Corey Farrell [Sun, 14 Oct 2018 12:58:59 +0000 (08:58 -0400)]
taskprocessor: Warn on unused result from pushing task.
Add attribute_warn_unused_result to ast_taskprocessor_push,
ast_taskprocessor_push_local and ast_threadpool_push. This will help
ensure we perform the necessary cleanup upon failure.
This patch is not in the upstream pjproject and does unsafe things with
the timer->_timer_id and timer->_grp_lock values in pj_timer_entry_reset()
outside of the timer heap lock. pj_timer_entry_reset() is also called for
timers that are not about to be rescheduled in a few places.
Corey Farrell [Thu, 11 Oct 2018 11:24:40 +0000 (07:24 -0400)]
threadpool: Eliminate pointless AO2 usage.
thread_worker_pair, set_size_data and task_pushed_data structures are
allocated with AO2 objects, passed to a taskprocessor, then released.
They never have multiple owners or use locking so AO2 only adds
overhead.
Alexei Gradinari [Fri, 12 Oct 2018 17:14:03 +0000 (13:14 -0400)]
res_pjsip: set callerid_tag to empty string
This patch sets the callerid_tag to empty string by default.
If the callerid_tag is set to NULL then the tag does not
become part of a connected line update.
For example:
Alice's tag is "Alice".
Bob's tag is empty.
Charlie's tag is "Charlie".
Alice calls Bob and then does attended transfer to Charlie.
When Alice hangs up the CONNECTEDLINE(tag) is "Alice"
on the interception routine on the Charlie's channel, but should be empty.
Ths patch also fix memory leaks if there are more then one options
"callerid", "callerid_tag", "voicemail_extension" and "contact_user"
in the pjsip.conf endpoint definition.
Corey Farrell [Fri, 12 Oct 2018 17:16:29 +0000 (13:16 -0400)]
main/astfd: Fix GCC8 format-truncation warning.
The field used to store call arguments was not large enough to hold the
arguments string that can be constructed for 'open'. Expand it to
prevent this warning/error.
Corey Farrell [Wed, 10 Oct 2018 07:54:08 +0000 (03:54 -0400)]
CI: Disable binary modules when REF_DEBUG is enabled.
In Asterisk 13 binary modules are linked to the non-REF_DEBUG variants
of ao2 functions. Cause false leaks to be reported when binary modules
are used.
Richard Mudgett [Tue, 9 Oct 2018 21:18:49 +0000 (16:18 -0500)]
res_statsd.c: Fix returned reload status.
The return status when there was no change in statsd.conf was incorrect.
This resulted in the wrong status message on the CLI when reloading the
module.
* Fixed cleanup on initial load if initializing statsd failed.
neutrino88 [Wed, 3 Oct 2018 21:51:49 +0000 (17:51 -0400)]
core/frame: generate correct T.140 payload in ast_sendtext_data()
ast_sendtext_data() would create an incorrect T.140 text frame which
length include the null terminator byte. It causes ultimately RTP
packets to be send with this trailing 0. The proposed fix just set the
correct length to the text frame
Sean Bright [Wed, 3 Oct 2018 12:56:34 +0000 (08:56 -0400)]
http.c: Reload TLS even if http.conf hasn't changed
There is currently no way to indicate to Asterisk that TLS certificates
and/or keys have been updated other than by modifying http.conf or
restarting Asterisk.
There is already code in main/tcptls.c that determines if a reload is
actually necessary based on the hashes of the certicate and dependent
files, so this change merely gives us a way to request a reload without
explicitly modifying http.conf.
Richard Mudgett [Tue, 2 Oct 2018 21:15:37 +0000 (16:15 -0500)]
res_smdi.c: Fix module ref counting and inverted test.
I think this module is so screwed up that it doesn't work anymore. Even
with these attempts to fix things it still won't gracefully shut down.
The module refs will not go to zero to allow unloading the module.
* Fix module ref counting dealing with the SMDI interface object. There
were several off-nominal paths that unbalanced the module ref count. Also
the destructor freed the ao2 object itself which is bad. Made the
smdi_read thread not hold its own ref to the SMDI interface object so when
all refs go away the destructor will stop the listener thread.
* Fixed the smdi_load() return code of 1 concerning the number of
listeners. The test was inverted.
core/frame: Fix ast_frdup() and ast_frisolate() for empty text frames
If a channel creates an AST_TEXT_FRAME with datalen == 0, the ast_frdup()
and ast_frisolate() functions could create a clone frame with an invalid
data.ptr which would cause a crash. The proposed fix is to make sure that
for such empty text frames, ast_frdup() and ast_frisolate() return cloned
text frames with a valid data.ptr.
Add a volatile flag to lock tracking structures so we only need to use
the global lock when first initializing tracking.
Additionally add support for DEBUG_THREADS_LOOSE_ABI. This is used by
astobj2.c to eliminate storage for tracking fields when DEBUG_THREADS is
not defined.
Ben Ford [Mon, 10 Sep 2018 16:28:09 +0000 (11:28 -0500)]
res_rtp_asterisk.c: Add "seqno" strictrtp option
When networks experience disruptions, there can be large gaps of time
between receiving packets. When strictrtp is enabled, this created
issues where a flood of packets could come in and be seen as an attack.
Another option - seqno - has been added to the strictrtp option that
ignores the time interval and goes strictly by sequence number for
validity.
res_pjsip: improve realtime performance on CLI 'pjsip show contacts'
CLI command 'pjsip show contacts' inefficiently make a lot of DB requests.
For example if there are 10k aors then asterisk requests these 10k records
of aor and then does 10k requests of contact - one request per aor.
Even if use 'like <pattern>' the asterisk requests all aor's and contact's
records and then filters them by itself.
This patch gathers contact's container by
- retrieving all dynamic contacts by regex (filtered by reg_server)
- retrieving all aors with permanent contacts
- finally filters container by regex
When REF_DEBUG and AO2_DEBUG are both enabled we closed the refs log
before we shutdown astobj2_container. This caused the AO2_DEBUG
container registration container to be reported as a leak.
Moritz Fain [Tue, 26 Jun 2018 14:17:37 +0000 (16:17 +0200)]
res_stasis: Fix stale data in ARI bridges
Fixed an issue that resulted in "Allocation failed" each time an ARI
request was made to start playing MOH on a bridge.
In bridge_moh_create() we were attaching the after bridge callbacks to
chan which is the ;1 channel of the unreal channel pair. We should have
attached them to the ;2 channel which is pushed into the bridge by
ast_unreal_channel_push_to_bridge(). The callbacks are called when the
specific channel leaves the bridging system. Since the ;1 channel is
never put into a bridge the callbacks never get called. The callbacks
then never remove the moh_wrapper from the app_bridges_moh container. As
a result we cannot find the channel associated with the wrapper to start
MOH because it has hungup. This is the reason causing the reported issue.
* Rather than using after bridge callbacks to cleanup, we now have
moh_channel_thread() doing the cleanup when the channel hangs up.
* Fixed moh_channel_thread() accumulating control frames on the stasis
bridge MOH channel until MOH is stopped. Control frames are no longer
accumulated while MOH is playing.
* Fixed channel ref counting issue. stasis_app_bridge_moh_channel() may
or may not return a channel ref. As a result ast_ari_bridges_start_moh()
wouldn't know it may have a channel ref to release.
stasis_app_bridge_moh_channel() will now return a ref with the channel it
returns.
app_queue: Fix Attended transfer hangup with removing pending member.
This issue related to setting of holdtime, announcements, member delays.
It works well if we set the member delays to "0" and no announcements
and no holdtime.This issue will happen if we set member delays to "1",
"2"... or announcements or holdtime and hangs up the call during
processing it.
And here is the reason:
(At the step of answering a phone.)
It takes care any holdtime, announcements, member delays,
or other options after a call has been answered if it exists.
Normally, After the call has been aswered,
and we wait for the processing one of the cases of the member delays
or hold time or announcements finished, "if (ast_check_hangup(peer))"
will be not executed, then queue will be updated at update_queue().
Here, pending member will be removed.
However, after the call has been aswered,
if we hangs up the call during one of the cases of the member delays
or hold time or announcements, "if (ast_check_hangup(peer))"
will be executed.
outgoing = NULL and at hangupcalls, pending members will not be removed.
* This fixed patch will remove the pending member from container
before hanging up the call with outgoing is NULL.
ASTERISK-27920
Reported by: Cao Minh Hiep
Tested by: Cao Minh Hiep
res_rtp_asterisk: Raise event when RTP port is allocated
This change raises a testsuite event to provide what port
Asterisk has actually allocated for RTP. This ensures that
testsuite tests can remove any assumption of ports and instead
use the actual port in use.
Kevin Harwell [Mon, 17 Sep 2018 20:35:05 +0000 (15:35 -0500)]
rtp_engine: rtcp_report_to_json can overflow the ssrc integer value
When writing an RTCP report to json the code attempts to pack the "ssrc" and
"source_ssrc" unsigned integer values as a signed int value type. This of course
means if the ssrc's unsigned value is greater than that which can fit into a
signed integer value it gets converted to a negative number. Subsequently, the
negative value goes out in the json report.
This patch now packs the value as a json_int_t, which is the widest integer type
available on a given system. This should make it so the value no longer
overflows.
Note, this was caught by two failing tests hep/rtcp-receiver/ and
hep/rtcp-sender.
On SQL error there is not diagnostic information about this error.
There is only
WARNING res_odbc.c: SQL Execute error -1!
The function ast_odbc_print_errors calls a SQLGetDiagField to get the number
of available diagnostic records, but the SQLGetDiagField returns 0.
However SQLGetDiagRec could return one diagnostic records in this case.
Looking at many example of getting diagnostics error information
I found out that the best way it's to use only SQLGetDiagRec
while it returns SQL_SUCCESS.
Also this patch adds calls of ast_odbc_print_errors on SQL_ERROR
to res_config_odbc.
George Joseph [Thu, 20 Sep 2018 15:15:48 +0000 (09:15 -0600)]
app_voicemail: Cleanup mailbox topic and cache
app_voicemail wasn't properly cleaning up the stasis cache or the
mwi topic pool when the module was unloaded or when a user was
deleted as a result of a reload. This resulted in leaks in both
areas.
* app_voicemail now calls ast_delete_mwi_state_full when it frees
a user structure and ast_delete_mwi_state_full in turn now calls
the new stasis_topic_pool_delete_topic function to clear the topic
from the pool.
The HTTP request processing in res_http_websocket allocates additional
space on the stack for various headers received during an Upgrade request.
An attacker could send a specially crafted request that causes this code
to overflow the stack, resulting in a crash.
* No longer allocate memory from the stack in a loop to parse the header
values. NOTE: There is a slight API change when using the passed in
strings as is. We now require the passed in strings to no longer have
leading or trailing whitespace. This isn't a problem as the only callers
have already done this before passing the strings to the affected
function.
George Joseph [Thu, 20 Sep 2018 14:41:15 +0000 (08:41 -0600)]
stasis: Add function to delete topic from pool
There's been a long standing leak when using topic pools. The
topics in the pool get cleaned up when the last pool reference is
released but you can't remove a topic specifically. If you reloaded
app_voicemail for instance, and mailboxes went away, their topics
were left in the pool.
* Added stasis_topic_pool_delete_topic() so modules can clean up
topics from pools.
* Registered the topic pool containers so it can be examined from
the CLI when AO2_DEBUG is enabled. They'll be named
"<topic_pool_name>-pool".
Richard Mudgett [Fri, 14 Sep 2018 20:51:41 +0000 (15:51 -0500)]
stasis: No need to keep a stasis type ref in a stasis msg or cache object.
Stasis message types are global ao2 objects and we make stasis messages
and cache entries hold references to them. Since there are currently
situations where cache objects are never deleted, the reference count on
the types can exceed 100000 and generate a FRACK assertion message. The
stasis message cache could conceivably also have that many messages
legitimately on large systems.
The only down side to not holding the message type ref in the stasis
message is it only makes a crash either at shutdown or when manually
unloading a busy module slightly more likely. However, this is more
exposing a pre-existing stasis shutdown ordering issue than a problem with
not holding a message type ref in stasis messages.
* Made stasis messages and cache entries no longer hold a ref to the
message type.
projects / thirdparty / asterisk.git / log
