asterisk/asterisk.git
8 months agoast_coredumper: Refactor the pid determination process
George Joseph [Mon, 24 Dec 2018 17:42:36 +0000 (10:42 -0700)]
ast_coredumper:  Refactor the pid determination process

In order to get a dump of the running process, we need to find the
pid of the main asterisk process.  This can be tricky if there are
also instances of "asterisk -r" running or if an alternate location
for asterisk.conf was specified on the command line with the -C
option that also specified an alternation location for the pid file.

So now...

1. We find the asterisk executable with "which" or the --asterisk-bin
   command line option.
2. If there's only 1 process with an executable path that matches,
   we use that pid.  If not...
3. We try "<asterisk-bin> -rx 'core show settings'" and parse the
   output to find the pidfile, then read that for the pid.  If that
   didn't work...
4. We get a list of all the pids matching <asterisk-bin> and look
   in /proc/<pid>/cmdline for a -C argument and retry the "core show
   settings" using the same -C option.  We can't parse the output
   of "ps" to get the -C path because it may contain spaces.  The
   contents of /proc/<pid>/cmdline are delimited by NULLs.  For BSDs
   we may have to mount /proc first. :(

ASTERISK-28221
Reported by: Andrew Nagy

Change-Id: I8aa1f3f912f949df2b5348908803c636bde1d57c

8 months agobacktrace.c: Fix casting pointer to/from integral type.
Richard Mudgett [Wed, 19 Dec 2018 18:39:08 +0000 (12:39 -0600)]
backtrace.c: Fix casting pointer to/from integral type.

The backtrace library bfd.h include file does not get the sizes of
pointers and ints right on some platforms.  On my old test box the size
of bfd_vma is 8 while the size of a pointer is 4.  gcc on the box
complains of the integer casting to/from pointers size mismatch.

* uintptr_t to the rescue by doing an appropriate two stage cast.

Change-Id: Icb2621583f50c8728de08a3c824d95fe53cc45d0

8 months agoMerge "res/res_ari: Add additional hangup reasons"
Friendly Automation [Wed, 19 Dec 2018 11:12:15 +0000 (05:12 -0600)]
Merge "res/res_ari: Add additional hangup reasons"

8 months agoMerge "app_voicemail: Don't delete mailbox state unless mailbox is deleted"
Friendly Automation [Wed, 19 Dec 2018 11:08:00 +0000 (05:08 -0600)]
Merge "app_voicemail:  Don't delete mailbox state unless mailbox is deleted"

9 months agoMerge "res_pjsip: Patch for res_pjsip_* module load/reload crash"
George Joseph [Tue, 18 Dec 2018 16:42:49 +0000 (10:42 -0600)]
Merge "res_pjsip: Patch for res_pjsip_* module load/reload crash"

9 months agoMerge "res_rtp_asterisk: Remove some unused structure fields."
George Joseph [Tue, 18 Dec 2018 16:42:26 +0000 (10:42 -0600)]
Merge "res_rtp_asterisk: Remove some unused structure fields."

9 months agoapp_voicemail: Don't delete mailbox state unless mailbox is deleted
George Joseph [Tue, 18 Dec 2018 16:33:50 +0000 (09:33 -0700)]
app_voicemail:  Don't delete mailbox state unless mailbox is deleted

The free_user function was automatically deleting the stasis mailbox
state but this only makes sense when the mailbox is actually
deleted, not just the structure freed.  This was causing issues
where leave_voicemail would publish the mwi message to stasis and
delete the state before the message could be processed by
res_pjsip_mwi.

* Removed the delete of state from free_user().

* Created a new free_user_final() function that both frees the data
  structure and deletes the state.  This function is only called
  during module load/unload where it's appropriate to delete the
  state.

ASTERISK-28215

Change-Id: I305e8b3c930e9ac41d901e5dc8a58fd7904d98dd

9 months agoMerge "res_format_attr_h264.c: Make sure profile-level-id fmtp attribute is set"
Joshua C. Colp [Mon, 17 Dec 2018 15:34:47 +0000 (09:34 -0600)]
Merge "res_format_attr_h264.c: Make sure profile-level-id fmtp attribute is set"

9 months agores_rtp_asterisk: Remove some unused structure fields.
Sean Bright [Fri, 14 Dec 2018 17:52:45 +0000 (12:52 -0500)]
res_rtp_asterisk: Remove some unused structure fields.

All of the fields that were removed were no longer referenced except for
'lastrxts' and 'rxseqno' which were only ever written to.

Change-Id: I5a5d31eb33e97663843698f58d0d97f22a76627c

9 months agoMerge "bridge_builtin_features.c: Set auto(mix)mon variables on both channels"
Joshua C. Colp [Fri, 14 Dec 2018 14:37:38 +0000 (08:37 -0600)]
Merge "bridge_builtin_features.c: Set auto(mix)mon variables on both channels"

9 months agores_format_attr_h264.c: Make sure profile-level-id fmtp attribute is set
Sean Bright [Thu, 13 Dec 2018 21:56:50 +0000 (16:56 -0500)]
res_format_attr_h264.c: Make sure profile-level-id fmtp attribute is set

The profile-iop octet (the 2nd) of profile-level-id can be zero
according to RFC 6184 Section 8.1. So we ignore its value when deciding
to include profile-level-id in the outgoing SDP.

ASTERISK-27959 #close
Reported by: David Kuehling

Change-Id: Id28cd916a3d7748058fe9609b585d07d9e243f73

9 months agoMerge "confbridge: announce to the marked users when they join an empty conference"
Joshua C. Colp [Thu, 13 Dec 2018 14:00:06 +0000 (08:00 -0600)]
Merge "confbridge: announce to the marked users when they join an empty conference"

9 months agobridge_builtin_features.c: Set auto(mix)mon variables on both channels
Sean Bright [Tue, 11 Dec 2018 20:49:03 +0000 (15:49 -0500)]
bridge_builtin_features.c: Set auto(mix)mon variables on both channels

This is how features behaved up through Asterisk 11, but was changed
when the new bridging framework was implemented in Asterisk 12.

Reported by rrittgarn in #asterisk.

Change-Id: I72cf86223947a8118c75f46e2c603dbc11e3125b

9 months agoMerge "utils: Don't set or clear flags that don't need setting or clearing"
Joshua C. Colp [Wed, 12 Dec 2018 19:12:19 +0000 (13:12 -0600)]
Merge "utils: Don't set or clear flags that don't need setting or clearing"

9 months agoMerge "stasis: Add statistics gathering in developer mode."
Friendly Automation [Wed, 12 Dec 2018 19:08:23 +0000 (13:08 -0600)]
Merge "stasis: Add statistics gathering in developer mode."

9 months agoMerge "Use non-blocking socket() and pipe() wrappers"
Joshua C. Colp [Wed, 12 Dec 2018 17:31:00 +0000 (11:31 -0600)]
Merge "Use non-blocking socket() and pipe() wrappers"

9 months agoconfbridge: announce to the marked users when they join an empty conference
Alexei Gradinari [Fri, 7 Dec 2018 20:22:29 +0000 (15:22 -0500)]
confbridge: announce to the marked users when they join an empty conference

Currently the file sound_only_person is not played when a marked
user (with announce_only_user=yes) joins an empty conference.

This patch fixes it.

ASTERISK-28201 #close

Change-Id: I85b67687e6b220939c3af8091d83a70a7b174cf4

9 months agostasis: Add statistics gathering in developer mode.
Joshua C. Colp [Fri, 30 Nov 2018 11:40:40 +0000 (07:40 -0400)]
stasis: Add statistics gathering in developer mode.

This change adds statistics gathering to Stasis topics,
subscriptions, and message types. These can be viewed using
CLI commands and provide insight into how Stasis is used
and how long certain operations take to execute.

These are only available when Asterisk is compiled in
developer mode and do not have any impact under normal
operation.

ASTERISK-28117

Change-Id: I94411b53767f89ee01714daaecf0c2f1666e863f

9 months agoMerge "stasis: Allow filtering by formatter"
Friendly Automation [Wed, 12 Dec 2018 17:09:19 +0000 (11:09 -0600)]
Merge "stasis:  Allow filtering by formatter"

9 months agoMerge "build: Update config.guess and config.sub"
Joshua C. Colp [Wed, 12 Dec 2018 17:05:30 +0000 (11:05 -0600)]
Merge "build: Update config.guess and config.sub"

9 months agoMerge "pjproject_bundled: check whether UPDATE is supported on outgoing calls"
George Joseph [Wed, 12 Dec 2018 16:51:57 +0000 (10:51 -0600)]
Merge "pjproject_bundled: check whether UPDATE is supported on outgoing calls"

9 months agoMerge "Revert "RTP: reset DTMF last seqno/timestamp on voice packet with marker bit""
George Joseph [Tue, 11 Dec 2018 20:18:25 +0000 (14:18 -0600)]
Merge "Revert "RTP: reset DTMF last seqno/timestamp on voice packet with marker bit""

9 months agoUse non-blocking socket() and pipe() wrappers
Sean Bright [Tue, 11 Dec 2018 14:54:43 +0000 (09:54 -0500)]
Use non-blocking socket() and pipe() wrappers

Change-Id: I050ceffe5a133d5add2dab46687209813d58f597

9 months agoutils: Don't set or clear flags that don't need setting or clearing
Sean Bright [Tue, 11 Dec 2018 15:06:15 +0000 (10:06 -0500)]
utils: Don't set or clear flags that don't need setting or clearing

Change-Id: I0e7fb507ac09b15e45e1ff8501ecfca67afa5217

9 months agoMerge "CI: Various updates to buildAsterisk.sh"
George Joseph [Tue, 11 Dec 2018 15:07:59 +0000 (09:07 -0600)]
Merge "CI: Various updates to buildAsterisk.sh"

9 months agoMerge "utils: Wrap socket() and pipe() to reduce syscalls"
Joshua C. Colp [Tue, 11 Dec 2018 15:01:38 +0000 (09:01 -0600)]
Merge "utils: Wrap socket() and pipe() to reduce syscalls"

9 months agobuild: Update config.guess and config.sub
Sean Bright [Tue, 11 Dec 2018 12:55:16 +0000 (07:55 -0500)]
build: Update config.guess and config.sub

Pulled from the authoritative respository at:

  https://git.savannah.gnu.org/cgit/config.git/tree/

Change-Id: I748708ce24d4d47ff1f395075d0b08d3da3355e0

9 months agoRevert "RTP: reset DTMF last seqno/timestamp on voice packet with marker bit"
George Joseph [Tue, 11 Dec 2018 14:28:48 +0000 (09:28 -0500)]
Revert "RTP: reset DTMF last seqno/timestamp on voice packet with marker bit"

This reverts commit 3f53041267234b21aedd522c1197ec57cca90845.

Pending resolution of ASTERISK_28200

Change-Id: Iad4f3614cac95b00fdbb2b799aab8ae6285ec988

9 months agores/res_ari: Add additional hangup reasons
Sebastian Damm [Thu, 6 Dec 2018 17:23:50 +0000 (18:23 +0100)]
res/res_ari: Add additional hangup reasons

The ARI DELETE /channels command takes a "reason" parameter
Previously, there were only five reasons implemented
This patch adds more reasons to choose from for more
complex setups

ASTERISK-28198 #close

Change-Id: I85996f1076c9946d65c778413f040a845a90fecc

9 months agoMerge "chan_sip: Fix leak using contact ACL"
Joshua C. Colp [Mon, 10 Dec 2018 13:05:21 +0000 (07:05 -0600)]
Merge "chan_sip: Fix leak using contact ACL"

9 months agoutils: Wrap socket() and pipe() to reduce syscalls
Sean Bright [Fri, 7 Dec 2018 12:57:48 +0000 (07:57 -0500)]
utils: Wrap socket() and pipe() to reduce syscalls

Some platforms provide an implementation of socket() and pipe2() that allow the
caller to specify that the resulting file descriptors should be non-blocking.

Using these allows us to potentially elide 3 calls into 1 by avoiding extraneous
calls to fcntl() to set the O_NONBLOCK flag afterwards.

In passing, change ast_alertpipe_init() to use pipe2() directly instead of the
wrapper if it is available.

Change-Id: I3ebe654fb549587537161506c6c950f4ab298bb0

9 months agostasis: Allow filtering by formatter
George Joseph [Thu, 29 Nov 2018 15:53:51 +0000 (08:53 -0700)]
stasis:  Allow filtering by formatter

A subscriber can now indicate that it only wants messages
that have formatters of a specific type.  For instance,
manager can indicate that it only wants messages that have a
"to_ami" formatter.  You can combine this with the existing
filter for message type to get only messages with specific
formatters or messages of specific types.

ASTERISK-28186

Change-Id: Ifdb7a222a73b6b56c6bb9e4ee93dc8a394a5494c

9 months agoRemoving registrar_expire from basic-pbx config
David M. Lee [Wed, 5 Dec 2018 21:28:03 +0000 (15:28 -0600)]
Removing registrar_expire from basic-pbx config

The module has been removed, so it shouldn't be in the default config any more.

Change-Id: Ie7e09f00f9c9a885574e29478250de4c2cefd9f1

9 months agochan_sip: Fix leak using contact ACL
Giuseppe Sucameli [Wed, 5 Dec 2018 00:00:40 +0000 (01:00 +0100)]
chan_sip: Fix leak using contact ACL

Free old peer's contactacl before overwrite it within build_peer.

ASTERISK-28194

Change-Id: Ie580db6494e50cee0e2a44b38e568e34116ff54c

9 months agoCI: Various updates to buildAsterisk.sh
George Joseph [Wed, 5 Dec 2018 15:37:45 +0000 (08:37 -0700)]
CI: Various updates to buildAsterisk.sh

* Added ---no-configure, --no-menuselect, --no-make and --no-alembic
  options that prevent those actions from being performed.  Useful
  for testing and re-running portions of the build after fixing
  earlier failures.

* Added "set -e" to abort the script on command failure.
  Not sure why this wasn't there in the first place.

* Fixed a few echos that were redirecting to stderr when they shouldn't
  have been.

* Catch more alembic failures by actually trying to generate the SQL.

Change-Id: I9f395fa4e9254be7299e7c1014f1a13db78faffb

9 months agoMerge "test_websocket_client.c: Disable websocket_client_create_and_connect test."
George Joseph [Wed, 5 Dec 2018 14:18:36 +0000 (08:18 -0600)]
Merge "test_websocket_client.c: Disable websocket_client_create_and_connect test."

9 months agopjsip_add_use_callerid_contact: fixed alembic script
Kevin Harwell [Mon, 3 Dec 2018 23:45:57 +0000 (17:45 -0600)]
pjsip_add_use_callerid_contact: fixed alembic script

Change-Id: I413f1583c797fb79651786cd8d0b003599f8ed10

9 months agocore: Add some documentation to the malloc_trim code
Sean Bright [Mon, 3 Dec 2018 22:41:56 +0000 (17:41 -0500)]
core: Add some documentation to the malloc_trim code

This adds documentation to handle_cli_malloc_trim() indicating how it
can be useful when debugging OOM conditions.

Change-Id: I1936185e78035bf123cd5e097b793a55eeebdc78

9 months agoMerge "core: Merge malloc_trim patch"
George Joseph [Mon, 3 Dec 2018 22:26:51 +0000 (16:26 -0600)]
Merge "core: Merge malloc_trim patch"

9 months agocore: Merge malloc_trim patch
Chris-Savinovich [Mon, 3 Dec 2018 20:01:01 +0000 (14:01 -0600)]
core: Merge malloc_trim patch

We've had multiple opportunities where Richard Mudgett's
malloc_trim patch has been useful. Let's get it
pushed up to gerrit and merged.

Since malloc_trim is only available in libc, an entry is
added to configure.ac to create a definition for
HAVE_MALLOC_TRIM.

Change-Id: Ia38308c550149d9d6eae4ca414a649957de9700c

9 months agores_pjsip: Patch for res_pjsip_* module load/reload crash
Sungtae Kim [Sun, 11 Nov 2018 16:29:10 +0000 (17:29 +0100)]
res_pjsip: Patch for res_pjsip_* module load/reload crash

The session_supplements for the pjsip makes crashes when the module
load/unload.

ASTERISK-28157

Change-Id: I5b82be3a75d702cf1933d8d1417f44aa10ad1029

9 months agoapp_queue: Revert broken queue channel reference patch
lvl [Mon, 22 Oct 2018 12:47:56 +0000 (14:47 +0200)]
app_queue: Revert broken queue channel reference patch

Revert commit 6409e7b11a2310196a9978b30a6b79e2760be592, and add
NULL checks for all app_queue event handling code.

Related issues: ASTERISK~25185, ASTERISK~27006, ASTERISK~25844

ASTERISK-28125

Change-Id: I37334ea184ebb56e54471496b82937d4927815a0

9 months agotest_websocket_client.c: Disable websocket_client_create_and_connect test.
Chris-Savinovich [Fri, 30 Nov 2018 20:00:14 +0000 (14:00 -0600)]
test_websocket_client.c: Disable websocket_client_create_and_connect test.

This test was occasionally failing, with:

  WARNING[5812]: http.c:1939 httpd_helper_thread: Failed to set
      TCP_NODELAY on HTTP connection: Bad file descriptor
  ERROR[5812]: iostream.c:91 ast_iostream_nonblock: Failed to get
      fcntl() flags for file descriptor: Bad file descriptor
  ERROR[5812]: iostream.c:569 ast_iostream_close: close() failed: Bad
      file descriptor

Disabled for now by making the test explicit only.

Change-Id: I778f6cbb6104c6b4e89737a2eaf1a9540888d351

9 months agopjproject_bundled: check whether UPDATE is supported on outgoing calls
Pirmin Walthert [Wed, 28 Nov 2018 07:14:12 +0000 (08:14 +0100)]
pjproject_bundled: check whether UPDATE is supported on outgoing calls

In ASTERISK-27095 an issue had been fixed because of which chan_pjsip was not
trying to send UPDATE messages when connected_line_method was set to invite.
However this only solved the issue for incoming INVITES. For outgoing INVITES
(important when transferring calls) the options variable needs to be updated
at a different place.

ASTERISK-28182 #close
Reported-by: nappsoft

Change-Id: I76cc06da4ca76ddd6dce814a8b97cc66b98aaf29

9 months agoMerge "Revert "app_voicemail: Remove need to subscribe to stasis""
George Joseph [Fri, 30 Nov 2018 13:30:35 +0000 (07:30 -0600)]
Merge "Revert "app_voicemail: Remove need to subscribe to stasis""

9 months agoMerge "bridges: Remove reliance on stasis caching"
George Joseph [Thu, 29 Nov 2018 21:05:33 +0000 (15:05 -0600)]
Merge "bridges:  Remove reliance on stasis caching"

9 months agoRevert "app_voicemail: Remove need to subscribe to stasis"
George Joseph [Thu, 29 Nov 2018 19:26:16 +0000 (12:26 -0700)]
Revert "app_voicemail: Remove need to subscribe to stasis"

This reverts commit 29115e23848cceee0e2763bc70e87cb311919cdd.

That commit closed a long standing hole which allowed subscriptions
to mailboxes that weren't configured in voicemail.conf.  This
caused an issue with FreePBX which depdended on that behavior.
The commit is being reverted until FreePBX can handle the new
behavior.

ASTERISK-28151
Reported by: Ronald Raikes

Change-Id: I57b7b85e75d7dd97c742b5c69d718a0f61260c15

9 months agoMerge "jansson: Upgrade to 2.12."
Kevin Harwell [Thu, 29 Nov 2018 18:57:32 +0000 (12:57 -0600)]
Merge "jansson: Upgrade to 2.12."

9 months agotest_cel: Plug a few ref leaks
George Joseph [Mon, 26 Nov 2018 22:18:00 +0000 (15:18 -0700)]
test_cel:  Plug a few ref leaks

These are only a few of the leaks.  The large number of macros
and return paths in this file would make a weeks worth of work
to plug them all.

Change-Id: Ie2369fa944023d44767871c5c30974cb077ffb56

9 months agobridges: Remove reliance on stasis caching
George Joseph [Wed, 19 Sep 2018 19:34:41 +0000 (13:34 -0600)]
bridges:  Remove reliance on stasis caching

* The bridging core no longer uses the stasis cache for bridge
  snapshots.  The latest bridge snapshot is now stored on the
  ast_bridge structure itself.

* The following APIs are no longer available since the stasis cache
  is no longer used:
    ast_bridge_topic_cached()
    ast_bridge_topic_all_cached()

* A topic pool is now used for individual bridge topics.

* The ast_bridge_cache() function was removed since there's no
  longer a separate container of snapshots.

* A new function "ast_bridges()" was created to retrieve the
  container of all bridges.  Users formerly calling
  ast_bridge_cache() can use the new function to iterate over
  bridges and retrieve the latest snapshot directly from the
  bridge.

* The ast_bridge_snapshot_get_latest() function was renamed to
  ast_bridge_get_snapshot_by_uniqueid().

* A new function "ast_bridge_get_snapshot()" was created to retrieve
  the bridge snapshot directly from the bridge structure.

* The ast_bridge_topic_all() function now returns a normal topic
  not a cached one so you can't use stasis cache functions on it
  either.

* The ast_bridge_snapshot_type() stasis message now has the
  ast_bridge_snapshot_update structure as it's data.  It contains
  the last snapshot and the new one.

* cdr, cel, manager and ari have been updated to use the new
  arrangement.

Change-Id: I7049b80efa88676ce5c4666f818fa18ad1985369

9 months agoMerge "stasis: Segment channel snapshot to reduce creation cost."
Jenkins2 [Mon, 26 Nov 2018 20:07:47 +0000 (14:07 -0600)]
Merge "stasis: Segment channel snapshot to reduce creation cost."

9 months agoMerge "astobj2: Create function to copy weak proxied objects from container."
Joshua Colp [Mon, 26 Nov 2018 19:48:00 +0000 (13:48 -0600)]
Merge "astobj2: Create function to copy weak proxied objects from container."

9 months agoMerge "RTP: need to reset DTMF last seqno/timestamp on voice packet with marker bit"
Joshua Colp [Mon, 26 Nov 2018 19:47:32 +0000 (13:47 -0600)]
Merge "RTP: need to reset DTMF last seqno/timestamp on voice packet with marker bit"

9 months agostasis: Segment channel snapshot to reduce creation cost.
Joshua Colp [Wed, 7 Nov 2018 17:18:34 +0000 (13:18 -0400)]
stasis: Segment channel snapshot to reduce creation cost.

When a channel snapshot was created it used to be done
from scratch, copying all data (many strings). This incurs
a cost when doing so.

This change segments the channel snapshot into different
components which can be reused if unchanged from the
previous snapshot creation, reducing the cost. In normal
cases this results in some pointers being copied with
reference count being bumped, some integers being set,
and a string or two copied. The other benefit is that it
is now possible to determine if a channel snapshot update
is redundant and thus stop it before a message is published
to stasis.

The specific segments in the channel snapshot were split up
based on whether they are changed together, how often they
are changed, and their general grouping. In practice only
1 (or 0) of the segments actually get changed in normal
operation.

Invalidation is done by setting a flag on the channel when
the segment source is changed, forcing creation of a new
segment when the channel snapshot is created.

ASTERISK-28119

Change-Id: I5d7ef3df963a88ac47bc187d73c5225c315f8423

9 months agostasis: Use an implementation specific channel snapshot cache.
Joshua Colp [Wed, 10 Oct 2018 14:28:18 +0000 (11:28 -0300)]
stasis: Use an implementation specific channel snapshot cache.

Channels no longer use the Stasis cache for channel snapshots. Instead
they are stored in a hash table in stasis_channels which reduces the
number of Stasis messages created and allows better storage.

As a result the following APIs are no longer available since the stasis
cache is no longer used:
ast_channel_topic_cached()
ast_channel_topic_all_cached()

The ast_channel_cache_all() and ast_channel_cache_by_name() functions
now return an ao2_container of ast_channel_snapshots rather than
a container of stasis_messages therefore you can't (and don't need
to) call stasis_cache functions on it.

The ast_channel_topic_all() function now returns a normal topic not
a cached one so you can't use stasis cache functions on it either.

The ast_channel_snapshot_type() stasis message now has the
ast_channel_snapshot_update structure as it's data. It contains the
last snapshot and the new one.

ast_channel_snapshot_get_latest() still returns the latest snapshot.

The latest snapshot is now stored on the channel itself to eliminate
cache hits when Stasis messages that have the snapshot as a payload
are created.

ASTERISK-28102

Change-Id: I9334febff60a82d7c39703e49059fa3a68825786

9 months agoMerge "func_strings: HASHKEY - negative array index can cause corruption"
Joshua Colp [Mon, 26 Nov 2018 13:44:11 +0000 (07:44 -0600)]
Merge "func_strings: HASHKEY - negative array index can cause corruption"

9 months agojansson: Upgrade to 2.12.
Corey Farrell [Mon, 26 Nov 2018 12:09:11 +0000 (07:09 -0500)]
jansson: Upgrade to 2.12.

This brings in jansson-2.12, removes all patches that were merged
upstream.  README is created in third-party/jansson/patches to explain
how to add patches but also because the patches folder must exist for
the build process to succeed.

Change-Id: If0f2d541c50997690660c21fb7b03d625a5cdadd

9 months agoRTP: need to reset DTMF last seqno/timestamp on voice packet with marker bit
Alexei Gradinari [Fri, 23 Nov 2018 15:40:50 +0000 (10:40 -0500)]
RTP: need to reset DTMF last seqno/timestamp on voice packet with marker bit

The marker bit set on the voice packet indicates the start
of a new stream and a new time stamp.
Need to reset the DTMF last sequence number and the timestamp
of the last END packet.

If the new time stamp is lower then the timestamp of the last DTMF END packet
the asterisk drops all DTMF frames as out of order.

This bug was caught using Cisco ip-phone SPA50X and codec g722.
On SIP session update the SPA50X resets stream indicating it with market bit
and a new timestamp is twice smaller then the previous.

ASTERISK-28162 #close

Change-Id: If9c5742158fa836ad549713a9814d46a5d2b1620

9 months agoastobj2: Remove legacy ao2_container_alloc routine.
Corey Farrell [Mon, 19 Nov 2018 20:10:02 +0000 (15:10 -0500)]
astobj2: Remove legacy ao2_container_alloc routine.

Replace usage of ao2_container_alloc with ao2_container_alloc_hash or
ao2_container_alloc_list.  Remove ao2_container_alloc macro.

Change-Id: I0907d78bc66efc775672df37c8faad00f2f6c088

9 months agoastobj2: Create function to copy weak proxied objects from container.
Corey Farrell [Wed, 14 Nov 2018 11:02:20 +0000 (06:02 -0500)]
astobj2: Create function to copy weak proxied objects from container.

Create ao2_container_dup_weakproxy_objs to perform a similar function to
ao2_container_dup.  This function expects the source container to have
weakproxy objects, inserts the associated non-weak objects into the
destination container.  Orphaned weakproxy objects are ignored.

Create test for this new function and for ao2_weakproxy_find.

Change-Id: I898387f058057e08696fe9070f8cd94ef3a27482

9 months agoMerge "stasis: Remove stringfields and lock from change message."
Joshua Colp [Tue, 20 Nov 2018 11:06:01 +0000 (05:06 -0600)]
Merge "stasis: Remove stringfields and lock from change message."

9 months agoMerge "app_queue: Cleanup queue_ref / queue_unref routines."
Joshua Colp [Tue, 20 Nov 2018 11:05:32 +0000 (05:05 -0600)]
Merge "app_queue: Cleanup queue_ref / queue_unref routines."

9 months agofunc_strings: HASHKEY - negative array index can cause corruption
Kevin Harwell [Fri, 16 Nov 2018 20:45:23 +0000 (14:45 -0600)]
func_strings: HASHKEY - negative array index can cause corruption

This patch makes it so only matching non-empty key names, and keys created by
the HASH function are eligible for inclusion in the comma separated string. It
also fixes a bug where it was possible to write to a negative index if the
result buffer was empty.

ASTERISK-28159
patches:
  ASTERISK-28159.diff submitted by Michael Walton (license 6502)

Change-Id: I6e57fe7307dfd856271753aed5ba64c59b511487

9 months agoCI: Get job timeouts from environment
George Joseph [Mon, 19 Nov 2018 17:59:07 +0000 (10:59 -0700)]
CI: Get job timeouts from environment

The job timeouts were hard coded in the jenkinsfiles which
means changes had to go through gerrit.  Now they are taken
from the following environment variables (and their defaults) that
can be set in Jenkins configuration...

TIMEOUT_GATES =      "60 MINUTES"
TIMEOUT_DAILIES =    "3 HOURS"
TIMEOUT_REF_DEBUG =  "24 HOURS"
TIMEOUT_UNITTESTS =  "30 MINUTES"

Change-Id: I673a551c1780bf665a3bc160b245da574aa4bbab

9 months agoMerge "test_res_pjsip_scheduler: Fix possible write after free in scheduler_policy."
Joshua Colp [Mon, 19 Nov 2018 15:38:50 +0000 (09:38 -0600)]
Merge "test_res_pjsip_scheduler: Fix possible write after free in scheduler_policy."

9 months agoMerge "res/res_ari: Fix null endpoint handle"
Joshua Colp [Mon, 19 Nov 2018 15:37:22 +0000 (09:37 -0600)]
Merge "res/res_ari: Fix null endpoint handle"

9 months agoMerge "bridge_native_rtp.c: Fail native bridge if no framing match."
Joshua Colp [Mon, 19 Nov 2018 15:36:17 +0000 (09:36 -0600)]
Merge "bridge_native_rtp.c: Fail native bridge if no framing match."

9 months agoMerge "res_pjsip_caller_id: Use static pj_str_t for fromto header names."
Joshua Colp [Mon, 19 Nov 2018 14:40:05 +0000 (08:40 -0600)]
Merge "res_pjsip_caller_id: Use static pj_str_t for fromto header names."

9 months agoMerge "stasis: Add internal filtering of messages."
Joshua Colp [Mon, 19 Nov 2018 14:36:50 +0000 (08:36 -0600)]
Merge "stasis: Add internal filtering of messages."

9 months agoapp_queue: Cleanup queue_ref / queue_unref routines.
Corey Farrell [Mon, 19 Nov 2018 13:00:03 +0000 (08:00 -0500)]
app_queue: Cleanup queue_ref / queue_unref routines.

This replaces the inline functions with macros.  This removes the need
to directly use __ao2_ref, opts instead for standard ao2_bump and
ao2_cleanup macros.

Change-Id: If4e04e9bab2e3c883188437cb9f487b3e498a21b

9 months agobacktrace: Refactor ast_bt_get_symbols so it doesn't crash
George Joseph [Thu, 8 Nov 2018 15:53:44 +0000 (08:53 -0700)]
backtrace:  Refactor ast_bt_get_symbols so it doesn't crash

We've been seeing crashes in libbfd when we attempt to generate
a stack trace from multiple threads.  It turns out that libbfd
is NOT thread-safe.  It can cache the bfd structure and give it to
multiple threads without protecting itself.  To get around this,
we've added a global mutex around the bfd functions and also have
refactored the use of those functions to be more efficient and
to provide more information about inlined functions.

Also added a few more tests to test_pbx.c.  One just calls
ast_assert() and the other calls ast_log_backtrace().  Neither are
run by default.

WARNING:  This change necessitated changing the return value of
ast_bt_get_symbols() from an array of strings to a VECTOR of
strings.  However, the use of this function outside Asterisk is not
likely.

ASTERISK-28140

Change-Id: I79d02862ddaa2423a0809caa4b3b85c128131621

9 months agostasis: Remove stringfields and lock from change message.
Joshua C. Colp [Sun, 18 Nov 2018 23:53:14 +0000 (19:53 -0400)]
stasis: Remove stringfields and lock from change message.

When a subscribe or unsubscribe occurs a message is published
containing this information. This change makes it so that the
message no longer uses stringfields or a lock, as both are not
really needed for the message.

Change-Id: I3f4831931d79f94fd979baf48048738df5dc1632

9 months agopjsip: New function PJSIP_PARSE_URI to parse URI and return part of URI
Alexei Gradinari [Tue, 13 Nov 2018 15:28:28 +0000 (10:28 -0500)]
pjsip: New function PJSIP_PARSE_URI to parse URI and return part of URI

New dialplan function PJSIP_PARSE_URI added to parse an URI and return
a specified part of the URI.

This is useful when need to get part of the URI instead of cutting it
using a CUT function.

For example to get 'user' part of Remote URI
${PJSIP_PARSE_URI(${CHANNEL(pjsip,remote_uri)},user)}

ASTERISK-28144 #close

Change-Id: I5d828fb87f6803b6c1152bb7b44835f027bb9d5a

9 months agoMerge "pjproject-bundled: Use AST_DEVMODE for conditional compilation."
George Joseph [Sun, 18 Nov 2018 20:11:58 +0000 (14:11 -0600)]
Merge "pjproject-bundled: Use AST_DEVMODE for conditional compilation."

9 months agostasis: Add internal filtering of messages.
Joshua Colp [Sun, 23 Sep 2018 20:50:01 +0000 (17:50 -0300)]
stasis: Add internal filtering of messages.

This change adds the ability for subscriptions to indicate
which message types they are interested in accepting. By
doing so the filtering is done before being dispatched
to the subscriber, reducing the amount of work that has
to be done.

This is optional and if a subscriber does not add
message types they wish to accept and set the subscription
to selective filtering the previous behavior is preserved
and they receive all messages.

There is also the ability to explicitly force the reception
of all messages for cases such as AMI or ARI where a large
number of messages are expected that are then generically
converted into a different format.

ASTERISK-28103

Change-Id: I99bee23895baa0a117985d51683f7963b77aa190

9 months agoCI: Add tmpfs to all jenkinsfiles
George Joseph [Sun, 18 Nov 2018 16:38:40 +0000 (09:38 -0700)]
CI:  Add tmpfs to all jenkinsfiles

Change-Id: Ida29d70d48d5f39aabf0b25c66b51f79324a8cba

9 months agoMerge "res/res_pjsip_nat: Fix logic for REINVITES"
George Joseph [Sun, 18 Nov 2018 14:47:10 +0000 (08:47 -0600)]
Merge "res/res_pjsip_nat: Fix logic for REINVITES"

10 months agoMerge "taskprocessor: Prevent race creating new taskprocessor."
George Joseph [Sat, 17 Nov 2018 23:28:41 +0000 (17:28 -0600)]
Merge "taskprocessor: Prevent race creating new taskprocessor."

10 months agoCI: Mount a tmpfs on /tmp for testsuite docker containers
George Joseph [Sat, 17 Nov 2018 21:40:46 +0000 (14:40 -0700)]
CI:  Mount a tmpfs on /tmp for testsuite docker containers

Change-Id: I0566d81b0852f22066cd76d58eae5f1fda5602aa
(cherry picked from commit 73efe86436427e5f43c532e5d42505ab4ec104d9)

10 months agoCI: Pass work directory to runTestsuite
George Joseph [Sat, 17 Nov 2018 19:07:32 +0000 (12:07 -0700)]
CI:  Pass work directory to runTestsuite

The testsuite can now use a user-specified work directory for
all it's temp files.  This allows the docker containers to use
a tmpfs backed directory for the temp files instead of it's
own write-layer image.

* runTestsuite.sh now accepts a --work-dir command line argument
  that gets exported as AST_WORK_DIR before running the testsuite.

* gates.jenkinsfile now specifies --work-dir to be
  <testsuite_dir>/astroot.

Since the Asterisk CI docker hosts now mount /srv/jenkins/workspace
on a tmpfs, asterisk should be compiled and the testsuite run all in
memory.

Change-Id: If5ee905a15821296c355bb84cda38950ad8edc45
(cherry picked from commit a335f4c9adb0a00211345634f61917bdf5b412c2)

10 months agoMerge "CI: Allow runUnittests to use 'expect' to run the tests"
George Joseph [Sat, 17 Nov 2018 17:30:30 +0000 (11:30 -0600)]
Merge "CI: Allow runUnittests to use 'expect' to run the tests"

10 months agores/res_ari: Fix null endpoint handle
Sungtae Kim [Sat, 17 Nov 2018 02:33:20 +0000 (03:33 +0100)]
res/res_ari: Fix null endpoint handle

The res_ari(POST /channels/create handler) deos not check the endpoint
parameter length. And it causes core
dump.
Fixed it to check the parameter length. Also fixed memory leak.

ASTERISK-28169

Change-Id: Ibf10a9eb8a2e3a9ee1e13fbe748b2ecf955c3993

10 months agoMerge "core: Fix handling of restart from remote console."
George Joseph [Fri, 16 Nov 2018 15:22:26 +0000 (09:22 -0600)]
Merge "core: Fix handling of restart from remote console."

10 months agoCI: Allow runUnittests to use 'expect' to run the tests
George Joseph [Thu, 15 Nov 2018 17:41:44 +0000 (10:41 -0700)]
CI: Allow runUnittests to use 'expect' to run the tests

There seems to be a race condition between starting the asterisk
daemon and attempting to use 'asterisk -r' that can cause the
control socket file to not be created.  Since all of the Jenkins
slaves have 'expect' installed, the runUnittests script can use
it to start asterisk in the forground and issue the commands
interactively.  This is much more reliable and it can also make
startup errors more visible since they'll be in the Jenkins console
output.

If 'expect' isn't installed, the original daemon/asterisk -r
process is used.

Also added a "core show settings" before running the tests
and added "notice,warning,error" to the console log.

Change-Id: Idd656085f854afede813ac241b9e312b31358160

10 months agotaskprocessor: Prevent race creating new taskprocessor.
Corey Farrell [Mon, 12 Nov 2018 18:23:34 +0000 (13:23 -0500)]
taskprocessor: Prevent race creating new taskprocessor.

Task processors are retrieved using a 'get or create' pattern.  The
singleton container was unlocked between the get and create steps so
it's possible that two threads could create task processors with the
same name at the same time.

Change-Id: Id64fae94a6a1e940ddf38fde622dcd4391635382

10 months agopjproject-bundled: Use AST_DEVMODE for conditional compilation.
Corey Farrell [Fri, 16 Nov 2018 12:20:11 +0000 (07:20 -0500)]
pjproject-bundled: Use AST_DEVMODE for conditional compilation.

We previously allowed resample and g711 codecs to be built when
TEST_FRAMEWORK was enabled.  This could cause errors if the testsuite
was run without this option enabled.  Switch the build system to allow
those codecs to be built when --enable-dev-mode is used.  This removes a
chance for strange testsuite errors from use of an inadequate pjsua
binary.

Change-Id: Iee8a3613cdb711fa7e7d217c5a775a575907ae22

10 months agores_pjsip_caller_id: Use static pj_str_t for fromto header names.
Corey Farrell [Thu, 15 Nov 2018 20:47:50 +0000 (15:47 -0500)]
res_pjsip_caller_id: Use static pj_str_t for fromto header names.

PJSIP assumes that these header names are not allocated, does not clone
the name strings when reusing headers.

Block unload of res_pjsip_caller_id until shutdown to ensure static
memory stays valid.  It was previously unsafe to unload while any
sessions are active.

Change-Id: I190854dea943d6e441cf03733f8a0da661aea27f

10 months agoMerge "pbx_config: Only the first [globals] section is seen."
George Joseph [Thu, 15 Nov 2018 13:48:20 +0000 (07:48 -0600)]
Merge "pbx_config: Only the first [globals] section is seen."

10 months agores/res_pjsip_nat: Fix logic for REINVITES
Torrey Searle [Wed, 24 Oct 2018 12:38:37 +0000 (14:38 +0200)]
res/res_pjsip_nat: Fix logic for REINVITES

The presence of Record-Route in re-invites is optional, thus it is
important to make sure the dialog doesn't have a routset before
rewriting the contact header.

ASTERISK-28129 #close

Change-Id: Ic8ceb54ccfc93f7e315e476c514a2c777f2da7dc

10 months agocore: Fix handling of restart from remote console.
Corey Farrell [Thu, 15 Nov 2018 11:33:11 +0000 (06:33 -0500)]
core: Fix handling of restart from remote console.

We cannot use need_el_end and SIGURG when restarting.  Instead we need
to run el_end within the SIGHUP restartnow handler.

ASTERISK-28158

Change-Id: Ia852276363c81bdcf1aa29eb4558c5c2fa1218a0

10 months agoAST-2018-010: Fix length of buffer needed for SRV and NAPTR results
George Joseph [Thu, 25 Oct 2018 15:25:58 +0000 (09:25 -0600)]
AST-2018-010: Fix length of buffer needed for SRV and NAPTR results

When dn_expand was being called on SRV and NAPTR results, the
return value was being used to calculate the size of the buffer
needed to store the host names.  Since dn_expand returns the
length of the COMPRESSED name the buffer could be too short
to hold the EXPANDED name.  The expanded name is NULL terminated
so using strlen() is the correct way to determine the length
actually needed for the buffer.

ASTERISK-28127
Reported by: Jan Hoffmann

patches:
  patch.diff submitted by janhoffmann (license 6986)

Change-Id: I4d35d6c431c6c6836cb61d37b1378cc47f0b414d

10 months agoMerge "core: Ensure that el_end is always run when needed."
Joshua Colp [Wed, 14 Nov 2018 13:06:38 +0000 (07:06 -0600)]
Merge "core: Ensure that el_end is always run when needed."

10 months agoMerge "taskprocessor: Do not use separate allocation for stats or name."
Joshua Colp [Wed, 14 Nov 2018 13:05:10 +0000 (07:05 -0600)]
Merge "taskprocessor: Do not use separate allocation for stats or name."

10 months agoMerge "jansson-bundled: Patch for off-nominal crash."
George Joseph [Tue, 13 Nov 2018 20:39:37 +0000 (14:39 -0600)]
Merge "jansson-bundled: Patch for off-nominal crash."

10 months agotest_res_pjsip_scheduler: Fix possible write after free in scheduler_policy.
Corey Farrell [Tue, 13 Nov 2018 16:51:00 +0000 (11:51 -0500)]
test_res_pjsip_scheduler: Fix possible write after free in scheduler_policy.

It's possible for a 4th task to be spawned before we cancel.  This
results in a write to the already freed test_data1.  Wait long enough to
verify success of the cancelation before freeing test_data1.

Change-Id: I057e2fcbe97f8a175e50890be89c28c20490a20f

10 months agobridge_native_rtp.c: Fail native bridge if no framing match.
Robert Cripps [Wed, 17 Oct 2018 13:48:13 +0000 (15:48 +0200)]
bridge_native_rtp.c: Fail native bridge if no framing match.

ASTERISK-28110 #close

Change-Id: Ic64b8fc6a140a93fbdb2f97550a40d0ff334e607

10 months agotaskprocessor: Do not use separate allocation for stats or name.
Corey Farrell [Mon, 12 Nov 2018 00:32:11 +0000 (19:32 -0500)]
taskprocessor: Do not use separate allocation for stats or name.

Merge storage for the stats object and name string into the main
allocation for struct ast_taskprocessor.

Change-Id: I74fe9a7f357f0e6d63152f163cf5eef6428218e1

10 months agoMerge "res_pjsip.c: Make taskprocessor scheduling algorithm pick the shortest queue"
Joshua Colp [Mon, 12 Nov 2018 11:38:44 +0000 (05:38 -0600)]
Merge "res_pjsip.c: Make taskprocessor scheduling algorithm pick the shortest queue"

10 months agocore: Ensure that el_end is always run when needed.
Corey Farrell [Sun, 11 Nov 2018 13:34:53 +0000 (08:34 -0500)]
core: Ensure that el_end is always run when needed.

* Ignore console=yes configuration option in remote console processes.
* Use new flag to tell consolethread to run el_end and exit when needed.

ASTERISK-28158

Change-Id: I9e23b31d4211417ddc88c6bbfd83ea4c9f3e5438

10 months agojansson-bundled: Patch for off-nominal crash.
Corey Farrell [Thu, 8 Nov 2018 21:37:34 +0000 (16:37 -0500)]
jansson-bundled: Patch for off-nominal crash.

pack_string crashed on non-NULL strings returned when s->has_error was
true if the string was the result of 's' format without '#', '%' or '+'.

Change-Id: Ic125df691d81ba2cbc413e37bdae657b304d20d0