I observed that the inline function suip2a() (ip_addr.h) contains a silly bug and therefor does not work for IPv6 conversions. This silly bug is preventing sip_trace() to work in the onsend_route block for IPv6 sip messages.
Rem.: The function suip2a() is used only by the siptrace module.

Test background (just for understanding the use case):

• using kamailio 4.1.3 (with an additional patch to let onsend_route be called also for replies taken from kamailio 4.2.x)
• kamailio is used as loadbalancer in stateless mode (using forward() only)
• sip tracing is used by explictly calling sip_trace() in onsend_route, route and onreply routing block to be able to implement an interface specific sip tracing (no SIP transaction based sip tracing is applied by using the siptrace's trace_flag)

The silly bug:

In suip2a() the function ip6tosbuf() is called with an odd buffer length argument of value "sizeof(buf)-4" which is equal to "IP6_MAX_STR_SIZE -1". ip6tosbuf() however needs at least a buffer length of IP6_MAX_STR_SIZE (otherwise it returns immediately without writing anything into the buffer).
So regarding IPv6 adresses the ascii output of suip2a() is alway an empty IPv6 reference ([]).

static inline int ip6tosbuf(unsigned char* ip6, char* buff, int len)
{
    int offset;
    register unsigned char a,b,c;
    register unsigned char d;
    register unsigned short hex4;
    int r;
    #define HEXDIG(x) (((x)>=10)?(x)-10+'A':(x)+'0')


    offset=0;
    if (unlikely(len<IP6_MAX_STR_SIZE))
        return 0;
    ...
}

#define SUIP2A_MAX_STR_SIZE  (IP6_MAX_STR_SIZE + 2 /* [] */ + 1 /* \0 */)
/* returns an asciiz string containing the ip
 *  (<ipv4_addr> or [<ipv6_addr>])
 */
static inline char* suip2a(union sockaddr_union* su, int su_len)
{
    static char buf[SUIP2A_MAX_STR_SIZE];
    int offs;

    if (unlikely(su->s.sa_family==AF_INET6)){
        if (unlikely(su_len<sizeof(su->sin6)))
            return "<addr. error>";
        buf[0]='[';
        offs=1+ip6tosbuf((unsigned char*)su->sin6.sin6_addr.s6_addr, &buf[1],
                            sizeof(buf)-4);
        buf[offs]=']';
        offs++;
    }else
    if (unlikely(su_len<sizeof(su->sin)))
        return "<addr. error>";
    else
        offs=ip4tosbuf((unsigned char*)&su->sin.sin_addr, buf, sizeof(buf)-2);
    buf[offs]=0;
    return buf;
}

A possible verfied patch looks like this (using this patch sip_trace() is working now as expected in this use case).

--- ip_addr.h   2014-04-24 12:02:35.000000000 +0200
+++ ip_addr.h   2015-10-22 18:48:40.553177206 +0200
@@ -737,14 +737,14 @@
            return "<addr. error>";
        buf[0]='[';
        offs=1+ip6tosbuf((unsigned char*)su->sin6.sin6_addr.s6_addr, &buf[1],
-                           sizeof(buf)-4);
+                           IP6_MAX_STR_SIZE);
        buf[offs]=']';
        offs++;
    }else
    if (unlikely(su_len<sizeof(su->sin)))
        return "<addr. error>";
    else
-       offs=ip4tosbuf((unsigned char*)&su->sin.sin_addr, buf, sizeof(buf)-2);
+       offs=ip4tosbuf((unsigned char*)&su->sin.sin_addr, buf, IP4_MAX_STR_SIZE);
    buf[offs]=0;
    return buf;
 }

—
Reply to this email directly or view it on GitHub.