Linux Kernel 7.0: What Changed in the New Major Version

Linus Torvalds announced Linux 7.0 on April 12, 2026. A jump to a new major number sounds dramatic, but Torvalds has long said that version bumps are not tied to milestones. He tends to move to a new number when the minor numbers start to get unwieldy. The release was described as a solid step forward rather than a feature-driven overhaul, though the list of changes is still long. It is the kernel in Ubuntu 26.04 LTS, and reports say Fedora 44 uses it as well.

Why the number changed

Kernel version numbers do not follow semantic versioning. Going from 6.19 to 7.0 does not signal a break in compatibility. Your applications, drivers and containers see the same system call interface as before, because the kernel maintains a strong promise not to break userspace programs. That promise is one reason the kernel can evolve for decades without forcing everyone to rewrite their software.

Filesystem improvements

• Btrfs gained support for direct I/O when the block size is larger than the page size, and initial support for the remap-tree feature.
• XFS received autonomous self-healing support, allowing it to repair certain problems without taking the filesystem offline.
• A new API lets filesystems report metadata corruption and file I/O errors to userspace in a standard way.
• A new immutable root filesystem called nullfs was added.

Better error reporting matters for monitoring tools, which can finally react to damage in a consistent fashion instead of parsing filesystem-specific messages. Self-healing XFS is aimed at servers where downtime is expensive.

Security updates

SELinux gained support for access control on BPF tokens, which govern how programs can use the kernel's powerful eBPF facility. The kernel can also verify ML-DSA signatures, a post-quantum digital signature scheme. Post-quantum cryptography is designed to remain secure even against future quantum computers, and building verification into the kernel prepares for signed modules and firmware that stay trustworthy over time.

Rust is no longer experimental

Release coverage notes that Rust support is no longer considered experimental, reflecting the decision kernel maintainers made at the end of 2025. We discuss what that decision does and does not mean in a separate article on Rust in the Linux kernel.

New hardware

As with every release, a large portion of the work is enabling new processors and devices. Linux 7.0 includes more groundwork for Intel's Nova Lake and Diamond Rapids chips and for AMD's Zen 6 architecture. On the RISC-V side it adds support for the Zicfiss and Zicfilp extensions, which provide control-flow protections, and LoongArch gained 128-bit atomic compare-and-exchange.

Performance

Reported highlights include notable gains for PostgreSQL on AMD EPYC servers, better sequential read performance for exFAT, and a fix so that non-blocking timestamp updates work correctly. Individual results depend heavily on hardware and workload, so benchmark your own use cases before expecting identical gains.

AI tools and kernel development

Torvalds attributed part of the busy development cycle to AI-assisted tools finding corner cases in the code, and said he expects that to continue for a while. Whatever your view of those tools, it shows how upstream development is adapting. The merge window for Linux 7.1 opened immediately after the release.

Getting it

You will most likely get Linux 7.0 through your distribution: Ubuntu 26.04 LTS ships it by default, and rolling distributions pick up new kernels quickly. Check your current version with uname, and prefer your distribution's packaged kernels over compiling your own unless you have a specific reason. Distribution kernels include the patches, signing and testing that keep your system bootable and secure.

uname -r

Common questions

Do you need to upgrade your kernel manually? Usually not. Your distribution delivers kernel updates through its normal updater, and you reboot to apply them. Will Linux 7.0 make my computer faster? Gains depend on your hardware and workload, and most desktop users will not notice a dramatic change, whereas servers with specific filesystems or new processors may see more. Is it safe to run the very newest kernel? Rolling distributions do so routinely, but long-term-support systems stay on older, heavily tested kernels by design. How do I go back if something breaks? Keep the previous kernel installed and pick it from the boot menu, which every mainstream distribution supports.