An Insight into the Advanced XIP Filesystem (AXFS)
summary
AXFS implements page-granular XIP plus per-page compression and uses an in‑kernel profiling pass that logs which pages become executable, feeding an image builder that marks those pages XIP. Images can do device spanning so NOR holds XIP pages and NAND holds compressed pages. A simple instrumentation tool emits a linker-script locality tool that clusters hot functions. On an ARMv6 board with a 500‑function synthetic workload (~1.6 KB each) XIP images show the fastest launch and smallest RAM footprint, compressed images show the smallest on‑flash size.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 0:30Background
- 1:10Embedded Systems
- 2:31Overview
- 3:49XIP Diagram
- 4:32Advantages
- 5:25Disadvantages
- 7:31Device Spanning
- 8:58XIP Profiling
- 11:15AXFS By Tables
- 12:57Code Snippet
- 14:02Format
- 15:33Regions
- 17:23Superblock Structure
- 17:37Regions XFS Images
- 18:41Node Index Region
- 20:57C Block
- 22:24Index
- 22:45AXFS Performance
- 23:35AXFS Performance Parameters
- 24:17Improving Code Locality
- 25:17Improving Page Faults
- 27:07XFS Performance
- 28:16Timing Measurement
- 31:06Total Size Measurement
- 32:32System Ram Footprint
- 33:54XFS RAM Footprint
- 34:22AXFS Key Points
related talks
43mEdge Computing with RISC-V Platforms Running XIP Linux
31mRunning Linux on Constrained IOT Device
53mIntel ME: Flash File System Explained
48mXen and the Art of Embedded Systems Virtualization
20mACES: Automatic Compartments for Embedded Systems
33mNext Generation Xen Architecture for Automotive
14mModernizing File System through In-Storage Indexing
47mHow To Dump, Parse, And Analyze i.MX Flash Memory Chips
