Network File System Version 4 C. Lever, Ed. Internet-Draft Oracle Intended status: Standards Track D. Noveck Expires: May 20, 2020 NetApp November 17, 2019 RPC-over-RDMA Version 2 Protocol draft-ietf-nfsv4-rpcrdma-version-two-00 Abstract This document specifies the second version of a protocol that conveys Remote Procedure Call (RPC) messages on transports capable of Remote Direct Memory Access (RDMA). This version of the protocol is extensible. Status of This Memo 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 May 20, 2020. Copyright Notice Copyright (c) 2019 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. Lever & Noveck Expires May 20, 2020 [Page 1] Internet-Draft RDMA Transport for RPC V2 November 2019 This document may contain material from IETF Documents or IETF Contributions published or made publicly available before November 10, 2008. The person(s) controlling the copyright in some of this material may not have granted the IETF Trust the right to allow modifications of such material outside the IETF Standards Process. Without obtaining an adequate license from the person(s) controlling the copyright in such materials, this document may not be modified outside the IETF Standards Process, and derivative works of it may not be created outside the IETF Standards Process, except to format it for publication as an RFC or to translate it into languages other than English. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4 2. Requirements Language . . . . . . . . . . . . . . . . . . . . 5 3. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 6 3.1. Remote Procedure Calls . . . . . . . . . . . . . . . . . 6 3.1.1. Upper-Layer Protocols . . . . . . . . . . . . . . . . 6 3.1.2. Requesters and Responders . . . . . . . . . . . . . . 6 3.1.3. RPC Transports . . . . . . . . . . . . . . . . . . . 7 3.1.4. External Data Representation . . . . . . . . . . . . 8 3.2. Remote Direct Memory Access . . . . . . . . . . . . . . . 9 3.2.1. Direct Data Placement . . . . . . . . . . . . . . . . 9 3.2.2. RDMA Transport Requirements . . . . . . . . . . . . . 9 4. RPC-over-RDMA Protocol Framework . . . . . . . . . . . . . . 11 4.1. Transfer Model . . . . . . . . . . . . . . . . . . . . . 11 4.2. Message Framing . . . . . . . . . . . . . . . . . . . . . 11 4.3. Managing Receiver Resources . . . . . . . . . . . . . . . 12 4.3.1. RPC-over-RDMA Version 2 Flow Control . . . . . . . . 12 4.3.2. Inline Threshold . . . . . . . . . . . . . . . . . . 14 4.3.3. Initial Connection State . . . . . . . . . . . . . . 14 4.4. XDR Encoding with Chunks . . . . . . . . . . . . . . . . 15 4.4.1. Reducing an XDR Stream . . . . . . . . . . . . . . . 15 4.4.2. DDP-Eligibility . . . . . . . . . . . . . . . . . . . 16 4.4.3. RDMA Segments . . . . . . . . . . . . . . . . . . . . 16 4.4.4. Chunks . . . . . . . . . . . . . . . . . . . . . . . 17 4.4.5. Read Chunks . . . . . . . . . . . . . . . . . . . . . 18 4.4.6. Write Chunks . . . . . . . . . . . . . . . . . . . . 19 4.5. Message Transfer Methods . . . . . . . . . . . . . . . . 20 4.5.1. Short Messages . . . . . . . . . . . . . . . . . . . 21 4.5.2. Continued Messages . . . . . . . . . . . . . . . . . 21 4.5.3. Chunked Messages . . . . . . . . . . . . . . . . . . 22 4.5.4. Long Messages . . . . . . . . . . . . . . . . . . . . 23 5. Transport Properties . . . . . . . . . . . . . . . . . . . . 24 5.1. Transport Properties Model . . . . . . . . . . . . . . . 25 5.2. Current Transport Properties . . . . . . . . . . . . . . 26 5.2.1. Maximum Send Size . . . . . . . . . . . . . . . . . . 27 Lever & Noveck Expires May 20, 2020 [Page 2] Internet-Draft RDMA Transport for RPC V2 November 2019 5.2.2. Receive Buffer Size . . . . . . . . . . . . . . . . . 28 5.2.3. Maximum RDMA Segment Size . . . . . . . . . . . . . . 28 5.2.4. Maximum RDMA Segment Count . . . . . . . . . . . . . 28 5.2.5. Reverse Request Support . . . . . . . . . . . . . . . 29 5.2.6. Host Authentication Message . . . . . . . . . . . . . 30 6. RPC-over-RDMA Version 2 Transport Messages . . . . . . . . . 30 6.1. Overall Transport Message Structure . . . . . . . . . . . 30 6.2. Transport Header Types . . . . . . . . . . . . . . . . . 30 6.3. RPC-over-RDMA Version 2 Headers and Chunks . . . . . . . 31 6.3.1. Common Transport Header Prefix . . . . . . . . . . . 31 6.3.2. RPC-over-RDMA Version 2 Transport Header Prefix . . . 32 6.3.3. Describing External Data Payloads . . . . . . . . . . 35 6.4. Header Types Defined in RPC-over-RDMA version 2 . . . . . 36 6.4.1. RDMA2_MSG: Convey RPC Message Inline . . . . . . . . 36 6.4.2. RDMA2_NOMSG: Convey External RPC Message . . . . . . 37 6.4.3. RDMA2_ERROR: Report Transport Error . . . . . . . . . 38 6.4.4. RDMA2_CONNPROP: Advertise Transport Properties . . . 41 6.5. Choosing a Reply Mechanism . . . . . . . . . . . . . . . 42 7. XDR Protocol Definition . . . . . . . . . . . . . . . . . . . 42 7.1. Code Component License . . . . . . . . . . . . . . . . . 43 7.2. Extraction and Use of XDR Definitions . . . . . . . . . . 45 7.3. XDR Definition for RPC-over-RDMA Version 2 Core Structures . . . . . . . . . . . . . . . . . . . . . . . 47 7.4. XDR Definition for RPC-over-RDMA Version 2 Base Header Types . . . . . . . . . . . . . . . . . . . . . . . . . . 49 7.5. Use of the XDR Description Files . . . . . . . . . . . . 50 8. RPC Bind Parameters . . . . . . . . . . . . . . . . . . . . . 52 9. Implementation Status . . . . . . . . . . . . . . . . . . . . 53 10. Security Considerations . . . . . . . . . . . . . . . . . . . 54 10.1. Memory Protection . . . . . . . . . . . . . . . . . . . 54 10.1.1. Protection Domains . . . . . . . . . . . . . . . . . 54 10.1.2. Handle (STag) Predictability . . . . . . . . . . . . 54 10.1.3. Memory Protection . . . . . . . . . . . . . . . . . 54 10.1.4. Denial of Service . . . . . . . . . . . . . . . . . 55 10.2. RPC Message Security . . . . . . . . . . . . . . . . . . 55 10.2.1. RPC-over-RDMA Protection at Lower Layers . . . . . . 56 10.2.2. RPCSEC_GSS on RPC-over-RDMA Transports . . . . . . . 56 10.3. Transport Properties . . . . . . . . . . . . . . . . . . 58 10.4. Host Authentication . . . . . . . . . . . . . . . . . . 59 11. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 59 12. References . . . . . . . . . . . . . . . . . . . . . . . . . 59 12.1. Normative References . . . . . . . . . . . . . . . . . . 59 12.2. Informative References . . . . . . . . . . . . . . . . . 61 Appendix A. ULB Specifications . . . . . . . . . . . . . . . . . 63 A.1. DDP-Eligibility . . . . . . . . . . . . . . . . . . . . . 63 A.2. Maximum Reply Size . . . . . . . . . . . . . . . . . . . 64 A.3. Additional Considerations . . . . . . . . . . . . . . . . 65 A.4. ULP Extensions . . . . . . . . . . . . . . . . . . . . . 65 Lever & Noveck Expires May 20, 2020 [Page 3] Internet-Draft RDMA Transport for RPC V2 November 2019 Appendix B. Extending the Version 2 Protocol . . . . . . . . . . 65 B.1. Adding New Header Types to RPC-over-RDMA Version 2 . . . 67 B.2. Adding New Header Flags to the Protocol . . . . . . . . . 68 B.3. Adding New Transport properties to the Protocol . . . . . 69 B.4. Adding New Error Codes to the Protocol . . . . . . . . . 70 Appendix C. Differences from the RPC-over-RDMA Version 1 Protocol . . . . . . . . . . . . . . . . . . . . . . 70 C.1. Relationship to the RPC-over-RDMA Version 1 XDR Definition . . . . . . . . . . . . . . . . . . . . . . . 70 C.2. Transport Properties . . . . . . . . . . . . . . . . . . 72 C.3. Credit Management Changes . . . . . . . . . . . . . . . . 72 C.4. Inline Threshold Changes . . . . . . . . . . . . . . . . 73 C.5. Message Continuation Changes . . . . . . . . . . . . . . 74 C.6. Host Authentication Changes . . . . . . . . . . . . . . . 75 C.7. Support for Remote Invalidation . . . . . . . . . . . . . 75 C.8. Error Reporting Changes . . . . . . . . . . . . . . . . . 76 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 76 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 77 1. Introduction Remote Direct Memory Access (RDMA) [RFC5040] [RFC5041] [IBA] is a technique for moving data efficiently between network nodes. By directing data into destination buffers as it is sent on a network and placing it using direct memory access implemented by hardware, the complementary benefits of faster transfers and reduced host overhead are obtained. Open Network Computing Remote Procedure Call (ONC RPC, often shortened in NFSv4 documents to RPC) [RFC5531] is a Remote Procedure Call protocol that runs over a variety of transports. Most RPC implementations today use UDP [RFC0768] or TCP [RFC0793]. On UDP, RPC messages are encapsulated inside datagrams, while on a TCP byte stream, RPC messages are delineated by a record marking protocol. An RDMA transport also conveys RPC messages in a specific fashion that must be fully described if RPC implementations are to interoperate when using RDMA to transport RPC transactions. RDMA transports present semantics that differ from either UDP or TCP. They retain message delineations like UDP but provide reliable and sequenced data transfer like TCP. They also provide an offloaded bulk transfer service not provided by UDP or TCP. RDMA transports are therefore appropriately treated as a new transport type by RPC. Although the RDMA transport described herein can provide relatively transparent support for any RPC application, this document also describes mechanisms that enable further optimization of data transfer, when RPC applications are structured to exploit awareness Lever & Noveck Expires May 20, 2020 [Page 4] Internet-Draft RDMA Transport for RPC V2 November 2019 of a transport's RDMA capability. In this context, the Network File System (NFS) protocols, as described in [RFC1094], [RFC1813], [RFC7530], [RFC5661], and subsequent NFSv4 minor versions, are all potential beneficiaries of RDMA transports. A complete problem statement is presented in [RFC5532]. The RPC-over-RDMA version 1 protocol specified in [RFC8166] is deployed and in use, although there are known shortcomings to this protocol: o The protocol's default size of Receive buffers forces the use of RDMA Read and Write transfers for small payloads, and limits the size of reverse direction messages. o It is difficult to make optimizations or protocol fixes that require changes to on-the-wire behavior. o For some RPC procedures, the maximum reply size is difficult or impossible for an RPC client to estimate in advance. To address these issues in a way that enables interoperation with existing RPC-over-RDMA version 1 deployments, a second version of the RPC-over-RDMA transport protocol is presented in this document. Version 2 of RPC-over-RDMA is extensible, enabling OPTIONAL extensions to be added without impacting existing implementations. To enable protocol extension, the XDR definition for RPC-over-RDMA version 2 is organized differently than the definition version 1. These changes, which are discussed in Appendix C.1, do not alter the on-the-wire format. In addition, RPC-over-RDMA version 2 contains a set of incremental changes that relieve certain performance constraints and enable recovery from abnormal corner cases. These changes are outlined in Appendix C and include a larger default inline threshold, the ability to convey a single RPC message using multiple RDMA Send operations, support for authentication of connection peers, richer error reporting, an improved credit-based flow control mechanism, and support for Remote Invalidation. 2. Requirements Language The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here. Lever & Noveck Expires May 20, 2020 [Page 5] Internet-Draft RDMA Transport for RPC V2 November 2019 3. Terminology 3.1. Remote Procedure Calls This section highlights key elements of the RPC protocol [RFC5531] and the External Data Representation (XDR) [RFC4506] used by it. RPC-over-RDMA version 2 enables the transmission of RPC messges built using XDR and also uses XDR internaly to describe its own header formats. An understanding of RPC and its use of XDR is assumed in this document. 3.1.1. Upper-Layer Protocols RPCs are an abstraction used to implement the operations of an Upper- Layer Protocol (ULP). "ULP" refers to an RPC Program and Version tuple, which is a versioned set of procedure calls that comprise a single well-defined API. One example of a ULP is the Network File System Version 4.0 [RFC7530]. In this document, the term "RPC consumer" refers to an implementation of a ULP running on an RPC client. 3.1.2. Requesters and Responders Like a local procedure call, every RPC procedure has a set of "arguments" and a set of "results". A calling context invokes a procedure, passing arguments to it, and the procedure subsequently returns a set of results. Unlike a local procedure call, the called procedure is executed remotely rather than in the local application's execution context. The RPC protocol as described in [RFC5531] is fundamentally a message-passing protocol between one or more clients, where RPC consumers are running, and a server, where a remote execution context is available to process RPC transactions on behalf of those consumers. ONC RPC transactions are made up of two types of messages: CALL A CALL message, or "Call", requests that work be done. An RPC Call message is designated by the value zero (0) in the message's msg_type field. An arbitrary unique value is placed in the message's XID field in order to match this RPC Call message to a corresponding RPC Reply message. REPLY Lever & Noveck Expires May 20, 2020 [Page 6] Internet-Draft RDMA Transport for RPC V2 November 2019 A REPLY message, or "Reply", reports the results of work requested by an RPC Call message. An RPC Reply message is designated by the value one (1) in the message's msg_type field. The value contained in an RPC Reply message's XID field is copied from the RPC Call message whose results are being reported. Each RPC client endpoint acts as a "Requester". It serializes the procedure's arguments and conveys them to a server endpoint via an RPC Call message. This message contains an RPC protocol header, a header describing the requested upper-layer operation, and all arguments. An RPC server endpoint acts as a "Responder". It deserializes the arguments and processes the requested operation. It then serializes the operation's results into another byte stream. This byte stream is conveyed back to the Requester via an RPC Reply message. This message contains an RPC protocol header, a header describing the upper-layer reply, and all results. The Requester deserializes the results and allows the RPC consumer to proceed. At this point, the RPC transaction designated by the XID in the RPC Call message is complete, and the XID is retired. In summary, Requesters send RPC Call messages to Responders to initiate RPC transactions. Responders send RPC Reply messages to Requesters to complete the processing on an RPC transaction. 3.1.3. RPC Transports The role of an "RPC transport" is to mediate the exchange of RPC messages between Requesters and Responders. An RPC transport bridges the gap between the RPC message abstraction and the native operations of a particular network transport. RPC-over-RDMA is a connection-oriented RPC transport. When a connection-oriented transport is used, clients initiate transport connections, while servers wait passively to accept incoming connection requests. Most commonly, the client end of the connection acts in the role of Requester, and the server end of the connection acts as a Responder. However, RPC transactions can also be sent in the reverse direction. In this case, the server end of the connection acts as a Requestor while the client end acts as a Responder. Lever & Noveck Expires May 20, 2020 [Page 7] Internet-Draft RDMA Transport for RPC V2 November 2019 3.1.4. External Data Representation One cannot assume that all Requesters and Responders represent data objects the same way internally. RPC uses External Data Representation (XDR) to translate native data types and serialize arguments and results [RFC4506]. The XDR protocol encodes data independently of the endianness or size of host-native data types, enabling unambiguous decoding of data by the receiver. RPC Programs are specified by writing an XDR definition of their procedures, argument data types, and result data types. XDR assumes only that the number of bits in a byte (octet) and their order are the same on both endpoints and on the physical network. The smallest indivisible unit of XDR encoding is a group of four octets. XDR can also flatten lists, arrays, and other complex data types so they can be conveyed as a stream of bytes. A serialized stream of bytes that is the result of XDR encoding is referred to as an "XDR stream". A sending endpoint encodes native data into an XDR stream and then transmits that stream to a receiver. A receiving endpoint decodes incoming XDR byte streams into its native data representation format. 3.1.4.1. XDR Opaque Data Sometimes, a data item is to be transferred as is: without encoding or decoding. The contents of such a data item are referred to as "opaque data". XDR encoding places the content of opaque data items directly into an XDR stream without altering it in any way. ULPs or applications perform any needed data translation in this case. Examples of opaque data items include the content of files or generic byte strings. 3.1.4.2. XDR Roundup The number of octets in a variable-length data item precedes that item in an XDR stream. If the size of an encoded data item is not a multiple of four octets, octets containing zero are added after the end of the item. This is the case so that the next encoded data item in the XDR stream always starts on a four-octet boundary. The encoded size of the item is not changed by the addition of the extra octets. These extra octets are never exposed to ULPs. This technique is referred to as "XDR roundup", and the extra octets are referred to as "XDR roundup padding". Lever & Noveck Expires May 20, 2020 [Page 8] Internet-Draft RDMA Transport for RPC V2 November 2019 3.2. Remote Direct Memory Access RPC Requesters and Responders can be made more efficient if large RPC messages are transferred by a third party, such as intelligent network-interface hardware (data movement offload), and placed in the receiver's memory so that no additional adjustment of data alignment has to be made (direct data placement or "DDP"). RDMA transports enable both optimizations. In the current document, "RDMA" refers to the physical mechanism an RDMA transport utilizes when moving data. 3.2.1. Direct Data Placement Typically, RPC implementations copy the contents of RPC messages into a buffer before being sent. An efficient RPC implementation sends bulk data without copying it into a separate send buffer first. However, socket-based RPC implementations are often unable to receive data directly into its final place in memory. Receivers often need to copy incoming data to finish an RPC operation: sometimes, only to adjust data alignment. Although it may not be efficient, before an RDMA transfer, a sender may copy data into an intermediate buffer. After an RDMA transfer, a receiver may copy that data again to its final destination. In this document, the term "DDP" refers to any optimized data transfer where it is unnecessary for a receiving host's CPU to copy transferred data to another location after it has been received. RPC-over-RDMA version 2 enables the use of RDMA Read and Write operations to achieve both data movement offload and DDP. However, not all RDMA-based data transfer qualifies as DDP, and DDP can be achieved using non-RDMA mechanisms. 3.2.2. RDMA Transport Requirements To achieve good performance during receive operations, RDMA transports require that RDMA consumers provision resources in advance in order to receive incoming messages. An RDMA consumer might provide Receive buffers in advance by posting an RDMA Receive Work Request for every expected RDMA Send from a remote peer. These buffers are provided before the remote peer posts RDMA Send Work Requests. Thus this is often referred to as "pre- posting" buffers. Lever & Noveck Expires May 20, 2020 [Page 9] Internet-Draft RDMA Transport for RPC V2 November 2019 An RDMA Receive Work Request remains outstanding until hardware matches it to an inbound Send operation. The resources associated with that Receive must be retained in host memory, or "pinned", until the Receive completes. Given these basic tenets of RDMA transport operation, the RPC-over- RDMA version 2 protocol assumes each transport provides the following abstract operations. A more complete discussion of these operations can be found in [RFC5040]. 3.2.2.1. Memory Registration Memory registration assigns a steering tag to a region of memory, permitting the RDMA provider to perform data-transfer operations. The RPC-over-RDMA version 2 protocol assumes that each registered memory region is identified with a steering tag of no more than 32 bits and memory addresses of up to 64 bits in length. 3.2.2.2. RDMA Send The RDMA provider supports an RDMA Send operation, with completion signaled on the receiving peer after data has been placed in a pre- posted buffer. Sends complete at the receiver in the order they were issued at the sender. The amount of data transferred by a single RDMA Send operation is limited by the size of the remote peer's pre- posted buffers. 3.2.2.3. RDMA Receive The RDMA provider supports an RDMA Receive operation to receive data conveyed by incoming RDMA Send operations. To reduce the amount of memory that must remain pinned awaiting incoming Sends, the amount of pre-posted memory is limited. Flow control to prevent overrunning receiver resources is provided by the RDMA consumer (in this case, the RPC-over-RDMA version 2 protocol). 3.2.2.4. RDMA Write The RDMA provider supports an RDMA Write operation to place data directly into a remote memory region. The local host initiates an RDMA Write, and completion is signaled there. No completion is signaled on the remote peer. The local host provides a steering tag, memory address, and the length of the remote peer's memory region. RDMA Writes are not ordered with respect to one another, but are ordered with respect to RDMA Sends. A subsequent RDMA Send completion obtained at the write initiator guarantees that prior RDMA Write data has been successfully placed in the remote peer's memory. Lever & Noveck Expires May 20, 2020 [Page 10] Internet-Draft RDMA Transport for RPC V2 November 2019 3.2.2.5. RDMA Read The RDMA provider supports an RDMA Read operation to place peer source data directly into the read initiator's memory. The local host initiates an RDMA Read, and completion is signaled there. No completion is signaled on the remote peer. The local host provides steering tags, memory addresses, and a length for the remote source and local destination memory region. The local host signals Read completion to the remote peer as part of a subsequent RDMA Send message. The remote peer can then invalidate steering tags and subsequently free associated source memory regions. 4. RPC-over-RDMA Protocol Framework 4.1. Transfer Model A "transfer model" designates which endpoint exposes its memory and which is responsible for initiating the transfer of data. To enable RDMA Read and Write operations, for example, an endpoint first exposes regions of its memory to a remote endpoint, which initiates these operations against the exposed memory. In RPC-over-RDMA version 2, Requesters expose their memory to the Responder, but the Responder does not expose its memory. The Responder pulls RPC arguments or whole RPC calls from each Requester. The Responder pushes RPC results or whole RPC replies to each Requester. 4.2. Message Framing Each RPC-over-RDMA version 2 message consists of at most two XDR streams: Transport Stream The "Transport stream" contains a header that describes and controls the transfer of the Payload stream in this RPC-over-RDMA message. Every RDMA Send message on an RPC-over-RDMA version 2 connection MUST begin with a Transport stream. RPC Payload Stream The "Payload stream" contains part or all of a single RPC message. The sender MAY divide an RPC message at any convenient boundary, but MUST send RPC message fragments in XDR stream order and MUST NOT interleave Payload streams from multiple RPC messages. The RPC-over-RDMA version 2 message carrying the final part of an RPC message is marked (see Section 6.3.2.2). Lever & Noveck Expires May 20, 2020 [Page 11] Internet-Draft RDMA Transport for RPC V2 November 2019 In its simplest form, an RPC-over-RDMA version 2 message conveying an RPC message payload consists of a Transport stream followed immediately by a Payload stream transmitted together via a single RDMA Send. RPC-over-RDMA framing replaces all other RPC framing (such as TCP record marking) when used atop an RPC-over-RDMA association, even when the underlying RDMA protocol may itself be layered atop a transport with a defined RPC framing (such as TCP). However, it is possible for RPC-over-RDMA to be dynamically enabled on a connection in the course of negotiating the use of RDMA via a ULP exchange. Because RPC framing delimits an entire RPC request or reply, the resulting shift in framing must occur between distinct RPC messages, and in concert with the underlying transport. 4.3. Managing Receiver Resources The longevity of an RDMA connection mandates that sending endpoints respect the resource limits of peer receivers. To ensure messages can be sent and received reliably, there are two operational parameters for each connection. It is critical to provide RDMA Send flow control for an RDMA connection. If any pre-posted Receive buffer on the connection is not large enough to accept an incoming RDMA Send, or if a pre-posted Receive buffer is not available to accept an incoming RDMA Send, the RDMA connection can be terminated. 4.3.1. RPC-over-RDMA Version 2 Flow Control Because RPC-over-RDMA requires reliable and in-order delivery of data payloads, RPC-over-RDMA transports MUST use the RDMA RC (Reliable Connected) Queue Pair (QP) type, which ensures in-transit data integrity and handles recovery from packet loss or misordering. However, RPC-over-RDMA transports provide their own flow control mechanism to prevent a sender from overwhelming receiver resources. RPC-over-RDMA transports employ an end-to-end credit-based flow control mechanism for this purpose [CBFC]. Credit-based flow control was chosen because it is relatively simple, provides robust operation in the face of bursty traffic, automated management of receive buffer allocation, and excellent buffer utilization. 4.3.1.1. Granting Credits An RPC-over-RDMA version 2 credit is the capability to receive one RPC-over-RDMA version 2 message. This enables RPC-over-RDMA version 2 to support asymmetrical operation, where a message in one direction Lever & Noveck Expires May 20, 2020 [Page 12] Internet-Draft RDMA Transport for RPC V2 November 2019 might be matched by zero, one, or multiple messages in the other direction. To achieve this, credits are assigned to each connection peer's posted Receive buffers. Each Requester has a set of Receive credits, and each Responder has a set of Receive credits. These credit values are managed independently of one another. Section 7 of [RFC8166] requires that the 32-bit field containing the credit grant is the third word in the transport header. To conform with that requirement, the two independent credit values are encoded into a single 32-bit field in the fixed portion of the transport header. After the field is XDR decoded, the receiver takes the low- order two bytes as the number of credits that are newly granted by the sender, and the high-order two bytes as the maximum number of credits that can be outstanding at the sender. In this approach, then, there are requester credits, sent in messages from the requester to the responder; and responder credits, sent in messages from the responder to the requester. A sender MUST NOT send RDMA messages in excess of the receiver's granted credit limit. If the granted value is exceeded, the RDMA layer may signal an error, possibly terminating the connection. The granted value MUST NOT be zero, since such a value would result in deadlock. The granted credit values MAY be adjusted to match the needs or policies in effect on either peer. For instance, a peer may reduce its granted credit value to accommodate the available resources in a Shared Receive Queue. Certain RDMA implementations may impose additional flow-control restrictions, such as limits on RDMA Read operations in progress at the Responder. Accommodation of such restrictions is considered the responsibility of each RPC-over-RDMA version 2 implementation. 4.3.1.2. Asynchronous Credit Grants A protocol convention is provided to enable one peer to refresh its credit grant to the other peer without sending a data payload. Messages of this type can also act as a keep-alive ping. See Section 6.4.2 for information about this convention. To prevent transport deadlock, receivers MUST always be in a position to receive one such credit grant update message, in addition to payload-bearing messages. One way a receiver can do this is to post one extra Receive more than the credit value it granted. Lever & Noveck Expires May 20, 2020 [Page 13] Internet-Draft RDMA Transport for RPC V2 November 2019 4.3.2. Inline Threshold An "inline threshold" value is the largest message size (in octets) that can be conveyed in one direction between peer implementations using RDMA Send and Receive operations. The inline threshold value is effectively the smaller of the largest number of bytes the sender can post via a single RDMA Send operation and the largest number of bytes the receiver can accept via a single RDMA Receive operation. Each connection has two inline threshold values: one for messages flowing from Requester-to-Responder, referred to as the "call inline threshold", and one for messages flowing from Responder-to-Requester, referred to as the "reply inline threshold". Inline threshold values can be advertised to peers via Transport Properties. Receiver implementations MUST support inline thresholds of 4096 bytes. In the absence of an exchange of Transport Properties, senders and receivers MUST assume both connection inline thresholds are 4096 bytes. 4.3.3. Initial Connection State When an RPC-over-RDMA version 2 client establishes a connection to a server, its first order of business is to determine the server's highest supported protocol version. Upon connection establishment a client MUST NOT send more than a single RPC-over-RDMA message at a time until it receives a valid non- error RPC-over-RDMA message from the server that grants client credits. The second word of each transport header is used to convey the transport protocol version. In the interest of simplicity, we refer to that word as rdma_vers even though in the RPC-over-RDMA version 2 XDR definition it is described as rdma_start.rdma_vers. First, the client sends a single valid RPC-over-RDMA message with the value two (2) in the rdma_vers field. Because the server might support only RPC-over-RDMA version 1, this initial message MUST NOT be larger than the version 1 default inline threshold of 1024 bytes. 4.3.3.1. Server Does Support RPC-over-RDMA Version 2 If the server does support RPC-over-RDMA version 2, it sends RPC- over-RDMA messages back to the client with the value two (2) in the rdma_vers field. Both peers may use the default inline threshold value for RPC-over-RDMA version 2 connections (4096 bytes). Lever & Noveck Expires May 20, 2020 [Page 14] Internet-Draft RDMA Transport for RPC V2 November 2019 4.3.3.2. Server Does Not Support RPC-over-RDMA Version 2 If the server does not support RPC-over-RDMA version 2, it MUST send an RPC-over-RDMA message to the client with the same XID, with RDMA2_ERROR in the rdma_start.rdma_htype field, and with the error code RDMA2_ERR_VERS. This message also reports a range of protocol versions that the server supports. To continue operation, the client selects a protocol version in the range of server-supported versions for subsequent messages on this connection. If the connection is lost immediately after an RDMA2_ERROR / RDMA2_ERR_VERS message is received, a client can avoid a possible version negotiation loop when re-establishing another connection by assuming that particular server does not support RPC-over-RDMA version 2. A client can assume the same situation (no server support for RPC-over-RDMA version 2) if the initial negotiation message is lost or dropped. Once the negotiation exchange is complete, both peers may use the default inline threshold value for the transport protocol version that has been selected. 4.3.3.3. Client Does Not Support RPC-over-RDMA Version 2 If the server supports the RPC-over-RDMA protocol version used in the first RPC-over-RDMA message received from a client, it MUST use that protocol version in all subsequent messages it sends on that connection. The client MUST NOT change the protocol version for the duration of the connection. 4.4. XDR Encoding with Chunks When a DDP capability is available, the transport places the contents of one or more XDR data items directly into the receiver's memory, separately from the transfer of other parts of the containing XDR stream. 4.4.1. Reducing an XDR Stream RPC-over-RDMA version 2 provides a mechanism for moving part of an RPC message via a data transfer distinct from an RDMA Send/Receive pair. The sender removes one or more XDR data items from the Payload stream. These items are conveyed via other mechanisms, such as one or more RDMA Read or Write operations. As the receiver decodes an incoming message, it skips over directly placed data items. The portion of an XDR stream that is split out and moved separately is referred to as a "chunk". In some contexts, data in an RPC-over- RDMA header that describes these split out regions of memory may also be referred to as a "chunk". Lever & Noveck Expires May 20, 2020 [Page 15] Internet-Draft RDMA Transport for RPC V2 November 2019 A Payload stream after chunks have been removed is referred to as a "reduced" Payload stream. Likewise, a data item that has been removed from a Payload stream to be transferred separately is referred to as a "reduced" data item. 4.4.2. DDP-Eligibility Not all XDR data items benefit from DDP. For example, small data items or data items that require XDR unmarshaling by the receiver do not benefit from DDP. In addition, it is impractical for receivers to prepare for every possible XDR data item in a protocol to be transferred in a chunk. To maintain practical interoperability on an RPC-over-RDMA transport, a determination must be made of which few XDR data items in each ULP are allowed to use DDP. This is done in additional specifications that describe how ULPs employ DDP. A "ULB specification" identifies which specific individual XDR data items in a ULP MAY be transferred via DDP. Such data items are referred to as "DDP-eligible". All other XDR data items MUST NOT be reduced. Detailed requirements for ULBs are provided in Appendix A. 4.4.3. RDMA Segments When encoding a Payload stream that contains a DDP-eligible data item, a sender may choose to reduce that data item. When it chooses to do so, the sender does not place the item into the Payload stream. Instead, the sender records in the RPC-over-RDMA Transport header the location and size of the memory region containing that data item. The Requester provides location information for DDP-eligible data items in both RPC Call and Reply messages. The Responder uses this information to retrieve arguments contained in the specified region of the Requester's memory or place results in that memory region. An "RDMA segment", or "plain segment", is an RPC-over-RDMA Transport header data object that contains the precise coordinates of a contiguous memory region that is to be conveyed separately from the Payload stream. Plain segments contain the following information: Handle Steering tag (STag) or R_key generated by registering this memory with the RDMA provider. Length Lever & Noveck Expires May 20, 2020 [Page 16] Internet-Draft RDMA Transport for RPC V2 November 2019 The length of the RDMA segment's memory region, in octets. An "empty segment" is an RDMA segment with the value zero (0) in its length field. Offset The offset or beginning memory address of the RDMA segment's memory region. See [RFC5040] for further discussion. 4.4.4. Chunks In RPC-over-RDMA version 2, a "chunk" refers to a portion of the Payload stream that is moved independently of the RPC-over-RDMA Transport header and Payload stream. Chunk data is removed from the sender's Payload stream, transferred via separate operations, and then reinserted into the receiver's Payload stream to form a complete RPC message. Each chunk is comprised of RDMA segments. Each RDMA segment represents a single contiguous piece of that chunk. A Requester MAY divide a chunk into RDMA segments using any boundaries that are convenient. The length of a chunk is exactly the sum of the lengths of the RDMA segments that comprise it. The RPC-over-RDMA version 2 transport protocol does not place a limit on chunk size. However, each ULP may cap the amount of data that can be transferred by a single RPC transaction. For example, NFS has "rsize" and "wsize", which restrict the payload size of NFS READ and WRITE operations. The Responder can use such limits to sanity check chunk sizes before using them in RDMA operations. 4.4.4.1. Counted Arrays If a chunk contains a counted array data type, the count of array elements MUST remain in the Payload stream, while the array elements MUST be moved to the chunk. For example, when encoding an opaque byte array as a chunk, the count of bytes stays in the Payload stream, while the bytes in the array are removed from the Payload stream and transferred within the chunk. Individual array elements appear in a chunk in their entirety. For example, when encoding an array of arrays as a chunk, the count of items in the enclosing array stays in the Payload stream, but each enclosed array, including its item count, is transferred as part of the chunk. Lever & Noveck Expires May 20, 2020 [Page 17] Internet-Draft RDMA Transport for RPC V2 November 2019 4.4.4.2. Optional-Data If a chunk contains an optional-data data type, the "is present" field MUST remain in the Payload stream, while the data, if present, MUST be moved to the chunk. 4.4.4.3. XDR Unions A union data type MUST NOT be made DDP-eligible, but one or more of its arms MAY be DDP-eligible, subject to the other requirements in this section. 4.4.4.4. Chunk Roundup Except in special cases (covered in Section 4.5.4), a chunk MUST contain exactly one XDR data item. This makes it straightforward to reduce variable-length data items without affecting the XDR alignment of data items in the Payload stream. When a variable-length XDR data item is reduced, the sender MUST remove XDR roundup padding for that data item from the Payload stream so that data items remaining in the Payload stream begin on four-byte alignment. 4.4.5. Read Chunks A "Read chunk" represents an XDR data item that is to be pulled from the Requester to the Responder. A Read chunk is a list of one or more RDMA read segments. Each RDMA read segment consists of a Position field followed by a plain segment. Position The byte offset in the unreduced Payload stream where the receiver reinserts the data item conveyed in a chunk. The Position value MUST be computed from the beginning of the unreduced Payload stream, which begins at Position zero. All RDMA read segments belonging to the same Read chunk have the same value in their Position field. While constructing an RPC Call message, a Requester registers memory regions that contain data to be transferred via RDMA Read operations. It advertises the coordinates of these regions in the RPC-over-RDMA Transport header of the RPC Call message. After receiving an RPC Call message sent via an RDMA Send operation, a Responder transfers the chunk data from the Requester using RDMA Read operations. The Responder reconstructs the transferred chunk data by concatenating the contents of each RDMA segment in list order Lever & Noveck Expires May 20, 2020 [Page 18] Internet-Draft RDMA Transport for RPC V2 November 2019 into the received Payload stream at the Position value recorded in that RDMA segment. Put another way, the Responder inserts the first RDMA segment in a Read chunk into the Payload stream at the byte offset indicated by its Position field. RDMA segments whose Position field value match this offset are concatenated afterwards, until there are no more RDMA segments at that Position value. The Position field in a read segment indicates where the containing Read chunk starts in the Payload stream. The value in this field MUST be a multiple of four. All segments in the same Read chunk share the same Position value, even if one or more of the RDMA segments have a non-four-byte-aligned length. 4.4.5.1. Decoding Read Chunks While decoding a received Payload stream, whenever the XDR offset in the Payload stream matches that of a Read chunk, the Responder initiates an RDMA Read to pull the chunk's data content into registered local memory. The Responder acknowledges its completion of use of Read chunk source buffers when it sends an RPC Reply message to the Requester. The Requester may then release Read chunks advertised in the request. 4.4.5.2. Read Chunk Roundup When reducing a variable-length argument data item, the Requester MUST NOT include the data item's XDR roundup padding in the chunk itself. The chunk's total length MUST be the same as the encoded length of the data item. 4.4.6. Write Chunks While constructing an RPC Call message, a Requester prepares memory regions in which to receive DDP-eligible result data items. A "Write chunk" represents an XDR data item that is to be pushed from a Responder to a Requester. It is made up of an array of zero or more plain segments. Write chunks are provisioned by a Requester long before the Responder has prepared the reply Payload stream. A Requester often does not know the actual length of the result data items to be returned, since the result does not yet exist. Thus, it MUST register Write chunks long enough to accommodate the maximum possible size of each returned data item. Lever & Noveck Expires May 20, 2020 [Page 19] Internet-Draft RDMA Transport for RPC V2 November 2019 In addition, the XDR position of DDP-eligible data items in the reply's Payload stream is not predictable when a Requester constructs an RPC Call message. Therefore, RDMA segments in a Write chunk do not have a Position field. For each Write chunk provided by a Requester, the Responder pushes one data item to the Requester, filling the chunk contiguously and in segment array order until that data item has been completely written to the Requester. The Responder MUST copy the segment count and all segments from the Requester-provided Write chunk into the RPC Reply message's Transport header. As it does so, the Responder updates each segment length field to reflect the actual amount of data that is being returned in that segment. The Responder then sends the RPC Reply message via an RDMA Send operation. An "empty Write chunk" is a Write chunk with a zero segment count. By definition, the length of an empty Write chunk is zero. An "unused Write chunk" has a non-zero segment count, but all of its segments are empty segments. 4.4.6.1. Decoding Write Chunks After receiving the RPC Reply message, the Requester reconstructs the transferred data by concatenating the contents of each segment in array order into the RPC Reply message's XDR stream at the known XDR position of the associated DDP-eligible result data item. 4.4.6.2. Write Chunk Roundup When provisioning a Write chunk for a variable-length result data item, the Requester MUST NOT include additional space for XDR roundup padding. A Responder MUST NOT write XDR roundup padding into a Write chunk, even if the result is shorter than the available space in the chunk. Therefore, when returning a single variable-length result data item, a returned Write chunk's total length MUST be the same as the encoded length of the result data item. 4.5. Message Transfer Methods A receiver of RDMA Send operations is required to have previously posted one or more adequately sized buffers. Memory savings are achieved on both Requesters and Responders by posting small Receive buffers. However, not all RPC messages are small. RPC-over-RDMA version 2 provides several mechanisms that enable RPC message payloads of any size to be conveyed efficiently. Lever & Noveck Expires May 20, 2020 [Page 20] Internet-Draft RDMA Transport for RPC V2 November 2019 4.5.1. Short Messages RPC message payloads are often smaller than typical inline thresholds. For example, an NFS version 3 GETATTR operation is only 56 octets: 20 octets of RPC header, a 32-octet file handle argument, and 4 octets for its length. The reply to this common request is about 100 octets. Since all RPC messages conveyed via RPC-over-RDMA version 2 require at least one RDMA Send operation, the most efficient way to send an RPC message that is smaller than the inline threshold is to append the Payload stream directly to the Transport stream. An RPC-over- RDMA header with a small RPC Call or Reply message immediately following is transferred using a single RDMA Send operation. No other operations are needed. An RPC-over-RDMA transaction using a Short Message: Requester Responder | RDMA Send (RDMA_MSG) | Call | ------------------------------> | | | | | Processing | | | RDMA Send (RDMA_MSG) | | <------------------------------ | Reply 4.5.2. Continued Messages If an RPC message is larger than the inline threshold, the sender can choose to split that message over multiple RPC-over-RDMA messages. The Payload stream of each RPC-over-RDMA message contains a part of the RPC message. The receiver reconstitutes the RPC message by concatenating the Payload streams of the sequence of RPC-over-RDMA messages together. Though the purpose of a Continued Message is to handle large RPC messages, senders MAY use a Continued Message at any time to convey an RPC message, and MAY split the RPC message payload on any convenient boundary. An RPC-over-RDMA transaction using a Continued Message: Lever & Noveck Expires May 20, 2020 [Page 21] Internet-Draft RDMA Transport for RPC V2 November 2019 Requester Responder | RDMA Send (RDMA_MSG) | Call | ------------------------------> | | RDMA Send (RDMA_MSG) | | ------------------------------> | | RDMA Send (RDMA_MSG) | | ------------------------------> | | | | | | | Processing | | | RDMA Send (RDMA_MSG) | | <------------------------------ | Reply 4.5.3. Chunked Messages If DDP-eligible data items are present in a Payload stream, a sender MAY reduce some or all of these items by removing them from the Payload stream. The sender then uses a separate mechanism to transfer the reduced data items. The Transport stream with the reduced Payload stream immediately following is then transferred using a single RDMA Send operation. After receiving the Transport and Payload streams of an RPC Call message accompanied by Read chunks, the Responder uses RDMA Read operations to move reduced data items in Read chunks. Before sending the Transport and Payload streams of an RPC Reply message containing Write chunks, the Responder uses RDMA Write operations to move reduced data items in Write and Reply chunks. An RPC-over-RDMA transaction with a Read chunk: Requester Responder | RDMA Send (RDMA_MSG) | Call | ------------------------------> | | RDMA Read | | <------------------------------ | | RDMA Response (arg data) | | ------------------------------> | | | | | Processing | | | RDMA Send (RDMA_MSG) | | <------------------------------ | Reply An RPC-over-RDMA transaction with a Write chunk: Lever & Noveck Expires May 20, 2020 [Page 22] Internet-Draft RDMA Transport for RPC V2 November 2019 Requester Responder | RDMA Send (RDMA_MSG) | Call | ------------------------------> | | | | | Processing | | | RDMA Write (result data) | | <------------------------------ | | RDMA Send (RDMA_MSG) | | <------------------------------ | Reply Chunking and Message Continuation can be combined. After reduction, the sender MAY split the reduced RPC message into multiple Payload streams and then send it via a Continued Message. 4.5.4. Long Messages When a Payload stream is larger than the receiver's inline threshold, the Payload stream is reduced by removing DDP-eligible data items and placing them in chunks to be moved separately. If there are no DDP- eligible data items in the Payload stream, or the Payload stream is still too large after it has been reduced, the sender uses either Message Continuation, or it can use RDMA Read or Write operations to convey the entire RPC message. The latter mechanism is referred to as a "Long Message". To transmit a Long Message, the sender conveys only the Transport stream with an RDMA Send operation. The Payload stream is not included in the Send buffer in this instance. Instead, the Requester provides chunks that the Responder uses to move the Payload stream. Long Call To send a Long Call message, the Requester provides a special Read chunk that contains the RPC Call message's Payload stream. Every RDMA read segment in this chunk MUST contain zero in its Position field. This type of chunk is known as a "Position Zero Read chunk". Long Reply To send a Long Reply, the Requester provides a single special Write chunk in advance, known as the "Reply chunk", that will contain the RPC Reply message's Payload stream. The Requester sizes the Reply chunk to accommodate the maximum expected reply size for that upper-layer operation. Though the purpose of a Long Message is to handle large RPC messages, Requesters MAY use a Long Message at any time to convey an RPC Call message. Lever & Noveck Expires May 20, 2020 [Page 23] Internet-Draft RDMA Transport for RPC V2 November 2019 A Responder chooses which form of reply to use based on the chunks provided by the Requester. If Write chunks were provided and the Responder has a DDP-eligible result, it first reduces the reply Payload stream. If a Reply chunk was provided and the reduced Payload stream is larger than the reply inline threshold, the Responder MUST use the Requester-provided Reply chunk for the reply. XDR data items may appear in these special chunks without regard to their DDP-eligibility. As these chunks contain a Payload stream, such chunks MUST include appropriate XDR roundup padding to maintain proper XDR alignment of their contents. An RPC-over-RDMA transaction using a Long Call: Requester Responder | RDMA Send (RDMA_NOMSG) | Call | ------------------------------> | | RDMA Read | | <------------------------------ | | RDMA Response (RPC call) | | ------------------------------> | | | | | Processing | | | RDMA Send (RDMA_MSG) | | <------------------------------ | Reply An RPC-over-RDMA transaction using a Long Reply: Requester Responder | RDMA Send (RDMA_MSG) | Call | ------------------------------> | | | | | Processing | | | RDMA Write (RPC reply) | | <------------------------------ | | RDMA Send (RDMA_NOMSG) | | <------------------------------ | Reply 5. Transport Properties RPC-over-RDMA version 2 provides a mechanism for connection endpoints to communicate information about implementation properties, enabling compatible endpoints to optimize data transfer. Initially only a small set of transport properties are defined and a single operation is provided to exchange transport properties (see Section 6.4.4). Lever & Noveck Expires May 20, 2020 [Page 24] Internet-Draft RDMA Transport for RPC V2 November 2019 Both the set of transport properties and the operations used to communicate may be extended. Within RPC-over-RDMA version 2, all such extensions are OPTIONAL. For information about existing transport properties, see Sections 5.1 through 5.2. For discussion of extensions to the set of transport properties, see Appendix B.3. 5.1. Transport Properties Model A basic set of receiver and sender properties is specified in this document. An extensible approach is used, allowing new properties to be defined in future Standards Track documents. Such properties are specified using: o A code point identifying the particular transport property being specified. o A nominally opaque array which contains within it the XDR encoding of the specific property indicated by the associated code point. The following XDR types are used by operations that deal with transport properties: typedef rpcrdma2_propid uint32; struct rpcrdma2_propval { rpcrdma2_propid rdma_which; opaque rdma_data<>; }; typedef rpcrdma2_propval rpcrdma2_propset<>; typedef uint32 rpcrdma2_propsubset<>; An rpcrdma2_propid specifies a particular transport property. In order to facilitate XDR extension of the set of properties by concatenating XDR definition files, specific properties are defined as const values rather than as elements in an enum. An rpcrdma2_propval specifies a value of a particular transport property with the particular property identified by rdma_which, while the associated value of that property is contained within rdma_data. Lever & Noveck Expires May 20, 2020 [Page 25] Internet-Draft RDMA Transport for RPC V2 November 2019 An rdma_data field which is of zero length is interpreted as indicating the default value or the property indicated by rdma_which. While rdma_data is defined as opaque within the XDR, the contents are interpreted (except when of length zero) using the XDR typedef associated with the property specified by rdma_which. As a result, when rpcrdma2_propval does not conform to that typedef, the receiver is REQUIRED to return the error RDMA2_ERR_BAD_XDR using the header type RDMA2_ERROR as described in Section 6.4.3. For example, the receiver of a message containing a valid rpcrdma2_propval returns this error if the length of rdma_data is such that it extends beyond the bounds of the message being transferred. In cases in which the rpcrdma2_propid specified by rdma_which is understood by the receiver, the receiver also MUST report the error RDMA2_ERR_BAD_XDR if either of the following occur: o The nominally opaque data within rdma_data is not valid when interpreted using the property-associated typedef. o The length of rdma_data is insufficient to contain the data represented by the property-associated typedef. Note that no error is to be reported if rdma_which is unknown to the receiver. In that case, that rpcrdma2_propval is not processed and processing continues using the next rpcrdma2_propval, if any. A rpcrdma2_propset specifies a set of transport properties. No particular ordering of the rpcrdma2_propval items within it is imposed. A rpcrdma2_propsubset identifies a subset of the properties in a previously specified rpcrdma2_propset. Each bit in the mask denotes a particular element in a previously specified rpcrdma2_propset. If a particular rpcrdma2_propval is at position N in the array, then bit number N mod 32 in word N div 32 specifies whether that particular rpcrdma2_propval is included in the defined subset. Words beyond the last one specified are treated as containing zero. 5.2. Current Transport Properties Although the set of transport properties may be extended, a basic set of transport properties is defined in Table 1. In that table, the columns contain the following information: o The column labeled "Property" identifies the transport property described by the current row. Lever & Noveck Expires May 20, 2020 [Page 26] Internet-Draft RDMA Transport for RPC V2 November 2019 o The column labeled "Code" specifies the rpcrdma2_propid value used to identify this property. o The column labeled "XDR type" gives the XDR type of the data used to communicate the value of this property. This data type overlays the data portion of the nominally opaque field rdma_data in a rpcrdma2_propval. o The column labeled "Default" gives the default value for the property which is to be assumed by those who do not receive, or are unable to interpret, information about the actual value of the property. o The column labeled "Sec" indicates the section within this document that explains the semantics and use of this transport property. +----------------------------+------+----------+---------+---------+ | Property | Code | XDR type | Default | Sec | +----------------------------+------+----------+---------+---------+ | Maximum Send Size | 1 | uint32 | 4096 | 5.2.1 | | Receive Buffer Size | 2 | uint32 | 4096 | 5.2.2 | | Maximum RDMA Segment Size | 3 | uint32 | 1048576 | 5.2.3 | | Maximum RDMA Segment Count | 4 | uint32 | 16 | 5.2.4 | | Reverse Request Support | 5 | uint32 | 1 | 5.2.5 | | Host Auth Message | 6 | opaque<> | N/A | 5.2.6 | +----------------------------+------+----------+---------+---------+ Table 1 5.2.1. Maximum Send Size The Maximum Send Size specifies the maximum size, in octets, of Send payloads. The endpoint sending this value ensures that it will not transmit a Send WR payload larger than this size, allowing the endpoint receiving this value to size its Receive buffers appropriately. const uint32 RDMA2_PROPID_SBSIZ = 1; typedef uint32 rpcrdma2_prop_sbsiz; Lever & Noveck Expires May 20, 2020 [Page 27] Internet-Draft RDMA Transport for RPC V2 November 2019 5.2.2. Receive Buffer Size The Receive Buffer Size specifies the minimum size, in octets, of pre-posted receive buffers. It is the responsibility of the endpoint sending this value to ensure that its pre-posted receive buffers are at least the size specified, allowing the endpoint receiving this value to send messages that are of this size. const uint32 RDMA2_PROPID_RBSIZ = 2; typedef uint32 rpcrdma2_prop_rbsiz; A sender may use his knowledge of the receiver's buffer size to determine when the message to be sent will fit in the preposted receive buffers that the receiver has set up. In particular, o Requesters may use the value to determine when it is necessary to provide a Position Zero Read chunk or Message Continuation when sending a request. o Requesters may use the value to determine when it is necessary to provide a Reply chunk when sending a request, based on the maximum possible size of the reply. o Responders may use the value to determine when it is necessary, given the actual size of the reply, to actually use a Reply chunk provided by the requester. 5.2.3. Maximum RDMA Segment Size The Maximum RDMA Segment Size specifies the maximum size, in octets, of an RDMA segment this endpoint is prepared to send or receive. const uint32 RDMA2_PROPID_RSSIZ = 3; typedef uint32 rpcrdma2_prop_rssiz; 5.2.4. Maximum RDMA Segment Count The Maximum RDMA Segment Count specifies the maximum number of RDMA segments that can appear in a requester's transport header. Lever & Noveck Expires May 20, 2020 [Page 28] Internet-Draft RDMA Transport for RPC V2 November 2019 const uint32 RDMA2_PROPID_RCSIZ = 4; typedef uint32 rpcrdma2_prop_rcsiz; 5.2.5. Reverse Request Support The value of this property is used to indicate a client implementation's readiness to accept and process messages that are part of reverse direction RPC requests. const uint32 RDMA_RVREQSUP_NONE = 0; const uint32 RDMA_RVREQSUP_INLINE = 1; const uint32 RDMA_RVREQSUP_GENL = 2; const uint32 RDMA2_PROPID_BRS = 5; typedef uint32 rpcrdma2_prop_brs; Multiple levels of support are distinguished: o The value RDMA2_RVREQSUP_NONE indicates that receipt of reverse direction requests and replies is not supported. o The value RDMA2_RVREQSUP_INLINE indicates that receipt of reverse direction requests or replies is only supported using inline messages and that use of explicit RDMA operations for reverse direction messages is not supported. o The value RDMA2_RVREQSUP_GENL that receipt of reverse direction requests or replies is supported in the same ways that forward direction requests or replies typically are. When information about this property is not provided, the support level of servers can be inferred from the reverse direction requests that they issue, assuming that issuing a request implicitly indicates support for receiving the corresponding reply. On this basis, support for receiving inline replies can be assumed when requests without Read chunks, Write chunks, or Reply chunks are issued, while requests with any of these elements allow the client to assume that general support for reverse direction replies is present on the server. Lever & Noveck Expires May 20, 2020 [Page 29] Internet-Draft RDMA Transport for RPC V2 November 2019 5.2.6. Host Authentication Message The value of this transport property is used as part of an exchange of host authentication material. This property can accommodate authentication handshakes that require multiple challenge-response interactions, and potentially large amounts of material. const uint32 RDMA2_PROPID_HOSTAUTH = 6; typedef opaque rpcrdma2_prop_hostauth<>; When this property is not provided, the peer(s) remain unauthenticated. Local security policy on each peer determines whether the connection is permitted to continue. 6. RPC-over-RDMA Version 2 Transport Messages 6.1. Overall Transport Message Structure Each transport message consists of multiple sections: o A transport header prefix, as defined in Section 6.3.2. Among other things, this structure indicates the header type. o The transport header proper, as defined by one of the sub-sections below. See Section 6.2 for the mapping between header types and the corresponding header structure. o Potentially, all or part of an RPC message payload being conveyed as an addendum to the transport header. This organization differs from that presented in the definition of RPC-over-RDMA version 1 [RFC8166], which presented the first and second of the items above as a single XDR item. The new organization is more in keeping with RPC-over-RDMA version 2's extensibility model in that new header types can be defined without modifying the existing set of header types. 6.2. Transport Header Types The new header types within RPC-over-RDMA version 2 are set forth in Table 2. In that table, the columns contain the following information: o The column labeled "Operation" specifies the particular operation. Lever & Noveck Expires May 20, 2020 [Page 30] Internet-Draft RDMA Transport for RPC V2 November 2019 o The column labeled "Code" specifies the value of header type for this operation. o The column labeled "XDR type" gives the XDR type of the data structure used to describe the information in this new message type. This data immediately follows the universal portion on the transport header present in every RPC-over-RDMA transport header. o The column labeled "Msg" indicates whether this operation is followed (or not) by an RPC message payload. o The column labeled "Sec" indicates the section (within this document) that explains the semantics and use of this operation. +-------------------------+------+-------------------+-----+--------+ | Operation | Code | XDR type | Msg | Sec | +-------------------------+------+-------------------+-----+--------+ | Convey Appended RPC | 0 | rpcrdma2_msg | Yes | 6.4.1 | | Message | | | | | | Convey External RPC | 1 | rpcrdma2_nomsg | No | 6.4.2 | | Message | | | | | | Report Transport Error | 4 | rpcrdma2_err | No | 6.4.3 | | Specify Properties at | 5 | rpcrdma2_connprop | No | 6.4.4 | | Connection | | | | | +-------------------------+------+-------------------+-----+--------+ Table 2 Suppport for the operations in Table 2 is REQUIRED. Support for additional operations will be OPTIONAL. RPC-over-RDMA version 2 implementations that receive an OPTIONAL operation that is not supported MUST respond with an RDMA2_ERROR message with an error code of RDMA2_ERR_INVAL_HTYPE. 6.3. RPC-over-RDMA Version 2 Headers and Chunks Most RPC-over-RDMA version 2 data structures are derived from corresponding structures in RPC-over-RDMA version 1. As is typical for new versions of an existing protocol, the XDR data structures have new names and there are a few small changes in content. In some cases, there have been structural re-organizations to enabled protocol extensibility. 6.3.1. Common Transport Header Prefix The rpcrdma_common prefix describes the first part of each RDMA-over- RPC transport header for version 2 and subsequent versions. Lever & Noveck Expires May 20, 2020 [Page 31] Internet-Draft RDMA Transport for RPC V2 November 2019 struct rpcrdma_common { uint32 rdma_xid; uint32 rdma_vers; uint32 rdma_credit; uint32 rdma_htype; }; RPC-over-RDMA version 2's use of these first four words matches that of version 1 as required by [RFC8166]. However, there are important structural differences in the way that these words are described by the respective XDR descriptions: o The header type is represented as a uint32 rather than as an enum that would need to be modified to reflect additions to the set of header types made by later extensions. o The header type field is part of an XDR structure devoted to representing the transport header prefix, rather than being part of a discriminated union, that includes the body of each transport header type. o There is now a prefix structure (see Section 6.3.2) of which the rpcrdma_common structure is the initial segment. This is a newly defined XDR object within the protocol description, in contrast with RPC-over-RDMA version 1, which limits the common portion of all header types to the four words in rpcrdma_common. These changes are part of a larger structural change in the XDR description of RPC-over-RDMA version 2 that enables a cleaner treatment of protocol extension. The XDR appearing in Section 7 reflects these changes, which are discussed in further detail in Appendix C.1. 6.3.2. RPC-over-RDMA Version 2 Transport Header Prefix The following prefix structure appears at the start of any RPC-over- RDMA version 2 transport header. Lever & Noveck Expires May 20, 2020 [Page 32] Internet-Draft RDMA Transport for RPC V2 November 2019 const RPCRDMA2_F_RESPONSE 0x00000001; const RPCRDMA2_F_MORE 0x00000002; struct rpcrdma2_hdr_prefix struct rpcrdma_common rdma_start; uint32 rdma_flags; }; The rdma_flags is new to RPC-over-RDMA version 2. Currently, the only flags defined within this word are the RPCRDMA2_F_RESPONSE flag and the RPCRDMA2_F_MORE flag. The other bits are reserved for future use as described in Appendix B.2. The sender MUST set these flags to zero. 6.3.2.1. RPCRDMA2_F_RESPONSE Flag The RPCRDMA2_F_RESPONSE flag qualifies the value contained in the transport header's rdma_start.rdma_xid field. The RPCRDMA2_F_RESPONSE flag enables a receiver to reliably avoid performing an XID lookup on incoming reverse direction Call messages. In general, when a message carries an XID that was generated by the message's receiver (that is, the receiver is acting as a requester), the message's sender sets the RPCRDMA2_F_RESPONSE flag. Otherwise that flag is clear. For example: o When the rdma_start.rdma_htype field has the value RDMA2_MSG or RDMA2_NOMSG, the value of the RPCRDMA2_F_RESPONSE flag MUST be the same as the value of the associated RPC message's msg_type field. o When the header type is anything else and a whole or partial RPC message payload is present, the value of the RPCRDMA2_F_RESPONSE flag MUST be the same as the value of the associated RPC message's msg_type field. o When no RPC message payload is present, a requester MUST set the value of RPCRDMA2_F_RESPONSE to reflect how the receiver is to interpret the rdma_start.rdma_xid field. o When the rdma_start.rdma_htype field has the value RDMA2_ERROR, the RPCRDMA2_F_RESPONSE flag MUST be set. Lever & Noveck Expires May 20, 2020 [Page 33] Internet-Draft RDMA Transport for RPC V2 November 2019 6.3.2.2. RPCRDMA2_F_MORE Flag The RPCRDMA2_F_MORE flag signifies that the RPC-over-RDMA message payload continues in the next message. This is referred to as Message Continuation, or Send chaining. When the RPCRDMA2_F_MORE flag is asserted, the receiver is to concatenate the data payload of the next received message to the end of the data payload of the current received message. The sender clears the RPCRDMA2_F_MORE flag in the final message in the sequence. All RPC-over-RDMA messages in such a sequence MUST have the same values in the rdma_start.rdma_xid and rdma_start.rdma_htype fields. If this constraint is not met, the receiver MUST respond with an RDMA2_ERROR message with the rdma_err field set to RDMA2_ERR_INVAL_FLAG. If a peer receives an RPC-over-RDMA message where the RPCRDMA2_F_MORE flag is set and the rdma_start.rdma_htype field does not contain RDMA2_MSG or RDMA2_CONNPROP, the receiver MUST respond with an RDMA2_ERROR message with the rdma_err field set to RDMA2_ERR_INVAL_FLAG. [ dnoveck: Both the above and your error in the existing third paragraph raise issues since they could be sent by a responder. Will need to fix RDMA2_ERROR so that this can be done when appropriate. ] When the RPCRDMA2_F_MORE flag is set in an individual message, that message's chunk lists MUST be empty. Chunks for a chained message may be conveyed in the final message in the sequence, whose RPCRDMA2_F_MORE flag is clear. There is no protocol-defined limit on the number of concatenated messages in a sequence. If the sender exhausts the receiver's credit grant before the final message is sent, the sender MUST wait for a further credit grant from the receiver before continuing to send messages. Credit exhaustion can occur at the receiver in the middle of a sequence of continued messages. To enable the sender to continue sending the remaining messages in the sequence, the receiver can grant more credits by sending an RPC message payload or an out-of- band credit grant (see Section 4.3.1.2). Lever & Noveck Expires May 20, 2020 [Page 34] Internet-Draft RDMA Transport for RPC V2 November 2019 6.3.3. Describing External Data Payloads The rpcrdma2_chunk_lists structure specifies how an RPC message is conveyed using explicit RDMA operations. struct rpcrdma2_chunk_lists { uint32 rdma_inv_handle; struct rpcrdma2_read_list *rdma_reads; struct rpcrdma2_write_list *rdma_writes; struct rpcrdma2_write_chunk *rdma_reply; }; For the most part this structure parallels its RPC-over-RDMA version 1 equivalent. That is, the rdma_reads, rdma_writes, rdma_reply fields provide, respectively, descriptions of the chunks used to read a Long message or directly placed data from the requester, to write directly placed response data into the requester's memory, and to write a long reply into the requester's memory. 6.3.3.1. Chunks and Chunk Lists The chunks and chunk list structures follow the same rules as in Section 3.4 of [RFC8166], with these exceptions: o In RPC-over-RDMA version 1, there were cases where XDR padding was allowed to appear in a reduced XDR data item. However, in RPC- over-RDMA version 2, requesters and responders MUST NOT include XDR padding in reduced Read and Write chunks, but chunks that make up Position Zero Read chunks and Reply chunks MUST include all XDR padding. o A responder MUST use Message Continuation if the requester does not provide a Reply chunk and the actual size of the reply is larger than the connection's inline threshold. A responder MAY use Message Continuation even if the requester has provided adequate Reply resources. This makes it unnecessary for RPC-over- RDMA version 2 requesters to have perfect reply size estimation. 6.3.3.2. Remote Invalidation An important addition relative to the corresponding RPC-over-RDMA version 1 rdma_header structures is the rdma_inv_handle field. This field supports remote invalidation of requester memory registrations via the RDMA Send With Invalidate operation. Lever & Noveck Expires May 20, 2020 [Page 35] Internet-Draft RDMA Transport for RPC V2 November 2019 To request Remote Invalidation, a requester sets the value of the rdma_inv_handle field in an RPC Call's transport header to a non-zero value that matches one of the rdma_handle fields in that header. If none of the rdma_handle values in the header conveying the Call may be invalidated by the responder, the requester sets the RPC Call's rdma_inv_handle field to the value zero. If the responder chooses not to use remote invalidation for this particular RPC Reply, or the RPC Call's rdma_inv_handle field contains the value zero, the responder uses RDMA Send to transmit the matching RPC reply. If a requester has provided a non-zero value in the RPC Call's rdma_inv_handle field and the responder chooses to use Remote Invalidation for the matching RPC Reply, the responder uses RDMA Send With Invalidate to transmit that RPC reply, and uses the value in the corresponding Call's rdma_inv_handle field to construct the Send With Invalidate Work Request. 6.4. Header Types Defined in RPC-over-RDMA version 2 The header types defined and used in RPC-over-RDMA version 1 are all carried over into RPC-over-RDMA version 2, although there may be limited changes in the definition of existing header types. In comparison with the header types of RPC-over-RDMA version 1, the changes can be summarized as follows: o To simplify interoperability with RPC-over-RDMA version 1, only the RDMA2_ERROR header (defined in Section 6.4.3) has an XDR definition that differs from that in RPC-over-RDMA version 1, and its modifications are all compatible extensions. o RDMA2_MSG and RDMA2_NOMSG (defined in Sections Section 6.4.1 and Section 6.4.2) have XDR definitions that match the corresponding RPC-over-RDMA version 1 header types. However, because of the changes to the header prefix, the version 1 and version 2 header types differ in on-the-wire format. o RDMA2_CONNPROP (defined in Section 6.4.4) is a completely new header type devoted to enabling connection peers to exchange information about their transport properties. 6.4.1. RDMA2_MSG: Convey RPC Message Inline RDMA2_MSG is used to convey an RPC message that immediately follows the Transport Header in the Send buffer. This is either an RPC Lever & Noveck Expires May 20, 2020 [Page 36] Internet-Draft RDMA Transport for RPC V2 November 2019 request that has no Position Zero Read chunk or an RPC reply that is not sent using a Reply chunk. const rpcrdma2_proc RDMA2_MSG = 0; struct rpcrdma2_msg { struct rpcrdma2_chunk_lists rdma_chunks; /* The rpc message starts here and continues * through the end of the transmission. */ uint32 rdma_rpc_first_word; }; 6.4.2. RDMA2_NOMSG: Convey External RPC Message RDMA2_NOMSG can convey an entire RPC message payload using explicit RDMA operations. When an RPC message payload is present, this message type is also known as a Long message. In particular, it is a Long call when the responder reads the RPC payload from a memory area specified by a Position Zero Read chunk; and it is a Long reply when the respond writes the RPC payload into a memory area specified by a Reply chunk. In both of these cases, the rdma_xid field is set to the same value as the xid of the RPC message payload. If all the chunk lists are empty (i.e., three 32-bit zeroes in the chunk list fields), the message conveys a credit grant refresh. The header prefix of this message contains a credit grant refresh in the rdma_credit field. In this case, the sender MUST set the rdma_xid field to zero. const rpcrdma2_proc RDMA2_NOMSG = 1; struct rpcrdma2_nomsg { struct rpcrdma2_chunk_lists rdma_chunks; }; In RPC-over-RDMA version 2, an alternative to using a Long message is to use Message Continuation. Lever & Noveck Expires May 20, 2020 [Page 37] Internet-Draft RDMA Transport for RPC V2 November 2019 6.4.3. RDMA2_ERROR: Report Transport Error RDMA2_ERROR provides a way of reporting the occurrence of transport errors on a previous transmission. This header type MUST NOT be transmitted by a requester. const rpcrdma2_proc RDMA2_ERROR = 4; struct rpcrdma2_err_vers { uint32 rdma_vers_low; uint32 rdma_vers_high; }; struct rpcrdma2_err_write { uint32 rdma_chunk_index; uint32 rdma_length_needed; }; union rpcrdma2_error switch (rpcrdma2_errcode rdma_err) { case RDMA2_ERR_VERS: rpcrdma2_err_vers rdma_vrange; case RDMA2_ERR_READ_CHUNKS: uint32 rdma_max_chunks; case RDMA2_ERR_WRITE_CHUNKS: uint32 rdma_max_chunks; case RDMA2_ERR_SEGMENTS: uint32 rdma_max_segments; case RDMA2_ERR_WRITE_RESOURCE: rpcrdma2_err_write rdma_writeres; case RDMA2_ERR_REPLY_RESOURCE: uint32 rdma_length_needed; default: void; }; Error reporting is addressed in RPC-over-RDMA version 2 in a fashion similar to RPC-over-RDMA version 1. Several new error codes, and error messages never flow from requester to responder. RPC-over-RDMA version 1 error reporting is described in Section 5 of [RFC8166]. Unless otherwise specified, in all cases below, the responder copies the values of the rdma_start.rdma_xid and rdma_start.rdma_vers fields from the incoming transport header that generated the error to transport header of the error response. The responder sets the Lever & Noveck Expires May 20, 2020 [Page 38] Internet-Draft RDMA Transport for RPC V2 November 2019 rdma_start.rdma_htype field of the transport header prefix to RDMA2_ERROR, and the rdma_start.rdma_credit field is set to the credit grant value for this connection. The receiver of this header type MUST ignore the value of the rdma_start.rdma_credit field. RDMA2_ERR_VERS This is the equivalent of ERR_VERS in RPC-over-RDMA version 1. The error code value, semantics, and utilization are the same. RDMA2_ERR_INVAL_HTYPE If a responder recognizes the value in the rdma_start.rdma_vers field, but it does not recognize the value in the rdma_start.rdma_htype field or does not support that header type, it MUST set the rdma_err field to RDMA2_ERR_INVAL_HTYPE. RDMA2_ERR_INVAL_FLAG If a receiver recognizes the value in the rdma_start.rdma_htype field but does not recognize the combination of flags in the rdma_flags field, it MUST set the rdma_err field to RDMA2_ERR_INVAL_HTYPE. RDMA2_ERR_BAD_XDR If a responder recognizes the values in the rdma_start.rdma_vers and rdma_start.rdma_proc fields, but the incoming RPC-over-RDMA transport header cannot be parsed, it MUST set the rdma_err field to RDMA2_ERR_BAD_XDR. This includes cases in which a nominally opaque property value field cannot be parsed using the XDR typedef associated with the transport property definition. The error code value of RDMA2_ERR_BAD_XDR is the same as the error code value of ERR_CHUNK in RPC-over-RDMA version 1. The responder MUST NOT process the request in any way except to send an error message. RDMA2_ERR_READ_CHUNKS If a requester presents more DDP-eligible arguments than the responder is prepared to Read, the responder MUST set the rdma_err field to RDMA2_ERR_READ_CHUNKS, and set the rdma_max_chunks field to the maximum number of Read chunks the responder can receive and process. If the responder implementation cannot handle any Read chunks for a request, it MUST set the rdma_max_chunks to zero in this response. The requester SHOULD resend the request using a Position Zero Read chunk. If this was a request using a Position Zero Read chunk, the requester MUST terminate the transaction with an error. RDMA2_ERR_WRITE_CHUNKS If a requester has constructed an RPC Call message with more DDP- eligible results than the server is prepared to Write, the Lever & Noveck Expires May 20, 2020 [Page 39] Internet-Draft RDMA Transport for RPC V2 November 2019 responder MUST set the rdma_err field to RDMA2_ERR_WRITE_CHUNKS, and set the rdma_max_chunks field to the maximum number of Write chunks the responder can process and return. If the responder implementation cannot handle any Write chunks for a request, it MUST return a response of RDMA2_ERR_REPLY_RESOURCE (below). The requester SHOULD resend the request with no Write chunks and a Reply chunk of appropriate size. RDMA2_ERR_SEGMENTS If a requester has constructed an RPC Call message with a chunk that contains more segments than the responder supports, the responder MUST set the rdma_err field to RDMA2_ERR_SEGMENTS, and set the rdma_max_segments field to the maximum number of segments the responder can process. RDMA2_ERR_WRITE_RESOURCE If a requester has provided a Write chunk that is not large enough to fully convey a DDP-eligible result, the responder MUST set the rdma_err field to RDMA2_ERR_WRITE_RESOURCE. The responder MUST set the rdma_chunk_index field to point to the first Write chunk in the transport header that is too short, or to zero to indicate that it was not possible to determine which chunk is too small. Indexing starts at one (1), which represents the first Write chunk. The responder MUST set the rdma_length_needed to the number of bytes needed in that chunk in order to convey the result data item. Upon receipt of this error code, a responder MAY choose to terminate the operation (for instance, if the responder set the index and length fields to zero), or it MAY send the request again using the same XID and more reply resources. RDMA2_ERR_REPLY_RESOURCE If an RPC Reply's Payload stream does not fit inline and the requester has not provided a large enough Reply chunk to convey the stream, the responder MUST set the rdma_err field to RDMA2_ERR_REPLY_RESOURCE. The responder MUST set the rdma_length_needed to the number of Reply chunk bytes needed to convey the reply. Upon receipt of this error code, a responder MAY choose to terminate the operation (for instance, if the responder set the index and length fields to zero), or it MAY send the request again using the same XID and larger reply resources. RDMA2_ERR_SYSTEM Lever & Noveck Expires May 20, 2020 [Page 40] Internet-Draft RDMA Transport for RPC V2 November 2019 If some problem occurs on a responder that does not fit into the above categories, the responder MAY report it to the sender by setting the rdma_err field to RDMA2_ERR_SYSTEM. This is a permanent error: a requester that receives this error MUST terminate the RPC transaction associated with the XID value in the rdma_start.rdma_xid field. 6.4.4. RDMA2_CONNPROP: Advertise Transport Properties The RDMA2_CONNPROP message type allows an RPC-over-RDMA endpoint, whether client or server, to indicate to its partner relevant transport properties that the partner might need to be aware of. The message definition for this operation is as follows: struct rpcrdma2_connprop { rpcrdma2_propset rdma_props; }; All relevant transport properties that the sender is aware of should be included in rdma_props. Since support of each of the properties is OPTIONAL, the sender cannot assume that the receiver will necessarily take note of these properties. The sender should be prepared for cases in which the receiver continues to assume that the default value for a particular property is still in effect. Generally, a participant will send a RDMA2_CONNPROP message as the first message after a connection is established. Given that fact, the sender should make sure that the message can be received by peers who use the default Receive Buffer Size. The connection's initial receive buffer size is typically 1KB, but it depends on the initial connection state of the RPC-over-RDMA version in use. Properties not included in rdma_props are to be treated by the peer endpoint as having the default value and are not allowed to change subsequently. The peer should not request changes in such properties. Those receiving an RDMA2_CONNPROP may encounter properties that they do not support or are unaware of. In such cases, these properties are simply ignored without any error response being generated. Lever & Noveck Expires May 20, 2020 [Page 41] Internet-Draft RDMA Transport for RPC V2 November 2019 6.5. Choosing a Reply Mechanism A requester provides any necessary registered memory resources for both an RPC Call message and its matching RPC Reply message. A requester forms each RPC Call itself, thus it can compute the exact memory resources needed to send every Call. However, the requester must allocate memory resources to receive the corresponding Reply before the responder has formed it. In some cases it is difficult for the requester to know in advance precisely what resources will be needed to receive the Reply. In RPC-over-RDMA version 2, a requester MAY provide a Reply chunk at any time. The responder MAY use the provided Reply chunk or decide to use another means to convey the RPC Reply. If the combination of the provided Write chunk list and Reply chunk is not adequate to convey a Reply, the responder SHOULD use Message Continuation (see Section 6.3.2.2 to send that Reply. If even that is not possible, the responder sends an RDMA2_ERROR message to the requester, as described in Section 6.4.3: o The responder MUST send a RDMA2_ERR_WRITE_RESOURCE error if the Write chunk list cannot accommodate the ULP's DDP-eligible data payload. o The responder MUST send a RDMA2_ERR_REPLY_RESOURCE error if the Reply chunk cannot accommodate the non DDP-eligible parts of the Reply. When receiving such errors, the requester SHOULD retry the ULP call using larger reply resources. In cases where retrying the ULP request is not possible, the requester terminates the RPC request and presents an error to the RPC consumer. 7. XDR Protocol Definition This section contains a description of the core features of the RPC- over-RDMA version 2 protocol expressed in the XDR language [RFC4506]. Because of the need to provide for protocol extensibility without modifying an existing XDR definition, this description has some important structural differences from the corresponding XDR description for RPC-over-RDMA version 1, which appears in [RFC8166]. This description is divided into three parts: o A code component license which appears in Section 7.1. Lever & Noveck Expires May 20, 2020 [Page 42] Internet-Draft RDMA Transport for RPC V2 November 2019 o An XDR description of the structures that are generally available for use by transport header types including both those defined in this document and those that may be defined as extensions. This includes definitions of the chunk-related structures derived from RPC-over-RDMA version 1, the transport property model introduced in this document, and a definition of the transport header prefixes that precede the various transport header types. This appears in Section 7.3. o An XDR description of the transport header types defined in this document, including those derived from RPC-over-RDMA version 1 and those introduced in RPC-over-RDMA version 2. This appears in Section 7.4. This description is provided in a way that makes it simple to extract into ready-to-compile form. To enable the combination of this description with the descriptions of subsequent extensions to RPC- over-RDMA version 2, the extracted description can be combined with similar descriptions published later, or those descriptions can be compiled separately. Refer to Section 7.2 for details. 7.1. Code Component License Code components extracted from this document must include the following license text. When the extracted XDR code is combined with other complementary XDR code which itself has an identical license, only a single copy of the license text need be preserved. Lever & Noveck Expires May 20, 2020 [Page 43] Internet-Draft RDMA Transport for RPC V2 November 2019 /// /* /// * Copyright (c) 2010-2018 IETF Trust and the persons /// * identified as authors of the code. All rights reserved. /// * /// * The authors of the code are: /// * B. Callaghan, T. Talpey, C. Lever, and D. Noveck. /// * /// * Redistribution and use in source and binary forms, with /// * or without modification, are permitted provided that the /// * following conditions are met: /// * /// * - Redistributions of source code must retain the above /// * copyright notice, this list of conditions and the /// * following disclaimer. /// * /// * - Redistributions in binary form must reproduce the above /// * copyright notice, this list of conditions and the /// * following disclaimer in the documentation and/or other /// * materials provided with the distribution. /// * /// * - Neither the name of Internet Society, IETF or IETF /// * Trust, nor the names of specific contributors, may be /// * used to endorse or promote products derived from this /// * software without specific prior written permission. /// * /// * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS /// * AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED /// * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE /// * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS /// * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO /// * EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE /// * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, /// * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT /// * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR /// * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS /// * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF /// * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, /// * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING /// * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF /// * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. /// */ /// Lever & Noveck Expires May 20, 2020 [Page 44] Internet-Draft RDMA Transport for RPC V2 November 2019 7.2. Extraction and Use of XDR Definitions The reader can apply the following sed script to this document to produce a machine-readable XDR description of the RPC-over-RDMA version 2 protocol without any OPTIONAL extensions. sed -n -e 's:^ */// ::p' -e 's:^ *///$::p' That is, if this document is in a file called "spec.txt" then the reader can do the following to extract an XDR description file and store it in the file rpcrdma-v2.x. sed -n -e 's:^ */// ::p' -e 's:^ *///$::p' \ < spec.txt > rpcrdma-v2.x Although this file is a usable description of the base protocol, when extensions are to supported, it may be desirable to divide into multiple files. The following script can be used for that purpose: Lever & Noveck Expires May 20, 2020 [Page 45] Internet-Draft RDMA Transport for RPC V2 November 2019 #!/usr/local/bin/perl open(IN,"rpcrdma-v2.x"); open(OUT,">temp.x"); while() { if (m/FILE ENDS: (.*)$/) { close(OUT); rename("temp.x", $1); open(OUT,">temp.x"); } else { print OUT $_; } } close(IN); close(OUT); Running the above script will result in two files: o The file common.x, containing the license plus the common XDR definitions which need to be made available to both the base operations and any subsequent extensions. o The file baseops.x containing the XDR definitions for the base operations, defined in this document. Optional extensions to RPC-over-RDMA version 2, published as Standards Track documents, will have similar means of providing XDR that describes those extensions. Once XDR for all desired extensions is also extracted, it can be appended to the XDR description file extracted from this document to produce a consolidated XDR description file reflecting all extensions selected for an RPC-over- RDMA implementation. Alternatively, the XDR descriptions can be compiled separately. In this case the combination of common.x and baseops.x serves to define the base transport, while using as XDR descriptions for extensions, the XDR from the document defining that extension, together with the file common.x, obtained from this document. Lever & Noveck Expires May 20, 2020 [Page 46] Internet-Draft RDMA Transport for RPC V2 November 2019 7.3. XDR Definition for RPC-over-RDMA Version 2 Core Structures /// /******************************************************************* /// * Transport Header Prefixes /// ******************************************************************/ /// /// struct rpcrdma_common { /// uint32 rdma_xid; /// uint32 rdma_vers; /// uint32 rdma_credit; /// uint32 rdma_htype; /// }; /// /// const RPCRDMA2_F_RESPONSE 0x00000001; /// const RPCRDMA2_F_MORE 0x00000002; /// /// struct rpcrdma2_hdr_prefix /// struct rpcrdma_common rdma_start; /// uint32 rdma_flags; /// }; /// /// /******************************************************************* /// * Chunks and Chunk Lists /// ******************************************************************/ /// /// struct rpcrdma2_segment { /// uint32 rdma_handle; /// uint32 rdma_length; /// uint64 rdma_offset; /// }; /// /// struct rpcrdma2_read_segment { /// uint32 rdma_position; /// struct rpcrdma2_segment rdma_target; /// }; /// /// struct rpcrdma2_read_list { /// struct rpcrdma2_read_segment rdma_entry; /// struct rpcrdma2_read_list *rdma_next; /// }; /// /// struct rpcrdma2_write_chunk { /// struct rpcrdma2_segment rdma_target<>; /// }; /// /// struct rpcrdma2_write_list { /// struct rpcrdma2_write_chunk rdma_entry; Lever & Noveck Expires May 20, 2020 [Page 47] Internet-Draft RDMA Transport for RPC V2 November 2019 /// struct rpcrdma2_write_list *rdma_next; /// }; /// /// struct rpcrdma2_chunk_lists { /// uint32 rdma_inv_handle; /// struct rpcrdma2_read_list *rdma_reads; /// struct rpcrdma2_write_list *rdma_writes; /// struct rpcrdma2_write_chunk *rdma_reply; /// }; /// /// /******************************************************************* /// * Transport Properties /// ******************************************************************/ /// /// /* /// * Types for transport properties model /// */ /// typedef rpcrdma2_propid uint32; /// /// struct rpcrdma2_propval { /// rpcrdma2_propid rdma_which; /// opaque rdma_data<>; /// }; /// /// typedef rpcrdma2_propval rpcrdma2_propset<>; /// typedef uint32 rpcrdma2_propsubset<>; /// /// /* /// * Transport propid values for basic properties /// */ /// const uint32 RDMA2_PROPID_SBSIZ = 1; /// const uint32 RDMA2_PROPID_RBSIZ = 2; /// const uint32 RDMA2_PROPID_RSSIZ = 3; /// const uint32 RDMA2_PROPID_RCSIZ = 4; /// const uint32 RDMA2_PROPID_BRS = 5; /// const uint32 RDMA2_PROPID_HOSTAUTH = 6; /// /// /* /// * Types specific to particular properties /// */ /// typedef uint32 rpcrdma2_prop_sbsiz; /// typedef uint32 rpcrdma2_prop_rbsiz; /// typedef uint32 rpcrdma2_prop_rssiz; /// typedef uint32 rpcrdma2_prop_rcsiz; /// typedef uint32 rpcrdma2_prop_brs; /// typedef opaque rpcrdma2_prop_hostauth<>; /// /// const uint32 RDMA_RVREQSUP_NONE = 0; Lever & Noveck Expires May 20, 2020 [Page 48] Internet-Draft RDMA Transport for RPC V2 November 2019 /// const uint32 RDMA_RVREQSUP_INLINE = 1; /// const uint32 RDMA_RVREQSUP_GENL = 2; /// /// /* FILE ENDS: common.x; */ 7.4. XDR Definition for RPC-over-RDMA Version 2 Base Header Types /// /******************************************************************* /// * Descriptions of RPC-over-RDMA Header Types /// ******************************************************************/ /// /// /* /// * Header Type Codes. /// */ /// const rpcrdma2_proc RDMA2_MSG = 0; /// const rpcrdma2_proc RDMA2_NOMSG = 1; /// const rpcrdma2_proc RDMA2_ERROR = 4; /// const rpcrdma2_proc RDMA2_CONNPROP = 5; /// /// /* /// * Header Types to Convey RPC Messages. /// */ /// struct rpcrdma2_msg { /// struct rpcrdma2_chunk_lists rdma_chunks; /// /// /* The rpc message starts here and continues /// * through the end of the transmission. */ /// uint32 rdma_rpc_first_word; /// }; /// /// struct rpcrdma2_nomsg { /// struct rpcrdma2_chunk_lists rdma_chunks; /// }; /// /// /* /// * Header Type to Report Errors. /// */ /// const uint32 RDMA2_ERR_VERS = 1; /// const uint32 RDMA2_ERR_BAD_XDR = 2; /// const uint32 RDMA2_ERR_INVAL_HTYPE = 3; /// const uint32 RDMA2_ERR_INVAL_FLAG = 4; /// const uint32 RDMA2_ERR_READ_CHUNKS = 5; /// const uint32 RDMA2_ERR_WRITE_CHUNKS = 6; /// const uint32 RDMA2_ERR_SEGMENTS = 7; /// const uint32 RDMA2_ERR_WRITE_RESOURCE = 8; Lever & Noveck Expires May 20, 2020 [Page 49] Internet-Draft RDMA Transport for RPC V2 November 2019 /// const uint32 RDMA2_ERR_REPLY_RESOURCE = 9; /// const uint32 RDMA2_ERR_SYSTEM = 10; /// /// struct rpcrdma2_err_vers { /// uint32 rdma_vers_low; /// uint32 rdma_vers_high; /// }; /// /// struct rpcrdma2_err_write { /// uint32 rdma_chunk_index; /// uint32 rdma_length_needed; /// }; /// /// union rpcrdma2_error switch (rpcrdma2_errcode rdma_err) { /// case RDMA2_ERR_VERS: /// rpcrdma2_err_vers rdma_vrange; /// case RDMA2_ERR_READ_CHUNKS: /// uint32 rdma_max_chunks; /// case RDMA2_ERR_WRITE_CHUNKS: /// uint32 rdma_max_chunks; /// case RDMA2_ERR_SEGMENTS: /// uint32 rdma_max_segments; /// case RDMA2_ERR_WRITE_RESOURCE: /// rpcrdma2_err_write rdma_writeres; /// case RDMA2_ERR_REPLY_RESOURCE: /// uint32 rdma_length_needed; /// default: /// void; /// }; /// /// /* /// * Header Type to Exchange Transport Properties. /// */ /// struct rpcrdma2_connprop { /// rpcrdma2_propset rdma_props; /// }; /// /// /* FILE ENDS: baseops.x; */ 7.5. Use of the XDR Description Files The three files common.x and baseops.x, when combined with the XDR descriptions for extension defined later, produce a human-readable and compilable description of the RPC-over-RDMA version 2 protocol with the included extensions. Lever & Noveck Expires May 20, 2020 [Page 50] Internet-Draft RDMA Transport for RPC V2 November 2019 Although this XDR description can be useful in generating code to encode and decode the transport and payload streams, there are elements of the structure of RPC-over-RDMA version 2 which are not expressible within the XDR language as currently defined. This requires implementations that use the output of the XDR processor to provide additional code to bridge the gaps. o The values of transport properties are represented within XDR as opaque values. However, the actual structures of each of the properties are represented by XDR typedefs, with the selection of the appropriate typedef described by text in this document. The determination of the appropriate typedef is not specified by XDR, which does not possess the facilities necessary for that determination to be specified in an extensible way. This is similar to the way in which NFSv4 attributes are handled [RFC7530] [RFC5661]. As in that case, implementations that need to encode and decode these nominally opaque entities need to use the protocol description to determine the actual XDR representation that underlays the items described as opaque. o The transport stream is not represented as a single XDR object. Instead, the header prefix is described by one XDR object while the rest of the header is described as another XDR object with the mapping between the header type in the header prefix and the XDR object representing the header type represented by tables contained in this document, with additional mappings being specifiable by a later extension document. This situation is similar to that in which RPC message headers contain program and procedure numbers, so that the XDR for those request and replies can be used to encode and decode the associated messages without requiring that all be present in a single XDR specification. As in that case, implementations need to use the header specification to select the appropriate XDR- generated code to be used in message processing. o The relationship between the transport stream and the payload stream is not specified in the XDR itself, although comments within the XDR text make clear where transported messages, described by their own XDR, need to appear. Such data by its nature is opaque to the transport, although its form differs XDR opaque arrays. Potential extensions allowing continuation of RPC messages across transport message boundaries will require that message assembly facilities, not specifiable within XDR, also be part of transport implementations. Lever & Noveck Expires May 20, 2020 [Page 51] Internet-Draft RDMA Transport for RPC V2 November 2019 To summarize, the role of XDR in this specification is more limited than for protocols which are themselves XDR programs, where the totality of the protocol is expressible within the XDR paradigm established for that purpose. This more limited role reflects the fact that XDR lacks facilities to represent the embedding of transported material within the transport framework. In addition, the need to cleanly accommodate extensions has meant that those using rpcgen in their applications need to take a more active role in providing the facilities that cannot be expressed within XDR. 8. RPC Bind Parameters In setting up a new RDMA connection, the first action by an RPC client is to obtain a transport address for the RPC server. The means used to obtain this address and to open an RDMA connection is dependent on the type of RDMA transport, and is the responsibility of each RPC protocol binding and its local implementation. RPC services normally register with a portmap or rpcbind service [RFC1833], which associates an RPC Program number with a service address. This policy is no different with RDMA transports. However, a different and distinct service address (port number) might sometimes be required for ULP operation with RPC-over-RDMA. When mapped atop the iWARP transport [RFC5040] [RFC5041], which uses IP port addressing due to its layering on TCP and/or SCTP, port mapping is trivial and consists merely of issuing the port in the connection process. The NFS/RDMA protocol service address has been assigned port 20049 by IANA, for both iWARP/TCP and iWARP/SCTP [RFC8267]. When mapped atop InfiniBand [IBA], which uses a service endpoint naming scheme based on a Group Identifier (GID), a translation MUST be employed. One such translation is described in Annexes A3 (Application Specific Identifiers), A4 (Sockets Direct Protocol (SDP)), and A11 (RDMA IP CM Service) of [IBA], which is appropriate for translating IP port addressing to the InfiniBand network. Therefore, in this case, IP port addressing may be readily employed by the upper layer. When a mapping standard or convention exists for IP ports on an RDMA interconnect, there are several possibilities for each upper layer to consider: o One possibility is to have the server register its mapped IP port with the rpcbind service under the netid (or netids) defined in [RFC8166]. An RPC-over-RDMA-aware RPC client can then resolve its desired service to a mappable port and proceed to connect. This Lever & Noveck Expires May 20, 2020 [Page 52] Internet-Draft RDMA Transport for RPC V2 November 2019 is the most flexible and compatible approach for those upper layers that are defined to use the rpcbind service. o A second possibility is to have the RPC server's portmapper register itself on the RDMA interconnect at a "well-known" service address (on UDP or TCP, this corresponds to port 111). An RPC client could connect to this service address and use the portmap protocol to obtain a service address in response to a program number; e.g., an iWARP port number or an InfiniBand GID. o Alternately, the RPC client could simply connect to the mapped well-known port for the service itself, if it is appropriately defined. By convention, the NFS/RDMA service, when operating atop such an InfiniBand fabric, uses the same 20049 assignment as for iWARP. Historically, different RPC protocols have taken different approaches to their port assignment. Therefore, the specific method is left to each RPC-over-RDMA-enabled ULB and is not addressed in this document. [RFC8166] defines two new netid values to be used for registration of upper layers atop iWARP [RFC5040] [RFC5041] and (when a suitable port translation service is available) InfiniBand [IBA]. Additional RDMA- capable networks MAY define their own netids, or if they provide a port translation, they MAY share the one defined in [RFC8166]. 9. Implementation Status 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 [RFC7942]. 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. At this time, no known implementations of the protocol described in this document exist. Lever & Noveck Expires May 20, 2020 [Page 53] Internet-Draft RDMA Transport for RPC V2 November 2019 10. Security Considerations 10.1. Memory Protection A primary consideration is the protection of the integrity and confidentiality of host memory by an RPC-over-RDMA transport. The use of an RPC-over-RDMA transport protocol MUST NOT introduce vulnerabilities to system memory contents nor to memory owned by user processes. It is REQUIRED that any RDMA provider used for RPC transport be conformant to the requirements of [RFC5042] in order to satisfy these protections. These protections are provided by the RDMA layer specifications, and in particular, their security models. 10.1.1. Protection Domains The use of Protection Domains to limit the exposure of memory regions to a single connection is critical. Any attempt by an endpoint not participating in that connection to reuse memory handles needs to result in immediate failure of that connection. Because ULP security mechanisms rely on this aspect of Reliable connected behavior, strong authentication of remote endpoints is recommended. 10.1.2. Handle (STag) Predictability Unpredictable memory handles should be used for any operation requiring advertised memory regions. Advertising a continuously registered memory region allows a remote host to read or write to that region even when an RPC involving that memory is not under way. Therefore, implementations should avoid advertising persistently registered memory. 10.1.3. Memory Protection Requesters should register memory regions for remote access only when they are about to be the target of an RPC operation that involves an RDMA Read or Write. Registered memory regions should be invalidated as soon as related RPC operations are complete. Invalidation and DMA unmapping of memory regions should be complete before message integrity checking is done and before the RPC consumer is allowed to continue execution and use or alter the contents of a memory region. An RPC transaction on a Requester might be terminated before a reply arrives if the RPC consumer exits unexpectedly (for example, it is signaled or a segmentation fault occurs). When an RPC terminates Lever & Noveck Expires May 20, 2020 [Page 54] Internet-Draft RDMA Transport for RPC V2 November 2019 abnormally, memory regions associated with that RPC should be invalidated appropriately before the regions are released to be reused for other purposes on the Requester. 10.1.4. Denial of Service A detailed discussion of denial-of-service exposures that can result from the use of an RDMA transport is found in Section 6.4 of [RFC5042]. A Responder is not obliged to pull Read chunks that are unreasonably large. The Responder can use an RDMA2_ERROR response to terminate RPCs with unreadable Read chunks. If a Responder transmits more data than a Requester is prepared to receive in a Write or Reply chunk, the RDMA Network Interface Cards (RNICs) typically terminate the connection. For further discussion, see Section 6.4.3. Such repeated chunk errors can deny service to other users sharing the connection from the errant Requester. An RPC-over-RDMA transport implementation is not responsible for throttling the RPC request rate, other than to keep the number of concurrent RPC transactions at or under the number of credits granted per connection. This is explained in Section 4.3.1. A sender can trigger a self denial of service by exceeding the credit grant repeatedly. When an RPC has been canceled due to a signal or premature exit of an application process, a Requester typically invalidates the RPC's Write and Reply chunks. Invalidation prevents the subsequent arrival of the Responder's reply from altering the memory regions associated with those chunks after the memory has been reused. On the Requester, a malfunctioning application or a malicious user can create a situation where RPCs are continuously initiated and then aborted, resulting in Responder replies that terminate the underlying RPC-over-RDMA connection repeatedly. Such situations can deny service to other users sharing the connection from that Requester. 10.2. RPC Message Security ONC RPC provides cryptographic security via the RPCSEC_GSS framework [RFC7861]. RPCSEC_GSS implements message authentication (rpc_gss_svc_none), per-message integrity checking (rpc_gss_svc_integrity), and per-message confidentiality (rpc_gss_svc_privacy) in the layer above the RPC-over-RDMA transport. The latter two services require significant computation and movement of data on each endpoint host. Some performance benefits enabled by RDMA transports can be lost. Lever & Noveck Expires May 20, 2020 [Page 55] Internet-Draft RDMA Transport for RPC V2 November 2019 10.2.1. RPC-over-RDMA Protection at Lower Layers For any RPC transport, utilizing RPCSEC_GSS integrity or privacy services has performance implications. Protection below the RPC transport is often more appropriate in performance-sensitive deployments, especially if it, too, can be offloaded. Certain configurations of IPsec can be co-located in RDMA hardware, for example, without change to RDMA consumers and little loss of data movement efficiency. Such arrangements can also provide a higher degree of privacy by hiding endpoint identity or altering the frequency at which messages are exchanged, at a performance cost. The use of protection in a lower layer MAY be negotiated through the use of an RPCSEC_GSS security flavor defined in [RFC7861] in conjunction with the Channel Binding mechanism [RFC5056] and IPsec Channel Connection Latching [RFC5660]. Use of such mechanisms is REQUIRED where integrity or confidentiality is desired and where efficiency is required. 10.2.2. RPCSEC_GSS on RPC-over-RDMA Transports Not all RDMA devices and fabrics support the above protection mechanisms. Also, per-message authentication is still required on NFS clients where multiple users access NFS files. In these cases, RPCSEC_GSS can protect NFS traffic conveyed on RPC-over-RDMA connections. RPCSEC_GSS extends the ONC RPC protocol without changing the format of RPC messages. By observing the conventions described in this section, an RPC-over-RDMA transport can convey RPCSEC_GSS-protected RPC messages interoperably. As part of the ONC RPC protocol, protocol elements of RPCSEC_GSS that appear in the Payload stream of an RPC-over-RDMA message (such as control messages exchanged as part of establishing or destroying a security context or data items that are part of RPCSEC_GSS authentication material) MUST NOT be reduced. 10.2.2.1. RPCSEC_GSS Context Negotiation Some NFS client implementations use a separate connection to establish a Generic Security Service (GSS) context for NFS operation. Such clients use TCP and the standard NFS port (2049) for context establishment. To enable the use of RPCSEC_GSS with NFS/RDMA, an NFS server MUST also provide a TCP-based NFS service on port 2049. Lever & Noveck Expires May 20, 2020 [Page 56] Internet-Draft RDMA Transport for RPC V2 November 2019 10.2.2.2. RPC-over-RDMA with RPCSEC_GSS Authentication The RPCSEC_GSS authentication service has no impact on the DDP- eligibility of data items in a ULP. However, RPCSEC_GSS authentication material appearing in an RPC message header can be larger than, say, an AUTH_SYS authenticator. In particular, when an RPCSEC_GSS pseudoflavor is in use, a Requester needs to accommodate a larger RPC credential when marshaling RPC Call messages and needs to provide for a maximum size RPCSEC_GSS verifier when allocating reply buffers and Reply chunks. RPC messages, and thus Payload streams, are made larger as a result. ULP operations that fit in a Short Message when a simpler form of authentication is in use might need to be reduced or conveyed via a Long Message when RPCSEC_GSS authentication is in use. It is more likely that a Requester provides both a Read list and a Reply chunk in the same RPC-over-RDMA Transport header to convey a Long Call and provision a receptacle for a Long Reply. In addition to this cost, the XDR encoding and decoding of each RPC message using RPCSEC_GSS authentication requires host compute resources to construct the GSS verifier. 10.2.2.3. RPC-over-RDMA with RPCSEC_GSS Integrity or Privacy The RPCSEC_GSS integrity service enables endpoints to detect modification of RPC messages in flight. The RPCSEC_GSS privacy service prevents all but the intended recipient from viewing the cleartext content of RPC arguments and results. RPCSEC_GSS integrity and privacy services are end-to-end. They protect RPC arguments and results from application to server endpoint, and back. The RPCSEC_GSS integrity and encryption services operate on whole RPC messages after they have been XDR encoded for transmit, and before they have been XDR decoded after receipt. Both sender and receiver endpoints use intermediate buffers to prevent exposure of encrypted data or unverified cleartext data to RPC consumers. After verification, encryption, and message wrapping has been performed, the transport layer MAY use RDMA data transfer between these intermediate buffers. The process of reducing a DDP-eligible data item removes the data item and its XDR padding from the encoded Payload stream. XDR padding of a reduced data item is not transferred in a normal RPC- over-RDMA message. After reduction, the Payload stream contains fewer octets than the whole XDR stream did beforehand. XDR padding octets are often zero bytes, but they don't have to be. Thus, Lever & Noveck Expires May 20, 2020 [Page 57] Internet-Draft RDMA Transport for RPC V2 November 2019 reducing DDP-eligible items affects the result of message integrity verification or encryption. Therefore, a sender MUST NOT reduce a Payload stream when RPCSEC_GSS integrity or encryption services are in use. Effectively, no data item is DDP-eligible in this situation, and Chunked Messages cannot be used. In this mode, an RPC-over-RDMA transport operates in the same manner as a transport that does not support DDP. When an RPCSEC_GSS integrity or privacy service is in use, a Requester provides both a Read list and a Reply chunk in the same RPC-over-RDMA header to convey a Long Call and provision a receptacle for a Long Reply. 10.2.2.4. Protecting RPC-over-RDMA Transport Headers Like the base fields in an ONC RPC message (XID, call direction, and so on), the contents of an RPC-over-RDMA message's Transport stream are not protected by RPCSEC_GSS. This exposes XIDs, connection credit limits, and chunk lists (but not the content of the data items they refer to) to malicious behavior, which could redirect data that is transferred by the RPC-over-RDMA message, result in spurious retransmits, or trigger connection loss. In particular, if an attacker alters the information contained in the chunk lists of an RPC-over-RDMA Transport header, data contained in those chunks can be redirected to other registered memory regions on Requesters. An attacker might alter the arguments of RDMA Read and RDMA Write operations on the wire to similar effect. If such alterations occur, the use of RPCSEC_GSS integrity or privacy services enable a Requester to detect unexpected material in a received RPC message. Encryption at lower layers, as described in Section 10.2.1 protects the content of the Transport stream. To address attacks on RDMA protocols themselves, RDMA transport implementations should conform to [RFC5042]. 10.3. Transport Properties Like other fields that appear in each RPC-over-RDMA header, property information is sent in the clear on the fabric with no integrity protection, making it vulnerable to man-in-the-middle attacks. For example, if a man-in-the-middle were to change the value of the Receive buffer size or the Requester Remote Invalidation boolean, it could reduce connection performance or trigger loss of connection. Repeated connection loss can impact performance or even prevent a new Lever & Noveck Expires May 20, 2020 [Page 58] Internet-Draft RDMA Transport for RPC V2 November 2019 connection from being established. Recourse is to deploy on a private network or use link-layer encryption. 10.4. Host Authentication Wherein we use the relevant sections of [RFC3552] to analyze the addition of host authentication to this RPC-over-RDMA transport. The authors refer readers to Appendix C of [RFC8446] for information on how to design and test a secure authentication handshake implementation. 11. IANA Considerations The RPC-over-RDMA family of transports have been assigned RPC netids by [RFC8166]. A netid is an rpcbind [RFC1833] string used to identify the underlying protocol in order for RPC to select appropriate transport framing and the format of the service addresses and ports. The following netid registry strings are already defined for this purpose: NC_RDMA "rdma" NC_RDMA6 "rdma6" The "rdma" netid is to be used when IPv4 addressing is employed by the underlying transport, and "rdma6" when IPv6 addressing is employed. The netid assignment policy and registry are defined in [RFC5665]. The current document does not alter these netid assignments. These netids MAY be used for any RDMA network that satisfies the requirements of Section 3.2.2 and that is able to identify service endpoints using IP port addressing, possibly through use of a translation service as described in Section 8. 12. References 12.1. Normative References [RFC1833] Srinivasan, R., "Binding Protocols for ONC RPC Version 2", RFC 1833, DOI 10.17487/RFC1833, August 1995, . Lever & Noveck Expires May 20, 2020 [Page 59] Internet-Draft RDMA Transport for RPC V2 November 2019 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, . [RFC4506] Eisler, M., Ed., "XDR: External Data Representation Standard", STD 67, RFC 4506, DOI 10.17487/RFC4506, May 2006, . [RFC5042] Pinkerton, J. and E. Deleganes, "Direct Data Placement Protocol (DDP) / Remote Direct Memory Access Protocol (RDMAP) Security", RFC 5042, DOI 10.17487/RFC5042, October 2007, . [RFC5056] Williams, N., "On the Use of Channel Bindings to Secure Channels", RFC 5056, DOI 10.17487/RFC5056, November 2007, . [RFC5280] Cooper, D., Santesson, S., Farrell, S., Boeyen, S., Housley, R., and W. Polk, "Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile", RFC 5280, DOI 10.17487/RFC5280, May 2008, . [RFC5531] Thurlow, R., "RPC: Remote Procedure Call Protocol Specification Version 2", RFC 5531, DOI 10.17487/RFC5531, May 2009, . [RFC5660] Williams, N., "IPsec Channels: Connection Latching", RFC 5660, DOI 10.17487/RFC5660, October 2009, . [RFC5665] Eisler, M., "IANA Considerations for Remote Procedure Call (RPC) Network Identifiers and Universal Address Formats", RFC 5665, DOI 10.17487/RFC5665, January 2010, . [RFC6125] Saint-Andre, P. and J. Hodges, "Representation and Verification of Domain-Based Application Service Identity within Internet Public Key Infrastructure Using X.509 (PKIX) Certificates in the Context of Transport Layer Security (TLS)", RFC 6125, DOI 10.17487/RFC6125, March 2011, . [RFC7861] Adamson, A. and N. Williams, "Remote Procedure Call (RPC) Security Version 3", RFC 7861, DOI 10.17487/RFC7861, November 2016, . Lever & Noveck Expires May 20, 2020 [Page 60] Internet-Draft RDMA Transport for RPC V2 November 2019 [RFC7942] Sheffer, Y. and A. Farrel, "Improving Awareness of Running Code: The Implementation Status Section", BCP 205, RFC 7942, DOI 10.17487/RFC7942, July 2016, . [RFC8166] Lever, C., Ed., Simpson, W., and T. Talpey, "Remote Direct Memory Access Transport for Remote Procedure Call Version 1", RFC 8166, DOI 10.17487/RFC8166, June 2017, . [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, May 2017, . [RFC8267] Lever, C., "Network File System (NFS) Upper-Layer Binding to RPC-over-RDMA Version 1", RFC 8267, DOI 10.17487/RFC8267, October 2017, . [RFC8446] Rescorla, E., "The Transport Layer Security (TLS) Protocol Version 1.3", RFC 8446, DOI 10.17487/RFC8446, August 2018, . 12.2. Informative References [CBFC] Kung, H., Blackwell, T., and A. Chapman, "Credit-Based Flow Control for ATM Networks: Credit Update Protocol, Adaptive Credit Allocation, and Statistical Multiplexing", Proc. ACM SIGCOMM '94 Symposium on Communications Architectures, Protocols and Applications, pp. 101-114., August 1994. [IBA] InfiniBand Trade Association, "InfiniBand Architecture Specification Volume 1", Release 1.3, March 2015. Available from https://www.infinibandta.org/ [RFC0768] Postel, J., "User Datagram Protocol", STD 6, RFC 768, DOI 10.17487/RFC0768, August 1980, . [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, RFC 793, DOI 10.17487/RFC0793, September 1981, . [RFC1094] Nowicki, B., "NFS: Network File System Protocol specification", RFC 1094, DOI 10.17487/RFC1094, March 1989, . Lever & Noveck Expires May 20, 2020 [Page 61] Internet-Draft RDMA Transport for RPC V2 November 2019 [RFC1813] Callaghan, B., Pawlowski, B., and P. Staubach, "NFS Version 3 Protocol Specification", RFC 1813, DOI 10.17487/RFC1813, June 1995, . [RFC3552] Rescorla, E. and B. Korver, "Guidelines for Writing RFC Text on Security Considerations", BCP 72, RFC 3552, DOI 10.17487/RFC3552, July 2003, . [RFC5040] Recio, R., Metzler, B., Culley, P., Hilland, J., and D. Garcia, "A Remote Direct Memory Access Protocol Specification", RFC 5040, DOI 10.17487/RFC5040, October 2007, . [RFC5041] Shah, H., Pinkerton, J., Recio, R., and P. Culley, "Direct Data Placement over Reliable Transports", RFC 5041, DOI 10.17487/RFC5041, October 2007, . [RFC5532] Talpey, T. and C. Juszczak, "Network File System (NFS) Remote Direct Memory Access (RDMA) Problem Statement", RFC 5532, DOI 10.17487/RFC5532, May 2009, . [RFC5661] Shepler, S., Ed., Eisler, M., Ed., and D. Noveck, Ed., "Network File System (NFS) Version 4 Minor Version 1 Protocol", RFC 5661, DOI 10.17487/RFC5661, January 2010, . [RFC5662] Shepler, S., Ed., Eisler, M., Ed., and D. Noveck, Ed., "Network File System (NFS) Version 4 Minor Version 1 External Data Representation Standard (XDR) Description", RFC 5662, DOI 10.17487/RFC5662, January 2010, . [RFC7530] Haynes, T., Ed. and D. Noveck, Ed., "Network File System (NFS) Version 4 Protocol", RFC 7530, DOI 10.17487/RFC7530, March 2015, . [RFC8167] Lever, C., "Bidirectional Remote Procedure Call on RPC- over-RDMA Transports", RFC 8167, DOI 10.17487/RFC8167, June 2017, . [RFC8178] Noveck, D., "Rules for NFSv4 Extensions and Minor Versions", RFC 8178, DOI 10.17487/RFC8178, July 2017, . Lever & Noveck Expires May 20, 2020 [Page 62] Internet-Draft RDMA Transport for RPC V2 November 2019 Appendix A. ULB Specifications An Upper-Layer Protocol (ULP) is typically defined independently of any particular RPC transport. An Upper-Layer Binding (ULB) specification provides guidance that helps the ULP interoperate correctly and efficiently over a particular transport. For RPC-over- RDMA version 2, a ULB may provide: o A taxonomy of XDR data items that are eligible for DDP o Constraints on which upper-layer procedures may be reduced and on how many chunks may appear in a single RPC request o A method for determining the maximum size of the reply Payload stream for all procedures in the ULP o An rpcbind port assignment for operation of the RPC Program and Version on an RPC-over-RDMA transport Each RPC Program and Version tuple that utilizes RPC-over-RDMA version 2 needs to have a ULB specification. A.1. DDP-Eligibility An ULB designates some XDR data items as eligible for DDP. As an RPC-over-RDMA message is formed, DDP-eligible data items can be removed from the Payload stream and placed directly in the receiver's memory. An XDR data item should be considered for DDP-eligibility if there is a clear benefit to moving the contents of the item directly from the sender's memory to the receiver's memory. Criteria for DDP-eligibility include: o The XDR data item is frequently sent or received, and its size is often much larger than typical inline thresholds. o If the XDR data item is a result, its maximum size must be predictable in advance by the requester. o Transport-level processing of the XDR data item is not needed. For example, the data item is an opaque byte array, which requires no XDR encoding and decoding of its content. o The content of the XDR data item is sensitive to address alignment. For example, a data copy operation would be required on the receiver to enable the message to be parsed correctly, or to enable the data item to be accessed. Lever & Noveck Expires May 20, 2020 [Page 63] Internet-Draft RDMA Transport for RPC V2 November 2019 o The XDR data item does not contain DDP-eligible data items. In addition to defining the set of data items that are DDP-eligible, a ULB may also limit the use of chunks to particular upper-layer procedures. If more than one data item in a procedure is DDP- eligible, the ULB may also limit the number of chunks that a requester can provide for a particular upper-layer procedure. Senders MUST NOT reduce data items that are not DDP-eligible. Such data items MAY, however, be moved as part of a Position Zero Read chunk or a Reply chunk. The programming interface by which an upper-layer implementation indicates the DDP-eligibility of a data item to the RPC transport is not described by this specification. The only requirements are that the receiver can re-assemble the transmitted RPC-over-RDMA message into a valid XDR stream, and that DDP-eligibility rules specified by the ULB are respected. There is no provision to express DDP-eligibility within the XDR language. The only definitive specification of DDP-eligibility is a ULB. In general, a DDP-eligibility violation occurs when: o A requester reduces a non-DDP-eligible argument data item. The Responder MUST NOT process this RPC Call message and MUST report the violation as described in Section 6.4.3. o A Responder reduces a non-DDP-eligible result data item. The requester MUST terminate the pending RPC transaction and report an appropriate permanent error to the RPC consumer. o A Responder does not reduce a DDP-eligible result data item into an available Write chunk. The requester MUST terminate the pending RPC transaction and report an appropriate permanent error to the RPC consumer. A.2. Maximum Reply Size When expecting small and moderately-sized Replies, a requester should typically rely on Message Continuation rather than provisioning a Reply chunk. For each ULP procedure where there is no clear Reply size maximum and the maximum can be large, the ULB should specify a dependable means for determining the maximum Reply size. Lever & Noveck Expires May 20, 2020 [Page 64] Internet-Draft RDMA Transport for RPC V2 November 2019 A.3. Additional Considerations There may be other details provided in a ULB. o An ULB may recommend inline threshold values or other transport- related parameters for RPC-over-RDMA version 2 connections bearing that ULP. o An ULP may provide a means to communicate these transport-related parameters between peers. Note that RPC-over-RDMA version 2 does not specify any mechanism for changing any transport-related parameter after a connection has been established and the initial transport properties have been exchanged. o Multiple ULPs may share a single RPC-over-RDMA version 2 connection when their ULBs allow the use of RPC-over-RDMA version 2 and the rpcbind port assignments for the Protocols allow connection sharing. In this case, the same transport parameters (such as inline threshold) apply to all Protocols using that connection. Each ULB needs to be designed to allow correct interoperation without regard to the transport parameters actually in use. Furthermore, implementations of ULPs must be designed to interoperate correctly regardless of the connection parameters in effect on a connection. A.4. ULP Extensions An RPC Program and Version tuple may be extensible. For instance, there may be a minor versioning scheme that is not reflected in the RPC version number, or the ULP may allow additional features to be specified after the original RPC Program specification was ratified. ULBs are provided for interoperable RPC Programs and Versions by extending existing ULBs to reflect the changes made necessary by each addition to the existing XDR. Appendix B. Extending the Version 2 Protocol This Appendix is not addressed to protocol implementers, but rather to authors of documents that intend to extend the protocol described earlier in this document. Subsequent RPC-over-RDMA versions are free to change the protocol in any way they choose as long as they leave unchanged those fields identified as "fixed for all versions" in Section 4.2.1 of [RFC8166]. Such changes might involve deletion or major re-organization of existing transport headers. However, the need for interoperability Lever & Noveck Expires May 20, 2020 [Page 65] Internet-Draft RDMA Transport for RPC V2 November 2019 between adjacent versions will often limit the scope of changes that can be made in a single version. In some cases it may prove desirable to transition to a new version by using the extension features described for use with RPC-over-RDMA version 2, by continuing the same basic extension model but allowing header types and properties that were OPTIONAL in one version to become REQUIRED in the subsequent version. RPC-over-RDMA version 2 is designed to be extensible in a way that enables the addition of OPTIONAL features that may subsequently be converted to REQUIRED status in a future protocol version. The protocol may be extended by Standards Track documents in a way analogous to that provided for Network File System Version 4 as described in [RFC8178]. This form of extensibility enables limited extensions to the base RPC-over-RDMA version 2 protocol presented in this document so that new optional capabilities can be introduced without a protocol version change, while maintaining robust interoperability with existing RPC-over-RDMA version 2 implementations. The design allows extensions to be defined, including the definition of new protocol elements, without requiring modification or recompilation of the existing XDR. A Standards Track document introduces each set of such protocol elements. Together these elements are considered an OPTIONAL feature. Each implementation is either aware of all the protocol elements introduced by that feature or is aware of none of them. Documents describing extensions to RPC-over-RDMA version 2 should contain: o An explanation of the purpose and use of each new protocol element added. o An XDR description including all of the new protocol elements, and a script to extract it. o A description of interactions with existing extensions. This includes possible requirements of other OPTIONAL features to be present for new protocol elements to work, or that a particular level of support for an OPTIONAL facility is required for the new extension to work. Implementers combine the XDR descriptions of the new features they intend to use with the XDR description of the base protocol in this Lever & Noveck Expires May 20, 2020 [Page 66] Internet-Draft RDMA Transport for RPC V2 November 2019 document. This may be necessary to create a valid XDR input file because extensions are free to use XDR types defined in the base protocol, and later extensions may use types defined by earlier extensions. The XDR description for the RPC-over-RDMA version 2 base protocol combined with that for any selected extensions should provide an adequate human-readable description of the extended protocol. The base protocol specified in this document may be extended within RPC-over-RDMA version 2 in two ways: o New OPTIONAL transport header types may be introduced by later Standards Track documents. Such transport header types will be documented as described in Appendix B.1. o New OPTIONAL transport properties may be defined in later Standards Track documents. Such transport properties will be documented as described in Appendix B.3. The following sorts of ancillary protocol elements may be added to the protocol to support the addition of new transport properties and header types. o New error codes may be created as described in Appendix B.4. o New flags to use within the rdma_flags field may be created as described in Appendix B.2. New capabilities can be proposed and developed independently of each other, and implementers can choose among them. This makes it straightforward to create and document experimental features and then bring them through the standards process. B.1. Adding New Header Types to RPC-over-RDMA Version 2 New transport header types are to defined in a manner similar to the way existing ones are described in Sections 6.4.1 through 6.4.4. Specifically what is needed is: o A description of the function and use of the new header type. o A complete XDR description of the new header type including a description of the use of all fields within the header. o A description of how errors are reported, including the definition of a mechanism for reporting errors when the error is outside the Lever & Noveck Expires May 20, 2020 [Page 67] Internet-Draft RDMA Transport for RPC V2 November 2019 available choices already available in the base protocol or in other existing extensions. o An indication of whether a Payload stream must be present, and a description of its contents and how such payload streams are used to construct RPC messages for processing. In addition, there needs to be additional documentation that is made necessary due to the Optional status of new transport header types. o Information about constraints on support for the new header types should be provided. For example, if support for one header type is implied or foreclosed by another one, this needs to be documented. o A preferred method by which a sender should determine whether the peer supports a particular header type needs to be provided. While it is always possible for a send a test invocation of a particular header type to see if support is available, when more efficient means are available (e.g. the value of a transport property, this should be noted. B.2. Adding New Header Flags to the Protocol New flag bits are to defined in a manner similar to the way existing ones are described in Sections 6.3.2.1 and 6.3.2.2. Each new flag definition should include: o An XDR description of the new flag. o A description of the function and use of the new flag. o An indication for which header types the flag value is meaningful and for which header types it is an error to set the flag or to leave it unset. o A means to determine whether receivers are prepared to receive transport headers with the new flag set. In addition, there needs to be additional documentation that is made necessary due to the Optional status of new transport header types. o Information about constraints on support for the new flags should be provided. For example, if support for one flag is implied or foreclosed by another one, this needs to be documented. Lever & Noveck Expires May 20, 2020 [Page 68] Internet-Draft RDMA Transport for RPC V2 November 2019 B.3. Adding New Transport properties to the Protocol The set of transport properties is designed to be extensible. As a result, once new properties are defined in standards track documents, the operations defined in this document may reference these new transport properties, as well as the ones described in this document. A standards track document defining a new transport property should include the following information paralleling that provided in this document for the transport properties defined herein. o The rpcrdma2_propid value used to identify this property. o The XDR typedef specifying the form in which the property value is communicated. o A description of the transport property that is communicated by the sender of RDMA2_CONNPROP. o An explanation of how this knowledge could be used by the peer receiving this information. The definition of transport property structures is such as to make it easy to assign unique values. There is no requirement that a continuous set of values be used and implementations should not rely on all such values being small integers. A unique value should be selected when the defining document is first published as an internet draft. When the document becomes a standards track document, the working group should ensure that: o rpcrdma2_propid values specified in the document do not conflict with those currently assigned or in use by other pending working group documents defining transport properties. o rpcrdma2_propid values specified in the document do not conflict with the range reserved for experimental use, as defined in Section 8.2. Documents defining new properties fall into a number of categories. o Those defining new properties and explaining (only) how they affect use of existing message types. o Those defining new OPTIONAL message types and new properties applicable to the operation of those new message types. o Those defining new OPTIONAL message types and new properties applicable both to new and existing message types. Lever & Noveck Expires May 20, 2020 [Page 69] Internet-Draft RDMA Transport for RPC V2 November 2019 When additional transport properties are proposed, the review of the associated standards track document should deal with possible security issues raised by those new transport properties. B.4. Adding New Error Codes to the Protocol New error codes to be returned when using new header types may be introduced in the same Standards Track document that defines the new header type. Cases in which a new error code is to be returned by an existing header type can be accommodated by defining the new error code in the same Standards Track document that defines the new transport property. For error codes that do not require that additional error information be returned with them, the existing RDMA_ERR2 header can be used to report the new error. The new error code is set as the value of rdma_err with the result that the default switch arm of the rpcrdma2_error (i.e. void) is selected. For error codes that do require the return of additional error- related information together with the error, a new header type should be defined for the purpose of returning the error together with needed additional information. It should be documented just like any other new header type. When a new header type is sent, the sender needs to be prepared to accept header types necessary to report associated errors. Appendix C. Differences from the RPC-over-RDMA Version 1 Protocol This section describes the substantive changes made in RPC-over-RDMA version 2. C.1. Relationship to the RPC-over-RDMA Version 1 XDR Definition There are a number of structural XDR changes whose goal is to enable within-version protocol extensibility. The RPC-over-RDMA version 1 transport header is defined as a single XDR object, with an RPC message proper potentially following it. In RPC-over-RDMA version 2, as described in Section 6.1 there are separate XDR definitions of the transport header prefix (see Section 6.3.2 which specifies the transport header type to be used, and the specific transport header, defined within one of the subsections of Section 6). This is similar to the way that an RPC message consists of an RPC header (defined in [RFC5531]) and an RPC request or reply, defined by the Upper-Layer protocol being conveyed. Lever & Noveck Expires May 20, 2020 [Page 70] Internet-Draft RDMA Transport for RPC V2 November 2019 As a new version of the RPC-over-RDMA transport protocol, RPC-over- RDMA version 2 exists within the versioning rules defined in [RFC8166]. In particular, it maintains the first four words of the protocol header as sent and received, as specified in Section 4.2 of [RFC8166], even though, as explained in Section 6.3.1 of this document, the XDR definition of those words is structured differently. Although each of the first four words retains its semantic function, there are important differences of field interpretation, besides the fact that the words have different names and different roles with the XDR constrict of they are parts. o The first word of the header, previously the rdma_xid field, retains the format and function that in had in RPC-over-RDMA version 1. Within RPC-over-RDMA version 2, this word is the rdma_xid field of the structure rdma_start. However, to accommodate the use of request-response pairing of non-RPC messages and the potential use of message continuation, it cannot be assumed that it will always have the same value it would have had in RPC-over-RDMA version 1. As a result, the contents of this field should not be used without consideration of the associated protocol version identification. o The second word of the header, previously the rdma_vers field, retains the format and function that it had in RPC-over-RDMA version 1. Within RPC-over-RDMA version 2, this word is the rdma_vers field of the structure rdma_start. To clearly distinguish version 1 and version 2 messages, senders MUST fill in the correct version (fixed after version negotiation) and receivers MUST check that the content of the rdma_vers is correct before using referencing any other header field. o The third word of the header, previously the rdma_credit field, retains the size and general purpose that it had in RPC-over-RDMA version 1. Within RPC-over-RDMA version 2, this word is the rdma_credit field of the structure rdma_start. o The fourth word of the header, previously the union discriminator field rdma_proc, retains its format and general function even though the set of valid values has changed. The value of this field is now considered an unsigned 32-bit integer rather than an enum. Within RPC-over-RDMA version 2, this word is the rdma_htype field of the structure rdma_start. Beyond conforming to the restrictions specified in [RFC8166], RPC- over-RDMA version 2 tightly limits the scope of the changes made in order to ensure interoperability. It makes no major structural Lever & Noveck Expires May 20, 2020 [Page 71] Internet-Draft RDMA Transport for RPC V2 November 2019 changes to the protocol, and all existing transport header types used in version 1 (as defined in [RFC8166]) are retained in version 2. Chunks are expressed using the same on-the-wire format and are used in the same way in both versions. C.2. Transport Properties RPC-over-RDMA version 2 provides a mechanism for exchanging the transport's operational properties. This mechanism allows connection endpoints to communicate the properties of their implementation at connection setup. The mechanism could be expanded to enable an endpoint to request changes in properties of the other endpoint and to notify peer endpoints of changes to properties that occur during operation. Transport properties are described in Section 5. C.3. Credit Management Changes RPC-over-RDMA transports employ credit-based flow control to ensure that a requester does not emit more RDMA Sends than the responder is prepared to receive. Section 3.3.1 of [RFC8166] explains the purpose and operation of RPC-over-RDMA version 1 credit management in detail. In the RPC-over-RDMA version 1 design, each RDMA Send from a requester contains an RPC Call with a credit request, and each RDMA Send from a responder contains an RPC Reply with a credit grant. The credit grant implies that enough Receives have been posted on the responder to handle the credit grant minus the number of pending RPC transactions (the number of remaining Receive buffers might be zero). In other words, each RPC Reply acts as an implicit ACK for a previous RPC Call from the requester, indicating that the responder has posted a Receive to replace the Receive consumed by the requester's RDMA Send. Without an RPC Reply message, the requester has no way to know that the responder is properly prepared for subsequent RPC Calls. Aside from being a bit of a layering violation, there are basic (but rare) cases where this arrangement is inadequate: o When a requester retransmits an RPC Call on the same connection as an earlier RPC Call for the same transaction. o When a requester transmits an RPC operation that requires no reply. o When more than one RPC-over-RDMA message is needed to complete the transaction (e.g., RDMA_DONE). Lever & Noveck Expires May 20, 2020 [Page 72] Internet-Draft RDMA Transport for RPC V2 November 2019 Typically, the connection must be replaced in these cases. This resets the credit accounting mechanism but has an undesirable impact on other ongoing RPC transactions on that connection. Because credit management accompanies each RPC message, there is a strict one-to-one ratio between RDMA Send and RPC message. There are interesting use cases that might be enabled if this relationship were more flexible: o RPC-over-RDMA operations which do not carry an RPC message; e.g., control plane operations. o A single RDMA Send that conveys more than one RPC message for the purpose of interrupt mitigation. o An RPC message that is conveyed via several sequential RDMA Sends to reduce the use of explicit RDMA operations for moderate-sized RPC messages. o An RPC transaction that needs multiple exchanges or an odd number of RPC-over-RDMA operations to complete. Bi-directional RPC operation also introduces an ambiguity. If the RPC-over-RDMA message does not carry an RPC message, then it is not possible to determine whether the sender is a requester or a responder, and thus whether the rdma_credit field contains a credit request or a credit grant. A more sophisticated credit accounting mechanism is provided in RPC- over-RDMA version 2 in an attempt to address some of these shortcomings. This new mechanism is detailed in Section 4.3.1. C.4. Inline Threshold Changes The term "inline threshold" is defined in Section 3.3.2 of [RFC8166]. An "inline threshold" value is the largest message size (in octets) that can be conveyed on an RDMA connection using only RDMA Send and Receive. Each connection has two inline threshold values: one for messages flowing from client-to-server (referred to as the "client- to-server inline threshold") and one for messages flowing from server-to-client (referred to as the "server-to-client inline threshold"). Note that [RFC8166] uses somewhat different terminology. This is because it was written with only forward- direction RPC transactions in mind. A connection's inline thresholds determine when RDMA Read or Write operations are required because the RPC message to be sent cannot be conveyed via a single RDMA Send and Receive pair. When an RPC Lever & Noveck Expires May 20, 2020 [Page 73] Internet-Draft RDMA Transport for RPC V2 November 2019 message does not contain DDP-eligible data items, a requester can prepare a Long Call or Reply to convey the whole RPC message using RDMA Read or Write operations. RDMA Read and Write operations require that each data payload resides in a region of memory that is registered with the RNIC. When an RPC is complete, that region is invalidated, fencing it from the responder. Memory registration and invalidation typically have a latency cost that is insignificant compared to data handling costs. When a data payload is small, however, the cost of registering and invalidating the memory where the payload resides becomes a relatively significant part of total RPC latency. Therefore the most efficient operation of RPC-over-RDMA occurs when explicit RDMA Read and Write operations are used for large payloads, and are avoided for small payloads. When RPC-over-RDMA version 1 was conceived, the typical size of RPC messages that did not involve a significant data payload was under 500 bytes. A 1024-byte inline threshold adequately minimized the frequency of inefficient Long messages. With NFS version 4.1 [RFC5661], the increased size of NFS COMPOUND operations resulted in RPC messages that are on average larger and more complex than previous versions of NFS. With 1024-byte inline thresholds, RDMA Read or Write operations are needed for frequent operations that do not bear a data payload, such as GETATTR and LOOKUP, reducing the efficiency of the transport. To reduce the need to use Long messages, RPC-over-RDMA version 2 increases the default size of inline thresholds. This also increases the maximum size of reverse-direction RPC messages. C.5. Message Continuation Changes In addition to a larger default inline threshold, RPC-over-RDMA version 2 introduces Message Continuation. Message Continuation is a mechanism that enables the transmission of a data payload using more than one RDMA Send. The purpose of Message Continuation is to provide relief in several important cases: o If a requester finds that it is inefficient to convey a moderately-sized data payload using Read chunks, the requester can use Message Continuation to send the RPC Call. o If a requester has provided insufficient Reply chunk space for a responder to send an RPC Reply, the responder can use Message Continuation to send the RPC Reply. Lever & Noveck Expires May 20, 2020 [Page 74] Internet-Draft RDMA Transport for RPC V2 November 2019 o If a sender has to convey a large non-RPC data payload (e.g, a large transport property), the sender can use Message Continuation to avoid using registered memory. C.6. Host Authentication Changes For general operation of NFS on open networks, we eventually intend to rely on RPC-on-TLS [citation needed] to provide cryptographic authentication of the two ends of each connection. In turn, this will improve the trustworthiness of AUTH_SYS-style user identities that flow on TCP, which are not cryptographic. We do not have a similar solution for RPC-over-RDMA, however. Here, the RDMA transport layer already provides a strong guarantee of message integrity. On some network fabrics, IPsec can be used to protect the privacy of in-transit data, or TLS itself could be used for transporting raw RDMA operations. However, this is not the case for all fabrics (e.g., InfiniBand [IBA]). Thus, it is sensible to add a mechanism in the RPC-over-RDMA transport itself for authenticating the connection peers. This mechanism is described in Section 5.2.6. And like GSS channel binding, there should also be a way to determine when the use of host authentication is superfluous and can be avoided. C.7. Support for Remote Invalidation An STag that is registered using the FRWR mechanism in a privileged execution context or is registered via a Memory Window in an unprivileged context may be invalidated remotely [RFC5040]. These mechanisms are available when a requester's RNIC supports MEM_MGT_EXTENSIONS. For the purposes of this discussion, there are two classes of STags. Dynamically-registered STags are used in a single RPC, then invalidated. Persistently-registered STags live longer than one RPC. They may persist for the life of an RPC-over-RDMA connection, or longer. An RPC-over-RDMA requester may provide more than one STag in one transport header. It may provide a combination of dynamically- and persistently-registered STags in one RPC message, or any combination of these in a series of RPCs on the same connection. Only dynamically-registered STags using Memory Windows or FRWR (i.e., registered via MEM_MGT_EXTENSIONS) may be invalidated remotely. There is no transport-level mechanism by which a responder can determine how a requester-provided STag was registered, nor whether Lever & Noveck Expires May 20, 2020 [Page 75] Internet-Draft RDMA Transport for RPC V2 November 2019 it is eligible to be invalidated remotely. A requester that mixes persistently- and dynamically-registered STags in one RPC, or mixes them across RPCs on the same connection, must therefore indicate which handles may be invalidated via a mechanism provided in the Upper-Layer Protocol. RPC-over-RDMA version 2 provides such a mechanism. The RDMA Send With Invalidate operation is used to invalidate an STag on a remote system. It is available only when a responder's RNIC supports MEM_MGT_EXTENSIONS, and must be utilized only when a requester's RNIC supports MEM_MGT_EXTENSIONS (can receive and recognize an IETH). Existing RPC-over-RDMA transport protocol specifications [RFC8166] [RFC8167] do not forbid direct data placement in the reverse direction, even though there is currently no Upper-Layer Protocol that makes data items in reverse direction operations elegible for direct data placement. When chunks are present in a reverse direction RPC request, Remote Invalidation allows the responder to trigger invalidation of a requester's STags as part of sending a reply, the same way as is done in the forward direction. However, in the reverse direction, the server acts as the requester, and the client is the responder. The server's RNIC, therefore, must support receiving an IETH, and the server must have registered the STags with an appropriate registration mechanism. C.8. Error Reporting Changes RPC-over-RDMA version 2 expands the repertoire of errors that may be reported by connection endpoints. This change, which is structured to enable extensibility, allows a peer to report overruns of specific resources and to avoid requester retries when an error is permanent. Acknowledgments The authors gratefully acknowledge the work of Brent Callaghan and Tom Talpey on the original RPC-over-RDMA version 1 specification (RFC 5666). The authors also wish to thank Bill Baker, Greg Marsden, and Matt Benjamin for their support of this work. The XDR extraction conventions were first described by the authors of the NFS version 4.1 XDR specification [RFC5662]. Herbert van den Bergh suggested the replacement sed script used in this document. Lever & Noveck Expires May 20, 2020 [Page 76] Internet-Draft RDMA Transport for RPC V2 November 2019 Special thanks go to Transport Area Director Magnus Westerlund, NFSV4 Working Group Chairs Spencer Shepler and Brian Pawlowski, and NFSV4 Working Group Secretary Thomas Haynes for their support. Authors' Addresses Charles Lever (editor) Oracle Corporation United States of America Email: chuck.lever@oracle.com David Noveck NetApp 1601 Trapelo Road Waltham, MA 02451 United States of America Phone: +1 781 572 8038 Email: davenoveck@gmail.com Lever & Noveck Expires May 20, 2020 [Page 77]