Current Status and Future of HWPOISON Subsystem

This talk introduces how Linux’s HWPOISON subsystem handles memory errors and outlines recent and planned improvements, including huge-page support, page-cache handling, and memory hotplug integration.

Syllabus

Intro
Memory error background
Hardware Error
level handling for memory error
Terminology Confusion
HWPOISON basic concept
Action Required/Optional Memory Error
Example of hard page offline
Example of soft page offline
HWPOISON Internal: hard-offline
Notification to Userspace
Recent development topics
Soft offline rework
2. More topics on pagecache handling
Folio related issues
Huge) Zero page
Persistent memory
Hugetlb pinning race
Subpage hwpoison tracking
1GB hugetib page
Memory hotplug integration
Machine Check safe memory copy
Management interface
Summary