[00/11] mm: page_vma_mapped_walk() cleanup and THP fixes

Message

Hugh Dickins June 10, 2021, 6:31 a.m. UTC

Here is the promised splitup of what I posted a few days ago as
5/7 mm/thp: fix page_vma_mapped_walk() if huge page mapped by ptes.

These are against 5.13-rc5 plus yesterday's mm/thp series:
https://lore.kernel.org/linux-mm/[e-mail]/T/#u
(sorry about the misthreading, I missed linux-mm and lkml at first).
03/10 and 04/10 of that series also touched mm/page_vma_mapped.c.

I've marked all of these for stable: many are merely cleanups,
but I think they are much better before the main fix than after.

01/11 makes the opposite cleanup to the earlier 5/7: that preferred
to use pvmw->page, this prefers to use page - it could go either way.
11/11 is a new fix that I noticed while splitting up.

01/11 mm: page_vma_mapped_walk(): use page for pvmw->page
02/11 mm: page_vma_mapped_walk(): settle PageHuge on entry
03/11 mm: page_vma_mapped_walk(): use pmd_read_atomic()
04/11 mm: page_vma_mapped_walk(): use pmde for *pvmw->pmd
05/11 mm: page_vma_mapped_walk(): prettify PVMW_MIGRATION block
06/11 mm: page_vma_mapped_walk(): crossing page table boundary
07/11 mm: page_vma_mapped_walk(): add a level of indentation
08/11 mm: page_vma_mapped_walk(): use goto instead of while (1)
09/11 mm: page_vma_mapped_walk(): get vma_address_end() earlier
10/11 mm/thp: fix page_vma_mapped_walk() if THP mapped by ptes
11/11 mm/thp: another PVMW_SYNC fix in page_vma_mapped_walk()

 mm/page_vma_mapped.c |  160 +++++++++++++++++++++++++++------------------
 1 file changed, 96 insertions(+), 64 deletions(-)

Hugh