Emacs TRAMP CVE, FFmpeg MVC, Rust DRM

Desktop and graphics work centered on an Emacs TRAMP security fix, H.264 multiview decoding in FFmpeg, and several Rust and DRM driver series. Reviewers also pressed Qualcomm on FastRPC duplication and debated AMDGPU SVM locking.

Emacs TRAMP user-field command injection

Emacs developers discussed CVE-2026-79992, a local shell command injection through the user field in TRAMP. The thread covered a fix commit and backport status. Remote-editing users should treat the update as a priority to close the injection path.

FFmpeg H.264 MVC multiview decoding

An 18-patch series on ffmpeg-devel implements H.264 MVC (Annex H) multiview decoding in h264dec. The work matches the existing MV-HEVC support path. Multiview and stereo pipelines gain a native decoder option inside FFmpeg.

Rust DRM driver for DisplayLink DL3

A v3 series for drm/vino, a Rust driver for DisplayLink DL3 docks, was posted to dri-devel. It is intended to replace an out-of-tree proprietary module and daemon. DisplayLink dock users get a clearer route to an in-tree open driver.

Rust KMS abstractions for display drivers

A v3 series on dri-devel adds KMS abstractions meant to support safe Rust display drivers. The patches supply shared infrastructure for DRM KMS work in Rust. Graphics developers tracking Rust in the kernel will care about the API shape.

Virtio devices that own virtqueue memory

A v2 patchset proposes VIRTIO_F_DMB so devices can own their virtqueue memory. The goal is confidential-compute and isolated vhost-user setups. Virtualization stacks that need stronger device-memory isolation get an explicit virtio feature for it.

flymake-harper and Flymake quick-fix API

A GNU ELPA proposal for flymake-harper led to extended maintainer discussion of the Flymake quick-fix API. Package inclusion questions mixed with broader design debate. Emacs Flymake users and package authors may see API clarification follow from the thread.

AMDGPU SVM phase 1 on DRM GPUSVM

Version 9 of the AMDGPU SVM series, phase 1 single GPU with XNACK on and built on the shared DRM GPUSVM core, drew review on dri-devel. Feedback included a NAK on a separate svm_lock versus eviction_lock. GPU compute watchers tracking shared virtual memory should note the locking objection.

Qualcomm DSP accel driver pushed back

Qualcomm posted a v2 accel/qda DRM driver for DSP FastRPC. Reviewers rejected duplication of the existing fastrpc stack and insisted on a single-driver approach. The exchange steers DSP acceleration toward consolidation instead of a parallel accel driver.