f2fs: set fsync mark only for the last dnode

In order to give atomic writes, we should consider power failure during
sync_node_pages in fsync.
So, this patch marks fsync flag only in the last dnode block.

Acked-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

1 files changed, 98 insertions, 8 deletions

diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index 8a1e21144ecb..de070a524fd2 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -1222,13 +1222,81 @@ iput_out:
 	iput(inode);
 }
 
+static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino)
+{
+	pgoff_t index, end;
+	struct pagevec pvec;
+	struct page *last_page = NULL;
+
+	pagevec_init(&pvec, 0);
+	index = 0;
+	end = ULONG_MAX;
+
+	while (index <= end) {
+		int i, nr_pages;
+		nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index,
+				PAGECACHE_TAG_DIRTY,
+				min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
+		if (nr_pages == 0)
+			break;
+
+		for (i = 0; i < nr_pages; i++) {
+			struct page *page = pvec.pages[i];
+
+			if (unlikely(f2fs_cp_error(sbi))) {
+				f2fs_put_page(last_page, 0);
+				pagevec_release(&pvec);
+				return ERR_PTR(-EIO);
+			}
+
+			if (!IS_DNODE(page) || !is_cold_node(page))
+				continue;
+			if (ino_of_node(page) != ino)
+				continue;
+
+			lock_page(page);
+
+			if (unlikely(page->mapping != NODE_MAPPING(sbi))) {
+continue_unlock:
+				unlock_page(page);
+				continue;
+			}
+			if (ino_of_node(page) != ino)
+				goto continue_unlock;
+
+			if (!PageDirty(page)) {
+				/* someone wrote it for us */
+				goto continue_unlock;
+			}
+
+			if (last_page)
+				f2fs_put_page(last_page, 0);
+
+			get_page(page);
+			last_page = page;
+			unlock_page(page);
+		}
+		pagevec_release(&pvec);
+		cond_resched();
+	}
+	return last_page;
+}
+
 int fsync_node_pages(struct f2fs_sb_info *sbi, nid_t ino,
-					struct writeback_control *wbc)
+			struct writeback_control *wbc, bool atomic)
 {
 	pgoff_t index, end;
 	struct pagevec pvec;
 	int ret = 0;
+	struct page *last_page = NULL;
+	bool marked = false;
 
+	if (atomic) {
+		last_page = last_fsync_dnode(sbi, ino);
+		if (IS_ERR_OR_NULL(last_page))
+			return PTR_ERR_OR_ZERO(last_page);
+	}
+retry:
 	pagevec_init(&pvec, 0);
 	index = 0;
 	end = ULONG_MAX;
@@ -1245,6 +1313,7 @@ int fsync_node_pages(struct f2fs_sb_info *sbi, nid_t ino,
 			struct page *page = pvec.pages[i];
 
 			if (unlikely(f2fs_cp_error(sbi))) {
+				f2fs_put_page(last_page, 0);
 				pagevec_release(&pvec);
 				return -EIO;
 			}
@@ -1264,33 +1333,54 @@ continue_unlock:
 			if (ino_of_node(page) != ino)
 				goto continue_unlock;
 
-			if (!PageDirty(page)) {
+			if (!PageDirty(page) && page != last_page) {
 				/* someone wrote it for us */
 				goto continue_unlock;
 			}
 
 			f2fs_wait_on_page_writeback(page, NODE, true);
 			BUG_ON(PageWriteback(page));
-			if (!clear_page_dirty_for_io(page))
-				goto continue_unlock;
 
-			set_fsync_mark(page, 1);
-			if (IS_INODE(page))
-				set_dentry_mark(page,
+			if (!atomic || page == last_page) {
+				set_fsync_mark(page, 1);
+				if (IS_INODE(page))
+					set_dentry_mark(page,
 						need_dentry_mark(sbi, ino));
+				/*  may be written by other thread */
+				if (!PageDirty(page))
+					set_page_dirty(page);
+			}
+
+			if (!clear_page_dirty_for_io(page))
+				goto continue_unlock;
 
 			ret = NODE_MAPPING(sbi)->a_ops->writepage(page, wbc);
 			if (ret) {
 				unlock_page(page);
+				f2fs_put_page(last_page, 0);
+				break;
+			}
+			if (page == last_page) {
+				f2fs_put_page(page, 0);
+				marked = true;
 				break;
 			}
 		}
 		pagevec_release(&pvec);
 		cond_resched();
 
-		if (ret)
+		if (ret || marked)
 			break;
 	}
+	if (!ret && atomic && !marked) {
+		f2fs_msg(sbi->sb, KERN_DEBUG,
+			"Retry to write fsync mark: ino=%u, idx=%lx",
+					ino, last_page->index);
+		lock_page(last_page);
+		set_page_dirty(last_page);
+		unlock_page(last_page);
+		goto retry;
+	}
 	return ret ? -EIO: 0;
 }