Desktop and graphics: procfs RFC, Emacs LLM bans, DRM cleanups

Kernel and desktop lists focused on a procfs overhaul, DRM subsystem cleanups, and contested Emacs rules on LLM-generated code. Patch series and policy threads set the pace for the day.

Procfs refactor modeled on kernfs

Christian Brauner posted an RFC series to restructure procfs along the lines of kernfs, targeting page-fault stalls that occur when entries are removed. Linus Torvalds replied with an alternative approach. The change would reshape how process information is exposed and matters for latency-sensitive kernel paths.

Emacs debate on banning LLM-generated code

Emacs maintainers and signatories argued over a permanent ban on LLM-generated contributions, citing ethics and community impact. The exchange on emacs-devel turned heated. Contributors and downstream packagers face a potential shift in what code the project will accept.

NonGNU ELPA clutch submission and LLM policy

An inquiry about adding clutch, an interactive database client, to NonGNU ELPA triggered disagreement over LLM-assisted packages. Richard Stallman opposed inclusion while Stefan and Eli pointed to NonGNU exemptions. The thread underscores unresolved governance around AI-assisted Emacs packages.

Explicit kobj release and __must_init

A nine-patch series introduces the __must_init attribute and requires every kobj_type initializer to supply a release method. The work aims to close latent use-after-free bugs across multiple subsystems. Kernel developers who manage object lifetimes should expect stricter initialization rules.

DRM panels gain atomic state access

A 35-patch series lets DRM panels use atomic state by routing through an embedded drm_bridge, remaining opt-in for drivers. The design modernizes panel handling inside the graphics stack. Driver authors can adopt the path to stay aligned with atomic modesetting.

Single DRM tree for linux-next

DRM maintainers proposed replacing the various per-subsystem linux-next trees with one unified drm-for-linux-next branch. The goal is fewer merge conflicts when material lands in linux-next. Contributors would target a single tip instead of several parallel trees.

Blender OpenPBR shader feedback

The Blender team shipped an OpenPBR shader node build for Cycles and EEVEE and asked for comments on SSS radius and thin-film units. Artists and renderer developers can exercise the new material model. Feedback on those parameters will guide final integration.

ekp package and further LLM discussion

A NonGNU ELPA submission for ekp, a Knuth-Plass line-breaking package, again raised LLM-generated code policy and naming questions, with RMS taking part. The recurring debate shows the project still lacks settled rules for AI-assisted packages. Emacs package authors should watch for emerging consensus.