asterisk/asterisk.git
3 years agochan_pjsip: ref leak when checking direct_media_glare
Kevin Harwell [Fri, 18 Mar 2016 19:31:12 +0000 (14:31 -0500)]
chan_pjsip: ref leak when checking direct_media_glare

Fix the reference leak introduced in the following commit:

c534bd58075e2e1a1e4f3b23c435186c71b155fd

ASTERISK-25849

Change-Id: I5cfefd5ee6c1c3a1715c050330aaa10e4d2a5e85

3 years agoMerge "chan_pjsip: transfers with direct media reinvite has wrong address/port"
zuul [Fri, 18 Mar 2016 17:47:14 +0000 (12:47 -0500)]
Merge "chan_pjsip: transfers with direct media reinvite has wrong address/port"

3 years agochan_pjsip: transfers with direct media reinvite has wrong address/port
Kevin Harwell [Wed, 16 Mar 2016 17:37:01 +0000 (12:37 -0500)]
chan_pjsip: transfers with direct media reinvite has wrong address/port

During a transfer involving direct media a race occurs between when the
transferer channel is swapped out, initiating rtp changes/updates, and the
subsequent reinvites.

When Alice, after speaking with Charlie (Bob is on hold), connects Bob and
Charlie invites are sent to each in order to establish the call between them.
Bob is taken off hold and Charlie is told to have his media flow through
Asterisk. However, if before those invites go out the bridge updates Bob's
and/or Charlie's rtp information with direct media data (i.e. address, port)
then the invite(s) will contain the remote data in the SDP instead of the
Asterisk data.

The race occurs in the native bridge glue code when updating the peer. The
direct_media_address can get set twice before sending out the first invite
during call connection. This can happen because the checking/setting of the
direct_media_address happened in one thread while the sending of the invite(s)
happened in another thread.

This fix removes the race condition by moving the checking/setting of the
direct_media_address to be in the same thread as the sending of the invites(s).
This serializes the checking/setting and sending so they can no longer happen
out of order.

ASTERISK-25849 #close

Change-Id: Idfea590175e74f401929a601dba0c91ca1a7f873

3 years agores_pjsip_refer.c: Fix seg fault in process of Refer-to header.
Sergio Medina Toledo [Thu, 3 Mar 2016 10:43:59 +0000 (10:43 +0000)]
res_pjsip_refer.c: Fix seg fault in process of Refer-to header.

The "Refer-to" header of an incoming REFER request is parsed by
pjsip_parse_uri().  That function requires the URI parameter to be NULL
terminated.  Unfortunately, the previous code added the NULL terminator by
overwriting memory that may not be safe.  The overwritten memory results
could be benign, memory corruption, or a segmentation fault.  Now the URI
is NULL terminated safely by copying the URI to a new chunk of memory with
the correct size to be NULL terminated.

ASTERISK-25814 #close

Change-Id: I32565496684a5a49c3278fce06474b8c94b37342

3 years agoMerge "Add initial support to build Docker images"
Joshua Colp [Thu, 17 Mar 2016 17:42:59 +0000 (12:42 -0500)]
Merge "Add initial support to build Docker images"

3 years agoMerge "chan_sip.c: Made sip_reinvite_retry() call sip_pvt_lock_full()."
Joshua Colp [Thu, 17 Mar 2016 16:19:22 +0000 (11:19 -0500)]
Merge "chan_sip.c: Made sip_reinvite_retry() call sip_pvt_lock_full()."

3 years agoAdd initial support to build Docker images
Leif Madsen [Thu, 25 Feb 2016 16:29:05 +0000 (11:29 -0500)]
Add initial support to build Docker images

This work-in-progress is the first step to being able to reliably
build Asterisk containers from the Asterisk source. I'm submitting
this based on feedback gained at AstriDevCon 2015.

Information about how to use this is provided in contrib/docker/README.md
and will result in a local Asterisk container being built right from
your source. I believe this can eventually be automated via
hub.docker.com.

Change-Id: Ifa070706d40e56755797097b6ed72c1e243bd0d1

3 years agochan_sip.c: Fix mwi resub deadlock potential.
Richard Mudgett [Fri, 11 Mar 2016 18:22:48 +0000 (12:22 -0600)]
chan_sip.c: Fix mwi resub deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

ASTERISK-25023 #close

Change-Id: I96d429c57a48861fd8bde63dd93db4e92dc3adb6

3 years agochan_sip.c: Fix registration timeout and expire deadlock potential.
Richard Mudgett [Thu, 10 Mar 2016 23:01:12 +0000 (17:01 -0600)]
chan_sip.c: Fix registration timeout and expire deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

ASTERISK-25023

Change-Id: I2e40de89efc8ae6e8850771d089ca44bc604b508

3 years agochan_sip.c: Fix waitid deadlock potential.
Richard Mudgett [Wed, 9 Mar 2016 22:26:26 +0000 (16:26 -0600)]
chan_sip.c: Fix waitid deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

* Made always run check_pendings() under the scheduler thread so scheduler
ids can be checked safely.

ASTERISK-25023

Change-Id: Ia834d6edd5bdb47c163e4ecf884428a4a8b17d52

3 years agochan_sip.c: Fix t38id deadlock potential.
Richard Mudgett [Thu, 10 Mar 2016 18:17:09 +0000 (12:17 -0600)]
chan_sip.c: Fix t38id deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

ASTERISK-25023

Change-Id: If595e4456cd059d7171880c7f354e844c21b5f5f

3 years agochan_sip.c: Fix session timers deadlock potential.
Richard Mudgett [Tue, 8 Mar 2016 21:08:19 +0000 (15:08 -0600)]
chan_sip.c: Fix session timers deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

ASTERISK-25023

Change-Id: I6d65269151ba95e0d8fe4e9e611881cde2ab4900

3 years agochan_sip.c: Fix reinviteid deadlock potential.
Richard Mudgett [Wed, 9 Mar 2016 22:34:53 +0000 (16:34 -0600)]
chan_sip.c: Fix reinviteid deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

ASTERISK-25023

Change-Id: I9c11b9d597468f63916c99e1dabff9f4a46f84c1

3 years agochan_sip.c: Fix autokillid deadlock potential.
Richard Mudgett [Mon, 7 Mar 2016 19:21:44 +0000 (13:21 -0600)]
chan_sip.c: Fix autokillid deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

* Fix clearing autokillid in __sip_autodestruct() even though we could
reschedule.

ASTERISK-25023

Change-Id: I450580dbf26e2e3952ee6628c735b001565c368f

3 years agochan_sip.c: Fix packet retransid deadlock potential.
Richard Mudgett [Wed, 9 Mar 2016 22:32:28 +0000 (16:32 -0600)]
chan_sip.c: Fix packet retransid deadlock potential.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

* Fix retrans_pkt() to call check_pendings() with both the owner channel
and the private objects locked as required.

* Refactor dialog retransmission packet list to safely remove packet
nodes.  The list nodes are now ao2 objects.  The list has a ref and the
scheduled entry has a ref.

ASTERISK-25023

Change-Id: I50926d81be53f4cd3d572a3292cd25f563f59641

3 years agochan_sip.c: Fix provisional_keepalive_sched_id deadlock.
Richard Mudgett [Tue, 8 Mar 2016 00:28:58 +0000 (18:28 -0600)]
chan_sip.c: Fix provisional_keepalive_sched_id deadlock.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Stopping a scheduled event can result in a deadlock if the scheduled event
is running when you try to stop the event.  If you hold a lock needed by
the scheduled event while trying to stop the scheduled event then a
deadlock can happen.  The general strategy for resolving the deadlock
potential is to push the actual starting and stopping of the scheduled
events off onto the scheduler/do_monitor() thread by scheduling an
immediate one shot scheduled event.  Some restructuring may be needed
because the code may assume that the start/stop of the scheduled events is
immediate.

ASTERISK-25023

Change-Id: I98a694fd42bc81436c83aa92de03226e6e4e3f48

3 years agochan_sip.c: Adjust how dialog_unlink_all() stops scheduled events.
Richard Mudgett [Wed, 9 Mar 2016 17:22:50 +0000 (11:22 -0600)]
chan_sip.c: Adjust how dialog_unlink_all() stops scheduled events.

This patch is part of a series to resolve deadlocks in chan_sip.c.

* Make dialog_unlink_all() unschedule all items at once in the sched
thread.

ASTERISK-25023

Change-Id: I7743072fb228836e8228b72f6dc46c8cc50b3fb4

3 years agochan_sip.c: Clear scheduled immediate events on unload.
Richard Mudgett [Fri, 11 Mar 2016 03:54:03 +0000 (21:54 -0600)]
chan_sip.c: Clear scheduled immediate events on unload.

This patch is part of a series to resolve deadlocks in chan_sip.c.

The reordering of chan_sip's shutdown is to handle any immediate events
that get put onto the scheduler so resources aren't leaked.  The typical
immediate events at this time are going to be concerned with stopping
other scheduled events.

ASTERISK-25023

Change-Id: I3f6540717634f6f2e84d8531a054976f2bbb9d20

3 years agosip/dialplan_functions.c: Fix /channels/chan_sip/test_sip_rtpqos crash.
Richard Mudgett [Tue, 15 Mar 2016 19:51:25 +0000 (14:51 -0500)]
sip/dialplan_functions.c: Fix /channels/chan_sip/test_sip_rtpqos crash.

This patch is part of a series to resolve deadlocks in chan_sip.c.

Delaying destruction of the chan_sip sip_pvt structures caused the
/channels/chan_sip/test_sip_rtpqos unit test to crash.  That test
registers a special test ast_rtp_engine with the rtp engine module.  When
the unit test completes it cleans up by unregistering the test
ast_rtp_engine and exits.  Since the delayed destruction of the sip_pvt
happens after the unit test returns, the destructor tries to call the rtp
engine destroy callback of the test ast_rtp_engine auto variable which no
longer exists on the stack.

* Change the test ast_rtp_engine auto variable to a static variable.  Now
the variable can still exist after the unit test exits so the delayed
sip_pvt destruction can complete successfully.

ASTERISK-25023

Change-Id: I61e34a12d425189ef7e96fc69ae14993f82f3f13

3 years agosched.c: Ensure oldest expiring entry runs first.
Richard Mudgett [Mon, 7 Mar 2016 21:50:22 +0000 (15:50 -0600)]
sched.c: Ensure oldest expiring entry runs first.

This patch is part of a series to resolve deadlocks in chan_sip.c.

* Updated sched unit test to check new behavior.

ASTERISK-25023

Change-Id: Ib69437327b3cda5e14c4238d9ff91b2531b34ef3

3 years agoMerge "app_stasis: Don't hang up if app is not registered"
zuul [Wed, 16 Mar 2016 19:15:44 +0000 (14:15 -0500)]
Merge "app_stasis: Don't hang up if app is not registered"

3 years agoMerge "chan_sip.c: Simplify sip_pvt destructor call levels."
zuul [Wed, 16 Mar 2016 17:14:17 +0000 (12:14 -0500)]
Merge "chan_sip.c: Simplify sip_pvt destructor call levels."

3 years agoapp_stasis: Don't hang up if app is not registered
Andrew Nagy [Tue, 15 Mar 2016 18:31:19 +0000 (11:31 -0700)]
app_stasis: Don't hang up if app is not registered

This prevents pbx_core from hanging up the channel if the app isn't
registered.

ASTERISK-25846 #close

Change-Id: I63216a61f30706d5362bc0906b50b6f0544aebce

3 years agoMerge "pjproject: Pass (dont_)optimize flags to pjproject and fix pjsua"
zuul [Tue, 15 Mar 2016 22:40:02 +0000 (17:40 -0500)]
Merge "pjproject:  Pass (dont_)optimize flags to pjproject and fix pjsua"

3 years agoMerge "build_system: Split COMPILE_DOUBLE from DONT_OPTIMIZE"
Joshua Colp [Tue, 15 Mar 2016 20:55:27 +0000 (15:55 -0500)]
Merge "build_system:  Split COMPILE_DOUBLE from DONT_OPTIMIZE"

3 years agoMerge "build: Add configure check for proto field of PJSIP TLS transport setting."
zuul [Tue, 15 Mar 2016 15:27:09 +0000 (10:27 -0500)]
Merge "build: Add configure check for proto field of PJSIP TLS transport setting."

3 years agoMerge "res_pjsip_refer.c: Delay sending the initial SIP Notify with frag 100"
Joshua Colp [Tue, 15 Mar 2016 13:47:43 +0000 (08:47 -0500)]
Merge "res_pjsip_refer.c: Delay sending the initial SIP Notify with frag 100"

3 years agochan_sip.c: Simplify sip_pvt destructor call levels.
Richard Mudgett [Tue, 8 Mar 2016 00:56:05 +0000 (18:56 -0600)]
chan_sip.c: Simplify sip_pvt destructor call levels.

Remove destructor calling destroy_it calling really_destroy_it
for no benefit.  Just make the destructor the really_destroy_it
function.

Change-Id: Idea0d47b27dd74f2488db75bcc7f353d8fdc614a

3 years agochan_sip.c: Made sip_reinvite_retry() call sip_pvt_lock_full().
Richard Mudgett [Sat, 5 Mar 2016 00:25:21 +0000 (18:25 -0600)]
chan_sip.c: Made sip_reinvite_retry() call sip_pvt_lock_full().

Change-Id: I90f04208a089f95488a2460185a8dbc3f6acca12

3 years agobuild: Add configure check for proto field of PJSIP TLS transport setting.
Joshua Colp [Mon, 14 Mar 2016 13:59:10 +0000 (10:59 -0300)]
build: Add configure check for proto field of PJSIP TLS transport setting.

Older versions of PJSIP do not have the proto field on the TLS transport
setting structure. This change adds a configure check so even if it is
not present we will still be able to build.

Change-Id: Ibf3f47befb91ed1b8194bf63888baa6fee05aba9

3 years agobuild_system: Split COMPILE_DOUBLE from DONT_OPTIMIZE
George Joseph [Sat, 12 Mar 2016 22:02:20 +0000 (15:02 -0700)]
build_system:  Split COMPILE_DOUBLE from DONT_OPTIMIZE

I can't ever recall actually needing the intermediate files or the checking
that a double compile produces.  What I CAN remember is every DONT_OPTIMIZE
build needing 3 invocations of gcc instead of 1 just to do the checks and
produce those intermediate files.

Having said that, Richard pointed out that the reason for the double compile
was that there were cases in the past where a submitted patch failed to compile
because the submitter never tried it with the optimizations turned on.

To get the best of both worlds, COMPILE_DOUBLE has been split into its own
option.  If DONT_OPTIMIZE is turned on, COMPILE_DOUBLE will also be selected
BUT you can then turn it off if all you need are the debugging symbols.  This
way you have to make an informed decision about disabling COMPILE_DOUBLE.

To allow COMPILE_DOUBLE to be both auto-selected and turned off, a new feature
was added to menuselect.  The <use> element can now contain an "autoselect"
attribute which will turn the used member on but not create a hard dependency.
The cflags.xml implementation for COMPILE_DOUBLE looks like this...

<member name="DONT_OPTIMIZE" displayname="Disable Optimizations ...">
<use autoselect="yes">COMPILE_DOUBLE</use>
<support_level>core</support_level>
</member>
<member name="COMPILE_DOUBLE" displayname="Pre-compile with ...>
<depend>DONT_OPTIMIZE</depend>
<support_level>core</support_level>
</member>

When DONT_OPTIMIZE is turned on, COMPILE_DOUBLE is turned on because
of the use.
When DONT_OPTIMIZE is turned off, COMPILE_DOUBLE is turned off because
of the depend.
When COMPILE_DOUBLE is turned on, DONT_OPTIMIZE is turned on because
of the depend.
When COMPILE_DOUBLE is turned off, DONT_OPTIMIZE is left as is because
it only uses COMPILE_DOUBLE, it doesn't depend on it.

I also made a few tweaks to the ncurses implementation to move things
left a bit to allow longer descriptions.

Change-Id: Id49ca930ac4b5ec4fc2d8141979ad888da7b1611

3 years agopjproject: Pass (dont_)optimize flags to pjproject and fix pjsua
George Joseph [Thu, 10 Mar 2016 19:09:13 +0000 (12:09 -0700)]
pjproject:  Pass (dont_)optimize flags to pjproject and fix pjsua

The pjproject Makefile now uses the Asterisk optimization flags which
are determined by the setting of the DONT_OPTMIZE menuselect flag.
The Makefile was also restructured so a change to the top level
menuselect.makeopts will result in a rebuild of pjproject.

Also, "--disable-resample" was removed from the pjproject configure
options.  Without resample, pjsua (which is used by the testsuite)
can't make audio calls.  When it can't, it segfaults.

Change-Id: I24b0a4d0872acef00ed89b3c527a713ee4c2ccd4

3 years agoapp_chanspy: Fix occasional deadlock with ChanSpy and Local channels.
Walter Doekes [Fri, 11 Mar 2016 22:03:08 +0000 (23:03 +0100)]
app_chanspy: Fix occasional deadlock with ChanSpy and Local channels.

Channel masquerading had a conflict with autochannel locking.

When locking autochannel->channel, the channel is fetched from the
autochannel and then locked. During the fetch, the autochannel -- which
has no locks itself -- can be modified by someone who owns the channel
lock. That means that the value of autochan->channel cannot be trusted
until you hold the lock.

In practice, this caused problems with Local channels getting
masqueraded away while the ChanSpy attempted to get info from that
channel. The old channel which was about to get removed got locked, but
the new (replaced) channel got unlocked (no-op). Because the replaced
channel was now locked (and would never get unlocked), it couldn't get
removed from the channel list in a timely manner, and would now cause
deadlocks when iterating over the channel list.

This change checks the autochannel after locking the channel for changes
to the autochannel. If the channel had been changed, the lock is
reobtained on the new channel.

In theory it seems possible that after this fix, the lock attempt on the
old (wrong) channel can be on an already destroyed lock, maybe causing
a crash. But that hasn't been observed in the wild and is harder induce
than the current deadlock.

Thanks go to Filip Frank for suggesting a fix similar to this and
especially to IRC user hexanol for pointing out why this deadlock was
possible and testing this fix. And to Richard for catching my rookie
while loop mistake ;)

ASTERISK-25321 #close

Change-Id: I293ae0014e531cd0e675c3f02d1d118a98683def

3 years agoMerge "install_prereq: Add packages for bundled pjproject"
zuul [Thu, 10 Mar 2016 12:44:25 +0000 (06:44 -0600)]
Merge "install_prereq: Add packages for bundled pjproject"

3 years agoMerge "res_pjsip_caller_id: Anonymize 'From' when caller id presentation is prohibited"
zuul [Wed, 9 Mar 2016 02:36:47 +0000 (20:36 -0600)]
Merge "res_pjsip_caller_id: Anonymize 'From' when caller id presentation is prohibited"

3 years agoMerge "pjproject_bundled: Remove --with-external-pa from configure options."
zuul [Tue, 8 Mar 2016 23:04:55 +0000 (17:04 -0600)]
Merge "pjproject_bundled: Remove --with-external-pa from configure options."

3 years agoMerge "res_pjsip: Strip spaces from items parsed from comma-separated lists"
zuul [Tue, 8 Mar 2016 18:27:15 +0000 (12:27 -0600)]
Merge "res_pjsip:  Strip spaces from items parsed from comma-separated lists"

3 years agoMerge "main/cli.c: Refactor function to print seconds formatted"
Joshua Colp [Tue, 8 Mar 2016 17:29:45 +0000 (11:29 -0600)]
Merge "main/cli.c: Refactor function to print seconds formatted"

3 years agoMerge "res_odbc_transaction: fix some format tab"
zuul [Tue, 8 Mar 2016 17:12:37 +0000 (11:12 -0600)]
Merge "res_odbc_transaction: fix some format tab"

3 years agopjproject_bundled: Remove --with-external-pa from configure options.
George Joseph [Tue, 8 Mar 2016 03:34:12 +0000 (20:34 -0700)]
pjproject_bundled: Remove --with-external-pa from configure options.

Not sure why it was there in the first place as we already specify
--disable-sound.

Change-Id: Ia80a40e8b1e1acc287955ab11ba1fbd0c7d4cff9

3 years agores_pjsip: Strip spaces from items parsed from comma-separated lists
George Joseph [Sun, 6 Mar 2016 20:38:41 +0000 (13:38 -0700)]
res_pjsip:  Strip spaces from items parsed from comma-separated lists

Configurations like "aors = a, b, c" were either ignoring everything after "a"
or trying to look up " b".  Same for mailboxes,  ciphers, contacts and a few
others.

To fix, all the strsep(&copy, ",") calls have been wrapped in ast_strip.  To
facilitate this, ast_strip, ast_skip_blanks and ast_skip_nonblanks were
updated to handle null pointers.

In some cases, an ast_strlen_zero() test was added to skip consecutive commas.

There was also an attempt to ast_free an ast_strdupa'd string in
ast_sip_for_each_aor which was causing a SEGV.  I removed it.

Although this issue was reported for realtime, the issue was in the res_pjsip
modules so all config mechanisms were affected.

ASTERISK-25829 #close
Reported-by: Mateusz Kowalski

Change-Id: I0b22a2cf22a7c1c50d4ecacbfa540155bec0e7a2

3 years agores_odbc_transaction: fix some format tab
Rodrigo Ramírez Norambuena [Mon, 7 Mar 2016 08:02:45 +0000 (05:02 -0300)]
res_odbc_transaction: fix some format tab

Change-Id: I265e4ac47c629c9a63dd86b59df82a7ab3c64384

3 years agomain/cli.c: Refactor function to print seconds formatted
Rodrigo Ramírez Norambuena [Thu, 18 Feb 2016 04:58:01 +0000 (01:58 -0300)]
main/cli.c: Refactor function to print seconds formatted

Refactor and created function ast_cli_print_timestr_fromseconds to print
seconds formatted:  year(s) week(s) day(s) hour(s) second(s)

This function now is used in addons/cdr_mysql.c,cdr_pgsql.c, main/cli.c,
res_config_ldap.c, res_config_pgsql.c.

Change-Id: Ibeb8634102cd11d3f8623398b279cb731bcde36c

3 years agoinstall_prereq: Add packages for bundled pjproject
George Joseph [Sat, 5 Mar 2016 02:37:44 +0000 (19:37 -0700)]
install_prereq: Add packages for bundled pjproject

RedHat/CentOS needs python-devel
Debian/Ubuntu needs automake, libsrtp-dev and python-dev

Ubuntu also needed libncurses5-dev for cmenuselect so while not
needed for pjproject, I adedd it anyway.

Change-Id: Idf5fa16e2d87c687439621507e122cb9461d7089

3 years agoMerge "third_party/Makefile.rules: Replace unsupported != operator with $(shell...
zuul [Fri, 4 Mar 2016 13:04:15 +0000 (07:04 -0600)]
Merge "third_party/Makefile.rules:  Replace unsupported != operator with $(shell ...)"

3 years agoMerge "config_transport: Fix objects returned by ast_sip_get_transport_states"
zuul [Fri, 4 Mar 2016 03:45:39 +0000 (21:45 -0600)]
Merge "config_transport:  Fix objects returned by ast_sip_get_transport_states"

3 years agores_pjsip_caller_id: Anonymize 'From' when caller id presentation is prohibited
George Joseph [Wed, 24 Feb 2016 23:25:09 +0000 (16:25 -0700)]
res_pjsip_caller_id: Anonymize 'From' when caller id presentation is prohibited

Per RFC3325, the 'From' header is now anonymized on outgoing calls when
caller id presentation is prohibited.

TID = trust_id_outbound
PRO = Set(CALLERID(pres)=prohib)
USR = endpoint/from_user
DOM = endpoint/from_domain
PAI = YES(privacy=off), NO(not sent), PRI(privacy=full) (assumes send_pai=yes)

Conditions          |Result
--------------------|----------------------------------------------------
TID PRO USR DOM     |PAI    FROM
--------------------|----------------------------------------------------
Y   Y   abc def.ghi |PRI    "Anonymous" <sip:abc@def.ghi>
Y   Y   abc         |PRI    "Anonymous" <sip:abc@anonymous.invalid>
Y   Y       def.ghi |PRI    "Anonymous" <sip:anonymous@def.ghi>
Y   Y               |PRI    "Anonymous" <sip:anonymous@anonymous.invalid>

Y   N   abc def.ghi |YES    <sip:abc@def.ghi>
Y   N   abc         |YES    <sip:abc@<ip_address>>
Y   N       def.ghi |YES    "Caller Name" <sip:<caller_exten>@def.ghi>
Y   N               |YES    "Caller Name" <sip:<caller_exten>@<ip_address>>

N   Y   abc def.ghi |NO     "Anonymous" <sip:abc@def.ghi>
N   Y   abc         |NO     "Anonymous" <sip:abc@anonymous.invalid>
N   Y       def.ghi |NO     "Anonymous" <sip:anonymous@def.ghi>
N   Y               |NO     "Anonymous" <sip:anonymous@anonymous.invalid>

N   N   abc def.ghi |YES    <sip:abc@def.ghi>
N   N   abc         |YES    <sip:abc@<ip_address>>
N   N       def.ghi |YES    "Caller Name" <sip:<caller_exten>@def.ghi>
N   N               |YES    "Caller Name" <sip:<caller_exten>@<ip_address>>

ASTERISK-25791 #close
Reported-by: Anthony Messina

Change-Id: I2c82a5ca1413c2c00fb62ea95b0ae8e97af54dc9

3 years agoMerge "alembic: Fix downgrade and tweak for sqlite"
zuul [Fri, 4 Mar 2016 02:01:15 +0000 (20:01 -0600)]
Merge "alembic: Fix downgrade and tweak for sqlite"

3 years agoMerge "loader: Retry dlopen when loading fails"
zuul [Fri, 4 Mar 2016 01:57:43 +0000 (19:57 -0600)]
Merge "loader: Retry dlopen when loading fails"

3 years agothird_party/Makefile.rules: Replace unsupported != operator with $(shell ...)
George Joseph [Thu, 3 Mar 2016 23:34:51 +0000 (16:34 -0700)]
third_party/Makefile.rules:  Replace unsupported != operator with $(shell ...)

Apparently the != operator is fairly new so I've replaced it with
the old $(shell ...) syntax.

Change-Id: I16b2e1878a4f91e7e9740abd427f9639f933c479
Reported-by: Richard Mudgett

3 years agoMerge "bridge.c: Crash during attended transfer when missing a local channel half"
zuul [Thu, 3 Mar 2016 23:37:45 +0000 (17:37 -0600)]
Merge "bridge.c: Crash during attended transfer when missing a local channel half"

3 years agoloader: Retry dlopen when loading fails
George Joseph [Sat, 23 Jan 2016 21:50:57 +0000 (14:50 -0700)]
loader: Retry dlopen when loading fails

Although we use the RTLD_LAZY flag when calling dlopen
the first time on a module, this only defers resolution
for function calls.  Pointer references to functions are
determined at link time so dlopen expects them to be there.
Since we don't cross-module link, pointers to functions
in other modules won't be available and dlopen will fail.

Doing a "hardened" build also causes problems because it
typically sets "-z now" on the ld command line which
overrides RTLD_LAZY at run time.

If the failing module isn't a GLOBAL_SYMBOLS module, then
dlopen will be called again after all the GLOBAL_SYMBOLS
modules have been loaded and they'll eventually resolve.

If the calling module IS a GLOBAL_SYMBOLS module itself
and a third module depends on it, then there's an issue
because the second time through the dlopen loop,
GLOBAL_SYMBOLS modules aren't given any special treatment
and since the order in which dlopen is called isn't
deterministic, the dependent may again be tried before the
module it needs is loaded.

Simple solution:  Save modules that fail load_resource
because of a dlopen error in a list and retry them
immediately after the first pass. Keep retrying until
the failed list is empty or we reach a #defined max
retries. Error messages are suppressed until the final
pass which also gets rid of those confusing error messages
about module failures that are later corrected.

Change-Id: Iddae1d97cd2f00b94e61662447432765755f64bb

3 years agoMerge "res_pjsip_dtmf_info: NULL terminate the message body."
zuul [Thu, 3 Mar 2016 20:51:10 +0000 (14:51 -0600)]
Merge "res_pjsip_dtmf_info: NULL terminate the message body."

3 years agobridge.c: Crash during attended transfer when missing a local channel half
Kevin Harwell [Tue, 1 Mar 2016 22:18:21 +0000 (16:18 -0600)]
bridge.c: Crash during attended transfer when missing a local channel half

It's possible for the transferer channel to get hung up early during the
attended transfer process. For instance, a phone may send a "bye" immediately
upon receiving a sip notify that contains a sip frag 100 (I'm looking at you
Jitsi). When this occurs a race begins between the transferer being hung up
and completion of the transfer code.

If the channel hangs up too early during a transfer involving stasis bridging
for instance, then when the created local channel goes to look up its swap
channel (and associated datastore) it can't find it (since it is no longer in
the bridge) thus it fails to enter the stasis application. Consequently, the
created local channel(s) hang up as well. If the timing is just right then the
bridging code attempts to add the message link with missing local channel(s).
Hence the crash.

Unfortunately, there is no great way to solve the problem of the unexpected
"bye". While we can't guarantee we won't receive an early hangup, and in this
case still fail to enter the stasis application, we can make it so asterisk
does not crash.

This patch does just that by locking the local channel structure, checking
that the local channel's peer has not been lost, and then continuing. This
keeps the local channel's peer from being ripped out from underneath it by
the local/unreal hangup code while attempting to set the stasis message link.

ASTERISK-25771

Change-Id: Ie6d6061e34c7c95f07116fffac9a09e5d225c880

3 years agores_pjsip_refer.c: Delay sending the initial SIP Notify with frag 100
Kevin Harwell [Wed, 2 Mar 2016 00:08:52 +0000 (18:08 -0600)]
res_pjsip_refer.c: Delay sending the initial SIP Notify with frag 100

During the transfer process, some phones (okay it was the Jitsi softphone,
but maybe others are out there) send a "bye" immediately after receiving a
SIP Notify. When a "bye" is received early for some types of transfers the
transferer channel may no longer be available during late stage transfer
processing.

For instance, during an attended transfer involving stasis bridging at one
point the created local channel looks for an associated swap channel in
order to retrieve the stasis application name. If the transferer has hung
up then the local channel will fail to find it. The local channel then has
no way to know which stasis app to enter, so it fails and hangs up as well.
Thus the transfer does not complete as expected.

This patch delays the sending of the initial notify in order to give the
transfer process enough time to gather the necessary data for a successful
transfer.

ASTERISK-25771

Change-Id: I09cfc9a5d6ed4c007bc70625e0972b470393bf16

3 years agoMerge "build-system: Allow building with static pjproject"
zuul [Thu, 3 Mar 2016 17:30:41 +0000 (11:30 -0600)]
Merge "build-system: Allow building with static pjproject"

3 years agores_pjsip_dtmf_info: NULL terminate the message body.
Joshua Colp [Thu, 3 Mar 2016 14:26:10 +0000 (10:26 -0400)]
res_pjsip_dtmf_info: NULL terminate the message body.

PJSIP does not ensure that when printing the message body the
buffer will be NULL terminated. This is problematic when searching
for the signal and duration values of the DTMF.

This change ensures the buffer is always NULL terminated.

Change-Id: I52653a1a60c93092d06af31a27408d569cc98968

3 years agoMerge "func_callerid.c: Update REDIRECTING reason documentation."
Joshua Colp [Thu, 3 Mar 2016 13:40:53 +0000 (07:40 -0600)]
Merge "func_callerid.c: Update REDIRECTING reason documentation."

3 years agoMerge "SIP diversion: Fix REDIRECTING(reason) value inconsistencies."
Joshua Colp [Thu, 3 Mar 2016 13:40:40 +0000 (07:40 -0600)]
Merge "SIP diversion: Fix REDIRECTING(reason) value inconsistencies."

3 years agoMerge "res_pjsip_send_to_voicemail.c: Allow either quoted or not send_to_vm reason."
Joshua Colp [Thu, 3 Mar 2016 11:32:59 +0000 (05:32 -0600)]
Merge "res_pjsip_send_to_voicemail.c: Allow either quoted or not send_to_vm reason."

3 years agoMerge "res_pjsip_send_to_voicemail.c: Fix off-nominal double channel unref."
zuul [Thu, 3 Mar 2016 03:24:14 +0000 (21:24 -0600)]
Merge "res_pjsip_send_to_voicemail.c: Fix off-nominal double channel unref."

3 years agoMerge "CHAOS: cleanup possible null vars on msg alloc failure"
zuul [Thu, 3 Mar 2016 00:02:38 +0000 (18:02 -0600)]
Merge "CHAOS: cleanup possible null vars on msg alloc failure"

3 years agoalembic: Fix downgrade and tweak for sqlite
George Joseph [Wed, 2 Mar 2016 02:03:04 +0000 (19:03 -0700)]
alembic: Fix downgrade and tweak for sqlite

Downgrade had a few issues.  First there was an errant 'update' statement in
add_auto_dtmf_mode that looks like it was a copy/paste error.  Second, we
weren't cleaning up the ENUMs so subsequent upgrades on postgres failed
because the types already existed.

For sqlite...  sqlite doesn't support ALTER or DROP COLUMN directly.
Fortunately alembic batch_operations takes care of this for us if we
use it so the alter and drops were converted to use batch operations.

Here's an example downgrade:

    with op.batch_alter_table('ps_endpoints') as batch_op:
        batch_op.drop_column('tos_audio')
        batch_op.drop_column('tos_video')
        batch_op.add_column(sa.Column('tos_audio', yesno_values))
        batch_op.add_column(sa.Column('tos_video', yesno_values))
        batch_op.drop_column('cos_audio')
        batch_op.drop_column('cos_video')
        batch_op.add_column(sa.Column('cos_audio', yesno_values))
        batch_op.add_column(sa.Column('cos_video', yesno_values))

    with op.batch_alter_table('ps_transports') as batch_op:
        batch_op.drop_column('tos')
        batch_op.add_column(sa.Column('tos', yesno_values))
    # Can't cast integers to YESNO_VALUES, so dropping and adding is required
        batch_op.drop_column('cos')
        batch_op.add_column(sa.Column('cos', yesno_values))

Upgrades from base to head and downgrades from head to base were tested
repeatedly for postgresql, mysql/mariadb, and sqlite3.

Change-Id: I862b0739eb3fd45ec3412dcc13c2340e1b7baef8

3 years agoconfig_transport: Fix objects returned by ast_sip_get_transport_states
George Joseph [Wed, 2 Mar 2016 21:55:48 +0000 (14:55 -0700)]
config_transport:  Fix objects returned by ast_sip_get_transport_states

ast_sip_get_transport_states was returning a container of internal_state
objects instead of ast_sip_transport_state objects.  This was causing
transport lookups to fail, most noticably in res_pjsip_nat, which
couldn't find the correct external addresses.  This was causing contacts
to go out with internal ip addresses.

ASTERISK-25830 #close
Reported-by: Sean Bright

Change-Id: I1aee6a2fd46c42e8dd0af72498d17de459ac750e

3 years agoCHAOS: cleanup possible null vars on msg alloc failure
Scott Griepentrog [Wed, 2 Mar 2016 17:17:54 +0000 (11:17 -0600)]
CHAOS: cleanup possible null vars on msg alloc failure

In message.c, if msg_alloc fails to init the string field,
vars may be null, so use a null tolerant cleanup.

In res_pjsip_messaging.c, if msg_data_create fails, mdata
will be null, so use a null tolerant cleanup.

ASTERISK-25323

Change-Id: Ic2d55c2c3750d5616e2a05ea92a19c717507ff56

3 years agoCHAOS: prevent crash on failed strdup
Scott Griepentrog [Wed, 2 Mar 2016 15:34:10 +0000 (09:34 -0600)]
CHAOS: prevent crash on failed strdup

This patch avoids crashing on a null pointer
if the strdup() allocation fails.

ASTERISK-25323

Change-Id: I3f67434820ba53b53663efd6cbb42749f4f6c0f5

3 years agofunc_callerid.c: Update REDIRECTING reason documentation.
Richard Mudgett [Tue, 1 Mar 2016 00:11:33 +0000 (18:11 -0600)]
func_callerid.c: Update REDIRECTING reason documentation.

Change-Id: I6e8d39b0711110a4bceafa652e58b30465e28386

3 years agoSIP diversion: Fix REDIRECTING(reason) value inconsistencies.
Richard Mudgett [Sat, 27 Feb 2016 00:57:17 +0000 (18:57 -0600)]
SIP diversion: Fix REDIRECTING(reason) value inconsistencies.

Previous chan_sip behavior:

Before this patch chan_sip would always strip any quotes from an incoming
reason and pass that value up as the REDIRECTING(reason).  For an outgoing
reason value, chan_sip would check the value against known values and
quote any it didn't recognize.  Incoming 480 response message reason text
was just assigned to the REDIRECTING(reason).

Previous chan_pjsip behavior:

Before this patch chan_pjsip would always pass the incoming reason value
up as the REDIRECTING(reason).  For an outgoing reason value, chan_pjsip
would send the reason value as passed down.

With this patch:

Both channel drivers match incoming reason values with values documented
by REDIRECTING(reason) and values documented by RFC5806 regardless of
whether they are quoted or not.  RFC5806 values are mapped to the
equivalent REDIRECTING(reason) documented value and is set in
REDIRECTING(reason).  e.g., an incoming RFC5806 'unconditional' value or a
quoted string version ('"unconditional"') is converted to
REDIRECTING(reason)'s 'cfu' value.  The user's dialplan only needs to deal
with 'cfu' instead of any of the aliases.

The incoming 480 response reason text supported by chan_sip checks for
known reason values and if not matched then puts quotes around the reason
string and assigns that to REDIRECTING(reason).

Both channel drivers send outgoing known REDIRECTING(reason) values as the
unquoted RFC5806 equivalent.  User custom values are either sent as is or
with added quotes if SIP doesn't allow a character within the value as
part of a RFC3261 Section 25.1 token.  Note that there are still
limitations on what characters can be put in a custom user value.  e.g.,
embedding quotes in the middle of the reason string is silly and just
going to cause you grief.

* Setting a REDIRECTING(reason) value now recognizes RFC5806 aliases.
e.g., Setting REDIRECTING(reason) to 'unconditional' is converted to the
'cfu' value.

* Added missing malloc() NULL return check in res_pjsip_diversion.c
set_redirecting_reason().

* Fixed potential read from a stale pointer in res_pjsip_diversion.c
add_diversion_header().  The reason string needed to be copied into the
tdata memory pool to ensure that the string would always be available.
Otherwise, if the reason string returned by reason_code_to_str() was a
user's reason string then the string could be freed later by another
thread.

Change-Id: Ifba83d23a195a9f64d55b9c681d2e62476b68a87

3 years agores_pjsip_send_to_voicemail.c: Allow either quoted or not send_to_vm reason.
Richard Mudgett [Sat, 27 Feb 2016 00:54:53 +0000 (18:54 -0600)]
res_pjsip_send_to_voicemail.c: Allow either quoted or not send_to_vm reason.

Change-Id: Id6350b3c7d4ec8df7ec89863566645e2b0f441fd

3 years agores_pjsip_send_to_voicemail.c: Fix off-nominal double channel unref.
Richard Mudgett [Tue, 1 Mar 2016 02:41:55 +0000 (20:41 -0600)]
res_pjsip_send_to_voicemail.c: Fix off-nominal double channel unref.

* Fix double unref of other_party channel in off nominal path.

* This is unlikely to be a real problem.  However, for safety,
in handle_incoming_request() keep the datastore ref with the
other_party channel ref until we are finished with the other_party
channel.

Change-Id: I78f22547bf0bb99fb20814ceab75952bd857f821

3 years agobuild-system: Allow building with static pjproject
George Joseph [Tue, 19 Jan 2016 03:54:28 +0000 (20:54 -0700)]
build-system: Allow building with static pjproject

Background here:
http://lists.digium.com/pipermail/asterisk-dev/2016-January/075266.html

From CHANGES:
 * To help insure that Asterisk is compiled and run with the same known
   version of pjproject, a new option (--with-pjproject-bundled) has been
   added to ./configure.  When specified, the version of pjproject specified
   in third-party/versions.mak will be downloaded and configured.  When you
   make Asterisk, the build process will also automatically build pjproject
   and Asterisk will be statically linked to it.  Once a particular version
   of pjproject is configured and built, it won't be configured or built
   again unless you run a 'make distclean'.

   To facilitate testing, when 'make install' is run, the pjsua and pjsystest
   utilities and the pjproject python bindings will be installed in
   ASTDATADIR/third-party/pjproject.

   The default behavior remains building with the shared pjproject
   installation, if any.

Building:

   All you have to do is include the --with-pjproject-bundled option on
   the ./configure command line (and remove any existing --with-pjproject
   option if specified).  Everything else is automatic.

Behind the scenes:

   The top-level Makefile was modified to include 'third-party' in the
   list of MOD_SUBDIRS.

   The third-party directory was created to contain any third party
   packages that may be needed in the future.  Its Makefile automatically
   iterates over any subdirectories passing on targets.

   The third-party/pjproject directory was created to house the pjproject
   source distribution.  Its Makefile contains targets to download, patch
   configure, generate dependencies, compile libs, apps and python bindings,
   sanitized build.mak and generate a symbols list.

   When bootstrap.sh is run, it automatically includes the configure.m4
   file in third-party/pjproject.  This file has a macro to download and
   conifgure pjproject and get and set PJPROJECT_INCLUDE, PJPROJECT_DIR
   and PJPROJECT_BUNDLED.  It also tests for the capabilities like
   PJ_TRANSACTION_GRP_LOCK by parsing preprocessor output as opposed to
   trying to compile.  Of course, bootstrap.sh is only run once and the
   configure file is incldued in the patch.

   When configure is run with the new options, the macro in configure.m4
   triggers the download, patch, conifgure and tests.  No compilation is
   performed at this time.  The downloaded tarball is cached in /tmp so
   it doesn't get downloaded again on a distclean.

   When make is run in the top-level Asterisk source directory, it will
   automatically descend all the subdirectories in third_party just as it
   does for addons, apps, etc.  The top-level Makefile makes sure that
   the 'third-party' is built before 'main' so that dependencies from the
   other directories are built first.

   When main does build, a new shared library (libasteriskpj) is created that
   links statically to the pjproject .a files and exports all their symbols.
   The asterisk binary links to that, just as it does with libasteriskssl.

   When Asterisk is installed, the pjsua and pjsystest apps, and the pjproject
   python bindings are installed in ASTDATADIR/third-party/pjproject.  This
   will facilitate testing, including running the testsuite which will be
   updated to check that directory for the pjsua module ahead of the system
   python library.

Modules should continue to depend on pjproject if they use pjproject APIs
directly.  They should not care about the implementation.  No changes to any
res_pjsip modules were made.

Change-Id: Ia7a60c28c2e9ba9537c5570f933c1ebcb20a3103

3 years agoMerge "chan_sip.c: Fix T.38 issues caused by leaving a bridge."
Joshua Colp [Tue, 1 Mar 2016 12:00:43 +0000 (06:00 -0600)]
Merge "chan_sip.c: Fix T.38 issues caused by leaving a bridge."

3 years agoMerge "res_pjsip_t38.c: Back out part of an earlier fix attempt."
Joshua Colp [Tue, 1 Mar 2016 12:00:35 +0000 (06:00 -0600)]
Merge "res_pjsip_t38.c: Back out part of an earlier fix attempt."

3 years agoMerge "bridge core: Add owed T.38 terminate when channel leaves a bridge."
Joshua Colp [Tue, 1 Mar 2016 12:00:27 +0000 (06:00 -0600)]
Merge "bridge core: Add owed T.38 terminate when channel leaves a bridge."

3 years agoMerge "channel api: Create is_t38_active accessor functions."
Joshua Colp [Tue, 1 Mar 2016 12:00:17 +0000 (06:00 -0600)]
Merge "channel api: Create is_t38_active accessor functions."

3 years agoMerge "bridge_channel: Don't settle owed events on an optimization."
Joshua Colp [Tue, 1 Mar 2016 12:00:03 +0000 (06:00 -0600)]
Merge "bridge_channel: Don't settle owed events on an optimization."

3 years agoMerge "channel.c: Route all control frames to a channel through the same code."
Joshua Colp [Tue, 1 Mar 2016 11:59:54 +0000 (05:59 -0600)]
Merge "channel.c: Route all control frames to a channel through the same code."

3 years agoMerge "res_pjsip_mwi: Turn some NOTICEs and WARNINGs into debug 1s."
zuul [Mon, 29 Feb 2016 22:55:33 +0000 (16:55 -0600)]
Merge "res_pjsip_mwi:  Turn some NOTICEs and WARNINGs into debug 1s."

3 years agochan_sip.c: Fix T.38 issues caused by leaving a bridge.
Richard Mudgett [Mon, 22 Feb 2016 22:59:40 +0000 (16:59 -0600)]
chan_sip.c: Fix T.38 issues caused by leaving a bridge.

chan_sip could not handle AST_T38_TERMINATED frames being sent to it when
the channel left the bridge.  The action resulted in overlapping outgoing
reINVITEs.  The testsuite tests/fax/sip/directmedia_reinvite_t38 was not
happy.

* Force T.38 to be remembered as locally bridged.  Now when the channel
leaves the native RTP bridge after T.38, the channel remembers that it has
already reINVITEed the media back to Asterisk.  It just needs to terminate
T.38 when the AST_T38_TERMINATED arrives.

* Prevent redundant AST_T38_TERMINATED from causing problems.  Redundant
AST_T38_TERMINATED frames could cause overlapping outgoing reINVITEs if
they happen before the T.38 state changes to disabled.  Now the T.38 state
is set to disabled before the reINVITE is sent.

ASTERISK-25582 #close

Change-Id: I53f5c6ce7d90b3f322a942af1a9bcab6d967b7ce

3 years agores_pjsip_t38.c: Back out part of an earlier fix attempt.
Richard Mudgett [Fri, 19 Feb 2016 00:27:02 +0000 (18:27 -0600)]
res_pjsip_t38.c: Back out part of an earlier fix attempt.

This backs out item 4 of the 4875e5ac32f5ccad51add6a4216947bfb385245d
commit.  Item 4 added the t38_bye_supplement.  Unfortunately, the frame
that it puts into the bridge may or may not be processed by the time the
bridged peer is kicked out of the bridge.  If it is processed then all is
well.  However, if it is not processed then that channel is stuck in fax
mode until it hangs up or maybe if it joins another bridge for T.38
faxing.

ASTERISK-25582

Change-Id: Ib20a03ecadf1bf8a0dcadfadf6c2f2e60919a9f7

3 years agobridge core: Add owed T.38 terminate when channel leaves a bridge.
Richard Mudgett [Mon, 22 Feb 2016 19:54:47 +0000 (13:54 -0600)]
bridge core: Add owed T.38 terminate when channel leaves a bridge.

The channel is now going to get T.38 terminated when it leaves the
bridging system and the bridged peers are going to get T.38 terminated as
well.

ASTERISK-25582

Change-Id: I77a9205979910210e3068e1ddff400dbf35c4ca7

3 years agochannel api: Create is_t38_active accessor functions.
Richard Mudgett [Fri, 19 Feb 2016 22:01:17 +0000 (16:01 -0600)]
channel api: Create is_t38_active accessor functions.

ASTERISK-25582

Change-Id: I69451920b122de7ee18d15bb231c80ea7067a22b

3 years agobridge_channel: Don't settle owed events on an optimization.
Richard Mudgett [Sat, 20 Feb 2016 01:06:14 +0000 (19:06 -0600)]
bridge_channel: Don't settle owed events on an optimization.

Local channel optimization could cause DTMF digits to be duplicated.
Pending DTMF end events would be posted to a bridge when the local channel
optimizes out and is replaced by the channel further down the chain.  When
the real digit ends, the channel would get another DTMF end posted to the
bridge.

A -- LocalA;1/n -- LocalA;2/n -- LocalB;1 -- LocalB;2 -- B

1) LocalA has the /n flag to prevent optimization.
2) B is sending DTMF to A through the local channel chain.
3) When LocalB optimizes out it can move B to the position of LocalB;1
4) Without this patch, when B swaps with LocalB;1 then LocalB;1 would
settle an owed DTMF end to the bridge toward LocalA;2.
5) When B finally ends its DTMF it sends the DTMF end down the chain.
6) Without this patch, A would hear the DTMF digit end when LocalB
optimizes out and when B ends the original digit.

ASTERISK-25582

Change-Id: I1bbd28b8b399c0fb54985a5747f330a4cd2aa251

3 years agochannel.c: Route all control frames to a channel through the same code.
Richard Mudgett [Mon, 22 Feb 2016 18:15:34 +0000 (12:15 -0600)]
channel.c: Route all control frames to a channel through the same code.

Frame hooks can conceivably return a control frame in exchange for an
audio frame inside ast_write().  Those returned control frames were not
handled quite the same as if they were sent to ast_indicate().  Now it
doesn't matter if you use ast_write() to send an AST_FRAME_CONTROL to a
channel or ast_indicate().

ASTERISK-25582

Change-Id: I5775f41421aca2b510128198e9b827bf9169629b

3 years agosorcery: Refactor create, update and delete to better deal with caches
George Joseph [Thu, 25 Feb 2016 21:13:19 +0000 (14:13 -0700)]
sorcery:  Refactor create, update and delete to better deal with caches

The ast_sorcery_create, update and delete function have been refactored
to better deal with caches and errors.

The action is now called on all non-caching wizards first. If ANY succeed,
the action is called on all caching wizards and the observers are notified.
This way we don't put something in the cache (or update or delete) before
knowing the action was performed in at least 1 backend and we only call the
observers once even if there were multiple writable backends.

ast_sorcery_create was never adding to caches in the first place which
was preventing contacts from getting added to a memory_cache when they
were created.  In turn this was causing memory_cache to emit errors if
the contact was deleted before being retrieved (which would have
populated the cache).

ASTERISK-25811 #close
Reported-by: Ross Beer

Change-Id: Id5596ce691685a79886e57b0865888458d6e7b46

3 years agores_pjsip_mwi: Turn some NOTICEs and WARNINGs into debug 1s.
George Joseph [Thu, 25 Feb 2016 21:39:54 +0000 (14:39 -0700)]
res_pjsip_mwi:  Turn some NOTICEs and WARNINGs into debug 1s.

There are a few cases where we're emitting notices or warnings
for things that really need neither, like a client retrying to subscribe
to mwi when they're not conifgured for it.  They get a 404 so there's no
need for non-debug messages.

Change-Id: I05e38a7ff6c2f2521146f4be6a79731b9864e61f

3 years agoMerge "res_pjsip/config_transport: Allow reloading transports."
Joshua Colp [Sat, 27 Feb 2016 16:18:26 +0000 (10:18 -0600)]
Merge "res_pjsip/config_transport: Allow reloading transports."

3 years agoMerge "res_sorcery_memory_cache: Fix SEGV in some CLI commands"
Joshua Colp [Sat, 27 Feb 2016 14:50:20 +0000 (08:50 -0600)]
Merge "res_sorcery_memory_cache:  Fix SEGV in some CLI commands"

3 years agoMerge "chan_sip: Optionally supply fromuser/fromdomain in SIP dial string."
zuul [Thu, 25 Feb 2016 23:56:42 +0000 (17:56 -0600)]
Merge "chan_sip: Optionally supply fromuser/fromdomain in SIP dial string."

3 years agores_sorcery_memory_cache: Fix SEGV in some CLI commands
George Joseph [Thu, 25 Feb 2016 20:17:04 +0000 (13:17 -0700)]
res_sorcery_memory_cache:  Fix SEGV in some CLI commands

A few of the CLI commands weren't checking for enough arguments
and were SEGVing.

Change-Id: Ie6494132ad2fe54b4f014bcdc112a37c36a9b413

3 years agoMerge "chan_sip.c: Suppress T.38 SDP c= line if addr is the same."
zuul [Thu, 25 Feb 2016 00:40:15 +0000 (18:40 -0600)]
Merge "chan_sip.c: Suppress T.38 SDP c= line if addr is the same."

3 years agoMerge "res_config_sqlite3: Fix crashes when reading peers from sqlite3 tables"
zuul [Thu, 25 Feb 2016 00:26:06 +0000 (18:26 -0600)]
Merge "res_config_sqlite3: Fix crashes when reading peers from sqlite3 tables"

3 years agoMerge "rtp_engine.h: Remove extraneous semicolons."
zuul [Wed, 24 Feb 2016 16:18:01 +0000 (10:18 -0600)]
Merge "rtp_engine.h: Remove extraneous semicolons."

3 years agoMerge "res_pjsip_config_wizard: Add command to export primitive objects"
zuul [Tue, 23 Feb 2016 23:41:35 +0000 (17:41 -0600)]
Merge "res_pjsip_config_wizard:  Add command to export primitive objects"

3 years agortp_engine.h: Remove extraneous semicolons.
Richard Mudgett [Tue, 23 Feb 2016 01:31:24 +0000 (19:31 -0600)]
rtp_engine.h: Remove extraneous semicolons.

Change-Id: Ib462633d396fa941379dfef648dcd2245e350084

3 years agochan_sip.c: Suppress T.38 SDP c= line if addr is the same.
Richard Mudgett [Tue, 23 Feb 2016 20:57:42 +0000 (14:57 -0600)]
chan_sip.c: Suppress T.38 SDP c= line if addr is the same.

Use the correct comparison function since we only care if the address
without the port is the same.

Change-Id: Ibf6c485f843a1be6dee58a47b33d81a7a8cbe3b0

3 years agoMerge "res_pjproject: Add ability to map pjproject log levels to Asterisk log levels"
Joshua Colp [Mon, 22 Feb 2016 16:55:03 +0000 (10:55 -0600)]
Merge "res_pjproject:  Add ability to map pjproject log levels to Asterisk log levels"

3 years agores_config_sqlite3: Fix crashes when reading peers from sqlite3 tables
Christof Lauber [Tue, 16 Feb 2016 14:14:15 +0000 (15:14 +0100)]
res_config_sqlite3: Fix crashes when reading peers from sqlite3 tables

Introduced realloaction of ast_str buf in sqlite3_escape functions in case
the returned buffer from threadstorage was actually too small.

Change-Id: I3c5eb43aaade93ee457943daddc651781954c445

3 years agores_pjsip/config_transport: Allow reloading transports.
George Joseph [Thu, 11 Feb 2016 17:01:05 +0000 (10:01 -0700)]
res_pjsip/config_transport: Allow reloading transports.

The 'reload' mechanism actually involves closing the underlying
socket and calling the appropriate udp, tcp or tls start functions
again.  Only outbound_registration, pubsub and session needed work
to reset the transport before sending requests to insure that the
pjsip transport didn't get pulled out from under them.

In my testing, no calls were dropped when a transport was changed
for any of the 3 transport types even if ip addresses or ports were
changed. To be on the safe side however, a new transport option was
added (allow_reload) which defaults to 'no'.  Unless it's explicitly
set to 'yes' for a transport, changes to that transport will be ignored
on a reload of res_pjsip.  This should preserve the current behavior.

Change-Id: I5e759850e25958117d4c02f62ceb7244d7ec9edf

3 years agochan_sip: Optionally supply fromuser/fromdomain in SIP dial string.
Walter Doekes [Fri, 19 Feb 2016 10:30:15 +0000 (11:30 +0100)]
chan_sip: Optionally supply fromuser/fromdomain in SIP dial string.

Previously you could add [!dnid] to the SIP dial string to alter the To:
header. This change allows you to alter the From header as well.

SIP dial string extra options now look like this:

    [![touser[@todomain]][![fromuser][@fromdomain]]]

INCOMPATIBLE CHANGE: If you were using an exclamation mark in your To:
header, that is no longer possible.

ASTERISK-25803 #close

Change-Id: I2457e9ba7a89eb1da22084bab5a4d4328e189db7