Barely two months after Linux 7.0, Linus Torvalds is about to release Linux 7.1, with the stable version expected in mid-June 2026. On the menu: a new NTFS implementation with full write support after four years of development, security hardening, a major cleanup of legacy hardware, and the end of an audio bug that had been haunting the Steam Deck OLED for nearly two years. But the 7.1 cycle will be remembered above all for a deeper debate: the arrival of AI tools in kernel development, and the friction it is starting to cause.
Here is an overview of what actually changes for system administrators, hybrid-storage users, and Linux gamers.
Schedule: a mid-June release
The pace is brisk. Linux 7.0 was released on 12 April 2026; the 7.1 RC1 followed on 26 April, closing a merge window that piled up roughly 13,000 non-merge commits. RC6 dropped at the end of May, and the stable release is expected in the second half of June: 14 June if there are only seven release candidates, 21 June if an eighth is needed. An amusing detail from the cycle: a single sync of AMD GPU register headers accounted on its own for nearly 25% of the RC1 patch.
The star: a new NTFS implementation
This is without question the most significant change in the 7.1 series. After four years of work, the kernel ships a new NTFS filesystem implementation, this time with full write support. The technical improvements are serious:
- Delayed allocation for better write performance.
- iomap and folio integration, bringing NTFS in line with the kernel's modern I/O and memory management mechanisms.
- Improved stability compared to the legacy implementations.
- A new userspace utility suite, ntfsprogs-plus.
For anyone juggling Windows and Linux — dual-boot, shared external drives, hybrid workstations — this is a tangible step forward: reading and writing NTFS reliably and efficiently from Linux had long been a chronic pain point. The kernel clears a hurdle that had been awaited for a long time.
Security and power: Landlock, pstate, exFAT
On the security front, Linux 7.1 adds a new Landlock right for UNIX domain sockets by path, thanks to a new LSM hook. Landlock lets an application restrict its own privileges in an unprivileged way — a lightweight sandboxing mechanism that is invaluable for confining services. This extra granularity over UNIX sockets fits into a defense-in-depth approach consistent with the best practices I detailed on securing a Linux server.
Power management also moves forward, with improvements to the amd-pstate and intel_idle drivers. And the exFAT filesystem gains the ability to preallocate clusters without zeroing them out, reducing fragmentation — handy on removable media and SD cards.
Spring cleaning: goodbye to the i486
RC1 kicks off a "spring cleaning" operation by dropping support for aging hardware, starting with the i486 processor and various obsolete network devices. It is the end of an era: the 486, released in 1989, bows out of the kernel. Beyond the nostalgia, removing this code lightens maintenance and allows more modern instructions to be used by default.
Gaming hardware: the Steam Deck OLED finally fixed
Good news for gamers: RC2 fixes an issue nearly two years old that left the Steam Deck OLED without audio support in the mainline kernel. The upstream fix has finally been merged, which should simplify support for Linux handheld consoles going forward. Later RCs also added support for the ASUS ROG Raikiri II and Nova 2 Lite controllers. On top of that comes the usual flood of improvements: USB/Thunderbolt drivers, EXT4 and F2FS, AMDGPU and i915 GPUs, sound and networking.
The real story of the cycle: AI in the kernel
Beyond the features, the 7.1 cycle crystallizes a trend that goes well beyond Linux. Torvalds noted that the latest cycles — both 7.0 and 7.1 — show abnormally high patch volumes, and he suspects AI tools are contributing to it. The networking subsystem in particular has seen its pull requests swell under the effect of agent-assisted contributions.
It is not all rosy, though. Torvalds voiced a specific concern: AI tools, "broadly excellent," are flooding the kernel security mailing list with duplicate reports, creating needless work for maintainers. That is the flip side of automation: when anyone can generate a plausible vulnerability report in a few seconds, the cost shifts onto the humans tasked with separating signal from noise. The topic ties directly into the governance questions raised by the rise of autonomous AI agents: raw productivity is only worth something if it does not suffocate the validation processes.
For the largest collaborative open-source project in the world, this is a real-world test: how do you absorb the speed of AI without diluting the rigor that built the kernel's reputation? Linux 7.1 does not provide the answer, but it clearly raises the question.
Should you upgrade?
As always, it depends on your profile:
- Production servers: do not rush onto a
.0. Wait for a few stabilization releases (7.1.3+), or stay on the 6.18 LTS branch, supported until the end of 2028. - Hybrid Windows/Linux workstations: the new writable NTFS is a serious reason to test 7.1, ideally on a rolling-release distribution.
- Gamers and handheld owners: the Steam Deck OLED audio fix and support for new controllers are worth a look.
- The curious and the enthusiasts: cutting-edge distributions such as CachyOS will pick up 7.1 quickly, if they haven't already.
FAQ
Is the new NTFS implementation in Linux 7.1 stable enough for daily use?
It is noticeably more mature than the legacy implementations, with full write support, delayed allocation, and iomap/folio integration. That said, this is its first appearance in a stable release: for critical data, make backups and let a few stabilization releases go by before entrusting a working drive to it. For dual-boot or external drives shared with Windows, it is already a tangible improvement — test on non-vital data first.
When exactly will Linux 7.1 be released?
The stable version is expected in mid-June 2026: 14 June if the series stops at seven release candidates, 21 June if an eighth is needed. The schedule follows Torvalds's usual rhythm — one RC per week after the merge window. After the stable release, expect several point releases (7.1.1, 7.1.2…) before a smooth production rollout.
Why remove support for the i486 processor?
The i486 dates back to 1989 and no longer has any real production use. Maintaining its code imposes constraints (avoiding certain modern instructions, keeping compatibility paths) that weigh down the kernel for zero benefit. Removing it simplifies maintenance and allows newer instructions to be used by default. It is a classic housekeeping operation: the kernel regularly sheds hardware that nobody uses anymore.
Do AI contributions threaten the quality of the Linux kernel?
Not the quality of the accepted code directly — every patch still goes through maintainer review. The problem Torvalds raised lies elsewhere: volume. AI tools generate a lot of patches and, above all, flood the security list with duplicate reports, which overloads the humans responsible for triage. The risk, then, is not bad code slipping through, but a human bottleneck. It is a matter of process and governance that the project will have to address as these tools become widespread.
Conclusion: a solid cycle and a warning sign
Linux 7.1 is a rich and useful release: the new writable NTFS fills a long-standing gap, Landlock refines sandboxing, exFAT and power management improve, and gaming hardware gains better support. Nothing revolutionary, but the accumulation of refinements that makes the kernel's development model so strong.
The real lesson of the cycle lies elsewhere. By publicly pointing at the overload caused by AI reports, Torvalds puts his finger on a tension that all of open source will have to manage: AI multiplies the ability to produce code and reports, but the human ability to validate them does not scale as fast. How do you preserve rigor without throttling innovation? The Linux kernel, by its scale and its standards, is the ideal place to invent that answer — and 7.1 is probably only the first cycle where the question is posed this starkly.
Comments