idnits 2.17.1 draft-andrews-dns-no-response-issue-11.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (August 19, 2015) is 3166 days in the past. Is this intentional? Checking references for intended status: Informational ---------------------------------------------------------------------------- ** Obsolete normative reference: RFC 5966 (Obsoleted by RFC 7766) Summary: 1 error (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group M. Andrews 3 Internet-Draft ISC 4 Intended status: Informational August 19, 2015 5 Expires: February 20, 2016 7 A Common Operational Problem in DNS Servers - Failure To Respond. 8 draft-andrews-dns-no-response-issue-11.txt 10 Abstract 12 The DNS is a query / response protocol. Failure to respond to 13 queries causes both immediate operational problems and long term 14 problems with protocol development. 16 This document identifies a number of common classes of queries that 17 some servers fail to respond too. This document also suggests 18 procedures for TLD and other similar zone operators to apply to help 19 reduce / eliminate the problem. 21 Status of this Memo 23 This Internet-Draft is submitted in full conformance with the 24 provisions of BCP 78 and BCP 79. 26 Internet-Drafts are working documents of the Internet Engineering 27 Task Force (IETF). Note that other groups may also distribute 28 working documents as Internet-Drafts. The list of current Internet- 29 Drafts is at http://datatracker.ietf.org/drafts/current/. 31 Internet-Drafts are draft documents valid for a maximum of six months 32 and may be updated, replaced, or obsoleted by other documents at any 33 time. It is inappropriate to use Internet-Drafts as reference 34 material or to cite them other than as "work in progress." 36 This Internet-Draft will expire on February 20, 2016. 38 Copyright Notice 40 Copyright (c) 2015 IETF Trust and the persons identified as the 41 document authors. All rights reserved. 43 This document is subject to BCP 78 and the IETF Trust's Legal 44 Provisions Relating to IETF Documents 45 (http://trustee.ietf.org/license-info) in effect on the date of 46 publication of this document. Please review these documents 47 carefully, as they describe your rights and restrictions with respect 48 to this document. Code Components extracted from this document must 49 include Simplified BSD License text as described in Section 4.e of 50 the Trust Legal Provisions and are provided without warranty as 51 described in the Simplified BSD License. 53 Table of Contents 55 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 56 2. Common queries class that result in non responses. . . . . . . 4 57 2.1. EDNS Queries - Version Independent . . . . . . . . . . . . 4 58 2.2. EDNS Queries - Version Specific . . . . . . . . . . . . . 4 59 2.3. EDNS Options . . . . . . . . . . . . . . . . . . . . . . . 4 60 2.4. EDNS Flags . . . . . . . . . . . . . . . . . . . . . . . . 5 61 2.5. DNS Flags . . . . . . . . . . . . . . . . . . . . . . . . 5 62 2.6. Unknown / Unsupported Type Queries . . . . . . . . . . . . 5 63 2.7. Unknown DNS opcodes . . . . . . . . . . . . . . . . . . . 5 64 2.8. TCP Queries . . . . . . . . . . . . . . . . . . . . . . . 5 65 3. Remediating . . . . . . . . . . . . . . . . . . . . . . . . . 6 66 4. Firewalls and Load Balancers . . . . . . . . . . . . . . . . . 7 67 5. Scrubbing Services . . . . . . . . . . . . . . . . . . . . . . 8 68 6. Response Code Selection . . . . . . . . . . . . . . . . . . . 8 69 7. Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 70 8. Security Considerations . . . . . . . . . . . . . . . . . . . 13 71 9. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 13 72 10. Normative References . . . . . . . . . . . . . . . . . . . . . 13 73 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 14 75 1. Introduction 77 The DNS [RFC1034], [RFC1035] is a query / response protocol. Failure 78 to respond to queries causes both immediate operational problems and 79 long term problems with protocol development. 81 Failure to respond to a query is indistinguishable from a packet loss 82 without doing a analysis of query response patterns and results in 83 unnecessary additional queries being made by DNS clients and 84 unnecessary delays being introduced to the resolution process. 86 Due to the inability to distinguish between packet loss and 87 nameservers dropping EDNS [RFC6891] queries, packet loss is sometimes 88 misclassified as lack of EDNS support which can lead to DNSSEC 89 validation failures. 91 Allowing servers which fail to respond to queries to remain results 92 in developers being afraid to deploy implementations of recent 93 standards. Such servers need to be identified and corrected / 94 replaced. 96 The DNS has response codes that cover almost any conceivable query 97 response. A nameserver should be able to respond to any conceivable 98 query using them. 100 Unless a nameserver is under attack, it should respond to all queries 101 directed to it as a result of following delegations. Additionally 102 code should not assume that there isn't a delegation to the server 103 even if it is not configured to serve the zone. Broken delegations 104 are a common occurrence in the DNS and receiving queries for zones 105 that you are not configured for is not a necessarily a indication 106 that you are under attack. Parent zone operators are supposed to 107 regularly check that the delegating NS records are consistent with 108 those of the delegated zone and to correct them when they are not 109 [RFC1034]. If this was being done regularly, the instances of broken 110 delegations would be much lower. 112 When a nameserver is under attack it may wish to drop packets. A 113 common attack is to use a nameserver as a amplifier by sending 114 spoofed packets. This is done because response packets are bigger 115 than the queries and big amplification factors are available 116 especially if EDNS is supported. Limiting the rate of responses is 117 reasonable when this is occurring and the client should retry. This 118 however only works if legitimate clients are not being forced to 119 guess whether EDNS queries are accept or not. While there is still a 120 pool of servers that don't respond to EDNS requests, clients have no 121 way to know if the lack of response is due to packet loss, EDNS 122 packets not being supported or rate limiting due to the server being 123 under attack. Mis-classifications of server characteristics are 124 unavoidable when rate limiting is done. 126 2. Common queries class that result in non responses. 128 There are three common query classes that result in non responses 129 today. These are EDNS queries, queries for unknown (unallocated) or 130 unsupported types, and filtering of TCP queries. 132 2.1. EDNS Queries - Version Independent 134 Identifying servers that fail to respond to EDNS queries can be done 135 by first identifying that the server responds to regular DNS queries, 136 followed by a series of otherwise identical responses using EDNS, 137 then making the original query again. A series of EDNS queries is 138 needed as at least one DNS implementation responds to the first EDNS 139 query with FORMERR but fails to respond to subsequent queries from 140 the same address for a period until a regular DNS query is made. The 141 EDNS query should specify a UDP buffer size of 512 bytes to avoid 142 false classification of not supporting EDNS due to response packet 143 size. 145 If the server responds to the first and last queries but fails to 146 respond to most or all of the EDNS queries, it is probably faulty. 147 The test should be repeated a number of times to eliminate the 148 likelihood of a false positive due to packet loss. 150 Firewalls may also block larger EDNS responses but there is no easy 151 way to check authoritative servers to see if the firewall is 152 misconfigured. 154 2.2. EDNS Queries - Version Specific 156 Some servers respond correctly to EDNS version 0 queries but fail to 157 respond to EDNS queries with version numbers that are higher than 158 zero. Servers should respond with BADVERS to EDNS queries with 159 version numbers that they do not support. 161 Some servers respond correctly to EDNS version 0 queries but fail to 162 set QR=1 when responding to EDNS versions they do not support. Such 163 answers are discarded or treated as requests. 165 2.3. EDNS Options 167 Some servers fail to respond to EDNS queries with EDNS options set. 168 Unknown EDNS options are supposed to be ignored by the server 169 [RFC6891]. 171 2.4. EDNS Flags 173 Some servers fail to respond to EDNS queries with EDNS flags set. 174 Server should ignore EDNS flags there do not understand and should 175 not add them to the response [RFC6891]. 177 2.5. DNS Flags 179 Some servers fail to respond to DNS queries with various DNS flags 180 set, regardless of whether they are defined or still reserved. At 181 the time of writing there are servers that fail to respond to queries 182 with the AD bit set to 1 and servers that fail to respond to queries 183 with the last reserved flag bit set. 185 2.6. Unknown / Unsupported Type Queries 187 Identifying servers that fail to respond to unknown or unsupported 188 types can be done by making an initial DNS query for an A record, 189 making a number of queries for an unallocated type, them making a 190 query for an A record again. IANA maintains a registry of allocated 191 types. 193 If the server responds to the first and last queries but fails to 194 respond to the queries for the unallocated type, it is probably 195 faulty. The test should be repeated a number of times to eliminate 196 the likelihood of a false positive due to packet loss. 198 2.7. Unknown DNS opcodes 200 The use of previously undefined opcodes is to be expected. Since the 201 DNS was first defined two new opcodes have been added, UPDATE and 202 NOTIFY. 204 NOTIMP is the expected rcode to an unknown / unimplemented opcode. 206 Note: while new opcodes will most probably use the current layout 207 structure for the rest of the message there is no requirement than 208 anything other than the DNS header match. 210 2.8. TCP Queries 212 All DNS servers are supposed to respond to queries over TCP 213 [RFC5966]. Firewalls that drop TCP connection attempts rather that 214 resetting the connect attempt or send a ICMP/ICMPv6 administratively 215 prohibited message introduce excessive delays to the resolution 216 process. 218 Whether a server accepts TCP connections can be tested by first 219 checking that it responds to UDP queries to confirm that it is up and 220 operating, then attempting the same query over TCP. An additional 221 query should be made over UDP if the TCP connection attempt fails to 222 confirm that the server under test is still operating. 224 3. Remediating 226 While the first step in remediating this problem is to get the 227 offending nameserver code corrected, there is a very long tail 228 problem with DNS servers in that it can often take over a decade 229 between the code being corrected and a nameserver being upgraded with 230 corrected code. With that in mind it is requested that TLD, and 231 other similar zone operators, take steps to identify and inform their 232 customers, directly or indirectly through registrars, that they are 233 running such servers and that the customers need to correct the 234 problem. 236 TLD operators should construct a list of servers child zones are 237 delegated to along with a delegated zone name. This name shall be 238 the query name used to test the server as it is supposed to exist. 240 For each server the TLD operator shall make an SOA query of the 241 delegated zone name. This should result in the SOA record being 242 returned in the answer section. If the SOA record is not returned 243 but some other response is returned, this is a indication of a bad 244 delegation and the TLD operator should take whatever steps it 245 normally takes to rectify a bad delegation. If more that one zone is 246 delegated to the server, it should choose another zone until it finds 247 a zone which responds correctly or it exhausts the list of zones 248 delegated to the server. 250 If the server fails to get a response to a SOA query, the TLD 251 operator should make an A query as some nameservers fail to respond 252 to SOA queries but respond to A queries. If it gets no response to 253 the A query, another delegated zone should be queried for as some 254 nameservers fail to respond to zones they are not configured for. If 255 subsequent queries find a responding zone, all delegation to this 256 server need to be checked and rectified using the TLD's normal 257 procedures. 259 Having identified a working tuple the TLD 260 operator should now check that the server responds to EDNS, Unknown 261 Query Type and TCP tests as described above. If the TLD operator 262 finds that server fails any of the tests, the TLD operator shall take 263 steps to inform the operator of the server that they are running a 264 faulty nameserver and that they need to take steps to correct the 265 matter. The TLD operator shall also record the 266 for follow-up testing. 268 If repeated attempts to inform and get the customer to correct / 269 replace the faulty server are unsuccessful the TLD operator shall 270 remove all delegations to said server from the zone. 272 It will also be necessary for TLD operators to repeat the scans 273 periodically. It is recommended that this be performed monthly 274 backing off to bi-annually once the numbers of faulty servers found 275 drops off to less than 1 in 100000 servers tested. Follow-up tests 276 for faulty servers still need to be performed monthly. 278 Some operators claim that they can't perform checks at registration 279 time. If a check is not performed at registration time, it needs to 280 be performed within a week of registration in order to detect faulty 281 servers swiftly. 283 Checking of delegations by TLD operators should be nothing new as 284 they have been required from the very beginnings of DNS to do this 285 [RFC1034]. Checking for compliance of nameserver operations should 286 just be a extension of such testing. 288 It is recommended that TLD operators setup a test web page which 289 performs the tests the TLD operator performs as part of their regular 290 audits to allow nameserver operators to test that they have correctly 291 fixed their servers. Such tests should be rate limited to avoid 292 these pages being a denial of service vector. 294 4. Firewalls and Load Balancers 296 Firewalls and load balancers can affect the externally visible 297 behaviour of a nameserver. Tests for conformance need to be done 298 from outside of any firewall so that the system as a whole is tested. 300 Firewalls and load balancers should not drop DNS packets that they 301 don't understand. They should either pass through the packets or 302 generate an appropriate error response. 304 Requests for unknown query types are not attacks and should not be 305 treated as such. 307 Requests with unassigned flags set (DNS or EDNS) are not attacks and 308 should not be treated as such. The behaviour for unassigned is to 309 ignore them in the request and to not set them in the response. All 310 dropping DNS / EDNS packets with unassigned flags does is make it 311 harder to deploy extensions that make use of them due to the need to 312 reconfigure / update firewalls. 314 Requests with unknown EDNS options are not an attack and should not 315 be treated as such. The correct behaviour for unknown EDNS options 316 is to ignore them. 318 Requests with unknown EDNS versions are not a attack and should not 319 be treated as such. The correct behaviour for unknown EDNS versions 320 is to return BADVERS along with the highest EDNS version the server 321 supports. All dropping EDNS packets does is break EDNS version 322 negotiation. 324 Firewalls should not assume that there will only be a single response 325 message to a requests. There have been proposals to use EDNS to 326 signal that multiple DNS messages be returned rather than a single 327 UDP message that is fragmented at the IP layer. 329 5. Scrubbing Services 331 Scrubbing services, like firewalls, can affect the externally visible 332 behaviour of a nameserver. If you use a scrubbing service, you 333 should check that legitimate queries are not being blocked. 335 Scrubbing services, unlike firewalls, are also turned on and off in 336 response to denial of service attacks. One needs to take care when 337 choosing a scrubbing service and ask questions like: 339 Do they pass unknown DNS query types? 340 Do they pass unknown EDNS versions? 341 Do they pass unknown EDNS options? 342 Do they pass unknown EDNS flags? 343 Do they pass requests with unknown DNS opcodes? 344 Do they pass requests with the remaining reserved DNS header flag 345 bit set? 347 All of these are not attack vectors but some scrubbing services treat 348 them as such. 350 6. Response Code Selection 352 Choosing the correct response code when fixing a nameserver is 353 important. Just because a type is not implemented does not mean that 354 NOTIMP is the correct response code to return. Response codes need 355 to be chosen considering how clients will handle them. 357 For unimplemented opcodes NOTIMP is the expected response code. 358 Additionally a new opcode could change the message format by 359 extending the header or changing the structure of the records etc. 361 This may result in FORMERR being returned though NOTIMP would be more 362 correct. 364 In general, for unimplemented type codes Name Error (NXDOMAIN) and 365 NOERROR (no data) are the expected response codes. A server is not 366 supposed to serve a zone which contains unsupported types ([RFC1034]) 367 so the only thing left is return if the QNAME exists or not. NOTIMP 368 and REFUSED are not useful responses as they force the clients to try 369 all the authoritative servers for a zone looking for a server which 370 will answer the query. 372 Meta queries type may be the exception but these need to be thought 373 about on a case by case basis. 375 If you support EDNS and get a query with an unsupported EDNS version, 376 the correct response is BADVERS [RFC6891]. 378 If you do not support EDNS at all, FORMERR and NOTIMP are the 379 expected error codes. That said a minimal EDNS server implementation 380 just requires parsing the OPT records and responding with an empty 381 OPT record. There is no need to interpret any EDNS options present 382 in the request as unsupported options are expected to be ignored 383 [RFC6891]. 385 7. Testing 387 Verify the server is configured for the zone: 389 dig +noedns +noad +norec soa $zone @$server 391 expect: status: NOERROR 392 expect: SOA record 394 Check that TCP queries work: 396 dig +noedns +noad +norec +tcp soa $zone @$server 398 expect: status: NOERROR 399 expect: SOA record 401 Check that queries for an unknown type to work: 403 dig +noedns +noad +norec type1000 $zone @$server 405 expect: status: NOERROR 406 expect: an empty answer section. 408 Check that queries with CD=1 work: 410 dig +noedns +noad +norec +cd soa $zone @$server 412 expect: status: NOERROR 413 expect: SOA record to be present 415 Check that queries with AD=1 work: 417 dig +noedns +norec +ad soa $zone @$server 419 expect: status: NOERROR 420 expect: SOA record to be present 422 Check that queries with the last unassigned DNS header flag to work: 424 dig +noedns +noad +norec +zflag soa $zone @$server 426 expect: status: NOERROR 427 expect: SOA record to be present 428 expect: MBZ to not be in the response 430 MBZ (Must Be Zero) presence indicates the flag bit has been copied. 432 Check that plain EDNS queries work: 434 dig +edns=0 +noad +norec soa $zone @$server 436 expect: status: NOERROR 437 expect: SOA record to be present 438 expect: OPT record to be present 439 expect: EDNS Version 0 in response 441 Check that EDNS version 1 queries work (EDNS supported): 443 dig +edns=1 +noednsneg +noad +norec soa $zone @$server 445 expect: status: BADVERS 446 expect: SOA record to not be present 447 expect: OPT record to be present 448 expect: EDNS Version 0 in response 450 (Only EDNS Verion 0 is currently defined so the response should 451 always be a 0 version. This will change when EDNS version 1 is 452 defined.) 454 Check that EDNS queries with an unknown option work (EDNS supported): 456 dig +edns=0 +noad +norec +ednsopt=100 soa $zone @$server 458 expect: status: NOERROR 459 expect: SOA record to be present 460 expect: OPT record to be present 461 expect: OPT=100 to not be present 462 expect: EDNS Version 0 in response 464 Check that EDNS queries with unknown flags work (EDNS supported): 466 dig +edns=0 +noad +norec +ednsflags=0x40 soa $zone @$server 468 expect: status: NOERROR 469 expect: SOA record to be present 470 expect: OPT record to be present 471 expect: MBZ not to be present 472 expect: EDNS Version 0 in response 474 MBZ (Must Be Zero) presence indicates the flag bit has been copied. 476 Check that EDNS version 1 queries with unknown flags work (EDNS 477 supported): 479 dig +edns=1 +noednsneg +noad +norec +ednsflags=0x40 soa \ 480 $zone @$server 482 expect: status: BADVERS 483 expect: SOA record to NOT be present 484 expect: OPT record to be present 485 expect: MBZ not to be present 486 expect: EDNS Version 0 in response 488 +noednsneg disables EDNS version negotiation in DiG; MBZ (Must Be 489 Zero) presence indicates the flag bit has been copied. 491 Check that EDNS version 1 queries with unknown options work (EDNS 492 supported): 494 dig +edns=1 +noednsneg +noad +norec +ednsopt=100 soa $zone @$server 496 expect: status: BADVERS 497 expect: SOA record to NOT be present 498 expect: OPT record to be present 499 expect: OPT=100 to NOT be present 500 expect: EDNS Version 0 in response 502 +noednsneg disables EDNS version negotiation in DiG. 504 Check that a DNSSEC queries work (EDNS supported): 506 dig +edns=0 +noad +norec +dnssec soa $zone @$server 508 expect: status: NOERROR 509 expect: SOA record to be present 510 expect: OPT record to be present 511 expect: DO=1 to be present if a RRSIG is in the response 512 expect: EDNS Version 0 in response 514 DO=1 should be present if RRSIGs are returned as they indicate that 515 the server supports DNSSEC. Servers that support DNSSEC are supposed 516 to copy the DO bit from the request to the response as per [RFC3225]. 518 Check that EDNS version 1 DNSSEC queries work (EDNS supported): 520 dig +edns=1 +noednsneg +noad +norec +dnssec soa \ 521 $zone @$server 523 expect: status: BADVERS 524 expect: SOA record to not be present 525 expect: OPT record to be present 526 expect: DO=1 to be present if the EDNS version 0 DNSSEC query test 527 returned DO=1 528 expect: EDNS Version 0 in response 530 +noednsneg disables EDNS version negotiation in DiG. 532 Check that new opcodes are handled: 534 dig +noedns +noad +opcode=15 +norec soa $zone @$server 536 expect: status: NOTIMP 537 expect: SOA record to not be present 539 If EDNS is not supported by the nameserver, we expect a response to 540 all the above queries. That response may be a FORMERR or NOTIMP 541 error response or the OPT record may just be ignored. 543 It is advisable to run all the above tests in parallel so as to 544 minimise the delays due to multiple timeouts when the servers do not 545 respond. 547 The above tests use dig from BIND 9.11.0 which is still in 548 development. 550 8. Security Considerations 552 Testing protocol compliance can potentially result in false reports 553 of attempts to break services from Intrusion Detection Services and 554 firewalls. None of the tests listed above should break nominally 555 EDNS compliant servers. None of the tests above should break non 556 EDNS servers. All the tests above are well formed, though not 557 necessarily common, DNS queries. 559 Relaxing firewall settings to ensure EDNS compliance could 560 potentially expose a critical implementation flaw in the nameserver. 561 Nameservers should be tested for conformance before relaxing firewall 562 settings. 564 9. IANA Considerations 566 IANA / ICANN needs to consider what tests, if any, from above that it 567 should add to the zone maintenance procedures for zones under its 568 control including pre-delegation checks. Otherwise this document has 569 no actions for IANA. 571 10. Normative References 573 [RFC1034] Mockapetris, P., "Domain names - concepts and facilities", 574 STD 13, RFC 1034, DOI 10.17487/RFC1034, November 1987, 575 . 577 [RFC1035] Mockapetris, P., "Domain names - implementation and 578 specification", STD 13, RFC 1035, DOI 10.17487/RFC1035, 579 November 1987, . 581 [RFC3225] Conrad, D., "Indicating Resolver Support of DNSSEC", 582 RFC 3225, DOI 10.17487/RFC3225, December 2001, 583 . 585 [RFC5966] Bellis, R., "DNS Transport over TCP - Implementation 586 Requirements", RFC 5966, DOI 10.17487/RFC5966, 587 August 2010, . 589 [RFC6891] Damas, J., Graff, M., and P. Vixie, "Extension Mechanisms 590 for DNS (EDNS(0))", STD 75, RFC 6891, DOI 10.17487/ 591 RFC6891, April 2013, 592 . 594 Author's Address 596 M. Andrews 597 Internet Systems Consortium 598 950 Charter Street 599 Redwood City, CA 94063 600 US 602 Email: marka@isc.org