March 2021 - FreeSWITCH-users - FreeSWITCH Mailing Lists

Installing Modules
by Rasheed Kalapurackal 10 Mar '21

10 Mar '21

Hello , I completed the installation of Freeswitch on Debian 10 . Now i need to install an additional module in to the Freeswitch . Is it possible to install Modules in to an already deployed Freeswitch instance ? Any quick help on this would be greatly appreciated. Thanks Abdul Rasheed

5 6

NOTIFYs Event
by Vishal Pai 10 Mar '21

10 Mar '21

Hello Everyone Few days back we had a issue with Notify request from park+*5903(a)abc.com the freeswitch was sending the notify request but the endpoint side was responding late on checking it out it was router on which sip alg was enabled so we disable the that and all started working to normal. My question is that is there any way to identify such issues and block the IP or unregistered that EXT automatically via fail2ban or any other way is possible. Thanks

1 0

Freeswitch DTMF issue - wrong timestamp base
by Arslan Saeed 10 Mar '21

10 Mar '21

Hi All, We have come across an issue where freeswitch ((Version 1.10.5 -release-17-25569c1631 64bit) seems to be doing something wrong with regards to RFC4733 DTMF event packets generation. What we see is that freeswitch would start using a new timestamp base for the RTP event packets (sometimes even for a few RTP packets preceeding the DTMF packetS) and then at the end of the DTMF event revert to older timestamp base. Although the marker bit is set on when new timestamp base is used but as far as I know this is not RFC compliant and causes issues for some of the PBX devices. RFC 4733, section 2.5.1.2 says: DTMF digits and other named telephone events are carried as part of the audio stream, and they MUST use the same sequence number and timestamp base as the regular audio channel to simplify the generation of audio waveforms at a gateway.``` https://tools.ietf.org/html/rfc4733#section-2.5.1 -- I think if freeswitch were to change the timestamp for DTMF event packets then it must also change the RTP SSRC at same time (which does not happen) Although I don't understand the need to change timestamp base at all for the start of DTMF event packets or even with the few proceeding RTP packets. I have tried to use following parameter but that does not help <param name="auto-rtp-bugs" data="CHANGE_SSRC_ON_MARKER"/> I have also tried to test with these additional parameters (in order to infleunce freeswitch to not use a new timestamp base for the start of dtmf event packets) but no change in freeswitch behaviour. <param name="pass-rfc2833" value="true"/> <param name="rtp-rewrite-timestamps"value="true"/> <param name="auto-rtp-bugs"value="CISCO_SKIP_MARK_BIT_2833"/> <param name="auto-rtp-bugs"value="clear"/> In addition I have tried to use some jitter related sofia profile parameters but no affect. Another DTMF related issue that I have seen with freeswitch is that it sends DTMF event packets in a short burst and not evenly spaced out per the codec sample rate. For example if RTP is being sent at 20ms, freeswitch would send dtmf event packets at different interval like it may send 8ms apart or 13 ms apat and at times it sends multiple event packets just a few ms apart. To fix that behaviour, when we explicity set the RTP-timer-name to "soft" using <param name="rtp-timer-name" value="soft"/> that fixes it and ensures evenly spaced DTMF packets say 20 ms apart (that fixes interop with some of the devices) Any help is highly appreciated. Thanks Arslan

2 1

using mod_unimrcp to transcribe voicemails
by Lewis Bergman 10 Mar '21

10 Mar '21

Does anyone have guidance or code to get this to work? I have looked at the transcription documentation in Freeswitch and it all seems to be related to command line type implementations using Google, Wartson, Azure, etc, with curl. using a FusionPBX Lua script ( /var/www/fusionpbx/app/scripts/resources/scripts/app/voicemail/resources/functions/record_message.lua ) If someone has made voicemail transcription work with mod_unimrcp I would be grateful in hearing how it was done. It seems like some Lua programming might be required at the minimum. If making it work is described an idea of where to start looking for documentation would be great. -- Lewis Bergman 325-439-0533 Cell

2 3

Problem with answered call and multiple re-invites
by Jose Fco. Irles Durá 09 Mar '21

09 Mar '21

Hi, I have an audio problem with some user agents (Alcatel OXO) that sends multiple re-invites. Example scenario: FreeSWITCH with public IP (example: 1.1.1.1) Alcatel behind NAT (private IP 192.168.20.200, public IP 2.2.2.2) 1. I send the call to Alcatel and the sdp negotiation works well. The remote port for RTP is 32014. In the capture I can see that FreeSWITCH sends first RTP packets to the IP 192.168.20.200 and port 32014. Later, the audio stream from Alcatel arrives from IP 2.2.2.2 and port 32014 and FreeSWITCH changes and sends RTP to 2.2.2.2:32014 (rtp auto adjust), so I can listen the audio (the Alcatel has an IVR menu). 2. In the IVR menu I press one option and the Alcatel sends me an INVITE changing media. It sends a new port (32016), FreeSWITCH accepts the re-invite without changing its source port. In the capture I can see that FreeSWITCH sends first RTP packets to the IP 192.168.20.200 and port 32016, but when the first packets from the real IP:PORT (2.2.2.2:32016) arrives, FreeSWITCH don't change its RTP stream to the new port, so I can't hear the audio from the Alcatel. 3. Later, FreeSWITCH receives another re-invite changing remote port to 32020. First, FreeSWITCH sends audio to the private IP:PORT (192.168.20.200:32020), but when it receives the real stream from 2.2.2.2:32020, it doesn't change to the new IP:PORT (2.2.2.2:32020). >From point 2, I don't hear anything. This problem happens with the last stable version (1.10.5). I think that it's a bug, but I don't know if there are some config that change that behaviour. I have uploaded the capture (png): https://i.imgur.com/emvnaCt.png Best regards -- Jose Fco. Irles Durá

1 0

Freeswitch loacally acks 200 in bypass media scenarios
by David VILLAUME 05 Mar '21

05 Mar '21

Hello, In bypass-media mode freeswitch locally acks the 200 ok (SDP), that leads to some Race condition when the party that answered the 200 sends a quick reinvite after Just after the Ack. I tried on 1.10.1 and 1.10.5, with and without late-negociation, but couldn’t figure how to avoid this. Has one of you experienced it or have any idea about this ? SBC FREESWITCH PBX +--------- ----------+--------- ----------+ | INVITE (SDP) | | | --------------------------> | | | 100 Trying | | | <-------------------------- | | | | INVITE (SDP) | | | --------------------------> | | | 100 Trying | | | <-------------------------- | | | 180 Ringing | | | <-------------------------- | | 180 Ringing | | | <-------------------------- | | | | 200 OK (SDP) | | | <-------------------------- | | | ACK | | | --------------------------> | | 200 OK (SDP) | | | <-------------------------- | | | | INVITE (SDP) | | | <-------------------------- | | | 100 | | | --------------------------> | | INVITE (SDP) | | | <-------------------------- | | | 100 | | | --------------------------> | | | 491 | | | --------------------------> | | | | 491 | | | --------------------------> | Regards, David

2 3

mod_sndfile error called from mod_tts_commandline - [WARNING] mod_sndfile.c:281 Error Opening File
by Lewis Bergman 05 Mar '21

05 Mar '21

Hello all. I am pretty new to Freeswitch and have an experimental container on Proxmox 6.3 Debian 10 to play with. The goal is to get mod_unimrcp working. To that end, I have mod_tts_commandline (installed from binary with FusionPBX as part of the default install) listed in modules.conf.xml /etc/freeswitch/autoload_configs/modules.conf.xml  <load module="mod_unimrcp"/> <load module="mod_tts_commandline"/> <load module="mod_flite"/> I have tried it with and without mod_flite enabled. Mod_tts_commandline refuses to pull a sound file and the debug log seems to blame mod_sndfile. This is using the example from the mod_tts_commandline page <https://freeswitch.org/confluence/display/FREESWITCH/mod_tts_commandline#mo…> as dialplan shows following the debug log output below: 2021-02-26 15:19:24.713040 [NOTICE] mod_dptools.c:1406 Channel [sofia/internal/102(a)66.151.243.45] has been answered 2021-02-26 15:19:24.713040 [DEBUG] switch_channel.c:3865 (sofia/internal/ 102(a)66.151.243.45) Callstate Change RINGING -> ACTIVE 2021-02-26 15:19:24.713040 [DEBUG] sofia.c:7326 Channel sofia/internal/ 102(a)66.151.243.45 entering state [completed][200] EXECUTE [depth=0] sofia/internal/102(a)66.151.243.45 speak(tts_commandline|pico|This is an example of using tts_commandline) 2021-02-26 15:19:24.713040 [DEBUG] switch_ivr_play_say.c:3023 OPEN TTS tts_commandline 2021-02-26 15:19:24.713040 [DEBUG] switch_ivr_play_say.c:3033 Raw Codec Activated 2021-02-26 15:19:24.713040 [DEBUG] mod_tts_commandline.c:160 Executing: echo 'This is an example of using tts_commandline' | text2wave -f 8000 > '/tmp/0665a30d-2bb8-40fe-b5a6-9625bd58306e.tmp.wav' 2021-02-26 15:19:25.433013 [WARNING] mod_sndfile.c:281 Error Opening File [/tmp/0665a30d-2bb8-40fe-b5a6-9625bd58306e.tmp.wav] [File contains data in an unknown format.] 2021-02-26 15:19:25.433013 [ERR] mod_tts_commandline.c:170 Failed to open file: /tmp/0665a30d-2bb8-40fe-b5a6-9625bd58306e.tmp.wav 2021-02-26 15:19:25.433013 [DEBUG] switch_ivr_play_say.c:2741 Speaking text: This is an example of using tts_commandline 2021-02-26 15:19:25.433013 [DEBUG] sofia.c:7326 Channel sofia/internal/ 102(a)66.151.243.45 entering state [ready][200] 2021-02-26 15:19:25.433013 [DEBUG] switch_ivr_play_say.c:2905 done speaking text 2021-02-26 15:19:25.433013 [NOTICE] switch_core_state_machine.c:386 sofia/internal/102(a)66.151.243.45 has executed the last dialplan instruction, hanging up. 2021-02-26 15:19:25.433013 [NOTICE] switch_core_state_machine.c:388 Hangup sofia/internal/102(a)66.151.243.45 [CS_EXECUTE] [NORMAL_CLEARING] The dialplan for extension I am testing looks like this: <extension name="tts_commandline_test"> <condition field="destination_number" expression="^561$"> <action application="speak" data="tts_commandline|pico|This is an example of using tts_commandline"/> </condition> </extension> The code lines mentioned in debug seem to be pretty mundane, not that I really know what I am looking for: >From mod_tts_commnandline line 160: switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Executing: %s\n", message); I tried setting the /tmp directory to 777 but that didn't help and the error output was identical. mod_tts_commandline never reports an error on creation and it looks like the next part of of mod_tts_commandline would print an error if it did like this: if (switch_system(message, SWITCH_TRUE) < 0) { switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Failed to execute command: %s\n", message); ret = SWITCH_STATUS_FALSE; goto done; } if (switch_core_file_open(info->fh, info->file, 0, //number_of_channels, info->rate, //samples_per_second, SWITCH_FILE_FLAG_READ | SWITCH_FILE_DATA_SHORT, NULL) !=SWITCH_STATUS_SUCCESS) { switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Failed to open file: %s\n", info->file); ret = SWITCH_STATUS_FALSE; goto done; } Starting at line 279 in mod_sndfile (sorry for the space formatting): if (!context->handle) { if (sndfile_perform_open(context, path, mode, handle) != SWITCH_STATUS_SUCCESS) { switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "Error Opening File [%s] [%s]\n", path, sf_strerror(context->handle)); status = SWITCH_STATUS_GENERR; goto end; } } According to the dptools docs <https://freeswitch.org/confluence/display/FREESWITCH/mod_dptools%3A+playbac…> and libsndfile wav has been supported forever and it would seem highly unlikely that is an issue. It seems like mod_tts isn't writing the file but also isn't reporting an error. I have started fresh with two new containers twice and a full VM once all with the same error resulting. I check the /tmp directory and no files are present but I don't know if they are being deleted by some cleanup process or just not created. Not knowing C, it seems like mod_sndfile never received a file handle but that is just a guess. Other Freeswitch apps are writing to the /tmp directory (at least there are some files in there). Maybe someone sees an error from mod_tts_commandline I don't see or knows I am doing something stupid. Any help would be appreciated. I only found 2 references to the error in the mail history and neither had anything to do with a similar issue. Freeswitch version reports: FreeSWITCH Version 1.10.5-release-17-25569c1631~64bit (-release-17-25569c1631 64bit) FusionPBX version, in case it matters, is : 4.5.21 OS is Debian 10 Buster -- Lewis Bergman

4 6

Starting dtmf_generation with mod_lcr
by Mimiko 04 Mar '21

04 Mar '21

Hello. I use lcr for outbound connection to providers based on phone call price. On one provider I need to use start_dtmf_generate to force recoding sip info from local extensions to inband dtmf. Setting variable from lcr execute_on_answer=start_dtmf_generate does not work. Also gateway profile have <variables> <variable direction="outbound" name="dtmf_type" value="none"/> </variables> <param name="dtmf-type" value="none"/> this also does not help. How can I accomplish per provider dtmf_generate for outbound calls?

1 0

Don't deprecate this option - useful for buggy user agents
by Martin Paterson 04 Mar '21

04 Mar '21

I had a problem which I solved by turning on a FreeSWITCH option that logs a warning: "Enabling reg-deny-binding-fetch-and-no-lookup - this functionality is deprecated and will be removed - let FS devs know if you think it should stay". So here is me letting the FS devs know. The problem is to do with buggy user agents and their handling of multiple Contact headers in the 200 OK response to their REGISTER. My FreeSWITCH system uses mainly Snom phones and Zoiper soft phones. I allows multiple registrations and I have two sip profiles that the user agents connect to, one for Snoms and one for Zoiper. They share a registration database with the dbname option. When multiple user agents register with the same credentials, the correct behaviour in the 200 OK to a REGISTER is to list all the user agents in multiple Contact headers with their individual registration expiry times. This is exactly what FreeSWITCH does. However both clients (Snom and Zoiper) fail in a similar way: For their own reasons (https://www.zoiper.com/en/support/home/article/205/Zoiper%20Push%20Proxy) Zoiper clients register with a 1 month expiry time. If in the 200 OK to a Snom the Zoiper contact header is first, then the Snom assumes that the header is meant for it, reads the large expiry time and uses it (and complains in its log if it's over 14 days). The FreeSWITCH registration for the Snom then expires after 1 hour, which what the Snom asked for in the first place. The Snom then doesn't receive calls. Annoyingly a similar thing happens the other way round - if the Snom Contact header is first in the 200 OK sent to Zoiper, then the Zoiper client just looks at the first header, and complains that the expiry time is too short with a message "You are connecting to a server with a low re-registration time. Battery life will not be optimal. Please contact your VoIP provider or PBX server administrator." (An aside: it would be interesting to know why returning multiple Contact headers is the defined behaviour, and what a user agent is supposed to do with the others headers, and therefore what the consequences of not having them might be.) We are contacting both Snom and Zoiper about this, but for the moment this is a useful FreeSWITCH option to cope with these user agents' behaviours. Martin Martin Paterson Development Team Phone: 0207 953 8840 Email: martin.paterson(a)technologywithin.com Chevron Business Park, Limekiln Lane, Southampton, Hampshire, SO45 2QL Registered Office: CP House, Otterspool Way, Watford, WD25 8JJ, U.K Registered in England No: 5964349 | VAT Number: GB 902 5369 37

1 0

mod lcr cid conditional replace.
by Mimiko 03 Mar '21

03 Mar '21

Hi. I use mod_lcr for dialout operators. On the freeswitch pbx we have extensions 500 to 899. When calling outside number the internal extension is converted ok to for external cid to be accepted by each operator using lcr_cid regular pattern: /(\d\d\d)/${operator_prefix}$1/ The problem lies with one operator where our extensions which start with 5 (500 to 599) must be converted to numbers which start with 9 (900 to 999) and then prefixed with operator's prefix. The rest of extensions does not have such a substitution of first number. How can I do this using regular expression in lcr_cid? Thank you.

1 0