Terry Wilson [Tue, 3 Jul 2012 16:58:16 +0000 (16:58 +0000)]
More improvements to re-INVITEs timing out after a provisional response
There is no need to call check_pendings() on a final response to an INVITE
when destroying the scheduler entry as it will be done later during normal
processing.
Terry Wilson [Tue, 3 Jul 2012 14:27:02 +0000 (14:27 +0000)]
Better handle re-INVITEs with provisional but no final repsonses
A previous attempt at fixing this issue had negative side effects related
to attended transfers which this patch should resolve. Many thanks to
Steve Davies for all of the good suggestions and testing.
(closes issue ASTERISK-19992)
Reported by: Steve Davies
Tested by: Steve Davies, Terry Wilson
Review: https://reviewboard.asterisk.org/r/2009/
Terry Wilson [Wed, 27 Jun 2012 20:58:51 +0000 (20:58 +0000)]
AST-2012-010: Clean up after a reinvite that never gets a final response
The basic problem is that if a re-INVITE is sent by Asterisk and it receives a
provisional response, but no final response, then the dialog is never torn
down. In addition to leaking memory, this also leaks file descriptors and will
eventually lead to Asterisk no longer being able to process calls.
This patch just keeps track of whether there is an outstanding re-INVITE, and if
there is goes ahead and cleans up everything as though there was no outstanding
reinvite.
Review: https://reviewboard.asterisk.org/r/2009/
(closes issue ASTERISK-19992)
Reported by: Steve Davies
Tested by: Steve Davies, Terry Wilson
Matthew Jordan [Tue, 26 Jun 2012 13:21:13 +0000 (13:21 +0000)]
Fix crash in unloading of res_adsi module
When res_adsi is unloaded, it removes the ADSI functions that it previously installed
by passing a NULL adsi_funcs pointer to ast_adsi_install_funcs. This function was not
checking whether or not the adsi_funcs pointer passed in was NULL before dereferencing
it to check whether or not the version of the functions matches what the core was
expecting it.
This patch makes it so that the version is only checked if a potentially valid adsi_funcs
pointer was passed in. Passing in NULL removes the installed functions, bypassing the
version check.
Matthew Jordan [Mon, 25 Jun 2012 19:24:55 +0000 (19:24 +0000)]
Tweak CDR change in r369351
As Tilghman pointed out on review 1996, the check to see if a CDR end time has
been set is sufficient to know whether or not the duration value can be used.
The check-in done for r369351 forgot to include this change.
Mark Michelson [Mon, 25 Jun 2012 19:13:31 +0000 (19:13 +0000)]
Re-fix how local tag is generated when sending a 481 to an INVITE.
Match our local tag to whatever to-tag was sent in the initial INVITE.
Because the size of the to-tag may not fit in the buffer in the sip_pvt,
it has been changed to a string field.
(closes issue ASTERISK-19892)
reported by Walter Doekes
Matthew Jordan [Mon, 25 Jun 2012 19:12:35 +0000 (19:12 +0000)]
Fix incorrect duration reporting in CDRs created in batch mode
Certain places in core/cdr.c would, if the duration value were 0, calculate the
duration as being the delta between the current time and the time at which the
CDR record was started. While this does not typically cause a problem in
non-batch mode, this can cause an issue in batch mode where CDR records are
gathered and written long after those calls have ended. In particular, this
affects calls that were never answered, as those are expected to have a duration
of 0. Often, this would result in CDR logs with a significant number of calls
with lengthy durations, but dispositions of "BUSY".
Note that this does not affect cdr_csv, as that backend does not use
ast_cdr_getvar and instead directly reports the duration value. The affected
core backends include cdr_apative_odbc and cdr_custom; other extended or
deprecated CDR backends may potentially still directly manipulate the duration
values.
(issue ASTERISK-19860)
Reported by: Thomas Arimont
(issue AST-883)
Reported by: Thomas Arimont
Tested by: Matt Jordan
Richard Mudgett [Mon, 25 Jun 2012 15:57:28 +0000 (15:57 +0000)]
Fix Bridge application occasionally returning to the wrong location.
* Fix do_bridge_masquerade() getting the resume location from the zombie
channel. The code must not touch a clone channel after it has masqueraded
it. The clone channel has become a zombie and is starting to hangup.
Mark Michelson [Mon, 25 Jun 2012 14:18:09 +0000 (14:18 +0000)]
Be more consistent with the return code for requests received from invalid domain.
When Asterisk receives an INVITE from an external domain when allowexternaldomains=no
send a 403 instead of a 404. This is consistent with Asterisk's behavior when receiving
a REGISTER in this situation.
(Closes issue ASTERISK-19601)
Reported by Matthew Jordan
Patches:
ASTERISK-19601-no401.patch uploaded by Mark Michelson (License #5049)
Terry Wilson [Fri, 22 Jun 2012 19:28:04 +0000 (19:28 +0000)]
Don't crash on a guest directmedia call
A sip_pvt may not have relatedpeer set if a call doesn't match up
with a peer. If there is no relatedpeer, there is no direct media
ACL to apply, so just return that it is allowed.
(closes issue ASTERISK-20040)
Reported by: Terry Wilson
Kinsey Moore [Fri, 22 Jun 2012 17:14:10 +0000 (17:14 +0000)]
Don't parse media stream state for SIP video streams
The sendonly/recvonly/sendrecv/inactive media stream attributes were
parsed for video, but nothing was ever done with them. With this code
removed, an UNSUPPORTED message is produced when these attributes are
used in conjunction with a video stream which is the better behavior
since they were never really supported in the first place.
Michael L. Young [Wed, 20 Jun 2012 02:03:22 +0000 (02:03 +0000)]
Fix NULL pointer segfault in ast_sockaddr_parse()
While working with ast_parse_arg() to perform a validity check, a segfault
occurred. The segfault occurred due to passing a NULL pointer to
ast_sockaddr_parse() from ast_parse_arg(). According to the documentation in
config.h, "result pointer to the result. NULL is valid here, and can be used to
perform only the validity checks."
This patch fixes the segfault by checking for a NULL pointer. This patch also
adds documentation to netsock2.h about why it is necessary to check for a NULL
pointer.
(Closes issue ASTERISK-20006)
Reported by: Michael L. Young
Tested by: Michael L. Young
Patches:
asterisk-20006-netsock-null-ptr.diff uploaded by Michael L. Young (license 5026)
Mark Michelson [Tue, 19 Jun 2012 15:30:58 +0000 (15:30 +0000)]
Fix request routing issue when outboundproxy is used.
Asterisk was incorrectly setting the destination of CANCELs
and ACKs for error responses to the URI of the initial INVITE.
This resulted in further requests, such as INVITEs with authentication
credentials, to be routed incorrectly. Instead, when these CANCEL
or ACKs are to be sent, we should simply keep the destination the
same as what it previously was. There is no need to alter it any.
(closes issue ASTERISK-20008)
Reported by Marcus Hunger
Patches:
ASTERISK-20008.patch uploaded by Mark Michelson (license #5049)
Kevin P. Fleming [Fri, 15 Jun 2012 15:56:08 +0000 (15:56 +0000)]
Add support-level indications to many more source files.
Since we now have tools that scan through the source tree looking for files
with specific support levels, we need to ensure that every file that is
a component of a 'core' or 'extended' module (or the main Asterisk binary)
is explicitly marked with its support level. This patch adds support-level
indications to many more source files in tree, but avoids adding them to
third-party libraries that are included in the tree and to source files
that don't end up involved in Asterisk itself.
Mark Michelson [Thu, 14 Jun 2012 15:23:10 +0000 (15:23 +0000)]
Revert Makefile change to remove embedding res_adsi.so
The change has resulted in a linking error for certain versions
of GCC. This is much worse than the original issue, so for now,
temporarily revert the change. A more thorough change will be
sought out.
Mark Michelson [Wed, 13 Jun 2012 20:59:01 +0000 (20:59 +0000)]
Fix a deadlock that occurs when func_volume is used on a local channel.
This was discovered by trying to perform a call forward to an extension
that makes use of func_volume. When the local channel is optimized away,
the datastore on the local;2 channel would have its audiohook destroyed
rather than detaching the audiohook from the channel and then destroying
it.
With this patch, func_volume's datastore destructor takes the proper
route of detaching the audiohook and then destroying it.
(closes issue ASTERISK-19611)
reported by Volker Sauer
Patches:
ASTERISK-19611.patch uploaded by Mark Michelson (license #5049)
Matthew Jordan [Wed, 13 Jun 2012 20:26:07 +0000 (20:26 +0000)]
Mark res_smdi/res_adsi as 'core' supported modules
Recently, various issues surrounding weak symbols have caused problems with
modules that rely on that feature to be enabled in menuselect. This includes
app_voicemail and chan_dahdi, as they both rely upon res_smdi and res_adsi,
which, in certain circumstances, may not be enabled by default in menuselect.
Because res_smdi/res_adsi are dependencies for chan_dahdi/app_voicemail, this
patch marks both as 'core' supported modules. This will allow both
app_voicemail and chan_dahdi to be enabled as well, regardless of whether or
not that system supports weak symbols.
(issue AST-900)
Reported by: Thomas Arimont
(issue AST-885)
Reported by: Denis Alberto Martinez
Matthew Jordan [Wed, 13 Jun 2012 14:27:57 +0000 (14:27 +0000)]
Do not install empty directories; add ASTLIBDIR
r368830 modified the installation script to only create a directory if that
directory does not exist. If some directory variable was empty, it would attempt
to create the empty location. It also failed to create the ASTLIBDIR directory.
This patch fixes it such that the correct directories are made and only created if
a value specifying them actually exists.
Matthew Jordan [Tue, 12 Jun 2012 18:23:01 +0000 (18:23 +0000)]
Do not perform install on existing directories
If a directory already exists, performing a 'make install' will remove the
permissions associated with the current directory and replace them with the
permissions of the user executing the install.
This patch changes this behavior to only perform an install on the directory
if the directory does not exist. Thus, if a user later changes the permissions
on that directory, those permissions will be preserved in subsequent installs.
Review: https://reviewboard.asterisk.org/r/1986
Review: https://reviewboard.asterisk.org/r/1864
(closes issue ASTERISK-19492)
Reported by: Karl Fife
Tested by: Paul Belanger, Tilghman Lesher
patches:
ASTERISK-19492 by pabelanger
(uploaded by mjordan)
Mark Michelson [Tue, 12 Jun 2012 15:36:34 +0000 (15:36 +0000)]
Set the Caller ID "tag" on peers even if remote party information is present.
On incoming calls, we were setting the cid_tag on the dialog only if there was
no remote party information (Remote-Party-ID or P-Asserted-Identity) present.
The Caller ID tag is an invented parameter, though, and should be set no matter
the circumstance.
(closes issue ASTERISK-19859)
Reported by Thomas Arimont
(closes issue AST-884)
Reported by Trey Blancher
Richard Mudgett [Wed, 6 Jun 2012 21:27:33 +0000 (21:27 +0000)]
Fix POTS flash hook to orignate a second call deadlock.
A deadlock can occur when a POTS phone tries to flash hook to originate a
second call for 3-way or transfer. If another process is scanning the
channels container when the POTS line flash hooks then a deadlock will
occur.
* Release the channel and private locks when creating a new channel as a
result of a flash hook.
Mark Michelson [Wed, 6 Jun 2012 19:13:45 +0000 (19:13 +0000)]
Fix a specific scenario where ACKs are not matched.
If a dialog-starting INVITE contains a to-tag, then Asterisk
will respond with a 481. In this case, the resulting incoming
ACK would not be matched, so Asterisk would continue retransmitting
the 481 until the transaction times out.
There were two issues. Asterisk, upon creating a sip_pvt would generate
a local tag. However, when the time came to transmit the 481, since there
was a to-tag in the INVITE, Asterisk would place this original to-tag
in the 481 response. When the ACK came in, Asterisk would attempt to
match the to-tag in the ACK to the generated local tag. Unfortunately,
Asterisk never actually transmitted a response with the generated local
tag, so the to-tag in the ACK would not match.
The other problem was that when the 481 was sent, nothing was set
on the sip_pvt to indicate what CSeq is expected in the ACK.
To fix the first problem, we zero out the to-tag seen in the incoming
INVITE. This way, Asterisk, when time to send a response, will send
its generated local tag instead.
To fix the second problem, we set the sip_pvt's pendinginvite to the
CSeq of the INVITE when we send a 481.
(closes issue ASTERISK-19892)
Reported by Mark Michelson
Matthew Jordan [Wed, 6 Jun 2012 17:20:07 +0000 (17:20 +0000)]
Add feature modifier to versions produced from branches
Certain branches, such as Certified Asterisk, may have a modifier added to
them that specifies the features available in that branch. For branches, this
modifier is expected to be reflected in the location of the branch in
subversion. For example, a subversion of URL of /certified/branches/1.8.11
would have a feature modifier of 'certified'. This is slightly different then
how features are determined for tags, where the feature is part of the actual
tag name, e.g., "10.5.0-digiumphones".
In keeping with the nomenclature used for tags, the feature specifier for
branches is translated and placed after the revision numbers. For the example
given previously, this would result in a branch version of
"Asterisk SVN-branch-1.8.11-cert-rXXXXXX".
Richard Mudgett [Wed, 6 Jun 2012 01:08:29 +0000 (01:08 +0000)]
Fix parked call performing a DTMF blind transfer after being retrieved.
When a parked call was retrieved from the parking lot, it could not do a
blind transfer because it caused the involved calls to be hung up
unconditionally.
* Made the ParkedCall application return the ast_bridge_call() return
value.
Kinsey Moore [Tue, 5 Jun 2012 15:15:57 +0000 (15:15 +0000)]
Ensure that pages and emails are sent using RFC822-compliant date format
When localization was added to app_voicemail, these headers were altered
when they should have remained in en_US format for RFC compliance. This
reverts the changes to those two lines.
Mark Michelson [Mon, 4 Jun 2012 21:56:05 +0000 (21:56 +0000)]
Relay proper SIP responses on calling side.
Revision 351130 broke corect HANGUPCAUSE setting
for the 404 case in chan_sip. Other cases were also
potentially broken. This patch fixes the relaying
of causes to be what they used to be.
(closes issue ASTERISK-19914)
Reported by Pavel Troller
Tested by Walter Doekes (via a reviewboard test to be committed later)
Patches:
chan_sip.diff uploaded by Pavel Troller (license #6302)
Richard Mudgett [Mon, 4 Jun 2012 18:41:18 +0000 (18:41 +0000)]
Fix potential deadlock between masquerade and chan_local.
* Restructure ast_do_masquerade() to not hold channel locks while it calls
ast_indicate().
* Simplify many calls to ast_do_masquerade() since it will never return a
failure now. If it does fail internally because a channel driver callback
operation failed, the only thing ast_do_masquerade() can do is generate a
warning message about strange things may happen and press on.
* Fixed the call to ast_bridged_channel() in ast_do_masquerade(). This
change fixes half of the deadlock reported in ASTERISK-19801 between
masquerades and chan_iax.
* 'Unsupported media type' is only reported when that is in fact the case,
not when a supported media type is included in an 'm' line that has an
invalid format.
* All warning messages related to parsing 'm' lines now include the 'm' line contents.
* (minor bugfix) newline added to port-number-zero warning messages.
* Warning messages improved to use RFC-specified terminology for various items.
* Warnings for offers that include more than one port for a single media type now
include the media type.
Add documentation to function CHANNEL for options echocan_mode and buffers
The ability to set "echocan_mode" and "buffers" through the dialplan was added
to chan_dahdi some time ago. This patch adds some documentation to
func_channel.
(Closes issue ASTERISK-19911)
Reported by: Dale Noll
Tested by: Michael L. Young
Patches:
asterisk-19911-branch18.diff uploaded by Michael L. Young (license 5026)
Richard Mudgett [Wed, 30 May 2012 17:21:43 +0000 (17:21 +0000)]
Fix deadlock when executing CLI "pri show channels" and "ss7 show channels" commands.
* Fix sig_pri_lock_owner() to avoid deadlock properly.
* Code pri_grab() better.
* Fix sig_ss7_lock_owner() to avoid deadlock properly.
* Code ss7_grab() better.
(closes issue ASTERISK-19854)
Reported by: Jaxon
Patches:
jira_asterisk_19854_v1.8.6.patch (license #5621) patch uploaded by rmudgett (Modified to do the same thing to sig_ss7)
Tested by: Jaxon
Richard Mudgett [Tue, 29 May 2012 22:25:21 +0000 (22:25 +0000)]
Coverity Report: Fix issues for error type REVERSE_INULL (deprecated modules)
* Fix only issue pointed out by deprecated_REVERSE_INULL.txt for
app_meetme.c in find_user().
* Change use of %i to %d in sscanf() in find_user(). The use of %i gives
unexpected parsing because it can accept hex, octal, and decimal integer
formats.
* Changed other uses of %i in app_meetme() to use %d for consistency.
Matthew Jordan [Tue, 29 May 2012 18:30:25 +0000 (18:30 +0000)]
AST-2012-008: Fix remote crash vulnerability in chan_skinny
When a skinny session is unregistered, the corresponding device pointer is set
to NULL in the channel private data. If the client was not in the on-hook state
at the time the connection was closed, the device pointer can later be
dereferenced if a message or channel event attempts to use a line's pointer to
said device.
The patches prevent this from occurring by checking the line's pointer in
message handlers and channel callbacks that can fire after an unregistration
attempt.
(closes issue ASTERISK-19905)
Reported by: Christoph Hebeisen
Tested by: mjordan, Damien Wedhorn
Patches:
AST-2012-008-1.8.diff uploaded by mjordan (license 6283)
AST-2012-008-10.diff uploaded by mjordan (license 6283)
Richard Mudgett [Fri, 25 May 2012 16:28:04 +0000 (16:28 +0000)]
AST-2012-007: Fix IAX receiving HOLD without suggested MOH class crash.
* Made schedule_delivery() set the received frame f->data.ptr to NULL if
the datalen is zero.
* Fix queue_signalling() memcpy() size error.
* Made queue_signalling() not use C++ keyword variable names.
(closes issue ASTERISK-19597)
Reported by: mgrobecker
Patches:
jira_asterisk_19597_v1.8.patch (license #5621) patch uploaded by rmudgett
Tested by: rmudgett, Michael L. Young
Michael L. Young [Fri, 25 May 2012 02:27:11 +0000 (02:27 +0000)]
Fix pvt_sip for inbound call to use peer's allowtransfer setting
The pvt_sip allowtransfer was not being set to that of the peer's setting.
Therefore, the global allowtransfer setting was being used instead which would
lead to calls not being transfered if the global setting was set to 'no' despite
the setting on the peer being 'yes' and vice versa, calls would be allowed to
transfer even if the peer's setting was 'no' but the global setting was 'yes'.
(Closes issue ASTERISK-19856)
Reported by: Jacek
Tested by: Michael L. Young, Jacek
Patches:
issue-asterisk-19856-branch10-v3.diff uploaded by
Michael L. Young (license 5026)
Richard Mudgett [Thu, 24 May 2012 22:21:18 +0000 (22:21 +0000)]
Fix Dial I option ignored if dial forked and one fork redirects.
The Dial and Queue I option is intended to block connected line updates
and redirecting updates. However, it is a feature that when a call is
locally redirected, the I option is disabled if the redirected call runs
as a local channel so the administrator can have an opportunity to setup
new connected line information. Unfortunately, the Dial and Queue I
option is disabled for *all* forked calls if one of those calls is
redirected.
* Make the Dial and Queue I option apply to each outgoing call leg
independently. Now if one outgoing call leg is locally redirected, the
other outgoing calls are not affected.
* Made Dial not pass any redirecting updates when forking calls.
Redirecting updates do not make sense for this scenario.
* Made Queue not pass any redirecting updates when using the ringall
strategy. Redirecting updates do not make sense for this scenario.
* Fixed deadlock potential with chan_local when Dial and Queue send
redirecting updates for a local redirect.
* Converted the Queue stillgoing flag to a boolean bitfield.
The AST_CONTROL_HOLD MOH class from the WaitExten application can now be
queued onto a channel, passed over local channels with the /m option, and
passed over IAX channels.
Matthew Jordan [Wed, 23 May 2012 13:06:08 +0000 (13:06 +0000)]
Update a peer's LastMsgsSent when the peer is notified of waiting messages
Previously, MWI logic utilized a counter called 'lastmsgssent' to know whether
or not MWI NOTIFY requests had been sent to a specific peer. When MWI
notifications were changed to use the internal event framework, this value was
no longer needed for its original purpose. Hence, it was no longer updated
with the new/old message counts for a peer. However, the value was still
presented when, either by AMI or CLI, a 'sip show peer [peer]' command
was executed. The output of the command would always display the erroneous
value of 32767/65535 for 'LastMsgsSent'.
This patch makes it so that the value of lastmsgssent is updated appropriately.
The value should now display the new/old message counts for a particular
peer.
(closes issue ASTERISK-17866)
Reported by: Steve Davies
patches by:
ast-17866-rb1272.patch (License #5041 by irroot)
Modified slightly for this commit
Terry Wilson [Tue, 22 May 2012 17:14:20 +0000 (17:14 +0000)]
Fix race condition for CEL LINKEDID_END event
This patch fixes to situations that could cause the CEL LINKEDID_END event to
be missed.
1) During a core stop gracefully, modules are unloaded when ast_active_channels
== 0. The LINKDEDID_END event fires during the channel destructor. This means
that occasionally, the cel_* module will be unloaded before the channel is
destroyed. It seemed generally useful to wait until the refcount of all
channels == 0 before unloading, so I added a channel counter and used it in the
shutdown code.
2) During a masquerade, ast_channel_change_linkedid is called. It calls
ast_cel_check_retire_linkedid which unrefs the linkedid in the linkedids
container in cel.c. It didn't ref the new linkedid. Now it does.
Terry Wilson [Tue, 22 May 2012 16:14:16 +0000 (16:14 +0000)]
Resolve crash in subscribing for MWI notifications
ASTOBJ_UNREF sets the variable to NULL after unreffing it, so the variable
should definitely not be used after that. To solve this in the two cases
that affect subscribing for MWI notifications, we instead save the ref
locally, and unref them in the error conditions.
(closes issue ASTERISK-19827)
Reported by: B. R
Review: https://reviewboard.asterisk.org/r/1940/
Mark Michelson [Fri, 18 May 2012 16:53:47 +0000 (16:53 +0000)]
Fix memory leak of SSL_CTX structures in TLS core.
SSL_CTX structures were allocated but never freed. This was a bigger
issue for clients than servers since new SSL_CTX structures could be
allocated for each connection. Servers, on the other hand, typically
set up a single SSL_CTX for their lifetime.
This is solved in two ways:
1. In __ssl_setup(), if a tcptls_cfg has an ssl_ctx on it, it is
freed so that a new one can take its place.
2. A companion to ast_ssl_setup() called ast_ssl_teardown() has
been added so that servers can properly free their SSL_CTXs.
Matthew Jordan [Fri, 18 May 2012 15:42:33 +0000 (15:42 +0000)]
Fix more memory leaks
This patch adds to what was fixed in r366880. Specifically, it addresses the
following:
* chan_sip: dispose of an allocated frame in off nominal code paths in
sip_rtp_read
* func_odbc: when disposing of an allocated resultset, ensure that any rows
that were appended to that resultset are also disposed of
* cli: free the created return string buffer in another off nominal code
path
Kinsey Moore [Fri, 18 May 2012 14:16:50 +0000 (14:16 +0000)]
Reorder and renumber tests appropriately
It appears that a patch did not apply properly when adding tests 12 and
13 and test 11 was duplicated. These tests have been reordered and
renumbered such that they make sense.
Matthew Jordan [Fri, 18 May 2012 13:58:23 +0000 (13:58 +0000)]
Fix a variety of memory leaks
This patch addresses a number of memory leaks in a variety of modules that were
found by a static analysis tool. A brief summary of the changes:
* app_minivm: free ast_str objects on off nominal paths
* app_page: free the ast_dial object if the requested channel technology
cannot be appended to the dialing structure
* app_queue: if a penalty rule failed to match any existing rule list
names, the created rule would not be inserted and its memory
would be leaked
* app_read: dispose of the created silence detector in the presence of
off nominal circumstances
* app_voicemail: dispose of an allocated unique ID field for MWI event
un-subscribe requests in off nominal paths; dispose of
configuration objects when using the secret.conf option
* chan_dahdi: dispose of the allocated frame produced by ast_dsp_process
* chan_iax2: properly unref peer in CLI command "iax2 unregister"
* chan_sip: dispose of the allocated frame produced by sip_rtp_read's
call of ast_dsp_process; free memory in parse unit tests
* func_dialgroup: properly deref ao2 object grhead in nominal path of
dialgroup_read
* func_odbc: free resultset in off nominal paths of odbc_read
* cli: free match_list in off nominal paths of CLI match completion
* config: free comment_buffer/list_buffer when configuration file load
is unchanged; free the same buffers any time they were
created and config files were processed
* data: free XML nodes in various places
* enum: free context buffer in off nominal paths
* features: free ast_call_feature in off nominal paths of applicationmap
config processing
* netsock2: users of ast_sockaddr_resolve pass in an ast_sockaddr struct
that is allocated by the method. Failures in
ast_sockaddr_resolve could result in the users of the method
not knowing whether or not the buffer was allocated. The
method will now not allocate the ast_sockaddr struct if it
will return failure.
* pbx: cleanup hash table traversals in off nominal paths; free
ignore pattern buffer if it already exists for the specified
context
* xmldoc: cleanup various nodes when we no longer need them
* main/editline: various cleanup of pointers not being freed before being
assigned to other memory, cleanup along off nominal paths
* menuselect/mxml: cleanup of value buffer for an attribute when that attribute
did not specify a value
* res_calendar*: responses are allocated via the various *_request method
returns and should not be allocated in the various
write_event methods; ensure attendee buffer is freed if no
data exists in the parsed node; ensure that calendar objects
are de-ref'd appropriately
* res_jabber: free buffer in off nominal path
* res_musiconhold: close the DIR* object in off nominal paths
* res_rtp_asterisk: if we run out of ports, close the rtp socket object and free
the rtp object
* res_srtp: if we fail to create the session in libsrtp, destroy the
temporary ast_srtp object
Jonathan Rose [Thu, 17 May 2012 14:40:07 +0000 (14:40 +0000)]
chan_sip: Fix missed locking of opposing pvt for directmedia acl from r366547
It also required deadlock avoidance since two sip_pvts structs needed to be
locked simultaneously. Trunk handles it differently, so this is a 1.8 and 10
patch only.
Matthew Jordan [Thu, 17 May 2012 12:51:36 +0000 (12:51 +0000)]
Fix checking bounds of array index after using it; improper sizeof
This patch fixes two problems pointed out by a static analysis tool.
* In chan_dahdi, when an event is handled the index of the sub channel is first
obtained. In very off nominal cases, the method that determines the index
can return a negative value. In the event handling code, whether or not
the index returned is valid was being checked after that value was used to
index into an array. This patch makes it so the value is checked before
any indexing is done.
* In res_calendar_ews, sizeof was being passed a pointer instead of the struct to
determine the amount of memory to allocate.
(issue ASTERISK-19651)
Reported by: Matt Jordan
(closes issue ASTERISK-19671)
Reported by: Matt Jordan
Jonathan Rose [Tue, 15 May 2012 20:14:05 +0000 (20:14 +0000)]
chan_sip: Check the right channel's host address for directmediapermit/deny
Prior to this patch, when checking the addresses for directmediapermit and
directmediadeny, Asterisk would check the host address of the channel
permit/deny was specified, which differs from the expectations of both
our users and the development team. Instead, directmediapermit/deny now
checks against the address of the channel that the peer with the ACL is
connected to.
Mark Michelson [Mon, 14 May 2012 19:57:42 +0000 (19:57 +0000)]
Fix two more coverity constant expression result findings.
These correspond to findings 0 and 1 in the core findings of
ASTERISK-19649.
After contacting Mark Spencer, he was unsure of what the intent
behind these lines of code were, so they are being axed.
For Asterisk 1.8 and 10, the output of debugging DUNDi frames
will not be changed, but for trunk the "Retry" portion will
be omitted since it does not properly distinguish retransmissions
from initial frames.
(closes issue ASTERISK-19649)
Reported by Matthew Jordan
Mark Michelson [Mon, 14 May 2012 19:10:20 +0000 (19:10 +0000)]
Fix broken reinvite glare scenario.
To make a long story short, reinvite glares were broken
because Asterisk would invert the To and From headers
when ACKing a 491 response.
The reason was because the initreq of the dialog was being
changed to the incoming glared reinvite instead of being
set to the outgoing glared reinvite. This change has three
parts
* In handle_incoming, we never will reject an ACK because it
has a to-tag present, even if we think the request may be out
of dialog.
* In handle_request_invite, we do not change the initreq when
receiving a reinvite to which we will respond with a 491.
* In handle_request_invite, several superflous settings up
pendinginvite have been removed since this is dones automatically
by transmit_response_reliable
Russell Bryant [Fri, 11 May 2012 23:53:38 +0000 (23:53 +0000)]
format_mp3: Fix a possible crash in mp3_read().
This patch fixes a potential crash in mp3_read() by not assuming that
dbuf has enough data to finish filling up the output buffer. The patch
also makes sure that the dbuf state gets reset after we know we read
everything out of it already.
In passing, this patch includes some other cleanups of this module,
including stripping trailing whitespace, formatting fixes based on
coding guidelines, and removing a number of unused members from the
private state struct.
(closes issue ASTERISK-19761)
Reported by: Chris Maciejewsk
Tested by: Chris Maciejewsk
Kinsey Moore [Thu, 10 May 2012 20:50:47 +0000 (20:50 +0000)]
Resolve FORWARD_NULL static analysis warnings
This resolves core findings from ASTERISK-19650 numbers 0-2, 6, 7, 9-11, 14-20,
22-24, 28, 30-32, 34-36, 42-56, 82-84, 87, 89-90, 93-102, 104, 105, 109-111,
and 115. Finding numbers 26, 33, and 29 were already resolved. Those skipped
were either extended/deprecated or in areas of code that shouldn't be
disturbed.
Jonathan Rose [Wed, 9 May 2012 19:10:17 +0000 (19:10 +0000)]
Block on frameout if the hardware has enough samples to complete a frame.
Fixes some problems with skipping audio in elaborate scenarios involving
multiple codecs by making codec_dahdi operate in a more synchronous
fashion similar to codec_g729. This change also fixes the use of file
conversion tools from Asterisk's CLI. This change may cause the thread
responsible for transcoding audio to block briefly (Shaun Ruffell describes
this as 'several milliseconds') while waiting for the hardware transcoder.
Mark Michelson [Wed, 9 May 2012 16:11:52 +0000 (16:11 +0000)]
Prevent sip_pvt refleak when an ast_channel outlasts its corresponding sip_pvt.
chan_sip was coded under the assumption that a SIP dialog with an owner channel
will always be destroyed after the owner channel has been hung up.
However, there are situations where the SIP dialog can time out and auto destruct
before the corresponding channel has hung up. A typical example of this would be
if the 'h' extension in the dialplan takes a long time to complete. In such cases,
__sip_autodestruct() would complain about the dialog being auto destroyed with
an owner channel still in place. The problem is that even once the owner channel
was hung up, the sip_pvt would still be linked in its ao2_container because nothing
would ever unlink it.
The fix for this is that if __sip_autodestruct() is called for a sip_pvt that still
has an owner channel in place, the destruction is rescheduled for 10 seconds in the
future. This will continue until the owner channel is finally hung up.
(closes issue ASTERISK-19425)
reported by David Cunningham
Patches:
ASTERISK-19425.patch uploaded by Mark Michelson (License #5049)
(closes issue ASTERISK-19455)
reported by Dean Vesvuio
Tested by Dean Vesvuio
Richard Mudgett [Tue, 8 May 2012 20:14:30 +0000 (20:14 +0000)]
* Fix FollowMe memory leak on error paths in app_exec().
* Fix FollowMe leaving recorded caller name file on error paths in
app_exec().
* Use correct buffer dimension define in struct call_followme.moh[] and
struct fm_args.namerecloc[]. This fixes unexpected namerecloc filename
length restriction.
Richard Mudgett [Tue, 8 May 2012 18:02:29 +0000 (18:02 +0000)]
* Fix accept/decline DTMF buffer overwrite in FollowMe.
* Made use MAX_YN_STRING define to make all accept/decline DTMF buffers
the same size. Just using 20 isn't good enough when someone didn't get
the memo.
* Fix stupid use of a global variable in FollowMe. (ynlongest)
Mark Michelson [Tue, 8 May 2012 15:48:10 +0000 (15:48 +0000)]
Send more accurate identification information in dialog-info SIP NOTIFYs.
This uses the calling channel's caller ID and connected line information
to populate the remote and local identities in the dialog-info NOTIFY when
an extension is ringing.
There is a bit of an oddity here, and that is that we seed the remote target
with the To header of the outbound call rather than the from header. This
is because it was reported that seeding with the from header caused hints
to be broken with certain SNOM devices. A comment has been added to the code
to explain this.
(closes issue ASTERISK-16735)
reported by Maciej Krajewski
patches:
local_remote_hint2.diff uploaded by Mark Michelson (license #5049)
16735_tweak1.diff uploaded by Mark Michelson (license #5049)
Tested by Niccolo Belli
Matthew Jordan [Mon, 7 May 2012 18:36:54 +0000 (18:36 +0000)]
Support VoiceMail d() option when extension does not exist in channel's context
The VoiceMail d([c]) option is documented to accept digits for a new extension
in context <c>, if played during the greeting. This option works fine if the
extension being redirected to has an extension with the same initial digit in
the channel's current context. If that digit did not happen to exist in some
extension, a dialplan match would fail and the user would not be redirected.
This patch fixes it such that if the <c> option is used, the extensions are
matched in that context as opposed to the caller's original context.
Kinsey Moore [Fri, 4 May 2012 22:12:55 +0000 (22:12 +0000)]
Fix many issues from the NULL_RETURNS Coverity report
Most of the changes here are trivial NULL checks. There are a couple
optimizations to remove the need to check for NULL and outboundproxy parsing
in chan_sip.c was rewritten to avoid use of strtok. Additionally, a bug was
found and fixed with the parsing of outboundproxy when "outboundproxy=," was
set.
Richard Mudgett [Fri, 4 May 2012 16:24:34 +0000 (16:24 +0000)]
Fix local channel chains optimizing themselves out of a call.
* Made chan_local.c:check_bridge() check the return value of
ast_channel_masquerade(). In long chains of local channels, the
masquerade occasionally fails to get setup because there is another
masquerade already setup on an adjacent local channel in the chain.
* Made the outgoing local channel (the ;2 channel) flush one voice or
video frame per optimization attempt.
* Made sure that the outgoing local channel also does not have any frames
in its queue before the masquerade.
* Made do the masquerade immediately to minimize the chance that the
outgoing channel queue does not get any new frames added and thus
unconditionally flushed.
* Made block indication -1 (Stop tones) event when the local channel is
going to optimize itself out. When the call is answered, a chain of local
channels pass down a -1 indication for each bridge. This blizzard of -1
events really slows down the optimization process.
(closes issue ASTERISK-16711)
Reported by: Alec Davis
Tested by: rmudgett, Alec Davis
Review: https://reviewboard.asterisk.org/r/1894/
Mark Michelson [Fri, 4 May 2012 15:48:44 +0000 (15:48 +0000)]
Fix core FINDING 2, FINDING 3, and FINDING 4 from Coverity's CONSTANT_EXPRESSION_RESULT report.
These three all are in RTP code that attempts to print the number of sequence number cycles
in an RTCP RR report. The code was masking out the upper 16 bits and then shifting the number
right by 16 bits. This led to an all zero result in all cases. The fix is to do the shift without
the bit masking.
Terry Wilson [Wed, 2 May 2012 17:02:39 +0000 (17:02 +0000)]
Don't leak a ref if out of memory and can't link the linkedid
If the ao2_link fails, we are most likely out of memory and bad things
are going to happen. Before those bad things happen, make sure to clean
up the linkedid references.
This patch also adds a comment explaining why linkedid can't be passed
to both local channel allocations and combines two ao2_ref calls into 1.
Terry Wilson [Wed, 2 May 2012 15:49:03 +0000 (15:49 +0000)]
Fix a CEL LINKEDID_END race and local channel linkedids
This patch has the ;2 channel inherit the linkedid of the ;1 channel and fixes
the race condition by no longer scanning the channel list for "other" channels
with the same linkedid. Instead, cel.c has an ao2 container of linkedid strings
and uses the refcount of the string as a counter of how many channels with the
linkedid exist. Not only does this eliminate the race condition, but it also
allows us to look up the linkedid by the hashed key instead of traversing the
entire channel list.
Mark Michelson [Tue, 1 May 2012 23:08:20 +0000 (23:08 +0000)]
Fix Coverity-reported ARRAY_VS_SINGLETON error.
As it turned out, this wasn't a huge deal. We were calling
ast_app_parse_options() for a set of options of which none
took arguments. The proper thing to do for this case is to
pass NULL for the "args" parameter here. We were instead passing
a seemingly-randomly chosen char * from the function. While this
would never get written to, you can rest assured things would
have gotten bad had new options (which took arguments) been added
to func_volume.
projects / thirdparty / asterisk.git / log
