Jonathan Rose [Mon, 1 Aug 2011 21:08:10 +0000 (21:08 +0000)]
Asterisk 18103 - Fix reload crash caused by destroying default parking lot
Default parking lot was being destroyed in reload and was not being rebuilt properly.
This patch keeps features.c reload from destroying the default parking lot in 1.6.2.
Bug was caused by a hasty backport which didn't test reload enough times to catch the
problem.
Kinsey Moore [Tue, 21 Jun 2011 16:10:09 +0000 (16:10 +0000)]
ConfBridge does not handle hangup properly
When playing back a prompt to a channel, confbridge neglects to check for
hangup events causing lockup condititions for hangups that occur before
actually joining the conference. This change ensures that the user is removed
from the conference in the event of a premature hangup.
DYNAMIC_FEATURES were broken by a recent DTMF change. This patch makes
sure that dynamic features are also checked when deciding whether or not
to pass DTMF through or store it for interpreting.
Resolve a segfault/bus error when we try to map memory that falls on a page
boundary.
The fix for ASTERISK-15359 was incorrect in that it added 1 to the length of the
mmap'd region. The problem with this is that reading/writing to that extra byte
outside of the bounds of the underlying fd causes a bus error.
The real issue is that we are working with both a FILE * and the raw fd
underneath it and not synchronizing between them. The code that was removed in
ASTERISK-15359 was correct, but we weren't flushing the FILE * before mapping
the fd.
Looking at the manager code in 1.4 reveals that the FILE * in 'struct
mansession' is never used except to create a temporary file that we immediately
fdopen. This means we just need to write a 0 byte to the fd and everything will
just work. The other branches require a call to fflush() which, while not a
guaranteed fix, should reduce the likelihood of a crash.
This all makes sense in my head.
(closes issue ASTERISK-16460)
Reported by: Ravelomanantsoa Hoby (hoby)
Patches:
issue17747_1.4_svn_markII.patch uploaded by Sean Bright (license #5060)
........
Alec L Davis [Fri, 27 May 2011 08:24:32 +0000 (08:24 +0000)]
Fix *8 directed pickup locks system during pickupsound play out
move playout from sip_pickup_thread to bridge using BRIDGE_PLAY_SOUND method,
This stop the clash of 2 threads trying to write audio to same channel.
In addition fixes choppy audio beep in issue 19177.
The meetme CLI command completion leaves conferences mutex locked.
When issuing a meetme kick CLI command and an invalid (non-existent)
conference number is specified, pressing Tab leaves the conferences mutex
locked and, therefore, all conferences deadlock.
This commit modifies the way polling is done on TLS sockets.
Because of the buffering the TLS layer does, polling is unreliable. If poll is
called while there is data waiting to be read in the TLS layer but not at the
network layer, the messaging processing engine will not proceed until something
else writes data to the socket, which may not occur. This change modifies the
logic around TLS sockets to only poll after a failed read on a non-blocking
socket. This way we know that there is no data waiting to be read from the
buffering layer.
(closes issue #19182)
Reported by: st
Patches:
ssl-poll-fix3.diff uploaded by mnicholson (license 96)
Tested by: mnicholson
Make sure everyone gets an unhold when a transfer succeeds
Some phones, like the Snom phones, send a hold to the transfer target after
before sending the REFER. We need to make sure that we unhold the parties
that are being connected after the masquerade. If Local channels with the /nm
option are used when dialing the parties, hold music would still be playing on
the transfer target, even after being connected with the transferee.
........
Merged revisions 318671 via svnmerge from
https://origsvn.digium.com/svn/asterisk/branches/1.8
* The applicable fixes for v1.4 are the SIP deadlock and the in progress
masquerade check for multiple parties trying to pickup the same call.
issue18654_v1.4.patch uploaded by rmudgett (license 664)
* Backported to v1.6.2.
issue18654_v1.6.2.patch uploaded by rmudgett (license 664)
Fix directed group pickup feature code *8 with pickupsounds enabled
Since 1.6.2, the new pickupsound and pickupfailsound in features.conf cause many issues.
1). chan_sip:handle_request_invite() shouldn't be playing out the fail/success audio, as it has 'netlock' locked.
2). dialplan applications for directed_pickups shouldn't beep.
3). feature code for directed pickup should beep on success/failure if configured.
Created a sip_pickup() thread to handle the pickup and playout the audio, spawned from handle_request_invite.
Moved app_directed:pickup_do() to features:ast_do_pickup().
Functions below, all now use the new ast_do_pickup()
app_directed_pickup.c:
pickup_by_channel()
pickup_by_exten()
pickup_by_mark()
pickup_by_part()
features.c:
ast_pickup_call()
Deadlock between ast_cel_report_event and ast_do_masquerade
chan_sip calls pbx_exec on a pvt's owner channel while only the
pvt lock is held. Since pbx_exec calls ast_cel_report_event which
attempts to lock the channel, invalid locking order occurs. Channels
should be locked before pvt's.
Terry Wilson [Wed, 11 May 2011 17:15:39 +0000 (17:15 +0000)]
Clean up several chan_sip reference leaks
Several situations in the code could lead to peers or sip_pvt references
being leaked. This would cause RTP ports to never be destroyed (leading
to exhaustion of all available RTP ports) and memory leaks.
The original patch for this issue from rgagnon was the result of an
obscene amount of testing and hard work, for which I am very grateful. I
did some cleanup and added a few additional refcount fixes that I found.
Terry Wilson [Mon, 9 May 2011 20:04:41 +0000 (20:04 +0000)]
Don't offer video to directmedia callee unless caller offered it as well
Make sure that when directmedia is enabled, that video is not offered to the
callee even if it supports it. p->vrtp will not exist since the caller didn't
offer video.
Don't start an autoservice in pbx_lua if pbx_lua already started one and don't
stop one if we didn't start one. Also start and stop the autoservice when
transferring control from and to the pbx.
This part of the change for r315596 was incorrect. No bridge occurs
when doing a roundrobin dial and no one answers, so this code shouldn't
have been removed.
........
Set SO_KEEPALIVE on SIP TCP sockets so that they eventually go away when a peer
abruptly disappears. This mostly occurs after a successful registration.
(closes issue #17544)
Reported by: marcelloceschia
Patches:
(modified) tcptls.patch uploaded by st (license 907)
David Vossel [Wed, 4 May 2011 13:40:41 +0000 (13:40 +0000)]
Fixes session-timers=refuse not being enforced for *caller*
During handle_request_invite, the session timer mode was retrieved from
a cached variable. This patch forces a peer lookup of the session timer
mode in the case of an incoming invite.
(closes issue #18804)
Reported by: wdoekes
Patches:
issue18804_session_timer_refuse_caller.patch uploaded by wdoekes (license 717)
issue_18804_v2.diff uploaded by dvossel (license 671)
This change optimizes the free_via() function and removes some redundant null
checking. It also fixes compliance with RFC 3261 section 18.2.2 by always using
the port specified in the Via header for routing responses (even when maddr is
not set). Also the htons() function is now used when setting the port.
Additional documentation comments have been added in various places to make the
logic in the code clearer.
Allow transfer loops without allowing forwarding loops
We try to avoid the situation where two phones may be forwarded to each other
causing an infinite loop by storing each dialed interface in a channel
datastore and checking the list before dialing out. This works, but currently
breaks situations like A calls B, A transfers B to C, B transfers C to A, and A
transfers C to B. Since human interaction is happening here and not an
automated forwarding loop, it should be allowed.
This patch removes the dialed_interfaces datastore when a call is bridged (a
suggestion from the brilliant mmichelson). If a call is being bridged, it
should be safe to assume that we aren't stuck in a loop.
Since we are now handling this is the bridge code, the previous attempts at
handling it in app_dial and app_queue are removed.
The code that set the bit within the select bitfield was correct, but the
bounds-checking code was not. The change to that line uses the new _bitsize
macro for clarity. Also, FD_ZERO macro did not zero-out anything but the
first word of the bitfield, so this could have caused problems with modules
using that macro with the expanded bitfield.
Initialize the buffers used to hold the result from GET VARIABLE or
GET VARIABLE FULL. The bug report shows func_read returning garbage in
the result. It assumed that the buffer passed in was initialized, like many
other functions do. In the more common code path (through the dialplan), it
is initialized, so just initialize it here too.
Russell Bryant [Fri, 22 Apr 2011 13:58:03 +0000 (13:58 +0000)]
Initialize buffers in getvar and getvarfull.
Initialize the buffers used to hold the result from GET VARIABLE or
GET VARIABLE FULL. The bug report shows func_read returning garbage in
the result. It assumed that the buffer passed in was initialized, like many
other functions do. In the more common code path (through the dialplan), it
is initialized, so just initialize it here too.
Added limits to the number of unauthenticated sessions TCP based protocols are allowed to have open simultaneously. Also added timeouts for unauthenticated sessions where it made sense to do so.
Unrelated, the manager interface now properly checks if the user has the "system" privilege before executing shell commands via the Originate action.
AST-2011-005
AST-2011-006
(closes issue #18787)
Reported by: kobaz
(related to issue #18996)
Reported by: tzafrir
........
Terry Wilson [Thu, 21 Apr 2011 00:17:34 +0000 (00:17 +0000)]
Don't allocate more space than necessary for a sip_pkt
This extra allocation is a hold-over from when pkt->data was a
character array. Now that it is an allocated string, just allocate
enough for the sip_pkt.
Asterisk does not hangup a channel after endpoint hangs up.
If the call that the dialplan started an AGI script for is hungup while
the AGI script is in the middle of a command then the AGI script is not
notified of the hangup. There are many AGI Exec commands that this can
happen with. The reported applications have been: Background, Wait, Read,
and Dial. Also the AGI Get Data command.
* Don't wait on the Asterisk channel after it has hung up. The channel is
likely to never need servicing again.
* Restored the AGI script's ability to return the AGI_RESULT_HANGUP value
in run_agi(). It previously only could return AGI_RESULT_SUCCESS or
AGI_RESULT_FAILURE after the DeadAGI and AGI applications were merged.
(closes issue #17954)
Reported by: mn3250
Patches:
issue17954_v1.8.patch uploaded by rmudgett (license 664)
issue17954_v1.6.2.patch uploaded by rmudgett (license 664)
issue17954_v1.4.patch uploaded by rmudgett (license 664)
Tested by: rmudgett
JIRA SWP-2171
chan_dahdi reloading with variables set via setvar in chan_dahdi.conf would
stay in the dahdi_pvt structs for individual channels (causing them to just
continue adding the new ones to the list) and also there was a memory leak
causes by the conf objects. This patch resolves both of these by using
ast_variables_destroy during the loading process.
Jonathan Rose [Tue, 12 Apr 2011 18:19:41 +0000 (18:19 +0000)]
white space change
........
reload Chan_dahdi memory leak caused by variables
chan_dahdi reloading with variables set via setvar in chan_dahdi.conf would
stay in the dahdi_pvt structs for individual channels (causing them to just
continue adding the new ones to the list) and also there was a memory leak
causes by the conf objects. This patch resolves both of these by using
ast_variables_destroy during the loading process.
Jonathan Rose [Tue, 12 Apr 2011 18:12:29 +0000 (18:12 +0000)]
fixes reload Chan_dahdi memory leak caused by variables
chan_dahdi reloading with variables set via setvar in chan_dahdi.conf would
stay in the dahdi_pvt structs for individual channels (causing them to just
continue adding the new ones to the list) and also there was a memory leak
causes by the conf objects. This patch resolves both of these by using
ast_variables_destroy during the loading process.
Stuck channel using FEATD_MF if caller hangs up at the right time.
The cause was actually a caller hanging up just at the end of the Feature
Group D DTMF tones that setup the call. The reason for this is a "guard
timer" that's implemented using ast_safe_sleep(100). If the caller
happens to hang up AFTER the final tone of the DTMF string but BEFORE the
end of that ast_safe_sleep(), then ast_safe_sleep() will return non-zero.
This causes the code to bounce to the end of ss_thread(), but it does NOT
tear down the call properly.
This should be a rare occurrence because the caller has to hang up at
EXACTLY the right time. Nonetheless, it was happening quite regularly on
the reporter's system. It's not easily reproducible, unless you purposely
increase the guard-time to 2000 or more. Once you do that, you can
reproduce it every time by watching the DTMF debug and hanging up just as
it ends.
Jonathan Rose [Thu, 7 Apr 2011 13:23:01 +0000 (13:23 +0000)]
Makes parking lots clear and rebuild properly when features reload is invoked from CLI
Before, default parkinglot in context parkedcalls with ext 700 would always be present and when reload was invoked, the previous parkinglots would not be cleared.
Issues with ISDN calls changing B channels during call negotiations.
The handling of the PROCEEDING message was not using the correct call
structure if the B channel was changed. (The same for PROGRESS.) The call
was also not hungup if the new B channel is not provisioned or is busy.
* Made all call connection messages (SETUP_ACKNOWLEDGE, PROCEEDING,
PROGRESS, ALERTING, CONNECT, CONNECT_ACKNOWLEDGE) ensure that they are
using the correct structure and B channel. If there is any problem with
the operations then the call is now hungup with an appropriate cause code.
* Made miscellaneous messages (INFORMATION, FACILITY, NOTIFY) find the
correct structure by looking for the call and not using the channel ID.
NOTIFY is an exception with versions of libpri before v1.4.11 because a
call pointer is not available for Asterisk to use.
* Made all hangup messages (DISCONNECT, RELEASE, RELEASE_COMPLETE) find
the correct structure by looking for the call and not using the channel
ID.
JIRA AST-437 (The issues fixed here are most likely causing this JIRA issue.)
JIRA DAHDI-406
JIRA LIBPRI-33 (Stuck resetting flag likely fixed)
........
voicemail: get real last_message_index and count_messages, ODBC resequence
change last_message_index to read the max msgnum stored in the database
change count_messages to actually count the number of messages.
last_message_index change:
This fixed overwriting of the last message if msgnum=0 was missing.
Previously every incoming message would overwrite msgnum=1.
count_messages change:
allows us to detect when requencing is required in opneA_mailbox.
resequence enabled for ODBC storage:
Assists with fixing up corrupt databases with gaps, but only when
a user actively opens there mailboxes.
(closes issue #18692,#18582,#19032)
Reported by: elguero
Patches:
based on odbc_resequence_mailbox2.1.diff uploaded by elguero (license 37)
Tested by: elguero, nivek, alecdavis
app_voicemail: close_mailbox needs to respect additional messages while mailbox is open.
close_mailbox leave gaps in message sequence if messages are deleted and new messages
arrive during this time, this is because the shuffle down to slot 0, only shuffles
the number of pre-existing messages when mailbox is opened, ignoring new arrivals.
Fix: in close_mailbox re-evaluate number of messages before the shuffle, this then includes new arrivals.
Jonathan Rose [Tue, 29 Mar 2011 13:17:01 +0000 (13:17 +0000)]
When comebacktoorigin=no, Asterisk no longer tries to dial extension <Originating Channe>@parkedcalltimeout and instead dials s without going through fallback.
Don't delay DTMF in core bridge while listening for DTMF features
This patch is mostly the work of Olle Johansson. I did some cleanup and
added the silence generating code if transmit_silence is set.
When a channel listens for DTMF in the core bridge, the outbound DTMF is not
sent until we have received DTMF_END. For a long DTMF, this is a disaster. We
send 4 seconds of DTMF to Asterisk, which sends no audio for those 4 seconds.
Some products see this delay and the time skew on RTP packets that results and
start ignoring the audio that is sent afterward.
With this change, the DTMF_BEGIN frame is inspected and checked. If it matches
a feature code, we wait for DTMF_END and activate the feature as before. If
transmit_silence=yes in asterisk.conf, silence is sent if we paritally match a
multi-digit feature. If it doesn't match a feature, the frame is forwarded
along with the DTMF_END without delay. By doing it this way, DTMF is not delayed.
(closes issue #15642)
Reported by: jasonshugart
Patches:
issue_15652_dtmf_ast-1.4.patch.txt uploaded by twilson (license 396)
Tested by: globalnetinc, jde
"Caller*ID failed checksum" on Wildcard TDM2400P and TDM410
The last character in the caller id message is getting a framing error.
The checksum is the last character in the message. A framing error in the
checksum could be because:
1) The sender did not send a full stop bit.
2) The sender cut off the FSK carrier too soon.
3) The sender opted to send zero of the specified zero to 10 trailing mark
bits and round-off errors in the code resulted in the code not being where
it thought it was in the demodulated bit stream.
Bit 8 of 'b' is set when parity error.
Bit 9 of 'b' is set when framing error.
Made ignore the framing and parity error bits if the errored character is
the checksum. We can tolerate a framing/parity error there. The checksum
character validates the message.
Jonathan Rose [Mon, 14 Mar 2011 13:56:22 +0000 (13:56 +0000)]
Adds 'p' as an option to func_volume. When it is on, the old behavior with DTMF controlling volume adjustment will be enforced.
When it is off, DTMF will not be processed by the function.
Programmed by Jonathan Rose
Reviewed by David Vossel, Leif Madsen, and Russell Bryant
Add AELSub, which provides a stable entry point into AEL subroutines.
This commit needs some explanation, given that we're adding a new application
into an existing release branch. This is generally a violation of our release
policy, except in very limited circumstances, and I believe this is one of
those circumstances.
The problem that this solves is one of the sanity of using multiple dialplan
languages to define a dialplan. In the case of the reporter, he or she is
using AEL is define subroutines, while using Realtime extensions to invoke
those subroutines. While you can do this, it's based upon the reality of AEL
using actual dialplan extensions; however, there is no guarantee that the
details of _how_ AEL is compiled into extensions will remain stable. In fact,
at the time of this commit, it has already changed twice, once in a
fundamental way.
Now normally, a new application would only be added to trunk. However, this
application is explicitly to create a stable user-level API between versions,
and adding it to trunk only will not solve the user's problem of switching
between 1.6.2 and 1.8, nor will it help anybody switching from 1.8 to 1.10.
Therefore, it needs to go into existing release branches. For the sake of
consistency, and also because one of the changes was between 1.4 and 1.6.x,
I am also electing to commit this to 1.4.
Check for errors from fseek() when loading config file, properly abort on errors from fread(), and supply a traceback for errors generated when loading the config file.
Also, prepend a newline to traceback output so that the main error message is on it's own line.
projects / thirdparty / asterisk.git / log
