unanswered channels when
another channel answers.  It overrides the default of

NOTE:  chan_sip does not support setting the cause on a CANCEL to
anything other than ANSWERED_ELSEWHERE.

ASTERISK-26446 #close

Change-Id: I71742e0919aaa16784c30a2b2e73fbeed7672e47

4 years agoMerge "res_rtp_asterisk: Fix infinite DTMF issue when switching to P2P bridge"
Joshua Colp [Tue, 11 Oct 2016 13:52:45 +0000 (08:52 -0500)]
Merge "res_rtp_asterisk: Fix infinite DTMF issue when switching to P2P bridge"

4 years agovector: After remove element recheck index
Badalyan Vyacheslav [Mon, 10 Oct 2016 21:59:58 +0000 (17:59 -0400)]
vector: After remove element recheck index

Small fix. It is necessary to double-check
the index that we just removed because there
is a new element.

ASTERISK-26453 #close

Change-Id: Ib947fa94dc91dcd9341f357f1084782c64434eb7

4 years agoMerge "cel_odbc: Fix memory leak on module unload"
Joshua Colp [Tue, 11 Oct 2016 00:26:37 +0000 (19:26 -0500)]
Merge "cel_odbc: Fix memory leak on module unload"

4 years agoMerge "pjproject_bundled: Add MALLOC_DEBUG capability"
Joshua Colp [Mon, 10 Oct 2016 23:06:30 +0000 (18:06 -0500)]
Merge "pjproject_bundled:  Add MALLOC_DEBUG capability"

4 years agores_rtp_asterisk: Fix infinite DTMF issue when switching to P2P bridge
Torrey Searle [Thu, 29 Sep 2016 17:52:45 +0000 (19:52 +0200)]
res_rtp_asterisk: Fix infinite DTMF issue when switching to P2P bridge

If a bridge switched to P2P when a DTMF was in progress it
was possible for the DTMF to continue being sent indefinitely.

Change-Id: I7e2a3efe0d59d4b214ed50cd0b5d0317e2d92e29

4 years agologger: Prevent output of verbose messages initiated from rasterisk.
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


Change-Id: Ie2a1bb3753ad2724c0349ec1a336f52f7117b52a

4 years agoaudiohooks: Remove redundant codec translations when using audiohooks
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
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


Change-Id: I6edd5771f0740e758e7eb42558b953f046c01f8f

4 years agores_pjsip_config_wizard: Memory leak in module_unload
Badalyan Vyacheslav [Mon, 10 Oct 2016 15:59:38 +0000 (11:59 -0400)]
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.

ASTERISK-26453 #close

Change-Id: I84508353463456d2495678f125738e20052da950

4 years agores_calendar: Add support for fetching calendars when reloading
Ludovic Gasc (GMLudo) [Thu, 29 Sep 2016 17:45:39 +0000 (19:45 +0200)]
res_calendar: Add support for fetching calendars when reloading

We use a lot res_calendar, we are very happy with that, especially
because you use libical, the almost alone opensource library that
supports really ical format with all types of recurrency.

Nevertheless, some features are missed for our business use cases.

This first patch adds a new option in calendar.conf:
fetch_again_at_reload. Be my guest for a better name.

If it's true, when you'll launch "module reload",
Asterisk will download again the calendar.

The business use case is that we have a WebUI with a scheduler planner,
we know when the calendars are modified.

For now, we need to define 1 minute of timeout to have a chance that
our user doesn't wait too long between the modification and the real
test.  But it generates a lot of useless HTTP traffic.

ASTERISK-26422 #close

Change-Id: I384b02ebfa42b142bbbd5b7221458c7f4dee7077

4 years agoMerge "Revert "Packet-Loss Concealment (PLC) for supporting codecs.""
Joshua Colp [Mon, 10 Oct 2016 11:01:07 +0000 (06:01 -0500)]
Merge "Revert "Packet-Loss Concealment (PLC) for supporting codecs.""

4 years agocel_odbc: Fix memory leak on module unload
Badalyan Vyacheslav [Mon, 10 Oct 2016 02:53:07 +0000 (22:53 -0400)]
cel_odbc: Fix memory leak on module unload

Change-Id: Ic7a1236eba2408090fdabb5f717b5fa455ead715

4 years agobundled_pjproject: Add tests for programs used by the Makefile, et al.
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

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 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.

ASTERISK-26416 #close
Reported-by: Corey Farrell

Change-Id: Id1d94251c0155f8dd41b7de7067f35cfbaafbb9b
(cherry picked from commit e6b0053d7561032b7adbf6f3afaecf30f5046605)
(cherry picked from commit a0d02f38322c2c4d7743504003fd376d32a133db)

4 years agoRevert "Packet-Loss Concealment (PLC) for supporting codecs."
Joshua Colp [Sun, 9 Oct 2016 23:54:53 +0000 (23:54 +0000)]
Revert "Packet-Loss Concealment (PLC) for supporting codecs."

This change introduced some fax test failures
that have not yet been addressed. So this is
not forgotten I'm submitting a change which
reverts it.

This reverts:


Change-Id: Ibc2f23c38643f5a2c89cf8915ae2d805b81bc3d5

4 years agopjproject_bundled: Add MALLOC_DEBUG capability
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.

Change-Id: Icc5e3d658fbfb00e0a46b44c66dcc2522d5171b0

4 years agoalembic: Allow cdr, config and voicemail to exist in the same schema
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 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

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

NOTE:  Since the scripts in each tree were identical, a common 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

4 years agoMerge "chan_sip: Honor support of Symmetric Response (rport) for SIP requests."
Joshua Colp [Fri, 7 Oct 2016 10:58:56 +0000 (05:58 -0500)]
Merge "chan_sip: Honor support of Symmetric Response (rport) for SIP requests."

4 years agochan_sip: Honor support of Symmetric Response (rport) for SIP requests.
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.

ASTERISK-26438 #close

Change-Id: If6e7891848aaf96666dee5305695f7c6667cd5a6

4 years agoMerge "app_queue: Update dynamic members ringinuse on reload."
Joshua Colp [Tue, 4 Oct 2016 17:45:54 +0000 (12:45 -0500)]
Merge "app_queue: Update dynamic members ringinuse on reload."

4 years agoBinaural synthesis (confbridge): interleaved two-channel audio.
frahaase [Fri, 12 Aug 2016 16:22:02 +0000 (18:22 +0200)]
Binaural synthesis (confbridge): interleaved two-channel audio.

Asterisk only supports mono audio at the moment.
This patch adds interleaved two-channel audio to Asterisk's channels.


Change-Id: I7a547cea0fd3c6d1e502709d9e7e39605035757a

4 years agoastobj2: Add backtrace to log_bad_ao2.
Corey Farrell [Fri, 16 Sep 2016 23:54:07 +0000 (19:54 -0400)]
astobj2: Add backtrace to log_bad_ao2.

* 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.

Change-Id: I48f1af44b2718ad74a421ff75cb6397b924a9751

4 years agoAdd text of cdr directory into for ast-db-manage
Rodrigo Ramírez Norambuena [Fri, 30 Sep 2016 21:29:37 +0000 (18:29 -0300)]
Add text of cdr directory into for ast-db-manage

Change-Id: I68321c4bea50730c39fdb486e5f23aeadd1ad636

4 years agoapp_queue: Update dynamic members ringinuse on reload.
Etienne Lessard [Fri, 9 Sep 2016 17:38:39 +0000 (13:38 -0400)]
app_queue: Update dynamic members ringinuse on reload.

Previously, when reloading the members of a queue, the members added statically
(i.e. defined in queues.conf) would see their "ringinuse" value updated but not
the members added dynamically.

This change makes dynamic members ringuse value to be updated on reload.

Note that it's impossible to add a dynamic member with a specific ringinuse
value. For both static and dynamic members, the ringinuse value can always be
changed later on with command like "queue set ringinuse" or with the AMI action
"QueueMemberRingInUse". So it's possible this commit could break a user workflow
if he was changing the ringinuse value of dynamic members via such commands and
was also relying on the fact that a queue reload would not update the dynamic
members ringinuse value.


Change-Id: I3745cc9a06ba7e02c399636f1ee9e58c04081f3f

4 years agoMerge "core: Remove ABI effects of LOW_MEMORY."
Joshua Colp [Fri, 30 Sep 2016 11:49:33 +0000 (06:49 -0500)]
Merge "core: Remove ABI effects of LOW_MEMORY."

4 years agoMerge "Remove "format_ogg_opus: New format""
Joshua Colp [Thu, 29 Sep 2016 21:14:28 +0000 (16:14 -0500)]
Merge "Remove "format_ogg_opus: New format""

4 years agoMerge "download_externals: Fix issue with re-install"
Joshua Colp [Thu, 29 Sep 2016 21:03:27 +0000 (16:03 -0500)]
Merge "download_externals: Fix issue with re-install"

4 years agoRemove "format_ogg_opus: New format"
Kevin Harwell [Thu, 29 Sep 2016 19:02:37 +0000 (14:02 -0500)]
Remove "format_ogg_opus: New format"

This reverts commit 40aa28131bc30b4516da2b20eb1a1e043920169c.

ASTERISK-26426 #close

Change-Id: I81e55c3c512f1dd6f49896f0c6b97a07d74fd8f5

4 years agocore: Remove ABI effects of LOW_MEMORY.
Corey Farrell [Mon, 19 Sep 2016 09:46:27 +0000 (05:46 -0400)]
core: Remove ABI effects of LOW_MEMORY.

This allows asterisk to compiled with LOW_MEMORY to load modules built
without LOW_MEMORY.

ASTERISK-26398 #close

Change-Id: I24b78ac9493ab933b11087a8b6794f3c96d4872d

4 years agodownload_externals: Fix issue with re-install
George Joseph [Tue, 27 Sep 2016 21:10:37 +0000 (15:10 -0600)]
download_externals: Fix issue with re-install

Needed to ignore an xmlstarlet return code for optional element.

Change-Id: I6a96f709b4b38c9a3f3dda4e8b07903787e16873
Reported-by: Dan Jenkins

4 years agologger: Output early verbose messages to console.
Corey Farrell [Tue, 27 Sep 2016 20:35:38 +0000 (16:35 -0400)]
logger: Output early verbose messages to console.

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

ASTERISK-26410 #close

Change-Id: Ia44235e30ec328aba92ea2c8a837b094e65c9a03

4 years agoMerge "chan_sip: Resolve externhost not to IPv6; instead go for IPv4."
zuul [Tue, 27 Sep 2016 19:30:46 +0000 (14:30 -0500)]
Merge "chan_sip: Resolve externhost not to IPv6; instead go for IPv4."

4 years agoMerge "codec_opus: Add download ability to menuselect"
George Joseph [Tue, 27 Sep 2016 19:13:04 +0000 (14:13 -0500)]
Merge "codec_opus: Add download ability to menuselect"

4 years agoMerge "codec_opus: Replace res_format_attr_opus with the one from codec_opus"
George Joseph [Tue, 27 Sep 2016 19:12:52 +0000 (14:12 -0500)]
Merge "codec_opus: Replace res_format_attr_opus with the one from codec_opus"

4 years agoMerge "format_ogg_opus: New format"
George Joseph [Tue, 27 Sep 2016 19:12:42 +0000 (14:12 -0500)]
Merge "format_ogg_opus: New format"

4 years agocodec_opus: Add download ability to menuselect
George Joseph [Thu, 22 Sep 2016 14:49:50 +0000 (08:49 -0600)]
codec_opus: Add download ability to menuselect

Updated codecs/codecs.xml to add codec_opus to the external
download list.


Change-Id: Ia07b36539f30e852125fb2b94147dc9774df31a4
(cherry picked from commit 2cdab0e36eec4997ca3bd85aa09efc477038e31c)
(cherry picked from commit e9684f3acd0e8def0df582c1505dd39dd3fd1610)

4 years agocodec_opus: Replace res_format_attr_opus with the one from codec_opus
George Joseph [Sat, 23 Jul 2016 19:50:37 +0000 (13:50 -0600)]
codec_opus: Replace res_format_attr_opus with the one from codec_opus



Change-Id: I9f20e7cce00c32464d9a180e81283d49d199d0a3
(cherry picked from commit 59f7662a93bf9c07204fb50e1020a0f5bfbbd5c9)

4 years agoformat_ogg_opus: New format
George Joseph [Sat, 23 Jul 2016 20:56:43 +0000 (14:56 -0600)]
format_ogg_opus: New format

Add Ogg/Opus playback support.

This uses libopusfile in order to be able to read .opus files and play
them back.

Writing/recording support is not present at this time.


Change-Id: I8815d23345108d8ca7c0bd640f6a1ce6b4f56955
(cherry picked from commit daee8bbd5209b4158bc1785eede845a26e6cbeaa)

4 years agobuild_tools: Add ability to download variants to download_externals
George Joseph [Sun, 25 Sep 2016 00:05:02 +0000 (18:05 -0600)]
build_tools:  Add ability to download variants to download_externals

Some external packages have multiple variants that apply to different
builds of asterisk.  The DPMA for instance has a "bundled" variant that
needs to be downloaded if asterisk was configured with

There are 2 ways to specify variants:

If you need the user to make the decision about which variant to
download, simply create multiple menuselect "member" entries like so...

<member name="res_digium_phone" displayname="..snipped..">

<member name="res_digium_phone-bundled" displayname="..snipped..">

Note that the second entry has "-<variant>" appended to the name.
You can then use the existing menuselect facilities to restrict which
members to enable or disable.  Youy probably don't want the user to
enable multiple at the same time.

If you want to hide the details of the variants, the better way to
do it is to create 1 member with "variant" elements.

<member name="res_digium_phone" displayname="..snipped..">
        <variant tag="bundled"
          condition='[[ "$PJPROJECT_BUNDLED" = "yes" ]]'/>

The condition must be a bash expression suitable for use with an "if"
statement.  Any environment variable can be used plus those available
in makeopts.

In this case, if asterisk was configured with --with-pjproject-bundled
the bundled variant will be automatically downloaded.  Otherwise the
normal version will be downloaded.

Change-Id: I4de23e06d4492b0a65e105c8369966547d0faa3e

4 years agoMerge "chan_sip: Address runaway when realtime peers subscribe to mailboxes"
zuul [Fri, 23 Sep 2016 21:59:59 +0000 (16:59 -0500)]
Merge "chan_sip:  Address runaway when realtime peers subscribe to mailboxes"

4 years agoMerge "channels/chan_pjsip: fix HANGUPCAUSE function bug."
zuul [Fri, 23 Sep 2016 20:38:53 +0000 (15:38 -0500)]
Merge "channels/chan_pjsip: fix HANGUPCAUSE function bug."

4 years agochan_sip: Resolve externhost not to IPv6; instead go for IPv4.
Alexander Traud [Fri, 23 Sep 2016 14:54:28 +0000 (16:54 +0200)]
chan_sip: Resolve externhost not to IPv6; instead go for IPv4.

For the channel driver chan_sip, you specify in sip.conf
when your Asterisk is behind a NAT and your IP address is assigned dynamically.
Or stated differently: You do not have a static IP address to use "externaddr"
directly. This NAT support is quite handy but just about IPv4. Previously,
Asterisk resolved "externhost" to any IP version. When the first DNS answer
resolved to an IPv6, Asterisk sent an IPv6 in SIP/SDP for origin (o=) and
connection (c=). This happened in outgoing SIP-REGISTER and while answering
SIP-INVITE. If the remote peer is IPv4-only, it might not handle o=/c= with an
IPv6. This change makes sure, no IPv6 is resolved anymore for "externhost".

ASTERISK-18232 #close
Reported by: Jacek Kowalski
Tested by: Alexander Traud
 changes.patch submitted by Alessandro Crespi

Change-Id: If68eedbeff65bd1c1d8a9ed921c02ba464b32dac

4 years agochan_sip: Address runaway when realtime peers subscribe to mailboxes
George Joseph [Tue, 20 Sep 2016 14:42:15 +0000 (08:42 -0600)]
chan_sip:  Address runaway when realtime peers subscribe to mailboxes

Users upgrading from asterisk 13.5 to a later version and who use
realtime with peers that have mailboxes were experiencing runaway
situations that manifested as a continuous stream of taskprocessor
congestion errors, memory leaks and an unresponsive chan_sip.

A related issue was that setting rtcachefriends=no NEVER worked in
asterisk 13 (since the move to stasis).  In 13.5 and earlier, when a
peer tried to register, all of the stasis threads would block and
chan_sip would again become unresponsive.  After 13.5, the runaway
would happen.

There were a number of causes...
* mwi_event_cb was (indirectly) calling build_peer even though calls to
  mwi_event_cb are often caused by build_peer.
* In an effort to prevent chan_sip from being unloaded while messages
  were still in flight, destroy_mailboxes was calling
  stasis_unsubscribe_and_join but in some cases waited forever for the
  final message.
* add_peer_mailboxes wasn't properly marking the existing mailboxes
  on a peer as "keep" so build_peer would always delete them all.
* add_peer_mwi_subs was unsubscribing existing mailbox subscriptions
  then just creating them again.

All of this was causing a flood of subscribes and unsubscribes on
multiple threads all for the same peer and mailbox.

* add_peer_mailboxes now marks mailboxes correctly and build_peer only
  deletes the ones that really are no longer needed by the peer.
* add_peer_mwi_subs now only adds subscriptions marked as "new" instead
  of unsubscribing and resubscribing everything.  It also adds the peer
  object's address to the mailbox instead of its name to the subscription
  userdata so mwi_event_cb doesn't have to call build_peer.

With these changes, with rtcachefriends=yes (the most common setting),
there are no leaks, locks, loops or crashes at shutdown.

rtcachefriends=no still causes leaks but at least it doesn't lock, loop
or crash.  Since making rtcachefriends=no work wasnt in scope for this
issue, further work will have to be deferred to a separate patch.

Side fixes...
 * The ast_lock_track structure had a member named "thread" which gdb
   doesn't like since it conflicts with it's "thread" command.  That
   member was renamed to "thread_id".

ASTERISK-25468 #close

Change-Id: I07519ef7f092629e1e844f855abd279d6475cdd0

4 years agoMerge "core: Ensure presencestate subtype and message are NULL."
Joshua Colp [Thu, 22 Sep 2016 13:45:38 +0000 (08:45 -0500)]
Merge "core: Ensure presencestate subtype and message are NULL."

4 years agoMerge "res_odbc: Make pooling option deprecation notice more useful."
Joshua Colp [Thu, 22 Sep 2016 12:10:47 +0000 (07:10 -0500)]
Merge "res_odbc: Make pooling option deprecation notice more useful."

4 years agoMerge "cdr_mysql: fix UTC support"
Joshua Colp [Thu, 22 Sep 2016 11:55:15 +0000 (06:55 -0500)]
Merge "cdr_mysql: fix UTC support"

4 years agochannels/chan_pjsip: fix HANGUPCAUSE function bug.
Aaron An [Thu, 22 Sep 2016 06:40:45 +0000 (14:40 +0800)]
channels/chan_pjsip: fix HANGUPCAUSE function bug.

HANGUPCAUSE not return 'SIP 200 Ok' when dialed channel answered.
This patch change the call order of ast_queue_control_data
and ast_queue_control in chan_pjsip_incoming_response.

ASTERISK-26396 #close
Reported by: AaronAn
Tested by: AaronAn

Change-Id: Ide2d31723d8d425961e985de7de625694580be61

4 years agoMerge "logger: Simplify ast_callid handling code."
zuul [Wed, 21 Sep 2016 20:15:14 +0000 (15:15 -0500)]
Merge "logger: Simplify ast_callid handling code."

4 years agoMerge "logger: Always enable verbose for console channel."
Joshua Colp [Wed, 21 Sep 2016 19:35:27 +0000 (14:35 -0500)]
Merge "logger: Always enable verbose for console channel."

4 years agocore: Ensure presencestate subtype and message are NULL.
Joshua Colp [Wed, 21 Sep 2016 19:24:08 +0000 (19:24 +0000)]
core: Ensure presencestate subtype and message are NULL.

When retrieving presence state information there is no
guarantee that the subtype and message passed in are
set to NULL. This change ensures they are.

ASTERISK-26397 #close

Change-Id: If38cd730e409e9a9b6eb9adef6591d15a9e61f86

4 years agoMerge "logger: Fix default console settings."
zuul [Wed, 21 Sep 2016 17:22:35 +0000 (12:22 -0500)]
Merge "logger: Fix default console settings."

4 years agoMerge "core: Fix LOW_MEMORY missing symbol ast_pbx_uuid_get."
zuul [Wed, 21 Sep 2016 16:31:54 +0000 (11:31 -0500)]
Merge "core: Fix LOW_MEMORY missing symbol ast_pbx_uuid_get."

4 years agores_odbc: Make pooling option deprecation notice more useful.
Joshua Colp [Wed, 21 Sep 2016 15:48:47 +0000 (15:48 +0000)]
res_odbc: Make pooling option deprecation notice more useful.

This changes the notice for the deprecation of the old
pooling options to point to the new option for doing
pooling. This gives a clearer direction as to what to
look into.

ASTERISK-26389 #close

Change-Id: I2ca9cdfdcd75aec170a7db9d5ff69a4cd25b7c10

4 years agoodbc: Remove options that are no longer applicable.
Joshua Colp [Wed, 21 Sep 2016 13:46:36 +0000 (13:46 +0000)]
odbc: Remove options that are no longer applicable.

The pooling, shared_connection, limit, and idlecheck options
are no longer used in res_odbc.


Change-Id: I2fde7b467d01f9d1c82cc0a339bb4f7e1dd6bbe6

4 years agoMerge "asterisk.c: Non-root users also get the astcanary after core restart."
zuul [Wed, 21 Sep 2016 12:10:09 +0000 (07:10 -0500)]
Merge "asterisk.c: Non-root users also get the astcanary after core restart."

4 years agologger: Simplify ast_callid handling code.
Corey Farrell [Tue, 16 Aug 2016 20:21:33 +0000 (16:21 -0400)]
logger: Simplify ast_callid handling code.

Routines responsible for managing ast_callid's are overly complicated.
This is left-over code from when ast_callid was an AO2 object.  Now that
it is an integer the code can be reduced.

ast_callid handler code no longer prints it's own error message upon failure
to allocate threadstorage as ast_calloc would have already printed a
message.  Debug messages that were printed when TEST_FRAMEWORK was
enabled have been also been removed.

Change-Id: I65a768a78dc6cf3cfa071e97f33ce3dce280258e

4 years agocore: Fix LOW_MEMORY missing symbol ast_pbx_uuid_get.
Corey Farrell [Tue, 20 Sep 2016 20:17:42 +0000 (16:17 -0400)]
core: Fix LOW_MEMORY missing symbol ast_pbx_uuid_get.

Move the function outside the conditional block that excludes

ASTERISK-26273 #close

Change-Id: Ic290fa128222c410c3531107e30efacabc8493b4

4 years agoMerge "res_pjsip_multihomed: Change Contact port to listening port."
zuul [Tue, 20 Sep 2016 17:45:16 +0000 (12:45 -0500)]
Merge "res_pjsip_multihomed: Change Contact port to listening port."

4 years agologger: Always enable verbose for console channel.
Corey Farrell [Tue, 20 Sep 2016 14:22:45 +0000 (10:22 -0400)]
logger: Always enable verbose for console channel.

Previous versions of Asterisk did not require verbose to be specified in
logger.conf for the console channel, if it was requested by command line
or asterisk.conf it just worked.  This change causes Asterisk to always
enable verbose in the console channel level mask.  Verbose is displayed
on consoles if requested by command line, option_verbose or 'core set

This also delays initialization of the logger until after threadstorage
is initialized.  Initializing too early can cause messages to be printed
multiple times to the console (stdout).

ASTERISK-26391 #close

Change-Id: I52187d67c2fcb3efd5561bf04b3e5e23e5ee8a04

4 years agologger: Fix default console settings.
Corey Farrell [Tue, 20 Sep 2016 15:16:42 +0000 (11:16 -0400)]
logger: Fix default console settings.

When logger.conf is missing or invalid we should be printing notices,
warnings and errors to the console.  The logmask was incorrectly

Change-Id: Ibaa9465a8682854bc1a5e9ba07079bea1bfb6bb3

4 years agoMerge "sd_notify (systemd status notifications) support"
zuul [Tue, 20 Sep 2016 16:19:02 +0000 (11:19 -0500)]
Merge "sd_notify (systemd status notifications) support"

4 years agoMerge "rtp: Only accept the first payload for a format in SDP."
zuul [Tue, 20 Sep 2016 14:34:58 +0000 (09:34 -0500)]
Merge "rtp: Only accept the first payload for a format in SDP."

4 years agoMerge "Fix showing of swap details when sysinfo() is available"
zuul [Mon, 19 Sep 2016 21:05:02 +0000 (16:05 -0500)]
Merge "Fix showing of swap details when sysinfo() is available"

4 years agoasterisk.c: Non-root users also get the astcanary after core restart.
Walter Doekes [Mon, 19 Sep 2016 19:21:23 +0000 (21:21 +0200)]
asterisk.c: Non-root users also get the astcanary after core restart.

Without this change, a 'core restart' would kill the astcanary forever
if you're not running as root. Both with and without this patch, the
scheduling priority was still SCHED_RR after restart.

Additionally, the astcanary is now spawned if you start with high
priority and Asterisk doesn't get a chance to lower it. For example
through: `chrt -r 10 sudo -u asterisk asterisk -c`

Also reap killed astcanary processes on core restart.

ASTERISK-26352 #close

Change-Id: Iacb49f26491a0717084ad46ed96b0bea5f627a55

4 years agoMerge "res_config_odbc.c: Fix buffer size limitation creating invalid SQL."
zuul [Mon, 19 Sep 2016 20:21:30 +0000 (15:21 -0500)]
Merge "res_config_odbc.c: Fix buffer size limitation creating invalid SQL."

4 years agoMerge "asterisk.c: When astcanary dies on linux, reset priority on all threads."
zuul [Mon, 19 Sep 2016 20:02:09 +0000 (15:02 -0500)]
Merge "asterisk.c: When astcanary dies on linux, reset priority on all threads."

4 years agoasterisk.c: When astcanary dies on linux, reset priority on all threads.
Walter Doekes [Mon, 19 Sep 2016 14:40:40 +0000 (16:40 +0200)]
asterisk.c: When astcanary dies on linux, reset priority on all threads.

Previously only the canary checking thread itself had its priority set
to SCHED_OTHER. Now all threads are traversed and adjusted.

ASTERISK-19867 #close
Reported by: Xavier Hienne

Change-Id: Ie0dd02a3ec42f66a78303e9c1aac28f7ed9aae39

4 years agores_config_odbc.c: Fix buffer size limitation creating invalid SQL.
Richard Mudgett [Mon, 12 Sep 2016 23:00:22 +0000 (18:00 -0500)]
res_config_odbc.c: Fix buffer size limitation creating invalid SQL.

Creating ODBC SQL queries resulted in queries too large to fit into the
supplied buffer.  The resulting truncated buffer contained an invalid SQL

* Made SQL query generation code use a thread storage buffer that can
increase in size as needed.

* Fixed bad multi-line warning messages.

ASTERISK-26263 #close
Reported by: Jeppe Ryskov Larsen

Change-Id: I23f3cdd43c2dac80bed3ded4dd77d18cb17f21ae

4 years agortp: Only accept the first payload for a format in SDP.
Joshua Colp [Wed, 14 Sep 2016 11:53:36 +0000 (07:53 -0400)]
rtp: Only accept the first payload for a format in SDP.

When receiving an SDP offer with multiple payloads for
the same format we would generate an answer with the first
payload, but during the payload crossover operation
(to set the payloads for receiving) we would remove all
payloads but the last. This would result in incoming
traffic being matched against the wrong format and outgoing
traffic being sent using the wrong payload.

This change makes it so that once a format has a payload
number put into the mapping all subsequent ones are ignored.
This ensures there is only ever one payload in the mapping
and that it is the payload placed into the answer SDP.

ASTERISK-26365 #close

Change-Id: I1e8150860a3518cab36d00b1fab50f9352b64e60

4 years agores_pjsip_multihomed: Change Contact port to listening port.
Joshua Colp [Wed, 14 Sep 2016 13:42:46 +0000 (09:42 -0400)]
res_pjsip_multihomed: Change Contact port to listening port.

The res_pjsip_multihomed module determines what interface and transport
a request is going out on and updates the SIP message accordingly with
the address information. This currently incorrectly updates the Contact
header for connectionful protocols to the ephemeral connection port,
instead of the bound address for the listening socket which can actually
accept the connection back. If the remote side attempts to connect back on
the epehemeral port it will fail.

This change makes it so the port is updated to the bound port on
connectionful protocols and is maintained on UDP (as there can be
multiple of those).

ASTERISK-26374 #close

Change-Id: I50f8dab65b9f75117d73ba5f6bbcf6c9871854ab

4 years agopjproject_bundled: Prevent SERVFAIL from marking name server bad
George Joseph [Wed, 7 Sep 2016 19:48:48 +0000 (13:48 -0600)]
pjproject_bundled:  Prevent SERVFAIL from marking name server bad

A name server that returns "Server Failure" is indicating only that
the server couldn't process that particular request.  We should NOT
assume that the name server is incapable of serving other requests.

Here's the scenario we've been encountering...

* 2 local name servers configured in resolv.conf.
* An OPTIONS request causes a request for A and AAAA records to go out
  to both nameservers.
* The A responses both come back successfully resolved.
* Because of an issue at some upstream nameserver, the AAAA responses
  for that particular query come back as "SERVFAIL" from both local
  name servers.
* Both local servers are marked as bad and no further queries can be
  sent until the 60 second ttl expires.  Only previously cached results
  can be used.
* In this case, 60 seconds is just enough time for another OPTIONS
  request to go out to the same host so the cycle repeats.

We could set the bad ttl really low but that also affects REFUSED and
NOTAUTH which probably DO signal a real server issue.  Besides, even
a really low bad ttl would be an issue on a pbx.

Although we use our own resolver in 14 and master and don't have this
issue there, Teluu has merged this patch upstream so it's appropriate
to cherry-pick to 14 and master to keep pjproject consistent.

Change-Id: Ie03ba902288e274aff23f9b9bb2786e1e8be09e0

4 years agocdr_mysql: fix UTC support
Tzafrir Cohen [Mon, 12 Sep 2016 12:37:30 +0000 (15:37 +0300)]
cdr_mysql: fix UTC support

* Make 'cdrzone=UTC' work properly.
* Fix the documentation of cdr_mysql.conf: it's cdrzone and not timezone

ASTERISK-26359 #close

Change-Id: I2a6f67b71bbbe77cac31a34d0bbfb1d67c933778

4 years agosd_notify (systemd status notifications) support
Tzafrir Cohen [Mon, 27 Jun 2016 19:26:54 +0000 (21:26 +0200)]
sd_notify (systemd status notifications) support

sd_notify() is used to notify systemd of changes to the status of the
process. This allows the systemd daemon to know when the process
finished loading (and thus only start another program after Asterisk has
finished loading).

To use this, use a systemd unit with 'Type=notify' for Asterisk.

This commit also adds the function ast_sd_notify(), a wrapper around
sd_notify that does nothing if not built with systemd support.

Also adds support for libsystemd detection in the configure script.

Change-Id: Ied6a59dafd5ef331c5c7ae8f3ccd2dfc94be7811

4 years agoFix showing of swap details when sysinfo() is available
Timo Teräs [Fri, 9 Sep 2016 11:35:43 +0000 (14:35 +0300)]
Fix showing of swap details when sysinfo() is available

If sysinfo() is available, but not sysctl() or swapctl() the
printing code for swap buffer sizes is incorrectly omitted.
The above condition happens with musl c-library.

Fix #if rule to consider defined(HAVE_SYSINFO). And also
remove the redundant || defined(HAVE_SYSCTL) which was
incorrectly there to start with. Now swap information is
displayed only if an actual libc function to get it is

This also fixes warnings previously seen with musl libc:

   [CC] asterisk.c -> asterisk.o
asterisk.c: In function 'handle_show_sysinfo':
asterisk.c:773:6: warning: variable 'totalswap' set but not used
  int totalswap = 0;
asterisk.c:770:11: warning: variable 'freeswap' set but not used
  uint64_t freeswap = 0;

Change-Id: I1fb21dad8f27e416c60f138c6f2bff03fb626eca

4 years agoMerge "res_pjsip_transport_management: Convert time in log message to seconds."
zuul [Thu, 15 Sep 2016 03:35:43 +0000 (22:35 -0500)]
Merge "res_pjsip_transport_management: Convert time in log message to seconds."

4 years agoMerge "chan_sip: Fix session timeout on retransmit of non-UDP packets"
zuul [Thu, 15 Sep 2016 00:42:21 +0000 (19:42 -0500)]
Merge "chan_sip: Fix session timeout on retransmit of non-UDP packets"

4 years agoMerge "rtp: Preserve timestamps on video frames."
zuul [Wed, 14 Sep 2016 22:21:12 +0000 (17:21 -0500)]
Merge "rtp: Preserve timestamps on video frames."

4 years agoMerge " Map legacy_useroption_parsing."
zuul [Wed, 14 Sep 2016 20:03:46 +0000 (15:03 -0500)]
Merge " Map legacy_useroption_parsing."

4 years agortp: Preserve timestamps on video frames.
Joshua Colp [Wed, 14 Sep 2016 12:59:51 +0000 (08:59 -0400)]
rtp: Preserve timestamps on video frames.

Currently when receiving video over RTP we store only
a calculated samples on the frame. When starting the video
it can take some time for this calculation to actually yield
a value as it requires constant changing timestamps. As well
if a video frame passes over multiple RTP packets this calculation
will fail as the timestamp is the same as the previous RTP
packet and the number of samples calculated will be 0.

This change preserves the timestamp on the frame and allows
it to pass through the core. When sending the video this timestamp
is used instead of a new one being calculated.

ASTERISK-26367 #close

Change-Id: Iba8179fb5c14c9443aee4baf670d2185da3ecfbd

4 years agoMerge "res_pjsip: Add ignore_uri_user_options option."
zuul [Wed, 14 Sep 2016 17:27:28 +0000 (12:27 -0500)]
Merge "res_pjsip: Add ignore_uri_user_options option."

4 years agores_pjsip_transport_management: Convert time in log message to seconds.
Joshua Colp [Wed, 14 Sep 2016 14:51:53 +0000 (10:51 -0400)]
res_pjsip_transport_management: Convert time in log message to seconds.

ASTERISK-26375 #close

Change-Id: I46496af5cae41413e76d44d2068a7431279f09dc

4 years agoMerge "res_pjsip: Don't assume a request will have any addresses."
zuul [Tue, 13 Sep 2016 23:24:44 +0000 (18:24 -0500)]
Merge "res_pjsip: Don't assume a request will have any addresses."

4 years agochan_sip: Fix session timeout on retransmit of non-UDP packets
Steve Davies [Tue, 13 Sep 2016 10:34:47 +0000 (11:34 +0100)]
chan_sip: Fix session timeout on retransmit of non-UDP packets

Change-Id I1cd33453c77c56c8e1394cd60a6f17bb61c1d957 Enable Session-Timers for
SIP over TCP (and TLS) also disables SIP retransmits in chan_sip for non-UDP
connections, allowing the TCP layer to handle the retransmits. Unfortunately,
this caused sessions to be terminated with a retransmit timeout becasue it
stopped at the point of the first retrans call.

This patch waits for the 64*T1 timer to expire instead.


Change-Id: I844f26801aada10bc94e9bebe6e151f0a8443204

4 years agoMerge "chan_sip: Allow target refresh (Contact update) on re-INVITE."
zuul [Tue, 13 Sep 2016 15:26:50 +0000 (10:26 -0500)]
Merge "chan_sip: Allow target refresh (Contact update) on re-INVITE."

4 years agoMerge "res_pjsip_messaging.c: Misc cleanups and fixes."
zuul [Tue, 13 Sep 2016 14:04:02 +0000 (09:04 -0500)]
Merge "res_pjsip_messaging.c: Misc cleanups and fixes."

4 years agores_pjsip: Don't assume a request will have any addresses.
Joshua Colp [Tue, 13 Sep 2016 11:08:18 +0000 (07:08 -0400)]
res_pjsip: Don't assume a request will have any addresses.

When performing DNS resolution the failover code present in
res_pjsip currently assumes that a request will always have
at least one viable address. In practice this is not true.
A domain may be used that has no records.

The code now checks that at least one address exists on the
request which prevents looping.

ASTERISK-26364 #close

Change-Id: Ic0761b0264864acd85915c94d878a81624940f4c

4 years agoapp_queue: Fix CLI "queue show" and AMI Queues action output truncation.
Richard Mudgett [Mon, 12 Sep 2016 17:25:54 +0000 (12:25 -0500)]
app_queue: Fix CLI "queue show" and AMI Queues action output truncation.

The output of CLI "queue show" and AMI Queues action is truncated and
"failed to extend from 240 to 327" messages are generated if the queue
member and interface names are lengthy.

* Increase the string buffer size from 240 to 512 in order to accommodate
for more information fields added to the output since v1.8.

ASTERISK-26360 #close
Reported by: Richard Mudgett

Change-Id: Id99c03cf5362453b80491a4b3b0434cb67aa966d

4 years agoMerge "contrib: Let safe_asterisk script continue without /dev/tty9."
zuul [Mon, 12 Sep 2016 13:42:18 +0000 (08:42 -0500)]
Merge "contrib: Let safe_asterisk script continue without /dev/tty9."

4 years agochan_sip: Allow target refresh (Contact update) on re-INVITE.
Walter Doekes [Mon, 12 Sep 2016 08:28:17 +0000 (10:28 +0200)]
chan_sip: Allow target refresh (Contact update) on re-INVITE.

Previously, the Contact was stored only on initial INVITE and on any
18X and 200. That meant that after re-INVITEs from *us* the Contact
could get updated, but after re-INVITEs from the *peer*, it did not.

This changeset fixes this inconsistency, properly allowing target
refreshes through re-INVITES (RFC3261, 12.2).

If your strictrtp setting allows it, this change allows you to switch
the source IP of a connected/calling device mid-call with a simple
re-INVITE from the new IP.

ASTERISK-26358 #close

Change-Id: Ibb8512054ab27c8c3d2514022568fde943bf2435

4 years Map legacy_useroption_parsing.
Richard Mudgett [Wed, 31 Aug 2016 20:22:01 +0000 (15:22 -0500)] Map legacy_useroption_parsing.

Map the sip.conf general section legacy_useroption_parsing to the
new pjsip.conf global ignore_uri_user_options.

Reported by: Kevin Harwell

Change-Id: I78108a31995db19d41f4e1a07b3324692c5363fc

4 years agores_pjsip: Add ignore_uri_user_options option.
Richard Mudgett [Mon, 29 Aug 2016 23:08:22 +0000 (18:08 -0500)]
res_pjsip: Add ignore_uri_user_options option.

This implements the chan_sip legacy_useroption_parsing option but with a
better name.

* Made the caller-id number and redirecting number strings obtained from
incoming SIP URI user fields always truncated at the first semicolon.
People don't care about anything after the semicolon showing up on their
displays even though the RFC allows the semicolon.

ASTERISK-26316 #close
Reported by: Kevin Harwell

Change-Id: Ib42b0e940dd34d84c7b14bc2e90d1ba392624f62