Intel Xe adds vRAM health checks and page offlining

Queued for Linux 7.4, the Intel Xe driver adds a debug-gated early-probe canary check for the last vRAM page, intended to catch BAR/CCS sizing and identity-map errors. It also adds vRAM page offlining after 21 review rounds, permanently keeping faulty pages out of future allocations and exposing bad ranges through DebugFS. The same pull adds per-engine idle status in DebugFS.

https://www.phoronix.com/news/Intel-Xe-vRAM-Health-Check

Primary patch discussion: https://lore.kernel.org/all/20260902124117.918018-12-matthew.auld@intel.com/

Notion’s production CRDT for collaborative rich text

Notion describes the CRDT/data-model redesign it deployed in July 2025 for concurrent editing and Offline Mode: an RGA-derived sequence structure with stable session/Lamport IDs, tombstones, Peritext-style rich-text annotations, and a text-slice/tree model that preserves edits across concurrent block splits. A search-label index limits the number of blocks fetched when a text instance has been split repeatedly. Notion says the system processes millions of CRDT operations per minute.

https://www.notion.com/blog/how-notion-handles-concurrent-editing-with-crdts

Gzip 1.15 fixes decades-old correctness and safety bugs

Gzip 1.15 (September 20, 2026) includes more than 100 commits, largely bug fixes accumulated over the past 18 months. The release fixes a race that could remove the wrong file after an ancestor rename, uninitialized-memory use, an LZH decompression buffer overflow, multiple LZH corruption cases, streamed ZIP handling, temporary-file races on limited platforms, and locale handling for diagnostics. It also drops several obsolete platform targets.

https://www.phoronix.com/news/Gzip-1.15-Released

Linux 7.4 may make read-only file opens ~39% faster

A VFS patch queued for Linux 7.4 lets do_dentry_open() consume an already-held terminal dentry reference instead of taking and dropping redundant references during pathname lookup. The five-revision series reports a 39% increase in read-only open() operations in will-it-scale on a 20-core VM.

That is a focused microbenchmark, not an end-to-end application result, but the change is unusually small and targets a hot path after more than two years of iteration.

Phoronix report

Btrfs, ZFS, and bcachefs under real workloads

A new benchmark suite compares multi-device CoW filesystems across core I/O, responsiveness, metadata work, and integrity—not just synthetic throughput. Its current cohort shows bcachefs scoring strongly on the aggregate workload mix, while Btrfs and ZFS vary substantially by layout.

Important caveat: the automated runs use loop devices on shared ephemeral VMs, so the author says to compare ratios rather than absolute MB/s; real-hardware runs are still being added. The benchmark sources and run history are public.

Benchmark · HN discussion

Wild vs. Mold: why linker benchmarks disagree

David Lattimore reproduced Mold’s recent linker results and traced the difference from Wild’s earlier numbers to benchmark setup: whether the output file already exists, tmpfs versus ext4, and whether startup uses fork or --no-fork. Under matched conditions, the linkers are much closer than the headline comparison suggests.

The post also identifies concrete follow-up work: Wild needs the same filesystem-specific optimizations Mold uses, including preallocation and huge-page mapping. It is a useful example of how benchmark methodology can dominate systems-tool comparisons.

Benchmark analysis

Linux 7.3-rc4 fixes two Btrfs regressions

The upcoming kernel release restores stable Btrfs filesystem IDs after a Linux 7.2 change made f_fsid vary across upgrades and reboots. That broke userspace expecting a stable ID—including NetworkManager OpenConnect users whose VPN key handling stopped working after upgrading.

The same fix batch repairs /dev/root detection for Btrfs root filesystems on initramfs-less GRUB2 boots. Both regressions are user-visible compatibility failures, not just cleanup.

Phoronix report

What Zig felt like, coming from Rust

A Rust developer with seven years of experience reimplemented an existing Rust JSONPath library in Zig, using the project to compare the languages in a non-toy but bounded setting. The report covers Zig’s sparse IDE support, build.zig as a compact test/compliance driver, the shift back toward CLI tooling, and the friction of carrying Rust habits into Zig. It is useful less as a language verdict than as a concrete account of the first serious port.

The Hacker News thread had 171 points and 209 comments when checked. Discussion pushed back on treating one project as representative, and several commenters highlighted Zig’s current stability and ecosystem tradeoffs for long-lived or archival code. The author also explicitly labels the experience as an early, potentially naive introduction to Zig.

Why building a Rust LSP is hard

A detailed tour of why a Rust language server is much more than JSON-RPC: startup must return useful answers before indexing finishes; the server needs a synchronized virtual filesystem, cancellable parallel queries, incremental semantic state, resilient parsing for incomplete code, and workspace/project discovery that reflects user intent.

The author contrasts rust-analyzer’s Salsa-driven lazy computation with Rust Glancer’s eager, filesystem-backed indexing and per-workspace engine processes. HN discussion focused on the tradeoff between LSP’s language-agnostic decoupling and its awkward shared-file state, with several commenters arguing that synchronous in-process designs would lose crash isolation and multi-client portability.

• HN discussion

Photon-Emission-Guided Laser Fault Injection Enables RP2350 Secure Debug

Ledger Donjon used photon-emission microscopy to localize the RP2350’s DEBUGEN logic, then laser pulses to set the two bits needed to restore Secure debug despite permanent debug-disable settings. A rescue reset kept firmware from reapplying a runtime OTP lock, allowing the researchers to recover a challenge secret. The attack requires destructive backside decapsulation, specialized equipment, and physical access, but exposes a real gap between OTP policy, mutable override registers, and reset-time behavior.

HN commenters debated the cost and practicality of the lab setup, and whether the result is a lesson about trusted-hardware threat models rather than a conventional software vulnerability.

• HN discussion

Needle 3: 8–29 MB automation models for tiny devices

Cactus’s Needle 3 targets tool calls and structured JSON rather than open-ended chat: deployable 2-bit subnetworks range from 25M to 121M parameters in 8–29 MB binaries, with reported Raspberry Pi 5 decode rates up to 4,000 tokens/s. The architecture uses a Monarch/Hadamard-factorized MLP, multilingual support, confidence scores, and runs across Linux, Android, iOS, RISC-V, MIPS32, microcontrollers, and browsers.

The important boundary is task scope: HN testers found the demo brittle on ambiguous smart-home commands, while the authors emphasize that the model is for narrow, grounded automation on hardware that cannot host a multi-billion-parameter model. That makes it interesting embedded inference, not a tiny general-purpose LLM.

• HN discussion

Saving another 100TB of RAM with math (and Rust)

Cloudflare reclaimed more than 100 TB of RAM globally by reducing the memory footprint of a Pingora-based backend router, combining consistent-hashing math with a tighter Rust representation. The post explains how its routing structure grew expensive at Cloudflare scale, then walks through the algorithmic and data-layout changes that made the savings practical across thousands of servers.

HN readers liked the calculus-driven optimization but questioned the sparse crate documentation and asked how the design compares with jump consistent hash; those are useful caveats when interpreting the headline number.

• HN discussion

Apple Video Decode Driver Posted For Upstream Linux Kernel Review

Asahi’s latest Apple-Silicon enablement work posts an AVD driver for upstream review, targeting accelerated video playback on Apple M1, M2, and M3 SoCs. It is another step in moving previously platform-specific media support into the upstream Linux graphics/media stack.

The Phoronix thread has four comments; detailed reactions were not exposed by the feed, so this post sticks to the driver and upstream-review milestone.

• Phoronix source

Google’s “Painful To Maintain” Binder C Linux Driver Being Removed In Favor Of Rust

Android’s long-standing C Binder IPC driver is set to retire after its Rust rewrite was upstreamed in Linux 6.18. The change is a concrete, high-impact Rust-in-the-kernel transition: Binder is a central Android IPC path, and the old implementation is scheduled to disappear during the Linux 7.4 cycle rather than merely coexist indefinitely.

Phoronix reports a 20-comment thread; the accessible feed exposed the implementation and migration context but not detailed comment text.

• Phoronix source

OpenJev

OpenJev is a browser-only experiment that runs a small local model and compares direct choice-logit reads with token-by-token JSON generation. The demo keeps the weights in the browser, asks the same model to score a fixed set of options, and exposes the latency/quality tradeoff instead of pretending structured generation is free.

The 239-point HN thread liked the low-latency, low-cost decision-model angle but questioned whether the implementation is mainly prefill plus restricted-token scoring. Others found incorrect classifications and prompt-injection behavior in the demo, so treat it as an interesting technique preview rather than a validated Jev replacement.

• Hacker News discussion

Inside ZCode: Silently Uploading Your Entire Git History to the Cloud

Reverse engineering of Zhipu’s ZCode desktop app reports that, while logged in, it packages a workspace—including .git history, LFS data, reflogs, and configs—encrypts the archive, and uploads it to Aliyun OSS. The reconstructed flow says the server supplies an RSA public key and keeps the private key, so the local client and user cannot decrypt the resulting archive; the author’s tests also found that the visible snapshot/indexing toggles did not stop capture. A filesystem-level immutable checkpoint directory is offered as a workaround, at the cost of rollback.

The author notes the write-up was AI-translated. HN discussion challenged the terminology and the reliability of a separate AI-paraphrased summary, but the original post includes the client-side evidence, archive measurements, and reconstructed upload flow.

• HN discussion

An Empirical Study of Harness Design for Coding Agents

A controlled study treats coding-agent harnesses as composable systems: 176 matched settings vary planning, action space, and context management across four models on SWE-Bench Verified and Terminal-Bench 2.1. The strongest result is practical: staged rule-based context elision before LLM summarization gives the best efficiency, while recoverable elision adds machinery without improving accuracy. Planning mainly scaffolds weaker models but saves cost for stronger ones; bash-capable models can often use a bash-only interface more cheaply than predefined tools.

• Hacker News discussion

A quartet of Linux local-root vulnerabilities: DirtyAH6, PPPoEject, TUNderflow, and DiagSpill

A coordinated disclosure describes four long-lived Linux kernel bugs that yield unprivileged local-to-root execution on affected configurations. The bugs span IPv6 AH routing-header validation, TUN receive-headroom arithmetic, a stale sk_buff pointer in PPPoE, and an SCTP transport-count overflow; the first three commonly need user namespaces or capabilities, while DiagSpill does not. Fixes are landing in stable kernels, and the disclosure includes mitigations such as disabling unused AH6, TUN, PPPoE, or SCTP paths.

• oss-sec disclosure and discussion
• Full technical write-up

SystemTap 5.6 adds BPF LSM/XDP probes and BTF tracepoints

SystemTap 5.6 turns its BPF runtime into a more capable Linux observability and policy tool: it adds BPF LSM hooks that can allow or deny operations, XDP packet probes with pass/drop/tx verdicts, and BTF-based tracepoint discovery without trace headers or tracefs. The release also brings statement tracing, faster parallel DWARF elaboration, richer runtime error context, Linux 7.2 compatibility, and ML-DSA module-signing support. BTF tracepoint and statement tracing support still have runtime limitations under --runtime=bpf.

• Release notes and source discussion