[RFC PATCH 15/15] mm: make MAX_ORDER a kernel boot time parameter.

[Zi Yan]

From: Zi Yan <ziy@xxxxxxxxxx>

With the new buddy_alloc_max_order, users can specify larger MAX_ORDER
than set in CONFIG_ARCH_MAX_ORDER or CONFIG_SET_MAX_ORDER.
It can be set any value >= CONFIG_ARCH_MAX_ORDER or CONFIG_SET_MAX_ORDER,
but < 256 (limited by vmscan scan_control and per-cpu free page list).

Signed-off-by: Zi Yan <ziy@xxxxxxxxxx>
Cc: Jonathan Corbet <corbet@xxxxxxx>
Cc: "Paul E. McKenney" <paulmck@xxxxxxxxxx>
Cc: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Vlastimil Babka <vbabka@xxxxxxx>
Cc: linux-doc@xxxxxxxxxxxxxxx
Cc: linux-mm@xxxxxxxxx
Cc: linux-kernel@xxxxxxxxxxxxxxx
---
 .../admin-guide/kernel-parameters.txt         |  5 +++
 include/linux/mmzone.h                        | 23 +++++++++++--
 mm/page_alloc.c                               | 34 ++++++++++++++++++-
 mm/vmscan.c                                   |  1 -
 4 files changed, 58 insertions(+), 5 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 5c59a5fb17c3..a37141aa28ae 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -465,6 +465,11 @@
 	bttv.pll=	See Documentation/admin-guide/media/bttv.rst
 	bttv.tuner=
 
+	buddy_alloc_max_order=	[KNL] This parameter adjusts the size of largest
+			pages that can be allocated from kernel buddy allocator. The largest
+			page size is 2^buddy_alloc_max_order * PAGE_SIZE.
+            Format: integer
+
 	bulk_remove=off	[PPC]  This parameter disables the use of the pSeries
 			firmware feature for flushing multiple hpte entries
 			at a time.
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index 379dada82d4b..9ca4d59722a1 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -26,14 +26,25 @@
 /* Free memory management - zoned buddy allocator.  */
 #ifndef CONFIG_ARCH_FORCE_MAX_ORDER
 #ifdef CONFIG_SET_MAX_ORDER
-#define MAX_ORDER CONFIG_SET_MAX_ORDER
+/* Defined in mm/page_alloc.c */
+extern int buddy_alloc_max_order;
+
+#define MAX_ORDER buddy_alloc_max_order
 #define MIN_MAX_ORDER CONFIG_SET_MAX_ORDER
 #else
 #define MAX_ORDER 11
 #define MIN_MAX_ORDER MAX_ORDER
 #endif /* CONFIG_SET_MAX_ORDER */
 #else
+
+#ifdef CONFIG_SPARSEMEM_VMEMMAP
+/* Defined in mm/page_alloc.c */
+extern int buddy_alloc_max_order;
+
+#define MAX_ORDER buddy_alloc_max_order
+#else
 #define MAX_ORDER CONFIG_ARCH_FORCE_MAX_ORDER
+#endif /* CONFIG_SPARSEMEM_VMEMMAP */
 #define MIN_MAX_ORDER CONFIG_ARCH_FORCE_MAX_ORDER
 #endif /* CONFIG_ARCH_FORCE_MAX_ORDER */
 #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
@@ -1557,8 +1568,14 @@ void sparse_init(void);
  * pfn_valid_within() should be used in this case; we optimise this away
  * when we have no holes within a MAX_ORDER_NR_PAGES block.
  */
-#if ((MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS)
-#define pfn_valid_within(pfn) pfn_valid(pfn)
+#if defined(CONFIG_ARCH_FORCE_MAX_ORDER) || defined(CONFIG_SET_MAX_ORDER)
+static inline bool pfn_valid_within(unsigned long pfn)
+{
+	if ((MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS)
+		return pfn_valid(pfn);
+
+	return 1;
+}
 #else
 #define pfn_valid_within(pfn) (1)
 #endif
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index bfa6962f7615..ea6f8d85a4cf 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1487,7 +1487,6 @@ static void free_pcppages_bulk(struct zone *zone, int count,
 			batch_free = count;
 
 		order = pindex_to_order(pindex);
-		BUILD_BUG_ON(MAX_ORDER >= (1<<NR_PCP_ORDER_WIDTH));
 		do {
 			page = list_last_entry(list, struct page, lru);
 			/* must delete to avoid corrupting pcp list */
@@ -9508,3 +9507,36 @@ bool take_page_off_buddy(struct page *page)
 	return ret;
 }
 #endif
+
+#if (defined(CONFIG_ARCH_FORCE_MAX_ORDER) && defined(CONFIG_SPARSEMEM_VMEMMAP)) \
+	|| defined(CONFIG_SET_MAX_ORDER)
+int buddy_alloc_max_order = MIN_MAX_ORDER;
+EXPORT_SYMBOL(buddy_alloc_max_order);
+
+static int __init buddy_alloc_set(char *val)
+{
+	int ret;
+	unsigned long max_order;
+
+	ret = kstrtoul(val, 10, &max_order);
+
+	if (ret < 0)
+		return -EINVAL;
+
+	/*
+	 * max_order is also limited at below locations:
+	 * 1. scan_control in mm/vmscan.c uses s8 field for order, max_order cannot
+	 * be bigger than S8_MAX before the field is changed.
+	 * 2. free_pcppages_bulk has max_order upper limit.
+	 */
+	if (max_order > MIN_MAX_ORDER && max_order < S8_MAX &&
+	    max_order < (1<<NR_PCP_ORDER_WIDTH))
+		buddy_alloc_max_order = max_order;
+	else
+		buddy_alloc_max_order = MIN_MAX_ORDER;
+
+	return 0;
+}
+
+early_param("buddy_alloc_max_order", buddy_alloc_set);
+#endif
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 403a175a720f..9a3963c6166e 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -3610,7 +3610,6 @@ unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
 	 * scan_control uses s8 fields for order, priority, and reclaim_idx.
 	 * Confirm they are large enough for max values.
 	 */
-	BUILD_BUG_ON(MAX_ORDER > S8_MAX);
 	BUILD_BUG_ON(DEF_PRIORITY > S8_MAX);
 	BUILD_BUG_ON(MAX_NR_ZONES > S8_MAX);
 
-- 
2.30.2