Russell Bryant [Wed, 22 Aug 2007 22:40:27 +0000 (22:40 +0000)]
When converting this code to use the list macros, I changed it so objects are
added to the head of a bucket instead of the tail. However, while looking over
code with mmichelson, we noticed that the algorithm used in ao2_iterator_next
requires that items are added to the tail. This wouldn't have caused any huge
problem, but it wasn't correct. It meant that if an object was added to a
container while you were iterating it, and it was added to the same bucket that
the current element is in, then the new object would be returned by
ao2_iterator_next, and any other objects in the bucket would be bypassed in
the traversal.
Russell Bryant [Wed, 22 Aug 2007 20:21:36 +0000 (20:21 +0000)]
Merge changes from team/russell/iax_refcount.
This set of changes fixes problems with the handling of iax2_user and iax2_peer
objects. It was very possible for a thread to still hold a reference to one of
these objects while a reload operation tries to delete them. The fix here is to
ensure that all references to these objects are tracked so that they can't go away
while still in use.
To accomplish this, I used the astobj2 reference counted object model. This
code has been in one of Luigi Rizzo's branches for a long time and was primarily
developed by one of his students, Marta Carbone. I wanted to go ahead and bring
this in to 1.4 because there are other problems similar to the ones fixed by these
changes, so we might as well go ahead and use the new astobj if we're going to go
through all of the work necessary to fix the problems.
As a nice side benefit of these changes, peer and user handling got more efficient.
Using astobj2 lets us not hold the container lock for peers or users nearly as long
while iterating. Also, by changing a define at the top of chan_iax2.c, the objects
will be distributed in a hash table, drastically increasing lookup speed in these
containers, which will have a very big impact on systems that have a large number of
users or peers.
The use of the hash table will be made the default in trunk. It is not the default
in 1.4 because it changes the behavior slightly. Previously, since peers and users
were stored in memory in the same order they were specified in the configuration file,
you could influence peer and user matching order based on the order they are specified
in the configuration. The hash table does not guarantee any order in the container,
so this behavior will be going away. It just means that you have to be a little
more careful ensuring that peers and users are matched explicitly and not forcing
chan_iax2 to have to guess which user is the right one based on secret, host, and
access list settings, instead of simply using the username.
If you have any questions, feel free to ask on the asterisk-dev list.
Russell Bryant [Wed, 22 Aug 2007 19:53:30 +0000 (19:53 +0000)]
Juggie in #asterisk-dev was reporting problems where fgets would return
without reading the whole line when using fastagi. When this happens,
errno was set to EINTR or EAGAIN. This patch accounts for the possibility
and lets fgets continue in that case.
Russell Bryant [Wed, 22 Aug 2007 16:21:58 +0000 (16:21 +0000)]
Honor the contents of the COPTS variable as custom target CFLAGS. Apparently
this is what openwrt does.
(reported by Brian Capouch on the asterisk-dev list, patch by me)
Russell Bryant [Tue, 21 Aug 2007 18:42:15 +0000 (18:42 +0000)]
Don't record SIP dialog history if it's not turned on. Also, put an upper
limit on how many history entires will be stored for each SIP dialog. It is
currently set to 50, but can be increased if deemed necessary.
(closes issue #10421, closes issue #10418, patches suggested by jmoldenhauer,
patches updated by me)
(Security implications documented in AST-2007-020)
Steve Murphy [Tue, 21 Aug 2007 16:36:34 +0000 (16:36 +0000)]
This patch solves problem 1 in 8126; it should not slow down the alaw codec, but should prevent signal degradation via multiple trips thru the codec. Fossil estimates the twice thru this codec will prevent fax from working. 4-6 times thru would result hearable, noticeable, voice degradation.
Mark Michelson [Mon, 20 Aug 2007 16:17:43 +0000 (16:17 +0000)]
Found a pointless ternary if. member->dynamic was set to 1 and has no opportunity to change
between then and this line, so "dynamic" will ALWAYS be output.
Tilghman Lesher [Sat, 18 Aug 2007 14:30:44 +0000 (14:30 +0000)]
Don't allocate vmu for messagecount when we could just use the stack instead (closes issue #10490)
Also, remove a useless (and leaky) SQLAllocHandle (closes issue #10480)
Russell Bryant [Fri, 17 Aug 2007 21:01:43 +0000 (21:01 +0000)]
Avoid a crash in the handling of DTMF based Caller ID. It is valid for
ast_read to return NULL in the case that the channel has been hung up.
(crash reported by anonymouz666 on IRC in #asterisk-dev)
Mark Michelson [Fri, 17 Aug 2007 19:14:05 +0000 (19:14 +0000)]
Patch allows for more seamless transition from file storage voicemail to ODBC storage voicemail.
If a retrieval of a greeting from the database fails, but the file is found on the file system, then
we go ahead an insert the greeting into the database. The result of this is that people who
switch from file storage to ODBC storage do not need to rerecord their voicemail greetings.
Russell Bryant [Fri, 17 Aug 2007 13:37:08 +0000 (13:37 +0000)]
Fix some crashes in chan_sip. This patch changes various places that add items
to the scheduler to ensure that they don't overwrite the ID of a previously
scheduled item. If there is one, it should be removed.
(closes issue #10391, closes issue #10256, probably others, patch by me)
sometimes we don't need to signal dtmf tones to asterisk, we just want them to go through as inband. Otherwise they might be generated by the other channel partner and then there is a double tone.
Russell Bryant [Thu, 16 Aug 2007 22:32:33 +0000 (22:32 +0000)]
Fix a little race condition that could cause a crash if two channels had MOH
stopped at the same time that were using a class that had been marked for
deletion when its use count hits zero.
Russell Bryant [Thu, 16 Aug 2007 22:24:25 +0000 (22:24 +0000)]
This patch fixes a bug where reloading the module with "module reload" did not
delete classes from memory that were no longer in the config. This patch fixes
that problem as well as another one. Previously, if you reloaded MOH using the
"moh reload" CLI command, which behaved differently than "module reload ...",
MOH had to be stopped on every channel and started again immediately. However,
there was no way to tell what class was being used, so they would all fall back
to the default class.
Russell Bryant [Thu, 16 Aug 2007 21:29:24 +0000 (21:29 +0000)]
Fix more deadlocks in chan_iax2 that were introduced by making frame handling
and scheduling multi-threaded. Unfortunately, we have to do some expensive
deadlock avoidance when queueing frames on to the ast_channel owner of the IAX2
pvt struct. This was already handled for regular frames, but ast_queue_hangup
and ast_queue_control were still used directly. Making these changes introduced
even more places where the IAX2 pvt struct can disappear in the context of a
function holding its lock due to calling a function that has to unlock/lock it
to avoid deadlocks. I went through and fixed all of these places to account for
this possibility.
(issue #10362, patch by me)
Philippe Sultan [Thu, 16 Aug 2007 09:37:10 +0000 (09:37 +0000)]
A fix for two critical problems detected while working with Daniel
McKeehan in issue #10184.
Upon priority change, the resource list is not NULL terminated when
moving an item to the end of the list. This makes Asterisk endlessy
loop whenever it needs to read the list. Jids with different resource and
priority values, like in Gmail's and GoogleTalk's jabber clients put
that problem in evidence.
Upon reception of a 'from' attribute with an empty resource string,
Asterisk crashes when trying to access the found->cap pointer if the
resource list for the given buddy is not empty. This situation is
perfectly valid and must be handled. The Gizmoproject's jabber client
put that problem in evidence.
Also added a few comments in the code as well as a handle for the
capabilities from Gmail's jabber client, which are stored in a caps:c tag
rather than the usual c tag.
Joshua Colp [Wed, 15 Aug 2007 14:40:23 +0000 (14:40 +0000)]
(closes issue #10440)
Reported by: irroot
(closes issue #10454)
Reported by: flo_turc
Increase maximum timestamp skew to 120. 20 was apparently far too low.
Joshua Colp [Wed, 15 Aug 2007 14:18:44 +0000 (14:18 +0000)]
(closes issue #10456)
Reported by: irroot
Patches:
sip_timeout.patch uploaded by irroot (license 52)
Change hardcoded timer value to defined value. I'm doing this in 1.4 as well so if it needs to be changed in the future this place would not have been forgotten.
Russell Bryant [Mon, 13 Aug 2007 20:18:30 +0000 (20:18 +0000)]
Release the pvt lock before calling find_peer in register_verify to avoid a
deadlock. Also, remove some unnecessary locking in auth_fail that was only done
recursively.
Russell Bryant [Mon, 13 Aug 2007 19:27:39 +0000 (19:27 +0000)]
I am fighting deadlocks in chan_iax2. I have tracked them down to a single
core issue. You can not call find_callno() while holding a pvt lock as this
function has to lock another (every) other pvt lock. Doing so can lead to a
classic deadlock. So, I am tracking down all of the code paths where this
can happen and fixing them.
The fix I committed earlier today was along the same theme. This patch fixes
some code down the path of authenticate_reply.
Russell Bryant [Mon, 13 Aug 2007 15:28:13 +0000 (15:28 +0000)]
Fix a potential deadlock in socket_process. check_provisioning can eventually
call find_callno. You can't hold a pvt lock while calling find_callno because
it goes through and locks every single one looking for a match.
Joshua Colp [Mon, 13 Aug 2007 14:18:04 +0000 (14:18 +0000)]
(closes issue #10437)
Reported by: haklin
Don't set the callerid name and number a second time on a newly created channel. ast_channel_alloc itself already sets it and setting it twice would cause a memory leak.
Russell Bryant [Fri, 10 Aug 2007 15:20:09 +0000 (15:20 +0000)]
The last set of changes that I made to "core show locks" made it not able to
track mutexes unless they were declared using AST_MUTEX_DEFINE_STATIC. Locks
initialized with ast_mutex_init() were not tracked. It should work now.
Joshua Colp [Fri, 10 Aug 2007 13:49:19 +0000 (13:49 +0000)]
(closes issue #10422)
Reported by: bhowell
Add note to sample configuration about module load order and how it can cause perfectly good queue members to be marked as invalid.
Mark Michelson [Thu, 9 Aug 2007 23:47:00 +0000 (23:47 +0000)]
Improved a bit of logic regarding comma-separated mailboxes in has_voicemail. Also added some braces to some compound if statements
since unbraced if statements scare me in general.
Mark Michelson [Thu, 9 Aug 2007 21:51:17 +0000 (21:51 +0000)]
Quite a few changes regarding IMAP storage.
1. instead of using inboxcount as the core message counting function, we use messagecount instead. This makes it possible to count messages in folders besides just INBOX and Old.
2. inboxcount and hasvoicemail now use messagecount as their means of determining return values.
3. Added a copy_message function for IMAP storage. Unfortunately I don't have the means to test it, but it seems like a pretty straightforward function.
4. Removed a #ifndef IMAP_STORAGE and matching #endif from leave_voicemail for a couple of reasons. One, we want to support copying mail to multiple IMAP boxes, and two, IMAP was
broken because a STORE macro had been moved into this section of code.
Mark Michelson [Thu, 9 Aug 2007 19:52:43 +0000 (19:52 +0000)]
I broke canreinvite...Now I'm fixing it. I put some new code in the wrong place and so I've reverted the canreinvite section to how it was and put my new code where it should be.
Russell Bryant [Thu, 9 Aug 2007 17:24:40 +0000 (17:24 +0000)]
Fix subscriptions to multiple mailboxes for ODBC_STORAGE. Also, leave a
comment for this to be fixed for IMAP_STORAGE, as well. I left IMAP alone
since I know MarkM was working on this code right now for another reason.
This is broken even worse in trunk, but for a different reason. The fact
that the mailbox option supported multiple mailboxes is completely not obvious
from the code in the channel drivers. Anyway, I will fix that in another
commit ...
Russell Bryant [Thu, 9 Aug 2007 16:12:57 +0000 (16:12 +0000)]
Fix a problem with the combination of the 'F' option to pass DTMF through a
conference and options that use DTMF to activate various features. The problem
was that the BEGIN frame would be passed through, but the END frame would get
intercepted to activate a feature. Then, the other conference members would hear
DTMF for forever, which they didn't seem to like very much.
(closes issue #10400, reported by stevefeinstein, fixed by me)
Mark Michelson [Wed, 8 Aug 2007 18:16:49 +0000 (18:16 +0000)]
Fixed some compiler warnings so that compiling with dev-mode and IMAP storage would not have any errors.
This section of code may get changed again shortly since my change uncovers a rather silly bit of logic.
Mark Michelson [Wed, 8 Aug 2007 14:26:36 +0000 (14:26 +0000)]
Changing a bit of logic so that someone will NEVER exit the queue on timeout unless they have enabled the 'n' option.
This commit relates to issue #10320. Thanks to jfitzgibbon for detailing the idea behind this code change.
Russell Bryant [Mon, 6 Aug 2007 20:44:09 +0000 (20:44 +0000)]
Fix an issue where dynamic threads can get free'd, but still exist in the
dynamic thread list.
(closes issue #10392, patch from Mihai, with credit to his colleague, Pete)
Russell Bryant [Mon, 6 Aug 2007 16:50:54 +0000 (16:50 +0000)]
Fix the return value of AST_LIST_REMOVE(). This shouldn't be causing any
problems, though, because the only code that uses the return value only checks
to see if it is NULL.
(closes issue #10390, pointed out by mihai)
Joshua Colp [Mon, 6 Aug 2007 16:32:44 +0000 (16:32 +0000)]
It is possible for a transfer to occur before the remote device has our tag in which case they send none in the transfer. In this case we need to not fail the transfer dialog lookup.
Joshua Colp [Mon, 6 Aug 2007 15:27:24 +0000 (15:27 +0000)]
(closes issue #10355)
Reported by: wdecarne
Now that we pass through RTP timestamp information we need to make the allowed timestamp skew considerably less. There are situations where a source may change and due to the timestamp difference the receiver will experience an audio gap since we did not indicate by setting the marker bit that the source changed.
Mark Michelson [Fri, 3 Aug 2007 20:25:22 +0000 (20:25 +0000)]
Changed the behavior of sip's realtime_peer function to match the corresponding way of matching for non-realtime peers.
Now matches are made on both the IP address and port number, or if the insecure setting is set to "port" then just match on the
IP address.
In order to accomplish this, I also added a new API call, ast_category_root, which returns the first variable of an ast_category struct
Fix a problem when you call Voicemail() with multiple mailboxes specified and
ODBC_STORAGE is in use. The audio part of the message was only given to the
first mailbox specified.
Russell Bryant [Fri, 3 Aug 2007 19:39:49 +0000 (19:39 +0000)]
Add some improvements to lock debugging. These changes take effect
with DEBUG_THREADS enabled and provide the following:
* This will keep track of which locks are held by which thread as well as
which lock a thread is waiting for in a thread-local data structure. A
reference to this structure is available on the stack in the dummy_start()
function, which is the common entry point for all threads. This information
can be easily retrieved using gdb if you switch to the dummy_start() stack
frame of any thread and print the contents of the lock_info variable.
* All of the thread-local structures for keeping track of this lock information
are also stored in a list so that the information can be dumped to the CLI
using the "core show locks" CLI command. This introduces a little bit of a
performance hit as it requires additional underlying locking operations
inside of every lock/unlock on an ast_mutex. However, the benefits of
having this information available at the CLI is huge, especially considering
this is only done in DEBUG_THREADS mode. It means that in most cases where
we debug deadlocks, we no longer have to request access to the machine to
analyze the contents of ast_mutex_t structures. We can now just ask them
to get the output of "core show locks", which gives us all of the information
we needed in most cases.
I also had to make some additional changes to astmm.c to make this work when
both MALLOC_DEBUG and DEBUG_THREADS are enabled. I disabled tracking of one
of the locks in astmm.c because it gets used inside the replacement memory
allocation routines, and the lock tracking code allocates memory. This caused
infinite recursion.
Russell Bryant [Fri, 3 Aug 2007 17:01:07 +0000 (17:01 +0000)]
Only pass through HOLD and UNHOLD control frames when the mohinterpret option
is set to "passthrough". This was pointed out by Kevin in the middle of a
training session.
Russell Bryant [Fri, 3 Aug 2007 02:04:22 +0000 (02:04 +0000)]
Don't reuse the timespec that was set to 0 in the previous timedwait as it
will just return immediately. Also, fix some logic so the thread's lock
isn't unlocked twice in the weird case of dynamic threads getting acquired
right after a timeout.
(pointed out by SteveK)
Russell Bryant [Thu, 2 Aug 2007 19:25:14 +0000 (19:25 +0000)]
Fix the case where a dynamic thread times out waiting for something to do
during the first time it runs. This shouldn't ever happen, but we should
account for it anyway.
(pointed out by pete, who works with mihai)
Steve Murphy [Thu, 2 Aug 2007 18:21:40 +0000 (18:21 +0000)]
Merged revisions 77942 via svnmerge from
https://origsvn.digium.com/svn/asterisk/branches/1.2
........
r77942 | murf | 2007-08-02 11:56:37 -0600 (Thu, 02 Aug 2007) | 1 line
This patch hopefully solves 10141; The user is running with it, and it doesn't appear to harm asterisk's operation, and may prevent a crash. I'll store it in 1.2, as we have shut down support on 1.2, but since I developed the patch before support finished, and it might affect 1.4 and trunk, I'm going ahead with it.
........
Russell Bryant [Thu, 2 Aug 2007 18:04:43 +0000 (18:04 +0000)]
Fix another race condition in the handling of dynamic threads. If the dynamic
thread timed out waiting for something to do, but was acquired to perform an
action immediately afterwords, then wait on the condition again to give the
other thread a chance to finish setting up the data for what action this thread
should perform. Otherwise, if it immediately continues, it will perform the
wrong action.
(reported on IRC by mihai, patch by me)
(related to issue #10289)
Russell Bryant [Thu, 2 Aug 2007 16:56:04 +0000 (16:56 +0000)]
Add another sanity check to vnak_retransmit(). This check ensures that frames
that have already been marked for deletion don't get retransmitted.
(closes issue #10361, patch from mihai)
projects / thirdparty / asterisk.git / log
