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.
Set the raw formats at the same time as the other formats.
(closes issue #13240)
Reported by: jvandal
Patches:
20080813__bug13240.diff.txt uploaded by Corydon76 (license 14)
Create rules for disallowing contacts at certain addresses, which may
improve the security of various installations. As this does not change
any default behavior, it is not classified as a direct security fix for
anything within Asterisk, but may help PBX admins better secure their
SIP servers.
(closes issue #11776)
Reported by: ibc
Patches:
20080829__bug11776.diff.txt uploaded by Corydon76 (license 14)
Tested by: Corydon76, blitzrage
Steve Murphy [Fri, 12 Sep 2008 04:29:34 +0000 (04:29 +0000)]
Tested by: sergee, murf, chris-mac, andrew, KNK
This is a "second attempt" to restore the previous "endbeforeh" behavior
in 1.4 and up. In order to capture information concerning all the
legs of transfers in all their infinite combinations, I was forced
to this particular solution by a chain of logical necessities, the
first being that I was not allowed to rewrite the CDR mechanism from
the ground up!
This change basically leaves the original machinery alone, which allows
IVR and local channel type situations to generate CDR's as normal, but
a channel flag can be set to suppress the normal running of the h exten.
That flag would be set by the code that runs the h exten from the
ast_bridge_call routine, to prevent the h exten from being run twice.
Also, a flag in the ast_bridge_config struct passed into ast_bridge_call
can be used to suppress the running of the h exten in that routine. This
would happen, for instance, if you use the 'g' option in the Dial app.
Running this routine 'early' allows not only the CDR() func to be used
in the h extension for reading CDR variables, but also allows them to
be modified before the CDR is posted to the backends.
While I dearly hope that this patch overcomes all problems, and
introduces no new problems, reality suggests that surely someone
will have problems. In this case, please re-open 13251 (or 13289),
and we'll see if we can't fix any remaining issues.
Steve Murphy [Thu, 11 Sep 2008 22:55:49 +0000 (22:55 +0000)]
(closes issue #13364)
Reported by: mdu113
Well, fundamentally, the problems revealed in 13364 are
because of the ForkCDR call that is done before the dial.
When the bridge is in place, it's dealing with the first
(and wrong) cdr in the list.
So, I wrote a little func to zip down to the first non-locked
cdr in the chain, and thru-out the ast_bridge_call, these
results are used instead of raw chan->cdr and peer->cdr pointers.
This shouldn't affect anyone who isn't forking cdrs before a
dial, and should correct the cdr's of those that do.
So, this change ends up correcting the dstchannel
and userfield; the disposition was fixed by a previous
patch, it was OK coming into this problem.
Steve Murphy [Wed, 10 Sep 2008 21:58:17 +0000 (21:58 +0000)]
(closes issue #12318)
Reported by: krtorio
I made a small change to the code that handles local channel situations.
In that code, I copy the answer time from the peer cdr, to the bridge_cdr,
but I wasn't also copying the disposition from the peer cdr.
So, Now I copy the disposition, and I've tested against
these cases:
1. phone 1 never answers the phone; no cdr is generated at all.
this should show up as a manager command failure or something.
2. phone 2 never answers. CDR is generated, says NO ANSWER
3. phone 2 is busy. CDR is generated, says BUSY
4. phone 2 answers: CDR is generated, times are correct; disposition
is ANSWERED, which is correct. The start time is the time that
the manager dialed the first phone. The answer time is the time
the second phone picks up.
I purposely left the cid and src fields blank; since this call really
originates from the manager, there is no 'easy' data to put in these
fields. If you feel strongly that these fields should be filled in,
re-open this bug and I'll dig further.
Russell Bryant [Wed, 10 Sep 2008 16:39:53 +0000 (16:39 +0000)]
It is a normal situation that a task gets put in the scheduler that should run
as soon as possible. Accept "0" as an acceptable time to run, and also treat
negative as "run now", and don't print a debug message about it.
(inspired by a message asking about the "request to schedule in the past"
debug message on the -dev list)
Mark Michelson [Tue, 9 Sep 2008 16:19:17 +0000 (16:19 +0000)]
When determining if codecs used by SIP peers allow
the media to be natively bridged, use the jointcapability
instead of the peercapability.
It seems that the intent of using the peercapability was to
expand the choice of codecs for the call to increase the
chances of being able to native bridge the channels. The
problem is that if a codec were settled on for the native
bridge and that wasn't a codec that was configured to be used
by Asterisk for that peer, then Asterisk would send a
REINVITE with no codecs in the SDP which is a bug no matter
how you slice it.
Mark Michelson [Mon, 8 Sep 2008 21:10:10 +0000 (21:10 +0000)]
Fix pedantic mode of chan_sip to only check the
remote tag of an endpoint once a dialog has
been confirmed. Up until that point, it is possible
and legal for the far-end to send provisional
responses with a different To: tag each time. With
this patch applied, these provisional messages
will not cause a matching problem.
Russell Bryant [Mon, 8 Sep 2008 21:02:36 +0000 (21:02 +0000)]
When doing an async goto, detect if the channel is already in the middle of a
masquerade. This can happen when chan_local is trying to optimize itself out.
If this happens, fail the async goto instead of bursting into flames.
Jason Parker [Mon, 8 Sep 2008 20:15:42 +0000 (20:15 +0000)]
Remove RPM package targets from Makefile (and all associated parts).
This has never worked in 1.4, and we decided that it makes no sense to be done here.
There are many distros out there that already have "proper" spec files that can be (re)used.
Steve Murphy [Sat, 6 Sep 2008 20:13:16 +0000 (20:13 +0000)]
This fix comes from Joshua Colp The Brilliant, who, given the trace, came up with a solution. This will most likely will close 13235 and 13409. I'll wait till Monday to verify, and then close these bugs.
Mark Michelson [Fri, 5 Sep 2008 16:00:24 +0000 (16:00 +0000)]
Commit 140417 had a logic flaw in it which
caused port 5060 to always be used when dialing
a peer if no explicit port was specified. This
broke the behavior of implicitly using the port
from which the peer registered if no port is
specified. This commit fixes the logic flaw.
If you find this message overly verbose, relax, it's probably
not meant for you. This message is meant for probably only
two people in the whole world: me, or the poor schnook that
has to maintain this code because I'm either dead or unavailable
at the moment.
This fix solves two reports, both having to do with embedding
a function call in a ${} construct. It was tricky because the
funccall syntax has parenthesis () in it. And up till now,
the 'word' token in the flex stuff didn't allow that, because
it would tend to steal the LP and RP tokens. To be truthful,
the "word" token was the trickiest, most unstable thing in
the whole lexer. I was lucky it made this long without complaints.
I had to choose every character in the pattern with extreme
care, and I knew that someday I'd have to revisit it. Well,
the day has come.
So, my brilliant idea (and I'm being modest), was to use the
surrounding ${} construct to make a state machine and capture
everything in it, no matter what it contains. But, I have to now
treat the word token like I did with comments, in that I turn
the whole thing into a state-machine sort of spec, with new
contexts "curlystate", "wordstate", and "brackstate".
Wait a minute, "brackstate"? Yes, well, it didn't take very many
regression tests to point out if I do this for ${} constructs,
I also have to do it with the $[] constructs, too.
I had to create a separate pcbstack2 and pcbstack3 because
these constructs can occur inside macro argument lists, and
when we have two state machines operating on the same structures
we'd get problems otherwise. I guess I could have stopped at
pcbstack2 and had the brackstate stuff share it, but it doesn't
hurt to be safe. So, the pcbpush and pcbpop routines also now
have versions for "2" and "3".
I had to add the {KEYWORD} construct to the initial pattern for
"word", because previously word would match stuff like "default7",
because it was a longer match than the keyword "default". But,
not any more, because the word pattern only matches only one or
two characters now, and it will always lose. So, I made it the
winner again by making an optional match on any of the keywords
before it's normal pattern.
I added another regression test to make sure we don't
lose this in future edits, and had to fix just one regression,
where it no longer reports a 'cascaded' error, which I guess
is a plus.
I've given some thought as to whether to apply these fixes to
1.4 and the 1.6.x releases, vs trunk; I decided to put it in
1.4 because one of the bug reports was against 1.4; and it
is unexpected that AEL cannot handle this situation. It actually
reduced the amount of useless "cascade" error messages that
appeared in the regressions (by one line, ehhem). There is
a possible side-effect in that it does now do more careful
checking of what's in those ${} constructs, as far as matching
parens, and brackets are concerned. Some users may find a an
insidious problem and correct it this way. This should be
exceedingly rare, I hope.
Jeff Peeler [Thu, 4 Sep 2008 17:00:29 +0000 (17:00 +0000)]
(closes issue #11979)
Fixes multiple parking problems:
Crash when executing a park on an extension dialed by AGI due to not returning the proper return code.
Crash when using a builtin feature that was a subset of a enabled dynamic feature.
Crash due to always hanging up the peer despite the fact that the peer was supposed to be parked.
Mark Michelson [Tue, 2 Sep 2008 23:47:49 +0000 (23:47 +0000)]
After adding the context checking to app_voicemail
for IMAP storage, I left out a crucial place to
copy the context to the vm_state structure. This
is the correction.
Steve Murphy [Tue, 2 Sep 2008 23:36:56 +0000 (23:36 +0000)]
I am turning the warnings generated in ast_cdr_free and post_cdr into verbose level 2 messages. Really, they matter little to end users. You either get the CDR's you wanted, or you don't, and it is a bug.
Steve Murphy [Tue, 2 Sep 2008 22:40:13 +0000 (22:40 +0000)]
After reconsidering, with respect to 13409, ast_cdr_detach should be OK, better in fact, than ast_cdr_free, which generates lots of useless warnings that will undoubtably generate complaints.
I basically spent the day, verifying that this patch
solves the problem, and doesn't hurt in non-problem
cases. Why valgrind did not plainly reveal this leak
absolutely mystifies and stuns me.
Many, many thanks to tomaso for finding and providing the fix.
Mark Michelson [Fri, 29 Aug 2008 17:34:17 +0000 (17:34 +0000)]
After working on the ao2_containers branch, I noticed
something a bit strange. In all cases where we provide
a callback function to ao2_container_alloc, the callback
function would only return 0 or CMP_MATCH. After inspecting
the ao2_callback() code carefully, I found that if you're
only looking for one specific item, then you should return
CMP_MATCH | CMP_STOP. Otherwise, astobj2 will continue
traversing the current bucket until the end searching for
more matches.
In cases like chan_iax2 where in 1.4, all the peers are
shoved into a single bucket, this makes for potentially
terrible performance since the entire bucket will be
traversed even if the peer is one of the first ones come
across in the bucket.
All the changes I have made were for cases where the
callback function defined was passed to ao2_container_alloc
so that calls to ao2_find could find a unique instance
of whatever object was being stored in the container.
Mark Michelson [Fri, 29 Aug 2008 16:01:07 +0000 (16:01 +0000)]
Add context checking when retrieving a vm_state.
This was causing a problem for people who had identically
named mailboxes in separate voicemail contexts.
This commit affects IMAP storage only.
Mark Michelson [Wed, 27 Aug 2008 19:49:20 +0000 (19:49 +0000)]
Fix tag checking in get_sip_pvt_byid_locked when
in pedantic mode. The problem was that the wrong
tags would be compared depending on the direction
of the call.
Russell Bryant [Tue, 26 Aug 2008 16:07:58 +0000 (16:07 +0000)]
Fix some bogus scheduler usage in chan_sip. This code used the return value
of a completely unrelated function to determine whether the scheduler should
be run or not. This would have caused the scheduler to not run in cases where
it should have. Also, leave a note about another scheduler issue that needs
to be addressed at some point.
projects / thirdparty / asterisk.git / log
