Luigi Rizzo [Sat, 21 Jul 2007 01:01:10 +0000 (01:01 +0000)]
Enhance NAT support as discussed on the -dev list, i.e.:
+ extensive documentation changes both in sip.conf.sample and in the source;
+ allow "externip" and "externhost" to include a port number as well;
+ allow "bindaddr" to have a port number (making bindport unnecessary,
even though it is still present for backward compatibility);
+ introduce the new "stunaddr" parameter to specify an STUN server to
be used from the main SIP socket;
+ extend the "sip show settings" output to show all the above.
Internally:
+ change related data structures from struct in_addr to struct sockaddr_in
to store the port numbers as well;
+ reorganize ast_sip_ouraddrfor() (should also be renamed to sip_ouraddrfor()
because it is not a generic API, though it might become so if called with
a socket as an additional argument, in which case it can be moved elsewhere).
As mentioned in the documentation, media sessions still do not use STUN so the
port numbers may still be incorrect when Asterisk is behind a NAT
On passing, some of the debugging messages printing media addresses are
probably using the wrong values, but this will be checked/fixed in a
subsequent commit if needed.
Part of the following chunk in the function that handles a "sip reload" is
probably needed on previous versions as well, to avoid leaking the memory
used for the "localaddr" list:
Jason Parker [Fri, 20 Jul 2007 22:25:41 +0000 (22:25 +0000)]
Add support for default "say mode" (whether to use the "old" method or "new" method. "new" method being config file)
Add support for autocomplete of "say load" CLI command.
When using users.conf for the entries in the directory, if multiple users had the same last name, only the first user listed would be available
in the directory.
(closes issue #10200, reported by mrskippy, patched by me)
Use the define that specifies the default length of an artificially created
DTMF digit in the ast_senddigit() function. The define is set to 100ms by
default, which is the same thing that this function was using. But, using
the define lets changes take effect in this case, as well as the others where
it was already used.
(closes issue #10247)
Reported by: fkasumovic
Patches:
chan_sip.patch uploaded by fkasumovic (license #101)
Drop any peer realm authentication entries when reloading so multiple entries do not get added to the peer.
(closes issue #10246)
Reported by: fkasumovic
Patches:
res_conver.patch uploaded by fkasumovic (license #101)
Use the last occurance of . to find the extension, not the first occurance.
Luigi Rizzo [Fri, 20 Jul 2007 14:54:01 +0000 (14:54 +0000)]
Extend the 'network settings' section with indication on the
localnet settings (requires the change in SVN 76034), and also
give an indication on whether/why/how the remapping of addresses
in SIP message is done or not.
I think this is especially useful for debugging the configuration,
as the address remapping depends on a combination of at least 3
parameters (localnet, externhost, externip) and successful DNS lookup.
An example of the output of this section is below:
I leave to the community the judgement if the above info is a
useful addition for 1.4. It is not a bugfix, but it is neither a
new feature, only a useful diagnostic tool.
Note that I would like to move there also the bindaddress/port
information, in the usual addr:port format e.g.
Luigi Rizzo [Fri, 20 Jul 2007 14:38:36 +0000 (14:38 +0000)]
expose struct ast_ha so external code can do things such as printing it
(e.g. chan_sip.c in a subsequent commit).
Obviously exposing the internals of a data structure is far from ideal
(especially in a case like this where the implementation is very
inefficient and will need to be changed at some point).
On the other hand, it was also unclear what additional APIs should
we provide instead, and because exposing the stucture has no impact
on source and binary compatibility, this seemed to me the best option at
this time.
Steve Murphy [Thu, 19 Jul 2007 23:24:27 +0000 (23:24 +0000)]
After some study, thought, comparing, etc. I've backed out the previous universal mod to make ast_flags a 64 bit thing. Instead, I added a 64-bit version of ast_flags (ast_flags64), and 64-bit versions of the test-flag, set-flag, etc. macros, and an app_parse_options64 routine, and I use these in app_dial alone, to eliminate the 30-option limit it had grown to meet. There is room now for 32 more options and flags. I was heavily tempted to implement some of the other ideas that were presented, but this solution does not intro any new versions of dial, doesn't have a different API, has a minimal/zero impact on code outside of dial, and doesn't seriously (I hope) affect the code structure of dial. It's the best I can think of right now. My goal was NOT to rewrite dial. I leave that to a future, coordinated effort.
Changes in handling return values of several functions in app_queue. This all started as a fix for issue #10008
but now includes all of the following changes:
1. Simplifying the code to handle positive return values from ast API calls.
2. Removing the background_file function.
3. The fix for issue #10008
(closes issue #10008, reported and patched by dimas)
Russell Bryant [Thu, 19 Jul 2007 15:59:19 +0000 (15:59 +0000)]
(closes issue #10210, reported and patched by juggie)
This merges the trunk only part of the patches from this issue. In 1.4, res_agi
will issue a warning if you try to use DeadAGI on a channel that is not hung up.
Now, in trunk, it just plain won't let you do it.
When processing full frames, take sequence number wraparound into account when
deciding whether or not we need to request retransmissions by sending a VNAK.
This code could cause VNAKs to be sent erroneously in some cases, and to not
be sent in other cases when it should have been.
(closes issue #10237, reported and patched by mihai)
Luigi Rizzo [Thu, 19 Jul 2007 08:07:04 +0000 (08:07 +0000)]
print more of the network settings (externip, externhost etc.)
in the "sip show settings" cli output. I have put these in a
separate section, probably even bindaddr and SIP port should go
there.
There are more things to add here e.g. localnet and so on.
Russell Bryant [Wed, 18 Jul 2007 22:52:54 +0000 (22:52 +0000)]
I thought I noticed a memory leak earlier when I saw that the contents of this
list were not destroyed when the module is unloaded. However, after reading
the code related to the use of this list a lot today, I realized that it isn't
necessary. So, I have added a comment to explain why it isn't necessary.
When traversing the queue of frames for possible retransmission after
receiving a VNAK, handle sequence number wraparound so that all frames that
should be retransmitted actually do get retransmitted.
(issue #10227, reported and patched by mihai)
Steve Murphy [Wed, 18 Jul 2007 14:35:07 +0000 (14:35 +0000)]
This corrects the problem with flags and %lld formats on 64-bit machines, where uint64_t is NOT acceptable for %lld, and also works on 32-bit machines. At least, with gcc.
Ensure that when encoding the contents of an ast_frame into an iax_frame, that
the size of the destination buffer is known in the iax_frame so that code
won't write past the end of the allocated buffer when sending outgoing frames.
(ASA-2007-014)
After parsing information elements in IAX frames, set the data length to zero,
so that code later on does not think it has data to copy.
(ASA-2007-015)
In ast_pbx_run(), mark a channel as hung up after an application returned -1,
or when it runs out of extensions to execute. This is so that code can detect
that this channel has been hung up for things like making sure DeadAGI is used
on actual dead channels, and is beneficial for other things, like making sure
someone doesn't try to start spying on a channel that is about to go away.
Luigi Rizzo [Tue, 17 Jul 2007 14:32:15 +0000 (14:32 +0000)]
Introduce ast_parse_arg() , a generic function to parse strings
in a consistent way. This is meant to replace the custom code
which is repeated all over the place in the various files when
parsing config files, CLI entries and other string information.
Right now the code supports parsing int32, uint32 and sockaddr_in with
optional default values and bound checks. It contains minimal error
checking, but that can be easily extended as the need arises.
Being a new API i am introducing this only in trunk, though I believe
that once the interface has been ironed out it might become a
worthwhile addition to 1.4 as well - basically, the first time
we will need to fix a piece of argument parsing code, we might as
well bring in this change and use the new API instead.
Luigi Rizzo [Tue, 17 Jul 2007 12:46:25 +0000 (12:46 +0000)]
Initialize a variable to avoid a warning when the compiler
(and/or the optimization level) may think it is used uninitialized.
The code was indeed correct, but unfortunately the result of
some compiler checks such as -Wunused and -Wuninitialized depends
heavily on the optimization level.
For my next trick I will make it so dialplan functions no longer need to call ast_module_user_add and ast_module_user_remove. These are now called in the ast_func_read and ast_func_write functions outside of the module.
Restoring functionality from 1.2 wherein Retrydial will not exit if there is no announce file specified.
This change makes it so that if there is no announce file specified, the application will continue until finished (or caller hangs up).
If a bogus announce file is specified, then a warning message will be printed saying that the file could not be found, but execution will
still continue.
(closes issue #10186, reported by jon, patched by me)
Applications no longer need to call ast_module_user_add and ast_module_user_remove. This is now taken care of in the pbx_exec function outside of the application.
It is no longer required for each module that deals with a channel to call ast_module_user_hangup_all in it's unload function. The loader will automatically perform this action for it.
Russell Bryant [Mon, 16 Jul 2007 02:51:56 +0000 (02:51 +0000)]
Merge a bunch of doxygen updates to header files. This includes changes to
use the \retval tag for documenting return values, fixing various warnings
when generating the documentation, and various other things.
(closes issue #10203, snuffy)
Russell Bryant [Mon, 16 Jul 2007 02:44:49 +0000 (02:44 +0000)]
Cast the 2nd argument to iconv() to a void *, as some systems define it as a
(const char *), while others define it as (char *). This is done to suppress
compiler warnings about it.
Fixed an issue where chanspy flags were uninitialized if no options were passed.
What triggered this investigation was an IRC chat where some people's quiet flags were
set while others' weren't even though none of them had specified the q option.
Ensure that adding a user to the list of users of a specific music on hold
class is not done at the same time as any of the other operations on this list
to prevent list corruption. Using the global moh_data lock for this is not
ideal, but it is what is used to protect these lists everywhere else in the
module, and I am only changing what is necessary to fix the bug.
When using a TDM400P (and probably other analog cards) there was a chance that
you could hang up and pick the phone back up where it has been long enough to
be not considered a flash hook, but too soon such that the device reports that
it is busy and the person on the phone will only hear silence. This patch
makes chan_zap more tolerant of this and gives the device a couple of seconds
to succeed so the person on the phone happily gets their dialtone.
which can be used to send an STUN request on a socket, and
optionally wait for a reply and store the STUN_MAPPED_ADDRESS
into the 'answer' argument (obviously, the version that
waits for a reply is blocking, but this is no different
from DNS resolutions).
Internally there are minor modifications to let stun_handle_packet()
be somewhat configurable on how to parse the body of responses.
At the moment i am not committing any change to the clients,
but adding STUN client support is extremely simple, e.g. chan_sip.c
could do something like this:
+ add a variable to store the stun server address;
+ add code to parse a config file of the form "stunaddr=my.stun.server.org:3478"
(not shown for brevity);
+ right after binding the main sip socket, talk to the stun server to
determine the externally visible address
if (stunaddr.sin_addr.s_addr != 0)
ast_stun_request(sipsock, &stunaddr, NULL, &externip);
so now 'externip' is set with the externally visible address.
so it is really trivial.
Similarly ast_stun_request could be called when creating the RTP
socket (possibly adding a struct sockaddr_in field in the struct
ast_rtp to store the externalip).
Luigi Rizzo [Thu, 12 Jul 2007 16:21:12 +0000 (16:21 +0000)]
more cleanup, this time to stun_handle_packet(). Among other things:
+ mark a potentially dangerous write-past-end-of-buffer
+ localize some variables in the block generating stun replies.
Luigi Rizzo [Thu, 12 Jul 2007 15:42:56 +0000 (15:42 +0000)]
a little bit of code cleanup to rtp.c, mostly to function
ast_rtp_new_with_bindaddr():
1. add comments to the logic of the main loop;
2. use a common exit point on failure so the cleanup is done only in one place;
3. handle failures in rtp_socket() in the main loop of the function;
No functional changes except for #3 above, so it is not yet
worthwhile merging this and other changes to 1.4
Once the cleanup work on this file will be complete (which among
other things should include some extensions to the stun support)
it might be a good thing to push all the changes to 1.4
The function make_trunk() can fail and return -1 instead of a valid new call
number. Fix the uses of this function to handle this instead of treating it
as the new call number. This would cause a deadlock and memory corruption.
(possible cause of issue #9614 and others, patch by me)
Make sure that the ESCAPE immediately follows the condition that uses LIKE.
This fixes realtime extensions with ODBC.
(closes issue #10175, reported by stuarth, patch by me)
projects / thirdparty / asterisk.git / log
