The new kernel version is out, this time with an increased major number. As usual, despite the perceived importance of the version number, this is just one release more. That means no spectacular announcements; just consistent work, more features, more fixes, better support.
A couple of non-technical changes stand out in this release: Rust is no longer considered experimental in the kernel, and there is now an official policy on tool-generated content. The kernel keeps evolving and this will surely pave the way for future development in the coming years.
For a detailed list of changes, you can check the KernelNewbies ChangeLog.
At Igalia, the Linux kernel is one of ever-present work areas and interests and we keep contributing to it on many fronts. Here’s our list of contributions and patches for this release.
Our work on the DRM (Direct Rendering Manager) continues. Being one of our biggest areas of contribution, in this release we once again provided improvements.
In the AMD display driver, we fixed unexpected color results reported by SteamOS/Gamescope users. These were related to incorrect mapping of color values made by the AMD color module when programming some color transformations into the hardware. We also improved the support for newer hardware families by exposing missing plane color blocks.
Further, we provided many assorted patches to do various cleanups in the TTM layer and multiple small (but important!) fixes in the AMD, Intel and v3d drivers, as well as a fix for the framebuffer probing on the Valve Steam Deck.
For the past couple of years, we’ve been consistently contributing to sched_ext and, in particular, to the scx_lavd scheduler to enhance the gaming experience in Linux, improving the latency and interactivity under gaming workloads.
In this release, we introduced a set of BPF helper APIs –bpf_in_nmi(), bpf_in_hardirq(), bpf_in_serving_softirq(), and bpf_in_task() – along with their corresponding self-tests. These helpers allow BPF programs to precisely detect their execution context. This capability is particularly valuable for sched_ext schedulers; for instance, a BPF scheduler like scx_lavd can now distinguish whether a task was woken up by another process or via an interrupt, enabling more informed scheduling decisions.
In our ongoing effort to improve Raspberry Pi devices, we addressed several issues related to the RPi 4 power domain, fixing system hangs during power transitions and correcting a broken reset status read. Additionally, on the RPi 3, we resolved a couple of memory leaks, race conditions, and PM reference imbalances, improving overall reliability for RPi users.
As usual, we helped with the never-ending task of fixing bugs, this time clearing up a few of them in the Bluetooth stack and the DRM TTM.
We determined the root cause of two race conditions in the thermal subsystem’s core that could potentially cause a system crash, and developed synthetic reproducers for it. Even though the approach selected to address the issue was not our proposed fix, we contributed to its code review and testing with our specific understanding of the issue and synthetic reproducers, improving the overall quality of the fix.
Another important and often overlooked type of contribution is the refactoring of code to make it more maintainable. In this case, by replacing deprecated APIs and legacy macros. We submitted some patches to do this kind of update in parts of the DRM subsystem and the v3d driver.
As with every release, we helped with assorted janitor tasks such as documentation fixes and cleanups — specifically improving the exportfs and DRM documentation — as well as the usual share of reviews and tests.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。