Tag: Hardware Support

  • Linux 6.18 LTS: Rust-Based Android Binder Lands as bcachefs is Removed from Mainline

    Linux 6.18 is the final major release of 2025, and it is highly likely to become the next long-term support (LTS) version. This means that, in the coming years, 6.18 will serve as the foundation for distributions that prioritize stability and predictable upgrades—from servers to desktop systems.

    As usual, the release brings a wealth of new drivers and refinements across an enormous span of hardware. While few changes qualify as revolutionary, several are undeniably significant. The most widely discussed news is Linus Torvalds’ decision to remove the experimental bcachefs file system from the mainline tree. Introduced in kernel 6.7 nearly two years ago, bcachefs has now fully “moved into the wild”: it is developed externally, offers DKMS packages for Debian, Ubuntu, Fedora, and openSUSE, and is already included in Arch and NixOS. The Linux kernel itself will no longer be responsible for it.

    By contrast, the more established file systems have received notable upgrades. XFS can now be checked and repaired in place, without unmounting—a major boon for massive server arrays where offline checks can stretch across hours or days. The exFAT driver, common on microSD cards and USB sticks, has been accelerated by up to 16× in certain operations. Btrfs gains improvements in parallel execution; ext4 receives functional enhancements; FUSE has been sped up for users of unconventional file systems; and caching behavior has been reworked. A curious detail: for volumes shared over NFSv4, caching can now be completely disabled.

    Hardware support has also grown in the customary fashion. Once distributions adopt the new kernel, many devices will work “out of the box,” eliminating the need for vendor modules or custom kernels. Support for portable consoles and handheld PCs from ASUS, Lenovo, and China’s GamePad Digital has been strengthened, as has compatibility with Sony’s DualSense controller. Systems from Dell and Alienware, the HP Omen line, and numerous ASUS ROG motherboards gain new monitoring capabilities and configuration options. Keyboard and touchpad handling has been refined—small improvements that nonetheless matter in everyday use.

    Kernel developers have also continued investing in a less common yet promising direction: persistent memory (PMEM), once championed through Intel’s Optane technology. Linux 6.18 introduces the dm-pcache subsystem, capable of using PMEM as an ultra-fast cache for conventional SSDs and hard drives—potentially offering dramatic I/O acceleration wherever PMEM is available.

    Under the hood, the release delivers an array of enhancements for x86-64 processors across profiles, especially in server and networking roles. Most will remain invisible to end users but will be welcomed by administrators and performance-minded engineers.

    Graphics support continues to advance as well. The open-source Nouveau driver for Nvidia GPUs now works with GSP firmware on Turing and Ampere cards and offers improved power management. A new Rust-based driver debuts for Arm Mali graphics—preliminary for now but significant given Mali’s ubiquity in ARM SoCs. Support has been added for Rockchip’s NPU accelerator used in tensor workloads. On the ARM side, Asahi Linux continues improving compatibility with Apple’s M2 family (including Pro, Max, and Ultra), while M3 and M4 support remains in development. Parallel progress continues for RISC-V and Loongson architectures.

    After two years of work, the kernel now includes a Rust-rewritten implementation of Binder, Android’s long-standing interprocess communication subsystem. Originating in BeOS, later appearing in Palm OS Cobalt, and present in Linux since 3.19 as a C module, Binder now has a modern, safer Rust incarnation intended to enhance security and maintainability.

    There is even a noteworthy intersection with the BSD world: Linux can now detect when it is running under FreeBSD’s built-in bhyve hypervisor and operate correctly in that configuration. This arrives just as FreeBSD 15.0 introduces support for over 255 virtual CPUs per VM—capability that Linux, as a guest OS, is now prepared to handle.

    The kernel’s pidfd mechanism—a file-descriptor-based interface for managing processes—has been significantly expanded. Christian Brauner, its original architect, has added support for kernel namespaces, allowing for finer control over container-like environments through pidfds.

    On the networking side, Linux gains support for the AccECN protocol (Accurate Explicit Congestion Notification), offering more precise congestion reporting and more efficient traffic management. The eBPF subsystem, long used to inject small programs into the kernel, now includes preliminary support for cryptographically signing those programs—another step toward stronger security.

    Kernel developers also continue reimagining memory management. After the introduction of folios in recent NTFS and 6.16 updates—a new way of handling disk pages in larger blocks—Linux 6.18 adds sheaves, a similar abstraction for RAM that layers atop traditional slabs and related structures. Though deeply internal, these shifts often emerge, over time, as improved responsiveness and resource efficiency.

    The new release will soon appear in rolling distributions such as Arch and openSUSE Tumbleweed. Since new versions of Debian and RHEL shipped earlier in 2025, it is unlikely they will adopt 6.18 within their current cycles. Ubuntu, however, may well include it in its next LTS release, Ubuntu 26.04 “Resolute Raccoon,” for which Canonical has already begun publishing monthly snapshots.

  • Linux Kernel 6.16 Released: Packed with Performance Boosts, New Features, and Core Improvements

    This past weekend marked the release of the final version of the Linux 6.16 kernel, traditionally announced by Linus Torvalds himself. The development process was calm and steady, though it lacked headline-grabbing features—making the release more technical than sensational. Nevertheless, it incorporates dozens of performance enhancements, support for new instruction sets, and foundational improvements in memory management.

    According to Phoronix, Linux 6.16 now comprises over 38 million lines of code spread across more than 78,000 files. Despite the absence of so-called “blockbusters,” the scope of work is extensive—from low-level optimizations to strengthened security mechanisms.

    One of the most notable additions is support for Intel’s Advanced Performance Extensions, introduced in 2023. These extensions enhance vector operations and double the number of general-purpose registers. However, they are not universally supported across all Intel processors, underscoring the company’s persistent fragmentation issues.

    File systems have also seen significant performance gains. XFS now supports advanced atomic writes, while ext4 benefits from bigalloc and large folio support—resulting in performance boosts of up to 30% in certain use cases. Improvements were also made to Btrfs and NFS.

    The core dump mechanism has undergone important changes: memory dumps can now be transmitted via AF\_SOCKET rather than being restricted to file directory storage. This update offers greater flexibility and security in debugging processes.

    For server systems utilizing NUMA architecture, the kernel now includes automatic self-tuning capabilities—an essential feature for optimized memory node distribution. Additionally, support for five-level page tables has been introduced, paving the way for enormous virtual memory spaces.

    A major advancement for embedded systems and resource-constrained devices is the long-awaited implementation of offloaded audio decoding to USB chips—particularly relevant for Qualcomm platforms.

    A comprehensive breakdown of the changes is available in two parts on LWN: Part One, Part Two, along with a summary article. For those seeking a quicker overview, kernelnewbies.org provides a concise rundown of the key updates.

    Torvalds also noted that he will be traveling during the Linux 6.17 merge window, which could result in delays. While this might sound like a routine scheduling notice, it also serves as a reminder that the fate of the Linux kernel still hinges heavily on a single individual.