[LinuxPPS] some more info on recent attempt at nmea patch

[Udo van den Heuvel]

Udo van den Heuvel wrote:
> So I started ntpd with some debugging and it finds a buffer overflow:
> 
> [root at epia ntpd]# ./ntpd -d -d -d

[...]

> ffffe000-fffff000 ---p 00000000 00:00 0          [vdso]
> Aborted

I found that the code probably stops bafore completing this piece of
code in the nmea refclock:

        /* Try to find the source */
        fd = time_pps_findpath(path, 70, id, 70);

(yes, I increased the size of the arrays a bit)

Some added logging in timepps.h reveals something goes wrong after
binding the socket, so probably here:

        /* Ask the kernel to destroy the PPS source */
        nlpps.cmd = PPS_FIND_PATH;
        strncpy(nlpps.path, path, pathlen);
        ret = netlink_msg(sock, &nlpps);
        if (ret < 0) {
                close(sock);
                return ret;
        }
        msyslog(LOG_ERR, "findpath: PPS source destroyed");

        strncpy(path, nlpps.path, pathlen);
        strncpy(idstring, nlpps.name, idlen);

        close(sock);
        msyslog(LOG_ERR, "findpath: PPS socket closed");
        return nlpps.source;
}

(the 2 msyslogs are never reached)

Conclusion?
Rodolfo?