This prevents use-after-scope issues when unwinding the stack,
which happens in reverse order. The varname variable needs to
remain alive for the destruction to be able to access it.
Issue was found using clang + address-sanitizer.
Richard Mudgett [Wed, 19 Dec 2018 19:02:35 +0000 (13:02 -0600)]
stasic.c: Fix printf format type mismatches with arguments.
An int64_t is not likely the same size as a long.
* Changed the int64_t values in the statistics structs to longs so casting
is not necessary when generating the formatted CLI output. The offending
members did not need to be int64_t anyway as they were only set by an int
type variable which was already truncating bits.
* Reordered the statistics structs to reduce potential padding bytes.
George Joseph [Mon, 24 Dec 2018 17:42:36 +0000 (10:42 -0700)]
ast_coredumper: Refactor the pid determination process
In order to get a dump of the running process, we need to find the
pid of the main asterisk process. This can be tricky if there are
also instances of "asterisk -r" running or if an alternate location
for asterisk.conf was specified on the command line with the -C
option that also specified an alternation location for the pid file.
So now...
1. We find the asterisk executable with "which" or the --asterisk-bin
command line option.
2. If there's only 1 process with an executable path that matches,
we use that pid. If not...
3. We try "<asterisk-bin> -rx 'core show settings'" and parse the
output to find the pidfile, then read that for the pid. If that
didn't work...
4. We get a list of all the pids matching <asterisk-bin> and look
in /proc/<pid>/cmdline for a -C argument and retry the "core show
settings" using the same -C option. We can't parse the output
of "ps" to get the -C path because it may contain spaces. The
contents of /proc/<pid>/cmdline are delimited by NULLs. For BSDs
we may have to mount /proc first. :(
Richard Mudgett [Wed, 19 Dec 2018 18:39:08 +0000 (12:39 -0600)]
backtrace.c: Fix casting pointer to/from integral type.
The backtrace library bfd.h include file does not get the sizes of
pointers and ints right on some platforms. On my old test box the size
of bfd_vma is 8 while the size of a pointer is 4. gcc on the box
complains of the integer casting to/from pointers size mismatch.
* uintptr_t to the rescue by doing an appropriate two stage cast.
George Joseph [Tue, 18 Dec 2018 16:33:50 +0000 (09:33 -0700)]
app_voicemail: Don't delete mailbox state unless mailbox is deleted
The free_user function was automatically deleting the stasis mailbox
state but this only makes sense when the mailbox is actually
deleted, not just the structure freed. This was causing issues
where leave_voicemail would publish the mwi message to stasis and
delete the state before the message could be processed by
res_pjsip_mwi.
* Removed the delete of state from free_user().
* Created a new free_user_final() function that both frees the data
structure and deletes the state. This function is only called
during module load/unload where it's appropriate to delete the
state.
Sean Bright [Thu, 13 Dec 2018 21:56:50 +0000 (16:56 -0500)]
res_format_attr_h264.c: Make sure profile-level-id fmtp attribute is set
The profile-iop octet (the 2nd) of profile-level-id can be zero
according to RFC 6184 Section 8.1. So we ignore its value when deciding
to include profile-level-id in the outgoing SDP.
Joshua C. Colp [Fri, 30 Nov 2018 11:40:40 +0000 (07:40 -0400)]
stasis: Add statistics gathering in developer mode.
This change adds statistics gathering to Stasis topics,
subscriptions, and message types. These can be viewed using
CLI commands and provide insight into how Stasis is used
and how long certain operations take to execute.
These are only available when Asterisk is compiled in
developer mode and do not have any impact under normal
operation.
Sebastian Damm [Thu, 6 Dec 2018 17:23:50 +0000 (18:23 +0100)]
res/res_ari: Add additional hangup reasons
The ARI DELETE /channels command takes a "reason" parameter
Previously, there were only five reasons implemented
This patch adds more reasons to choose from for more
complex setups
Sean Bright [Fri, 7 Dec 2018 12:57:48 +0000 (07:57 -0500)]
utils: Wrap socket() and pipe() to reduce syscalls
Some platforms provide an implementation of socket() and pipe2() that allow the
caller to specify that the resulting file descriptors should be non-blocking.
Using these allows us to potentially elide 3 calls into 1 by avoiding extraneous
calls to fcntl() to set the O_NONBLOCK flag afterwards.
In passing, change ast_alertpipe_init() to use pipe2() directly instead of the
wrapper if it is available.
George Joseph [Thu, 29 Nov 2018 15:53:51 +0000 (08:53 -0700)]
stasis: Allow filtering by formatter
A subscriber can now indicate that it only wants messages
that have formatters of a specific type. For instance,
manager can indicate that it only wants messages that have a
"to_ami" formatter. You can combine this with the existing
filter for message type to get only messages with specific
formatters or messages of specific types.
George Joseph [Wed, 5 Dec 2018 15:37:45 +0000 (08:37 -0700)]
CI: Various updates to buildAsterisk.sh
* Added ---no-configure, --no-menuselect, --no-make and --no-alembic
options that prevent those actions from being performed. Useful
for testing and re-running portions of the build after fixing
earlier failures.
* Added "set -e" to abort the script on command failure.
Not sure why this wasn't there in the first place.
* Fixed a few echos that were redirecting to stderr when they shouldn't
have been.
* Catch more alembic failures by actually trying to generate the SQL.
WARNING[5812]: http.c:1939 httpd_helper_thread: Failed to set
TCP_NODELAY on HTTP connection: Bad file descriptor
ERROR[5812]: iostream.c:91 ast_iostream_nonblock: Failed to get
fcntl() flags for file descriptor: Bad file descriptor
ERROR[5812]: iostream.c:569 ast_iostream_close: close() failed: Bad
file descriptor
Disabled for now by making the test explicit only.
Pirmin Walthert [Wed, 28 Nov 2018 07:14:12 +0000 (08:14 +0100)]
pjproject_bundled: check whether UPDATE is supported on outgoing calls
In ASTERISK-27095 an issue had been fixed because of which chan_pjsip was not
trying to send UPDATE messages when connected_line_method was set to invite.
However this only solved the issue for incoming INVITES. For outgoing INVITES
(important when transferring calls) the options variable needs to be updated
at a different place.
That commit closed a long standing hole which allowed subscriptions
to mailboxes that weren't configured in voicemail.conf. This
caused an issue with FreePBX which depdended on that behavior.
The commit is being reverted until FreePBX can handle the new
behavior.
George Joseph [Wed, 19 Sep 2018 19:34:41 +0000 (13:34 -0600)]
bridges: Remove reliance on stasis caching
* The bridging core no longer uses the stasis cache for bridge
snapshots. The latest bridge snapshot is now stored on the
ast_bridge structure itself.
* The following APIs are no longer available since the stasis cache
is no longer used:
ast_bridge_topic_cached()
ast_bridge_topic_all_cached()
* A topic pool is now used for individual bridge topics.
* The ast_bridge_cache() function was removed since there's no
longer a separate container of snapshots.
* A new function "ast_bridges()" was created to retrieve the
container of all bridges. Users formerly calling
ast_bridge_cache() can use the new function to iterate over
bridges and retrieve the latest snapshot directly from the
bridge.
* The ast_bridge_snapshot_get_latest() function was renamed to
ast_bridge_get_snapshot_by_uniqueid().
* A new function "ast_bridge_get_snapshot()" was created to retrieve
the bridge snapshot directly from the bridge structure.
* The ast_bridge_topic_all() function now returns a normal topic
not a cached one so you can't use stasis cache functions on it
either.
* The ast_bridge_snapshot_type() stasis message now has the
ast_bridge_snapshot_update structure as it's data. It contains
the last snapshot and the new one.
* cdr, cel, manager and ari have been updated to use the new
arrangement.
Joshua Colp [Wed, 7 Nov 2018 17:18:34 +0000 (13:18 -0400)]
stasis: Segment channel snapshot to reduce creation cost.
When a channel snapshot was created it used to be done
from scratch, copying all data (many strings). This incurs
a cost when doing so.
This change segments the channel snapshot into different
components which can be reused if unchanged from the
previous snapshot creation, reducing the cost. In normal
cases this results in some pointers being copied with
reference count being bumped, some integers being set,
and a string or two copied. The other benefit is that it
is now possible to determine if a channel snapshot update
is redundant and thus stop it before a message is published
to stasis.
The specific segments in the channel snapshot were split up
based on whether they are changed together, how often they
are changed, and their general grouping. In practice only
1 (or 0) of the segments actually get changed in normal
operation.
Invalidation is done by setting a flag on the channel when
the segment source is changed, forcing creation of a new
segment when the channel snapshot is created.
Joshua Colp [Wed, 10 Oct 2018 14:28:18 +0000 (11:28 -0300)]
stasis: Use an implementation specific channel snapshot cache.
Channels no longer use the Stasis cache for channel snapshots. Instead
they are stored in a hash table in stasis_channels which reduces the
number of Stasis messages created and allows better storage.
As a result the following APIs are no longer available since the stasis
cache is no longer used:
ast_channel_topic_cached()
ast_channel_topic_all_cached()
The ast_channel_cache_all() and ast_channel_cache_by_name() functions
now return an ao2_container of ast_channel_snapshots rather than
a container of stasis_messages therefore you can't (and don't need
to) call stasis_cache functions on it.
The ast_channel_topic_all() function now returns a normal topic not
a cached one so you can't use stasis cache functions on it either.
The ast_channel_snapshot_type() stasis message now has the
ast_channel_snapshot_update structure as it's data. It contains the
last snapshot and the new one.
ast_channel_snapshot_get_latest() still returns the latest snapshot.
The latest snapshot is now stored on the channel itself to eliminate
cache hits when Stasis messages that have the snapshot as a payload
are created.
Corey Farrell [Mon, 26 Nov 2018 12:09:11 +0000 (07:09 -0500)]
jansson: Upgrade to 2.12.
This brings in jansson-2.12, removes all patches that were merged
upstream. README is created in third-party/jansson/patches to explain
how to add patches but also because the patches folder must exist for
the build process to succeed.
Alexei Gradinari [Fri, 23 Nov 2018 15:40:50 +0000 (10:40 -0500)]
RTP: need to reset DTMF last seqno/timestamp on voice packet with marker bit
The marker bit set on the voice packet indicates the start
of a new stream and a new time stamp.
Need to reset the DTMF last sequence number and the timestamp
of the last END packet.
If the new time stamp is lower then the timestamp of the last DTMF END packet
the asterisk drops all DTMF frames as out of order.
This bug was caught using Cisco ip-phone SPA50X and codec g722.
On SIP session update the SPA50X resets stream indicating it with market bit
and a new timestamp is twice smaller then the previous.
Corey Farrell [Wed, 14 Nov 2018 11:02:20 +0000 (06:02 -0500)]
astobj2: Create function to copy weak proxied objects from container.
Create ao2_container_dup_weakproxy_objs to perform a similar function to
ao2_container_dup. This function expects the source container to have
weakproxy objects, inserts the associated non-weak objects into the
destination container. Orphaned weakproxy objects are ignored.
Create test for this new function and for ao2_weakproxy_find.
Kevin Harwell [Fri, 16 Nov 2018 20:45:23 +0000 (14:45 -0600)]
func_strings: HASHKEY - negative array index can cause corruption
This patch makes it so only matching non-empty key names, and keys created by
the HASH function are eligible for inclusion in the comma separated string. It
also fixes a bug where it was possible to write to a negative index if the
result buffer was empty.
ASTERISK-28159
patches:
ASTERISK-28159.diff submitted by Michael Walton (license 6502)
George Joseph [Mon, 19 Nov 2018 17:59:07 +0000 (10:59 -0700)]
CI: Get job timeouts from environment
The job timeouts were hard coded in the jenkinsfiles which
means changes had to go through gerrit. Now they are taken
from the following environment variables (and their defaults) that
can be set in Jenkins configuration...
This replaces the inline functions with macros. This removes the need
to directly use __ao2_ref, opts instead for standard ao2_bump and
ao2_cleanup macros.
George Joseph [Thu, 8 Nov 2018 15:53:44 +0000 (08:53 -0700)]
backtrace: Refactor ast_bt_get_symbols so it doesn't crash
We've been seeing crashes in libbfd when we attempt to generate
a stack trace from multiple threads. It turns out that libbfd
is NOT thread-safe. It can cache the bfd structure and give it to
multiple threads without protecting itself. To get around this,
we've added a global mutex around the bfd functions and also have
refactored the use of those functions to be more efficient and
to provide more information about inlined functions.
Also added a few more tests to test_pbx.c. One just calls
ast_assert() and the other calls ast_log_backtrace(). Neither are
run by default.
WARNING: This change necessitated changing the return value of
ast_bt_get_symbols() from an array of strings to a VECTOR of
strings. However, the use of this function outside Asterisk is not
likely.
Joshua C. Colp [Sun, 18 Nov 2018 23:53:14 +0000 (19:53 -0400)]
stasis: Remove stringfields and lock from change message.
When a subscribe or unsubscribe occurs a message is published
containing this information. This change makes it so that the
message no longer uses stringfields or a lock, as both are not
really needed for the message.
This change adds the ability for subscriptions to indicate
which message types they are interested in accepting. By
doing so the filtering is done before being dispatched
to the subscriber, reducing the amount of work that has
to be done.
This is optional and if a subscriber does not add
message types they wish to accept and set the subscription
to selective filtering the previous behavior is preserved
and they receive all messages.
There is also the ability to explicitly force the reception
of all messages for cases such as AMI or ARI where a large
number of messages are expected that are then generically
converted into a different format.
George Joseph [Sat, 17 Nov 2018 19:07:32 +0000 (12:07 -0700)]
CI: Pass work directory to runTestsuite
The testsuite can now use a user-specified work directory for
all it's temp files. This allows the docker containers to use
a tmpfs backed directory for the temp files instead of it's
own write-layer image.
* runTestsuite.sh now accepts a --work-dir command line argument
that gets exported as AST_WORK_DIR before running the testsuite.
* gates.jenkinsfile now specifies --work-dir to be
<testsuite_dir>/astroot.
Since the Asterisk CI docker hosts now mount /srv/jenkins/workspace
on a tmpfs, asterisk should be compiled and the testsuite run all in
memory.
Sungtae Kim [Sat, 17 Nov 2018 02:33:20 +0000 (03:33 +0100)]
res/res_ari: Fix null endpoint handle
The res_ari(POST /channels/create handler) deos not check the endpoint
parameter length. And it causes core
dump.
Fixed it to check the parameter length. Also fixed memory leak.
George Joseph [Thu, 15 Nov 2018 17:41:44 +0000 (10:41 -0700)]
CI: Allow runUnittests to use 'expect' to run the tests
There seems to be a race condition between starting the asterisk
daemon and attempting to use 'asterisk -r' that can cause the
control socket file to not be created. Since all of the Jenkins
slaves have 'expect' installed, the runUnittests script can use
it to start asterisk in the forground and issue the commands
interactively. This is much more reliable and it can also make
startup errors more visible since they'll be in the Jenkins console
output.
If 'expect' isn't installed, the original daemon/asterisk -r
process is used.
Also added a "core show settings" before running the tests
and added "notice,warning,error" to the console log.
Corey Farrell [Mon, 12 Nov 2018 18:23:34 +0000 (13:23 -0500)]
taskprocessor: Prevent race creating new taskprocessor.
Task processors are retrieved using a 'get or create' pattern. The
singleton container was unlocked between the get and create steps so
it's possible that two threads could create task processors with the
same name at the same time.
Corey Farrell [Fri, 16 Nov 2018 12:20:11 +0000 (07:20 -0500)]
pjproject-bundled: Use AST_DEVMODE for conditional compilation.
We previously allowed resample and g711 codecs to be built when
TEST_FRAMEWORK was enabled. This could cause errors if the testsuite
was run without this option enabled. Switch the build system to allow
those codecs to be built when --enable-dev-mode is used. This removes a
chance for strange testsuite errors from use of an inadequate pjsua
binary.
Corey Farrell [Thu, 15 Nov 2018 20:47:50 +0000 (15:47 -0500)]
res_pjsip_caller_id: Use static pj_str_t for fromto header names.
PJSIP assumes that these header names are not allocated, does not clone
the name strings when reusing headers.
Block unload of res_pjsip_caller_id until shutdown to ensure static
memory stays valid. It was previously unsafe to unload while any
sessions are active.
Torrey Searle [Wed, 24 Oct 2018 12:38:37 +0000 (14:38 +0200)]
res/res_pjsip_nat: Fix logic for REINVITES
The presence of Record-Route in re-invites is optional, thus it is
important to make sure the dialog doesn't have a routset before
rewriting the contact header.
George Joseph [Thu, 25 Oct 2018 15:25:58 +0000 (09:25 -0600)]
AST-2018-010: Fix length of buffer needed for SRV and NAPTR results
When dn_expand was being called on SRV and NAPTR results, the
return value was being used to calculate the size of the buffer
needed to store the host names. Since dn_expand returns the
length of the COMPRESSED name the buffer could be too short
to hold the EXPANDED name. The expanded name is NULL terminated
so using strlen() is the correct way to determine the length
actually needed for the buffer.
ASTERISK-28127
Reported by: Jan Hoffmann
patches:
patch.diff submitted by janhoffmann (license 6986)
projects / thirdparty / asterisk.git / log
