Jeff Peeler [Wed, 12 Nov 2008 17:53:44 +0000 (17:53 +0000)]
(closes issue #13173)
Reported by: pep
This change adds an announce_thread responsible for playing announcements to an existing conference. This allows all announcing to be immediately stopped if necessary but more importantly allows other threads that need to play something to not block. There are multiple benefits to this, but the actual bug is for solving the scenario for a channel to be unusable after hang up for the entire duration of the parting announcement. The parting announcement can be extremely long depending on what the user recorded upon joining the conference.
Reviewed by Russell on Review Board:
http://reviewboard.digium.com/r/25/
Mark Michelson [Wed, 12 Nov 2008 17:38:33 +0000 (17:38 +0000)]
When doing some tests, I was having a crash at the end of every call
if an attended transfer occurred during the call. I traced the cause to
the CDR on one of the channels being NULL. murf suggested a check in
the end bridge callback to be sure the CDR is non-NULL before proceeding,
so that's what I'm adding.
Russell Bryant [Wed, 12 Nov 2008 17:29:52 +0000 (17:29 +0000)]
Move the sanity check that makes sure "always fork" is not set along with the
console option to be after the code that reads options from asterisk.conf.
This resolves a situation where Asterisk can start taking up 100% when
misconfigured.
(Thanks to Bryce Porter (x86 on IRC) for letting me log in to his system to
figure out what was causing the 100% CPU problem.)
Mark Michelson [Mon, 10 Nov 2008 21:07:39 +0000 (21:07 +0000)]
Channel drivers assume that when their indicate callback
is invoked, that the channel on which the callback was called
is locked. This patch corrects an instance in chan_agent where
a channel's indicate callback is called directly without first
locking the channel.
This was leading to some observed locking issues in chan_local,
but considering that all channel drivers operate under the
same expectations, the generic fix in chan_agent is the right
way to go.
Sean Bright [Sun, 9 Nov 2008 01:08:07 +0000 (01:08 +0000)]
Use static functions here instead of nested ones. This requires a small
change to the ast_bridge_config struct as well. To understand the reason
for this change, see the following post:
Mark Michelson [Thu, 6 Nov 2008 19:45:52 +0000 (19:45 +0000)]
The documentation listed the ability to set 'maxmsg' per
context. The truth is that you can only set this in the general section
or per mailbox. Thus I am updating the sample config file to be more
accurate.
Thanks to sasargen on IRC for bringing up this issue.
Tilghman Lesher [Tue, 4 Nov 2008 20:49:33 +0000 (20:49 +0000)]
On busy systems, it's possible for the values checked within a single line
of code to change, unless the structure is locked to ensure a consistent
state.
(closes issue #13717)
Reported by: kowalma
Patches:
20081102__bug13717.diff.txt uploaded by Corydon76 (license 14)
Tested by: kowalma
Tilghman Lesher [Mon, 3 Nov 2008 22:27:10 +0000 (22:27 +0000)]
Attempting to expunge a mailbox when the mailstream is NULL will crash Asterisk.
(Closes issue #13829)
Reported by: jaroth
Patch by: me (modified jaroth's patch)
fix a bunch of potential problems found by gcc 4.3.x, primarily bare strings being passed to printf()-like functions and ignored results from read()/write() and friends
Terry Wilson [Fri, 31 Oct 2008 15:45:29 +0000 (15:45 +0000)]
Recent CDR fixes moved execution of the 'h' exten into the bridging code, so variables that were set after ast_bridge_call was called would not show up in the 'h' exten. Added a callback function to handle setting variables, etc. from w/in the bridging code. Calls back into a nested function within the function calling ast_bridge_call
Steve Murphy [Wed, 29 Oct 2008 05:19:04 +0000 (05:19 +0000)]
A little documentation cross-ref between features and
dial and queue... I wasted some time (stupidly) trying
to get the one-touch parking stuff working, because it
didn't occur to me that I had to also have the corresponding
options in the dial command! Duh! (In all this time, I never
set this up before!)
So, to keep some poor fool from suffering the same fate,
I made the features.conf.sample file mention the corresponding
opts in dial/queue; and the docs for dial/app specifically
mention the corresponding decls in the feature.conf file.
I hope this doesn't spoil some vast, eternal plan...
Steve Murphy [Wed, 29 Oct 2008 04:36:32 +0000 (04:36 +0000)]
The magic trick to avoid this crash is not to
try to find the channel by name in the list,
which is slow and resource consuming, but rather
to pay attention to the result codes from the
ast_bridge_call, to which I added the
AST_PBX_NO_HANGUP_PEER_PARKED value, which
now are returned when a channel is parked.
If you get AST_PBX_KEEPALIVE,
then don't touch the channel pointer.
If you get AST_PBX_NO_HANGUP_PEER, or
AST_PBX_NO_HANGUP_PEER_PARKED, then don't
touch the peer pointer.
Updated the several places where the results
from a bridge were not being properly obeyed,
and fixed some code I had introduced so that
the results of the bridge were not overridden
(in trunk).
All the places that previously tested for
AST_PBX_NO_HANGUP_PEER now have to check for
both AST_PBX_NO_HANGUP_PEER and AST_PBX_NO_HANGUP_PEER_PARKED.
I tested this against the 4 common parking
scenarios:
1. A calls B; B answers; A parks B; B hangs up while A is getting the parking
slot announcement, immediately after being put on hold.
2. A calls B; B answers; A parks B; B hangs up after A has been hung up, but
before the park times out.
3. A calls B; B answers; B parks A; A hangs up while B is getting the parking slot announcement, immediately after being put on hold.
4. A calls B; B answers; B parks A; A hangs up after B has been hung up, but before the park times out.
No crash.
I also ran the scenarios above against valgrind, and accesses looked good.
Tilghman Lesher [Tue, 28 Oct 2008 17:04:56 +0000 (17:04 +0000)]
Reset all DIAL variables back to blank, in case Dial is called multiple times
per call (which could otherwise lead to inconsistent status reports).
(closes issue #13216)
Reported by: ruddy
Patches:
20081014__bug13216.diff.txt uploaded by Corydon76 (license 14)
Tested by: ruddy
Tilghman Lesher [Mon, 27 Oct 2008 21:32:00 +0000 (21:32 +0000)]
Inherit ALL elements of CallerID across a local channel.
(closes issue #13368)
Reported by: Peter Schlaile
Patches:
20080826__bug13368.diff.txt uploaded by Corydon76 (license 14)
Sean Bright [Sun, 26 Oct 2008 20:23:36 +0000 (20:23 +0000)]
Since passing \0 as the second argument to strchr is valid (and will
match the trailing \0 of a string) we need to check that first, otherwise
we end up with incorrect results. Fix suggested by reporter.
Russell Bryant [Sat, 25 Oct 2008 10:59:02 +0000 (10:59 +0000)]
Move AMI initialization to occur after loading modules. This prevents a
deadlock when someone tries to initiate a module reload from the AMI just
as Asterisk is starting.
(closes issue #13778)
Reported by: hotsblanc
Fix suggested by hotsblanc
Terry Wilson [Thu, 23 Oct 2008 16:04:42 +0000 (16:04 +0000)]
Backport fix from 1.6.0 that allows you to set parkedcalltransfers=no|caller|callee|both, but default to both which would be the equivalent of the existing behaviour.
The problem was that if someone parked a call, the callee and caller would both get assigned the builtin transfer feature, which would not only be potentially giving someone the ability to transfer themselves when they shouldn't have it, but would also dissallow reinviting the media off of the call.
(closes issue #12854)
Reported by: davidw
Patches:
parkingfix4.diff.txt uploaded by otherwiseguy
Tested by: davidw, otherwiseguy
Kevin P. Fleming [Mon, 20 Oct 2008 04:45:56 +0000 (04:45 +0000)]
break up acinclude.m4 into individual files, which will make it easier to maintain, easier to add new macros (less patching) and will ease maintenance of these macros across Asterisk branches
BJ Weschke [Sun, 19 Oct 2008 19:51:16 +0000 (19:51 +0000)]
As per kpfleming's comments to the prior commit, I'm reverting some of the changes here.
A comment was made in bug #13726
"3. The same mistake as in (2) is done in a few other places in the code that check for: #if defined(HAVE_ZAPTEL) || defined(HAVE_DAHDI)
Harmless, but still incorrect."
In the case of main/asterisk.c, this is not incorrect because without HAVE_ZAPTEL defined, we're missing
the include for ioctl and the namespace that defines DAHDI_TIMERCONFIG which is still required when
using Zaptel with the 1.4 branch.
BJ Weschke [Sat, 18 Oct 2008 01:42:23 +0000 (01:42 +0000)]
Using the GetVar handler in AMI is potentially dangerous (insta-crash [tm]) when you use a dialplan function that requires a channel and then you don't provide one or provide an invalid one in the Channel: parameter. We'll handle this situation exactly the same way it was handled in pbx.c back on r61766.
We'll create a bogus channel for the function call and destroy it when we're done. If we have trouble allocating the bogus channel then we're not going to try executing the function call at all and run the risk of crashing.
(closes issue #13715)
reported by: makoto
patch by: bweschke
Jason Parker [Fri, 17 Oct 2008 15:31:35 +0000 (15:31 +0000)]
Correctly allow chan_dahdi to compile against older versions of Zaptel.
Don't always define HAVE_ZAPTEL_CHANALARMS (since we check if it's defined..)
Minor cleanup to make things clear.
Mark Michelson [Thu, 16 Oct 2008 23:40:54 +0000 (23:40 +0000)]
Reverting changes from commits 150298 and 150301 since
I was mistakenly under the assumption that dialplan functions
*always* required that a channel be present. I need to go
home earlier, I think :)
Mark Michelson [Thu, 16 Oct 2008 23:34:37 +0000 (23:34 +0000)]
Don't try to call a dialplan function's read callback from
the manager's GetVar handler if an invalid channel has
been specified. Several dialplan functions, including
CHANNEL and SIP_HEADER, do not check for NULL-ness of
the channel being passed in.
Steve Murphy [Thu, 16 Oct 2008 15:26:10 +0000 (15:26 +0000)]
This patch is relevant to:
ABE-1628 and RYM-150398 and AST-103 in internal Digium
bug trackers.
These fixes address a really subtle memory corruption
problem that would happen in machines heavily loaded
in production environments. The corruption would
always take the form of the STMT object getting
nulled out and one of the unixODBC calls would
crash trying to access statement->connection.
It isn't fully proven yet, but the server has
now been running 2.5 days without appreciable
memory growth, or any gain of %cpu, and no
crashes. Whether this is the problem or not
on that server, these fixes are still warranted.
As it turns out, **I** introduced these errors
unwittingly, when I corrected another crash earlier.
I had formed the build_query routine, and failed
to remove mutex_unlock calls in 3 places in the
transplanted code. These unlocks would only
happen in error situations, but unlocking the
mutex early set the code up for a catastrophic
failure, it appears. It would happen only once
every 100K-200K or more calls, under heavy load...
but that is enough.
If another crash occurs, with the same MO,
I'll come back and remove my confession from the log, and
we'll keep searching, but the fact that we
have Asterisk dying from an asynchronous
wiping of the STMT object, only on some connection
error, and that the server has lived for 2.5
days on this code without a crash, sure make
it look like this was the problem!
Also, in several points, Statement handles are
set to NULL after SQLFreeHandle. This was mainly
for insurance, to guarantee a crash. As it turns
out, the code does not appear to be attempting
to use these freed pointers.
Asterisk owes a debt of gratitude to Federico Alves
and Frediano Ziglio for their untiring efforts in
finding this bug, among others.
BJ Weschke [Wed, 15 Oct 2008 18:28:54 +0000 (18:28 +0000)]
An update to the documentation/example of agents.conf.sample with the correct parameter for this feature as defined in chan_agent.c
(closes issue #13709)
Mark Michelson [Tue, 14 Oct 2008 23:00:01 +0000 (23:00 +0000)]
Add a tolerance period for sync-triggered audiohooks
so that if packetization of audio is close (but not equal)
we don't end up flushing the audiohooks over small
inconsistencies in synchronization.
Related to issue #13005, and solves the issue
for most people who were experiencing the problem.
However, a small number of people are still experiencing
the problem on long calls, so I am not closing
the issue yet
Mark Michelson [Tue, 14 Oct 2008 22:40:42 +0000 (22:40 +0000)]
Update the queue with the correct number of calls and
whether the call was completed within the service level
when a transfer takes place. This way, we do not "break"
the leastrecent and fewestcalls strategies by not logging
a call until after the transferred call has ended.
Tilghman Lesher [Tue, 14 Oct 2008 20:09:06 +0000 (20:09 +0000)]
Check correct values in the return of ast_waitfor(); also, get rid of a
possible memory leak.
(closes issue #13658)
Reported by: explidous
Patch by: me
Kevin P. Fleming [Tue, 14 Oct 2008 10:30:54 +0000 (10:30 +0000)]
on Ubuntu (at least), recent versions of ld in binutils delete all debugging symbols when -x is supplied; since the reasons why -x is being passed are lost in the mists of time, remove it so debugging will work properly
Tilghman Lesher [Fri, 10 Oct 2008 16:25:31 +0000 (16:25 +0000)]
User not notified of temporary greeting, if ODBC storage is in use.
(closes issue #13659)
Reported by: moliveras
Patches:
20081009__bug13659.diff.txt uploaded by Corydon76 (license 14)
Tested by: moliveras
This change prevents a call that is placed in the parkinglot to be picked up before the PBX is finished. If another extension dials the parking extension before the PBX thread has completed at minimum warnings will occur about the PBX not properly being terminated. At worst, a crash could occur.
when parsing a text configuration option, ensure that the buffer on the stack is actually large enough to hold the legal values of that option, and also ensure that sscanf() knows to stop parsing if it would overrun the buffer (without these changes, specifying "buffers=...,immediate" would overflow the buffer on the stack, and could not have worked as expected)
don't start a PBX on incoming PRI call channels until after we're done setting channel variables and other things on the channel, otherwise the channel might go away (if the dialplan hangs up quickly) before we are done, which results in a spectacular crash
Tilghman Lesher [Mon, 6 Oct 2008 20:52:04 +0000 (20:52 +0000)]
Dialplan functions should not actually return 0, unless they have modified the
workspace. To signal an error (and no change to the workspace), -1 should be
returned instead.
(closes issue #13340)
Reported by: kryptolus
Patches:
20080827__bug13340__2.diff.txt uploaded by Corydon76 (license 14)
Tilghman Lesher [Mon, 6 Oct 2008 16:51:21 +0000 (16:51 +0000)]
Check whether an extension exists in the _call method, rather than the _alloc
method, because we need to evaluate the callerid (since that data affects
whether an extension exists).
(closes issue #13343)
Reported by: efutch
Patches:
20080915__bug13343.diff.txt uploaded by Corydon76 (license 14)
Tested by: efutch
Tilghman Lesher [Thu, 2 Oct 2008 16:39:56 +0000 (16:39 +0000)]
Backport support for some of the keyword modifications used in 1.6 (while warning that
some options aren't really supported) and add some warning messages. Some credit to
oej, who was complaining in #asterisk-dev.
channels/chan_misdn.c
* Made bearer2str() use allowed_bearers_array[]
* Made use the causes.h defines instead of hardcoded numbers.
* Made use Asterisk presentation indicator values if either of the
mISDN presentation or screen options are negative.
* Updated the misdn_set_opt application option descriptions.
* Renamed the awkward Caller ID presentation misdn_set_opt
application option value not_screened to restricted.
Deprecated the not_screened option value.
channels/misdn/isdn_lib.c
* Made use the causes.h defines instead of hardcoded numbers.
* Fixed some spelling errors and typos.
* Added all defined facility code strings to fac2str().
channels/misdn/isdn_lib.h
* Added doxygen comments to struct misdn_bchannel.
channels/misdn/isdn_lib_intern.h
* Added doxygen comments to struct misdn_stack.
channels/misdn_config.c
configs/misdn.conf.sample
* Updated the mISDN presentation and screen parameter descriptions.
doc/misdn.txt (doc/tex/misdn.tex)
* Updated the misdn_set_opt application option descriptions.
* Fixed some spelling errors and typos.
Kevin P. Fleming [Sat, 27 Sep 2008 15:00:48 +0000 (15:00 +0000)]
improve header inclusion process in a few small ways:
- it is no longer necessary to forcibly include asterisk/autoconfig.h; every module already includes asterisk.h as its first header (even before system headers), which serves the same purpose
- astmm.h is now included by asterisk.h when needed, instead of being forced by the Makefile; this means external modules will build properly against installed headers with MALLOC_DEBUG enabled
- simplify the usage of some of these headers in the AEL-related stuff in the utils directory
Mark Michelson [Fri, 26 Sep 2008 22:14:59 +0000 (22:14 +0000)]
This patch was applied to 1.4 but it completely
does not apply since the "found" pointer is not
passed in to this function. If this is going to
be backported, it needs to be done differently or
a deeper backport needs to be done.
This patch was mainly meant to apply to trunk and 1.6.x,
but I'm applying it to 1.4 also, which should be a perfectly
harmless fix to the vast majority of users who are not using
external switches, but the few who might be affected
will not have to go to the pain of filing a bug report.
Many thanks to neutrino88 for this patch, which
solves a problem whereby channels get a CANCEL
request, respond to it properly, but end up
in a hung state, infinitely being rescheduled.
This fix is a bit crude, in that catches the
problem at a rather late phase, but it may
prevent infinite rescheduling problems that
might still arise.
It might have been better to find out why,
in the course of protocol handling, the channel
was not destroyed, but we leave that to
future generations.
Many thanks to urzedo and thiagofernandes for
their work in verifying that the patch code
indeed is being executing, and averting the
problem.
This crash happens because we are unsafely handling old pointers.
The channel whose cdr is being handled, has been hung up and
destroyed already. I reorganized the code a bit, and tried not
to lose the fork-cdr-chain concepts of the previous code.
I now verify that the 'previous' channel (the channel we
had when the bridge was started), still exists, by looking it up
by name in the channel list. I also do not try to reset the
CDR's of channels involved in bridges.
Testing shows it solves the crash problem, and should not
negatively impact previous fixes involving CDR's generated
during/after blind transfers. (The reason we need to reset
the CDR's on the "beginning" channels in the first place).
Steve Murphy [Tue, 23 Sep 2008 14:22:10 +0000 (14:22 +0000)]
In at least one machine, we noted that the timestr
was not getting set in the STMT; it was coming out,
usually, as binary garbage to an mssql server.
These changes fixed the problem. The only thing
I can venture forth as a guess, is that the pointer
is being stored in the interface, not a copy of the
string. Because we ripped the build process into a
subroutine, the timestr became a temp. stack variable,
and between the time the STMT got built and the
time it was executed on the server, the string being
pointed to was damaged. At any rate, even if this
theory is false, and some mechanism was at fault,
this fix worked reliably where it didn't before.
Why this bug didn't bite last week, I have no idea.
This change basically defines the timestr buffer
in the calling function, extending the life of the
buffer to cover both the STMT's building and
processing to the server.
Steve Murphy [Fri, 19 Sep 2008 21:07:05 +0000 (21:07 +0000)]
This fix comes from a debugging session on a test box that has been getting hung channels when the mssqlserver bounces. All the connections then become invalid, and must be reconnected. The cdr_odbc backend had code to do it, but depended on re-establishing the connection, but re-using the STMT that had been built. By trial and error, we determined that the STMT could not be re-used after the connection was re-established. and must be rebuilt. These changes accomplish this.
When callerid is blank, we want to use "unknown caller" in those cases, too.
(closes issue #13486)
Reported by: tomo1657
Patches:
20080917__bug13486.diff.txt uploaded by Corydon76 (license 14)
Mark Michelson [Wed, 17 Sep 2008 18:24:15 +0000 (18:24 +0000)]
Allow for "G.729" if offered in an SDP even though
it is not RFC 3551 compliant. Some Cisco switches
will send this in an SDP, and it doesn't hurt to
be able to accept this.
Changed park_call_full to hold the parkinglot lock a little longer, which protects the parkeduser struct from being freed out from underneath. Made sure that the parking extension is added to the parking context while holding the lock thereby ensuring that there are no spurious warnings from removal attempts when a hangup occurs while the parking lot is being announced.
The main change here was to masquerade the channel if the channel that was to be parked was running a PBX on it. The PBX thread can then maintain full control of the channel (the zombie) as it expects to while allowing the parking thread full control of the real (parked) channel.
projects / thirdparty / asterisk.git / log
