Linux 7.2, someone moved 3 lines of code and storage is now faster

Software optimization is a process that often requires a thorough review of existing architectures, with weeks of testing and endless debugging sessions to achieve purely marginal improvements.

However, sometimes the most effective solutions hide in the finest details and require an extremely linear approach. It is exactly what happened during the current development cycle of the Linux kernel 7.2, where a seemingly trivial intervention produced a 5% increase in storage performance.

The result does not come from a complete rewrite of complex modules, but from the simple repositioning of 3 lines of code moved down a few lines within the script: an action sufficient to guarantee higher responsiveness of the entire system.

Linux, moved 3 lines of code and storage is faster

The discovery and subsequent implementation of this ingenious fix bear the signature of engineer Fengnan Chang, while the official insertion into the source code was handled by Christian Brauner.

The commit notes explain precisely the nature of the intervention: the modification consists in avoiding the execution of a “memset” operation of the structure “iomap” inside the function “iomap_iter()” once the iteration has completed.

Before this update, the code performed this operation, wasting precious memory bandwidth in writes, a bottleneck factor in scenarios characterized by a high number of input/output operations per second (IOPS).

In high-load scenarios, such as 4-kilobyte random reads on NVMe drives via the io_uring interface, the hardware resource savings achieved produced a net gain of 5% on the ext4 and xfs file systems.

Moving three instructions to the end of a process rather than leaving them at the start has eliminated an unnecessary workload, demonstrating how a well-targeted debug can make a tangible difference.

The release and wait times

The rollout of this improvement will still require a few months of waiting. The 7.1 kernel version made its public debut only last weekend, establishing the structural foundations.

Consequently, Chang’s technical insight will be made available to all users with the debut of release 7.2, whose arrival on servers and personal computers is currently planned for the month of August 2026.

Although it will take some patience before measuring these performance gains on our machines, the efficiency demonstrated in this pre-release phase again confirms how open-source development can reap enormous benefits from a careful and rational reading of preexisting code.