Improvements throughout – what's new in Linux 2.6.25

In association with heise online

Lockout

Locked out

Information source: source code administration – Many of the links in this article and in the [anchorlink Appendix]brief description of less important changes[/anchorlink] appended to it, point to some related commit or other in the web front end of the Linux source code that Linus Torvald administers using Git. A great deal of additional information can be found there dealing with individual changes and background information on them. The patches used in various commits can be found in the web front end under the Commitdiff link and the accompanying comments and documentation changes often make them a good source of additional information, even for interested non-programmers. On the lower part of displayed page, the Git web front end also shows a list of files that have been changed in connection with the corresponding commit. To the right of the file names are links that can be used to display the changes to the individual files separately. This makes it possible to view individual changes to the documentation, for instance, or to the Kconfig files that are responsible for the kernel configuration.

People using proprietary USB drivers in the form of kernel modules should delay the switch to 2.6.25 and later kernel versions, since such drivers can no longer be compiled with an unmodified version of Linux 2.6.25. The reason for this is a change, introduced by Greg Kroah-Hartman, that causes the kernel to export certain interfaces that are absolutely necessary for USB kernel drivers via EXPORT_SYMBOL_GPL. The long-time kernel developer and USB subsystem administrator did this to make it clear that only drivers under the GPL2 or compatible licenses are able to directly use this API and the code behind it – code that in no small part originates from Kroah-Hartman himself.

Two years ago this change was already in the main development line for a few weeks, but it was removed shortly before 2.6.16 was released. This was intended to give companies like AVM adequate time to adapt their Linux drivers to work like normal closed or open source programs in userspace and from there to access USB devices via interfaces designated for userspace applications. But AVM did not take advantage of this opportunity, claming high development costs and technical difficulties. Other firms have even developed new proprietary USB kernel drivers in the past two years, despite clear warnings in the kernel since version 2.6.16 that this change was coming.

In&Out

In and quickly back out again

The change to the USB system and an unrelated module loader patch cut off Ndiswrapper access to the USB subsystem. Following a number of long debates on the Linux-Kernel Mailing List (LKML), the second patch was for the most part withdrawn to allow the Ndiswrapper to operate smoothly with kernel versions 2.6.25, fourth pre-release and higher.

The pre-releases of 2.6.25 also briefly contained a patch that had caused controversy for years, but was nonetheless contained in many distribution kernels, which allows the kernel to access a hand-adapted DSTD (Differentiated System Description Table) from the initial RAM disk (Initrd), rather than using the vendor-programmed DSTD in the ACPI BIOS. The point of contention was not whether users should be given the opportunity to do this, but rather the technical implementation, persistently criticised by a long time kernel developer shortly after the patch was integrated because it caused the kernel to initialise in a different sequence if the Initrd contained a modified DSTD.

That fuelled a debate that petered out on its own, but without a consensus or results. Some weeks later, during the 2.6.25 development cycle, Torvalds suddenly removed the patch; the code was just not fully developed and would do "ugly things". But he does not seem to be leaning toward a revised and smoothly running variant of the patch. Until there is such a patch and it makes its way into the kernel, distributors will have to maintain the other patch themselves despite the developers' rejecting it for poor quality.

[More: new and improved drivers - what's coming up in 2.6.25.]

« previous 1 2 3 4 5 6 7 next »