asterisk/main/autoservice.c

385 lines
9.5 KiB
C
Raw Normal View History

/*
* Asterisk -- An open source telephony toolkit.
*
* Copyright (C) 1999 - 2008, Digium, Inc.
*
* Mark Spencer <markster@digium.com>
* Russell Bryant <russell@digium.com>
*
* See http://www.asterisk.org for more information about
* the Asterisk project. Please do not directly contact
* any of the maintainers of this project for assistance;
* the project provides a web site, mailing lists and IRC
* channels for your use.
*
* This program is free software, distributed under the terms of
* the GNU General Public License Version 2. See the LICENSE file
* at the top of the source tree.
*/
/*! \file
*
* \brief Automatic channel service routines
*
* \author Mark Spencer <markster@digium.com>
* \author Russell Bryant <russell@digium.com>
*/
/*** MODULEINFO
<support_level>core</support_level>
***/
#include "asterisk.h"
#include <sys/time.h>
#include <signal.h>
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
#include "asterisk/_private.h" /* prototype for ast_autoservice_init() */
#include "asterisk/pbx.h"
#include "asterisk/frame.h"
#include "asterisk/sched.h"
#include "asterisk/channel.h"
#include "asterisk/file.h"
#include "asterisk/translate.h"
#include "asterisk/manager.h"
#include "asterisk/chanvars.h"
#include "asterisk/linkedlists.h"
#include "asterisk/indications.h"
#include "asterisk/lock.h"
#include "asterisk/utils.h"
#define MAX_AUTOMONS 1500
struct asent {
struct ast_channel *chan;
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
/*! This gets incremented each time autoservice gets started on the same
* channel. It will ensure that it doesn't actually get stopped until
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
* it gets stopped for the last time. */
unsigned int use_count;
unsigned int orig_end_dtmf_flag:1;
unsigned int video_update:1;
unsigned int ignore_frame_types;
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
/*! Frames go on at the head of deferred_frames, so we have the frames
* from newest to oldest. As we put them at the head of the readq, we'll
* end up with them in the right order for the channel's readq. */
AST_LIST_HEAD_NOLOCK(, ast_frame) deferred_frames;
AST_LIST_ENTRY(asent) list;
};
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
static AST_LIST_HEAD_STATIC(aslist, asent);
static ast_cond_t as_cond;
static pthread_t asthread = AST_PTHREADT_NULL;
static volatile int asexit = 0;
static int as_chan_list_state;
static void *autoservice_run(void *ign)
{
ast_callid callid = 0;
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
struct ast_frame hangup_frame = {
.frametype = AST_FRAME_CONTROL,
.subclass.integer = AST_CONTROL_HANGUP,
};
while (!asexit) {
struct ast_channel *mons[MAX_AUTOMONS];
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
struct asent *ents[MAX_AUTOMONS];
struct ast_channel *chan;
struct asent *as;
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
int i, x = 0, ms = 50;
struct ast_frame *f = NULL;
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
struct ast_frame *defer_frame = NULL;
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_LOCK(&aslist);
/* At this point, we know that no channels that have been removed are going
* to get used again. */
as_chan_list_state++;
if (AST_LIST_EMPTY(&aslist)) {
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
ast_cond_wait(&as_cond, &aslist.lock);
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_TRAVERSE(&aslist, as, list) {
if (!ast_check_hangup(as->chan)) {
if (x < MAX_AUTOMONS) {
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
ents[x] = as;
mons[x++] = as->chan;
} else {
ast_log(LOG_WARNING, "Exceeded maximum number of automatic monitoring events. Fix autoservice.c\n");
}
}
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_UNLOCK(&aslist);
if (!x) {
/* If we don't sleep, this becomes a busy loop, which causes
* problems when Asterisk runs at a different priority than other
* user processes. As long as we check for new channels at least
* once every 10ms, we should be fine. */
usleep(10000);
continue;
}
chan = ast_waitfor_n(mons, x, &ms);
if (!chan) {
continue;
}
callid = ast_channel_callid(chan);
ast_callid_threadassoc_change(callid);
f = ast_read(chan);
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
if (!f) {
/* No frame means the channel has been hung up.
* A hangup frame needs to be queued here as ast_waitfor() may
* never return again for the condition to be detected outside
* of autoservice. So, we'll leave a HANGUP queued up so the
* thread in charge of this channel will know. */
defer_frame = &hangup_frame;
} else if (ast_is_deferrable_frame(f)) {
defer_frame = f;
} else {
/* Can't defer. Discard and continue with next. */
ast_frfree(f);
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
continue;
}
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
for (i = 0; i < x; i++) {
struct ast_frame *dup_f;
if (mons[i] != chan) {
continue;
}
if (!f) { /* defer_frame == &hangup_frame */
if ((dup_f = ast_frdup(defer_frame))) {
AST_LIST_INSERT_HEAD(&ents[i]->deferred_frames, dup_f, frame_list);
}
} else {
if (defer_frame->frametype == AST_FRAME_CONTROL &&
defer_frame->subclass.integer == AST_CONTROL_VIDUPDATE) {
/* If a video update is already queued don't needlessly queue another */
if (ents[i]->video_update) {
ast_frfree(defer_frame);
break;
}
ents[i]->video_update = 1;
}
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
if ((dup_f = ast_frisolate(defer_frame))) {
AST_LIST_INSERT_HEAD(&ents[i]->deferred_frames, dup_f, frame_list);
}
if (dup_f != defer_frame) {
ast_frfree(defer_frame);
}
}
break;
}
/* The ast_waitfor_n() call will only read frames from
* the channels' file descriptors. If ast_waitfor_n()
* returns non-NULL, then one of the channels in the
* mons array must have triggered the return. It's
* therefore impossible that we got here while (i >= x).
* If we did, we'd need to ast_frfree(f) if (f). */
}
ast_callid_threadassoc_change(0);
asthread = AST_PTHREADT_NULL;
return NULL;
}
int ast_autoservice_start(struct ast_channel *chan)
{
int res = 0;
struct asent *as;
if (ast_thread_is_user_interface()) {
/* User interface threads do not handle channel media. */
ast_debug(1, "Thread is a user interface, not putting channel %s into autoservice\n",
ast_channel_name(chan));
return 0;
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_LOCK(&aslist);
AST_LIST_TRAVERSE(&aslist, as, list) {
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
if (as->chan == chan) {
as->use_count++;
break;
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
}
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_UNLOCK(&aslist);
if (as) {
/* Entry exists, autoservice is already handling this channel */
return 0;
}
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
if (!(as = ast_calloc(1, sizeof(*as))))
return -1;
/* New entry created */
as->chan = chan;
as->use_count = 1;
ast_channel_lock(chan);
as->orig_end_dtmf_flag = ast_test_flag(ast_channel_flags(chan), AST_FLAG_END_DTMF_ONLY) ? 1 : 0;
if (!as->orig_end_dtmf_flag)
ast_set_flag(ast_channel_flags(chan), AST_FLAG_END_DTMF_ONLY);
ast_channel_unlock(chan);
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_LOCK(&aslist);
if (AST_LIST_EMPTY(&aslist) && asthread != AST_PTHREADT_NULL) {
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
ast_cond_signal(&as_cond);
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_INSERT_HEAD(&aslist, as, list);
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
if (asthread == AST_PTHREADT_NULL) { /* need start the thread */
if (ast_pthread_create_background(&asthread, NULL, autoservice_run, NULL)) {
ast_log(LOG_WARNING, "Unable to create autoservice thread :(\n");
/* There will only be a single member in the list at this point,
the one we just added. */
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_REMOVE(&aslist, as, list);
ast_free(as);
asthread = AST_PTHREADT_NULL;
res = -1;
} else {
pthread_kill(asthread, SIGURG);
}
}
AST_LIST_UNLOCK(&aslist);
return res;
}
int ast_autoservice_stop(struct ast_channel *chan)
{
int res = -1;
struct asent *as, *removed = NULL;
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
struct ast_frame *f;
int chan_list_state;
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
if (ast_thread_is_user_interface()) {
/* User interface threads do not handle channel media. */
ast_debug(1, "Thread is a user interface, not removing channel %s from autoservice\n",
ast_channel_name(chan));
return 0;
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_LOCK(&aslist);
/* Save the autoservice channel list state. We _must_ verify that the channel
* list has been rebuilt before we return. Because, after we return, the channel
* could get destroyed and we don't want our poor autoservice thread to step on
* it after its gone! */
chan_list_state = as_chan_list_state;
/* Find the entry, but do not free it because it still can be in the
autoservice thread array */
AST_LIST_TRAVERSE_SAFE_BEGIN(&aslist, as, list) {
if (as->chan == chan) {
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
as->use_count--;
if (as->use_count < 1) {
AST_LIST_REMOVE_CURRENT(list);
removed = as;
}
break;
}
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_TRAVERSE_SAFE_END;
if (removed && asthread != AST_PTHREADT_NULL) {
pthread_kill(asthread, SIGURG);
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
AST_LIST_UNLOCK(&aslist);
if (!removed) {
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
return 0;
}
/* Wait while autoservice thread rebuilds its list. */
while (chan_list_state == as_chan_list_state) {
usleep(1000);
}
/* Now autoservice thread should have no references to our entry
and we can safely destroy it */
if (!ast_channel_softhangup_internal_flag(chan)) {
res = 0;
}
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
ast_channel_lock(chan);
if (!as->orig_end_dtmf_flag) {
ast_clear_flag(ast_channel_flags(chan), AST_FLAG_END_DTMF_ONLY);
}
Frame deferral: Revert API refactoring. There are several issues with deferring frames that are caused by the refactoring. 1) The code deferring frames mishandles adding a deferred frame to the deferred queue. As a result the deferred queue can only be one frame long. 2) Deferrable frames can come directly from the channel driver as well as the read queue. These frames need to be added to the deferred queue. 3) Whoever is deferring frames is really only doing the __ast_read() to collect deferred frames and doesn't care about the returned frames except to detect a hangup event. When frame deferral is completed we must make the normal frame processing see the hangup as a frame anyway. As such, there is no need to have varying hangup frame deferral methods. We also need to be aware of the AST_SOFTHANGUP_ASYNCGOTO hangup that isn't real. That fake hangup is to cause the PBX thread to break out of loops to go execute a new dialplan location. 4) To properly deal with deferrable frames from the channel driver as pointed out by (2) above, means that it is possible to process a dialplan interception routine while frames are deferred because of the AST_CONTROL_READ_ACTION control frame. Deferring frames is not implemented as a re-entrant operation so you could have the unsupported case of two sections of code thinking they have control of the media stream. A worse problem is because of the bad implementation of the AMI PlayDTMF action. It can cause two threads to be deferring frames on the same channel at the same time. (ASTERISK_25940) * Rather than fix all these problems simply revert the API refactoring as there is going to be only autoservice and safe_sleep deferring frames anyway. ASTERISK-26343 ASTERISK-26716 #close Change-Id: I45069c779aa3a35b6c863f65245a6df2c7865496
2017-01-31 22:32:18 +00:00
while ((f = AST_LIST_REMOVE_HEAD(&as->deferred_frames, frame_list))) {
if (!((1 << f->frametype) & as->ignore_frame_types)) {
ast_queue_frame_head(chan, f);
}
ast_frfree(f);
}
ast_channel_unlock(chan);
Merged revisions 89790 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r89790 | russell | 2007-11-27 15:45:51 -0600 (Tue, 27 Nov 2007) | 41 lines Merge changes from team/russell/autoservice_1.4 This set of changes fixes an issue that was reported to me on IRC yesterday. The user, d1mas, was using chan_zap for incoming calls and was having DTMF recognition issues in some situations. Specifically, he noticed that the problem occurred when using DISA or WaitExten. He also noticed that when using Read, the problem did not occur. His system also used DUNDi for dialplan lookups. So, he theorized that if the DUNDi lookups blocked for some period of time, that audio from the zap channel could get lost. If the audio got lost, then it wouldn't be run through the DTMF detector, and digits could get lost. He was correct, and the following set of changes fixes the problem. However, the changes go a little bit further than what was necessary to fix this exact problem. 1) I updated pbx_extension_helper() to autoservice the associated channel to handle cases where extension lookups may take a long time. This would normally be a dialplan switch that does some lookup over the network, such as the DUNDi or IAX2 switches. This ensures that even while a DUNDi lookup is blocking, the channel will be continuously serviced. 2) I made a change to the autoservice code. This is actually something that has bothered me for a long time. When a channel is in autoservice, _all_ frames get thrown away. However, some frames really shouldn't be thrown away. The most notable examples are signalling (CONTROL) frames, and DTMF. So, this patch queues up important frames while a channel is in autoservice. When autoservice is stopped on the channel, the queued up frames get stuck back on the channel so that they can get processed instead of thrown away. 3) I made another change to the autoservice code to handle the case where autoservice is started on channels recursively. Previously, you could call ast_autoservice_start() multiple times on a channel, and it would stop the first time ast_autoservice_stop() gets called. Now, it will ensure that autoservice doesn't actually stop until the final call to ast_autoservice_stop(). ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@89791 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2007-11-27 22:05:36 +00:00
ast_free(as);
return res;
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
void ast_autoservice_chan_hangup_peer(struct ast_channel *chan, struct ast_channel *peer)
{
if (chan && !ast_autoservice_start(chan)) {
ast_hangup(peer);
ast_autoservice_stop(chan);
} else {
ast_hangup(peer);
}
}
int ast_autoservice_ignore(struct ast_channel *chan, enum ast_frame_type ftype)
{
struct asent *as;
int res = -1;
AST_LIST_LOCK(&aslist);
AST_LIST_TRAVERSE(&aslist, as, list) {
if (as->chan == chan) {
res = 0;
as->ignore_frame_types |= (1 << ftype);
break;
}
}
AST_LIST_UNLOCK(&aslist);
return res;
}
static void autoservice_shutdown(void)
{
pthread_t th = asthread;
asexit = 1;
if (th != AST_PTHREADT_NULL) {
ast_cond_signal(&as_cond);
pthread_kill(th, SIGURG);
pthread_join(th, NULL);
}
}
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
void ast_autoservice_init(void)
{
ast_register_cleanup(autoservice_shutdown);
3) In addition to merging the changes below, change trunk back to a regular LIST instead of an RWLIST. The way this list works makes it such that a RWLIST provides no additional benefit. Also, a mutex is needed for use with the thread condition. Merged revisions 105563 via svnmerge from https://origsvn.digium.com/svn/asterisk/branches/1.4 ........ r105563 | russell | 2008-03-03 09:50:43 -0600 (Mon, 03 Mar 2008) | 24 lines Merge in some changes from team/russell/autoservice-nochans-1.4 These changes fix up some dubious code that I came across while auditing what happens in the autoservice thread when there are no channels currently in autoservice. 1) Change it so that autoservice thread doesn't keep looping around calling ast_waitfor_n() on 0 channels twice a second. Instead, use a thread condition so that the thread properly goes to sleep and does not wake up until a channel is put into autoservice. This actually fixes an interesting bug, as well. If the autoservice thread is already running (almost always is the case), then when the thread goes from having 0 channels to have 1 channel to autoservice, that channel would have to wait for up to 1/2 of a second to have the first frame read from it. 2) Fix up the code in ast_waitfor_nandfds() for when it gets called with no channels and no fds to poll() on, such as was the case with the previous code for the autoservice thread. In this case, the code would call alloca(0), and pass the result as the first argument to poll(). In this case, the 2nd argument to poll() specified that there were no fds, so this invalid pointer shouldn't actually get dereferenced, but, this code makes it explicit and ensures the pointers are NULL unless we have valid data to put there. (related to issue #12116) ........ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@105564 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2008-03-03 15:59:50 +00:00
ast_cond_init(&as_cond, NULL);
}