Network File System Version 4 C. Lever
Internet-Draft Oracle
Intended status: Informational April 8, 2016
Expires: October 10, 2016

RPC-over-RDMA Version One Implementation Experience


This document details experiences and challenges implementing the RPC-over-RDMA Version One protocol. Specification changes are recommended to address avoidable interoperability failures.

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

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 October 10, 2016.

Copyright Notice

Copyright (c) 2016 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 ( 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.

Table of Contents

1. Introduction

This document summarizes implementation experience with the RPC-over-RDMA Version One protocol [RFC5666], and proposes improvements to the protocol specification based on implementer experience, frequently-asked questions, and interviews with a co-author of RFC 5666.

1.1. Purpose Of This Document

A key contribution of this document is to highlight areas of RFC 5666 where independent good faith readings could result in distinct implementations that do not interoperate with each other. Correcting these specification issues is critical: fresh implementations of RPC-over-RDMA Version One continue to arise.

Recommendations are limited to the following areas:

1.2. Updating RFC 5666

During IETF 92, several alternatives for updating RFC 5666 were discussed with the RFC Editor and with the assembled members of the nfsv4 Working Group. Among them were:

An additional possibility which is sometimes chosen by other Working Groups would be to update RFC 5666 as it transitions from Proposed Standard to Draft Standard.

There was general agreement during the meeting regarding the need to update and obsolete RFC 5666 while retaining a high degree of interoperability with current RPC-over-RDMA Version One implementations. This approach would avoid changes to on-the-wire behavior without burdening implementers, who could continue to reference a single specification of the protocol. In addition, this alternative extends the life of current interoperable RPC-over-RDMA Version One implementations in the field.

Subsequent discussion within the nfsv4 Working Group has focused on resolving specification ambiguities that make the construction of interoperable implementations unduly difficult. Subsequent Versions of RPC-over-RDMA, where deeper changes can be made and new functionality introduced, remain a possibility.

1.3. 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 [RFC2119].

2. RPC-Over-RDMA Essentials

The following sections summarize the state of affairs defined in RFC 5666. This is a distillation of text from RFC 5666, dialog with a co-author of RFC 5666, and implementer experience. The XDR definitions are copied from RFC 5666 Section 4.3.

2.1. Arguments And Results

Like a local function call, every Remote Procedure Call (RPC) operation has a set of one or more "arguments" and a set of one or more "results." The calling context is not allowed to proceed until the function's results are available. Unlike a local function call, the called function is executed remotely rather than in the local application's context.

A client endpoint, or "requester", serializes an RPC call's arguments into a byte stream using XDR [RFC4506]. This "XDR stream" is conveyed to a server endpoint via an RPC call message (sometimes referred to as an "RPC request").

The server endpoint, or "responder", deserializes the arguments and processes the requested operation. It then serializes the operation's results into another XDR stream. This stream is conveyed back to the client endpoint via an RPC reply message. The client deserializes the results and allows the original caller to proceed.

The remainder of this document assumes a working knowledge of the RPC protocol [RFC5531] and especially XDR [RFC4506].

2.2. Remote Direct Memory Access

RPC messages may be very large. For example, NFS READ and WRITE operations are often 100KB or larger.

An RPC client system can be made more efficient if RPC messages are transferred by a third party such as intelligent network interface hardware. Remote Direct Memory Access (RDMA) and Direct Data Placement (DDP) enables offloading data movement to avoid the negative performance effects of using traditional host CPU-based network operations to move bulk data.

RFC 5666 describes how to use only the Send, Receive, RDMA Read, and RDMA Write operations described in [RFC5040] and [RFC5041] to move RPC calls and replies between requesters and responders.

2.2.1. Direct Data Placement

RFC 5666 makes an important distinction between RDMA and Direct Data Placement (DDP).

Very often, RPC implementations copy the contents of RPC messages into a buffer before being sent. A good RPC implementation may be able to send bulk data without having to copy 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.

In RFC 5666, "RDMA" refers to the physical mechanism an RDMA transport utilizes when moving data. Though it may not be optimal, before an RDMA transfer, the sender may still copy data into place. After an RDMA transfer, the receiver may copy that data again to its final destination.

RFC 5666 uses the term "direct data placement" to refer to an optimization that makes it unnecessary for a host CPU to copy data to be transferred. RPC-over-RDMA Version One utilizes RDMA Read and Write operations to enable DDP. Not every RDMA-based transfer in RPC-over-RDMA Version One is DDP, however.

2.2.2. Channel Operation

A Send operation initiates the transfer of a message from a local endpoint to a remote endpoint, similar to a datagram send operation.

The remote endpoint pre-posts Receive operations to catch incoming messages. Send operations are flow-controlled to prevent overrunning receive resources. To reduce the amount of memory that must remain pinned awaiting incoming messages, receive buffers are limited in size and number.

This transfer mode is utilized to convey size-limited RPC operations, and advertisements of buffer coordinates for explicit RDMA data transfer. Buffers involved in Send and Receive operations are usually left unexposed.

2.2.3. Explicit RDMA Operation

A local endpoint tags memory areas to be involved in RDMA, exposes the areas, then advertises the coordinates of those areas to a remote endpoint via a Send operation.

The remote endpoint transfers data into or out of those areas using RDMA Read and Write operations. The remote registers large sink buffers as needed, and invalidates them when data transfer is complete.

Finally the remote endpoint signals that its work is done, and the local endpoint ensures remote access to the memory areas is no longer allowed.

This transfer mode can be utilized to convey large whole RPC messages, although typically only one data item within a message is large. Explicit RDMA is most often used to move large argument or result data items directly into place. The remaining portions of the message are conveyed via a channel operation.

2.3. Transfer Models

A "transfer model" describes which endpoint is responsible for performing RDMA Read and Write operations. The opposite endpoint must expose part or all of its memory, and advertise the coordinates of that memory.

2.3.1. Read-Read

Requesters expose their memory to the responder, and the responder exposes its memory to requesters. The responder employs RDMA Read operations to convey RPC arguments or whole RPC calls. Requesters employ RDMA Read operations to convey RPC results or whole RPC relies.

Although this model is specified in RFC 5666, no current RPC-over-RDMA Version One implementation uses the Read-Read transfer model.

2.3.2. Write-Write

Requesters expose their memory to the responder, and the responder exposes its memory to requesters. Requesters employ RDMA Write operations to convey RPC arguments or whole RPC calls. The responder employs RDMA Write operations to convey RPC results or whole RPC relies.

The Write-Write transfer model is not considered in RFC 5666.

2.3.3. Read-Write

Requesters expose their memory to the responder, but the responder does not expose its memory. The responder employs RDMA Read operations to convey RPC arguments or whole RPC calls. The responder employs RDMA Write operations to convey RPC results or whole RPC relies.

This model is specified in RFC 5666. All known RPC-over-RDMA Version One implementations employ this model. For clarity, the remainder of this document considers only the Read-Write transfer model.

2.4. Upper Layer Binding Specifications

RFC 5666 provides a framework for conveying RPC requests and replies on RDMA transports. By itself this is insufficient to enable an RPC program, referred to as an "Upper Layer Protocol" or ULP, to operate over an RDMA transport.

Arguments and results come in different sizes and have different serialization requirements, all depending on the Upper Layer Protocol. Some arguments and results are appropriate for Direct Data Placement, while others are not. Thus RFC 5666 requires additional separate specifications that describe how each ULP may use explicit RDMA operations to enable Direct Data Placement. The set of requirements for a ULP to use an RDMA transport is known as an "Upper Layer Binding" specification, or ULB.

An Upper Layer Binding states which specific individual RPC arguments and results are permitted to be transferred via RDMA Read and Write for the purpose of Direct Data Placement. RFC 5666 refers to such arguments and results as "RDMA-eligible." RDMA-eligibility restrictions do not apply when a whole RPC call or reply is transferred via an RDMA Read or Write (long messages).

A ULB is required for each RPC program and version tuple that may operate on an RDMA transport. A ULB may be part of another specification, or it may be a stand-alone document, similar to [RFC5667].

2.5. On-The-Wire Protocol

2.5.1. Inline Operation

Each RPC call or reply message conveyed on an RDMA transport starts with an RPC-over-RDMA header, which is encoded into an XDR stream. A requester uses a Send operation to convey the RPC-over-RDMA header to a responder. A responder does likewise to convey RPC replies back to a requester. All message contents sent via Send, including the RPC-over-RDMA header and possibly an RPC message proper, are referred to as "inline content."


   struct rdma_msg {
           uint32    rdma_xid;     /* Mirrors the RPC header xid */
           uint32    rdma_vers;    /* Version of this protocol */
           uint32    rdma_credit;  /* Buffers requested/granted */
           rdma_body rdma_body;


The RPC-over-RDMA header starts with three uint32 fields:


   enum rdma_proc {
           RDMA_MSG=0,   /* An RPC call or reply msg */
           RDMA_NOMSG=1, /* An RPC call or reply msg -
                            separate body */
           . . .
           RDMA_ERROR=4  /* An RPC RDMA encoding error */

   union rdma_body switch (rdma_proc proc) {
           case RDMA_MSG:
             rpc_rdma_header rdma_msg;
           case RDMA_NOMSG:
             rpc_rdma_header_nomsg rdma_nomsg;
           . . .
           case RDMA_ERROR:
             rpc_rdma_error rdma_error;

   struct rpc_rdma_header {
           struct xdr_read_list   *rdma_reads;
           struct xdr_write_list  *rdma_writes;
           struct xdr_write_chunk *rdma_reply;
           /* rpc body follows */

   struct rpc_rdma_header_nomsg {
           struct xdr_read_list   *rdma_reads;
           struct xdr_write_list  *rdma_writes;
           struct xdr_write_chunk *rdma_reply;


Following these three fields is a union:

The difference between these two cases is whether or not the traditional RPC header itself is included in this Send operation (RDMA_MSG), or not (RDMA_NOMSG). In the former case, the RPC header follows immediately after the rdma_reply field. In the latter case, the RPC header is transfered via another mechanism (typically a separate RDMA Read operation).

A requester may use either type of message to send an RPC call message, depending on the requirements of the RPC call message being conveyed. A responder may use RDMA_NOMSG only when the requester provides a Reply chunk (see Section 4.3). A responder is free to use RDMA_MSG instead in that case, depending on the requirements of the RPC reply message.

2.5.2. RDMA Segment

An "RDMA segment", or just "segment", is a part of the RPC-over-RDMA header that contains the co-ordinates of a contiguous memory region that is to be conveyed via an RDMA Read or RDMA Write operation.


   struct xdr_rdma_segment {
           uint32 handle;
           uint32 length;
           uint64 offset;


The region defined by a segment is advertised in an RPC-over-RDMA header to enable the receiving endpoint to drive subsequent RDMA access of the data in that memory region. The RPC-over-RDMA Version One XDR represents an RDMA segment with the xdr_rdma_segment struct: [RFC5040] for a discussion of what the content of these fields means.

2.5.3. Chunk

A "chunk" refers to XDR stream data that is moved via an RDMA Read or Write operation. Chunk data is removed from the sender's XDR stream, is transferred by a separate RDMA operation, and is re-inserted into the XDR stream by the receiver.

Each chunk is made up of one or more segments. Each segment represents a single contiguous piece of that chunk.

If a chunk is to move a whole counted array, the count of array elements is left in the XDR stream, while the array elements appear in the chunk. Individual array elements appear in the chunk in their entirety.

2.5.4. Read Chunk

One or more "read chunks" are used to advertise the coordinates of XDR stream data to be transferred via RDMA Read operations.


   struct xdr_read_chunk {
           uint32 position;
           struct xdr_rdma_segment target;


Each read chunk is represented by the xdr_read_chunk struct:

The transferred data might be contained in one contiguous memory region. That data can be represented by a single read chunk (segment).

Alternately, the transferred data might reside in multiple discontiguous memory regions. The data is represented by a list of read chunks (segments). The Position field in each segment in this list contains the same value.

The receiver reconstructs the transferred data by concatenating the contents of each segment in list order into the receiver's XDR stream. The first segment begins at the XDR position in the Position field, and subsequent segments are concatenated afterwards until there are no more segments left at that XDR Position. This enables gathering data from multiple buffers on the sender.

2.5.5. Write Chunk

A "Write chunk" advertises the coordinates of XDR stream data to be transferred via RDMA Write operations.


   struct xdr_write_chunk {
           struct xdr_rdma_segment target<>;


A write chunk is represented by the xdr_write_chunk struct:

Typically the exact size of the data cannot be predicted before the responder has formed its reply. Thus the requester must provide enough space in the write chunk for the largest result the responder might generate for this RPC operation. The responder updates the size field of each segment in the Write chunk when it returns the Write list to the requester via a matching RPC reply message. If a segment is not used, the responder sets the segment size field to zero.

Because the requester must pre-allocate the area in which the responder writes the result before the responder has formed the reply, giving a position and size to the data, the requester cannot know the XDR stream position of the result data. Thus write chunks do not have a Position field.

2.5.6. Read List

Each RPC-over-RDMA Version One call has one "Read list," provided by the requester. A requester provides the locations of RDMA-eligible argument data via read chunks. Via a Position Zero read chunk, a requester may provide an entire RPC request message as a chunk in this list.


   struct xdr_read_list {
           struct xdr_read_chunk entry;
           struct xdr_read_list  *next;


A Read list is represented by the xdr_read_list struct:

The Read list may be empty if the RPC call has no argument data that is RDMA-eligible and the Position Zero Read chunk is not being used.

2.5.7. Write List

Each RPC-over-RDMA Version One call has one "Write list," provided by the requester. A requester provides write chunks in this list as receptacles for RDMA-eligible result data.


   struct xdr_write_list {
           struct xdr_write_chunk entry;
           struct xdr_write_list  *next;


A Write list is represented by the xdr_write_list struct:

The Write list may be empty if there is no RDMA-eligible result data to return.

The requester provides as many Write chunks as the Upper Layer Binding allows for the particular operation. The responder fills in each Write chunk with an RDMA-eligible result until the Write list is exhausted or there are no more RDMA-eligible results.

2.5.8. Position Zero Read Chunk

A requester may use a "Position Zero read chunk" to convey part or all of an entire RPC call, rather than including the RPC call message inline. A Position Zero read chunk is necessary if the RPC call message is too large to fit inline. RFC 5666 Section 5.1 defines the operation of a "Position Zero read chunk."

To support gathering a large RPC call message from multiple locations on the requester, a Position Zero read chunk may be comprised of more than one xdr_read_chunk. Each read chunk that belongs to the Position Zero read chunk has the value zero in its Position field.

2.5.9. Reply Chunk

Each RPC-over-RDMA Version One call may have one "Reply chunk," provided by the requester. A responder may convey an entire RPC reply message in this chunk.

A Reply chunk is a write chunk, thus it is an array of one or more RDMA segments. This enables a requester to control where the responder scatters the parts of an RPC reply message. In current implementations, there is only one RDMA segment in a Reply chunk.

A requester provides the Reply chunk whenever it predicts the responder's reply might not fit inline. It may choose to provide the Reply chunk even when the responder can return only a small reply. A responder may use a "Reply chunk" to convey most or all of an entire RPC reply, rather than including the RPC reply message inline.

3. Specification Issues

3.1. Extensibility Considerations

RPC-based protocols are defined solely by their XDR definitions. They are independent of the transport mechanism used to convey base RPC messages. Protocols defined this way often have signifcant extensibility restrictions placed on them.

Not all restrictions on RPC-based Upper Layer Protocols may be appropriate for an RPC transport protocol, however. TCP [RFC0793], for example, is an RPC transport protocol that has been extended many times independently of the RPC and XDR standards.

RPC-over-RDMA is partially specified by XDR, and it provides a version field in its headers. However, it is distinct from other RPC- and XDR-based protocols in some key ways:

There are still reasonable restrictions, of course, that are necessary to maintain interoperability within a single Version of RPC-over-RDMA. But they are left largely unstated in RFC 5666.

3.1.1. Recommendations

RFC 5666bis should not alter the basic physical operations that are in use by current implementations. It should not alter the on-the-wire appearance of RPC-over-RDMA Version One headers, and never by an explicit RDMA operations.

Although it is implied, RFC 5666bis should state explicitly that all items in an RPC-over-RDMA Version One header must be conveyed via Send and Receive operations (ie, none of these items is ever RDMA-eligible).

RFC 5666bis should discuss when a Version bump is required. Any significant changes to the way RDMA operations are used should require a Version bump, for instance. Certain limited XDR changes might be allowed, as long as the standards-specified set of wire behaviors remains intact.

3.2. XDR Clarifications

Even seasoned NFS/RDMA implementers have had difficulty agreeing on precisely what a "chunk" is, and had challenges distinguishing the structure of the Read list from structure of the Write list.

On occasion, the text of RFC 5666 uses the term "chunk" to represent either read chunks or write chunks, even though these are different data types and have different semantics.

For example, RFC 5666 Section 3.4 uses the term "chunk list entry" even though the discussion is referring to an array element (a segment). It implies all chunk types have a Position field, even though only read chunks have this field.

Near the end of Section 3.4, it says:

The Read list is not an XDR array, it is always an XDR list. A Write chunk is an XDR array.

RFC 5666 Section 3.7, third paragraph uses the terms "chunked element" and "chunk segment." Neither term is defined or used anywhere else. The fourth paragraph refers to a "sequence of chunks" but likely means a sequence of RDMA segments.

The XDR definition for a read chunk is an RDMA segment with a position field. RFC 5666 Section 3.4 states that multiple xdr_read_chunk structs can make up a single RPC argument if they share the same Position in the XDR stream. Some implementations depend on using multiple RDMA segments in the same XDR Position, particularly for sending Position Zero read chunks efficiently by gathering an RPC call message from multiple discontiguous memory locations. Other implementations do not support sending or receiving multiple Read chunks with the same Position.

An Upper Layer Binding may limit the number of Read list entries allowed for a particular operation. In that case, the Upper Layer Binding is not restricting the total number of read chunks in the list, but rather the number of distinct Positions that appear in the list.

RFC 5666 does not restrict the boundaries of a chunk other than to imply that a chunk's starting position and its length is a multiple of an XDR data unit. However, implementations have observed a practical restriction to facilitate straightforward integration of RDMA support into existing XDR infrastructure: A chunk containing RDMA-eligible data must be encoded or decoded as a single XDR object.

In addition, Upper Layer Bindings make RDMA-eligibility statements about specific arguments and results (or portions thereof which still are whole XDR objects). The implication is that chunks contain only whole XDR objects, even though RFC 5666 is not explicit about this.

A Position Zero read chunk typically contains an entire RPC request message, and a Reply chunk contains an entire RPC reply message. These are exceptions to the above restriction.

The Write list is especially confusing because it is a list of arrays of RDMA segments, rather than a simple list of xdr_read_chunk objects. What is referred to as a Read list entry often means one xdr_read_chunk, or one segment. That segment can be either a portion of or a whole XDR object. A Write list entry is an array, and always represents a single XDR object in its entirety.

An Upper Layer Binding may limit the number of chunks in a Write list allowed for a particular operation. That strictly limits the number of Write list entries.

Not having a firm one-to-one correspondence between read chunks and XDR objects is sometimes awkward. The two chunk types should be more symmetrical to avoid confusion, although that might be difficult to pull off without altering the RPC-over-RDMA Version One XDR definition. As we will see later, the XDR roundup rules also appear to apply asymmetrically to read chunks and write chunks.


   * Encoding key for single-list chunks
   *         (HLOO = Handle32 Length32 Offset64):
   *  Read chunklist (a linked list):
   *   N elements, position P (same P for all chunks of same arg!):
   *    1 - PHLOO - 1 - PHLOO - ... - 1 - PHLOO - 0
   *  Write chunklist (a list of (one) counted array):
   *   N elements:
   *    1 - N - HLOO - HLOO - ... - HLOO - 0
   *  Reply chunk (a counted array):
   *   N elements:
   *    1 - N - HLOO - HLOO - ... - HLOO


Implementers have been aided by the ASCII art block comments in the Linux kernel in net/sunrpc/xprtrdma/rpcrdma.c, excerpted here. This diagram shows exactly how the Read list and Write list are constructed in an XDR stream.

3.2.1. Recommendations

To aid in reader understanding, RFC 5666bis should expand the glossary that explains and distinguishes the various elements in the protocol. Upper Layer Binding specifications refer to these terms. RFC 5666bis should utilize and capitalize these glossary terms consistently.

RFC 5666bis should introduce additional diagrams that supplement the XDR definition in RFC 5666 Section 4.3. RFC 5666bis should explain the structure of the XDR and how it is used. RFC 5666bis should contain an explicit but brief rationalization for the structural differences between the Read list and the Write list.

RFC 5666bis should explicitly restrict chunks containing RDMA-eligible data so that a chunk represents exactly a single XDR object in its entirety.

RFC 5666bis should use a consistent naming convention for all XDR definitions. For example, all structures and union names should use an "rpcrdma1_" prefix.

To address conflation of a read chunk that is a single xdr_read_chunk and a read chunk that is a list of xdr_read_chunk elements with identical Position field values, the following specification changes should be made:

With these changes, there would no longer be a simple XDR object that explicitly represents a read chunk, but a read chunk and a write chunk are now equivalent objects that both map to a whole XDR object. All discussion should take care to use the terms "segment" and "read segment" instead of the term "read chunk" where appropriate.

As a clean up, RFC 5666bis should remove the rpc_rdma_header_nomsg struct, and use the rpc_rdma_header struct in its place. Since rpc_rdma_header does not comprise the entire RPC-over-RDMA header, it should be renamed rpcrdma1_chunks to avoid confusion.

XDR definitions should be enclosed in CODE BEGINS and CODE ENDS delimiters. An appropriate copyright block should accompany the XDR definitions in RFC 5666bis. An XDR extraction shell script should be provided in the text.

See Section 10 for a full listing of the proposed XDR definitions.

3.3. Additional XDR Issues

3.3.1. Mechanical Issues

There are some mechanical problems with the XDR language definition of RPC-over-RDMA Version One provided in Section 4.3 of [RFC5666]:

3.3.2. XDR Definition Recursiveness

The usual practice when defining an XDR-based protocol is that there is one encompassing data type that represents one message in the protocol.

This is not true for RPC-over-RDMA. The header is defined by one data type (struct rdma_msg) but the RPC message payload is not formally represented in the XDR definition in Section 4.3. The presence or absence of the RPC message payload is indicated by the message type, and the body of that payload is noted only with a code comment.

3.3.3. Recommendations

The XDR presented in RFC5666bis should correct the deficiencies described above.

To correct the lack of formal recursiveness issue without forcing an on-the-wire behavior change, RFC5666bis should place the RPC-over-RDMA header and the RPC message payload in separate XDR streams.

3.4. The Position Zero Read Chunk

RFC 5666 Section 5.1 defines the operation of the Position Zero read chunk. A requester uses the Position Zero read chunk in place of inline content. A requester is required to use the Position Zero read chunk when the total size of an RPC call message exceeds the size of the responder's receive buffers, and RDMA-eligible data has already been removed from the message.

RFC 5666 Section 3.4 says:

The intention of this text is to spell out that RDMA-eligibility applies only to individual XDR data objects in the Upper Layer Protocol. RDMA-eligibility criteria are specified within a separate specification, rather than in RFC 5666.

The Position Zero read chunk is an exception to both of these guidelines. The Position Zero read chunk, by virtue of the fact that it typically conveys an entire RPC call message, may contain multiple arguments, independent of whether any particular argument in the RPC call is RDMA-eligible.

Unlike the read chunks described in the RFC 5666 excerpt above, the content of a Position Zero read chunk is typically marshaled and copied on both ends of the transport, so it cannot benefit from Direct Data Placement. In particular, the Position Zero read chunk is not for conveying performance critical Upper Layer operations.

Thus the requirements for what may or may not appear in the Position Zero read chunk are indeed specified by RFC 5666, in contradiction to the second paragraph quoted above. Upper Layer Binding specifications may have something to say about what may appear in the Position Zero read chunk, but the basic definition of Position Zero should be made clear in RFC 5666bis as distinct from a read chunk whose Position field is non-zero.

Because a read chunk is defined as one RDMA segment with a Position field, at least one implementation allows only a single chunk segment in Position zero read chunks. This is a problem for two reasons:

RFC 5666 does not limit the number of read segments in a read chunk, nor does it limit the number of chunks that can appear in the Read list. The Position Zero read chunk, despite its name, is not limited to a single xdr_read_chunk.

3.4.1. Recommendations

RFC 5666bis should state that the guidelines in RFC 5666 Section 3.4 apply only to RDMA_MSG type calls. When the Position Zero read chunk is introduced in RFC 5666 Section 5.1, enumerate the differences between it and the read chunks previously described in RFC 5666 Section 3.4.

RFC 5666bis should describe what restrictions an Upper Layer Binding may make on Position Zero read chunks.

3.5. RDMA_NOMSG Call Messages

The second paragraph of RFC 5667 Section 4 says, in reference to NFSv2 and NFSv3 WRITE and SYMLINK operations:

However, large non-write NFS operations are conveyed via a Read list containing at least a Position Zero read chunk. Strictly speaking, the above requirement means large non-write NFS operations may never be conveyed because the responder MUST ignore the read chunk in such requests.

It is likely the authors of RFC 5667 intended this limit to apply only to RDMA_MSG type calls. If that is true, however, an NFS implementation could legally skirt the stated restriction simply by using an RDMA_NOMSG type call that conveys both a Position Zero and a non-zero position read chunk to send a non-write NFS operation.

Unless either RFC 5666 or the protocol's Upper Layer Binding explicitly prohibits it, allowing a read chunk in a non-zero Position in an RDMA_NOMSG type call means an Upper Layer Protocol may ignore Binding requirements like the above.

Typically there is no benefit to allowing multiple read chunks for RDMA_NOMSG type calls. Any non-zero Position read segments can always be conveyed as part of the Position Zero read chunk.

However, there is a class of RPC operations where RDMA_NOMSG with multiple read chunks is useful: when the body of an RPC call message is larger than the inline buffer size, even after RDMA-eligible argument data has been moved to read chunks.

A similar discussion applies to RDMA_NOMSG replies with large reply bodies and RDMA-eligible result data. Such replies would use both the Write list and the Reply chunk simultaneously. However, write chunks do not have Position fields.

3.5.1. Recommendations

RFC 5666bis should continue to allow RDMA_NOMSG type calls with additional read chunks. The rules about RDMA-eligibility in RFC 5666bis should discuss when the use of this construction is beneficial, and when it should be avoided.

Authors of Upper Layer Bindings should be warned about ignoring these cases. RPC 5666bis should provide a default behavior that applies when Upper Layer Bindings omit this discussion.

3.6. RDMA_MSG Call with Position Zero Read Chunk

The first item in the header of both RPC calls and RPC replies is the XID field [RFC5531]. RFC 5666 Section 4.1 says:

This is a strong implication that the RPC header in an RDMA_MSG type message starts at XDR position zero. Assume for a moment that, by definition, the RPC header in an RPC-over-RDMA XDR stream starts at XDR position zero.

An RDMA_MSG type call message includes the RPC header and zero or more read chunks. Recall the definition of a read chunk as a list of read segments whose Position field contains the same value. The value of the Position field determines where the read chunk appears in the XDR stream that comprises an RPC call message.

A Position Zero read chunk, therefore, starts at XDR position zero, just like RPC header does. In an RDMA_NOMSG type call message, which does not include an RPC header, a Position Zero read chunk conveys the RPC header.

There is no prohibition in RFC 5666 against an RDMA_MSG type call messsage with a Position Zero read chunk. However, it's not clear how a responder should interpret such a message. RFC 5666 requires the RPC header to start at XDR position zero, but there is a Position Zero read chunk, which also starts at XDR position zero.

3.6.1. Recommendations

RPC 5666bis should clearly define what is meant by an XDR stream. RFC 5666bis should state that the value in the xdr_read_chunk "position" field is measured relative to the start of the RPC header, which is the first byte of the header's XID field.

RFC 5666bis should prohibit requesters from providing a Position Zero read chunk in RDMA_MSG type calls. Likewise, RFC 5666bis should prohibit responders from utilizing a Reply chunk in RDMA_MSG type replies.

The diagrams in RFC 5666 Section 3.8 which number chunks starting with 1 should be revised. Readers confuse this number with an XDR position.

3.7. Padding Inline Content After A Chunk

To help clarify the discussion in this section, the term "read chunk" here always means the new definition where one or more read segments that have identical values in their Position fields represents exactly one RDMA-eligible XDR object.

A read chunk conveys a large argument payload via one or more RDMA transfers. For instance, the data payload of an NFS WRITE operation may be be transferred using a read chunk [RFC5667].

NFSv3 WRITE operations place the data payload at the end of an RPC call message [RFC1813]. The RPC call's XDR stream starts in an inline buffer, continues in a read chunk, then ends there.

An NFSv4 WRITE operation may occur as a middle operation in an NFSv4 COMPOUND [RFC5661]. The read chunk containing the data payload argument of the WRITE operation might finish before the RPC call's XDR stream does. In this case, the RPC call's XDR stream starts in an inline buffer, continues in the Read list, then finishes back in the inline buffer.

The length of a chunk is the sum of the lengths of the segments that make up that chunk. The data payload in a chunk may have a length that is not evenly divisible by four. One or more of the segments may have an unaligned length.

RFC 5666 Section 3.7 describes how to manage XDR roundup in a read chunk when its length is not XDR-aligned. The sender is not required to send the extra pad bytes at the end of a chunk because a) the receiver never references their content, therefore it is wasteful to transmit them, and b) each read chunk has a Position field and length that determines exactly where that chunk starts and ends in the XDR stream.

A question arises, however, when considering where the next XDR object after a read chunk should appear. XDR requires each object to begin on 4-byte alignment [RFC4506]. But a read chunk's XDR padding is optional (see above) and thus may not appear in the chunk as actual zero bytes.

The next read chunk's position field determines where it is placed in the XDR stream, so in that case there is no ambiguity. Inline content following a read chunk does not have a Position field to guide the receiver in the reassembly of the XDR stream, however.

Paragraph 4 of RFC 5666 Section 3.7 says:

This paragraph adequately describes XDR padding requirements when a read chunk is followed by another read chunk. But it leaves unspoken any requirements for XDR padding and alignment when a read chunk is followed in the XDR stream by more inline content.

Applying the rules of XDR, the XDR pad for the read chunk must not appear in the inline content, even if it was also not included in the chunk itself. This is because the inline content that preceded the read chunk will have been padded to 4-byte alignment. The next position in the inline buffer is already on a 4-byte boundary, thus no padding is necessary.

3.7.1. Recommendations

State the above requirement in RFC 5666bis in its equivalent of RFC 5666 Section 3.7. When a responder forms a reply, the same restriction applies to inline content interleaved with write chunks.

Because all XDR objects must start on an XDR alignment boundary, all read and write chunks and all inline XDR objects in any XDR stream must start on an XDR alignment boundary. This has implications for the values allowed in read chunk Position fields, for how XDR roundup works for chunks, and for how XDR objects are placed in inline buffers. XDR alignment in inline buffers is always relative to Position Zero (or, where the RPC header starts).

3.8. Write Chunk XDR Roundup

The final paragraph of RFC 5666 Section 3.7 says:

A responder should avoid writing XDR pad bytes, as the requester's upper layer does not reference them, though the language does not fully prohibit writing these bytes. A requester always provides the extra space for XDR padding anyway.

A problem arises if the data item written into a Write chunk is shorter than the chunk and requires an XDR pad. A responder may write the XDR pad past the end of the data content. For a short directly-placed write, the pad bytes are then exposed in the RPC consumer's data buffer.

In addition, for the chunk length to be rounded up as described, the requester must provide adequate extra space in the chunk for the XDR pad. A requester can provide space for the XDR pad using one of two approaches:

  1. It can extend the last segment in the chunk.
  2. It can provide another segment after the segments that receive RDMA Write payloads.

Case 1 is adequate when there is no danger that the responder's RDMA Write operations will overwrite existing data on the requester in memory following the advertised receive buffers.

In Direct Data Placement scenarios, an extra segment must be provided separately to avoid overwriting existing data that follows the sink buffer (case 2). Thus, an extra registration is needed for just a handful of bytes that may not be written by the responder, and are ignored by the requester. Even so, this does not force the responder to direct the XDR pad bytes into this extra segment, should the data item in that chunk be shorter than the chunk itself.

Registering the extra buffer is a needless cost. It would be more efficient if the XDR pad at the end of a write chunk were treated the same as it is for Read chunks. Because RPC result data must begin on an XDR alignment boundary, the result following the write chunk in the reply's XDR stream must begin on an XDR alignment boundary. There is no need for a XDR pad to be present for the receiver to re-assemble the RPC reply's XDR stream properly.

One responder implementation requires the requester to provide the extra buffer space in the Write chunk, but does not write to it. This follows the letter of the last paragraph of Section 3.7 of [RFC5666].

Another responder implementation does not rely on having the extra space (operation proceeds if it is missing) but when the extra space is present, this responder does write zeroes to it. While the intention of Section 3.7 is that the responder does not write the pad, it is not strictly forbidden.

Client implementations all appear to provide the extra buffer space needed to accommodate the XDR pad. However, one implementation does not register this extra buffer, since the responder is not expected to write into it, while another implementation does.

These implementations may not be 100% interoperable. The language of Section 3.7 of [RFC5666] appears to allow all of this behavior (in particular, it does not prohibit a responder from writing the XDR pad using RFC2119-style keywords, and does not require that requesters register the extra space to accommodate the XDR pad).

Note that because the Reply chunk is a write chunk, these roundup rules also apply to it.

3.8.1. Recommendations

The current specification allows XDR pad bytes to leak into user buffers, and none of the current implementations prevent this leak. There may be room to adjust the protocol specification independently of current implementation behavior.

RFC 5666bis should explicitly discuss the requirements around write chunk roundup separately from the discussion of read chunk roundup.

Explicit RFC2119-style interoperability requirements should be provided for write chunks. Responders MUST NOT write XDR pad bytes at the end of a Write chunk.

Allocating and registering extra space for XDR pad bytes that are never written is wasteful. RFC 5666bis should forbid it. Responders should not expect requesters to provide space for XDR pad bytes.

3.9. Write List Error Cases

RFC 5666 Section 3.6 says:

This requires the responder to use the Write list when it is provided. Another way to say it is a responder is not permitted to return bulk data inline or in the Reply chunk when the requester has provided a Write list.


   struct READ3resok {
           post_op_attr file_attributes;
           count3       count;
           bool         eof;
           opaque       data<>;

   struct READ3resfail {
           post_op_attr file_attributes;

   union READ3res switch (nfsstat3 status) {
   case NFS3_OK:
           READ3resok resok;
           READ3resfail resfail;


This requirement is less clear when it comes to situations where a particular RPC reply is allowed to use a provided Write list, but does not have a bulk data payload to return. For example, RFC 5667 Section 4 permits requester to provide a Write list for NFS READ operations. However, NFSv3 READ operations have a union reply [RFC1813]:

RFC 5666 does not prescribe how a responder should behave when RDMA-eligible result data for which the Write list is provided does not appear in the reply. RFC 5666 Section 3.4 says:

It also says:

The disposition of the advertised write buffers is therefore clear. The requirements for how the Write list must appear in an RPC reply are somewhat less than clear.

Here we are concerned with two cases:

The language above suggests the proper value for the Write chunk's segment count is always the same value that the requester sent, even when the chunk is not used in the reply. The proper value for the length of an unused segment in a Write chunk is always zero.

Inspection of one existing server implementation shows that when an NFS READ operation fails, the returned Write list contains one entry: a chunk array containing zero elements. Another server implementation returns the original Write list chunk in this case.

In either case, requesters appear to ignore the Write list when no bulk data payload is expected. Thus it appears that, currently, responders may put whatever they like in the Write list.

Current NFSv4 client implementations behave like legacy NFS implementations in the sense that each READ COMPOUND requests only one contiguous data payload that is never larger than the rsize setting of the mount. However it is legal for an NFSv4 COMPOUND to contain more than one READ operation. Each READ request in a COMPOUND may have an RDMA-eligible result in the COMPOUND reply.

In general, a complex Upper Layer Binding may wish to return more than one RDMA-eligible result in a single RPC reply. Depending on the RPC program, there may be nested or sequential switched unions in the reply. There is no Position field in the segments making up a Write chunk, so both sender and receiver must be careful about how the reply message is re-assembled.

It should always be unambiguous which Write chunk matches with which result. To ensure interoperability, the responder associates the first RDMA-eligible result with the first chunk in the Write list, and so on, until either results or Write chunks are exhausted. The receiver makes the same associations while parsing the XDR stream of the reply. It should be the responsibility of the Upper Layer Binding to avoid ambiguous situations by appropriately restricting RDMA-eligible data items.

Remember that a responder MUST use the Write list if the requester provided it and the responder has RDMA-eligible result data. If the requester has not provided enough Write chunks in the Write list, the responder may have to use a long message as well, depending on the remaining size of the RPC reply.

3.9.1. Recommendations

RFC 5666bis should explicitly discuss responder behavior when an RPC reply does not need to use a Write list entry provided by a requester. This is generic behavior, independent of any Upper Layer Binding. The explanation can be partially or wholly copied from RFC 5667 Section 5's discussion of NFSv4 COMPOUND.

A number of places in RFC 5666 Section 3.6 hint at how a responder behaves when it is to return data that does not use every byte of every provided Write chunk segment. RFC 5666bis should state specific requirements about how a responder should form the Write list in RPC replies, and/or it should explicitly require requesters to ignore the Write list in these cases. RFC 5666bis should require that the responder not alter the count of segments in the Write chunk. One or more explicit examples should be provided in RFC 5666bis.

RFC 5666bis should provide clear instructions on how Upper Layer Bindings are to be written to take care of switched unions.

4. Operational Considerations

4.1. Computing Request Buffer Requirements

The size maximum of a single Send operation includes both the RPC-over-RDMA header and the RPC header. Combined, those two headers must not exceed the size of one receive buffer.

Senders often construct the RPC-over-RDMA header and the RPC call or reply message in separate buffers, then combine them via an iovec into a single Send. This does not mean each element of that iovec can be as large as the inline threshold.

An HCA or RNIC may have a small limit on the size of a registered memory region. In that case, RDMA-eligible data may be comprised of many chunk segments.

This has implications for the size of the Read and Write lists, which take up a variable amount of space in the RPC-over-RDMA header. The sum of the size of the RPC-over-RDMA header, including the Read and Write lists, and the size of the RPC header must not exceed the inline threshold. This limits the maximum Upper Layer payload size.

4.1.1. Recommendations

RFC 5666bis should provide implementation guidance on how the inline threshold (the maximum send size) is computed.

4.2. Default Inline Buffer Size

Section 6 of RFC 5666 specifies an out-of-band protocol that allows an endpoint to discover a peer endpoint's receive buffer size, to avoid overrunning the receiving buffer, causing a connection loss.

Not all RPC-over-RDMA Version One implementations also implement CCP, as it is optional. Given the importance of knowing the receiving end's receive buffer size, there should be some way that a sender can choose a size that is guaranteed to work with no CCP interaction.

RFC 5666 Section 6.1 describes a 1KB receive buffer limit for the first operation on a connection with an unfamiliar responder. In the absence of CCP, the client cannot discover that responder's true limit without risking the loss of the transport connection.

4.2.1. Recommendations

RFC 5666bis should specify a fixed send/receive buffer size as part of the RPC-over-RDMA Version One protocol, to use when CCP is not available. For example, the following could be added to the RFC 5666bis equivalent of RFC 5666 Section 6.1: "In the absence of CCP, requesters and responders MUST assume 1KB receive buffers for all Send operations."

It should be safe for Upper Layer Binding specifications to provide a different default inline threshold. Care must be taken when an endpoint is associated with multiple RPC programs that have different default inline thresholds.

4.3. When To Use Reply Chunks

RFC 5666 Section 3.6 says:

It is not clear whether the authors of RFC 5666 intended the above requirement to apply only to the Write list, or to both the Write list and to the Reply chunk, which is not a list.

Implementation experience has shown that setting up an explicit RDMA operation to move a few hundred bytes of data is inefficient, especially if there is no DDP opportunity. Channel operations are nearly always the best choice when handling a small RPC reply.

If a requester does not provide a Reply chunk when one is needed, the responder must reply with ERR_CHUNK (see RFC 5666, Section 4.2). The requester simply has to send the request again, this time with a Reply chunk. However ERR_CHUNK a generic failure mode. The requester may have some difficulty identifying the problem as a missing Reply chunk.

To maintain 100% interoperability, a requester should always provide a Reply chunk, and the responder should always use it. However, as noted, this is likely to be inefficient.

4.3.1. Recommendations

To provide a stronger guarantee of interoperation while ensuring efficient operation, RFC 5666bis should explicitly specify when a requester must offer a Reply chunk, and when the responder must use an offered Reply chunk.

Mandating a default buffer size would allow both sides to choose correctly with an in-advance CCP exchange.

4.4. Computing Credit Values

The third paragraph of Section 3.3 of RFC 5666 leaves open the exact mechanism of how often the requested and granted credit limits are supposed to be adjusted. A reader might believe that these values are adjusted whenever an RPC call or reply is received, to reflect the number of posted receive buffers on each side.

Although adjustments are allowed by RFC 5666 due to changing availability of resources on either endpoint, current implementations use a fixed value. Advertised credit values are always the sum of the in-process receive buffers and the ready-to-use receive buffers.

4.4.1. Recommendations

RFC 5666bis should clarify the method used to calculate these values. RFC 5666bis might also discuss how flow control is impacted when a server endpoint utilizes a shared receive queue.

4.5. Race Windows

The second paragraph of RFC 5666 Section 3.3 says:

It is true that the RPC server must always be able to reply, and that therefore the client must provide an adequate number of receive buffers. The dependent clause "whether or not new buffers have been posted to accept future receives" is problematic, however.

It's not clear whether this clause refers to a server leaving even a small window where the sum of posted and in-process receive buffers is less than the credit limit; or refers to a client leaving a window where the sum of posted and in-process receive buffers is less than its advertised credit limit. In either case, such a window could result in lost messages or be catastrophic for the transport connection.

4.5.1. Recommendations

Clarify or remove the dependent clause in the section in RFC 5666bis that is equivalent to RFC 5666 Section 3.3.

4.6. Detection Of Unsupported Protocol Versions

Section 4.2 of [RFC5666] is explicit about how a responder must handle RPC-over-RDMA messages that carry an unrecognized RPC-over-RDMA protocol version:

However, one widely deployed RPC-over-RDMA Version One server implementation is known to discard requests that do not contain the value one (1) in their rdma_vers field. This server implementation does not reply with RDMA_ERROR / RDMA_ERR_VERS in this case.

Without a proper protocol version detection mechanism, it is not possible for RPC-over-RDMA Version One implementations to interoperate with implementations that support newer protocol versions.

4.6.1. Recommendations

RPC-over-RDMA Version One implementations that discard non-Version One requests without an error response are considered non-compliant with [RFC5666]. No changes to the specification are needed.

5. Pre-requisites For NFSv4

5.1. Bi-directional Operation

NFSv4.1 moves the backchannel onto the same transport as forward requests [RFC5661]. Typically RPC client endpoints do not expect to receive RPC call messages. To support NFSv4.1 callback operations, client and server implementations must be updated to support bi-directional operation.

Because of RDMA's requirement to pre-post unadvertised receive buffers, special considerations are needed for bi-directional operation. Conventions have been provided to allow bi-direction, with a limit on backchannel message size, such that no changes to the RPC-over-RDMA Version One protocol are needed [I-D.ietf-nfsv4-rpcrdma-bidirection].

5.1.1. Recommendations

RFC 5666bis should cite or include the bulk of [I-D.ietf-nfsv4-rpcrdma-bidirection].

6. Considerations For Upper Layer Binding Specifications

RFC 5666 requires a Binding specification for any RPC program wanting to use RPC-over-RDMA. The requirement appears in two separate places: The fourth paragraph of Section 3.4, and the final paragraph of Section 3.6. As critical as it is to have a Binding specification, RFC 5666's text regarding these specifications is sparse and not easy to find.

6.1. Organization Of Binding Specification Requirements

Throughout RPC 5666, various Binding requirements appear, such as:

A similar specific requirement for read list entries is missing.

Usually these statements are followed by a reference to the NFS Binding specification [RFC5667]. There is no summary of these requirements, however.

Additional advice appears in the middle of Section 3.4:

This requirement, being in the middle of a dense paragraph about how write lists are formed, is easy for an author of Upper Layer Binding specifications to miss.

6.1.1. Recommendations

RFC 5666bis should summarize explicit generic requirements for the contents of an Upper Layer Binding specification in one separate section, perhaps in an Appendix. In particular, move the third, fourth and fifth paragraph of RFC 5666 Section 3.4 to this new section discussing Binding specification requirements.

6.2. RDMA-Eligibility

Any RPC message that fits in an inline buffer is conveyed via a Send operation. Any RPC message that is too large to fit in an inline buffer is conveyed by transferring the whole RPC message via an RDMA Read (i.e., a Position Zero Read chunk) or an RDMA Write (i.e., a Reply chunk).

RPC-over-RDMA also allows a mixture of these two mechanisms, where argument or result data is removed from the XDR stream and conveyed via a separate RDMA transfer. The receiving end assembles the disparate buffers into a single XDR stream that represents the whole RPC message.

RFC 5666 uses the term "RDMA eligibility" to mean that an particular argument or result object is allowed to be moved as a separate chunk for the purpose of Direct Data Placement. The RPC program's Upper Layer Binding makes eligibility statements permitting particular RPC argument or result objects to be directly placed.

The third paragraph of Section 3.4 states that any XDR object MAY be RDMA-eligible in any given message, but that:

Any large XDR object that can benefit from Direct Data Placement is a good candidate for being moved in a chunk. When data alignment matters, or when the NFS stack on either end of the connection does not need to manipulate the transferred data, the Upper Layer Binding should make that object eligible for Direct Data Placement.

Section 3.4 is specifically not discussing long messages, where a whole RPC message is moved via RDMA. When an RPC message is too large to fit inline, even after RDMA-eligible arguments or results are removed, the message is always moved via a long message. All arguments or results in the message are moved via RDMA in this case.

For instance, an NFSv3 READDIR result can be large. However, an NFS server assembles this result in place, encoding each section individually. The NFS client must perform the converse actions. Though there is potentially a large amount of data, the benefit of direct data placement is lost because of the need for both host CPUs to be involved in marshaling and decoding.

Thus the NFSv3 Upper Layer Binding [RFC5667] does not make any part of an NFSv3 READDIR reply RDMA-eligible. However, any NFS READDIR reply that is larger than an inline buffer is still moved via RDMA (a Reply chunk, in this case).

6.2.1. Recommendations

RFC 5666bis should define the term "Upper Layer Binding", and explain what it specifies. RFC 5666bis should explicitly require an Upper Layer Binding for every RPC program that may operate on RDMA transports. Separate bindings may be required for different versions of that program.

The term "RDMA eligibility" should be retired. It is easy to confuse the use of RDMA for Direct Data Placement with the use of RDMA in long messages. Instead, RFC 5666bis should use a more precise term such as DDP-eligibility, which should be clearly defined before it is used.

RFC 5666bis should provide generic guidance about what makes an XDR object or data type eligible for Direct Data Placement. RFC 5666bis should state that the DDP-eligibility of any XDR object not mentioned explicitly in an Upper Layer Binding is "not eligible."

RFC 5666bis should note that Position Zero read chunks and Reply chunks may contain any and all argument and results regardless of their DDP-eligibility. RFC 5666bis should remind authors of Upper Layer Bindings that the Reply chunk and Position Zero read chunks are expressly not for performance-critical Upper Layer operations.

It is the responsibility of the Upper Layer Binding to specify RDMA-eligibity rules so that if an RDMA-eligible XDR object is embedded within another, only one of these two objects is to be represented by a chunk. This ensures that the mapping from XDR position to the XDR object represented is unambiguous.

6.3. Inline Threshold Requirements

An RPC-over-RDMA connection has two connection parameters that affect the operation of Upper Layer Protocols: The credit limit, which is how many outstanding RPCs are allowed on that connection; and the inline threshold, which is the maximum payload size of an RDMA Send on that connection. All ULPs sharing a connection also share the same credits and inline threshold values.

The inline threshold is set when a connection is established. The base RPC-over-RDMA protocol does not provide a mechanism for altering the inline threshold of a connection once it has been established.

[RFC5667] places normative requirements on the inline threshold value for a connection. There is no guidance provided on how implementations should behave when two ULPs that have different inline threshold requirements share the same connection.

Further, current NFS implementations ignore the inline threshold requirements stated in [RFC5667]. It is unlikely that they would interoperate successfully with any new implementation that followed the letter of [RFC5667].

6.3.1. Recommendations

Upper Layer Protocols should be able to operate no matter what inline threshold is in use.

An Upper Layer Binding might provide informative guidance about optimal values of an inline threshold, but normative requirements are difficult to enforce unless connection sharing is explicitly not permitted.

6.4. Violations Of Binding Rules

Section 3.4 of RFC 5666 introduces the idea of an Upper Layer Binding specification to state which Upper Layer operations are allowed to use explicit RDMA to transfer a bulk payload item.

The fifth paragraph of this section states:

If the Upper Layer on a receiver is not aware of the presence and operation of an RPC-over-RDMA transport under it, it could be challenging to discover when a sender has violated an Upper Layer Binding rule.

If a violation does occur, RFC 5666 does not define an unambiguous mechanism for reporting the violation. The violation of Binding rules is an Upper Layer Protocol issue, but it is likely that there is nothing the Upper Layer can do but reply with the equivalent of BAD XDR.

When an erroneously-constructed reply reaches a requester, there is no recourse but to drop the reply, and perhaps the transport connection as well.

6.4.1. Recommendations

Policing DDP-eligibility must be done in co-operation with the Upper Layer Protocol by its receive endpoint implementation.

It is the Upper Layer Binding's responsibility to specify how a responder must reply if a requester violates a DDP-eligibilty rule. The Binding specification should provide similar guidance for requesters about handling invalid RPC-over-RDMA replies.

6.5. Binding Specification Completion Assessment

RFC 5666 Section 3.4 states:

Authors of Upper Layer Binding specifications should consider each data type in the Upper Layer's XDR definition, in particular compound types such as arrays and lists, when restricting what XDR objects are eligible for Direct Data Placement.

In addition, there are requirements related to using NFS with RPC-over-RDMA in [RFC5667], and there are some in [RFC5661]. It could be helpful to have guidance about what kind of requirements belong in an Upper Layer Binding specification versus what belong in the Upper Layer Protocol specification.

6.5.1. Recommendations

RFC 5666bis should describe what makes a Binding specification complete (i.e. ready for publication).

7. Unimplemented Protocol Features

There are features of RPC-over-RDMA Version One that remain unimplemented in current implementations. Some are candidates to be removed from the protocol because they have proven unnecessary or were not properly specified.

Other features are unimplemented, unspecified, or have only one implementation (thus interoperability remains unproven). These are candidates to be retained and properly specified.

7.1. Unimplemented Features To Be Removed

7.1.1. Connection Configuration Protocol

No implementation has seen fit to support the Connection Configuration Protocol. While a need to exchange pertinent connection information remains, the preference is to exchange that information as part of the set up of each connection, rather than as settings that apply to all connections (and thus all ULPs) between two peers. Recommendations

CCP should be removed from RFC 5666bis.

7.1.2. Read-Read Transfer Model

All existing RPC-over-RDMA Version One implementations use a Read-Write data transfer model. The server endpoint is responsible for initiating all RDMA data transfers. The Read-Read transfer model has been deprecated, but because it appears in RFC 5666, implementations are still responsible for supporting it. By removing the specification and discussion of Read-Read, the protocol and specification can be made simpler and more clear.

Once the Read-Read transfer model is no longer supported, a responder would no longer be allowed to send a Read list to a requester. Sending a Read list would be needed if a requester has not provided enough memory space in the form of a Reply chunk or Write list to receive a large RPC Reply.

There is currently no mechanism in the RPC-over-RDMA Version One protocol for a responder to indicate that inadequate reply buffer resources were provided by a requester. Therefore, requesters should be fully responsible for providing all necessary memory resources to receive each RPC reply, including a properly populated Write list and/or a Reply chunk. Recommendations

Remove Read-Read from RFC 5666bis, in particular from its equivalent of RFC 5666 Section 3.8. RFC 5666bis should require implementations not to send RDMA_DONE; an implementation receiving it should ignore it. The XDR definition should reserve RDMA_DONE. RFC 5666bis should explicitly state requirements for requesters to allocate and prepare reply buffer resources for each RPC-over-RDMA message.

7.1.3. RDMA_MSGP

It has been observed that the current specification of RDMA_MSGP is not clear enough to result in interoperable implementations. Possibly as a result, current receive endpoints do recognize and process RDMA_MSGP messages, though they do not take advantage of the passed alignment parameters. Receivers treat RDMA_MSGP messages like RDMA_MSG messages.

Currently senders do not use RDMA_MSGP messages. RDMA_MSGP depends on bulk payload occurring at the end of RPC messages, which is often not true of NFSv4 COMPOUND requests. Most NFSv3 requests are small enough not to need RDMA_MSGP.

To be effective, RDMA_MSGP depends on getting alignment preferences in advance via CCP. There are no CCP implementations to date. Without CCP, there is no way for peers to discover a receiver endpoint's preferred alignment parameters, unless the implementation provides an administrative interface for specifying a remote's alignment parameters. RDMA_MSGP is useless without that knowledge. Recommendations

To maintain backward-compatibility, RDMA_MSGP must remain in the protocol. RFC 5666bis should require implementations to not send RDMA_MSGP messages. If an RDMA_MSGP message is seen by a receiver, it should ignore the alignment parameters and treat RDMA_MSGP messages as RDMA_MSG messages. The XDR definition should reserve RDMA_MSGP.

7.2. Unimplemented Features To Be Retained

7.2.1. RDMA_ERROR Type Messages

Server implementations the author is familiar with can send RDMA_ERROR type messages, but only when an RPC-over-RDMA version mismatch occurs. There is no facility to return the ERR_CHUNK error. These implementations treat unrecognized message types and other parsing errors as an RDMA_MSG type message. Obviously this behavior does not comply with RFC 5666, but it is also recognized that this behavior is not an improvement over the specification. Recommendations

RFC 5666bis should provide stronger guidance for error checking, and in particular, when a connection must be broken.

Implementations that do not adequately check incoming RPC-over-RDMA headers must be updated.

7.2.2. RPCSEC_GSS On RPC-over-RDMA

The second paragraph of RFC 5666 Section 11 says:

However, consider:

Therefore, despite its performance impact, RPCSEC_GSS can add important function to RPC-over-RDMA deployments.

Currently there is an InfiniBand-only client and server implementation of RPCSEC_GSS on RPC-over-RDMA that supports the authentication, integrity, and privacy services. This pair of implementations was created without the benefit of normative guidance from RFC 5666. This client and server pair interoperates with each other, but there are no independent implementations to test with.

RPC-over-RDMA requesters are responsible for providing adequate reply resources to responders. These resources require special treatment when an integrity or privacy service is in use. Direct data placement cannot be used with software integrity checking or encryption. Thus standards guidance is imperative to ensure that independent RPCSEC_GSS implementations can interoperate on RPC-over-RDMA transports. Recommendations

RFC 5666bis should continue to require the use of link layer protection when facilities are available to support it.

At the least, RPCSEC_GSS per-message authentiction is valuable, even if link layer protection is in use. Integrity and privacy should also be made available even if they do not perform well, because there is no link layer protection for some fabrics.

Therefore, RFC 5666bis should provide a specification for RPCSEC_GSS on RPC-over-RDMA, codifying the one existing implementation so that others may interoperate with it.

8. Security Considerations

To enable RDMA Read and Write operations, an RPC-over-RDMA Version One requester exposes some or all of its memory to other hosts. RFC 5666bis should suggest best implementation practices to minimize exposure to careless or potentially malicious implementations that share the same fabric. Important considerations include:

9. IANA Considerations

This document does not require actions by IANA.

10. Appendix A: XDR Language Description

Revised XDR definition of RPC-over-RDMA Version One. The original definition is in Section 4.3 of RFC 5666.

The XDR stream position of the fields and their use are not altered by this revision. The significant changes are:

  1. Copyright boilerplate has been provided
  2. The structure, field, and enum names have been made consistent with other standard XDR definitions
  3. The xdr_read_chunk structure is now called an rpcrdma1_read_segment because that structure functions the same way that an rpcrdma1_segment element in a Write chunk array does
  4. Duplicate definitions of the chunk list fields have been removed
  5. As the Read-Read transfer model is deprecated, RDMA_DONE is now a reserved value
  6. As RDMA_MSGP messages are deprecated, RDMA_MSGP is now a reserved value


    * Copyright (c) 2010, 2015 IETF Trust and the persons
    * identified as authors of the code.  All rights reserved.
    * The authors of the code are:
    * B. Callaghan, T. Talpey, and C. Lever.
    * 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.

   struct rpcrdma1_segment {
           uint32 rdma_handle;
           uint32 rdma_length;
           uint64 rdma_offset;

   struct rpcrdma1_read_segment {
           uint32                  rdma_position;
           struct rpcrdma1_segment rdma_target;

   struct rpcrdma1_read_list {
           struct rpcrdma1_read_segment rdma_entry;
           struct rpcrdma1_read_list    *rdma_next;

   struct rpcrdma1_write_chunk {
           struct rpcrdma1_segment rdma_target<>;

   struct rpcrdma1_write_list {
           struct rpcrdma1_write_chunk rdma_entry;
           struct rpcrdma1_write_list  *rdma_next;

   struct rpcrdma1_msg {
           uint32        rdma_xid;
           uint32        rdma_vers;
           uint32        rdma_credit;
           rpcrdma1_body rdma_body;

   enum rpcrdma1_proc {
           RDMA_MSG = 0,
           RDMA_NOMSG = 1,
           RDMA_MSGP = 2,  /* Reserved */
           RDMA_DONE = 3,  /* Reserved */
           RDMA_ERROR = 4

   struct rpcrdma1_chunks {
           struct rpcrdma1_read_list   *rdma_reads;
           struct rpcrdma1_write_list  *rdma_writes;
           struct rpcrdma1_write_chunk *rdma_reply;

   enum rpcrdma1_errcode {
           RDMA_ERR_VERS = 1,
           RDMA_ERR_CHUNK = 2

   union rpcrdma1_error switch (rpcrdma1_errcode err) {
           case RDMA_ERR_VERS:
             uint32 rdma_vers_low;
             uint32 rdma_vers_high;
           case RDMA_ERR_CHUNK:

   union rdma_body switch (rpcrdma1_proc proc) {
           case RDMA_MSG:
           case RDMA_NOMSG:
             rpcrdma1_chunks rdma_chunks;
           case RDMA_MSGP:
             uint32          rdma_align;
             uint32          rdma_thresh;
             rpcrdma1_chunks rdma_achunks;
           case RDMA_DONE:
           case RDMA_ERROR:
             rpcrdma1_error rdma_error;


Code components extracted from this document must include the following license:

11. Appendix B: Binding Requirement Summary

This appendix collects the known generic Binding Requirements from RFC 5666 and this document. This might not be an exhaustive list. Note that RFC 5666 uses RFC 2119-style terms to specify binding requirements, even though the requirement statements apply to protocol specifications rather than to a particular protocol.

  1. "Chunk eligibility criteria MUST be determined by each upper-layer in order to provide for an interoperable specification." (RFC 5666 Section 3.4)
  2. More specifically, an Upper Layer Binding is required for every RPC program interested in using RPC-over-RDMA. Separate bindings may be required for different versions of that program.
  3. Upper Layer Bindings make DDP-eligibility statements about specific arguments and results (or portions thereof which still are whole XDR objects). A chunk must contain only one whole XDR object.
  4. DDP-eligibility of any XDR object not mentioned explicitly in an Upper Layer Binding is "not eligible."
  5. Any XDR object may appear in a Position Zero read chunk or a Reply chunk regardless of its DDP-eligibility.
  6. An Upper Layer Binding may limit the number of unique read chunk Positions allowed for a particular operation. An Upper Layer Binding may limit the number of chunks in a Write list allowed for a particular operation.
  7. An Upper Layer Binding must take care not to allow abuses of the Position Zero read chunk to avoid DDP-eligibility restrictions.
  8. "It is NOT RECOMMENDED that upper-layer RPC client protocol specifications omit write chunk lists for eligible replies, due to the lower performance of the additional handshaking to perform data transfer, and the requirement that the RPC server must expose (and preserve) the reply data for a period of time." (RFC 5666 Section 3.4)
  9. "The mapping of write chunk list entries to procedure arguments MUST be determined for each protocol." (RFC 5666 Section 3.6)
  10. More specifically: by default, the requester provides as many Write chunks as the Upper Layer Binding allows for the particular operation. The responder fills in each Write chunk with an RDMA-eligible result until the Write list is exhausted or there are no more RDMA-eligible results. If this default behavior leads to ambiguity when the requester re-assembles the XDR stream, the Binding must explain how to resolve the ambiguity, or restrict DDP-eligibility to ensure confusion cannot occur.
  11. It is the responsibility of the Upper Layer Binding to specify DDP-eligibity rules so that if an DDP-eligible XDR object is embedded within another, only one of these two objects is to be represented by a chunk.
  12. The Upper Layer Binding must specify how a responder should reply if a requester violates a DDP-eligibilty rule. The Binding specification should provide guidance for requesters about handling invalid RPC-over-RDMA replies.

12. Acknowledgements

The author gratefully acknowledges the contributions of Dai Ngo, Karen Deitke, Chunli Zhang, Mahesh Siddheshwar, Dominique Martinet, and William Simpson.

The author also wishes to thank Dave Noveck and Bill Baker for their support of this work. Special thanks go to nfsv4 Working Group Chair Spencer Shepler and nfsv4 Working Group Secretary Tom Haynes for their support.

13. References

13.1. Normative References

[RFC0793] Postel, J., "Transmission Control Protocol", STD 7, RFC 793, DOI 10.17487/RFC0793, September 1981.
[RFC1813] Callaghan, B., Pawlowski, B. and P. Staubach, "NFS Version 3 Protocol Specification", RFC 1813, DOI 10.17487/RFC1813, June 1995.
[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., "XDR: External Data Representation Standard", STD 67, RFC 4506, DOI 10.17487/RFC4506, May 2006.
[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.
[RFC5056] Williams, N., "On the Use of Channel Bindings to Secure Channels", RFC 5056, DOI 10.17487/RFC5056, November 2007.
[RFC5403] Eisler, M., "RPCSEC_GSS Version 2", RFC 5403, DOI 10.17487/RFC5403, February 2009.
[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.
[RFC5661] Shepler, S., Eisler, M. and D. Noveck, "Network File System (NFS) Version 4 Minor Version 1 Protocol", RFC 5661, DOI 10.17487/RFC5661, January 2010.
[RFC5666] Talpey, T. and B. Callaghan, "Remote Direct Memory Access Transport for Remote Procedure Call", RFC 5666, DOI 10.17487/RFC5666, January 2010.
[RFC5667] Talpey, T. and B. Callaghan, "Network File System (NFS) Direct Data Placement", RFC 5667, DOI 10.17487/RFC5667, January 2010.

13.2. Informative References

[I-D.ietf-nfsv4-rpcrdma-bidirection] Lever, C., "Size-Limited Bi-directional Remote Procedure Call On Remote Direct Memory Access Transports", Internet-Draft draft-ietf-nfsv4-rpcrdma-bidirection-01, September 2015.

Author's Address

Charles Lever Oracle Corporation 1015 Granger Avenue Ann Arbor, MI 48104 US Phone: +1 734 274 2396 EMail: