Linux 7.4 EDAC: four old drivers unmaintained

On October 19, EDAC maintainer Borislav Petkov asked Linus Torvalds to orphan four old Linux drivers (i82975x, x38, xgene and i3200), citing what he called an “AI-generated fixes frenzy.” EDAC, short for Error Detection and Correction, is the kernel subsystem that reports hardware error events such as memory faults. The request arrived in the Linux 7.4 merge window, which opened the day after Linux 7.3 shipped on October 18. The same window carries a Rust pull request from Miguel Ojeda that touches 248 files.

Key points

The EDAC pull request touches 14 files, adding 567 lines and removing 24. In the request, the maintainer change is in four commits titled “MAINTAINERS: Orphan drivers/edac/…”. The driver code stays in the tree until a later release removes it.
The Rust pull request adds 17,297 lines and removes 3,925. Most of that is an update of the vendored zerocopy crate to v0.8.62, released October 8, which moves eight releases forward at once.
As of October 20 (UTC), neither pull request is in torvalds/linux master. The file rust/kernel/mem.rs does not exist there, and i82975x is still listed as Maintained.

Background

Linux 7.3 shipped on October 18, and the 7.4 merge window opened the next day (our 7.3 report). The EDAC request came in that window.

Pruning old drivers is not new this year. On April 21, Phoronix reported that Andrew Lunn had posted a patch series removing a batch of ISA and PCMCIA-era Ethernet drivers, about 27.6k lines in total. His reason: “Now there are more newbies using AI and fuzzers finding issues, resulting in more work for Maintainers.” Phoronix says EDAC is following similar logic, with one difference. It orphans the drivers first and leaves removal to a later release.

The same mailing list carries another thread. On October 13, Linus Torvalds posted a page-cache patch series written with Claude Code, and a reviewer soon found a race condition in it (our report on Linus’s kernel patch). The same week, Phoronix reported a patch to remove the “Assisted-by” tag requirement for AI tools, and said developers reportedly agreed at the Maintainers Summit in Prague, in early October, to drop it. Google faces similar pressure: it paused product-vulnerability submissions to its open-source bounty program from October 1, saying the vast majority of automated reports were not valid (our report on Google’s freeze).

The facts

The table covers both pull requests. Diffstats come from the mail bodies. Mainline status was checked against the torvalds/linux repository on GitHub on October 20.

	Rust pull request	EDAC pull request
Sent	October 18	October 19
Sender	Miguel Ojeda	Borislav Petkov
Diffstat	248 files, +17,297 / −3,925	14 files, +567 / −24
Main content	zerocopy to v0.8.62; new mem and ratelimit modules; KUnit doctest generation	Four Orphan commits; a dummy EDAC driver; the intel-bff driver
Status on master (October 20)	Not merged; `rust/kernel/mem.rs` is absent	Not merged; i82975x still Maintained

The “17k” in Phoronix’s headline matches the added-lines count. The same pull request deletes 3,925, so net growth is about 13,372 lines. A Chinese report headlines it as about 17,000 lines of Rust kernel code and leaves out the deletions.

Phoronix maps the four drivers to hardware. i82975x covers the Intel 975X Express chipset, x38 covers the Intel X38 MCH, and i3200 covers the Intel 3200 and 3210 chipsets. xgene covers AppliedMicro’s X-Gene SoC; the X-Gene 3 later went into Ampere’s eMAG.

What others say

Petkov’s evidence sits in the request itself: “At least there are no traces of usage on the mailing lists.” That is silence on a list, not usage data. The request gives no count of AI-generated patches, and “AI-generated” is the maintainer’s label, which we could not check patch by patch. The same pull also calls its other fixes “the now-usual AI-powered fixes.”

Phoronix’s Michael Larabel writes that the drivers are “likely not seeing any real-world use,” and he places the EDAC move beside the networking one.

Coverage is thin. The Chinese IT之家 report from October 20 covers only the Rust pull request, and it credits Phoronix. We found no EDAC coverage outside Phoronix. A Hacker News search for EDAC from October 17 onward returned 29 fuzzy matches, and none had a title or link about EDAC. The one thread on the Phoronix Rust story had five points and no comments.

The Assisted-by debate is a separate question. It asks whether contributors should disclose AI help, which is a labeling issue. The EDAC request is about volume and testability: fixes keep arriving for hardware nobody can test. It does not argue that AI-written code is worse.

Our take

If you run one of these chipsets on a current kernel, nothing changes on day one. The drivers stay in the tree, unchanged. What changes is that fixes for them will have no maintainer to review them, and removal comes later with no date. If your hardware depends on one of these platforms, count that in your upgrade plan.

For kernel developers, the useful unit is testing capacity, not line count. Ojeda says most of the Rust diff is a mechanical zerocopy import, so 17,297 added lines say little about review effort. In the EDAC pull, the real signal is the four MAINTAINERS commits, not the 567 added lines.

The gap is in the numbers. We found no count of AI-generated patches in EDAC and no data on how many machines still run these four drivers. “Nobody uses them” is an inference from mailing-list silence, and it could be wrong.

What to watch

Whether Linus merges either pull request during the merge window. Neither is in master yet, and the usual rhythm puts the rc1 release at the close of the window.
When the four drivers are removed. Phoronix says that comes in a later release and gives no date.
Whether the Assisted-by change goes through. Phoronix reported on October 13 that it was still under discussion, and we did not check its progress after that.