A Common Operational Problem in DNS Servers

The DNS , is a query / response protocol. Failure to respond to queries or to respond incorrectly causes both immediate operational problems and long term problems with protocol development. Failure to respond to a query is indistinguishable from a packet loss without doing a analysis of query response patterns and results in unnecessary additional queries being made by DNS clients and unnecessary delays being introduced to the resolution process. Due to the inability to distinguish between packet loss and nameservers dropping EDNS queries, packet loss is sometimes misclassified as lack of EDNS support which can lead to DNSSEC validation failures. Allowing servers which fail to respond to queries to remain results in developers being afraid to deploy implementations of recent standards. Such servers need to be identified and corrected / replaced. The DNS has response codes that cover almost any conceivable query response. A nameserver should be able to respond to any conceivable query using them. Unless a nameserver is under attack, it should respond to all queries directed to it as a result of following delegations. Additionally code should not assume that there isn't a delegation to the server even if it is not configured to serve the zone. Broken delegations are a common occurrence in the DNS and receiving queries for zones that the server is not configured for is not necessarily an indication that the server is under attack. Parent zone operators are supposed to regularly check that the delegating NS records are consistent with those of the delegated zone and to correct them when they are not . If this was being done regularly, the instances of broken delegations would be much lower. When a nameserver is under attack it may wish to drop packets. A common attack is to use a nameserver as a amplifier by sending spoofed packets. This is done because response packets are bigger than the queries and big amplification factors are available especially if EDNS is supported. Limiting the rate of responses is reasonable when this is occurring and the client should retry. This however only works if legitimate clients are not being forced to guess whether EDNS queries are accepted or not. While there is still a pool of servers that don't respond to EDNS requests, clients have no way to know if the lack of response is due to packet loss, EDNS packets not being supported or rate limiting due to the server being under attack. Mis-classifications of server characteristics are unavoidable when rate limiting is done.

There are number common query kinds that result in non responses today. These are EDNS queries with and without extensions, queries for unknown (unallocated) or unsupported types, and filtering of TCP queries.

Identifying servers that fail to respond to unknown or unsupported types can be done by making an initial DNS query for an A record, making a number of queries for an unallocated type, then making a query for an A record again. IANA maintains a registry of allocated types. If the server responds to the first and last queries but fails to respond to the queries for the unallocated type, it is probably faulty. The test should be repeated a number of times to eliminate the likelihood of a false positive due to packet loss.

Some servers fail to respond to DNS queries with various DNS flags set, regardless of whether they are defined or still reserved. At the time of writing there are servers that fail to respond to queries with the AD bit set to 1 and servers that fail to respond to queries with the last reserved flag bit set.

The use of previously undefined opcodes is to be expected. Since the DNS was first defined two new opcodes have been added, UPDATE and NOTIFY. NOTIMP is the expected rcode to an unknown / unimplemented opcode. Note: while new opcodes will most probably use the current layout structure for the rest of the message there is no requirement than anything other than the DNS header match.

All DNS servers are supposed to respond to queries over TCP . Firewalls that drop TCP connection attempts rather that resetting the connect attempt or send a ICMP/ICMPv6 administratively prohibited message introduce excessive delays to the resolution process. Whether a server accepts TCP connections can be tested by first checking that it responds to UDP queries to confirm that it is up and operating, then attempting the same query over TCP. An additional query should be made over UDP if the TCP connection attempt fails to confirm that the server under test is still operating.

Identifying servers that fail to respond to EDNS queries can be done by first identifying that the server responds to regular DNS queries, followed by a series of otherwise identical queries using EDNS, then making the original query again. A series of EDNS queries is needed as at least one DNS implementation responds to the first EDNS query with FORMERR but fails to respond to subsequent queries from the same address for a period until a regular DNS query is made. The EDNS query should specify a UDP buffer size of 512 bytes to avoid false classification of not supporting EDNS due to response packet size. If the server responds to the first and last queries but fails to respond to most or all of the EDNS queries, it is probably faulty. The test should be repeated a number of times to eliminate the likelihood of a false positive due to packet loss. Firewalls may also block larger EDNS responses but there is no easy way to check authoritative servers to see if the firewall is misconfigured.

Some servers respond correctly to EDNS version 0 queries but fail to respond to EDNS queries with version numbers that are higher than zero. Servers should respond with BADVERS to EDNS queries with version numbers that they do not support. Some servers respond correctly to EDNS version 0 queries but fail to set QR=1 when responding to EDNS versions they do not support. Such answers are discarded or treated as requests.

Some servers fail to respond to EDNS queries with EDNS options set. Unknown EDNS options are supposed to be ignored by the server .

Some servers fail to respond to EDNS queries with EDNS flags set. Server should ignore EDNS flags they do not understand and should not add them to the response .

While the first step in remediating this problem is to get the offending nameserver code corrected, there is a very long tail problem with DNS servers in that it can often take over a decade between the code being corrected and a nameserver being upgraded with corrected code. With that in mind it is requested that TLD, and other similar zone operators, take steps to identify and inform their customers, directly or indirectly through registrars, that they are running such servers and that the customers need to correct the problem. TLD operators are being asked to do this as they, due to the nature of running a TLD and the hierarchical nature of the DNS, have access to a large numbers of nameserver names as well as contact details for the registrants of those nameservers. While it is possible to construct lists of nameservers from other sources, and that has been done to survey the state of the Internet, that doesn't give the tester the contact details necessary to inform the operators. The SOA RNAME is often invalid and whois data is obscured and / or not available which makes it infeasible for others to do this. While this section talks about TLD operators performing this work, it may be done by registrars on behalf of the TLD operator. The intent is to ensure that the testing happens and that operators of non-compliant nameservers be informed, rather than to prescribe who does the actual testing and communication. Note: having registrars perform this testing and reporting is likely to result in duplicate reports for the same server being issued by multiple registrars. TLD operators should construct a list of servers child zones are delegated to along with a delegated zone name. This name shall be the query name used to test the server as it is supposed to exist. For each server the TLD operator shall make an SOA query of the delegated zone name. This should result in the SOA record being returned in the answer section. If the SOA record is not returned but some other response is returned, this is a indication of a bad delegation and the TLD operator should take whatever steps it normally takes to rectify a bad delegation. If more that one zone is delegated to the server, it should choose another zone until it finds a zone which responds correctly or it exhausts the list of zones delegated to the server. If the server fails to get a response to a SOA query, the TLD operator should make an A query as some nameservers fail to respond to SOA queries but respond to A queries. If it gets no response to the A query, another delegated zone should be queried for as some nameservers fail to respond to zones they are not configured for. If subsequent queries find a responding zone, all delegation to this server need to be checked and rectified using the TLD's normal procedures. Having identified a working <server, query name> tuple the TLD operator should now check that the server responds to EDNS, Unknown Query Type and TCP tests as described above. If the TLD operator finds that server fails any of the tests, the TLD operator shall take steps to inform the operator of the server that they are running a faulty nameserver and that they need to take steps to correct the matter. The TLD operator shall also record the <server, query name> for follow-up testing. If repeated attempts to inform and get the customer to correct / replace the faulty server are unsuccessful the TLD operator shall remove all delegations to said server from the zone. Removal of delegations is the step of last resort in handling complaints as specified in COMPLAINTS. It will also be necessary for TLD operators to repeat the scans periodically. It is recommended that this be performed monthly backing off to bi-annually once the numbers of faulty servers found drops off to less than 1 in 100000 servers tested. Follow-up tests for faulty servers still need to be performed monthly. Some operators claim that they can't perform checks at registration time. If a check is not performed at registration time, it needs to be performed within a week of registration in order to detect faulty servers swiftly. Checking of delegations by TLD operators should be nothing new as they have been required from the very beginnings of DNS to do this . Checking for compliance of nameserver operations should just be a extension of such testing. It is recommended that TLD operators setup a test web page which performs the tests the TLD operator performs as part of their regular audits to allow nameserver operators to test that they have correctly fixed their servers. Such tests should be rate limited to avoid these pages being a denial of service vector. Nothing in this section precludes others testing servers for protocol compliance. DNS operators should test their servers to ensure that their vendors have shipped protocol compliant products. Nameserver vendors can use these tests as a part of this release processes. Registrants can use these tests to check their DNS operators servers.

Firewalls and load balancers can affect the externally visible behaviour of a nameserver. Tests for conformance need to be done from outside of any firewall so that the system as a whole is tested. Firewalls and load balancers should not drop DNS packets that they don't understand. They should either pass through the packets or generate an appropriate error response. Requests for unknown query types is normal client behaviour and should not be construed as an attack. Nameservers have always been expected to be able to handle such queries. Requests for unknown query classes is normal client behaviour and should not be construed as an attack. Nameservers have always been expected to be able to handle such queries. Requests with unknown opcodes is normal client behaviour and should not be construed as an attack. Nameservers have always been expected to be able to handle such queries. Requests with unassigned flags set (DNS or EDNS) is expected client behaviour and should not be construed as an attack. The behaviour for unassigned is to ignore them in the request and to not set them in the response. All dropping DNS / EDNS packets with unassigned flags does is make it harder to deploy extensions that make use of them due to the need to reconfigure / update firewalls. Requests with unknown EDNS options is expected client behaviour and should not be construed as an attack. The correct behaviour for unknown EDNS options is to ignore there presence when constructing a reply. Requests with unknown EDNS versions is expected client behaviour and should not be construed as an attack. The correct behaviour for unknown EDNS versions is to return BADVERS along with the highest EDNS version the server supports. All dropping EDNS packets does is break EDNS version negotiation. Firewalls should not assume that there will only be a single response message to a requests. There have been proposals to use EDNS to signal that multiple DNS messages be returned rather than a single UDP message that is fragmented at the IP layer. With the above said, there will be times when a nameserver mishandles messages with a particular flag, EDNS option, EDNS version field, opcode, type or class field or combination there of to the point where the integrity of the nameserver is compromised. Firewalls should have the ability to selectively reject messages with an appropriately constructed response based on all these fields while awaiting a fix from the nameserver vendor. DNS and EDNS in particular is designed to allow clients to be able to use new features against older servers without having to ask a thousand questions to determine what the server supports which takes time clients do not have. Indiscriminate blocking of messages breaks that design.

Scrubbing services, like firewalls, can affect the externally visible behaviour of a nameserver. If a operator uses a scrubbing service, they should check that legitimate queries are not being blocked. Scrubbing services, unlike firewalls, are also turned on and off in response to denial of service attacks. One needs to take care when choosing a scrubbing service and ask questions like: Do they pass unknown DNS query types? Do they pass unknown EDNS versions? Do they pass unknown EDNS options? Do they pass unknown EDNS flags? Do they pass requests with unknown DNS opcodes? Do they pass requests with the remaining reserved DNS header flag bit set? None of these are attack vectors but some scrubbing services treat them as such.

Whole answer caches take a previously constructed answer and return it to a subsequent query for the same qname, qtype and qclass, just updating the query id field and possibly the qname to match the incoming query to avoid constructing each response individually. Whole answer caches can return the wrong response to a query if they do not take all of the attributes of the query into account, rather than just some of them e.g. qname, qtype and qclass. This has implications when testing and with overall protocol compliance. e.g. There are whole answer caches that ignore the EDNS version field which results in incorrect answers to non EDNS version 0 queries being returned if they were preceded by a EDNS version 0 query for the same name and type. e.g. There are caches that ignore the EDNS options in the query resulting in options only working some of the time and/or options being returned when not requested.

Choosing the correct response code when fixing a nameserver is important. Just because a type is not implemented does not mean that NOTIMP is the correct response code to return. Response codes need to be chosen considering how clients will handle them. For unimplemented opcodes NOTIMP is the expected response code. Additionally a new opcode could change the message format by extending the header or changing the structure of the records etc. This may result in FORMERR being returned though NOTIMP would be more correct. In general, for unimplemented type codes Name Error (NXDOMAIN) and NOERROR (no data) are the expected response codes. A server is not supposed to serve a zone which contains unsupported types () so the only thing left is return if the QNAME exists or not. NOTIMP and REFUSED are not useful responses as they force the clients to try all the authoritative servers for a zone looking for a server which will answer the query. Meta queries type may be the exception but these need to be thought about on a case by case basis. If the server supports EDNS and get a query with an unsupported EDNS version, the correct response is BADVERS . If the server do not support EDNS at all, FORMERR and NOTIMP are the expected error codes. That said a minimal EDNS server implementation just requires parsing the OPT records and responding with an empty OPT record. There is no need to interpret any EDNS options present in the request as unsupported options are expected to be ignored .

Testing is divided into two sections. Basic DNS which all servers should meet and Extended DNS which should be met by all servers that support EDNS (a server is deemed to support EDNS if it gives a valid EDNS response to any EDNS query). If a server does not support EDNS it should still respond to all the tests. It is advisable to run all of the tests below in parallel so as to minimise the delays due to multiple timeouts when the servers do not respond. There are 16 queries directed to each nameserver assuming no packet loss testing different aspects of Basic DNS and EDNS. The tests below use dig from BIND 9.11.0 which is still in development.

This first set of tests cover basic DNS server behaviour and all servers should pass these tests.

Verify the server is configured for the zone: dig +noedns +noad +norec soa $zone @$server expect: status: NOERROR expect: SOA record expect: flag: aa to be present

Check that queries for an unknown type work: dig +noedns +noad +norec type1000 $zone @$server expect: status: NOERROR expect: an empty answer section. expect: flag: aa to be present That new types are to be expected is specified in Section 3.6, . Servers that don't support a new type are expected to reject a zone that contains a unsupported type as per Section 5.2, . This means that a server that does load a zone can answer questions for unknown types with NOERROR or NXDOMAIN as per Section 4.3.2, . later reserved distinct ranges for meta and data types which allows servers to be definitive about whether a query should be answerable from zone content or not.

Check that queries with CD=1 work: dig +noedns +noad +norec +cd soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: flag: aa to be present CD use in queries is defined in .

Check that queries with AD=1 work: dig +noedns +norec +ad soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: flag: aa to be present AD use in queries is defined in .

Check that queries with the last unassigned DNS header flag work and that the flag bit is not copied to the response: dig +noedns +noad +norec +zflag soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: MBZ to not be in the response expect: flag: aa to be present MBZ (Must Be Zero) presence indicates the flag bit has been incorrectly copied. See Section 4.1.1, "Z Reserved for future use. Must be zero in all queries and responses."

Check that new opcodes are handled: dig +noedns +noad +opcode=15 +norec +header-only @$server expect: status: NOTIMP expect: SOA record to not be present expect: flag: aa to NOT be present As unknown opcodes have no definition, including packet format other than there must be a DNS header present, there is only one possible rcode that make sense to return to a request with a unknown opcode and that is NOTIMP.

Check that TCP queries work: dig +noedns +noad +norec +tcp soa $zone @$server expect: status: NOERROR expect: SOA record expect: flag: aa to be present The requirement that TCP be supported is defined in .

The next set of test cover various aspects of EDNS behaviour. If any of these tests succeed, then all of them should succeed. There are servers that support EDNS but fail to handle plain EDNS queries correctly so a plain EDNS query is not a good indicator of lack of EDNS support.

Check that plain EDNS queries work: dig +nocookie +edns=0 +noad +norec soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: OPT record to be present expect: EDNS Version 0 in response expect: flag: aa to be present +nocookie disables sending a EDNS COOKIE option in which is on by default.

Check that EDNS version 1 queries work (EDNS supported): dig +nocookie +edns=1 +noednsneg +noad +norec soa $zone @$server expect: status: BADVERS expect: SOA record to not be present expect: OPT record to be present expect: EDNS Version 0 in response expect: flag: aa to NOT be present Only EDNS Version 0 is currently defined so the response should always be a 0 version. This will change when EDNS version 1 is defined. BADVERS is the expected rcode if EDNS is supported as per Section 6.1.3, .

Check that EDNS queries with an unknown option work (EDNS supported): dig +nocookie +edns=0 +noad +norec +ednsopt=100 soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: OPT record to be present expect: OPT=100 to not be present expect: EDNS Version 0 in response expect: flag: aa to be present Unknown EDNS options are supposed to be ignored, Section 6.1.2, .

Check that EDNS queries with unknown flags work (EDNS supported): dig +nocookie +edns=0 +noad +norec +ednsflags=0x40 soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: OPT record to be present expect: MBZ not to be present expect: EDNS Version 0 in response expect: flag: aa to be present MBZ (Must Be Zero) presence indicates the flag bit has been incorrectly copied as per Section 6.1.4, .

Check that EDNS version 1 queries with unknown flags work (EDNS supported): dig +nocookie +edns=1 +noednsneg +noad +norec +ednsflags=0x40 soa \ $zone @$server expect: status: BADVERS expect: SOA record to NOT be present expect: OPT record to be present expect: MBZ not to be present expect: EDNS Version 0 in response expect: flag: aa to NOT be present +noednsneg disables EDNS version negotiation in DiG; MBZ (Must Be Zero) presence indicates the flag bit has been incorrectly copied.

Check that EDNS version 1 queries with unknown options work (EDNS supported): dig +nocookie +edns=1 +noednsneg +noad +norec +ednsopt=100 soa \ $zone @$server expect: status: BADVERS expect: SOA record to NOT be present expect: OPT record to be present expect: OPT=100 to NOT be present expect: EDNS Version 0 in response expect: flag: aa to be present +noednsneg disables EDNS version negotiation in DiG.

Check that a DNSSEC queries work (EDNS supported): dig +nocookie +edns=0 +noad +norec +dnssec soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: OPT record to be present expect: DO=1 to be present if a RRSIG is in the response expect: EDNS Version 0 in response expect: flag: aa to be present DO=1 should be present if RRSIGs are returned as they indicate that the server supports DNSSEC. Servers that support DNSSEC are supposed to copy the DO bit from the request to the response as per .

Check that EDNS version 1 DNSSEC queries work (EDNS supported): dig +nocookie +edns=1 +noednsneg +noad +norec +dnssec soa \ $zone @$server expect: status: BADVERS expect: SOA record to not be present expect: OPT record to be present expect: DO=1 to be present if the EDNS version 0 DNSSEC query test returned DO=1 expect: EDNS Version 0 in response expect: flag: aa to NOT be present +noednsneg disables EDNS version negotiation in DiG.

Check that EDNS queries with multiple defined EDNS options work: dig +edns=0 +noad +norec +cookie +nsid +expire +subnet=0.0.0.0/0 \ soa $zone @$server expect: status: NOERROR expect: SOA record to be present expect: OPT record to be present expect: EDNS Version 0 in response expect: flag: aa to be present

If EDNS is not supported by the nameserver, we expect a response to all the above queries. That response may be a FORMERR or NOTIMP error response or the OPT record may just be ignored. Some nameservers only return a EDNS response when a particular EDNS option or flag (e.g. DO=1) is present in the request. This behaviour is not compliant behaviour and may hide other incorrect behaviour from the above tests. Re-testing with the triggering option / flag present will expose this misbehaviour.

Testing protocol compliance can potentially result in false reports of attempts to break services from Intrusion Detection Services and firewalls. None of the tests listed above should break nominally EDNS compliant servers. None of the tests above should break non EDNS servers. All the tests above are well formed, though not necessarily common, DNS queries. Relaxing firewall settings to ensure EDNS compliance could potentially expose a critical implementation flaw in the nameserver. Nameservers should be tested for conformance before relaxing firewall settings. When removing delegations for non-compliant servers there can be a knock on effect on other zones that require these zones to be operational for the nameservers addresses to be resolved.

There are no actions for IANA.