Sasha Levin from Google, who is involved in the maintenance of the LTS branches of the Linux kernel and is a member of the advisory board of the Linux Foundation, has announced the creation of a new Linux kernel branch — 'linus-next'. This branch, analogous to 'linux-next', is intended for testing changes proposed for inclusion in the upcoming mainline Linux kernel branch, but it focuses on changes ready for acceptance into the mainline and provides a more stable and testing-friendly base.
The previously used 'linux-next' branch is well-suited for testing integration changes in the kernel, evaluating conflicts between different maintainers' sub-branches, and identifying related build errors and warnings. However, it is problematic to use for testing the functionality of the proposed changes. The 'linux-next' branch contains a mix of changes ready to be merged into the mainline kernel and experimental add-ons, which may take a long time to refine and possibly may never be accepted into the main kernel branch.
The presence of unstable experimental code prevents the 'linux-next' kernel from being used for full-fledged functional testing, as experimental code can adversely affect other kernel subsystems. In the new 'linus-next' branch, the criteria for accepting changes will be stricter — only pull requests addressed to Linus Torvalds, which the developers assume Linus will accept into the main kernel branch, will be included.
The emergence of the 'linus-next' branch will allow not only testing the compatibility of proposed changes at the kernel build level but also evaluating their functionality directly. This will lead to an increase in the quality of testing changes at the stage of reviewing pull requests that have already been submitted but not yet reviewed by Linus. The new branch will also encourage submitting pull requests as early as possible, even before the acceptance window opens, to allow for testing in operation. Furthermore, the 'linus-next' branch will evolve according to a continuous integration model rather than daily updates with new tag creation, simplifying the analysis of emerging issues and identifying specific changes that caused failures.
Additionally, it is worth noting Linus Torvalds' dissatisfaction with the process of blocking hardware vulnerabilities at the kernel level, expressed during discussions about issues and regressions in the code aimed at countering attacks that exploit vulnerabilities in the CPU's speculative execution mechanism. The changes necessary to block hardware vulnerabilities complicate development, lead to regressions, and negatively impact performance, given that most attacks exploiting such vulnerabilities are purely theoretical and never occur in practice outside of laboratory conditions.
Considering the proposal to modify the kernel to bypass potential issues in countering hardware vulnerabilities, Linus suggested placing the responsibility on hardware manufacturers, as the culprits behind the emergence of hardware vulnerabilities, and making it clear that they should address their own problems. The proposed change was suggested as a precaution, as the uncertainty in CPU behavior could potentially render protection ineffective on some new Intel CPUs.
Source: opennet.ru
