- Linux 7.2 has been released as stable by Linus Torvalds, marking a successful and timely launch for this feature-packed kernel.
- This version is targeted for Ubuntu 26.10 and other upcoming Linux distribution releases.
- Cache-aware scheduling has been introduced, allowing the scheduler to recognize when threads belong to the same process and keep them on cores that share a cache.
- The development cycle set a record with over 2,100 contributors involved.
- During the cycle, AI-assisted commits accounted for roughly 5% of all commits.
- A fair(er) GPU scheduler was reverted to FIFO due to performance regressions on certain AMD graphics cards.
Linux 7.2 has officially launched, introducing significant enhancements including cache-aware scheduling and performance improvements for the ext4 and Btrfs filesystems. This release is set to power upcoming distributions like Ubuntu 26.10 and showcases a record of over 2,100 individual contributors during its nine-week development cycle.134
The new cache-aware scheduling allows the kernel to recognize when threads belong to the same process, optimizing CPU core usage and reducing cache bouncing. This change is expected to enhance performance by 6-28% and decrease write latency by 5-22% under memory pressure. Additionally, a locking fix for pipes improves concurrent access during writes.

Performance gains are also noted in the ext4 filesystem, where a change to the fast commit feature reduces lock contention and deadlocks, while a directory lookup modification speeds up file searches in dense folders. The Multi-Generational LRU (MGLRU) algorithm has been refined, promising up to a 30% increase in MongoDB performance with YCSB and minimizing unexpected out-of-memory kills.
Furthermore, the update enhances NTFS handling with proper Windows symbolic link support and fixes a bug that caused these links to appear as zero-byte files. Overall, Linux 7.2 represents a substantial step forward in kernel performance and hardware compatibility.
“The kernel's new cache-aware scheduler limits cache bouncing by keeping threads on shared-cache cores, with guardrails preventing overload. A last-minute revert switched the GPU scheduler back to FIFO due to regressions on some AMD cards, while AI-assisted commits made up about 5% of the cycle.”