Linux 7.3

Original Article ↗ Hacker News Discussion ↗

Linux 7.3 is described as a record-sized kernel release that introduces features like the failfs pseudo-filesystem for stronger sandboxing, major improvements to the in-kernel SMB server including Apple Time Machine support, and initial Wi‑Fi HaLow (900 MHz) driver support. Commenters debate the value and risks of an in-kernel SMB server versus user-space Samba, the pace and quality of reverse‑engineered support for Apple Silicon, and how well Linux serves as a “fire‑and‑forget” OS for non-technical users. A prominent thread examines the rising use of AI-assisted coding in kernel development, noting both the project’s evolving attribution policies and concerns about “slop” code versus maintainers’ insistence that human submitters remain fully responsible for patch quality.

Failfs and Filesystem Sandboxing

  • Failfs is discussed as a pseudo-filesystem where all path lookups fail.
  • Intended use: hardening/sandboxing “purely network” or data-processing daemons that only use pre-opened file descriptors.
  • Strength: a process chrooted into failfs cannot open new files, limiting post-exploit movement.
  • Limitations: dynamically linked binaries generally break, since they need filesystem lookups for libraries.
  • Some compare it to more fine-grained jail/chroot controls; others argue this complexity exists because Linux never adopted strong capability systems like Capsicum.
  • Questions arise about whether seccomp rules could achieve similar effects.

In-Kernel SMB Server and Time Machine

  • Many are surprised Linux has an in-kernel SMB server (ksmbd), added for performance and because Samba is GPLv3.
  • 7.3 adds opt‑in Apple Time Machine support to ksmbd via Apple-specific SMB extensions.
  • Mixed experiences with Time Machine over SMB: some report years of reliable NAS backups; others say it is fragile or slow.
  • Apple now only supports SMB (not AFP) for Time Machine. Official docs emphasize locally attached drives; NAS support is acknowledged but vendor-specific.
  • Apple’s own SMB implementation (SMBX) is seen as buggy, and SMB interoperability with macOS (and Linux) is widely described as painful.
  • Some question whether an in‑kernel SMB server is worth the security risk versus userspace Samba, especially since user-space daemons can saturate most real-world links; others counter that at very high speeds or with RDMA, kernel implementations still matter.

Apple Ecosystem, Reverse Engineering, and Linux on Apple Silicon

  • Linux 7.3 adds device-tree support for more Apple M‑series SoCs; progress is described as steady but not fast due to lack of vendor docs and missing GPU/Thunderbolt support.
  • Discussion contrasts Apple’s self-focused hardware/software stack with Intel/AMD’s Linux-first practices.
  • Several people describe the difficulty of using Apple hardware with non-Apple services (storage, photos, backups) and advise either “go all‑Apple” or move to Linux/Windows.
  • Reverse engineering is seen as easier with modern tools and AI, but others stress that tooling doesn’t remove the need for clear goals and ongoing work to keep hacks working across updates.

AI-Assisted Kernel Development and Process

  • 7.3 has record commit and contributor counts; some suspect a rise in AI-assisted patches.
  • Kernel policy as described: contributors may use AI, but are personally responsible for code quality; previously required “Assisted-by” tags for AI, but maintainers have now voted to drop mandatory disclosure.
  • Supporters say maintainers should judge only technical quality and interaction, not tools used.
  • Critics worry about “slopspam” from low-quality AI patches and about losing provenance signals; they argue that AI usage should always be disclosed.
  • Several maintainers’ perspectives emphasize that bad contributors (AI-assisted or not) simply lose trust and attention.

Bugs, CVEs, and LTS Selection

  • A linked graph shows the number of kernel CVEs fixed reaching an all‑time high.
  • Some want a complementary graph of when CVEs were introduced per kernel version.
  • Longterm (LTS) kernels are not tied to major version numbers; releases are numbered sequentially and LTS choices are informal, often falling on the last release of the year, though this is not guaranteed.

New Features and Other Notes

  • First in‑tree driver for Wi‑Fi HaLow (900 MHz Wi‑Fi) lands in 7.3, enabling low‑power, long‑range Wi‑Fi; users are excited about IoT and niche connectivity use cases.
  • Discussion notes upcoming hostapd/wpa_supplicant and iw support being worked on.
  • Some mention 7.3’s changes enabling relocatable Nix packages and hope this will improve mixing Nix and non‑Nix software.
  • A few users share practical issues (e.g., input event devices missing in an RC) and expectations of an easy “download and install” path for new kernels, which currently requires distro-specific steps.