Thought Leadership
After 6.19 releases the version count had run its course. The new kernel brings more mature Rust integration, modern security signatures, and targeted improvements across filesystems, networking, and virtualization.
Anyone expecting a revolution from “7.0” will be disappointed. Linus Torvalds simply bumped the version number because the minor version had grown too large. The Linux kernel, which serves as the central link between hardware and software at the heart of every Linux system, continues to evolve incrementally as always.
Rust leaves experimental status
The most symbolically significant change concerns the Rust programming language. It has carried the “experimental” label in the kernel for a long time — that is now over. Rust, which unlike the long-established C prevents faulty memory accesses at compile time, can now be used on equal footing. It will not replace C, but the barrier to writing new Rust components drops considerably.
Security
The most notable security change is a switch in how kernel modules are authenticated. Linux 7.0 drops SHA-1, a checksum scheme that has been considered broken for years, replacing it with ML-DSA, a next-generation signature algorithm resistant to quantum computer attacks. Further changes include:
- For io_uring, the fast asynchronous interface for file operations, BPF-based filtering now allows administrators to precisely control which operations are permitted in restricted environments.
- BTF type lookups, internal kernel type data queries, have been sped up using binary search.
Boot process and kernel internals
The outdated linuxrc script, which was executed before the actual operating system during startup, is gone. The more modern initramfs now handles this role exclusively. New to this release is NULLFS, a deliberately empty placeholder filesystem for systems that mount their actual storage only after booting. A number of smaller but continuous improvements round things out:
- Preemption handling, the mechanism by which the kernel interrupts running processes to give others a turn, has been simplified on most platforms.
- Further improvements cover workqueues, RCU, the slab allocator, and restartable sequences.
Filesystems
Several of Linux’s most important filesystems receive notable updates in version 7.0:
- Btrfs, the feature-rich default filesystem of many modern Linux distributions, now supports direct I/O even when block sizes exceed the system page size.
- XFS, the proven high-performance filesystem common on servers, gains autonomous self-healing and loses laptop mode.
- EROFS, the lean read-only filesystem used on embedded devices, enables LZMA, DEFLATE, and Zstandard compression by default.
- F2FS, the filesystem optimized for flash storage, is modernized internally.
- Non-blocking timestamp updates now work reliably; filesystems must explicitly opt in to leases rather than receiving them by default.
Memory management
zram, a virtual memory area in RAM that buffers data in compressed form, can now write back paged-out content directly without decompressing it first. This saves CPU cycles. The swap subsystem, the area on disk the kernel falls back to when RAM is full, also receives a simplified table structure.
Networking
Several networking improvements land at once. AccECN, an improved congestion control mechanism for TCP connections, is released for general use. The first groundwork for Wi-Fi 8 (802.11bn), the upcoming wireless standard promising higher speeds in dense radio environments, is also included. Further highlights:
- The traffic shaper CAKE, which distributes outgoing traffic evenly across available bandwidth, now supports multiple CPU cores simultaneously.
- VSOCK sockets in virtual machines, through which guest and host systems communicate, gain network namespace support.
NFS and kernel filesystems
NFS, the widely used protocol for network-based file sharing, gets a dynamically adjustable thread pool and switches to version 4.1 as the new default. Experimental support for POSIX access control lists, which allow fine-grained control over who can access which shared files, is added as well. The special kernel filesystems pidfs and nsfs are no longer exportable over the network.
Virtualization
KVM, the kernel’s built-in virtualization layer that underpins countless cloud infrastructures worldwide, receives several targeted improvements in this release:
- CPU data for guests on LoongArch hardware is now reported correctly.
- AMD ERAPS is now supported.
- Virtual machines can be granted full access to the processor’s performance monitoring unit, making profiling tools significantly more precise.
- Hyper-V, Microsoft’s own virtualization layer, gains a new diagnostics interface for hypervisor statistics.
