Age | Commit message (Collapse) | Author |
|
Previously, if context A was including context B and context B was including
context A, i.e. if there was a circular dependency between contexts, then
calling manager_show_dialplan_helper could lead to an infinite recursion,
resulting in a crash.
This commit applies the same solution as the one implemented in the
show_dialplan_helper function. The manager_show_dialplan_helper and
show_dialplan_helper functions contain lots of code in common, but the former
was missing the "infinite recursion avoidance" code.
ASTERISK-26226 #close
Change-Id: I1aea85133c21787226f4f8442253a93000aa0897
|
|
|
|
|
|
In your Diaplan, if you specify
same => n,Set(CHANNEL(secure_bridge_media)=1)
same => n,Set(CHANNEL(secure_bridge_signaling)=1)
only the SIP channel driver chan_sip supports this. All other channels drivers
like res_pjsip fail. In case of failure, the original sRTP source code released
the whole channel, even if not hung-up, yet. This change does not release the
channel but instead hangs-up the channel.
ASTERISK-26306
Change-Id: I0489f0cb660fab6673b0db8af027d116e70a66db
|
|
We may check a global config option hundreds of times a second or more.
Asking sorcery for the global configuration from the config files backend
involves several allocations and container traversals. Using realtime
without a memory cache is a lot worse because you have to lookup in the
realtime database each time to reconstitute the sorcery object. With a
memory cache for realtime, there is about the same amount of overhead as
for config files. Either way, it is still fairly expensive to access the
sorcery object that much.
* Cache the global config options so we can access them faster. You must
now always perform a res_pjsip reload to change the global options.
Change-Id: Ice16c7a4cbca4614da344aaea21a072b86263ef7
|
|
ast_channel_get_t38_state() calls ast_channel_queryoption() with
AST_OPTION_T38_STATE. If the passed in channel is a local channel then a
deadlock can happen if a channel lock is held when called.
* Made ast_channel_get_t38_state() callers not hold a channel lock before
calling.
* Update ast_channel_get_t38_state() doxygen to note that no channel locks
can be held when calling the function.
ASTERISK-26203 #close
Reported by: Etienne Lessard
ASTERISK-24822 #close
Reported by: David Brillert
ASTERISK-22732 #close
Reported by: Richard Mudgett
Change-Id: I49fd76fa9af628b4198009b5c0b82c8b03681214
|
|
ASTERISK-25980 added the FAXMODE channel variable to res_fax.c.
Unfortunately, it also introduced a deadlock potential because
set_channel_variables() which sets FAXMODE can be called during a
masquerade. The ast_channel_get_t38_state() which gets the value used to
set FAXMODE cannot be called with the channel locked. As a result, local
channels can deadlock because of how they must acquire the locks necessary
to operate.
The intent of FAXMODE is for dialplan to know how a fax was transferred
after the fax completes. However, the previous patch sets FAXMODE to the
channel's current T.38 state AFTER the fax has completed and where T.38
may have already disconnected.
* Set FAXMODE based upon T.38 negotiations exchanged either with the fax
applications or the fax framehooks.
ASTERISK-26203
Reported by: Etienne Lessard
ASTERISK-24822
Reported by: David Brillert
ASTERISK-22732
Reported by: Richard Mudgett
Change-Id: Id525747254b64c1efe8b1b5973d52ff9719c2ae1
|
|
fax_gateway_indicate_t38() calls ast_indicate_data() which cannot be
called with any channel locks already held. A deadlock can happen if the
function is operating on a local channel.
* Made fax_gateway_indicate_t38() unlock the channel before calling
ast_indicate_data() since fax_gateway_indicate_t38() is always called with
the channel locked.
* Made fax_gateway_indicate_t38() return void since nothing cared about
its return value.
ASTERISK-26203
Reported by: Etienne Lessard
ASTERISK-24822
Reported by: David Brillert
ASTERISK-22732
Reported by: Richard Mudgett
Change-Id: I701ff2d26c5fc23e0d5a48a3fd98759a9fd09407
|
|
Change-Id: Ic10ae434536bbf7fb7055d6ab36cc50b8748a4e7
|
|
The framehook container could become corrupted if the channel lock is not
held before calling.
Change-Id: If0a1c7ba0484ed3a191106a7516526b905952584
|
|
The framehook container could become corrupted if the channel lock is not
held before calling.
Change-Id: I1a6b957a1f7b899eb29a186915f8cccab886a438
|
|
|
|
ast_multicast_rtp_create_options now checks for NULL or empty options
Change-Id: Ib845eae46a67a9787e89a87ebd1027344e5e0362
|
|
MALLOC_DEBUG should not be used to check if debugging is actually
enabled, __AST_DEBUG_MALLOC should be used instead. MALLOC_DEBUG only
indicates that debugging is requested, __AST_DEBUG_MALLOC indicates it
is active.
Change-Id: I3ce9cdb6ec91b74ee1302941328462231be1ea53
|
|
NOTE: This patch was submitted earlier and reverted because of a failing
test. The test has been patched so that it adjusts for the changes here,
so this is being resubmitted for review.
One feature that confbridge has is the ability to play sounds to all
participants in the conference. Prior to this commit, the algorithm for
this was as follows:
* Grab the playback lock
* Push the conference announcer channel into the bridge
* Play back the sound
* Pull the conference announcer channel from the bridge
* Release the playback lock
The issue here is that the act of adding the playback channel to the
bridge and removing it for each announcement is expensive. Amongst the
expenses:
* The announcer channel is imparted into the bridge, meaning a new
thread is spun up for each playback.
* When the announcer is added or removed from the bridge, it results
in the BRIDGEPEER channel variable being set on all channels in the
bridge. This requires keeping the bridge locked and locking each
individual channel in order to set it.
* There's also just the general overhead of adding the channel and
removing it from the bridge. The bridge potentially has to reconfigure
every single time
With this commit, the paradigm for playing back announcements has
shifted.
* The announcer channel is now added to the bridge when the conference
is allocated, and it is hung up when the conference is destroyed.
* A taskprocessor is used to queue playbacks onto the announcer channel.
This keeps the behavior from before where playbacks do not overlap.
* The announcer channel is no longer placed into the bridge as
departable. Since we are not constantly removing the channel from
the bridge, it is safe to add the channel using an independent thread
and simply hang the channel up when it is time for the conference to
be destroyed.
The use of the taskprocessor for playbacks opens up the interesting
possibility of having asynchronous announcements played. In this commit,
however, the behavior is still exactly the same as it previously was.
ASTERISK-26289
Reported by Mark Michelson
Change-Id: Ica9fa4907c2f3728cdd1cf0bc564ef4eb40754a0
|
|
|
|
This reverts commit 0cdeb2bfb0f4203384c08858951af3c77be8b9b3.
Change-Id: I18ba73b6d4dc0b994f4ffb01ae0b6cfad36ac636
|
|
|
|
|
|
|
|
When res_odbc_transaction depended on res_odbc, it got the generic_odbc
headers and libs implicitly. Now that it no longer depends on res_odbc,
its dependency on generic_odbc must be explicit.
Change-Id: I9db88f7af7388437f49903d3008ba8d4890d5911
|
|
PJProject supports a lot of platforms even Windows, some with different defaults
when it comes to IPv6. In many Linux platforms like Ubuntu 16.04 LTS,
"/proc/sys/net/ipv6/bindv6only" is set to 0 (false). Different than in Windows.
Because of this, if configured with just an IPv6 address/transport, PJProject
listens to both IPv4 and IPv6. However, this is not supported by the PJProject
team. As consequence, you end-up with IPv4-mapped IPv6 addresses in SDP,
incompatible with IPv4-only clients. Technically, you end-up with an IPv6-only
server which accepts incoming connections on IPv4.
If you try to configure two transports, one with IPv4 and one with IPv6 on the
same interface, as expected by the PJProject team, the IPv4 transport is not
able to bind because the IPv6 transport listens to both already.
One solution would be to change "/proc/sys/net/ipv6/bindv6only" system-wide.
Then, you are able to configure two transports, one for each IP version on the
same interface. That way, you get a server which works with IPv4 clients and
IPv6 clients at the same time over the same interface.
Here, this change sets this parameter directly within PJProject to match the
expectations of the PJProject team in any case. This allows IPv4/IPv6 Dual Stack
servers out of the box like in chan_sip. This change was accepted by the
PJProject team as <http://trac.pjsip.org/repos/changeset/5403> and is expected
to arrive in the next version, PJProject 2.6.0. Until then, this change is
incorporated in the bundled PJProject of Asterisk.
ASTERISK-26309
Change-Id: I3335d8718f79f4b2feae91b5b005a3ce684a63ae
|
|
|
|
|
|
res_odbc_transaction" into 13
|
|
|
|
One feature that confbridge has is the ability to play sounds to all
participants in the conference. Prior to this commit, the algorithm for
this was as follows:
* Grab the playback lock
* Push the conference announcer channel into the bridge
* Play back the sound
* Pull the conference announcer channel from the bridge
* Release the playback lock
The issue here is that the act of adding the playback channel to the
bridge and removing it for each announcement is expensive. Amongst the
expenses:
* The announcer channel is imparted into the bridge, meaning a new
thread is spun up for each playback.
* When the announcer is added or removed from the bridge, it results
in the BRIDGEPEER channel variable being set on all channels in the
bridge. This requires keeping the bridge locked and locking each
individual channel in order to set it.
* There's also just the general overhead of adding the channel and
removing it from the bridge. The bridge potentially has to reconfigure
every single time
With this commit, the paradigm for playing back announcements has
shifted.
* The announcer channel is now added to the bridge when the conference
is allocated, and it is hung up when the conference is destroyed.
* A taskprocessor is used to queue playbacks onto the announcer channel.
This keeps the behavior from before where playbacks do not overlap.
* The announcer channel is no longer placed into the bridge as
departable. Since we are not constantly removing the channel from
the bridge, it is safe to add the channel using an independent thread
and simply hang the channel up when it is time for the conference to
be destroyed.
The use of the taskprocessor for playbacks opens up the interesting
possibility of having asynchronous announcements played. In this commit,
however, the behavior is still exactly the same as it previously was.
ASTERISK-26289
Reported by Mark Michelson
Change-Id: Ic5cd2c4b98a1eaa1715eb7a5b35d62f1a76d78a5
|
|
brief" into 13
|
|
|
|
|
|
The compilation failed for devmode
--enable DONT_OPTIMIZE
--enable BETTER_BACKTRACES
--enable DO_CRASH
--enable TEST_FRAMEWORK
res_pjsip/pjsip_configuration.c: In function dtls_handler:
res_pjsip/pjsip_configuration.c:974:20: error:
back may be used uninitialized in this function [-Werror=maybe-uninitialized]
int size = strlen(front);
^
cc1: all warnings being treated as errors
Change-Id: I7f082ead0312792a577ec7c73015ba64dabca580
|
|
|
|
|
|
When using the migration script sip_to_pjsip.py, cert_file was not migrated to
pjsip.conf. A previous change regarding this contained a copy/paste error.
ASTERISK-22374
Change-Id: I0fa72e9412117d53b4284fc6b83fa5b2b95ba03b
|
|
When 'tlsclientmethod' is not specified in sip.conf, chan_sip uses the OpenSSL
SSLv23_method. This was documented incorrectly in the file sip.conf.sample.
SSLv23_method got its name in the 90s. Today, with OpenSSL 1.0.2, this method
enables (just) the secure TLSv1.0 and TLSv1.2. Or stated differently, that
function should have been called 'secure_method' or 'automatic_method' back in
the 90s.
Consequently please, specify 'tlsclientmethod=tlsv1' in your sip.conf only if
you face a server which has problems like not falling back to TLSv1.0
automatically.
ASTERISK-24425
Change-Id: I502ce6146b4504cadfd3973af8d6ec3994f54fa3
|
|
|
|
Given resource paths did not have 'json' substituted in for the '{format}'. For
some auto generated documentation/comment strings it resulted in something like
the following:
"... REST handler for /api-docs/sounds.{format}"
This patch makes sure the resource api's path is properly substituted.
ASTERISK-25472 #close
Change-Id: Ie3e950a35db4043e284019d6c9061f3b03922e23
|
|
Historically, Asterisk has always specified annexb=no for the g729 format.
However, when using res_pjsip no format attribute was specified. This patch
makes it so the SDP now contains a format attribute line with annexb=no.
Note, that this means only g729a is negotiated. Even for pass through support.
According to rfc7261 the type of annex used (a or b) is dependent upon the
answerer. However, Asterisk being a back to back user agent makes this tricky
to support at this time, thus we only allow annex 'a' for now.
ASTERISK-26228 #close
patches:
res_format_attr_g729.c submitted by Jason Parker (license 4993)
Change-Id: I76bc20cc0a01af01536e9915afef319c269c22d0
|
|
The MODULEINFO dependencies between these 2 modules was reversed.
res_odbc should depend on res_odbc_transaction, not the other way
around.
ASTERISK-25984 #close
Change-Id: Ifcfbb49c0b51cf6640a5446d47cd6c48caf1331f
|
|
A recent update had a copy/paste error where the unused variable 'val' was
being passed to the set_value function instead of the 'method' value itself.
This patch passes in the right variable.
ASTERISK-22374
Change-Id: I895b7b3779ce4442bc58b8ec40d59dd29bb43f06
|
|
|
|
contact_permit." into 13
|
|
|
|
|
|
|
|
|
|
into 13
|
|
|
|
|
|
When using the migration script sip_to_pjsip.py and tlsclientmethod is not set
in sip.conf, the default value of chan_sip (sslv23) is copied to pjsip.conf, to
overwrite the default of the PJProject (tlsv1). This makes sure, res_pjsip is
offering/using not just TLSv1.0 but TLSv1.2 as well.
ASTERISK-22374
Change-Id: Ie530a3dae9926ae14f3920a21be1e2edb15bda4f
|