| dnsext | P. Wouters |
| Internet-Draft | Red Hat |
| Intended status: Standards Track | February 14, 2014 |
| Expires: August 18, 2014 |
Chain Query requests in DNS
draft-wouters-edns-chain-query-00
This document defines an EDNS0 extension that can be used by a DNSSEC enabled Recursive Nameserver configured as a forwarder to send a single DNS query requesting to receive a complete validation path along with the regular DNS answer, without the need to rapid-fire many UDP requests in an attempt to attain a low latency.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on August 18, 2014.
Copyright (c) 2014 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Traditionally, clients operate in stub-mode for DNS. For each DNS question the client needs to resolve, it sends a single query to an upstream DNS resolver to obtain a single DNS answer. When DNSSEC [RFC4033] is deployed on such clients, validation requires that the client obtains all the (intermediate) information from the DNS root down to the queried-for hostname so it can perform DNSSEC validation on the complete chain of trust. This process increases the number of DNS queries and answers required, and thus increases the latency before a validated DNS answer has been obtained.
Currently, applications use a rapid-fire approach to send out many UDP requests concurrently. This requires more resources on the DNS client with respect to state (cpu, memory, battery) and bandwidth. There is also no guarantee that the initial burst of UDP questions will result in all the records required for DNSSEC validation, and more round trips could be required depending on the resulting DNS answers. This especially affects high-latency links.
This document specifies an EDNS0 extension that allows a validating recursive name server running as a forwarder to request another recursive name server for a DNS chain answer using one DNS query/answer pair. This reduces the number of round-trip times ("RTT") to two. If combined with [TCP-KEEPALIVE] there is only 1 RTT. While the upstream DNS resolver still needs to perform all the individual queries required for the complete answer, it usually has a much bigger cache and does not experience significant slowdown from last-mile latency.
This EDNS0 extension allows the Forwarder to indicate which part of the DNS hierarchy it already contains in its cache. This reduces the amount of data required to be transferred and reduces the work the upstream Resolving Nameserver has to perform.
This EDNS0 extension is only intended for Forwarders. It can (and should be) ignored by Authoritative Nameservers and by Recursive Nameservers that do not support this EDNS0 option.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
When DNSSEC is deployed on the client, it can no longer delegate all DNS work to the upstream Resolving Nameserer. Obtaining just the DNS answer itself is not enough to validate that answer using DNSSEC. For DNSSEC validation, the client requires a locally running validating DNS server configured as Resolving Nameserver so it can confirm DNSSEC validation of all intermediary DNS answers. It can configure itself as a Forwarder if the DHCP server has indicated that one or more Resolving Nameservers are available. Regardless, generating the required queries for validation adds a significant delay in answering the DNS question of the locally running application. The application has to wait while the Forwarder on the client is querying for all the intermediate work. Each round-trip adds to the total time waiting on DNS resolving to complete. This makes DNSSEC resolving impractical on networks with a high latency.
The edns-chain-query option allows the client to request all intermediate DNS data it requires to resolve and validate a particular DNS answer in a single round-trip DNS query and answer.
Servers answering with chain query data exceeding 512 bytes should ensure that the transport is TCP or source IP address verified UDP. See Section 8. This avoids abuse in DNS amplification attacks.
The format of this option is described in Section 4.
As described in Section 5.3, a recursive nameserver could use this EDNS0 option to include additional data required by the client in the Authority Section of the DNS answer packet when using a source IP verified transport. The Answer Section remains unchanged from a traditional DNS answer and contains the answer and related DNSSEC entries.
An empty edns-chain-query EDNS0 option MAY be sent over any transport as a discovery method. A DNS server receiving such an empty edns-chain-query option SHOULD add an empty edns-chain-query option in its answer to indicate that it supports edns-chain-query for source IP address verified transports.
The mechanisms provided by edns-chain-query raise various security related concerns, related to the additional work, bandwidth, amplification attacks as well as privacy issues with the cache. These concerns are described in Section 8.
This draft uses an EDNS0 ([RFC2671]) option to include client IP information in DNS messages. The option is structured as follows:
1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-------------------------------+-------------------------------+
! OPTION-CODE ! OPTION-LENGTH !
+-------------------------------+-------------------------------+
~ Last Known Query Name (FQDN) ~
+---------------------------------------------------------------+
A Forwarder may include a zero-length edns-chain-query option in queries over UDP or TCP to discover the DNS server capability for edns-chain-query. DNS Servers that support and are willing to accept chain queries over TCP SHOULD respond to a zero-length edns-chain-query received over UDP or TCP queries by including a zero-length edns-chain-query option in the answer. A Forwarder MAY then switch to the TCP transport and sent a non-zero edns-chain-query value to request a chain-query response from the DNS server.
The edns-chain-query option should generally be deployed by Forwarders, as described in Section 5.4.
In this option value, the Forwarder sets the last known entry point in the chain - furthest from the root - that it already has a DNSSEC validated (secure or not) answer for in its cache. The upstream Recursive Resolver does not need to include any part of the chain from the root down to this option's FQDN. A complete example is described in Section 9.
Depending on the size of the labels of the last known entry point value, a DNS Query packet could be arbitrarily large. If using the last known entry point would result in a query size of more then 512 bytes, the last known entry point should be replaced with its parent entry until the query size would be 512 bytes or less.
When a query containing a non-zero edns-chain-query option is received over a TCP connection from a Forwarder, the upstream Recursive Resolver supporting edns-chain-query MAY respond by confirming that it is returning a DNS Query Chain. To do so, it MUST set the edns-chain-query option with an OPTION-LENGTH of zero to indicate the DNS answer contains a Chain Query. It extends the Authority Section for the DNS answer packet with the required DNS RRSets resulting in an Authority Section that contains a complete chain of DNS RRsets that start with the first chain element below the received Last Known Query Name upto and including the NS and DS RRsets that represent the zone cut (authoritative servers) of the QNAME. The actual DNS answer to the question in the Query Section is placed in the DNS Answer Section identical to traditional DNS answers. If the received query has the DNSSEC OK flag set, all required DNSSEC related records must be added to their appropriate sections. This includes records required for proof of non-existence of regular and/or wildcard records, such as NSEC or NSEC3 records.
Recursive Resolvers that have not implemented or enabled support for the edns-chain-query option, or are otherwise unwilling to perform the additional work for a Chain Query due to work load, may safely ignore the option in the incoming queries. Such a server MUST NOT include an edns-chain-query option when sending DNS answer replies back, thus indicating it is not able to support Chain Queries at this time.
Requests with wrongly formatted options (i.e. bogus FQDN) MUST be rejected and a FORMERR response must be returned to the sender, as described by [RFC2671], Transport Considerations.
Requests resulting in chains that the receiving resolver is unwilling to serve can be rejected by sending a REFUSED response to the sender, as described by [RFC2671], Transport Considerations. This refusal can be used for chains that would be too big or chains that would reveal too much information considered private.
At any time, a DNS server that has determined that it is running low on resources can refuse to acknowledge a Chain Query by omitting the edns-chain-query option. It may do so even if it conveyed support to a DNS client previously. If [TCP-KEEPALIVE] is used, it may even change its support for edns-chain-query within the same TCP session.
If the DNS request results in an CNAME or DNAME for the Answer Section, the DNS server MUST return these records in the Answer Section similar to regular DNS processing. The CNAME or DNAME target MAY be placed in the Additional Section only if all supporting records for DNSSEC validation of the CNAME or DNAME target is also added to the Authority Section.
In any case, the response from the receiving resolver to the client resolver MUST NOT contain the edns-chain-query option if none was present in the client's resolver original request.
When edns-chain-query is available, the downstream Resolving Nameserver can adjust its query strategy based on the desired queries and its cache contents.
A Forwarder can request the edns-chain-query option with every outgoing DNS query. However, it is RECOMMENDED that Forwarders remember which upstream Resolving Nameservers did not return the option (and additional data) with their response. The Forwarder SHOULD fallback to regular DNS for subsequent queries to those Recursive Nameservers. It MAY switch to another Resolving Nameserver that does support the edns-chain-query option or try again later to see if the server has become less loaded and is now willing to answer with Query Chains.
The presence or absence of an OPT resource record containing an edns-chain-query option in a DNS query does not change the usage of those resource records and mechanisms used to provide data origin authentication and data integrity to the DNS, as described in [RFC4033], [RFC4034] and [RFC4035].
edns-chain-query responses MUST include the NS RRset from the child zone, which includes DNSSEC RRSIG records required for validation.
When a DNSSEC chain is supplied via edns-chain-query, the Forwarder no longer requires to use the NS RRset, as it can construct the validation path via the DNSKEY and DS RRsets without using the NS RRset. However, it is prefered that the Forwarder can populate its cache with this information regardless, to avoid requiring queries in the future just to obtain the missing NS records. This can happen on a roaming device that needs to swich from using a DHCP obtained DNS server as forwarder to running in full autonomous resolver mode, for example when the DHCP obtained DNS server is broken in some way.
It is recommended that TCP Chain Queries are used in combination with [TCP-KEEPALIVE].
Both DNS clients and servers are subject to resource constraints which will limit the extent to which TCP Chain Queries can be executed. Effective limits for the number of active sessions that can be maintained on individual clients and servers should be established, either as configuration options or by interrogation of process limits imposed by the operating system.
In the event that there is greater demand for TCP Chain Queries than can be accommodated, DNS servers may stop advertising the edns-query-chain option in successive DNS messages. This allows, for example, clients with other candidate servers to query to establish new TCP sessions with different servers in expectation that those servers might still allow TCP Chain Queries.
Many paths between DNS clients and servers suffer from poor hygiene, limiting the free flow of DNS messages that include particular EDNS0 options, or messages that exceed a particular size. A fallback strategy similar to that described in [RFC6891] section 6.2.2 SHOULD be employed to avoid persistent interference due to non-clean paths.
DNS servers of various types are commonly deployed using anycast [RFC4786].
Successive DNS transactions between a client and server using UDP transport may involve responses generated by different anycast nodes, and the use of anycast in the implementation of a DNS server is effectively undetectable by the client. The edns-chain-query option SHOULD NOT be included in responses using UDP transport from servers provisioned using anycast unless all anycast server nodes are capable of processing the edns-query-chain option.
Changes in network topology between clients and anycast servers may cause disruption to TCP sessions making use of edns-chain-query more often than with TCP sessions that omit it, since the TCP sessions are expected to be longer-lived. Anycast servers MAY make use of TCP multipath [RFC6824] to anchor the server side of the TCP connection to an unambiguously-unicast address in order to avoid disruption due to topology changes.
This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in [RFC6982]. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist.
According to [RFC6982], "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit".
[While there is some interest, no work has started yet]
Chain Queries can potentially send very large DNS answers. Attackers could abuse this using spoofed source IP addresses to inflict large Distributed Denial of Service attacks using query-chains as an amplification vector in their attack. While TCP is not vulnerable for this type of abuse, the UDP protocol is vulnerable to this.
A recursive nameserver MUST NOT return Query Chain answers to clients over UDP without source IP address verification, for instance using [EASTLAKE-COOKIES]. A recursive nameserver SHOULD signal support in response to a Query Chain request over UDP by responding using a zero-length edns-chain-query option over UDO even without source IP address verification.
If the answer does not exist, and the zone uses DNSSEC, it also adds the proof of non-existance, such as NSEC or NSEC3 records, to the Authority Section.
If the answer does not exist (no-data or NXDOMAIN), the Answer Section remains empty. For the NXDOMAIN case, the RCode of the DNS answer packet is set to NXDOMAIN. Otherwise it remains NOERROR.
A Recursive Resolver receives a query for the A record for example.com. It includes the edns-chain-query option with the following parameters:
As there is no chain that leads from "unrelated.ca." to "example.com", the Resolving Nameserver answers with RCODE "FormErr". It includes the edns-chain-query with the following parameters:
A Recursive Resolver receives a query for the A record for "ipv6.toronto.redhat.ca". It includes the edns-chain-query option with the following parameters:
Using regular UDP queries towards Authoritative Nameservers, it locates the NS RRset for "toronto.redhat.ca.". When querying for the A record it receives a reply with RCODE "NoError" and an empty Answer Section. The Authority Section contains NSEC3 and RRSIG records proving there is no A RRtype for the QNAME "ipv6.toronto.redhat.ca".
The Recursive Resolver constructs a DNS reply with the following edns-chain-query option parameters:
The RCODE is set to "NoError". The Authority Section is filled in with:
The Answer Section is empty. The RCode is set to NOERROR.
IANA has assigned option code [TBD] in the "DNS EDNS0 Option Codes (OPT)" registry to edns-chain-query.
Andrew Sullivan pointed out that we do not need any new data formats to support DNS chains. Olafur Gudmundsson ensured the RRsets are returned in the proper Sections.