[SR-Users] Re: Depending on topoh.mask_key Kamailio cannot parse RURIs or modifies magically the SIP uri by changing some bytes in it (2)
Hello Henning, thanks for your answer. The symptoms are that the BYE cannot be handled properly, when the caller is behind websockets. When the caller has established TLS connection, then BYE from caller is delivered properly to the TCP callee. In December in this thread https://lists.kamailio.org/mailman3/hyperkitty/list/sr-users@lists.kamailio.org/thread/J5PHX5O6HVOZH5DMQZIQRJYHXMI3GXUQ/ it was again reported that with topoh just BYE does not work. But it says nothing about websockets. I enable system (normal) malloc/free by adding to src/Makefile.defs: diff --git a/src/Makefile.defs b/src/Makefile.defs index eb71d6cd0b..648102d63d 100644 --- a/src/Makefile.defs +++ b/src/Makefile.defs @@ -118,6 +118,8 @@ MEMMNG ?= 0 # 0 - off (no-debug mode) # 1 - on (debug mode) MEMDBG ?= 1 +MEMPKG = sys +MEMDBGSYS = 1 VERSIONVAL = $(shell expr $(VERSION) \* 100 + $(PATCHLEVEL) \* 1000 + \ $(SUBLEVEL) ) kamailio -I prints: Print out of kamailio internals Version: kamailio 5.6.3 (x86_64/linux) 641148 Default config: /etc/kamailio/kamailio.cfg Default paths to modules: /lib64/kamailio/modules Compile flags: USE_TCP, USE_TLS, USE_SCTP, TLS_HOOKS, USE_RAW_SOCKS, DISABLE_NAGLE, USE_MCAST, DNS_IP_HACK, SHM_MMAP, Q_MALLOC, F_MALLOC, TLSF_MALLOC, DBG_SR_MEMORY, USE_FUTEX, FAST_LOCK-ADAPTIVE_WAIT, USE_DNS_CACHE, USE_DNS_FAILOVER, USE_NAPTR, USE_DST_BLOCKLIST, HAVE_RESOLV_RES, TLS_PTHREAD_MUTEX_SHARED MAX_RECV_BUFFER_SIZE=262144 MAX_URI_SIZE=1024 BUF_SIZE=65535 DEFAULT PKG_SIZE=8MB DEFAULT SHM_SIZE=64MB ADAPTIVE_WAIT_LOOPS=1024 TCP poll methods: poll, epoll_lt, epoll_et, sigio_rt, select Source code revision ID: 641148 Compiled with: gcc 12.2.1 Compiled architecture: x86_64 Compiled on: 15:50:37 Feb 3 2023 Thank you for flying kamailio! as the output includes Q_MALLOC, F_MALLOC, TLSF_MALLOC, DBG_SR_MEMORY how can I know that the system malloc/free are used? Running under valgrind shows no problems, so my changes to src/Makefile.defs might not trigger system malloc/free use. Any suggestions? To sum up: - without TOPOH, there are no problems - with TOPOH, caller is behind TLS, callee is TCP-connected, there are no problems - with TOPOH, caller is behind websocket, calle is TCP-connected, there are problems: some bytes in the URL of the destination are replaced with unprintable characters, or the R-URI is reported as unparsable (depending on the value of topoh.mask_key). Moreover I have both loaded outbound-module and have configured the websocket provisions, when websockets is used without outbound. As last speciality, the websocket client connects over secured-TLS (https/wss) to the webserver and the latter connects to Kamailio over insecure HTTP/TCP. That is the webclient includes Via:/WSS, but Kamailio adds transport=ws (not transport=wss). Any hints? Greetings Dilyan -Original Message- From: Henning Westerholt To: Kamailio (SER) - Users Mailing List Cc: Дилян Палаузов Subject: RE: [SR-Users] Depending on topoh.mask_key Kamailio cannot parse RURIs or modifies magically the SIP uri by changing some bytes in it (2) Date: 02/02/2023 11:27:47 PM Hello, regarding shared memory, Kamailio uses usually shared memory, regardless of how many workers are configured. About compiling Kamailio without its internal memory pool, its possible. This was discussed last month in a thread about memory performance, if I remember correctly, on the list. Cheers, Henning -- Henning Westerholt – https://skalatan.de/blog/ Kamailio services – https://gilawa.com -Original Message- From: Дилян Палаузов Sent: Wednesday, February 1, 2023 4:11 PM To: Kamailio - Users Mailing List Subject: [SR-Users] Depending on topoh.mask_key Kamailio cannot parse RURIs or modifies magically the SIP uri by changing some bytes in it (2) Hello, does somebody has success, when using websockets and topoh? If I run a single Kamailio process, I guess I do not need real shared memory. Can I adjust Kamailio somehow to use just malloc()/free() in this case, so that debugging is easier? As it turned out dns_cache_init=off leads to crashes - https://github.com/kamailio/kamailio/issues/3350 . With dns_cache_init=on and use_dns_cache=on, does not lead to reading invalid data. Running Kamailio under valgrind: When I do not set topoh.mask_key (so the default is used), for the yesterday mentioned workflow, for incoming packet: 18(19) ERROR: [core/kemi.c:96]: sr_kemi_core_err(): ksr_sipdump_event 1- sipdump:msg src 144.76.142.78:48294 dst 144.76.142.78:5060 ACK sip:127.3.4.84;line=sr-n6iazhainwp1g6s5orj136tdouwy3etacsf73ejxgwvqp9e7zhkucrp- gbeewrfl36313.0e3j4amxpworeiw.y1wlpagevs3eabw.3lz6tdcrtlh.cwobffzgzfwx37m.jlibmlzxczibv* SIP/2.0 Route: Route: Via: SIP/2.0/WSS lvqj9uhih64a.invalid;branch=z9hG4bK962314 Max-Forwards: 70 To: ;tag=5Ml7bJz From: "Online https://sip.bapha.be"; ;tag=kab7lqjdgi Call-ID: i7jjtim7h775c6vqqudg CSeq
[SR-Users] topoh: th_mask_decode (th_mask_encode (x)) ≠ x
Hello,
after receiving
INVITE sip:test-gnome-ca...@bapha.be SIP/2.0
Via: SIP/2.0/WSS bapha.be;branch=z9hG4bK974100
Max-Forwards: 70
To:
From: "Online https://sip.bapha.be"; ;tag=12q2efc423
Call-ID: fkhe8faq0fh80glmq59i
CSeq: 5622 INVITE
Contact:
Allow: ACK,CANCEL,INVITE,MESSAGE,BYE,OPTIONS,INFO,NOTIFY,REFER
Supported: outbound
User-Agent: SIP.js/0.7.8
Content-Type: application/sdp
Content-Length: 1964
v=0
o=mozilla...THIS_IS_SDPARTA-99.0 8263261229339246488 0 IN IP4 0.0.0.0
s=-
…
kamailio calls
th_mask_encode(
char *in = "sip:test-gnome-ca...@bapha.be;gr=urn:uuid:f2f5a3cf-a0fb-0047-be50-
740fb9bdc562;alias=87.118.146.153~60722~2>;+sip.instance=\"";+org.linphone.specs=\"conference/1.0,ephemeral/1.1,groupchat/1.1,groupchat/1.2,lime\n"\
"Content-Type: application/sdp and so on",
int ilen = 107,
str prefix = { .len = 23, .s ="sip:127.3.4.84;line=sr-" },
int *olen).
It returns olen=167, and the string "sip:127.3.4.84;line=sr-
if7s1mg7i36PNf0AbdwPpfzlbqWEpLzsSGItpLwyN39ZMY4t1mCTSd6DNo4LWdIOpfpPp5FLpUQscX63Kd47W5EPWO6sNL90pLgoW5p-1fzlSdzO25n3KoIk1vHkWXptc5wOeopsWO9-eo9*".
Then Kamailio forwards (sends):
Contact:
;+sip.instance="";+org.linphone.specs="conference/1.0,ephemeral/1.1,groupchat/1.1,groupchat/1.2,lime"
which at some point leads to
ACK
sip:127.3.4.84;line=sr-if7s1mg7i36pnf0abdwppfzlbqweplzssgitplwyn39zmy4t1mctsd6dno4lwdiopfppp5flpuqscx63kd47w5epwo6snl90plgow5p-
1fzlsdzo25n3koik1vhkwxptc5woeopswo9-eo9* SIP/2.0
and
BYE
sip:127.3.4.84;line=sr-if7s1mg7i36pnf0abdwppfzlbqweplzssgitplwyn39zmy4t1mctsd6dno4lwdiopfppp5flpuqscx63kd47w5epwo6snl90plgow5p-
1fzlsdzo25n3koik1vhkwxptc5woeopswo9-eo9* SIP/2.0
Below you can find th_test.c . The functions th_mask_encode() and
th_mask_decode() there are identical to the same functions in
src/modules/topoh/th_mask.c . main() provides some tests:
Passing «sip:test-gnome-ca...@bapha.be;gr=urn:uuid:f2f5a3cf-a0fb-0047-be50-
740fb9bdc562;alias=87.118.146.153~60472~2>;+sip.instance="";+org.linphone.specs="conference/1.0,ephemeral/1.1,groupchat/1.1,groupchat/1.2,lime»
to th_mask_encode() returns
sip:127.3.4.84;line=sr-if7s1mg7i36PNf0AbdwPpfzlbqWEpLzsSGItpLwyN39ZMY4t1mCTSd6DNo4LWdIOpfpPp5FLpUQscX63Kd47W5EPWO6sNL90pLgoW5p-
1fzlSdzO25n3KoIk1vHkWXptc5wOeopsWXi-eo9*
(this is what Kamailio does send out). Passing the last string to
th_mask_decode() returns
sip:test-gnome-ca...@bapha.be;gr=urn:uuid:f2f5a3cf-a0fb-0047-be50-740fb9bdc562;alias=87.118.1
the ;alias= is incomplete, compared to the origin.
The above ACK leads to passing
"sr-if7s1mg7i36pnf0abdwppfzlbqweplzssgitplwyn39zmy4t1mctsd6dno4lwdiopfppp5flpuqscx63kd47w5epwo6snl90plgow5p-1fzlsdzo25n3koik1vhkwxptc5woeopswo9-eo9*
SIP/2.0\n"\
"Route:
\n"\
"Route:
\n"\
to th_mask_decode() and the result is
callub.p.b;rn:s.d2lQg#cfa=bA07beQ7R409bcQ62;alac=87718*1P.15#~60R22~2 with the
dots being unprintable
characters. (result is garbage).
I have seen that callub.p.b in the unparsable R-URI reports from Kamailio.
That is an indication that th_mask_decode() does not decode the Contact encoded
by th_mask_encode() correctly.
Valgrind does not report improper memory usage.
Greetings
Dilyan
gcc -g -o th-test ./th-test.c && ./th-test
th-test.c
#include
#include
#include
#define TH_EB64I \
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789.-"
char _th_EB64[65] =
"EFvXIzGq94xKcW126gV5wCdYpNSbiM8enRUou7LmhajBlPtAsk-OQTf3H0DyrZ.J";
int _th_DB64[256];
char *_th_PD64 = "*";
struct str_ {
char* s; /**< Pointer to the first character of the string */
int len; /**< Length of the string */
};
typedef struct str_ str;
str prefix = { .len = 23, .s ="sip:127.3.4.84;line=sr-" };
char* th_mask_encode(char *in, int ilen, const str *prefix, int *olen);
char* th_mask_decode(char *in, int ilen, const str *prefix, int extra, int
*olen);
char* th_mask_encode(char *in, int ilen, const str *prefix, int *olen)
{
char *out;
int left;
int idx;
int i;
int r;
char *p;
int block;
*olen = (((ilen+2)/3)<<2) +
((prefix!=NULL&&prefix->len>0)?prefix->len:0);
out = (char*)malloc((*olen+1)*sizeof(char));
if(out==NULL)
{
fprintf(stderr, "malloc error\n");
*olen = 0;
return NULL;
}
memset(out, 0, (*olen+1)*sizeof(char));
if(prefix!=NULL&&prefix->len>0)
memcpy(out, prefix->s, prefix->len);
p = out + (int)((prefix!=NULL&&prefix->len>0)?prefix->len:0);
for(idx=0; idx1)?2:left;
block = 0;
for(i=0, r=16; i<=left; i++, r-=8)
block += ((unsigned char)in[idx+i]) << r;
*(p++) = _th_EB64[(block >> 18) & 0x3f];
*(p++) = _th_EB64[(block >> 12) & 0x3f];
*(p++) = (left>0)?_th_EB64[(block >> 6) & 0x3f]:_th_PD64[0];
*(p++) = (left>1)?_th_EB64[block & 0x3f]:_th_PD64[0];
}
r
[SR-Users] no autocomplete for kamcmd for latest debian packages
Hello, The latest debian packages are built without autocomplete for kamcmd. Do we have the libreadline-dev package installed on the build machine? I know that we have kamctl rpc and kamcli, but if we are still shipping kamcmd, it should have the autocomplete feature enabled. Regards, Ovidiu Sas -- VoIP Embedded, Inc. http://www.voipembedded.com __ Kamailio - Users Mailing List - Non Commercial Discussions To unsubscribe send an email to sr-users-le...@lists.kamailio.org Important: keep the mailing list in the recipients, do not reply only to the sender! Edit mailing list options or unsubscribe:
