« previous 1 2 3 4 5 6 7 8 next »

Main development phase of ext4 completed

When he began developing version 2.6.28, Torvalds integrated a large collection of patches prepared by Theodore Ts'o (tytso), a developer of (ext) file systems. Besides error corrections and improvements for the successor of ext3, this contains a patch to make the kernel cease referring to ext4 as ext4dev ("ext4 for developers"), and it is no longer referred to as "experimental". These changes are intended to make it clear that the file system structures used on data media will not incompatibly change, so future users won't need to make sure, as they once did, that the kernel version, the ext4 utilities (e2fsprogs), and the data media format will work together. The successor of ext3 thus leaves its "hot" development phase with the changes accepted in 2.6.28. The kernel developers accepted an early variant of ext4 Linux 2.6.19 so that it would enjoy further joint development and maturation in the main development tree.

The development of ext4 is in no way finished: on the contrary, the developers will probably keep on developing it over time, as they did with ext3, not only making error corrections, but also adding the occasional new function. It would also be premature to say that ext4 is now stable, for its developers are well aware of some errors it still contains, in the same way that modern CPUs appear with errata, though they probably won't show up in normal use. It won't be possible to describe ext4 as really stable until one of the larger mainstream distributions uses it as standard and it proves itself with a large number of users in a wide variety of configurations and hardware environments. Only then will the normally conservative enterprise distributions begin to support ext4 officially.

According to the rough notions of some important Linux filesystem developers, however, Ext4 is in any case only a temporary solution on the way to Btrfs, which is favoured as the next generation Linux file system. Some believe it will be included in the official Linux kernel in the coming months via the linux-next tree, to undergo further development and ripening there, more or less as happened with ext4 over the last two years.

Faster Better Wider

Some changes in 2.6.28 are exceedingly important but some are hidden away. For example, better kernel memory operation should provide a performance boost. After several years of development, the kernel development team has adopted a number of "VM pageout scalability improvements", driven primarily by Rik van Riel, which revamp and optimise parts of the memory subsystem (including 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12). This is intended to reduce the administrative overhead for memory-intensive applications and for systems with large amounts of RAM. These changes may not result in any performance changes for desktops or small servers; but with large servers and HPC systems, they may yeild improvements. Details and background on the changes can be found in an LWN.net article. A number of changes introduced by Nick Piggin will also specifically reduce the CPU load arising from memory management (including 1, 2, 3, 4, 5, 6) – LWN.net also has background information on this.

In 2.6.28 the new Graphics Execution Manager (GEM) takes over some of the graphics drivers' memory management tasks and co-ordinates the access to the graphics chip's processing units; this is designed to take care of some limitations in the current driver design and to improve performance. GEM was started several months ago by Intel developers as a short term alternative to Translation Table Maps (TTM), the memory manager by Tungsten Graphics that was originally intended for this job. (see also vs. GEM. TTM on LWN.net).

Important New Drivers

Many Linux 2.6.28 drivers appreciably improve the support of modern PC components. Version 2.6.28 will include a special driver for notebooks equipped with Elantech touchpads – many of Asus' EeePCs among them – that, along with current versions of the Synaptics driver for X, enable extended touchpad features such as gestures, see – (documentation). Also eagerly anticipated by Asus users is the atl2 driver for the Atheros L2 which has been appearing on several, mostly cheaper, motherboards with 100Mbit LAN ports built in.

Although not widespread, but of importance for the long term in 2.6.28, is the Intel developed support for UWB/Ultra Wide Band and Wireless USB. Unlike many previous kernel versions, no new Wi-Fi drivers are being added to this version. Instead, more popular Wi-fi drivers have been worked on. The ath5k driver for Atheros WLAN chips now supports mesh networks and the Atheros AR2417 v2 chip; the rt2x00 driver now makes use of the encryption technology of various RaLink Wi-Fi chips (1, 2, 3, 4).

The new generic Wireless Regulatory Infrastructure should, in future, be more flexible and better ensure that Wi-Fi hardware does not infringe local rules and regulations on wireless connections. In a change from the previous kernel code, the new infrastructure makes considerable use of a program running in userspace, which can be updated much more easily and rapidly by the distributors than modifying the kernel. Details of how it works and lots of background information on this issue can be found in the commit notes, kernel documentation, an article on LWN.net and a description of the Central Regulatory Domain Agent in the Linux wireless wiki

Next: Improvements for Notebooks and SSDs

« previous 1 2 3 4 5 6 7 8 next »