23 months agores_pjproject: Fix cleanup of buildopts vector.
Corey Farrell [Tue, 10 Oct 2017 03:51:03 +0000 (23:51 -0400)]
res_pjproject: Fix cleanup of buildopts vector.


Change-Id: I3bed0edf3f55b1d4adcbabb25ec14f11dc766c72

23 months agoMerge "cdr.c: Replace redundant check with an ast_assert()"
Jenkins2 [Tue, 10 Oct 2017 01:04:33 +0000 (20:04 -0500)]
Merge "cdr.c: Replace redundant check with an ast_assert()"

23 months agoMerge "cdr.h: Fix doxygen comments."
Jenkins2 [Tue, 10 Oct 2017 00:01:19 +0000 (19:01 -0500)]
Merge "cdr.h: Fix doxygen comments."

23 months agoMerge "cdr.c: Use current ao2 flag names"
Joshua Colp [Mon, 9 Oct 2017 23:08:26 +0000 (18:08 -0500)]
Merge "cdr.c: Use current ao2 flag names"

23 months agoMerge "res_pjsip: Fix issues that prevented shutdown of modules."
Jenkins2 [Mon, 9 Oct 2017 22:29:26 +0000 (17:29 -0500)]
Merge "res_pjsip: Fix issues that prevented shutdown of modules."

23 months agoMerge "cdr.c: Replace inlined code with ao2_t_replace()"
Jenkins2 [Mon, 9 Oct 2017 21:52:29 +0000 (16:52 -0500)]
Merge "cdr.c: Replace inlined code with ao2_t_replace()"

23 months agoMerge "res_pjsip_session: Fix format_cap leak."
Jenkins2 [Mon, 9 Oct 2017 21:23:01 +0000 (16:23 -0500)]
Merge "res_pjsip_session: Fix format_cap leak."

23 months agoMerge "res_config_sqlite: Don't enable SQLite CDRs when running 'make samples'"
Jenkins2 [Mon, 9 Oct 2017 21:20:55 +0000 (16:20 -0500)]
Merge "res_config_sqlite: Don't enable SQLite CDRs when running 'make samples'"

23 months agoMerge "res_pjsip: Fix leak of persistent endpoint references."
Jenkins2 [Mon, 9 Oct 2017 20:24:49 +0000 (15:24 -0500)]
Merge "res_pjsip: Fix leak of persistent endpoint references."

23 months agocdr.c: Replace redundant check with an ast_assert()
Richard Mudgett [Tue, 3 Oct 2017 21:09:58 +0000 (16:09 -0500)]
cdr.c: Replace redundant check with an ast_assert()

The only caller of cdr_object_fn_table.process_party_b() explicitly does
the check before calling.

Change-Id: Ib0c53cdf5048227842846e0df9d2c19117c45618

23 months agocdr.c: Replace inlined code with ao2_t_replace()
Richard Mudgett [Mon, 2 Oct 2017 22:41:12 +0000 (17:41 -0500)]
cdr.c: Replace inlined code with ao2_t_replace()

Change-Id: I9f424f5282ca7d833592f958d95f1b2bafb549b0

23 months agocdr.c: Use current ao2 flag names
Richard Mudgett [Fri, 29 Sep 2017 17:07:14 +0000 (12:07 -0500)]
cdr.c: Use current ao2 flag names

Change-Id: Ib59d7d2f2a4a822754628f2c48a308d6791a6e6e

23 months agocdr.h: Fix doxygen comments.
Richard Mudgett [Fri, 29 Sep 2017 17:31:18 +0000 (12:31 -0500)]
cdr.h: Fix doxygen comments.

* Also some misc formatting in cdr.c.

Change-Id: Ied89a28802a662c37c43326a1aafdce596e0df4a

23 months agoMerge "res_sdp_translator_pjmedia: Fix test unregistration."
Jenkins2 [Mon, 9 Oct 2017 17:10:34 +0000 (12:10 -0500)]
Merge "res_sdp_translator_pjmedia: Fix test unregistration."

23 months agores_config_sqlite: Don't enable SQLite CDRs when running 'make samples'
Sean Bright [Mon, 9 Oct 2017 13:15:54 +0000 (09:15 -0400)]
res_config_sqlite: Don't enable SQLite CDRs when running 'make samples'

Change-Id: I65a5190b2732b2246d67472db70dd37db64ddad4

23 months agores/res_ari.c Fix: Memory leaks in ARI when using Content-Type: application/json
hajekd [Sun, 8 Oct 2017 19:05:56 +0000 (21:05 +0200)]
res/res_ari.c Fix: Memory leaks in ARI when using Content-Type: application/json

Reported by: David Hajek
Tested by: David Hajek

Change-Id: Ife3e289062e6cf7d0e7d342dbf79ed96feff441e

23 months agores_pjsip_session: Fix format_cap leak.
Corey Farrell [Sat, 7 Oct 2017 20:47:53 +0000 (16:47 -0400)]
res_pjsip_session: Fix format_cap leak.


Change-Id: I2c8d3fc148f9f53715c958314e1146f9611741f3

23 months agoMerge "vector: multiple evaluation of elem in AST_VECTOR_ADD_SORTED."
Joshua Colp [Sat, 7 Oct 2017 19:46:49 +0000 (14:46 -0500)]
Merge "vector: multiple evaluation of elem in AST_VECTOR_ADD_SORTED."

23 months agores_pjsip: Fix leak of persistent endpoint references.
Corey Farrell [Thu, 5 Oct 2017 21:26:14 +0000 (17:26 -0400)]
res_pjsip: Fix leak of persistent endpoint references.

Do not manually call sip_endpoint_apply_handler from load_all_endpoints.
This is not necessary and causes memory leaks.

Additionally reinitialize persistent->aors when we reuse a persistent
object with a new endpoint.


Change-Id: I59bbfc8da8a14d5f4af8c5bb1e71f8592ae823eb

23 months agoMerge "main/strings: Fix uninitialized value."
Jenkins2 [Fri, 6 Oct 2017 19:49:11 +0000 (14:49 -0500)]
Merge "main/strings: Fix uninitialized value."

23 months agoMerge "res_pjsip: Fix leak of fake_auth references."
Jenkins2 [Fri, 6 Oct 2017 19:16:56 +0000 (14:16 -0500)]
Merge "res_pjsip: Fix leak of fake_auth references."

23 months agovector: multiple evaluation of elem in AST_VECTOR_ADD_SORTED.
Corey Farrell [Thu, 5 Oct 2017 22:59:06 +0000 (18:59 -0400)]
vector: multiple evaluation of elem in AST_VECTOR_ADD_SORTED.

Use temporary variable to prevent multiple evaluations of elem argument.
This resolves a memory leak in res_pjproject startup.

ASTERISK-27317 #close

Change-Id: Ib960d7f5576f9e1a3c478ecb48995582a574e06d

23 months agores_pjsip: Fix leak of fake_auth references.
Corey Farrell [Thu, 5 Oct 2017 20:54:12 +0000 (16:54 -0400)]
res_pjsip: Fix leak of fake_auth references.

pjsip_distributor leaks references to fake_auth when the default realm
has not changed.


Change-Id: I3fcf103b3680ad2d1d4610dcd6738eeaebf4d202

23 months agomain/strings: Fix uninitialized value.
Corey Farrell [Fri, 6 Oct 2017 01:23:31 +0000 (21:23 -0400)]
main/strings: Fix uninitialized value.

ast_strings_match uses sscanf and checks for non-zero return to verify a
token was parsed. This is incorrect as sscanf returns EOF (-1) for errors.

ASTERISK-27318 #close

Change-Id: Ifcece92605f58116eff24c5a0a3b0ee08b3c87b1

23 months agores_sdp_translator_pjmedia: Fix test unregistration.
Corey Farrell [Fri, 6 Oct 2017 00:55:31 +0000 (20:55 -0400)]
res_sdp_translator_pjmedia: Fix test unregistration.


Change-Id: Ib3ed47167cb697ab7bd0a56cab589893f491651b

23 months agores_pjsip_caller_id chan_sip: Comply to RFC 3323 values for privacy
Daniel Tryba [Mon, 2 Oct 2017 12:48:41 +0000 (14:48 +0200)]
res_pjsip_caller_id chan_sip: Comply to RFC 3323 values for privacy

Currently privacy requests are only granted if the Privacy header
value is exactly "id" (defined in RFC 3325). It ignores any other
possible value (or a combination there of). This patch reverses the
logic from testing for "id" to grant privacy, to testing for "none" and
granting privacy for any other value. "none" must not be used in
combination with any other value (RFC 3323 section 4.2).

ASTERISK-27284 #close

Change-Id: If438a21f31a962da32d7a33ff33bdeb1e776fe56

23 months agoMerge "app_queue.c: Fix announcements when announce-to-first-user not enabled."
Joshua Colp [Wed, 4 Oct 2017 20:06:49 +0000 (15:06 -0500)]
Merge "app_queue.c: Fix announcements when announce-to-first-user not enabled."

23 months agores_pjsip: Fix issues that prevented shutdown of modules.
Corey Farrell [Wed, 4 Oct 2017 15:46:44 +0000 (11:46 -0400)]
res_pjsip: Fix issues that prevented shutdown of modules.

res_pjsip and res_pjsip_session had circular references, preventing both
modules from shutting down.
* Move session supplement registration to res_pjsip.
* Use create internal functions for use by pjsip_message_filter.c.


Change-Id: Ifbd5c19ec848010111afeab2436f9699da06ba6b

23 months agores_calendar_icalendar: Filter out occurrences superceded by another VEVENT
krells [Thu, 28 Sep 2017 07:56:14 +0000 (09:56 +0200)]
res_calendar_icalendar: Filter out occurrences superceded by another VEVENT

When we are loading the calendars, we call libical's
icalcomponent_foreach_recurrence method for each VEVENT component that
we have in our calendar.

That method has no knowledge concerning the existence of the other
VEVENT components and will feed our callback with all ocurrences
matching the requested time span.

The occurrences generated by icalcomponent_foreach_recurrence while
expanding a recurring VEVENT's RRULE and RDATE properties can be
superceded by an other VEVENT sharing the same UID.

I use an external iterator (in libical terminology) to avoid messing
with the internal ones from the calling function, and search for
VEVENTS which could supersede the current occurrence.

The event which can invalidate this occurence needs to have:

- the same UID as our recurrent component (comp)
- a RECURRENCE-ID property, which represents the start time of this

If one component is found, just clean and return.

ASTERISK-27296 #close
Reported by: Benoît Dereck-Tricot

Change-Id: I8587ae3eaa765af7cb21eda3b6bf84e8a1c87af8

23 months agoMerge "heap.c: No need to calloc heap pointer array."
Jenkins2 [Wed, 4 Oct 2017 01:41:08 +0000 (20:41 -0500)]
Merge "heap.c: No need to calloc heap pointer array."

23 months agoMerge "logger: Bring back ability to turn debug on by source file"
Joshua Colp [Wed, 4 Oct 2017 00:33:32 +0000 (19:33 -0500)]
Merge "logger:  Bring back ability to  turn debug on by source file"

23 months agoapp_queue.c: Fix announcements when announce-to-first-user not enabled.
Richard Mudgett [Thu, 28 Sep 2017 22:37:15 +0000 (17:37 -0500)]
app_queue.c: Fix announcements when announce-to-first-user not enabled.

The previous patch for ASTERISK-27216 made it so you wouldn't get any
position or periodic announcements unless you had announce-to-first-user
enabled.  The announce-to-first-user feature was added by ASTERISK_21782
as a result of the patch which introduced the redundant announcements that
ASTERISK-27216 removes.

* By noting that the makeannouncement variable is used to suppresses the
first user announcement, we set its initial value to the
announce-to-first-user enable setting.


Change-Id: Ieaeb7dbea8ae7073086b775fbafe0625b000b10a

23 months agoheap.c: No need to calloc heap pointer array.
Richard Mudgett [Thu, 21 Sep 2017 19:43:09 +0000 (14:43 -0500)]
heap.c: No need to calloc heap pointer array.

Change-Id: I5ae2f316229f336eb90d99c7af7ed07a33097e68

23 months agoMerge "pjsip_message_filter: Fix regression causing bad contact address"
Jenkins2 [Thu, 28 Sep 2017 18:36:28 +0000 (13:36 -0500)]
Merge "pjsip_message_filter: Fix regression causing bad contact address"

23 months agoMerge "res_stasis: Add 'video_sfu' as a requested bridge type."
Joshua Colp [Thu, 28 Sep 2017 18:13:31 +0000 (13:13 -0500)]
Merge "res_stasis: Add 'video_sfu' as a requested bridge type."

23 months agoMerge "res_pjsip_session: outgoing call did not offer all configured codecs"
Joshua Colp [Thu, 28 Sep 2017 17:24:11 +0000 (12:24 -0500)]
Merge "res_pjsip_session: outgoing call did not offer all configured codecs"

23 months agologger: Bring back ability to turn debug on by source file
George Joseph [Wed, 27 Sep 2017 18:45:21 +0000 (12:45 -0600)]
logger:  Bring back ability to  turn debug on by source file

Somewhere along the way we lost the ability to debug individual
source files.  For modules, this wasn't a big deal but all the
source files in ./main are in the one "core" module so debugging
individual core capabilities was almost impossible.

* Added a test to DEBUG_ATLEAST that also checks __FILE__ instead
of just module name.  Any source file will work even if it's in
a module subdirectory.

Change-Id: Icc0af41837f3b1679dec7af21fa32cd1f7469f6e

23 months agoMerge "pjproject: Patch to correct STUN FINGERPRINT usage"
Jenkins2 [Thu, 28 Sep 2017 13:30:54 +0000 (08:30 -0500)]
Merge "pjproject: Patch to correct STUN FINGERPRINT usage"

23 months agoMerge "res_rtp_asterisk.c: Fix bridge_p2p_rtp_write() reentrancy potential."
Joshua Colp [Thu, 28 Sep 2017 12:08:33 +0000 (07:08 -0500)]
Merge "res_rtp_asterisk.c: Fix bridge_p2p_rtp_write() reentrancy potential."

23 months agoMerge "res_rtp_asterisk: Trim trailing byte off of SDES packet"
Joshua Colp [Thu, 28 Sep 2017 11:45:28 +0000 (06:45 -0500)]
Merge "res_rtp_asterisk: Trim trailing byte off of SDES packet"

23 months agores_stasis: Add 'video_sfu' as a requested bridge type.
Joshua Colp [Thu, 28 Sep 2017 10:33:00 +0000 (10:33 +0000)]
res_stasis: Add 'video_sfu' as a requested bridge type.

This change adds 'video_sfu' as a requested bridge type when
creating a bridge. By specifying this a mixing type bridge is
created that exchanges video in an SFU fashion.

Change-Id: I2ada47cf5f3fc176518b647c0b4aa39d55339606

23 months agores_pjsip_outbound_publish.c: Fix misplaced parenthesis.
Richard Mudgett [Wed, 27 Sep 2017 16:16:16 +0000 (11:16 -0500)]
res_pjsip_outbound_publish.c: Fix misplaced parenthesis.

The pjsip_publishc_init() call was referenced with a misplaced
parentheses.  As a result, outbound publication messages went out with an
expiration of 1 second.


Change-Id: I93622eabc8ee83e7a22e98c107f921284c605a08

23 months agopjsip_message_filter: Fix regression causing bad contact address
George Joseph [Tue, 26 Sep 2017 16:01:48 +0000 (10:01 -0600)]
pjsip_message_filter: Fix regression causing bad contact address

The "res_pjsip:  Filter out non SIP(S) requests" commit moved the
filtering of messages to pjproject's PJSIP_MOD_PRIORITY_TRANSPORT_LAYER
in order to filter out incoming bad uri schemes as early as possible.
Since the change affected outgoing messages as well and the TRANSPORT
layer is the last to be run on outgoing messages, we were overwriting
the setting of external_signaling_address (which is set earlier by
res_pjsip_nat) with an internal address.

* pjsip_message_filter now registers itself as a pjproject module
twice.  Once in the TSX layer for the outgoing messages (as it was
originally), then a second time in the TRANSPORT layer for the
incoming messages to catch the invalid uri schemes.

Reported by: Sean Bright

Change-Id: I2c90190c43370f8a9d1c4693a19fd65840689c8c

23 months agores_rtp_asterisk.c: Fix bridge_p2p_rtp_write() reentrancy potential.
Richard Mudgett [Thu, 14 Sep 2017 02:31:52 +0000 (21:31 -0500)]
res_rtp_asterisk.c: Fix bridge_p2p_rtp_write() reentrancy potential.

The bridge_p2p_rtp_write() has potential reentrancy problems.

* Accessing the bridged RTP members must be done with the instance1 lock
held.  The DTMF and asymmetric codec checks must be split to be done with
the correct RTP instance struct locked.  i.e., They must be done when
working on the appropriate side of the point to point bridge.

* Forcing the RTP mark bit was referencing the wrong side of the point to
point bridge.  The set mark bit is used everywhere else to set the mark
bit when sending not receiving.

The patches for ASTERISK_26745 and ASTERISK_27158 did not take into
account that not everything carried by RTP uses a codec.  The telephony
DTMF events are not exchanged with a codec.  As a result when
RFC2833/RFC4733 sent digits you would crash if "core set debug 1" is
enabled, the DTMF digits would always get passed to the core even though
the local native RTP bridge is active, and the DTMF digits would go out
using the wrong SSRC id.

* Add protection for non-format payload types like DTMF when updating the
lastrxformat and lasttxformat.  Also protect against non-format payload
types when checking for asymmetric codecs.


Change-Id: I6344ab7de21e26f84503c4d1fca1a41579364186

23 months agores_rtp_asterisk: Trim trailing byte off of SDES packet
Sean Bright [Tue, 26 Sep 2017 15:55:29 +0000 (11:55 -0400)]
res_rtp_asterisk: Trim trailing byte off of SDES packet

This could have been fixed by subtracting 1 from the final value of
'len' but the way the packet was being constructed was confusing so I
took the opportunity to (I think) make it more clear.

We were sending 1 extra byte at the end of the SDES RTCP packet which
caused Chrome to complain (in its debug log):

    Too little data (1 byte) remaining in buffer to parse
    RTCP header (4 bytes).

We now send the correct number of bytes.

Change-Id: I9dcf087cdaf97da0374ae0acb7d379746a71e81b

23 months agoMerge "webrtc: Allow 'webrtc' to be set on endpoints without dtls_ca_file"
Joshua Colp [Tue, 26 Sep 2017 12:37:22 +0000 (07:37 -0500)]
Merge "webrtc: Allow 'webrtc' to be set on endpoints without dtls_ca_file"

23 months agoMerge "channel.c: Fix invalid reference in conditionaled out code."
Joshua Colp [Tue, 26 Sep 2017 12:29:02 +0000 (07:29 -0500)]
Merge "channel.c: Fix invalid reference in conditionaled out code."

23 months agoMerge "app_queue: Only do announcement logic between ringing cycles"
Joshua Colp [Tue, 26 Sep 2017 11:36:47 +0000 (06:36 -0500)]
Merge "app_queue: Only do announcement logic between ringing cycles"

23 months agoMerge "res_pjsip_session: Reduce (and improve) SDP renegotiation."
Joshua Colp [Mon, 25 Sep 2017 20:35:11 +0000 (15:35 -0500)]
Merge "res_pjsip_session: Reduce (and improve) SDP renegotiation."

23 months agowebrtc: Allow 'webrtc' to be set on endpoints without dtls_ca_file
Sean Bright [Mon, 25 Sep 2017 18:00:53 +0000 (14:00 -0400)]
webrtc: Allow 'webrtc' to be set on endpoints without dtls_ca_file

If using a legitimate certificate from a trusted certificate authority,
you don't need to provide CA file.

Change-Id: I8623973b4209b44889243716d7880274caed8a6d

23 months agopjproject: Patch to correct STUN FINGERPRINT usage
Sean Bright [Mon, 25 Sep 2017 18:09:33 +0000 (14:09 -0400)]
pjproject: Patch to correct STUN FINGERPRINT usage

Change-Id: I0e453253dff1388b0186b36c754457c1d0d12db6

23 months agoMerge "build: A few gcc 7 error fixes"
Joshua Colp [Mon, 25 Sep 2017 18:04:22 +0000 (13:04 -0500)]
Merge "build:  A few gcc 7 error fixes"

23 months agores_pjsip_session: outgoing call did not offer all configured codecs
Kevin Harwell [Mon, 25 Sep 2017 17:30:56 +0000 (12:30 -0500)]
res_pjsip_session: outgoing call did not offer all configured codecs

For some scenarios when an outgoing call was made only a subset of the
configured codecs were offered. If the codecs being offered happened to
not have a codec supported by the phone then the call would fail.

For instance Alice and Bob both are configured in Asterisk for g722 and ulaw(
allow=!all,g722,ulaw). Alice's endpoint however only supports g722 while Bob's
only supports ulaw. When Alice calls Bob, Alice negotiates g722 fine with
Asterisk. But when Asterisk sends the outgoing offer to Bob it only contains
g722 and not both g722 and ulaw, so the call ends.

This patch makes it so all the audio codecs configured on the endpoint always
get sent, and not just a subset. However priority is given to those codecs that
are compatible with the "other side".

ASTERISK-27259 #close

Change-Id: Iffabc373bd94cd1dc700925dcfe406e12918c696

23 months agoMerge "res_pjsip_session: Don't end session when receiving a 500 on a reinvite"
Joshua Colp [Mon, 25 Sep 2017 17:23:53 +0000 (12:23 -0500)]
Merge "res_pjsip_session: Don't end session when receiving a 500 on a reinvite"

23 months agochannel.c: Fix invalid reference in conditionaled out code.
Richard Mudgett [Mon, 25 Sep 2017 15:59:17 +0000 (10:59 -0500)]
channel.c: Fix invalid reference in conditionaled out code.


Change-Id: I7a415948116493050614d9f4fa91ffbe0c21ec4c

23 months agoMerge "app_stream_echo: Don't echo declined streams"
Joshua Colp [Mon, 25 Sep 2017 13:00:58 +0000 (08:00 -0500)]
Merge "app_stream_echo: Don't echo declined streams"

23 months agobuild: A few gcc 7 error fixes
George Joseph [Mon, 25 Sep 2017 12:25:06 +0000 (06:25 -0600)]
build:  A few gcc 7 error fixes

Change-Id: I7b5300fbf1af7d88d47129db13ad6dbdc9b553ec

23 months agoMerge "res_pjsip: Use ast_sip_is_content_type() where appropriate"
Joshua Colp [Mon, 25 Sep 2017 12:28:41 +0000 (07:28 -0500)]
Merge "res_pjsip: Use ast_sip_is_content_type() where appropriate"

23 months agoapp_queue: Only do announcement logic between ringing cycles
StefanEng86 [Fri, 15 Sep 2017 07:59:59 +0000 (09:59 +0200)]
app_queue: Only do announcement logic between ringing cycles

This patch reverts the change by patch 2263 from old reviewboard.
Note that reverting that 2263-patch still preserves the behaviour that
the commit log of the 2263-patch claimed to add. The reason for this is:

The function wait_for_answer is only called from try_calling which
in turn is only called from the main for loop in queue_exec, and
earlier in that loop we already check the things that's removed by
this patch. There's no need to check those things twice each loop
iteration, and I think the proper place to check it is before each
ringing cycle. By checking it in wait_for_answer, you allow the issue
explained in the jira - that the head caller hears announcements while
the agents' sip phones are actively ringing.

Reported-by: Stefan Engström
Tested-by: Stefan Engström
ASTERISK-27216 #close

Change-Id: Ic4290dc75256f9743900c6762ee1bb915f672db0

23 months agoapp_stream_echo: Don't echo declined streams
Sean Bright [Sat, 23 Sep 2017 17:32:26 +0000 (13:32 -0400)]
app_stream_echo: Don't echo declined streams

Discovered while experimenting with Cyber Mega Phone 2K Ultimate Dynamic
Edition after accepting the audio request but declining the video one.

Change-Id: Iaa86d41fccfbc1b559a30ccf740d78a3b5f8a98c

23 months agores_pjsip_session: Reduce (and improve) SDP renegotiation.
Joshua Colp [Fri, 22 Sep 2017 22:49:21 +0000 (22:49 +0000)]
res_pjsip_session: Reduce (and improve) SDP renegotiation.

When pruning a request to change the topology of a channel be
more intelligent about the resulting topology that is actually
used for SDP renegotiation.

In a case where a stream has not already been negotiated we
don't need to renegotiate and offer a declined stream. This can
occur if something in Asterisk (such as ConfBridge) requests
to add video to a PJSIP channel that has no video codecs configured.
In this case since the stream did not already exist we can safely
remove the stream from the requested topology, resulting in no
renegotiation occurring.

In a case where a renegotiation is requested with a codec that is
not supported we can reuse the formats of the existing stream if
it exists to ensure that the stream continues to flow, instead of
removing it.

Change-Id: I636540798d55922377318fe619c510fb6ed125fb

23 months agores_pjsip_session: Don't end session when receiving a 500 on a reinvite
Kevin Harwell [Fri, 22 Sep 2017 20:29:24 +0000 (15:29 -0500)]
res_pjsip_session: Don't end session when receiving a 500 on a reinvite

During a reinvite, if a remote endpoint error occurs and it returns a 500 the
session would end. This patch makes it so the session is not terminated, but
continues as it was.

The reason for this is because some endpoints may send non session terminating
"server errors" like a failed codec negotiation. So in this case instead of
ending the call it can hopefully continue. In the case of a real server error
the session is already "doomed", will be known soon enough and appropriately
ended by Asterisk later.

Change-Id: Ifeedae86b8cb44b92d52c79046522ec5f0aff1d5

23 months agoMerge "res_pjsip_session/BUNDLE: Handle no audio codecs on endpoint"
Joshua Colp [Fri, 22 Sep 2017 20:35:29 +0000 (15:35 -0500)]
Merge "res_pjsip_session/BUNDLE:  Handle no audio codecs on endpoint"

23 months agoMerge "res_pjsip_session: Change some asserts to warning/debug messages"
Joshua Colp [Fri, 22 Sep 2017 16:10:11 +0000 (11:10 -0500)]
Merge "res_pjsip_session:  Change some asserts to warning/debug messages"

23 months agores_pjsip: Use ast_sip_is_content_type() where appropriate
Sean Bright [Fri, 22 Sep 2017 15:02:11 +0000 (11:02 -0400)]
res_pjsip: Use ast_sip_is_content_type() where appropriate

Change-Id: If3ab0d73d79ac4623308bd48508af2bfd554937d

23 months agores_pjsip_session/BUNDLE: Handle no audio codecs on endpoint
George Joseph [Thu, 21 Sep 2017 14:47:11 +0000 (08:47 -0600)]
res_pjsip_session/BUNDLE:  Handle no audio codecs on endpoint

When an INVITE came in with both audio and video streams but there
were no audio codecs defined for the endpoint, we weren't declining
the audio stream.  Since it's usually the first/transport stream,
when the video stream was processed and tried to use the transport,
it was empty and caused a crash.  We now decline the the stream if
there are no matching codecs so when the video stream is processed,
it's now the first/transport stream and processes normally.

Change-Id: Ic854eda54c95031e66b076ecfae3041d34daa692

23 months agoMerge "res_rtp_asterisk.c: Fix bundled SSRC handling."
Joshua Colp [Fri, 22 Sep 2017 11:42:30 +0000 (06:42 -0500)]
Merge "res_rtp_asterisk.c: Fix bundled SSRC handling."

23 months agoMerge "res_config_pgsql: Fix removed support to previous for versions PostgreSQL...
Joshua Colp [Fri, 22 Sep 2017 10:42:43 +0000 (05:42 -0500)]
Merge "res_config_pgsql: Fix removed support to previous for versions PostgreSQL 9.1"

2 years agoMerge "bridge: Change participant SFU streams when source streams change."
Joshua Colp [Thu, 21 Sep 2017 20:56:47 +0000 (15:56 -0500)]
Merge "bridge: Change participant SFU streams when source streams change."

2 years agores_rtp_asterisk.c: Fix bundled SSRC handling.
Richard Mudgett [Tue, 19 Sep 2017 19:28:37 +0000 (14:28 -0500)]
res_rtp_asterisk.c: Fix bundled SSRC handling.

Assertions in the v15+ AST-2017-008 patches found that we were not
handling the case if the incoming SDP did not specify the required SSRC
attributes for bundled to work.

* Be strict on matching SSRC for bundled instances including the parent
instance.  If the SSRC doesn't match then discard the packet.  Bundled has
to tell us in the SDP signaling what SSRC to expect.  Otherwise, we will
not know how to find the bundled instance structure.

Change-Id: I152830bbff71c662408909042068fada39e617f9

2 years agoMerge "AST-2017-008: Improve RTP and RTCP packet processing."
Joshua Colp [Thu, 21 Sep 2017 19:48:17 +0000 (14:48 -0500)]
Merge "AST-2017-008: Improve RTP and RTCP packet processing."

2 years agoMerge "res_config_pgsql: Add missing \n in debug log and update copyright year"
Jenkins2 [Thu, 21 Sep 2017 18:06:18 +0000 (13:06 -0500)]
Merge "res_config_pgsql: Add missing \n in debug log and update copyright year"

2 years agoMerge "res_pjsip_session: Check for removed stream state."
Joshua Colp [Thu, 21 Sep 2017 17:30:18 +0000 (12:30 -0500)]
Merge "res_pjsip_session: Check for removed stream state."

2 years agobridge: Change participant SFU streams when source streams change.
Joshua Colp [Sat, 16 Sep 2017 14:19:59 +0000 (11:19 -0300)]
bridge: Change participant SFU streams when source streams change.

Some endpoints do not like a stream being reused for a new
media stream. The frame/jitterbuffer can rely on underlying
attributes of the media stream in order to order the packets.
When a new stream takes its place without any notice the
buffer can get confused and the media ends up getting dropped.

This change uses the SSRC change to determine that a new source
is reusing an existing stream and then bridge_softmix renegotiates
each participant such that they see a new media stream. This
causes the frame/jitterbuffer to start fresh and work as expected.


Change-Id: I30ccbdba16ca073d7f31e0e59ab778c153afae07

2 years agoMerge "res_srtp: lower log level of auth failures"
Jenkins2 [Thu, 21 Sep 2017 17:07:57 +0000 (12:07 -0500)]
Merge "res_srtp: lower log level of auth failures"

2 years agoMerge "chan_sip: Expose read-only access to the full SIP INVITE Request-URI"
Jenkins2 [Thu, 21 Sep 2017 16:11:15 +0000 (11:11 -0500)]
Merge "chan_sip: Expose read-only access to the full SIP INVITE Request-URI"

2 years agoMerge "bridge : Fix one-way direct-media when early bridging with native_rtp"
Joshua Colp [Thu, 21 Sep 2017 15:57:56 +0000 (10:57 -0500)]
Merge "bridge : Fix one-way direct-media when early bridging with native_rtp"

2 years agores_pjsip_session: Change some asserts to warning/debug messages
George Joseph [Wed, 20 Sep 2017 15:45:16 +0000 (09:45 -0600)]
res_pjsip_session:  Change some asserts to warning/debug messages

There was an issue reported where an SDP received on a 183 Session
Progress message caused a crash because the pending streams had
already been processed when the OK was received.  In that case the
pending topology was legitimately NULL.  There was an assert for an
incorrect number of streams in the topology but not one for
topology being NULL.  In any case, if you're not in dev-mode the
asserts don't do anything and since the scenario is legit, the
asserts weren't appropriate anyway.

* Changed several asserts to warning or debug messages and return
codes as appropriate.

Reported by: Daniel Heckl

Change-Id: I58daaa9d2938fa980857ab3ec41925ab5ff9c848

2 years agores_config_pgsql: Fix removed support to previous for versions PostgreSQL 9.1
Rodrigo Ramírez Norambuena [Tue, 19 Sep 2017 10:22:50 +0000 (07:22 -0300)]
res_config_pgsql: Fix removed support to previous for versions PostgreSQL 9.1

In PostgreSQL 9.1 the backslash are string literals and not the escape
of characters.

In previous issue ASTERISK_26057 was fixed the use of escape LIKE but the
support for old version of Postgresql than 9.1 was dropped. The sentence
before make was "ESCAPE '\'" but in version before than 9.1  need it to be
as follow "ESCAPE '\\'".


Change-Id: I96d9ee1ed7693ab17503cb36a9cd72847165f949

2 years agores_pjsip_session: Check for removed stream state.
Ben Ford [Fri, 15 Sep 2017 14:43:21 +0000 (09:43 -0500)]
res_pjsip_session: Check for removed stream state.

When a sip session is refreshed, the stream topology is looped
through, checking each stream for compatible formats. This would
cause a crash if the stream state was AST_STREAM_STATE_REMOVED,
since the formats would never be set for this stream, causing
a NULL value to be returned from ast_stream_get_formats. This
commit adds a check for streams with removed states.

Also removed a stray semicolon.

Change-Id: Ic86f8b65a4a26a60885b28b8b1a0b22e1b471d42

2 years agochan_pjsip: Ignore AST_CONTROL_STREAM_TOPOLOGY_CHANGED for now
George Joseph [Tue, 19 Sep 2017 10:44:28 +0000 (04:44 -0600)]

chan_pjsip_indicate was missing a case for the recently added
AST_CONTROL_STREAM_TOPOLOGY_CHANGED condition and was returning an
error and causing the call to be hung up instead of just ignoring

Reported by: Daniel Heckl

Change-Id: I4fecbb00a0b8a853da85155065c1a6bddf235e80

2 years agobridge : Fix one-way direct-media when early bridging with native_rtp
Jean Aunis [Thu, 7 Sep 2017 09:41:09 +0000 (11:41 +0200)]
bridge : Fix one-way direct-media when early bridging with native_rtp

When two channels were early bridged in a native_rtp bridge, the RTP description
on one side was not updated when the other side answered.
This patch forbids non-answered channels to enter a native_rtp bridge, and
triggers a bridge reconfiguration when an ANSWER frame is received.


Change-Id: If1aaee1b4ed9658a1aa91ab715ee0a6413b878df

2 years agoMerge "res_calendar: Plug memory leak and micro-optimization"
Joshua Colp [Wed, 20 Sep 2017 14:44:53 +0000 (09:44 -0500)]
Merge "res_calendar: Plug memory leak and micro-optimization"

2 years agoMerge "res_pjsip_pubsub: Check for Content-Type header in rx_notify_request"
Jenkins2 [Wed, 20 Sep 2017 14:14:12 +0000 (09:14 -0500)]
Merge "res_pjsip_pubsub:  Check for Content-Type header in rx_notify_request"

2 years agores_srtp: lower log level of auth failures
Alexander Traud [Mon, 18 Sep 2017 14:51:15 +0000 (16:51 +0200)]
res_srtp: lower log level of auth failures

Previously, sRTP authentication failures were reported on log level WARNING.
When such failures happen, each RT(C)P packet is affected, spamming the log.
Now, those failures are reported at log level VERBOSE 2. Furthermore, the
amount is further reduced (previously all two seconds, now all three seconds).
Additionally, the new log entry informs whether media (RTP) or statistics (RTCP)
are affected.

ASTERISK-16898 #close

Change-Id: I6c98d46b711f56e08655abeb01c951ab8e8d7fa0

2 years agores_pjsip_pubsub: Check for Content-Type header in rx_notify_request
George Joseph [Tue, 19 Sep 2017 15:38:30 +0000 (09:38 -0600)]
res_pjsip_pubsub:  Check for Content-Type header in rx_notify_request

pubsub_on_rx_notify_request wasn't checking for a null
Content-Type header before checking that it was

Reported by: Ross Beer

Change-Id: Iec2a6c4d2e74af37ff779ecc9fd35644c5c4ea52

2 years agochan_sip: Expose read-only access to the full SIP INVITE Request-URI
David J. Pryke [Tue, 19 Sep 2017 14:34:01 +0000 (10:34 -0400)]
chan_sip: Expose read-only access to the full SIP INVITE Request-URI

Provide a way to get the contents of the the Request URI from the initial SIP
INVITE in dial plan function call. (In this case "${CHANNEL(ruri)}")

Reported by: David J. Pryke
Tested by: David J. Pryke

Change-Id: I1dd4d6988eed1b6c98a9701e0e833a15ef0dac3e

2 years agoapp_confbridge: Only create a channel that records audio.
Joshua Colp [Tue, 19 Sep 2017 12:53:41 +0000 (12:53 +0000)]
app_confbridge: Only create a channel that records audio.

This change makes it so that the conference recorder channel
that is created only contains audio formats and an audio stream.
This is because the underlying application used by ConfBridge to
record, MixMonitor, only allows recording audio.

Having additional streams (and in particular a video stream) can
result in clients needlessly renegotiating to add a video stream
that will never receive video.

Change-Id: I89d38aedc9205eca7741d5435e73e73bb9de97a0

2 years agores_config_pgsql: Add missing \n in debug log and update copyright year
Rodrigo Ramírez Norambuena [Tue, 19 Sep 2017 11:34:28 +0000 (08:34 -0300)]
res_config_pgsql: Add missing \n in debug log and update copyright year

Change-Id: I4ba338ecbdecc6a814a902eddc4121c8ef3cda58

2 years agores_calendar: Plug memory leak and micro-optimization
Sean Bright [Wed, 13 Sep 2017 19:14:25 +0000 (15:14 -0400)]
res_calendar: Plug memory leak and micro-optimization

ast_variables_destroy is NULL safe, so there is no need to check its
argument before passing it.

ASTERISK-25524 #close
Reported by: Jesper

Change-Id: Ib0f8057642e9d471960f1a79fd42e5a3ce587d3b

2 years agocdr_mysql.c: Apply cdrzone to start and answer
alex [Wed, 13 Sep 2017 08:46:27 +0000 (11:46 +0300)]
cdr_mysql.c: Apply cdrzone to start and answer

Change-Id: I7de0a5adc89824a5f2b696fc22c80fc22dff36b0

2 years agoAST-2017-008: Improve RTP and RTCP packet processing.
Richard Mudgett [Fri, 25 Aug 2017 22:01:57 +0000 (17:01 -0500)]
AST-2017-008: Improve RTP and RTCP packet processing.

Validate RTCP packets before processing them.

* Validate that the received packet is of a minimum length and apply the
RFC3550 RTCP packet validation checks.

* Fixed potentially reading garbage beyond the received RTCP record data.

* Fixed rtp->themssrc only being set once when the remote could change
the SSRC.  We would effectively stop handling the RTCP statistic records.

* Fixed rtp->themssrc to not treat a zero value as special by adding
rtp->themssrc_valid to indicate if rtp->themssrc is available.


Make strict RTP learning more flexible.

Direct media can cause strict RTP to attempt to learn a remote address
again before it has had a chance to learn the remote address the first
time.  Because of the rapid relearn requests, strict RTP could latch onto
the first remote address and fail to latch onto the direct media remote
address.  As a result, you have one way audio until the call is placed on
and off hold.

The new algorithm learns remote addresses for a set time (1.5 seconds)
before locking the remote address.  In addition, we must see a configured
number of remote packets from the same address in a row before switching.

* Fixed strict RTP learning from always accepting the first new address
packet as the new stream.

* Fixed strict RTP to initialize the expected sequence number with the
last received sequence number instead of the last transmitted sequence

* Fixed the predicted next sequence number calculation in
rtp_learning_rtp_seq_update() to handle overflow.


Change-Id: Ia2d3aa6e0f22906c25971e74f10027d96525f31c

2 years agoMerge "res_pjsip: Filter out non SIP(S) requests"
Jenkins2 [Fri, 15 Sep 2017 20:37:04 +0000 (15:37 -0500)]
Merge "res_pjsip:  Filter out non SIP(S) requests"

2 years agoMerge "res_calendar: Various fixes"
Joshua Colp [Fri, 15 Sep 2017 13:20:45 +0000 (08:20 -0500)]
Merge "res_calendar: Various fixes"

2 years agores_pjsip: Filter out non SIP(S) requests
George Joseph [Wed, 13 Sep 2017 21:23:54 +0000 (15:23 -0600)]
res_pjsip:  Filter out non SIP(S) requests

Incoming requests with non sip(s) URIs in the Request, To, From
or Contact URIs are now rejected with
PJSIP_SC_UNSUPPORTED_URI_SCHEME (416).  This is performed in
pjsip_message_filter (formerly pjsip_message_ip_updater) and is
done at pjproject's "TRANSPORT" layer before a request can even
reach the distributor.

URIs read by res_pjsip_outbound_publish from pjsip.conf are now
also checked for both length and sip(s) scheme.  Those URIs read
by outbound registration and aor were already being checked for
scheme but their error messages needed to be updated to include
scheme failure as well as length failure.

Change-Id: Ibb2f9f1d2dc7549da562af4cbd9156c44ffdd460

2 years agoMerge "chan_rtp: Use μ-law by default instead of signed linear"
Jenkins2 [Thu, 14 Sep 2017 17:37:43 +0000 (12:37 -0500)]
Merge "chan_rtp: Use μ-law by default instead of signed linear"

2 years agoMerge "tcptls: Change error message to debug."
Joshua Colp [Thu, 14 Sep 2017 17:11:38 +0000 (12:11 -0500)]
Merge "tcptls: Change error message to debug."

2 years agotcptls: Change error message to debug.
Joshua Colp [Thu, 14 Sep 2017 12:54:40 +0000 (12:54 +0000)]
tcptls: Change error message to debug.

The Websocket implementation will steal the underlying stream of
TCP/TLS sessions. This results in an error message being output
about a stream not being present when in reality this is actually

This change moves it to a debug message instead.

Change-Id: I66cc639080b4b4599beadb4faa7d313f2721d094

2 years agores_calendar: Various fixes
Sean Bright [Wed, 13 Sep 2017 19:08:39 +0000 (15:08 -0400)]
res_calendar: Various fixes

* The way that we were looking at XML elements for CalDAV was extremely
  fragile, so use SAX2 for increased robustness.

* Don't complain about a 'channel' not be specified if autoreminder is
  not set. Assume that if 'channel' is not set, we don't want to be

* Fix some truncated CLI output in 'calendar show calendar' and make the
  'Autoreminder' description a bit more clear

ASTERISK-24588 #close
Reported by: Stefan Gofferje

ASTERISK-25523 #close
Reported by: Jesper

Change-Id: I200d11afca6a47e7d97888f286977e2e69874b2c

2 years agochan_rtp: Use μ-law by default instead of signed linear
Sean Bright [Wed, 13 Sep 2017 14:38:11 +0000 (10:38 -0400)]
chan_rtp: Use μ-law by default instead of signed linear

Multicast/Unicast RTP do not use SDP so we need to use a format that
cleanly maps to one of the static RTP payload types. Without this
change, an Originate to a Multicast or Unicast channel without a format
specified would produce no audio on the receiving device.

ASTERISK-21399 #close
Reported by: Tzafrir Cohen

Change-Id: I97e332b566e85da04b0004b9b0daae746cfca0e3