Linux Plumbers Conference 2025
2025-12-11..2025-12-13
-- online --
[Linux Plumbers Conference 2022, Linux Plumbers Conference 2023, Linux Plumbers Conference 2025]
| title | author | length |
|---|---|---|
| Evolving genpd’s - sync_state() support | Saravana Kannan, Ulf Hansson | 19m |
Generic power domain sync_state support lands in v6.17-rc1 and provides a unified consumer-probe notification, but it leaves several concrete failures: domains… | ||
| Multiple system-wide low power-states | Kevin Hilman, Ulf Hansson | 25m |
Kernel now exposes a user-space latency knob and a boolean flag so suspend-to-idle can consider per-device PMQoS resume-latency votes (values are microseconds… | ||
| Generic voting mechanism to build on top PM QoS | Daniel Lezcano | 26m |
Proposal builds a generic kernel voting subsystem that aggregates kernel and userspace ballots, exposes a planned /dev user interface, and auto-cleans votes… | ||
| Updating Energy Model from Thermal | Lukasz Luba | 20m |
The Energy Model can be updated at runtime and the Thermal framework already provides SoC temperatures, so the proposal is to adjust static power (leakage)… | ||
| Thermal framework issues and limitations | Daniel Lezcano | 19m |
Kernel change in 6.10 moves mitigation decision out of the step_wise governor into the thermal core, so crossing a cooling trip stop immediately, which causes… | ||
| Rex and its integration with Rust-for-Linux | Jinghao Jia, Ruowen Qin | 30m |
Rex implements a Rust-based kernel extension framework that compiles a safe subset of Rust to native code, ships a Rex kernel crate with safe wrappers over… | ||
| Kernel TEE subsystem BoF | Jens Wiklander, Sumit Garg | 44m |
Kernel TEE subsystem now contains merged support for protected DMA buffers (reference patchset landed but not yet deployed), and an upstream RPMB in-kernel… | ||
| ARM Laptop BoF | Stefan Schmidt, Vinod Koul | 48m |
Survey of concrete pain points for ARM laptops: bootchain work centers on device-tree vs ACPI tradeoffs with practical tools like UKI+stubble matching, EFI… | ||
| Agentic AI, confidential computing and attested TLS | Muhammad Usama Sardar | 41m |
Argues that combining agentic AI with confidential computing requires attestation of model weights as executable code, source code, system prompts, tool lists… | ||
| 32bit and-or noMMU Linux BoF | ChuanTzu Tsai, Mr D.Jeff Dionne, Geert Uytterhoeven, Rob Landley | 46m |
Active upstream work keeps 32-bit and noMMU Linux viable: recent tapeouts like HPMicro hpm6360/hpm6700/hpm6800 and an Allwinner V821 MCU boot mainline kernels… | ||
| Kernel Sanitizers Office Hours | Aleksandr N, Alexander P, Dmitry V, Kees C, Marco E | 38m |
Overview of what each kernel sanitizer actually finds, how it’s used, and the real tradeoffs: KASAN hardware tag mode uses ARM MTE to cut runtime cost, while… | ||
| Syzbot BoF: Open Discussion | Aleksandr N, Alexander P, Dmitry V, Marco E, Taras M | 44m |
Syzbot reports 13,500+ kernel bugs since 2017 and already ~1,700 findings in 2025; dashboard moves to syzbot.org and coverage reports are now aggregated and… | ||
| Solving Zombie Cgroups - Simple Fix or Smart Solution? | Harry Yoo, Kamalesh Babulal | 38m |
Kernel fixes reparent stuck pages to parent memory cgroup to free zombie cgroup metadata, implemented by moving page charges and updating folio locking so… | ||
| Towards Real-Time NVMe Monitoring (nvme-top) | Daniel Wagner, Nilay Shroff | 47m |
Proposes nvme-top: a real-time NVMe-oF multipath monitor built as an nvme-cli-integrated proof of concept that polls sysfs once per second to refresh a… | ||
| Enabling AMD 2D Hardware Acceleration for Video API and Applications | Solomon Chiu | 46m |
AMD brings up its VPE 2D fixed‑function path with kernel driver plumbing and a shared C VP lib that builds and submits the VPC/VPP command stream; user‑space… | ||
| From Projects to Ecosystems: Lessons from the eBPF Foundation | Bill Mulligan | 31m |
eBPF Foundation reports concrete investments: contracted Boot to land BTF improvements, low‑overhead trampolines, ARM64 feature parity and CI/test… | ||
| “Is Upstream Really Enough?” — Practical Realities of Using eBPF in Long-Term Supporte... | Kenta Tada | 16m |
Upstream eBPF innovation is not enough for LTS and embedded products: kernel internals and tracepoint behavior change, verifier incompatibilities and shifting… | ||
| Python-BPF: Writing eBPF programs in pure Python | Pragyansh Chaturvedi, Mr Varun Mallya | 33m |
Python-BPF lets developers write eBPF logic and userspace in one reduced Python grammar file, compiles by generating kernel structs from vm_linux.py, emits… | ||
| Extending eBPF to GPU Device and Driver Contexts | Yusheng Zheng, Tong Yu | 34m |
Prototypes offload eBPF into GPU device contexts by compiling eBPF to PTX/SPIR-V and adding a verifier, helpers and map support built on bpftime, with… | ||
| Fingeprinting GPU workloads with eBPF | Jiri Gogela | 20m |
Uses eBPF kprobes on the NVIDIA kernel driver to record timestamps for 41 kprobe hooks across 60 seconds per run, tested on A10 and V100 GPUs against… | ||
| Lessons from scaling BPF to detect RDMA Device Drivers Bugs in real time | Prankur Gupta | 30m |
RDMA tracer uses eBPF fexit probes on a focused subset (≈8 of ~40) of IB control-path functions to capture nonzero return contexts, stores events in a 256 KB… | ||
| AI flame graphs with eBPF | Brendan Gregg, Ben Olson | 31m |
An open source AI flame graphs tool is built that merges eBPF context with Intel EU stall hardware sampling to produce unified CPU+GPU flame graphs showing GPU… | ||
| cache_ext: Customizing the Page Cache with eBPF | Tal Zussman | 29m |
An eBPF-based framework called cache_ext injects hooks for page insertion, access, and eviction so users load kernel-verified policies per cgroup without… | ||
| Supply true signatures to vmlinux BTF | Yonghong Song | 35m |
Effort adds true function signatures into vmlinux.BTF by changing clang/pahole/libbpf/kernel: clang emits extra IR debug CU entries (artificial compilation… | ||
| uprobe optimization and override | Jiri Olsa | 25m |
Kernel now implements an x86-only uprobe optimization that replaces a five‑byte NOP probe with a user‑space trampoline and uprobe syscall, avoiding exception… | ||
| bpf tracing multi-link supporting | Menglong Dong | 29m |
Implements a global trampoline that uses an rhashtable to map target function addresses to metadata (BPF programs, cookie, arg count), letting one BPF program… | ||
| Advanced BPF profiling techniques: how to escape the curse of NMI and fetch ELF... | Andrii Nakryiko | 38m |
Shows how a profiler moves from sampling to reliable user-memory reads by combining perf-event sampling + ring buffer postprocessing with new kernel-side… | ||
| Hibernation Testing Challenges and Optimization Opportunities | Srinivas Pandruvada | 13m |
Hibernation validation runs thousands of cycles and each cycle can take days; CPU online/offline steps cost ~100–200 ms and ~70–100 ms per core so limit CPUs… | ||
| Thermal binning support | Amit Kucheria | 48m |
Proposal adds a new device tree property to encode thermal variants so one DTB covers multiple SKUs, reading a two-bit fuse (stored via NVM cells) that selects… | ||
| Exploring a real life RCU use case for Rust | Vitaly Wool | 17m |
A real allocator backend for Zswap called Z block divides blocks into equal slots and keeps a list of partially empty blocks; C uses an RCU list variant that… | ||
| Tackling challenges with HID and related device driver support in Rust | Rahul Rameshbabu | 22m |
Upstream patches already enable a specific mouse device so its buttons work around a hardware bug, and out‑of‑tree work exposes many Razer Blade controls; the… | ||
| Initialization in Rust with pin-init | Benno Lossin | 38m |
Rust code in the kernel uses pin-init to perform safe, ergonomic in-place initialization of address-sensitive types, avoiding 1MB stack overflows by returning… | ||
| External locking for internally synchronized data structures | Andreas Hindborg | 19m |
Proposes a Rust API for XArray that lets an external lock token permit the XArray to momentarily drop locks to perform allocations, avoiding C-only tricks like… | ||
| Rust language evolutions for better kernel developer experience | Xiangfei Ding | 20m |
Short update on arbitrary_self_types and its projected arrival on stable Rust, showing how receivers can now be expressed for smart-pointer-like types (Pin… | ||
| Overflowing with Fear: Detecting and Mitigating Implicit Panics in Rust | Alexandre Courbot | 30m |
Rust integer overflow and slice indexing generate implicit panics inserted by the compiler with no warning, creating blind spots for kernel code that handles… | ||
| How do we make a Steamdeck scheduler work on large servers | David Dai, Ryan Newton | 16m |
Meta evaluates SCX_LAVD on large servers and finds concrete problems and fixes: 50+ CPUs per CCX causes heavy DSQ lock contention and task thrashing so a… | ||
| The Current Status and Future Direction of the LAVD Scheduler | Dr Changwoo Min | 17m |
LAVD is a sched_ext scheduler that targets gaming latency and energy by picking CPU sets that meet a measured computing demand with minimum power. It reads… | ||
| Accelerating AI training fleets with sched_ext | Patrick Lu, Valentin Andrei, Pat Somaru | 36m |
Deploys a BPF-based userland scheduler sched_ext (scx_layered) across Reality Labs trainers with tens of thousands of GPUs to protect latency‑critical… | ||
| Taming the Chiplet: High Performance CCX scheduling via BPF | Aniket Gattani, Josh Don | 19m |
Implements fine-grained CCX-aware scheduling using BPF plus a userspace agent that maintains per-CCX runqueue state and runs an asynchronous bin-packing… | ||
| Scheduler composability with arena-based data structures | Emil Tsalapatis | 17m |
Progress on BPF arenas shows they bypass the BPF verifier enough to implement recursive data structures (trees, priority queues, hash tables) in-prog, removing… | ||
| scx_chaos: Pushing the sched_ext API to schedule badly | Jake Hillion | 16m |
scx_chaos is a deliberately non work‑conserving scheduler built on top of scx_p2dq that injects concrete faults: randomly delaying scheduling events… | ||
| Can ProxyExec and sched_ext get along? | John Stultz | 16m |
Proxy Execution implements split contexts so the scheduler-selected task becomes a runq donor while the lock owner runs; benchmarks show a 15 µs average… | ||
| From Fragmentation to Integration: Enhancing sched_ext BPF Scheduler Interoperabili... | Daniel Hodges | 16m |
Work over a year on multiple schedulers reveals duplicated effort, user-space topology parsing, and five different ways of accessing topology across… | ||
| sched_ext: current status and future plans | Andrea Righi | 19m |
sched_ext is upstream and already lets BPF replace scheduler logic, but real-time tasks can starve schedulers and trigger watchdog-driven unloads; the proposed… | ||
| Secure Interrupt delivery: Lessons Learned from Alternate Injection Enablement | Melody Wang | 17m |
SEV‑SNP Alternate Injection is enabled end‑to‑end: KVM gets a separate alternate injection enabled pass to isolate SVSM-managed interrupts, SVSM implements… | ||
| A Linux Bus for SVSM Services: Build New, Reuse VIRTIO, or Both? | Stefano Garzarella | 19m |
Proposal compares three architectures for exposing SVSM services to guests: a new minimalist kernel bus, reusing VIRTIO, or a hybrid. Key specifics: SVSM… | ||
| Discussion: TDISP, VM migration, and paravisors | John Starks | 24m |
Compares two concrete architectures for TDISP-based PCI device trust and VM migration: a paravisor-provided TIS device that handles attestation versus pushing… | ||
| Standardization of Attested TLS Protocols | Muhammad Usama Sardar | 31m |
IETF workgroup SEAT forms to standardize attested TLS and the team brings a candidate draft, ProVerif symbolic analysis, and industry partners (Arm, Intel… | ||
| PCI device authentication & encryption | Alexey Kardashevskiy, Mathieu Poirier | 27m |
Two kernel patch sets converge on a concrete plan: one brings SPDM artifacts via CISFS plus a netlink delivery path for signatures, the other implements the… | ||
| Toward a standard device attestation token for device assignment | Mathieu Poirier, Thomas Fossati | 32m |
Proposal defines a portable device-attestation token for device assignment that models SPDM artifacts and legacy PCIe claims using seven CDDL files and… | ||
| Debugging RISC-V Kernel Crashes: KDUMP, vmcore, and Crash Utility Enhancements | Austin Kim | 20m |
KDUMP on a real VisionFive2 board needs specific fixes: applying two patches enables actual vmcore extraction (author collects >10 vmcores), then loading those… | ||
| Solving Pre-silicon Kernel Upstream for RISC-V First Ever | Yuning Liang | 18m |
Claims that the Linux kernel for RISC-V can be upstreamed before silicon exist using simulation, FPGA emulation and CI/CD, and shows a real repository (K3)… | ||
| RISC-V QoS: CBQRI, RQSC and resctrl | Drew Fustini | 12m |
RISC-V adds an srmcfg CSR that tags requests with an RCID and MCID so shared controllers can enforce and measure allocation, and the CBQRI spec maps that into… | ||
| Finer-grained kernel control flow integrity and challenges | Deepak Gupta | 27m |
Kernel RFCs add riscv hardware-assisted forward CFI using a 20-bit label in LPAD matched against a caller-set X7 register and finer-grained KCFI-like hashing… | ||
| Schism: Splitting RV32 from `arch/riscv` - - We Could Both... "Ruinland" | ChuanTzu Tsai, ChunWei Shu | 30m |
Proposal splits arch/riscv32 and arch/riscv64 so 32-bit RISC‑V can carry different code, backed by a minimal RV32 profile ("B" profile) that drops the A… | ||
| Declaring WARL on the ISA string | Radim Krčmář | 23m |
WARL behavior in RISC-V causes KVM to probe CSRs by writing and reading back values, which breaks virtualization performance and consistency; the concrete fix… | ||
| Preparing RISC-V Linux for RVA23 | Charlie Jenkins | 25m |
Proposal compiles the kernel for the RVA23 profile so C code can use performance extensions (bitmanip, vector, hypervisor, CFI) instead of relegating them to… | ||
| Introduction | Atish Patra, Björn Töpel, Palmer Dabbelt | 03m |
RISC V Linux community needs concrete fixes: require vendor hardware emulation for pre-silicon testing, supply detailed test reports that name platforms and… | ||
| seccomp listeners for nested containers | Aleksandr Mikhalitsyn | 27m |
Kernel currently restricts seccomp listeners to single listener per process even though filters are nested; this patchset proposes letting listeners be nested… | ||
| Files on Unmounted Mounts | Bhavik Sachdev | 19m |
CRIU fails to checkpoint files opened on mounts detached with MNT_DETACH because /proc lacks mountinfo for those mounts; two kernel changes are proposed: a… | ||
| EROFS and containers | Xiang Gao | 31m |
EROFS is a block‑based, immutable image filesystem with fixed block alignment, chunk deduplication, per‑file compression and optional metadata compression; it… | ||
| Guarded Control Stack on arm64: Challenges in Enabling Shadow Stack Support... | Igor Svilenkov Bozic | 24m |
Adds Arm64 GCS support to CRIU by adapting x86 shadow‑stack work, using a FVP simulator and Fedora Rawhide with GCS patches for testing. Implements… | ||
| Optimizing Checkpoints with Built-in Memory Page Compression | Radostin Stoyanov | 20m |
Integrates LZ4 in-flight compression into CRIU so pages are compressed before being written, eliminating the extra write-read-compress cycle and cutting I/O… | ||
| Clean-up on Aisle "fsconfig(2)" | Aleksa Sarai | 38m |
Man pages for the “new” mount API finally land, and the talk calls out concrete gaps: FSCONFIG_SET_PATH and FSCONFIG_SET_PATH_EMPTY are effectively unused but… | ||
| A fast path to benchmarking | Mark Brown | 28m |
A new open source tool called Fastpath runs benchmarks interactively or in CI, stores raw runs in local or remote databases, and drives tests inside… | ||
| KernelCI kci-dev: Closing the developer feedback loop | Arisu Tachibana | 12m |
kci-dev brings KernelCI access into developer workflows with concrete features: query and filter the huge dataset (about 7,000 builds in recent days), trigger… | ||
| The singularity is here for testing: genai advancements on kdevops | Luis Chamberlain | 20m |
kdevops reports that since Q3 it uses generative AI to extend features and add workflows, accelerating development and improving accuracy; the whole dashboard… | ||
| Automating Scope-Based Resource Cleanup with Coccinelle | Erick Muthama | 10m |
Automates conversion of lock/unlock pairs into scope-based cleanup using Coccinelle SmPL, replacing explicit unlocks with scope guard or guard where safe… | ||
| syzbot ci: continuous patch series fuzzing | Aleksandr Nogikh | 21m |
syzbot ci is a prototype that shifts fuzzing left by monitoring kernel mailing lists, automatically applies patch series to an inferred base tree, and runs… | ||
| SYZOS: Practical KVM fuzzing | Alexander Potapenko | 19m |
SYZOS loads an immutable C library into the guest and fuzzes sequences of API calls instead of raw instruction blobs, exposing primitives for MSR access… | ||
| KFuzzTest: Targeted Fuzzing of Internal Kernel Functions | Ethan Graham | 20m |
KFuzzTest exposes stateless or low‑state internal kernel functions to userspace fuzzers by embedding metadata into vmlinux (DWARF/ELF sections) and invoking… | ||
| Kselftests augmented: running kernelspace tests from userspace | Ricardo Cañuelo | 16m |
Proposal implements a proof of concept kernel feature that lets kernel code register tests as a ktest object (function, name, parameter size) and exposes a… | ||
| UAPI testing with KUnit | Thomas Weißschuh | 21m |
Proposes building UAPI usertests as kernel userprogs that are embedded into the kernel image and executed under KUnit, reusing KUnit’s runner and preserving… | ||
| BASIL: Traceability as Code | Luigi Pellecchia | 24m |
BASIL implements Traceability as Code by letting a single YAML configuration list repositories, extraction rules and relationships and then auto-generate… | ||
| Exploring possibilities for integrating StrictDoc with ELISA's requirements temp... | Tobias Deiminger | 14m |
A working GitHub repo demonstrates StrictDoc parsing Linux source and sidecar files, merging nodes, running custom validations (example: requirement has test)… | ||
| KUnit Testing Insufficiencies | Matthew Whitehead | 24m |
Argues that KUnit lacks scalability and mocking: KUnit embeds tests in the kernel, so tests require full kernel builds and boots, forcing multiple kernel… | ||
| Defining and maintaining requirements in the Linux... | Chuck Wolber, Gabriele Paoloni, Kate Stewart | 37m |
ELISA Architecture WG drafts a requirements template and applies it to the tracing subsystem, finding and fixing a redundant flag and documenting a non‑atomic… | ||
| Applying Program Verification to Linux Kernel Code: Challenges, Practices, and... | Keisuke Nishimura | 25m |
Shows concrete results from applying deductive verification to Linux scheduler code, focusing on the task scheduler and functions like is_core_idle where… | ||
| NVIDIA Approach for Achieving ASIL B Qualified Linux: minimizing expectations from up... | Igor Stoppa | 24m |
NVIDIA claims a path to ASIL B by keeping most kernel code unqualified and concentrating verification on a small set of mechanisms that can be merged or… | ||
| Aspects of Dependable Linux Systems | Kate Stewart, Philipp Ahmann | 09m |
Linux already runs in automotive, satellites and robotics but industry standards assume traceable requirements and controlled artifacts that the kernel lacks… | ||
| Page-level and Fleet-wide Data Access Monitoring for Meta | SeongJae Park | 34m |
DAMON now exposes page-level monitoring (filters for huge pages, anon/file-backed/active lists) and a kernel agent damon_stat so a node can export idle time… | ||
| NVIDIA's Approach for Achieving ASIL B Qualified Linux | Mr Bryan Huntsman | 40m |
NVIDIA claims a practical path to ISO 26262 ASIL B for Linux by treating Linux as an SEooC and isolating safety code with context-specific page tables that… | ||
| TPMs and the Linux Kernel: unlocking a better path to hardware security | Ignat Korchagin | 42m |
Proposal adds a new kernel key type that uses TPM2 CreatePrimary in counter mode to derive reproducible per-application keys rather than storing wrapped blobs… | ||
| SFrame for Arm64 Reliable Stacktrace | Dylan Hatch, Weinan Liu | 33m |
Proposal implements a compiler-generated SFrame section that stores FDE/FRE tables mapping PC ranges to CFA, return-address and frame-pointer offsets so the… | ||
| Userspace Assisted Scheduling via Sc... | John Stultz, Mr Qais Yousef, Steven Rostedt, Vincent Guittot | 44m |
Proposes a userspace-to-kernel QoS stack that maps four classes (user interactive, user initiated, utility, background) through a middleware library using… | ||
| More Effective Approach to Detecting Potential Deadlocks, DEPT(DEPendency Track... | Mr Byungchul Park | 48m |
DEPT is a runtime dependency tracker that tracks waits and events instead of only lock acquisition order, and it finds deadlocks lockdep misses — for example… | ||
| One year into the new KernelCI | Mr Gustavo Padovan | 40m |
One year after launching the rewritten KernelCI system the project reports concrete gains: improved test stability, automated triage for some regressions, and… | ||
| TAB Q&A | Dan W, Dave H, Greg K, Jonathan C, Kees C, Miguel O, Sasha L, Shuah K, Steven R, Theodore T | 45m |
Open forum about TAB work and concrete actions: AI-assisted review is already reducing reviewer burden in prototypes but dependence on proprietary models risks… | ||
| Firmware ABI stability | Marek Vasut | 37m |
Embedded ARM systems expose bootloader firmware ABIs (notably PSCI and SCMI) that Linux uses for core start, clocks, resets and power domains, and vendors… | ||
| Fixing Clock Tree Propagation in the Common Clk Framework | Mr Brian Masney | 24m |
The Common Clk Framework currently lets a child call clk_set_rate that silently changes an ancestor via CLK_SET_RATE_PARENT, breaking sibling rates… | ||
| Improving stability for TCPM using boards that are not self-powered | Sebastian Reichel | 26m |
Boards that use cheap FUSB302-like controllers and have no backup power hit reboot loops because USB‑PD sources time out after ~5s and trigger a hard reset… | ||
| How to extract a bare metal flavor of code out of Zephyr to use in RTOS? | Khasim Syed Mohammed | 40m |
Argues for a “Zephyr‑first” workflow and shows a case of extracting a Zephyr peripheral driver into a bare‑metal/other‑RTOS build by using the Zephyr device… | ||
| Boot time optimization for embedded devices | Saravana Kannan | 33m |
Switching gzip to LZ4 for kernel and initrd decompression saves roughly 0.5–1s on modern flash, and stripping module debug symbols cuts several seconds from… | ||
| Reducing Android Boot Times: Evaluating Proactive File Prefetching | Akilesh Kailash | 25m |
A Rust user-space daemon implements a two-phase "Record and Replay" prefetch that records page insertions via the mm_file_map_add_to_page_cache tracepoint and… | ||
| Guider: Lightweight Real-Time Performance & Fault Monitoring Framework for Embedded... | Mr Peace Lee | 20m |
Guider is a self-contained monitoring toolkit for embedded Linux that continuously runs from boot, captures ftrace/atrace/ptrace/kprobe/uprobe/DWARF-backed… | ||
| Improving page_owner for profiling and monitoring memory usage per all... | Mauricio Faria de Oliveira | 18m |
page_owner now records allocation stack handles (not just text) via a new show_handles file so userspace can sample stack IDs cheaply, then resolve them once… | ||
| Scaling Kernel Production Monitoring @ Meta | Vlad Poenaru | 22m |
Meta monitors millions of servers running bleeding‑edge kernels and collects kernel logs via netconsole into regional netcond listeners, captures full dumps… | ||
| Module Error Injection with eBPF | Daniel Gomez | 23m |
moderr is an eBPF-based error injector that uses libbpf to attach to annotated kernel functions and force specific errno returns, targeting the module-loading… | ||
| Actionable Data Access Monitoring Output Data and Format | SeongJae Park | 17m |
DAMON samples pages every 5 ms and reports per-page access location, frequency, stability and recency by default every 100 ms, claiming samples pages every 5… | ||
| Reading memcg stats more efficiently | JP Kobryn | 21m |
Periodic memcg reads spend most time formatting text so the team implements BPF kfuncs to iterate cgroups, map a binary data section to user space and fetch… | ||
| From Tool To ToolBox: How bpftrace is evolving to become more composa... | Adin Scannell, Jordan Rome | 22m |
bpftrace is moving from a single-tool language to a composable toolbox: the project reports ~10,000 lines of user scripts vs ~30,000 lines of bpftrace code and… | ||
| Proactive and crash-time data collection on Steam Deck | Guilherme G. Piccoli | 20m |
Steam Deck uses a pstore default path for fast persistent crash logs (a few megabytes) and falls back to full-memory kdumpst VMcore captures (historically ~200… | ||
| Methodology and Practice in Observing Kernel Networking | Jason Xing | 18m |
Tencent builds a continuous, low-overhead observability platform for older kernels (3.10, 4.4) that targets latency as root metric and drills down to… | ||
| Challenges with Kernels built with LTO | Song Liu, Yonghong Song | 23m |
Using CONFIG_LTO_CLANG_THIN in production increases inlining surprises (about 6,800 inline-but-still-present functions without LTO vs >8,000 with LTO) and… | ||
| GCC BPF and Linux kernel | Jose E. Marchesi | 26m |
Status update on GCC BPF support: GCC already compiled all kernel selftests last year and aims to become the primary BPF compiler; BPF declaration/type… | ||
| Enforcing PI locks by default | Mr Qais Yousef | 25m |
Argues for making PI the default by changing libc to flip PTHREAD_PRIO_INHERIT default and shipping a runtime systemwide tunable or environment switch to roll… | ||
| Adding C library wrappers for all Linux syscalls | Mr Carlos O'Donell | 24m |
Calls for adding C library wrappers for every Linux syscall, arguing this makes kernel features easy to use and test, and proposing prioritized work on futex… | ||
| Adaptive futexes: spin if the owner is still running on another CPU | Steven Rostedt | 25m |
Proposal: store the lock owner (owner TID) in the futex word so the kernel can observe the owner’s run state and spin if owner is running on another CPU… | ||
| Distributed ThinLTO Build Support in Kbuild | Rong Xu | 20m |
Kbuild adds a distributed ThinLTO mode that extracts the LTO index from VMLinux.A and generates backend compile actions and a new makefile so the build system… | ||
| Exciting new compiler flags for kernel security | Florent Revest | 18m |
Describes two concrete toolchain paths: selective sanitization using -lower-allow-check-percentile-cutoff-hot plus -fsanitize-ignorelist that reuses FDO/PGO… | ||
| Security Features status update | Justin Stitt, Kees Cook | 21m |
GCC, Clang, and Rust reach concrete parity on several items: per-thread stack protector guard locations are implemented across RISC-V, PowerPC and s390, and… | ||
| Improving Build Time SBoMs | Joshua Watt | 28m |
Build‑time SBoMs aim to record the exact inputs, compiler invocations and outputs of a build so consumers can triage CVEs and support attestation and… | ||
| OpenWrt (One) build system: lessons in *all* the compliance and how to broadly... | Denver Gingerich | 29m |
OpenWrt One produces a 1.2 GB tarball that bundles toolchain plus all upstream packages, feeds and DL archives so the full image builds offline; workflow uses… | ||
| One Config to Rule Them All: From systemd to Reliable Embedded Upgrades | Behan Webster | 29m |
Argues for a concrete, versioned layering: /usr defaults (vendor-provided), /etc overrides (fragments/conf.d) and /run ephemeral, using tools like systemd… | ||
| Enabling UEFI Secure Boot Across Modern Build Systems | Frank Vasquez | 30m |
Modern distros rely on shim signed by Microsoft to boot on commodity x86 hardware, inserting a distro key while preserving Microsoft as the initial root; that… | ||
| Debian Official Debug Kernel for All? | Yunseong Kim, Seonghee Jin | 26m |
Proposal adds official Debian debug-kernel flavors starting with amd64 and arm64, delivering a lightweight, midweight, and heavyweight config family that… | ||
| Build System Analysis for Kernel Development and Customization | Mr Alejandro Hernandez Samaniego | 28m |
Compares concrete kernel workflows and shows what actually builds and boots: uses Fedora (fedpkg, SRPMs / KernelArc source tree) to produce RPMs and apply… | ||
| News from PREEMPT_RT | Sebastian Siewior | 38m |
Local bottom halves get an explicit lock implemented as lock_nested_bh (merged in 6.18) which removes implicit PCPU locking in PREEMPT_RT and eliminates some… | ||
| Revocable: a mechanism for preventing "classic" use-after-free bugs | Tzung-Bi Shih | 27m |
Proposes revocable as a kernel indirection that decouples resource and consumer lifetimes, adding provider and consumer handles and using RCU/SRCU (calls like… | ||
| Re-think of richACLs in AI/LLM era | Coly Li | 44m |
fnnas ships an AI‑native NAS with over one‑million active installations and runs compact LLM agents that read photos, write metadata and create hard links but… | ||
| drm and cgroup memory management - two great tastes | David Airlie | 30m |
Integration work patches add memcg counters for GPU active and reclaimable memory, port TTM pools to list_lru and make the TTM shrinker NUMA-aware, and account… | ||
| Highmem deprecation planning | Arnd Bergmann | 46m |
Proposal: deprecate highmem on 32-bit kernels over the next decade by shrinking what actually uses it, landing on a few supported paths and then removing… | ||
| Moving kernel swapping infrastructure to Rust | Mr Vitaly Wool | 33m |
Rust implementation shows equal or slightly better performance than the C version on kernel builds, and a complete Rust allocator prototype called ZBlock… | ||
| Implement simple local DNS server with BPF | Raman Shukhau | 23m |
Claims a working BPF-based local DNS responder that parses UDP DNS queries and rewrites packets in-place, implemented as XDP and TC BPF programs. It swaps… | ||
| Challenges with implementing in-kernel FQDN policies using eBPF | Hemanth Malla | 25m |
Cilium replaces its user-space DNS proxy by parsing DNS in-kernel with eBPF, removing the single-point-of-failure and reducing dataplane/control-plane… | ||
| Kernel-Resident Regex and Jails: DFA-powered eBPF filtering and certificate... | Justin Ngai | 34m |
Compiles regexes to DFAs and runs them in-kernel via eBPF at LSM and fentry to achieve no-emit security, cutting user-space filtering and I/O; integrates with… | ||
| Tracking Files across the operating system using eBPF | Carl El Khoury | 23m |
Metal Armor hooks eBPF syscall events to detect file creation and then track renames, moves, deletes, compression/decompression and uploads, tagging sensitive… | ||
| BpfJailer: eBPF based Mandatory Access Control | Liam Wisehart | 32m |
BpfJailer is an eBPF LSM that Meta uses to jail AI-generated code, block network/filesystem/IPC access, enforce signed binaries and command-line checks for… | ||
| Discussion: What's next for BPF Signing | KP Singh | 25m |
Initial code lands upstream adding a detached signature field in the BPF attr and kernel keyring verification so stable-instruction programs verify at load… | ||
| Introduction to eBPF Load-Acquire and Store-Release Instructions | Peilin Ye | 37m |
Adds two BPF_ATOMIC ops providing load-acquire and store-release semantics, available in LLVM 21+ and merged into mainline kernel with support for arm64… | ||
| Fuzzing the Verifier with a Test Oracle | Paul Chaignon | 18m |
Proposal implements a test oracle that saves verifier expectations into a BPF map of verifier states and at runtime performs interpreter runtime checks of… | ||
| ePass: A Framework for Enhancing Flexibility and Runtime Safety of eBPF Programs | Yiming Xiang | 29m |
ePass lifts eBPF bytecode into an SSA IR and provides a kernel-side compiler pipeline that runs verifier-cooperative passes; 12 example passes (most under 100… | ||
| Loops handling by BPF verifier and what should we do about it? | Eduard Zingerman | 28m |
Scalar evolution analysis infers algebraic formulas for loop-varying registers and solves for iteration counts, but works on assembly so stack spills and… | ||
| Making Sense of State Pruning | Mahé Tardy, Paul Chaignon | 30m |
State pruning reduces BPF verifier path explosion and currently accounts for about 15% of the verifier by saving and comparing states at designated pruning… | ||
| BPF Verifier Visualizer | Ihor Solodrai, Jordan Rome | 28m |
BPF Verifier Visualizer (bpfvv) parses verifier log text and builds a model of instructions and states, offering a browser UI that shows C source mapping… | ||
| Famfs Update: Status, DAX Challenges & Use Cases | John Groves | 15m |
Famfs formats devdax-backed memory as a scale-out POSIX file system and aims for upstream inclusion next year, while navigating a recent DAX refactor that… | ||
| Protected DMAbufs and its dynamic memory assignment woes | Sumit Garg | 20m |
Protected DMA-buf support targets secure media pipelines using a GStreamer pipeline with OP-TEE and must handle allocations of hundreds of MBs while keeping… | ||
| DAMON-based Pages Migration for {C,G,X}PU [un]attached NUMA nodes | SeongJae Park | 19m |
DAMON monitors page recency, frequency and stability and DAMOS automates promotion/demotion; SK Hynix reports up to 94% Alama CP CPP benchmark speed up for a… | ||
| Parallel Paths to High-Bandwidth Memory for ML/AI: Specific Purpose Me... | Mr Rajneesh Bhardwaj | 23m |
UEFI tags HBM as EFI_MEMORY_SP (UEFI 2.8), the kernel maps that to E820_TYPE_SOFT_RESERVED and calls memblock_reserve so Linux’s DAX/HMEM auto‑claims the… | ||
| Kernel CXL Cache (safe) support | Alejandro Lucero | 22m |
CXL Type‑2 devices can cache host memory through a snoop cache per host bridge so the kernel must ensure a device’s advertised cache size fits the cache and… | ||
| Unifying sources of page hotness information | Bharata Bhasker Rao | 29m |
A centralized PG hot subsystem collects access reports from NUMA hint faults, PTE access-bit scans (KLRU/KLRU scanD), device samples like AMD IBS and CXL HMU… | ||
| CXL HDM-DB support for Linux | Davidlohr Bueso | 29m |
Linux gets upstreamable CXL HDM-DB plumbing: a kernel-side BI discovery flag lets the stack detect device-coherent Back-Invalidate capability, and QEMU… | ||
| Introduction and Kickoff | Adam Manzanares, Dan Williams | 01m |
Welcome and quick orientation for a packed microconference on device memory, CXL and Optane-derived topics, with many accepted submissions so sessions run… | ||
| Memory descriptors for DEVICE_PRIVATE memory | Alistair Popple | 15m |
Proposal to remove struct pages for DEVICE_PRIVATE/ZONE_DEVICE memory and replace PFNs with an index into a preallocated descriptor array accessed via a… | ||
| Folio Restructuring for Huge Pages in guest_memfd | Ackerley Tng | 17m |
Adds HugeTLB and THP support to guest_memfd by making folio restructuring (split/merge) orthogonal to THP and HugeTLB, using existing helpers like… | ||
| Decoupling Large Folios from Transparent Huge Pages | Pankaj Raghav | 14m |
Goal is to let file-backed large folios work without depending on CONFIG_TRANSPARENT_HUGEPAGE by removing the huge_memory.c splitting dependency. One quick… | ||
| Taming Zombie Cgroups: Should Reparenting Be Simple or Smart? | Harry Yoo, Kamalesh Babulal | 16m |
Proposes upstreaming a concrete fix that reassigns LRU page charges from memory cgroup to the object cgroup API so zombie cgroups free their metadata, with a… | ||
| type-based slab allocation: kmalloc_obj family | Kees Cook | 15m |
Proposal changes slab allocator to be type-aware by introducing kmalloc_obj() v4/v5 so allocations return typed pointers instead of void, and it converts the… | ||
| Per-cgroup Swap Device Control | Chris Li, YoungJun Park | 14m |
Proposal implements per-cgroup swap tiers so cgroups pick from named groups of swap devices mapped to priority ranges (example ranges: network -1–40, HDD… | ||
| Deprecating zone_reclaim_mode | Joshua Hahn | 12m |
zone_reclaim_mode, added in 2005, forces local direct reclaim when a local zone runs out instead of falling back to remote allocation; on a two‑node 125 GB… | ||
| The life cycle of the mm_struct | Liam R. Howlett | 15m |
Multiple regressions where an mm_struct is observed half-initialized or lingering cause real crashes: tasks with zero VMAs, incomplete setup that drops locks… | ||
| CMA allocations fail due to pinned MOVABLE pages | Juan Yescas, Kalesh Singh | 12m |
Android reserves ~20% RAM as CMA and large O_DIRECT workloads (OTA reads) create a cascade of short-term pinned MIGRATE_MOVABLE pages that exhaust MIGRATE_CMA… | ||
| Mempolicy is dead, long live memory policy! | Gregory Price | 14m |
Proposes changing the allocator so special-purpose memory sits off the default fallback path and is only reachable with a driver-visible flag and node… | ||
| Memory Allocation Profiling upcoming features | Sourav Panda, Suren Baghdasaryan | 16m |
Kernel allocations become visible via a tracepoint-backed context capture implementation that records call stacks into BPF and aggregates bytes per stack… | ||
| Wrap up | Vincent Guittot | 01m |
Schedule session closes and attendees get a brief announcement: an OPM conference is set for Cambridge April focusing on scheduling in power management and… | ||
| Paravirt Scheduling: Framework for better physical CPU utilizat... | Ilya Leoshkevich, Shrikanth Hegde | 25m |
Defines a kernel framework that marks idle-but-constrained vCPUs as a paravirt CPU mask and makes the load‑balancer and wakeups respect it so no new tasks land… | ||
| CPU Isolation and IPI interference | Valentin Schneider | 19m |
CPU isolation fails because housekeeping CPUs still send IPIs for things like static key updates and TLB invalidations; observed with a kernel compile plus… | ||
| RV and real-time properties | Gabriele Monaco | 22m |
A runtime verifier module now uses hybrid automata (timed constraints on nanoseconds) to validate the kernel’s real‑time behavior, applied to the deadline… | ||
| Runtime verification monitors for real-time | Nam Cao | 19m |
Mainline PREEMPT_RT makes the kernel real-time capable but userspace still causes surprises; runtime verification monitors use kernel tracepoints and currently… | ||
| Push based load-balancing for fair tasks | Prateek Nayak | 22m |
Prototype RFC sent after OSPM implements a push-based balancer that replaces global idle masks with a global counter plus a chain of per-LLC nohz masks stored… | ||
| Enabling Push call back in Fair class | Vincent Guittot | 21m |
Proposal adds a push callback mechanism to the fair scheduler so the system calls select_task_fair outside wakeup, mainly when a task goes back to runnable or… | ||
| Rethinking Android's Priority Inheritance | Carlos Llamas | 22m |
Binder currently inherits CFS niceness inheritance, which feeds static nice values into CFS/EVDF and produces inconsistent scheduling behavior; an out‑of‑tree… | ||
| Cache Aware Scheduling | Mr Tim Chen, Mr Yu Chen | 21m |
Proposed RFC patches add cache aware scheduling that keeps threads of a process inside the same last-level cache (LLC) domains by adding a new migration type… | ||
| Introduction to Scheduler and RT micro-conference | Vincent Guittot | 01m |
Micro-conference runs eight sessions with a rhythm of 22-minute slots and a schedule of four sessions then a 30-minute break then four more. Each slot… | ||
| KernelSBOM: Reconstructing Linux Kernel Builds into Trusted SPDX Bills of Materi... | Maximilian Huber | 34m |
KernelSBOM reconstructs the Linux kernel build graph by parsing .cmd files to recover exact commands, inputs and outputs, augments that with .S include parsing… | ||
| Automating Linux Kernel Documentation for Safety-Criti... | Grant S, Justin S, Morgan R, Toby H, Tom I | 31m |
Claims an automated pipeline that uses locally hosted LLMs to generate safety-oriented documentation for Linux kernel functions, applied to ftrace and… | ||
| Adding Testable Code Specifications in the Linux Ke... | Chuck Wolber, Gabriele Paoloni, Kate Stewart | 45m |
Proposal creates a machine‑readable kernel-doc format for testable expectations that ties declared intent to tests and coverage via a checksum/hash linking so… | ||
| Path Safety in the Trenches | Aleksa Sarai | 44m |
Walks through concrete fixes for filesystem confused‑deputy bugs that produced CVE-2025-31133, CVE-2025-52565 and CVE-2025-52881, showing a migration of runc’s… | ||
| RV and the deadline scheduler: how to verify your subsystem at runti... | Gabriele Monaco, Juri Lelli | 42m |
Runtime Verification runs inside the kernel to check the deadline scheduler by compiling models into kernel monitors, adding timed automata and per-object… | ||
| Tyr: a new Rust GPU driver | Daniel Almeida | 30m |
Tyr is a new Rust kernel‑mode GPU driver for ARM Mali (tested on the G16) that implements IO‑MMU mappings, DRM VM and buffer management, CSF microcontroller… | ||
| Rust for Linux | Miguel Ojeda | 37m |
Project reports 2,500 patch series so far with about 300 unique submitters, roughly 20 maintainer entries (≈30 people when expanded) and seven active core… | ||
| Multikernel Architecture: Kernel-to-Kernel Isolation and Resource Management | Cong Wang | 50m |
Claims a working multikernel that runs multiple isolated Linux kernel instances on one machine without a hypervisor, using enhanced kexec handover to spawn… | ||
| Beyond the Demo Kernel: Mapping BSPs to Mainline | Alessandro Carminati | 45m |
Shows a practical method to shrink a vendor BSP to the real upstream delta: use code fingerprinting and fuzzy matching on source and history, run artifact… | ||
| Rethinking Linux tools with Compact Debuginfo Formats | Stephen Brennan | 35m |
Compact debuginfo formats let tools ship small, targeted metadata alongside kernels and binaries so you get useful runtime features without full DWARF… | ||
| Perf tools updates and beyond | Namhyung Kim | 45m |
Adds memory-oriented output and code annotations to data type profiling and improves instruction-pointer tracking for pointer arithmetic; current… | ||
| Meminspect - Kernel memory selection and inspection mechanism | Eugen Hristev | 46m |
Meminspect provides a compact table of annotated kernel regions so consumers can produce a tiny core image instead of a full kdump; it implements a… | ||
| Linux Tracing updates | Mr Masami Hiramatsu | 35m |
Persistent ring buffer keeps pre-crash kernel traces across reboot by reserving memory and exposes a bootmap backup to avoid overwrite; a real case shows it… | ||
| A (Simple) Hazard Pointer Update | Boqun Feng, Paul E. McKenney | 51m |
A variant called simple hazard pointers (shazptr) offers an intuitive API that uses a shared per-CPU hazard pointer slot and a wildcard mechanism to avoid… | ||
| Demystifying Linux NPU Subsystem: From Vision to LLM at Edge | Mr Jagan Teki | 23m |
Linux lacks a unified NPU subsystem so current drivers are vendor-specific, tuned for vision, and rely on DRM for DMA-buf and fences while hiding job… | ||
| Power Management and USB... | Mr DARRION RAMOS, Ms VICTORIA SIVER, Dr Ken Yihang Bai, Dr Tuba Yavuz | 37m |
A modular fuzzing pipeline combines LibAFL, S2E basic‑block coverage, and an SSH-driven input processor to fuzz kernel power‑management races during suspend… | ||
| Android Boot: Next Steps | Mr Dmitrii Merkurev | 14m |
Google pushes GBL UEFI app as a reusable Android bootflow to reduce vendor work and enable features like bootconfig and PVMF loading. Current problems are… | ||
| Bootloader-Based S2D(suspend to disk) Mechanism on mobile | Tengfei Fan | 15m |
Claims a bootloader-side S2D restore that loads snapshot in blocks rather than page-by-page, bypasses the temporary kernel and thus avoids temporary kernel… | ||
| Ferrochrome: A Linux VM on Android host via AVF | Jeongik Cha | 15m |
A Debian VM runs on Android via AVF using virtio block/input and a gfxstream-backed virtio‑gpu with explicit cache flushes and zero‑copy mapping to Android… | ||
| Ramdump for Trusted VMs: Efficient Crash Debugging Without Reboot Device | Cong Zhang | 13m |
A ramdump system captures only registered critical regions from a secure VM so the hypervisor can encrypt the memory and hand back small, targeted dumps… | ||
| BPF Based Telemetry, Metrics, and Accounting on Android | T.J. Mercier | 16m |
BPF iterators replace procfs/ci sfs scans to collect DMA-buf, wakeup-source and CPU metrics, cutting a global read that used to enumerate thousands of CISFS… | ||
| AOSP on taped-out RISC-V SoC without RVA22/23, I came back to finish... "Ruinland" | ChuanTzu Tsai | 11m |
Running AOSP on taped-out Qilai silicon that predates RVA22/23 requires concrete fixes: patched build system, added an Andes CPU variant, modified… | ||
| Pixel 6 support upstream and on android-mainline | William McVicker | 15m |
Upstream Pixel 6 support already lands an ACPM power coprocessor driver that unblocks CPU frequency, mailbox and regulator control and enables PMX NVMe-based… | ||
| How do we make KUnit work for us? | Tiffany Yang | 16m |
KUnit converts a slow, hard-to-run binder selftest into a fast suite that compiles and runs in ~25s on UML and ~35s on QEMU and exercises 75,000 combinations… | ||
| Cooperation between CPU and system level cache by using MPAM | Yiwei Huang | 14m |
MPAM PartID tags get translated at a small hardware-backed translation layer that maps incoming PartIDs to SLC-specific IDs via mapping tables programmed over… | ||
| Android Ashmem to Memfd | Isaac Manjarres | 15m |
Android replaces the legacy ashmem driver (removed from staging in kernel 5.18) by switching allocations to memfd_create via an ashared interface, temporarily… | ||
| The Challenge of Loading 4kB-Aligned ELFs on 16kB Systems | Juan Yescas, Kalesh Singh | 15m |
Devices gain ~30% app launch and ~5–7.5% power savings by moving to 16k pages, but legacy 4k-aligned ELFs put RX and RW segments in one page. A modified… | ||
| HW/SW design recommendations for 16kb devices | Juan Yescas, Kalesh Singh | 13m |
Recommendation is to keep MMU page size = CPU page size (benchmarks show ~2% better throughput) and to make control-status registers 16KB-aligned so users can… | ||
| Kexec Handover Performance Improvements | Jason Miu | 16m |
Replaces the xarray+serialize linked‑list handover with a rad‑tree key encoding that packs physical address and page order so the kernel only passes a single… | ||
| KSTATE as [de]serializer of kernel's state | Andrey Ryabinin | 15m |
Proposal creates a common framework KSTATE to serialize kernel structs by describing preserved fields and storing field offsets so generic code can copy them… | ||
| Managing serialization versions for LUO objects | Pratyush Yadav | 14m |
Proposal adds a machine-readable list of live update versions into an ELF section so userland tooling can read current and next kernels, compute overlaps, and… | ||
| Memory preservation using LUO | Pratyush Yadav | 13m |
LUO is extended to preserve memory via memfd by serializing file map folios and supporting both normal folios and huge TLB backing; patches already preserve 1… | ||
| Supporting Live Update in VFIO | David Matlack, Josh Hilke | 16m |
Work to let VFIO PCI devices keep running across kexec centers on concrete steps: a live-update orchestrator already accepts a vfio cdev fd preservation… | ||
| IDPF live-update support | Brian Vazquez | 17m |
IDPF live-update work demonstrates a concrete approach: a prototype saves and later restores ring head/tail pointers for the IDPF firmware channel and… | ||
| Supporting Hypervisor Kexec with Modern Devices | Adithya Jayachandran, Saeed Mahameed | 11m |
Demonstrates a working approach that preserves state across hypervisor kexec by moving a minimal mlx5 PF driver into user space as a VFIO user-space mlx5… | ||
| Blackout Reduction | Pasha Tatashin | 15m |
Aim is to cut kexec live-update blackout toward under a second by specific changes: parallelize hundreds of device shutdown callbacks and reboot notifiers… | ||
| PCSC: Caching PCI Config Space Accesses for faster Live Updates | Evangelos Petrongonas | 12m |
At scale PCI enumeration and VF creation trigger a flood of hardware config reads, so PCSC implements a per-device cache attached to struct pci_dev, intercepts… | ||
| Live Update Orchestrator | Pasha Tatashin | 15m |
LUO is a kernel framework plus daemon LUOD that coordinates live updates by preserving chosen kernel pages across a kexec reboot (KO added in 6.16) while… | ||
| IOMMU State Preservation for Kernel Liveupdate | Samiullah Khawaja | 15m |
Kernel liveupdate preserves full IOMMU state across kexec by saving IOMMUFDs, domain and device context into the FLB so DMA mappings survive early boot. It… | ||
| PCI liveupdate subsystem and PCI device | Chris Li | 14m |
Describes a prototype PCI liveupdate subsystem built on the Live Update Orchestrator that preserves PCI state across kexec by serializing a folio of pci_dev… | ||
| Address Space Isolation is ready | Brendan Jackman | 48m |
Claims ASI now defends against native attackers as well as KVM guests and shows a working prototype that fixes the page-cache slowdown that previously killed… | ||
| Leveraging New General-Purpose Registers in the Kernel | Chang Bae | 33m |
APX adds 16 new GPRs in 64‑bit mode and exposes state via XSAVE component 19; compilers already accept an APX flag (GCC mostly ready, LLVM fixing bugs). New… | ||
| The new kernel CPUID API | Ahmed S. Darwish | 21m |
A centralized kernel CPUID API exposes typed, generated headers (from the x86 CPUID database XML→CSV→C pipeline) and replaces hundreds of ad‑hoc call sites… | ||
| Dynamic mitigations | David Kaplan | 34m |
Cloud VMs can change CPU mitigation policy at runtime via a writable sysfs reconfiguration interface that accepts bug-style kernel options and causes the… | ||
| Nova GPU Workshop: Nova: status update + 2026 plans | Alexandre Courbot | 05m |
Nova is a Rust-based upstream kernel driver for NVIDIA GPUs that grows from 387 lines in 6.15 to ~10,000 lines on master in under a year; the GPU System… | ||
| Nova GPU Workshop: GPU Firmware: status and 2026 plans | John Hubbard | 15m |
Nova targets Turing and later GPUs and locked on to 570.144 for a stable GSP firmware ABI, with a planned migration in 2026 to roughly move to ~600 in 2026 and… | ||
| Nova GPU Workshop: vGPU + VFIO: status and 2026 plans | Zhi Wang | 08m |
Nova’s Rust-based GPU driver (Nova) already lands two RFCs published and plans a staged vGPU rollout: stage one adds PCI config/SRV cap reads into GSP init and… | ||
| Nova GPU Workshop: User mode command submission for GPUs | John Hubbard | 58m |
Nova’s Rust driver explores keeping user mode command submission for performance while solving VRAM overcommit: current user submission offers no… | ||
| Ultraviolet: A Code Integrity Model for Minimal Container Hosts | James Morris | 21m |
Ultraviolet proposes an end-to-end, policy-driven launch-time code integrity model for minimal container hosts that extends secure boot and a hardware root of… | ||
| The Future of Platform Security Measurement in Linux | Maciej Pijanowski | 25m |
LVFS/ fwupd’s Host Security ID (HSI) becomes the de facto Linux platform score but currently targets proprietary UEFI and leaves coreboot/edk2 vendors with… | ||
| Reconcilable Differences: Booting Linux on IBM PowerVM LPARs | George Wilson | 21m |
IBM PowerVM LPAR boot differs from UEFI: partition firmware (Open Firmware) loads a client program from a prep partition or TFTP, negotiates client… | ||
| ReLaunch Revisited: A Refresher on TrenchBoot Late Launch | Daniel Kiper, Daniel Smith | 20m |
Secure ReLaunch integrates D-RTM “late launch” into Linux by adding a new GXE platform for GRUB that kexecs to a multiboot ELF, uses a SLRT struct to hand… | ||
| Design Space and Challenges in Design of Attested TLS Protocols | Muhammad Usama Sardar | 19m |
Formal symbolic analysis using ProVerif shows pre- and intra-handshake designs leave many security properties unmet and even the best intra-handshake variant… | ||
| Oak stage0: a minimal firmware for (confidential) virtual machines | Mr Andri Saar, Kevin Hui | 21m |
Minimal VM firmware targeting QEMU microvm and Q35 machines that boots Linux with far less code than OVMF/SeaBIOS, supports AMD SEV-SNP and Intel TDX, and is… | ||
| Who Authenticates Linux? Rethinking PAM & NSS in the Age of Cloud Identity | Mr David Mulder | 26m |
An open-source daemon called Himmelblau glues Azure Entra ID into PAM and NSS by reverse-engineering web flows, scraping JavaScript configs, polling… | ||
| Android Boot, DRTM, UKIs | Dmitrii Merkurev, Leif Lindholm, Ram Muthiah | 33m |
Android introduces a UEFI application called GBL Rust UEFI app that runs no_std, compiles with Bazel, and already boots Cuttlefish and several vendor… | ||
| Welcome to the Gaming MC! | André Almeida, David Carl Vernet | 01m |
A short status report on Linux gaming work that lists concrete kernel and userland gaps being tackled and who is already shipping on Linux, with clear calls… | ||
| Resource management on embedded gaming devices | Mr Stefan Bossbaly | 33m |
Meta builds Orchestrator, a Rust policy framework that centralizes cgroup v2 control, runs BPF collectors into BPF ring buffers, and enforces per-process… | ||
| Level Up Your Game: OS Kernel and Game Interactions revealed with Perfetto | Mr Ramesh Peri | 28m |
Perfetto collects ftrace events (sched_switch, sched_waking, sched_wakeup, IRQ and softirq entry/exit) and a minimal config yields ~2–3% tracing overhead while… | ||
| A new API for robust futex list | André Almeida | 25m |
Proposal adds a new get/set_robust_list API that returns a kernel-assigned list index and accepts an explicit bitness flag (32 or 64) so emulators like FEX can… | ||
| Steps Towards a Gaming-Optimized Scheduler | Dr Changwoo Min | 42m |
Proposal shows a working sched_ext prototype called LAVD that targets Steam Deck/Wine+Proton stacks and claims measurable reduction in stutter by prioritizing… | ||
| Proactive and crash-time data collection on Steam Deck | Guilherme G. Piccoli | 21m |
Steam Deck collects both proactive metrics and crash-time artifacts using pstore default (≈60 MB RAM) for fast persistent logs and kdump for full vmcore when… | ||
| Confidential Devices (TEE I/O): A series of modest proposals | Dan Williams | 28m |
Phase one lands in 6.19 with PCI, VFIO, DMABUF and IOMMUFD pieces merged and the repo shows 75% complete for the host stack, but KVM integration and the guest… | ||
| iommu page table consolidation and iommufd | Mr Jason Gunthorpe | 30m |
Generic IO page table code lands in 6.19 with AMD and VTD formats working, Risk5 on the mailing list and about three quarters of ARM in-tree, replacing six… | ||
| vfio, iommufd: Enabling user space drivers to vend more granular access to client pro... | Alex Mastro | 29m |
Runs a user space driver on legacy VFIO containers today and plans to move to MMUFD imminently, but currently shares whole VFIO/MMU file descriptors so clients… | ||
| VFIO: Very Frightening I/O? Taming Wild Guests and their PCIe Config-Space Abuse | Hubertus Franke | 21m |
Experiments fuzzing PCIe config space across about ten GPUs, NVMe and NICs find single writes into unassigned config offsets repeatedly trigger PCIe errors… | ||
| A Hyper-V Based pvIOMMU Built on Generic I/O Page Table | Yu Zhang | 21m |
A Hyper-V pvIOMMU prototype for Linux guests uses the community Generic I/O Page Table framework and drives DMA remapping from the guest’s stage-1 nested… | ||
| Plumbing SDXI into Linux: From DMA Engine to User‑Space Offloads | Wei Huang | 20m |
SDXI extends DMA engines to handle multiple address spaces and user‑space ownership via PASID/AKEY validation and 64‑byte descriptors with vendor extension… | ||
| Turning PCIe Hints into Cache Hits: Enabling Smart Data Cache Injection in Linux | Wei Huang | 16m |
AMD’s Smart Data Cache Injection (SDCI uses PCIe TLP Processing Hints to steer inbound DMA into the consumer CPU’s L2 cache; Linux adds a new ACPI _DSM… | ||
| PCI Pwrctrl woes | Manivannan Sadhasivam | 14m |
Framework currently fails to control PERST#/POR timing, multi-endpoint supplies and D3 suspend; proposed patchset adds three host-controller APIs… | ||
| Status of DTS validation in the Linux kernel (and how to improve it) | Mr Krzysztof Kozlowski | 21m |
A one-year audit compares DTS validation results (bscheck) between kernel snapshots by building the schema from linux-next and measuring total warnings… | ||
| Optimizing dependency tracking through DTC changes | Saravana Kannan | 22m |
fw_devlink currently parses ~20 DT properties to find phandles, which scales poorly and misses many vendor-specific properties. Proposed fixes: have DTC emit a… | ||
| "Exclusive" resources that are not exclusive - the case of shared GPIOs in... | Mr Bartosz Golaszewski | 19m |
Device-tree often lists the same GPIO for multiple devices but the GPIO API treats a line as exclusive; current workaround uses a non‑exclusive flag that lets… | ||
| Power Sequencing for Enumerable Busses - What's Still Missing From the Device Tree... | Chen-Yu Tsai | 19m |
Enumerates real gaps: SDIO power-sequence bindings exist but are now deprecated, USB host Type-A power is described in DT yet USB Type-A VBUS binding… | ||
| DeviceTrees - MIPI SoundWire® Device Class for Audio (SDCA) and classic ACPI-DT... | Srini Kandagatla | 26m |
SDCA targets ACPI but a merged Intel generic codec driver today expects firmware-described init/topology via the kernel fwnode APIs; the choice is either… | ||
| DT formatter - a maintainer's and contributor's elusive dream? | Konrad Dybcio | 22m |
A formatter that normalizes style aims to cut weeks of review churn by enforcing a common subset of DTS syntax and ordering rather than changing semantics; the… | ||
| Device Tree Metadata: How Bootloaders Pick DTBs and Apply Overlays | Chen-Yu Tsai | 29m |
Bootloaders need concrete metadata to pick a base DTB and the exact DTBOs: FIT image uses a configuration that maps a compatible string to kernel + DTB +… | ||
| Networking track kick-off | Jakub Kicinski | 24m |
Netdev maintainers report concrete infrastructure changes: the project forms the NetDev Foundation, buys five servers (HAProxy-hosted) and migrates CI off a… | ||
| SUPERp: Scale UP EtherRnet Protocol | Tom Herbert | 33m |
Proposes SUPERp, an Ethernet-based Scale Up protocol for RMA and HPC ops that compresses IP into a SunH network header with 16-bit addresses (fits ~65k hosts)… | ||
| Experiments in Using AI to Enhance the Linux Spanning Tree Protocol (STP) | Stephen Hemminger | 24m |
Experiments produce kernel patches, documentation and tests: JSON output for iproute2 is implemented with minor cleanup, working STP conformance test is… | ||
| Locality-Aware, Self-Adjusting Data Structures: A Case Study in Linux Firewalls | Jonas Köppeler | 24m |
Prototype implements the Move-Recursively-Forward algorithm inside nftables using a per-CPU pointer-array view so rule swaps are cheap, and it extracts IP/port… | ||
| Congestion Signaling (CSIG) for Linux TCP Data Center Networking | Sagarika Sharma | 26m |
Adds CSIG support to the Linux TCP stack: initializes a 4-byte compact L2 tag on transmit, switches compare-and-replace the tag for the worst bottleneck, and… | ||
| Running PTP at scale | Vadim Fedorenko | 24m |
Meta runs PTP to 10 microseconds precision across millions of servers and demonstrates a PTP stack handling 1.5M QPS to a modified PTP (ATSC) grandmaster and… | ||
| Packet Metadata - Where Are Thee? | Jakub Sitnicki, Mr Willem Ferguson, Mr Tiago Lam | 38m |
Proposal reuses XDP headroom and ships a bpf_dynptr indirection (merged for 6.8) to decouple metadata from skb->data, but still needs a new skb field to track… | ||
| The case for zero-copy in containers: Accelerating... | Anton Protopopov, Daniel Borkmann, David Wei | 50m |
Built a kernel-side queue leasing mechanism that lets a container bind virtual NIC queues which proxy to a physical NIC queue, enabling AF_XDP and… | ||
| Kernel-Native Packet Processing on AMD GPUs: XDP Offload without Userspace Runtimes | Taehee Yoo | 22m |
Claims kernel-native XDP offload on AMD hardware by building BPF JITed GPU kernels and submitting them from kernel space via AQL descriptors, avoiding… | ||
| Linux Networking with MANA: RX Path Optimization and Netsh... | Mr Dipayan Roy, Sri Satya Vennela Erni | 44m |
MANA's RX path changes allocation from one page per packet to page_pool fragments plus pre-DMA-mapped page pools per RX queue, letting up to 35 1500‑byte… | ||
| ML applications in Linux kernel | Viacheslav Dubeyko | 41m |
Argues for a lightweight ML infrastructure that links kernel subsystems to user space via sysfs + eBPF bridge and netlink so models train off‑host and deliver… | ||
| Linux CVE workgroup | Damiano Melotti | 43m |
A community workgroup publishes structured kernel CVE assessments in a GitHub repo (cloud-lts/linux-cve-analysis), analyzes roughly 700+ CVEs analyzed, meets… | ||
| KDLP: Expanding the Linux Kernel Talent Pipeline | Denis Aleksandrov, Mr Rado Vrbovksy | 44m |
KDLP creates a repeatable university-to-industry pipeline that started 4–5 years ago at UMass Lowell and now runs in multiple countries, publishes all course… | ||
| Microconference for resctrl | Fenghua Yu | 01h 05m |
Microconference reports concrete kernel work: resctrl is adding ARM MPAM integration, a new schemata layout that separates resources from controls, and reads… | ||
| Rust for Linux Office Hours | Alice R, Mr Andreas H, Benno L, Boqun F, Danilo K, Miguel O | 35m |
Discussion documents that likely/unlikely hints currently wait for stabilization but can be shimmed in the kernel crate and a GitHub tracking issue already… | ||
| Installing and Using the Linux-Kernel Memory Model (LKMM) | Boqun Feng | 27m |
Explains how to install LKMM from the kernel tree (use tools/memory-model and the distro package for a faster path), run lit tests and the verifier’s eval… | ||
| COCONUT-SVSM BoF | Joerg Roedel | 37m |
COCONUT‑SVSM now ships regular development releases, with VTPM usable on SEV SNP and much improved interrupt handling, and Fedora already packages the dev… | ||
| Extending Flamegraphs for Multi-Dimensional Performance Analysis | Gavin Guo | 17m |
Claims ability to generate flamegraphs from latency, memory, crashdump and kern.log traces and shows concrete results: a memory framegraph that captures an… | ||
| Kernel documentation: update and future directions | Jonathan Corbet | 41m |
Documentation build time drops once a recent Sphinx update is used; PDF output still takes ~30 minutes and drags in thousands of LaTeX packages because the… | ||
| Tooling and Sharing Traceability Discussion | Luigi Pellecchia, Matthew Whitehead, Tobias Deiminger | 07m |
Discussion focuses on replacing brittle line contains and long regex rules with syntax-aware approaches using tree-sitter or AST parsing so tools can find… | ||
| Safe Systems with Linux MC Wrap up and next steps | Kate Stewart, Philipp Ahmann | 02m |
Group recap stresses concrete next steps: build tooling to identify code snippets and link tests to code, add machine-readable metadata so each change has… | ||
| Hardware in the loop pipelines for testing - for everyone! | Vipul Gupta | 22m |
A Hardware‑in‑the‑Loop pipeline that runs tests on real DUTs and ties into code review and CI systems is described, with deployments automating releases for… | ||
| Improving fw_devlink dependency tracking | Saravana Kannan | 22m |
Current fw_devlink parses every device tree property with a dedicated parser in drivers/of/property (around 40 today) and that won't scale to hundreds, so the… | ||
| To online or not to online CXL memory | Hannes Reinecke, Gregory Price | 28m |
Kernel currently requires per-block sysfs writes to change CXL memory type, forcing hundreds-to-thousands of writes (e.g. 640+ blocks) with per-write error… | ||
| nothing matches | ||