[PATCH v2 10/32] mm/pagewalk: walk_pte_range() allow for pte_offset_map()

[Hugh Dickins <hughd@xxxxxxxxxx>]

walk_pte_range() has a no_vma option to serve walk_page_range_novma().
I don't know of any problem, but it looks safer to check for init_mm,
and use pte_offset_kernel() rather than pte_offset_map() in that case:
pte_offset_map()'s pmdval validation is intended for userspace.

Allow for its pte_offset_map() or pte_offset_map_lock() to fail, and
retry with ACTION_AGAIN if so.  Add a second check for ACTION_AGAIN
in walk_pmd_range(), to catch it after return from walk_pte_range().

Remove the pmd_trans_unstable() check after split_huge_pmd() in
walk_pmd_range(): walk_pte_range() now handles those cases safely
(and they must fail powerpc's is_hugepd() check).

Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
---
 mm/pagewalk.c | 33 +++++++++++++++++++++++----------
 1 file changed, 23 insertions(+), 10 deletions(-)

diff --git a/mm/pagewalk.c b/mm/pagewalk.c
index cb23f8a15c13..64437105fe0d 100644
--- a/mm/pagewalk.c
+++ b/mm/pagewalk.c
@@ -46,15 +46,27 @@ static int walk_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end,
 	spinlock_t *ptl;
 
 	if (walk->no_vma) {
-		pte = pte_offset_map(pmd, addr);
-		err = walk_pte_range_inner(pte, addr, end, walk);
-		pte_unmap(pte);
+		/*
+		 * pte_offset_map() might apply user-specific validation.
+		 */
+		if (walk->mm == &init_mm)
+			pte = pte_offset_kernel(pmd, addr);
+		else
+			pte = pte_offset_map(pmd, addr);
+		if (pte) {
+			err = walk_pte_range_inner(pte, addr, end, walk);
+			if (walk->mm != &init_mm)
+				pte_unmap(pte);
+		}
 	} else {
 		pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
-		err = walk_pte_range_inner(pte, addr, end, walk);
-		pte_unmap_unlock(pte, ptl);
+		if (pte) {
+			err = walk_pte_range_inner(pte, addr, end, walk);
+			pte_unmap_unlock(pte, ptl);
+		}
 	}
-
+	if (!pte)
+		walk->action = ACTION_AGAIN;
 	return err;
 }
 
@@ -141,11 +153,8 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
 		    !(ops->pte_entry))
 			continue;
 
-		if (walk->vma) {
+		if (walk->vma)
 			split_huge_pmd(walk->vma, pmd, addr);
-			if (pmd_trans_unstable(pmd))
-				goto again;
-		}
 
 		if (is_hugepd(__hugepd(pmd_val(*pmd))))
 			err = walk_hugepd_range((hugepd_t *)pmd, addr, next, walk, PMD_SHIFT);
@@ -153,6 +162,10 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end,
 			err = walk_pte_range(pmd, addr, next, walk);
 		if (err)
 			break;
+
+		if (walk->action == ACTION_AGAIN)
+			goto again;
+
 	} while (pmd++, addr = next, addr != end);
 
 	return err;
-- 
2.35.3