Changeset 794

Timestamp:

10/22/20 14:05:25 (4 years ago)

Author:

Kris Deugau

Message:

/trunk

Commit critical-failure-shortstopping fix from production around RPC delete.

If an RPC request is made to delete a nested subzone, records outside of that
range in the "parent" could be deleted as well.

File:

• trunk/dns-rpc.cgi (modified) (2 diffs)

trunk/dns-rpc.cgi

@@ -1683 +1683 @@
     # Overlapping reverse zones shouldn't be possible, so if we're here we've got a CIDR
     # that spans multiple reverse zones (eg, /23 CIDR -> 2 /24 rzones)
+    # 2018/09/18 found an edge case, of course;  if you've hacked IPDB to allow branched master
+    # blocks you *can* end up with nested reverse zones, in which case deleting a record in one
+    # may axe records in the other.  dunno if it affects cidr-in-large axes recs-in-smaller, but
+    # I have an active failure for cidr-in-smaller axes recs-in-larger.  eeep.
     foreach my $zdata (@$zonelist) {
       my $reclist = $dnsdb->getRecList(rpc => 1, defrec => 'n', revrec => 'y', id => $zdata->{rdns_id});
@@ -1697 +1701 @@
           # Template types are only useful when attached to a reverse zone.
 ##fixme  ..... or ARE THEY?
+          # edge case:  if we have nested zones, make sure that we do not delete records outside of
+          # the passed $cidr.  This is horrible-ugly-bad, especially when said out-of-scope records
+          # constitute key core network names...
+##fixme:  should this check be moved into getRecList as a search restriction of some kind?
+#  cf args{filter}, but we really need to leverage the DB's IP type handling for this to be worthwhile
+          my $rcidr = new NetAddr::IP $rec->{val};
+          next unless $cidr->contains($rcidr);
           if ($args{delforward} ||
               $rec->{type} == 12 || $rec->{type} == 65282 ||