Merge tag 'for-f2fs-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs

Pull f2fs updates from Jaegeuk Kim:
 "This series includes patches to:
   - add nobarrier mount option
   - support tmpfile and rename2
   - enhance the fdatasync behavior
   - fix the error path
   - fix the recovery routine
   - refactor a part of the checkpoint procedure
   - reduce some lock contentions"

* tag 'for-f2fs-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (40 commits)
  f2fs: use for_each_set_bit to simplify the code
  f2fs: add f2fs_balance_fs for expand_inode_data
  f2fs: invalidate xattr node page when evict inode
  f2fs: avoid skipping recover_inline_xattr after recover_inline_data
  f2fs: add tracepoint for f2fs_direct_IO
  f2fs: reduce competition among node page writes
  f2fs: fix coding style
  f2fs: remove redundant lines in allocate_data_block
  f2fs: add tracepoint for f2fs_issue_flush
  f2fs: avoid retrying wrong recovery routine when error was occurred
  f2fs: test before set/clear bits
  f2fs: fix wrong condition for unlikely
  f2fs: enable in-place-update for fdatasync
  f2fs: skip unnecessary data writes during fsync
  f2fs: add info of appended or updated data writes
  f2fs: use radix_tree for ino management
  f2fs: add infra for ino management
  f2fs: punch the core function for inode management
  f2fs: add nobarrier mount option
  f2fs: fix to put root inode in error path of fill_super
  ...

Documentation/filesystems/f2fs.txt

@@ -126,6 +126,11 @@ flush_merge	       Merge concurrent cache_flush commands as much as possible
                        to eliminate redundant command issues. If the underlying
 		       device handles the cache_flush command relatively slowly,
 		       recommend to enable this option.
+nobarrier              This option can be used if underlying storage guarantees
+                       its cached data should be written to the novolatile area.
+		       If this option is set, no cache_flush commands are issued
+		       but f2fs still guarantees the write ordering of all the
+		       data writes.
 
 ================================================================================
 DEBUGFS ENTRIES

fs/f2fs/acl.c

@@ -203,12 +203,6 @@ static int __f2fs_set_acl(struct inode *inode, int type,
 	size_t size = 0;
 	int error;
 
-	if (acl) {
-		error = posix_acl_valid(acl);
-		if (error < 0)
-			return error;
-	}
-
 	switch (type) {
 	case ACL_TYPE_ACCESS:
 		name_index = F2FS_XATTR_INDEX_POSIX_ACL_ACCESS;

fs/f2fs/checkpoint.c

@@ -22,7 +22,7 @@
 #include "segment.h"
 #include <trace/events/f2fs.h>
 
-static struct kmem_cache *orphan_entry_slab;
+static struct kmem_cache *ino_entry_slab;
 static struct kmem_cache *inode_entry_slab;
 
 /*
@@ -282,72 +282,120 @@ const struct address_space_operations f2fs_meta_aops = {
 	.set_page_dirty	= f2fs_set_meta_page_dirty,
 };
 
+static void __add_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
+{
+	struct ino_entry *e;
+retry:
+	spin_lock(&sbi->ino_lock[type]);
+
+	e = radix_tree_lookup(&sbi->ino_root[type], ino);
+	if (!e) {
+		e = kmem_cache_alloc(ino_entry_slab, GFP_ATOMIC);
+		if (!e) {
+			spin_unlock(&sbi->ino_lock[type]);
+			goto retry;
+		}
+		if (radix_tree_insert(&sbi->ino_root[type], ino, e)) {
+			spin_unlock(&sbi->ino_lock[type]);
+			kmem_cache_free(ino_entry_slab, e);
+			goto retry;
+		}
+		memset(e, 0, sizeof(struct ino_entry));
+		e->ino = ino;
+
+		list_add_tail(&e->list, &sbi->ino_list[type]);
+	}
+	spin_unlock(&sbi->ino_lock[type]);
+}
+
+static void __remove_ino_entry(struct f2fs_sb_info *sbi, nid_t ino, int type)
+{
+	struct ino_entry *e;
+
+	spin_lock(&sbi->ino_lock[type]);
+	e = radix_tree_lookup(&sbi->ino_root[type], ino);
+	if (e) {
+		list_del(&e->list);
+		radix_tree_delete(&sbi->ino_root[type], ino);
+		if (type == ORPHAN_INO)
+			sbi->n_orphans--;
+		spin_unlock(&sbi->ino_lock[type]);
+		kmem_cache_free(ino_entry_slab, e);
+		return;
+	}
+	spin_unlock(&sbi->ino_lock[type]);
+}
+
+void add_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type)
+{
+	/* add new dirty ino entry into list */
+	__add_ino_entry(sbi, ino, type);
+}
+
+void remove_dirty_inode(struct f2fs_sb_info *sbi, nid_t ino, int type)
+{
+	/* remove dirty ino entry from list */
+	__remove_ino_entry(sbi, ino, type);
+}
+
+/* mode should be APPEND_INO or UPDATE_INO */
+bool exist_written_data(struct f2fs_sb_info *sbi, nid_t ino, int mode)
+{
+	struct ino_entry *e;
+	spin_lock(&sbi->ino_lock[mode]);
+	e = radix_tree_lookup(&sbi->ino_root[mode], ino);
+	spin_unlock(&sbi->ino_lock[mode]);
+	return e ? true : false;
+}
+
+static void release_dirty_inode(struct f2fs_sb_info *sbi)
+{
+	struct ino_entry *e, *tmp;
+	int i;
+
+	for (i = APPEND_INO; i <= UPDATE_INO; i++) {
+		spin_lock(&sbi->ino_lock[i]);
+		list_for_each_entry_safe(e, tmp, &sbi->ino_list[i], list) {
+			list_del(&e->list);
+			radix_tree_delete(&sbi->ino_root[i], e->ino);
+			kmem_cache_free(ino_entry_slab, e);
+		}
+		spin_unlock(&sbi->ino_lock[i]);
+	}
+}
+
 int acquire_orphan_inode(struct f2fs_sb_info *sbi)
 {
 	int err = 0;
 
-	spin_lock(&sbi->orphan_inode_lock);
+	spin_lock(&sbi->ino_lock[ORPHAN_INO]);
 	if (unlikely(sbi->n_orphans >= sbi->max_orphans))
 		err = -ENOSPC;
 	else
 		sbi->n_orphans++;
-	spin_unlock(&sbi->orphan_inode_lock);
+	spin_unlock(&sbi->ino_lock[ORPHAN_INO]);
 
 	return err;
 }
 
 void release_orphan_inode(struct f2fs_sb_info *sbi)
 {
-	spin_lock(&sbi->orphan_inode_lock);
+	spin_lock(&sbi->ino_lock[ORPHAN_INO]);
 	f2fs_bug_on(sbi->n_orphans == 0);
 	sbi->n_orphans--;
-	spin_unlock(&sbi->orphan_inode_lock);
+	spin_unlock(&sbi->ino_lock[ORPHAN_INO]);
 }
 
 void add_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
 {
-	struct list_head *head;
-	struct orphan_inode_entry *new, *orphan;
-
-	new = f2fs_kmem_cache_alloc(orphan_entry_slab, GFP_ATOMIC);
-	new->ino = ino;
-
-	spin_lock(&sbi->orphan_inode_lock);
-	head = &sbi->orphan_inode_list;
-	list_for_each_entry(orphan, head, list) {
-		if (orphan->ino == ino) {
-			spin_unlock(&sbi->orphan_inode_lock);
-			kmem_cache_free(orphan_entry_slab, new);
-			return;
-		}
-
-		if (orphan->ino > ino)
-			break;
-	}
-
-	/* add new orphan entry into list which is sorted by inode number */
-	list_add_tail(&new->list, &orphan->list);
-	spin_unlock(&sbi->orphan_inode_lock);
+	/* add new orphan ino entry into list */
+	__add_ino_entry(sbi, ino, ORPHAN_INO);
 }
 
 void remove_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
 {
-	struct list_head *head;
-	struct orphan_inode_entry *orphan;
-
-	spin_lock(&sbi->orphan_inode_lock);
-	head = &sbi->orphan_inode_list;
-	list_for_each_entry(orphan, head, list) {
-		if (orphan->ino == ino) {
-			list_del(&orphan->list);
-			f2fs_bug_on(sbi->n_orphans == 0);
-			sbi->n_orphans--;
-			spin_unlock(&sbi->orphan_inode_lock);
-			kmem_cache_free(orphan_entry_slab, orphan);
-			return;
-		}
-	}
-	spin_unlock(&sbi->orphan_inode_lock);
+	/* remove orphan entry from orphan list */
+	__remove_ino_entry(sbi, ino, ORPHAN_INO);
 }
 
 static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
@@ -401,14 +449,14 @@ static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
 	unsigned short orphan_blocks = (unsigned short)((sbi->n_orphans +
 		(F2FS_ORPHANS_PER_BLOCK - 1)) / F2FS_ORPHANS_PER_BLOCK);
 	struct page *page = NULL;
-	struct orphan_inode_entry *orphan = NULL;
+	struct ino_entry *orphan = NULL;
 
 	for (index = 0; index < orphan_blocks; index++)
 		grab_meta_page(sbi, start_blk + index);
 
 	index = 1;
-	spin_lock(&sbi->orphan_inode_lock);
-	head = &sbi->orphan_inode_list;
+	spin_lock(&sbi->ino_lock[ORPHAN_INO]);
+	head = &sbi->ino_list[ORPHAN_INO];
 
 	/* loop for each orphan inode entry and write them in Jornal block */
 	list_for_each_entry(orphan, head, list) {
@@ -448,7 +496,7 @@ static void write_orphan_inodes(struct f2fs_sb_info *sbi, block_t start_blk)
 		f2fs_put_page(page, 1);
 	}
 
-	spin_unlock(&sbi->orphan_inode_lock);
+	spin_unlock(&sbi->ino_lock[ORPHAN_INO]);
 }
 
 static struct page *validate_checkpoint(struct f2fs_sb_info *sbi,
@@ -714,10 +762,10 @@ static void block_operations(struct f2fs_sb_info *sbi)
 	 * until finishing nat/sit flush.
 	 */
 retry_flush_nodes:
-	mutex_lock(&sbi->node_write);
+	down_write(&sbi->node_write);
 
 	if (get_pages(sbi, F2FS_DIRTY_NODES)) {
-		mutex_unlock(&sbi->node_write);
+		up_write(&sbi->node_write);
 		sync_node_pages(sbi, 0, &wbc);
 		goto retry_flush_nodes;
 	}
@@ -726,7 +774,7 @@ static void block_operations(struct f2fs_sb_info *sbi)
 
 static void unblock_operations(struct f2fs_sb_info *sbi)
 {
-	mutex_unlock(&sbi->node_write);
+	up_write(&sbi->node_write);
 	f2fs_unlock_all(sbi);
 }
 
@@ -748,6 +796,7 @@ static void wait_on_all_pages_writeback(struct f2fs_sb_info *sbi)
 static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
 {
 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
+	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
 	nid_t last_nid = 0;
 	block_t start_blk;
 	struct page *cp_page;
@@ -761,7 +810,7 @@ static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
 	 * This avoids to conduct wrong roll-forward operations and uses
 	 * metapages, so should be called prior to sync_meta_pages below.
 	 */
-	discard_next_dnode(sbi);
+	discard_next_dnode(sbi, NEXT_FREE_BLKADDR(sbi, curseg));
 
 	/* Flush all the NAT/SIT pages */
 	while (get_pages(sbi, F2FS_DIRTY_META))
@@ -885,8 +934,9 @@ static void do_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
 	/* Here, we only have one bio having CP pack */
 	sync_meta_pages(sbi, META_FLUSH, LONG_MAX);
 
-	if (unlikely(!is_set_ckpt_flags(ckpt, CP_ERROR_FLAG))) {
+	if (!is_set_ckpt_flags(ckpt, CP_ERROR_FLAG)) {
 		clear_prefree_segments(sbi);
+		release_dirty_inode(sbi);
 		F2FS_RESET_SB_DIRT(sbi);
 	}
 }
@@ -932,31 +982,37 @@ void write_checkpoint(struct f2fs_sb_info *sbi, bool is_umount)
 	trace_f2fs_write_checkpoint(sbi->sb, is_umount, "finish checkpoint");
 }
 
-void init_orphan_info(struct f2fs_sb_info *sbi)
+void init_ino_entry_info(struct f2fs_sb_info *sbi)
 {
-	spin_lock_init(&sbi->orphan_inode_lock);
-	INIT_LIST_HEAD(&sbi->orphan_inode_list);
-	sbi->n_orphans = 0;
+	int i;
+
+	for (i = 0; i < MAX_INO_ENTRY; i++) {
+		INIT_RADIX_TREE(&sbi->ino_root[i], GFP_ATOMIC);
+		spin_lock_init(&sbi->ino_lock[i]);
+		INIT_LIST_HEAD(&sbi->ino_list[i]);
+	}
+
 	/*
 	 * considering 512 blocks in a segment 8 blocks are needed for cp
 	 * and log segment summaries. Remaining blocks are used to keep
 	 * orphan entries with the limitation one reserved segment
 	 * for cp pack we can have max 1020*504 orphan entries
 	 */
+	sbi->n_orphans = 0;
 	sbi->max_orphans = (sbi->blocks_per_seg - 2 - NR_CURSEG_TYPE)
 				* F2FS_ORPHANS_PER_BLOCK;
 }
 
 int __init create_checkpoint_caches(void)
 {
-	orphan_entry_slab = f2fs_kmem_cache_create("f2fs_orphan_entry",
-			sizeof(struct orphan_inode_entry));
-	if (!orphan_entry_slab)
+	ino_entry_slab = f2fs_kmem_cache_create("f2fs_ino_entry",
+			sizeof(struct ino_entry));
+	if (!ino_entry_slab)
 		return -ENOMEM;
 	inode_entry_slab = f2fs_kmem_cache_create("f2fs_dirty_dir_entry",
 			sizeof(struct dir_inode_entry));
 	if (!inode_entry_slab) {
-		kmem_cache_destroy(orphan_entry_slab);
+		kmem_cache_destroy(ino_entry_slab);
 		return -ENOMEM;
 	}
 	return 0;
@@ -964,6 +1020,6 @@ int __init create_checkpoint_caches(void)
 
 void destroy_checkpoint_caches(void)
 {
-	kmem_cache_destroy(orphan_entry_slab);
+	kmem_cache_destroy(ino_entry_slab);
 	kmem_cache_destroy(inode_entry_slab);
 }

fs/f2fs/data.c

@@ -139,6 +139,9 @@ void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi,
 	/* change META to META_FLUSH in the checkpoint procedure */
 	if (type >= META_FLUSH) {
 		io->fio.type = META_FLUSH;
+		if (test_opt(sbi, NOBARRIER))
+			io->fio.rw = WRITE_FLUSH | REQ_META | REQ_PRIO;
+		else
 			io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO;
 	}
 	__submit_merged_bio(io);
@@ -626,8 +629,10 @@ static int __get_data_block(struct inode *inode, sector_t iblock,
 	if (check_extent_cache(inode, pgofs, bh_result))
 		goto out;
 
-	if (create)
+	if (create) {
+		f2fs_balance_fs(sbi);
 		f2fs_lock_op(sbi);
+	}
 
 	/* When reading holes, we need its node page */
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
@@ -784,9 +789,11 @@ int do_write_data_page(struct page *page, struct f2fs_io_info *fio)
 			!is_cold_data(page) &&
 			need_inplace_update(inode))) {
 		rewrite_data_page(page, old_blkaddr, fio);
+		set_inode_flag(F2FS_I(inode), FI_UPDATE_WRITE);
 	} else {
 		write_data_page(page, &dn, &new_blkaddr, fio);
 		update_extent_cache(new_blkaddr, &dn);
+		set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE);
 	}
 out_writepage:
 	f2fs_put_dnode(&dn);
@@ -914,6 +921,16 @@ static int f2fs_write_data_pages(struct address_space *mapping,
 	return 0;
 }
 
+static void f2fs_write_failed(struct address_space *mapping, loff_t to)
+{
+	struct inode *inode = mapping->host;
+
+	if (to > inode->i_size) {
+		truncate_pagecache(inode, inode->i_size);
+		truncate_blocks(inode, inode->i_size);
+	}
+}
+
 static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 		loff_t pos, unsigned len, unsigned flags,
 		struct page **pagep, void **fsdata)
@@ -931,11 +948,13 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 repeat:
 	err = f2fs_convert_inline_data(inode, pos + len);
 	if (err)
-		return err;
+		goto fail;
 
 	page = grab_cache_page_write_begin(mapping, index, flags);
-	if (!page)
-		return -ENOMEM;
+	if (!page) {
+		err = -ENOMEM;
+		goto fail;
+	}
 
 	/* to avoid latency during memory pressure */
 	unlock_page(page);
@@ -949,10 +968,9 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
 	err = f2fs_reserve_block(&dn, index);
 	f2fs_unlock_op(sbi);
-
 	if (err) {
 		f2fs_put_page(page, 0);
-		return err;
+		goto fail;
 	}
 inline_data:
 	lock_page(page);
@@ -982,19 +1000,20 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 			err = f2fs_read_inline_data(inode, page);
 			if (err) {
 				page_cache_release(page);
-				return err;
+				goto fail;
 			}
 		} else {
 			err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr,
 							READ_SYNC);
 			if (err)
-				return err;
+				goto fail;
 		}
 
 		lock_page(page);
 		if (unlikely(!PageUptodate(page))) {
 			f2fs_put_page(page, 1);
-			return -EIO;
+			err = -EIO;
+			goto fail;
 		}
 		if (unlikely(page->mapping != mapping)) {
 			f2fs_put_page(page, 1);
@@ -1005,6 +1024,9 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 	SetPageUptodate(page);
 	clear_cold_data(page);
 	return 0;
+fail:
+	f2fs_write_failed(mapping, pos + len);
+	return err;
 }
 
 static int f2fs_write_end(struct file *file,
@@ -1016,7 +1038,6 @@ static int f2fs_write_end(struct file *file,
 
 	trace_f2fs_write_end(inode, pos, len, copied);
 
-	SetPageUptodate(page);
 	set_page_dirty(page);
 
 	if (pos + copied > i_size_read(inode)) {
@@ -1050,7 +1071,10 @@ static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb,
 		struct iov_iter *iter, loff_t offset)
 {
 	struct file *file = iocb->ki_filp;
-	struct inode *inode = file->f_mapping->host;
+	struct address_space *mapping = file->f_mapping;
+	struct inode *inode = mapping->host;
+	size_t count = iov_iter_count(iter);
+	int err;
 
 	/* Let buffer I/O handle the inline data case. */
 	if (f2fs_has_inline_data(inode))
@@ -1062,8 +1086,15 @@ static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb,
 	/* clear fsync mark to recover these blocks */
 	fsync_mark_clear(F2FS_SB(inode->i_sb), inode->i_ino);
 
-	return blockdev_direct_IO(rw, iocb, inode, iter, offset,
-				  get_data_block);
+	trace_f2fs_direct_IO_enter(inode, offset, count, rw);
+
+	err = blockdev_direct_IO(rw, iocb, inode, iter, offset, get_data_block);
+	if (err < 0 && (rw & WRITE))
+		f2fs_write_failed(mapping, offset + count);
+
+	trace_f2fs_direct_IO_exit(inode, offset, count, rw, err);
+
+	return err;
 }
 
 static void f2fs_invalidate_data_page(struct page *page, unsigned int offset,

fs/f2fs/debug.c

@@ -167,7 +167,7 @@ static void update_mem_info(struct f2fs_sb_info *sbi)
 	si->cache_mem += npages << PAGE_CACHE_SHIFT;
 	npages = META_MAPPING(sbi)->nrpages;
 	si->cache_mem += npages << PAGE_CACHE_SHIFT;
-	si->cache_mem += sbi->n_orphans * sizeof(struct orphan_inode_entry);
+	si->cache_mem += sbi->n_orphans * sizeof(struct ino_entry);
 	si->cache_mem += sbi->n_dirty_dirs * sizeof(struct dir_inode_entry);
 }
 
@@ -345,21 +345,14 @@ void __init f2fs_create_root_stats(void)
 
 	f2fs_debugfs_root = debugfs_create_dir("f2fs", NULL);
 	if (!f2fs_debugfs_root)
-		goto bail;
+		return;
 
 	file = debugfs_create_file("status", S_IRUGO, f2fs_debugfs_root,
 			NULL, &stat_fops);
-	if (!file)
-		goto free_debugfs_dir;
-
-	return;
-
-free_debugfs_dir:
+	if (!file) {
 		debugfs_remove(f2fs_debugfs_root);
-
-bail:
 		f2fs_debugfs_root = NULL;
-	return;
+	}
 }
 
 void f2fs_destroy_root_stats(void)

fs/f2fs/dir.c

@@ -77,8 +77,8 @@ static unsigned long dir_block_index(unsigned int level,
 	return bidx;
 }
 
-static bool early_match_name(const char *name, size_t namelen,
-			f2fs_hash_t namehash, struct f2fs_dir_entry *de)
+static bool early_match_name(size_t namelen, f2fs_hash_t namehash,
+				struct f2fs_dir_entry *de)
 {
 	if (le16_to_cpu(de->name_len) != namelen)
 		return false;
@@ -90,7 +90,7 @@ static bool early_match_name(const char *name, size_t namelen,
 }
 
 static struct f2fs_dir_entry *find_in_block(struct page *dentry_page,
-			const char *name, size_t namelen, int *max_slots,
+			struct qstr *name, int *max_slots,
 			f2fs_hash_t namehash, struct page **res_page)
 {
 	struct f2fs_dir_entry *de;
@@ -109,9 +109,10 @@ static struct f2fs_dir_entry *find_in_block(struct page *dentry_page,
 			continue;
 		}
 		de = &dentry_blk->dentry[bit_pos];
-		if (early_match_name(name, namelen, namehash, de)) {
+		if (early_match_name(name->len, namehash, de)) {
 			if (!memcmp(dentry_blk->filename[bit_pos],
-							name, namelen)) {
+							name->name,
+							name->len)) {
 				*res_page = dentry_page;
 				goto found;
 			}
@@ -120,6 +121,13 @@ static struct f2fs_dir_entry *find_in_block(struct page *dentry_page,
 			*max_slots = max_len;
 			max_len = 0;
 		}
+
+		/*
+		 * For the most part, it should be a bug when name_len is zero.
+		 * We stop here for figuring out where the bugs are occurred.
+		 */
+		f2fs_bug_on(!de->name_len);
+
 		bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
 	}
 
@@ -132,10 +140,10 @@ static struct f2fs_dir_entry *find_in_block(struct page *dentry_page,
 }
 
 static struct f2fs_dir_entry *find_in_level(struct inode *dir,
-		unsigned int level, const char *name, size_t namelen,
+			unsigned int level, struct qstr *name,
 			f2fs_hash_t namehash, struct page **res_page)
 {
-	int s = GET_DENTRY_SLOTS(namelen);
+	int s = GET_DENTRY_SLOTS(name->len);
 	unsigned int nbucket, nblock;
 	unsigned int bidx, end_block;
 	struct page *dentry_page;
@@ -160,8 +168,8 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
 			continue;
 		}
 
-		de = find_in_block(dentry_page, name, namelen,
-					&max_slots, namehash, res_page);
+		de = find_in_block(dentry_page, name, &max_slots,
+					namehash, res_page);
 		if (de)
 			break;
 
@@ -187,8 +195,6 @@ static struct f2fs_dir_entry *find_in_level(struct inode *dir,
 struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
 			struct qstr *child, struct page **res_page)
 {
-	const char *name = child->name;
-	size_t namelen = child->len;
 	unsigned long npages = dir_blocks(dir);
 	struct f2fs_dir_entry *de = NULL;
 	f2fs_hash_t name_hash;
@@ -200,12 +206,11 @@ struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
 
 	*res_page = NULL;
 
-	name_hash = f2fs_dentry_hash(name, namelen);
+	name_hash = f2fs_dentry_hash(child);
 	max_depth = F2FS_I(dir)->i_current_depth;
 
 	for (level = 0; level < max_depth; level++) {
-		de = find_in_level(dir, level, name,
-				namelen, name_hash, res_page);
+		de = find_in_level(dir, level, child, name_hash, res_page);
 		if (de)
 			break;
 	}
@@ -298,14 +303,13 @@ static int make_empty_dir(struct inode *inode,
 	struct page *dentry_page;
 	struct f2fs_dentry_block *dentry_blk;
 	struct f2fs_dir_entry *de;
-	void *kaddr;
 
 	dentry_page = get_new_data_page(inode, page, 0, true);
 	if (IS_ERR(dentry_page))
 		return PTR_ERR(dentry_page);
 
-	kaddr = kmap_atomic(dentry_page);
-	dentry_blk = (struct f2fs_dentry_block *)kaddr;
+
+	dentry_blk = kmap_atomic(dentry_page);
 
 	de = &dentry_blk->dentry[0];
 	de->name_len = cpu_to_le16(1);
@@ -323,7 +327,7 @@ static int make_empty_dir(struct inode *inode,
 
 	test_and_set_bit_le(0, &dentry_blk->dentry_bitmap);
 	test_and_set_bit_le(1, &dentry_blk->dentry_bitmap);
-	kunmap_atomic(kaddr);
+	kunmap_atomic(dentry_blk);
 
 	set_page_dirty(dentry_page);
 	f2fs_put_page(dentry_page, 1);
@@ -333,11 +337,12 @@ static int make_empty_dir(struct inode *inode,
 static struct page *init_inode_metadata(struct inode *inode,
 		struct inode *dir, const struct qstr *name)
 {
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	struct page *page;
 	int err;
 
 	if (is_inode_flag_set(F2FS_I(inode), FI_NEW_INODE)) {
-		page = new_inode_page(inode, name);
+		page = new_inode_page(inode);
 		if (IS_ERR(page))
 			return page;
 
@@ -362,6 +367,7 @@ static struct page *init_inode_metadata(struct inode *inode,
 		set_cold_node(inode, page);
 	}
 
+	if (name)
 		init_dent_inode(name, page);
 
 	/*
@@ -370,6 +376,12 @@ static struct page *init_inode_metadata(struct inode *inode,
 	 */
 	if (is_inode_flag_set(F2FS_I(inode), FI_INC_LINK)) {
 		file_lost_pino(inode);
+		/*
+		 * If link the tmpfile to alias through linkat path,
+		 * we should remove this inode from orphan list.
+		 */
+		if (inode->i_nlink == 0)
+			remove_orphan_inode(sbi, inode->i_ino);
 		inc_nlink(inode);
 	}
 	return page;
@@ -453,7 +465,7 @@ int __f2fs_add_link(struct inode *dir, const struct qstr *name,
 	int err = 0;
 	int i;
 
-	dentry_hash = f2fs_dentry_hash(name->name, name->len);
+	dentry_hash = f2fs_dentry_hash(name);
 	level = 0;
 	current_depth = F2FS_I(dir)->i_current_depth;
 	if (F2FS_I(dir)->chash == dentry_hash) {
@@ -529,6 +541,27 @@ int __f2fs_add_link(struct inode *dir, const struct qstr *name,
 	return err;
 }
 
+int f2fs_do_tmpfile(struct inode *inode, struct inode *dir)
+{
+	struct page *page;
+	int err = 0;
+
+	down_write(&F2FS_I(inode)->i_sem);
+	page = init_inode_metadata(inode, dir, NULL);
+	if (IS_ERR(page)) {
+		err = PTR_ERR(page);
+		goto fail;
+	}
+	/* we don't need to mark_inode_dirty now */
+	update_inode(inode, page);
+	f2fs_put_page(page, 1);
+
+	clear_inode_flag(F2FS_I(inode), FI_NEW_INODE);
+fail:
+	up_write(&F2FS_I(inode)->i_sem);
+	return err;
+}
+
 /*
  * It only removes the dentry from the dentry page,corresponding name
  * entry in name page does not need to be touched during deletion.
@@ -541,14 +574,13 @@ void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
 	struct address_space *mapping = page->mapping;
 	struct inode *dir = mapping->host;
 	int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
-	void *kaddr = page_address(page);
 	int i;
 
 	lock_page(page);
 	f2fs_wait_on_page_writeback(page, DATA);
 
-	dentry_blk = (struct f2fs_dentry_block *)kaddr;
-	bit_pos = dentry - (struct f2fs_dir_entry *)dentry_blk->dentry;
+	dentry_blk = page_address(page);
+	bit_pos = dentry - dentry_blk->dentry;
 	for (i = 0; i < slots; i++)
 		test_and_clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
 
@@ -603,7 +635,6 @@ bool f2fs_empty_dir(struct inode *dir)
 	unsigned long nblock = dir_blocks(dir);
 
 	for (bidx = 0; bidx < nblock; bidx++) {
-		void *kaddr;
 		dentry_page = get_lock_data_page(dir, bidx);
 		if (IS_ERR(dentry_page)) {
 			if (PTR_ERR(dentry_page) == -ENOENT)
@@ -612,8 +643,8 @@ bool f2fs_empty_dir(struct inode *dir)
 				return false;
 		}
 
-		kaddr = kmap_atomic(dentry_page);
-		dentry_blk = (struct f2fs_dentry_block *)kaddr;
+
+		dentry_blk = kmap_atomic(dentry_page);
 		if (bidx == 0)
 			bit_pos = 2;
 		else
@@ -621,7 +652,7 @@ bool f2fs_empty_dir(struct inode *dir)
 		bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
 						NR_DENTRY_IN_BLOCK,
 						bit_pos);
-		kunmap_atomic(kaddr);
+		kunmap_atomic(dentry_blk);
 
 		f2fs_put_page(dentry_page, 1);
 

fs/f2fs/f2fs.h

@@ -41,6 +41,7 @@
 #define F2FS_MOUNT_INLINE_XATTR		0x00000080
 #define F2FS_MOUNT_INLINE_DATA		0x00000100
 #define F2FS_MOUNT_FLUSH_MERGE		0x00000200
+#define F2FS_MOUNT_NOBARRIER		0x00000400
 
 #define clear_opt(sbi, option)	(sbi->mount_opt.opt &= ~F2FS_MOUNT_##option)
 #define set_opt(sbi, option)	(sbi->mount_opt.opt |= F2FS_MOUNT_##option)
@@ -99,8 +100,15 @@ enum {
 	META_SSA
 };
 
-/* for the list of orphan inodes */
-struct orphan_inode_entry {
+/* for the list of ino */
+enum {
+	ORPHAN_INO,		/* for orphan ino list */
+	APPEND_INO,		/* for append ino list */
+	UPDATE_INO,		/* for update ino list */
+	MAX_INO_ENTRY,		/* max. list */
+};
+
+struct ino_entry {
 	struct list_head list;	/* list head */
 	nid_t ino;		/* inode number */
 };
@@ -256,6 +264,8 @@ struct f2fs_nm_info {
 	unsigned int nat_cnt;		/* the # of cached nat entries */
 	struct list_head nat_entries;	/* cached nat entry list (clean) */
 	struct list_head dirty_nat_entries; /* cached nat entry list (dirty) */
+	struct list_head nat_entry_set;	/* nat entry set list */
+	unsigned int dirty_nat_cnt;	/* total num of nat entries in set */
 
 	/* free node ids management */
 	struct radix_tree_root free_nid_root;/* root of the free_nid cache */
@@ -442,14 +452,17 @@ struct f2fs_sb_info {
 	struct inode *meta_inode;		/* cache meta blocks */
 	struct mutex cp_mutex;			/* checkpoint procedure lock */
 	struct rw_semaphore cp_rwsem;		/* blocking FS operations */
-	struct mutex node_write;		/* locking node writes */
+	struct rw_semaphore node_write;		/* locking node writes */
 	struct mutex writepages;		/* mutex for writepages() */
 	bool por_doing;				/* recovery is doing or not */
 	wait_queue_head_t cp_wait;
 
-	/* for orphan inode management */
-	struct list_head orphan_inode_list;	/* orphan inode list */
-	spinlock_t orphan_inode_lock;		/* for orphan inode list */
+	/* for inode management */
+	struct radix_tree_root ino_root[MAX_INO_ENTRY];	/* ino entry array */
+	spinlock_t ino_lock[MAX_INO_ENTRY];		/* for ino entry lock */
+	struct list_head ino_list[MAX_INO_ENTRY];	/* inode list head */
+
+	/* for orphan inode, use 0'th array */
 	unsigned int n_orphans;			/* # of orphan inodes */
 	unsigned int max_orphans;		/* max orphan inodes */
 
@@ -768,7 +781,7 @@ static inline void *__bitmap_ptr(struct f2fs_sb_info *sbi, int flag)
 		if (flag == NAT_BITMAP)
 			return &ckpt->sit_nat_version_bitmap;
 		else
-			return ((unsigned char *)ckpt + F2FS_BLKSIZE);
+			return (unsigned char *)ckpt + F2FS_BLKSIZE;
 	} else {
 		offset = (flag == NAT_BITMAP) ?
 			le32_to_cpu(ckpt->sit_ver_bitmap_bytesize) : 0;
@@ -983,10 +996,14 @@ enum {
 	FI_NO_EXTENT,		/* not to use the extent cache */
 	FI_INLINE_XATTR,	/* used for inline xattr */
 	FI_INLINE_DATA,		/* used for inline data*/
+	FI_APPEND_WRITE,	/* inode has appended data */
+	FI_UPDATE_WRITE,	/* inode has in-place-update data */
+	FI_NEED_IPU,		/* used fo ipu for fdatasync */
 };
 
 static inline void set_inode_flag(struct f2fs_inode_info *fi, int flag)
 {
+	if (!test_bit(flag, &fi->flags))
 		set_bit(flag, &fi->flags);
 }
 
@@ -997,6 +1014,7 @@ static inline int is_inode_flag_set(struct f2fs_inode_info *fi, int flag)
 
 static inline void clear_inode_flag(struct f2fs_inode_info *fi, int flag)
 {
+	if (test_bit(flag, &fi->flags))
 		clear_bit(flag, &fi->flags);
 }
 
@@ -1136,6 +1154,7 @@ void f2fs_set_link(struct inode *, struct f2fs_dir_entry *,
 int update_dent_inode(struct inode *, const struct qstr *);
 int __f2fs_add_link(struct inode *, const struct qstr *, struct inode *);
 void f2fs_delete_entry(struct f2fs_dir_entry *, struct page *, struct inode *);
+int f2fs_do_tmpfile(struct inode *, struct inode *);
 int f2fs_make_empty(struct inode *, struct inode *);
 bool f2fs_empty_dir(struct inode *);
 
@@ -1155,7 +1174,7 @@ void f2fs_msg(struct super_block *, const char *, const char *, ...);
 /*
  * hash.c
  */
-f2fs_hash_t f2fs_dentry_hash(const char *, size_t);
+f2fs_hash_t f2fs_dentry_hash(const struct qstr *);
 
 /*
  * node.c
@@ -1173,7 +1192,7 @@ int truncate_inode_blocks(struct inode *, pgoff_t);
 int truncate_xattr_node(struct inode *, struct page *);
 int wait_on_node_pages_writeback(struct f2fs_sb_info *, nid_t);
 void remove_inode_page(struct inode *);
-struct page *new_inode_page(struct inode *, const struct qstr *);
+struct page *new_inode_page(struct inode *);
 struct page *new_node_page(struct dnode_of_data *, unsigned int, struct page *);
 void ra_node_page(struct f2fs_sb_info *, nid_t);
 struct page *get_node_page(struct f2fs_sb_info *, pgoff_t);
@@ -1185,6 +1204,7 @@ void alloc_nid_done(struct f2fs_sb_info *, nid_t);
 void alloc_nid_failed(struct f2fs_sb_info *, nid_t);
 void recover_node_page(struct f2fs_sb_info *, struct page *,
 		struct f2fs_summary *, struct node_info *, block_t);
+void recover_inline_xattr(struct inode *, struct page *);
 bool recover_xattr_data(struct inode *, struct page *, block_t);
 int recover_inode_page(struct f2fs_sb_info *, struct page *);
 int restore_node_summary(struct f2fs_sb_info *, unsigned int,
@@ -1206,7 +1226,7 @@ void destroy_flush_cmd_control(struct f2fs_sb_info *);
 void invalidate_blocks(struct f2fs_sb_info *, block_t);
 void refresh_sit_entry(struct f2fs_sb_info *, block_t, block_t);
 void clear_prefree_segments(struct f2fs_sb_info *);
-void discard_next_dnode(struct f2fs_sb_info *);
+void discard_next_dnode(struct f2fs_sb_info *, block_t);
 int npages_for_summary_flush(struct f2fs_sb_info *);
 void allocate_new_segments(struct f2fs_sb_info *);
 struct page *get_sum_page(struct f2fs_sb_info *, unsigned int);
@@ -1240,6 +1260,9 @@ struct page *grab_meta_page(struct f2fs_sb_info *, pgoff_t);
 struct page *get_meta_page(struct f2fs_sb_info *, pgoff_t);
 int ra_meta_pages(struct f2fs_sb_info *, int, int, int);
 long sync_meta_pages(struct f2fs_sb_info *, enum page_type, long);
+void add_dirty_inode(struct f2fs_sb_info *, nid_t, int type);
+void remove_dirty_inode(struct f2fs_sb_info *, nid_t, int type);
+bool exist_written_data(struct f2fs_sb_info *, nid_t, int);
 int acquire_orphan_inode(struct f2fs_sb_info *);
 void release_orphan_inode(struct f2fs_sb_info *);
 void add_orphan_inode(struct f2fs_sb_info *, nid_t);
@@ -1251,7 +1274,7 @@ void add_dirty_dir_inode(struct inode *);
 void remove_dirty_dir_inode(struct inode *);
 void sync_dirty_dir_inodes(struct f2fs_sb_info *);
 void write_checkpoint(struct f2fs_sb_info *, bool);
-void init_orphan_info(struct f2fs_sb_info *);
+void init_ino_entry_info(struct f2fs_sb_info *);
 int __init create_checkpoint_caches(void);
 void destroy_checkpoint_caches(void);
 
@@ -1295,7 +1318,6 @@ bool space_for_roll_forward(struct f2fs_sb_info *);
 struct f2fs_stat_info {
 	struct list_head stat_list;
 	struct f2fs_sb_info *sbi;
-	struct mutex stat_lock;
 	int all_area_segs, sit_area_segs, nat_area_segs, ssa_area_segs;
 	int main_area_segs, main_area_sections, main_area_zones;
 	int hit_ext, total_ext;

fs/f2fs/file.c

@@ -127,12 +127,30 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 		return 0;
 
 	trace_f2fs_sync_file_enter(inode);
+
+	/* if fdatasync is triggered, let's do in-place-update */
+	if (datasync)
+		set_inode_flag(fi, FI_NEED_IPU);
+
 	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
+	if (datasync)
+		clear_inode_flag(fi, FI_NEED_IPU);
 	if (ret) {
 		trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
 		return ret;
 	}
 
+	/*
+	 * if there is no written data, don't waste time to write recovery info.
+	 */
+	if (!is_inode_flag_set(fi, FI_APPEND_WRITE) &&
+		!exist_written_data(sbi, inode->i_ino, APPEND_INO)) {
+		if (is_inode_flag_set(fi, FI_UPDATE_WRITE) ||
+			exist_written_data(sbi, inode->i_ino, UPDATE_INO))
+			goto flush_out;
+		goto out;
+	}
+
 	/* guarantee free sections for fsync */
 	f2fs_balance_fs(sbi);
 
@@ -188,6 +206,13 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
 		ret = wait_on_node_pages_writeback(sbi, inode->i_ino);
 		if (ret)
 			goto out;
+
+		/* once recovery info is written, don't need to tack this */
+		remove_dirty_inode(sbi, inode->i_ino, APPEND_INO);
+		clear_inode_flag(fi, FI_APPEND_WRITE);
+flush_out:
+		remove_dirty_inode(sbi, inode->i_ino, UPDATE_INO);
+		clear_inode_flag(fi, FI_UPDATE_WRITE);
 		ret = f2fs_issue_flush(F2FS_SB(inode->i_sb));
 	}
 out:
@@ -206,8 +231,9 @@ static pgoff_t __get_first_dirty_index(struct address_space *mapping,
 
 	/* find first dirty page index */
 	pagevec_init(&pvec, 0);
-	nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs, PAGECACHE_TAG_DIRTY, 1);
-	pgofs = nr_pages ? pvec.pages[0]->index: LONG_MAX;
+	nr_pages = pagevec_lookup_tag(&pvec, mapping, &pgofs,
+					PAGECACHE_TAG_DIRTY, 1);
+	pgofs = nr_pages ? pvec.pages[0]->index : LONG_MAX;
 	pagevec_release(&pvec);
 	return pgofs;
 }
@@ -272,8 +298,7 @@ static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 			}
 		}
 
-		end_offset = IS_INODE(dn.node_page) ?
-			ADDRS_PER_INODE(F2FS_I(inode)) : ADDRS_PER_BLOCK;
+		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
 
 		/* find data/hole in dnode block */
 		for (; dn.ofs_in_node < end_offset;
@@ -380,13 +405,15 @@ static void truncate_partial_data_page(struct inode *inode, u64 from)
 		return;
 
 	lock_page(page);
-	if (unlikely(page->mapping != inode->i_mapping)) {
-		f2fs_put_page(page, 1);
-		return;
-	}
+	if (unlikely(!PageUptodate(page) ||
+			page->mapping != inode->i_mapping))
+		goto out;
+
 	f2fs_wait_on_page_writeback(page, DATA);
 	zero_user(page, offset, PAGE_CACHE_SIZE - offset);
 	set_page_dirty(page);
+
+out:
 	f2fs_put_page(page, 1);
 }
 
@@ -645,6 +672,8 @@ static int expand_inode_data(struct inode *inode, loff_t offset,
 	loff_t off_start, off_end;
 	int ret = 0;
 
+	f2fs_balance_fs(sbi);
+
 	ret = inode_newsize_ok(inode, (len + offset));
 	if (ret)
 		return ret;

fs/f2fs/gc.c

@@ -186,7 +186,6 @@ static unsigned int get_max_cost(struct f2fs_sb_info *sbi,
 static unsigned int check_bg_victims(struct f2fs_sb_info *sbi)
 {
 	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
-	unsigned int hint = 0;
 	unsigned int secno;
 
 	/*
@@ -194,11 +193,9 @@ static unsigned int check_bg_victims(struct f2fs_sb_info *sbi)
 	 * selected by background GC before.
 	 * Those segments guarantee they have small valid blocks.
 	 */
-next:
-	secno = find_next_bit(dirty_i->victim_secmap, TOTAL_SECS(sbi), hint++);
-	if (secno < TOTAL_SECS(sbi)) {
+	for_each_set_bit(secno, dirty_i->victim_secmap, TOTAL_SECS(sbi)) {
 		if (sec_usage_check(sbi, secno))
-			goto next;
+			continue;
 		clear_bit(secno, dirty_i->victim_secmap);
 		return secno * sbi->segs_per_sec;
 	}

fs/f2fs/hash.c

@@ -69,12 +69,14 @@ static void str2hashbuf(const char *msg, size_t len, unsigned int *buf, int num)
 		*buf++ = pad;
 }
 
-f2fs_hash_t f2fs_dentry_hash(const char *name, size_t len)
+f2fs_hash_t f2fs_dentry_hash(const struct qstr *name_info)
 {
 	__u32 hash;
 	f2fs_hash_t f2fs_hash;
 	const char *p;
 	__u32 in[8], buf[4];
+	const char *name = name_info->name;
+	size_t len = name_info->len;
 
 	if ((len <= 2) && (name[0] == '.') &&
 		(name[1] == '.' || name[1] == '\0'))

fs/f2fs/inline.c

@@ -172,6 +172,7 @@ int f2fs_write_inline_data(struct inode *inode,
 		stat_inc_inline_inode(inode);
 	}
 
+	set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE);
 	sync_inode_page(&dn);
 	f2fs_put_dnode(&dn);
 

fs/f2fs/inode.c

@@ -267,13 +267,14 @@ int f2fs_write_inode(struct inode *inode, struct writeback_control *wbc)
 void f2fs_evict_inode(struct inode *inode)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
+	nid_t xnid = F2FS_I(inode)->i_xattr_nid;
 
 	trace_f2fs_evict_inode(inode);
 	truncate_inode_pages_final(&inode->i_data);
 
 	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
 			inode->i_ino == F2FS_META_INO(sbi))
-		goto no_delete;
+		goto out_clear;
 
 	f2fs_bug_on(get_dirty_dents(inode));
 	remove_dirty_dir_inode(inode);
@@ -295,6 +296,13 @@ void f2fs_evict_inode(struct inode *inode)
 
 	sb_end_intwrite(inode->i_sb);
 no_delete:
-	clear_inode(inode);
 	invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino, inode->i_ino);
+	if (xnid)
+		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);
+	if (is_inode_flag_set(F2FS_I(inode), FI_APPEND_WRITE))
+		add_dirty_inode(sbi, inode->i_ino, APPEND_INO);
+	if (is_inode_flag_set(F2FS_I(inode), FI_UPDATE_WRITE))
+		add_dirty_inode(sbi, inode->i_ino, UPDATE_INO);
+out_clear:
+	clear_inode(inode);
 }

fs/f2fs/namei.c

@@ -13,6 +13,7 @@
 #include <linux/pagemap.h>
 #include <linux/sched.h>
 #include <linux/ctype.h>
+#include <linux/dcache.h>
 
 #include "f2fs.h"
 #include "node.h"
@@ -22,14 +23,13 @@
 
 static struct inode *f2fs_new_inode(struct inode *dir, umode_t mode)
 {
-	struct super_block *sb = dir->i_sb;
-	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	nid_t ino;
 	struct inode *inode;
 	bool nid_free = false;
 	int err;
 
-	inode = new_inode(sb);
+	inode = new_inode(dir->i_sb);
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
 
@@ -102,8 +102,7 @@ static inline void set_cold_files(struct f2fs_sb_info *sbi, struct inode *inode,
 static int f2fs_create(struct inode *dir, struct dentry *dentry, umode_t mode,
 						bool excl)
 {
-	struct super_block *sb = dir->i_sb;
-	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	struct inode *inode;
 	nid_t ino = 0;
 	int err;
@@ -146,8 +145,7 @@ static int f2fs_link(struct dentry *old_dentry, struct inode *dir,
 		struct dentry *dentry)
 {
 	struct inode *inode = old_dentry->d_inode;
-	struct super_block *sb = dir->i_sb;
-	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	int err;
 
 	f2fs_balance_fs(sbi);
@@ -207,8 +205,7 @@ static struct dentry *f2fs_lookup(struct inode *dir, struct dentry *dentry,
 
 static int f2fs_unlink(struct inode *dir, struct dentry *dentry)
 {
-	struct super_block *sb = dir->i_sb;
-	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	struct inode *inode = dentry->d_inode;
 	struct f2fs_dir_entry *de;
 	struct page *page;
@@ -242,8 +239,7 @@ static int f2fs_unlink(struct inode *dir, struct dentry *dentry)
 static int f2fs_symlink(struct inode *dir, struct dentry *dentry,
 					const char *symname)
 {
-	struct super_block *sb = dir->i_sb;
-	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	struct inode *inode;
 	size_t symlen = strlen(symname) + 1;
 	int err;
@@ -330,8 +326,7 @@ static int f2fs_rmdir(struct inode *dir, struct dentry *dentry)
 static int f2fs_mknod(struct inode *dir, struct dentry *dentry,
 				umode_t mode, dev_t rdev)
 {
-	struct super_block *sb = dir->i_sb;
-	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
 	struct inode *inode;
 	int err = 0;
 
@@ -369,8 +364,7 @@ static int f2fs_mknod(struct inode *dir, struct dentry *dentry,
 static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			struct inode *new_dir, struct dentry *new_dentry)
 {
-	struct super_block *sb = old_dir->i_sb;
-	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct f2fs_sb_info *sbi = F2FS_SB(old_dir->i_sb);
 	struct inode *old_inode = old_dentry->d_inode;
 	struct inode *new_inode = new_dentry->d_inode;
 	struct page *old_dir_page;
@@ -393,8 +387,6 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 			goto out_old;
 	}
 
-	f2fs_lock_op(sbi);
-
 	if (new_inode) {
 
 		err = -ENOTEMPTY;
@@ -407,6 +399,8 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 		if (!new_entry)
 			goto out_dir;
 
+		f2fs_lock_op(sbi);
+
 		err = acquire_orphan_inode(sbi);
 		if (err)
 			goto put_out_dir;
@@ -435,9 +429,13 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 		update_inode_page(old_inode);
 		update_inode_page(new_inode);
 	} else {
+		f2fs_lock_op(sbi);
+
 		err = f2fs_add_link(new_dentry, old_inode);
-		if (err)
+		if (err) {
+			f2fs_unlock_op(sbi);
 			goto out_dir;
+		}
 
 		if (old_dir_entry) {
 			inc_nlink(new_dir);
@@ -472,6 +470,7 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	return 0;
 
 put_out_dir:
+	f2fs_unlock_op(sbi);
 	kunmap(new_page);
 	f2fs_put_page(new_page, 0);
 out_dir:
@@ -479,7 +478,151 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 		kunmap(old_dir_page);
 		f2fs_put_page(old_dir_page, 0);
 	}
+out_old:
+	kunmap(old_page);
+	f2fs_put_page(old_page, 0);
+out:
+	return err;
+}
+
+static int f2fs_cross_rename(struct inode *old_dir, struct dentry *old_dentry,
+			     struct inode *new_dir, struct dentry *new_dentry)
+{
+	struct super_block *sb = old_dir->i_sb;
+	struct f2fs_sb_info *sbi = F2FS_SB(sb);
+	struct inode *old_inode = old_dentry->d_inode;
+	struct inode *new_inode = new_dentry->d_inode;
+	struct page *old_dir_page, *new_dir_page;
+	struct page *old_page, *new_page;
+	struct f2fs_dir_entry *old_dir_entry = NULL, *new_dir_entry = NULL;
+	struct f2fs_dir_entry *old_entry, *new_entry;
+	int old_nlink = 0, new_nlink = 0;
+	int err = -ENOENT;
+
+	f2fs_balance_fs(sbi);
+
+	old_entry = f2fs_find_entry(old_dir, &old_dentry->d_name, &old_page);
+	if (!old_entry)
+		goto out;
+
+	new_entry = f2fs_find_entry(new_dir, &new_dentry->d_name, &new_page);
+	if (!new_entry)
+		goto out_old;
+
+	/* prepare for updating ".." directory entry info later */
+	if (old_dir != new_dir) {
+		if (S_ISDIR(old_inode->i_mode)) {
+			err = -EIO;
+			old_dir_entry = f2fs_parent_dir(old_inode,
+							&old_dir_page);
+			if (!old_dir_entry)
+				goto out_new;
+		}
+
+		if (S_ISDIR(new_inode->i_mode)) {
+			err = -EIO;
+			new_dir_entry = f2fs_parent_dir(new_inode,
+							&new_dir_page);
+			if (!new_dir_entry)
+				goto out_old_dir;
+		}
+	}
+
+	/*
+	 * If cross rename between file and directory those are not
+	 * in the same directory, we will inc nlink of file's parent
+	 * later, so we should check upper boundary of its nlink.
+	 */
+	if ((!old_dir_entry || !new_dir_entry) &&
+				old_dir_entry != new_dir_entry) {
+		old_nlink = old_dir_entry ? -1 : 1;
+		new_nlink = -old_nlink;
+		err = -EMLINK;
+		if ((old_nlink > 0 && old_inode->i_nlink >= F2FS_LINK_MAX) ||
+			(new_nlink > 0 && new_inode->i_nlink >= F2FS_LINK_MAX))
+			goto out_new_dir;
+	}
+
+	f2fs_lock_op(sbi);
+
+	err = update_dent_inode(old_inode, &new_dentry->d_name);
+	if (err)
+		goto out_unlock;
+
+	err = update_dent_inode(new_inode, &old_dentry->d_name);
+	if (err)
+		goto out_undo;
+
+	/* update ".." directory entry info of old dentry */
+	if (old_dir_entry)
+		f2fs_set_link(old_inode, old_dir_entry, old_dir_page, new_dir);
+
+	/* update ".." directory entry info of new dentry */
+	if (new_dir_entry)
+		f2fs_set_link(new_inode, new_dir_entry, new_dir_page, old_dir);
+
+	/* update directory entry info of old dir inode */
+	f2fs_set_link(old_dir, old_entry, old_page, new_inode);
+
+	down_write(&F2FS_I(old_inode)->i_sem);
+	file_lost_pino(old_inode);
+	up_write(&F2FS_I(old_inode)->i_sem);
+
+	update_inode_page(old_inode);
+
+	old_dir->i_ctime = CURRENT_TIME;
+	if (old_nlink) {
+		down_write(&F2FS_I(old_dir)->i_sem);
+		if (old_nlink < 0)
+			drop_nlink(old_dir);
+		else
+			inc_nlink(old_dir);
+		up_write(&F2FS_I(old_dir)->i_sem);
+	}
+	mark_inode_dirty(old_dir);
+	update_inode_page(old_dir);
+
+	/* update directory entry info of new dir inode */
+	f2fs_set_link(new_dir, new_entry, new_page, old_inode);
+
+	down_write(&F2FS_I(new_inode)->i_sem);
+	file_lost_pino(new_inode);
+	up_write(&F2FS_I(new_inode)->i_sem);
+
+	update_inode_page(new_inode);
+
+	new_dir->i_ctime = CURRENT_TIME;
+	if (new_nlink) {
+		down_write(&F2FS_I(new_dir)->i_sem);
+		if (new_nlink < 0)
+			drop_nlink(new_dir);
+		else
+			inc_nlink(new_dir);
+		up_write(&F2FS_I(new_dir)->i_sem);
+	}
+	mark_inode_dirty(new_dir);
+	update_inode_page(new_dir);
+
 	f2fs_unlock_op(sbi);
+	return 0;
+out_undo:
+	/* Still we may fail to recover name info of f2fs_inode here */
+	update_dent_inode(old_inode, &old_dentry->d_name);
+out_unlock:
+	f2fs_unlock_op(sbi);
+out_new_dir:
+	if (new_dir_entry) {
+		kunmap(new_dir_page);
+		f2fs_put_page(new_dir_page, 0);
+	}
+out_old_dir:
+	if (old_dir_entry) {
+		kunmap(old_dir_page);
+		f2fs_put_page(old_dir_page, 0);
+	}
+out_new:
+	kunmap(new_page);
+	f2fs_put_page(new_page, 0);
 out_old:
 	kunmap(old_page);
 	f2fs_put_page(old_page, 0);
@@ -487,6 +630,71 @@ static int f2fs_rename(struct inode *old_dir, struct dentry *old_dentry,
 	return err;
 }
 
+static int f2fs_rename2(struct inode *old_dir, struct dentry *old_dentry,
+			struct inode *new_dir, struct dentry *new_dentry,
+			unsigned int flags)
+{
+	if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE))
+		return -EINVAL;
+
+	if (flags & RENAME_EXCHANGE) {
+		return f2fs_cross_rename(old_dir, old_dentry,
+					 new_dir, new_dentry);
+	}
+	/*
+	 * VFS has already handled the new dentry existence case,
+	 * here, we just deal with "RENAME_NOREPLACE" as regular rename.
+	 */
+	return f2fs_rename(old_dir, old_dentry, new_dir, new_dentry);
+}
+
+static int f2fs_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode)
+{
+	struct f2fs_sb_info *sbi = F2FS_SB(dir->i_sb);
+	struct inode *inode;
+	int err;
+
+	inode = f2fs_new_inode(dir, mode);
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
+
+	inode->i_op = &f2fs_file_inode_operations;
+	inode->i_fop = &f2fs_file_operations;
+	inode->i_mapping->a_ops = &f2fs_dblock_aops;
+
+	f2fs_lock_op(sbi);
+	err = acquire_orphan_inode(sbi);
+	if (err)
+		goto out;
+
+	err = f2fs_do_tmpfile(inode, dir);
+	if (err)
+		goto release_out;
+
+	/*
+	 * add this non-linked tmpfile to orphan list, in this way we could
+	 * remove all unused data of tmpfile after abnormal power-off.
+	 */
+	add_orphan_inode(sbi, inode->i_ino);
+	f2fs_unlock_op(sbi);
+
+	alloc_nid_done(sbi, inode->i_ino);
+	d_tmpfile(dentry, inode);
+	unlock_new_inode(inode);
+	return 0;
+
+release_out:
+	release_orphan_inode(sbi);
+out:
+	f2fs_unlock_op(sbi);
+	clear_nlink(inode);
+	unlock_new_inode(inode);
+	make_bad_inode(inode);
+	iput(inode);
+	alloc_nid_failed(sbi, inode->i_ino);
+	return err;
+}
+
 const struct inode_operations f2fs_dir_inode_operations = {
 	.create		= f2fs_create,
 	.lookup		= f2fs_lookup,
@@ -497,6 +705,8 @@ const struct inode_operations f2fs_dir_inode_operations = {
 	.rmdir		= f2fs_rmdir,
 	.mknod		= f2fs_mknod,
 	.rename		= f2fs_rename,
+	.rename2	= f2fs_rename2,
+	.tmpfile	= f2fs_tmpfile,
 	.getattr	= f2fs_getattr,
 	.setattr	= f2fs_setattr,
 	.get_acl	= f2fs_get_acl,

fs/f2fs/node.c

@@ -25,6 +25,7 @@
 
 static struct kmem_cache *nat_entry_slab;
 static struct kmem_cache *free_nid_slab;
+static struct kmem_cache *nat_entry_set_slab;
 
 bool available_free_memory(struct f2fs_sb_info *sbi, int type)
 {
@@ -90,12 +91,8 @@ static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid)
 
 	/* get current nat block page with lock */
 	src_page = get_meta_page(sbi, src_off);
-
-	/* Dirty src_page means that it is already the new target NAT page. */
-	if (PageDirty(src_page))
-		return src_page;
-
 	dst_page = grab_meta_page(sbi, dst_off);
+	f2fs_bug_on(PageDirty(src_page));
 
 	src_addr = page_address(src_page);
 	dst_addr = page_address(dst_page);
@@ -845,7 +842,7 @@ void remove_inode_page(struct inode *inode)
 	truncate_node(&dn);
 }
 
-struct page *new_inode_page(struct inode *inode, const struct qstr *name)
+struct page *new_inode_page(struct inode *inode)
 {
 	struct dnode_of_data dn;
 
@@ -1234,12 +1231,12 @@ static int f2fs_write_node_page(struct page *page,
 	if (wbc->for_reclaim)
 		goto redirty_out;
 
-	mutex_lock(&sbi->node_write);
+	down_read(&sbi->node_write);
 	set_page_writeback(page);
 	write_node_page(sbi, page, &fio, nid, ni.blk_addr, &new_addr);
 	set_node_addr(sbi, &ni, new_addr, is_fsync_dnode(page));
 	dec_page_count(sbi, F2FS_DIRTY_NODES);
-	mutex_unlock(&sbi->node_write);
+	up_read(&sbi->node_write);
 	unlock_page(page);
 	return 0;
 
@@ -1552,7 +1549,7 @@ void recover_node_page(struct f2fs_sb_info *sbi, struct page *page,
 	clear_node_page_dirty(page);
 }
 
-static void recover_inline_xattr(struct inode *inode, struct page *page)
+void recover_inline_xattr(struct inode *inode, struct page *page)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
 	void *src_addr, *dst_addr;
@@ -1591,8 +1588,6 @@ bool recover_xattr_data(struct inode *inode, struct page *page, block_t blkaddr)
 	nid_t new_xnid = nid_of_node(page);
 	struct node_info ni;
 
-	recover_inline_xattr(inode, page);
-
 	if (!f2fs_has_xattr_block(ofs_of_node(page)))
 		return false;
 
@@ -1744,7 +1739,90 @@ int restore_node_summary(struct f2fs_sb_info *sbi,
 	return err;
 }
 
-static bool flush_nats_in_journal(struct f2fs_sb_info *sbi)
+static struct nat_entry_set *grab_nat_entry_set(void)
+{
+	struct nat_entry_set *nes =
+			f2fs_kmem_cache_alloc(nat_entry_set_slab, GFP_ATOMIC);
+
+	nes->entry_cnt = 0;
+	INIT_LIST_HEAD(&nes->set_list);
+	INIT_LIST_HEAD(&nes->entry_list);
+	return nes;
+}
+
+static void release_nat_entry_set(struct nat_entry_set *nes,
+						struct f2fs_nm_info *nm_i)
+{
+	f2fs_bug_on(!list_empty(&nes->entry_list));
+
+	nm_i->dirty_nat_cnt -= nes->entry_cnt;
+	list_del(&nes->set_list);
+	kmem_cache_free(nat_entry_set_slab, nes);
+}
+
+static void adjust_nat_entry_set(struct nat_entry_set *nes,
+						struct list_head *head)
+{
+	struct nat_entry_set *next = nes;
+
+	if (list_is_last(&nes->set_list, head))
+		return;
+
+	list_for_each_entry_continue(next, head, set_list)
+		if (nes->entry_cnt <= next->entry_cnt)
+			break;
+
+	list_move_tail(&nes->set_list, &next->set_list);
+}
+
+static void add_nat_entry(struct nat_entry *ne, struct list_head *head)
+{
+	struct nat_entry_set *nes;
+	nid_t start_nid = START_NID(ne->ni.nid);
+
+	list_for_each_entry(nes, head, set_list) {
+		if (nes->start_nid == start_nid) {
+			list_move_tail(&ne->list, &nes->entry_list);
+			nes->entry_cnt++;
+			adjust_nat_entry_set(nes, head);
+			return;
+		}
+	}
+
+	nes = grab_nat_entry_set();
+
+	nes->start_nid = start_nid;
+	list_move_tail(&ne->list, &nes->entry_list);
+	nes->entry_cnt++;
+	list_add(&nes->set_list, head);
+}
+
+static void merge_nats_in_set(struct f2fs_sb_info *sbi)
+{
+	struct f2fs_nm_info *nm_i = NM_I(sbi);
+	struct list_head *dirty_list = &nm_i->dirty_nat_entries;
+	struct list_head *set_list = &nm_i->nat_entry_set;
+	struct nat_entry *ne, *tmp;
+
+	write_lock(&nm_i->nat_tree_lock);
+	list_for_each_entry_safe(ne, tmp, dirty_list, list) {
+		if (nat_get_blkaddr(ne) == NEW_ADDR)
+			continue;
+		add_nat_entry(ne, set_list);
+		nm_i->dirty_nat_cnt++;
+	}
+	write_unlock(&nm_i->nat_tree_lock);
+}
+
+static bool __has_cursum_space(struct f2fs_summary_block *sum, int size)
+{
+	if (nats_in_cursum(sum) + size <= NAT_JOURNAL_ENTRIES)
+		return true;
+	else
+		return false;
+}
+
+static void remove_nats_in_journal(struct f2fs_sb_info *sbi)
 {
 	struct f2fs_nm_info *nm_i = NM_I(sbi);
 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
@@ -1752,12 +1830,6 @@ static bool flush_nats_in_journal(struct f2fs_sb_info *sbi)
 	int i;
 
 	mutex_lock(&curseg->curseg_mutex);
-
-	if (nats_in_cursum(sum) < NAT_JOURNAL_ENTRIES) {
-		mutex_unlock(&curseg->curseg_mutex);
-		return false;
-	}
-
 	for (i = 0; i < nats_in_cursum(sum); i++) {
 		struct nat_entry *ne;
 		struct f2fs_nat_entry raw_ne;
@@ -1767,23 +1839,21 @@ static bool flush_nats_in_journal(struct f2fs_sb_info *sbi)
 retry:
 		write_lock(&nm_i->nat_tree_lock);
 		ne = __lookup_nat_cache(nm_i, nid);
-		if (ne) {
-			__set_nat_cache_dirty(nm_i, ne);
-			write_unlock(&nm_i->nat_tree_lock);
-			continue;
-		}
+		if (ne)
+			goto found;
+
 		ne = grab_nat_entry(nm_i, nid);
 		if (!ne) {
 			write_unlock(&nm_i->nat_tree_lock);
 			goto retry;
 		}
 		node_info_from_raw_nat(&ne->ni, &raw_ne);
+found:
 		__set_nat_cache_dirty(nm_i, ne);
 		write_unlock(&nm_i->nat_tree_lock);
 	}
 	update_nats_in_cursum(sum, -i);
 	mutex_unlock(&curseg->curseg_mutex);
-	return true;
 }
 
 /*
@@ -1794,65 +1864,68 @@ void flush_nat_entries(struct f2fs_sb_info *sbi)
 	struct f2fs_nm_info *nm_i = NM_I(sbi);
 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
 	struct f2fs_summary_block *sum = curseg->sum_blk;
-	struct nat_entry *ne, *cur;
-	struct page *page = NULL;
-	struct f2fs_nat_block *nat_blk = NULL;
-	nid_t start_nid = 0, end_nid = 0;
-	bool flushed;
+	struct nat_entry_set *nes, *tmp;
+	struct list_head *head = &nm_i->nat_entry_set;
+	bool to_journal = true;
 
-	flushed = flush_nats_in_journal(sbi);
+	/* merge nat entries of dirty list to nat entry set temporarily */
+	merge_nats_in_set(sbi);
 
-	if (!flushed)
-		mutex_lock(&curseg->curseg_mutex);
-
-	/* 1) flush dirty nat caches */
-	list_for_each_entry_safe(ne, cur, &nm_i->dirty_nat_entries, list) {
-		nid_t nid;
-		struct f2fs_nat_entry raw_ne;
-		int offset = -1;
-
-		if (nat_get_blkaddr(ne) == NEW_ADDR)
-			continue;
-
-		nid = nat_get_nid(ne);
-
-		if (flushed)
-			goto to_nat_page;
+	/*
+	 * if there are no enough space in journal to store dirty nat
+	 * entries, remove all entries from journal and merge them
+	 * into nat entry set.
+	 */
+	if (!__has_cursum_space(sum, nm_i->dirty_nat_cnt)) {
+		remove_nats_in_journal(sbi);
 
-		/* if there is room for nat enries in curseg->sumpage */
-		offset = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 1);
-		if (offset >= 0) {
-			raw_ne = nat_in_journal(sum, offset);
-			goto flush_now;
-		}
-to_nat_page:
-		if (!page || (start_nid > nid || nid > end_nid)) {
-			if (page) {
-				f2fs_put_page(page, 1);
-				page = NULL;
+		/*
+		 * merge nat entries of dirty list to nat entry set temporarily
+		 */
+		merge_nats_in_set(sbi);
 	}
-			start_nid = START_NID(nid);
-			end_nid = start_nid + NAT_ENTRY_PER_BLOCK - 1;
+
+	if (!nm_i->dirty_nat_cnt)
+		return;
 
 	/*
-			 * get nat block with dirty flag, increased reference
-			 * count, mapped and lock
+	 * there are two steps to flush nat entries:
+	 * #1, flush nat entries to journal in current hot data summary block.
+	 * #2, flush nat entries to nat page.
 	 */
+	list_for_each_entry_safe(nes, tmp, head, set_list) {
+		struct f2fs_nat_block *nat_blk;
+		struct nat_entry *ne, *cur;
+		struct page *page;
+		nid_t start_nid = nes->start_nid;
+
+		if (to_journal && !__has_cursum_space(sum, nes->entry_cnt))
+			to_journal = false;
+
+		if (to_journal) {
+			mutex_lock(&curseg->curseg_mutex);
+		} else {
 			page = get_next_nat_page(sbi, start_nid);
 			nat_blk = page_address(page);
+			f2fs_bug_on(!nat_blk);
 		}
 
-		f2fs_bug_on(!nat_blk);
-		raw_ne = nat_blk->entries[nid - start_nid];
-flush_now:
-		raw_nat_from_node_info(&raw_ne, &ne->ni);
+		/* flush dirty nats in nat entry set */
+		list_for_each_entry_safe(ne, cur, &nes->entry_list, list) {
+			struct f2fs_nat_entry *raw_ne;
+			nid_t nid = nat_get_nid(ne);
+			int offset;
 
-		if (offset < 0) {
-			nat_blk->entries[nid - start_nid] = raw_ne;
-		} else {
-			nat_in_journal(sum, offset) = raw_ne;
+			if (to_journal) {
+				offset = lookup_journal_in_cursum(sum,
+							NAT_JOURNAL, nid, 1);
+				f2fs_bug_on(offset < 0);
+				raw_ne = &nat_in_journal(sum, offset);
 				nid_in_journal(sum, offset) = cpu_to_le32(nid);
+			} else {
+				raw_ne = &nat_blk->entries[nid - start_nid];
 			}
+			raw_nat_from_node_info(raw_ne, &ne->ni);
 
 			if (nat_get_blkaddr(ne) == NULL_ADDR &&
 				add_free_nid(sbi, nid, false) <= 0) {
@@ -1865,9 +1938,17 @@ void flush_nat_entries(struct f2fs_sb_info *sbi)
 				write_unlock(&nm_i->nat_tree_lock);
 			}
 		}
-	if (!flushed)
+
+		if (to_journal)
 			mutex_unlock(&curseg->curseg_mutex);
+		else
 			f2fs_put_page(page, 1);
+
+		release_nat_entry_set(nes, nm_i);
+	}
+
+	f2fs_bug_on(!list_empty(head));
+	f2fs_bug_on(nm_i->dirty_nat_cnt);
 }
 
 static int init_node_manager(struct f2fs_sb_info *sbi)
@@ -1896,6 +1977,7 @@ static int init_node_manager(struct f2fs_sb_info *sbi)
 	INIT_RADIX_TREE(&nm_i->nat_root, GFP_ATOMIC);
 	INIT_LIST_HEAD(&nm_i->nat_entries);
 	INIT_LIST_HEAD(&nm_i->dirty_nat_entries);
+	INIT_LIST_HEAD(&nm_i->nat_entry_set);
 
 	mutex_init(&nm_i->build_lock);
 	spin_lock_init(&nm_i->free_nid_list_lock);
@@ -1976,19 +2058,30 @@ int __init create_node_manager_caches(void)
 	nat_entry_slab = f2fs_kmem_cache_create("nat_entry",
 			sizeof(struct nat_entry));
 	if (!nat_entry_slab)
-		return -ENOMEM;
+		goto fail;
 
 	free_nid_slab = f2fs_kmem_cache_create("free_nid",
 			sizeof(struct free_nid));
-	if (!free_nid_slab) {
+	if (!free_nid_slab)
+		goto destory_nat_entry;
+
+	nat_entry_set_slab = f2fs_kmem_cache_create("nat_entry_set",
+			sizeof(struct nat_entry_set));
+	if (!nat_entry_set_slab)
+		goto destory_free_nid;
+	return 0;
+
+destory_free_nid:
+	kmem_cache_destroy(free_nid_slab);
+destory_nat_entry:
 	kmem_cache_destroy(nat_entry_slab);
+fail:
 	return -ENOMEM;
-	}
-	return 0;
 }
 
 void destroy_node_manager_caches(void)
 {
+	kmem_cache_destroy(nat_entry_set_slab);
 	kmem_cache_destroy(free_nid_slab);
 	kmem_cache_destroy(nat_entry_slab);
 }

fs/f2fs/node.h

@@ -89,6 +89,13 @@ enum mem_type {
 	DIRTY_DENTS	/* indicates dirty dentry pages */
 };
 
+struct nat_entry_set {
+	struct list_head set_list;	/* link with all nat sets */
+	struct list_head entry_list;	/* link with dirty nat entries */
+	nid_t start_nid;		/* start nid of nats in set */
+	unsigned int entry_cnt;		/* the # of nat entries in set */
+};
+
 /*
  * For free nid mangement
  */

fs/f2fs/recovery.c

@@ -300,6 +300,8 @@ static int do_recover_data(struct f2fs_sb_info *sbi, struct inode *inode,
 	struct node_info ni;
 	int err = 0, recovered = 0;
 
+	recover_inline_xattr(inode, page);
+
 	if (recover_inline_data(inode, page))
 		goto out;
 
@@ -434,7 +436,9 @@ static int recover_data(struct f2fs_sb_info *sbi,
 
 int recover_fsync_data(struct f2fs_sb_info *sbi)
 {
+	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
 	struct list_head inode_list;
+	block_t blkaddr;
 	int err;
 	bool need_writecp = false;
 
@@ -447,6 +451,9 @@ int recover_fsync_data(struct f2fs_sb_info *sbi)
 
 	/* step #1: find fsynced inode numbers */
 	sbi->por_doing = true;
+
+	blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
+
 	err = find_fsync_dnodes(sbi, &inode_list);
 	if (err)
 		goto out;
@@ -462,8 +469,21 @@ int recover_fsync_data(struct f2fs_sb_info *sbi)
 out:
 	destroy_fsync_dnodes(&inode_list);
 	kmem_cache_destroy(fsync_entry_slab);
+
+	if (err) {
+		truncate_inode_pages_final(NODE_MAPPING(sbi));
+		truncate_inode_pages_final(META_MAPPING(sbi));
+	}
+
 	sbi->por_doing = false;
-	if (!err && need_writecp)
+	if (err) {
+		discard_next_dnode(sbi, blkaddr);
+
+		/* Flush all the NAT/SIT pages */
+		while (get_pages(sbi, F2FS_DIRTY_META))
+			sync_meta_pages(sbi, META, LONG_MAX);
+	} else if (need_writecp) {
 		write_checkpoint(sbi, false);
+	}
 	return err;
 }

fs/f2fs/segment.c

@@ -239,6 +239,12 @@ int f2fs_issue_flush(struct f2fs_sb_info *sbi)
 	struct flush_cmd_control *fcc = SM_I(sbi)->cmd_control_info;
 	struct flush_cmd cmd;
 
+	trace_f2fs_issue_flush(sbi->sb, test_opt(sbi, NOBARRIER),
+					test_opt(sbi, FLUSH_MERGE));
+
+	if (test_opt(sbi, NOBARRIER))
+		return 0;
+
 	if (!test_opt(sbi, FLUSH_MERGE))
 		return blkdev_issue_flush(sbi->sb->s_bdev, GFP_KERNEL, NULL);
 
@@ -272,13 +278,13 @@ int create_flush_cmd_control(struct f2fs_sb_info *sbi)
 		return -ENOMEM;
 	spin_lock_init(&fcc->issue_lock);
 	init_waitqueue_head(&fcc->flush_wait_queue);
-	sbi->sm_info->cmd_control_info = fcc;
+	SM_I(sbi)->cmd_control_info = fcc;
 	fcc->f2fs_issue_flush = kthread_run(issue_flush_thread, sbi,
 				"f2fs_flush-%u:%u", MAJOR(dev), MINOR(dev));
 	if (IS_ERR(fcc->f2fs_issue_flush)) {
 		err = PTR_ERR(fcc->f2fs_issue_flush);
 		kfree(fcc);
-		sbi->sm_info->cmd_control_info = NULL;
+		SM_I(sbi)->cmd_control_info = NULL;
 		return err;
 	}
 
@@ -287,13 +293,12 @@ int create_flush_cmd_control(struct f2fs_sb_info *sbi)
 
 void destroy_flush_cmd_control(struct f2fs_sb_info *sbi)
 {
-	struct flush_cmd_control *fcc =
-				sbi->sm_info->cmd_control_info;
+	struct flush_cmd_control *fcc = SM_I(sbi)->cmd_control_info;
 
 	if (fcc && fcc->f2fs_issue_flush)
 		kthread_stop(fcc->f2fs_issue_flush);
 	kfree(fcc);
-	sbi->sm_info->cmd_control_info = NULL;
+	SM_I(sbi)->cmd_control_info = NULL;
 }
 
 static void __locate_dirty_segment(struct f2fs_sb_info *sbi, unsigned int segno,
@@ -377,11 +382,8 @@ static int f2fs_issue_discard(struct f2fs_sb_info *sbi,
 	return blkdev_issue_discard(sbi->sb->s_bdev, start, len, GFP_NOFS, 0);
 }
 
-void discard_next_dnode(struct f2fs_sb_info *sbi)
+void discard_next_dnode(struct f2fs_sb_info *sbi, block_t blkaddr)
 {
-	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
-	block_t blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
-
 	if (f2fs_issue_discard(sbi, blkaddr, 1)) {
 		struct page *page = grab_meta_page(sbi, blkaddr);
 		/* zero-filled page */
@@ -437,17 +439,12 @@ static void add_discard_addrs(struct f2fs_sb_info *sbi,
 static void set_prefree_as_free_segments(struct f2fs_sb_info *sbi)
 {
 	struct dirty_seglist_info *dirty_i = DIRTY_I(sbi);
-	unsigned int segno = -1;
+	unsigned int segno;
 	unsigned int total_segs = TOTAL_SEGS(sbi);
 
 	mutex_lock(&dirty_i->seglist_lock);
-	while (1) {
-		segno = find_next_bit(dirty_i->dirty_segmap[PRE], total_segs,
-				segno + 1);
-		if (segno >= total_segs)
-			break;
+	for_each_set_bit(segno, dirty_i->dirty_segmap[PRE], total_segs)
 		__set_test_and_free(sbi, segno);
-	}
 	mutex_unlock(&dirty_i->seglist_lock);
 }
 
@@ -974,14 +971,12 @@ void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
 {
 	struct sit_info *sit_i = SIT_I(sbi);
 	struct curseg_info *curseg;
-	unsigned int old_cursegno;
 
 	curseg = CURSEG_I(sbi, type);
 
 	mutex_lock(&curseg->curseg_mutex);
 
 	*new_blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
-	old_cursegno = curseg->segno;
 
 	/*
 	 * __add_sum_entry should be resided under the curseg_mutex
@@ -1002,7 +997,6 @@ void allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
 	 * since SSR needs latest valid block information.
 	 */
 	refresh_sit_entry(sbi, old_blkaddr, *new_blkaddr);
-	locate_dirty_segment(sbi, old_cursegno);
 
 	mutex_unlock(&sit_i->sentry_lock);
 
@@ -1532,7 +1526,7 @@ void flush_sit_entries(struct f2fs_sb_info *sbi)
 	struct page *page = NULL;
 	struct f2fs_sit_block *raw_sit = NULL;
 	unsigned int start = 0, end = 0;
-	unsigned int segno = -1;
+	unsigned int segno;
 	bool flushed;
 
 	mutex_lock(&curseg->curseg_mutex);
@@ -1544,7 +1538,7 @@ void flush_sit_entries(struct f2fs_sb_info *sbi)
 	 */
 	flushed = flush_sits_in_journal(sbi);
 
-	while ((segno = find_next_bit(bitmap, nsegs, segno + 1)) < nsegs) {
+	for_each_set_bit(segno, bitmap, nsegs) {
 		struct seg_entry *se = get_seg_entry(sbi, segno);
 		int sit_offset, offset;
 
@@ -1703,7 +1697,7 @@ static int build_curseg(struct f2fs_sb_info *sbi)
 	struct curseg_info *array;
 	int i;
 
-	array = kzalloc(sizeof(*array) * NR_CURSEG_TYPE, GFP_KERNEL);
+	array = kcalloc(NR_CURSEG_TYPE, sizeof(*array), GFP_KERNEL);
 	if (!array)
 		return -ENOMEM;
 

fs/f2fs/segment.h

@@ -347,8 +347,8 @@ static inline void __set_test_and_free(struct f2fs_sb_info *sbi,
 	if (test_and_clear_bit(segno, free_i->free_segmap)) {
 		free_i->free_segments++;
 
-		next = find_next_bit(free_i->free_segmap, TOTAL_SEGS(sbi),
-								start_segno);
+		next = find_next_bit(free_i->free_segmap,
+				start_segno + sbi->segs_per_sec, start_segno);
 		if (next >= start_segno + sbi->segs_per_sec) {
 			if (test_and_clear_bit(secno, free_i->free_secmap))
 				free_i->free_sections++;
@@ -486,6 +486,10 @@ static inline bool need_inplace_update(struct inode *inode)
 	if (S_ISDIR(inode->i_mode))
 		return false;
 
+	/* this is only set during fdatasync */
+	if (is_inode_flag_set(F2FS_I(inode), FI_NEED_IPU))
+		return true;
+
 	switch (SM_I(sbi)->ipu_policy) {
 	case F2FS_IPU_FORCE:
 		return true;

fs/f2fs/super.c

@@ -52,6 +52,7 @@ enum {
 	Opt_inline_xattr,
 	Opt_inline_data,
 	Opt_flush_merge,
+	Opt_nobarrier,
 	Opt_err,
 };
 
@@ -69,6 +70,7 @@ static match_table_t f2fs_tokens = {
 	{Opt_inline_xattr, "inline_xattr"},
 	{Opt_inline_data, "inline_data"},
 	{Opt_flush_merge, "flush_merge"},
+	{Opt_nobarrier, "nobarrier"},
 	{Opt_err, NULL},
 };
 
@@ -339,6 +341,9 @@ static int parse_options(struct super_block *sb, char *options)
 		case Opt_flush_merge:
 			set_opt(sbi, FLUSH_MERGE);
 			break;
+		case Opt_nobarrier:
+			set_opt(sbi, NOBARRIER);
+			break;
 		default:
 			f2fs_msg(sb, KERN_ERR,
 				"Unrecognized mount option \"%s\" or missing value",
@@ -544,6 +549,8 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 		seq_puts(seq, ",inline_data");
 	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
 		seq_puts(seq, ",flush_merge");
+	if (test_opt(sbi, NOBARRIER))
+		seq_puts(seq, ",nobarrier");
 	seq_printf(seq, ",active_logs=%u", sbi->active_logs);
 
 	return 0;
@@ -615,7 +622,7 @@ static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 	 * Previous and new state of filesystem is RO,
 	 * so skip checking GC and FLUSH_MERGE conditions.
 	 */
-	if ((sb->s_flags & MS_RDONLY) && (*flags & MS_RDONLY))
+	if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
 		goto skip;
 
 	/*
@@ -642,8 +649,7 @@ static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 	 */
 	if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
 		destroy_flush_cmd_control(sbi);
-	} else if (test_opt(sbi, FLUSH_MERGE) &&
-					!sbi->sm_info->cmd_control_info) {
+	} else if (test_opt(sbi, FLUSH_MERGE) && !SM_I(sbi)->cmd_control_info) {
 		err = create_flush_cmd_control(sbi);
 		if (err)
 			goto restore_gc;
@@ -947,7 +953,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 	mutex_init(&sbi->gc_mutex);
 	mutex_init(&sbi->writepages);
 	mutex_init(&sbi->cp_mutex);
-	mutex_init(&sbi->node_write);
+	init_rwsem(&sbi->node_write);
 	sbi->por_doing = false;
 	spin_lock_init(&sbi->stat_lock);
 
@@ -997,7 +1003,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 	INIT_LIST_HEAD(&sbi->dir_inode_list);
 	spin_lock_init(&sbi->dir_inode_lock);
 
-	init_orphan_info(sbi);
+	init_ino_entry_info(sbi);
 
 	/* setup f2fs internal modules */
 	err = build_segment_manager(sbi);
@@ -1034,8 +1040,9 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 		goto free_node_inode;
 	}
 	if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
+		iput(root);
 		err = -EINVAL;
-		goto free_root_inode;
+		goto free_node_inode;
 	}
 
 	sb->s_root = d_make_root(root); /* allocate root dentry */
@@ -1082,7 +1089,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 	 * If filesystem is not mounted as read-only then
 	 * do start the gc_thread.
 	 */
-	if (!(sb->s_flags & MS_RDONLY)) {
+	if (!f2fs_readonly(sb)) {
 		/* After POR, we can run background GC thread.*/
 		err = start_gc_thread(sbi);
 		if (err)

include/trace/events/f2fs.h

@@ -587,6 +587,69 @@ TRACE_EVENT(f2fs_fallocate,
 		__entry->ret)
 );
 
+TRACE_EVENT(f2fs_direct_IO_enter,
+
+	TP_PROTO(struct inode *inode, loff_t offset, unsigned long len, int rw),
+
+	TP_ARGS(inode, offset, len, rw),
+
+	TP_STRUCT__entry(
+		__field(dev_t,	dev)
+		__field(ino_t,	ino)
+		__field(loff_t,	pos)
+		__field(unsigned long,	len)
+		__field(int,	rw)
+	),
+
+	TP_fast_assign(
+		__entry->dev	= inode->i_sb->s_dev;
+		__entry->ino	= inode->i_ino;
+		__entry->pos	= offset;
+		__entry->len	= len;
+		__entry->rw	= rw;
+	),
+
+	TP_printk("dev = (%d,%d), ino = %lu pos = %lld len = %lu rw = %d",
+		show_dev_ino(__entry),
+		__entry->pos,
+		__entry->len,
+		__entry->rw)
+);
+
+TRACE_EVENT(f2fs_direct_IO_exit,
+
+	TP_PROTO(struct inode *inode, loff_t offset, unsigned long len,
+		 int rw, int ret),
+
+	TP_ARGS(inode, offset, len, rw, ret),
+
+	TP_STRUCT__entry(
+		__field(dev_t,	dev)
+		__field(ino_t,	ino)
+		__field(loff_t,	pos)
+		__field(unsigned long,	len)
+		__field(int,	rw)
+		__field(int,	ret)
+	),
+
+	TP_fast_assign(
+		__entry->dev	= inode->i_sb->s_dev;
+		__entry->ino	= inode->i_ino;
+		__entry->pos	= offset;
+		__entry->len	= len;
+		__entry->rw	= rw;
+		__entry->ret	= ret;
+	),
+
+	TP_printk("dev = (%d,%d), ino = %lu pos = %lld len = %lu "
+		"rw = %d ret = %d",
+		show_dev_ino(__entry),
+		__entry->pos,
+		__entry->len,
+		__entry->rw,
+		__entry->ret)
+);
+
 TRACE_EVENT(f2fs_reserve_new_block,
 
 	TP_PROTO(struct inode *inode, nid_t nid, unsigned int ofs_in_node),
@@ -926,6 +989,30 @@ TRACE_EVENT(f2fs_issue_discard,
 		(unsigned long long)__entry->blkstart,
 		(unsigned long long)__entry->blklen)
 );
+
+TRACE_EVENT(f2fs_issue_flush,
+
+	TP_PROTO(struct super_block *sb, bool nobarrier, bool flush_merge),
+
+	TP_ARGS(sb, nobarrier, flush_merge),
+
+	TP_STRUCT__entry(
+		__field(dev_t,	dev)
+		__field(bool, nobarrier)
+		__field(bool, flush_merge)
+	),
+
+	TP_fast_assign(
+		__entry->dev	= sb->s_dev;
+		__entry->nobarrier = nobarrier;
+		__entry->flush_merge = flush_merge;
+	),
+
+	TP_printk("dev = (%d,%d), %s %s",
+		show_dev(__entry),
+		__entry->nobarrier ? "skip (nobarrier)" : "issue",
+		__entry->flush_merge ? " with flush_merge" : "")
+);
 #endif /* _TRACE_F2FS_H */
 
  /* This part must be outside protection */