On Nov 8, 2022, at 11:41 AM, Linus Torvalds [off-list ref] wrote:

When we remove a page table entry, we are very careful to only free the
page after we have flushed the TLB, because other CPUs could still be
using the page through stale TLB entries until after the flush.

The patches (all 4) look fine to me.

I mean there are minor issues here and there, like s390’s tlb_flush_rmaps()
that can have VM_WARN_ON(1);the generic tlb_flush_rmaps() that is missing an
empty line after the ‘page' variable definition; or perhaps using __bitwise
for sparse (as David pointed) — but it can all be addressed later.