On Mon, 3 Dec 2018, Michal Hocko wrote: > I have merely said that a better THP locality needs more work and during > the review discussion I have even volunteered to work on that. There > are other reclaim related fixes under work right now. All I am saying > is that MADV_TRANSHUGE having numa locality implications cannot satisfy > all the usecases and it is particurarly KVM that suffers from it.
I think extending functionality so thp can be allocated remotely if truly desired is worthwhile just so long as it does not cause regressions for other users. I think that is separate from the swap storm regression that Andrea is reporting, however, since that would also exist even if we allowed remote thp allocations when the host is fully fragmented rather than only locally fragmented.
