Hugh Dickins wrote: > On Mon, 29 Jan 2018, Zi Yan wrote: >> From: Zi Yan <zi.yan@xxxxxxxxxxxxxx> >> >> migrate_pages() requires at least down_read(mmap_sem) to protect >> related page tables and VMAs from changing. Let's do it in > > Page tables are protected by their locks. VMAs may change while > migration is active on them, but does that need locking against? > >> do_page_moves() for both do_move_pages_to_node() and >> add_page_for_migration(). >> >> Also add this lock requirement in the comment of migrate_pages(). >> >> Signed-off-by: Zi Yan <zi.yan@xxxxxxxxxxxxxx> >> --- >> mm/migrate.c | 13 +++++++++++-- >> 1 file changed, 11 insertions(+), 2 deletions(-) >> >> diff --git a/mm/migrate.c b/mm/migrate.c >> index 5d0dc7b85f90..52d029953c32 100644 >> --- a/mm/migrate.c >> +++ b/mm/migrate.c >> @@ -1354,6 +1354,9 @@ static int unmap_and_move_huge_page(new_page_t get_new_page, >> * or free list only if ret != 0. >> * >> * Returns the number of pages that were not migrated, or an error code. >> + * >> + * The caller must hold at least down_read(mmap_sem) for to-be-migrated pages >> + * to protect related page tables and VMAs from changing. > > I have not been keeping up with Michal's recent migration changes, > but migrate_pages() never used to need mmap_sem held (despite being > called with an mmap_sem held from some of its callsites), and it > would be a backward step to require that now. > > There is not even an mm argument to migrate_pages(), so which > mm->mmap_sem do you think would be required for it? There may be > particular cases in which it is required (when the new_page function > involves the old_page's vma - is that so below?), but in general not. mmap_sem is held during migrate_pages() in current implementation. http://elixir.free-electrons.com/linux/latest/source/mm/migrate.c#L1576 -- Best Regards, Yan Zi
Attachment:
signature.asc
Description: OpenPGP digital signature
