16 Apr
2008
16 Apr
'08
6:03 p.m.
Hi,
Im'm cross-posting this from OpenSER-Devel because I am not 100% sure
this is a bug.
I am trying to create openser configuration for proxy that connects
private (VPN) network with public network. I started with a config
similar to http://voipembedded.com/resources/openser_cr.cfg, added a
call to force_send_socket() to send from public IP and
rtpproxy/nathelper in bridge mode. As per config loose_route() is called
from openser.cfg to route a sequential request within a dialog through
record-routing.
Now when a BYE from callee comes in, it contains two Route headers:
Route: <a.b.c.d;r2=on;lr;ftag=...> where a.b.c.d is server's public IP,
Route: <10.0.0.100;r2=on;lr;ftag=...> where 10.0.0.100 is a private IP
and username(a)a.b.c.d in Request-URI.
Take a look at loose_route():
/*
* Do loose routing as defined in RFC3261
*/
int loose_route(struct sip_msg* _m, char* _s1, char* _s2)
{
int ret;
if (find_first_route(_m) != 0) {
LM_DBG("There is no Route HF\n");
return -1;
}
if (parse_sip_msg_uri(_m)<0) {
LM_ERR("failed to parse Request URI\n");
return -1;
}
ret = is_preloaded(_m);
if (ret < 0) {
return -1;
} else if (ret == 1) {
return after_loose(_m, 1);
} else {
#ifdef ENABLE_USER_CHECK
if (is_myself(&_m->parsed_uri.user, &_m->parsed_uri.host,
_m->parsed_uri.port_no)) {
#else
if (is_myself(&_m->parsed_uri.host, _m->parsed_uri.port_no)) {
#endif
return after_strict(_m);
} else {
return after_loose(_m, 0);
}
}
}
Obviously, is_myself() check passes and after_strict() is called.
when after_strict() is called, it looks if the URI given in Route
contains lr parameter which marks loose routers, if so this block of
code is executed:
{
LM_DBG("Next hop: '%.*s' is loose router\n",
uri.len, ZSW(uri.s));
if(get_maddr_uri(&uri, &puri)!=0) {
LM_ERR("failed to check maddr\n");
return RR_ERROR;
}
if (set_dst_uri(_m, &uri) < 0) {
LM_ERR("failed to set dst_uri\n");
return RR_ERROR;
}
/* Next hop is a loose router - Which means that is is not
endpoint yet
* In This case we have to recover from previous strict
routing, that
* means we have to find the last Route URI and put in in R-URI and
* remove the last Route URI. */
if (rt != hdr->parsed) {
/* There is a previous route uri which was 2nd uri of mine
* and must be removed here */
rem_off = hdr->body.s;
rem_len = rt->nameaddr.name.s - hdr->body.s;
if (!del_lump(_m, rem_off - _m->buf, rem_len, 0)) {
LM_ERR("failed to remove Route HF\n");
return RR_ERROR;
}
}
res = find_rem_target(_m, &hdr, &rt, &prev);
if (res < 0) {
LM_ERR("searching for last Route URI failed\n");
return RR_ERROR;
} else if (res > 0) {
/* No remote target is an error */
return RR_ERROR;
}
uri = rt->nameaddr.uri;
if(get_maddr_uri(&uri, 0)!=0) {
LM_ERR("checking maddr failed\n");
return RR_ERROR;
}
if (rewrite_uri(_m, &uri) < 0) {
LM_ERR("failed to rewrite R-URI\n");
return RR_ERROR;
}
/* The first character if uri will be either '<' when it is the
* only URI in a Route header field or ',' if there is more than
* one URI in the header field */
LM_DBG("The last route URI: '%.*s'\n",
rt->nameaddr.uri.len,
ZSW(rt->nameaddr.uri.s));
if (prev) {
rem_off = prev->nameaddr.name.s + prev->len;
rem_len = rt->nameaddr.name.s + rt->len - rem_off;
} else {
rem_off = hdr->name.s;
rem_len = hdr->len;
}
if (!del_lump(_m, rem_off - _m->buf, rem_len, 0)) {
LM_ERR("failed to remove Route HF\n");
return RR_ERROR;
}
}
Therefore recover from previous strict routing is triggered: contents of
request line is destroyed and 10.0.0.100 is put instead:
Apr 15 16:16:24 vpn-proxy oser[11027]: DBG:core:rewrite_uri: rewriting
Request-URI with
'sip:10.0.0.100;r2=on;lr;ftag=aa40c2a5-54fb-1810-970d-0015c5bf2da5'
Apr 15 16:16:24 vpn-proxy oser[11032]: [1] Got in dialog request not
catched by loose_route(): BYE
sip:10.0.0.100;r2=on;lr;ftag=aa40c2a5-54fb-1810-970d-0015c5bf2da5
SIP/2.0^M Via:...
And we respond with 404 to in-dialog requests not caught by
loose_route(). Previous hop is SER and I don't know why OpenSER
considered it a strict router. Looks like a peculiarity of ALG mode.
Please advice what should I do to solve this problem. Is it a
configuration issue or perhaps loose_route() should not call
after_strict() in case of double record-routing?
--
Thank you,
Andrew Pogrebennyk