Commit a482289d authored by Nick Piggin's avatar Nick Piggin Committed by Linus Torvalds
Browse files

[PATCH] hugepage allocator cleanup 



Insert "fresh" huge pages into the hugepage allocator by the same means as
they are freed back into it.  This reduces code size and allows
enqueue_huge_page to be inlined into the hugepage free fastpath.

Eliminate occurances of hugepages on the free list with non-zero refcount.
This can allow stricter refcount checks in future.  Also required for
lockless pagecache.

Signed-off-by: default avatarNick Piggin <npiggin@suse.de>

"This patch also eliminates a leak "cleaned up" by re-clobbering the
refcount on every allocation from the hugepage freelists.  With respect to
the lockless pagecache, the crucial aspect is to eliminate unconditional
set_page_count() to 0 on pages with potentially nonzero refcounts, though
closer inspection suggests the assignments removed are entirely spurious."

Acked-by: default avatarWilliam Irwin <wli@holomorphy.com>
Signed-off-by: default avatarAndrew Morton <akpm@osdl.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@osdl.org>
parent 545b1ea9

mm/hugetlb.c

+8 −16
Original line number Original line Diff line number Diff line
@@ -64,7 +64,7 @@ static struct page *dequeue_huge_page(struct vm_area_struct *vma, 
	return page;

	return page;

}

}





static struct page *alloc_fresh_huge_page(void)

static int alloc_fresh_huge_page(void)

{

{

	static int nid = 0;

	static int nid = 0;

	struct page *page;

	struct page *page;
@@ -72,12 +72,15 @@ static struct page *alloc_fresh_huge_page(void) 
					HUGETLB_PAGE_ORDER);

					HUGETLB_PAGE_ORDER);

	nid = (nid + 1) % num_online_nodes();

	nid = (nid + 1) % num_online_nodes();

	if (page) {

	if (page) {

		page[1].lru.next = (void *)free_huge_page;	/* dtor */

		spin_lock(&hugetlb_lock);

		spin_lock(&hugetlb_lock);

		nr_huge_pages++;

		nr_huge_pages++;

		nr_huge_pages_node[page_to_nid(page)]++;

		nr_huge_pages_node[page_to_nid(page)]++;

		spin_unlock(&hugetlb_lock);

		spin_unlock(&hugetlb_lock);

		put_page(page); /* free it into the hugepage allocator */

		return 1;

	}

	}

	return page;

	return 0;

}

}





void free_huge_page(struct page *page)

void free_huge_page(struct page *page)
@@ -85,7 +88,6 @@ void free_huge_page(struct page *page) 
	BUG_ON(page_count(page));

	BUG_ON(page_count(page));





	INIT_LIST_HEAD(&page->lru);

	INIT_LIST_HEAD(&page->lru);

	page[1].lru.next = NULL;			/* reset dtor */





	spin_lock(&hugetlb_lock);

	spin_lock(&hugetlb_lock);

	enqueue_huge_page(page);

	enqueue_huge_page(page);
@@ -105,7 +107,6 @@ struct page *alloc_huge_page(struct vm_area_struct *vma, unsigned long addr) 
	}

	}

	spin_unlock(&hugetlb_lock);

	spin_unlock(&hugetlb_lock);

	set_page_count(page, 1);

	set_page_count(page, 1);

	page[1].lru.next = (void *)free_huge_page;	/* set dtor */

	for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); ++i)

	for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); ++i)

		clear_user_highpage(&page[i], addr);

		clear_user_highpage(&page[i], addr);

	return page;

	return page;
@@ -114,7 +115,6 @@ struct page *alloc_huge_page(struct vm_area_struct *vma, unsigned long addr) 
static int __init hugetlb_init(void)

static int __init hugetlb_init(void)

{

{

	unsigned long i;

	unsigned long i;

	struct page *page;





	if (HPAGE_SHIFT == 0)

	if (HPAGE_SHIFT == 0)

		return 0;

		return 0;
@@ -123,12 +123,8 @@ static int __init hugetlb_init(void) 
		INIT_LIST_HEAD(&hugepage_freelists[i]);

		INIT_LIST_HEAD(&hugepage_freelists[i]);





	for (i = 0; i < max_huge_pages; ++i) {

	for (i = 0; i < max_huge_pages; ++i) {

		page = alloc_fresh_huge_page();

		if (!alloc_fresh_huge_page())

		if (!page)

			break;

			break;

		spin_lock(&hugetlb_lock);

		enqueue_huge_page(page);

		spin_unlock(&hugetlb_lock);

	}

	}

	max_huge_pages = free_huge_pages = nr_huge_pages = i;

	max_huge_pages = free_huge_pages = nr_huge_pages = i;

	printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);

	printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);
@@ -154,8 +150,8 @@ static void update_and_free_page(struct page *page) 
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error | 1 << PG_referenced |

		page[i].flags &= ~(1 << PG_locked | 1 << PG_error | 1 << PG_referenced |

				1 << PG_dirty | 1 << PG_active | 1 << PG_reserved |

				1 << PG_dirty | 1 << PG_active | 1 << PG_reserved |

				1 << PG_private | 1<< PG_writeback);

				1 << PG_private | 1<< PG_writeback);

		set_page_count(&page[i], 0);

	}

	}

	page[1].lru.next = NULL;

	set_page_count(page, 1);

	set_page_count(page, 1);

	__free_pages(page, HUGETLB_PAGE_ORDER);

	__free_pages(page, HUGETLB_PAGE_ORDER);

}

}
@@ -188,12 +184,8 @@ static inline void try_to_free_low(unsigned long count) 
static unsigned long set_max_huge_pages(unsigned long count)

static unsigned long set_max_huge_pages(unsigned long count)

{

{

	while (count > nr_huge_pages) {

	while (count > nr_huge_pages) {

		struct page *page = alloc_fresh_huge_page();

		if (!alloc_fresh_huge_page())

		if (!page)

			return nr_huge_pages;

			return nr_huge_pages;

		spin_lock(&hugetlb_lock);

		enqueue_huge_page(page);

		spin_unlock(&hugetlb_lock);

	}

	}

	if (count >= nr_huge_pages)

	if (count >= nr_huge_pages)

		return nr_huge_pages;

		return nr_huge_pages;

CRA Git | Maintained and supported by SUSTech CRA and CCSE