[Pkg-sysvinit-devel] Bug#612823: sysvinit: Regarding ipAddr printed by utmpdump

Navdeep Bhatia navdeep at aristanetworks.com
Thu Feb 10 22:09:19 UTC 2011


Package: sysvinit
Version: 2.88

X-Debbugs-CC: "Navdeep Bhatia" <navdeep at aristanetworks.com>, "Fred
Xia" <fxia at aristanetworks.com>, "Henrique de Moraes Holschuh"
<hmh at debian.org>,



*Bug Description:*

The addr field of the output from utmpdump contains the ip address of the
connection. This is obtained by calling inet_ntoa on the ut.ut_addr field
(as depicted in the code below, utmpdump.c, Release 2.88). But in case the
ip address is an IPV6 address, the output contains the wrong ip address
(decimal value of first 4 octets only).

This code should be updated to call inet_ntop and should be checked whether
the ut_addr field contains IPV4 or IPV6 address. The check can be made
either by checking ut_addr_v6[1:3] for zero (this is based on the assumption
that login program zeroes out fields before filling utmp structure, though
it is not mentioned in man entry for utmp) or by introducing a flag in the
utmp structure that depicts whether the address field contains IPV4 or IPV6
address.


void
print_utline(struct utmp ut)
{
        char *addr_string, *time_string;
        struct in_addr in;

        in.s_addr = ut.ut_addr;
        *addr_string = inet_ntoa(in);   //Code statement causes issue in
case of IPV6 addresses*
        time_string = timetostr(ut.ut_time);
        cleanse(ut.ut_id);
        cleanse(ut.ut_user);
        cleanse(ut.ut_line);
        cleanse(ut.ut_host);

        /*            pid    id       user     line     host     addr
time */
        printf("[%d] [%05d] [%-4.4s] [%-*.*s] [%-*.*s] [%-*.*s] [%-15.15s]
[%-28.28s]\n",
               ut.ut_type, ut.ut_pid, ut.ut_id, 8, UT_NAMESIZE, ut.ut_user,
               12, UT_LINESIZE, ut.ut_line, 20, UT_HOSTSIZE, ut.ut_host,
               addr_string, time_string);
}


*Solution:*

a) One solution could be to look for ut_addr_v6[1:3] for zero. This is based
on the assumption that login program zeroes out the fields before filling
utmp structure. In this case the updated program would be:

void
print_utline(struct utmp ut)
{
        char addr_string[INET6_ADDRSTRLEN], *time_string;
        int addr_family = AF_INET6;

        if( ut.ut_addr_v6[1] == 0 &&
              ut.ut_addr_v6[2] == 0 &&
              ut.ut_addr_v6[3] == 0 )
        {
           addr_family = AF_INET;
        }
        if( NULL == inet_ntop( addr_family, ut.ut_addr_v6, addr_string,
                 INET6_ADDRSTRLEN ) )
        {
           perror( "inet_ntop" );
           exit(EXIT_FAILURE);
        }

        time_string = timetostr(ut.ut_time);
        cleanse(ut.ut_id);
        cleanse(ut.ut_user);
        cleanse(ut.ut_line);
        cleanse(ut.ut_host);

        /*            pid    id       user     line     host     addr
time */
        printf("[%d] [%05d] [%-4.4s] [%-*.*s] [%-*.*s] [%-*.*s] [%-.45s]
 [%-28.28s]\n",
               ut.ut_type, ut.ut_pid, ut.ut_id, 8, UT_NAMESIZE, ut.ut_user,
               12, UT_LINESIZE, ut.ut_line, 20, UT_HOSTSIZE, ut.ut_host,
               addr_string, time_string);
}


b) The other solution could be to introduce a new flag in utmp structure
that fills whether the address is IPV4 or IPV6.


Thanks.

Regards,
Navdeep
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-sysvinit-devel/attachments/20110210/3ae7de81/attachment.htm>


More information about the Pkg-sysvinit-devel mailing list