Changeset 824

Timestamp:

12/15/20 17:28:55 (3 years ago)

Author:

Kris Deugau

Message:

/trunk

Refresh leading chunk of bind-import from bind2hosts, remove multiple
previous development comment chunks

File:

• trunk/bind-import (modified) (11 diffs)

trunk/bind-import

@@ -188 +188 @@
                                 # arguably should do some more targeted voodoo when parsing the SOA details
 
-##fixme:  would prefer to break the loop below *AND* next; the while starting above
     # check skiplist.  do this early since it's (mostly) a simple string match against the raw record line
     my $skipflag = 0;
     foreach (@skipdefs) {
-#print "skipdbg: $_ =~ $rec\n" if $rec =~ /207/;
       if ($rec =~ /\Q$_\E/) {
         $skipflag = 1;
-#        print "skip: $rec\n";
+        # might want to do something with the skipped records someday
       }
     }
     next if $skipflag;
 
-##fixme:  use external skiplist
-#  # skip stale records that have no value
-#  next if /^ip-192-168-1(12|20)-\d+/;
-#  next if /ip.add.re.\d+\s*$/;
 
 $i++;
@@ -218 +212 @@
         }
       } elsif ($macro eq 'ORIGIN') {
-##fixme:  going to skip the stupid case of "$ORIGIN com." and the like that lie
-# between . and the root domain we were told we're importing;  anyone using such
-# a mess outside the root servers is clearly insane
-
-# $ORIGIN supports cascading/nesting, by watching for fully-qualified names vs partial names.
-
-print "origin ($mdetail)\n";
+        # $ORIGIN supports cascading/nesting, by watching for fully-qualified names vs partial names.
         if ($mdetail =~ /\.$/) {
           $origin = $mdetail;
@@ -231 +219 @@
           $origin = "$mdetail.$origin";
         }
-
-#      if ($mdetail eq '.' || $mdetail =~ /$zname\.$/ || $zname =~ /$mdetail\.$/) {
-#        $origin = $mdetail;
-#      } else {
-#        # if we continue, we either use an $ORIGIN that's out of zone, or ignore it and potentially publish incorrect records.
-#        die "bad \$ORIGIN: $_\n";
-#      }
-
       }
       elsif ($macro eq 'GENERATE') {
 # needs to generate CIDR range(s) as needed to match the start/stop points
       }
-    # not handling $INCLUDE or $GENERATE (altho the latter seems to be mostly a less-flexible version of the template types)
+##fixme: should arguably handle $INCLUDE
       next;
     }
@@ -257 +237 @@
     }
 
-#print "  found '$curlabel'\n";
+    # yay for special cases
+    $origin = '' if $origin eq '.';
+
+    # leading whitespace indicates "same label as last record"
+    if ($rec =~ /^\s/) {
+      $curlabel = $prevlabel;
+    } else {
+      ($curlabel) = ($rec =~ /^([\w\@_.-]+)\s/);
+    }
 
     # magic name!
@@ -264 +252 @@
     # append $ORIGIN if name is not fully qualified.
     if ($curlabel !~ /\.$/) {
-      $curlabel .= ($origin eq '.' ? '.' : ".$origin");
-    }
-#print "  expanded '$curlabel'\n";
-
-# hack pthbptt
-#$curlabel =~ s/\.\.$/./;
+      $curlabel .= ".$origin";
+    }
+
     # check for zone scope.  skip bad records.
     if ($curlabel !~ /$zname.$/) {
@@ -281 +266 @@
     $rec =~ s/^([\w\@_.-]*)\s+//;
 
-#  # records must begin in the first column, no leading whitespace
-#  my ($name) = /^([\w\@_.-]+)\s/;
-
-# foo IN A 1.2.3.4
-#   IN A 2.3.4.5
-# =
-# foo.zone. IN A 1.2.3.4
-# foo.zone. IN A 2.3.4.5
-
-#  # "empty" label records inherit the previous label
-#  # RRs start in the first column by definition, so leading whitespace indicates an inherited label
-#  if (/^\s+/) {
-#    # fatal error.  if there is no previous label, we can by definition not set
-#    # the current label based on it.  this can only happen on the very first
-#    # record, following records will *ALWAYS* have a previous label
-#    die "bad first record ($_):  no previous label\n" if !$prevlabel;
-#    $name = $prevlabel;
-#  }
-
-#print "$i ($rec)\n";#\t$curlabel";
-
-
-
-
-#  # append zone name to record name if missing AND not dot-terminated;
-#  # this happens automagically for forward zones, but not reverse because Reasons.  (fixme?)
-#  # suck up and deal with the error if the dot-termiated name is out of zone;  should be
-#  # impossible with valid BIND zone file but...
-#  if ($name !~ /\.$/) {
-#    $name .= ".$zname" if $name !~ /$zname$/;
-#  } else {
-#    warn "skipping out-of-zone record:\n\t($_)\n" if $name !~ /$zname\.$/;
-#    next;
-#  }
-
-
+    # now that we've collected and trimmed off the record's label, unpack the class, TTL, and type.
+    # class and TTL may be omitted, and may appear in either class,TTL or TTL,class order.
     my $nc = 0;
+    # we don't actually use these but we have to recognize them
     my $class = 'IN';
+    # not preset as we need to detect whether it's present in the record
+    my $ttl;
     my $type;
-    my $ttl;
-    my $distance;
-    my $weight;
-    my $port;
     my $badrec;
     my $curatom = 'class';
-
-    # unpack the class, TTL, and type
     eval {
       for (; $nc < 3; $nc++) {
+        last if $type;  # short-circuit if we've got a type, further data is record-specific.
         my ($atom) = ($rec =~ /^([\w\d.]+)\s/);
         # should be safe?
         last if !$atom;
-        last if $type;
-#print "nc:$nc: $atom\n";
         if ($atom =~ /^\d+$/) {
           if (defined($ttl)) {
+            # we already have a TTL, so another all-numeric field is invalid.
             die "bad record ($origrec)\n";
-#            warn "bad record ($origrec)\n";
-#            $badrec = 1;
-#            last;
           } else {
             if ($curatom ne 'class' && $curatom ne 'ttl') {
               die "bad record ($origrec)\n";
-#              warn "bad record ($origrec)\n";
-#              $badrec = 1;
-#              last;
             }
             $curatom = 'ttl';
@@ -352 +294 @@
           }
         }
-
         elsif ($atom =~ /^IN|CS|CH|HS$/) {
-#print "a$nc: d2: atom [$atom]\n        $rec\n" if $i == $debugid;
           if ($atom =~ /CS|CH|HS/) {
             die "unsupported class $atom in record ($origrec)\n";
-#            warn "unsupported class $atom in record ($origrec)\n";
-#            $badrec = 1;
-#            last;
           }
           $curatom = 'class';
           $class = $atom;
         }
-
-        elsif ($atom =~ /^[A-Z]+/) {
-#      print "dbg: type $atom\n";
+        elsif ($atom =~ /^[A-Z\d-]+/) {
+          # check against dnsadmin's internal list of known DNS types.
           if ($reverse_typemap{$atom}) {
             $type = $atom;
@@ -372 +308 @@
             die "unknown type $atom in record ($origrec)\n";
           }
+          $curatom = 'type';
         }
         $rec =~ s/^$atom\s*//;
       }
-    }; # record class/type/TTL parse
+    }; # record class/type/TTL loop
     if ($@) {
       warn $@;
@@ -387 +324 @@
 #next if $badrec;
 
+
+    # Just In Case we need the original rdata after we've sliced off more pieces
+    my $rdata = $rec;
     $prevlabel = $curlabel;
 
-
-## by convention the optional TTL leads the optional class, but they're apparently swappable.
-#  my ($ttl) = /^(\d+)?\s/;
-#  if (defined $ttl) {
-#    # TTL may be zero
-#    s/(\d+)?\s+//;
-#  } else {
-#    # Fall back to zone default TTL
-#    $ttl = $zonettl;
-#  }
-#  my ($class) = /^(IN|CS|CH|HS|\d+)\s/;
-#  if (defined $class) {
-#    if ($class =~ /\d+/) {
-#      
-#    }
-#    if ($class ne 'IN') {
-#      warn "Non-Internet class ($class) records not supported:\n\t$origrec\n";
-#      next;
-#    }
-#    s/(IN|CS|CH|HS)\s+//;
-#  } else {
-#    $class = 'IN';
-#  }
-#  my ($type) = /([A-Z-]+)\s/;
-#  if (!$reverse_typemap{$type}) {
-#    warn "Unknown type $type, skipping\n\t($rec)\n";
-#    next;
-#  }
-#  s/([A-Z-]+)\s+//;
-#  chomp;
-
+    # part of the record data, when present
+    my $distance;
+    my $weight;
+    my $port;
 
     my $itype = $reverse_typemap{$type};
-    my $rdata = $rec;
-
-  # SOA is the only type that may span multiple lines.  Probably.  Note even AXFRed zones write multiline SOA records:
-  #@       IN      SOA     test.example.invalid.   test.example.invalid.   (2020082500 7200 900 604800 3600)
-  #        IN      NS      olddns.example.com.
-  #        IN      MX      1 fred.foo.bar.invalid.
-  #foo     IN      A       192.168.16.45
-  # AXFR'ed zone file gets written as
-  #$ORIGIN .
-  #$TTL 3600       ; 1 hour
-  #example.invalid         IN SOA  test.example.invalid. test.example.invalid. (
-  #                                2020082500 ; serial
-  #                                7200       ; refresh (2 hours)
-  #                                900        ; retry (15 minutes)
-  #                                604800     ; expire (1 week)
-  #                                3600       ; minimum (1 hour)
-  #                                )
-  #                        NS      olddns.example.com.
-  #                        MX      1 fred.foo.bar.invalid.
-  #$ORIGIN example.invalid.
-  #foo                     A       192.168.16.45
-  $foundtypes{$type}++;
+
+
+# See RFC1035 and successors for the canonical zone file format reference.  We'll
+# ignore a number of edge cases because they're quite horrible to parse.
+# Of particular note is use of () to continue entries across multiple lines.  Use
+# outside of SOA records is quite rare, although some compliant zone file
+# *writers* may use it on TXT records.
+# We'll also ignore the strict interpretation in SOA records in favour of spotting
+# the more standard pattern where the SOA serial, refresh, retry, expire, and minttl
+# numbers are in ():
+
+#example.invalid         IN SOA  test.example.invalid. test.example.invalid. (
+#                                2020082500 ; serial
+#                                7200       ; refresh (2 hours)
+#                                900        ; retry (15 minutes)
+#                                604800     ; expire (1 week)
+#                                3600       ; minimum (1 hour)
+#                                )
+
+    $foundtypes{$type}++;
 
 ##fixme:  strip trailing . here?  dnsadmin's normalized internal format omits it, some validation fails or may go funky
@@ -451 +362 @@
       die "Can't parse gibberish SOAish record: $rec\n" if !$ns;
       $rdata =~ s/([\w.]+)\s+([\w.]+)\s+\(\s*//;
-
-      # There are probably more efficient ways to do this but the SOA record
-      # format is essentially character based, not line-based.
-      # In theory the SOA serial etc may be spread over up to 5 lines, in any combination.
 
       # Parse fields from $rdata if present
@@ -488 +395 @@
         $dnsdb->{dbh}->do("INSERT INTO records (domain_id,host,type,val,ttl,location) VALUES (?,?,?,?,?,?)", undef, @soavals);
       }
-#  $dnsdb->{dbh}->do("INSERT INTO records () VALUES ()");
-#  next;
-#Zfqdn:mname:rname:ser:ref:ret:exp:min:ttl:timestamp:lo
-#print "Z$zname:$ns:$adminmail:$soabits[0]:$soabits[1]:$soabits[2]:$soabits[3]:$soabits[4]:$ttl\n";
       # skip insert at end of loop;  SOA records are not handled by DNSDB::addRec()
       next;