[PATCH v2 06/12] mm/hugetlb: make hugetlb migration target allocation APIs CMA aware

[js1304@gmail.com]

From: Joonsoo Kim <iamjoonsoo.kim@lge.com>

There is a user who do not want to use CMA memory for migration. Until
now, it is implemented by caller side but it's not optimal since there
is limited information on caller. This patch implements it on callee side
to get better result.

Acked-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
---
 include/linux/hugetlb.h |  2 --
 mm/gup.c                |  9 +++------
 mm/hugetlb.c            | 21 +++++++++++++++++----
 mm/internal.h           |  1 +
 4 files changed, 21 insertions(+), 12 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index f482563..3d05f7d 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -503,8 +503,6 @@ struct huge_bootmem_page {
 	struct hstate *hstate;
 };
 
-struct page *alloc_migrate_huge_page(struct hstate *h,
-				struct alloc_control *ac);
 struct page *alloc_huge_page_nodemask(struct hstate *h,
 				struct alloc_control *ac);
 struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
diff --git a/mm/gup.c b/mm/gup.c
index 6b78f11..87eca79 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -1617,14 +1617,11 @@ static struct page *new_non_cma_page(struct page *page, unsigned long private)
 		struct alloc_control ac = {
 			.nid = nid,
 			.nmask = NULL,
-			.gfp_mask = gfp_mask,
+			.gfp_mask = __GFP_NOWARN,
+			.skip_cma = true,
 		};
 
-		/*
-		 * We don't want to dequeue from the pool because pool pages will
-		 * mostly be from the CMA region.
-		 */
-		return alloc_migrate_huge_page(h, &ac);
+		return alloc_huge_page_nodemask(h, &ac);
 	}
 
 	if (PageTransHuge(page)) {
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 8132985..e465582 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1033,13 +1033,19 @@ static void enqueue_huge_page(struct hstate *h, struct page *page)
 	h->free_huge_pages_node[nid]++;
 }
 
-static struct page *dequeue_huge_page_node_exact(struct hstate *h, int nid)
+static struct page *dequeue_huge_page_node_exact(struct hstate *h,
+						int nid, bool skip_cma)
 {
 	struct page *page;
 
-	list_for_each_entry(page, &h->hugepage_freelists[nid], lru)
+	list_for_each_entry(page, &h->hugepage_freelists[nid], lru) {
+		if (skip_cma && is_migrate_cma_page(page))
+			continue;
+
 		if (!PageHWPoison(page))
 			break;
+	}
+
 	/*
 	 * if 'non-isolated free hugepage' not found on the list,
 	 * the allocation fails.
@@ -1080,7 +1086,7 @@ static struct page *dequeue_huge_page_nodemask(struct hstate *h,
 			continue;
 		node = zone_to_nid(zone);
 
-		page = dequeue_huge_page_node_exact(h, node);
+		page = dequeue_huge_page_node_exact(h, node, ac->skip_cma);
 		if (page)
 			return page;
 	}
@@ -1937,7 +1943,7 @@ static struct page *alloc_surplus_huge_page(struct hstate *h, gfp_t gfp_mask,
 	return page;
 }
 
-struct page *alloc_migrate_huge_page(struct hstate *h,
+static struct page *alloc_migrate_huge_page(struct hstate *h,
 				struct alloc_control *ac)
 {
 	struct page *page;
@@ -1999,6 +2005,13 @@ struct page *alloc_huge_page_nodemask(struct hstate *h,
 	}
 	spin_unlock(&hugetlb_lock);
 
+	/*
+	 * clearing __GFP_MOVABLE flag ensure that allocated page
+	 * will not come from CMA area
+	 */
+	if (ac->skip_cma)
+		ac->gfp_mask &= ~__GFP_MOVABLE;
+
 	return alloc_migrate_huge_page(h, ac);
 }
 
diff --git a/mm/internal.h b/mm/internal.h
index 6e613ce..159cfd6 100644
--- a/mm/internal.h
+++ b/mm/internal.h
@@ -618,6 +618,7 @@ struct alloc_control {
 	int nid;		/* preferred node id */
 	nodemask_t *nmask;
 	gfp_t gfp_mask;
+	bool skip_cma;
 };
 
 #endif	/* __MM_INTERNAL_H */
-- 
2.7.4