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:
<CODE BEGINS>
typedef rpcrdma2_propid uint32;
struct rpcrdma2_propval {
rpcrdma2_propid rdma_which;
opaque rdma_data<>;
};
typedef rpcrdma2_propval rpcrdma2_propset<>;
typedef uint32 rpcrdma2_propsubset<>;
<CODE ENDS>
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.
<CODE BEGINS>
const uint32 RDMA2_PROPID_SBSIZ = 1;
typedef uint32 rpcrdma2_prop_sbsiz;
<CODE ENDS>
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.
<CODE BEGINS>
const uint32 RDMA2_PROPID_RBSIZ = 2;
typedef uint32 rpcrdma2_prop_rbsiz;
<CODE ENDS>
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.
<CODE BEGINS>
const uint32 RDMA2_PROPID_RSSIZ = 3;
typedef uint32 rpcrdma2_prop_rssiz;
<CODE ENDS>
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
<CODE BEGINS>
const uint32 RDMA2_PROPID_RCSIZ = 4;
typedef uint32 rpcrdma2_prop_rcsiz;
<CODE ENDS>
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.
<CODE BEGINS>
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;
<CODE ENDS>
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.
<CODE BEGINS>
const uint32 RDMA2_PROPID_HOSTAUTH = 6;
typedef opaque rpcrdma2_prop_hostauth<>;
<CODE ENDS>
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
<CODE BEGINS>
struct rpcrdma_common {
uint32 rdma_xid;
uint32 rdma_vers;
uint32 rdma_credit;
uint32 rdma_htype;
};
<CODE ENDS>
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
<CODE BEGINS>
const RPCRDMA2_F_RESPONSE 0x00000001;
const RPCRDMA2_F_MORE 0x00000002;
struct rpcrdma2_hdr_prefix
struct rpcrdma_common rdma_start;
uint32 rdma_flags;
};
<CODE ENDS>
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.
<CODE BEGINS>
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;
};
<CODE ENDS>
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.
<CODE BEGINS>
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;
};
<CODE ENDS>
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.
<CODE BEGINS>
const rpcrdma2_proc RDMA2_NOMSG = 1;
struct rpcrdma2_nomsg {
struct rpcrdma2_chunk_lists rdma_chunks;
};
<CODE ENDS>
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.
<CODE BEGINS>
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;
};
<CODE ENDS>
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:
<CODE BEGINS>
struct rpcrdma2_connprop {
rpcrdma2_propset rdma_props;
};
<CODE ENDS>
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
<CODE BEGINS>
/// /*
/// * 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.
/// */
///
<CODE ENDS>
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.
<CODE BEGINS>
sed -n -e 's:^ */// ::p' -e 's:^ *///$::p'
<CODE ENDS>
That is, if this document is in a file called "spec.txt" then the
reader can do the following to extract an XDR description file and
store it in the file rpcrdma-v2.x.
<CODE BEGINS>
sed -n -e 's:^ */// ::p' -e 's:^ *///$::p' \
< spec.txt > rpcrdma-v2.x
<CODE ENDS>
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
<CODE BEGINS>
#!/usr/local/bin/perl
open(IN,"rpcrdma-v2.x");
open(OUT,">temp.x");
while(<IN>)
{
if (m/FILE ENDS: (.*)$/)
{
close(OUT);
rename("temp.x", $1);
open(OUT,">temp.x");
}
else
{
print OUT $_;
}
}
close(IN);
close(OUT);
<CODE ENDS>
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
<CODE BEGINS>
/// /*******************************************************************
/// * 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; */
<CODE ENDS>
7.4. XDR Definition for RPC-over-RDMA Version 2 Base Header Types
<CODE BEGINS>
/// /*******************************************************************
/// * 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; */
<CODE ENDS>
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,
<https://www.rfc-editor.org/info/rfc1833>.
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,
<https://www.rfc-editor.org/info/rfc2119>.
[RFC4506] Eisler, M., Ed., "XDR: External Data Representation
Standard", STD 67, RFC 4506, DOI 10.17487/RFC4506, May
2006, <https://www.rfc-editor.org/info/rfc4506>.
[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, <https://www.rfc-editor.org/info/rfc5042>.
[RFC5056] Williams, N., "On the Use of Channel Bindings to Secure
Channels", RFC 5056, DOI 10.17487/RFC5056, November 2007,
<https://www.rfc-editor.org/info/rfc5056>.
[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,
<https://www.rfc-editor.org/info/rfc5280>.
[RFC5531] Thurlow, R., "RPC: Remote Procedure Call Protocol
Specification Version 2", RFC 5531, DOI 10.17487/RFC5531,
May 2009, <https://www.rfc-editor.org/info/rfc5531>.
[RFC5660] Williams, N., "IPsec Channels: Connection Latching",
RFC 5660, DOI 10.17487/RFC5660, October 2009,
<https://www.rfc-editor.org/info/rfc5660>.
[RFC5665] Eisler, M., "IANA Considerations for Remote Procedure Call
(RPC) Network Identifiers and Universal Address Formats",
RFC 5665, DOI 10.17487/RFC5665, January 2010,
<https://www.rfc-editor.org/info/rfc5665>.
[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, <https://www.rfc-editor.org/info/rfc6125>.
[RFC7861] Adamson, A. and N. Williams, "Remote Procedure Call (RPC)
Security Version 3", RFC 7861, DOI 10.17487/RFC7861,
November 2016, <https://www.rfc-editor.org/info/rfc7861>.
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,
<https://www.rfc-editor.org/info/rfc7942>.
[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,
<https://www.rfc-editor.org/info/rfc8166>.
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
May 2017, <https://www.rfc-editor.org/info/rfc8174>.
[RFC8267] Lever, C., "Network File System (NFS) Upper-Layer Binding
to RPC-over-RDMA Version 1", RFC 8267,
DOI 10.17487/RFC8267, October 2017,
<https://www.rfc-editor.org/info/rfc8267>.
[RFC8446] Rescorla, E., "The Transport Layer Security (TLS) Protocol
Version 1.3", RFC 8446, DOI 10.17487/RFC8446, August 2018,
<https://www.rfc-editor.org/info/rfc8446>.
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,
<https://www.rfc-editor.org/info/rfc768>.
[RFC0793] Postel, J., "Transmission Control Protocol", STD 7,
RFC 793, DOI 10.17487/RFC0793, September 1981,
<https://www.rfc-editor.org/info/rfc793>.
[RFC1094] Nowicki, B., "NFS: Network File System Protocol
specification", RFC 1094, DOI 10.17487/RFC1094, March
1989, <https://www.rfc-editor.org/info/rfc1094>.
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,
<https://www.rfc-editor.org/info/rfc1813>.
[RFC3552] Rescorla, E. and B. Korver, "Guidelines for Writing RFC
Text on Security Considerations", BCP 72, RFC 3552,
DOI 10.17487/RFC3552, July 2003,
<https://www.rfc-editor.org/info/rfc3552>.
[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, <https://www.rfc-editor.org/info/rfc5040>.
[RFC5041] Shah, H., Pinkerton, J., Recio, R., and P. Culley, "Direct
Data Placement over Reliable Transports", RFC 5041,
DOI 10.17487/RFC5041, October 2007,
<https://www.rfc-editor.org/info/rfc5041>.
[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,
<https://www.rfc-editor.org/info/rfc5532>.
[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,
<https://www.rfc-editor.org/info/rfc5661>.
[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,
<https://www.rfc-editor.org/info/rfc5662>.
[RFC7530] Haynes, T., Ed. and D. Noveck, Ed., "Network File System
(NFS) Version 4 Protocol", RFC 7530, DOI 10.17487/RFC7530,
March 2015, <https://www.rfc-editor.org/info/rfc7530>.
[RFC8167] Lever, C., "Bidirectional Remote Procedure Call on RPC-
over-RDMA Transports", RFC 8167, DOI 10.17487/RFC8167,
June 2017, <https://www.rfc-editor.org/info/rfc8167>.
[RFC8178] Noveck, D., "Rules for NFSv4 Extensions and Minor
Versions", RFC 8178, DOI 10.17487/RFC8178, July 2017,
<https://www.rfc-editor.org/info/rfc8178>.
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]
