Preserving guest memory across kexec
summary
Proposes carving a large persistent PFN pool out of host RAM, mounting a new muse filesystem backed by that memory, and letting a user-space control process program page mappings, handle on‑demand faults for overcommit, and replay mappings after kexec so guest RAM survives a kernel upgrade. Contrasts that with PKRAM and Xen breadcrumbs, discusses using DAX-like backing and implementing swap and MMIO/page‑table persistence in user space; idea is conceptual, not yet implemented.
AI-written summary · about
chapters (6)
- 0:00Introduction to live updates
- 3:31Memory management requirements
- 8:06Kernel vs user space solutions
- 11:42Muse file system proposal
- 17:04Comparing with Xen breadcrumbs
- 21:14Q&A and closing remarks
related talks
43mpkernfs: Persisting guest memory and kernel/device state safely...
11mAdvanced Parallel Memory Virtualization
21mSupporting guest private memory in Protected KVM on Android
48mPersistent Memory Usage within Linux Environment
26mLet's demystify memory
25mShared CXL 3 memory what will be required
19mP2CACHE: Exploring Tiered Memory for In-Kernel File Systems Caching
27mWhen kdump is way too much