0001-hugetlb-force-bootmem-prealloc-with-continuous-membl.patch

From 155f3ef694ee54fc2f9a3ad8f75e3a4e81477e27 Mon Sep 17 00:00:00 2001
From: Yukinari Toyota <[email protected]>
Date: Thu, 13 Feb 2014 09:28:59 +0900
Subject: [PATCH] hugetlb: force bootmem prealloc with continuous memblock alloc

---
 arch/x86/kernel/setup.c |    2 +
 mm/hugetlb.c            |   53 ++++++++++++++++-------------------------------
 2 files changed, 20 insertions(+), 35 deletions(-)

diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index ce72964..6d0a681 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -403,6 +403,7 @@ static void __init reserve_initrd(void)
 	printk(KERN_INFO "RAMDISK: [mem %#010llx-%#010llx]\n", ramdisk_image,
 			ramdisk_end - 1);
 
+#if 0 /* force relocate to clear the way to allocate large continuous memblock */
 	if (pfn_range_is_mapped(PFN_DOWN(ramdisk_image),
 				PFN_DOWN(ramdisk_end))) {
 		/* All are mapped, easy case */
@@ -410,6 +411,7 @@ static void __init reserve_initrd(void)
 		initrd_end = initrd_start + ramdisk_size;
 		return;
 	}
+#endif
 
 	relocate_initrd();
 
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index c01cb9f..3698f0c 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1263,16 +1263,16 @@ struct page *alloc_huge_page_noerr(struct vm_area_struct *vma,
 	return page;
 }
 
-int __weak alloc_bootmem_huge_page(struct hstate *h)
+static int __init alloc_bootmem_huge_pages_atonce(struct hstate *h, int nr_pages)
 {
 	struct huge_bootmem_page *m;
 	int nr_nodes, node;
+	unsigned long i;
+	void *addr;
 
 	for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) {
-		void *addr;
-
 		addr = memblock_virt_alloc_try_nid_nopanic(
-				huge_page_size(h), huge_page_size(h),
+				huge_page_size(h)*nr_pages, huge_page_size(h),
 				0, BOOTMEM_ALLOC_ACCESSIBLE, node);
 		if (addr) {
 			/*
@@ -1284,22 +1284,25 @@ int __weak alloc_bootmem_huge_page(struct hstate *h)
 			goto found;
 		}
 	}
+	WARN_ON(1);
 	return 0;
 
 found:
-	BUG_ON((unsigned long)virt_to_phys(m) & (huge_page_size(h) - 1));
-	/* Put them into a private list first because mem_map is not up yet */
-	list_add(&m->list, &huge_boot_pages);
-	m->hstate = h;
+	for(i = 0; i < nr_pages; i++) {
+		m = addr + huge_page_size(h)*i;
+		BUG_ON((unsigned long)virt_to_phys(m) & (huge_page_size(h) - 1));
+		/* Put them into a private list first because mem_map is not up yet */
+		list_add(&m->list, &huge_boot_pages);
+		m->hstate = h;
+	}
 	return 1;
 }
 
+
 static void prep_compound_huge_page(struct page *page, int order)
 {
-	if (unlikely(order > (MAX_ORDER - 1)))
-		prep_compound_gigantic_page(page, order);
-	else
-		prep_compound_page(page, order);
+	// since we always use bootmem, call the below even if order is not gigantic
+	prep_compound_gigantic_page(page, order);
 }
 
 /* Put bootmem huge pages into the standard lists after mem_map is up */
@@ -1335,28 +1338,13 @@ static void __init gather_bootmem_prealloc(void)
 
 static void __init hugetlb_hstate_alloc_pages(struct hstate *h)
 {
-	unsigned long i;
-
-	for (i = 0; i < h->max_huge_pages; ++i) {
-		if (h->order >= MAX_ORDER) {
-			if (!alloc_bootmem_huge_page(h))
-				break;
-		} else if (!alloc_fresh_huge_page(h,
-					 &node_states[N_MEMORY]))
-			break;
+	if (!alloc_bootmem_huge_pages_atonce(h, h->max_huge_pages)) {
+		h->max_huge_pages = 0;
 	}
-	h->max_huge_pages = i;
 }
 
 static void __init hugetlb_init_hstates(void)
 {
-	struct hstate *h;
-
-	for_each_hstate(h) {
-		/* oversize hugepages were init'ed in early boot */
-		if (h->order < MAX_ORDER)
-			hugetlb_hstate_alloc_pages(h);
-	}
 }
 
 static char * __init memfmt(char *buf, unsigned long n)
@@ -2028,12 +2016,7 @@ static int __init hugetlb_nrpages_setup(char *s)
 	if (sscanf(s, "%lu", mhp) <= 0)
 		*mhp = 0;
 
-	/*
-	 * Global state is always initialized later in hugetlb_init.
-	 * But we need to allocate >= MAX_ORDER hstates here early to still
-	 * use the bootmem allocator.
-	 */
-	if (hugetlb_max_hstate && parsed_hstate->order >= MAX_ORDER)
+	if (hugetlb_max_hstate)
 		hugetlb_hstate_alloc_pages(parsed_hstate);
 
 	last_mhp = mhp;
-- 
1.7.1