Daniel-Constantin Mierla writes:

Do you have the rr modparam related socket name mode set?

Yes, sorry, forgot to mention it:

modparam("rr", "sockname_mode", 1)

I have added some debug to loose.c and some converted DBGs to INFOs in rr_do_force_send_socket() function:

if(rr_sockname_mode!=0 && puri->params.len>0) { s = puri->params; if(s.s[s.len-1]==';') { s.len--; } LM_INFO("Looking ;sn param from %.*s\n", s.len, s.s); ... LM_INFO("use of sockname parameter enabled, but failed to find it\n"); free_params(plist); }

This is what I got to syslog:

eb 12 14:53:07 lab /usr/bin/sip-proxy[8662]: INFO: rr [loose.c:759]: rr_do_force_send_socket(): Looking ;sn param from transport=tls;r2=on;sn=ext_tls;lr;n1;savp=avp;pm=0 Feb 12 14:53:07 lab /usr/bin/sip-proxy[8662]: INFO: rr [loose.c:783]: rr_do_force_send_socket(): use of sockname parameter enabled, but failed to find it Feb 12 14:53:07 lab /usr/bin/sip-proxy[8662]: WARNING: rr [loose.c:800]: rr_do_force_send_socket(): no socket found to match second RR (sip:tenantX.teams.tutpro.com:8007;transport=tls;r2=on;sn=ext_tls;lr;n1;savp=avp;pm=0)

So looks like some bug in the looking, since ;sn is there.

-- Juha

Juha Heinanen writes:

I think I found the bug:

./src/core/config.h:#define SOCKNAME_PARAM ";sn="

but in parse_params() result params are without ";".

Also, there is no "=". So instead of ";sn=", the code should look for "sn".

-- Juha

On 12.02.21 15:24, Daniel-Constantin Mierla wrote:

On 12.02.21 15:20, Juha Heinanen wrote:

...

Juha Heinanen writes:

...

I think I found the bug:

./src/core/config.h:#define SOCKNAME_PARAM ";sn="

but in parse_params() result params are without ";".

Also, there is no "=". So instead of ";sn=", the code should look for "sn".

Indeed, I will push a fix for it.

See if works now with latest master.

Cheers, Daniel

On 12.02.21 15:44, Juha Heinanen wrote:

Daniel-Constantin Mierla writes:

...

...

Also, there is no "=". So instead of ";sn=", the code should look for "sn".

Indeed, I will push a fix for it.

I also did one, but yours is cleaner than mine. Thanks. Tested and it works, no more warnings to syslog.

Looks like no-one had ever before used this socket name feature in R-R URIs.

I added it for convenience, after building a configuration for an older kamailio version where I use check_rout_param() function to figure out what socket to force. But that deployment was not yet upgraded to use this new mode, it runs well with the old approach.

I tested to see if the sn parameter is added in Record-Route, not the callbacks on Route headers.

I will backport to 5.4.

Cheers, Daniel