SMB3 kernel status: Difference between revisions

From SambaWiki
 
(19 intermediate revisions by 5 users not shown)
Line 1: Line 1:
[[Category:CIFSclient]]

= Introduction =
= Introduction =


This page describes the plan, design and work in progress of the efforts to implement SMB3 and later dialects in the kernel cifs/smb3 client (cifs.ko).
This page describes the plan, design and work in progress of the efforts to implement SMB3 and later dialects in the kernel cifs/smb3 client (cifs.ko).


The minimum version for use of the Linux kernel SMB3 support is kernel version 3.12 (or backport of cifs.ko module version 2.02 or later) but kernel version 3.18 or later (or equivalent, ie cifs module version 2.04 or later) is recommended for best SMB3 support.
The minimum version for use of the Linux kernel SMB3 support is kernel version 3.12 (or backport of cifs.ko module version 2.02 or later) but kernel version 4.11 or later is recommended due to the addition of the SMB3 share encryption security feature. The default dialect was changed to SMB2.1 or later in the 4.14 Kernel.


Kernel client features/fixes by release can be found at https://wiki.samba.org/index.php/LinuxCIFSKernel
Kernel client features/fixes by release can be found at https://wiki.samba.org/index.php/LinuxCIFSKernel
Line 12: Line 14:
** Features done:
** Features done:
*** multi credit/large MTU
*** multi credit/large MTU
*** leases, resilient file handles
leases

** Features TODO:
** Features TODO:
*** resilient file handles
*** branch cache
*** branch cache
* SMB 3 (previously known as SMB2.2 dialect) was introduced with Windows 8 and Windows Server 2012. SMB3 support in the kernel wsa much improved in kernel version 3.12. SMB3 dialect defines the following features:
* SMB 3 (previously known as SMB2.2 dialect) was introduced with Windows 8 and Windows Server 2012. SMB3 support in the kernel was much improved in kernel version 3.12. SMB3 dialect defines the following features:
** Basic support for SMB3 is included, as are security improvements (improved faster more secure packet signing, secure negotiate protection against downgrade attacks
** Support for SMB3 is included, as are security improvements (improved faster more secure packet signing, secure negotiate protection against downgrade attacks
** In addition the client can now do network interface discovery (a new FSCTL)
**Still need to do:
***share level encryption
*** directory leases
*** directory leases
*** persistent file handles
*** multi channel
*** multi channel
****and also SMB direct (SMB3 over RDMA)
*** witness notification protocol (a new RPC service)
*** witness notification protocol (a new RPC service)
** In addition the client can now do network interface discovery (a new FSCTL),
** Still need to do:
*** Support for a misc. set of loosely related storage features for virtualization (new fsctls, T10 block copy offload, TRIM etc.)
*** Support for a misc. set of loosely related storage features for virtualization (new fsctls, T10 block copy offload, TRIM etc.)
*** remote shadow copy support
*** remote shadow copy support
*** branch cache v2
*** branch cache v2
* SMB3.02 was introduced in Windows 8.1 (Windows 'Blue') and Windows Server 2012 R2. Among the new protocol features are those particularly useful for virtualization (HyperV):
* SMB3.02 was introduced in Windows 8.1 (Windows 'Blue') and Windows Server 2012 R2. Among the new protocol features are those particularly useful for virtualization (HyperV):
** SMB3.02 dialect is not yet negotiated by Samba servers
** SMB3.02 dialect can be requested by the Linux cifs client ("vers=3.02" on mount)
** SMB3.02 dialect can be requested by the Linux cifs client ("vers=3.02" on mount) but the new optional features, unique to SMB3.02, are not requested.
** Unbuffered I/O flags (ie a 'no cache flag' which may be sent on read or write)
** Unbuffered I/O flags (ie a 'no cache flag' which may be sent on read or write)
** New RDMA remote invalidate flag
** New RDMA remote invalidate flag
Line 39: Line 35:
* SMB3.1.1 was introduced in Windows 10. Among the new features defined:
* SMB3.1.1 was introduced in Windows 10. Among the new features defined:
** Improvements to security negotiation ("negotiate contexts") and dynamically selectable Cipher and Hash Algorithms.
** Improvements to security negotiation ("negotiate contexts") and dynamically selectable Cipher and Hash Algorithms.
** New FSCTL for server side copying of file ranges (DUPLICATE_EXTENTS)
** Also the cifs client has added support for the new FSCTL for server side copying of file ranges (DUPLICATE_EXTENTS)


= Prerequisite / accompanying work =
= Prerequisite / accompanying work =
Line 63: Line 59:


== Leases ==
== Leases ==
We request leases on opens (unless "cache=none" mount option is used, or unless oplock is disabled in the module parameters when cifs.ko is loaded). There is no mechanism to do request lease upgrades currently.
We request leases on opens (unless "cache=none" mount option is used, or unless oplock is disabled in the module parameters when cifs.ko is loaded). There is no mechanism to do request lease upgrades currently. Leases continue to provide significant performance benefits in SMB3, just as oplock did in the cifs protocol.


== Resilient File Handles ==
== Resilient File Handles ==
mount with option "resilienthandles" (starting with 4.4 kernel)


== Branch Cache ==
== Branch Cache ==
Line 73: Line 70:
= SMB 3.0 =
= SMB 3.0 =
== Security Features ==
== Security Features ==
* Encryption and improved packet signing
* Encryption and improved packet signing (in progress)
* Secure negotiate (complete)
* Secure negotiate (complete)


Line 79: Line 76:


== Directory Leases ==
== Directory Leases ==

This does speed up revalidating path names since the root directory existing etc. no longer needs to be requeried when opening a file or subdirectory. And directory contents for subdirectories can be cached.


=== Ideas/concepts ===
=== Ideas/concepts ===
Line 91: Line 90:
Persistent file handles are a like durable file handles with strong guarantees.
Persistent file handles are a like durable file handles with strong guarantees.
They are requested with the durable v2 create request blob with the persistent
They are requested with the durable v2 create request blob with the persistent
flag set to true. The server only grants persistent handles on shares that are
flag set to true. Persistent file handles are supported with mount option
"persistenthandles" (starting with Linux kernel 4.4), but implementation is
marked CA (continuously available).
not quite complete as the "ChannelSequenceNumber" is not yet set

on reconnect.
There is no finished design yet for the implementation of persistent handles.
The foundations have been layed with the introduction of durable handles.
The challenge is to implement the additional guarantees.


[https://www.youtube.com/watch?v=fA78hdeYi6k Persistent Handles]


== Multi Channel ==
== Multi Channel ==
Multichannel allows use of multiple network interfaces on the server (and optionally on the client) to significantly improve available network bandwidth. Experimental multichannel support was added to the kernel client in the 5.5 kernel, but it was limited in scalability for large i/o due to locking. Improvements to locking to allow significantly more scalability for large file i/o will be in the 5.8 kernel.


== Witness Notification Protocol ==


The witness service is an RPC service that allows a client to be actively notified about the state change of resources. The client asks the node it is connected to for a list of interfaces and registers itself on a different node with the witness service for notification about a resource, which might be a network name (typically the DNS name of the file server), or an interface group and an IP address. Afterward it can request notification.
== Witness Notification Protocol ==


== Client Side Witness Daemon ==
Gunther has a prototype of this, but it is not packaged with cifs-utils yet


=== Admin Tools ===
=== Admin Tools ===
Line 109: Line 111:
* we need a tool to display the witness registrations
* we need a tool to display the witness registrations
* we need a tool to move client to a different node
* we need a tool to move client to a different node



== SMB Direct ==
== SMB Direct ==


aka ''SMB 3.0 over RDMA''
aka ''SMB 3.0 over RDMA''

Implemented in the 4.16 kernel and marked non-experimental in the 5.3 kernel. It provides significant performance benefits (and CPU reduction) for large i/o.


Requires Multi-Channel
Requires Multi-Channel
Line 121: Line 124:


=== RDMA Read/Write support in the client ===
=== RDMA Read/Write support in the client ===
Provides significant benefit in large i/o and reduction in CPU utilization.




== Remote Shadow Copy (FSRVP) ==
== Remote Shadow Copy (FSRVP) ==
Line 135: Line 137:
= SMB3.02 =
= SMB3.02 =
See http://www.snia.org/sites/default/files2/SDC2013/presentations/SMB3/DavidKruse_SMB3_Update.pdf
See http://www.snia.org/sites/default/files2/SDC2013/presentations/SMB3/DavidKruse_SMB3_Update.pdf
SMB3.02 is very similar to SMB3 but with some optional features added. Note that the Linux CIFS client can negotiate SMB3.02 dialect (with these optional features disabled) by specifying vers=3.02 on mount. Samba server can not currently negotiate SMB3.02 as it does not have support for the new READ/WRITE flags (and the RDMA and Witness protocol improvements for SMB3.02 are not possible until the corresponding prerequisite optional SMB3.0 features that they are based on are added).
SMB3.02 is very similar to SMB3 but with some optional features added. Note that the Linux CIFS client can negotiate SMB3.02 dialect (with these optional features disabled) by specifying vers=3.02 on mount. Samba server can not currently negotiate SMB3.02 as it does not have support for the new READ/WRITE flags (and the Witness protocol improvements for SMB3.02 are not possible until the corresponding prerequisite optional SMB3.0 features that they are based on are added).


Currently cifs.ko can negotiate SMB3.02 dialect (vers=3.02) but does not request the optional features listed below so a vers=3.02 mount acts much like a vers=3.0 mount.
Currently cifs.ko can negotiate SMB3.02 dialect (vers=3.02) but does not request the optional features listed below so a vers=3.02 mount acts much like a vers=3.0 mount.

Latest revision as of 17:16, 11 November 2022


Introduction

This page describes the plan, design and work in progress of the efforts to implement SMB3 and later dialects in the kernel cifs/smb3 client (cifs.ko).

The minimum version for use of the Linux kernel SMB3 support is kernel version 3.12 (or backport of cifs.ko module version 2.02 or later) but kernel version 4.11 or later is recommended due to the addition of the SMB3 share encryption security feature. The default dialect was changed to SMB2.1 or later in the 4.14 Kernel.

Kernel client features/fixes by release can be found at https://wiki.samba.org/index.php/LinuxCIFSKernel

  • SMB 2.0 (SMB2.02 dialect) was introduced with Windows Vista/2008 and includes a useful data integrity feature ("durable file handles"). Implementation in the kernel client is complete.
  • SMB 2.1 was introduced with Windows 7/Windows 2008R2.
    • Basic support for SMB 2.1 was added in kernel version 3.07
    • Features done:
      • multi credit/large MTU
      • leases, resilient file handles
    • Features TODO:
      • branch cache
  • SMB 3 (previously known as SMB2.2 dialect) was introduced with Windows 8 and Windows Server 2012. SMB3 support in the kernel was much improved in kernel version 3.12. SMB3 dialect defines the following features:
    • Support for SMB3 is included, as are security improvements (improved faster more secure packet signing, secure negotiate protection against downgrade attacks
      • directory leases
      • multi channel
      • witness notification protocol (a new RPC service)
    • In addition the client can now do network interface discovery (a new FSCTL),
    • Still need to do:
      • Support for a misc. set of loosely related storage features for virtualization (new fsctls, T10 block copy offload, TRIM etc.)
      • remote shadow copy support
      • branch cache v2
  • SMB3.02 was introduced in Windows 8.1 (Windows 'Blue') and Windows Server 2012 R2. Among the new protocol features are those particularly useful for virtualization (HyperV):
    • SMB3.02 dialect can be requested by the Linux cifs client ("vers=3.02" on mount)
    • Unbuffered I/O flags (ie a 'no cache flag' which may be sent on read or write)
    • New RDMA remote invalidate flag
    • MS-RSVD (a set of remoteable FSCTLs that improve "SCSI over SMB3")
    • Asymmetric Shares (extensions to Witness protocol to allow moving users of one share to a different server, eg for load balancing or maintenance - previously witness protocol could only do this on a per server rather than per-share basis).
  • SMB3.1.1 was introduced in Windows 10. Among the new features defined:
    • Improvements to security negotiation ("negotiate contexts") and dynamically selectable Cipher and Hash Algorithms.
    • Also the cifs client has added support for the new FSCTL for server side copying of file ranges (DUPLICATE_EXTENTS)

Prerequisite / accompanying work

SMB 2.0

durable handles

implement durable open and durable reconnect with reopening files

Complete

Durable handle cross-node

No mechanism currently in place to reconnect to server other than the one that you have mounted against (e.g. we do not fail over to an alternate DFS referral if two servers export the same DFS path, and we do not support witness protocol failover so can't reconnect to a different server yet).

Unbuffered Write

Supported by mount option cache=strict, or disabling oplocks - but no Linux mechanism exists for control of buffering on a per-write granularity

Multi Credit / Large MTU

Supported


Leases

We request leases on opens (unless "cache=none" mount option is used, or unless oplock is disabled in the module parameters when cifs.ko is loaded). There is no mechanism to do request lease upgrades currently. Leases continue to provide significant performance benefits in SMB3, just as oplock did in the cifs protocol.

Resilient File Handles

mount with option "resilienthandles" (starting with 4.4 kernel)

Branch Cache

SMB 3.0

Security Features

  • Encryption and improved packet signing (in progress)
  • Secure negotiate (complete)

Replay/Retry Detection

Directory Leases

This does speed up revalidating path names since the root directory existing etc. no longer needs to be requeried when opening a file or subdirectory. And directory contents for subdirectories can be cached.

Ideas/concepts

  • Directory leases are a mechanism for caching metadata read operations/directory listings of child objects of a directory (File leases are a mechanism for caching the data operations.)
  • The client maintains separate caches for each user context, but still using just one lease to invalidate the cache. This is needed because access based enumeration may cause different directory listing depending on the user context.

Persistent File Handles

Introduction

Persistent file handles are a like durable file handles with strong guarantees. They are requested with the durable v2 create request blob with the persistent flag set to true. Persistent file handles are supported with mount option "persistenthandles" (starting with Linux kernel 4.4), but implementation is not quite complete as the "ChannelSequenceNumber" is not yet set on reconnect.

Persistent Handles

Multi Channel

Multichannel allows use of multiple network interfaces on the server (and optionally on the client) to significantly improve available network bandwidth. Experimental multichannel support was added to the kernel client in the 5.5 kernel, but it was limited in scalability for large i/o due to locking. Improvements to locking to allow significantly more scalability for large file i/o will be in the 5.8 kernel.

Witness Notification Protocol

The witness service is an RPC service that allows a client to be actively notified about the state change of resources. The client asks the node it is connected to for a list of interfaces and registers itself on a different node with the witness service for notification about a resource, which might be a network name (typically the DNS name of the file server), or an interface group and an IP address. Afterward it can request notification.

Client Side Witness Daemon

Gunther has a prototype of this, but it is not packaged with cifs-utils yet

Admin Tools

  • we need a tool to display the witness registrations
  • we need a tool to move client to a different node

SMB Direct

aka SMB 3.0 over RDMA

Implemented in the 4.16 kernel and marked non-experimental in the 5.3 kernel. It provides significant performance benefits (and CPU reduction) for large i/o.

Requires Multi-Channel

SMB-Direct backend for smb_transport abstraction

RDMA Read/Write support in the client

Provides significant benefit in large i/o and reduction in CPU utilization.

Remote Shadow Copy (FSRVP)

Not an SMB 3.0 specific feature per se.

  • Need to add:
    • add rpcclient support for FSRVP commands
    • implement user interface (/proc or /sys or ioctl) and tools for this.

Branch Cache v2

Branch Cache is a wide area network caching protocol implemented in Windows 7 and later. It allows the server to return hashes of the data to the client, and then the client can use these hashes to request copies of the actual data from nearby systems, optimizing network bandwidth. Although Branch Cache is not SMB3 specific (e.g. HTTP etc) it is useful in conjunction with SMB2.1 and SMB3 file serving to improve WAN performance and better optimize bandwidth usage. See MS-PCCRC, MS-PCCRD, MS-PCCRR.

SMB3.02

See http://www.snia.org/sites/default/files2/SDC2013/presentations/SMB3/DavidKruse_SMB3_Update.pdf SMB3.02 is very similar to SMB3 but with some optional features added. Note that the Linux CIFS client can negotiate SMB3.02 dialect (with these optional features disabled) by specifying vers=3.02 on mount. Samba server can not currently negotiate SMB3.02 as it does not have support for the new READ/WRITE flags (and the Witness protocol improvements for SMB3.02 are not possible until the corresponding prerequisite optional SMB3.0 features that they are based on are added).

Currently cifs.ko can negotiate SMB3.02 dialect (vers=3.02) but does not request the optional features listed below so a vers=3.02 mount acts much like a vers=3.0 mount.

RDMA Improvements

SMB Direct Remote Invalidation. Improves performance.

New ReadWrite Flags

SMB2_READFLAG_UNBUFFERED and SMB2_WRITEFLAG_UNBUFFERED allow the client to indicate whether or not any particular individual i/o request (read or write) should be cached by the server or not. There are no interfaces in the Linux kernel VFS for per i/o flags yet, so support for this on the wire would require a private ioctl on the client.

Asymmetric Shares

The Witness protocol can now signal to Windows clients to 'move' from one share to another, to allow more flexible migration, allowing taking a volume offline without taking the whole server down, with applications continuing to run even as the storage which that application uses is moved. Previous versions of the witness protocol allowed users of one server to be moved to another server, but this allows more granular movement - those using a particular share now can be redirected on the fly to another share.

Cluster-Wide Durable Handles

Work in progress

Talks

Demos