mm/rmap: Fix handling of hugetlbfs pages in page_vma_mapped_walk

page_mapped_in_vma() sets nr_pages to 1, which is usually correct as we
only want to know about the precise page and not about other pages in
the folio.  However, hugetlbfs does want to know about the entire hpage,
and using nr_pages to get the size of the hpage is wrong.  We could
change page_mapped_in_vma() to special-case hugetlbfs pages, but it's
better to ignore nr_pages in page_vma_mapped_walk() and get the size
from the VMA instead.

Fixes: 2aff7a4755bed ("mm: Convert page_vma_mapped_walk to work on PFNs")
Signed-off-by: zhenwei pi <pizhenwei@bytedance.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
[edit commit message, use hstate directly]

diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
index 1187f9c1ec5b1048a2c6a8cbd2a0e6341f19d400..14a5cda73dee62759d5c3dc52a26f13f6e624e43 100644 (file)
--- a/mm/page_vma_mapped.c
+++ b/mm/page_vma_mapped.c
@@ -163,7 +163,8 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
                return not_found(pvmw);
 
        if (unlikely(is_vm_hugetlb_page(vma))) {
-               unsigned long size = pvmw->nr_pages * PAGE_SIZE;
+               struct hstate *hstate = hstate_vma(vma);
+               unsigned long size = huge_page_size(hstate);
                /* The only possible mapping was handled on last iteration */
                if (pvmw->pte)
                        return not_found(pvmw);
@@ -173,8 +174,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
                if (!pvmw->pte)
                        return false;
 
-               pvmw->ptl = huge_pte_lockptr(size_to_hstate(size), mm,
-                                               pvmw->pte);
+               pvmw->ptl = huge_pte_lockptr(hstate, mm, pvmw->pte);
                spin_lock(pvmw->ptl);
                if (!check_pte(pvmw))
                        return not_found(pvmw);