Kinsey Moore [Thu, 16 Oct 2014 14:24:55 +0000 (14:24 +0000)]
PJSIP: Enforce module load dependencies
This enforces that res_pjsip, res_pjsip_session, and res_pjsip_pubsub
have loaded properly before attempting to load any modules that depend
on them since the module loader system is not currently capable of
resolving module dependencies on its own.
ASTERISK-24312 #close
Reported by: Dafi Ni
Review: https://reviewboard.asterisk.org/r/4062/
Fix loss of voice after second call drops (on a second line) in case using multiple lines on unistim phones. There is regression was introduced in r391379.
George Joseph [Tue, 14 Oct 2014 20:45:53 +0000 (20:45 +0000)]
config: Fix SEGV in unit test with MALLOC_DEBUG
With MALLOC_DEBUG the /main/config config_basic_ops test was causing a
SEGV while doing an ast_category_delete in an ast_category_browse loop.
Apparently this never worked but was also never tested. I removed the
test, added 2 notes to config.h indicating that it's not supported and
added a few lines of code to ast_category_delete to prevent the SEGV
should someone attempt it in the future.
Tested-by: George Joseph
Review: https://reviewboard.asterisk.org/r/4078/
Jonathan Rose [Tue, 14 Oct 2014 18:49:25 +0000 (18:49 +0000)]
Scheduler: Fix a nasty scheduler caching bug which makes new tasks not execute
Tasks that were marked for pending deletion in the scheduler would be moved to
the cache for later reuse, but after being recycled the deleted mark wouldn't
be removed resulting in fresh tasks being deleted without reason... and
immediately moved back into the cache where they could be reused again. This
could cause horrendous things to happen in just about anything that used a
scheduler.
ASTERISK-24321 #close
Reported by: Steve Pitts
Review: https://reviewboard.asterisk.org/r/4071/
Corey Farrell [Tue, 14 Oct 2014 16:45:14 +0000 (16:45 +0000)]
res_fax: Fix reference leak caused by gateway sessions
Fax gateway session objects can be re-used, causing the
same gateway session to be added to faxregistry.container
more than once. This change causes fax_session_new to
remove the reserved session from the container before
it's id is changed, ensuring it's possible for the
session to be freed.
Richard Mudgett [Tue, 14 Oct 2014 16:24:32 +0000 (16:24 +0000)]
stasis_channels.c: Resolve unfinished Dials when doing masquerades (Part 2)
Masquerades into and out of channels that are involved in a dial operation
don't create the expected dial end event. The missing dial end event goes
against the model for things like CDRs and generating Dial end manager
actions and such.
There are four cases:
1) A channel masquerades into the caller channel. The case happens when
performing a blonde transfer using the channel driver's protocol.
2) A channel masquerades into a callee channel. The case happens when
performing a directed call pickup.
3) The caller channel masquerades out of dial. The case happens when
using the Bridge application on the caller channel.
4) A callee channel masquerades out of dial. The case happens when using
the Bridge application on a peer channel.
As it turned out, all four cases need to be handled instead of just the
first one.
ASTERISK-24237
Reported by: Richard Mudgett
ASTERISK-24394 #close
Reported by: Richard Mudgett
Corey Farrell [Tue, 14 Oct 2014 16:18:46 +0000 (16:18 +0000)]
res_fax: Resolve module reference leak caused by reserved sessions
Remove reference to module providing reserved session after
adding a reference to the final module. This re-reference
is done to ensure that module references are correct even
if the final session selects a different module than the
reserved session.
George Joseph [Mon, 13 Oct 2014 16:07:56 +0000 (16:07 +0000)]
manager/config: Support templates and non-unique category names via AMI
This patch provides the capability to manipulate templates and categories
with non-unique names via AMI.
Summary of changes:
GetConfig and GetConfigJSON: Added "Filter" parameter: A comma separated list
of name_regex=value_regex expressions which will cause only categories whose
variables match all expressions to be considered. The special variable name
TEMPLATES can be used to control whether templates are included. Passing
'include' as the value will include templates along with normal categories.
Passing 'restrict' as the value will restrict the operation to ONLY templates.
Not specifying a TEMPLATES expression results in the current default behavior
which is to not include templates.
UpdateConfig: NewCat now includes options for allowing duplicate category
names, indicating if the category should be created as a template, and
specifying templates the category should inherit from. The rest of the
actions now accept a filter string as defined above. If there are non-unique
category names, you can now update specific ones based on variable values.
To facilitate the new capabilities in manager, corresponding changes had to be
made to config, most notably the addition of filter criteria to many of the
APIs. In some cases it was easy to change the references to use the new
prototype but others would have required touching too many files for this
patch so a wrapper with the original prototype was created. Macros couldn't
be used in this case because it would break binary compatibility with modules
such as res_digium_phone that are linked to real symbols.
Tested-by: George Joseph
Review: https://reviewboard.asterisk.org/r/4033/
Walter Doekes [Sun, 12 Oct 2014 08:14:04 +0000 (08:14 +0000)]
chan_sip: Fix so asterisk won't send reINVITE after a BYE.
After a reINVITE glare situation, Asterisk would re-send the reINVITE
even though the call had been hung up in the mean time. This patch
unschedules the reinvite when handling the BYE.
ASTERISK-22791 #close
Reported by: Paolo Compagnini
Tested by: Paolo Compagnini
Review: https://reviewboard.asterisk.org/r/4056/
(testcase is in review r4055)
........
Merged revisions 425296 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 425297 from http://svn.asterisk.org/svn/asterisk/branches/11
Walter Doekes [Sun, 12 Oct 2014 07:52:58 +0000 (07:52 +0000)]
build: Relax badshell tilde test to allow for ~ in middle of DESTDIR.
The main Makefile has a target test called 'badshell' that tests if
DESTDIR does not happen to have an an-expanded tilde (~). This might
be the case if you run: make install DESTDIR=~/somewhere/
That test also disallowed valid tildes in directory names. The test is
now changed to only trigger on a tilde at the start of the path.
George Joseph [Sat, 11 Oct 2014 21:07:36 +0000 (21:07 +0000)]
res_phoneprov: Cleanup module load error handling
Tested module load/reload interaction between res_phoneprov and
res_pjsip_phoneprov_provider in cases where res_phoneprov didn't
load correctly (usually misconfiguration or missing phoneprov.conf)
Tested-by: George Joseph
Review: https://reviewboard.asterisk.org/r/4069/
Joshua Colp [Fri, 10 Oct 2014 20:47:21 +0000 (20:47 +0000)]
bridge: During a smart bridge operation provide a more complete bridge to the old technology.
When a smart bridge operation occurs and a bridge transitions from one
technology to another the old technology is provided the channels formerly
in it and told that they are leaving. Unfortunately the bridge provided
along with them is incomplete. The bridge, despite there being channels in it,
contains none. This forces technology implementations to have additional
logic when channels are leaving or to store their own duplicated
state.
This change makes the bridge more complete so it contains the expected
channels. Now that the bridge is complete special logic within
bridge_native_rtp is no longer needed and has been removed.
Matthew Jordan [Fri, 10 Oct 2014 14:30:43 +0000 (14:30 +0000)]
res/res_phoneprov: Bail on registration if res_phoneprov didn't load
If res_phoneprov failed to fully load (due to not being configured), the
providers container will be NULL. If a module attempts to register a phone
provisioning provider, it should check for the presence of the container.
If there is no providers container, it should return an error.
This patch makes the ast_phoneprov_provider_register function do that...
otherwise this would be a silly commit message.
Kinsey Moore [Fri, 10 Oct 2014 12:58:52 +0000 (12:58 +0000)]
CallerID: Fix parsing regression
This fixes a regression in callerid parsing introduced when another bug
was fixed. This bug occurred when the name was composed entirely of
DTMF keys and quoted without a number section (<>).
ASTERISK-24406 #close
Reported by: Etienne Lessard
Tested by: Etienne Lessard
Patches:
callerid_fix.diff uploaded by Kinsey Moore
Review: https://reviewboard.asterisk.org/r/4067/
........
Merged revisions 425152 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 425153 from http://svn.asterisk.org/svn/asterisk/branches/11
Joshua Colp [Fri, 10 Oct 2014 12:09:39 +0000 (12:09 +0000)]
res_pjsip_nat: Place source port into rport of responses if 'force_rport' is on.
When the 'force_rport' option is enabled the behavior should be the same
as if the remote side placed rport into the message themselves. Therefore
any responses we send should include the source port of the request in the
rport of the Via header.
Walter Doekes [Fri, 10 Oct 2014 07:27:46 +0000 (07:27 +0000)]
chan_sip: Fix dialog leak resulting from missing ACK to re-INVITE.
If a device re-INVITEs at the same time as the dialog is hung up, and
if then the ACK to the re-INVITE never reaches Asterisk, chan_sip would
fail to destroy the dialog after a while. This resulted in (most
prominently) file handle leaks.
Kevin Harwell [Thu, 9 Oct 2014 21:28:15 +0000 (21:28 +0000)]
res_rtp_asterisk: Crash if no candidates received for component
When starting ice if there is not at least one remote ice candidate with an RTP
component asterisk will crash. This is due to an assertion in pjnath as it
expects at least one candidate with an RTP component. Added a check to make
sure at least one candidate contains an RTP component and at least one candidate
has an RTCP component.
George Joseph [Thu, 9 Oct 2014 17:41:12 +0000 (17:41 +0000)]
res_phoneprov: Refactor phoneprov to allow pluggable config providers
This patch makes res_phoneprov more modular so other modules (like pjsip)
can provide configuration information instead of res_phoneprov relying solely
on users.conf and sip.conf. To accomplish this a new ast_phoneprov public API
is now exposed which allows config providers to register themselves, set
defaults (server profile, etc) and add user extensions.
* ast_phoneprov_provider_register registers the provider and provides callbacks
for loading default settings and loading users.
* ast_phoneprov_provider_unregister clears the defaults and users.
* ast_phoneprov_add_extension should be called once for each user/extension
by the provider's load_users callback to add them.
* ast_phoneprov_delete_extension deletes one extension.
* ast_phoneprov_delete_extensions deletes all extensions for the provider.
Tested-by: George Joseph
Review: https://reviewboard.asterisk.org/r/3970/
Walter Doekes [Thu, 9 Oct 2014 08:07:27 +0000 (08:07 +0000)]
safe_asterisk: Don't automatically exceed MAXFILES value of 2^20.
On systems with lots of RAM (e.g. 24GB) /proc/sys/fs/file-max divided
by two can exceed the per-process file limit of 2^20. This patch
ensures the value is capped.
(Patch cleaned up by me.)
ASTERISK-24011 #close
Reported by: Michael Myles
Patches:
safe_asterisk-ulimit.diff uploaded by Michael Myles (License #6626)
........
Merged revisions 424875 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 424878 from http://svn.asterisk.org/svn/asterisk/branches/11
Joshua Colp [Wed, 8 Oct 2014 18:45:19 +0000 (18:45 +0000)]
res_rtp_asterisk: Allow only UDP ICE candidates.
The underlying library, pjnath, that res_rtp_asterisk uses for ICE
support does not have support for ICE-TCP. As candidates are
passed through directly to it this can cause error messages to occur
when it receives something unexpected (such as a TCP candidate).
This change merely ignores all non-UDP candidates so they never
reach pjnath.
When ao2_callback is run with OBJ_MULTIPLE and not OBJ_NODATA
it allocates a temporary container in a way that does not
record REF_DEBUG log entries. This changes that container
to correctly record unref's when the container is freed.
Prior to this patch, the auth_reject_permanent parameter was not initialized on
the registration client state, leading to the parameter being disabled
regardless of the value specified in pjsip.conf.
This patch initialized the setting on the registration client state to the
provided configuration value.
Matthew Jordan [Mon, 6 Oct 2014 18:37:13 +0000 (18:37 +0000)]
message: Don't close an AMI connection on SendMessage action error
If SendMessage encounters an error (such as incorrect input provided to the
action), it will currently return -1. Actions should only return -1 if the
connection to the AMI client should be closed. In this case, SendMessage
causing the client to disconnect is inappropriate.
This patch causes the action to return 0, which simply causes the action to
fail.
Review: https://reviewboard.asterisk.org/r/4024
ASTERISK-24354 #close
Reported by: Peter Katzmann
patches:
sendMessage.patch uploaded by Peter Katzmann (License 5968)
........
Merged revisions 424690 from http://svn.asterisk.org/svn/asterisk/branches/11
Richard Mudgett [Mon, 6 Oct 2014 15:36:32 +0000 (15:36 +0000)]
features.c: Fix lingering channel ref while Bridge() application is active.
Using the Bridge application to bridge a channel that is executing an
applicaiton such as Wait results in a lingering Surrogate channel in the
CLI "core show channels" output even though it has already hungup.
* Fix bridge_exec() to not hold onto the current_dest_chan ref once it has
been put into the bridge.
Matthew Jordan [Mon, 6 Oct 2014 00:59:12 +0000 (00:59 +0000)]
res_pjsip/pjsip_options: Do not 404 an OPTIONS request not sent to an endpoint
An OPTIONS request that is sent to Asterisk but not to a specific endpoint is
currently sent a 404 in response. This is because, not surprisingly, an empty
extension is never going to be found in the dialplan.
This patch makes it so that we only attempt to look up the endpoint in the
dialplan if it is specified in the OPTIONS request URI.
Matthew Jordan [Mon, 6 Oct 2014 00:51:43 +0000 (00:51 +0000)]
pjsip/dialplan_functions: Handle PJSIP_MEDIA_OFFER called on non-PJSIP channels
Calling PJSIP_MEDIA_OFFER on a non-PJSIP channel is hazardous to your health.
It will treat the channels as a PJSIP channel, eventually hitting an ao2 error,
FRACKing on assertion error, and quite likely crashing.
This patch adds checks to the read/write callbacks that ensure that the channel
technology is of type 'PJSIP' before attempting to operate on the channel.
Matthew Jordan [Mon, 6 Oct 2014 00:30:34 +0000 (00:30 +0000)]
res_pjsip: Prevent crashes when PJPROJECT presents an rdata with no message
When a message that exceeds the PJ_MAX_PKT_SIZE is sent over a reliable
transport, it is possible (although it shouldn't occur) for pjproject to pass
up an rdata object with a NULL msg in the msg_info. Needless to say, things
that attempt to dereference this are in for a rough ride.
In particular, this caused crashes in three different locations, all of which
are 'low level' enough to intercept an rdata object early in processing:
Richard Mudgett [Fri, 3 Oct 2014 19:36:09 +0000 (19:36 +0000)]
audiohooks: Reevaluate the bridge technology when an audiohook is added or removed.
Adding a mixmonitor to a channel causes the bridge to change technologies
from native to simple_bridge so the call can be recorded. However, when
the mixmonitor is stopped the bridge does not switch back to the native
technology.
* Added unbridge requests to reevaluate the bridge when a channel
audiohook is removed.
* Moved the unbridge request into ast_audiohook_attach() ensure that the
bridge reevaluates whenever an audiohook is attached. This simplified the
mixmonitor and chan_spy start code as well.
* Added defensive code to stop_mixmonitor_full() in case additional
arguments are ever added to the StopMixMonitor application.
* Made ast_framehook_detach() not do an unbridge request if the framehook
does not exist.
* Made ast_framehook_list_fixup() do an unbridge request if there are any
framehooks. Also simplified the loop.
Richard Mudgett [Fri, 3 Oct 2014 17:30:07 +0000 (17:30 +0000)]
chan_pjsip: Fix deadlock when masquerading PJSIP channels.
Performing a directed call pickup resulted in a deadlock when PJSIP
channels were involved.
A masquerade needs to hold onto the channel locks while it swaps channel
information between the two channels involved in the masquerade. With
PJSIP channels, the fixup routine needed to push a fixup task onto the
PJSIP channel's serializer. Unfortunately, if the serializer was also
processing a task that needed to lock the channel, you get deadlock.
* Added a new control frame that is used to notify the channels that a
masquerade is about to start and when it has completed.
* Added the ability to query taskprocessors if the current thread is the
taskprocessor thread.
* Added the ability to suspend/unsuspend the PJSIP serializer thread so a
masquerade could fixup the PJSIP channel without using the serializer.
George Joseph [Fri, 3 Oct 2014 15:53:59 +0000 (15:53 +0000)]
sorcery: Prevent SEGV in sorcery_wizard_create when there's no create function
When you call ast_sorcery_create() you don't necessarily know which wizard is
going to be invoked. If it happens to be a wizard like 'config' that doesn't
have a 'create' virtual function you get a segfault in the
sorcery_wizard_create callback. This patch catches the null function pointer,
does an ast_assert, and logs an error.
Kinsey Moore [Fri, 3 Oct 2014 13:57:24 +0000 (13:57 +0000)]
PJSIP: Restore functional default for callerid_privacy
The pjsip config option default fixups from r424263 altered the
functional default from "allowed_not_screened" to "allowed". This
change restores the functional default value when none is provided.
Kinsey Moore [Fri, 3 Oct 2014 13:24:21 +0000 (13:24 +0000)]
Manager: Add missing fields and documentation for CoreShowChannels
This corrects some issues introduced in the responses to the
CoreShowChannels AMI command as well as adding documentation for the
responses. The command in Asterisk 12 was missing the following fields:
Duration, Application, ApplicationData, and BridgedChannel and
BridgedUniqueID (replaced with BridgeId).
Joshua Colp [Fri, 3 Oct 2014 07:52:54 +0000 (07:52 +0000)]
res_pjsip_session: Reduce SDP size by removing duplicate connection lines.
Due to the architecture of how media streams are handled each individual
handler adds connection details (IP address) for it. The first media stream
is then used as the top level SDP connection line. In practice each
line ends up being the same so to reduce the SDP size stream-level connection
information is also added to the SDP if it differs from the top level SDP
connection line.
Richard Mudgett [Thu, 2 Oct 2014 21:45:58 +0000 (21:45 +0000)]
res_pjsip: Make transport cipher option accept a comma separated list of cipher names.
Improvements to the res_pjsip transport cipher option.
* Made the cipher option accept a comma separated list of OpenSSL cipher
names. Users of realtime will be glad if they have more than one name to
list.
* Added the CLI command 'pjsip list ciphers' so a user can know what
OpenSSL names are available for the cipher option.
* Updated the cipher option online XML documentation to specify what is
expected for the value.
* Updated pjsip.conf.sample to not indicate that ALL is acceptable since
ALL does not imply a preference order for the ciphers and PJSIP does not
simply pass the string to OpenSSL for interpretation.
Jonathan Rose [Thu, 2 Oct 2014 19:58:08 +0000 (19:58 +0000)]
Alembic: Add enumerator value to sippeers -> directmedia - 'outgoing'
The 'outgoing' value was left off of the enumerator when first creating the
column. This patch adds it, and should gracefully upgrade keeping the existing
data in tact.
ASTERISK-23781 #close
Reported by: Stephen More
Review: https://reviewboard.asterisk.org/r/4013/
res_pjsip: document use of rewrite_contact in sample conf
Without setting rewrite_contact, an invite to an endpoint
behind NAT will not reach it - unless the endpoint itself
uses STUN or TURN to discover it's public URI. Thus, the
use of this should be in the sample documentation.
Joshua Colp [Wed, 1 Oct 2014 16:35:02 +0000 (16:35 +0000)]
res_pjsip: Add 'dtls_fingerprint' option to configure DTLS fingerprint hash.
During the latest update to DTLS-SRTP support the ability to configure
the hash used for fingerprints was added. This gave us two supported ones:
SHA-1 and SHA-256. The default was accordingly updated to SHA-256.
Unfortunately this configuration ability was not exposed within res_pjsip.
This change adds a dtls_fingerprint option that controls it.
Kinsey Moore [Wed, 1 Oct 2014 12:24:12 +0000 (12:24 +0000)]
PJSIP: Handle defaults properly
This updates the code behind PJSIP configuration options with custom
handlers to deal with the assigned default values properly where it
makes sense and adjusting the default value where it doesn't. Before
applying this patch, there were several cases where the default value
for an option would prevent that config section from loading properly.
Reported by: Thomas Thompson
Review: https://reviewboard.asterisk.org/r/4019/
Kinsey Moore [Wed, 1 Oct 2014 12:13:20 +0000 (12:13 +0000)]
PJSIP: Force transport on contact rewrite
If contact rewriting is enabled but the contact differs in transport
from what is actually being used, messages after the initial INVITE
transaction can be sent to an incorrect transport/port combination. In
the case where this bug occurred the remote party never received a BYE
since it was sent to the remote party's TCP port over UDP.
res_rtp_asterisk: Ensure that the base and mapped address for candidates is present in SDP.
This change fixes an issue where ICE candidates put into the SDP did not contain
the 'raddr' and 'rport' information for server reflexive and relay candidates.
#SIPit31
........
Merged revisions 424151 from http://svn.asterisk.org/svn/asterisk/branches/11
res_pjsip_session: Add additional checks for delaying session refreshes.
There are certain situations which no checks existed for which need to prevent
session refreshes. This includes sending a session refresh with SDP before SDP
negotiation has completed and sending a session refresh before the dialog itself
has been established. Checks for these have been added.
Additionally COLP related UPDATEs were including SDP when it is not needed.
Richard Mudgett [Fri, 26 Sep 2014 15:19:39 +0000 (15:19 +0000)]
res_fax: Fix out of bounds error in update_modem_bits().
ASTERISK-24357 #close
Reported by: Jeremy Laine
Patches:
res_fax_bounds.patch (license #6561) patch uploaded by Jeremy Laine
Modified patch to not use magic numbers.
........
Merged revisions 423979 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 423983 from http://svn.asterisk.org/svn/asterisk/branches/11
Richard Mudgett [Wed, 24 Sep 2014 18:28:18 +0000 (18:28 +0000)]
pjsip_options.c: Fix race condition stopping periodic out of dialog OPTIONS request.
The crash on the issues is a result of an invalid transport configuration
change when asterisk is restarted. The attempt to send the qualify
request fails and we cleaned up. However, the callback is also called
which results in a double unref of the objects involved.
* Put a wrapper around pjsip_endpt_send_request() to detect when the
passed in callback is called because of an error so callers can know to
not cleanup.
* Made send_request_cb() able to handle repeated challenges (Up to 10).
* Fix periodic endpoint qualify OPTIONS sched deletion race by avoiding
it. The sched entry will no longer self stop and must be externally
stopped.
* Added REF_DEBUG description tags to struct sched_data in
pjsip_options.c.
* Fix some off-nominal ref leaks in schedule_qualify(),
qualify_and_schedule().
* Reordered pjsip_options.c module start/stop code to cleanup better on
error.
Prior to the release of Swagger 1.2, the attribute 'extends' was being
promoted as a possible way to show that a particular object extends an existing
object. Instead, the Swagger specification went with the 'subTypes' attribute
in the base object. This patch removes the unsupported attribute; the object
that the offending objects proposed to extend already lists them in its
'subTypes' attribute.
Matthew Jordan [Sat, 20 Sep 2014 23:40:31 +0000 (23:40 +0000)]
rest-api/api-docs: Correct basePath in resources to match top resources file
The resources.json file that defines the resource JSON files used with ARI
references a basePath of 'http://localhost:8088/ari'. This does not match what
is defined in the resource files themselves, 'http://localhost:8088/stasis'.
The correct base path is the one that includes 'ari' in the URL; this patch
updates the various resource JSON files to have the correct basePath.
res_pjsip_notify: Fix crash on unload/load and don't say the module doesn't exist on reload.
When unloading the module did not unregister the CLI commands causing a crash upon
load when they were registered again.
When reloading the module the return value from the config options framework was not
checked to determine if an error occurred or not. This caused a message to be output
saying the module did not exist when reloading if no changes were present.
Jonathan Rose [Fri, 19 Sep 2014 15:10:50 +0000 (15:10 +0000)]
Stasis_channels: Resolve unfinished Dials when doing masquerades
Masquerades into channels that are in the dialing state don't end their dial
and this goes against the model for things like CDRs and generating Dial end
manager actions and such.
ASTERISK-24237 #close
Reported by: Richard Mudgett
Review: https://reviewboard.asterisk.org/r/3990/
Kinsey Moore [Fri, 19 Sep 2014 12:30:39 +0000 (12:30 +0000)]
PJSIP: Prevent T38 framehook being put on wrong channel
This change gives framehooks a reverse-direction masquerade callback in
addition to chan_fixup_cb similar to the callback added to datastores
to handle the same situation. The new callback provides the same
parameters as the fixup callback, but is called on the new channel's
framehooks before moving framehooks from the old channel to the new
channel. This gives the framehooks an oppurtunity to decide whether
they should remain on the new channel or be removed.
This new callback is used to prevent the PJSIP T.38 framehook from
remaining on a masqueraded channel if the new channel is not also a
PJSIP channel. This was causing a crash when a local channel was
masqueraded into a PJSIP channel and the framehook was executed on the
local channel since the channel's tech private data was not structured
as expected.
George Joseph [Thu, 18 Sep 2014 19:21:56 +0000 (19:21 +0000)]
utils: Create ast_strsep function that ignores separators inside quotes
This function acts like strsep with three exceptions...
* The separator is a single character instead of a string.
* Separators inside quotes are treated literally instead of like separators.
* You can elect to have leading and trailing whitespace and quotes
stripped from the result and have '\' sequences unescaped.
Like strsep, ast_strsep maintains no internal state and you can call it
recursively using different separators on the same storage.
Also like strsep, for consistent results, consecutive separators are not
collapsed so you may get an empty string as a valid result.
Tested by: George Joseph
Review: https://reviewboard.asterisk.org/r/3989/
Richard Mudgett [Thu, 18 Sep 2014 16:39:06 +0000 (16:39 +0000)]
astobj2.c/refcounter.py: Fix to deal with invalid object refs.
* Make astob2 REF_DEBUG output an invalid object line when an invalid ao2
object ref/unref is attempted. This is similar to the
constructor/destructor lines.
* Fixed refcounter.py to handle skewed objects that have
constructor/destructor states.
* Made refcounter.py highlight the invalid ao2 object refs by putting them
in their own section of the processed output file.
* Made refcounter.py highlight unreffing an object by more than one that
results in a negative ref count and the object being destroyed. The
abnormally destroyed object is reported in the invalid and finalized
object sections of the output.
Mark Michelson [Thu, 18 Sep 2014 15:40:47 +0000 (15:40 +0000)]
res_pjsip_pubsub: Add some type safety when generating NOTIFY bodies.
res_pjsip_pubsub has two separate checks that it makes when a SUBSCRIBE
arrives.
* It checks that there is a subscription handler for the Event
* It checks that there are body generators for the types in the Accept header
The problem is, there's nothing that ensures that these two things will
actually mesh with each other. For instance, Asterisk will accept a subscription
to MWI that accepts pidf+xml bodies. That doesn't make sense.
With this commit, we add some type information to the mix. Subscription
handlers state they generate data of type X, and body generators state
that they consume data of type X. This way, Asterisk doesn't end up in
some hilariously mismatched situation like the one in the previous paragraph.
George Joseph [Thu, 18 Sep 2014 15:01:11 +0000 (15:01 +0000)]
res_pjsip: ami: Fix error in AMI output when an endpoint has no transport
When no transport is associated to an endpoint, the AMI output for
PJSIPShowEndpoint indicates an error instead of silently ignoring the
missing transport.
This patch causes the error to appear only if a transport was specified
on the endpoint and the transport doesn't exist. It also fixes an issue
with counting the objects that were actually found.
ASTERISK-24161 #close
ASTERISK-24331 #close
Tested by: George Joseph
Review: https://reviewboard.asterisk.org/r/3998/
George Joseph [Thu, 18 Sep 2014 14:43:41 +0000 (14:43 +0000)]
config: bug: Fix SEGV in ast_category_insert when matching category isn't found
If you call ast_category_insert with a match category that doesn't exist, the
list traverse runs out of 'next' categories and you get a SEGV. This patch
adds check for the end-of-list condition and changes the signature to return
an int for success/failure indication instead of a void.
The only consumer of this function is manager and it was also changed to use
the return value.
Tested by: George Joseph
Review: https://reviewboard.asterisk.org/r/3993/
........
Merged revisions 423276 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 423277 from http://svn.asterisk.org/svn/asterisk/branches/11
res_rtp_asterisk: Fix a myriad of TURN client issues.
1. The number of file descriptors an ioqueue instance can handle is fixed, so we
now spawn the required number to handle the load.
2. Our transport identifiers were exceeding the range supported by pjnath.
3. The TURN client did not set up client binding causing needless bandwidth usage.
4. The code no longer updates address information on each packet.
5. STUN traffic was getting looped back to Asterisk instead of going through the
TURN server.
6. Synchronization now ensures things are completely setup or destroyed.
7. Logging now reflects the target the TURN server is sending to/receiving from
on our behalf.
Kinsey Moore [Fri, 12 Sep 2014 18:19:48 +0000 (18:19 +0000)]
Blocked revisions 423010
........
Bridging: Fix bouncing native bridge
This fixes a situation in Asterisk 1.8 and 11 where ast_channel_bridge
could cause a bouncing native bridge. In the case of the
dial_LS_options test, this was a remote RTP bridge which caused the
audio path to continually cycle between Asterisk and the remote
endpoints generating a large number of SIP messages and delaying the
test long enough to cause it to fail (checking timing was part of the
test). The root cause was that the code to decide whether to use native
bridging was expecting a time-remaining value of 0 to be the default
instead of the actual default value of -1. A value of 0 or negative
numbers could also be generated by preceding code in some
circumstances. Both issues are addressed in this patch.
ASTERISK-24211 #close
Reported by: Matt Jordan
Review: https://reviewboard.asterisk.org/r/3987/
........
Merged revisions 423006 from http://svn.asterisk.org/svn/asterisk/branches/1.8
Jonathan Rose [Fri, 12 Sep 2014 16:01:36 +0000 (16:01 +0000)]
Realtime: Fix a bug that caused realtime destroy command to crash
Also has could affect with anything that goes through ast_destroy_realtime.
If a CLI user used the command 'realtime destroy <family>' with only a single
column/value pair, Asterisk would crash when trying to create a variable list
from a NULL value.
Mark Michelson [Thu, 11 Sep 2014 22:16:03 +0000 (22:16 +0000)]
Remove undocumented default behavior of ast_play_and_record_full acceptdtmf.
ast_play_and_record_full() has a parameter called "acceptdtmf" that is a
string of acceptable DTMF digits that may be pressed by a caller to end
and accept the recording.
ARI uses this function in order to perform recording, and it provides
options for what is passed as acceptdtmf to ast_play_and_record_full().
By default, ARI passes an empty string, with the intention that no DTMF
can be used to end the recording.
The problem is that ast_play_and_record_full() attempts to be "helpful"
by setting "#" as the acceptdtmf if an empty string or NULL pointer
has been passed in. With ARI, this results in unexpected behavior
occurring if you have attempted to intercept "#" yourself in order
to perform some other manipulation of the live recording.
This change removes the "helpful" behavior by no longer accepting
"#" as a default acceptdtmf if none is specified by the caller of
ast_play_and_record_full(). This makes the ARI scenario work as
expected.
The other callers of ast_play_and_record_full() are app_voicemail
and app_minivm, and in both cases, they pass an explicit "#" to
ast_play_and_record_full() as acceptdtmf, so they are unaffected
by this change.
George Joseph [Wed, 10 Sep 2014 16:02:54 +0000 (16:02 +0000)]
config: bug: fix truncation of included config files on permissions error
ast_config_text_file_save() currently truncates include files as they
are processed. If a subsequent include file or the main config file has
a permissions error that prevents writing, earlier include files are left
truncated resulting in a frantic search for backups.
This patch causes ast_config_text_file_save to check for write access
on all files before it truncates any of them.
Will be applied 1.8 > trunk.
Tested by: George Joseph
Review: https://reviewboard.asterisk.org/r/3986/
........
Merged revisions 422900 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 422903 from http://svn.asterisk.org/svn/asterisk/branches/11
Rusty Newton [Sun, 7 Sep 2014 00:09:54 +0000 (00:09 +0000)]
Sounds/BuildSystem: Modifications to include new releases and Japanese language.
Modifying Makefile and sounds.xml to include new core 1.4.26 and extra 1.4.15
sound prompt releases, plus the new Japanese core sound prompts contributed
by QLOOG.
ASTERISK-23324
Reported by: Kevin McCoy
Tested by: Rusty Newton
........
Merged revisions 422789 from http://svn.asterisk.org/svn/asterisk/branches/1.8
........
Merged revisions 422790 from http://svn.asterisk.org/svn/asterisk/branches/11
Matthew Jordan [Sat, 6 Sep 2014 22:48:24 +0000 (22:48 +0000)]
main/cdr: Copy over location information during a fork
When a CDR is forked, a new CDR is created and appended to the CDR chain for
the Party A. The forked CDR starts life off as a clone of the last
non-finalized for the particular Party A. In the past, merely copying over
the snapshots for Party A/Party B would be sufficient. However, as the CDRs
now contain cached information from Party A - specifically application/data,
context, and extension - we need to copy that over during a fork as well.
Huzzah for unit tests catching this when the context/extension were derived
from a cached value on the CDR instead of on Party A.
Matthew Jordan [Sat, 6 Sep 2014 22:20:27 +0000 (22:20 +0000)]
main/rtp_engine: Format NTP timestamps as unsigned ints
On some systems, a timeval's tv_sec/tv_usec will be unsigned lont ints, as
opposed to long ints. When the RTP engine formats these as strings, it was
previously formatting them as signed integers, which can result in some
odd negative timestamp values (particularly on 32-bit systems). This patch
formats the values as unsigned long integers.
projects / thirdparty / asterisk.git / log
