Roadmap: Difference between revisions

From SambaWiki
 
(49 intermediate revisions by 9 users not shown)
Line 1: Line 1:
== Introduction ==
= Introduction =


This page describes the bigger next steps in the development of Samba. The purpose is to point out the broader direction into which Samba is heading.
This page describes the bigger next steps in the development of Samba. The purpose is to point out the broader direction into which Samba is heading.
Line 5: Line 5:
If a feature listed below is flagged as '''FUNDED''', this means that someone is currently being paid to work on it. Hence there are realistic chances that this feature might be completed in a reasonably short time frame. For all other features, further involvement is needed: Otherwise it could even take years to complete even if a feature is flagged as work in progress (WIP), since these are usually being worked on in someone's spare time.
If a feature listed below is flagged as '''FUNDED''', this means that someone is currently being paid to work on it. Hence there are realistic chances that this feature might be completed in a reasonably short time frame. For all other features, further involvement is needed: Otherwise it could even take years to complete even if a feature is flagged as work in progress (WIP), since these are usually being worked on in someone's spare time.


Involvement is highly welcome and can come in various guises: '''Manpower for coding, testing, documentation, ...'''
Involvement is highly welcome and can come in various guises: manpower for coding, testing, documentation, ...


Contact the Samba Team go get involved!
'''Contact the Samba Team go get involved!'''

= Features =


== File Server (smbd) ==
== File Server (smbd) ==

=== Infrastructure ===

* s4u2self (Correct authorization in complex AD topologies on Linux AD member for non-SMB acccess)
* ksmbd integration


=== [[Samba3/SMB2|SMB2/SMB3]] ===
=== [[Samba3/SMB2|SMB2/SMB3]] ===


* Witness service - Prototype
* '''FUNDED''': Implment Multi-channel ([[User:Obnox|Michael]], [[User:Metze|Metze]])
* Clustering (Continuous availability, Scale-out)
* '''FUNDED''': Add support for Witness service (Günther, [[User:Metze|Metze]])
* Persistent file handles - Prototype ([[User:slowfranklin|slow]])
** Prerequisite: A asynchronous RPC server. See [[DCERPC]]
* Clustering (continuous availability, scale-out) - Planning ([[User:Obnox|Michael]])
* RDMA (SMB direct) - Prototype ([[User:Metze|Metze]])
* Persistent file handles - Planning (Ira, [[User:Obnox|Michael]], ...)
* Directory leases - Prototype ([[User:slowfranklin|slow]])
* SMB2 UNIX Extensions
* RDMA (SMB direct) - Planning ([[User:Metze|Metze]], [[User:Rsharpe|Richard]], [[User:Obnox|Michael]])
* Directory leases


=== Clustering - CTDB ===
=== Clustering - CTDB ===


* Split up locking.tdb for better scaling of per-file concurrent access
* integrate the clustered file server into selftest/autobuild - WIP ([[User:Obnox|Michael]])

=== File Systems ===

Support for special features of various file systems, especially cluster file systems, typically through VFS modules.

* gpfs ('''FUNDED''')
* GlusterFS ('''FUNDED''')
* CephFS


=== Performance ===
=== Performance ===


Performance tuning and optimization is an important reoccurring topic. It is difficult
* Performance on small-CPU platforms (like ARM)
to really track the current issues...
** reduce CPU usage

* Database/TDB-Performance
* Use iouring with splice() - Prototype ('''FUNDED''')
* performance in clusters, TDB/CTDB
* Performance in clusters, TDB/CTDB
* Parallel, small I/O (HyperV) workload
* Parallel, small I/O (HyperV) workload


== Print Server (smbd|spoolssd) ==
== Print Server (smbd|spoolssd) ==


* 'MS-PAR'
* [[Spoolss|SPOOLSS]]
* [[Spoolss|SPOOLSS]]
** Improve Spoolss Server performance
** Improve Spoolss Server performance
Line 42: Line 58:
== Active Directory Server ==
== Active Directory Server ==


* S4U2Self, S4U2Proxy, PKINIT ... You can find more details at: [[Roadmap_MIT_KDC]]
* '''FUNDED:''' MIT Kerberos support (Andreas)
* '''FUNDED:''' Two-way forest trusts (Metze)
* Two-way forest trusts (Metze)
* [[The_Samba_AD_DNS_Back_Ends]]
* '''FUNDED:''' [[Samba4/LDAP Backend|Use the OpenLDAP Database engine and LDAP protocol handlers in AD server]]
* '''FUNDED:''' Knowledge Consistency Checker (KCC)

* [[DNS]]
* [[Samba4/DRS_TODO_List|Directory Replication Service (DRS)]]
* [[Samba4/DRS_TODO_List|Directory Replication Service (DRS)]]
* Sysvolume Replication
* sysvol Replication (FRS/DFSR)
* Read-only Domain Controller (RODC)
* Subdomain support
* Subdomain support
* One way trusts
* One way trusts
Line 58: Line 72:
The RPC server infrastructure component is of crucial importance for both the file server and the active directory server. A few tasks in for the RPC server are prerequisites for higher level features in the file server and the active directory server. See [[DCERPC]] for details.
The RPC server infrastructure component is of crucial importance for both the file server and the active directory server. A few tasks in for the RPC server are prerequisites for higher level features in the file server and the active directory server. See [[DCERPC]] for details.


* Merge source3 and source4 server and client implementations (Metze)
* Merge source3 and source4 client implementations (Metze)
* Make RPC server (and client) implementation fully asynchronuous (Metze)
* Make RPC client implementation fully asynchronuous (Metze)
* Merge [[Endpoint_Mapper|endpoint mapper]] implementations
* Merge [[Endpoint_Mapper|endpoint mapper]] implementations
* Implement Association groups
* Implement Association groups
* async schannel (NETLOGON) client (Metze)
* merged crypto handling for samlogon cred validation (Günther/Metze)
* merged libnetjoin interfaces (Günther/Metze)


== Testing ==
== Testing ==
Line 68: Line 85:
* Rewrite and improve the Selftest Suite
* Rewrite and improve the Selftest Suite


= Completed tasks =
------------------------------------------------------------------------------------------------


* [[DNS]]
* [[Samba4/DRS_TODO_List]]
* sysvol replication (file system replication)
** FRS / DFSR
*** See [[DCERPC]]
** use source3/winbindd in AD server
** subdomains
** forest trusts
* MIT Kerberos support in AD server (IN_PROGRESS, '''HELP NEEDED, TESTING NEEDED'''! Talk to Günther and Andreas to help with resources)
** Create preloadable wrappers (DONE)
** Re-test existing MIT KDC glue around the hdb backend (this has been completely rewritten)

* [[Samba4/LDAP Backend|Use the OpenLDAP Database engine and LDAP protocol handlers in AD server]]

=== Trust support ===

A lot of DCE/RPC work needs to be done before we can really finish this task.

* async schannel (NETLOGON) client (Metze)
* merged crypto handling for samlogon cred validation (Günther/Metze)
* merged libnetjoin interfaces (Günther/Metze)
* DCE/RPC client API changes (Günther/Metze)
** tevent based async infrastructure (new services: witness, replication, snapshot, PAR)
** client context (get rid of lp_ usage and globals for client details)
** association group implementation
* DCE/RPC server API merge and cleanup
* See [[DCERPC]]

==== Abstract data model ====

We might want to implement the abstract data model as outlined in MS-ADPS and other specs. This would allow to abstract the existing interfaces (dsdb/passdb) which both are not the best match for trust handling. Once we have that, and we have the common DCE/RPC infrastructure, we can easily plug implementations for trusts (e.g. from s3) into s4 (and vice-versa).

==== Testing ====

* We need to find a way to setup multi-trust environments during make test to test more compelx scenarios (e.g. the forst trust test from Sumit Bose).
* We need to test trust calls on MEMBER and DC configurations.



== Completed ==


=== File Server ===
=== File Server ===
Line 118: Line 93:
* SMB 2.1 Multi-Credit
* SMB 2.1 Multi-Credit
* SMB 3.0 protocol support (including encryption)
* SMB 3.0 protocol support (including encryption)
* Multichannel


* Transparent file compression
* Transparent file compression
* Serverside copy using COPYCHUNK
* Serverside copy using COPYCHUNK


* Improved performance on small-CPUs
* Improved performance on small-CPUs
* Improved TDB Database performance (using robust mutex locking)
* Improved TDB database performance (using robust mutex locking)


=== Clustering - CTDB ===
=== Clustering - CTDB ===


* integrate CTDB master into samba master:
* integrate CTDB master into samba master:
** integrate the code under ctdb/ - DONE
** integrate the code under ctdb/
** integrate the build into the top level waf build - DONE
** integrate the build into the top level waf build


=== Active Directory Server ===
=== Active Directory Server ===

* internal dns server
* use smbd as file server
* use winbindd for id-mapping
* Correct non-mesh inter-site and intra-site replication via Knowledge Consistency Checker (KCC) (Andrew Bartlett, Garming Sam, Douglas Bagnall)


=== DCERPC Infrastructure ===
=== DCERPC Infrastructure ===


* common secure channel implementation
*
* Merge source3 and source4 server implementations
* Make RPC server implementation fully asynchronuous

=== Testing ===

* Implement preloadable wrappers for better testing - [https://cwrap.org The cwrap project]

Latest revision as of 15:38, 20 January 2022

Introduction

This page describes the bigger next steps in the development of Samba. The purpose is to point out the broader direction into which Samba is heading.

If a feature listed below is flagged as FUNDED, this means that someone is currently being paid to work on it. Hence there are realistic chances that this feature might be completed in a reasonably short time frame. For all other features, further involvement is needed: Otherwise it could even take years to complete even if a feature is flagged as work in progress (WIP), since these are usually being worked on in someone's spare time.

Involvement is highly welcome and can come in various guises: manpower for coding, testing, documentation, ...

Contact the Samba Team go get involved!

Features

File Server (smbd)

Infrastructure

  • s4u2self (Correct authorization in complex AD topologies on Linux AD member for non-SMB acccess)
  • ksmbd integration

SMB2/SMB3

  • Witness service - Prototype
  • Clustering (Continuous availability, Scale-out)
  • Persistent file handles - Prototype (slow)
  • RDMA (SMB direct) - Prototype (Metze)
  • Directory leases - Prototype (slow)
  • SMB2 UNIX Extensions

Clustering - CTDB

  • Split up locking.tdb for better scaling of per-file concurrent access

File Systems

Support for special features of various file systems, especially cluster file systems, typically through VFS modules.

  • gpfs (FUNDED)
  • GlusterFS (FUNDED)
  • CephFS

Performance

Performance tuning and optimization is an important reoccurring topic. It is difficult to really track the current issues...

  • Use iouring with splice() - Prototype (FUNDED)
  • Performance in clusters, TDB/CTDB
  • Parallel, small I/O (HyperV) workload

Print Server (smbd|spoolssd)

Active Directory Server

DCERPC infrastructure

The RPC server infrastructure component is of crucial importance for both the file server and the active directory server. A few tasks in for the RPC server are prerequisites for higher level features in the file server and the active directory server. See DCERPC for details.

  • Merge source3 and source4 client implementations (Metze)
  • Make RPC client implementation fully asynchronuous (Metze)
  • Merge endpoint mapper implementations
  • Implement Association groups
  • async schannel (NETLOGON) client (Metze)
  • merged crypto handling for samlogon cred validation (Günther/Metze)
  • merged libnetjoin interfaces (Günther/Metze)

Testing

  • Multi-trust environments setup to test trusts
  • Rewrite and improve the Selftest Suite

Completed tasks

File Server

  • SMB 2.0 durable file handles
  • SMB 2.1 Leases
  • SMB 2.1 Multi-Credit
  • SMB 3.0 protocol support (including encryption)
  • Multichannel
  • Transparent file compression
  • Serverside copy using COPYCHUNK
  • Improved performance on small-CPUs
  • Improved TDB database performance (using robust mutex locking)

Clustering - CTDB

  • integrate CTDB master into samba master:
    • integrate the code under ctdb/
    • integrate the build into the top level waf build

Active Directory Server

  • internal dns server
  • use smbd as file server
  • use winbindd for id-mapping
  • Correct non-mesh inter-site and intra-site replication via Knowledge Consistency Checker (KCC) (Andrew Bartlett, Garming Sam, Douglas Bagnall)

DCERPC Infrastructure

  • common secure channel implementation
  • Merge source3 and source4 server implementations
  • Make RPC server implementation fully asynchronuous

Testing