s390/mm: handle PTE-mapped tail pages in fast gup

With the THP refcounting rework it is possible to see THP compound tail
pages mapped with PTEs during a THP split. This needs to be considered
when using page_cache_get_speculative(), which will always fail on tail
pages because ->_count is always zero. commit 7aef4172 "mm: handle
PTE-mapped tail pages in gerneric fast gup implementaiton" fixed it for
the generic fast gup code by using compound_head(page) instead of page,
but not for s390.

This patch is a 1:1 adaption of commit 7aef4172 for the s390 fast gup
code. Without this fix, gup will fall back to the slow path or fail
in the unlikely scenario that we hit a THP under splitting in-between
the page table split and the compound page split.

Cc: stable@vger.kernel.org # v4.5
Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/mm/gup.c

@@ -20,9 +20,9 @@
 static inline int gup_pte_range(pmd_t *pmdp, pmd_t pmd, unsigned long addr,
 		unsigned long end, int write, struct page **pages, int *nr)
 {
+	struct page *head, *page;
 	unsigned long mask;
 	pte_t *ptep, pte;
-	struct page *page;
 
 	mask = (write ? _PAGE_PROTECT : 0) | _PAGE_INVALID | _PAGE_SPECIAL;
 
@@ -37,12 +37,14 @@ static inline int gup_pte_range(pmd_t *pmdp, pmd_t pmd, unsigned long addr,
 			return 0;
 		VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
 		page = pte_page(pte);
-		if (!page_cache_get_speculative(page))
+		head = compound_head(page);
+		if (!page_cache_get_speculative(head))
 			return 0;
 		if (unlikely(pte_val(pte) != pte_val(*ptep))) {
-			put_page(page);
+			put_page(head);
 			return 0;
 		}
+		VM_BUG_ON_PAGE(compound_head(page) != head, page);
 		pages[*nr] = page;
 		(*nr)++;