$Revision: 2382 $ $Date: 2007-06-29 20:44:51 +0300 (Fri, 29 Jun 2007) $

TM module enables stateful processing of SIP transactions. The main use of stateful logic, which is costly in terms of memory and CPU, is some services inherently need state. For example, transaction-based accounting (module acc) needs to process transaction state as opposed to individual messages, and any kinds of forking must be implemented statefully. Other use of stateful processing is it trading CPU caused by retransmission processing for memory. That makes however only sense if CPU consumption per request is huge. For example, if you want to avoid costly DNS resolution for every retransmission of a request to an unresolvable destination, use stateful mode. Then, only the initial message burdens server by DNS queries, subsequent retransmissions will be dropped and will not result in more processes blocked by DNS resolution. The price is more memory consumption and higher processing latency. From user's perspective, the major function is t_relay(). It setup transaction state, absorb retransmissions from upstream, generate downstream retransmissions and correlate replies to requests. In general, if TM is used, it copies clones of received SIP messages in shared memory. That costs the memory and also CPU time (memcpys, lookups, shmem locks, etc.) Note that non-TM functions operate over the received message in private memory, that means that any core operations will have no effect on statefully processed messages after creating the transactional state. For example, calling record_route after t_relay is pretty useless, as the RR is added to privately held message whereas its TM clone is being forwarded. TM is quite big and uneasy to program--lot of mutexes, shared memory access, malloc & free, timers--you really need to be careful when you do anything. To simplify TM programming, there is the instrument of callbacks. The callback mechanisms allow programmers to register their functions to specific event. See t_hooks.h for a list of possible events. Other things programmers may want to know is UAC--it is a very simplistic code which allows you to generate your own transactions. Particularly useful for things like NOTIFYs or IM gateways. The UAC takes care of all the transaction machinery: retransmissions , FR timeouts, forking, etc. See t_uac prototype in uac.h for more details. Who wants to see the transaction result may register for a callback.

First what is the idea with the branch concept: branch route is a route to be execute separately for each branch before being sent out - changes in that route should reflect only on that branch. There are several types of flags in &ser; : message/transaction flags - they are visible everywhere in the transaction (in all routes and in all sequential replies/request). branch flags - flags that are visible only from a specific branch - in all replies and routes connected to this branch. script flags - flags that exist only during script execution. They are not store anywhere and are lost once the top level route was left. For example: I have a call parallel forking to GW and to a user. And I would like to know from which branch I will get the final negative reply (if so). I will set a branch route before relaying the calls (with the 2 branches). The branch route will be separately executed for each branch; in the branch going to GW (I can identified it by looking to RURI), I will set a branch flag. This flag will appear only in the onreply route run for replied from GW. It will be also be visible in failure route if the final elected reply belongs to the GW branch. This flags will not be visible in the other branch (in routes executing replies from the other branch). For how to define branch flags and use via script, see and the setbflag(), resetbflag() and isbflagset() script functions. Also, modules may set branch flags before transaction creation (for the moment this feature is not available in script). The REGISTRAR module was the first to use this type of flags. The NAT flag is pushed in branch flags instead in message flags

Timers can be used to trigger failover behavior. E.g. if we send a call a gateway and the gateway does not send a provisional response within 3 seconds, we want to cancel this call and send the call to another gateway. Another example is to ring a SIP client only for 30 seconds and then redirect the call to the voicemail. There are two timers in &ser; : fr_timer - this timer is used when no response was received yet. If there is no response after fr_timer seconds the timer triggers (and failure route will be executed if t_on_failure() was called). If a provisional response was received, the timer is set to fr_inv_timer for INVITE transactions, and RT_T2 for all other transactions. If a final reponse is received, the transaction has finished. fr_inv_timer - this timer is used when a provisional reponse was received for an INVITE transaction. For example: You want to have failover if there is no provisional response after 3 seconds, but you want to ring for 60 seconds. Thuse, set the fr_timer to 3 and fr_inv_timer to 60.

DNS based failover can be use when relaying stateful requests. According to RFC 3263, DNS failover should be done on transport level or transaction level. TM module supports them both. Failover at transport level may be triggered by a failure of sending out the request message. A failure occures if the coresponding interface was found for sending the request, if the TCP connection was refused ot if a geenric internal error happend during send. There is no ICMP error report support. Failover at transaction level may be triggered when the transaction completed either with a 503 reply, either with a timeout without any received reply. In such a case, automatically, a new branch will be forked if any other destination IPs can be used to deliver the requests. The new branch will be a clone of the winning branch. The set of destinations IPs is step-by-step build (on demand) based on the NAPTR, SRV and A records available for the destination domain. DNS-based failover is by default applied excepting when this failover is globally disabled (see the core parameter disable_dns_failover) or when the relay flag (per transaction) is set (see the t_relay() function).

The following modules must be loaded before this module: No dependencies on other &ser; modules.

The following libraries or applications must be installed before running &ser; with this module loaded: None.

Timer which hits if no final reply for a request or ACK for a negative INVITE reply arrives (in seconds). Default value is 30 seconds. ... modparam("tm", "fr_timer", 10) ...

Timer which hits if no final reply for an INVITE arrives after a provisional message was received (in seconds). This timer is started after the first provisional response. Thus, fast failover (no 100 trying from gateway) can be achieved by setting fr_timer to low values. See example below Default value is 120 seconds. ... modparam("tm", "fr_inv_timer", 200) ...

Time for which a transaction stays in memory to absorb delayed messages after it completed; also, when this timer hits, retransmission of local cancels is stopped (a puristic but complex behavior would be not to enter wait state until local branches are finished by a final reply or FR timer--we simplified). Default value is 5 seconds. ... modparam("tm", "wt_timer", 10) ...

Time after which a to-be-deleted transaction currently ref-ed by a process will be tried to be deleted again. Default value is 2 seconds. ... modparam("tm", "delete_timer", 5) ...

Retransmission T1 period, in milliseconds. Default value is 500 milliseconds. ... modparam("tm", "T1_timer", 700) ...

Maximum retransmission period, in milliseconds. Default value is 4000 milliseconds. ... modparam("tm", "T2_timer", 8000) ...

If set, on FR timer INVITE transactions will be explicitly canceled if possible, silently dropped otherwise. Preferably, it is turned off to allow very long ringing. This behavior is overridden if a request is forked, or some functionality explicitly turned it off for a transaction (like acc does to avoid unaccounted transactions due to expired timer). Default value is 0 (false). ... modparam("tm", "noisy_ctimer", 1) ...

Should be request-uri matching used as a part of pre-3261 transaction matching as the standard wants us to do so? Turn only off for better interaction with devices that are broken and send different r-uri in CANCEL/ACK than in original INVITE. Default value is 1 (true). ... modparam("tm", "ruri_matching", 0) ...

Should be top most VIA matching used as a part of pre-3261 transaction matching as the standard wants us to do so? Turn only off for better interaction with devices that are broken and send different top most VIA in CANCEL/ACK than in original INVITE. Default value is 1 (true). ... modparam("tm", "via1_matching", 0) ...

Send timeout to be used by function which use UNIX sockets (as t_write_unix). Default value is 2 seconds. ... modparam("tm", "unix_tx_timeout", 5) ...

If true (non null value), the final response timer will be re-triggered for each received provisional reply. In this case, final response timeout may occure after a time longe than fr_inv_timer (if UAS keeps sending provisional replies) Default value is 1 (true). ... modparam("tm", "restart_fr_on_each_reply", 0) ...

Full specification (NAME, ID, Alias) of an AVP which contains a final response timeout value. If present, ths value will overeide the static fr_timer parameter. If set to empty string, the whole mechanism for variable timeout will be disabled, falling back to the static value. Default value is "NULL" (feature disabled). ... modparam("tm", "fr_timer_avp", "$avp(i:24)") ...

Full specification (NAME, ID, Alias) of an AVP which contains a final INVITE response timeout value. If present, ths value will overeide the static fr_inv_timer parameter. If set to empty string, the whole mechanism for variable timeout will be disabled, falling back to the static value. Default value is "NULL" (feature disabled). ... modparam("tm", "fr_inv_timer_avp", "$avp(i:25)") ...

List of additional information to be appended by t_write_fifo and t_write_unix functions. Default value is null string. Syntax of the parameter is: tw_append = append_name':' element (';'element)* element = ( [name '='] pseudo_variable) The full list of supported pseudo-variables in &ser; is availabe at: http://openser.org/docs/pseudo-variables-1.1.x.html Each element will be appended per line in name: value format. Element $rb (message body) is the only one which does not accept name; the body it will be printed all the time at the end, disregarding its position in the definition string. ... modparam("tm", "tw_append", "test: ua=$hdr(User-Agent) ;avp=$avp(i:10);$rb;time=$Ts") ...

Enable/disable passing of provisional replies to FIFO applications. Default value is 0. ... modparam("tm", "pass_provisional_replies", 1) ...

Enable/disable the usage of stateful synonym branch IDs in the generated Via headers. They are faster but not reboot-safe. Default value is 1 (use synonym branches). ... modparam("tm", "syn_branch", 0) ...

Describes how the AVPs should be handled in reply route: 0 - the AVPs will be per message only; they will not interfere with the AVPS stored in transaction; initially there will be an empty list and at the end of the route, all AVPs that were created will be discarded. 1 - the AVPs will be the transaction AVPs; initially the transaction AVPs will be visible; at the end of the route, the list will attached back to transaction (with all the changes) In mode 1, you can see the AVPs yo set in request route, branch route or failure route. The side efect is performance as more locking is required in order to keep the AVP's list integrity. Default value is 0. ... modparam("tm", "onreply_avp_mode", 1) ...

Creates a new transaction, returns a negative value on error. This is the only way a script can add a new transaction in an atomic way. Typically, it is used to deploy a &uas;. NOTE that the changes on the request that are made after this function call will not be saved into transaction!!! This function can be used from REQUEST_ROUTE. ... if (t_newtran()) { log("UAS logic"); t_reply("999","hello"); } else sl_reply_error(); ...

Relay a message statefully to destination indicated in current URI. (If the original URI was rewritten by UsrLoc, RR, strip/prefix, etc., the new URI will be taken). Returns a negative value on failure--you may still want to send a negative reply upstream statelessly not to leave upstream UAC in lurch. The coresponding transaction may or may not be already created. If not yet created, the function will automatically create it. The function may take as parameter an optional set of flags for controlling the internal behaviour. The flags may be given in decimal or hexa format; supported flags are: 0x01 - do not generate an 100 trying provisional reply when building the transaction. By default one is generated. Useful if you already pushed an stateless 100 reply from script. 0x02 - do not internally send a negative reply in case of failure. It applies only when the transaction is created. By default one is sent. Useful if you want to implement a serial forking in case of failure. 0x04 - disable the DNS failover for the transaction. Only first IP will be used. It disable the failover both at transaport and transaction level. This function can be used from REQUEST_ROUTE, FAILURE_ROUTE. ... if (!t_relay()) { sl_reply_error(); exit; } ...

Relay a message statefully to a fixed destination. The destination is specified as [proto:]host[:port]. The function may take as parameter an optional set of flags for controlling the internal behaviour - for details see the above t_relay([flags]) function. This functions can be used from REQUEST_ROUTE, FAILURE_ROUTE. ... t_relay("tcp:192.168.1.10:5060"); t_relay("mydomain.com:5070","0x1"); t_relay("udp:mydomain.com"); ...

Sends a stateful reply after a transaction has been established. See t_newtran for usage. Meaning of the parameters is as follows: code - Reply code number. reason_phrase - Reason string. This function can be used from REQUEST_ROUTE, FAILURE_ROUTE. ... t_reply("404", "Not found"); ...

Replicates a request to another destination. No information due the replicated request (like reply code) will be forwarded to the original SIP UAC. The destination is specified by a SIP URI. If multiple destinations are to be used, the additional SIP URIs have to be set as branches. The function may take as parameter an optional set of flags for controlling the internal behaviour - for description see the above t_relay([flags]) function. Note that only 0x4 is applicable here. This functions can be used from REQUEST_ROUTE. ... t_replicate("sip:1.2.3.4:5060"); t_replicate("sip:1.2.3.4:5060;transport=tcp"); t_replicate("sip:1.2.3.4","0x4"); ...

Remove transaction from memory (it will be first put on a wait timer to absorb delayed messages). This function can be used from REQUEST_ROUTE. ... t_release(); ...

Returns true if the regualr expresion re match the reply code of the response message as follows: in routing block - the code of the last sent reply. in on_reply block - the code of the current received reply. in on_failure block - the code of the selected negative final reply. This function can be used from REQUEST_ROUTE, ONREPLY_ROUTE, FAILURE_ROUTE and BRANCH_ROUTE . ... if (t_check_status("(487)|(408)")) { log("487 or 408 negative reply\n"); } ...

Flush the flags from current request into the already created transaction. It make sens only in routing block if the trnasaction was created via t_newtran() and the flags have been altered since. This function can be used from REQUEST_ROUTE and BRANCH_ROUTE . ... t_flush_flags(); ...

Returns true if all or last (depending of the parameter) reply(es) were local generated (and not received). Parameter may be all or last. This function can be used from REQUEST_ROUTE, BRANCH_ROUTE, FAILURE_ROUTE and ONREPLY_ROUTE. ... if (t_local_replied("all")) { log ("no reply received\n"); } ...

Write via FIFO file or UNIX socket a lot of information regarding the request. Which information should be written may be control via the tw_append parameter. This functions can be used from REQUEST_ROUTE, FAILURE_ROUTE and BRANCH_ROUTE. ... modparam("tm","tw_append","append1:Email=avp[i:12];UA=hdr[User-Agent]") modparam("tm","tw_append","append2:body=msg[body]") ... t_write_fifo("voicemail/append1","/tmp/appx_fifo"); ... t_write_unix("logger/append2","/var/run/logger.sock"); ...

Returns true if the current request is associated to a transaction. The relationship between the request ans transaction is defined as follow: non-CANCEL/non-ACK requests - is true if the request belongs to a transaction; if so, it means that the request is a retransmision. CANCEL request - true if the cancelled INVITE transaction exists. ACK request - true if the ACK is a local end-to-end ACK for an existent INVITE transaction. This function can be used from REQUEST_ROUTE and BRANCH_ROUTE. ... if ( is_method("CANCEL") ) { if ( t_check_trans() ) t_relay(); exit; } ...

Retuns true if called for an INVITE transaction that was explicitly cancelled by UAC side via a CANCEL request. This function can be used from ONREPLY_ROUTE, FAILURE_ROUTE. ... if (t_was_cancelled()) { log("transaction was cancelled by UAC\n"); } ...

Sets reply routing block, to which control is passed after a transaction completed with a negative result but before sending a final reply. In the referred block, you can either start a new branch (good for services such as forward_on_no_reply) or send a final reply on your own (good for example for message silo, which received a negative reply from upstream and wants to tell upstream 202 I will take care of it). As not all functions are available from failure route, please check the documentation for each function to see the permissions. Any other commands may result in unpredictable behavior and possible server failure. Only one failure_route can be armed for a request. If you use many times t_on_failure(), only the last one has effect. Note that whenever failure_route is entered, RURI is set to value of the winning branch. Meaning of the parameters is as follows: failure_route - Reply route block to be called. This function can be used from REQUEST_ROUTE, BRANCH_ROUTE, ONREPLY_ROUTE and FAILURE_ROUTE. ... route { t_on_failure("1"); t_relay(); } failure_route[1] { seturi("sip:user@voicemail"); append_branch(); t_relay(); } ...

Sets reply routing block, to which control is passed each time a reply (provisional or final) for the transaction is received. The route is not called for local generated replies! In the referred block, you can inspect the reply and perform text operations on it. As not all functions are available from this type of route, please check the documentation for each function to see the permissions. Any other commands may result in unpredictable behavior and possible server failure. Only one onreply_route can be armed for a request. If you use many times t_on_reply(), only the last one has effect. If the processed reply is provisionla reply (1xx code), by calling the drop() function (exported by core), the execution of the route will end and the reply will not be forwarded further. Meaning of the parameters is as follows: reply_route - Reply route block to be called. This function can be used from REQUEST_ROUTE, BRANCH_ROUTE, ONREPLY_ROUTE and FAILURE_ROUTE. ... route { t_on_reply("1"); t_relay(); } onreply_route[1] { if (t_check_status("1[0-9][0-9]")) { setflag(1); log("provisional reply received\n"); if (t_check_status("183")) drop; } } ...

Sets a branch route to be execute separately for each branch of the transaction before being sent out - changes in that route should reflect only on that branch. As not all functions are available from this type of route, please check the documentation for each function to see the permissions. Any other commands may result in unpredictable behavior and possible server failure. Only one branch_route can be armed for a request. If you use many time t_on_branch(), only the last one has effect. By calling the drop() function (exported by core), the execution of the branch route will end and the branch will not be forwarded further. Meaning of the parameters is as follows: branch_route - Branch route block to be called. This function can be used from REQUEST_ROUTE, BRANCH_ROUTE, ONREPLY_ROUTE and FAILURE_ROUTE. ... route { t_on_branch("1"); t_relay(); } branch_route[1] { if (uri=~"bad_uri") { xlog("dropping branch $ru \n"); drop; } if (uri=~"GW_uri") { append_rpid(); } } ...

Exported pseudo-variables are listed in the next sections.

$T_branch_idx - the index (starting with 1 for the first branch) of the branch for which is executed the branch_route[]. If used outside of branch_route[] block, the value is '0'.

$T_reply_code - the code of the reply, as follows: in request_route will be the last stateful sent reply; in reply_route will be the current processed reply; in failure_route will be the negative winning reply. In case of no-reply or error, '0' value is returned.

Generates and sends a local SIP request. Parameters: method - request method RURI - request SIP URI NEXT HOP - next hop SIP URI (OBP); use . if no value. socket - local socket to be used for sending the request; use . if no value. headers - set of additional headers to be added to the request; at least From and To headers must be specify) body - (optional, may not be present) request body (if present, requires the Content-Type and Content-length headers)

Generates and sends a CANCEL for an existing local SIP request. Parameters: callid - callid of the INVITE request to be cancelled. cseq - cseq of the INVITE request to be cancelled.

Gets information about the load of TM internal hash table. Parameters: none

Generates and sends a reply for an existing inbound SIP transaction. Parameters: code - reply code reason - reason phrase. trans_id - transaction identifier (has the hash_entry:label format) to_tag - To tag to be added to TO header new_headers - extra headers to be appended to the reply body - (optional, may not be present) reply body (if present, requires the Content-Type and Content-length headers)

Exported statistics are listed in the next sections.

Total number of total replies received by TM module - can be resetted.

Total number of replied received and relayed by TM module - can be resetted.

Total number of replied local generated by TM module - can be resetted.

Total number of transaction created by received requests - can be resetted.

Total number of transaction created by local generated requests - can be resetted.

Total number of transaction completed with 2xx replies - can be resetted.

Total number of transaction completed with 3xx replies - can be resetted.

Total number of transaction completed with 4xx replies - can be resetted.

Total number of transaction completed with 5xx replies - can be resetted.

Total number of transaction completed with 6xx replies - can be resetted.

Number of transaction existing in memeory at current time - can not be resetted.