Sharing Linux kernel-related cutting-edge technology, technical articles, technical news, and curated tutorials
The article walks through the Flash Friendly File System’s on‑disk layout—zones, sections, segments, blocks, SuperBlock, checkpoint, SIT, NAT, and SSA—showing how formatting, mounting, creating, modifying, and deleting files updates these structures, offering a practical guide for kernel and storage engineers.
The Linux kernel’s extensible scheduler class sched_ext introduces a new ext_sched_class with eBPF‑driven callbacks and dispatch queues, allowing developers to plug custom scheduling policies via struct sched_ext_ops without recompiling the kernel, while integrating into the existing hierarchy and exposing trade‑offs such as central‑CPU load and community adoption challenges.
The article examines Linux 6.1’s preemption mechanism, explaining latency sources, the three preemption configurations (none, voluntary, full), the TIF_NEED_RESCHED flag and preempt_count tracking, and how preempt_enable/disable affect real‑time responsiveness, illustrated by a case where RT threads cannot preempt CFS due to disabled preemption in critical driver code.
V4L2 is the Linux kernel framework that unifies video device access, abstracts complex multi‑IC hardware into sub‑devices, and provides core, platform, and sensor driver layers, while supporting capture, output, and M2M codec operations with sophisticated buffer management—including mmap, userptr, and DMA‑Buf for zero‑copy processing on Android.
Large folios in the Linux kernel combine multiple pages to reduce TLB misses, page faults, and reclamation cost while enabling more efficient compression; they are supported by filesystems like XFS and bcachefs, and recent patches add multi‑size THP, swap‑in/out handling, TAO allocation, NUMA balancing, and debug tools, with OPPO’s production deployment showing performance gains and motivating broader adoption and fragmentation mitigation.
The article explains how AArch64 stack unwinding works using both frame‑pointer based methods and DWARF .eh_frame/.debug_frame CFI data, detailing the calling convention, compiler effects, CIE/FDE structures, and their integration into Android’s simpleperf profiler for accurate back‑traces.
The guide walks through fine‑tuning InternLM‑Chat‑7B with XTuner, showing how to build a DataLoader from a HuggingFace Dataset, convert a LoRA .pth checkpoint to HuggingFace format, merge the adapter into the base model, run inference, and adapt the process for custom datasets and 4‑bit quantization experiments.
This tutorial walks through fine‑tuning Shanghai AI Lab’s open‑source InternLM models with XTuner, explaining chat‑format conventions, loading and inference (including multimodal InternLM‑XComposer), dataset preparation, configuration sections, DeepSpeed acceleration, and memory‑efficient QLoRA details for 7‑B‑parameter chat models.
From legacy spinning-disk optimizations to modern SSD-focused QoS, Linux’s I/O scheduler landscape has evolved through noop, deadline, cfq and now multiqueue designs such as Kyber, MQ-Deadline, and BFQ, each employing distinct latency, deadline, and budget-fairness algorithms, supporting cgroup/ionice priorities, and complemented by numerous experimental out-of-tree implementations.
Using Qualcomm’s Adreno GPU as a case study, the article explains how the Linux devfreq framework enables GPU frequency scaling by creating a kgsl devfreq device and an msm‑adreno‑tz governor, detailing their initialization, event handling, target‑frequency computation, and the kernel callbacks that apply the new rates.
