I forgot to hit "record", so unfortunately there's no video evidence this time
around :-(
LPC:
- Format will utilize flexible timeslots; tentatively planning on ~3 "big"
slots of ~40 minutes each, and ~6 "small" slots of 10-15 minutes each.
- Schedule will be announced around September 15th (though that may get pushed
backed to October 1st).
- Schedule is flexibile up until the conference, e.g. if code is merged between
September 15th and early November and obviates the need for a discussion.
guest_memfd:
Topic Branch:
- Work will be coordinated through a dedicated topic branch (likely
kvm-x86/guest_memfd).
- Mostly stable branch; plan is to not rebase it for at least the next month
- Fixes (use Fixes: tag!!!) will be applied on top and squashed when the
branch is rebased.
- TBD will post a series for broader review in a month or so once
development has settled down.
TODO:
- Documentation [Sean?]
- Changelogs [Sean]
- Non-filemap implementation [Paolo]
- Hugepage tests (verify KVM actually installs hugepages as expected)
- Memory failure tests (e.g. synthetic #MC injection)
- Get Input from IOMMU / VFIO [Paolo]
Opens:
- .release_folio and .invalidate_folio versus .evict_inode [Mike]
- Fully anonymous inode vs. proper filesystem [Sean / Paolo]
- HugeTLB (or something similar) support [Future]
- Intrahost migration support [Google GCE]
- NUMA support [ TBD - implement fbind() ]
- Memory Accounting [Sean - should Just Work, keyword should ]
- Page Table Accounting - Should we? [Nikunj]
Recording:
/facepalm
