diff --git a/arch/mips/mm/gup.c b/arch/mips/mm/gup.c
index d8c3c15..71a19d2 100644
--- a/arch/mips/mm/gup.c
+++ b/arch/mips/mm/gup.c
@@ -271,7 +271,14 @@
 		next = pgd_addr_end(addr, end);
 		if (pgd_none(pgd))
 			goto slow;
-		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
+		/*
+		 * The FAST_GUP case requires FOLL_WRITE even for pure reads,
+		 * because get_user_pages() may need to cause an early COW in
+		 * order to avoid confusing the normal COW routines. So only
+		 * targets that are already writable are safe to do by just
+		 * looking at the page tables.
+		 */
+		if (!gup_pud_range(pgd, addr, next, 1, pages, &nr))
 			goto slow;
 	} while (pgdp++, addr = next, addr != end);
 	local_irq_enable();
diff --git a/arch/s390/mm/gup.c b/arch/s390/mm/gup.c
index cf045f5..be1e2ed 100644
--- a/arch/s390/mm/gup.c
+++ b/arch/s390/mm/gup.c
@@ -261,7 +261,14 @@
 
 	might_sleep();
 	start &= PAGE_MASK;
-	nr = __get_user_pages_fast(start, nr_pages, write, pages);
+	/*
+	 * The FAST_GUP case requires FOLL_WRITE even for pure reads,
+	 * because get_user_pages() may need to cause an early COW in
+	 * order to avoid confusing the normal COW routines. So only
+	 * targets that are already writable are safe to do by just
+	 * looking at the page tables.
+	 */
+	nr = __get_user_pages_fast(start, nr_pages, 1, pages);
 	if (nr == nr_pages)
 		return nr;
 
diff --git a/arch/sh/mm/gup.c b/arch/sh/mm/gup.c
index 063c298..7fec66e 100644
--- a/arch/sh/mm/gup.c
+++ b/arch/sh/mm/gup.c
@@ -239,7 +239,14 @@
 		next = pgd_addr_end(addr, end);
 		if (pgd_none(pgd))
 			goto slow;
-		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
+		/*
+		 * The FAST_GUP case requires FOLL_WRITE even for pure reads,
+		 * because get_user_pages() may need to cause an early COW in
+		 * order to avoid confusing the normal COW routines. So only
+		 * targets that are already writable are safe to do by just
+		 * looking at the page tables.
+		 */
+		if (!gup_pud_range(pgd, addr, next, 1, pages, &nr))
 			goto slow;
 	} while (pgdp++, addr = next, addr != end);
 	local_irq_enable();
diff --git a/arch/sparc/mm/gup.c b/arch/sparc/mm/gup.c
index cd0e32b..685679f 100644
--- a/arch/sparc/mm/gup.c
+++ b/arch/sparc/mm/gup.c
@@ -218,7 +218,14 @@
 		next = pgd_addr_end(addr, end);
 		if (pgd_none(pgd))
 			goto slow;
-		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
+		/*
+		 * The FAST_GUP case requires FOLL_WRITE even for pure reads,
+		 * because get_user_pages() may need to cause an early COW in
+		 * order to avoid confusing the normal COW routines. So only
+		 * targets that are already writable are safe to do by just
+		 * looking at the page tables.
+		 */
+		if (!gup_pud_range(pgd, addr, next, 1, pages, &nr))
 			goto slow;
 	} while (pgdp++, addr = next, addr != end);
 
diff --git a/arch/x86/mm/gup.c b/arch/x86/mm/gup.c
index 82f727f..549f89f 100644
--- a/arch/x86/mm/gup.c
+++ b/arch/x86/mm/gup.c
@@ -454,7 +454,14 @@
 		next = pgd_addr_end(addr, end);
 		if (pgd_none(pgd))
 			goto slow;
-		if (!gup_pud_range(pgd, addr, next, write, pages, &nr))
+		/*
+		 * The FAST_GUP case requires FOLL_WRITE even for pure reads,
+		 * because get_user_pages() may need to cause an early COW in
+		 * order to avoid confusing the normal COW routines. So only
+		 * targets that are already writable are safe to do by just
+		 * looking at the page tables.
+		 */
+		if (!gup_pud_range(pgd, addr, next, 1, pages, &nr))
 			goto slow;
 	} while (pgdp++, addr = next, addr != end);
 	local_irq_enable();
diff --git a/mm/gup.c b/mm/gup.c
index 6bb7a8e..0b80bf3 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -61,13 +61,22 @@
 }
 
 /*
- * FOLL_FORCE can write to even unwritable pte's, but only
- * after we've gone through a COW cycle and they are dirty.
+ * FOLL_FORCE or a forced COW break can write even to unwritable pte's,
+ * but only after we've gone through a COW cycle and they are dirty.
  */
 static inline bool can_follow_write_pte(pte_t pte, unsigned int flags)
 {
-	return pte_write(pte) ||
-		((flags & FOLL_FORCE) && (flags & FOLL_COW) && pte_dirty(pte));
+	return pte_write(pte) || ((flags & FOLL_COW) && pte_dirty(pte));
+}
+
+/*
+ * A (separate) COW fault might break the page the other way and
+ * get_user_pages() would return the page from what is now the wrong
+ * VM. So we need to force a COW break at GUP time even for reads.
+ */
+static inline bool should_force_cow_break(struct vm_area_struct *vma, unsigned int flags)
+{
+	return is_cow_mapping(vma->vm_flags) && (flags & FOLL_GET);
 }
 
 static struct page *follow_page_pte(struct vm_area_struct *vma,
@@ -577,12 +586,18 @@
 			if (!vma || check_vma_flags(vma, gup_flags))
 				return i ? : -EFAULT;
 			if (is_vm_hugetlb_page(vma)) {
+				if (should_force_cow_break(vma, foll_flags))
+					foll_flags |= FOLL_WRITE;
 				i = follow_hugetlb_page(mm, vma, pages, vmas,
 						&start, &nr_pages, i,
-						gup_flags);
+						foll_flags);
 				continue;
 			}
 		}
+
+		if (should_force_cow_break(vma, foll_flags))
+			foll_flags |= FOLL_WRITE;
+
 retry:
 		/*
 		 * If we have a pending SIGKILL, don't keep faulting pages and
@@ -1503,6 +1518,10 @@
 /*
  * Like get_user_pages_fast() except it's IRQ-safe in that it won't fall back to
  * the regular GUP. It will only return non-negative values.
+ *
+ * Careful, careful! COW breaking can go either way, so a non-write
+ * access can get ambiguous page results. If you call this function without
+ * 'write' set, you'd better be sure that you're ok with that ambiguity.
  */
 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
 			  struct page **pages)
@@ -1532,6 +1551,12 @@
 	 *
 	 * We do not adopt an rcu_read_lock(.) here as we also want to
 	 * block IPIs that come from THPs splitting.
+	 *
+	 * NOTE! We allow read-only gup_fast() here, but you'd better be
+	 * careful about possible COW pages. You'll get _a_ COW page, but
+	 * not necessarily the one you intended to get depending on what
+	 * COW event happens after this. COW may break the page copy in a
+	 * random direction.
 	 */
 
 	local_irq_save(flags);
@@ -1580,7 +1605,14 @@
 	int nr, ret;
 
 	start &= PAGE_MASK;
-	nr = __get_user_pages_fast(start, nr_pages, write, pages);
+	/*
+	 * The FAST_GUP case requires FOLL_WRITE even for pure reads,
+	 * because get_user_pages() may need to cause an early COW in
+	 * order to avoid confusing the normal COW routines. So only
+	 * targets that are already writable are safe to do by just
+	 * looking at the page tables.
+	 */
+	nr = __get_user_pages_fast(start, nr_pages, 1, pages);
 	ret = nr;
 
 	if (nr < nr_pages) {
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 1c84c5c..0fd4bc8 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -1137,13 +1137,12 @@
 }
 
 /*
- * FOLL_FORCE can write to even unwritable pmd's, but only
- * after we've gone through a COW cycle and they are dirty.
+ * FOLL_FORCE or a forced COW break can write even to unwritable pmd's,
+ * but only after we've gone through a COW cycle and they are dirty.
  */
 static inline bool can_follow_write_pmd(pmd_t pmd, unsigned int flags)
 {
-	return pmd_write(pmd) ||
-	       ((flags & FOLL_FORCE) && (flags & FOLL_COW) && pmd_dirty(pmd));
+	return pmd_write(pmd) || ((flags & FOLL_COW) && pmd_dirty(pmd));
 }
 
 struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,