[PATCH 4/4] mm: hugetlb: allocate frozen pages in alloc_gigantic_folio()

[Kefeng Wang <wangkefeng.wang@huawei.com>]

The alloc_gigantic_folio() allocates a folio by alloc_contig_range()
with refcount increated and then freeze it, convert to allocate a frozen
folio directly to remove the atomic operation about folio refcount, also
saving atomic operation during __update_and_free_hugetlb_folio too.

Rename some functions to make them more self-explanatory,

  folio_alloc_gigantic            -> folio_alloc_frozen_gigantic
  cma_{alloc,free}_folio          -> cma_{alloc,free}_frozen_folio
  hugetlb_cma_{alloc,free}_folio  -> hugetlb_cma_{alloc,free}_frozen_folio

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
---
 include/linux/cma.h |  8 ++++----
 include/linux/gfp.h | 15 +++++++-------
 mm/cma.c            | 34 +++++++++++++++++++-----------
 mm/hugetlb.c        | 50 +++++++++------------------------------------
 mm/hugetlb_cma.c    | 13 ++++++------
 mm/hugetlb_cma.h    | 10 ++++-----
 6 files changed, 55 insertions(+), 75 deletions(-)

diff --git a/include/linux/cma.h b/include/linux/cma.h
index e5745d2aec55..cceec7b25bae 100644
--- a/include/linux/cma.h
+++ b/include/linux/cma.h
@@ -57,16 +57,16 @@ extern bool cma_intersects(struct cma *cma, unsigned long start, unsigned long e
 extern void cma_reserve_pages_on_error(struct cma *cma);
 
 #ifdef CONFIG_CMA
-struct folio *cma_alloc_folio(struct cma *cma, int order, gfp_t gfp);
-bool cma_free_folio(struct cma *cma, const struct folio *folio);
+struct folio *cma_alloc_frozen_folio(struct cma *cma, int order, gfp_t gfp);
+bool cma_free_frozen_folio(struct cma *cma, const struct folio *folio);
 bool cma_validate_zones(struct cma *cma);
 #else
-static inline struct folio *cma_alloc_folio(struct cma *cma, int order, gfp_t gfp)
+static inline struct folio *cma_alloc_frozen_folio(struct cma *cma, int order, gfp_t gfp)
 {
 	return NULL;
 }
 
-static inline bool cma_free_folio(struct cma *cma, const struct folio *folio)
+static inline bool cma_free_frozen_folio(struct cma *cma, const struct folio *folio)
 {
 	return false;
 }
diff --git a/include/linux/gfp.h b/include/linux/gfp.h
index ed9445e6fe38..38a7be7792a2 100644
--- a/include/linux/gfp.h
+++ b/include/linux/gfp.h
@@ -454,26 +454,27 @@ void free_contig_range(unsigned long pfn, unsigned long nr_pages);
 #endif
 
 #ifdef CONFIG_CONTIG_ALLOC
-static inline struct folio *folio_alloc_gigantic_noprof(int order, gfp_t gfp,
-							int nid, nodemask_t *node)
+static inline struct folio *folio_alloc_frozen_gigantic_noprof(int order,
+		gfp_t gfp, int nid, nodemask_t *node)
 {
 	struct page *page;
 
 	if (WARN_ON(!order || !(gfp & __GFP_COMP)))
 		return NULL;
 
-	page = alloc_contig_pages_noprof(1 << order, gfp, nid, node);
+	page = alloc_contig_frozen_pages_noprof(1 << order, gfp, nid, node);
 
 	return page ? page_folio(page) : NULL;
 }
 #else
-static inline struct folio *folio_alloc_gigantic_noprof(int order, gfp_t gfp,
-							int nid, nodemask_t *node)
+static inline struct folio *folio_alloc_frozen_gigantic_noprof(int order,
+		gfp_t gfp, int nid, nodemask_t *node)
 {
 	return NULL;
 }
 #endif
-/* This should be paired with folio_put() rather than free_contig_range(). */
-#define folio_alloc_gigantic(...) alloc_hooks(folio_alloc_gigantic_noprof(__VA_ARGS__))
+/* This should be paired with free_frozen_pages() rather than free_contig_range_frozen(). */
+#define folio_alloc_frozen_gigantic(...)	\
+	alloc_hooks(folio_alloc_frozen_gigantic_noprof(__VA_ARGS__))
 
 #endif /* __LINUX_GFP_H */
diff --git a/mm/cma.c b/mm/cma.c
index 2af8c5bc58dd..bca95b8a1015 100644
--- a/mm/cma.c
+++ b/mm/cma.c
@@ -778,7 +778,7 @@ static void cma_debug_show_areas(struct cma *cma)
 
 static int cma_range_alloc(struct cma *cma, struct cma_memrange *cmr,
 				unsigned long count, unsigned int align,
-				struct page **pagep, gfp_t gfp)
+				struct page **pagep, gfp_t gfp, bool frozen)
 {
 	unsigned long bitmap_maxno, bitmap_no, bitmap_count;
 	unsigned long start, pfn, mask, offset;
@@ -836,7 +836,12 @@ static int cma_range_alloc(struct cma *cma, struct cma_memrange *cmr,
 		spin_unlock_irq(&cma->lock);
 
 		mutex_lock(&cma->alloc_mutex);
-		ret = alloc_contig_range(pfn, pfn + count, ACR_FLAGS_CMA, gfp);
+		if (frozen)
+			ret = alloc_contig_range_frozen(pfn, pfn + count,
+							ACR_FLAGS_CMA, gfp);
+		else
+			ret = alloc_contig_range(pfn, pfn + count,
+						 ACR_FLAGS_CMA, gfp);
 		mutex_unlock(&cma->alloc_mutex);
 		if (!ret)
 			break;
@@ -857,7 +862,7 @@ static int cma_range_alloc(struct cma *cma, struct cma_memrange *cmr,
 }
 
 static struct page *__cma_alloc(struct cma *cma, unsigned long count,
-		       unsigned int align, gfp_t gfp)
+		       unsigned int align, gfp_t gfp, bool frozen)
 {
 	struct page *page = NULL;
 	int ret = -ENOMEM, r;
@@ -879,7 +884,7 @@ static struct page *__cma_alloc(struct cma *cma, unsigned long count,
 		page = NULL;
 
 		ret = cma_range_alloc(cma, &cma->ranges[r], count, align,
-				       &page, gfp);
+				       &page, gfp, frozen);
 		if (ret != -EBUSY || page)
 			break;
 	}
@@ -927,23 +932,24 @@ static struct page *__cma_alloc(struct cma *cma, unsigned long count,
 struct page *cma_alloc(struct cma *cma, unsigned long count,
 		       unsigned int align, bool no_warn)
 {
-	return __cma_alloc(cma, count, align, GFP_KERNEL | (no_warn ? __GFP_NOWARN : 0));
+	return __cma_alloc(cma, count, align,
+			   GFP_KERNEL | (no_warn ? __GFP_NOWARN : 0), false);
 }
 
-struct folio *cma_alloc_folio(struct cma *cma, int order, gfp_t gfp)
+struct folio *cma_alloc_frozen_folio(struct cma *cma, int order, gfp_t gfp)
 {
 	struct page *page;
 
 	if (WARN_ON(!order || !(gfp & __GFP_COMP)))
 		return NULL;
 
-	page = __cma_alloc(cma, 1 << order, order, gfp);
+	page = __cma_alloc(cma, 1 << order, order, gfp, true);
 
 	return page ? page_folio(page) : NULL;
 }
 
 static bool __cma_release(struct cma *cma, const struct page *pages,
-			  unsigned long count)
+			  unsigned long count, bool frozen)
 {
 	unsigned long pfn, end;
 	int r;
@@ -966,7 +972,11 @@ static bool __cma_release(struct cma *cma, const struct page *pages,
 	if (r == cma->nranges)
 		return false;
 
-	free_contig_range(pfn, count);
+	if (frozen)
+		free_contig_range_frozen(pfn, count);
+	else
+		free_contig_range(pfn, count);
+
 	cma_clear_bitmap(cma, cmr, pfn, count);
 	cma_sysfs_account_release_pages(cma, count);
 	trace_cma_release(cma->name, pfn, pages, count);
@@ -987,15 +997,15 @@ static bool __cma_release(struct cma *cma, const struct page *pages,
 bool cma_release(struct cma *cma, const struct page *pages,
 		 unsigned long count)
 {
-	return __cma_release(cma, pages, count);
+	return __cma_release(cma, pages, count, false);
 }
 
-bool cma_free_folio(struct cma *cma, const struct folio *folio)
+bool cma_free_frozen_folio(struct cma *cma, const struct folio *folio)
 {
 	if (WARN_ON(!folio_test_large(folio)))
 		return false;
 
-	return __cma_release(cma, &folio->page, folio_nr_pages(folio));
+	return __cma_release(cma, &folio->page, folio_nr_pages(folio), true);
 }
 
 int cma_for_each_area(int (*it)(struct cma *cma, void *data), void *data)
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index f0e701e455ed..100e840894f1 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -125,16 +125,6 @@ static void hugetlb_unshare_pmds(struct vm_area_struct *vma,
 		unsigned long start, unsigned long end, bool take_locks);
 static struct resv_map *vma_resv_map(struct vm_area_struct *vma);
 
-static void hugetlb_free_folio(struct folio *folio)
-{
-	if (folio_test_hugetlb_cma(folio)) {
-		hugetlb_cma_free_folio(folio);
-		return;
-	}
-
-	folio_put(folio);
-}
-
 static inline bool subpool_is_free(struct hugepage_subpool *spool)
 {
 	if (spool->count)
@@ -1472,43 +1462,20 @@ static int hstate_next_node_to_free(struct hstate *h, nodemask_t *nodes_allowed)
 		nr_nodes--)
 
 #ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
-#ifdef CONFIG_CONTIG_ALLOC
 static struct folio *alloc_gigantic_folio(int order, gfp_t gfp_mask,
 		int nid, nodemask_t *nodemask)
 {
 	struct folio *folio;
-	bool retried = false;
-
-retry:
-	folio = hugetlb_cma_alloc_folio(order, gfp_mask, nid, nodemask);
-	if (!folio) {
-		if (hugetlb_cma_exclusive_alloc())
-			return NULL;
-
-		folio = folio_alloc_gigantic(order, gfp_mask, nid, nodemask);
-		if (!folio)
-			return NULL;
-	}
 
-	if (folio_ref_freeze(folio, 1))
+	folio = hugetlb_cma_alloc_frozen_folio(order, gfp_mask, nid, nodemask);
+	if (folio)
 		return folio;
 
-	pr_warn("HugeTLB: unexpected refcount on PFN %lu\n", folio_pfn(folio));
-	hugetlb_free_folio(folio);
-	if (!retried) {
-		retried = true;
-		goto retry;
-	}
-	return NULL;
-}
+	if (hugetlb_cma_exclusive_alloc())
+		return NULL;
 
-#else /* !CONFIG_CONTIG_ALLOC */
-static struct folio *alloc_gigantic_folio(int order, gfp_t gfp_mask, int nid,
-					  nodemask_t *nodemask)
-{
-	return NULL;
+	return folio_alloc_frozen_gigantic(order, gfp_mask, nid, nodemask);
 }
-#endif /* CONFIG_CONTIG_ALLOC */
 
 #else /* !CONFIG_ARCH_HAS_GIGANTIC_PAGE */
 static struct folio *alloc_gigantic_folio(int order, gfp_t gfp_mask, int nid,
@@ -1641,9 +1608,12 @@ static void __update_and_free_hugetlb_folio(struct hstate *h,
 	if (unlikely(folio_test_hwpoison(folio)))
 		folio_clear_hugetlb_hwpoison(folio);
 
-	folio_ref_unfreeze(folio, 1);
+	VM_BUG_ON_FOLIO(folio_ref_count(folio), folio);
 
-	hugetlb_free_folio(folio);
+	if (folio_test_hugetlb_cma(folio))
+		hugetlb_cma_free_frozen_folio(folio);
+	else
+		free_frozen_pages(&folio->page, folio_order(folio));
 }
 
 /*
diff --git a/mm/hugetlb_cma.c b/mm/hugetlb_cma.c
index e8e4dc7182d5..24738a1d6098 100644
--- a/mm/hugetlb_cma.c
+++ b/mm/hugetlb_cma.c
@@ -18,29 +18,28 @@ static unsigned long hugetlb_cma_size_in_node[MAX_NUMNODES] __initdata;
 static bool hugetlb_cma_only;
 static unsigned long hugetlb_cma_size __initdata;
 
-void hugetlb_cma_free_folio(struct folio *folio)
+void hugetlb_cma_free_frozen_folio(struct folio *folio)
 {
 	int nid = folio_nid(folio);
 
-	WARN_ON_ONCE(!cma_free_folio(hugetlb_cma[nid], folio));
+	WARN_ON_ONCE(!cma_free_frozen_folio(hugetlb_cma[nid], folio));
 }
 
-
-struct folio *hugetlb_cma_alloc_folio(int order, gfp_t gfp_mask,
-				      int nid, nodemask_t *nodemask)
+struct folio *hugetlb_cma_alloc_frozen_folio(int order, gfp_t gfp_mask,
+		int nid, nodemask_t *nodemask)
 {
 	int node;
 	struct folio *folio = NULL;
 
 	if (hugetlb_cma[nid])
-		folio = cma_alloc_folio(hugetlb_cma[nid], order, gfp_mask);
+		folio = cma_alloc_frozen_folio(hugetlb_cma[nid], order, gfp_mask);
 
 	if (!folio && !(gfp_mask & __GFP_THISNODE)) {
 		for_each_node_mask(node, *nodemask) {
 			if (node == nid || !hugetlb_cma[node])
 				continue;
 
-			folio = cma_alloc_folio(hugetlb_cma[node], order, gfp_mask);
+			folio = cma_alloc_frozen_folio(hugetlb_cma[node], order, gfp_mask);
 			if (folio)
 				break;
 		}
diff --git a/mm/hugetlb_cma.h b/mm/hugetlb_cma.h
index 2c2ec8a7e134..71db3544816e 100644
--- a/mm/hugetlb_cma.h
+++ b/mm/hugetlb_cma.h
@@ -3,8 +3,8 @@
 #define _LINUX_HUGETLB_CMA_H
 
 #ifdef CONFIG_CMA
-void hugetlb_cma_free_folio(struct folio *folio);
-struct folio *hugetlb_cma_alloc_folio(int order, gfp_t gfp_mask,
+void hugetlb_cma_free_frozen_folio(struct folio *folio);
+struct folio *hugetlb_cma_alloc_frozen_folio(int order, gfp_t gfp_mask,
 				      int nid, nodemask_t *nodemask);
 struct huge_bootmem_page *hugetlb_cma_alloc_bootmem(struct hstate *h, int *nid,
 						    bool node_exact);
@@ -14,12 +14,12 @@ unsigned long hugetlb_cma_total_size(void);
 void hugetlb_cma_validate_params(void);
 bool hugetlb_early_cma(struct hstate *h);
 #else
-static inline void hugetlb_cma_free_folio(struct folio *folio)
+static inline void hugetlb_cma_free_frozen_folio(struct folio *folio)
 {
 }
 
-static inline struct folio *hugetlb_cma_alloc_folio(int order, gfp_t gfp_mask,
-		int nid, nodemask_t *nodemask)
+static inline struct folio *hugetlb_cma_alloc_frozen_folio(int order,
+		gfp_t gfp_mask, int nid, nodemask_t *nodemask)
 {
 	return NULL;
 }
-- 
2.43.0