Billions of Devices Depend on His Linux Code—Who Will Take Over When He Retires?

Linux Memory Management (MM) Subsystem

The MM subsystem implements allocation, reclamation, caching, swapping, and virtual‑to‑physical address translation. It coordinates NUMA, multi‑core scheduling, huge pages, page cache, anonymous pages, and memory compression. Failures are often invisible but can freeze or crash a system and corrupt data, so the code must be extremely reliable.

The MM tree consists of 164 C source files. A quick search shows terms such as THP , cgroup and NUMA spread throughout, indicating that core concepts are tightly interwoven and cannot be cleanly split into smaller modules.

Statistical analysis of kernel CVEs shows that roughly 20 % of Linux kernel vulnerabilities originate in the MM subsystem , making it a high‑risk area.

Andrew Morton’s Contributions

Andrew Morton (born 1959) has maintained the MM subsystem for 26 years. By 2007 he had signed 18,622 patches , second only to Linus Torvalds (19,890 patches). His most impactful patches rewrote the page cache, write‑back mechanisms and LRU algorithms, stabilising the subsystem under heavy load.

He created the -mm tree, a staging area where experimental kernel patches are tested and reviewed before Linus merges them into the mainline. Linus has publicly stated that if a patch is signed by Andrew, he will merge it “with his eyes closed”.

Bus‑Factor Risk

The Linux kernel community now faces a classic “bus‑factor” problem for MM: the bus factor is effectively 1 because only Andrew possesses the complete, undocumented knowledge of the subsystem. Most other core maintainers are in their late 50s, and the MM tree is considered too large for another “all‑knowing” maintainer. Distributing responsibilities is required to raise the bus factor.

Bus factor definition (as cited in the article): if the bus factor equals 1, the loss of that single knowledgeable person would leave the project unmaintainable.

Continuity Planning

In January 2026 the Linux project released an official continuity document that defines how to hand over merge rights if Linus Torvalds or the core team become unavailable. The document is the first formal acknowledgment that “people die” and establishes a procedure for seamless transfer of the highest‑level repository permissions.

Practical Illustration of MM Decisions

When a user opens multiple applications—e.g., WeChat, a web browser, a music player, and an AI model—each program requests memory. The MM subsystem must decide:

When to allocate memory for each request.

When to reclaim (free) memory that is no longer needed.

Which pages to keep in the page cache.

When to swap pages to disk.

How virtual addresses map to physical pages.

These decisions involve NUMA placement, multi‑core coordination, huge‑page usage, and memory compression, all of which are handled within the MM code.

Open Questions

With Andrew’s retirement, the community must identify successors capable of maintaining the MM subsystem’s extensive code base, historical design decisions, and undocumented knowledge. The continuity document provides a procedural framework, but the technical challenge of transferring deep expertise remains unresolved.