Kernel devs review two LLM-assisted memory management patch sets
Two LLM-assisted Linux kernel patch sets—1GB huge page allocation and VM memory tracking—reveal how maintainers judge AI-assisted code.
The Linux memory-management community is currently evaluating two substantial patch series developed with LLM assistance, both submitted by established, well-respected kernel developers. Rik van Riel's 40-part series aims to make reliable allocation of 1GB (PUD-level) huge pages possible without relying on hugetlbfs's rigid boot-time reservations, introducing a new 'super page block' abstraction that segregates movable and unmovable allocations at the 1GB scale.
The series, credited with Claude Opus assistance, drew sharp criticism over code organization and quality despite broad support for its underlying design goal — Lorenzo Stoakes called it 'completely unmergeable.' Van Riel responded that he had hoped for design-level feedback before a proper reimplementation, but the sheer volume of LLM-generated code made that harder for reviewers to parse.
A second, separately submitted series by Kiryl Shutsemau adds a new userfaultfd() registration mode to expose virtual-machine guest memory usage to the hypervisor, helping host-level reclaim decisions. Together, the two efforts illustrate how the kernel community is beginning to navigate LLM-assisted contributions, judging them on code quality and process rather than tooling alone.