Turn off the periodic sending of CRLNCRLN. Default is on (90 seconds),
which conflicts with the global section's keep_alive_interval option in
pjsip.conf.
patches:
pjsip_keep_not_alive.patch submitted by Alexander Traud (License 6520)
George Joseph [Wed, 11 Jul 2018 11:14:49 +0000 (05:14 -0600)]
CI: Initial commit for moving CI into source repo
Create tests/CI directory and add files used by Jenkins to
build and test Asterisk.
With this commit, Jenkins will run the Asterisk Unit Tests using
the Jenkinsfile at tests/CI/unittests.jenkinsfile. Bash scripts
to do the actual building and testing are also in the same directory.
Output is placed in tests/CI/output so that directory has been
added to .gitignore.
George Joseph [Tue, 10 Jul 2018 18:28:09 +0000 (12:28 -0600)]
app_confbridge: Use the SDP 'label' attribute to correlate users
Previously, the msid "label" attribute was used to correlate
participant info but because streams could be reused, the msid
wasn't being updated correctly when someone left the bridge and
another joined.
Now, instead of looking for the msid attribute on a channel's streams,
app_confbridge sets an "SDP:LABEL" attribute on the stream which
res_pjsip_sdp_rtp looks for. If it finds it, it adds a "label"
attribute to the current sdp.
Kevin Harwell [Fri, 6 Jul 2018 20:05:47 +0000 (15:05 -0500)]
res_pjsip_session: sdp group:BUNDLE attribute being truncated
When setting/appending the media id's to the bundle group attribute a '-1' was
being passed to the 'ast_str_set/append' function for the 'max_len' parameter.
This essentially capped the length of the string to what it was originally
allocated with. In this case 64 bytes.
This patch makes it so a '0' is passed as in for the 'max_len', which means
"no maximum length".
res_pjsip_pubsub: segfault in function publish_expire
The function pubsub_on_rx_publish_request incorrectly uses
of AST_SCHED_REPLACE_UNREF.
The AST_SCHED_REPLACE_UNREF should unref old '_data'.
Because of this, there may be a double unref
of variable 'publication' when ast_sched_del is unsuccessful
that leads to use after free of the 'publication' in publish_expire.
George Joseph [Fri, 6 Jul 2018 14:04:56 +0000 (08:04 -0600)]
test.c: Make output jUnit compatible
Separate "name" into "classname" and "name".
Use '.' for classname separator instead of '/'.
Prefix reserved words with '_'.
Wrap output with a top-level "testsuites" element.
George Joseph [Fri, 6 Jul 2018 12:57:37 +0000 (06:57 -0600)]
res_pjsip: Add 'suppress_q850_reason_headers' option to endpoint
A new option 'suppress_q850_reason_headers' has been added to the
endpoint object. Some devices can't accept multiple Reason headers and
get confused when both 'SIP' and 'Q.850' Reason headers are received.
This option allows the 'Q.850' Reason header to be suppressed.
The default value is 'no'.
ASTERISK-27949 Reported-by: Ross Beer
Change-Id: I54cf37a827d77de2079256bb3de7e90fa5e1deb1
res_pjsip_t38: Decline T.38 stream on failure case.
When negotiating an incoming T.38 stream the code incorrectly
returned failure instead of a decline for the stream when a
problem occurred or the configuration didn't allow it. This
resulted in SDP offers being rejected with a 488 response
in all cases, even when another valid stream was present.
This change makes it so the stream is now declined. If no
streams are accepted a 488 response is sent while if at least
one stream is accepted all the declined streams are, well,
declined.
Richard Mudgett [Mon, 2 Jul 2018 23:43:10 +0000 (18:43 -0500)]
res_pjsip_t38.c: Be smarter about how we respond when T.38 is disabled.
We were blindly responding with AST_T38_REFUSED when ANY T.38 control
frame came accross the bridge. This causes T.38 Gateway to get confused
and the T.38 session to get in a strange state.
* Made the T.38 framehook only respond to request frames and ignore
response frames.
Richard Mudgett [Tue, 3 Jul 2018 17:10:36 +0000 (12:10 -0500)]
res_pjsip/pjsip_transport_management.c: Fix deadlock with transport keep alive.
Using the keep_alive_interval option can result in a deadlock between the
pjproject transport manager group lock and the monitored transports ao2
container lock. The pjproject transport manager group lock has to be
superior in the locking order to the monitored transports ao2 container
lock because of pjproject callbacks called when already holding the group
lock. The lock inversion happens when Asterisk attempts to send a keep
alive packet over the reliable transports.
* Made keepalive_transport_thread() iterate over the monitored transports
container rather than use the ao2_callback() method. This avoids holding
the container lock when sending the keep alive packet.
Richard Mudgett [Fri, 29 Jun 2018 23:28:26 +0000 (18:28 -0500)]
res_pjsip_t38.c: Fix crash by ignoring 1xx messages.
If we initiated a T.38 reINVITE, we would crash if we received any other
1xx response message except 100 if it were followed by a 200 response.
* Made ignore any 1xx response so we do not close out the T.38 negotiation
too early. For good measure we'll now accept any 2xx response as
acceptance of the reINVITE T.38 offer.
res_pjsip_pubsub: Hold module reference for publications.
Incoming publications need to ensure that the module remains
loaded for the lifetime of them. This is now done by holding
a reference to the module while the publication exists. This
mirrors that of inbound subscriptions.
Robert Mordec [Mon, 21 May 2018 12:24:51 +0000 (14:24 +0200)]
app_confbridge: Bridge and announcers not removed if conference ends quickly
If a conference is ended very quickly after it was created (i.e., the
first user immediately hangs up) then the conference bridge and announcer
channels are not removed.
When a conference is created, the push_announcer() function is added to
the playback queue task processor and the conference object reference is
bumped. If a conference is ended while the push_announcer() function is
still going then the ao2_cleanup(conference) at the end of
push_announcer() will call the destructor function -
destroy_conference_bridge().
The destroy_conference_bridge() function will then add the
hangup_playback() task to the playback queue and will wait for it to end.
Since it is already a current task of the playback queue it will wait
forever.
This patch makes the conference thread call push_announcer() directly.
This way the conference object reference bump is not needed. Since the
playback queue task processor is only used by the conference thread
itself, there is no danger of trying to play announcements before the
announcer is pushed to the bridge.
Richard Mudgett [Thu, 28 Jun 2018 17:07:01 +0000 (12:07 -0500)]
AMI SendText action: Fix to use correct thread to send the text.
The AMI action was directly sending the text to the channel driver.
However, this makes two threads attempt to handle media and runs afowl of
CHECK_BLOCKING.
* Queue a read action to make the channel's media handling thread actually
send the text message. This changes the AMI actions success/fail response
to just mean the text was queued to be sent not that the text actually got
sent. The channel driver may not even support sending text messages.
George Joseph [Mon, 25 Jun 2018 20:42:14 +0000 (14:42 -0600)]
app_confbridge: Move participant info code to confbridge_manager.
With the participant info code in app_confbridge, we were still
in the process of adding the channel to the bridge when trying to send
an in-dialog MESSAGE. This caused 2 threads to grab the channel
blocking flag at the same time. To mitigate this, the participant
info code was moved to confbridge_manager so it runs after all
channel/bridge actions have finished.
There is a rare case (do to the infrequent timing involved) where
CDR submission threads in batch mode can deadlock with a currently
running CDR batch process. This patch should remove the need for
holding the lock in the scheduler and should clean a few code
paths up that inconsistently submitted new work to the CDR batch
processor.
George Joseph [Tue, 19 Jun 2018 02:22:17 +0000 (20:22 -0600)]
res_pjsip_session: Add ability to accept multiple sdp answers
pjproject by default currently will follow media forked during an INVITE
on outbound calls if the To tag is different on a subsequent response as
that on an earlier response. We handle this correctly. There have
been reported cases where the To tag is the same but we still need to
follow the media. The pjproject patch in this commit adds the
capability to sip_inv and also adds the capability to control it at
runtime. The original "different tag" behavior was always controllable
at runtime but we never did anything with it and left it to default to
TRUE.
So, along with the pjproject patch, this commit adds options to both the
system and endpoint objects to control the two behaviors, and a small
logic change to session_inv_on_media_update in res_pjsip_session to
control the behavior at the endpoint level.
The default behavior for "different tags" remains the same at TRUE and
the default for "same tag" is FALSE.
Change-Id: I64d071942b79adb2f0a4e13137389b19404fe3d6
ASTERISK-27936 Reported-by: Ross Beer
Richard Mudgett [Thu, 21 Jun 2018 21:39:45 +0000 (16:39 -0500)]
VECTOR: Passing parameters with side effects to macros is dangerous.
* Fix several instances where we were bumping a ref in the parameter and
then unrefing the object if it failed. The way the AST_VECTOR_APPEND()
and AST_VECTOR_REPLACE() macros are implemented means if it fails the new
value was never evaluated.
Alexander Traud [Thu, 21 Jun 2018 10:08:56 +0000 (12:08 +0200)]
codecs/ilbc: Compile in Solaris 11.
The symbol FS is the sampling frequency. That symbol is not used in Asterisk at
all and was a copy-and-paste of the iLBC reference code from the IETF RFC.
However, in Solaris, that symbol is defined by another header already. To
compile in Solaris, that symbol has to go.
Alexander Traud [Thu, 21 Jun 2018 10:01:53 +0000 (12:01 +0200)]
utils: Avoid an unused variable in Solaris 11.
With ./configure --enable-dev-mode[=noisy], the build fails because every
warning gets an error. Therefore, Asterisk has to be free of warnings and this
variable must go.
Alexander Traud [Wed, 20 Jun 2018 18:24:53 +0000 (20:24 +0200)]
BuildSystem: Enable autotools in Solaris 11.
Because this was the last operating system which required a special case, a
version appended to the autotools, the whole version stuff is removed by this
change. This simplifies the script ./bootstrap.sh. Hopefully, this gives even
broader platform compatibility.
Richard Mudgett [Tue, 12 Jun 2018 19:09:54 +0000 (14:09 -0500)]
channel.c: Fix usage of CHECK_BLOCKING()
The CHECK_BLOCKING() macro is used to indicate if a channel's handling
thread is about to do a blocking operation (poll, read, or write) of
media. A few operations such as ast_queue_frame(), soft hangup, and
masquerades use the indication to wake up the blocked thread to reevaluate
what is going on.
Richard Mudgett [Mon, 18 Jun 2018 23:04:54 +0000 (18:04 -0500)]
autoservice: Don't start channel autoservice if the thread is a user interface.
Executing dialplan functions from either AMI or ARI by getting a variable
could place the channel into autoservice. However, these user interface
threads do not handle the channel's media so we wind up with two threads
attempting to handle the media.
There can be one and only one thread handling a channel's media at a time.
Otherwise, we don't know which thread is going to handle the media frames.
Richard Mudgett [Mon, 18 Jun 2018 21:07:47 +0000 (16:07 -0500)]
Dialplan functions: Fix some channel autoservice misuse.
* Fix off nominal paths leaving the channel in autoservice.
* Remove unnecessary start/stop channel autoservice.
* Fix channel locking around a channel datastore search.
Acccording to the man page for sprintf, using the same buffer for
output as one used as an input yields undefined behavior.
This patch should work around this problem.
ASTERISK-27903 Reported-by: Alexander Traud
Change-Id: I2213dcb454aff26457e2e4cc9c6821276463ae3a
projects / thirdparty / asterisk.git / log
