mm: Fix mremap not considering huge pmd devmap

The original code in mm/mremap.c checks huge pmd by:

		if (is_swap_pmd(*old_pmd) || pmd_trans_huge(*old_pmd)) {

However, a DAX mapped nvdimm is mapped as huge page (by default) but it
is not transparent huge page (_PAGE_PSE | PAGE_DEVMAP).  This commit
changes the condition to include the case.

This addresses CVE-2020-10757.

Fixes: 5c7fb56e ("mm, dax: dax-pmd vs thp-pmd vs hugetlbfs-pmd")
Cc: <stable@vger.kernel.org>
Reported-by: NFan Yang <Fan_Yang@sjtu.edu.cn>
Signed-off-by: NFan Yang <Fan_Yang@sjtu.edu.cn>
Tested-by: NFan Yang <Fan_Yang@sjtu.edu.cn>
Tested-by: NDan Williams <dan.j.williams@intel.com>
Reviewed-by: NDan Williams <dan.j.williams@intel.com>
Acked-by: NKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

arch/x86/include/asm/pgtable.h

@@ -257,6 +257,7 @@ static inline int pmd_large(pmd_t pte)
 }
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
+/* NOTE: when predicate huge page, consider also pmd_devmap, or use pmd_large */
 static inline int pmd_trans_huge(pmd_t pmd)
 {
 	return (pmd_val(pmd) & (_PAGE_PSE|_PAGE_DEVMAP)) == _PAGE_PSE;

mm/mremap.c

@@ -266,7 +266,7 @@ unsigned long move_page_tables(struct vm_area_struct *vma,
 		new_pmd = alloc_new_pmd(vma->vm_mm, vma, new_addr);
 		if (!new_pmd)
 			break;
-		if (is_swap_pmd(*old_pmd) || pmd_trans_huge(*old_pmd)) {
+		if (is_swap_pmd(*old_pmd) || pmd_trans_huge(*old_pmd) || pmd_devmap(*old_pmd)) {
 			if (extent == HPAGE_PMD_SIZE) {
 				bool moved;
 				/* See comment in move_ptes() */