| Internet-Draft | NFSv4 Security Needs | December 2020 |
| Noveck & Lever | Expires 19 June 2021 | [Page] |
This document discusses the existing inadequacies in the handling of security issues within the NFSv4 protocols and the steps that need to be taken to correct the situation, including the preparation of new standards-track documents updating existing treatments of NFSv4 security. Because the security approach has been and should remain the same for all minor versions, it is expected that the basics of security for all the NFSv4 protocols will be addressed together in a single NFSv4-wide standards-track document.¶
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 https://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 19 June 2021.¶
Copyright (c) 2020 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 (https://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.¶
This document will discuss the changes that need to be made in the security facilities to be used with NFSv4. These changes are made necessary by the fact that the original goal for NFSv4, that of enabling secure use on the internet has not been effectively realized and because the current approach to security does not address current needs for secure operation in many environments. This document is intended to serve as a resource for the working group, as it creates new standards-track documents to address the security issues that exist within NFSv4. These potential documents are listed and the likely processes expected to be used in producing them are discussed in Section 9.¶
Needed changes are expected to involve use of new transport-level security facilities provided within the RPC layer (See Section 7). These will involve the encryption of NFSv4 traffic and the probable use of client host authentication to improve security when AUTH_SYS is used, replacing the security approach specified in [RFC7530] and [RFC8881]. For a discussion of issues connected to shifting the security approach for existing protocols, see Section 4.2.¶
Although most use of NFSv4 is not within the internet, but in more restricted local area network environments, it appears that NFSv4 security, as actually used, needs updating to provide security appropriate to the current needs of these environments as well. Although restricting physical access to the network on which the client and server interact can result in limiting access to users associated with a particular organization, this step does not eliminate the need for protocol support for secure operation. The specific environments and the corresponding security needs are discussed in Section 4.1.¶
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 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
Because this is an informational document, it will generally not make normative statements including the keywords discussed in Section 2. However, in discussing the existing treatment of security in the NFSv4 protocols, there will be occasions in which we include quotations (most often direct but sometimes indirect) from existing standards-track documents that use these keywords. In such cases, the definitions within BCP 14 [RFC2119] [RFC8174] are to be adhered to when the terms appear in all-capitals.¶
In addition, we will often use these terms in exploring possible changes that might be made in possible future standards-track documents dealing with NFSv4 and RPC security. While these terms do not have a normative effect in this context, the meaning of these terms remains the same. These statements will differ from and may contradict corresponding existing normative statements for one or more of the following reasons:¶
Any such changes in the set of normative statements applying to implementations of a particular NFSv4 minor version require special care to avoid creating later difficulties. Problems might arise because:¶
We will consider a range of network environments and how the type of environment affects the need for the features being introduced to provide for secure operation.¶
Within a data center local area network, there may well be sufficient administrative controls over the software running on machines with physical access to the network as to make additional security features within NFS unnecessary.¶
If such a network is physically isolated or has all access to it controlled using an appropriately configured firewall, it may be acceptable, from a security point of view, to use AUTH_SYS without client host authentication, and without encryption, assuming all attackers have been excluded from access.¶
Although such networks typically do have firewalls, they are generally configured to allow significant access, including NFSv4 access, to traffic originating outside the data center, although not outside the owning organization. In such cases, we effectively have a within-organization network, dealt with in item 2 below.¶
Within a network devoted to a single organization or a single site within an organization. Such networks often have associated firewalls configured to exclude access from outside the organization, and restrict it inside the organization.¶
It had until recently been assumed that, by excluding access to those outside the organization, it could be assumed that attackers would also be excluded. However, this approach ignores the possibility of insider threats. Since we feel these threats cannot be ignored, the security of NFSv4 needs to be upgraded to prevent attacks by insiders, making the security needs in this case more like those on the internet.¶
Use on the internet is the most challenging. While secure use on the internet was a goal of NFSv4 and steps were taken to achieve that goal, the approach required a number of steps, including a significant performance penalty and a significantly different approach to systems administration. Given that those administering such systems were unwilling to adapt in this way, we now need to meet the goal of secure use on the internet in a new way.¶
Secure use on the internet requires not only protection of user data in-flight from unauthorized disclosure or modification, as in case 2 above, but also effort to deal with the possibility of extensive external monitoring and denial-of-service attacks.¶
As will be discussed below, NFSv4 security, as currently defined, has significant security issues in many of the above environments, although not in all. No special provisions will be made for the more restricted environments, except where necessary to avoid incompatibilities when interacting with existing implementations.¶
By focusing on a single set of requirements for all of these environments, there will necessarily be occasions in which some security-related work is required that is not, strictly speaking, needed for the particular environment in which it is used. However, we have endeavored to keep the required effort small, eliminating requirements for major administrative changes or compute-intensive additions to the processing path.¶
The fact that serious security issues exist in many of the environment discussed above has made it necessary to consider taking the drastic step of modifying security-related recommendations for existing protocols. This situation seems to call for an explanation and we will attempt to provide one.¶
One possible form this might take would be to explain why [RFC7530] and [RFC5661] made erroneous choices, with accompanying commentary explaining the choices that should have been made instead.¶
Such an explanation does not appear to be possible, since it would not have been possible for the authors and approvers of these documents to reach the conclusions we have adopted without an implausible level of foresight on their part. This is so even though we would not adopt the choices made in [RFC7530] and [RFC5661] today. Knowing what was known then, we might well have made the same choices, although it is hard not to fault the lack of attention, in the relevant documents, to the security consequences of the decisions made.¶
Although we will call attention in sections below to things which might have been done differently, this does indicate that we believe that the authors and approvers of these documents, could have arrived at conclusions similar to the ones we expect to choose now. Our choices are so different from the ones made previously because they take into account the needs of NFSv4 today and take advantage of security facilities not available earlier.¶
In order to better understand this divergence, we have to look at the history of NFS and take note of the context in which assumptions were established with regard to security or changes in security guidance occurred. In understanding this history and the choices of RFC2119-defined keywords, the constraints on the working group need to be understood, as described in Section 6.2¶
When NFSv4 was developed, around twenty years ago, serious steps were taken to improve the security situation by requiring the implementation of RPCSEC_GSS. This included support for encryption, necessary to support use on the internet. AUTH_SYS was retained as an additional optional means of authentication, which would have allowed implementations to drop support at a later point, although, as things developed, that became less and less likely as use of AUTH_SYS remained quite common.¶
While it is possible to interpret the treatment of the matter in [RFC3530] as implying that these two (i.e. RPCSEC_GSS and AUTH_SYS) were equivalent and to find fault with that implication, it is not clear how different choices would have led to a better result. In particular, attempts to deprecate AUTH_SYS (e.g. by using the phrase "SHOULD NOT") would have been problematic in that they might have retarded adoption of NFSv4 rather than enhancing adoption of RPCSEC_GSS.¶
When NFSv4.1 was developed, there appeared to be no reason to change the security decisions made for NFSv4.0. Although it might have been possible to change the status of AUTH_SYS in NFSv4.1, there was no reason to tie the session architecture to changes in security that many implementers would have been unwilling to make.¶
The only significant security-related extension was the provision made for state protection, which required the server to protect clients from one another, apart from the users on whose behalf requests were being made.¶
With the above history in mind, we will look at why NFSv4 security needs are different today and take advantage of some options that were not available on earlier occasions in which NFSv4 security decisions were made.¶
These options have been made available by the work at the RPC layer described in Section 7 while the specifics of having NFSv4 take advantage of these options are described in Section 8.¶
Providing a new security approach for multiple existing protocols is potentially disruptive and due care will need to be taken to avoid damaging implementation incompatibilities. However, as will be discussed below, the existing situation, in which serious security inadequacies need to be addressed, requires that significant changes be made. It is our expectation that the situation will be resolved by the eventual publication of a Standards-track document updating [RFC7530] and [RFC8881] and much of this document will concern itself with determining how the needed changes can resolve existing security issues without undue disruption.¶
The problems to be addressed concern the way that security information has been conveyed in earlier specifications (discussed in Section 5.1) as well as two sets of substantive security weaknesses discussed in Sections 5.2 and 5.3.¶
Although the inadequacies in the presentation of security issues have contributed to prolongation of the substantive weaknesses and will be discussed in the latter two sections, there is no reason to believe that correction of the presentation problems, will, by itself, improve the security situation, since the substantive problems still need to be addressed. Nevertheless, correction of the presentation issues is necessary so that the proposed solutions to the substantive security issues receive the proper attention and analysis, both now in connection with the changes to be proposed, and also going forward, as needs change.¶
While attention has previously focused on the deficiencies of the Security Considerations sections (e.g. Lack of a threat analysis), this is not the only presentation issue that needs to be addressed.¶
Problems with the overall presentation of NFSv4 security, particularly the discussion of the security architecture will be dealt with in Section 5.1.1.¶
Problems with the evaluation of NFSv4 security including the lack of a threat analysis and the contents of the Security considerations sections will be dealt with in Section 5.1.2.¶
The presentation of the NFSv4 security architecture (for both minor versions) focuses on the work done to make RPCSEC_GSS usable in view of the basic architectural decisions made in going from NFSv3 to NFSv4. These include the use of NFS4ERR_WRONGSEC and SECINFO to allow the server namespace to be carved into regions with the use of particular security flavors and services (associated with particular quality-of- protection values) required in each one.¶
Unfortunately, features added later in the NFSv4.0 development cycle were not integrated into the description of the security architecture. A list of such noteworthy features follows.¶
Another consequence of the approach taken to writing Security Considerations sections has been that these sections may have not appropriately highlighted the security implications of new features being added to the protocols.¶
The addition of callbacks to NFS was not accompanied by an explanation of how security for callbacks was to be dealt with, It was never made clear whether the existing mandate to provide support for RPCSEC_GSS applied to reverse-direction operation. The fact that there is no reverse-direction complement to NFS4ERR_WRONGSEC and SECINFO means that there is no way that the client, as a responder, could force use of another flavor if it did not provide support for RPCSEC_GSS authentication, in the role of a responder.¶
The fact that callbacks are normally issued by the server itself and not on behalf a specific user, means that the typical function of RPCSEC_GSS authentication is not relevant to the application. However, if RPCSEC_GSS is used in the forward direction, then the mutual authentication of client and server should be adequate to provide authentication of the server to the client, although the possibility that an attacker might inject a spurious callback to the reverse-direction request stream remains a concern.¶
In any case, the absence of mention of reverse-direction authentication in a description of the NFSv4 security approach is troubling. The lack of mention in the Security Considerations section is noteworthy because this creates a large set of requests and responses for which there is no available facility to mandate encryption, and that lack could be exploited by monitoring attacks.¶
The addition of pNFS to NFSv4.1 [RFC5661] required that there be a means of providing that the proper quality of protection by provided by the a pNFS data server implementing the files mapping type. Given that SECINFO is namespace-oriented, it was necessary to define SECINFO_NONAME, implementable by the data server, to provide this functionality.¶
This security-related change, while explained adequately in [RFC8881], is not mentioned in the description of the security architecture or the Security Considerations section. As this feature does create new threats, it should have been mentioned in the Security Considerations section. Also, this change while compatible with the existing NFSv4 security architecture, does represent a considerable change to it. It seems to call for special mention in a standards-track document devoted to NFSv4 security issues. It makes sense to include it in an introductory section providing an overview of the NFSv4 security architecture.¶
The implementation of sessions in NFSv4.1 [RFC5661] created a requirement to protect clients from one another, even when they were making requests on behalf of the same user. Although adequate means of dealing with this issue were described in [RFC5661], they were OPTIONAL features and implementation has been very limited.¶
In this case, the additional exposure makes the absence of mention of the issue in the Security Considerations section troublesome, although its role in the limited implementation of the corresponding security features is hard to evaluate at this point. In any case, this was a major architectural change to the security architecture which requires more attention. Over the years, there has been considerable discussion on the NFSv4 mailing list of the lack of authentication of the client, as opposed to the client users on whose behalf it is acting, although no work has been undertaken to make this a basic element of NFSv4 security. Despite the delay, the issue now seems on its way to effective resolution, using the work done in [I-D.ietf-nfsv4-rpc-tls], as described in Section 8.5.¶
As we have seen, the description of the security architecture of NFSv4 is quite limited and a new security document will have some gaps to fill.¶
The principal difficulty in the overall approach to security taken in the Security Considerations section of [RFC7530] and [RFC8881] concerns the absence of a threat analysis within these documents. The absence of such an analysis has meant that:¶
While [RFC3530], [RFC7530], and [RFC8881] provide for the MANDATORY implementation of RPCSEC_GSS, they do not require its use, allowing an opportunity for unauthenticated requests to be issued by attackers targeting the many NFSv4 servers in which RPCSEC_GSS is not the only form of authentication provided, which appears to be the vast majority in operation.¶
Although there are a number of nominal authentication flavors that allow unauthenticated requests to be presented in circumstances in which they will typically be executed by the server, AUTH_SYS is of greatest importance for a number of reasons:¶
The common use of AUTH_SYS, which has continued until now, arose from the early history of NFS and its development, as part of the RPC authentication function, together with NFSv2 [RFC1094] and NFSv3 [RFC1813]. That approach, in which security was made the responsibility of large multi-user clients trusted by the servers, might have made sense when it was developed, but, as things changed, became less relevant to NFSv4 security needs.¶
In the shift to NFSv4, RPCSEC_GSS was added on a mandatory-to-implement basis, while AUTH_SYS was left in place as an alternative without any effective consideration of the corresponding consequences for security.¶
A number of common practices which servers have typically used in determining whether AUTH_SYS requests would be accepted are mentioned in associated standards-track documents (i.e. in [RFC5531], [RFC7530], RFC8881 [RFC8881]) often without clear discussion of their obvious security weaknesses.¶
As a result of the lack of attention to AUTH_SYS, there is no standards-track description of its implementation in connection with NFSv4. Nevertheless, the following practices seem to have been adhered to in providing whatever security does exist.¶
A common server practice, deriving from use with earlier versions of NFS, has been to limit acceptance of such requests to requests from particular known clients, as determined by the source address of the request.¶
The possibility of address spoofing is not discussed, although the Security Considerations section of [RFC7530] does state that this is "certainly not a safe model", most likely alluding to this possibility. This recognition is not followed up on to call into question the use of AUTH_SYS as an optional security flavor (stated earlier in the form "MAY be implemented"). Instead, it is used as justification for the mandatory nature of RPCSEC_GSS implementation, with the lack of security when AUTH_SYS is used not further mentioned.¶
Another common server practice is to restrict use of AUTH_SYS based on the associated client source port, assuming that this assures that the kernel is making the request and that the kernel can be trusted. That assumption has, over time, become more and more dubious, undercutting the underlying logic of AUTH_SYS, without any recognition of the consequent effect on the reasonableness of continued use of AUTH_SYS or other effective action to address the issue.¶
Appendix A of [RFC5531] refers to this practice as follows:¶
Many servers do not execute AUTH_SYS requests specifying a user id of zero (i.e. root) under that user id or limit such execution to a small set of known clients using the source address checking referred to above.¶
No mention of this practice (generally called "root squashing") has been found so far in IETF documents. Nevertheless, it continues to be implemented in connection with IETF-defined protocols such as NFSv4, despite this fact. While it might make an attacker's task more complicated, there is little reason to assume that an attacker able to assume any identity he chooses other than that of root will be effectively prevented from taking advantage of the ability to get unauthenticated requests executed, when AUTH_SYS is implemented by a target server.¶
As can be seen from the above, important elements of AUTH_SYS implementations are not specified within current standards-track documents in sufficient detail that interoperable implementations could be developed using these documents or subjected to a threat analysis, even if one were to be attempted.¶
The Security Considerations section of [RFC5531] states that AUTH_SYS is "known to be insecure" and further states "AUTH_SYS SHOULD NOT be used for services that permit clients to modify data. Nevertheless, AUTH_SYS has been commonly used by NFSv4 clients modifying data since NFSv4 was first implemented, while the consequences for security have never been addressed.¶
Given the security difficulties that use of AUTH_SYS gives rise to, we face a choice between two alternatives:¶
Take steps to eliminate or otherwise minimize use of AUTH_SYS as a valid authentication flavor, either for NFSv4 as a whole or for minor versions based on NFSv4.1, whose description was published as a Proposed Standard after [RFC5531], which stated that AUTH_SYS was "known to be insecure".¶
Given the many clients using AUTH_SYS, such steps might need to take the form of recommendations, with the difficulty of switching authentication models considered a valid reason to continue to use AUTH_SYS, as long as one is aware of the security consequences.¶
Other alternatives are possible and will be discussed later. However, regardless of the working group's choice of normative terms, it is important that the security consequences of using AUTH_SYS are clearly explained.¶
Revise AUTH_SYS to enable secure implementation of an authentication model in which authenticated client hosts are assigned the responsibility to determine, within appropriate limits, the user ids to be used for each request sent.¶
To do this we would authenticate client hosts issuing AUTH_SYS requests taking advantage of the authentication provided by [I-D.ietf-nfsv4-rpc-tls] as described in Section 7.2. There is also a more detailed discussion of issues to be addressed in establishing the rules for this form of authentication in Section 8.4.6.¶
Although both of these alternatives should be considered, there is reason to believe that the first of these is not feasible, for reasons similar to those that have prevented such a step being attempted earlier, either as part of the transition to NFSv4 or to NFSv4.1. The use of AUTH_SYS has continued despite the fact that alternatives were made available, and there is no reason to expect that the use of the words "SHOULD NOT" or "MUST NOT" would prevent implementers from continuing to support it, or administrators depending on such support, despite the negative effect on the security of NFSv4 implementations¶
Because of the current predominant role of AUTH_SYS in existing NFSv4 implementations, providing better security while remaining within the AUTH_SYS framework will also be difficult because of the issues discussed in Section 6.2. Nevertheless, we feel it needs to be attempted, as a new security framework cannot address AUTH_SYS without clearly discussing its security weaknesses. Once that is done, we need to take advantage of the ability to effect client host authentication provided by [I-D.ietf-nfsv4-rpc-tls] to provide a modicum of security while allowing administrators to delegate credential checking to trusted clients as they are now accustomed to doing.¶
While this approach is far from ideal, in that it requires a trust relationship between client and servers, adopting it provides a reasonable approach to support NFSv4 in the near-term while allowing development of a better alternative (e.g. a new authentication flavor) to proceed in parallel. With regard to our immediate need for better NFSv4 security, the obvious alternatives might not be realizable:¶
Despite this summary of the author's views on this matter, this question is ultimately a matter for working group decision. For a further discussion of related issues, see Section 8.4.3. Although the rest of this document is written assuming that the author's choice will be selected, Section 8.4.3 explains the changes to the eventual standards-track document that would be required if different choices were to be made.¶
Confidentiality is currently provided within NFSv4 by the use of RPCSEC_GSS, with the server charged with enforcing any administrator-specified needs to use these facilities on appropriate portions of the server namespace. Requirements for the use of integrity protection are handled similarly. This approach is capable of providing confidentiality when accessing certain directories or file systems, assuming that files that require such protection can be isolated in certain regions of the server namespace.¶
An important difficulty with regard to this approach to confidentiality is that there is significant non-encrypted data sent on NFSv4 connections which can allow extensive data to be gathered on the part of those engaged in monitoring attacks¶
Despite the availability of encryption in NFSv4, it is very rarely used, which makes its formal sufficiency essentially irrelevant. In understanding why confidentiality is not more generally used, we need look at the issues below in order to understand how to address the problem going forward.¶
The cost is such that its use has a noticeable effect on performance, given that the design (by requiring different keys for different users) makes it difficult or impossible for the work of encryption to be offloaded.¶
It is likely that, given increasing network speeds, this factor is more important today than it was when this approach was settled upon.¶
Looking at the relevant Security Considerations sections, we find the following:¶
All of the issues discussed above may have contributed to the lack of use of encryption with NFSv4, but the issues of poor performance due to the a non-offloadable nature of encryption appears to have had a critical role in creating an unsatisfactory situation for a protocol where high performance is often of critical importance.¶
Once the choice was made to limit encryption to specific file systems or directories, there was commonly a perceived need to avoid the cost of encryption and the difficulties that approach to the matter led to were exacerbated by the lack within the Security Considerations sections of information about the damage thus done. Because there was insufficient attention to these issues, effective action to address them was delayed, until the problem could no longer be ignored.¶
Given these weaknesses with regard to encryption, those needing to use NFSv4 outside local area networks often implemented NFSv4 over secure tunnels. As this work proceeded, it served as an inspiration for the work done in [I-D.ietf-nfsv4-rpc-tls] that provided the opportunity to provide encryption uniformly and potentially with high performance. An overview of steps necessary to address these issues appears in Section 7.1¶
This section addresses the fundamental issues of the organization and presentation of a new security framework. The complementary issues involved in making substantive improvements in NFSv4 security will be dealt with in Section 7.¶
Of prime importance is the inclusion of appropriate threat analyses, even apart from the requirement (in BCP72 [RFC3552]) that such analyses appear in Security Considerations sections. As we have already seen, without such an analysis, there is no way to determine whether any security changes adopted are adequate to address NFSv4's security difficulties. In the case of NFSv4, because of its complicated history and the conflict between the status of AUTH_SYS as specified by the NFSv4 specification documents ([RFC7530], [RFC8881]) and that in [RFC5531], there might be multiple analyses, located in different places:¶
The appropriate location of a threat analysis for AUTH_SYS as currently used depends on the question of whether to consider AUTH_SYS as part of RPC, as it formally is, or to transfer responsibility to NFSv4, since it was deprecated as insecure in [RFC5531], while at the same time embedded as a heavily used optional feature for all minor versions of NFSv4, including a new protocol, NFSv4.1 published as a Proposed Standard subsequently, in [RFC5661].¶
Also relevant to the question of the appropriate location for this threat analysis is the question of whether there might be a need for a revised treatment of AUTH_SYS including client host authentication. The possible existence of such a revised treatment, together with issues related to a threat analysis necessary for its inclusion is discussed in Section 8.4.4.¶
Another important issue concerns the potential need for adjustment of inappropriate use of terms defined by RFC2119 [RFC2119], particularly any suggestion that use of AUTH_SYS may be validly used without important security consequences, as suggested by the current understanding that its use is optional.¶
However, as desirable as such clarifications might be, it is necessary that we be realistic about what such changes can accomplish, even if the particular issues cited in Section 3 are properly attended to. Documents can be written to define certain implementations as non-conformant, but the ability of any such designation to affect implementer behavior is strongly affected by circumstances, particularly when the term is a hard one to pin down such as "SHOULD" or "SHOULD NOT".¶
The ability of such term choices to affect implementer behavior is at its maximum when a new protocol is defined. In this case, the specification document formalizes a contract between implementations, so that the consequence of not following these terms is a lack of interoperability rather the fact that one's implementation may be, formally speaking, non-conformant. In the case of a protocol such as NFSv4,intended to build and extend the work of earlier NFS versions, the ability of the IETF to affect implementer choices is significantly reduced, as interoperability with the existing protocol and the existing infrastructure to support it will be of greater importance to implementers. Further, it needs to be recognized that when these terms are used to constrain security-related behavior rather than interoperability per se, their effect on implementer choice is likely to be similarly reduced.¶
Although [RFC2119] provides definitions of "SHOULD" and "SHOULD NOT", it is difficult to determine the proper interpretation unless it is clear what the "valid reasons" that might supersede the recommendation or the expected consequences of doing so. For this reason, when these terms are used in proposed text, we will make such matters explicit, where they are not otherwise clear, as suggested by Section 7 of [RFC2119].¶
The work done to provide TLS support for RPC (in [I-D.ietf-nfsv4-rpc-tls]) presents an important opportunity to address the substantive security problems described in Sections 5.2 and 5.3. While there will be a need to specify appropriate policies for its use by NFSv4 as a ULP (see Sections 8.3.2 and 8.4.6 for details), the following opportunities present themselves and are likely to be taken advantage of:¶
Although [I-D.ietf-nfsv4-rpc-tls]) does not provide support for RDMA transports, parallel services to provide encryption and authentication are expected to be available, as discussed in Section 7.3.¶
[I-D.ietf-nfsv4-rpc-tls] provides the ability to encrypt all traffic on the connection used between an NFSv4 client and server. This requires that that both the client and server provide support for RPC-over-TLS. Appropriate requirements/recommendations are discussed in Section 8.3.2.¶
Because this facility is based on an opportunistic use of TLS, there is a need for ULP-defined policies to deal with situations in which the server rejected the request for a TLS connection. These matters are discussed in Section 8.3.2 as well, as is the policy that servers are to adopt when dealing with a connection on which RPC-over-TLS is not requested.¶
Because this new form of encryption is being added to an existing protocol, all of these policies are more complicated than would be the case for a new protocol, since there might be a need to accommodate earlier implementations, which might not have had time to be enhanced, while providing the ability to take advantage of whatever confidentiality support that might be present for a given client and server.¶
When establishing a connection, as provided for by [I-D.ietf-nfsv4-rpc-tls], authentication material is provided by the client and could be used to authenticate the client host to the server.¶
By using this authentication material to authenticate the client host, it would be possible to correct some of the issues discussed in Section 5.2. This would involve clearly specifying how that material is to be used, in contrast to the situation for the existing AUTH_SYS, for which current documentation (in Appendix A of [RFC5531]) is quite limited. Any such respecification would need to be acceptable to those used to using AUTH_SYS while avoiding the security issues that make its further use, in the old form, inadvisable.¶
This could allow the following improvements:¶
The details of how the authentication material could be used are discussed Section 8.4.6.¶
As noted above, [I-D.ietf-nfsv4-rpc-tls]) is not supported on RDMA transports, making it necessary to provide appropriate support for encryption and client host authentication in other ways.¶
It should be noted that our need to normatively reference [I-D.ietf-nfsv4-rpcrdma-version-two] would affect the schedule of a standards-track document dealing with NFSv4 security. Given that the current milestone for requesting publication is December 2020, this not expected to pose an obstacle.¶
These sections discuss the issues that need to be addressed by new standards-track documents including both issues of the presentation/evaluation of NFSv4 (discussed in Section 5.1) and the substantive security weaknesses discussed in Sections 5.2 and 5.3. For each issue there is a section providing background followed by information suggesting how the issue would be addressed or issues that would need to be resolved before the issue could be addressed.¶
For reasons that remain unclear, none of the specification documents for the existing NFSv4 minor versions contain a threat analysis. As a result, for the reasons discussed in Section 5.1, we need to provide an appropriate threat analysis for all NFSv4 minor versions. Because of the high degree of overlap between pairs of NFSv4 minor versions, most of the analysis will be common to all versions. However, there are some areas where features in later minor versions have significant security implications:¶
In NFSv4.1 [RFC5661], SECINFO_NO_NAME was added, in large part because it would be necessary to allow the data server, when the pNFS file mapping type is used, to communicate to the client that either integrity or encryption would be required when access to file data was provided through the data server (as opposed to the metadata server).¶
The use of TLS-based encryption, which we expect to be recommended (see Section 8.3.2, may limit the need for this feature, but it will remain REQUIRED, when file-based pNFS is implemented.¶
NFSv4.1 [RFC5661] facilities were added to protect locking and session-related state from modification by other servers (i.e. SP4_MACHCRED and SP4_SSV).¶
These features have had very limited implementation work, so it would be desirable to provide alternative means to achieve the goal, as described in Section 8.5.¶
While the bulk of the threat analysis would be minor-version- independent, differences between minor version will have to be noted, as will differences that reflect uses of the facilities discussed in Section 7.¶
Beyond the lack of NFsv4 threat analyses, there is a further problem in that there is no threat analysis dealing with the AUTH_SYS case in the existing RPC specification, RFC5531 [RFC5531]. In this case, the gap is easier to understand since RPC deals with security on a per-flavor basis and declares AUTH_SYS as "insecure". It may well be that the authors, the working group and the IESG thought this was adequate, and it would have been if this had resulted in its not being used subsequently. As things happened however, AUTH_SYS continued to be used extensively, including for NFsv4, so that there still remains a gap in this regard.¶
The following facts are relevant to the continued use of AUTH_SYS despite its security problems:¶
It is expected that a threat analysis dealing with all NFSv4 minor will be dealt with in new standards-track document dealing with NFSv4 security. Because this document will address security for all minor versions, it will update [RFC7530], [RFC8881], and [RFC7862].¶
There will, in addition, be a need for a threat analysis dealing with AUTH_SYS, which might or might not appear in the NFSv4-wide security document for reasons explained in Section 6.1. The question of where this best be done is discussed in Section 8.4.4.¶
The basic framework for the extension of NFSv4 was established by [RFC8178]. Given that the anticipated standards-track document dealing with NFSv4 security will, in a sense, extend NFSv4, there might be some uncertainty about whether the changes anticipated are in line with that extension framework and whether there might be a need for that document to update [RFC8178].¶
The following questions need to be addressed:¶
We anticipate that the material would be in a section entitled, "NFSv4 Security Changes and the Existing NFSv4 Extension Model"¶
The following introductory material seems appropriate:¶
The following paragraphs discuss the core issues:¶
In order to limit the possibility of interoperability issues, the security changes are made in the form of recommendations, but, because there exist implementations built before these recommendations were in effect, clients and server have to be aware of the possibility that these recommendations might not be followed. As a result, the effect of these recommendations, made to limit security issues, is, from an interoperability point of view, similar to those that might arise from OPTIONAL features. With regard to the NFSv4 extension framework,¶
The use of the new RPC-level facilities does not raise any extension issues, since the existing specification do not limit the transport used, except to require reliable and in-order delivery. For this reason, these facilities can be used without updating the existing minor version specifications.¶
Use of these facilities does not raise additional interoperability issues since [I-D.ietf-nfsv4-rpc-tls], by using the opportunistic TLS model, provided for interoperability between those that do and do not support these extensions.¶
The policies to use these facilities, made the responsibility of the ULP by [I-D.ietf-nfsv4-rpc-tls] are significant additions to existing minor versions, requiring that the new standards-track document dealing with NFSv4 security update [RFC7530] and [RFC8881].¶
These changes do not raise interoperability issues since the policies apply to all uses of these transport-level facilities for NFSV4. That addresses the case in which client and server both provide support while other cases are dealt with as indicated in the previous bullet.¶
The new recommendations regarding use of existing features (e.g. encryption, AUTH_SYS) often contradict existing guidance and so also require that new standards-track document dealing with NFSv4 security update [RFC7530] and [RFC8881].¶
These change do potentially raise interoperability issues which need to be dealt with by giving sufficient scope, within the framework of new recommendations, to accommodate existing behavior. As a result, from an interoperability standpoint, these changes are compatible with the previous designation of OPTIONAL use, even though the security consequences make it necessary that the use of these previously OPTIONAL facilities (e.g. transferring data in the clear, use of AUTH_SYS without client host authentication) be something to be warned against.¶
As discussed in Section 5.3, NFSv4 as currently defined, has serious problems providing the appropriate level of confidentiality for its transmissions. Providing encryption at the transport layer, whether as described in Section 7.1 or otherwise, can be expected to resolve this issue by providing encryption in a potentially offloadable way, making use of RPC-level privacy and integrity services unnecessary.¶
If we were defining a new protocol, such transport-level encryption could be REQUIRED without difficulty. However, given that we are upgrading the security for an existing protocol, it might not be possible to designate this encryption as "REQUIRED", since this might give rise to extensive interoperability problems between new and old implementations. In any case, existing implementers of the existing protocol would not have the same interest in specification compliance as in the corresponding situation with a new protocol, leading to a situation in which one group of implementations essentially ignored the issue of specification-compliance with regard to the latest protocol specifications.¶
We need to be aware that,¶
As we consider how to discuss the need for encryption in Section 8.3.2, we will be mindful of the above. Since it not possible to REQUIRE encryption, we will RECOMMEND it. However, in discussing the valid reasons that encryption might not be provided, we will try to restrict them to the degree we can reasonably do so. Further, we will try to clearly state the security consequences which implementers and those configuring clients and server need to be aware of when choosing to not provide the RECOMMENDED encryption.¶
There is a need to recommend the implementation and use of appropriate encryption by the server and client, as is done, for example, by the following paragraphs.¶
The lack of encryption for the connection as a whole can be addressed by encryption of individual requests using the facilities within RPCSEC_GSS specified in [RFC7530] and [RFC8881] as providing "privacy". When doing so, the following complicating issues need to be taken account of:¶
Although, as indicated in [I-D.ietf-nfsv4-rpc-tls], that there should be no need to negotiate security flavors to be used to provide integrity and confidentiality, the exact effect on existing clients and servers needs to be made clearer.¶
The text below is a suggested way of doing this.¶
When TLS is used to provide encryption, as specified in [I-D.ietf-nfsv4-rpc-tls], the client and server will, as indicated by that document, be aware that encryption, is present. In the case of RPC-over-RDMA, when the RNICs provide encryption transparently, the server and client have the opportunity to each make their own judgment about the adequacy of the encryption provided and communicate this to their peer as a transport property as provided for by [I-D.ietf-nfsv4-rpcrdma-version-two]). In either case, the client and server will only act, as described below, on knowledge about the existence of encryption shared by both parties:¶
The server treats all requests as being made requesting encryption, even if a different pseudo-flavor is used. As a result, the server will never return, NFS4ERR_WRONGSEC because of a quality-of-protection issue, despite a possible mismatch between the pseudo-flavor specified in SECINFO response and the one actually used.¶
The server will still return NFS4ERR_WRONGSEC when an inappropriate security flavor or oid is used and clients need to be prepared to use SECINFO to determine the security to be used.¶
During the development of NFSv2 [RFC1094], little attention was directed to network security and no attention was directed to the possibility that a machine on the network might represent itself as a client, with the ability to make requests on behalf of non-existent client processes it identified by user id. During this time, NFS clients were multi-user machines and it made sense to many for the client kernel to have the same responsibilities to verify user credentials and keep track of process user ids as it had successfully been doing with local file systems. This was the origin of AUTH_UNIX (later renamed AUTH_SYS) which was based on a model of co-operating UNIX kernels and provided no protection against an attacker with kernel privileges. Despite the later name change, actual implementations were strongly focused on UNIX, deriving the necessary implementation requirements from shared code and from informal inter-developer communications, rather than standards documents. As a result, as discussed in Section 5.2, when it was necessary to discuss the security of AUTH_SYS in connection with RPC (in [RFC5531], there was insufficient information on which to base a threat analysis, although it was correctly concluded that the result of use was a lack of security.¶
With regard to the following aspects of AUTH_SYS, this UNIX orientation was significant. In many cases as the computing environment changed, it was not possible to change the details to keep up with evolving needs.¶
Later, with the development of NFSv3 [RFC1813], some of these assumptions started to seem less justifiable and alternate authentication models were developed. However, because AUTH_SYS was so suited to the Unix environment, making NFS administration so similar to local file system administration, its use remained predominant despite its obvious weaknesses from the security standpoint.¶
When NFSv4 was first defined (in [RFC3530]), RPCSEC_GSS was added as a MANDATORY to implement means of authentication which was optional to use, presumably under the assumption that other optional-to-implement authentication flavors, such as AUTH_SYS would be used by some clients¶
The continued use of AUTH_SYS is troubling and hard to explain if one is focused on network security, but the following possibilities are worthy of note:¶
Regardless of why this happened, the important fact is that it did happen, despite the serious security problems that it gave rise to. Later subsections of Section 8.4 will discuss ways to address the resulting situation.¶
Given the serious security weaknesses described in Section 5.2, there is clearly a need to discourage its further use in its existing form, since the difficulties with it will not be eliminated by the adoption of encryption of NFSv4 connections alone¶
There are two basic forms that such an effort might take:¶
While the author is strongly of the opinion that the second choice is more likely to be successful, and this document reflects that view, there is no intention to foreclose the first option, should the working group choose it. If that option is chosen, then the material in Section 8.4.6 becomes irrelevant and can be ignored while much of that in Section 8.4.5 can be simply modified to expand its scope, as described below in that section.¶
As described in Section 8.1 there is a need for an appropriate threat analysis for NFSv4. For a number of reasons, issues with AUTH_SYS that would make it difficult to refer to in an NFSv4-wide security document without some additional work preparatory work:¶
Under other circumstances, it might have been possible to dispense with further analysis of AUTH_SYS because of the following facts:¶
Despite the above, we will need a more complete of treatment of AUTH_SYS security for a number of reasons.¶
Given that the ongoing security issues with AUTH_SYS might give rise to efforts to address them, an accurate understanding of these issues is even more important, regardless of how the working group chooses to resolve the issue discussed in Section 8.4.3>.¶
If the decision is made to improve the security characteristics of AUTH_SYS, it is important to understand existing security issues and how they might be best addressed.¶
If the decision is made to eliminate the use of AUTH_SYS, an adequate understanding of the security issues that made that decision necessary would seem to be required. In that context, the current statement in [RFC5531], regarding the insecurity of AUTH_SYS needs to be clarified.¶
An important question is where such a treatment should appear. In deciding this question, the important question is whether, at this point, it is best addressed as a feature associated with NFSv4 or with RPC.¶
The continued use of AUTH_SYS by all NFSv4 minor versions might well be considered to justify a transfer of responsibility to NFSv4, with the authentication flavor being mentioned, in RPC specifications, only as a historical artifact.¶
The potential continued use of AUTH_SYS with client-host authentication strengthens the NFSv4 connection. Since [I-D.ietf-nfsv4-rpc-tls] assigns the specification of policies related to authentication to the ULP, it seems that AUTH_SYS needs to be discussed by NFSv4 documents in a manner more substantial than simply listed as another authentication flavor.¶
AUTH_SYS, both with and without client host authentication could be discussed in its own document, including an appropriate threat analysis.¶
That document could be referenced by a new document updating or obsoleting [RFC5531]. This would provide an opportunity to address existing text which was essentially ignored by NFSv4.¶
That document could also be referenced by a new standards-track document dealing with NFSv4 security. The threat analysis would provide a basis for the whatever approach the working group decides to take with regard to the discussion of AUTH_SYS with client-host authentication. See Section 8.4.6 for further details.¶
The last of these is most likely to be adopted regardless of how the working group decides the issue described in Section 8.4.3. However, if the working group decides to strongly discourage the use of AUTH_SYS even with client-host authentication, the first is a possibility that should be considered.¶
This section presents material that might be included in a standards-track document in an effort to suitably discourage use of AUTH_SYS in its current (insecure) form.¶
While it is sometimes the case that such a task can be accomplished by substituting one of the terms defined in [RFC2119] by another, that is not the case here While the designation of AUTH_SYS as optional (to implement) with [RFC7530] and [RFC8881] using the language "MAY be implemented", is wrong and probably needs to be fixed, the situation is more complex than generally recognized. It should be kept in mind that these terms are, in many cases, ignored or followed while ignoring the underlying intent. As a result, such changes, without supporting explanations might not be effective in changing implementer behavior.¶
To summarize the confusing situation with AUTH_SYS and NFS and the role of the RPC specification, [RFC5531].¶
While there may well be sufficient loopholes within [RFC5531] to avoid an outright contradiction of the rules established there, it is certainly troubling that the security implications of implementing and using AUTH_SYS are not properly recommended against.¶
While we may need to adjust the normative language, the effect of doing so is limited for reasons that have already been discussed. As a result, we need to put substantial emphasis on clearly stating the consequences of doing what "SHOULD NOT" (or "should not") be done, as is suggested by [RFC2119], at least for the former case.¶
The following text is one way to present the situation:¶
Although, as a result of these previous approaches to AUTH_SYS without client host authentication, it might be used extensively despite the security issues that cause us to specify that it "SHOULD NOT" be used are sufficient to justify this shift. However, due to the difficulty of making such a shift, the need to maintain continuity of support is a valid reason to continue to use it in this way, as long as this decision is made with awareness of the security consequences:¶
The availability of client-host authentication makes possible a revised approach to AUTH_SYS addressing some of its security vulnerabilities. A revised approach to AUTH_SYS might provide:¶
The ability to prevent, using client host-authentication, another machine masquerading as the expected client.¶
The authentication features described in [I-D.ietf-nfsv4-rpc-tls], would be assure that you could identify trusted client machines, eliminating a major security issue with AUTH_SYS. On the other hand, it is not clear whether servers would be wise to trust clients to this degree, especially if the items below (#2 and #3) are not satisfactorily dealt with.¶
It would be possible to include, as part of the client authentication material, information known only to the client and the server. The secret value would be chosen by the client, and saved so it is usable by later client instances. If this were done, it would prevent any process with root privileges on a trusted machine from impersonating the trusted client. If this were not done, then any compromise of a single trusted client would compromise any server which trusted it to present AUTH_SYS requests.¶
To make this approach viable, servers would have to maintain, in persistent storage, a record of the secret value assigned to each authenticated client. When an authenticated client connected with a different value it would indicate that a root process on the client-host was attempting to impersonate the NFSv4 client.¶
The power granted to the authenticated client-host, to make a request on behalf of user or group has troubling consequences. Although most implementations have the ability to exclude requests from the root user, that may not, as has been noted previously, be sufficient to satisfactorily deal with the problem, since many deployments contain security-critical files that can be accessed and modified but users other than root.¶
It would be desirable to exclude a wider range of users such as all users within a given group (or alternatively include only users within a given group). However, neither of these is viable since servers receiving AUTH_SYS requests typically have no knowledge of what users belong to what groups, relying on the AUTH_SYS client to provide this information within the request.¶
Given this lack of knowledge on the part of the servers, any restrictions on the ability of AUTH_SYS requests to access or modify files designate a set of protected files in some way. Since designating a list of files is complicated and hard to maintain, it might be possible to describe the files whose access and/or modification by means of an ACL. When a request is made to access a file, that request is suppressed (i.e. treated as a request by "nobody") iff it would be allowed to access by the designated ACL and similarly in the case of modifying requests.¶
How to address the question of the potential use of AUTH_SYS with client-host authentication would depend on the working group's judgement about how likely such an arrangement would be likely to be in providing security when AUTH_SYS is used. This in turn would depend on the working group's judgment as to whether solutions for #2 and #3 could be defined and effectively deployed, as well as the results of the threat analysis¶
I am assuming that AUTH_SYS with client-host authentication needs to be considered inferior to RPCSEC_GSS but sufficiently superior to AUTH_SYS without client authentication, to make it desirable to encourage its use, given that it is impossible to eliminate use of AUTH_SYS, even with a "MUST NOT". A number of potential introductions to the topic, of varying tone are explored below:¶
If the working group wants to focus on the improvement relative to AUTH_SYS without client-host authentication, the following paragraphs might be an appropriate introduction:¶
When AUTH_SYS is used with client-host authentication, then the server, by allowing use of AUTH_SYS for that specific client, is essentially delegating the user authentication task to that client who is trusted by the server to perform that task locally, just as typically done when authenticating users for the purpose of local file access, even though the scope of damage that could be enabled by any false authentication is likely to be much wider in this case.¶
By accepting AUTH_SYS requests from a particular client, a server is relying on the security of the client for the security of all data stored on the server. This is in contrast to RPCSC_GSS, in which authentication is provided for each request issued. Because of this, the use of AUTH_SYS should only be allowed from client hosts that can reasonably be trusted.¶
If the working group wants to adopt a more cautionary tone regarding use of AUTH_SYS with client-host authentication, the following paragraphs might be an appropriate introduction:¶
By accepting AUTH_SYS requests from a particular client, a server is relying on the security of the client for the security of all data stored on the server. This is in contrast to RPCSC_GSS, in which authentication is provided for each request issued. Because of this, the use of AUTH_SYS should only be allowed from client hosts that can reasonably be trusted and whose internal security is sufficiently robust that the additional risk to server data can be considered reasonable.¶
If the working group wants to focus more on encouraging use of RPCSEC_GSS, then on improving the security of those choosing to use AUTH_SYS, a paragraph like the following might be appropriate.¶
While the use of AUTH_SYS with client-host authentication eliminates some of the major security issues that arise from use of AUTH_SYS, it is still inferior, from a security point of view, to RPCSEC_GSS. As a result, its use should be limited to situations in which, use of RPCSEC_GSS is not a practical possibility.¶
The following paragraph might be added if item #2 is not successfully addressed:¶
The following paragraph might be added if item #2 is successfully addressed:¶
The following paragraph might be added if item #3 is not successfully addressed:¶
The following paragraph might be added if item #3 is successfully addressed:¶
The following closing paragraphs are proposed for an eventual introduction to AUTH_SYS, as revised. In these paragraphs, material to be added only if a particular item above is addressed or not addressed will appear in brackets with the prefix "if-#x:" or "if-not-#x:".¶
The inclusion of state management within NFSv4 created new security issues in which the objects requiring protection were not files associated with particular users but locks, opens, and delegations, associated with particular clients.¶
In NFSv4.0, some protection could be provided when RPCSEC_GSS was used by limiting modification of state objects to users having access to the associated file. Nevertheless, the fundamental issue remained unresolved.¶
When sessions were introduced into NFSv4 (in NFSv4.1) there were additional vulnerabilities that needed to be protected against. There needed to be protection to prevent a session established by one client being bound to a connection established by another client, leading to the following possibilities:¶
Facilities were defined in [RFC5661] to allow such client-based protection to be implemented: SP4_MACHCRED and SP4_SSV but implementation has been quite limited. Client-host authentication makes it possible to apply such protection more generally, as described below.¶
While transport-level encryption would make it harder for attackers to mount attacks based on these vulnerabilities, the authentication material provided when a TLS connection is established could enable the vulnerabilities discussed above to be eliminated without the work of implementing SP4_MACHCRED or SP4_SSV, and irrespective of whether the client is using AUTH_SYS or RPCSEC_GSS.¶
If one has a TLS connection including suitable authentication material, then the server is in a position to reject attempts to bind to sessions established under this connection by other connections that are not TLS connections or are not established by the same client, unless SP4_MACHCRED or SP4_SSV is in effect. This would make state protection available to those using SP4_NONE, i.e. the vast majority, whether they used RPCSEC_GSS or AUTH_SYS with client-host authentication.¶
The fact that this is an NFSv4.1-only security feature make it difficult to fully it address solely in the NFSv4-wide security document which we writing. The following plan is intended to allow the security document to be published before proceeding to publish the anticipated rfc5661bis:¶
It should be noted that this use of client-host authentication provides applicable protection even when AUTH_SYS is used. This is so however the issues discussed in Section 8.4.6 are addressed, and independent of the decision made regarding the choice discussed in Section 8.4.3 unless it is decided that AUTH_SYS MUST NOT be used. Issues regarding the impersonation of a large set of users are not relevant because user ids are not referenced. Furthermore, the possibility of a privileged user-level client implementing a denial-of-service attach (if issue #2 in Section 8.4.6 is not addressed), is not of concern since such a privileged process would find it far easier to deny service directly on the client host.¶
This document is expected to facilitate concrete discussion of how to address security in NFSv4. It is expected to be modified to reflect working group comments. If the working group is amenable to this approach, at some point a request will be made to adopt it as a working group item.¶
A large part of the discussion is expected to concern how AUTH_SYS will be dealt with. However, it is not expected that the working group needs to reach consensus on this issue immediately. This document could be adopted as a working group item if it is an appropriate vehicle to organize the discussion of related issues, such as formulating a consensus regarding the treatment of AUTH_SYS.¶
It is anticipated that, once the document is adopted as a working group item, this document will be used to discuss the writing of multiple standards-track documents, which may need co-ordination since they address a set of related issues:¶
This document is being written to guide the working group in preparing future standards-track documents. There are no plans to publish it as an RFC. If a milestone is to be associated with it, the appropriate goal would be completion of WGLC.¶
The current document does not require any actions by IANA.¶
The convention of requiring a Security Considerations Section within an I-D encounters difficulties when applied to a document dealing solely with security, making it most unclear what such a section might contain for such documents.¶
In addition, the nature of this particular document poses further issues regarding the possible content of a Security Consideration Section:¶
In light of the above, this section will summarize the security-related message of earlier sections and will, in subsections, discuss the appropriate contents of Security Considerations Sections for a eventual standards-track documents to be written¶
The message of this document with regard to security issues can be summarized as follows¶
Although it is clear that such a document would need to include a threat analysis, as provided for by RFC3552/BCP72 [RFC3552]. However, the length of such an analysis might well be so large that it cannot be reasonably contained within a Security Considerations Section. In that case, the Security Considerations Section will summarize the results of the threat analysis, referring to those parts that appear in other sections of the document.¶
The general form of this section in [RFC5531] in which references are made to documents for each security flavor, seem appropriate. However, the following changes are likely to be required:¶
The treatment of AUTH_SYS needs to reference the new document describing AUTH_SYS, rather than the incomplete description in Appendix A, which needs to be removed.¶
In addition, the last two sentences of the initial paragraph of section 14 need to be reworked to respond to the fact that while, there was no clear violation of them, this insecure authentication flavor was used by NFSv4 for over a decade subsequently. Although the following attempt at loophole removal might not survive the editing and review process, the issues it raises are worth considering.¶
Given that the final paragraph of the Security Considerations section was drafted without consideration of reverse direction operation and without consideration the possibility that either TLs encryption or client host authentication might be available (and potentially REQUIRED), redrafting it to be more appropriate for the environment made possible by [I-D.ietf-nfsv4-rpc-tls] is likely.¶
The following replacement text is suggested to deal with these issues:¶
As mentioned above, a prime goal of the threat analysis for AUTH_SYS would be to clarify the degree to which the security weaknesses of AUTH_SYS can be successfully addressed using transport-level security facilities. Those pieces of the threat analysis might well appear outside the Security Consideration section proper.¶
The Security Considerations needs to summarize these and provide a basis for ULPs that allow AUTH_SYS to specify the appropriate conditions for use and the consequences of not enforcing these.¶
The authors would like to thank all the people contributed to rpc-tls RFC-to-be for their important work providing security at the RPC layer, enabling us to break the NFSv4 security logjam.¶
The authors would like to thank Tom Haynes (of Hammerspace) for introducing the idea of dealing with security for all minor versions in a single document.¶