2451
Comment: Complete FS-Cache
|
4571
NILFS
|
Deletions are marked like this. | Additions are marked like this. |
Line 1: | Line 1: |
Linux 2.6.30 is currently in development. The release candidate -rc1 was released the 7 April 2009, closing the merge window. | Linux 2.6.30 is currently in development. The last prepatch version is -rc2, released the 15th April. |
Line 9: | Line 9: |
== NILFS2 log-structured filesystem == [http://www.nilfs.org NILFS2] (new implementation of a log-structured file system 2) is a log-structured filesystem, which allows continuous snapshotting. |
== NILFS2, a log-structured filesystem == NILFS2 is a new filesystem, contributed by [http://en.wikipedia.org/wiki/Nippon_Telegraph_and_Telephone_Corporation NTT] (Nippon Telegraph and Telephone Corporation). It uses a [http://en.wikipedia.org/wiki/Log-structured_file_system log-structured] design. What makes log-structured filesystems different is that they treat the whole disk as a consecutive list of blocks (called log). All the operations append data at the end of the log, they never rewrite blocks (except when there's no space left - in that case, new blocks are reclaimed from the tail of the log). The advantage of this approach is that writes are always sequential. Crashes can't corrupt the filesystem. On mount, the filesystem detects the real end of the log, and continues working from that point. Another advantage of this approach is that the log offers a coherent historical view of all the operations done in the disk in the past. This is called "continuous snapshotting" - snapshots of modifications done in all the filesystem at any time are created automatically due to the log-structure design, with no requeriment of intervention from an admin, and with the filesystem size as the only limit. NILFS2 allows to access those snapshots and even mount them (on read-only mode). NILFS2 is [http://www.nilfs.org/en/current_status.html under development]. Code: [http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=tree;f=fs/nilfs2;hb=HEAD fs/nilfs2]. Filesystem web page: [http://www.nilfs.org www.nilfs.org] |
Line 17: | Line 23: |
== Local Filesystem Cache == | == Local Filesystem Caching Infrastructure == |
Line 19: | Line 25: |
== Filesystems performance improvement == Shortly after the 2.6.29 release, lots of discussions occurred on LKML about disk I/O (summary available [http://lwn.net/Articles/326471/ at LWN]), and how (and why) they can stall processes for minutes. Some measure have been taken to limit this, like setting the default mount-option to relatime, with 24 hours refresh time ([http://valhenson.livejournal.com/36519.html Detailed article in Valerie Aurora's blog]), and imporving fsync() for ext3. == Integrity Management Architecture == This is an infrastructure which use the Trusted Platform Module (TPM) to check that the OS has not been tampered with. ([LWN's article http://lwn.net/Articles/227937/] |
|
Line 27: | Line 42: |
* Integrity Management Architecture has been merged. |
Linux 2.6.30 is currently in development. The last prepatch version is -rc2, released the 15th April.
This page is, obviously, a work in progress.
Summary:
Prominent features (the cool stuff)
NILFS2, a log-structured filesystem
NILFS2 is a new filesystem, contributed by [http://en.wikipedia.org/wiki/Nippon_Telegraph_and_Telephone_Corporation NTT] (Nippon Telegraph and Telephone Corporation). It uses a [http://en.wikipedia.org/wiki/Log-structured_file_system log-structured] design. What makes log-structured filesystems different is that they treat the whole disk as a consecutive list of blocks (called log). All the operations append data at the end of the log, they never rewrite blocks (except when there's no space left - in that case, new blocks are reclaimed from the tail of the log). The advantage of this approach is that writes are always sequential. Crashes can't corrupt the filesystem. On mount, the filesystem detects the real end of the log, and continues working from that point.
Another advantage of this approach is that the log offers a coherent historical view of all the operations done in the disk in the past. This is called "continuous snapshotting" - snapshots of modifications done in all the filesystem at any time are created automatically due to the log-structure design, with no requeriment of intervention from an admin, and with the filesystem size as the only limit. NILFS2 allows to access those snapshots and even mount them (on read-only mode).
NILFS2 is [http://www.nilfs.org/en/current_status.html under development]. Code: [http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=tree;f=fs/nilfs2;hb=HEAD fs/nilfs2]. Filesystem web page: [http://www.nilfs.org www.nilfs.org]
Support for Object-Based Storage Device
Object-Based Storage Device (OSD) are basically "smart" disks, which do not store the data as blocks, but more elaborated objects. 2.6.30 adds support to these devices in the SCSI layer, as well as the addition of a filesystem (exofs, Extended Object File System) able to use them.
([http://lwn.net/Articles/305740/ LWN's Article], where exofs is still known by its old name, osdfs)
Local Filesystem Caching Infrastructure
FS-Cache is an infrastructure to locally cache remote (or slow) filesystems. It is described in more details in [http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob;f=Documentation/filesystems/caching/fscache.txt].
Filesystems performance improvement
Shortly after the 2.6.29 release, lots of discussions occurred on LKML about disk I/O (summary available [http://lwn.net/Articles/326471/ at LWN]), and how (and why) they can stall processes for minutes. Some measure have been taken to limit this, like setting the default mount-option to relatime, with 24 hours refresh time ([http://valhenson.livejournal.com/36519.html Detailed article in Valerie Aurora's blog]), and imporving fsync() for ext3.
Integrity Management Architecture
This is an infrastructure which use the Trusted Platform Module (TPM) to check that the OS has not been tampered with. ([LWN's article http://lwn.net/Articles/227937/]
Various core changes
- Scheduler
- Memory management
Wi-Fi
Security
- Integrity Management Architecture has been merged.
Networking
Tracing
Filesystems
- ReiserFS
- Lots of clean-ups.
- Ext3
- fsync() improvements.
- AFS
- Now makes use of the FSCache infrastructure.
- NFS
- Now makes use of the FSCache infrastructure.
- btrfs
- Stack usage improvements.
- NILFS2
- The filesystem has been merged for this release. (see above)
- exofs
- The filesystem has been merged for this release.(see above)
Crypto
DM/MD
Virtualization
Architecture-specific changes
- X86
- ARM
- PPC
- SH
- S390
- Blackfin
- MIPS
- SPARC
- AVR32
Drivers
Storage
- SATA:
- IDE
- SCSI
- Add OSD support. (see above)
Graphics
Network
Input
Sound
V4L/DVB
USB
HWMON
Watchdog
RTC
HID
MTD
MFD
Power
Serial
Various
Other sources about 2.6.30 kernel
- Heise Online's Kernel Log :
[http://www.h-online.com/open/Kernel-Log-Linux-2-6-30-is-taking-shape--/news/113036 Linux 2.6.30 is taking shape]
- Linux Weekly news :
[http://lwn.net/Articles/325921/ 2.6.30 merge window, part I]