Posts

A deep dive into Linux Page Cache — how VFS, inode, and address_space organize cached file content, how Buffered I/O and mmap create and release cache pages, the four mmap mapping types, and the tools to observe cache hit rates.

Why modern multi-socket systems moved from UMA to NUMA — the FSB bottleneck, per-socket memory controllers, local vs remote access, and how Linux handles NUMA (plus the classic MySQL swap-insanity problem fixed by interleaving).

Zero-copy techniques in Linux — starting from the traditional read/write path (4 context switches, 2 DMA + 2 CPU copies) and walking through Direct I/O, mmap, sendfile, DMA gather, splice, and COW.

How Linux organizes physical memory — comparing FLATMEM, DISCONTIGMEM, and SPARSEMEM, and how each model implements pfn_to_page / page_to_pfn.

Memory consistency models define which reorderings are legal. A walk through SC, TSO (x86), PSO, and relaxed models (ARM / POWER), plus the memory barriers used to enforce ordering when it matters.

How multi-core CPUs keep a consistent view of memory — the MESI protocol, bus snooping, cache-to-cache transfers, and the store-buffer / invalid-queue optimizations (plus the memory-ordering headaches they introduce).

An introduction to Linux Kernel Library (LKL) and User Interrupts (UINTR) — how LKL runs the Linux kernel as a library for userspace applications, and how UINTR enables low-latency cross-process notifications without syscall overhead.

Welcome to my blog — a place to collect notes on operating systems, virtualization, agents, and whatever else I end up learning.