7 Commits (59881fbb9988be2f4e07ca750f45a404e79cb115)

Author	SHA1	Message	Date
Mark Michelson	b3cc2bd7df	res_pjsip_pubsub: Solidify lifetime and ownership of objects. ... There have been crashes and general instability seen in the pubsub code, so this patch introduces three changes to increase the stability. First, the ownership model for subscriptions has been modified. Due to RLS, subscriptions are stored in memory as a tree structure. Prior to my patch, the PJSIP subscription was the owner of the subscription tree. When the PJSIP subscription told us that it was terminating, we started destroying the subscription tree along with all of the individual leaf subscriptions that belong to the tree. The problem with this model is that the two actors in play here, the PJSIP subscription and the individual leaf subscriptions, need to have joint ownership of the subscription tree. So now, the PJSIP subscription and the individual leaf subscriptions each have a reference to the subscription tree. This way, we will not actually free memory until no players are left that care. The PJSIP subscription is a bigger stakeholder, in that if the PJSIP subscription's reference to the subscription tree is removed, the subscription tree instructs the leaf subscriptions to shut down and drop their references to the subscription tree when possible. The individual leaf subscriptions, upon being told to shut down, can drop their stasis subscriptions or whatever they use to learn of new state, and then drop their reference to the subscription tree once they are ready to die. Second, the lifetime of a PJSIP subscription's reference to our subscription tree has been altered. As I learned from doing a deep dive, the PJSIP evsub code can tell Asterisk multiple times that the subscription has been terminated, and not all of these times are especially helpful. I have altered the message flow that we use for SIP subscriptions such that we will always drop the PJSIP subscription's reference to the subscription tree when we send the NOTIFY that terminates a SIP subscription. This also means that we will now queue NOTIFY requests to be sent after responding to incoming SUBSCRIBEs so that we can have predictable state changes from the PJSIP evsub code. Third, the synchronization of operations has been improved. PJSIP can call into our code from a serializer thread (e.g. upon receiving an incoming request) or from the monitor thread (e.g. when a subscription times out). Because of this, there is the possibility of competing threads stepping on each other. PJSIP attempts to do some synchronization on its own by always keeping the dialog lock held when it calls into us. However, since we end up pushing tasks into the serializer, the result was that serialized operations were not grabbing the dialog lock and could, as a result, step on something that was being attempted by a different thread. Now we ensure that serialized operations grab the dialog lock, then check for extenuating circumstances, then proceed with their operation if they can. Change-Id: Iff2990c40178dad9cc5f6a5c7f76932ec644b2e5	10 years ago
Joshua Colp	d649d682c4	res_pjsip_exten_state: Fix race condition between sending NOTIFY and termination ... The res_pjsip_exten_state module currently has a race condition between processing the extension state callback from the PBX core and processing the subscription shutdown callback from res_pjsip_pubsub. There is currently no synchronization between the two. This can present a problem as while the SIP subscription will remain valid the tree it points to may not. This is in particular a problem as a task to send a NOTIFY may get queued which will try to use the tree that may no longer be valid. This change does the following to fix this problem: 1. All access to the subscription tree is done within the task that sends the NOTIFY to ensure that no other thread is modifying or destroying the tree. This task executes on the serializer for the subscriptions. 2. A reference to the subscription serializer is kept to ensure it remains valid for the lifetime of the extension state subscription. 3. The NOTIFY task has been changed so it will no longer attempt to send a NOTIFY if the subscription has already been terminated. ASTERISK-25057 #close Reported by: Matt Jordan Change-Id: I0b3cd2fac5be8d9b3dc5e693aaa79846eeaf5643	11 years ago
Joshua Colp	a8829490b6	res_pjsip_publish_asterisk: Add support for exchanging device and mailbox state using SIP. ... This module uses the inbound and outbound PUBLISH support to exchange device and mailbox state between Asterisk instances. Each instance is configured to publish to the other and requires no intermediary server. The functionality provided is similar to the XMPP and Corosync support. Review: https://reviewboard.asterisk.org/r/3780/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@420315 65c4cc65-6c06-0410-ace0-fbb531ad65f3	11 years ago
Mark Michelson	bc8c08c609	Abstract PJSIP-specific elements from the pubsub API. ... This helps to pave the way for RLS work that is to come. Since this is a self-contained change and subscription tests still pass, this work is being committed directly to trunk instead of a working branch. ASTERISK-23865 #close Review: https://reviewboard.asterisk.org/r/3628 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@417233 65c4cc65-6c06-0410-ace0-fbb531ad65f3	12 years ago
Joshua Colp	58f4c18ab6	res_pjsip_pubsub: Persist subscriptions in sorcery so they are recreated on startup. ... This change makes res_pjsip_pubsub persist inbound subscriptions in sorcery. By default this uses the local astdb but it can also be configured to store within an outside database. When Asterisk is started these subscriptions are recreated if they have not expired. Notifications are sent to the devices which have subscribed and they are none the wiser that the system has restarted. Review: https://reviewboard.asterisk.org/r/3598/ ........ Merged revisions 415766 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@415767 65c4cc65-6c06-0410-ace0-fbb531ad65f3	12 years ago
Mark Michelson	f55abe9cf1	Decouple subscription handling from NOTIFY/PUBLISH body generation. ... When the PJSIP pubsub framework was created, subscription handlers were required to state what event they handled along with what body types they knew how to generate. While this serves well when implementing a base RFC, it has problems when trying to extend the body to support non-standard or proprietary body elements. The code also was NOTIFY-specific, meaning that when the time comes that we start writing code to send out PUBLISH requests with MWI or presence bodies, we would likely find ourselves duplicating code that had previously been written. This changeset introduces the concept of body generators and body supplements. A body generator is responsible for allocating a native structure for a given body type, providing the primary body content, converting the native structure to a string, and deallocating resources. A body supplement takes the primary body content (the native structure, not a string) generated by the body generator and adds nonstandard elements to the body. With these elements living in their own module, it becomes easy to extend our support for body types and to re-use resources when sending a PUBLISH request. Body generators and body supplements register themselves with the pubsub core, similar to how subscription and publish handlers had done. Now, subscription handlers do not need to know what type of body content they generate, but they still need to inform the pubsub core about what the default body type for a given event package is. The pubsub core keeps track of what body generators and body supplements have been registered. When a SUBSCRIBE arrives, the pubsub core will check that there is a subscription handler for the event in the SUBSCRIBE, then it will check that there is a body generator that can provide the content specified in the Accept header(s). Because of the nature of body generators and supplements, it means res_pjsip_exten_state and res_pjsip_mwi have been completely gutted. They no longer worry about body types, instead calling ast_sip_pubsub_generate_body_content() when they need to generate a NOTIFY body. Review: https://reviewboard.asterisk.org/r/3150 ........ Merged revisions 407016 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@407030 65c4cc65-6c06-0410-ace0-fbb531ad65f3	12 years ago
Mark Michelson	735b30ad71	The large GULP->PJSIP renaming effort. ... The general gist is to have a clear boundary between old SIP stuff and new SIP stuff by having the word "SIP" for old stuff and "PJSIP" for new stuff. Here's a brief rundown of the changes: * The word "Gulp" in dialstrings, functions, and CLI commands is now "PJSIP" * chan_gulp.c is now chan_pjsip.c * Function names in chan_gulp.c that were "gulp_*" are now "chan_pjsip_*" * All files that were "res_sip*" are now "res_pjsip*" * The "res_sip" directory is now "res_pjsip" * Files in the "res_pjsip" directory that began with "sip_*" are now "pjsip_*" * The configuration file is now "pjsip.conf" instead of "res_sip.conf" * The module info for all PJSIP-related files now uses "PJSIP" instead of "SIP" * CLI and AMI commands created by Asterisk's PJSIP modules now have "pjsip" as the starting word instead of "sip" git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@395764 65c4cc65-6c06-0410-ace0-fbb531ad65f3	12 years ago