Sean Bright [Thu, 7 Oct 2021 17:50:08 +0000 (13:50 -0400)]
Makefile: Use basename in a POSIX-compliant way.
If you aren't using GNU coreutils, chances are that your basename
doesn't know about the -s argument. Luckily for us, basename does what
we need it do even without the -s argument.
Naveen Albert [Mon, 24 May 2021 18:04:12 +0000 (14:04 -0400)]
chan_iax2: Add encryption for RSA authentication
Adds support for encryption to RSA-authenticated
calls. Also prevents crashes if an RSA IAX2 call
is initiated to a switch requiring encryption
but no secret is provided.
Matthew Kern [Mon, 19 Jul 2021 16:34:00 +0000 (10:34 -0600)]
res_pjsip_t38: bind UDPTL sessions like RTP
In res_pjsip_sdp_rtp, the bind_rtp_to_media_address option and the
fallback use of the transport's bind address solve problems sending
media on systems that cannot send ipv4 packets on ipv6 sockets, and
certain other situations. This change extends both of these behaviors
to UDPTL sessions as well in res_pjsip_t38, to fix fax-specific
problems on these systems, introducing a new option
endpoint/t38_bind_udptl_to_media_address.
Naveen Albert [Wed, 29 Sep 2021 17:58:14 +0000 (17:58 +0000)]
app_read: Fix null pointer crash
If the terminator character is not explicitly specified
and an indications tone is used for reading a digit,
there is no null pointer check so Asterisk crashes.
This prevents null usage from occuring.
main/say.c: Support future dates with Q and q format params
The current versions do not support future dates in all say application when using the 'Q' or 'q' format parameter and says "today" for everything that is greater than today
Joseph Nadiv [Wed, 21 Jul 2021 21:36:03 +0000 (17:36 -0400)]
res_pjsip_registrar: Remove unavailable contacts if exceeds max_contacts
The behavior of max_contacts and remove_existing are connected. If
remove_existing is enabled, the soonest expiring contacts are removed.
This may occur when there is an unavailable contact. Similarly,
when remove_existing is not enabled, registrations from good
endpoints are rejected in favor of retaining unavailable contacts.
This commit adds a new AOR option remove_unavailable, and the effect
of this setting will depend on remove_existing. If remove_existing
is set to no, we will still remove unavailable contacts when they
exceed max_contacts, if there are any. If remove_existing is set to
yes, we will prioritize the removal of unavailable contacts before
those that are expiring soonest.
Sean Bright [Tue, 21 Sep 2021 14:58:09 +0000 (10:58 -0400)]
message.c: Support 'To' header override with AMI's MessageSend.
The MessageSend AMI action has been updated to allow the Destination
and the To addresses to be provided separately. This brings the
MessageSend manager command in line with the capabilities of the
MessageSend dialplan application.
Naveen Albert [Sun, 5 Sep 2021 18:11:57 +0000 (18:11 +0000)]
app_queue: Fix hint updates for included contexts
Previously, if custom hints were used with the hint:
format in app_queue, when device state changes occured,
app_queue would only do a literal string comparison of
the context used for the hint in app_queue and the context
of the hint which just changed state. This caused hints
to not update and become stale if the context associated
with the agent included the context which actually changes
state, essentially completely breaking device state for
any such agents defined in this manner.
This fix adds an additional check to ensure that included
contexts are also compared against the context which changed
state, so that the behavior is correct no matter whether the
context is specified to app_queue directly or indirectly.
Rather than stripping parameters from Content-Type headers before
comparison, first try to compare the whole string. If no match is
found, strip the parameters and try that way.
George Joseph [Tue, 14 Sep 2021 17:02:01 +0000 (11:02 -0600)]
pjproject: Add patch to fix trailing whitespace issue in rtpmap
An issue was found where a particular manufacturer's phones add a
trailing space to the end of the rtpmap attribute when specifying
a payload type that has a "param" after the format name and clock
rate. For example:
a=rtpmap:120 opus/48000/2 \r\n
Because pjmedia_sdp_attr_get_rtpmap currently takes everything after
the second '/' up to the line end as the param, the space is
included in future comparisons, which then fail if the param being
compared to doesn't also have the space.
We now use pj_scan_get() to parse the param part of rtpmap so
trailing whitespace is automatically stripped.
Carlos Oliva [Mon, 13 Sep 2021 15:18:11 +0000 (17:18 +0200)]
app_mp3: Force output to 16 bits in mpg123
In new mpg123 versions (since 1.26) the default output is 32 bits
Asterisk expects the output in 16 bits, so we force the output to be on 16 bits.
It will work wit new and old versions of mpg123.
Thanks Thomas Orgis <thomas-forum@orgis.org> for giving the key!
Naveen Albert [Thu, 2 Sep 2021 23:20:43 +0000 (23:20 +0000)]
app_stack: Include current location if branch fails
Previously, the error emitted when app_stack tries
to branch to a dialplan location that doesn't exist
has included only the information about the attempted
branch in the error log. This adds the current location
as well so users can see where the branch failed in
the logs.
Naveen Albert [Thu, 2 Sep 2021 23:57:03 +0000 (23:57 +0000)]
func_strings: Add STRBETWEEN function
Adds the STRBETWEEN function, which can be used to insert a
substring between each character in a string. For instance,
this can be used to insert pauses between DTMF tones in a
string of digits.
Naveen Albert [Fri, 3 Sep 2021 00:00:50 +0000 (00:00 +0000)]
func_env: Add DIRNAME and BASENAME functions
Adds the DIRNAME and BASENAME functions, which are
wrappers around the corresponding C library functions.
These can be used to safely and conveniently work with
file paths and names in the dialplan.
Naveen Albert [Mon, 26 Jul 2021 17:46:44 +0000 (17:46 +0000)]
func_sayfiles: Retrieve say file names
Up until now, all of the logic used to translate
arguments to the Say applications has been
directly coupled to playback, preventing other
modules from using this logic.
This refactors code in say.c and adds a SAYFILES
function that can be used to retrieve the file
names that would be played. These can then be
used in other applications or for other purposes.
Additionally, a SayMoney application and a SayOrdinal
application are added. Both SayOrdinal and SayNumber
are also expanded to support integers greater than
one billion.
Naveen Albert [Mon, 9 Aug 2021 17:41:24 +0000 (17:41 +0000)]
res_tonedetect: Tone detection module
dsp.c contains arbitrary tone detection functionality
which is currently only used for fax tone recognition.
This change makes this functionality publicly
accessible so that other modules can take advantage
of this.
Additionally, a WaitForTone and TONE_DETECT app and
function are included to allow users to do their
own tone detection operations in the dialplan.
Sean Bright [Sat, 4 Sep 2021 17:07:14 +0000 (13:07 -0400)]
term.c: Add support for extended number format terminfo files.
ncurses 6.1 introduced an extended number format for terminfo files
which the terminfo parsing in Asterisk is not able to parse. This
results in some TERM values that do support color (screen-256color on
Ubuntu 20.04 for example) to not get a color console.
Sean Bright [Tue, 7 Sep 2021 17:32:33 +0000 (13:32 -0400)]
app_voicemail.c: Ability to silence instructions if greeting is present.
There is an option to silence voicemail instructions but it does not
take into consideration if a recorded greeting exists or not. Add a
new 'S' option that does that.
When compiled without extended srtp crypto suites also disable parsing
these from received SDP. This prevents using these, as some client
implementations are not stable.
Sean Bright [Wed, 25 Aug 2021 15:21:07 +0000 (11:21 -0400)]
config_options: Handle ACO arrays correctly in generated XML docs.
There are 3 separate changes here but they are all closely related:
* Only try to set matchfield attributes on 'field' nodes
* We need to adjust how we treat the category pointer based on the
value of the category_match, to avoid memory corruption. We now
generate a regex-like string when match types other than
ACO_WHITELIST and ACO_BLACKLIST are used.
* Switch app_agent_pool from ACO_BLACKLIST_ARRAY to
ACO_BLACKLIST_EXACT since we only have one category we need to
ignore, not two.
Mark Murawski [Tue, 24 Aug 2021 18:25:50 +0000 (14:25 -0400)]
pbx_ael: Fix crash and lockup issue regarding 'ael reload'
Currently pbx_ael does not check if a reload is currently pending
before proceeding with a reload. This can cause multiple threads to
operate at the same time on what should be mutex protected data. This
change adds protection to reloading to ensure only one ael reload is
executing at a time.
Naveen Albert [Wed, 25 Aug 2021 11:49:06 +0000 (11:49 +0000)]
app_read: Allow reading # as a digit
Allows for the digit # to be read as a digit,
just like any other DTMF digit, as opposed to
forcing it to be used as an end of input
indicator. The default behavior remains
unchanged.
Naveen Albert [Wed, 25 Aug 2021 01:04:31 +0000 (01:04 +0000)]
bridge_basic: Change warning to verbose if transfer cancelled
The attended transfer feature will emit a warning if the user
cancels the transfer or the attended transfer doesn't complete
for any reason. Changes the warning to a verbose message,
since nothing is actually wrong here.
Sarah Autumn [Sun, 20 Jun 2021 04:36:34 +0000 (21:36 -0700)]
sig_analog: Changes to improve electromechanical signalling compatibility
This changeset is intended to address compatibility issues encountered
when interfacing Asterisk to electromechanical telephone switches that
implement ANI-B, ANI-C, or ANI-D.
In particular the behaviours that this impacts include:
- FGC-CAMA did not work at all when using MF signaling. Modified the
switch case block to send calls to the correct part of the
signaling-handling state machine.
- For FGC-CAMA operation, the delay between called number ST and
second wink for ANI spill has been made configurable; previously
all calls were made to wait for one full second.
- After the ANI spill, previous behavior was to require a 'ST' tone
to advance the call. This has been changed to allow 'STP' 'ST2P'
or 'ST3P' as well, for compatibility with ANI-D.
- Store ANI2 (ANI INFO) digits in the CALLERID(ANI2) channel variable.
- For calls with an ANI failure, No. 1 Crossbar switches will send
forward a single-digit failure code, with no calling number digits
and no ST pulse to terminate the spill. I've made the ANI timeout
configurable so to reduce dead air time on calls with ANI fail.
- ANI info digits configurable. Modern digital switches will send 2
digits, but ANI-B sends only a single info digit. This caused the
ANI reported by Asterisk to be misaligned.
- Changed a confusing log message to be more informative.
Andre Barbosa [Thu, 5 Aug 2021 16:55:33 +0000 (17:55 +0100)]
media_cache: Don't lock when curl the remote file
When playing a remote sound file, which is not in cache, first we need
to download it with ast_bucket_file_retrieve.
This can take a while if the remote host is slow. The current CURL
timeout is 180secs, so in extreme situations, it can take 3 minutes to
return.
Because ast_media_cache_retrieve has a lock on all function, while we
are waiting for the delayed download, Asterisk is not able to play any
more files, even the files already cached locally.
George Joseph [Mon, 16 Aug 2021 13:25:44 +0000 (07:25 -0600)]
res_pjproject: Allow mapping to Asterisk TRACE level
Allow mapping pjproject log messages to the Asterisk TRACE
log level. The defaults were also changes to log pjproject
levels 3,4 to DEBUG and 5,6 to TRACE. Previously 3,4,5,6
all went to DEBUG.
Naveen Albert [Thu, 12 Aug 2021 21:02:27 +0000 (21:02 +0000)]
app_milliwatt: Timing fix
The Milliwatt application uses incorrect tone timings
that cause it to play the 1004 Hz tone constantly.
This adds an option to enable the correct timing
behavior, so that the Milliwatt application can
be used for milliwatt test lines. The default behavior
remains unchanged for compatability reasons, even
though it is incorrect.
Naveen Albert [Wed, 4 Aug 2021 14:46:42 +0000 (14:46 +0000)]
app_morsecode: Add American Morse code
Previously, the Morsecode application only supported international
Morse code. This adds support for American Morse code and adds an
option to configure the frequency used in off intervals.
Additionally, the application checks for hangup between tones
to prevent application execution from continuing after hangup.
Naveen Albert [Mon, 28 Jun 2021 14:25:24 +0000 (14:25 +0000)]
func_math: Return integer instead of float if possible
The MIN, MAX, and ABS functions all support float
arguments, but currently return floats even if the
arguments are all integers and the response is
a whole number, in which case the user is likely
expecting an integer. This casts the float to an integer
before printing into the response buffer if possible.
Naveen Albert [Wed, 4 Aug 2021 19:16:59 +0000 (19:16 +0000)]
func_scramble: Audio scrambler function
Adds a function to scramble audio on a channel using
whole spectrum frequency inversion. This can be used
as a privacy enhancement with applications like
ChanSpy or other potentially sensitive audio.
Naveen Albert [Thu, 5 Aug 2021 00:28:46 +0000 (00:28 +0000)]
app_originate: Add ability to set codecs
A list of codecs to use for dialplan-originated calls can
now be specified in Originate, similar to the ability
in call files and the manager action.
Additionally, we now default to just using the slin codec
for originated calls, rather than all the slin* codecs up
through slin192, which has been known to cause issues
and inconsistencies from AMI and call file behavior.
Alexander Traud [Mon, 16 Aug 2021 16:11:56 +0000 (18:11 +0200)]
BuildSystem: Remove two dead exceptions for compiler Clang.
Commit 305ce3d added -Wno-parentheses-equality to Makefile.rules,
turning the previous two warning suppressions from commit e9520db
redundant. Let us remove the latter.
Joshua C. Colp [Wed, 11 Aug 2021 11:15:16 +0000 (08:15 -0300)]
policy: Add deprecation and removal versions to modules.
app_meetme is deprecated in 19, to be removed in 21.
app_osplookup is deprecated in 19, to be removed in 21.
chan_alsa is deprecated in 19, to be removed in 21.
chan_mgcp is deprecated in 19, to be removed in 21.
chan_skinny is deprecated in 19, to be removed in 21.
res_pktccops is deprecated in 19, to be removed in 21.
cdr_mysql was deprecated in 1.8, to be removed in 19.
app_mysql was deprecated in 1.8, to be removed in 19.
app_ices was deprecated in 16, to be removed in 19.
app_macro was deprecated in 16, to be removed in 21.
app_fax was deprecated in 16, to be removed in 19.
app_url was deprecated in 16, to be removed in 19.
app_image was deprecated in 16, to be removed in 19.
app_nbscat was deprecated in 16, to be removed in 19.
app_dahdiras was deprecated in 16, to be removed in 19.
cdr_syslog was deprecated in 16, to be removed in 19.
chan_oss was deprecated in 16, to be removed in 19.
chan_phone was deprecated in 16, to be removed in 19.
chan_sip was deprecated in 17, to be removed in 21.
chan_nbs was deprecated in 16, to be removed in 19.
chan_misdn was deprecated in 16, to be removed in 19.
chan_vpb was deprecated in 16, to be removed in 19.
res_config_sqlite was deprecated in 16, to be removed in 19.
res_monitor was deprecated in 16, to be removed in 21.
conf2ael was deprecated in 16, to be removed in 19.
muted was deprecated in 16, to be removed in 19.
Alexander Traud [Mon, 2 Aug 2021 17:33:07 +0000 (19:33 +0200)]
aelparse: Accept an included context with timings.
With Asterisk 1.6.0, in the main parser for the configuration file
extensions.conf, the separator was changed from vertical bar to comma.
However, the first separator was not changed in aelparse; it still had
to be a vertical bar, and no comma was allowed.
Additionally, this change allows the vertical bar for the first and
last parameter again, even in the main parser, because the vertical bar
was still accepted for the other parameters.
Naveen Albert [Mon, 28 Jun 2021 13:48:16 +0000 (13:48 +0000)]
cdr_adaptive_odbc: Prevent filter warnings
Previously, if CDR filters were used so that
not all CDR records used all sections defined
in cdr_adaptive_odbc.conf, then warnings will
always be emitted (if each CDR record is unique
to a particular section, n-1 warnings to be
specific).
This turns the offending warning log into
a verbose message like the other one, since
this behavior is intentional and not
indicative of anything wrong.
Allows multiple files comprising an agent announcement
to be played by separating on the ampersand, similar
to the multi-file support in other Asterisk applications.
res_pjsip_header_funcs: Add PJSIP_HEADERS() ability to read header by pattern
PJSIP currently does not provide a function to replace SIP_HEADERS() function to get a list of headers from INVITE request.
It may be used to get all X- headers in case the actual set and names of headers unknown.
Meter types are not well supported,
lacking support in telegraf, datadog and the official statsd servers.
We deprecate meters and provide a compliant fallback for any existing usages.
A flag has been introduced to allow meters to fallback to counters.
Naveen Albert [Wed, 16 Jun 2021 20:26:46 +0000 (16:26 -0400)]
app_dtmfstore: New application to store digits
Adds application to asynchronously collect digits
dialed on a channel in the TX or RX direction
using a framehook and stores them in a specified
variable, up to a configurable number of digits.
under [Thu, 22 Jul 2021 16:39:11 +0000 (19:39 +0300)]
codec_builtin.c: G729 audio gets corrupted by Asterisk due to smoother
If Asterisk gets G.729 6-byte VAD frames inbound, then at outbound Asterisk sends this G.729 stream with non-continuous timestamps.
This makes the audio stream not-playable at the receiver side.
Linphone isn't able to play such an audio - lots of disruptions are heard.
Also I had complains of bad audio from users which use other types of phones.
After debugging, I found this is a regression connected with RTP Smoother (main/smoother.c).
Smoother has a special code to handle G.729 VAD frames (search for AST_SMOOTHER_FLAG_G729 in smoother.c).
However, this flag is never set in Asterisk-12 and newer.
Previously it has been set (see Asterisk-11).
Kevin Harwell [Mon, 14 Jun 2021 18:28:25 +0000 (13:28 -0500)]
AST-2021-009 - pjproject-bundled: Avoid crash during handshake for TLS
If an SSL socket parent/listener was destroyed during the handshake,
depending on timing, it was possible for the handling callback to
attempt access of it after the fact thus causing a crash.
Kevin Harwell [Mon, 10 May 2021 22:59:00 +0000 (17:59 -0500)]
AST-2021-008 - chan_iax2: remote crash on unsupported media format
If chan_iax2 received a packet with an unsupported media format, for
example vp9, then it would set the frame's format to NULL. This could
then result in a crash later when an attempt was made to access the
format.
This patch makes it so chan_iax2 now ignores/drops frames received
with unsupported media format types.
AST-2021-007 - res_pjsip_session: Don't offer if no channel exists.
If a re-INVITE is received after we have sent a BYE request then it
is possible for no channel to be present on the session. If this
occurs we allow PJSIP to produce the offer instead. Since the call
is being hung up if it produces an incorrect offer it doesn't
actually matter. This also ensures that code which produces SDP
does not need to handle if a channel is not present.
Andre Barbosa [Tue, 29 Jun 2021 16:07:44 +0000 (17:07 +0100)]
res_stasis_playback: Check for chan hangup on play_on_channels
Verify `ast_check_hangup` before looping to the next sound file.
If the call is already hangup we just break the cycle.
It also ensures that the PlaybackFinished event is sent if the call was hangup.
This is also use-full when we are playing a big list of file for a channel that is hangup.
Before this patch Asterisk will give a warning for every sound not played and fire a PlaybackStart for every sound file on the list tried to be played.
With the patch we just break the playback cycle when the chan is hangup.
Sean Bright [Thu, 15 Jul 2021 20:04:01 +0000 (16:04 -0400)]
res_pjsip_stir_shaken: RFC 8225 compliance and error message cleanup.
From RFC 8225 Section 5.2.1:
The "dest" claim is a JSON object with the claim name of "dest"
and MUST have at least one identity claim object. The "dest"
claim value is an array containing one or more identity claim JSON
objects representing the destination identities of any type
(currently "tn" or "uri"). If the "dest" claim value array
contains both "tn" and "uri" claim names, the JSON object should
list the "tn" array first and the "uri" array second. Within the
"tn" and "uri" arrays, the identity strings should be put in
lexicographical order, including the scheme-specific portion of
the URI characters.
Additionally, make it clear that there was a failure to sign the JWT
payload and not necessarily a memory allocation failure.
Naveen Albert [Sat, 22 May 2021 14:31:22 +0000 (10:31 -0400)]
app_waitforcond: New application
While several applications exist to wait for
a certain event to occur, none allow waiting
for any generic expression to become true.
This application allows for waiting for a condition
to become true, with configurable timeout and
checking interval.
Andre Barbosa [Fri, 4 Jun 2021 11:11:10 +0000 (12:11 +0100)]
res_stasis_playback: Send PlaybackFinish event only once for errors
When we try to play a list of sound files in the same Play command,
we get only one PlaybackFinish event, after all sounds are played.
But in the case where the Play fails (because channel is destroyed
for example), Asterisk will send one PlaybackFinish event for each
sound file still to be played. If the list is big, Asterisk is
sending many events.
This patch adds a failed state so we can understand that the play
failed. On that case we don't send the event, if we still have a
list of sounds to be played.
When we reach the last sound, we send the PlaybackFinish with
the failed state.
If the system time has stepped backwards because of a time
adjustment between the time a frame is timestamped and the
time we check the timestamps in abstract_jb:hook_event_cb(),
we get a negative interval, but we don't check for that there.
abstract_jb:hook_event_cb() then calls
fixedjitterbuffer:fixed_jb_get() (via abstract_jb:jb_get_fixed)
and the first thing that does is assert(interval >= 0).
There are several issues with this...
* abstract_jb:hook_event_cb() saves the interval in a variable
named "now" which is confusing in itself.
* "now" is defined as an unsigned int which converts the negative
value returned from ast_tvdiff_ms() to a large positive value.
* fixed_jb_get()'s parameter is defined as a signed int so the
interval gets converted back to a negative value.
* fixed_jb_get()'s assert is NOT an ast_assert but a direct define
that points to the system assert() so it triggers even in
production mode.
So...
* hook_event_cb()'s "now" was renamed to "relative_frame_start" and
changed to an int64_t.
* hook_event_cb() now checks for a negative value right after
retrieving both the current and framedata timestamps and just
returns the frame if the difference is negative.
* fixed_jb_get()'s local define of ASSERT() was changed to call
ast_assert() instead of the system assert().
Naveen Albert [Sat, 22 May 2021 00:08:58 +0000 (20:08 -0400)]
app_dial: Expanded A option to add caller announcement
Hitherto, the A option has made it possible to play
audio upon answer to the called party only. This option
is expanded to allow for playback of an audio file to
the caller instead of or in addition to the audio
played to the answerer.
Joshua C. Colp [Mon, 21 Jun 2021 11:31:41 +0000 (08:31 -0300)]
core: Don't play silence for Busy() and Congestion() applications.
When using the Busy() and Congestion() applications the
function ast_safe_sleep is used by wait_for_hangup to safely
wait on the channel. This function may send silence if Asterisk
is configured to do so using the transmit_silence option.
In a scenario where an answered channel dials a Local channel
either directly or through call forwarding and the Busy()
or Congestion() dialplan applications were executed with the
transmit_silence option enabled the busy or congestion
tone would not be heard.
This is because inband generation of tones (such as busy
and congestion) is stopped when other audio is sent to
the channel they are being played to. In the given
scenario the transmit_silence option would result in
silence being sent to the channel, thus stopping the
inband generation.
This change adds a variant of ast_safe_sleep which can be
used when silence should not be played to the channel. The
wait_for_hangup function has been updated to use this
resulting in the tones being generated as expected.
Bernd Zobl [Fri, 7 May 2021 06:18:40 +0000 (08:18 +0200)]
res_pjsip_sdp_rtp: Evaluate remotely held for Session Progress
With the fix for ASTERISK_28754 channels are no longer put on hold if an
outbound INVITE is answered with a "Session Progress" containing
"inactive" audio.
The previous change moved the evaluation of the media attributes to
`negotiate_incoming_sdp_stream()` to have the `remotely_held` status
available when building the SDP in `create_outgoing_sdp_stream()`.
This however means that an answer to an outbound INVITE, which does not
traverse `negotiate_incoming_sdp_stream()`, cannot set the
`remotely_held` status anymore.
This change moves the check so that both, `negotiate_incoming_sdp_stream()` and
`apply_negotiated_sdp_stream()` can do the checks.
George Joseph [Wed, 16 Jun 2021 13:50:54 +0000 (07:50 -0600)]
res_pjsip_messaging: Overwrite user in existing contact URI
When the MessageSend destination is in the form
PJSIP/<number>@<endpoint> and the endpoint's contact
URI already has a user component, that user component
will now be replaced with <number> when creating the
request URI.
Bernd Zobl [Tue, 16 Mar 2021 16:45:23 +0000 (17:45 +0100)]
res_pjsip/pjsip_message_filter: set preferred transport in pjsip_message_filter
Set preferred transport when querying the local address to use in
filter_on_tx_messages(). This prevents the module to erroneously select
the wrong transport if more than one transports of the same type (TCP or
TLS) are configured.
Naveen Albert [Thu, 10 Jun 2021 14:34:53 +0000 (10:34 -0400)]
pbx_builtins: Corrects SayNumber warning
Previously, SayNumber always emitted a warning if the caller hung up
during execution. Usually this isn't correct, so check if the channel
hung up and, if so, don't emit a warning.
Jaco Kroon [Sat, 22 May 2021 12:29:22 +0000 (14:29 +0200)]
func_lock: Add "dialplan locks show" cli command.
For example:
arthur*CLI> dialplan locks show
func_lock locks:
Name Requesters Owner
uls-autoref 0 (unlocked)
1 total locks listed.
Obviously other potentially useful stats could be added (eg, how many
times there was contention, how many times it failed etc ... but that
would require keeping the stats and I'm not convinced that's worth the
effort. This was useful to troubleshoot some other issues so submitting
it.
The scenario where a channel still has an associated datastore we
cannot unload since there is a function pointer to the destroy and fixup
functions in play. Thus increase the module ref count whenever we
allocate a datastore, and decrease it during destroy.
In order to tighten the race that still exists in spite of this (below)
add some extra failure cases to prevent allocations in these cases.
Race:
If module ref is zero, an LOCK or TRYLOCK is invoked (near)
simultaneously on a channel that has NOT PREVIOUSLY taken a lock, and if
in such a case the datastore is created *prior* to unloading being set
to true (first step in module unload) then it's possible that the module
will unload with the destructor being called (and segfault) post the
module being unloaded. The module will however wait for such locks to
release prior to unloading.
If post that we can recheck the module ref before returning the we can
(in theory, I think) eliminate the last of the race. This race is
mostly theoretical in nature.
Jaco Kroon [Sat, 22 May 2021 12:42:04 +0000 (14:42 +0200)]
func_lock: Fix memory corruption during unload.
AST_TRAVERSE accessess current as current = current->(field).next ...
and since we free current (and ast_free poisons the memory) we either
end up on a ast_mutex_lock to a non-existing lock that can never be
obtained, or a segfault.
Incidentally add logging in the "we have to wait for a lock to release"
case, and remove an ineffective statement that sets memory that was just
cleared by ast_calloc to zero.
Jaco Kroon [Sat, 22 May 2021 12:48:18 +0000 (14:48 +0200)]
func_lock: Fix requesters counter in error paths.
In two places we bail out with failure after we've already incremented
the requesters counter, if this occured then it would effectively result
in unload to wait indefinitely, thus preventing clean shutdown.
Naveen Albert [Tue, 25 May 2021 15:36:04 +0000 (11:36 -0400)]
app_originate: Allow setting Caller ID and variables
Caller ID can now be set on the called channel and
Variables can now be set on the destination
using the Originate application, just as
they can be currently using call files
or the Manager Action.
Naveen Albert [Mon, 24 May 2021 00:20:28 +0000 (20:20 -0400)]
app_confbridge: New ConfKick() application
Adds a new ConfKick() application, which may
be used to kick a specific channel, all channels,
or all non-admin channels from a specified
conference bridge, similar to existing CLI and
AMI commands.
Naveen Albert [Wed, 2 Jun 2021 13:11:24 +0000 (09:11 -0400)]
sip_to_pjsip: Fix missing cases
Adds the "auto" case which is valid with
both chan_sip dtmfmode and chan_pjsip's
dtmf_mode, adds subscribecontext to
subscribe_context conversion, and accounts
for cipher = ALL being invalid.
projects / thirdparty / asterisk.git / log
