[PATCH v2 3/5] mm/hugetlb: fix races when looking up a CONT-PMD size hugetlb page

[Baolin Wang]

On some architectures (like ARM64), it can support CONT-PTE/PMD size
hugetlb, which means it can support not only PMD/PUD size hugetlb
(2M and 1G), but also CONT-PTE/PMD size(64K and 32M) if a 4K page size
specified.

When looking up a CONT-PMD size hugetlb page by follow_page(), it will
always use the PMD page lock to protect the pmd entry in follow_huge_pmd().
However this is not the correct lock for CONT-PMD size hugetlb, and the
pmd entry will be unstable under the incorrect lock, which means it still
can be migrated or poisoned, and can not get the correct CONT-PMD size
page.

Thus changing to use huge_pte_lock() to get the correct pmd entry lock
for CONT-PMD size hugetlb to fix the potential race.

Signed-off-by: Baolin Wang <baolin.wang@xxxxxxxxxxxxxxxxx>
---
 include/linux/hugetlb.h | 4 ++--
 mm/gup.c                | 2 +-
 mm/hugetlb.c            | 7 ++++---
 3 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 4b172a7..3a96f67 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -209,7 +209,7 @@ struct page *follow_huge_pd(struct vm_area_struct *vma,
 			    int flags, int pdshift);
 struct page *follow_huge_pte(struct vm_area_struct *vma, unsigned long address,
 			     pmd_t *pmd, int flags);
-struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
+struct page *follow_huge_pmd(struct vm_area_struct *vma, unsigned long address,
 				pmd_t *pmd, int flags);
 struct page *follow_huge_pud(struct mm_struct *mm, unsigned long address,
 				pud_t *pud, int flags);
@@ -320,7 +320,7 @@ static inline struct page *follow_huge_pte(struct vm_area_struct *vma,
 	return NULL;
 }
 
-static inline struct page *follow_huge_pmd(struct mm_struct *mm,
+static inline struct page *follow_huge_pmd(struct vm_area_struct *vma,
 				unsigned long address, pmd_t *pmd, int flags)
 {
 	return NULL;
diff --git a/mm/gup.c b/mm/gup.c
index 87a94f5..014accd 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -673,7 +673,7 @@ static struct page *follow_pmd_mask(struct vm_area_struct *vma,
 	if (pmd_none(pmdval))
 		return no_page_table(vma, flags);
 	if (pmd_huge(pmdval) && is_vm_hugetlb_page(vma)) {
-		page = follow_huge_pmd(mm, address, pmd, flags);
+		page = follow_huge_pmd(vma, address, pmd, flags);
 		if (page)
 			return page;
 		return no_page_table(vma, flags);
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index cf742d1..2c4048a 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -7035,9 +7035,11 @@ struct page * __weak
 }
 
 struct page * __weak
-follow_huge_pmd(struct mm_struct *mm, unsigned long address,
+follow_huge_pmd(struct vm_area_struct *vma, unsigned long address,
 		pmd_t *pmd, int flags)
 {
+	struct mm_struct *mm = vma->vm_mm;
+	struct hstate *hstate = hstate_vma(vma);
 	struct page *page = NULL;
 	spinlock_t *ptl;
 	pte_t pte;
@@ -7050,8 +7052,7 @@ struct page * __weak
 		return NULL;
 
 retry:
-	ptl = pmd_lockptr(mm, pmd);
-	spin_lock(ptl);
+	ptl = huge_pte_lock(hstate, mm, (pte_t *)pmd);
 	/*
 	 * make sure that the address range covered by this pmd is not
 	 * unmapped from other threads.
-- 
1.8.3.1