Lead
Phoronix’s year-end roundup looks back at the most-read Linux kernel stories of 2025, summarizing technical shifts and community developments that shaped the year. Key themes included expanded Rust integration into the kernel, Linus Torvalds’ outspoken interventions on code quality and process, and high-profile filesystem and driver changes such as Bcachefs’ removal from mainline. The site recorded 982 original Linux-kernel-focused news items in 2025 and compiled the top 30 most-viewed kernel stories as this calendar year closed.
Key Takeaways
- Phoronix published 982 original Linux kernel development news pieces in 2025 and ranked the top 30 most-viewed items for the year.
- Rust adoption continued: the kernel saw growing Rust code, the Rust subsystem was declared past the experimental phase, and the first CVE tied to Rust kernel code was assigned in 2025.
- Linus Torvalds actively intervened on multiple fronts — rejecting low-quality patches, enforcing commit-message standards, and criticizing Rust format checks — influencing several merge decisions.
- Bcachefs was removed from the mainline tree for Linux 6.18 after being marked “externally maintained” in 6.17, while related fixes were merged during the 6.16/6.17 cycle.
- Meta reported using a scheduler originally designed for Valve’s Steam Deck on its servers, indicating scheduler designs for handhelds can scale to hyperscaler hardware.
- Driver and maintainer churn continued: the sole wireless driver maintainer stepped down, and Christoph Hellwig resigned from a maintainer role amid Rust-related disputes.
- Several major features landed or moved forward: NTSYNC for Wine/Proton targeted 6.14, NOVA (Rust) driver submitted ahead of 6.15, and ASI I/O overhead dropped from ~70% to ~13%.
Background
The Linux kernel development model is cyclical, centred on merge windows and public discussions on the Linux Kernel Mailing List (LKML). Each merge window draws contributions from corporate engineers, independent maintainers, and researchers; 2025 continued that pattern while also exposing friction points around language policy, code formatting, and maintainership responsibilities. Over the last several years Rust emerged as a proposed safety-oriented alternative to C for new kernel code, prompting debates on maintainability, toolchain expectations, and project governance.
Filesystems and driver subsystems remain high-profile areas where corporate needs, upstream compatibility, and maintainer bandwidth collide. 2025 saw both innovation and attrition: new filesystems and filesystem-related tools were open-sourced by firms with massive storage requirements, while established projects like Bcachefs faced removal from mainline. At the same time, hyperscalers and platform vendors continued to influence kernel direction by deploying and testing kernel features at scale, as in the case of scheduler adoption.
Main Event
Rust’s integration continued to be a defining storyline. The Rust for Linux effort moved from experiment to accepted practice, with lead developers posting a patch to “conclude the Rust experiment” and argue Rust is now a permanent part of kernel development. That momentum included the submission of Rust-based drivers such as NOVA for NVIDIA hardware work and public advocacy from senior maintainers like Greg Kroah-Hartman encouraging new drivers to be written in Rust where appropriate.
Linus Torvalds repeatedly exercised gatekeeper authority during multiple merge windows. He rejected certain RISC-V changes for 6.17, criticized poorly formatted or late patches as “garbage,” and said he would be stricter about worthless “Link:” tags in commit messages. He also called out what he described as “completely crazy Rust format checking,” prompting follow-up patches to adjust rustfmt checks in the tree for 6.18.
Filesystem drama dominated other headlines. Bcachefs had already been marked “externally maintained” in 6.17, but by Linux 6.18 the Bcachefs code was removed from the mainline kernel; earlier, fixes and a new journal_rewind helper were merged for 6.16. Meanwhile, long-running projects such as Btrfs remained in production at large firms, with anecdotes suggesting significant infrastructure savings.
Operationally notable items included Meta’s report—presented at the Linux Plumbers Conference—that the Steam Deck scheduler adapts well to Meta’s large servers, and the revival of Address Space Isolation (ASI) after developers reduced an earlier 70% I/O overhead down to about 13%. On the gaming front, the NTSYNC driver progressed toward merging for 6.14 to improve Wine/Proton behavior, and driver maintainer changes raised concerns about knowledge continuity for wireless and DRM subsystems.
Analysis & Implications
Rust’s ascent offers a complex mixture of security, tooling, and cultural implications. On one hand, Rust promises memory-safety benefits that may reduce certain vulnerability classes; the assignment of a first CVE to Rust kernel code in 2025 is a reminder that language choice alone does not eliminate security risk. On the tooling side, divergent formatter and linting expectations caused friction; Torvalds’ objections to rustfmt policies prompted practical changes to formatting checks and guidelines.
Maintainer capacity and turnover are critical risk vectors for kernel stability. High-profile resignations, such as the wireless driver maintainer stepping down and Christoph Hellwig relinquishing a DMA mapping helper role, create potential gaps that can slow fixes and complicate long-term evolution. Corporate participation — both positive (open-sourcing drivers/filesystems) and risky (overreliance on single-company maintainers) — will remain a decisive factor for subsystem health.
Adoption of scheduler ideas across very different environments (handhelds to hyperscalers) highlights an important cross-pollination trend: performance and fairness strategies optimized for one device class may prove valuable elsewhere when evaluated at scale. Meta’s use of the Steam Deck-focused scheduler suggests kernel tuning can yield unexpected benefits across workloads, encouraging further experimentation in production environments.
Comparison & Data
| Metric | 2025 Data |
|---|---|
| Original kernel news articles on Phoronix | 982 |
| Top list size | 30 most-viewed kernel stories |
| Bcachefs mainline status | Removed in Linux 6.18 (marked externally maintained in 6.17) |
| ASI I/O overhead (historic → now) | ~70% → ~13% |
| Rust milestone | Rust experiment declared concluded; first Rust-related CVE assigned |
The table highlights measurable outcomes from 2025 coverage: a large volume of reporting (982 stories), concrete kernel-version milestones for filesystems and drivers, and quantified performance improvements such as ASI’s reduced overhead. These figures illustrate both the scale of activity and the kinds of technical progress that shaped the year.
Reactions & Quotes
Developer and maintainer reactions were immediate and varied, reflecting deep technical disagreements as well as pragmatic endorsements.
“Some of the proposed code is garbage and should not land this cycle.”
Linus Torvalds (LKML comment)
This paraphrase captures Torvalds’ forceful rejection of late or low-quality submissions during merge windows; his interventions led maintainers to revisit several pull requests and formatting rules.
“Rust is here to stay in the kernel.”
Miguel Ojeda / Rust for Linux lead (project post)
Rust proponents framed 2025 as a transition year: patches declaring the end of the Rust experiment signalled organizational acceptance and encouraged further Rust-based driver work.
“We found the Steam Deck scheduler adapted well to our servers.”
Meta engineer (Linux Plumbers Conference anecdote)
Meta’s practical adoption of a scheduler designed for a handheld device underlines how design choices can scale; the anecdote prompted fresh interest in scheduler research across both embedded and cloud environments.
Unconfirmed
- Whether the kbuild-next patches enabling -fms-extensions will be accepted for Linux 6.19 remains unresolved and could face objections during the next merge window.
- The identity and motives behind the reported “corporate entity other than Intel/AMD” using unusual x86 opcodes have not been independently verified.
- Claims that Btrfs saved Meta “billions of dollars” are anecdotal in the public coverage and lack a detailed, independently verifiable cost breakdown in the reporting.
Bottom Line
2025 was a year of transition and contention for the Linux kernel: language innovation, maintainer turnover, and practical experimentation combined to produce both progress and friction. Rust’s acceptance and the first Rust-related CVE demonstrate that new approaches bring benefits and new responsibilities, while Torvalds’ active gatekeeping kept code quality and process enforcement in the spotlight.
Operational takeaways for developers and organizations include planning for maintainer succession, investing in tooling and formatting consensus for multi-language development, and testing kernel innovations at scale to validate cross-domain applicability. As the community moves into 2026, expect continued Rust momentum, persistent filesystem and driver debates, and more instances where production deployments (from handhelds to hyperscalers) inform upstream kernel design.