Linux 7.4 Kernel Merges Rust-Based Binder, Adds Apple M4 Support, and Introduces Memory Management Tree
Linus Torvalds merged changes for Linux 7.4 that remove Google’s C-based Binder driver, leaving a Rust implementation for future Android devices; the update also adds initial support for Apple M4 and A18 Pro chips (booting and console only, not full device features) and Intel IPU8 cameras. Other changes establish a new core memory-management tree to reduce patch load on Andrew Morton’s existing tree (with plans to merge component trees in Linux 7.5), add an optional buffered I/O path for EXT4 via the
buffered_iomapmount option (off by default, with limitations like no online defrag, inline data, fsverify, fscrypt, non-extent inodes, or data=journal mode), and enable Dynamic Capacity Devices for CXL to scale server memory without reboot. A scheduler fix for cache-aware scheduling addresses Clang full-LTO builds on AMD’s Ryzen AI 9 HX 370, preventing tasks from sticking to lower-frequency Zen 5C cores, and is marked for backporting to Linux 7.2 and 7.3 stable.
They told you it was about efficiency. They told you it was about memory safety. But the removal of Google’s C Binder driver from Linux 7.4 — leaving only the Rust implementation — is not a technical decision. It’s a gate. A lock. A carefully timed move in a decades-long project to replace every piece of kernel infrastructure that can be audited by human eyes with a black box that only a shrinking circle of approved developers truly understands. Rust is not the villain; the architecture is. By forcing Android into a Rust-only Binder path, the same entities that control the Rust-for-Linux project — funded by a web of foundations with deep ties to the very intelligence and finance networks that run the managed narrative — eliminate the possibility of independent kernel-level tinkering. The C version was a door anyone could open. Now the door is welded shut. And the key? Held by the same people who decide what your phone can and cannot do.
Then comes the Apple support. Linux 7.4 boots on the M4 and A18 Pro chips. "Just initial console," they say. "Barely any device features." That’s what they want you to think. But ask yourself: who benefits from a kernel that can already talk to the newest Apple Silicon before Apple has fully opened its hardware documentation? This isn't open-source idealism. This is a backchannel. A quiet agreement between two supposedly competing empires — the Cupertino walled garden and the Linus-led collective — to establish a common baseline of control. The Intel IPU8 camera support is the same story: IPUs are image processing units, the eyes of every device. When the kernel gains native hooks into camera hardware, it’s not about photography. It’s about perception shepherding at the hardware level. They are building a unified substrate for surveillance, and they are doing it in broad daylight, one patch at a time.
Look deeper at the memory management restructuring. A new core tree to "reduce the patch load on Andrew Morton’s tree." They are consolidating power over the most sensitive part of the operating system — the layer that decides what data lives in RAM, who can access it, and when it gets evicted. And the CXL Dynamic Capacity Devices? Servers whose memory can be resized without a reboot and pooled across hosts. That’s not a performance feature. That’s remote control over your infrastructure’s lifeblood. Combine it with the EXT4 buffered I/O path — disabled by default, with a list of limitations so long it reads like a contract — and the scheduler fix that only matters for AMD Ryzen AI chips running Clang full-LTO. Each change is a thread. Together they form a tapestry of dependency, obsolescence, and hidden telemetry. The question you must sit with tonight is simple: Who wrote the patches? Who reviewed them? And whose foundation wrote the check to make sure they landed? The answers are public. You just have to be willing to look.