Bug 2070193

Description of problem:

Issue:- when sending a trap, the second trap in the list does not get the proper source IP and cause erroneous determinations by the trap receiver and statistics and not aggregated properly.

Version-Release number of selected component (if applicable):

# rpm -qi net-snmp
Name        : net-snmp
Epoch       : 1
Version     : 5.8
Release     : 14.el8
Architecture: x86_64
Install Date: Mon 19 Oct 2020 02:32:09 PM UTC
Group       : Unspecified
Size        : 893577
License     : BSD
Signature   : RSA/SHA256, Mon 16 Mar 2020 11:41:14 AM UTC, Key ID 199e2f91fd431d51
Source RPM  : net-snmp-5.8-14.el8.src.rpm
Build Date  : Mon 16 Mar 2020 10:21:24 AM UTC
Build Host  : x86-vm-08.build.eng.bos.redhat.com
Relocations : (not relocatable)
Packager    : Red Hat, Inc. <http://bugzilla.redhat.com/bugzilla>
Vendor      : Red Hat, Inc.
URL         : http://net-snmp.sourceforge.net/
Summary     : A collection of SNMP protocol tools and libraries


So there should be a BZ opened for the issue and the fix appears to be roughly:

"re-initializing the new copy's source member (the original 'source' member is already squirrel away in a stack variable) seems to have done the trick"

attaching the RedHatCase03177898Diagnostic.tar.gz so that engineering can see how the problem happens


=====

I also downloaded the source from sourceforge for net-snmp 5.8 which is the version we are using and I think I found something in the net-snmp code, in snmplib/snmp_transport.c in netsnmp_tdomain_transport_tspec() there's a for loop

 for(;;) {
        if (match) {
            netsnmp_transport *t = NULL;
            const char* addr2;

            any_found = 1;
            /*
             * Ok, we know what domain to try, lets see what default data
             * should be used with it
             */
            if (default_target != NULL)
                addr2 = default_target;
            else
                addr2 = netsnmp_lookup_default_target(application,
                                                      match->prefix[0]);
            DEBUGMSGTL(("tdomain",
                        "trying domain \"%s\" address \"%s\" "
                        "default address \"%s\"\n",
                        match->prefix[0], addr ? addr : "[NIL]",
                        addr2 ? addr2 : "[NIL]"));
            if (match->f_create_from_tspec) {
                netsnmp_tdomain_spec tspec_tmp;
                memcpy(&tspec_tmp, tspec, sizeof(tspec_tmp));

so it creates essentially a local copy of the passed in tspec in course of processing the 'source' member of tspec can get initialized from NULL (if that's the case) and so the second time around through this loop, the new copy inherits the updated 'source' and then tries to bind and fails saying 'Address in use' so then this trapsink (the second one in the list regardless of any specifics)
doesn't do the right thing. so reinitializing the new copy's source member (the original 'source' member is already squirrel away in a stack variable) seems to have done the trick