getaddrinfo fails to parse IPv6 address on some systems

[alistairking]

As reported by @vgiotsas

It seems that the way we use getaddrinfo (https://github.com/CAIDA/bgpstream/blob/master/lib/utils/bgpstream_utils_addr.c#L209) will fail to parse an IPv6 address on (some?) IPv6-only systems. Probably omitting AI_ADDRCONFIG from hints would fix this, but it seems there can be issues with getaddrinfo (see https://blog.powerdns.com/2014/05/21/a-surprising-discovery-on-converting-ipv6-addresses-we-no-longer-prefer-getaddrinfo/), so we may want to consider switching to inet_pton...

[alistairking]

I just personally ran into this problem on one of our nodes, and whipped up the following fix. It'll need to be tested more, but it seems to work...

bgpstream_addr_storage_t *bgpstream_str2addr(char *addr_str,
                                             bgpstream_addr_storage_t *addr)
{
  if (addr_str == NULL || addr == NULL) {
    return NULL;
  }

  if (strchr(addr_str, ':') != NULL) {
    /* this looks like it will be an IPv6 address */
    if (inet_pton(AF_INET6, addr_str, &addr->ipv6) != 1) {
      fprintf(stderr, "ERROR: Could not parse address string %s\n", addr_str);
      return NULL;
    }
    addr->version = BGPSTREAM_ADDR_VERSION_IPV6;
  } else {
    /* probably a v4 address */
    if (inet_pton(AF_INET, addr_str, &addr->ipv4) != 1) {
      fprintf(stderr, "ERROR: Could not parse address string %s\n", addr_str);
      return NULL;
    }
    addr->version = BGPSTREAM_ADDR_VERSION_IPV4;
  }

  return addr;
}

[digizeph]

getaddrinfo does DNS lookups. It seems that in multi-threading workflow, it can cause problem, likely because of the DNS lookups.

"The getaddrinfo() function combines the functionality provided by the gethostbyname(3) and getservbyname(3) functions into a single interface"

The domain name queries carried out by gethostbyname() and
       gethostbyaddr() rely on the Name Service Switch (nsswitch.conf(5))
       configured sources or a local name server (named(8)).  The default
       action is to query the Name Service Switch (nsswitch.conf(5))
       configured sources, failing that, a local name server (named(8)).

The usage of Name Service Switch was pointed out to be problematic in here and here.

Addding AI_NUMERICHOST flag could potentially resolve the problem, but I don't think getaddrinfo is the right tool to use here.

inet_pton() does not do anything "complicated", thus looks safer to use (https://beej.us/guide/bgnet/html/multi/inet_ntopman.html, "These functions don't do DNS lookups—you'll need getaddrinfo() for that.")

In any case, we don't seem to use DNS at all in the context, thus using inet_pton() is better even without the crashing issue. The revised wrapper function also looks good.

[digizeph]

the issue should be fixed by #69 .