[PATCH] handle radix_tree_node allocation failures

This patch uses the radix_tree_preload() API in add_to_page_cache().

A new gfp_mask argument is added to add_to_page_cache(), which is then passed
on to radix_tree_preload().   It's pretty simple.

In the case of adding pages to swapcache we're still using GFP_ATOMIC, so
these addition attempts can still fail.  That's OK, because the error is
handled and, unlike file pages, it will not cause user applicaton failures.
This codepath (radix-tree node exhaustion on swapout) was well tested in the
days when the swapper_space radix tree was fragmented all over the place due
to unfortunate swp_entry bit layout.

arch/i386/mm/hugetlbpage.c

@@ -234,8 +234,12 @@ int hugetlb_prefault(struct address_space *mapping, struct vm_area_struct *vma)
 				ret = -ENOMEM;
 				goto out;
 			}
-			add_to_page_cache(page, mapping, idx);
+			ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC);
 			unlock_page(page);
+			if (ret) {
+				free_huge_page(page);
+				goto out;
+			}
 		}
 		set_huge_pte(mm, vma, page, pte, vma->vm_flags & VM_WRITE);
 	}

arch/sparc64/mm/hugetlbpage.c

@@ -476,9 +476,16 @@ static int alloc_shared_hugetlb_pages(int key, unsigned long addr, unsigned long
 			page = alloc_hugetlb_page();
 			if (page == NULL) {
 				pte_unmap(pte);
+				retval = -ENOMEM;
+				goto out;
+			}
+			retval = add_to_page_cache(page, mapping,
+						idx, GFP_ATOMIC);
+			if (retval) {
+				pte_unmap(pte);
+				free_hugetlb_page(page);
 				goto out;
 			}
-			add_to_page_cache(page, mapping, idx);
 		}
 		set_huge_pte(mm, vma, page, pte,
 			     (vma->vm_flags & VM_WRITE));

arch/x86_64/mm/hugetlbpage.c

@@ -232,8 +232,12 @@ int hugetlb_prefault(struct address_space *mapping, struct vm_area_struct *vma)
 				ret = -ENOMEM;
 				goto out;
 			}
-			add_to_page_cache(page, mapping, idx);
+			ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC);
 			unlock_page(page);
+			if (ret) {
+				free_huge_page(page);
+				goto out;
+			}
 		}
 		set_huge_pte(mm, vma, page, pte, vma->vm_flags & VM_WRITE);
 	}

fs/mpage.c

@@ -275,7 +275,8 @@ mpage_readpages(struct address_space *mapping, struct list_head *pages,
 
 		prefetchw(&page->flags);
 		list_del(&page->list);
-		if (!add_to_page_cache(page, mapping, page->index)) {
+		if (!add_to_page_cache(page, mapping,
+					page->index, GFP_KERNEL)) {
 			bio = do_mpage_readpage(bio, page,
 					nr_pages - page_idx,
 					&last_block_in_bio, get_block);

include/linux/pagemap.h

@@ -66,10 +66,10 @@ extern struct page * read_cache_page(struct address_space *mapping,
 extern int read_cache_pages(struct address_space *mapping,
 		struct list_head *pages, filler_t *filler, void *data);
 
-extern int add_to_page_cache(struct page *page,
-		struct address_space *mapping, unsigned long index);
-extern int add_to_page_cache_lru(struct page *page,
-		struct address_space *mapping, unsigned long index);
+int add_to_page_cache(struct page *page, struct address_space *mapping,
+				unsigned long index, int gfp_mask);
+int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
+				unsigned long index, int gfp_mask);
 extern void remove_from_page_cache(struct page *page);
 extern void __remove_from_page_cache(struct page *page);
 

mm/filemap.c

@@ -202,28 +202,31 @@ int filemap_fdatawait(struct address_space * mapping)
  *
  * This function does not add the page to the LRU.  The caller must do that.
  */
-int add_to_page_cache(struct page *page,
-		struct address_space *mapping, pgoff_t offset)
+int add_to_page_cache(struct page *page, struct address_space *mapping,
+		pgoff_t offset, int gfp_mask)
 {
-	int error;
+	int error = radix_tree_preload(gfp_mask & ~__GFP_HIGHMEM);
 
-	page_cache_get(page);
-	write_lock(&mapping->page_lock);
-	error = radix_tree_insert(&mapping->page_tree, offset, page);
-	if (!error) {
-		SetPageLocked(page);
-		___add_to_page_cache(page, mapping, offset);
-	} else {
-		page_cache_release(page);
+	if (error == 0) {
+		page_cache_get(page);
+		write_lock(&mapping->page_lock);
+		error = radix_tree_insert(&mapping->page_tree, offset, page);
+		if (!error) {
+			SetPageLocked(page);
+			___add_to_page_cache(page, mapping, offset);
+		} else {
+			page_cache_release(page);
+		}
+		write_unlock(&mapping->page_lock);
+		radix_tree_preload_end();
 	}
-	write_unlock(&mapping->page_lock);
 	return error;
 }
 
-int add_to_page_cache_lru(struct page *page,
-		struct address_space *mapping, pgoff_t offset)
+int add_to_page_cache_lru(struct page *page, struct address_space *mapping,
+				pgoff_t offset, int gfp_mask)
 {
-	int ret = add_to_page_cache(page, mapping, offset);
+	int ret = add_to_page_cache(page, mapping, offset, gfp_mask);
 	if (ret == 0)
 		lru_cache_add(page);
 	return ret;
@@ -432,7 +435,8 @@ struct page *find_or_create_page(struct address_space *mapping,
 			if (!cached_page)
 				return NULL;
 		}
-		err = add_to_page_cache_lru(cached_page, mapping, index);
+		err = add_to_page_cache_lru(cached_page, mapping,
+					index, gfp_mask);
 		if (!err) {
 			page = cached_page;
 			cached_page = NULL;
@@ -488,6 +492,7 @@ struct page *
 grab_cache_page_nowait(struct address_space *mapping, unsigned long index)
 {
 	struct page *page = find_get_page(mapping, index);
+	int gfp_mask;
 
 	if (page) {
 		if (!TestSetPageLocked(page))
@@ -495,8 +500,9 @@ grab_cache_page_nowait(struct address_space *mapping, unsigned long index)
 		page_cache_release(page);
 		return NULL;
 	}
-	page = alloc_pages(mapping->gfp_mask & ~__GFP_FS, 0);
-	if (page && add_to_page_cache_lru(page, mapping, index)) {
+	gfp_mask = mapping->gfp_mask & ~__GFP_FS;
+	page = alloc_pages(gfp_mask, 0);
+	if (page && add_to_page_cache_lru(page, mapping, index, gfp_mask)) {
 		page_cache_release(page);
 		page = NULL;
 	}
@@ -648,7 +654,8 @@ void do_generic_mapping_read(struct address_space *mapping,
 				break;
 			}
 		}
-		error = add_to_page_cache_lru(cached_page, mapping, index);
+		error = add_to_page_cache_lru(cached_page, mapping,
+						index, GFP_KERNEL);
 		if (error) {
 			if (error == -EEXIST)
 				goto find_page;
@@ -940,7 +947,7 @@ static int page_cache_read(struct file * file, unsigned long offset)
 	if (!page)
 		return -ENOMEM;
 
-	error = add_to_page_cache_lru(page, mapping, offset);
+	error = add_to_page_cache_lru(page, mapping, offset, GFP_KERNEL);
 	if (!error) {
 		error = mapping->a_ops->readpage(file, page);
 		page_cache_release(page);
@@ -1327,7 +1334,8 @@ static inline struct page *__read_cache_page(struct address_space *mapping,
 			if (!cached_page)
 				return ERR_PTR(-ENOMEM);
 		}
-		err = add_to_page_cache_lru(cached_page, mapping, index);
+		err = add_to_page_cache_lru(cached_page, mapping,
+					index, GFP_KERNEL);
 		if (err == -EEXIST)
 			goto repeat;
 		if (err < 0) {
@@ -1406,7 +1414,8 @@ __grab_cache_page(struct address_space *mapping, unsigned long index,
 			if (!*cached_page)
 				return NULL;
 		}
-		err = add_to_page_cache(*cached_page, mapping, index);
+		err = add_to_page_cache(*cached_page, mapping,
+					index, GFP_KERNEL);
 		if (err == -EEXIST)
 			goto repeat;
 		if (err == 0) {

mm/readahead.c

@@ -65,7 +65,7 @@ int read_cache_pages(struct address_space *mapping, struct list_head *pages,
 	while (!list_empty(pages)) {
 		page = list_entry(pages->prev, struct page, list);
 		list_del(&page->list);
-		if (add_to_page_cache(page, mapping, page->index)) {
+		if (add_to_page_cache(page, mapping, page->index, GFP_KERNEL)) {
 			page_cache_release(page);
 			continue;
 		}
@@ -93,7 +93,8 @@ static int read_pages(struct address_space *mapping, struct file *filp,
 	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
 		struct page *page = list_entry(pages->prev, struct page, list);
 		list_del(&page->list);
-		if (!add_to_page_cache(page, mapping, page->index)) {
+		if (!add_to_page_cache(page, mapping,
+					page->index, GFP_KERNEL)) {
 			mapping->a_ops->readpage(filp, page);
 			if (!pagevec_add(&lru_pvec, page))
 				__pagevec_lru_add(&lru_pvec);

mm/shmem.c

@@ -889,7 +889,7 @@ static int shmem_getpage(struct inode *inode, unsigned long idx, struct page **p
 			}
 			if (error || swap.val ||
 			    (error = add_to_page_cache_lru(
-					filepage, mapping, idx))) {
+					filepage, mapping, idx, GFP_ATOMIC))) {
 				spin_unlock(&info->lock);
 				page_cache_release(filepage);
 				shmem_free_block(inode);

mm/swap_state.c

@@ -78,7 +78,7 @@ int add_to_swap_cache(struct page *page, swp_entry_t entry)
 		INC_CACHE_INFO(noent_race);
 		return -ENOENT;
 	}
-	error = add_to_page_cache(page, &swapper_space, entry.val);
+	error = add_to_page_cache(page, &swapper_space, entry.val, GFP_ATOMIC);
 	/*
 	 * Anon pages are already on the LRU, we don't run lru_cache_add here.
 	 */
@@ -149,7 +149,8 @@ int add_to_swap(struct page * page)
 		/*
 		 * Add it to the swap cache and mark it dirty
 		 */
-		err = add_to_page_cache(page, &swapper_space, entry.val);
+		err = add_to_page_cache(page, &swapper_space,
+					entry.val, GFP_ATOMIC);
 
 		if (pf_flags & PF_MEMALLOC)
 			current->flags |= PF_MEMALLOC;