draft-ietf-nfsv4-minorversion2-04.txt   draft-ietf-nfsv4-minorversion2-05.txt 
NFSv4 T. Haynes NFSv4 T. Haynes
Internet-Draft Editor Internet-Draft Editor
Intended status: Standards Track August 24, 2011 Intended status: Standards Track September 06, 2011
Expires: February 25, 2012 Expires: March 9, 2012
NFS Version 4 Minor Version 2 NFS Version 4 Minor Version 2
draft-ietf-nfsv4-minorversion2-04.txt draft-ietf-nfsv4-minorversion2-05.txt
Abstract Abstract
This Internet-Draft describes NFS version 4 minor version two, This Internet-Draft describes NFS version 4 minor version two,
focusing mainly on the protocol extensions made from NFS version 4 focusing mainly on the protocol extensions made from NFS version 4
minor version 0 and NFS version 4 minor version 1. Major extensions minor version 0 and NFS version 4 minor version 1. Major extensions
introduced in NFS version 4 minor version two include: Server-side introduced in NFS version 4 minor version two include: Server-side
Copy, Space Reservations, and Support for Sparse Files. Copy, Space Reservations, and Support for Sparse Files.
Requirements Language Requirements Language
skipping to change at page 1, line 40 skipping to change at page 1, line 40
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet- working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/. Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
This Internet-Draft will expire on February 25, 2012. This Internet-Draft will expire on March 9, 2012.
Copyright Notice Copyright Notice
Copyright (c) 2011 IETF Trust and the persons identified as the Copyright (c) 2011 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of (http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
skipping to change at page 3, line 13 skipping to change at page 3, line 13
than English. than English.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 6 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1. The NFS Version 4 Minor Version 2 Protocol . . . . . . . . 6 1.1. The NFS Version 4 Minor Version 2 Protocol . . . . . . . . 6
1.2. Scope of This Document . . . . . . . . . . . . . . . . . . 6 1.2. Scope of This Document . . . . . . . . . . . . . . . . . . 6
1.3. NFSv4.2 Goals . . . . . . . . . . . . . . . . . . . . . . 6 1.3. NFSv4.2 Goals . . . . . . . . . . . . . . . . . . . . . . 6
1.4. Overview of NFSv4.2 Features . . . . . . . . . . . . . . . 6 1.4. Overview of NFSv4.2 Features . . . . . . . . . . . . . . . 6
1.5. Differences from NFSv4.1 . . . . . . . . . . . . . . . . . 6 1.5. Differences from NFSv4.1 . . . . . . . . . . . . . . . . . 6
2. pNFS LAYOUTRETURN Error Handling . . . . . . . . . . . . . . . 6 2. NFS Server-side Copy . . . . . . . . . . . . . . . . . . . . . 6
2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 7 2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 7
2.2. Changes to Operation 51: LAYOUTRETURN . . . . . . . . . . 7 2.2. Protocol Overview . . . . . . . . . . . . . . . . . . . . 7
2.2.1. ARGUMENT . . . . . . . . . . . . . . . . . . . . . . . 7 2.2.1. Intra-Server Copy . . . . . . . . . . . . . . . . . . 9
2.2.2. RESULT . . . . . . . . . . . . . . . . . . . . . . . . 8 2.2.2. Inter-Server Copy . . . . . . . . . . . . . . . . . . 10
2.2.3. DESCRIPTION . . . . . . . . . . . . . . . . . . . . . 8 2.2.3. Server-to-Server Copy Protocol . . . . . . . . . . . . 13
2.2.4. IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . 8 2.3. Operations . . . . . . . . . . . . . . . . . . . . . . . . 15
3. Sharing change attribute implementation details with NFSv4 2.3.1. netloc4 - Network Locations . . . . . . . . . . . . . 15
clients . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.3.2. Copy Offload Stateids . . . . . . . . . . . . . . . . 16
3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 10 2.4. Security Considerations . . . . . . . . . . . . . . . . . 16
3.2. Definition of the 'change_attr_type' per-file system 2.4.1. Inter-Server Copy Security . . . . . . . . . . . . . . 16
attribute . . . . . . . . . . . . . . . . . . . . . . . . 10 3. Sparse Files . . . . . . . . . . . . . . . . . . . . . . . . . 24
4. NFS Server-side Copy . . . . . . . . . . . . . . . . . . . . . 11 3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 24
4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 12 3.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 25
4.2. Protocol Overview . . . . . . . . . . . . . . . . . . . . 12 3.3. Overview of Sparse Files and NFSv4 . . . . . . . . . . . . 25
4.2.1. Intra-Server Copy . . . . . . . . . . . . . . . . . . 14 3.4. Operation 65: READ_PLUS . . . . . . . . . . . . . . . . . 26
4.2.2. Inter-Server Copy . . . . . . . . . . . . . . . . . . 15 3.4.1. ARGUMENT . . . . . . . . . . . . . . . . . . . . . . . 26
4.2.3. Server-to-Server Copy Protocol . . . . . . . . . . . . 18 3.4.2. RESULT . . . . . . . . . . . . . . . . . . . . . . . . 27
4.3. Operations . . . . . . . . . . . . . . . . . . . . . . . . 20 3.4.3. DESCRIPTION . . . . . . . . . . . . . . . . . . . . . 27
4.3.1. netloc4 - Network Locations . . . . . . . . . . . . . 20 3.4.4. IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . 29
4.3.2. Copy Offload Stateids . . . . . . . . . . . . . . . . 21 3.4.5. READ_PLUS with Sparse Files Example . . . . . . . . . 30
4.4. Security Considerations . . . . . . . . . . . . . . . . . 21 3.5. Related Work . . . . . . . . . . . . . . . . . . . . . . . 31
4.4.1. Inter-Server Copy Security . . . . . . . . . . . . . . 21 3.6. Other Proposed Designs . . . . . . . . . . . . . . . . . . 31
5. Application Data Block Support . . . . . . . . . . . . . . . . 29 3.6.1. Multi-Data Server Hole Information . . . . . . . . . . 31
5.1. Generic Framework . . . . . . . . . . . . . . . . . . . . 30 3.6.2. Data Result Array . . . . . . . . . . . . . . . . . . 32
5.1.1. Data Block Representation . . . . . . . . . . . . . . 31 3.6.3. User-Defined Sparse Mask . . . . . . . . . . . . . . . 32
5.1.2. Data Content . . . . . . . . . . . . . . . . . . . . . 31 3.6.4. Allocated flag . . . . . . . . . . . . . . . . . . . . 32
5.2. pNFS Considerations . . . . . . . . . . . . . . . . . . . 31 3.6.5. Dense and Sparse pNFS File Layouts . . . . . . . . . . 33
5.3. An Example of Detecting Corruption . . . . . . . . . . . . 32 4. Space Reservation . . . . . . . . . . . . . . . . . . . . . . 33
5.4. Example of READ_PLUS . . . . . . . . . . . . . . . . . . . 34 4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 33
5.5. Zero Filled Holes . . . . . . . . . . . . . . . . . . . . 34 4.2. Operations and attributes . . . . . . . . . . . . . . . . 35
6. Space Reservation . . . . . . . . . . . . . . . . . . . . . . 34 4.3. Attribute 77: space_reserved . . . . . . . . . . . . . . . 35
6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 34 4.4. Attribute 78: space_freed . . . . . . . . . . . . . . . . 36
6.2. Use Cases . . . . . . . . . . . . . . . . . . . . . . . . 35 4.5. Attribute 79: max_hole_punch . . . . . . . . . . . . . . . 36
6.2.1. Space Reservation . . . . . . . . . . . . . . . . . . 36 5. Application Data Block Support . . . . . . . . . . . . . . . . 36
6.2.2. Space freed on deletes . . . . . . . . . . . . . . . . 36 5.1. Generic Framework . . . . . . . . . . . . . . . . . . . . 37
6.2.3. Operations and attributes . . . . . . . . . . . . . . 37 5.1.1. Data Block Representation . . . . . . . . . . . . . . 38
6.2.4. Attribute 77: space_reserved . . . . . . . . . . . . . 37 5.1.2. Data Content . . . . . . . . . . . . . . . . . . . . . 38
6.2.5. Attribute 78: space_freed . . . . . . . . . . . . . . 38 5.2. pNFS Considerations . . . . . . . . . . . . . . . . . . . 38
6.2.6. Attribute 79: max_hole_punch . . . . . . . . . . . . . 38 5.3. An Example of Detecting Corruption . . . . . . . . . . . . 39
6.2.7. Operation 64: HOLE_PUNCH - Zero and deallocate 5.4. Example of READ_PLUS . . . . . . . . . . . . . . . . . . . 40
blocks backing the file in the specified range. . . . 38 5.5. Zero Filled Holes . . . . . . . . . . . . . . . . . . . . 41
7. Sparse Files . . . . . . . . . . . . . . . . . . . . . . . . . 39 6. Labeled NFS . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 39 6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 41
7.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 40 6.2. Definitions . . . . . . . . . . . . . . . . . . . . . . . 42
7.3. Applications and Sparse Files . . . . . . . . . . . . . . 41 6.3. MAC Security Attribute . . . . . . . . . . . . . . . . . . 43
7.4. Overview of Sparse Files and NFSv4 . . . . . . . . . . . . 42 6.3.1. Interpreting FATTR4_SEC_LABEL . . . . . . . . . . . . 44
7.5. Operation 65: READ_PLUS . . . . . . . . . . . . . . . . . 43 6.3.2. Delegations . . . . . . . . . . . . . . . . . . . . . 44
7.5.1. ARGUMENT . . . . . . . . . . . . . . . . . . . . . . . 43 6.3.3. Permission Checking . . . . . . . . . . . . . . . . . 45
7.5.2. RESULT . . . . . . . . . . . . . . . . . . . . . . . . 44 6.3.4. Object Creation . . . . . . . . . . . . . . . . . . . 45
7.5.3. DESCRIPTION . . . . . . . . . . . . . . . . . . . . . 44 6.3.5. Existing Objects . . . . . . . . . . . . . . . . . . . 45
7.5.4. IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . 46 6.3.6. Label Changes . . . . . . . . . . . . . . . . . . . . 45
7.5.5. READ_PLUS with Sparse Files Example . . . . . . . . . 47 6.4. pNFS Considerations . . . . . . . . . . . . . . . . . . . 46
7.6. Related Work . . . . . . . . . . . . . . . . . . . . . . . 48 6.5. Discovery of Server LNFS Support . . . . . . . . . . . . . 47
7.7. Other Proposed Designs . . . . . . . . . . . . . . . . . . 48 6.6. MAC Security NFS Modes of Operation . . . . . . . . . . . 47
7.7.1. Multi-Data Server Hole Information . . . . . . . . . . 48 6.6.1. Full Mode . . . . . . . . . . . . . . . . . . . . . . 47
7.7.2. Data Result Array . . . . . . . . . . . . . . . . . . 49 6.6.2. Smart Client Mode . . . . . . . . . . . . . . . . . . 49
7.7.3. User-Defined Sparse Mask . . . . . . . . . . . . . . . 49 6.6.3. Smart Server Mode . . . . . . . . . . . . . . . . . . 49
7.7.4. Allocated flag . . . . . . . . . . . . . . . . . . . . 49 6.7. Security Considerations . . . . . . . . . . . . . . . . . 50
7.7.5. Dense and Sparse pNFS File Layouts . . . . . . . . . . 50 7. Sharing change attribute implementation details with NFSv4
8. Labeled NFS . . . . . . . . . . . . . . . . . . . . . . . . . 50 clients . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 50 7.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . 51
8.2. Definitions . . . . . . . . . . . . . . . . . . . . . . . 51 7.2. Definition of the 'change_attr_type' per-file system
8.3. MAC Security Attribute . . . . . . . . . . . . . . . . . . 51 attribute . . . . . . . . . . . . . . . . . . . . . . . . 51
8.3.1. Interpreting FATTR4_SEC_LABEL . . . . . . . . . . . . 52 8. Security Considerations . . . . . . . . . . . . . . . . . . . 53
8.3.2. Delegations . . . . . . . . . . . . . . . . . . . . . 53 9. Operations: REQUIRED, RECOMMENDED, or OPTIONAL . . . . . . . . 53
8.3.3. Permission Checking . . . . . . . . . . . . . . . . . 53 10. NFSv4.2 Operations . . . . . . . . . . . . . . . . . . . . . . 56
8.3.4. Object Creation . . . . . . . . . . . . . . . . . . . 54 10.1. Operation 59: COPY - Initiate a server-side copy . . . . . 56
8.3.5. Existing Objects . . . . . . . . . . . . . . . . . . . 54 10.2. Operation 60: COPY_ABORT - Cancel a server-side copy . . . 64
8.3.6. Label Changes . . . . . . . . . . . . . . . . . . . . 54 10.3. Operation 61: COPY_NOTIFY - Notify a source server of
8.4. pNFS Considerations . . . . . . . . . . . . . . . . . . . 55 a future copy . . . . . . . . . . . . . . . . . . . . . . 65
8.5. Discovery of Server LNFS Support . . . . . . . . . . . . . 55 10.4. Operation 62: COPY_REVOKE - Revoke a destination
8.6. MAC Security NFS Modes of Operation . . . . . . . . . . . 56 server's copy privileges . . . . . . . . . . . . . . . . . 68
8.6.1. Full Mode . . . . . . . . . . . . . . . . . . . . . . 56 10.5. Operation 63: COPY_STATUS - Poll for status of a
8.6.2. Smart Client Mode . . . . . . . . . . . . . . . . . . 57 server-side copy . . . . . . . . . . . . . . . . . . . . . 69
8.6.3. Smart Server Mode . . . . . . . . . . . . . . . . . . 58 10.6. Modification to Operation 42: EXCHANGE_ID -
8.7. Security Considerations . . . . . . . . . . . . . . . . . 59 Instantiate Client ID . . . . . . . . . . . . . . . . . . 70
9. Security Considerations . . . . . . . . . . . . . . . . . . . 59 10.7. Operation 64: INITIALIZE . . . . . . . . . . . . . . . . . 71
10. Operations: REQUIRED, RECOMMENDED, or OPTIONAL . . . . . . . . 59 10.8. Changes to Operation 51: LAYOUTRETURN . . . . . . . . . . 74
11. NFSv4.2 Operations . . . . . . . . . . . . . . . . . . . . . . 63 10.8.1. Introduction . . . . . . . . . . . . . . . . . . . . . 75
11.1. Operation 59: COPY - Initiate a server-side copy . . . . . 63 10.8.2. ARGUMENT . . . . . . . . . . . . . . . . . . . . . . . 75
11.2. Operation 60: COPY_ABORT - Cancel a server-side copy . . . 71 10.8.3. RESULT . . . . . . . . . . . . . . . . . . . . . . . . 76
11.3. Operation 61: COPY_NOTIFY - Notify a source server of 10.8.4. DESCRIPTION . . . . . . . . . . . . . . . . . . . . . 76
a future copy . . . . . . . . . . . . . . . . . . . . . . 72 10.8.5. IMPLEMENTATION . . . . . . . . . . . . . . . . . . . . 76
11.4. Operation 62: COPY_REVOKE - Revoke a destination 10.9. Operation 65: READ_PLUS . . . . . . . . . . . . . . . . . 78
server's copy privileges . . . . . . . . . . . . . . . . . 75 11. NFSv4.2 Callback Operations . . . . . . . . . . . . . . . . . 80
11.5. Operation 63: COPY_STATUS - Poll for status of a 11.1. Procedure 16: CB_ATTR_CHANGED - Notify Client that the
server-side copy . . . . . . . . . . . . . . . . . . . . . 76 File's Attributes Changed . . . . . . . . . . . . . . . . 80
11.6. Operation 64: INITIALIZE . . . . . . . . . . . . . . . . . 77 11.2. Operation 15: CB_COPY - Report results of a
11.7. Modification to Operation 42: EXCHANGE_ID - server-side copy . . . . . . . . . . . . . . . . . . . . . 80
Instantiate Client ID . . . . . . . . . . . . . . . . . . 79 12. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 82
11.8. Operation 65: READ_PLUS . . . . . . . . . . . . . . . . . 81 13. References . . . . . . . . . . . . . . . . . . . . . . . . . . 82
12. NFSv4.2 Callback Operations . . . . . . . . . . . . . . . . . 83 13.1. Normative References . . . . . . . . . . . . . . . . . . . 82
12.1. Procedure 16: CB_ATTR_CHANGED - Notify Client that the 13.2. Informative References . . . . . . . . . . . . . . . . . . 83
File's Attributes Changed . . . . . . . . . . . . . . . . 83 Appendix A. Acknowledgments . . . . . . . . . . . . . . . . . . . 84
12.2. Operation 15: CB_COPY - Report results of a Appendix B. RFC Editor Notes . . . . . . . . . . . . . . . . . . 85
server-side copy . . . . . . . . . . . . . . . . . . . . . 83 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 85
13. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 85
14. References . . . . . . . . . . . . . . . . . . . . . . . . . . 85
14.1. Normative References . . . . . . . . . . . . . . . . . . . 85
14.2. Informative References . . . . . . . . . . . . . . . . . . 86
Appendix A. Acknowledgments . . . . . . . . . . . . . . . . . . . 87
Appendix B. RFC Editor Notes . . . . . . . . . . . . . . . . . . 88
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 88
1. Introduction 1. Introduction
1.1. The NFS Version 4 Minor Version 2 Protocol 1.1. The NFS Version 4 Minor Version 2 Protocol
The NFS version 4 minor version 2 (NFSv4.2) protocol is the third The NFS version 4 minor version 2 (NFSv4.2) protocol is the third
minor version of the NFS version 4 (NFSv4) protocol. The first minor minor version of the NFS version 4 (NFSv4) protocol. The first minor
version, NFSv4.0, is described in [10] and the second minor version, version, NFSv4.0, is described in [10] and the second minor version,
NFSv4.1, is described in [2]. It follows the guidelines for minor NFSv4.1, is described in [2]. It follows the guidelines for minor
versioning that are listed in Section 11 of [10]. versioning that are listed in Section 11 of [10].
skipping to change at page 6, line 49 skipping to change at page 6, line 49
[[Comment.1: This needs fleshing out! --TH]] [[Comment.1: This needs fleshing out! --TH]]
1.4. Overview of NFSv4.2 Features 1.4. Overview of NFSv4.2 Features
[[Comment.2: This needs fleshing out! --TH]] [[Comment.2: This needs fleshing out! --TH]]
1.5. Differences from NFSv4.1 1.5. Differences from NFSv4.1
[[Comment.3: This needs fleshing out! --TH]] [[Comment.3: This needs fleshing out! --TH]]
2. pNFS LAYOUTRETURN Error Handling 2. NFS Server-side Copy
2.1. Introduction 2.1. Introduction
In the pNFS description provided in [2], the client is not enabled to
relay an error code from the DS to the MDS. In the specification of
the Objects-Based Layout protocol [4], use is made of the opaque
lrf_body field of the LAYOUTRETURN argument to do such a relaying of
error codes. In this section, we define a new data structure to
enable the passing of error codes back to the MDS and provide some
guidelines on what both the client and MDS should expect in such
circumstances.
There are two broad classes of errors, transient and persistent. The
client SHOULD strive to only use this new mechanism to report
persistent errors. It MUST be able to deal with transient issues by
itself. Also, while the client might consider an issue to be
persistent, it MUST be prepared for the MDS to consider such issues
to be persistent. A prime example of this is if the MDS fences off a
client from either a stateid or a filehandle. The client will get an
error from the DS and might relay either NFS4ERR_ACCESS or
NFS4ERR_STALE_STATEID back to the MDS, with the belief that this is a
hard error. The MDS on the other hand, is waiting for the client to
report such an error. For it, the mission is accomplished in that
the client has returned a layout that the MDS had most likley
recalled.
2.2. Changes to Operation 51: LAYOUTRETURN
The existing LAYOUTRETURN operation is extended by introducing a new
data structure to report errors, layoutreturn_device_error4. Also,
layoutreturn_device_error4 is introduced to enable an array of errors
to be reported.
2.2.1. ARGUMENT
The ARGUMENT specification of the LAYOUTRETURN operation in section
18.44.1 of [2] is augmented by the following XDR code [11]:
struct layoutreturn_device_error4 {
deviceid4 lrde_deviceid;
nfsstat4 lrde_status;
nfs_opnum4 lrde_opnum;
};
struct layoutreturn_error_report4 {
layoutreturn_device_error4 lrer_errors<>;
};
2.2.2. RESULT
The RESULT of the LAYOUTRETURN operation is unchanged; see section
18.44.2 of [2].
2.2.3. DESCRIPTION
The following text is added to the end of the LAYOUTRETURN operation
DESCRIPTION in section 18.44.3 of [2].
When a client used LAYOUTRETURN with a type of LAYOUTRETURN4_FILE,
then if the lrf_body field is NULL, it indicates to the MDS that the
client experienced no errors. If lrf_body is non-NULL, then the
field references error information which is layout type specific.
I.e., the Objects-Based Layout protocol can continue to utilize
lrf_body as specified in [4]. For both Files-Based Layouts, the
field references a layoutreturn_device_error4, which contains an
array of layoutreturn_device_error4.
Each individual layoutreturn_device_error4 descibes a single error
associated with a DS, which is identfied via lrde_deviceid. The
operation which returned the error is identified via lrde_opnum.
Finally the NFS error value (nfsstat4) encountered is provided via
lrde_status and may consist of the following error codes:
NFS4_OKAY: No issues were found for this device.
NFS4ERR_NXIO: The client was unable to establish any communication
with the DS.
NFS4ERR_*: The client was able to establish communication with the
DS and is returning one of the allowed error codes for the
operation denoted by lrde_opnum.
2.2.4. IMPLEMENTATION
The following text is added to the end of the LAYOUTRETURN operation
IMPLEMENTATION in section 18.4.4 of [2].
A client that expects to use pNFS for a mounted filesystem SHOULD
check for pNFS support at mount time. This check SHOULD be performed
by sending a GETDEVICELIST operation, followed by layout-type-
specific checks for accessibility of each storage device returned by
GETDEVICELIST. If the NFS server does not support pNFS, the
GETDEVICELIST operation will be rejected with an NFS4ERR_NOTSUPP
error; in this situation it is up to the client to determine whether
it is acceptable to proceed with NFS-only access.
Clients are expected to tolerate transient storage device errors, and
hence clients SHOULD NOT use the LAYOUTRETURN error handling for
device access problems that may be transient. The methods by which a
client decides whether an access problem is transient vs. persistent
are implementation-specific, but may include retrying I/Os to a data
server under appropriate conditions.
When an I/O fails to a storage device, the client SHOULD retry the
failed I/O via the MDS. In this situation, before retrying the I/O,
the client SHOULD return the layout, or the affected portion thereof,
and SHOULD indicate which storage device or devices was problematic.
If the client does not do this, the MDS may issue a layout recall
callback in order to perform the retried I/O.
The client needs to be cognizant that since this error handling is
optional in the MDS, the MDS may silently ignore this functionality.
Also, as the MDS may consider some issues the client reports to be
expected (see Section 2.1), the client might find it difficult to
detect a MDS which has not implemented error handling via
LAYOUTRETURN.
If an MDS is aware that a storage device is proving problematic to a
client, the MDS SHOULD NOT include that storage device in any pNFS
layouts sent to that client. If the MDS is aware that a storage
device is affecting many clients, then the MDS SHOULD NOT include
that storage device in any pNFS layouts sent out. Clients must still
be aware that the MDS might not have any choice in using the storage
device, i.e., there might only be one possible layout for the system.
Another interesting complication is that for existing files, the MDS
might have no choice in which storage devices to hand out to clients.
The MDS might try to restripe a file across a different storage
device, but clients need to be aware that not all implementations
have restriping support.
An MDS SHOULD react to a client return of layouts with errors by not
using the problematic storage devices in layouts for that client, but
the MDS is not required to indefinitely retain per-client storage
device error information. An MDS is also not required to
automatically reinstate use of a previously problematic storage
device; administrative intervention may be required instead.
A client MAY perform I/O via the MDS even when the client holds a
layout that covers the I/O; servers MUST support this client
behavior, and MAY recall layouts as needed to complete I/Os.
3. Sharing change attribute implementation details with NFSv4 clients
3.1. Introduction
Although both the NFSv4 [10] and NFSv4.1 protocol [2], define the
change attribute as being mandatory to implement, there is little in
the way of guidance. The only feature that is mandated by them is
that the value must change whenever the file data or metadata change.
While this allows for a wide range of implementations, it also leaves
the client with a conundrum: how does it determine which is the most
recent value for the change attribute in a case where several RPC
calls have been issued in parallel? In other words if two COMPOUNDs,
both containing WRITE and GETATTR requests for the same file, have
been issued in parallel, how does the client determine which of the
two change attribute values returned in the replies to the GETATTR
requests corresponds to the most recent state of the file? In some
cases, the only recourse may be to send another COMPOUND containing a
third GETATTR that is fully serialised with the first two.
NFSv4.2 avoids this kind of inefficiency by allowing the server to
share details about how the change attribute is expected to evolve,
so that the client may immediately determine which, out of the
several change attribute values returned by the server, is the most
recent.
3.2. Definition of the 'change_attr_type' per-file system attribute
enum change_attr_typeinfo {
NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR = 0,
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER = 1,
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS = 2,
NFS4_CHANGE_TYPE_IS_TIME_METADATA = 3,
NFS4_CHANGE_TYPE_IS_UNDEFINED = 4
};
+------------------+----+---------------------------+-----+
| Name | Id | Data Type | Acc |
+------------------+----+---------------------------+-----+
| change_attr_type | XX | enum change_attr_typeinfo | R |
+------------------+----+---------------------------+-----+
The solution enables the NFS server to provide additional information
about how it expects the change attribute value to evolve after the
file data or metadata has changed. 'change_attr_type' is defined as a
new recommended attribute, and takes values from enum
change_attr_typeinfo as follows:
NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR: The change attribute value MUST
monotonically increase for every atomic change to the file
attributes, data or directory contents.
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER: The change attribute value MUST
be incremented by one unit for every atomic change to the file
attributes, data or directory contents. This property is
preserved when writing to pNFS data servers.
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS: The change attribute
value MUST be incremented by one unit for every atomic change to
the file attributes, data or directory contents. In the case
where the client is writing to pNFS data servers, the number of
increments is not guaranteed to exactly match the number of
writes.
NFS4_CHANGE_TYPE_IS_TIME_METADATA: The change attribute is
implemented as suggested in the NFSv4 spec [10] in terms of the
time_metadata attribute.
NFS4_CHANGE_TYPE_IS_UNDEFINED: The change attribute does not take
values that fit into any of these categories.
If either NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR,
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER, or
NFS4_CHANGE_TYPE_IS_TIME_METADATA are set, then the client knows at
the very least that the change attribute is monotonically increasing,
which is sufficient to resolve the question of which value is the
most recent.
If the client sees the value NFS4_CHANGE_TYPE_IS_TIME_METADATA, then
by inspecting the value of the 'time_delta' attribute it additionally
has the option of detecting rogue server implementations that use
time_metadata in violation of the spec.
Finally, if the client sees NFS4_CHANGE_TYPE_IS_VERSION_COUNTER, it
has the ability to predict what the resulting change attribute value
should be after a COMPOUND containing a SETATTR, WRITE, or CREATE.
This again allows it to detect changes made in parallel by another
client. The value NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS permits
the same, but only if the client is not doing pNFS WRITEs.
4. NFS Server-side Copy
4.1. Introduction
This section describes a server-side copy feature for the NFS This section describes a server-side copy feature for the NFS
protocol. protocol.
The server-side copy feature provides a mechanism for the NFS client The server-side copy feature provides a mechanism for the NFS client
to perform a file copy on the server without the data being to perform a file copy on the server without the data being
transmitted back and forth over the network. transmitted back and forth over the network.
Without this feature, an NFS client copies data from one location to Without this feature, an NFS client copies data from one location to
another by reading the data from the server over the network, and another by reading the data from the server over the network, and
then writing the data back over the network to the server. Using then writing the data back over the network to the server. Using
skipping to change at page 12, line 31 skipping to change at page 7, line 31
location to another on the server. It is particularly useful when location to another on the server. It is particularly useful when
copying the contents of a file from a backup. Backup-versions of a copying the contents of a file from a backup. Backup-versions of a
file are copied for a number of reasons, including restoring and file are copied for a number of reasons, including restoring and
cloning data. cloning data.
If the source object and destination object are on different file If the source object and destination object are on different file
servers, the file servers will communicate with one another to servers, the file servers will communicate with one another to
perform the copy operation. The server-to-server protocol by which perform the copy operation. The server-to-server protocol by which
this is accomplished is not defined in this document. this is accomplished is not defined in this document.
4.2. Protocol Overview 2.2. Protocol Overview
The server-side copy offload operations support both intra-server and The server-side copy offload operations support both intra-server and
inter-server file copies. An intra-server copy is a copy in which inter-server file copies. An intra-server copy is a copy in which
the source file and destination file reside on the same server. In the source file and destination file reside on the same server. In
an inter-server copy, the source file and destination file are on an inter-server copy, the source file and destination file are on
different servers. In both cases, the copy may be performed different servers. In both cases, the copy may be performed
synchronously or asynchronously. synchronously or asynchronously.
Throughout the rest of this document, we refer to the NFS server Throughout the rest of this document, we refer to the NFS server
containing the source file as the "source server" and the NFS server containing the source file as the "source server" and the NFS server
skipping to change at page 13, line 6 skipping to change at page 8, line 6
server are the same server. Therefore in the context of an intra- server are the same server. Therefore in the context of an intra-
server copy, the terms source server and destination server refer to server copy, the terms source server and destination server refer to
the single server performing the copy. the single server performing the copy.
The operations described below are designed to copy files. Other The operations described below are designed to copy files. Other
file system objects can be copied by building on these operations or file system objects can be copied by building on these operations or
using other techniques. For example if the user wishes to copy a using other techniques. For example if the user wishes to copy a
directory, the client can synthesize a directory copy by first directory, the client can synthesize a directory copy by first
creating the destination directory and then copying the source creating the destination directory and then copying the source
directory's files to the new destination directory. If the user directory's files to the new destination directory. If the user
wishes to copy a namespace junction [12] [13], the client can use the wishes to copy a namespace junction [11] [12], the client can use the
ONC RPC Federated Filesystem protocol [13] to perform the copy. ONC RPC Federated Filesystem protocol [12] to perform the copy.
Specifically the client can determine the source junction's Specifically the client can determine the source junction's
attributes using the FEDFS_LOOKUP_FSN procedure and create a attributes using the FEDFS_LOOKUP_FSN procedure and create a
duplicate junction using the FEDFS_CREATE_JUNCTION procedure. duplicate junction using the FEDFS_CREATE_JUNCTION procedure.
For the inter-server copy protocol, the operations are defined to be For the inter-server copy protocol, the operations are defined to be
compatible with a server-to-server copy protocol in which the compatible with a server-to-server copy protocol in which the
destination server reads the file data from the source server. This destination server reads the file data from the source server. This
model in which the file data is pulled from the source by the model in which the file data is pulled from the source by the
destination has a number of advantages over a model in which the destination has a number of advantages over a model in which the
source pushes the file data to the destination. The advantages of source pushes the file data to the destination. The advantages of
skipping to change at page 14, line 13 skipping to change at page 9, line 13
COPY: Used by the client to request a file copy. COPY: Used by the client to request a file copy.
COPY_ABORT: Used by the client to abort an asynchronous file copy. COPY_ABORT: Used by the client to abort an asynchronous file copy.
COPY_STATUS: Used by the client to poll the status of an COPY_STATUS: Used by the client to poll the status of an
asynchronous file copy. asynchronous file copy.
CB_COPY: Used by the destination server to report the results of an CB_COPY: Used by the destination server to report the results of an
asynchronous file copy to the client. asynchronous file copy to the client.
These operations are described in detail in Section 4.3. This These operations are described in detail in Section 2.3. This
section provides an overview of how these operations are used to section provides an overview of how these operations are used to
perform server-side copies. perform server-side copies.
4.2.1. Intra-Server Copy 2.2.1. Intra-Server Copy
To copy a file on a single server, the client uses a COPY operation. To copy a file on a single server, the client uses a COPY operation.
The server may respond to the copy operation with the final results The server may respond to the copy operation with the final results
of the copy or it may perform the copy asynchronously and deliver the of the copy or it may perform the copy asynchronously and deliver the
results using a CB_COPY operation callback. If the copy is performed results using a CB_COPY operation callback. If the copy is performed
asynchronously, the client may poll the status of the copy using asynchronously, the client may poll the status of the copy using
COPY_STATUS or cancel the copy using COPY_ABORT. COPY_STATUS or cancel the copy using COPY_ABORT.
A synchronous intra-server copy is shown in Figure 1. In this A synchronous intra-server copy is shown in Figure 1. In this
example, the NFS server chooses to perform the copy synchronously. example, the NFS server chooses to perform the copy synchronously.
skipping to change at page 15, line 25 skipping to change at page 10, line 25
| . | Multiple COPY_STATUS | . | Multiple COPY_STATUS
| . | operations may be sent. | . | operations may be sent.
| . | | . |
| | | |
|<-- CB_COPY --------------------------| Server reports results |<-- CB_COPY --------------------------| Server reports results
|\------------------------------------>| |\------------------------------------>|
| | | |
Figure 2: An asynchronous intra-server copy. Figure 2: An asynchronous intra-server copy.
4.2.2. Inter-Server Copy 2.2.2. Inter-Server Copy
A copy may also be performed between two servers. The copy protocol A copy may also be performed between two servers. The copy protocol
is designed to accommodate a variety of network topologies. As shown is designed to accommodate a variety of network topologies. As shown
in Figure 3, the client and servers may be connected by multiple in Figure 3, the client and servers may be connected by multiple
networks. In particular, the servers may be connected by a networks. In particular, the servers may be connected by a
specialized, high speed network (network 192.168.33.0/24 in the specialized, high speed network (network 192.168.33.0/24 in the
diagram) that does not include the client. The protocol allows the diagram) that does not include the client. The protocol allows the
client to setup the copy between the servers (over network client to setup the copy between the servers (over network
10.11.78.0/24 in the diagram) and for the servers to communicate on 10.11.78.0/24 in the diagram) and for the servers to communicate on
the high speed network if they choose to do so. the high speed network if they choose to do so.
skipping to change at page 18, line 39 skipping to change at page 13, line 39
| | . | | | . |
| | | | | |
| | | | | |
| | | | | |
|<-- CB_COPY --------------------------| Destination reports |<-- CB_COPY --------------------------| Destination reports
|\------------------------------------>| results |\------------------------------------>| results
| | | | | |
Figure 5: An asynchronous inter-server copy. Figure 5: An asynchronous inter-server copy.
4.2.3. Server-to-Server Copy Protocol 2.2.3. Server-to-Server Copy Protocol
During an inter-server copy, the destination server reads the file During an inter-server copy, the destination server reads the file
data from the source server. The source server and destination data from the source server. The source server and destination
server are not required to use a specific protocol to transfer the server are not required to use a specific protocol to transfer the
file data. The choice of what protocol to use is ultimately the file data. The choice of what protocol to use is ultimately the
destination server's decision. destination server's decision.
4.2.3.1. Using NFSv4.x as a Server-to-Server Copy Protocol 2.2.3.1. Using NFSv4.x as a Server-to-Server Copy Protocol
The destination server MAY use standard NFSv4.x (where x >= 1) to The destination server MAY use standard NFSv4.x (where x >= 1) to
read the data from the source server. If NFSv4.x is used for the read the data from the source server. If NFSv4.x is used for the
server-to-server copy protocol, the destination server can use the server-to-server copy protocol, the destination server can use the
filehandle contained in the COPY request with standard NFSv4.x filehandle contained in the COPY request with standard NFSv4.x
operations to read data from the source server. Specifically, the operations to read data from the source server. Specifically, the
destination server may use the NFSv4.x OPEN operation's CLAIM_FH destination server may use the NFSv4.x OPEN operation's CLAIM_FH
facility to open the file being copied and obtain an open stateid. facility to open the file being copied and obtain an open stateid.
Using the stateid, the destination server may then use NFSv4.x READ Using the stateid, the destination server may then use NFSv4.x READ
operations to read the file. operations to read the file.
4.2.3.2. Using an alternative Server-to-Server Copy Protocol 2.2.3.2. Using an alternative Server-to-Server Copy Protocol
In a homogeneous environment, the source and destination servers In a homogeneous environment, the source and destination servers
might be able to perform the file copy extremely efficiently using might be able to perform the file copy extremely efficiently using
specialized protocols. For example the source and destination specialized protocols. For example the source and destination
servers might be two nodes sharing a common file system format for servers might be two nodes sharing a common file system format for
the source and destination file systems. Thus the source and the source and destination file systems. Thus the source and
destination are in an ideal position to efficiently render the image destination are in an ideal position to efficiently render the image
of the source file to the destination file by replicating the file of the source file to the destination file by replicating the file
system formats at the block level. Another possibility is that the system formats at the block level. Another possibility is that the
source and destination might be two nodes sharing a common storage source and destination might be two nodes sharing a common storage
area network, and thus there is no need to copy any data at all, and area network, and thus there is no need to copy any data at all, and
instead ownership of the file and its contents might simply be re- instead ownership of the file and its contents might simply be re-
assigned to the destination. To allow for these possibilities, the assigned to the destination. To allow for these possibilities, the
destination server is allowed to use a server-to-server copy protocol destination server is allowed to use a server-to-server copy protocol
of its choice. of its choice.
In a heterogeneous environment, using a protocol other than NFSv4.x In a heterogeneous environment, using a protocol other than NFSv4.x
(e.g,. HTTP [14] or FTP [15]) presents some challenges. In (e.g,. HTTP [13] or FTP [14]) presents some challenges. In
particular, the destination server is presented with the challenge of particular, the destination server is presented with the challenge of
accessing the source file given only an NFSv4.x filehandle. accessing the source file given only an NFSv4.x filehandle.
One option for protocols that identify source files with path names One option for protocols that identify source files with path names
is to use an ASCII hexadecimal representation of the source is to use an ASCII hexadecimal representation of the source
filehandle as the file name. filehandle as the file name.
Another option for the source server is to use URLs to direct the Another option for the source server is to use URLs to direct the
destination server to a specialized service. For example, the destination server to a specialized service. For example, the
response to COPY_NOTIFY could include the URL response to COPY_NOTIFY could include the URL
skipping to change at page 19, line 51 skipping to change at page 14, line 51
destination server receives the source server's URL, it would use destination server receives the source server's URL, it would use
"_FH/0x12345" as the file name to pass to the FTP server listening on "_FH/0x12345" as the file name to pass to the FTP server listening on
port 9999 of s1.example.com. On port 9999 there would be a special port 9999 of s1.example.com. On port 9999 there would be a special
instance of the FTP service that understands how to convert NFS instance of the FTP service that understands how to convert NFS
filehandles to an open file descriptor (in many operating systems, filehandles to an open file descriptor (in many operating systems,
this would require a new system call, one which is the inverse of the this would require a new system call, one which is the inverse of the
makefh() function that the pre-NFSv4 MOUNT service needs). makefh() function that the pre-NFSv4 MOUNT service needs).
Authenticating and identifying the destination server to the source Authenticating and identifying the destination server to the source
server is also a challenge. Recommendations for how to accomplish server is also a challenge. Recommendations for how to accomplish
this are given in Section 4.4.1.2.4 and Section 4.4.1.4. this are given in Section 2.4.1.2.4 and Section 2.4.1.4.
4.3. Operations 2.3. Operations
In the sections that follow, several operations are defined that In the sections that follow, several operations are defined that
together provide the server-side copy feature. These operations are together provide the server-side copy feature. These operations are
intended to be OPTIONAL operations as defined in section 17 of [2]. intended to be OPTIONAL operations as defined in section 17 of [2].
The COPY_NOTIFY, COPY_REVOKE, COPY, COPY_ABORT, and COPY_STATUS The COPY_NOTIFY, COPY_REVOKE, COPY, COPY_ABORT, and COPY_STATUS
operations are designed to be sent within an NFSv4 COMPOUND operations are designed to be sent within an NFSv4 COMPOUND
procedure. The CB_COPY operation is designed to be sent within an procedure. The CB_COPY operation is designed to be sent within an
NFSv4 CB_COMPOUND procedure. NFSv4 CB_COMPOUND procedure.
Each operation is performed in the context of the user identified by Each operation is performed in the context of the user identified by
the ONC RPC credential of its containing COMPOUND or CB_COMPOUND the ONC RPC credential of its containing COMPOUND or CB_COMPOUND
request. For example, a COPY_ABORT operation issued by a given user request. For example, a COPY_ABORT operation issued by a given user
indicates that a specified COPY operation initiated by the same user indicates that a specified COPY operation initiated by the same user
be canceled. Therefore a COPY_ABORT MUST NOT interfere with a copy be canceled. Therefore a COPY_ABORT MUST NOT interfere with a copy
of the same file initiated by another user. of the same file initiated by another user.
An NFS server MAY allow an administrative user to monitor or cancel An NFS server MAY allow an administrative user to monitor or cancel
copy operations using an implementation specific interface. copy operations using an implementation specific interface.
4.3.1. netloc4 - Network Locations 2.3.1. netloc4 - Network Locations
The server-side copy operations specify network locations using the The server-side copy operations specify network locations using the
netloc4 data type shown below: netloc4 data type shown below:
enum netloc_type4 { enum netloc_type4 {
NL4_NAME = 0, NL4_NAME = 0,
NL4_URL = 1, NL4_URL = 1,
NL4_NETADDR = 2 NL4_NETADDR = 2
}; };
union netloc4 switch (netloc_type4 nl_type) { union netloc4 switch (netloc_type4 nl_type) {
case NL4_NAME: utf8str_cis nl_name; case NL4_NAME: utf8str_cis nl_name;
case NL4_URL: utf8str_cis nl_url; case NL4_URL: utf8str_cis nl_url;
case NL4_NETADDR: netaddr4 nl_addr; case NL4_NETADDR: netaddr4 nl_addr;
}; };
If the netloc4 is of type NL4_NAME, the nl_name field MUST be If the netloc4 is of type NL4_NAME, the nl_name field MUST be
specified as a UTF-8 string. The nl_name is expected to be resolved specified as a UTF-8 string. The nl_name is expected to be resolved
to a network address via DNS, LDAP, NIS, /etc/hosts, or some other to a network address via DNS, LDAP, NIS, /etc/hosts, or some other
means. If the netloc4 is of type NL4_URL, a server URL [5] means. If the netloc4 is of type NL4_URL, a server URL [4]
appropriate for the server-to-server copy operation is specified as a appropriate for the server-to-server copy operation is specified as a
UTF-8 string. If the netloc4 is of type NL4_NETADDR, the nl_addr UTF-8 string. If the netloc4 is of type NL4_NETADDR, the nl_addr
field MUST contain a valid netaddr4 as defined in Section 3.3.9 of field MUST contain a valid netaddr4 as defined in Section 3.3.9 of
[2]. [2].
When netloc4 values are used for an inter-server copy as shown in When netloc4 values are used for an inter-server copy as shown in
Figure 3, their values may be evaluated on the source server, Figure 3, their values may be evaluated on the source server,
destination server, and client. The network environment in which destination server, and client. The network environment in which
these systems operate should be configured so that the netloc4 values these systems operate should be configured so that the netloc4 values
are interpreted as intended on each system. are interpreted as intended on each system.
4.3.2. Copy Offload Stateids 2.3.2. Copy Offload Stateids
A server may perform a copy offload operation asynchronously. An A server may perform a copy offload operation asynchronously. An
asynchronous copy is tracked using a copy offload stateid. Copy asynchronous copy is tracked using a copy offload stateid. Copy
offload stateids are included in the COPY, COPY_ABORT, COPY_STATUS, offload stateids are included in the COPY, COPY_ABORT, COPY_STATUS,
and CB_COPY operations. and CB_COPY operations.
Section 8.2.4 of [2] specifies that stateids are valid until either Section 8.2.4 of [2] specifies that stateids are valid until either
(A) the client or server restart or (B) the client returns the (A) the client or server restart or (B) the client returns the
resource. resource.
A copy offload stateid will be valid until either (A) the client or A copy offload stateid will be valid until either (A) the client or
server restart or (B) the client returns the resource by issuing a server restart or (B) the client returns the resource by issuing a
COPY_ABORT operation or the client replies to a CB_COPY operation. COPY_ABORT operation or the client replies to a CB_COPY operation.
A copy offload stateid's seqid MUST NOT be 0 (zero). In the context A copy offload stateid's seqid MUST NOT be 0 (zero). In the context
of a copy offload operation, it is ambiguous to indicate the most of a copy offload operation, it is ambiguous to indicate the most
recent copy offload operation using a stateid with seqid of 0 (zero). recent copy offload operation using a stateid with seqid of 0 (zero).
Therefore a copy offload stateid with seqid of 0 (zero) MUST be Therefore a copy offload stateid with seqid of 0 (zero) MUST be
considered invalid. considered invalid.
4.4. Security Considerations 2.4. Security Considerations
The security considerations pertaining to NFSv4 [10] apply to this The security considerations pertaining to NFSv4 [10] apply to this
document. document.
The standard security mechanisms provide by NFSv4 [10] may be used to The standard security mechanisms provide by NFSv4 [10] may be used to
secure the protocol described in this document. secure the protocol described in this document.
NFSv4 clients and servers supporting the the inter-server copy NFSv4 clients and servers supporting the the inter-server copy
operations described in this document are REQUIRED to implement [6], operations described in this document are REQUIRED to implement [5],
including the RPCSEC_GSSv3 privileges copy_from_auth and including the RPCSEC_GSSv3 privileges copy_from_auth and
copy_to_auth. If the server-to-server copy protocol is ONC RPC copy_to_auth. If the server-to-server copy protocol is ONC RPC
based, the servers are also REQUIRED to implement the RPCSEC_GSSv3 based, the servers are also REQUIRED to implement the RPCSEC_GSSv3
privilege copy_confirm_auth. These requirements to implement are not privilege copy_confirm_auth. These requirements to implement are not
requirements to use. NFSv4 clients and servers are RECOMMENDED to requirements to use. NFSv4 clients and servers are RECOMMENDED to
use [6] to secure server-side copy operations. use [5] to secure server-side copy operations.
4.4.1. Inter-Server Copy Security 2.4.1. Inter-Server Copy Security
4.4.1.1. Requirements for Secure Inter-Server Copy 2.4.1.1. Requirements for Secure Inter-Server Copy
Inter-server copy is driven by several requirements: Inter-server copy is driven by several requirements:
o The specification MUST NOT mandate an inter-server copy protocol. o The specification MUST NOT mandate an inter-server copy protocol.
There are many ways to copy data. Some will be more optimal than There are many ways to copy data. Some will be more optimal than
others depending on the identities of the source server and others depending on the identities of the source server and
destination server. For example the source and destination destination server. For example the source and destination
servers might be two nodes sharing a common file system format for servers might be two nodes sharing a common file system format for
the source and destination file systems. Thus the source and the source and destination file systems. Thus the source and
destination are in an ideal position to efficiently render the destination are in an ideal position to efficiently render the
skipping to change at page 22, line 36 skipping to change at page 17, line 36
destination first have a "copying relationship" increases the destination first have a "copying relationship" increases the
administrative burden. However the specification MUST NOT administrative burden. However the specification MUST NOT
preclude implementations that require pre-configuration. preclude implementations that require pre-configuration.
o The specification MUST NOT mandate a trust relationship between o The specification MUST NOT mandate a trust relationship between
the source and destination server. The NFSv4 security model the source and destination server. The NFSv4 security model
requires mutual authentication between a principal on an NFS requires mutual authentication between a principal on an NFS
client and a principal on an NFS server. This model MUST continue client and a principal on an NFS server. This model MUST continue
with the introduction of COPY. with the introduction of COPY.
4.4.1.2. Inter-Server Copy with RPCSEC_GSSv3 2.4.1.2. Inter-Server Copy with RPCSEC_GSSv3
When the client sends a COPY_NOTIFY to the source server to expect When the client sends a COPY_NOTIFY to the source server to expect
the destination to attempt to copy data from the source server, it is the destination to attempt to copy data from the source server, it is
expected that this copy is being done on behalf of the principal expected that this copy is being done on behalf of the principal
(called the "user principal") that sent the RPC request that encloses (called the "user principal") that sent the RPC request that encloses
the COMPOUND procedure that contains the COPY_NOTIFY operation. The the COMPOUND procedure that contains the COPY_NOTIFY operation. The
user principal is identified by the RPC credentials. A mechanism user principal is identified by the RPC credentials. A mechanism
that allows the user principal to authorize the destination server to that allows the user principal to authorize the destination server to
perform the copy in a manner that lets the source server properly perform the copy in a manner that lets the source server properly
authenticate the destination's copy, and without allowing the authenticate the destination's copy, and without allowing the
skipping to change at page 23, line 10 skipping to change at page 18, line 10
An approach that sends delegated credentials of the client's user An approach that sends delegated credentials of the client's user
principal to the destination server is not used for the following principal to the destination server is not used for the following
reasons. If the client's user delegated its credentials, the reasons. If the client's user delegated its credentials, the
destination would authenticate as the user principal. If the destination would authenticate as the user principal. If the
destination were using the NFSv4 protocol to perform the copy, then destination were using the NFSv4 protocol to perform the copy, then
the source server would authenticate the destination server as the the source server would authenticate the destination server as the
user principal, and the file copy would securely proceed. However, user principal, and the file copy would securely proceed. However,
this approach would allow the destination server to copy other files. this approach would allow the destination server to copy other files.
The user principal would have to trust the destination server to not The user principal would have to trust the destination server to not
do so. This is counter to the requirements, and therefore is not do so. This is counter to the requirements, and therefore is not
considered. Instead an approach using RPCSEC_GSSv3 [6] privileges is considered. Instead an approach using RPCSEC_GSSv3 [5] privileges is
proposed. proposed.
One of the stated applications of the proposed RPCSEC_GSSv3 protocol One of the stated applications of the proposed RPCSEC_GSSv3 protocol
is compound client host and user authentication [+ privilege is compound client host and user authentication [+ privilege
assertion]. For inter-server file copy, we require compound NFS assertion]. For inter-server file copy, we require compound NFS
server host and user authentication [+ privilege assertion]. The server host and user authentication [+ privilege assertion]. The
distinction between the two is one without meaning. distinction between the two is one without meaning.
RPCSEC_GSSv3 introduces the notion of privileges. We define three RPCSEC_GSSv3 introduces the notion of privileges. We define three
privileges: privileges:
skipping to change at page 24, line 36 skipping to change at page 19, line 36
struct copy_confirm_auth_priv { struct copy_confirm_auth_priv {
/* equal to GSS_GetMIC() of cfap_shared_secret */ /* equal to GSS_GetMIC() of cfap_shared_secret */
opaque ccap_shared_secret_mic<>; opaque ccap_shared_secret_mic<>;
/* the NFSv4 user name that the user principal maps to */ /* the NFSv4 user name that the user principal maps to */
utf8str_mixed ccap_username; utf8str_mixed ccap_username;
/* equal to seq_num of rpc_gss_cred_vers_3_t */ /* equal to seq_num of rpc_gss_cred_vers_3_t */
unsigned int ccap_seq_num; unsigned int ccap_seq_num;
}; };
4.4.1.2.1. Establishing a Security Context 2.4.1.2.1. Establishing a Security Context
When the user principal wants to COPY a file between two servers, if When the user principal wants to COPY a file between two servers, if
it has not established copy_from_auth and copy_to_auth privileges on it has not established copy_from_auth and copy_to_auth privileges on
the servers, it establishes them: the servers, it establishes them:
o The user principal generates a secret it will share with the two o The user principal generates a secret it will share with the two
servers. This shared secret will be placed in the servers. This shared secret will be placed in the
cfap_shared_secret and ctap_shared_secret fields of the cfap_shared_secret and ctap_shared_secret fields of the
appropriate privilege data types, copy_from_auth_priv and appropriate privilege data types, copy_from_auth_priv and
copy_to_auth_priv. copy_to_auth_priv.
skipping to change at page 26, line 35 skipping to change at page 21, line 35
rpc_gss3_assertion server_assertions<>; rpc_gss3_assertion server_assertions<>;
rpc_gss3_extension extensions<>; rpc_gss3_extension extensions<>;
} rpc_gss3_create_res; } rpc_gss3_create_res;
The field "handle" is the RPCSEC_GSSv3 handle that the client will The field "handle" is the RPCSEC_GSSv3 handle that the client will
use on COPY requests involving the source and destination server. use on COPY requests involving the source and destination server.
The field granted_assertions[0].privs will be equal to The field granted_assertions[0].privs will be equal to
"copy_to_auth". The server will return a GSS_Wrap() of "copy_to_auth". The server will return a GSS_Wrap() of
copy_to_auth_priv. copy_to_auth_priv.
4.4.1.2.2. Starting a Secure Inter-Server Copy 2.4.1.2.2. Starting a Secure Inter-Server Copy
When the client sends a COPY_NOTIFY request to the source server, it When the client sends a COPY_NOTIFY request to the source server, it
uses the privileged "copy_from_auth" RPCSEC_GSSv3 handle. uses the privileged "copy_from_auth" RPCSEC_GSSv3 handle.
cna_destination_server in COPY_NOTIFY MUST be the same as the name of cna_destination_server in COPY_NOTIFY MUST be the same as the name of
the destination server specified in copy_from_auth_priv. Otherwise, the destination server specified in copy_from_auth_priv. Otherwise,
COPY_NOTIFY will fail with NFS4ERR_ACCESS. The source server COPY_NOTIFY will fail with NFS4ERR_ACCESS. The source server
verifies that the privilege <"copy_from_auth", user id, destination> verifies that the privilege <"copy_from_auth", user id, destination>
exists, and annotates it with the source filehandle, if the user exists, and annotates it with the source filehandle, if the user
principal has read access to the source file, and if administrative principal has read access to the source file, and if administrative
policies give the user principal and the NFS client read access to policies give the user principal and the NFS client read access to
skipping to change at page 27, line 24 skipping to change at page 22, line 24
If the client sends a COPY_REVOKE to the source server to rescind the If the client sends a COPY_REVOKE to the source server to rescind the
destination server's copy privilege, it uses the privileged destination server's copy privilege, it uses the privileged
"copy_from_auth" RPCSEC_GSSv3 handle and the cra_destination_server "copy_from_auth" RPCSEC_GSSv3 handle and the cra_destination_server
in COPY_REVOKE MUST be the same as the name of the destination server in COPY_REVOKE MUST be the same as the name of the destination server
specified in copy_from_auth_priv. The source server will then delete specified in copy_from_auth_priv. The source server will then delete
the <"copy_from_auth", user id, destination> privilege and fail any the <"copy_from_auth", user id, destination> privilege and fail any
subsequent copy requests sent under the auspices of this privilege subsequent copy requests sent under the auspices of this privilege
from the destination server. from the destination server.
4.4.1.2.3. Securing ONC RPC Server-to-Server Copy Protocols 2.4.1.2.3. Securing ONC RPC Server-to-Server Copy Protocols
After a destination server has a "copy_to_auth" privilege established After a destination server has a "copy_to_auth" privilege established
on it, and it receives a COPY request, if it knows it will use an ONC on it, and it receives a COPY request, if it knows it will use an ONC
RPC protocol to copy data, it will establish a "copy_confirm_auth" RPC protocol to copy data, it will establish a "copy_confirm_auth"
privilege on the source server, using nfs@<destination> as the privilege on the source server, using nfs@<destination> as the
initiator principal, and nfs@<source> as the target principal. initiator principal, and nfs@<source> as the target principal.
The value of the field ccap_shared_secret_mic is a GSS_VerifyMIC() of The value of the field ccap_shared_secret_mic is a GSS_VerifyMIC() of
the shared secret passed in the copy_to_auth privilege. The field the shared secret passed in the copy_to_auth privilege. The field
ccap_username is the mapping of the user principal to an NFSv4 user ccap_username is the mapping of the user principal to an NFSv4 user
skipping to change at page 28, line 16 skipping to change at page 23, line 16
policies can be overridden in case the destination server as-an- policies can be overridden in case the destination server as-an-
NFS-client is not authorized NFS-client is not authorized
o manual configuration to allow a copy relationship between the o manual configuration to allow a copy relationship between the
source and destination is not needed. source and destination is not needed.
If the attempt to establish a "copy_confirm_auth" privilege fails, If the attempt to establish a "copy_confirm_auth" privilege fails,
then when the user principal sends a COPY request to destination, the then when the user principal sends a COPY request to destination, the
destination server will reject it with NFS4ERR_PARTNER_NO_AUTH. destination server will reject it with NFS4ERR_PARTNER_NO_AUTH.
4.4.1.2.4. Securing Non ONC RPC Server-to-Server Copy Protocols 2.4.1.2.4. Securing Non ONC RPC Server-to-Server Copy Protocols
If the destination won't be using ONC RPC to copy the data, then the If the destination won't be using ONC RPC to copy the data, then the
source and destination are using an unspecified copy protocol. The source and destination are using an unspecified copy protocol. The
destination could use the shared secret and the NFSv4 user id to destination could use the shared secret and the NFSv4 user id to
prove to the source server that the user principal has authorized the prove to the source server that the user principal has authorized the
copy. copy.
For protocols that authenticate user names with passwords (e.g., HTTP For protocols that authenticate user names with passwords (e.g., HTTP
[14] and FTP [15]), the nfsv4 user id could be used as the user name, [13] and FTP [14]), the nfsv4 user id could be used as the user name,
and an ASCII hexadecimal representation of the RPCSEC_GSSv3 shared and an ASCII hexadecimal representation of the RPCSEC_GSSv3 shared
secret could be used as the user password or as input into non- secret could be used as the user password or as input into non-
password authentication methods like CHAP [16]. password authentication methods like CHAP [15].
4.4.1.3. Inter-Server Copy via ONC RPC but without RPCSEC_GSSv3 2.4.1.3. Inter-Server Copy via ONC RPC but without RPCSEC_GSSv3
ONC RPC security flavors other than RPCSEC_GSSv3 MAY be used with the ONC RPC security flavors other than RPCSEC_GSSv3 MAY be used with the
server-side copy offload operations described in this document. In server-side copy offload operations described in this document. In
particular, host-based ONC RPC security flavors such as AUTH_NONE and particular, host-based ONC RPC security flavors such as AUTH_NONE and
AUTH_SYS MAY be used. If a host-based security flavor is used, a AUTH_SYS MAY be used. If a host-based security flavor is used, a
minimal level of protection for the server-to-server copy protocol is minimal level of protection for the server-to-server copy protocol is
possible. possible.
In the absence of strong security mechanisms such as RPCSEC_GSSv3, In the absence of strong security mechanisms such as RPCSEC_GSSv3,
the challenge is how the source server and destination server the challenge is how the source server and destination server
skipping to change at page 29, line 30 skipping to change at page 24, line 30
from 192.168.33.56 over this network using NFSv4.1, it does the from 192.168.33.56 over this network using NFSv4.1, it does the
following: following:
COMPOUND { PUTROOTFH, LOOKUP "_COPY" ; LOOKUP "10.11.78.56"; LOOKUP COMPOUND { PUTROOTFH, LOOKUP "_COPY" ; LOOKUP "10.11.78.56"; LOOKUP
"_FH" ; OPEN "0x12345" ; GETFH } "_FH" ; OPEN "0x12345" ; GETFH }
The source server will therefore know that these NFSv4.1 operations The source server will therefore know that these NFSv4.1 operations
are being issued by the destination server identified in the are being issued by the destination server identified in the
COPY_NOTIFY. COPY_NOTIFY.
4.4.1.4. Inter-Server Copy without ONC RPC and RPCSEC_GSSv3 2.4.1.4. Inter-Server Copy without ONC RPC and RPCSEC_GSSv3
The same techniques as Section 4.4.1.3, using unique URLs for each
destination server, can be used for other protocols (e.g., HTTP [14]
and FTP [15]) as well.
5. Application Data Block Support
At the OS level, files are contained on disk blocks. Applications
are also free to impose structure on the data contained in a file and
we can define an Application Data Block (ADB) to be such a structure.
From the application's viewpoint, it only wants to handle ADBs and
not raw bytes (see [17]). An ADB is typically comprised of two
sections: a header and data. The header describes the
characteristics of the block and can provide a means to detect
corruption in the data payload. The data section is typically
initialized to all zeros.
The format of the header is application specific, but there are two
main components typically encountered:
1. An ADB Number (ADBN), which allows the application to determine
which data block is being referenced. The ADBN is a logical
block number and is useful when the client is not storing the
blocks in contiguous memory.
2. Fields to describe the state of the ADB and a means to detect
block corruption. For both pieces of data, a useful property is
that allowed values be unique in that if passed across the
network, corruption due to translation between big and little
endian architectures are detectable. For example, 0xF0DEDEF0 has
the same bit pattern in both architectures.
Applications already impose structures on files [17] and detect
corruption in data blocks [18]. What they are not able to do is
efficiently transfer and store ADBs. To initialize a file with ADBs,
the client must send the full ADB to the server and that must be
stored on the server. When the application is initializing a file to
have the ADB structure, it could compress the ADBs to just the
information to necessary to later reconstruct the header portion of
the ADB when the contents are read back. Using sparse file
techniques, the disk blocks described by would not be allocated.
Unlike sparse file techniques, there would be a small cost to store
the compressed header data.
In this section, we are going to define a generic framework for an
ADB, present one approach to detecting corruption in a given ADB
implementation, and describe the model for how the client and server
can support efficient initialization of ADBs, reading of ADB holes,
punching holes in ADBs, and space reservation. Further, we need to
be able to extend this model to applications which do not support
ADBs, but wish to be able to handle sparse files, hole punching, and
space reservation.
5.1. Generic Framework
We want the representation of the ADB to be flexible enough to
support many different applications. The most basic approach is no
imposition of a block at all, which means we are working with the raw
bytes. Such an approach would be useful for storing holes, punching
holes, etc. In more complex deployments, a server might be
supporting multiple applications, each with their own definition of
the ADB. One might store the ADBN at the start of the block and then
have a guard pattern to detect corruption [19]. The next might store
the ADBN at an offset of 100 bytes within the block and have no guard
pattern at all. The point is that existing applications might
already have well defined formats for their data blocks.
The guard pattern can be used to represent the state of the block, to
protect against corruption, or both. Again, it needs to be able to
be placed anywhere within the ADB.
We need to be able to represent the starting offset of the block and
the size of the block. Note that nothing prevents the application
from defining different sized blocks in a file.
5.1.1. Data Block Representation
struct app_data_block4 {
offset4 adb_offset;
length4 adb_block_size;
length4 adb_block_count;
length4 adb_reloff_blocknum;
count4 adb_block_num;
length4 adb_reloff_pattern;
opaque adb_pattern<>;
};
The app_data_block4 structure captures the abstraction presented for
the ADB. The additional fields present are to allow the transmission
of adb_block_count ADBs at one time. We also use adb_block_num to
convey the ADBN of the first block in the sequence. Each ADB will
contain the same adb_pattern string.
As both adb_block_num and adb_pattern are optional, if either
adb_reloff_pattern or adb_reloff_blocknum is set to NFS4_UINT64_MAX,
then the corresponding field is not set in any of the ADB.
5.1.2. Data Content
/*
* Use an enum such that we can extend new types.
*/
enum data_content4 {
NFS4_CONTENT_DATA = 0,
NFS4_CONTENT_APP_BLOCK = 1,
NFS4_CONTENT_HOLE = 2
};
New operations might need to differentiate between wanting to access
data versus an ADB. Also, future minor versions might want to
introduce new data formats. This enumeration allows that to occur.
5.2. pNFS Considerations
While this document does not mandate how sparse ADBs are recorded on
the server, it does make the assumption that such information is not
in the file. I.e., the information is metadata. As such, the
INITIALIZE operation is defined to be not supported by the DS - it
must be issued to the MDS. But since the client must not assume a
priori whether a read is sparse or not, the READ_PLUS operation MUST
be supported by both the DS and the MDS. I.e., the client might
impose on the MDS to asynchronously read the data from the DS.
Furthermore, each DS MUST not report to a client either a sparse ADB
or data which belongs to another DS. One implication of this
requirement is that the app_data_block4's adb_block_size MUST be
either be the stripe width or the stripe width must be an even
multiple of it.
The second implication here is that the DS must be able to use the
Control Protocol to determine from the MDS where the sparse ADBs
occur. [[Comment.4: Need to discuss what happens if after the file
is being written to and an INITIALIZE occurs? --TH]] Perhaps instead
of the DS pulling from the MDS, the MDS pushes to the DS? Thus an
INITIALIZE causes a new push? [[Comment.5: Still need to consider
race cases of the DS getting a WRITE and the MDS getting an
INITIALIZE. --TH]]
5.3. An Example of Detecting Corruption
In this section, we define an ADB format in which corruption can be
detected. Note that this is just one possible format and means to
detect corruption.
Consider a very basic implementation of an operating system's disk
blocks. A block is either data or it is an indirect block which
allows for files to be larger than one block. It is desired to be
able to initialize a block. Lastly, to quickly unlink a file, a
block can be marked invalid. The contents remain intact - which
would enable this OS application to undelete a file.
The application defines 4k sized data blocks, with an 8 byte block
counter occurring at offset 0 in the block, and with the guard
pattern occurring at offset 8 inside the block. Furthermore, the
guard pattern can take one of four states:
0xfeedface - This is the FREE state and indicates that the ADB
format has been applied.
0xcafedead - This is the DATA state and indicates that real data
has been written to this block.
0xe4e5c001 - This is the INDIRECT state and indicates that the
block contains block counter numbers that are chained off of this
block.
0xba1ed4a3 - This is the INVALID state and indicates that the block
contains data whose contents are garbage.
Finally, it also defines an 8 byte checksum [20] starting at byte 16
which applies to the remaining contents of the block. If the state
is FREE, then that checksum is trivially zero. As such, the
application has no need to transfer the checksum implicitly inside
the ADB - it need not make the transfer layer aware of the fact that
there is a checksum (see [18] for an example of checksums used to
detect corruption in application data blocks).
Corruption in each ADB can be detected thusly:
o If the guard pattern is anything other than one of the allowed
values, including all zeros.
o If the guard pattern is FREE and any other byte in the remainder
of the ADB is anything other than zero.
o If the guard pattern is anything other than FREE, then if the
stored checksum does not match the computed checksum.
o If the guard pattern is INDIRECT and one of the stored indirect
block numbers has a value greater than the number of ADBs in the
file.
o If the guard pattern is INDIRECT and one of the stored indirect
block numbers is a duplicate of another stored indirect block
number.
As can be seen, the application can detect errors based on the
combination of the guard pattern state and the checksum. But also,
the application can detect corruption based on the state and the
contents of the ADB. This last point is important in validating the
minimum amount of data we incorporated into our generic framework.
I.e., the guard pattern is sufficient in allowing applications to
design their own corruption detection.
Finally, it is important to note that none of these corruption checks
occur in the transport layer. The server and client components are
totally unaware of the file format and might report everything as
being transferred correctly even in the case the application detects
corruption.
5.4. Example of READ_PLUS
The hypothetical application presented in Section 5.3 can be used to
illustrate how READ_PLUS would return an array of results. A file is
created and initialized with 100 4k ADBs in the FREE state:
INITIALIZE {0, 4k, 100, 0, 0, 8, 0xfeedface}
Further, assume the application writes a single ADB at 16k, changing
the guard pattern to 0xcafedead, we would then have in memory:
0 -> (16k - 1) : 4k, 4, 0, 0, 8, 0xfeedface
16k -> (20k - 1) : 00 00 00 05 ca fe de ad XX XX ... XX XX
20k -> 400k : 4k, 95, 0, 6, 0xfeedface
And when the client did a READ_PLUS of 64k at the start of the file,
it would get back a result of an ADB, some data, and a final ADB:
ADB {0, 4, 0, 0, 8, 0xfeedface}
data 4k
ADB {20k, 4k, 59, 0, 6, 0xfeedface}
5.5. Zero Filled Holes
As applications are free to define the structure of an ADB, it is
trivial to define an ADB which supports zero filled holes. Such a
case would encompass the traditional definitions of a sparse file and
hole punching. For example, to punch a 64k hole, starting at 100M,
into an existing file which has no ADB structure:
INITIALIZE {100M, 64k, 1, NFS4_UINT64_MAX,
0, NFS4_UINT64_MAX, 0x0}
6. Space Reservation
6.1. Introduction
This section describes a set of operations that allow applications
such as hypervisors to reserve space for a file, report the amount of
actual disk space a file occupies and freeup the backing space of a
file when it is not required.
In virtualized environments, virtual disk files are often stored on
NFS mounted volumes. Since virtual disk files represent the hard
disks of virtual machines, hypervisors often have to guarantee
certain properties for the file.
One such example is space reservation. When a hypervisor creates a
virtual disk file, it often tries to preallocate the space for the
file so that there are no future allocation related errors during the
operation of the virtual machine. Such errors prevent a virtual
machine from continuing execution and result in downtime.
Another useful feature would be the ability to report the number of
blocks that would be freed when a file is deleted. Currently, NFS
reports two size attributes:
size The logical file size of the file.
space_used The size in bytes that the file occupies on disk
While these attributes are sufficient for space accounting in
traditional filesystems, they prove to be inadequate in modern
filesystems that support block sharing. Having a way to tell the
number of blocks that would be freed if the file was deleted would be
useful to applications that wish to migrate files when a volume is
low on space.
Since virtual disks represent a hard drive in a virtual machine, a
virtual disk can be viewed as a filesystem within a file. Since not
all blocks within a filesystem are in use, there is an opportunity to
reclaim blocks that are no longer in use. A call to deallocate
blocks could result in better space efficiency. Lesser space MAY be
consumed for backups after block deallocation.
We propose the following operations and attributes for the
aforementioned use cases:
space_reserved This attribute specifies whether the blocks backing
the file have been preallocated.
space_freed This attribute specifies the space freed when a file is
deleted, taking block sharing into consideration.
max_hole_punch This attribute specifies the maximum sized hole that
can be punched on the filesystem.
HOLE_PUNCH This operation zeroes and/or deallocates the blocks
backing a region of the file.
6.2. Use Cases
6.2.1. Space Reservation
Some applications require that once a file of a certain size is
created, writes to that file never fail with an out of space
condition. One such example is that of a hypervisor writing to a
virtual disk. An out of space condition while writing to virtual
disks would mean that the virtual machine would need to be frozen.
Currently, in order to achieve such a guarantee, applications zero
the entire file. The initial zeroing allocates the backing blocks
and all subsequent writes are overwrites of already allocated blocks.
This approach is not only inefficient in terms of the amount of I/O
done, it is also not guaranteed to work on filesystems that are log
structured or deduplicated. An efficient way of guaranteeing space
reservation would be beneficial to such applications.
If the space_reserved attribute is set on a file, it is guaranteed
that writes that do not grow the file will not fail with
NFSERR_NOSPC.
6.2.2. Space freed on deletes
Currently, files in NFS have two size attributes:
size The logical file size of the file.
space_used The size in bytes that the file occupies on disk.
While these attributes are sufficient for space accounting in
traditional filesystems, they prove to be inadequate in modern
filesystems that support block sharing. In such filesystems,
multiple inodes can point to a single block with a block reference
count to guard against premature freeing.
If space_used of a file is interpreted to mean the size in bytes of
all disk blocks pointed to by the inode of the file, then shared
blocks get double counted, over-reporting the space utilization.
This also has the adverse effect that the deletion of a file with
shared blocks frees up less than space_used bytes.
On the other hand, if space_used is interpreted to mean the size in
bytes of those disk blocks unique to the inode of the file, then
shared blocks are not counted in any file, resulting in under-
reporting of the space utilization.
For example, two files A and B have 10 blocks each. Let 6 of these
blocks be shared between them. Thus, the combined space utilized by
the two files is 14 * BLOCK_SIZE bytes. In the former case, the
combined space utilization of the two files would be reported as 20 *
BLOCK_SIZE. However, deleting either would only result in 4 *
BLOCK_SIZE being freed. Conversely, the latter interpretation would
report that the space utilization is only 8 * BLOCK_SIZE.
Adding another size attribute, space_freed, is helpful in solving
this problem. space_freed is the number of blocks that are allocated
to the given file that would be freed on its deletion. In the
example, both A and B would report space_freed as 4 * BLOCK_SIZE and
space_used as 10 * BLOCK_SIZE. If A is deleted, B will report
space_freed as 10 * BLOCK_SIZE as the deletion of B would result in
the deallocation of all 10 blocks.
The addition of this problem doesn't solve the problem of space being
over-reported. However, over-reporting is better than under-
reporting.
6.2.3. Operations and attributes
In the sections that follow, one operation and three attributes are
defined that together provide the space management facilities
outlined earlier in the document. The operation is intended to be
OPTIONAL and the attributes RECOMMENDED as defined in section 17 of
[2].
6.2.4. Attribute 77: space_reserved
The space_reserve attribute is a read/write attribute of type
boolean. It is a per file attribute. When the space_reserved
attribute is set via SETATTR, the server must ensure that there is
disk space to accommodate every byte in the file before it can return
success. If the server cannot guarantee this, it must return
NFS4ERR_NOSPC.
If the client tries to grow a file which has the space_reserved
attribute set, the server must guarantee that there is disk space to
accommodate every byte in the file with the new size before it can
return success. If the server cannot guarantee this, it must return
NFS4ERR_NOSPC.
It is not required that the server allocate the space to the file
before returning success. The allocation can be deferred, however,
it must be guaranteed that it will not fail for lack of space.
The value of space_reserved can be obtained at any time through
GETATTR.
In order to avoid ambiguity, the space_reserve bit cannot be set
along with the size bit in SETATTR. Increasing the size of a file
with space_reserve set will fail if space reservation cannot be
guaranteed for the new size. If the file size is decreased, space
reservation is only guaranteed for the new size and the extra blocks
backing the file can be released.
6.2.5. Attribute 78: space_freed
space_freed gives the number of bytes freed if the file is deleted.
This attribute is read only and is of type length4. It is a per file
attribute.
6.2.6. Attribute 79: max_hole_punch
max_hole_punch specifies the maximum size of a hole that the
HOLE_PUNCH operation can handle. This attribute is read only and of
type length4. It is a per filesystem attribute. This attribute MUST
be implemented if HOLE_PUNCH is implemented.
6.2.7. Operation 64: HOLE_PUNCH - Zero and deallocate blocks backing
the file in the specified range.
WARNING: Most of this section is now obsolete. Parts of it need to
be scavanged for the ADB discussion, but for the most part, it cannot
be trusted.
6.2.7.1. DESCRIPTION
Whenever a client wishes to deallocate the blocks backing a
particular region in the file, it calls the HOLE_PUNCH operation with
the current filehandle set to the filehandle of the file in question,
start offset and length in bytes of the region set in hpa_offset and
hpa_count respectively. All further reads to this region MUST return
zeros until overwritten. The filehandle specified must be that of a
regular file.
Situations may arise where hpa_offset and/or hpa_offset + hpa_count
will not be aligned to a boundary that the server does allocations/
deallocations in. For most filesystems, this is the block size of
the file system. In such a case, the server can deallocate as many
bytes as it can in the region. The blocks that cannot be deallocated
MUST be zeroed. Except for the block deallocation and maximum hole
punching capability, a HOLE_PUNCH operation is to be treated similar
to a write of zeroes.
The server is not required to complete deallocating the blocks
specified in the operation before returning. It is acceptable to
have the deallocation be deferred. In fact, HOLE_PUNCH is merely a
hint; it is valid for a server to return success without ever doing
anything towards deallocating the blocks backing the region
specified. However, any future reads to the region MUST return
zeroes.
HOLE_PUNCH will result in the space_used attribute being decreased by
the number of bytes that were deallocated. The space_freed attribute
may or may not decrease, depending on the support and whether the
blocks backing the specified range were shared or not. The size
attribute will remain unchanged.
The HOLE_PUNCH operation MUST NOT change the space reservation
guarantee of the file. While the server can deallocate the blocks
specified by hpa_offset and hpa_count, future writes to this region
MUST NOT fail with NFSERR_NOSPC.
The HOLE_PUNCH operation may fail for the following reasons (this is
a partial list):
NFS4ERR_NOTSUPP The Hole punch operations are not supported by the
NFS server receiving this request.
NFS4ERR_DIR The current filehandle is of type NF4DIR.
NFS4ERR_SYMLINK The current filehandle is of type NF4LNK.
NFS4ERR_WRONG_TYPE The current filehandle does not designate an
ordinary file.
7. Sparse Files The same techniques as Section 2.4.1.3, using unique URLs for each
destination server, can be used for other protocols (e.g., HTTP [13]
and FTP [14]) as well.
WARNING: Some of this section needs to be reworked because of the 3. Sparse Files
work going on in the ADB section.
7.1. Introduction 3.1. Introduction
A sparse file is a common way of representing a large file without A sparse file is a common way of representing a large file without
having to utilize all of the disk space for it. Consequently, a having to utilize all of the disk space for it. Consequently, a
sparse file uses less physical space than its size indicates. This sparse file uses less physical space than its size indicates. This
means the file contains 'holes', byte ranges within the file that means the file contains 'holes', byte ranges within the file that
contain no data. Most modern file systems support sparse files, contain no data. Most modern file systems support sparse files,
including most UNIX file systems and NTFS, but notably not Apple's including most UNIX file systems and NTFS, but notably not Apple's
HFS+. Common examples of sparse files include Virtual Machine (VM) HFS+. Common examples of sparse files include Virtual Machine (VM)
OS/disk images, database files, log files, and even checkpoint OS/disk images, database files, log files, and even checkpoint
recovery files most commonly used by the HPC community. recovery files most commonly used by the HPC community.
If an application reads a hole in a sparse file, the file system must If an application reads a hole in a sparse file, the file system must
returns all zeros to the application. For local data access there is return all zeros to the application. For local data access there is
little penalty, but with NFS these zeroes must be transferred back to little penalty, but with NFS these zeroes must be transferred back to
the client. If an application uses the NFS client to read data into the client. If an application uses the NFS client to read data into
memory, this wastes time and bandwidth as the application waits for memory, this wastes time and bandwidth as the application waits for
the zeroes to be transferred. the zeroes to be transferred.
A sparse file is typically created by initializing the file to be all A sparse file is typically created by initializing the file to be all
zeros - nothing is written to the data in the file, instead the hole zeros - nothing is written to the data in the file, instead the hole
is recorded in the metadata for the file. So a 8G disk image might is recorded in the metadata for the file. So a 8G disk image might
be represented initially by a couple hundred bits in the inode and be represented initially by a couple hundred bits in the inode and
nothing on the disk. If the VM then writes 100M to a file in the nothing on the disk. If the VM then writes 100M to a file in the
middle of the image, there would now be two holes represented in the middle of the image, there would now be two holes represented in the
metadata and 100M in the data. metadata and 100M in the data.
Other applications want to initialize a file to patterns other than This section introduces a new operation READ_PLUS which supports all
zero. The problem with initializing to zero is that it is often
difficult to distinguish a byte-range of initialized to all zeroes
from data corruption, since a pattern of zeroes is a probable pattern
for corruption. Instead, some applications, such as database
management systems, use pattern consisting of bytes or words of non-
zero values.
Besides reading sparse files and initializing them, applications
might want to hole punch, which is the deallocation of the data
blocks which back a region of the file. At such time, the affected
blocks are reinitialized to a pattern.
This section introduces a new operation to read patterns from a file,
READ_PLUS, and a new operation to both initialize patterns and to
punch pattern holes into a file, WRITE_PLUS. READ_PLUS supports all
the features of READ but includes an extension to support sparse the features of READ but includes an extension to support sparse
pattern files. READ_PLUS is guaranteed to perform no worse than pattern files. READ_PLUS is guaranteed to perform no worse than
READ, and can dramatically improve performance with sparse files. READ, and can dramatically improve performance with sparse files.
READ_PLUS does not depend on pNFS protocol features, but can be used READ_PLUS does not depend on pNFS protocol features, but can be used
by pNFS to support sparse files. by pNFS to support sparse files.
7.2. Terminology 3.2. Terminology
Regular file: An object of file type NF4REG or NF4NAMEDATTR. Regular file: An object of file type NF4REG or NF4NAMEDATTR.
Sparse file: A Regular file that contains one or more Holes. Sparse file: A Regular file that contains one or more Holes.
Hole: A byte range within a Sparse file that contains regions of all Hole: A byte range within a Sparse file that contains regions of all
zeroes. For block-based file systems, this could also be an zeroes. For block-based file systems, this could also be an
unallocated region of the file. unallocated region of the file.
Hole Threshold The minimum length of a Hole as determined by the Hole Threshold The minimum length of a Hole as determined by the
server. If a server chooses to define a Hole Threshold, then it server. If a server chooses to define a Hole Threshold, then it
would not return hole information (nfs_readplusreshole) with a would not return hole information (nfs_readplusreshole) with a
hole_offset and hole_length that specify a range shorter than the hole_offset and hole_length that specify a range shorter than the
Hole Threshold. Hole Threshold.
7.3. Applications and Sparse Files 3.3. Overview of Sparse Files and NFSv4
Applications may cause an NFS client to read holes in a file for
several reasons. This section describes three different application
workloads that cause the NFS client to transfer data unnecessarily.
These workloads are simply examples, and there are probably many more
workloads that are negatively impacted by sparse files.
The first workload that can cause holes to be read is sequential
reads within a sparse file. When this happens, the NFS client may
perform read requests ("readahead") into sections of the file not
explicitly requested by the application. Since the NFS client cannot
differentiate between holes and non-holes, the NFS client may
prefetch empty sections of the file.
This workload is exemplified by Virtual Machines and their associated
file system images, e.g., VMware .vmdk files, which are large sparse
files encapsulating an entire operating system. If a VM reads files
within the file system image, this will translate to sequential NFS
read requests into the much larger file system image file. Since NFS
does not understand the internals of the file system image, it ends
up performing readahead file holes.
The second workload is generated by copying a file from a directory
in NFS to either the same NFS server, to another file system, e.g.,
another NFS or Samba server, to a local ext3 file system, or even a
network socket. In this case, bandwidth and server resources are
wasted as the entire file is transferred from the NFS server to the
NFS client. Once a byte range of the file has been transferred to
the client, it is up to the client application, e.g., rsync, cp, scp,
on how it writes the data to the target location. For example, cp
supports sparse files and will not write all zero regions, whereas
scp does not support sparse files and will transfer every byte of the
file.
The third workload is generated by applications that do not utilize
the NFS client cache, but instead use direct I/O and manage cached
data independently, e.g., databases. These applications may perform
whole file caching with sparse files, which would mean that even the
holes will be transferred to the clients and cached.
7.4. Overview of Sparse Files and NFSv4
This proposal seeks to provide sparse file support to the largest
number of NFS client and server implementations, and as such proposes
to add a new return code to the mandatory NFSv4.1 READ_PLUS operation
instead of proposing additions or extensions of new or existing
optional features (such as pNFS).
As well, this document seeks to ensure that the proposed extensions
are simple and do not transfer data between the client and server
unnecessarily. For example, one possible way to implement sparse
file read support would be to have the client, on the first hole
encountered or at OPEN time, request a Data Region Map from the
server. A Data Region Map would specify all zero and non-zero
regions in a file. While this option seems simple, it is less useful
and can become inefficient and cumbersome for several reasons:
o Data Region Maps can be large, and transferring them can reduce
overall read performance. For example, VMware's .vmdk files can
have a file size of over 100 GBs and have a map well over several
MBs.
o Data Region Maps can change frequently, and become invalidated on
every write to the file. NFSv4 has a single change attribute,
which means any change to any region of a file will invalidate all
Data Region Maps. This can result in the map being transferred
multiple times with each update to the file. For example, a VM
that updates a config file in its file system image would
invalidate the Data Region Map not only for itself, but for all
other clients accessing the same file system image.
o Data Region Maps do not handle all zero-filled sections of the
file, reducing the effectiveness of the solution. While it may be
possible to modify the maps to handle zero-filled sections (at
possibly great effort to the server), it is almost impossible with
pNFS. With pNFS, the owner of the Data Region Map is the metadata
server, which is not in the data path and has no knowledge of the
contents of a data region.
Another way to handle holes is compression, but this not ideal since
it requires all implementations to agree on a single compression
algorithm and requires a fair amount of computational overhead.
Note that supporting writing to a sparse file does not require This section provides sparse file support to the largest number of
changes to the protocol. Applications and/or NFS implementations can NFS client and server implementations, and as such proposes to add a
choose to ignore WRITE requests of all zeroes to the NFS server new return code to the READ_PLUS operation instead of proposing
without consequence. additions or extensions of new or existing optional features (such as
pNFS).
7.5. Operation 65: READ_PLUS 3.4. Operation 65: READ_PLUS
The section introduces a new read operation, named READ_PLUS, which The section introduces a new read operation, named READ_PLUS, which
allows NFS clients to avoid reading holes in a sparse file. allows NFS clients to avoid reading holes in a sparse file.
READ_PLUS is guaranteed to perform no worse than READ, and can READ_PLUS is guaranteed to perform no worse than READ, and can
dramatically improve performance with sparse files. dramatically improve performance with sparse files.
READ_PLUS supports all the features of the existing NFSv4.1 READ READ_PLUS supports all the features of the existing NFSv4.1 READ
operation [2] and adds a simple yet significant extension to the operation [2] and adds a simple yet significant extension to the
format of its response. The change allows the client to avoid format of its response. The change allows the client to avoid
returning all zeroes from a file hole, wasting computational and returning all zeroes from a file hole, wasting computational and
skipping to change at page 43, line 31 skipping to change at page 26, line 31
contain information about a file that may not even be read in its contain information about a file that may not even be read in its
entirely. entirely.
A new read operation is required due to NFSv4.1 minor versioning A new read operation is required due to NFSv4.1 minor versioning
rules that do not allow modification of existing operation's rules that do not allow modification of existing operation's
arguments or results. READ_PLUS is designed in such a way to allow arguments or results. READ_PLUS is designed in such a way to allow
future extensions to the result structure. The same approach could future extensions to the result structure. The same approach could
be taken to extend the argument structure, but a good use case is be taken to extend the argument structure, but a good use case is
first required to make such a change. first required to make such a change.
7.5.1. ARGUMENT 3.4.1. ARGUMENT
struct READ_PLUS4args { struct READ_PLUS4args {
/* CURRENT_FH: file */ /* CURRENT_FH: file */
stateid4 rpa_stateid; stateid4 rpa_stateid;
offset4 rpa_offset; offset4 rpa_offset;
count4 rpa_count; count4 rpa_count;
}; };
7.5.2. RESULT 3.4.2. RESULT
union read_plus_content switch (data_content4 content) { union read_plus_content switch (data_content4 content) {
case NFS4_CONTENT_DATA: case NFS4_CONTENT_DATA:
opaque rpc_data<>; opaque rpc_data<>;
case NFS4_CONTENT_APP_BLOCK: case NFS4_CONTENT_APP_BLOCK:
app_data_block4 rpc_block; app_data_block4 rpc_block;
case NFS4_CONTENT_HOLE: case NFS4_CONTENT_HOLE:
hole_info4 rpc_hole; hole_info4 rpc_hole;
default: default:
void; void;
skipping to change at page 44, line 33 skipping to change at page 27, line 33
read_plus_content rpr_contents<>; read_plus_content rpr_contents<>;
}; };
union READ_PLUS4res switch (nfsstat4 status) { union READ_PLUS4res switch (nfsstat4 status) {
case NFS4_OK: case NFS4_OK:
read_plus_res4 resok4; read_plus_res4 resok4;
default: default:
void; void;
}; };
7.5.3. DESCRIPTION 3.4.3. DESCRIPTION
The READ_PLUS operation is based upon the NFSv4.1 READ operation [2], The READ_PLUS operation is based upon the NFSv4.1 READ operation [2],
and similarly reads data from the regular file identified by the and similarly reads data from the regular file identified by the
current filehandle. current filehandle.
The client provides an offset of where the READ_PLUS is to start and The client provides an offset of where the READ_PLUS is to start and
a count of how many bytes are to be read. An offset of zero means to a count of how many bytes are to be read. An offset of zero means to
read data starting at the beginning of the file. If offset is read data starting at the beginning of the file. If offset is
greater than or equal to the size of the file, the status NFS4_OK is greater than or equal to the size of the file, the status NFS4_OK is
returned with nfs_readplusrestype4 set to READ_OK, data length set to returned with nfs_readplusrestype4 set to READ_OK, data length set to
skipping to change at page 46, line 15 skipping to change at page 29, line 15
For a READ_PLUS with a stateid value of all bits equal to zero, the For a READ_PLUS with a stateid value of all bits equal to zero, the
server MAY allow the READ_PLUS to be serviced subject to mandatory server MAY allow the READ_PLUS to be serviced subject to mandatory
byte-range locks or the current share deny modes for the file. For a byte-range locks or the current share deny modes for the file. For a
READ_PLUS with a stateid value of all bits equal to one, the server READ_PLUS with a stateid value of all bits equal to one, the server
MAY allow READ_PLUS operations to bypass locking checks at the MAY allow READ_PLUS operations to bypass locking checks at the
server. server.
On success, the current filehandle retains its value. On success, the current filehandle retains its value.
7.5.4. IMPLEMENTATION 3.4.4. IMPLEMENTATION
If the server returns a "short read" (i.e., fewer data than requested If the server returns a "short read" (i.e., fewer data than requested
and eof is set to FALSE), the client should send another READ_PLUS to and eof is set to FALSE), the client should send another READ_PLUS to
get the remaining data. A server may return less data than requested get the remaining data. A server may return less data than requested
under several circumstances. The file may have been truncated by under several circumstances. The file may have been truncated by
another client or perhaps on the server itself, changing the file another client or perhaps on the server itself, changing the file
size from what the requesting client believes to be the case. This size from what the requesting client believes to be the case. This
would reduce the actual amount of data available to the client. It would reduce the actual amount of data available to the client. It
is possible that the server reduce the transfer size and so return a is possible that the server reduce the transfer size and so return a
short read result. Server resource exhaustion may also occur in a short read result. Server resource exhaustion may also occur in a
skipping to change at page 47, line 5 skipping to change at page 30, line 5
being read, the delegation must be recalled, and the operation cannot being read, the delegation must be recalled, and the operation cannot
proceed until that delegation is returned or revoked. Except where proceed until that delegation is returned or revoked. Except where
this happens very quickly, one or more NFS4ERR_DELAY errors will be this happens very quickly, one or more NFS4ERR_DELAY errors will be
returned to requests made while the delegation remains outstanding. returned to requests made while the delegation remains outstanding.
Normally, delegations will not be recalled as a result of a READ_PLUS Normally, delegations will not be recalled as a result of a READ_PLUS
operation since the recall will occur as a result of an earlier OPEN. operation since the recall will occur as a result of an earlier OPEN.
However, since it is possible for a READ_PLUS to be done with a However, since it is possible for a READ_PLUS to be done with a
special stateid, the server needs to check for this case even though special stateid, the server needs to check for this case even though
the client should have done an OPEN previously. the client should have done an OPEN previously.
7.5.4.1. Additional pNFS Implementation Information 3.4.4.1. Additional pNFS Implementation Information
With pNFS, the semantics of using READ_PLUS remains the same. Any With pNFS, the semantics of using READ_PLUS remains the same. Any
data server MAY return a READ_HOLE result for a READ_PLUS request data server MAY return a READ_HOLE result for a READ_PLUS request
that it receives. that it receives.
When a data server chooses to return a READ_HOLE result, it has the When a data server chooses to return a READ_HOLE result, it has the
option of returning hole information for the data stored on that data option of returning hole information for the data stored on that data
server (as defined by the data layout), but it MUST not return a server (as defined by the data layout), but it MUST not return a
nfs_readplusreshole structure with a byte range that includes data nfs_readplusreshole structure with a byte range that includes data
managed by another data server. managed by another data server.
skipping to change at page 47, line 34 skipping to change at page 30, line 34
A data server should do its best to return as much information about A data server should do its best to return as much information about
a hole as is feasible without having to contact the metadata server. a hole as is feasible without having to contact the metadata server.
If communication with the metadata server is required, then every If communication with the metadata server is required, then every
attempt should be taken to minimize the number of requests. attempt should be taken to minimize the number of requests.
If mandatory locking is enforced, then the data server must also If mandatory locking is enforced, then the data server must also
ensure that to return only information for a Hole that is within the ensure that to return only information for a Hole that is within the
owner's locked byte range. owner's locked byte range.
7.5.5. READ_PLUS with Sparse Files Example 3.4.5. READ_PLUS with Sparse Files Example
To see how the return value READ_HOLE will work, the following table To see how the return value READ_HOLE will work, the following table
describes a sparse file. For each byte range, the file contains describes a sparse file. For each byte range, the file contains
either non-zero data or a hole. In addition, the server in this either non-zero data or a hole. In addition, the server in this
example uses a hole threshold of 32K. example uses a hole threshold of 32K.
+-------------+----------+ +-------------+----------+
| Byte-Range | Contents | | Byte-Range | Contents |
+-------------+----------+ +-------------+----------+
| 0-15999 | Hole | | 0-15999 | Hole |
skipping to change at page 48, line 30 skipping to change at page 31, line 30
3. READ_PLUS(s, 256K, 64K) --> NFS_OK, readplusrestype4 = READ_OK, 3. READ_PLUS(s, 256K, 64K) --> NFS_OK, readplusrestype4 = READ_OK,
eof = false, data<>[32K]. Return a short read, as the last half eof = false, data<>[32K]. Return a short read, as the last half
of the request was all zeroes. of the request was all zeroes.
4. READ_PLUS(s, 288K, 64K) --> NFS_OK, readplusrestype4 = READ_HOLE, 4. READ_PLUS(s, 288K, 64K) --> NFS_OK, readplusrestype4 = READ_HOLE,
nfs_readplusreshole(HOLE_INFO)(288K, 66K). nfs_readplusreshole(HOLE_INFO)(288K, 66K).
5. READ_PLUS(s, 354K, 64K) --> NFS_OK, readplusrestype4 = READ_OK, 5. READ_PLUS(s, 354K, 64K) --> NFS_OK, readplusrestype4 = READ_OK,
eof = true, data<>[64K]. eof = true, data<>[64K].
7.6. Related Work 3.5. Related Work
Solaris and ZFS support an extension to lseek(2) that allows Solaris and ZFS support an extension to lseek(2) that allows
applications to discover holes in a file. The values, SEEK_HOLE and applications to discover holes in a file. The values, SEEK_HOLE and
SEEK_DATA, allow clients to seek to the next hole or beginning of SEEK_DATA, allow clients to seek to the next hole or beginning of
data, respectively. data, respectively.
XFS supports the XFS_IOC_GETBMAP extended attribute, which returns XFS supports the XFS_IOC_GETBMAP extended attribute, which returns
the Data Region Map for a file. Clients can then use this the Data Region Map for a file. Clients can then use this
information to avoid reading holes in a file. information to avoid reading holes in a file.
NTFS and CIFS support the FSCTL_SET_SPARSE attribute, which allows NTFS and CIFS support the FSCTL_SET_SPARSE attribute, which allows
applications to control whether empty regions of the file are applications to control whether empty regions of the file are
preallocated and filled in with zeros or simply left unallocated. preallocated and filled in with zeros or simply left unallocated.
7.7. Other Proposed Designs 3.6. Other Proposed Designs
7.7.1. Multi-Data Server Hole Information 3.6.1. Multi-Data Server Hole Information
The current design prohibits pnfs data servers from returning hole The current design prohibits pnfs data servers from returning hole
information for regions of a file that are not stored on that data information for regions of a file that are not stored on that data
server. Having data servers return information regarding other data server. Having data servers return information regarding other data
servers changes the fundamental principal that all metadata servers changes the fundamental principal that all metadata
information comes from the metadata server. information comes from the metadata server.
Here is a brief description if we did choose to support multi-data Here is a brief description if we did choose to support multi-data
server hole information: server hole information:
For a data server that can obtain hole information for the entire For a data server that can obtain hole information for the entire
file without severe performance impact, it MAY return HOLE_INFO and file without severe performance impact, it MAY return HOLE_INFO and
the byte range of the entire file hole. When a pNFS client receives the byte range of the entire file hole. When a pNFS client receives
a READ_HOLE result and a non-empty nfs_readplusreshole structure, it a READ_HOLE result and a non-empty nfs_readplusreshole structure, it
MAY use this information in conjunction with a valid layout for the MAY use this information in conjunction with a valid layout for the
file to determine the next data server for the next region of data file to determine the next data server for the next region of data
that is not in a hole. that is not in a hole.
7.7.2. Data Result Array 3.6.2. Data Result Array
If a single read request contains one or more Holes with a length If a single read request contains one or more Holes with a length
greater than the Sparse Threshold, the current design would return greater than the Sparse Threshold, the current design would return
results indicating a short read to the client. A client would then results indicating a short read to the client. A client would then
send a series of read requests to the server to retrieve information send a series of read requests to the server to retrieve information
for the Holes and the remaining data. To avoid turning a single read for the Holes and the remaining data. To avoid turning a single read
request into several exchanges between the client and server, the request into several exchanges between the client and server, the
server may need to choose a relatively large Sparse Threshold in server may need to choose a relatively large Sparse Threshold in
order to decrease the number of short reads it creates. A large order to decrease the number of short reads it creates. A large
Sparse Threshold may miss many smaller holes, which in turn may Sparse Threshold may miss many smaller holes, which in turn may
skipping to change at page 49, line 40 skipping to change at page 32, line 40
To avoid this situation, one option is to have the READ_PLUS To avoid this situation, one option is to have the READ_PLUS
operation return information for multiple holes in a single return operation return information for multiple holes in a single return
value. This would allow several small holes to be described in a value. This would allow several small holes to be described in a
single read response without requiring multliple exchanges between single read response without requiring multliple exchanges between
the client and server. the client and server.
One important item to consider with returning an array of data chunks One important item to consider with returning an array of data chunks
is its impact on RDMA, which may use different block sizes on the is its impact on RDMA, which may use different block sizes on the
client and server (among other things). client and server (among other things).
7.7.3. User-Defined Sparse Mask 3.6.3. User-Defined Sparse Mask
Add mask (instead of just zeroes). Specified by server or client? Add mask (instead of just zeroes). Specified by server or client?
7.7.4. Allocated flag 3.6.4. Allocated flag
A Hole on the server may be an allocated byte-range consisting of all A Hole on the server may be an allocated byte-range consisting of all
zeroes or may not be allocated at all. To ensure this information is zeroes or may not be allocated at all. To ensure this information is
properly communicated to the client, it may be beneficial to add a properly communicated to the client, it may be beneficial to add a
'alloc' flag to the HOLE_INFO section of nfs_readplusreshole. This 'alloc' flag to the HOLE_INFO section of nfs_readplusreshole. This
would allow an NFS client to copy a file from one file system to would allow an NFS client to copy a file from one file system to
another and have it more closely resemble the original. another and have it more closely resemble the original.
7.7.5. Dense and Sparse pNFS File Layouts 3.6.5. Dense and Sparse pNFS File Layouts
The hole information returned form a data server must be understood The hole information returned form a data server must be understood
by pNFS clients using both Dense or Sparse file layout types. Does by pNFS clients using both Dense or Sparse file layout types. Does
the current READ_PLUS return value work for both layout types? Does the current READ_PLUS return value work for both layout types? Does
the data server know if it is using dense or sparse so that it can the data server know if it is using dense or sparse so that it can
return the correct hole_offset and hole_length values? return the correct hole_offset and hole_length values?
8. Labeled NFS 4. Space Reservation
8.1. Introduction 4.1. Introduction
This section describes a set of operations that allow applications
such as hypervisors to reserve space for a file, report the amount of
actual disk space a file occupies and freeup the backing space of a
file when it is not required. In virtualized environments, virtual
disk files are often stored on NFS mounted volumes. Since virtual
disk files represent the hard disks of virtual machines, hypervisors
often have to guarantee certain properties for the file.
One such example is space reservation. When a hypervisor creates a
virtual disk file, it often tries to preallocate the space for the
file so that there are no future allocation related errors during the
operation of the virtual machine. Such errors prevent a virtual
machine from continuing execution and result in downtime.
Currently, in order to achieve such a guarantee, applications zero
the entire file. The initial zeroing allocates the backing blocks
and all subsequent writes are overwrites of already allocated blocks.
This approach is not only inefficient in terms of the amount of I/O
done, it is also not guaranteed to work on filesystems that are log
structured or deduplicated. An efficient way of guaranteeing space
reservation would be beneficial to such applications.
If the space_reserved attribute is set on a file, it is guaranteed
that writes that do not grow the file will not fail with
NFSERR_NOSPC.
Another useful feature would be the ability to report the number of
blocks that would be freed when a file is deleted. Currently, NFS
reports two size attributes:
size The logical file size of the file.
space_used The size in bytes that the file occupies on disk
While these attributes are sufficient for space accounting in
traditional filesystems, they prove to be inadequate in modern
filesystems that support block sharing. In such filesystems,
multiple inodes can point to a single block with a block reference
count to guard against premature freeing. Having a way to tell the
number of blocks that would be freed if the file was deleted would be
useful to applications that wish to migrate files when a volume is
low on space.
Since virtual disks represent a hard drive in a virtual machine, a
virtual disk can be viewed as a filesystem within a file. Since not
all blocks within a filesystem are in use, there is an opportunity to
reclaim blocks that are no longer in use. A call to deallocate
blocks could result in better space efficiency. Lesser space MAY be
consumed for backups after block deallocation.
The following operations and attributes can be used to resolve this
issues:
space_reserved This attribute specifies whether the blocks backing
the file have been preallocated.
space_freed This attribute specifies the space freed when a file is
deleted, taking block sharing into consideration.
max_hole_punch This attribute specifies the maximum sized hole that
can be punched on the filesystem.
INITIALIZED This operation zeroes and/or deallocates the blocks
backing a region of the file.
If space_used of a file is interpreted to mean the size in bytes of
all disk blocks pointed to by the inode of the file, then shared
blocks get double counted, over-reporting the space utilization.
This also has the adverse effect that the deletion of a file with
shared blocks frees up less than space_used bytes.
On the other hand, if space_used is interpreted to mean the size in
bytes of those disk blocks unique to the inode of the file, then
shared blocks are not counted in any file, resulting in under-
reporting of the space utilization.
For example, two files A and B have 10 blocks each. Let 6 of these
blocks be shared between them. Thus, the combined space utilized by
the two files is 14 * BLOCK_SIZE bytes. In the former case, the
combined space utilization of the two files would be reported as 20 *
BLOCK_SIZE. However, deleting either would only result in 4 *
BLOCK_SIZE being freed. Conversely, the latter interpretation would
report that the space utilization is only 8 * BLOCK_SIZE.
Adding another size attribute, space_freed, is helpful in solving
this problem. space_freed is the number of blocks that are allocated
to the given file that would be freed on its deletion. In the
example, both A and B would report space_freed as 4 * BLOCK_SIZE and
space_used as 10 * BLOCK_SIZE. If A is deleted, B will report
space_freed as 10 * BLOCK_SIZE as the deletion of B would result in
the deallocation of all 10 blocks.
The addition of this problem doesn't solve the problem of space being
over-reported. However, over-reporting is better than under-
reporting.
4.2. Operations and attributes
In the sections that follow, one operation and three attributes are
defined that together provide the space management facilities
outlined earlier in the document. The operation is intended to be
OPTIONAL and the attributes RECOMMENDED as defined in section 17 of
[2].
4.3. Attribute 77: space_reserved
The space_reserve attribute is a read/write attribute of type
boolean. It is a per file attribute. When the space_reserved
attribute is set via SETATTR, the server must ensure that there is
disk space to accommodate every byte in the file before it can return
success. If the server cannot guarantee this, it must return
NFS4ERR_NOSPC.
If the client tries to grow a file which has the space_reserved
attribute set, the server must guarantee that there is disk space to
accommodate every byte in the file with the new size before it can
return success. If the server cannot guarantee this, it must return
NFS4ERR_NOSPC.
It is not required that the server allocate the space to the file
before returning success. The allocation can be deferred, however,
it must be guaranteed that it will not fail for lack of space.
The value of space_reserved can be obtained at any time through
GETATTR.
In order to avoid ambiguity, the space_reserve bit cannot be set
along with the size bit in SETATTR. Increasing the size of a file
with space_reserve set will fail if space reservation cannot be
guaranteed for the new size. If the file size is decreased, space
reservation is only guaranteed for the new size and the extra blocks
backing the file can be released.
4.4. Attribute 78: space_freed
space_freed gives the number of bytes freed if the file is deleted.
This attribute is read only and is of type length4. It is a per file
attribute.
4.5. Attribute 79: max_hole_punch
max_hole_punch specifies the maximum size of a hole that the
INITIALIZE operation can handle. This attribute is read only and of
type length4. It is a per filesystem attribute. This attribute MUST
be implemented if INITIALIZE is implemented. [[Comment.4:
max_hole_punch when doing ADB initialization? --TH]]
5. Application Data Block Support
At the OS level, files are contained on disk blocks. Applications
are also free to impose structure on the data contained in a file and
we can define an Application Data Block (ADB) to be such a structure.
From the application's viewpoint, it only wants to handle ADBs and
not raw bytes (see [16]). An ADB is typically comprised of two
sections: a header and data. The header describes the
characteristics of the block and can provide a means to detect
corruption in the data payload. The data section is typically
initialized to all zeros.
The format of the header is application specific, but there are two
main components typically encountered:
1. An ADB Number (ADBN), which allows the application to determine
which data block is being referenced. The ADBN is a logical
block number and is useful when the client is not storing the
blocks in contiguous memory.
2. Fields to describe the state of the ADB and a means to detect
block corruption. For both pieces of data, a useful property is
that allowed values be unique in that if passed across the
network, corruption due to translation between big and little
endian architectures are detectable. For example, 0xF0DEDEF0 has
the same bit pattern in both architectures.
Applications already impose structures on files [16] and detect
corruption in data blocks [17]. What they are not able to do is
efficiently transfer and store ADBs. To initialize a file with ADBs,
the client must send the full ADB to the server and that must be
stored on the server. When the application is initializing a file to
have the ADB structure, it could compress the ADBs to just the
information to necessary to later reconstruct the header portion of
the ADB when the contents are read back. Using sparse file
techniques, the disk blocks described by would not be allocated.
Unlike sparse file techniques, there would be a small cost to store
the compressed header data.
In this section, we are going to define a generic framework for an
ADB, present one approach to detecting corruption in a given ADB
implementation, and describe the model for how the client and server
can support efficient initialization of ADBs, reading of ADB holes,
punching holes in ADBs, and space reservation. Further, we need to
be able to extend this model to applications which do not support
ADBs, but wish to be able to handle sparse files, hole punching, and
space reservation.
5.1. Generic Framework
We want the representation of the ADB to be flexible enough to
support many different applications. The most basic approach is no
imposition of a block at all, which means we are working with the raw
bytes. Such an approach would be useful for storing holes, punching
holes, etc. In more complex deployments, a server might be
supporting multiple applications, each with their own definition of
the ADB. One might store the ADBN at the start of the block and then
have a guard pattern to detect corruption [18]. The next might store
the ADBN at an offset of 100 bytes within the block and have no guard
pattern at all. The point is that existing applications might
already have well defined formats for their data blocks.
The guard pattern can be used to represent the state of the block, to
protect against corruption, or both. Again, it needs to be able to
be placed anywhere within the ADB.
We need to be able to represent the starting offset of the block and
the size of the block. Note that nothing prevents the application
from defining different sized blocks in a file.
5.1.1. Data Block Representation
struct app_data_block4 {
offset4 adb_offset;
length4 adb_block_size;
length4 adb_block_count;
length4 adb_reloff_blocknum;
count4 adb_block_num;
length4 adb_reloff_pattern;
opaque adb_pattern<>;
};
The app_data_block4 structure captures the abstraction presented for
the ADB. The additional fields present are to allow the transmission
of adb_block_count ADBs at one time. We also use adb_block_num to
convey the ADBN of the first block in the sequence. Each ADB will
contain the same adb_pattern string.
As both adb_block_num and adb_pattern are optional, if either
adb_reloff_pattern or adb_reloff_blocknum is set to NFS4_UINT64_MAX,
then the corresponding field is not set in any of the ADB.
5.1.2. Data Content
/*
* Use an enum such that we can extend new types.
*/
enum data_content4 {
NFS4_CONTENT_DATA = 0,
NFS4_CONTENT_APP_BLOCK = 1,
NFS4_CONTENT_HOLE = 2
};
New operations might need to differentiate between wanting to access
data versus an ADB. Also, future minor versions might want to
introduce new data formats. This enumeration allows that to occur.
5.2. pNFS Considerations
While this document does not mandate how sparse ADBs are recorded on
the server, it does make the assumption that such information is not
in the file. I.e., the information is metadata. As such, the
INITIALIZE operation is defined to be not supported by the DS - it
must be issued to the MDS. But since the client must not assume a
priori whether a read is sparse or not, the READ_PLUS operation MUST
be supported by both the DS and the MDS. I.e., the client might
impose on the MDS to asynchronously read the data from the DS.
Furthermore, each DS MUST not report to a client either a sparse ADB
or data which belongs to another DS. One implication of this
requirement is that the app_data_block4's adb_block_size MUST be
either be the stripe width or the stripe width must be an even
multiple of it.
The second implication here is that the DS must be able to use the
Control Protocol to determine from the MDS where the sparse ADBs
occur. [[Comment.5: Need to discuss what happens if after the file
is being written to and an INITIALIZE occurs? --TH]] Perhaps instead
of the DS pulling from the MDS, the MDS pushes to the DS? Thus an
INITIALIZE causes a new push? [[Comment.6: Still need to consider
race cases of the DS getting a WRITE and the MDS getting an
INITIALIZE. --TH]]
5.3. An Example of Detecting Corruption
In this section, we define an ADB format in which corruption can be
detected. Note that this is just one possible format and means to
detect corruption.
Consider a very basic implementation of an operating system's disk
blocks. A block is either data or it is an indirect block which
allows for files to be larger than one block. It is desired to be
able to initialize a block. Lastly, to quickly unlink a file, a
block can be marked invalid. The contents remain intact - which
would enable this OS application to undelete a file.
The application defines 4k sized data blocks, with an 8 byte block
counter occurring at offset 0 in the block, and with the guard
pattern occurring at offset 8 inside the block. Furthermore, the
guard pattern can take one of four states:
0xfeedface - This is the FREE state and indicates that the ADB
format has been applied.
0xcafedead - This is the DATA state and indicates that real data
has been written to this block.
0xe4e5c001 - This is the INDIRECT state and indicates that the
block contains block counter numbers that are chained off of this
block.
0xba1ed4a3 - This is the INVALID state and indicates that the block
contains data whose contents are garbage.
Finally, it also defines an 8 byte checksum [19] starting at byte 16
which applies to the remaining contents of the block. If the state
is FREE, then that checksum is trivially zero. As such, the
application has no need to transfer the checksum implicitly inside
the ADB - it need not make the transfer layer aware of the fact that
there is a checksum (see [17] for an example of checksums used to
detect corruption in application data blocks).
Corruption in each ADB can be detected thusly:
o If the guard pattern is anything other than one of the allowed
values, including all zeros.
o If the guard pattern is FREE and any other byte in the remainder
of the ADB is anything other than zero.
o If the guard pattern is anything other than FREE, then if the
stored checksum does not match the computed checksum.
o If the guard pattern is INDIRECT and one of the stored indirect
block numbers has a value greater than the number of ADBs in the
file.
o If the guard pattern is INDIRECT and one of the stored indirect
block numbers is a duplicate of another stored indirect block
number.
As can be seen, the application can detect errors based on the
combination of the guard pattern state and the checksum. But also,
the application can detect corruption based on the state and the
contents of the ADB. This last point is important in validating the
minimum amount of data we incorporated into our generic framework.
I.e., the guard pattern is sufficient in allowing applications to
design their own corruption detection.
Finally, it is important to note that none of these corruption checks
occur in the transport layer. The server and client components are
totally unaware of the file format and might report everything as
being transferred correctly even in the case the application detects
corruption.
5.4. Example of READ_PLUS
The hypothetical application presented in Section 5.3 can be used to
illustrate how READ_PLUS would return an array of results. A file is
created and initialized with 100 4k ADBs in the FREE state:
INITIALIZE {0, 4k, 100, 0, 0, 8, 0xfeedface}
Further, assume the application writes a single ADB at 16k, changing
the guard pattern to 0xcafedead, we would then have in memory:
0 -> (16k - 1) : 4k, 4, 0, 0, 8, 0xfeedface
16k -> (20k - 1) : 00 00 00 05 ca fe de ad XX XX ... XX XX
20k -> 400k : 4k, 95, 0, 6, 0xfeedface
And when the client did a READ_PLUS of 64k at the start of the file,
it would get back a result of an ADB, some data, and a final ADB:
ADB {0, 4, 0, 0, 8, 0xfeedface}
data 4k
ADB {20k, 4k, 59, 0, 6, 0xfeedface}
5.5. Zero Filled Holes
As applications are free to define the structure of an ADB, it is
trivial to define an ADB which supports zero filled holes. Such a
case would encompass the traditional definitions of a sparse file and
hole punching. For example, to punch a 64k hole, starting at 100M,
into an existing file which has no ADB structure:
INITIALIZE {100M, 64k, 1, NFS4_UINT64_MAX,
0, NFS4_UINT64_MAX, 0x0}
6. Labeled NFS
6.1. Introduction
Access control models such as Unix permissions or Access Control Access control models such as Unix permissions or Access Control
Lists are commonly referred to as Discretionary Access Control (DAC) Lists are commonly referred to as Discretionary Access Control (DAC)
models. These systems base their access decisions on user identity models. These systems base their access decisions on user identity
and resource ownership. In contrast Mandatory Access Control (MAC) and resource ownership. In contrast Mandatory Access Control (MAC)
models base their access control decisions on the label on the models base their access control decisions on the label on the
subject (usually a process) and the object it wishes to access. subject (usually a process) and the object it wishes to access.
These labels may contain user identity information but usually These labels may contain user identity information but usually
contain additional information. In DAC systems users are free to contain additional information. In DAC systems users are free to
specify the access rules for resources that they own. MAC models specify the access rules for resources that they own. MAC models
skipping to change at page 50, line 51 skipping to change at page 42, line 19
The second change is to provide a method for the server to notify the The second change is to provide a method for the server to notify the
client that the attribute changed on an open file on the server. If client that the attribute changed on an open file on the server. If
the file is closed, then during the open attempt, the client will the file is closed, then during the open attempt, the client will
gather the new attribute value. The server MUST not communicate the gather the new attribute value. The server MUST not communicate the
new value of the attribute, the client MUST query it. This new value of the attribute, the client MUST query it. This
requirement stems from the need for the client to provide sufficient requirement stems from the need for the client to provide sufficient
access rights to the attribute. access rights to the attribute.
The final change necessary is a modification to the RPC layer used in The final change necessary is a modification to the RPC layer used in
NFSv4 in the form of a new version of the RPCSEC_GSS [7] framework. NFSv4 in the form of a new version of the RPCSEC_GSS [6] framework.
In order for an NFSv4 server to apply MAC checks it must obtain In order for an NFSv4 server to apply MAC checks it must obtain
additional information from the client. Several methods were additional information from the client. Several methods were
explored for performing this and it was decided that the best explored for performing this and it was decided that the best
approach was to incorporate the ability to make security attribute approach was to incorporate the ability to make security attribute
assertions through the RPC mechanism. RPCSECGSSv3 [6] outlines a assertions through the RPC mechanism. RPCSECGSSv3 [5] outlines a
method to assert additional security information such as security method to assert additional security information such as security
labels on gss context creation and have that data bound to all RPC labels on gss context creation and have that data bound to all RPC
requests that make use of that context. requests that make use of that context.
8.2. Definitions 6.2. Definitions
Label Format Specifier (LFS): is an identifier used by the client to Label Format Specifier (LFS): is an identifier used by the client to
establish the syntactic format of the security label and the establish the syntactic format of the security label and the
semantic meaning of its components. These specifiers exist in a semantic meaning of its components. These specifiers exist in a
registry associated with documents describing the format and registry associated with documents describing the format and
semantics of the label. semantics of the label.
Label Format Registry: is the IANA registry containing all Label Format Registry: is the IANA registry containing all
registered LFS along with references to the documents that registered LFS along with references to the documents that
describe the syntactic format and semantics of the security label. describe the syntactic format and semantics of the security label.
skipping to change at page 51, line 45 skipping to change at page 43, line 15
Object: is a passive resource within the system that we wish to be Object: is a passive resource within the system that we wish to be
protected. Objects can be entities such as files, directories, protected. Objects can be entities such as files, directories,
pipes, sockets, and many other system resources relevant to the pipes, sockets, and many other system resources relevant to the
protection of the system state. protection of the system state.
Subject: A subject is an active entity usually a process which is Subject: A subject is an active entity usually a process which is
requesting access to an object. requesting access to an object.
Multi-Level Security (MLS): is a traditional model where objects are Multi-Level Security (MLS): is a traditional model where objects are
given a sensitivity level (Unclassified, Secret, Top Secret, etc) given a sensitivity level (Unclassified, Secret, Top Secret, etc)
and a category set [21]. and a category set [20].
8.3. MAC Security Attribute 6.3. MAC Security Attribute
MAC models base access decisions on security attributes bound to MAC models base access decisions on security attributes bound to
subjects and objects. This information can range from a user subjects and objects. This information can range from a user
identity for an identity based MAC model, sensitivity levels for identity for an identity based MAC model, sensitivity levels for
Multi-level security, or a type for Type Enforcement. These models Multi-level security, or a type for Type Enforcement. These models
base their decisions on different criteria but the semantics of the base their decisions on different criteria but the semantics of the
security attribute remain the same. The semantics required by the security attribute remain the same. The semantics required by the
security attributes are listed below: security attributes are listed below:
o Must provide flexibility with respect to MAC model. o Must provide flexibility with respect to MAC model.
skipping to change at page 52, line 23 skipping to change at page 43, line 42
o Must provide the ability to enforce access control decisions both o Must provide the ability to enforce access control decisions both
on the client and the server on the client and the server
o Must not expose an object to either the client or server name o Must not expose an object to either the client or server name
space before its security information has been bound to it. space before its security information has been bound to it.
NFSv4 implements the security attribute as a recommended attribute. NFSv4 implements the security attribute as a recommended attribute.
These attributes have a fixed format and semantics, which conflicts These attributes have a fixed format and semantics, which conflicts
with the flexible nature of the security attribute. To resolve this with the flexible nature of the security attribute. To resolve this
the security attribute consists of two components. The first the security attribute consists of two components. The first
component is a LFS as defined in [22] to allow for interoperability component is a LFS as defined in [21] to allow for interoperability
between MAC mechanisms. The second component is an opaque field between MAC mechanisms. The second component is an opaque field
which is the actual security attribute data. To allow for various which is the actual security attribute data. To allow for various
MAC models NFSv4 should be used solely as a transport mechanism for MAC models NFSv4 should be used solely as a transport mechanism for
the security attribute. It is the responsibility of the endpoints to the security attribute. It is the responsibility of the endpoints to
consume the security attribute and make access decisions based on consume the security attribute and make access decisions based on
their respective models. In addition, creation of objects through their respective models. In addition, creation of objects through
OPEN and CREATE allows for the security attribute to be specified OPEN and CREATE allows for the security attribute to be specified
upon creation. By providing an atomic create and set operation for upon creation. By providing an atomic create and set operation for
the security attribute it is possible to enforce the second and the security attribute it is possible to enforce the second and
fourth requirements. The recommended attribute FATTR4_SEC_LABEL will fourth requirements. The recommended attribute FATTR4_SEC_LABEL will
be used to satisfy this requirement. be used to satisfy this requirement.
8.3.1. Interpreting FATTR4_SEC_LABEL 6.3.1. Interpreting FATTR4_SEC_LABEL
The XDR [11] necessary to implement Labeled NFSv4 is presented below: The XDR [22] necessary to implement Labeled NFSv4 is presented below:
const FATTR4_SEC_LABEL = 81; const FATTR4_SEC_LABEL = 81;
typedef uint32_t policy4; typedef uint32_t policy4;
Figure 6 Figure 6
struct labelformat_spec4 { struct labelformat_spec4 {
policy4 lfs_lfs; policy4 lfs_lfs;
policy4 lfs_pi; policy4 lfs_pi;
skipping to change at page 53, line 21 skipping to change at page 44, line 32
labelformat_spec4 slai_lfs; labelformat_spec4 slai_lfs;
opaque slai_data<>; opaque slai_data<>;
}; };
The FATTR4_SEC_LABEL contains an array of two components with the The FATTR4_SEC_LABEL contains an array of two components with the
first component being an LFS. It serves to provide the receiving end first component being an LFS. It serves to provide the receiving end
with the information necessary to translate the security attribute with the information necessary to translate the security attribute
into a form that is usable by the endpoint. Label Formats assigned into a form that is usable by the endpoint. Label Formats assigned
an LFS may optionally choose to include a Policy Identifier field to an LFS may optionally choose to include a Policy Identifier field to
allow for complex policy deployments. The LFS and Label Format allow for complex policy deployments. The LFS and Label Format
Registry are described in detail in [22]. The translation used to Registry are described in detail in [21]. The translation used to
interpret the security attribute is not specified as part of the interpret the security attribute is not specified as part of the
protocol as it may depend on various factors. The second component protocol as it may depend on various factors. The second component
is an opaque section which contains the data of the attribute. This is an opaque section which contains the data of the attribute. This
component is dependent on the MAC model to interpret and enforce. component is dependent on the MAC model to interpret and enforce.
In particular, it is the responsibility of the LFS specification to In particular, it is the responsibility of the LFS specification to
define a maximum size for the opaque section, slai_data<>. When define a maximum size for the opaque section, slai_data<>. When
creating or modifying a label for an object, the client needs to be creating or modifying a label for an object, the client needs to be
guaranteed that the server will accept a label that is sized guaranteed that the server will accept a label that is sized
correctly. By both client and server being part of a specific MAC correctly. By both client and server being part of a specific MAC
model, the client will be aware of the size. model, the client will be aware of the size.
8.3.2. Delegations 6.3.2. Delegations
In the event that a security attribute is changed on the server while In the event that a security attribute is changed on the server while
a client holds a delegation on the file, the client should follow the a client holds a delegation on the file, the client should follow the
existing protocol with respect to attribute changes. It should flush existing protocol with respect to attribute changes. It should flush
all changes back to the server and relinquish the delegation. all changes back to the server and relinquish the delegation.
8.3.3. Permission Checking 6.3.3. Permission Checking
It is not feasible to enumerate all possible MAC models and even It is not feasible to enumerate all possible MAC models and even
levels of protection within a subset of these models. This means levels of protection within a subset of these models. This means
that the NFSv4 client and servers cannot be expected to directly make that the NFSv4 client and servers cannot be expected to directly make
access control decisions based on the security attribute. Instead access control decisions based on the security attribute. Instead
NFSv4 should defer permission checking on this attribute to the host NFSv4 should defer permission checking on this attribute to the host
system. These checks are performed in addition to existing DAC and system. These checks are performed in addition to existing DAC and
ACL checks outlined in the NFSv4 protocol. Section 8.6 gives a ACL checks outlined in the NFSv4 protocol. Section 6.6 gives a
specific example of how the security attribute is handled under a specific example of how the security attribute is handled under a
particular MAC model. particular MAC model.
8.3.4. Object Creation 6.3.4. Object Creation
When creating files in NFSv4 the OPEN and CREATE operations are used. When creating files in NFSv4 the OPEN and CREATE operations are used.
One of the parameters to these operations is an fattr4 structure One of the parameters to these operations is an fattr4 structure
containing the attributes the file is to be created with. This containing the attributes the file is to be created with. This
allows NFSv4 to atomically set the security attribute of files upon allows NFSv4 to atomically set the security attribute of files upon
creation. When a client is MAC aware it must always provide the creation. When a client is MAC aware it must always provide the
initial security attribute upon file creation. In the event that the initial security attribute upon file creation. In the event that the
server is the only MAC aware entity in the system it should ignore server is the only MAC aware entity in the system it should ignore
the security attribute specified by the client and instead make the the security attribute specified by the client and instead make the
determination itself. A more in depth explanation can be found in determination itself. A more in depth explanation can be found in
Section 8.6. Section 6.6.
8.3.5. Existing Objects 6.3.5. Existing Objects
Note that under the MAC model, all objects must have labels. Note that under the MAC model, all objects must have labels.
Therefore, if an existing server is upgraded to include LNFS support, Therefore, if an existing server is upgraded to include LNFS support,
then it is the responsibility of the security system to define the then it is the responsibility of the security system to define the
behavior for existing objects. For example, if the security system behavior for existing objects. For example, if the security system
is LFS 0, which means the server just stores and returns labels, then is LFS 0, which means the server just stores and returns labels, then
existing files should return labels which are set to an empty value. existing files should return labels which are set to an empty value.
8.3.6. Label Changes 6.3.6. Label Changes
As per the requirements, when a file's security label is modified, As per the requirements, when a file's security label is modified,
the server must notify all clients which have the file opened of the the server must notify all clients which have the file opened of the
change in label. It does so with CB_ATTR_CHANGED. There are change in label. It does so with CB_ATTR_CHANGED. There are
preconditions to making an attribute change imposed by NFSv4 and the preconditions to making an attribute change imposed by NFSv4 and the
security system might want to impose others. In the process of security system might want to impose others. In the process of
meeting these preconditions, the server may chose to either serve the meeting these preconditions, the server may chose to either serve the
request in whole or return NFS4ERR_DELAY to the SETATTR operation. request in whole or return NFS4ERR_DELAY to the SETATTR operation.
If there are open delegations on the file belonging to client other If there are open delegations on the file belonging to client other
than the one making the label change, then the process described in than the one making the label change, then the process described in
Section 8.3.2 must be followed. Section 6.3.2 must be followed.
As the server is always presented with the subject label from the As the server is always presented with the subject label from the
client, it does not necessarily need to communicate the fact that the client, it does not necessarily need to communicate the fact that the
label has changed to the client. In the cases where the change label has changed to the client. In the cases where the change
outright denies the client access, the client will be able to quickly outright denies the client access, the client will be able to quickly
determine that there is a new label in effect. It is in cases where determine that there is a new label in effect. It is in cases where
the client may share the same object between multiple subjects or a the client may share the same object between multiple subjects or a
security system which is not strictly hierarchical that the security system which is not strictly hierarchical that the
CB_ATTR_CHANGED callback is very useful. It allows the server to CB_ATTR_CHANGED callback is very useful. It allows the server to
inform the clients that the cached security attribute is now stale. inform the clients that the cached security attribute is now stale.
skipping to change at page 55, line 13 skipping to change at page 46, line 26
server has a very simple security system which just stores the server has a very simple security system which just stores the
labels. In this system, the MAC label check always allows access, labels. In this system, the MAC label check always allows access,
regardless of the subject label. regardless of the subject label.
The way in which MAC labels are enforced is by the smart client. So The way in which MAC labels are enforced is by the smart client. So
if client A changes a security label on a file, then the server MUST if client A changes a security label on a file, then the server MUST
inform all clients that have the file opened that the label has inform all clients that have the file opened that the label has
changed via CB_ATTR_CHANGED. Then the clients MUST retrieve the new changed via CB_ATTR_CHANGED. Then the clients MUST retrieve the new
label and MUST enforce access via the new attribute values. label and MUST enforce access via the new attribute values.
[[Comment.6: Describe a LFS of 0, which will be the means to indicate [[Comment.7: Describe a LFS of 0, which will be the means to indicate
such a deployment. In the current LFR, 0 is marked as reserved. If such a deployment. In the current LFR, 0 is marked as reserved. If
we use it, then we define the default LFS to be used by a LNFS aware we use it, then we define the default LFS to be used by a LNFS aware
server. I.e., it lets smart clients work together in the face of a server. I.e., it lets smart clients work together in the face of a
dumb server. Note that will supporting this system is optional, it dumb server. Note that will supporting this system is optional, it
will make for a very good debugging mode during development. I.e., will make for a very good debugging mode during development. I.e.,
even if a server does not deploy with another security system, this even if a server does not deploy with another security system, this
mode gets your foot in the door. --TH]] mode gets your foot in the door. --TH]]
8.4. pNFS Considerations 6.4. pNFS Considerations
This section examines the issues in deploying LNFS in a pNFS This section examines the issues in deploying LNFS in a pNFS
community of servers. community of servers.
8.4.1. MAC Label Checks 6.4.1. MAC Label Checks
The new FATTR4_SEC_LABEL attribute is metadata information and as The new FATTR4_SEC_LABEL attribute is metadata information and as
such the DS is not aware of the value contained on the MDS. such the DS is not aware of the value contained on the MDS.
Fortunately, the NFSv4.1 protocol [2] already has provisions for Fortunately, the NFSv4.1 protocol [2] already has provisions for
doing access level checks from the DS to the MDS. In order for the doing access level checks from the DS to the MDS. In order for the
DS to validate the subject label presented by the client, it SHOULD DS to validate the subject label presented by the client, it SHOULD
utilize this mechanism. utilize this mechanism.
If a file's FATTR4_SEC_LABEL is changed, then the MDS should utilize If a file's FATTR4_SEC_LABEL is changed, then the MDS should utilize
CB_ATTR_CHANGED to inform the client of that fact. If the MDS is CB_ATTR_CHANGED to inform the client of that fact. If the MDS is
maintaining maintaining
8.5. Discovery of Server LNFS Support 6.5. Discovery of Server LNFS Support
The server can easily determine that a client supports LNFS when it The server can easily determine that a client supports LNFS when it
queries for the FATTR4_SEC_LABEL label for an object. Note that it queries for the FATTR4_SEC_LABEL label for an object. Note that it
cannot assume that the presence of RPCSEC_GSSv3 indicates LNFS cannot assume that the presence of RPCSEC_GSSv3 indicates LNFS
support. The client might need to discover which LFS the server support. The client might need to discover which LFS the server
supports. supports.
A server which supports LNFS MUST allow a client with any subject A server which supports LNFS MUST allow a client with any subject
label to retrieve the FATTR4_SEC_LABEL attribute for the root label to retrieve the FATTR4_SEC_LABEL attribute for the root
filehandle, ROOTFH. The following compound must always succeed as filehandle, ROOTFH. The following compound must always succeed as
skipping to change at page 56, line 15 skipping to change at page 47, line 28
PUTROOTFH, GETATTR {FATTR4_SEC_LABEL} PUTROOTFH, GETATTR {FATTR4_SEC_LABEL}
Note that the server might have imposed a security flavor on the root Note that the server might have imposed a security flavor on the root
that precludes such access. I.e., if the server requires kerberized that precludes such access. I.e., if the server requires kerberized
access and the client presents a compound with AUTH_SYS, then the access and the client presents a compound with AUTH_SYS, then the
server is allowed to return NFS4ERR_WRONGSEC in this case. But if server is allowed to return NFS4ERR_WRONGSEC in this case. But if
the client presents a correct security flavor, then the server MUST the client presents a correct security flavor, then the server MUST
return the FATTR4_SEC_LABEL attribute with the supported LFS filled return the FATTR4_SEC_LABEL attribute with the supported LFS filled
in. in.
8.6. MAC Security NFS Modes of Operation 6.6. MAC Security NFS Modes of Operation
A system using Labeled NFS may operate in three modes. The first A system using Labeled NFS may operate in three modes. The first
mode provides the most protection and is called "full mode". In this mode provides the most protection and is called "full mode". In this
mode both the client and server implement a MAC model allowing each mode both the client and server implement a MAC model allowing each
end to make an access control decision. The remaining two modes are end to make an access control decision. The remaining two modes are
variations on each other and are called "smart client" and "smart variations on each other and are called "smart client" and "smart
server" modes. In these modes one end of the connection is not server" modes. In these modes one end of the connection is not
implementing a MAC model and because of this these operating modes implementing a MAC model and because of this these operating modes
offer less protection than full mode. offer less protection than full mode.
8.6.1. Full Mode 6.6.1. Full Mode
Full mode environments consist of MAC aware NFSv4 servers and clients Full mode environments consist of MAC aware NFSv4 servers and clients
and may be composed of mixed MAC models and policies. The system and may be composed of mixed MAC models and policies. The system
requires that both the client and server have an opportunity to requires that both the client and server have an opportunity to
perform an access control check based on all relevant information perform an access control check based on all relevant information
within the network. The file object security attribute is provided within the network. The file object security attribute is provided
using the mechanism described in Section 8.3. The security attribute using the mechanism described in Section 6.3. The security attribute
of the subject making the request is transported at the RPC layer of the subject making the request is transported at the RPC layer
using the mechanism described in RPCSECGSSv3 [6]. using the mechanism described in RPCSECGSSv3 [5].
8.6.1.1. Initial Labeling and Translation 6.6.1.1. Initial Labeling and Translation
The ability to create a file is an action that a MAC model may wish The ability to create a file is an action that a MAC model may wish
to mediate. The client is given the responsibility to determine the to mediate. The client is given the responsibility to determine the
initial security attribute to be placed on a file. This allows the initial security attribute to be placed on a file. This allows the
client to make a decision as to the acceptable security attributes to client to make a decision as to the acceptable security attributes to
create a file with before sending the request to the server. Once create a file with before sending the request to the server. Once
the server receives the creation request from the client it may the server receives the creation request from the client it may
choose to evaluate if the security attribute is acceptable. choose to evaluate if the security attribute is acceptable.
Security attributes on the client and server may vary based on MAC Security attributes on the client and server may vary based on MAC
skipping to change at page 57, line 11 skipping to change at page 48, line 28
identify the format and meaning of the opaque portion of the security identify the format and meaning of the opaque portion of the security
attribute. A full mode environment may contain hosts operating in attribute. A full mode environment may contain hosts operating in
several different LFSs and DOIs. In this case a mechanism for several different LFSs and DOIs. In this case a mechanism for
translating the opaque portion of the security attribute is needed. translating the opaque portion of the security attribute is needed.
The actual translation function will vary based on MAC model and The actual translation function will vary based on MAC model and
policy and is out of the scope of this document. If a translation is policy and is out of the scope of this document. If a translation is
unavailable for a given LFS and DOI then the request SHOULD be unavailable for a given LFS and DOI then the request SHOULD be
denied. Another recourse is to allow the host to provide a fallback denied. Another recourse is to allow the host to provide a fallback
mapping for unknown security attributes. mapping for unknown security attributes.
8.6.1.2. Policy Enforcement 6.6.1.2. Policy Enforcement
In full mode access control decisions are made by both the clients In full mode access control decisions are made by both the clients
and servers. When a client makes a request it takes the security and servers. When a client makes a request it takes the security
attribute from the requesting process and makes an access control attribute from the requesting process and makes an access control
decision based on that attribute and the security attribute of the decision based on that attribute and the security attribute of the
object it is trying to access. If the client denies that access an object it is trying to access. If the client denies that access an
RPC call to the server is never made. If however the access is RPC call to the server is never made. If however the access is
allowed the client will make a call to the NFS server. allowed the client will make a call to the NFS server.
When the server receives the request from the client it extracts the When the server receives the request from the client it extracts the
skipping to change at page 57, line 34 skipping to change at page 49, line 5
trying to access to make an access control decision. If the server's trying to access to make an access control decision. If the server's
policy allows this access it will fulfill the client's request, policy allows this access it will fulfill the client's request,
otherwise it will return NFS4ERR_ACCESS. otherwise it will return NFS4ERR_ACCESS.
Implementations MAY validate security attributes supplied over the Implementations MAY validate security attributes supplied over the
network to ensure that they are within a set of attributes permitted network to ensure that they are within a set of attributes permitted
from a specific peer, and if not, reject them. Note that a system from a specific peer, and if not, reject them. Note that a system
may permit a different set of attributes to be accepted from each may permit a different set of attributes to be accepted from each
peer. peer.
8.6.2. Smart Client Mode 6.6.2. Smart Client Mode
Smart client environments consist of NFSv4 servers that are not MAC Smart client environments consist of NFSv4 servers that are not MAC
aware but NFSv4 clients that are. Clients in this environment are aware but NFSv4 clients that are. Clients in this environment are
may consist of groups implementing different MAC models policies. may consist of groups implementing different MAC models policies.
The system requires that all clients in the environment be The system requires that all clients in the environment be
responsible for access control checks. Due to the amount of trust responsible for access control checks. Due to the amount of trust
placed in the clients this mode is only to be used in a trusted placed in the clients this mode is only to be used in a trusted
environment. environment.
8.6.2.1. Initial Labeling and Translation 6.6.2.1. Initial Labeling and Translation
Just like in full mode the client is responsible for determining the Just like in full mode the client is responsible for determining the
initial label upon object creation. The server in smart client mode initial label upon object creation. The server in smart client mode
does not implement a MAC model, however, it may provide the ability does not implement a MAC model, however, it may provide the ability
to restrict the creation and labeling of object with certain labels to restrict the creation and labeling of object with certain labels
based on different criteria as described in Section 8.6.1.2. based on different criteria as described in Section 6.6.1.2.
In a smart client environment a group of clients operate in a single In a smart client environment a group of clients operate in a single
DOI. This removes the need for the clients to maintain a set of DOI DOI. This removes the need for the clients to maintain a set of DOI
translations. Servers should provide a method to allow different translations. Servers should provide a method to allow different
groups of clients to access the server at the same time. However it groups of clients to access the server at the same time. However it
should not let two groups of clients operating in different DOIs to should not let two groups of clients operating in different DOIs to
access the same files. access the same files.
8.6.2.2. Policy Enforcement 6.6.2.2. Policy Enforcement
In smart client mode access control decisions are made by the In smart client mode access control decisions are made by the
clients. When a client accesses an object it obtains the security clients. When a client accesses an object it obtains the security
attribute of the object from the server and combines it with the attribute of the object from the server and combines it with the
security attribute of the process making the request to make an security attribute of the process making the request to make an
access control decision. This check is in addition to the DAC checks access control decision. This check is in addition to the DAC checks
provided by NFSv4 so this may fail based on the DAC criteria even if provided by NFSv4 so this may fail based on the DAC criteria even if
the MAC policy grants access. As the policy check is located on the the MAC policy grants access. As the policy check is located on the
client an access control denial should take the form that is native client an access control denial should take the form that is native
to the platform. to the platform.
8.6.3. Smart Server Mode 6.6.3. Smart Server Mode
Smart server environments consist of NFSv4 servers that are MAC aware Smart server environments consist of NFSv4 servers that are MAC aware
and one or more MAC unaware clients. The server is the only entity and one or more MAC unaware clients. The server is the only entity
enforcing policy, and may selectively provide standard NFS services enforcing policy, and may selectively provide standard NFS services
to clients based on their authentication credentials and/or to clients based on their authentication credentials and/or
associated network attributes (e.g., IP address, network interface). associated network attributes (e.g., IP address, network interface).
The level of trust and access extended to a client in this mode is The level of trust and access extended to a client in this mode is
configuration-specific. configuration-specific.
8.6.3.1. Initial Labeling and Translation 6.6.3.1. Initial Labeling and Translation
In smart server mode all labeling and access control decisions are In smart server mode all labeling and access control decisions are
performed by the NFSv4 server. In this environment the NFSv4 clients performed by the NFSv4 server. In this environment the NFSv4 clients
are not MAC aware so they cannot provide input into the access are not MAC aware so they cannot provide input into the access
control decision. This requires the server to determine the initial control decision. This requires the server to determine the initial
labeling of objects. Normally the subject to use in this calculation labeling of objects. Normally the subject to use in this calculation
would originate from the client. Instead the NFSv4 server may choose would originate from the client. Instead the NFSv4 server may choose
to assign the subject security attribute based on their to assign the subject security attribute based on their
authentication credentials and/or associated network attributes authentication credentials and/or associated network attributes
(e.g., IP address, network interface). (e.g., IP address, network interface).
In smart server mode security attributes are contained solely within In smart server mode security attributes are contained solely within
the NFSv4 server. This means that all security attributes used in the NFSv4 server. This means that all security attributes used in
the system remain within a single LFS and DOI. Since security the system remain within a single LFS and DOI. Since security
attributes will not cross DOIs or change format there is no need to attributes will not cross DOIs or change format there is no need to
provide any translation functionality above that which is needed provide any translation functionality above that which is needed
internally by the MAC model. internally by the MAC model.
8.6.3.2. Policy Enforcement 6.6.3.2. Policy Enforcement
All access control decisions in smart server mode are made by the All access control decisions in smart server mode are made by the
server. The server will assign the subject a security attribute server. The server will assign the subject a security attribute
based on some criteria (e.g., IP address, network interface). Using based on some criteria (e.g., IP address, network interface). Using
the newly calculated security attribute and the security attribute of the newly calculated security attribute and the security attribute of
the object being requested the MAC model makes the access control the object being requested the MAC model makes the access control
check and returns NFS4ERR_ACCESS on a denial and NFS4_OK on success. check and returns NFS4ERR_ACCESS on a denial and NFS4_OK on success.
This check is done transparently to the client so if the MAC This check is done transparently to the client so if the MAC
permission check fails the client may be unaware of the reason for permission check fails the client may be unaware of the reason for
the permission failure. When operating in this mode administrators the permission failure. When operating in this mode administrators
attempting to debug permission failures should be aware to check the attempting to debug permission failures should be aware to check the
MAC policy running on the server in addition to the DAC settings. MAC policy running on the server in addition to the DAC settings.
8.7. Security Considerations 6.7. Security Considerations
This entire document deals with security issues. This entire document deals with security issues.
Depending on the level of protection the MAC system offers there may Depending on the level of protection the MAC system offers there may
be a requirement to tightly bind the security attribute to the data. be a requirement to tightly bind the security attribute to the data.
When only one of the client or server enforces labels, it is When only one of the client or server enforces labels, it is
important to realize that the other side is not enforcing MAC important to realize that the other side is not enforcing MAC
protections. Alternate methods might be in use to handle the lack of protections. Alternate methods might be in use to handle the lack of
MAC support and care should be taken to identify and mitigate threats MAC support and care should be taken to identify and mitigate threats
from possible tampering outside of these methods. from possible tampering outside of these methods.
An example of this is that a server that modifies READDIR or LOOKUP An example of this is that a server that modifies READDIR or LOOKUP
results based on the client's subject label might want to always results based on the client's subject label might want to always
construct the same subject label for a client which does not present construct the same subject label for a client which does not present
one. This will prevent a non-LNFS client from mixing entries in the one. This will prevent a non-LNFS client from mixing entries in the
directory cache. directory cache.
9. Security Considerations 7. Sharing change attribute implementation details with NFSv4 clients
10. Operations: REQUIRED, RECOMMENDED, or OPTIONAL 7.1. Introduction
Although both the NFSv4 [10] and NFSv4.1 protocol [2], define the
change attribute as being mandatory to implement, there is little in
the way of guidance. The only feature that is mandated by them is
that the value must change whenever the file data or metadata change.
While this allows for a wide range of implementations, it also leaves
the client with a conundrum: how does it determine which is the most
recent value for the change attribute in a case where several RPC
calls have been issued in parallel? In other words if two COMPOUNDs,
both containing WRITE and GETATTR requests for the same file, have
been issued in parallel, how does the client determine which of the
two change attribute values returned in the replies to the GETATTR
requests corresponds to the most recent state of the file? In some
cases, the only recourse may be to send another COMPOUND containing a
third GETATTR that is fully serialised with the first two.
NFSv4.2 avoids this kind of inefficiency by allowing the server to
share details about how the change attribute is expected to evolve,
so that the client may immediately determine which, out of the
several change attribute values returned by the server, is the most
recent.
7.2. Definition of the 'change_attr_type' per-file system attribute
enum change_attr_typeinfo {
NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR = 0,
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER = 1,
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS = 2,
NFS4_CHANGE_TYPE_IS_TIME_METADATA = 3,
NFS4_CHANGE_TYPE_IS_UNDEFINED = 4
};
+------------------+----+---------------------------+-----+
| Name | Id | Data Type | Acc |
+------------------+----+---------------------------+-----+
| change_attr_type | XX | enum change_attr_typeinfo | R |
+------------------+----+---------------------------+-----+
The solution enables the NFS server to provide additional information
about how it expects the change attribute value to evolve after the
file data or metadata has changed. 'change_attr_type' is defined as a
new recommended attribute, and takes values from enum
change_attr_typeinfo as follows:
NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR: The change attribute value MUST
monotonically increase for every atomic change to the file
attributes, data or directory contents.
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER: The change attribute value MUST
be incremented by one unit for every atomic change to the file
attributes, data or directory contents. This property is
preserved when writing to pNFS data servers.
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS: The change attribute
value MUST be incremented by one unit for every atomic change to
the file attributes, data or directory contents. In the case
where the client is writing to pNFS data servers, the number of
increments is not guaranteed to exactly match the number of
writes.
NFS4_CHANGE_TYPE_IS_TIME_METADATA: The change attribute is
implemented as suggested in the NFSv4 spec [10] in terms of the
time_metadata attribute.
NFS4_CHANGE_TYPE_IS_UNDEFINED: The change attribute does not take
values that fit into any of these categories.
If either NFS4_CHANGE_TYPE_IS_MONOTONIC_INCR,
NFS4_CHANGE_TYPE_IS_VERSION_COUNTER, or
NFS4_CHANGE_TYPE_IS_TIME_METADATA are set, then the client knows at
the very least that the change attribute is monotonically increasing,
which is sufficient to resolve the question of which value is the
most recent.
If the client sees the value NFS4_CHANGE_TYPE_IS_TIME_METADATA, then
by inspecting the value of the 'time_delta' attribute it additionally
has the option of detecting rogue server implementations that use
time_metadata in violation of the spec.
Finally, if the client sees NFS4_CHANGE_TYPE_IS_VERSION_COUNTER, it
has the ability to predict what the resulting change attribute value
should be after a COMPOUND containing a SETATTR, WRITE, or CREATE.
This again allows it to detect changes made in parallel by another
client. The value NFS4_CHANGE_TYPE_IS_VERSION_COUNTER_NOPNFS permits
the same, but only if the client is not doing pNFS WRITEs.
8. Security Considerations
9. Operations: REQUIRED, RECOMMENDED, or OPTIONAL
The following tables summarize the operations of the NFSv4.2 protocol The following tables summarize the operations of the NFSv4.2 protocol
and the corresponding designation of REQUIRED, RECOMMENDED, and and the corresponding designation of REQUIRED, RECOMMENDED, and
OPTIONAL to implement or MUST NOT implement. The designation of MUST OPTIONAL to implement or MUST NOT implement. The designation of MUST
NOT implement is reserved for those operations that were defined in NOT implement is reserved for those operations that were defined in
either NFSv4.0 or NFSV4.1 and MUST NOT be implemented in NFSv4.2. either NFSv4.0 or NFSV4.1 and MUST NOT be implemented in NFSv4.2.
For the most part, the REQUIRED, RECOMMENDED, or OPTIONAL designation For the most part, the REQUIRED, RECOMMENDED, or OPTIONAL designation
for operations sent by the client is for the server implementation. for operations sent by the client is for the server implementation.
The client is generally required to implement the operations needed The client is generally required to implement the operations needed
skipping to change at page 63, line 29 skipping to change at page 56, line 29
| CB_RECALL_ANY | OPT | FDELG, DDELG, pNFS | | CB_RECALL_ANY | OPT | FDELG, DDELG, pNFS |
| | | (REQ) | | | | (REQ) |
| CB_RECALL_SLOT | REQ | | | CB_RECALL_SLOT | REQ | |
| CB_RECALLABLE_OBJ_AVAIL | OPT | DDELG, pNFS (REQ) | | CB_RECALLABLE_OBJ_AVAIL | OPT | DDELG, pNFS (REQ) |
| CB_SEQUENCE | OPT | FDELG, DDELG, pNFS | | CB_SEQUENCE | OPT | FDELG, DDELG, pNFS |
| | | (REQ) | | | | (REQ) |
| CB_WANTS_CANCELLED | OPT | FDELG, DDELG, pNFS | | CB_WANTS_CANCELLED | OPT | FDELG, DDELG, pNFS |
| | | (REQ) | | | | (REQ) |
+-------------------------+-------------------+---------------------+ +-------------------------+-------------------+---------------------+
11. NFSv4.2 Operations 10. NFSv4.2 Operations
11.1. Operation 59: COPY - Initiate a server-side copy 10.1. Operation 59: COPY - Initiate a server-side copy
11.1.1. ARGUMENT 10.1.1. ARGUMENT
const COPY4_GUARDED = 0x00000001; const COPY4_GUARDED = 0x00000001;
const COPY4_METADATA = 0x00000002; const COPY4_METADATA = 0x00000002;
struct COPY4args { struct COPY4args {
/* SAVED_FH: source file */ /* SAVED_FH: source file */
/* CURRENT_FH: destination file or */ /* CURRENT_FH: destination file or */
/* directory */ /* directory */
offset4 ca_src_offset; offset4 ca_src_offset;
offset4 ca_dst_offset; offset4 ca_dst_offset;
length4 ca_count; length4 ca_count;
uint32_t ca_flags; uint32_t ca_flags;
component4 ca_destination; component4 ca_destination;
netloc4 ca_source_server<>; netloc4 ca_source_server<>;
}; };
11.1.2. RESULT 10.1.2. RESULT
union COPY4res switch (nfsstat4 cr_status) { union COPY4res switch (nfsstat4 cr_status) {
case NFS4_OK: case NFS4_OK:
stateid4 cr_callback_id<1>; stateid4 cr_callback_id<1>;
default: default:
length4 cr_bytes_copied; length4 cr_bytes_copied;
}; };
11.1.3. DESCRIPTION 10.1.3. DESCRIPTION
The COPY operation is used for both intra-server and inter-server The COPY operation is used for both intra-server and inter-server
copies. In both cases, the COPY is always sent from the client to copies. In both cases, the COPY is always sent from the client to
the destination server of the file copy. The COPY operation requests the destination server of the file copy. The COPY operation requests
that a file be copied from the location specified by the SAVED_FH that a file be copied from the location specified by the SAVED_FH
value to the location specified by the combination of CURRENT_FH and value to the location specified by the combination of CURRENT_FH and
ca_destination. ca_destination.
The SAVED_FH must be a regular file. If SAVED_FH is not a regular The SAVED_FH must be a regular file. If SAVED_FH is not a regular
file, the operation MUST fail and return NFS4ERR_WRONG_TYPE. file, the operation MUST fail and return NFS4ERR_WRONG_TYPE.
skipping to change at page 66, line 15 skipping to change at page 59, line 15
If the ca_source_server list is specified, then this is an inter- If the ca_source_server list is specified, then this is an inter-
server copy operation and the source file is on a remote server. The server copy operation and the source file is on a remote server. The
client is expected to have previously issued a successful COPY_NOTIFY client is expected to have previously issued a successful COPY_NOTIFY
request to the remote source server. The ca_source_server list request to the remote source server. The ca_source_server list
SHOULD be the same as the COPY_NOTIFY response's cnr_source_server SHOULD be the same as the COPY_NOTIFY response's cnr_source_server
list. If the client includes the entries from the COPY_NOTIFY list. If the client includes the entries from the COPY_NOTIFY
response's cnr_source_server list in the ca_source_server list, the response's cnr_source_server list in the ca_source_server list, the
source server can indicate a specific copy protocol for the source server can indicate a specific copy protocol for the
destination server to use by returning a URL, which specifies both a destination server to use by returning a URL, which specifies both a
protocol service and server name. Server-to-server copy protocol protocol service and server name. Server-to-server copy protocol
considerations are described in Section 4.2.3 and Section 4.4.1. considerations are described in Section 2.2.3 and Section 2.4.1.
The ca_flags argument allows the copy operation to be customized in The ca_flags argument allows the copy operation to be customized in
the following ways using the guarded flag (COPY4_GUARDED) and the the following ways using the guarded flag (COPY4_GUARDED) and the
metadata flag (COPY4_METADATA). metadata flag (COPY4_METADATA).
If the guarded flag is set and the destination exists on the server, If the guarded flag is set and the destination exists on the server,
this operation will fail with NFS4ERR_EXIST. this operation will fail with NFS4ERR_EXIST.
If the guarded flag is not set and the destination exists on the If the guarded flag is not set and the destination exists on the
server, the behavior is implementation dependent. server, the behavior is implementation dependent.
skipping to change at page 71, line 23 skipping to change at page 64, line 23
NFS4ERR_DELAY: The server does not have the resources to perform the NFS4ERR_DELAY: The server does not have the resources to perform the
copy operation at the current time. The client should retry the copy operation at the current time. The client should retry the
operation sometime in the future. operation sometime in the future.
NFS4ERR_METADATA_NOTSUPP: The destination file cannot support the NFS4ERR_METADATA_NOTSUPP: The destination file cannot support the
same metadata as the source file. same metadata as the source file.
NFS4ERR_WRONGSEC: The security mechanism being used by the client NFS4ERR_WRONGSEC: The security mechanism being used by the client
does not match the server's security policy. does not match the server's security policy.
11.2. Operation 60: COPY_ABORT - Cancel a server-side copy 10.2. Operation 60: COPY_ABORT - Cancel a server-side copy
11.2.1. ARGUMENT 10.2.1. ARGUMENT
struct COPY_ABORT4args { struct COPY_ABORT4args {
/* CURRENT_FH: desination file */ /* CURRENT_FH: desination file */
stateid4 caa_stateid; stateid4 caa_stateid;
}; };
11.2.2. RESULT 10.2.2. RESULT
struct COPY_ABORT4res { struct COPY_ABORT4res {
nfsstat4 car_status; nfsstat4 car_status;
}; };
11.2.3. DESCRIPTION 10.2.3. DESCRIPTION
COPY_ABORT is used for both intra- and inter-server asynchronous COPY_ABORT is used for both intra- and inter-server asynchronous
copies. The COPY_ABORT operation allows the client to cancel a copies. The COPY_ABORT operation allows the client to cancel a
server-side copy operation that it initiated. This operation is sent server-side copy operation that it initiated. This operation is sent
in a COMPOUND request from the client to the destination server. in a COMPOUND request from the client to the destination server.
This operation may be used to cancel a copy when the application that This operation may be used to cancel a copy when the application that
requested the copy exits before the operation is completed or for requested the copy exits before the operation is completed or for
some other reason. some other reason.
The request contains the filehandle and copy stateid cookies that act The request contains the filehandle and copy stateid cookies that act
skipping to change at page 72, line 27 skipping to change at page 65, line 27
NFS4ERR_RETRY: The abort failed, but a retry at some time in the NFS4ERR_RETRY: The abort failed, but a retry at some time in the
future MAY succeed. future MAY succeed.
NFS4ERR_COMPLETE_ALREADY: The abort failed, and a callback will NFS4ERR_COMPLETE_ALREADY: The abort failed, and a callback will
deliver the results of the copy operation. deliver the results of the copy operation.
NFS4ERR_SERVERFAULT: An error occurred on the server that does not NFS4ERR_SERVERFAULT: An error occurred on the server that does not
map to a specific error code. map to a specific error code.
11.3. Operation 61: COPY_NOTIFY - Notify a source server of a future 10.3. Operation 61: COPY_NOTIFY - Notify a source server of a future
copy copy
11.3.1. ARGUMENT 10.3.1. ARGUMENT
struct COPY_NOTIFY4args { struct COPY_NOTIFY4args {
/* CURRENT_FH: source file */ /* CURRENT_FH: source file */
netloc4 cna_destination_server; netloc4 cna_destination_server;
}; };
11.3.2. RESULT 10.3.2. RESULT
struct COPY_NOTIFY4resok { struct COPY_NOTIFY4resok {
nfstime4 cnr_lease_time; nfstime4 cnr_lease_time;
netloc4 cnr_source_server<>; netloc4 cnr_source_server<>;
}; };
union COPY_NOTIFY4res switch (nfsstat4 cnr_status) { union COPY_NOTIFY4res switch (nfsstat4 cnr_status) {
case NFS4_OK: case NFS4_OK:
COPY_NOTIFY4resok resok4; COPY_NOTIFY4resok resok4;
default: default:
void; void;
}; };
11.3.3. DESCRIPTION 10.3.3. DESCRIPTION
This operation is used for an inter-server copy. A client sends this This operation is used for an inter-server copy. A client sends this
operation in a COMPOUND request to the source server to authorize a operation in a COMPOUND request to the source server to authorize a
destination server identified by cna_destination_server to read the destination server identified by cna_destination_server to read the
file specified by CURRENT_FH on behalf of the given user. file specified by CURRENT_FH on behalf of the given user.
The cna_destination_server MUST be specified using the netloc4 The cna_destination_server MUST be specified using the netloc4
network location format. The server is not required to resolve the network location format. The server is not required to resolve the
cna_destination_server address before completing this operation. cna_destination_server address before completing this operation.
skipping to change at page 75, line 11 skipping to change at page 68, line 11
present on the source server. The client can determine the present on the source server. The client can determine the
correct location and reissue the operation with the correct correct location and reissue the operation with the correct
location. location.
NFS4ERR_NOTSUPP: The copy offload operation is not supported by the NFS4ERR_NOTSUPP: The copy offload operation is not supported by the
NFS server receiving this request. NFS server receiving this request.
NFS4ERR_WRONGSEC: The security mechanism being used by the client NFS4ERR_WRONGSEC: The security mechanism being used by the client
does not match the server's security policy. does not match the server's security policy.
11.4. Operation 62: COPY_REVOKE - Revoke a destination server's copy 10.4. Operation 62: COPY_REVOKE - Revoke a destination server's copy
privileges privileges
11.4.1. ARGUMENT 10.4.1. ARGUMENT
struct COPY_REVOKE4args { struct COPY_REVOKE4args {
/* CURRENT_FH: source file */ /* CURRENT_FH: source file */
netloc4 cra_destination_server; netloc4 cra_destination_server;
}; };
11.4.2. RESULT 10.4.2. RESULT
struct COPY_REVOKE4res { struct COPY_REVOKE4res {
nfsstat4 crr_status; nfsstat4 crr_status;
}; };
11.4.3. DESCRIPTION 10.4.3. DESCRIPTION
This operation is used for an inter-server copy. A client sends this This operation is used for an inter-server copy. A client sends this
operation in a COMPOUND request to the source server to revoke the operation in a COMPOUND request to the source server to revoke the
authorization of a destination server identified by authorization of a destination server identified by
cra_destination_server from reading the file specified by CURRENT_FH cra_destination_server from reading the file specified by CURRENT_FH
on behalf of given user. If the cra_destination_server has already on behalf of given user. If the cra_destination_server has already
begun copying the file, a successful return from this operation begun copying the file, a successful return from this operation
indicates that further access will be prevented. indicates that further access will be prevented.
The cra_destination_server MUST be specified using the netloc4 The cra_destination_server MUST be specified using the netloc4
skipping to change at page 76, line 16 skipping to change at page 69, line 16
a partial list): a partial list):
NFS4ERR_MOVED: The file system which contains the source file is not NFS4ERR_MOVED: The file system which contains the source file is not
present on the source server. The client can determine the present on the source server. The client can determine the
correct location and reissue the operation with the correct correct location and reissue the operation with the correct
location. location.
NFS4ERR_NOTSUPP: The copy offload operation is not supported by the NFS4ERR_NOTSUPP: The copy offload operation is not supported by the
NFS server receiving this request. NFS server receiving this request.
11.5. Operation 63: COPY_STATUS - Poll for status of a server-side copy 10.5. Operation 63: COPY_STATUS - Poll for status of a server-side copy
11.5.1. ARGUMENT 10.5.1. ARGUMENT
struct COPY_STATUS4args { struct COPY_STATUS4args {
/* CURRENT_FH: destination file */ /* CURRENT_FH: destination file */
stateid4 csa_stateid; stateid4 csa_stateid;
}; };
11.5.2. RESULT 10.5.2. RESULT
struct COPY_STATUS4resok { struct COPY_STATUS4resok {
length4 csr_bytes_copied; length4 csr_bytes_copied;
nfsstat4 csr_complete<1>; nfsstat4 csr_complete<1>;
}; };
union COPY_STATUS4res switch (nfsstat4 csr_status) { union COPY_STATUS4res switch (nfsstat4 csr_status) {
case NFS4_OK: case NFS4_OK:
COPY_STATUS4resok resok4; COPY_STATUS4resok resok4;
default: default:
void; void;
}; };
11.5.3. DESCRIPTION 10.5.3. DESCRIPTION
COPY_STATUS is used for both intra- and inter-server asynchronous COPY_STATUS is used for both intra- and inter-server asynchronous
copies. The COPY_STATUS operation allows the client to poll the copies. The COPY_STATUS operation allows the client to poll the
server to determine the status of an asynchronous copy operation. server to determine the status of an asynchronous copy operation.
This operation is sent by the client to the destination server. This operation is sent by the client to the destination server.
If this operation is successful, the number of bytes copied are If this operation is successful, the number of bytes copied are
returned to the client in the csr_bytes_copied field. The returned to the client in the csr_bytes_copied field. The
csr_bytes_copied value indicates the number of bytes copied but not csr_bytes_copied value indicates the number of bytes copied but not
which specific bytes have been copied. which specific bytes have been copied.
skipping to change at page 77, line 23 skipping to change at page 70, line 23
If the server supports asynchronous copies, the server is REQUIRED to If the server supports asynchronous copies, the server is REQUIRED to
support the COPY_STATUS operation. support the COPY_STATUS operation.
The COPY_STATUS operation may fail for the following reasons (this is The COPY_STATUS operation may fail for the following reasons (this is
a partial list): a partial list):
NFS4ERR_NOTSUPP: The copy status operation is not supported by the NFS4ERR_NOTSUPP: The copy status operation is not supported by the
NFS server receiving this request. NFS server receiving this request.
NFS4ERR_BAD_STATEID: The stateid is not valid (see Section 4.3.2 NFS4ERR_BAD_STATEID: The stateid is not valid (see Section 2.3.2
below). below).
NFS4ERR_EXPIRED: The stateid has expired (see Copy Offload Stateid NFS4ERR_EXPIRED: The stateid has expired (see Copy Offload Stateid
section below). section below).
11.6. Operation 64: INITIALIZE 10.6. Modification to Operation 42: EXCHANGE_ID - Instantiate Client ID
10.6.1. ARGUMENT
/* new */
const EXCHGID4_FLAG_SUPP_FENCE_OPS = 0x00000004;
10.6.2. RESULT
Unchanged
10.6.3. MOTIVATION
Enterprise applications require guarantees that an operation has
either aborted or completed. NFSv4.1 provides this guarantee as long
as the session is alive: simply send a SEQUENCE operation on the same
slot with a new sequence number, and the successful return of
SEQUENCE indicates the previous operation has completed. However, if
the session is lost, there is no way to know when any in progress
operations have aborted or completed. In hindsight, the NFSv4.1
specification should have mandated that DESTROY_SESSION abort/
complete all outstanding operations.
10.6.4. DESCRIPTION
A client SHOULD request the EXCHGID4_FLAG_SUPP_FENCE_OPS capability
when it sends an EXCHANGE_ID operation. The server SHOULD set this
capability in the EXCHANGE_ID reply whether the client requests it or
not. If the client ID is created with this capability then the
following will occur:
o The server will not reply to DESTROY_SESSION until all operations
in progress are completed or aborted.
o The server will not reply to subsequent EXCHANGE_ID invoked on the
same Client Owner with a new verifier until all operations in
progress on the Client ID's session are completed or aborted.
o When DESTROY_CLIENTID is invoked, if there are sessions (both idle
and non-idle), opens, locks, delegations, layouts, and/or wants
(Section 18.49) associated with the client ID are removed.
Pending operations will be completed or aborted before the
sessions, opens, locks, delegations, layouts, and/or wants are
deleted.
o The NFS server SHOULD support client ID trunking, and if it does
and the EXCHGID4_FLAG_SUPP_FENCE_OPS capability is enabled, then a
session ID created on one node of the storage cluster MUST be
destroyable via DESTROY_SESSION. In addition, DESTROY_CLIENTID
and an EXCHANGE_ID with a new verifier affects all sessions
regardless what node the sessions were created on.
10.7. Operation 64: INITIALIZE
This operation can be used to initialize the structure imposed by an
application onto a file and to punch a hole into a file.
The server has no concept of the structure imposed by the The server has no concept of the structure imposed by the
application. It is only when the application writes to a section of application. It is only when the application writes to a section of
the file does order get imposed. In order to detect corruption even the file does order get imposed. In order to detect corruption even
before the application utilizes the file, the application will want before the application utilizes the file, the application will want
to initialize a range of ADBs. It uses the INITIALIZE operation to to initialize a range of ADBs. It uses the INITIALIZE operation to
do so. do so.
11.6.1. ARGUMENT 10.7.1. ARGUMENT
/* /*
* We use data_content4 in case we wish to * We use data_content4 in case we wish to
* extend new types later. Note that we * extend new types later. Note that we
* are explicitly disallowing data. * are explicitly disallowing data.
*/ */
union initialize_arg4 switch (data_content4 content) { union initialize_arg4 switch (data_content4 content) {
case NFS4_CONTENT_APP_BLOCK: case NFS4_CONTENT_APP_BLOCK:
app_data_block4 ia_adb; app_data_block4 ia_adb;
case NFS4_CONTENT_HOLE: case NFS4_CONTENT_HOLE:
skipping to change at page 78, line 28 skipping to change at page 72, line 28
void; void;
}; };
struct INITIALIZE4args { struct INITIALIZE4args {
/* CURRENT_FH: file */ /* CURRENT_FH: file */
stateid4 ia_stateid; stateid4 ia_stateid;
stable_how4 ia_stable; stable_how4 ia_stable;
initialize_arg4 ia_data<>; initialize_arg4 ia_data<>;
}; };
11.6.2. RESULT 10.7.2. RESULT
struct INITIALIZE4resok { struct INITIALIZE4resok {
count4 ir_count; count4 ir_count;
stable_how4 ir_committed; stable_how4 ir_committed;
verifier4 ir_writeverf; verifier4 ir_writeverf;
data_content4 ir_sparse; data_content4 ir_sparse;
}; };
union INITIALIZE4res switch (nfsstat4 status) { union INITIALIZE4res switch (nfsstat4 status) {
case NFS4_OK: case NFS4_OK:
INITIALIZE4resok resok4; INITIALIZE4resok resok4;
default: default:
void; void;
}; };
11.6.3. DESCRIPTION 10.7.3. DESCRIPTION
When the client invokes the INITIALIZE operation, it has two desired When the client invokes the INITIALIZE operation, it has two desired
results: results:
1. The structure described by the app_data_block4 be imposed on the 1. The structure described by the app_data_block4 be imposed on the
file. file.
2. The contents described by the app_data_block4 be sparse. 2. The contents described by the app_data_block4 be sparse.
If the server supports the INITIALIZE operation, it still might not If the server supports the INITIALIZE operation, it still might not
support sparse files. So if it receives the INITIALIZE operation, support sparse files. So if it receives the INITIALIZE operation,
then it MUST populate the contents of the file with the initialized then it MUST populate the contents of the file with the initialized
ADBs. In other words, if the server supports INITIALIZE, then it ADBs. In other words, if the server supports INITIALIZE, then it
supports the concept of ADBs. [[Comment.7: Do we want to support an supports the concept of ADBs. [[Comment.8: Do we want to support an
asynchronous INITIALIZE? Do we have to? --TH]] asynchronous INITIALIZE? Do we have to? --TH]]
If the data was already initialized, There are two interesting If the data was already initialized, There are two interesting
scenarios: scenarios:
1. The data blocks are allocated. 1. The data blocks are allocated.
2. Initializing in the middle of an existing ADB. 2. Initializing in the middle of an existing ADB.
If the data blocks were already allocated, then the INITIALIZE is a If the data blocks were already allocated, then the INITIALIZE is a
hole punch operation. If INITIALIZE supports sparse files, then the hole punch operation. If INITIALIZE supports sparse files, then the
data blocks are to be deallocated. If not, then the data blocks are data blocks are to be deallocated. If not, then the data blocks are
to be rewritten in the indicated ADB format. [[Comment.8: Need to to be rewritten in the indicated ADB format. [[Comment.9: Need to
document interaction between space reservation and hole punching? document interaction between space reservation and hole punching?
--TH]] --TH]]
Since the server has no knowledge of ADBs, it should not report Since the server has no knowledge of ADBs, it should not report
misaligned creation of ADBs. Even while it can detect them, it misaligned creation of ADBs. Even while it can detect them, it
cannot disallow them, as the application might be in the process of cannot disallow them, as the application might be in the process of
changing the size of the ADBs. Thus the server must be prepared to changing the size of the ADBs. Thus the server must be prepared to
handle an INITIALIZE into an existing ADB. handle an INITIALIZE into an existing ADB.
This document does not mandate the manner in which the server stores This document does not mandate the manner in which the server stores
ADBs sparsely for a file. It does assume that if ADBs are stored ADBs sparsely for a file. It does assume that if ADBs are stored
sparsely, then the server can detect when an INITIALIZE arrives that sparsely, then the server can detect when an INITIALIZE arrives that
will force a new ADB to start inside an existing ADB. For example, will force a new ADB to start inside an existing ADB. For example,
assume that ADBi has a adb_block_size of 4k and that an INITIALIZE assume that ADBi has a adb_block_size of 4k and that an INITIALIZE
starts 1k inside ADBi. The server should [[Comment.9: Need to flesh starts 1k inside ADBi. The server should [[Comment.10: Need to flesh
this out. --TH]] this out. --TH]]
11.7. Modification to Operation 42: EXCHANGE_ID - Instantiate Client ID 10.7.3.1. Hole punching
11.7.1. ARGUMENT Whenever a client wishes to deallocate the blocks backing a
particular region in the file, it calls the INITIALIZE operation with
the current filehandle set to the filehandle of the file in question,
start offset and length in bytes of the region set in hpa_offset and
hpa_count respectively. All further reads to this region MUST return
zeros until overwritten. The filehandle specified must be that of a
regular file.
/* new */ Situations may arise where ia_hole.hi_offset and/or ia_hole.hi_offset
const EXCHGID4_FLAG_SUPP_FENCE_OPS = 0x00000004; + ia_hole.hi_length will not be aligned to a boundary that the server
does allocations/ deallocations in. For most filesystems, this is
the block size of the file system. In such a case, the server can
deallocate as many bytes as it can in the region. The blocks that
cannot be deallocated MUST be zeroed. Except for the block
deallocation and maximum hole punching capability, a INITIALIZE
operation is to be treated similar to a write of zeroes.
11.7.2. RESULT The server is not required to complete deallocating the blocks
specified in the operation before returning. It is acceptable to
have the deallocation be deferred. In fact, INITIALIZE is merely a
hint; it is valid for a server to return success without ever doing
anything towards deallocating the blocks backing the region
specified. However, any future reads to the region MUST return
zeroes.
Unchanged If used to hole punch, INITIALIZE will result in the space_used
attribute being decreased by the number of bytes that were
deallocated. The space_freed attribute may or may not decrease,
depending on the support and whether the blocks backing the specified
range were shared or not. The size attribute will remain unchanged.
11.7.3. MOTIVATION The INITIALIZE operation MUST NOT change the space reservation
guarantee of the file. While the server can deallocate the blocks
specified by hpa_offset and hpa_count, future writes to this region
MUST NOT fail with NFSERR_NOSPC.
Enterprise applications require guarantees that an operation has The INITIALIZE operation may fail for the following reasons (this is
either aborted or completed. NFSv4.1 provides this guarantee as long a partial list):
as the session is alive: simply send a SEQUENCE operation on the same
slot with a new sequence number, and the successful return of
SEQUENCE indicates the previous operation has completed. However, if
the session is lost, there is no way to know when any in progress
operations have aborted or completed. In hindsight, the NFSv4.1
specification should have mandated that DESTROY_SESSION abort/
complete all outstanding operations.
11.7.4. DESCRIPTION NFS4ERR_NOTSUPP The Hole punch operations are not supported by the
NFS server receiving this request.
A client SHOULD request the EXCHGID4_FLAG_SUPP_FENCE_OPS capability NFS4ERR_DIR The current filehandle is of type NF4DIR.
when it sends an EXCHANGE_ID operation. The server SHOULD set this
capability in the EXCHANGE_ID reply whether the client requests it or
not. If the client ID is created with this capability then the
following will occur:
o The server will not reply to DESTROY_SESSION until all operations NFS4ERR_SYMLINK The current filehandle is of type NF4LNK.
in progress are completed or aborted.
o The server will not reply to subsequent EXCHANGE_ID invoked on the NFS4ERR_WRONG_TYPE The current filehandle does not designate an
same Client Owner with a new verifier until all operations in ordinary file.
progress on the Client ID's session are completed or aborted.
o When DESTROY_CLIENTID is invoked, if there are sessions (both idle 10.8. Changes to Operation 51: LAYOUTRETURN
and non-idle), opens, locks, delegations, layouts, and/or wants 10.8.1. Introduction
(Section 18.49) associated with the client ID are removed.
Pending operations will be completed or aborted before the
sessions, opens, locks, delegations, layouts, and/or wants are
deleted.
o The NFS server SHOULD support client ID trunking, and if it does In the pNFS description provided in [2], the client is not enabled to
and the EXCHGID4_FLAG_SUPP_FENCE_OPS capability is enabled, then a relay an error code from the DS to the MDS. In the specification of
session ID created on one node of the storage cluster MUST be the Objects-Based Layout protocol [7], use is made of the opaque
destroyable via DESTROY_SESSION. In addition, DESTROY_CLIENTID lrf_body field of the LAYOUTRETURN argument to do such a relaying of
and an EXCHANGE_ID with a new verifier affects all sessions error codes. In this section, we define a new data structure to
regardless what node the sessions were created on. enable the passing of error codes back to the MDS and provide some
guidelines on what both the client and MDS should expect in such
circumstances.
11.8. Operation 65: READ_PLUS There are two broad classes of errors, transient and persistent. The
client SHOULD strive to only use this new mechanism to report
persistent errors. It MUST be able to deal with transient issues by
itself. Also, while the client might consider an issue to be
persistent, it MUST be prepared for the MDS to consider such issues
to be persistent. A prime example of this is if the MDS fences off a
client from either a stateid or a filehandle. The client will get an
error from the DS and might relay either NFS4ERR_ACCESS or
NFS4ERR_STALE_STATEID back to the MDS, with the belief that this is a
hard error. The MDS on the other hand, is waiting for the client to
report such an error. For it, the mission is accomplished in that
the client has returned a layout that the MDS had most likley
recalled.
The existing LAYOUTRETURN operation is extended by introducing a new
data structure to report errors, layoutreturn_device_error4. Also,
layoutreturn_device_error4 is introduced to enable an array of errors
to be reported.
10.8.2. ARGUMENT
The ARGUMENT specification of the LAYOUTRETURN operation in section
18.44.1 of [2] is augmented by the following XDR code [22]:
struct layoutreturn_device_error4 {
deviceid4 lrde_deviceid;
nfsstat4 lrde_status;
nfs_opnum4 lrde_opnum;
};
struct layoutreturn_error_report4 {
layoutreturn_device_error4 lrer_errors<>;
};
10.8.3. RESULT
The RESULT of the LAYOUTRETURN operation is unchanged; see section
18.44.2 of [2].
10.8.4. DESCRIPTION
The following text is added to the end of the LAYOUTRETURN operation
DESCRIPTION in section 18.44.3 of [2].
When a client used LAYOUTRETURN with a type of LAYOUTRETURN4_FILE,
then if the lrf_body field is NULL, it indicates to the MDS that the
client experienced no errors. If lrf_body is non-NULL, then the
field references error information which is layout type specific.
I.e., the Objects-Based Layout protocol can continue to utilize
lrf_body as specified in [7]. For both Files-Based Layouts, the
field references a layoutreturn_device_error4, which contains an
array of layoutreturn_device_error4.
Each individual layoutreturn_device_error4 descibes a single error
associated with a DS, which is identfied via lrde_deviceid. The
operation which returned the error is identified via lrde_opnum.
Finally the NFS error value (nfsstat4) encountered is provided via
lrde_status and may consist of the following error codes:
NFS4_OKAY: No issues were found for this device.
NFS4ERR_NXIO: The client was unable to establish any communication
with the DS.
NFS4ERR_*: The client was able to establish communication with the
DS and is returning one of the allowed error codes for the
operation denoted by lrde_opnum.
10.8.5. IMPLEMENTATION
The following text is added to the end of the LAYOUTRETURN operation
IMPLEMENTATION in section 18.4.4 of [2].
A client that expects to use pNFS for a mounted filesystem SHOULD
check for pNFS support at mount time. This check SHOULD be performed
by sending a GETDEVICELIST operation, followed by layout-type-
specific checks for accessibility of each storage device returned by
GETDEVICELIST. If the NFS server does not support pNFS, the
GETDEVICELIST operation will be rejected with an NFS4ERR_NOTSUPP
error; in this situation it is up to the client to determine whether
it is acceptable to proceed with NFS-only access.
Clients are expected to tolerate transient storage device errors, and
hence clients SHOULD NOT use the LAYOUTRETURN error handling for
device access problems that may be transient. The methods by which a
client decides whether an access problem is transient vs. persistent
are implementation-specific, but may include retrying I/Os to a data
server under appropriate conditions.
When an I/O fails to a storage device, the client SHOULD retry the
failed I/O via the MDS. In this situation, before retrying the I/O,
the client SHOULD return the layout, or the affected portion thereof,
and SHOULD indicate which storage device or devices was problematic.
If the client does not do this, the MDS may issue a layout recall
callback in order to perform the retried I/O.
The client needs to be cognizant that since this error handling is
optional in the MDS, the MDS may silently ignore this functionality.
Also, as the MDS may consider some issues the client reports to be
expected (see Section 10.8.1), the client might find it difficult to
detect a MDS which has not implemented error handling via
LAYOUTRETURN.
If an MDS is aware that a storage device is proving problematic to a
client, the MDS SHOULD NOT include that storage device in any pNFS
layouts sent to that client. If the MDS is aware that a storage
device is affecting many clients, then the MDS SHOULD NOT include
that storage device in any pNFS layouts sent out. Clients must still
be aware that the MDS might not have any choice in using the storage
device, i.e., there might only be one possible layout for the system.
Another interesting complication is that for existing files, the MDS
might have no choice in which storage devices to hand out to clients.
The MDS might try to restripe a file across a different storage
device, but clients need to be aware that not all implementations
have restriping support.
An MDS SHOULD react to a client return of layouts with errors by not
using the problematic storage devices in layouts for that client, but
the MDS is not required to indefinitely retain per-client storage
device error information. An MDS is also not required to
automatically reinstate use of a previously problematic storage
device; administrative intervention may be required instead.
A client MAY perform I/O via the MDS even when the client holds a
layout that covers the I/O; servers MUST support this client
behavior, and MAY recall layouts as needed to complete I/Os.
10.9. Operation 65: READ_PLUS
If the client sends a READ operation, it is explicitly stating that If the client sends a READ operation, it is explicitly stating that
it is not supporting sparse files. So if a READ occurs on a sparse it is not supporting sparse files. So if a READ occurs on a sparse
ADB, then the server must expand such ADBs to be raw bytes. If a ADB, then the server must expand such ADBs to be raw bytes. If a
READ occurs in the middle of an ADB, the server can only send back READ occurs in the middle of an ADB, the server can only send back
bytes starting from that offset. bytes starting from that offset.
Such an operation is inefficient for transfer of sparse sections of Such an operation is inefficient for transfer of sparse sections of
the file. As such, READ is marked as OBSOLETE in NFSv4.2. Instead, the file. As such, READ is marked as OBSOLETE in NFSv4.2. Instead,
a client should issue READ_PLUS. Note that as the client has no a a client should issue READ_PLUS. Note that as the client has no a
priori knowledge of whether an ADB is present or not, it should priori knowledge of whether an ADB is present or not, it should
always use READ_PLUS. always use READ_PLUS.
11.8.1. ARGUMENT 10.9.1. ARGUMENT
struct READ_PLUS4args { struct READ_PLUS4args {
/* CURRENT_FH: file */ /* CURRENT_FH: file */
stateid4 rpa_stateid; stateid4 rpa_stateid;
offset4 rpa_offset; offset4 rpa_offset;
count4 rpa_count; count4 rpa_count;
}; };
11.8.2. RESULT 10.9.2. RESULT
union read_plus_content switch (data_content4 content) { union read_plus_content switch (data_content4 content) {
case NFS4_CONTENT_DATA: case NFS4_CONTENT_DATA:
opaque rpc_data<>; opaque rpc_data<>;
case NFS4_CONTENT_APP_BLOCK: case NFS4_CONTENT_APP_BLOCK:
app_data_block4 rpc_block; app_data_block4 rpc_block;
case NFS4_CONTENT_HOLE: case NFS4_CONTENT_HOLE:
hole_info4 rpc_hole; hole_info4 rpc_hole;
default: default:
void; void;
skipping to change at page 82, line 33 skipping to change at page 79, line 33
read_plus_content rpr_contents<>; read_plus_content rpr_contents<>;
}; };
union READ_PLUS4res switch (nfsstat4 status) { union READ_PLUS4res switch (nfsstat4 status) {
case NFS4_OK: case NFS4_OK:
read_plus_res4 resok4; read_plus_res4 resok4;
default: default:
void; void;
}; };
11.8.3. DESCRIPTION 10.9.3. DESCRIPTION
Over the given range, READ_PLUS will return all data and ADBs found Over the given range, READ_PLUS will return all data and ADBs found
as an array of read_plus_content. It is possible to have consecutive as an array of read_plus_content. It is possible to have consecutive
ADBs in the array as either different definitions of ADBs are present ADBs in the array as either different definitions of ADBs are present
or as the guard pattern changes. or as the guard pattern changes.
Edge cases exist for ABDs which either begin before the rpa_offset Edge cases exist for ABDs which either begin before the rpa_offset
requested by the READ_PLUS or end after the rpa_count requested - requested by the READ_PLUS or end after the rpa_count requested -
both of which may occur as not all applications which access the file both of which may occur as not all applications which access the file
are aware of the main application imposing a format on the file are aware of the main application imposing a format on the file
contents, i.e., tar, dd, cp, etc. READ_PLUS MUST retrieve whole contents, i.e., tar, dd, cp, etc. READ_PLUS MUST retrieve whole
ADBs, but it need not retrieve an entire sequences of ADBs. ADBs, but it need not retrieve an entire sequences of ADBs.
The server MUST return a whole ADB because if it does not, it must The server MUST return a whole ADB because if it does not, it must
expand that partial ADB before it sends it to the client. E.g., if expand that partial ADB before it sends it to the client. E.g., if
an ADB had a block size of 64k and the READ_PLUS was for 128k an ADB had a block size of 64k and the READ_PLUS was for 128k
starting at an offset of 32k inside the ADB, then the first 32k would starting at an offset of 32k inside the ADB, then the first 32k would
be converted to data. be converted to data.
12. NFSv4.2 Callback Operations 11. NFSv4.2 Callback Operations
12.1. Procedure 16: CB_ATTR_CHANGED - Notify Client that the File's 11.1. Procedure 16: CB_ATTR_CHANGED - Notify Client that the File's
Attributes Changed Attributes Changed
12.1.1. ARGUMENTS 11.1.1. ARGUMENTS
struct CB_ATTR_CHANGED4args { struct CB_ATTR_CHANGED4args {
nfs_fh4 acca_fh; nfs_fh4 acca_fh;
bitmap4 acca_critical; bitmap4 acca_critical;
bitmap4 acca_info; bitmap4 acca_info;
}; };
12.1.2. RESULTS 11.1.2. RESULTS
struct CB_ATTR_CHANGED4res { struct CB_ATTR_CHANGED4res {
nfsstat4 accr_status; nfsstat4 accr_status;
}; };
12.1.3. DESCRIPTION 11.1.3. DESCRIPTION
The CB_ATTR_CHANGED callback operation is used by the server to The CB_ATTR_CHANGED callback operation is used by the server to
indicate to the client that the file's attributes have been modified indicate to the client that the file's attributes have been modified
on the server. The server does not convey how the attributes have on the server. The server does not convey how the attributes have
changed, just that they have been modified. The server can inform changed, just that they have been modified. The server can inform
the client about both critical and informational attribute changes in the client about both critical and informational attribute changes in
the bitmask arguments. The client SHOULD query the server about all the bitmask arguments. The client SHOULD query the server about all
attributes set in acca_critical. For all changes reflected in attributes set in acca_critical. For all changes reflected in
acca_info, the client can decide whether or not it wants to poll the acca_info, the client can decide whether or not it wants to poll the
server. server.
The CB_ATTR_CHANGED callback operation with the FATTR4_SEC_LABEL set The CB_ATTR_CHANGED callback operation with the FATTR4_SEC_LABEL set
in acca_critical is the method used by the server to indicate that in acca_critical is the method used by the server to indicate that
the MAC label for the file referenced by acca_fh has changed. In the MAC label for the file referenced by acca_fh has changed. In
many ways, the server does not care about the result returned by the many ways, the server does not care about the result returned by the
client. client.
12.2. Operation 15: CB_COPY - Report results of a server-side copy 11.2. Operation 15: CB_COPY - Report results of a server-side copy
12.2.1. ARGUMENT 11.2.1. ARGUMENT
union copy_info4 switch (nfsstat4 cca_status) { union copy_info4 switch (nfsstat4 cca_status) {
case NFS4_OK: case NFS4_OK:
void; void;
default: default:
length4 cca_bytes_copied; length4 cca_bytes_copied;
}; };
struct CB_COPY4args { struct CB_COPY4args {
nfs_fh4 cca_fh; nfs_fh4 cca_fh;
stateid4 cca_stateid; stateid4 cca_stateid;
copy_info4 cca_copy_info; copy_info4 cca_copy_info;
}; };
12.2.2. RESULT 11.2.2. RESULT
struct CB_COPY4res { struct CB_COPY4res {
nfsstat4 ccr_status; nfsstat4 ccr_status;
}; };
12.2.3. DESCRIPTION 11.2.3. DESCRIPTION
CB_COPY is used for both intra- and inter-server asynchronous copies. CB_COPY is used for both intra- and inter-server asynchronous copies.
The CB_COPY callback informs the client of the result of an The CB_COPY callback informs the client of the result of an
asynchronous server-side copy. This operation is sent by the asynchronous server-side copy. This operation is sent by the
destination server to the client in a CB_COMPOUND request. The copy destination server to the client in a CB_COMPOUND request. The copy
is identified by the filehandle and stateid arguments. The result is is identified by the filehandle and stateid arguments. The result is
indicated by the status field. If the copy failed, cca_bytes_copied indicated by the status field. If the copy failed, cca_bytes_copied
contains the number of bytes copied before the failure occurred. The contains the number of bytes copied before the failure occurred. The
cca_bytes_copied value indicates the number of bytes copied but not cca_bytes_copied value indicates the number of bytes copied but not
which specific bytes have been copied. which specific bytes have been copied.
skipping to change at page 85, line 8 skipping to change at page 82, line 8
If the client supports the COPY operation, the client is REQUIRED to If the client supports the COPY operation, the client is REQUIRED to
support the CB_COPY operation. support the CB_COPY operation.
The CB_COPY operation may fail for the following reasons (this is a The CB_COPY operation may fail for the following reasons (this is a
partial list): partial list):
NFS4ERR_NOTSUPP: The copy offload operation is not supported by the NFS4ERR_NOTSUPP: The copy offload operation is not supported by the
NFS client receiving this request. NFS client receiving this request.
13. IANA Considerations 12. IANA Considerations
This section uses terms that are defined in [23]. This section uses terms that are defined in [23].
14. References 13. References
14.1. Normative References 13.1. Normative References
[1] Bradner, S., "Key words for use in RFCs to Indicate Requirement [1] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", March 1997. Levels", March 1997.
[2] Shepler, S., Eisler, M., and D. Noveck, "Network File System [2] Shepler, S., Eisler, M., and D. Noveck, "Network File System
(NFS) Version 4 Minor Version 1 Protocol", RFC 5661, (NFS) Version 4 Minor Version 1 Protocol", RFC 5661,
January 2010. January 2010.
[3] Haynes, T., "Network File System (NFS) Version 4 Minor Version [3] Haynes, T., "Network File System (NFS) Version 4 Minor Version
2 External Data Representation Standard (XDR) Description", 2 External Data Representation Standard (XDR) Description",
March 2011. March 2011.
[4] Halevy, B., Welch, B., and J. Zelenka, "Object-Based Parallel [4] Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform
NFS (pNFS) Operations", RFC 5664, January 2010.
[5] Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform
Resource Identifier (URI): Generic Syntax", STD 66, RFC 3986, Resource Identifier (URI): Generic Syntax", STD 66, RFC 3986,
January 2005. January 2005.
[6] Haynes, T. and N. Williams, "Remote Procedure Call (RPC) [5] Haynes, T. and N. Williams, "Remote Procedure Call (RPC)
Security Version 3", draft-williams-rpcsecgssv3 (work in Security Version 3", draft-williams-rpcsecgssv3 (work in
progress), 2011. progress), 2011.
[7] Eisler, M., Chiu, A., and L. Ling, "RPCSEC_GSS Protocol [6] Eisler, M., Chiu, A., and L. Ling, "RPCSEC_GSS Protocol
Specification", RFC 2203, September 1997. Specification", RFC 2203, September 1997.
[7] Halevy, B., Welch, B., and J. Zelenka, "Object-Based Parallel
NFS (pNFS) Operations", RFC 5664, January 2010.
[8] Shepler, S., Eisler, M., and D. Noveck, "Network File System [8] Shepler, S., Eisler, M., and D. Noveck, "Network File System
(NFS) Version 4 Minor Version 1 External Data Representation (NFS) Version 4 Minor Version 1 External Data Representation
Standard (XDR) Description", RFC 5662, January 2010. Standard (XDR) Description", RFC 5662, January 2010.
[9] Black, D., Glasgow, J., and S. Fridella, "Parallel NFS (pNFS) [9] Black, D., Glasgow, J., and S. Fridella, "Parallel NFS (pNFS)
Block/Volume Layout", RFC 5663, January 2010. Block/Volume Layout", RFC 5663, January 2010.
14.2. Informative References 13.2. Informative References
[10] Haynes, T. and D. Noveck, "Network File System (NFS) version 4 [10] Haynes, T. and D. Noveck, "Network File System (NFS) version 4
Protocol", draft-ietf-nfsv4-rfc3530bis-09 (Work In Progress), Protocol", draft-ietf-nfsv4-rfc3530bis-09 (Work In Progress),
March 2011. March 2011.
[11] Eisler, M., "XDR: External Data Representation Standard", [11] Lentini, J., Everhart, C., Ellard, D., Tewari, R., and M. Naik,
RFC 4506, May 2006.
[12] Lentini, J., Everhart, C., Ellard, D., Tewari, R., and M. Naik,
"NSDB Protocol for Federated Filesystems", "NSDB Protocol for Federated Filesystems",
draft-ietf-nfsv4-federated-fs-protocol (Work In Progress), draft-ietf-nfsv4-federated-fs-protocol (Work In Progress),
2010. 2010.
[13] Lentini, J., Everhart, C., Ellard, D., Tewari, R., and M. Naik, [12] Lentini, J., Everhart, C., Ellard, D., Tewari, R., and M. Naik,
"Administration Protocol for Federated Filesystems", "Administration Protocol for Federated Filesystems",
draft-ietf-nfsv4-federated-fs-admin (Work In Progress), 2010. draft-ietf-nfsv4-federated-fs-admin (Work In Progress), 2010.
[14] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, L., [13] Fielding, R., Gettys, J., Mogul, J., Frystyk, H., Masinter, L.,
Leach, P., and T. Berners-Lee, "Hypertext Transfer Protocol -- Leach, P., and T. Berners-Lee, "Hypertext Transfer Protocol --
HTTP/1.1", RFC 2616, June 1999. HTTP/1.1", RFC 2616, June 1999.
[15] Postel, J. and J. Reynolds, "File Transfer Protocol", STD 9, [14] Postel, J. and J. Reynolds, "File Transfer Protocol", STD 9,
RFC 959, October 1985. RFC 959, October 1985.
[16] Simpson, W., "PPP Challenge Handshake Authentication Protocol [15] Simpson, W., "PPP Challenge Handshake Authentication Protocol
(CHAP)", RFC 1994, August 1996. (CHAP)", RFC 1994, August 1996.
[17] Strohm, R., "Chapter 2, Data Blocks, Extents, and Segments, of [16] Strohm, R., "Chapter 2, Data Blocks, Extents, and Segments, of
Oracle Database Concepts 11g Release 1 (11.1)", January 2011. Oracle Database Concepts 11g Release 1 (11.1)", January 2011.
[18] Ashdown, L., "Chapter 15, Validating Database Files and [17] Ashdown, L., "Chapter 15, Validating Database Files and
Backups, of Oracle Database Backup and Recovery User's Guide Backups, of Oracle Database Backup and Recovery User's Guide
11g Release 1 (11.1)", August 2008. 11g Release 1 (11.1)", August 2008.
[19] McDougall, R. and J. Mauro, "Section 11.4.3, Detecting Memory [18] McDougall, R. and J. Mauro, "Section 11.4.3, Detecting Memory
Corruption of Solaris Internals", 2007. Corruption of Solaris Internals", 2007.
[20] Bairavasundaram, L., Goodson, G., Schroeder, B., Arpaci- [19] Bairavasundaram, L., Goodson, G., Schroeder, B., Arpaci-
Dusseau, A., and R. Arpaci-Dusseau, "An Analysis of Data Dusseau, A., and R. Arpaci-Dusseau, "An Analysis of Data
Corruption in the Storage Stack", Proceedings of the 6th USENIX Corruption in the Storage Stack", Proceedings of the 6th USENIX
Symposium on File and Storage Technologies (FAST '08) , 2008. Symposium on File and Storage Technologies (FAST '08) , 2008.
[21] "Section 46.6. Multi-Level Security (MLS) of Deployment Guide: [20] "Section 46.6. Multi-Level Security (MLS) of Deployment Guide:
Deployment, configuration and administration of Red Hat Deployment, configuration and administration of Red Hat
Enterprise Linux 5, Edition 6", 2011. Enterprise Linux 5, Edition 6", 2011.
[22] Quigley, D. and J. Lu, "Registry Specification for MAC Security [21] Quigley, D. and J. Lu, "Registry Specification for MAC Security
Label Formats", draft-quigley-label-format-registry (work in Label Formats", draft-quigley-label-format-registry (work in
progress), 2011. progress), 2011.
[22] Eisler, M., "XDR: External Data Representation Standard",
RFC 4506, May 2006.
[23] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA [23] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA
Considerations Section in RFCs", BCP 26, RFC 5226, May 2008. Considerations Section in RFCs", BCP 26, RFC 5226, May 2008.
[24] Nowicki, B., "NFS: Network File System Protocol specification", [24] Nowicki, B., "NFS: Network File System Protocol specification",
RFC 1094, March 1989. RFC 1094, March 1989.
[25] Callaghan, B., Pawlowski, B., and P. Staubach, "NFS Version 3 [25] Callaghan, B., Pawlowski, B., and P. Staubach, "NFS Version 3
Protocol Specification", RFC 1813, June 1995. Protocol Specification", RFC 1813, June 1995.
[26] Srinivasan, R., "Binding Protocols for ONC RPC Version 2", [26] Srinivasan, R., "Binding Protocols for ONC RPC Version 2",
 End of changes. 173 change blocks. 
1112 lines changed or deleted 977 lines changed or added

This html diff was produced by rfcdiff 1.41. The latest version is available from http://tools.ietf.org/tools/rfcdiff/