asterisk/asterisk.git
2 years agores_rtp_asterisk.c: Add "seqno" strictrtp option
Ben Ford [Mon, 10 Sep 2018 16:28:09 +0000 (11:28 -0500)]
res_rtp_asterisk.c: Add "seqno" strictrtp option

When networks experience disruptions, there can be large gaps of time
between receiving packets. When strictrtp is enabled, this created
issues where a flood of packets could come in and be seen as an attack.
Another option - seqno - has been added to the strictrtp option that
ignores the time interval and goes strictly by sequence number for
validity.

Change-Id: I8a42b8d193673899c8fc22fe7f98ea87df89be71

2 years agoMerge "jansson: Backport fixes to bundled, use json_vsprintf if available."
George Joseph [Wed, 26 Sep 2018 16:09:50 +0000 (11:09 -0500)]
Merge "jansson: Backport fixes to bundled, use json_vsprintf if available."

2 years agoMerge "chan_sip: SipNotify on Chan_Sip vi AMI behave different to CLI"
George Joseph [Wed, 26 Sep 2018 14:34:10 +0000 (09:34 -0500)]
Merge "chan_sip: SipNotify on Chan_Sip vi AMI behave different to CLI"

2 years agoMerge "rtp_engine: rtcp_report_to_json can overflow the ssrc integer value"
George Joseph [Wed, 26 Sep 2018 13:02:28 +0000 (08:02 -0500)]
Merge "rtp_engine: rtcp_report_to_json can overflow the ssrc integer value"

2 years agochan_sip: SipNotify on Chan_Sip vi AMI behave different to CLI
pk16208 [Tue, 18 Sep 2018 13:01:02 +0000 (15:01 +0200)]
chan_sip: SipNotify on Chan_Sip vi AMI behave different to CLI

With tls and udp enabled asterisk generates a warning about sending
message via udp instead of tls.
sip notify command via cli works as expected and without warning.

asterisk has to set the connection information accordingly to connection
and not on presumption

ASTERISK-28057 #close

Change-Id: Ib43315aba1f2c14ba077b52d8c5b00be0006656e

2 years agojansson: Backport fixes to bundled, use json_vsprintf if available.
Corey Farrell [Tue, 17 Jul 2018 03:55:02 +0000 (23:55 -0400)]
jansson: Backport fixes to bundled, use json_vsprintf if available.

Use json_vsprintf from versions which contain fix for va_copy leak.

Apply fixes from jansson master:
* va_copy leak fix.
* Avoid potential invalid memory read in json_pack.
* Rename variable that shadowed another.

Change-Id: I7522e462d2a52f53010ffa1e7d705c666ec35539

2 years agojson: Take advantage of new API's.
Corey Farrell [Tue, 17 Jul 2018 03:55:02 +0000 (23:55 -0400)]
json: Take advantage of new API's.

* Use "o*" format specifier for optional fields in ast_json_party_id.
* Stop using ast_json_deep_copy on immutable objects, it is now thread
  safe to just use ast_json_ref.

Additional changes to ast_json_pack calls in the vicinity:
* Use "O" when an object needs to be bumped.  This was previously
  avoided as it was not thread safe.
* Use "o?" and "O?" to replace NULL with ast_json_null().  The
  "?" is a new feature of ast_json_pack starting with Asterisk 16.

Change-Id: I8382d28d7d83ee0ce13334e51ae45dbc0bdaef48

2 years agoMerge "app_voicemail: Fix stack overrun in append_mailbox"
George Joseph [Mon, 24 Sep 2018 18:50:36 +0000 (13:50 -0500)]
Merge "app_voicemail:  Fix stack overrun in append_mailbox"

2 years agoMerge "app_voicemail: Cleanup mailbox topic and cache"
George Joseph [Mon, 24 Sep 2018 16:46:50 +0000 (11:46 -0500)]
Merge "app_voicemail:  Cleanup mailbox topic and cache"

2 years agoMerge "chan_sip.c: chan_sip unstable with TLS after asterisk start or reloads"
George Joseph [Mon, 24 Sep 2018 15:44:51 +0000 (10:44 -0500)]
Merge "chan_sip.c: chan_sip unstable with TLS after asterisk start or reloads"

2 years agoMerge "res_remb_modifier: Add module for controlling REMB from CLI."
George Joseph [Mon, 24 Sep 2018 15:11:54 +0000 (10:11 -0500)]
Merge "res_remb_modifier: Add module for controlling REMB from CLI."

2 years agoapp_voicemail: Cleanup mailbox topic and cache
George Joseph [Thu, 20 Sep 2018 15:15:48 +0000 (09:15 -0600)]
app_voicemail:  Cleanup mailbox topic and cache

app_voicemail wasn't properly cleaning up the stasis cache or the
mwi topic pool when the module was unloaded or when a user was
deleted as a result of a reload.  This resulted in leaks in both
areas.

* app_voicemail now calls ast_delete_mwi_state_full when it frees
  a user structure and ast_delete_mwi_state_full in turn now calls
  the new stasis_topic_pool_delete_topic function to clear the topic
  from the pool.

Change-Id: Ide23144a4a810e7e0faad5a8e988d15947965df8

2 years agoMerge "stasis: Add function to delete topic from pool"
George Joseph [Mon, 24 Sep 2018 14:29:14 +0000 (09:29 -0500)]
Merge "stasis:  Add function to delete topic from pool"

2 years agoMerge "res_rtp_asterisk: Fix crash on ast_rtp_new failure."
George Joseph [Mon, 24 Sep 2018 14:27:01 +0000 (09:27 -0500)]
Merge "res_rtp_asterisk: Fix crash on ast_rtp_new failure."

2 years agoMerge "channel.c: Address stack overflow in does_id_conflict()"
George Joseph [Mon, 24 Sep 2018 14:23:10 +0000 (09:23 -0500)]
Merge "channel.c:  Address stack overflow in does_id_conflict()"

2 years agortp_engine: rtcp_report_to_json can overflow the ssrc integer value
Kevin Harwell [Mon, 17 Sep 2018 20:35:05 +0000 (15:35 -0500)]
rtp_engine: rtcp_report_to_json can overflow the ssrc integer value

When writing an RTCP report to json the code attempts to pack the "ssrc" and
"source_ssrc" unsigned integer values as a signed int value type. This of course
means if the ssrc's unsigned value is greater than that which can fit into a
signed integer value it gets converted to a negative number. Subsequently, the
negative value goes out in the json report.

This patch now packs the value as a json_int_t, which is the widest integer type
available on a given system. This should make it so the value no longer
overflows.

Note, this was caught by two failing tests hep/rtcp-receiver/ and
hep/rtcp-sender.

Change-Id: I2af275286ee5e795b79f0c3d450d9e4b28e958b0

2 years agoapp_voicemail: Fix stack overrun in append_mailbox
George Joseph [Fri, 21 Sep 2018 19:32:52 +0000 (13:32 -0600)]
app_voicemail:  Fix stack overrun in append_mailbox

The append_mailbox function wasn't calculating the correct length
to pass to ast_alloca and it wasn't handling the case where context
might be empty.

Found by the Address Sanitizer.

Change-Id: I7eb51c7bd18a7a8dbdba261462a95cc69e84f161

2 years agochannel.c: Address stack overflow in does_id_conflict()
George Joseph [Fri, 21 Sep 2018 20:23:34 +0000 (14:23 -0600)]
channel.c:  Address stack overflow in does_id_conflict()

does_id_conflict() was passing a pointer to an int to a callback
that expected a pointer to a size_t.

Found by the Address Sanitizer.

Change-Id: I0ff542067eef63a14a60301654d65d34fe2ad503

2 years agores_rtp_asterisk: Fix crash on ast_rtp_new failure.
Corey Farrell [Fri, 21 Sep 2018 15:19:52 +0000 (11:19 -0400)]
res_rtp_asterisk: Fix crash on ast_rtp_new failure.

ast_rtp_new free'd rtp upon failure, but rtp_engine.c would also call
the destroy callback.  Remove call to ast_free from ast_rtp_new, leave
it to rtp_engine.c to initiate the full cleanup.  Add error detection
for the ssrc_mapping vector initialization.  In rtp_allocate_transport
set rtp->s = -1 in the failure path where we close that FD to ensure we
don't try closing it twice.

ASTERISK-27854 #close

Change-Id: Ie02aecbb46228ca804e24b19cec2bb6f7b94e451

2 years agores_rtp_asterisk: Reset all settings on module reload
Sean Bright [Thu, 20 Sep 2018 20:26:55 +0000 (16:26 -0400)]
res_rtp_asterisk: Reset all settings on module reload

'rtpchecksums' and 'rtcpinterval' are not being reset to their defaults
if they are not present in the updated configuration file.

Change-Id: I1162e40199314d46cf3225d5e1271c4c81176670

2 years agostasis: Add function to delete topic from pool
George Joseph [Thu, 20 Sep 2018 14:41:15 +0000 (08:41 -0600)]
stasis:  Add function to delete topic from pool

There's been a long standing leak when using topic pools.  The
topics in the pool get cleaned up when the last pool reference is
released but you can't remove a topic specifically.  If you reloaded
app_voicemail for instance, and mailboxes went away, their topics
were left in the pool.

* Added stasis_topic_pool_delete_topic() so modules can clean up
  topics from pools.
* Registered the topic pool containers so it can be examined from
  the CLI when AO2_DEBUG is enabled.  They'll be named
  "<topic_pool_name>-pool".

Change-Id: Ib7957951ee5c9b9b4482af7b9b4349112d62bc25

2 years agoMerge "stasis: No need to keep a stasis type ref in a stasis msg or cache object."
George Joseph [Thu, 20 Sep 2018 18:08:45 +0000 (13:08 -0500)]
Merge "stasis: No need to keep a stasis type ref in a stasis msg or cache object."

2 years agoAST-2018-009: Fix crash processing websocket HTTP Upgrade requests
Sean Bright [Thu, 16 Aug 2018 15:45:53 +0000 (11:45 -0400)]
AST-2018-009: Fix crash processing websocket HTTP Upgrade requests

The HTTP request processing in res_http_websocket allocates additional
space on the stack for various headers received during an Upgrade request.
An attacker could send a specially crafted request that causes this code
to overflow the stack, resulting in a crash.

* No longer allocate memory from the stack in a loop to parse the header
values.  NOTE: There is a slight API change when using the passed in
strings as is.  We now require the passed in strings to no longer have
leading or trailing whitespace.  This isn't a problem as the only callers
have already done this before passing the strings to the affected
function.

ASTERISK-28013 #close

Change-Id: Ia564825a8a95e085fd17e658cb777fe1afa8091a

2 years agoMerge "stasis_cache: Stop caching stasis subscription change messages"
Joshua Colp [Thu, 20 Sep 2018 14:43:26 +0000 (09:43 -0500)]
Merge "stasis_cache:  Stop caching stasis subscription change messages"

2 years agochan_sip.c: chan_sip unstable with TLS after asterisk start or reloads
hajekd [Mon, 3 Sep 2018 14:55:04 +0000 (16:55 +0200)]
chan_sip.c: chan_sip unstable with TLS after asterisk start or reloads

Fixes random asterisk crash on start or reload with TLS phones.

ASTERISK-28034 #close
Reported-by: David Hajek

Change-Id: I2a859f97dc80c348e2fa56e918214ee29521c4ac

2 years agoMerge "pjproject: Update initial 2.8 patches to apply cleanly."
Joshua Colp [Thu, 20 Sep 2018 11:49:52 +0000 (06:49 -0500)]
Merge "pjproject: Update initial 2.8 patches to apply cleanly."

2 years agores_remb_modifier: Add module for controlling REMB from CLI.
Joshua Colp [Thu, 20 Sep 2018 09:48:38 +0000 (09:48 +0000)]
res_remb_modifier: Add module for controlling REMB from CLI.

This adds a module which registers a CLI command that can set the
REMB bitrate value for REMB as it enters or exits Asterisk. This
allows you to ignore what Asterisk or a client produces and is
useful for demonstrations.

This does not generate REMB frames, however, but just modifies
them as they flow to or from a channel.

Change-Id: Ib089427c46a4a36d645cecfe02406adb38c17bec

2 years agoMerge "app_voicemail: Remove need to subscribe to stasis"
Joshua Colp [Thu, 20 Sep 2018 09:53:18 +0000 (04:53 -0500)]
Merge "app_voicemail: Remove need to subscribe to stasis"

2 years agoMerge "install_prereq: Remove unpackaged version of jansson."
Richard Mudgett [Wed, 19 Sep 2018 19:11:40 +0000 (14:11 -0500)]
Merge "install_prereq: Remove unpackaged version of jansson."

2 years agostasis: No need to keep a stasis type ref in a stasis msg or cache object.
Richard Mudgett [Fri, 14 Sep 2018 20:51:41 +0000 (15:51 -0500)]
stasis: No need to keep a stasis type ref in a stasis msg or cache object.

Stasis message types are global ao2 objects and we make stasis messages
and cache entries hold references to them.  Since there are currently
situations where cache objects are never deleted, the reference count on
the types can exceed 100000 and generate a FRACK assertion message.  The
stasis message cache could conceivably also have that many messages
legitimately on large systems.

The only down side to not holding the message type ref in the stasis
message is it only makes a crash either at shutdown or when manually
unloading a busy module slightly more likely.  However, this is more
exposing a pre-existing stasis shutdown ordering issue than a problem with
not holding a message type ref in stasis messages.

* Made stasis messages and cache entries no longer hold a ref to the
message type.

Change-Id: Ibaa28efa8d8ad3836f0c65957192424c7f561707

2 years agopjproject: Update initial 2.8 patches to apply cleanly.
Richard Mudgett [Tue, 18 Sep 2018 18:59:21 +0000 (13:59 -0500)]
pjproject: Update initial 2.8 patches to apply cleanly.

ASTERISK-28059

Change-Id: I027472f2753391646dde594a709a75f14422db93

2 years agoMerge "alembic: fix suppress_q850_reason_headers column name"
Joshua Colp [Wed, 19 Sep 2018 14:36:35 +0000 (09:36 -0500)]
Merge "alembic: fix suppress_q850_reason_headers column name"

2 years agoMerge "res_pjsip_session: Don't add declined stream if one does not exist."
Joshua Colp [Wed, 19 Sep 2018 13:42:37 +0000 (08:42 -0500)]
Merge "res_pjsip_session: Don't add declined stream if one does not exist."

2 years agoMerge "pjproject: Upgrade to 2.8."
George Joseph [Wed, 19 Sep 2018 13:06:03 +0000 (08:06 -0500)]
Merge "pjproject: Upgrade to 2.8."

2 years agostasis_message.c: Don't create immutable stasis objects with locks.
Richard Mudgett [Fri, 14 Sep 2018 20:48:24 +0000 (15:48 -0500)]
stasis_message.c: Don't create immutable stasis objects with locks.

* Create the stasis message object without a lock as it is immutable.
* Create the stasis message type object without a lock as it is immutable.
* Creating the stasis message type could crash if the passed in type name
is NULL and REF_DEBUG is enabled.  Added missing NULL check when passing
the ao2 object tag string.

Change-Id: I28763c58bb9f0b427c11971d0103bf94055e7b32

2 years agopjproject: Upgrade to 2.8.
Joshua Colp [Mon, 17 Sep 2018 16:38:19 +0000 (16:38 +0000)]
pjproject: Upgrade to 2.8.

This change brings in PJSIP 2.8, removes all the patches
that were merged upstream, and makes a minor change to
support a breaking change that was done.

ASTERISK-28059

Change-Id: I5097772b11b0f95c3c1f52df6400158666f0a189

2 years agoalembic: fix suppress_q850_reason_headers column name
Florian Floimair [Tue, 18 Sep 2018 14:39:05 +0000 (16:39 +0200)]
alembic: fix suppress_q850_reason_headers column name

In the original commit introducing the feature the column in the alembic
script was called 'suppress_q850_reason_header'.
In the code however the option is called 'suppress_q850_reason_headers'
(trailing 's'). This leads to errors when ARI push configuration is used.

Change-Id: Ie84808adbca6fcc9136556e4f5d741adbef5d14f

2 years agoapp_voicemail: Remove need to subscribe to stasis
George Joseph [Thu, 13 Sep 2018 12:55:20 +0000 (06:55 -0600)]
app_voicemail: Remove need to subscribe to stasis

app_voicemail was using the stasis cache to build and maintain a
list of mailboxes that had subscribers.  It then used this list
to determine if a mailbox should be polled for new messages if
polling was enabled.  For this to work, stasis had to cache every
subscription and unsubscription to the mailbox which caused a lot of
overhead, both cpu and memory related.

Since polling is only required when changes are being made to
mailboxes outside of app_voicemail and since the number of mailboxes
that don't have any subscribers is likely to be very low, all
mailboxes are now polled instead of just the ones with subscribers.

This paves the way for disabling the caching of stasis subscription
change messages.

Also fixed cleanup in some of the unit tests that not only left
test users in the users list but also caused segfaults if the tests
were run more than once.

ASTERISK-27121

Change-Id: I5cceb737246949f9782955c64425b8bd25a9e9ee

2 years agores_pjsip_session: Don't add declined stream if one does not exist.
Joshua Colp [Tue, 18 Sep 2018 11:08:24 +0000 (11:08 +0000)]
res_pjsip_session: Don't add declined stream if one does not exist.

Given a scenario where a session refresh was done with a removed
stream we would always add a removed stream to the outgoing SDP
even if one did not already exist.

This change makes it so that a removed stream is only placed into
the SDP if one already exists.

ASTERISK-28047

Change-Id: Ibb97d21cdeb87a8acae0c720861b0ff255708442

2 years agoinstall_prereq: Remove unpackaged version of jansson.
Corey Farrell [Mon, 10 Sep 2018 15:12:55 +0000 (11:12 -0400)]
install_prereq: Remove unpackaged version of jansson.

This is removed in favor of ./configure --with-jansson-bundled.  The
install-unpackaged command would only install jansson once, so once
installed it would never update, where the bundled copy will be kept up
to date.

Change-Id: Ideab1f65419608d3795aa608e9da873823cc42d3

2 years agoautoconf: Check for srtp_get_version_string() before using it
Sean Bright [Mon, 17 Sep 2018 15:38:28 +0000 (11:38 -0400)]
autoconf: Check for srtp_get_version_string() before using it

Change-Id: Id2a916ff9448706090e72ff2c7fb3f5ba24a05df

2 years agoMerge "res_srtp.c: Show linked version of libsrtp on module init"
George Joseph [Mon, 17 Sep 2018 14:23:52 +0000 (09:23 -0500)]
Merge "res_srtp.c: Show linked version of libsrtp on module init"

2 years agoMerge "res_pjsip: Log IPv6 addresses correctly"
George Joseph [Mon, 17 Sep 2018 13:34:02 +0000 (08:34 -0500)]
Merge "res_pjsip: Log IPv6 addresses correctly"

2 years agoCI: Fix typo in testsuite git checkout
George Joseph [Mon, 17 Sep 2018 12:10:18 +0000 (06:10 -0600)]
CI: Fix typo in testsuite git checkout

Change-Id: I30024515e5b00a5044fd39fbff27d818f016b719

2 years agores_srtp.c: Show linked version of libsrtp on module init
Sean Bright [Sun, 16 Sep 2018 11:08:29 +0000 (07:08 -0400)]
res_srtp.c: Show linked version of libsrtp on module init

Change-Id: Ib0a645d6985de5757cc4399ed2524b2d02c4f342

2 years agores_pjsip: Log IPv6 addresses correctly
Sean Bright [Fri, 7 Sep 2018 14:40:05 +0000 (10:40 -0400)]
res_pjsip: Log IPv6 addresses correctly

Both pjsip_tx_data.tp_info.dst_name and pjsip_rx_data.pkt_info.src_name
store IPv6 addresses without enclosing brackets. This causes some log
output to be confusing because it is difficult to separate the IPv6
address from a port specification.

* Use pj_sockaddr_print() along with pjsip_tx_data.tp_info.dst_addr and
  pjsip_rx_data.pkt_info.src_addr where possible for consistent IPv6
  output.

* When a pj_sockaddr is not available, explicitly wrap IPv6 addresses
  in brackets.

* When assigning pjsip_rx_data.pkt_info.src_name ourselves, make sure
  to also set pjsip_rx_data.pkt_info.src_addr.

Change-Id: I5cfe997ced7883862a12b9c7d8551d76ae02fcf8

2 years agoCI: Use proper credentials for Security testsuite checkout
George Joseph [Fri, 14 Sep 2018 17:31:28 +0000 (11:31 -0600)]
CI: Use proper credentials for Security testsuite checkout

Can't do anonymous http checkout from Security-testsuite.
Need to use same credentials as the gerrit review checkout.

Change-Id: I87af68c995cb8926f5e87f9af245600d76984f05

2 years agoMerge "res_musiconhold.c: Restart MOH if previous hold just reached end-of-file"
George Joseph [Fri, 14 Sep 2018 16:11:47 +0000 (11:11 -0500)]
Merge "res_musiconhold.c: Restart MOH if previous hold just reached end-of-file"

2 years agostasis_cache: Stop caching stasis subscription change messages
George Joseph [Thu, 13 Sep 2018 16:06:00 +0000 (10:06 -0600)]
stasis_cache:  Stop caching stasis subscription change messages

Since app_voicemail no longer uses the cache to maintain its state
there is no longer a need to cache these messages.

ASTERISK-27121

Change-Id: I321c708505f5ad8d00e1b0afc4c27dc2ac12ecb4

2 years agoMerge "optional_api: Remove unused nonoptreq fields"
Jenkins2 [Thu, 13 Sep 2018 18:08:10 +0000 (13:08 -0500)]
Merge "optional_api: Remove unused nonoptreq fields"

2 years agoMerge "CI: Use .gitreview to default BRANCH_NAME."
Joshua Colp [Thu, 13 Sep 2018 14:00:15 +0000 (09:00 -0500)]
Merge "CI: Use .gitreview to default BRANCH_NAME."

2 years agoMerge "res_pjproject: Fix sockaddr conversion routines for non-bundled PJSIP"
Joshua Colp [Thu, 13 Sep 2018 12:11:40 +0000 (07:11 -0500)]
Merge "res_pjproject: Fix sockaddr conversion routines for non-bundled PJSIP"

2 years agoMerge "Build System: Resolve conflict between DESTDIR and bundled jansson."
Joshua Colp [Wed, 12 Sep 2018 22:19:24 +0000 (17:19 -0500)]
Merge "Build System: Resolve conflict between DESTDIR and bundled jansson."

2 years agoCI: Use .gitreview to default BRANCH_NAME.
Corey Farrell [Wed, 12 Sep 2018 17:39:23 +0000 (13:39 -0400)]
CI: Use .gitreview to default BRANCH_NAME.

This ensures that binary modules are avoided in the master branch even
if BRANCH_NAME is not set.

Change-Id: I79162d2063f22fa9d6b31fde4827ace2dd5bf0da

2 years agooptional_api: Remove unused nonoptreq fields
Walter Doekes [Tue, 11 Sep 2018 12:22:18 +0000 (14:22 +0200)]
optional_api: Remove unused nonoptreq fields

As they're not actively used, they only grow stale. The moduleinfo field itself
is kept in Asterisk 13/15 for ABI compatibility.

ASTERISK-28046 #close

Change-Id: I8df66a7007f807840414bb348511a8c14c05a9fc

2 years agoMerge "manager: Set AMI event "Newexten" to the EVENT_FLAG_DIALPLAN class"
Joshua Colp [Wed, 12 Sep 2018 16:01:25 +0000 (11:01 -0500)]
Merge "manager: Set AMI event "Newexten" to the EVENT_FLAG_DIALPLAN class"

2 years agomanager: Set AMI event "Newexten" to the EVENT_FLAG_DIALPLAN class
lvl [Mon, 3 Sep 2018 11:50:07 +0000 (13:50 +0200)]
manager: Set AMI event "Newexten" to the EVENT_FLAG_DIALPLAN class

The documentation already specified EVENT_FLAG_DIALPLAN for this
event, but the implementation was using EVENT_FLAG_CALL.

Using EVENT_FLAG_DIALPLAN allows AMI clients to opt out of receiving
this highly verbose event.

ASTERISK-28033

Change-Id: I45b3119f30e4dbc17b49831f2b1a4f2c1beadafe

2 years agores_pjproject: Fix sockaddr conversion routines for non-bundled PJSIP
Sean Bright [Wed, 12 Sep 2018 12:18:07 +0000 (08:18 -0400)]
res_pjproject: Fix sockaddr conversion routines for non-bundled PJSIP

The bundled version of pjproject has a patch for Solaris compatability
that changes the definition of various socket structures which we need
to account for when compiling against a non-bundled version.

ASTERISK-28049 #close

Change-Id: Ia1ea47c433fc2d915115193ee889a752373925f0

2 years agoBuild System: Resolve conflict between DESTDIR and bundled jansson.
Corey Farrell [Tue, 11 Sep 2018 03:28:04 +0000 (23:28 -0400)]
Build System: Resolve conflict between DESTDIR and bundled jansson.

If Asterisk is built using a DESTDIR this will cause the bundled jansson
to be installed to an unexpected location and we will fail to find it.

Change-Id: Id033e2813261e0d45232383d44c6391122169548

2 years agores_musiconhold.c: Restart MOH if previous hold just reached end-of-file
Frederic LE FOLL [Thu, 30 Aug 2018 08:42:18 +0000 (10:42 +0200)]
res_musiconhold.c: Restart MOH if previous hold just reached end-of-file

On MOH activation, moh_files_readframe() is called while the current
stream attached to the channel is NULL and it calls ast_moh_files_next()
immediately.  However, it won't call ast_moh_files_next() again if sample
reading fails.  The failure may occur because res_musiconhold retains the
last sample reading position in the channel data and MOH during the
previous hold/retrieve just reached EOF.  Obviously, a bit of bad luck is
required here.

* Restructured moh_files_readframe() to try a second time to start MOH if
there was no stream setup and the saved position was at EOF.  Also added
comments describing what is going on for each step.

ASTERISK-28029

Change-Id: I1508cf2c094f8feca22d6f76deaa9fdfa9944860

2 years agoMerge "core: Don't stop generators when writing RTCP frames."
Jenkins2 [Fri, 7 Sep 2018 12:02:38 +0000 (07:02 -0500)]
Merge "core: Don't stop generators when writing RTCP frames."

2 years agoMerge "stasis_cache: Prune stasis_subscription_change messages"
Joshua Colp [Fri, 7 Sep 2018 10:40:36 +0000 (05:40 -0500)]
Merge "stasis_cache: Prune stasis_subscription_change messages"

2 years agoMerge "app_queue: Update realtime queuemembers after wait_a_bit(), not before"
Joshua Colp [Fri, 7 Sep 2018 09:48:30 +0000 (04:48 -0500)]
Merge "app_queue: Update realtime queuemembers after wait_a_bit(), not before"

2 years agocore: Don't stop generators when writing RTCP frames.
Joshua Colp [Wed, 5 Sep 2018 11:39:40 +0000 (11:39 +0000)]
core: Don't stop generators when writing RTCP frames.

Generators provide such functionality as tone generation or
silence generation. RTCP frames provide RTCP information and
should not stop generators from operating.

ASTERISK-28005

Change-Id: Ieadada07b068a7aa426e8763f1b73a18e1ac34a9

2 years agoapp_queue: Update realtime queuemembers after wait_a_bit(), not before
lvl [Mon, 3 Sep 2018 11:28:26 +0000 (13:28 +0200)]
app_queue: Update realtime queuemembers after wait_a_bit(), not before

This ensures the most up-to-date information is used for the next
call attempt.

ASTERISK-28032

Change-Id: I02fc17c6ffb50bb60ea97c2d2e6023e8061815ce

2 years agores_pjproject: Add utility functions to convert between socket structures
Sean Bright [Tue, 28 Aug 2018 13:42:13 +0000 (09:42 -0400)]
res_pjproject: Add utility functions to convert between socket structures

Currently, to convert from a pj_sockaddr to an ast_sockaddr, the address
needs to be rendered to a string and then parsed into the correct
structure. This also involves a call to getaddrinfo(3). The same is true
for the inverse operation.

Instead, because we know the internal structure of both ast_sockaddr and
pj_sockaddr, we can translate directly between the two without the
need for an intermediate string.

Change-Id: If0fc4bba9643f755604c6ffbb0d7cc46020bc761

2 years agoMerge "http.c: Give HTTP error response when received lines are too long."
George Joseph [Thu, 6 Sep 2018 16:49:25 +0000 (11:49 -0500)]
Merge "http.c: Give HTTP error response when received lines are too long."

2 years agoMerge "iostream.c: Fix ast_iostream_gets() needlessly returning failure."
Jenkins2 [Wed, 5 Sep 2018 19:29:13 +0000 (14:29 -0500)]
Merge "iostream.c: Fix ast_iostream_gets() needlessly returning failure."

2 years agostasis_cache: Prune stasis_subscription_change messages
George Joseph [Thu, 30 Aug 2018 18:08:05 +0000 (12:08 -0600)]
stasis_cache: Prune stasis_subscription_change messages

The stasis cache provides a way to reconstruct the current state
of topic subscribers.  Unfortunately, since every subscribe and
unsubscribe is cached, the cache continues to grow unabated while
asterisk is running.  This patch removes subscribe messages from
the cache when the corresponding unsubscribe is received.

This patch also registers the cache containers with ao2 so that if
AO2_DEBUG is turned on, you can list the container and get its
stats from the CLI.

ASTERISK-27121

Change-Id: I3d18905e477f3721815da91f30da8d3fbb2d4f56

2 years agoMerge "app_dial: set the comment for OPT_ARG_ANNOUNCE to really what is done"
George Joseph [Wed, 5 Sep 2018 16:00:11 +0000 (11:00 -0500)]
Merge "app_dial: set the comment for OPT_ARG_ANNOUNCE to really what is done"

2 years agoMerge "res_pjsip: Fix mwi_subscribe_replaces_unsolicited type mismatch"
George Joseph [Wed, 5 Sep 2018 14:56:21 +0000 (09:56 -0500)]
Merge "res_pjsip: Fix mwi_subscribe_replaces_unsolicited type mismatch"

2 years agoapp_dial: set the comment for OPT_ARG_ANNOUNCE to really what is done
Rodrigo Ramírez Norambuena [Mon, 3 Sep 2018 14:27:07 +0000 (11:27 -0300)]
app_dial: set the comment for OPT_ARG_ANNOUNCE to really what is done

Change-Id: I08f88adb09f7e5813f37e70fecd787468cdb32c8

2 years agopbx_config.c: Fix reloading module if initially declined to load
Chris-Savinovich [Wed, 15 Aug 2018 19:27:52 +0000 (15:27 -0400)]
pbx_config.c: Fix reloading module if initially declined to load

Added decline if extensions.conf file not available
when loading pbx_config, and also made sure everything
gets properly unregistered and/or destroyed on unload.

Change-Id: Ib00665106043b1be5148ffa7a477396038915854

2 years agoMerge "make config: os-release output error."
Joshua Colp [Fri, 31 Aug 2018 09:55:01 +0000 (04:55 -0500)]
Merge "make config: os-release output error."

2 years agohttp.c: Give HTTP error response when received lines are too long.
Richard Mudgett [Thu, 30 Aug 2018 19:42:06 +0000 (14:42 -0500)]
http.c: Give HTTP error response when received lines are too long.

Added a check when we receive a HTTP request line or header line that is
too long.  We now return an error response to the sender because we are
not able to process the request.

Change-Id: I6df2705435fd7dde4d5d3bdf7acec859cfb7c12d

2 years agoiostream.c: Fix ast_iostream_gets() needlessly returning failure.
Richard Mudgett [Wed, 29 Aug 2018 21:14:46 +0000 (16:14 -0500)]
iostream.c: Fix ast_iostream_gets() needlessly returning failure.

Providing a buffer larger than the internal buffer of ast_iostream_gets()
fails to get lines longer than the internal buffer.

* Made ast_iostream_gets() fill the supplied buffer with read data until
either a '\n' is found or the supplied buffer is filled just like fgets().

Change-Id: If18b3f6ee500e22f0633a68779ed09f7e0f305ed

2 years agoMerge "res_fax: Handle fax gateway being started more than once."
Joshua Colp [Thu, 30 Aug 2018 10:44:02 +0000 (05:44 -0500)]
Merge "res_fax: Handle fax gateway being started more than once."

2 years agoMerge "res_pjsip_transport_websocket: Properly set src_name for IPv6"
Joshua Colp [Thu, 30 Aug 2018 10:08:34 +0000 (05:08 -0500)]
Merge "res_pjsip_transport_websocket: Properly set src_name for IPv6"

2 years agores_pjsip: Fix mwi_subscribe_replaces_unsolicited type mismatch
Richard Mudgett [Mon, 6 Aug 2018 20:37:05 +0000 (15:37 -0500)]
res_pjsip: Fix mwi_subscribe_replaces_unsolicited type mismatch

ASTERISK-27988

Change-Id: Iccafdd0552ea8aaed647620fb14499f1bf341843

2 years agoMerge "Create --disable-binary-modules option."
George Joseph [Wed, 29 Aug 2018 11:31:54 +0000 (06:31 -0500)]
Merge "Create --disable-binary-modules option."

2 years agores_fax: Handle fax gateway being started more than once.
Joshua Colp [Wed, 29 Aug 2018 10:18:08 +0000 (07:18 -0300)]
res_fax: Handle fax gateway being started more than once.

The T.38 fax gateway state machine can cause the fax gateway
to be started more than once on a channel depending on the
responses of the remote endpoint. This would previously leak
the channel name, channel unique id, and underlying fax engine
state. This change instead makes it so that if the fax gateway
session is already present and not reserved the fax gateway
is not started again.

ASTERISK-27981

Change-Id: I552d95086860cb18f2522ee40ef47b13b6da2e0e

2 years agoMerge "alembic: increase uri column size"
Joshua Colp [Wed, 29 Aug 2018 10:20:01 +0000 (05:20 -0500)]
Merge "alembic: increase uri column size"

2 years agores_pjsip_transport_websocket: Properly set src_name for IPv6
Sean Bright [Tue, 28 Aug 2018 13:01:19 +0000 (09:01 -0400)]
res_pjsip_transport_websocket: Properly set src_name for IPv6

SIP responses over WebSockets when the client is using IPv6 have invalid
Via headers according to RFC 3261. The 'received' header parameter
should not be wrapped in brackets if it is an IPv6 address.

When src_name is populated by the built-in PJSIP transports, the code
uses pj_sockaddr_print() with 'flags' set to 0, meaning that the
brackets are not rendered around IPv6 addresses.

This may be related to ASTERISK~27101.

See also: https://github.com/onsip/SIP.js/pull/594

ASTERISK-28020 #close

Change-Id: I8ea9d289901b837512bee2ca2535e3dc14f04d77

2 years agoCreate --disable-binary-modules option.
Corey Farrell [Sun, 26 Aug 2018 18:18:42 +0000 (14:18 -0400)]
Create --disable-binary-modules option.

This new option can be passed for ./configure or
./tests/CI/buildAsterisk.sh to prevent download/install of binary
modules.

Normally enabling the categories MENUSELECT_CODECS or MENUSELECT_RES
will result in binary modules being enabled even if the build target is
incompatible with those modules.  This includes CI scripts which enable
categories before disabling specific modules.

If more binary modules are offered in the future this will help avoid
accidentally downloading them if unwanted or incompatible.  Adding a
binary module will only require creating a new menuselect entry similar
to the existing ones, it will not be necessary to modify the CI scripts.

Change-Id: I6b1bd1c75a2e48f05b8b8a45b7a7a2d00a079166

2 years agores/res_rtp_asterisk: remove debug traces generated by an empty frame
neutrino88 [Tue, 21 Aug 2018 12:59:08 +0000 (08:59 -0400)]
res/res_rtp_asterisk: remove debug traces generated by an empty frame

The realtime text timer pops regularly and sends text frames even if
the buffer is empty. This causes a lot of unecessary debug logging.

* Made red_write() test if we need to send a frame before calling
ast_rtp_write()

ASTERISK-28002
Reported by: Emmanuel BUU
Tested by: Emmanuel BUU

Change-Id: Icf81310c3b8080b615a42060afc02ab41f9523dd

2 years agoMerge "pbx_dundi: Added IPv6 support for dundi"
Jenkins2 [Mon, 27 Aug 2018 14:38:15 +0000 (09:38 -0500)]
Merge "pbx_dundi: Added IPv6 support for dundi"

2 years agoMerge "chan_sip: improved ip:port finding of peers for non-UDP transports."
George Joseph [Mon, 27 Aug 2018 12:17:39 +0000 (07:17 -0500)]
Merge "chan_sip: improved ip:port finding of peers for non-UDP transports."

2 years agochan_sip: improved ip:port finding of peers for non-UDP transports.
Jaco Kroon [Mon, 13 Aug 2018 13:12:21 +0000 (15:12 +0200)]
chan_sip: improved ip:port finding of peers for non-UDP transports.

Also remove function peer_ipcmp_cb since it's not used (according to
rmudgett).

Prior to b2c4e8660a9c89d07041271371151779b7ec75f6 (ASTERISK_27457)
insecure=port was the defacto standard.  That commit also prevented
insecure=port from being applied for sip/tcp or sip/tls.

Into consideration there are three sets of behaviour:

1.  "previous" - before the above commit.
2.  "current" - post above commit, pre this one.
3.  "new" - post this commit.

The problem that the above commit tried to address was guests over TCP.
It succeeded in doing that but broke transport!=udp with host!=dynamic.

This commit attempts to restore sane behaviour with respect to
transport!=udp for host!=dynamic whilst still retaining the guest users
over tcp.

It should be noted that when looking for a peer, two passes are made, the
first pass doesn't have SIP_INSECURE_PORT set for the searched-for peer,
thus looking for full matches (IP + Port), the second pass sets
SIP_INSECURE_PORT, thus expecting matches on IP only where the matched
peer allows for that (in the author's opinion:  UDP with insecure=port,
or any TCP based, non-dynamic host).

In previous behaviour there was special handling for transport=tcp|tls
whereby a peer would match during the first pass if the utilized
transport was TCP|TLS (and the peer allowed that specific transport).

This behaviour was wrong, or dubious at best.  Consider two dynamic tcp
peers, both registering from the same IP (NAT), in this case either peer
could match for connections from an IP.  It's also this behaviour that
prevented SIP guests over tcp.

The above referenced commit removed this behaviour, but kept applying
the SIP_INSECURE_PORT only to WS|WSS|UDP.  Since WS and WSS is also TCP
based, the logic here should fall into the TCP category.

This patch updates things such that the previously non-explicit (TCP
behaviour) transport test gets performed explicitly (ie, matched peer
must allow for the used transport), as well as the indeterministic
source-port nature of the TCP protocol is taken into account.  The new
match algorithm now looks like:

1.  As per previous behaviour, IP address is matched first.

2.  Explicit filter with respect to transport protocol, previous
    behaviour was semi-implied in the test for TCP pure IP match - this now
    made explicit.

3.  During first pass (without SIP_INSECURE_PORT), always match on port.

4.  If doing UDP, match if matched against peer also has
    SIP_INSECURE_PORT, else don't match.

5.  Match if not a dynamic host (for non-UDP protocols)

6.  Don't match if this is WS|WSS, or we can't trust the Contact address
    (presumably due to NAT)

7.  Match (we have a valid Contact thus if the IP matches we have no
    choice, this will likely only apply to non-NAT).

To logic-test this we need a few different scenarios.  Towards this end,
I work with a set number of peers defined in sip.conf:

[peer1]
host=1.1.1.1
transport=tcp

[peer2]
host=1.1.1.1
transport=udp

[peer3]
host=1.1.1.1
port=5061
insecure=port
transport=udp

[peer4]
host=1.1.1.2
transport=udp,tcp

[peer5]
host=dynamic
transport=udp,tcp

Test cases for UDP:

1 - incoming UDP request from 1.1.1.1:
  - previous:
    - pass 1:
      * peer1 or peer2 if from port 5060 (indeterminate, depends on peer
        ordering)
      * peer3 if from port 5061
      * peer5 if registered from 1.1.1.1 and source port matches
    - pass 2:
      * peer3
  - current: as per previous.
  - new:
    - pass 1:
      * peer2 if from port 5060
      * peer3 if from port 5061
      * peer5 if registered from 1.1.1.1 and source port matches
    - pass 2:
      * peer3

2 - incoming UDP request from 1.1.1.2:
  - previous:
    - pass 1:
      * peer5 if registered from 1.1.1.2 and port matches
      * peer4 if source port is 5060
    - pass 2:
      * no match (guest)
  - current: as previous.
  - new as previous (with the variation that if peer5 didn't have udp as
          allowed transport it would not match peer5 whereas previous
          and current code could).

3 - incoming UDP request from anywhere else:
  - previous:
    - pass 1:
      * peer5 if registered from that address and source port matches.
    - pass 2:
      * peer5 if insecure=port is additionally set.
      * no match (guest)
  - current - as per previous
  - new - as per previous

Test cases for TCP based transports:

4 - incoming TCP request from 1.1.1.1
  - previous:
    - pass 1 (indeterministic, depends on ordering of peers in memory):
      * peer1; or
      * peer5 if peer5 registered from 1.1.1.1 (irrespective of source port); or
      * peer2 if the source port happens to be 5060; or
      * peer3 if the source port happens to be 5061.
    - pass 2: cannot happen since pass 1 will always find a peer.
  - current:
    - pass 1:
      * peer1 or peer2 if from source port 5060
      * peer3 if from source port 5060
      * peer5 if registered as 1.1.1.1 and source port matches
    - pass 2:
      * no match (guest)
  - new:
    - pass 1:
      * peer 1 if from port 5060
      * peer 5 if registered and source port matches
    - pass 2:
      * peer 1

5 - incoming TCP request from 1.1.1.2
  - previous (indeterminate, depends on ordering):
    - pass 1:
      * peer4; or
      * peer5 if peer5 registered from 1.1.1.2
    - pass 2: cannot happen since pass 1 will always find a peer.
  - current:
    - pass 1:
      * peer4 if source port is 5060
      * peer5 if peer5 registered as 1.1.1.2 and source port matches
    - pass 2:
      * no match (guest).
  - new:
    - pass 1:
      * peer4 if source port is 5060
      * peer5 if peer5 registered as 1.1.1.2 and source port matches
    - pass 2:
      * peer4

6 - incoming TCP request from anywhere else:
  - previous:
    - pass 1:
      * peer5 if registered from that address
    - pass 2: cannot happen since pass 1 will always find a peer.
  - current:
    - pass 1:
      * peer5 if registered from that address and port matches.
    - pass 2:
      * no match (guest)
  - new: as per current.

It should be noted the test cases don't make explicit mention of TLS, WS
or WSS.  WS and WSS previously followed UDP semantics, they will now
enforce source port matching.  TLS follow TCP semantics.

The previous commit specifically tried to address test-case 6, but broke
test-cases 4 and 5 in the process.

ASTERISK-27881 #close

Change-Id: I61a9804e4feba9c7224c481f7a10bf7eb7c7f2a2

2 years agoAMI: be less verbose when adding HTTP headers to AMI/HTTP messages.
Jaco Kroon [Mon, 20 Aug 2018 12:23:38 +0000 (14:23 +0200)]
AMI: be less verbose when adding HTTP headers to AMI/HTTP messages.

All HTTP/AMI message headers are being sent to the verbose channel.
There are multiple places this is happening.  Consolidate the loop into
a function.  Drop the debug/verbose message.

Convert to using ast_asprintf to perform the length calculation, memory
allocation and snprintf all in one step.

Change-Id: Ic45e673fde05bd544be95ad5cdbc69518207c1a1

2 years agoMerge "sample_configs: noload res_hep.so by default"
Jenkins2 [Thu, 23 Aug 2018 13:55:44 +0000 (08:55 -0500)]
Merge "sample_configs: noload res_hep.so by default"

2 years agoalembic: increase uri column size
Florian Floimair [Thu, 23 Aug 2018 11:57:31 +0000 (13:57 +0200)]
alembic: increase uri column size

When mobile SIP clients register with Asterisk that use some sort of
push notifications, the URI can get quite lengthy due to the
additional push-service annotations (things like tokens, pn-type, etc.)
contained in it.

ASTERISK-28022 #close

Change-Id: I4c7ceadc3bb405f3daf722641c8cd5ca4188cc37

2 years agosample_configs: noload res_hep.so by default
Matthew Fredrickson [Wed, 22 Aug 2018 15:50:55 +0000 (10:50 -0500)]
sample_configs: noload res_hep.so by default

Change disables loading of res_hep.so in default installation.  Loading
res_hep has a performance impact whether it's used or not.  This disables
loading of it in sample config files.

Change-Id: I5ec150cf941634fabc72973e5bf1a965cb0ef9d0

2 years agoMerge "res_pjsip: Reduce processing when a Contact is updated."
Joshua Colp [Wed, 22 Aug 2018 17:42:46 +0000 (12:42 -0500)]
Merge "res_pjsip: Reduce processing when a Contact is updated."

2 years agoapp_queue: Silence GCC 8 compiler warning
Sean Bright [Tue, 21 Aug 2018 18:50:33 +0000 (14:50 -0400)]
app_queue: Silence GCC 8 compiler warning

I'm only seeing an error in 14+, so I assume it is due to different
compiler options:

app_queue.c: In function ‘handle_queue_add_member’:
app_queue.c:10234:19: error: ‘%d’ directive writing between 1 and 11
    bytes into a region of size 3 [-Werror=format-overflow=]
     sprintf(num, "%d", state);
                   ^~
app_queue.c:10234:18: note: directive argument in the range
    [-2147483648, 99]
     sprintf(num, "%d", state);
                  ^~~~

Compiler: gcc version 8.0.1 20180414 (experimental)
    [trunk revision 259383] (Ubuntu 8-20180414-1ubuntu2)

Change-Id: I18577590da46829c1ea7d8b82e41d69f105baa10

2 years agoMerge "AMI: Remove docs for nonexistent AMI ContactStatus event headers"
Joshua Colp [Tue, 21 Aug 2018 23:53:11 +0000 (18:53 -0500)]
Merge "AMI: Remove docs for nonexistent AMI ContactStatus event headers"

2 years agoMerge "pbx_dundi: Fix debug frame decode string."
Joshua Colp [Tue, 21 Aug 2018 23:52:46 +0000 (18:52 -0500)]
Merge "pbx_dundi: Fix debug frame decode string."

2 years agoMerge "pbx_dundi.c: Handle thread shutdown better."
George Joseph [Tue, 21 Aug 2018 12:26:01 +0000 (07:26 -0500)]
Merge "pbx_dundi.c: Handle thread shutdown better."

2 years agoMerge "pbx_dundi.c: Misc memory management fixes when destroying peers"
Joshua Colp [Tue, 21 Aug 2018 11:27:23 +0000 (06:27 -0500)]
Merge "pbx_dundi.c: Misc memory management fixes when destroying peers"

2 years agoAMI: Remove docs for nonexistent AMI ContactStatus event headers
Richard Mudgett [Mon, 20 Aug 2018 16:23:21 +0000 (11:23 -0500)]
AMI: Remove docs for nonexistent AMI ContactStatus event headers

Change-Id: I5736965c64c44338f7330e85a24bb46818607f19

2 years agoMerge "res_rtp_asterisk.c: Fix unused variable warnings"
George Joseph [Mon, 20 Aug 2018 16:31:20 +0000 (11:31 -0500)]
Merge "res_rtp_asterisk.c: Fix unused variable warnings"