Erin Spiceland [Fri, 19 Nov 2010 19:24:05 +0000 (19:24 +0000)]
Add extra functionality to AGI command WAIT FOR DIGIT.
Add the ability to play a sound file, listen for more than just one digit, specify
escape characters. Backwards compatible (to work with only timeout specified).
Jeff Peeler [Tue, 16 Nov 2010 21:29:29 +0000 (21:29 +0000)]
Ensure original message duration is preserved when prepending a message.
It seems the fix to issue 17103 was a little overzealous and removed the code
that backed up the textfile containing the original message duration. This
code has now been restored.
Jeff Peeler [Fri, 12 Nov 2010 20:49:09 +0000 (20:49 +0000)]
Fix regression causing abort in voicemail after opening a mailbox with no mesgs.
In order to be more safe, some error handling code was changed to respect more
error conditions including the potential memory allocation failure for deleted
and heard message tracking introduced in 293004. However, last_message_index
returns -1 for zero messages (perhaps as expected) and was triggering the
stricter error checking. Because last_message_index is only called directly
in one place, just return 0 from open_mailbox (for file based storage) when no
messages are detected unless a real error has occurred.
Jeff Peeler [Thu, 11 Nov 2010 21:12:27 +0000 (21:12 +0000)]
Fix problem with qualify option packets for realtime peers never stopping.
The option packets not only never stopped, but if a realtime peer was not in
the peer list multiple options dialogs could accumulate over time. This
scenario has the potential to progress to the point of saturating a link just
from options packets. The fix was to ensure that the poke scheduler checks to
see if a peer is in the peer list before continuing to poke. The reason a peer
must be in the peer list to be able to properly manage an options dialog is
because otherwise the call pointer is lost when the peer is regenerated from
the database, which is how existing qualify dialogs are detected.
Jeff Peeler [Tue, 9 Nov 2010 17:37:59 +0000 (17:37 +0000)]
Fix a deadlock in device state change processing.
Copied from some notes from the original author (Russell):
Deadlock scenario:
Thread 1: device state change thread
Holds - rdlock on contexts
Holds - hints lock
Waiting on channels container lock
Thread 2: SIP monitor thread
Holds the "iflock"
Holds a sip_pvt lock
Holds channel container lock
Waiting for a channel lock
Thread 3: A channel thread (chan_local in this case)
Holds 2 channel locks acquired within app_dial
Holds a 3rd channel lock it got inside of chan_local
Holds a local_pvt lock
Waiting on a rdlock of the contexts lock
A bunch of other threads waiting on a wrlock of the contexts lock
To address this deadlock, some locking order rules must be put in place and
enforced. Existing relevant rules:
1) channel lock before a pvt lock
2) contexts lock before hints lock
3) channels container before a channel
What's missing is some enforcement of the order when you involve more than any
two. To fix this problem, I put in some code that ensures that (at least in the
code paths involved in this bug) the locks in (3) come before the locks in (2).
To change the operation of thread 1 to comply, I converted the storage of hints
to an astobj2 container. This allows processing of hints without holding the
hints container lock. So, in the code path that led to thread 1's state, it no
longer holds either the contexts or hints lock while it attempts to lock the
channels container.
Modify our handling of 491 responses to drop any pending reinvite retry scheduler entries if we get a new 491.
This prevents a scheduler entry from leaking if we receive a 491 response when one is pending. If a scheduler entry leaks, the pvt it is associated my get destroyed before the scheduler entry fires, and then memory corruption and crashes can occur when the scheduled reinvite attempts to access and modify the memory of the destroyed pvt.
Shaun Ruffell [Fri, 5 Nov 2010 00:02:53 +0000 (00:02 +0000)]
codecs/codec_dahdi: Prevent "choppy" audio when receiving unexpected frame sizes.
dahdi-linux 2.4.0 (specifically commit 9034) added the capability for
the wctc4xxp to return more than a single packet of data in response to
a read. However, when decoding packets, codec_dahdi was still assuming
that the default number of samples was in each read.
In other words, each packet your provider sent you, regardless of size,
would result in 20 ms of decoded data (30 ms if decoding G723). If your
provider was sending 60 ms packets then codec_dahdi would end up
stripping 40 ms of data from each transcoded frame resulting in "choppy"
audio.
This would only affect systems where G729 packets are arriving in sizes
greater than 20ms or G723 packets arriving in sizes greater than 30ms.
Richard Mudgett [Wed, 3 Nov 2010 18:23:04 +0000 (18:23 +0000)]
Party A in an analog 3-way call would continue to hear ringback after party C answers.
All parties are analog FXS ports.
1) A calls B.
2) A flash hooks to call C.
3) A flash hooks to bring C into 3-way call before C answers. (A and B hear ringback)
4) C answers
5) A continues to hear ringback during the 3-way call. (All parties can hear each other.)
* Fixed use of wrong variable in dahdi_bridge() that stopped ringback on
the wrong subchannel.
* Made several debug messages have more information.
A similar issue happens if B and C are SIP channels. B continues to hear
ringback. For some reason this only affects v1.8 and trunk.
* Don't start ringback on the real and 3-way subchannels when creating the
3-way conference. Removing this code is benign on v1.6.2 and earlier.
Richard Mudgett [Tue, 2 Nov 2010 21:24:13 +0000 (21:24 +0000)]
Make warning message have more useful information in it.
Change "Unable to get index, and nullok is not asserted" to "Unable to get
index for '<channel-name>' on channel <number> (<function>(), line
<number>)".
Tilghman Lesher [Thu, 28 Oct 2010 19:44:37 +0000 (19:44 +0000)]
"!00" evaluated as false, which is incorrect. Fixing.
Reported (though the reporter did not understand he was reporting a bug) on the asterisk-users list:
http://lists.digium.com/pipermail/asterisk-users/2010-October/255505.html
Jeff Peeler [Mon, 25 Oct 2010 22:55:28 +0000 (22:55 +0000)]
Fix inprocess_container in voicemail to correctly restrict max messages.
The comparison function logic was off, so the number of sessions for a given
mailbox were not being incremented properly. This problem caused the maximum
number of messages per folder to not be respected when simultaneously leaving
multiple voicemails just below the threshold.
These problems should be fixed by the above, but just in case:
Fixed resequence_mailbox to rely on the actual number of detected number of
files in a directory rather than just assuming only 10 messages more than the
maximum had been left. Also if more messages than the maximum are deleted they
are actually removed now.
The second purpose of this commit should have been separated out probably, but
is related to the above. Again, if the number of messages in a given voicemail
folder exceeds the maximum set limit make sure to allocate enough space for the
deleted and heard index tracking array.
A few random fixes:
There was a forgotten decrement of the inprocess count in imap_store_file.
When using IMAP storage, do not look in the directory where file based storage
messages may still reside and influence the message count.
David Vossel [Mon, 25 Oct 2010 19:05:07 +0000 (19:05 +0000)]
This patch turns chan_local pvts into astobj2 objects.
chan_local does some dangerous things involving deadlock avoidance.
tech_pvt functions like hangup and queue_frame are provided with a
locked channel upon entry. Those functions are completely safe as
long as you don't attempt to give up that channel lock, but that is
impossible to guarantee due to the required deadlock avoidance necessary
to lock both the tech_pvt and both channels involved.
In the past, we have tried to account for this by doing things like
setting a "glare" flag that indicates what function should destroy the
pvt. This was used in local_hangup and local_queue_frame to decided
who should destroy the pvt if they collided in separate threads. I
have removed the need to do this by converting all chan_local tech_pvts
to astobj2. This means we can ref a pvt before deadlock avoidance
and not have to worry about that pvt possibly getting destroyed under
us. It also cleans up where we destroy the tech_pvt. The only unlink
from the tech_pvt container occurs in local_hangup now, which is where
it should occur.
Since there still may be thread collisions on some functions like
local_hangup after deadlock avoidance, I have added some checks to detect
those collisions and exit appropriately. I think this patch is going to
solve quite a bit of weirdness we have had with local channels in the past.
Jeff Peeler [Mon, 18 Oct 2010 21:50:30 +0000 (21:50 +0000)]
Fix improper operator key acceptance and clean up temp recording files.
This is a fix for when pressing the operator key after recording an unavailable,
busy, name, or temporary message in mailbox options. The operator key should not
be accepted here, but should be allowed during the message recording. If the
operator key is pressed during ensure the file is saved or deleted as
apporopriate. Also, ensure removal of temporary recorded files after an early
hang up or when message acceptance confirmation times out.
Richard Mudgett [Wed, 13 Oct 2010 23:29:58 +0000 (23:29 +0000)]
Deadlock between dahdi_exception() and dahdi_indicate().
There is a deadlock between dahdi_exception() and dahdi_indicate() for
analog ports. The call-waiting and three-way-calling feature can
experience deadlock if these features are trying to do something and an
event from the bridged channel happens at the same time.
Deadlock avoidance code added to obtain necessary channel locks before
attemting an operation with call-waiting and three-way-calling.
(closes issue #16847)
Reported by: shin-shoryuken
Patches:
issue_16847_v1.4.patch uploaded by rmudgett (license 664)
issue_16847_v1.6.2.patch uploaded by rmudgett (license 664)
issue_16847_v1.8_v2.patch uploaded by rmudgett (license 664)
Tested by: alecdavis, rmudgett
Terry Wilson [Wed, 13 Oct 2010 22:45:15 +0000 (22:45 +0000)]
Don't ignore frames that have been queued when softhangup'd
When an outgoing call is answered and hung up by the far end *very* quickly, we
may not read any frames and therefor end up with a call that displays the wrong
disposition/DIALSTATUS. The reason is because ast_queue_hangup() immediately
sets the _softhangup flag on the channel and then queues the HANGUP control
frame, but __ast_read refuses to read any frames if ast_check_hangup() indicates
that a hangup request has been made (which it will if _softhangup is set). So,
we end up losing control frames.
This change makes __ast_read continue to read frames even if a soft hangup has
been requested. It queues a hangup frame to make sure that __ast_read() will
still eventually return NULL.
Much thanks to David Vossel for all of the reviews, discussion, and help!
Jeff Peeler [Fri, 8 Oct 2010 02:35:29 +0000 (02:35 +0000)]
Ensure editline cleanup occurs when Ctrl-C is pressed at control console.
A recent change was made to avoid a race condition on shutdown which only called
the end functions from the console thread. However, when pressing Ctrl-C the
quit handler is called from the signal handler thread.
Jeff Peeler [Fri, 1 Oct 2010 22:58:38 +0000 (22:58 +0000)]
Change RFC2833 DTMF event duration on end to report actual elapsed time.
The scenario here is with a non P2P early media session. The reported time
length of DTMF presses are coming up short when sending to the remote side.
Currently the event duration is a running total that is incremented when sending
continuation packets. These continuation packets are only triggered upon
incoming media from the remote side, which means that the running total probably
is not going to end up matching the actual length of time Asterisk received
DTMF. This patch changes the end event duration to be lengthened if it is
detected that the end event is going to come up short.
Jeff Peeler [Fri, 1 Oct 2010 16:20:00 +0000 (16:20 +0000)]
Ensure user portion of SIP URI matches dialplan when using encoded characters.
This commit takes a simliar approach to 288112 and checks the dialplan to
determine the proper action for an incoming contact header as to whether or not
it should be decoded or not. sip_new was blindly always decoding the extension,
which also caused the outgoing contact header to be incorrect as well as failing
to match the encoded extension in the dialplan.
Stefan Schmidt [Fri, 1 Oct 2010 09:42:22 +0000 (09:42 +0000)]
don't iterate through all dialogs to find and delete old subscribes
On every incoming subscribe there is a iteration through all dialogs to find old subscribes and delete them. This is slow and not RFC conform. This was only needed in 1.2 cause a subscribe was not deleted when a dialog was destroyed, after 1.4 a subscribe get removed when its dialog is destroyed.
Russell Bryant [Thu, 30 Sep 2010 15:34:29 +0000 (15:34 +0000)]
Fix a crash in app_sms.
Since the data being passed to the generator callback is on the stack of the
SMS() application, we must ensure that the generator is stopped before the
application exits.
Jason Parker [Wed, 29 Sep 2010 20:56:26 +0000 (20:56 +0000)]
Allow a manager originate to succeed on forwarded devices.
The timeout to wait for an answer was being set to 0 when a device forwarded to another
extension. We don't always need the timeout set like this, so make it an optional
parameter, and don't use it in this case.
Fixes an issue with the Newchannel AMI event during the Masquerading process.
Fixes an issue with the Newchannel AMI event during the Masquerading process,
where no Newchannel AMI event was generated for the psuedo channel used during
the masquerading process.
Terry Wilson [Wed, 22 Sep 2010 23:00:30 +0000 (23:00 +0000)]
Don't let a Local channel get bridged to itself
If a local channel gets bridged to itself, it becomes orphaned with no devices
left to actually tell it to hang up. This patch modifies local_fixup() to detect
this case and deny it.
Russell Bryant [Wed, 22 Sep 2010 16:39:16 +0000 (16:39 +0000)]
Fix a 100% CPU consumption problem when setting console=yes in asterisk.conf.
The handling of -c and console=yes should be the same, but they were not.
When you specify -c, it sets both a flag for console module and for asterisk
not to fork() off into the background. The handling of console=yes only set
console mode, so you would end up with a background process() trying to run
the Asterisk console and freaking out since it didn't have anything to read
input from.
Thanks to beagles for reporting and helping debug the problem!
Richard Mudgett [Tue, 21 Sep 2010 23:55:58 +0000 (23:55 +0000)]
In chan_iax2.c:schedule_delivery() calls ast_bridged_channel() on an unlocked channel.
Near the beginning of schedule_delivery(), ast_bridged_channel() is called
on iaxs[fr->callno]->owner. However, the channel is not locked, which can
result in ast_bridged_channel() crashing should owner->tech change to a
technology that doesn't implement bridged_channel.
I also fixed the other calls to ast_bridged_channel() in chan_iax2.c since
the owner lock was not held there either.
Converted the existing channel deadlock avoidance to use
iax2_lock_owner(). Using the new function simplified some awkward code.
In the process of fixing the locking on ast_bridged_channel(), I also
found a memory leak in socket_process() for v1.6.2 and v1.8. The local
struct variable ies.vars is not freed on early/abnormal function exits.
(closes issue #17919)
Reported by: rain
Patches:
issue17919_v1.4.patch uploaded by rmudgett (license 664)
issue17919_w_leak_v1.6.2.patch uploaded by rmudgett (license 664)
issue17919_w_leak_v1.8.patch uploaded by rmudgett (license 664)
Try both the encoded and unencoded subscription URI for a match in hints.
When a phone sends an encoded URI for a subscription, the URI is not matched
with the actual hint that is in decoded format. For example, if we have an
extension with a hint that is named: "#5601" or "*5601", the subscription will
work fine if the phone subscribes with an already decoded URI, but when it's
decoded like "%255601" or "%2A5601", Asterisk is unable to match it with the
correct hint.
Fix misvalidation of meetme pins in conjunction with the 'a' MeetMe flag.
When using the 'a' MeetMe flag and having a user and admin pin setup for your
conference, using the user pin would gain you admin priviledges. Also, when no
user pin was set, an admin pin was, the 'a' MeetMe flag wasn't used, and the
user tried to enter a conference then they were still prompted for a pin and
forced to hit #.
Check all 4 combinations of (original/clonechan) * (masq/masqr).
Initially original->masq and clonechan->masqr were only checked.
It's possible with multiple masq's planned - and not yet executed, that
the 'original' chan could already have another masq'd into it - thus original->masqr
would be set, that masqr would lost.
Likewise for the clonechan->masq.
Only drop duplicate answer frames if the channel is bridged.
Back in r3710 ast_read() was modified to drop answer frames on channels that were in the UP state. This modification prevented bridges that were up before the answer from being broken and reestablished by an ANSWER control frame. That change also prevents pickup of channels called from the ast_dial framework from working properly. The ast_dial framework expects to see an ANSWER frame after dialing and the pickup code queues one but ast_read() drops it. This new change only drops ANSWER frames when the channel is bridged, allowing the answer queued by the pickup code to properly pass through ast_read() on to the ast_dial framework.
Richard Mudgett [Fri, 10 Sep 2010 20:33:16 +0000 (20:33 +0000)]
An outgoing call may not get hung up if a pre-connect incoming ISDN call is disconnected.
If the ISDN link a pre-connect incoming call is using fails or is reset,
the outgoing leg may not hang up or be delayed in hanging up. (Causes:
PRI_CAUSE_NETWORK_OUT_OF_ORDER, PRI_CAUSE_DESTINATION_OUT_OF_ORDER, and
PRI_CAUSE_NORMAL_TEMPORARY_FAILURE.)
Just hang up the call if the incoming call leg hangs up before connecting
for any reason. It makes no sense to send a BUSY or CONGESTION control
frame to the outgoing call leg under these circumstances.
David Vossel [Fri, 10 Sep 2010 20:03:50 +0000 (20:03 +0000)]
Fixes sip extension state update DEADLOCK
PROBLEM:
In chan_sip, and all the other channel drivers, it is common for
us to hold the tech_pvt lock while we ask the Asterisk core about
an extension and context. Every time we do this the locking
order becomes, (1. tech_pvt lock ---> 2. global context lock). In
chan_sip when a dialog subscribes to a hint, that locking order
is reversed in the extensionstate callback which will occur outside
of the channel_driver's monitor loop. So, on an extension state
update we have (1. global context lock ----> 2. tech_pvt lock).
Typically when we have to do a reversed locking order like this
we'd just do some sort of deadlock avoidance to fix the problem...
That will not work here. There are more locks involved here than
just the context and tech_pvt. Those are the two that are colliding,
but it is impossible to give up the context lock because the global
hints list lock MUST be held as well and we can not give that lock
up during the extensionstate callback traversal... The locking order
for the context and hints are (1. global context lock ----> 2.
hints list lock). Deadlock avoidance is not an option here.
SOLUTION:
The solution this patch implements is to queue the extension state updates
into a list and send the NOTIFY messages out during the do_monitor pvt
traversal. This clears out the problem of having to hold the context
lock before the tech_pvt lock entirely.
Terry Wilson [Fri, 10 Sep 2010 19:25:08 +0000 (19:25 +0000)]
Inherit CHANNEL() writes to both sides of a Local channel
Having Local (/n) channels as queue members and setting the language in the
extension with Set(CHANNEL(language)=fr) sets the language on the Local/...,2
channel. Hold time report playbacks happen on the Local/...,1 channel and
therefor do not play in the specified language.
This patch modifies func_channel_write to call the setoption callback and pass
the CHANNEL() write info to the callback. chan_local uses this information to
look up the other side of the channel and apply the same changes to it.
This also fixes a rather grievous calculation error for the offset of
ast_fdset, which was masked on Linux and FreeBSD, because these platforms
check the first 256 FDs regardless of the bitmask setting (due to backwards
compatibility).
Fixes an issue with MOH where it doesn't recover cleanly when it can't play a file and would just stop, instead of continuing to find the next playable file in the MOH class.
Ensure that all areas that previously used select(2) now use poll(2), with implementations that need poll(2) implemented with select(2) safe against 1024-bit overflows.
This is a followup to the fix for the pthread timer in 1.6.2 and beyond, fixing
a potential crash bug in all supported releases.
(closes issue #17678)
Reported by: russell
Branch: https://origsvn.digium.com/svn/asterisk/team/tilghman/ast_select
projects / thirdparty / asterisk.git / log
