Fewer bug reports than I was expecting from enabling large folios.
One that doesn't show up on x86 but does on arm64, one that shows up
with hugetlbfs memory failure testing and one that shows up with page
migration, which it turns out I wasn't testing because my last NUMA
machine died. Need to set up a qemu fake NUMA machine so I don't skip
testing that in future.
The following changes since commit 3e732ebf7316ac83e8562db7e64cc68aec390a18:
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost (2022-04-05 10:40:52 -0700)
are available in the Git repository at:
git://git.infradead.org/users/willy/pagecache.git tags/folio-5.18e
for you to fetch changes up to 98ea02597b9967c0817d29fee2f96d21b9e59ca5:
mm/rmap: Fix handling of hugetlbfs pages in page_vma_mapped_walk (2022-04-07 10:11:20 -0400)
----------------------------------------------------------------
Three folio-related fixes for 5.18:
- Remove the migration code's assumptions about large pages being PMD sized
- Don't call pmd_page() on a non-leaf PMD
- Fix handling of hugetlbfs pages in page_vma_mapped_walk
----------------------------------------------------------------
Matthew Wilcox (Oracle) (5):
mm/huge_memory: Avoid calling pmd_page() on a non-leaf PMD
mm/migrate: Use a folio in alloc_migration_target()
mm/migrate: Use a folio in migrate_misplaced_transhuge_page()
mm: Add vma_alloc_folio()
mm/mempolicy: Use vma_alloc_folio() in new_page()
zhenwei pi (1):
mm/rmap: Fix handling of hugetlbfs pages in page_vma_mapped_walk
include/linux/gfp.h | 8 ++++--
mm/huge_memory.c | 11 ++++----
mm/mempolicy.c | 38 ++++++++++++++++---------
mm/migrate.c | 80 +++++++++++++++++-----------------------------------
mm/page_vma_mapped.c | 6 ++--
5 files changed, 64 insertions(+), 79 deletions(-)
