[PATCH] mm/hugetlb: add support for mempolicy MPOL_PREFERRED_MANY

[Ben Widawsky]

Implement the missing huge page allocation functionality while obeying the
preferred node semantics.  This is similar to the implementation for
general page allocation, as it uses a fallback mechanism to try multiple
preferred nodes first, and then all other nodes. 

To avoid adding too many "#ifdef CONFIG_NUMA" check, add a helper function
in mempolicy.h to check whether a mempolicy is MPOL_PREFERRED_MANY.

[akpm: fix compling issue when merging with other hugetlb patch]
[Thanks to 0day bot for catching the !CONFIG_NUMA compiling issue]
[Michal Hocko: suggest to remove the #ifdef CONFIG_NUMA check]
Link: https://lore.kernel.org/r/20200630212517.308045-12-ben.widawsky@xxxxxxxxx
Link: https://lkml.kernel.org/r/1627970362-61305-4-git-send-email-feng.tang@xxxxxxxxx
Suggested-by: Michal Hocko <mhocko@xxxxxxxx>
Signed-off-by: Ben Widawsky <ben.widawsky@xxxxxxxxx>
Co-developed-by: Feng Tang <feng.tang@xxxxxxxxx>
Signed-off-by: Feng Tang <feng.tang@xxxxxxxxx>
--
 include/linux/mempolicy.h | 12 ++++++++++++
 mm/hugetlb.c              | 28 ++++++++++++++++++++++++----
 2 files changed, 36 insertions(+), 4 deletions(-)

diff --git a/include/linux/mempolicy.h b/include/linux/mempolicy.h
index 0117e1e..60d5e6c 100644
--- a/include/linux/mempolicy.h
+++ b/include/linux/mempolicy.h
@@ -187,6 +187,12 @@ extern void mpol_put_task_policy(struct task_struct *);
 
 extern bool numa_demotion_enabled;
 
+static inline bool mpol_is_preferred_many(struct mempolicy *pol)
+{
+	return  (pol->mode == MPOL_PREFERRED_MANY);
+}
+
+
 #else
 
 struct mempolicy {};
@@ -297,5 +303,11 @@ static inline nodemask_t *policy_nodemask_current(gfp_t gfp)
 }
 
 #define numa_demotion_enabled	false
+
+static inline bool mpol_is_preferred_many(struct mempolicy *pol)
+{
+	return  false;
+}
+
 #endif /* CONFIG_NUMA */
 #endif
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 95714fb..75ea8bc 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1145,7 +1145,7 @@ static struct page *dequeue_huge_page_vma(struct hstate *h,
 				unsigned long address, int avoid_reserve,
 				long chg)
 {
-	struct page *page;
+	struct page *page = NULL;
 	struct mempolicy *mpol;
 	gfp_t gfp_mask;
 	nodemask_t *nodemask;
@@ -1166,7 +1166,17 @@ static struct page *dequeue_huge_page_vma(struct hstate *h,
 
 	gfp_mask = htlb_alloc_mask(h);
 	nid = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
-	page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
+
+	if (mpol_is_preferred_many(mpol)) {
+		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
+
+		/* Fallback to all nodes if page==NULL */
+		nodemask = NULL;
+	}
+
+	if (!page)
+		page = dequeue_huge_page_nodemask(h, gfp_mask, nid, nodemask);
+
 	if (page && !avoid_reserve && vma_has_reserves(vma, chg)) {
 		SetHPageRestoreReserve(page);
 		h->resv_huge_pages--;
@@ -2147,9 +2157,19 @@ struct page *alloc_buddy_huge_page_with_mpol(struct hstate *h,
 	nodemask_t *nodemask;
 
 	nid = huge_node(vma, addr, gfp_mask, &mpol, &nodemask);
-	page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask, false);
-	mpol_cond_put(mpol);
+	if (mpol_is_preferred_many(mpol)) {
+		gfp_t gfp = gfp_mask | __GFP_NOWARN;
 
+		gfp &=  ~(__GFP_DIRECT_RECLAIM | __GFP_NOFAIL);
+		page = alloc_surplus_huge_page(h, gfp, nid, nodemask, false);
+
+		/* Fallback to all nodes if page==NULL */
+		nodemask = NULL;
+	}
+
+	if (!page)
+		page = alloc_surplus_huge_page(h, gfp_mask, nid, nodemask, false);
+	mpol_cond_put(mpol);
 	return page;
 }
 
-- 
2.7.4