Michael L. Young [Thu, 18 Jul 2013 12:51:22 +0000 (12:51 +0000)]
Properly indicate failure to open an audio stream in res_agi
If there is an error streaming an audio file, the current return status makes it
difficult for an AGI script to determine that there was an error with the audio
file.
This patches changes the result to return -1 and the function returns
RESULT_FAILURE instead of RESULT_SUCCESS. From looking at other parts of
res_agi, this would appear to be the proper way to handle an error.
(closes issue ASTERISK-21903)
Reported by: Ariel Wainer
Tested by: Ariel Wainer
Patches:
asterisk-21903-return-stream-res_1.8.diff
by Michael L. Young (license 5026)
Matthew Jordan [Sun, 14 Jul 2013 01:53:15 +0000 (01:53 +0000)]
Clarify documentation for function PASSTHRU
It is not apparent to the average user that the PASSTHRU function should not
be passed as ${PASSTHRU(string)} but just as PASSTHRU(string) to functions
which take a variable name and not its contents.
This patch clarifies the behavior in the documentation and provides an example.
(closes issue ASTERISK-21717)
Reported by: Richard Miller
patches:
func_strings.diff uploaded by Richard Miller (license 5685)
Fix a longstanding issue with MFC-R2 configuration that prevented users
from mixing different variants or general MFC-R2 settings within the same E1 line.
Most users do not have a problem with this since MFC-R2 lines are usually fractional E1s, or
the whole E1 has the same country variant and R2 settings.
In Venezuela however is common to have inbound MFC-R2 and outbound DTMF-R2 within the same E1.
This fix now properly parses the chan_dahdi.conf file to generate a new openr2 context every
time a new channel => section is found and the configuration was changed.
(closes issue ASTERISK-21117)
Reported by: Rafael Angulo
Related Elastix issue: http://bugs.elastix.org/view.php?id=1612
Russell Bryant [Wed, 10 Jul 2013 01:41:42 +0000 (01:41 +0000)]
astobj2-ify the SLA code
The SLA code within app_meetme was written before asotbj2 had been
merged into Asterisk. Worse, support for reloads did not exist at first
and was added later as a bolt-on feature. I knew at the time that
reloading was not safe at all while SLA was in use, so the reload would
be queued up to execute when the system was idle. Unfortunately, this
approach was still prone to errors beyond the fact that this was the
only place in Asterisk where configuration was not reloaded
instantly when requested.
This patch converts various SLA objects to be reference counted objects
using astobj2. This allows reloads to be processed while the system is
in use. The code ensures that the objects will not disappear while one
of the other threads is using them. However, they will be immediately
removed from the global trunk and station containers so no new calls
will use them if removed from configuration.
Jonathan Rose [Fri, 14 Jun 2013 16:14:48 +0000 (16:14 +0000)]
app_mixmonitor: Fix crashes caused by unloading app_mixmonitor
Unloading app_mixmonitor while active mixmonitors were running would
cause a segfault. This patch fixes that by making it impossible to
unload app_mixmonitor while mixmonitors are active.
Matthew Jordan [Wed, 12 Jun 2013 02:19:52 +0000 (02:19 +0000)]
Fix memory leak while loading priority modules
When we load a module with the LOAD_PRIORITY flag, we remove its entry from
the load order list. Unfortunately, we don't free the memory associated with
entry in the list. This patch corrects that and properly frees the memory for
the module in the list.
Matthew Jordan [Mon, 10 Jun 2013 14:15:19 +0000 (14:15 +0000)]
Add announce-to-first-user option for app_queue
In r386792, the ability to play prompts to the first caller in a call queue was
added. While this is arguably a bug fix for those who expect the first caller
to continue receiving prompts while the agent is dialed, it has the side effect
of preventing the first caller from hearing the agent immediately upon
bridging. This may not be a problem for those who really want this option, but
for those who didn't care whether or not the first caller in queue heard their
position, it was an issue.
This patch disables the ability for the first caller in the queue to hear
prompts and adds a new option, announce-to-first-user, to queues.conf. Those
who the behavior can enable it by setting this value to True.
Note that if we ever implement the ability to have the prompts be stopped
upon bridging, this option can be removed.
Alec L Davis [Mon, 10 Jun 2013 07:30:35 +0000 (07:30 +0000)]
IAX2: fix race condition with nativebridge transfers.
1). When touching the bridgecallno, we need to lock it.
2). stop_stuff() which calls iax2_destroy_helper()
Assumes the lock on the pvt is already held, when iax2_destroy_helper() is called.
Thus we need to lock the bridgecallno pvt before we call stop_stuff(iaxs[fr->callno]->bridgecallno);
3). When evaluating the state of 'callno->transferring' of the current leg,
we can't change it to READY unless the bridgecallno is locked.
Why, if we are interrupted by the other call leg before 'transferring = TRANSFER_RELEASED',
the interrupt will find that it is READY and that the bridgecallno is also READY so Releases the legs.
Richard Mudgett [Wed, 29 May 2013 20:10:04 +0000 (20:10 +0000)]
Fix segfault when dealing with chan_agent channels.
Check the returned bridged pointer for NULL to avoid a crash. It looks
like chan_agent is returning a NULL pointer when it probably should be
returning a pointer to the channel the Agent channel is pretending to be.
(closes issue ASTERISK-21793)
Reported by: Rodrigo P. Telles
Patches:
jira_asterisk_21793_v1.8.patch (license #5621) patch uploaded by rmudgett
Tested by: Rodrigo P. Telles
Jonathan Rose [Tue, 28 May 2013 17:35:12 +0000 (17:35 +0000)]
Fix a memory copying bug in slinfactory which was causing mixmonitor issues.
Reported by: Michael Walton
Tested by: Jonathan Rose
Patches:
slinfactory.c.ASTERISK-21799.patch uploaded by Michael Walton (license 6502)
(closes issue ASTERISK-21799)
Matthew Jordan [Fri, 24 May 2013 11:42:38 +0000 (11:42 +0000)]
Print all logger messages on shutdown
When Asterisk shuts down and shuts down the loggin gsubsystem, any
messages currently in flight will not get logged. This patch prevents the
loop writing messages from breaking out prematurely, such that all of the
messages are logged.
Kevin Harwell [Wed, 15 May 2013 15:54:50 +0000 (15:54 +0000)]
Fix for segfault in __ast_rwlock_destroy with DEBUG_THREADS
If DEBUG_THREADS is enabled __ast_rwlock_destroy causes a segfault while trying
to access a possible NULL t->track object. A NULL check has been added before
trying to access the memory.
Richard Mudgett [Fri, 10 May 2013 22:09:32 +0000 (22:09 +0000)]
Allow mISDN to send PROGRESS messsage.
* Made isdn_msg_parser.c build a progress message with the mandatory
progress indicator IE. (The mISDNuser NT state machine rejected sending
the incomplete message.)
Note: The associated mISDN and mISDNuser patches respectively are viewable
here:
http://svnview.digium.com/svn/thirdparty?view=rev&rev=200
http://svnview.digium.com/svn/thirdparty?view=rev&rev=201
Fix The Payload Being Set On CN Packets And Do Not Set Marker Bit
When we send out a CN packet (for instance, in the case of using rtpkeepalives),
we are not setting the payload code properly. Also, we are setting the marker
bit when we shouldn't be according to RFC 3389, section 4.
AST_RTP_CN is not defined by AST_FORMAT codes. Therefore, we should be using
ast_rtp_codecs_payload_code() rather than ast_rtp_codecs_payload_lookup().
11 and trunk already use the appropriate function.
* In 1.8, use ast_rtp_codecs_payload_code()
* Remove the setting of the marker bit
* Fix the debug message by incrementing the seqno after the debug message is set
in order to display the correct seqno that was sent out
(closes issue ASTERISK-21246)
Reported by: Peter Katzmann
Tested by: Peter Katzmann, Michael L. Young
Patches:
asterisk-21246-rtp-cng-payload-error_1.8_v2.diff
uploaded by Michael L. Young (license 5026)
Alec L Davis [Wed, 8 May 2013 07:17:17 +0000 (07:17 +0000)]
chan_sip: NOTIFYs for BLF start queuing up and fail to be sent out after retries fail
RFC6665 4.2.2: ... after a failed State NOTIFY transaction remove the subscription
The problem is that the State Notify requests rely on the 200OK reponse for pacing control
and to not confuse the notify susbsystem.
The issue is, the pendinginvite isn't cleared if a response isn't received,
thus further notify's are never sent.
The solution, follow RFC 6665 4.2.2's 'SHOULD' and remove the subscription after failure.
(closes issue ASTERISK-21677)
Reported by: Dan Martens
Tested by: Dan Martens, David Brillert, alecdavis
alecdavis (license 585)
Russell Bryant [Mon, 6 May 2013 15:52:16 +0000 (15:52 +0000)]
Make SLA reload more paranoid.
Reload support was originally not included for SLA. It was added later,
but in a fairly non-traditional way. It basically sets a flag
indicating that a reload is pending, and then waits for a time where it
thinks everything SLA related is idle and unused, and *then* executes
the reload. It does this because the reload process is destructive. It
starts by throwing everything away and starting over.
There are a number of problems with this approach. One of them is that
the check to see if anything in use was incomplete. This patch makes it
more complete and thus less likely for a crash to occur during reload
processing. However, this approach still has problems so some much more
significant reworking of this code will need to come in as a next step.
Matthew Jordan [Thu, 2 May 2013 17:11:20 +0000 (17:11 +0000)]
Update utils Makefile to handle r387294
Alec's patch that added the Asterisk version to 'core show locks' angered the
items in utils, as they exist somewhat outside of the Asterisk build system.
Some day, this Makefile should get nuked from high orbit, but for now, include
version.c in its list of stuff to pile in.
Alec L Davis [Thu, 2 May 2013 07:53:55 +0000 (07:53 +0000)]
chan_sip: Session-Expires: Set timer to correctly expire at (~2/3) of the interval when not the refresher
RFC 4028 Section 10
if the side not performing refreshes does not receive a
session refresh request before the session expiration, it SHOULD send
a BYE to terminate the session, slightly before the session
expiration. The minimum of 32 seconds and one third of the session
interval is RECOMMENDED.
Prior to this asterisk would refresh at 1/2 the Session-Expires interval,
or if the remote device was the refresher, asterisk would timeout at interval end.
Now, when not refresher, timeout as per RFC noted above.
Alec L Davis [Thu, 2 May 2013 07:20:54 +0000 (07:20 +0000)]
chan_sip: Honor Session-Expires in 200OK response when it's a RE-INVITE when asterisk is the refresher.
RFC 4028 Section 7.2
"UACs MUST be prepared to receive a Session-Expires header field in a
response, even if none were present in the request."
What changed
After ASTERISK-20787, inbound calls to asterisk with no Session-Expires in the INVITE are now are offered
a Session-Expires (1800 asterisk default) in the response, with asterisk as the refresher.
Symptom:
After 900 seconds (asterisk default refresher period 1800), asterisk RE-INVITEs the device, the device
may respond with a much lower Session-Expires (180 in our case) value that it is now using.
Asterisk ignores this response, as it's deemed both an INBOUND CALL, and a RE-INVITE.
After 180 seconds the device times out and sends BYE (hangs up), asterisk is still working with the
refresher period of 1800 as it ignored the 'Session Expires: 180' in the previous 200OK response.
Fix:
handle_response_invite() when 200OK, remove check for outbound and reinvite.
Matthew Jordan [Wed, 1 May 2013 21:15:46 +0000 (21:15 +0000)]
Clear the DTMF sending digit tracking on off nominal paths
In certain situations, when the RTP engine goes to send a DTMF end digit
it may be in a situation where the remote address is no longer available,
or the digit that was supposed to be sent is invalid. In such cases, we
need to clear the RTP counters appropriately. Otherwise, when the RTP
source is set again, we'll continue to think that we're in the middle of
sending a DTMF digit, which can confuse the remote party (signficantly).
Matthew Jordan [Wed, 1 May 2013 18:34:44 +0000 (18:34 +0000)]
Prevent crash in 'sip show peers' when the number of peers on a system is large
When you have lots of SIP peers (according to the issue reporter, around 3500),
the 'sip show peers' CLI command or AMI action can crash due to a poorly placed
string duplication that occurs on the stack. This patch refactors the command
to not allocate the string on the stack, and handles the formatting of a single
peer in a separate function call.
Matthew Jordan [Tue, 30 Apr 2013 22:34:05 +0000 (22:34 +0000)]
Fix CDR not being created during an externally initiated blind transfer
Way back when in the dark days of Asterisk 1.8.9, blind transferring a call
in a context that included the 'h' extension would inadvertently execute the
hangup code logic on the transferred channel. This was a "bad thing". The fix
was to properly check for the softhangup flags on the channel and only execute
the 'h' extension logic (and, in later versions, hangup handler logic) if the
channel was well and truly dead (Jim).
Unfortunately, CDRs are fickle. Setting the softhangup flag when we detected
that the channel was leaving the bridge (but not to die) caused some crucial
snippet of CDR code, lying in ambush in the middle of the bridging code, to
not get executed. This had the effect of blowing away one of the CDRs that is
typically created during a blind transfer.
While we live and die by the adage "don't touch CDRs in release branches", this
was our bad. The attached patch restores the CDR behavior, and still manages to
not run the 'h' extension during a blind transfer (at least not when it's
supposed to).
Thanks to Steve Davies for diagnosing this and providing a fix.
Review: https://reviewboard.asterisk.org/r/2476
(closes issue ASTERISK-21394)
Reported by: Ishfaq Malik
Tested by: Ishfaq Malik, mjordan
patches:
fix_missing_blindXfer_cdr2 uploaded by one47 (License 5012)
Matthew Jordan [Fri, 26 Apr 2013 21:26:39 +0000 (21:26 +0000)]
Clean up memory leak in config file on off nominal paths when glob is allowed
If a system allows for its usage, Asterisk will use glob to help parse
Asterisk .conf files. The config file loading routine was leaking the memory
allocated by the glob() routine when the config file was in an unmodified
or invalid state.
This patch properly calls globfree in those off nominal paths.
Matthew Jordan [Fri, 26 Apr 2013 21:11:33 +0000 (21:11 +0000)]
Clean up resources in features on exit
This patch cleans up two things features:
* It properly unregisters the CLI commands that features registered
* It cancels and performs a pthread_join on the created parking thread. This
not only properly joins a non-detached thread, but also prevents disposing
of the parking lots prior to the parking thread completely exiting.
Richard Mudgett [Mon, 22 Apr 2013 16:10:29 +0000 (16:10 +0000)]
Fix crash when AMI redirect action redirects two channels out of a bridge.
The two party bridging loops were changing the bridge peer pointers
without the channel locks held. Thus when ast_channel_massquerade()
tested and used the pointer there is a small window of opportunity for the
pointers to become NULL even though the masquerade code has the channels
locked.
(closes issue ASTERISK-21356)
Reported by: William luke
Patches:
jira_asterisk_21356_v11.patch (license #5621) patch uploaded by rmudgett
Tested by: William luke
Matthew Jordan [Fri, 19 Apr 2013 15:59:25 +0000 (15:59 +0000)]
Prevent res_timing_pthread from blocking callers
There were several reports of deadlock when using
res_timing_pthread. Backtraces indicated that one thread was blocked
waiting for the write to the pipe to complete and this thread held
the container lock for the timers. Therefore any thread that wanted
to create a new timer or read an existing timer would block waiting
for either the timer lock or the container lock and deadlock ensued.
This patch changes the way the pipe is used to eliminate this source
of deadlocks:
1) The pipe is placed in non-blocking mode so that it would never
block even if the following changes someone fail...
2) Instead of writing bytes into the pipe for each "tick" that's
fired the pipe now has two states--signaled and unsignaled. If
signaled, the pipe is hot and any pollers of the read side
filedescriptor will be woken up. If unsigned the pipe is idle. This
eliminates even the chance of filling up the pipe and reduces the
potential overhead of calling unnecessary writes.
3) Since we're tracking the signaled / unsignaled state, we can
eliminate the exta poll system call for every firing because we know
that there is data to be read.
(closes issue ASTERISK-21389)
Reported by: Matt Jordan
Tested by: Shaun Ruffell, Matt Jordan, Tony Lewis
patches:
0001-res_timing_pthread-Reduce-probability-of-deadlocking.patch uploaded by sruffell (License 5417)
(closes issue ASTERISK-19754)
Reported by: Nikola Ciprich
(closes issue ASTERISK-20577)
Reported by: Kien Kennedy
(closes issue ASTERISK-17436)
Reported by: Henry Fernandes
Alec L Davis [Tue, 16 Apr 2013 23:11:46 +0000 (23:11 +0000)]
Distributed Device State broken at sites using res_xmpp or res_jabber where Secuity Advisory AST-2012-015 is inplace
res_jabber/res_xmpp were not adding AST_EVENT_IE_CACHABLE to the event as each message came in,
then devstate_change_collector_cb() was unable to find AST_EVENT_IE_CACHABLE in the event,
so defaulted incorrectly to AST_DEVSTATE_NOT_CACHABLE.
David M. Lee [Mon, 15 Apr 2013 14:38:24 +0000 (14:38 +0000)]
Fix the svn:keywords property on several files.
Normally I think keyword expansion is silly, but the one time it would have
been good, it didn't work because the property had quotes in it. This patch
fixes obviously busted svn:keywords properties.
Matthew Jordan [Sun, 14 Apr 2013 02:58:57 +0000 (02:58 +0000)]
Calculate the timestamp for outbound RTP if we don't have timing information
This patch calculates the timestamp for outbound RTP when we don't have timing
information. This uses the same approach in res_rtp_asterisk. Thanks to both
Pietro and Tzafrir for providing patches.
(closes issue ASTERISK-19883)
Reported by: Giacomo Trovato
Tested by: Pietro Bertera, Tzafrir Cohen
patches:
rtp-timestamp-1.8.patch uploaded by tzafrir (License 5035)
rtp-timestamp.patch uploaded by pbertera (License 5943)
Matthew Jordan [Sun, 14 Apr 2013 02:21:30 +0000 (02:21 +0000)]
Don't attempt to create a voice frame on a read error
Prior to this patch, a read error in snd_pcm_readi would still be treated as a
nominal result when constructing a voice frame from the expected data. Since
the value returned is negative, as opposed to the number of samples read,
this could result in a crash. With this patch, we now return a null frame
when a read error is detected.
Note that the patch on ASTERISK-21329 was modified slightly for this commit,
in that we bail immediately on detecting the read error, rather than bypassing
the construction of the voice frame.
Michael L. Young [Fri, 12 Apr 2013 22:34:46 +0000 (22:34 +0000)]
Fix Manager Segfault When app_queue Is Unloaded
When app_queue is unloaded, some manager commands are not being unregistered
which result in a segfault. This patch corrects this.
(closes issue ASTERISK-21397)
Reported by: Peter Katzmann, Corey Farrell
Tested by: Corey Farrell
Patches:
asterisk-21397-missing-unreg-manager-cmd_1.8.diff
Michael L. Young (license 5026)
asterisk-21397-missing-unreg-manager-cmd_11.diff
Michael L. Young (license 5026)
Michael L. Young [Fri, 12 Apr 2013 22:14:31 +0000 (22:14 +0000)]
Fix app_voicemail Segfault And A Few Memory Leaks
The original report was that app_voicemail would crash. This was caused by
ast_config_load() returning CONFIG_STATUS_FILEINVALID but no checks being
performed for that return status. After adding the initial patch to fix this
issue, Jaco Kroon (jkroon) added some fixes to memory leaks he had discovered.
During review, Walter Doekes (wdoekes) suggested adding a helper function in
order to determine if we had a valid configuration or not.
This patch does the following:
* Creates a helper function to check if the configuration is valid
* Adds calls to the new helper function where appropiate
* Fixes memory leaks where the code returned without running
ast_config_destroy() on the configuration that was loaded
(closes issue ASTERISK-21302)
Reported by: Jaco Kroon
Tested by: Jaco Kroon, Michael L. Young
Patches:
asterisk-11.3.0-app_voicemail-ast_config-fixes.patch
Jaco Kroon (license 5671)
asterisk-21302-valid_cfg_and_mem_leaks_v3-1.8.diff
Michael L. Young (license 5026)
Alec L Davis [Fri, 12 Apr 2013 08:14:20 +0000 (08:14 +0000)]
IAX2, prevent network thread starting before all helper threads are ready
On startup, it's possible for a frame to arrive before the processing threads were ready.
In iax2_process_thread() the first pass through falls into ast_cond_wait, should a frame arrive
before we are at ast_cond_wait, the signal will be ignored.
The result iax2_process_thread stays at ast_cond_wait forever, with deferred frames being queued.
Fix: When creating initial idle iax2_process_threads, wait for init_cond to be signalled
after each thread is started.
Matthew Jordan [Wed, 10 Apr 2013 14:22:45 +0000 (14:22 +0000)]
Use LDAP memory management functions instead of Asterisk's
When MALLOC_DEBUG is enabled with res_config_ldap, issues (munmap_chunk:
invalid pointer errors) can occur as the memory is being allocated with
Asterisk's wrappers around malloc/calloc/free/strdup, as opposed to the
LDAP library's wrappers.
This patch uses the LDAP library's wrappers where appropriate, so that
compiling with MALLOC_DEBUG doesn't cause more problems than it solves.
Note that the patch listed below was modified slightly for this commit
to account for some additional memory allocation/deallocations.
(closes issue ASTERISK-17386)
Reported by: John Covert
Tested by: Andrew Latham
patches:
issue18789-1.8-r316873.patch uploaded by seanbright (License 5060)
Matthew Jordan [Wed, 10 Apr 2013 14:03:05 +0000 (14:03 +0000)]
Fix crash in chan_sip when a core initiated op occurs at the same time as a BYE
When a BYE request is processed in chan_sip, the current SIP dialog is detached
from its associated Asterisk channel structure. The tech_pvt pointer in the
channel object is set to NULL, and the dialog persists for an RFC mandated
period of time to handle re-transmits.
While this process occurs, the channel is locked (which is good).
Unfortunately, operations that are initiated externally have no way of knowing
that the channel they've just obtained (which is still valid) and that they are
attempting to lock is about to have its tech_pvt pointer removed. By the time
they obtain the channel lock and call the channel technology callback, the
tech_pvt is NULL.
This patch adds a few checks to some channel callbacks that make sure the
tech_pvt isn't NULL before using it. Prime offenders were the DTMF digit
callbacks, which would crash if AMI initiated a DTMF on the channel at the
same time as a BYE was received from the UA. This patch also adds checks on
sip_transfer (as AMI can also cause a callback into this function), as well
as sip_indicate (as lots of things can queue an indication onto a channel).
Review: https://reviewboard.asterisk.org/r/2434/
(closes issue ASTERISK-20225)
Reported by: Jeff Hoppe
Rusty Newton [Mon, 8 Apr 2013 23:34:59 +0000 (23:34 +0000)]
Modified the list of keys for the driver backends for sake of sample clarity
Added a line showing the mapping of "mysql" to res_config_mysql available in add-ons. We used "mysql" as an example driver key in the sample, but didn't show what module it mapped too. Also added a subtitle above the list of keys for driver backends.
Fix For Not Overriding The Default Settings In chan_sip
The initial report was that the "nat" setting in the [general] section was not
having any effect in overriding the default setting. Upon confirming that this
was happening and looking into what was causing this, it was discovered that
other default settings would not be overriden as well.
This patch works similar to what occurs in build_peer(). We create a temporary
ast_flags structure and using a mask, we override the default settings with
whatever is set in the [general] section.
In the bug report, the reporter who helped to test this patch noted that the
directmedia settings were being overriden properly as well as the nat settings.
(closes issue ASTERISK-21225)
Reported by: Alexandre Vezina
Tested by: Alexandre Vezina, Michael L. Young
Patches:
asterisk-21225-handle-options-default-prob_1.8_v4.diff.diff
Michael L. Young (license 5026)
In ASTERISK-20904, the focus was around the changes to NAT that took place in
Asterisk 11. Since the report stated that 1.8 was fine, we didn't take a look
at 1.8 at the time.
While working on ASTERISK-21225, I could see that 1.8 would benefit from having
some of those changes applied to it.
This patch does the following:
* The important part of this patch is that it sets the peer's flags earlier in
build_peer so that the code properly uses the peer's flags based on the peer's
configuration.
* constify req parameter in check_via()
* update realtime schemas under the contrib directory to handle properly the NAT
settings available in 1.8 as well as to handle the changes made in 11 to make
upgrading easier when installing newer versions of Asterisk
(closes issue ASTERISK-21243)
Reported by: Michael L. Young
Patches:
asterisk-20904-changes_for_1.8.diff Michael L. Young (license 5026)
The new inband_on_proceeding option causes Asterisk to assume inband audio
may be present when a PROCEEDING message is received.
Q.931 Section 5.1.2 says the network cannot assume that the CPE side has
attached to the B channel at this time without explicitly sending the
progress indicator ie informing the CPE side to attach to the B channel
for audio. However, some non-compliant ISDN switches send a PROCEEDING
without the progress indicator ie indicating inband audio is available and
assume that the CPE device has connected the media path for listening to
ringback and other messages.
ASTERISK-17834 which causes this issue was dealing with a non-compliant
network switch.
David M. Lee [Tue, 2 Apr 2013 17:33:53 +0000 (17:33 +0000)]
Fixed spurious rebuilds of func_version.
func_version.so was being rebuilt every time, because build.h was
changing every build, because of the cleantest dependency that was
added in r384410 to fix parallel make bugs.
Now build.h will only be created if it does not exist, which was the
original behavior of the Makefile.
David M. Lee [Mon, 1 Apr 2013 13:18:15 +0000 (13:18 +0000)]
Fix parallel make problems.
Occasionally, make -j would fail due to missing includes, or other
unusual errors.
This was due to the 'cleantest' target, which was designed to force a
make clean when some change in the code would cause the typical
depedency checking to fail. Several targets in the main Makefile did
not depend upon cleantest, hence would run in parallel to it. By
adding the dependency, make -j runs happily now.
Jonathan Rose [Fri, 29 Mar 2013 16:23:03 +0000 (16:23 +0000)]
app_voicemail: Add blank argument to externnotify if no context argument
At least one call to run_externnotify provides a NULL context parameter and
because the snprintf statement doesn't account for a NULL context parameter,
it simply writes '(null)' to the arguments string instead. This patch makes
it write two quotes back to back for that argument instead in the event of
a NULL context.
(closes issue ASTERISK-18207)
Reported by: Barry L. Kline
Patches:
modified from patch-20130306 uploaded by Karsten Wemheuer (License 5930)
Matthew Jordan [Wed, 27 Mar 2013 18:49:17 +0000 (18:49 +0000)]
Fix a file descriptor leak in off nominal path
While looking at the security vulnerability in ASTERISK-20967, Walter noticed
a file descriptor leak and some other issues in off nominal code paths. This
patch corrects them.
Note that this patch is not related to the vulnerability in ASTERISK-20967,
but the patch was placed on that issue.
Kinsey Moore [Wed, 27 Mar 2013 17:02:32 +0000 (17:02 +0000)]
Fix white noise on SRTP decryption
When res_rtp_asterisk.c was altered to avoid attempting to apply
unprotect algorithms to non-audio RTP packets, the test used was
incorrect. This caused the audio packets to not be decrypted and
resulted in loud white noise on the other endpoint (or both endpoints
depending on the call legs involved). The test now properly checks the
version field in the RTP header to ensure that RTP and RTCP are
decrypted while other types of packets are not.
(closes issue ASTERISK-21323)
Reported by: andrea
Tested by: Kinsey Moore, andrea, John Bigelow
Patches:
whitenoise_fix.diff uploaded by Kinsey Moore
Matthew Jordan [Wed, 27 Mar 2013 14:53:13 +0000 (14:53 +0000)]
AST-2013-003: Prevent username disclosure in SIP channel driver
When authenticating a SIP request with alwaysauthreject enabled, allowguest
disabled, and autocreatepeer disabled, Asterisk discloses whether a user
exists for INVITE, SUBSCRIBE, and REGISTER transactions in multiple ways. The
information is disclosed when:
* A "407 Proxy Authentication Required" response is sent instead of a
"401 Unauthorized" response
* The presence or absence of additional tags occurs at the end of "403
Forbidden" (such as "(Bad Auth)")
* A "401 Unauthorized" response is sent instead of "403 Forbidden" response
after a retransmission
* Retransmission are sent when a matching peer did not exist, but not when a
matching peer did exist.
This patch resolves these various vectors by ensuring that the responses sent
in all scenarios is the same, regardless of the presence of a matching peer.
This issue was reported by Walter Doekes, OSSO B.V. A substantial portion of
the testing and the solution to this problem was done by Walter as well - a
huge thanks to his tireless efforts in finding all the ways in which this
setting didn't work, providing automated tests, and working with Kinsey on
getting this fixed.
Matthew Jordan [Wed, 27 Mar 2013 14:35:11 +0000 (14:35 +0000)]
AST-2013-002: Prevent denial of service in HTTP server
AST-2012-014, fixed in January of this year, contained a fix for Asterisk's
HTTP server for a remotely-triggered crash. While the fix put in place fixed
the possibility for the crash to be triggered, a denial of service vector still
exists with that solution if an attacker sends one or more HTTP POST requests
with very large Content-Length values. This patch resolves this by capping
the Content-Length at 1024 bytes. Any attempt to send an HTTP POST with
Content-Length greater than this cap will not result in any memory allocation.
The POST will be responded to with an HTTP 413 "Request Entity Too Large"
response.
This issue was reported by Christoph Hebeisen of TELUS Security Labs
(closes issue ASTERISK-20967)
Reported by: Christoph Hebeisen
patches:
AST-2013-002-1.8.diff uploaded by mmichelson (License 5049)
AST-2013-002-10.diff uploaded by mmichelson (License 5049)
AST-2013-002-11.diff uploaded by mmichelson (License 5049)
Matthew Jordan [Tue, 26 Mar 2013 02:23:01 +0000 (02:23 +0000)]
Resolve deadlock between SIP registration and channel based functions
In r373424, several reentrancy problems in chan_sip were addressed. As a
result, the SIP channel driver is now properly locking the channel driver
private information in certain operations that it wasn't previously. This
exposed two latent problems either in register_verify or by functions called
by register_verify. This includes:
* Holding the private lock while calling sip_send_mwi_to_peer. This can create
a new sip_pvt via sip_alloc, which will obtain the channel container lock.
This is a locking inversion, as any channel related lock must be obtained
prior to obtaining the SIP channel technology private lock.
* Holding the privat elock while calling sip_poke_peer. In the same vein as
sip_send_mwi_to_peer, sip_poke_peer can create a new SIP private, causing
the same locking inversion.
Note that this locking inversion typically occured when CLI commands were run
while a SIP REGISTER request was being processed, as many CLI commands (such
as 'sip show channels', 'core show channels', etc.) have to obtain the channel
container lock.
(issue ASTERISK-21068)
Reported by: Nicolas Bouliane
(issue ASTERISK-20550)
Reported by: David Brillert
Matthew Jordan [Tue, 26 Mar 2013 01:50:28 +0000 (01:50 +0000)]
Resolve deadlock between pending CDR and batch CDR locks
r375757 attempted to resolve a race condition between multiple submissions of
CDRs while in batch mode from attempting to destroy the scheduled batch
submission by extending the batch CDR lock. Unfortunately, this causes a
deadlock between the pending CDR lock and the batch CDR lock. This patch
resolves the intent of r375757 by simply providing a new lock that protects
the scheduling of the batches. The original batch CDR lock is kept to protect
manipulation of the batch CDR settings, but has been placed such that it
is not held when the pending lock is held.
Thanks to Chase Venters for providing lock analysis on the issue.
Russell Bryant [Tue, 26 Mar 2013 01:32:27 +0000 (01:32 +0000)]
Fix multi-station answer race condition.
When an SLA trunk is ringing (inbound call on the trunk) Asterisk will
make outbound calls to the stations that have that trunk. If more than
one station answers the call at the same time, all channels other than
the first one to answer are left in a bad state. The channel gets
leaked, is not connected to anything, and there's no way to get rid of
it.
We now properly clean up these losing channels by hanging up on them.
Since they lost the race, as we process their answer, there is no
ringing trunk for them to answer.
Richard Mudgett [Mon, 25 Mar 2013 23:19:06 +0000 (23:19 +0000)]
Set the CALLERID(dnid-num-plan) for incoming ISDN calls.
The CALLEDTON channel variable is set for incoming ISDN calls to the lower
7 bits of the Q.931 type-of-number/numbering-plan octet. The
CALLERID(dnid-num-plan) should have the same value.
Kinsey Moore [Fri, 15 Mar 2013 12:49:59 +0000 (12:49 +0000)]
tcptls: Prevent unsupported options from being set
AMI, HTTP, and chan_sip all support TLS in some way, but none of them
support all the options that Asterisk's TLS core is capable of
interpreting. This prevents consumers of the TLS/SSL layer from setting
TLS/SSL options that they do not support.
This also gets tlsverifyclient closer to a working state by requesting
the client certificate when tlsverifyclient is set. Currently, there is
no consumer of main/tcptls.c in Asterisk that supports this feature and
so it can not be properly tested.
Review: https://reviewboard.asterisk.org/r/2370/ Reported-by: John Bigelow Patch-by: Kinsey Moore
(closes issue AST-1093)
Matthew Jordan [Fri, 15 Mar 2013 01:32:06 +0000 (01:32 +0000)]
When a session timer expires during a T.38 call, re-invite with correct SDP
When a session timer expires during a dialog that has re-negotiated to T.38
and Asterisk is the refresher, Asterisk will send a re-INVITE with an SDP
containing audio media only. This causes some hilarity with the poor fax
session under weigh.
This patch corrects that by sending T.38 parameters if we are in the middle of
a T.38 session.
Matthew Jordan [Fri, 15 Mar 2013 01:21:55 +0000 (01:21 +0000)]
Fix processing of call files when using KQueue on OS X
In certain situations, call files are not processed when using KQueue with
pbx_spool. Asterisk was sending an invalid timeout value when the spool
directory is empty, causing the call to kevent to error immediately. This
can create a tight loop, increasing the CPU load on the system.
Michael L. Young [Tue, 12 Mar 2013 21:15:52 +0000 (21:15 +0000)]
Fix Sorting Order For Parking Lots Stored In Static Realtime
When retrieving the parking lots from a MySQL database table, the current order
is "filename, cat_metric desc, var_metric asc, category". If there are multiple
parking lots with the same cat_metric but different categories, everything is
being sorted on cat_metric first resulting in errors when loading the parking
lots.
This patch fixes the problem by sorting on the category field first, then the
cat_metric field.
(closes issue ASTERISK-21035)
Reported by: Alex Epshteyn
Patches:
asterisk-21035-orderby.diff Michael L. Young (license 5026)
Matthew Jordan [Tue, 12 Mar 2013 16:20:30 +0000 (16:20 +0000)]
Include the Username field in SIP Registry events when Status is registered
In ASTERISK-17888, the AMI Registry event during SIP registrations was supposed
to include the Username field. Somehow, one of the events was missed. This
patch corrects that - the Username field should be included in all AMI Registry
events involving SIP registrations.
Matthew Jordan [Thu, 28 Feb 2013 17:09:25 +0000 (17:09 +0000)]
Prevent deadlock in chan_iax2 when attempting to set caller ID
A deadlock can occur in chan_iax2 when it attempts to set the caller ID, as it
already holds the iax2 private lock and improperly fails to obtain the channel
lock before calling ast_set_callerid. By not safely obtaining the channel lock,
a locking inversion can take place, causing a deadlock.
This patch solves this by calling the required deadlock avoidance functions
that obtain the channel lock before setting the caller ID.
Thanks to Pavel for fixing my syntax errors and testing this patch out.
(closes issue ASTERISK-21128)
Reported by: Pavel Troller
Tested by: Pavel Troller
patches:
ASTERISK-21128-1.8.diff uploaded by mjordan (license 6283)
ASTERISK-21128-modified-1.8.diff uploaded by Pavel Troller (license 6302)
Matthew Jordan [Thu, 28 Feb 2013 16:40:31 +0000 (16:40 +0000)]
Let channels joining a MeetMe conference opt out of the denoiser
For some channel drivers, specifically those that have a varying rate in the
number of audio samples, the audio quality for a MeetMe conference can be
exceedingly poor. This is due to a unilateral application of the DENOISE
function in func_speex to channels joining the conference.
The denoiser function in the speex library is initialized with the number of
audio samples in each sample that will be provided to it. If the number of
audio samples changes, the denoiser has to be thrown away and re-initialized.
While this could be worked around by removing func_speex, that doesn't help
if you actually use the denoiser with other channels on the system.
This patches does the following:
* Checks for the presence of func_speex as opposed to codec_speex when
determining if the DENOISE function is present (which is where the function
is actually implemented)
* Adds an option to MeetMe 'n' that causes the denoiser to not be applied
to a channel when it joins. This keeps the current behavior the default, but
let's users disable the denoiser if it causes problems on their system.
Review: https://reviewboard.asterisk.org/r/2358
(closes issue AST-1062)
Reported by: Thomas Arimont
Tzafrir Cohen [Tue, 26 Feb 2013 19:37:33 +0000 (19:37 +0000)]
Consider linux-gnuspe as linux-gnu
* The powerpcspe Linux port uses linux-gnuspe as the OS string.
* Our build system shouldn't really care for that, so just call it linux-gnu.
* Original report: Roland Stigge , http://bugs.debian.org/701505
Matthew Jordan [Mon, 25 Feb 2013 12:48:52 +0000 (12:48 +0000)]
Clean up use of va_end/va_args in res_config_mysql
There were several problems using variadic argument macros in res_config_mysql.
* Improper use of va_end. Multiple calls to va_end were possible resulting in
an unbalanced matching of va_start/va_end.
* Calls to va_arg after a possible encounter of a SENTINEL value.
Matthew Jordan [Sun, 24 Feb 2013 22:59:11 +0000 (22:59 +0000)]
Set the sin_family on the bind address socket during initialization
Somehow, chan_jingle has managed to operate for years without setting the
sin_family on its bindaddr socket. This patch properly sets the field during
initial module load to AF_INET.
Note that the patch on the issue was modified slightly to change the
initialization of the socket from allocation of a chan_jingle private to the
module initialization, as the bindaddr object (which is static) only needs to
have the address set once.
(closes issue ASTERISK-19341)
Reported by: andre valentin
patches:
0105-chan_jingle.patch uploaded by avalentin (License 6064)
Matthew Jordan [Sun, 24 Feb 2013 16:25:08 +0000 (16:25 +0000)]
Don't display the AMI ALL class authorization for users if they don't have it
When converting AMI class authorizations to a string representation, the
method always appends the ALL class authorization. This is especially
important for events, as they should always communicate that class
authorization - even if the event itself does not specify ALL as a class
authorization for itself. (Events have always assumed that the ALL class
authorization is implied when they are raised)
Unfortunately, this did mean that specifying a user with restricted class
authorizations would show up in the 'manager show user' CLI command as
having the ALL class authorization.
Rather then modifying the existing string manipulation function, this patch
adds a function that will only return a string if the field being compared
explicitly matches class authorization field it is being compared against.
This prevents ALL from being returned unless it is actually specified for
the user.
(closes issue ASTERISK-20397)
Reported by: Johan Wilfer
Matthew Jordan [Sun, 24 Feb 2013 15:33:32 +0000 (15:33 +0000)]
Make ParkAndAnnounce return to priority + 1 when return context is not defined
The ParkAndAnnounce application documentation for the optional return_context
parameter states the following:
return_context
The goto-style label to jump the call back into after timeout. Default
'priority+1'.
Unfortunately, the application was sending the channel back into the dialplan
at 'priority', which is the ParkAndAnnounce application call. This causes an
infinite loop of the channel constantly being parked, announced, timed out,
parked, announced, timed out... while fun, especially for those callers you
wish to drive to the end of madness, this was not the intent of the
application.
Matthew Jordan [Thu, 21 Feb 2013 22:44:59 +0000 (22:44 +0000)]
Properly detect launchd
Asterisk was a little too pro-active in claiming that it found launchd. On
systems without launchd - such as FreeBSD - this resulted in certain items
in Asterisk that conflict with launchd to not be selectable, such as
res_timing_kqueue.
Kevin Harwell [Tue, 19 Feb 2013 19:16:44 +0000 (19:16 +0000)]
Write the correct callid to the data1 field in queue_log for transfer events.
The incorrect callid was being written to the "data1" field in queue_log table
for transfer events. The callid of the queue was being written instead of the
transfer target's callid. This now gets the correct "transfer to" number and
places that in the "data1" field of the queue_log table when a transfer event
is triggered.
(closes issue ASTERISK-19960)
Reported by: vladimir shmagin
projects / thirdparty / asterisk.git / log
