2.6 is out, but what does it bring?

Finally, the 2.6 kernel is an official release. What does it have that you want? When is your Linux Distro Likely to have it? Heres a very brief 2.6 Rundown.

The 2.6 Kernel includes some super new features. If you want a fully loaded description, read the article by Joseph Pranevich here.

Firstly, Hardware support: Major improvements here: Support for a large number of new CPU’s and architectures. All the way from the likes of Hitachi’s H8/300 series, the NEC v850 processor right down to the latest Dragonball and ColdFire chips on eval boards from Motorola, Lineo, Arcturus, and others.

With Embedded Linux firmly in mind the 2.6 kernel is the first propper release with the uClinux work merged back in.

At the opposite end of the scale, NUMA support means much bigger SMP boxes are within grasp without curtailed memory access.

Internally a new sub-architecture system allows the processor type to be independant from the architecture: in previous versions special code was needed for the the same CPU in slightly different architectures. The sub architecture systems fixes the need for this code.

Hyperthreading support (first available in Pentium 4’s) is supported: This allows a single physical processor to masquerade (at the hardware level) as two or more processors. This in turn allows for performance boosts in some circumstances, but also adds scheduling complexity and other issues.

There are other scaleability improvement too: 2.6 has other changes for Intel servers at the top of the food chain. First and foremost is improved support for other new Intel hardware features including Intel’s PAE (“Physical Address Extension”) which allows most newer 32-bit x86 systems to access up to 64GB of RAM, but in a paged mode. In addition,

To give the user a more responsive feel (Critical for Linux-On-The-Desktop projects), the kernel is finally pre-emptible. This means that under Linux 2.6, the kernel now can be interrupted mid-task, so that other applications can continue to run even when something low-level and complicated is going on in the background. Although we are only talking about tiny fractions of a second: some users will see considerable improvements.

The IO subsystem generally has been revamped to give good performance accross a wide range of systems and hardware. It’s also far less likely to get lock-ups or race conditions whilst waiting for a resource.

At the module level, a significant rethink to the module description code and the module loading/unloading process has meant that there is now the promise of far better support for hot plug hardware such as found in Laptops etc.

In fact, the APM code has become ACPI compliant: this in turn brings linux up to the current state of the art in hardware power management support. Laptops are the goal here again although we should not be blinkered into thinking that laptop users are the only beneficiaries. Talking of laptops, there is better support for the hardware suspend modes offered from the main vendors.

Ext 3 has been revamped to allow the use of extended ACL’s: this is a clear requirement if Linux (with a native F/S) is to achieve NT’esque file serving. Don’t hold your breath though: many user space tools will need rewriting.

At the multimedia colaface, ALSA has finally replaced OSS: Improvements this brings are many: better SMP support for multimedia machines, proper full duplex sound, multiple sound cards in a single machine. The list goes on.

On the security front, Linux can fianally support Hardware Random Number Generators (Vital for strong cryptography at speed). Also Binary modules no longer have the ability to overload the system with calls. Most significant though is the granularisation of the old superuser priveledges into something modularized.

The virtualisation code is merged back in allowing a stock linux box to run a virtual linux kernel. This significantly simplifies security architecture tasks: honeypots fdor example. Also other development tasks are made much easier with this included.

All in all, 2.6 looks set to raise the benchmark yet again in terms of Performance, scaleability and breadth of hardware supported.

So what about mainstream availability?. Well if mainstream to you includes Fedora, then the release date is around April 2004. Check your preferred vendor for their release schedules.

TurboTas 2003.

Leave a Reply