1 views collected around this technical thread.
This article explains how Linux handles I/O operations, covering the virtual file system, inode and dentry structures, superblock layout, ZFS features, disk types, the generic block layer, I/O scheduling strategies, and key performance metrics for storage.
The article thoroughly examines the Linux SCSI subsystem architecture, illustrating how abstract hardware topology—hosts, channels, IDs, and LUNs—is modeled in software through structs such as scsi_host and scsi_device, and detailing initialization, scanning, path establishment, power management, and block‑layer integration to guide kernel storage driver design.
The Linux blk-mq framework replaces the legacy single-queue block layer with a two-queue architecture—per-CPU software queues and hardware dispatch queues—eliminating lock contention and interrupt overhead, pre-allocating request tags, and supporting modern multi-queue I/O schedulers to fully exploit high-IOPS SSD performance.
The Linux block layer, positioned between the VFS and device drivers, evolved from a single‑queue design to a multi‑queue architecture (kernel 3.13 onward) to reduce lock contention and exploit modern hardware queues, managing bio/request lifecycles, dispatching through layered queues, and employing various I/O schedulers for balanced throughput and latency.