BIND 9.7.2-P3 is a maintenance release for BIND 9.7.
		
    
			This document summarizes changes from BIND 9.7.1 to BIND 9.7.2-P3.
			Please see the CHANGES file in the source code release for a
			complete list of all changes.
		
   
  
    
    
			The latest release of BIND 9 software can always be found
	 		on our web site at
      http://www.isc.org/software/bind.
  		There you will find additional information about each release,
 			source code, and some pre-compiled versions for certain operating
 			systems.
		
   
  
  
    
		
			
	    - 
					Zones may be dynamically added and removed with the
					“rndc addzone” and “rndc delzone” commands.  These
					dynamically added zones are written to a per-view
					configuration file.  Do not rely on the configuration
					file name nor contents as this will change in a future
					release.  This is an experimental feature at this time.
				
 - 
					Added new “filter-aaaa-on-v4” access control list to
					select which IPv4 clients have AAAA record filtering
					applied.
				
 - 
					A new command “rndc secroots” was added to dump a combined
					summary of the currently managed keys combined with statically
					configured trust anchors.
				
 - 
					Added support to load new keys into managed zones without
					signing immediately with "rndc loadkeys".  Added support
					to link keys with "dnssec-keygen -S" and
					"dnssec-settime -S".
				
 
 
		 
		
		
		
   
  
    
		
			
	    - 
					Documentation improvements
				
 - 
					ORCHID prefixes were removed from the automatic empty
					zone list.
				
 - 
					Improved handling of GSSAPI security contexts.  Specifically,
					better memory management of cached contexts, limited lifetime
					of a context to 1 hour, and added a “realm” command to
					nsupdate to allow selection of a non-default realm name.
				
 - 
					The contributed tool “zkt” was updated to version 1.0.
				
 
 
		 
		
		
		
   
  
    
		
		
			
	    - 
					If BIND, acting as a DNSSEC validating server, has two or more trust
	        anchors configured in named.conf for the same zone (such as
	        example.com) and the response for a record in that zone from the
	        authoritative server includes a bad signature, the validating server
	        will crash while trying to validate that query.
				
 
 
		 
		
			
			- 
					A flaw where the wrong ACL was applied was fixed.  This flaw
					allowed access to a cache via recursion even though the ACL
					disallowed it.
				
 
 
		 
		
			
			- 
				 	Adding a NO DATA signed negative response to cache failed to clear
				  any matching RRSIG records already in cache. A subsequent lookup
				  of the cached NO DATA entry could crash named (INSIST) when the
				  unexpected RRSIG was also returned with the NO DATA cache entry.
				  [RT #22288] [CVE-2010-3613] [VU#706148]
				
 - 
					BIND, acting as a DNSSEC validator, was determining if the NS RRset
				  is insecure based on a value that could mean either that the RRset
				  is actually insecure or that there wasn't a matching key for the RRSIG
				  in the DNSKEY RRset when resuming from validating the DNSKEY RRset.
				  This can happen when in the middle of a DNSKEY algorithm rollover,
				  when two different algorithms were used to sign a zone but only the
				  new set of keys are in the zone DNSKEY RRset.
					[RT #22309] [CVE-2010-3614] [VU#837744]
				
 - 
					
				 		When BIND is running as an authoritative server for a zone and
				   	receives a query for that zone data, it first checks for allow-query
				   	acls in the zone statement, then in that view, then in global
				   	options. If none of these exist, it defaults to allowing any query
				   	(allow-query {"any"};).
					
					
			   		With this bug, if the allow-query is not set in the zone statement,
				   	it failed to check in view or global options and fell back to the
				   	default of allowing any query. This means that queries that the zone
				   	owner did not wish to allow were incorrectly allowed.
						[RT #22418] [CVE-2010-3615] [VU#510208]
					
				 
 
		 
   
  
    
		
			
	    - 
					Removed a warning message when running BIND 9 under Windows
					for when a TCP connection was aborted.  This is a common
					occurrence and the warning was extraneous.
				
 - 
					Worked around a race condition in the cache database memory
					handling.  Without this fix a DNS cache DB or ADB could
					incorrectly stay in an over memory state, effectively refusing
					further caching, which subsequently made a BIND 9 caching
					server unworkable.
				
 - 
					Partially disabled change 2864 because it would cause
					infinite attempts of RRSIG queries.
				
 - 
					BIND did not properly handle non-cacheable negative responses
					from insecure zones. This caused several non-protocol-compliant
					zones to become unresolvable.  BIND is now more accepting of
					responses it receives from less strict servers.
				
 
 
		 
		
			
			- 
					A bug, introduced in BIND 9.7.2, caused named to fail to start
					if a master zone file was unreadable or missing.  This has
					been corrected in 9.7.2-P1.
				
 - 
	        BIND previously accepted answers from authoritative servers that did
	        not provide a "proper" response, such as not setting AA bit.  BIND was
	        changed to be more strict in what it accepted but this caused
	        operational issues.  This new strictness has been backed out in
	        9.7.2-P1.
	      
 
 
		 
		
		
			
			- 
					Microsoft changed the behavior of sockets between NT/XP based
				  stacks vs Vista/windows7 stacks. Server 2003/2008 have the older
				  behavior, 2008r2 has the new behavior. With the change, different
				  error results are possible, so ISC adapted BIND to handle the new
				  error results.
				  This resolves an issue where sockets would shut down on
				  Windows servers causing named to stop responding to queries.
					[RT #21906]
				
 - 
				 	Windows has non-POSIX compliant behavior in its rename() and unlink()
				  calls. This caused journal compaction to fail on Windows BIND servers
				  with the log error: "dns_journal_compact failed: failure".
					[RT #22434]
				
 
 
		 
   
  Known issues in this release
 
    
    - 
        
          "make test" will fail on OSX and possibly other operating systems.
          The failure occurs in a new test to check for allow-query ACLs.
          The failure is caused because the source address is not specified on
          the dig commands issued in the test.
        
        
          If running "make test" is part of your usual acceptance process,
          please edit the file bin/tests/system/allow_query/test.sh
          and add
          
            -b 10.53.0.2
          
          to the DIGOPTS line.
        
       
 
   
  
    
    
      Thank you to everyone who assisted us in making this release possible.
      If you would like to contribute to ISC to assist us in continuing to make
      quality open source software, please visit our donations page at
      http://www.isc.org/supportisc.