f2fs: use extent_cache by default

We don't need to handle the duplicate extent information.

The integrated rule is:
 - update on-disk extent with largest one tracked by in-memory extent_cache
 - destroy extent_tree for the truncation case
 - drop per-inode extent_cache by shrinker
Reviewed-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

fs/f2fs/f2fs.h

@@ -349,6 +349,7 @@ struct extent_tree {
 	nid_t ino;			/* inode number */
 	struct rb_root root;		/* root of extent info rb-tree */
 	struct extent_node *cached_en;	/* recently accessed extent node */
+	struct extent_info largest;	/* largested extent info */
 	rwlock_t lock;			/* protect extent info rb-tree */
 	atomic_t refcount;		/* reference count of rb-tree */
 	unsigned int count;		/* # of extent node in rb-tree*/
@@ -420,14 +421,14 @@ struct f2fs_inode_info {
 	unsigned int clevel;		/* maximum level of given file name */
 	nid_t i_xattr_nid;		/* node id that contains xattrs */
 	unsigned long long xattr_ver;	/* cp version of xattr modification */
-	struct extent_info ext;		/* in-memory extent cache entry */
-	rwlock_t ext_lock;		/* rwlock for single extent cache */
 	struct inode_entry *dirty_dir;	/* the pointer of dirty dir */
 
 	struct radix_tree_root inmem_root;	/* radix tree for inmem pages */
 	struct list_head inmem_pages;	/* inmemory pages managed by f2fs */
 	struct mutex inmem_lock;	/* lock for inmemory pages */
 
+	struct extent_tree *extent_tree;	/* cached extent_tree entry */
+
 #ifdef CONFIG_F2FS_FS_ENCRYPTION
 	/* Encryption params */
 	struct f2fs_crypt_info *i_crypt_info;
@@ -1548,6 +1549,17 @@ static inline bool is_dot_dotdot(const struct qstr *str)
 	return false;
 }
 
+static inline bool f2fs_may_extent_tree(struct inode *inode)
+{
+	mode_t mode = inode->i_mode;
+
+	if (!test_opt(F2FS_I_SB(inode), EXTENT_CACHE) ||
+			is_inode_flag_set(F2FS_I(inode), FI_NO_EXTENT))
+		return false;
+
+	return S_ISREG(mode);
+}
+
 #define get_inode_mode(i) \
 	((is_inode_flag_set(F2FS_I(i), FI_ACL_MODE)) ? \
 	 (F2FS_I(i)->i_acl_mode) : ((i)->i_mode))
@@ -1755,10 +1767,10 @@ void set_data_blkaddr(struct dnode_of_data *);
 int reserve_new_block(struct dnode_of_data *);
 int f2fs_reserve_block(struct dnode_of_data *, pgoff_t);
 unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *, int);
+void f2fs_init_extent_tree(struct inode *, struct f2fs_extent *);
+unsigned int f2fs_destroy_extent_node(struct inode *);
 void f2fs_destroy_extent_tree(struct inode *);
-void f2fs_init_extent_cache(struct inode *, struct f2fs_extent *);
 void f2fs_update_extent_cache(struct dnode_of_data *);
-void f2fs_preserve_extent_tree(struct inode *);
 struct page *get_read_data_page(struct inode *, pgoff_t, int);
 struct page *find_data_page(struct inode *, pgoff_t);
 struct page *get_lock_data_page(struct inode *, pgoff_t);

fs/f2fs/inode.c

@@ -139,7 +139,7 @@ static int do_read_inode(struct inode *inode)
 	fi->i_pino = le32_to_cpu(ri->i_pino);
 	fi->i_dir_level = ri->i_dir_level;
 
-	f2fs_init_extent_cache(inode, &ri->i_ext);
+	f2fs_init_extent_tree(inode, &ri->i_ext);
 
 	get_inline_info(fi, ri);
 
@@ -237,10 +237,11 @@ void update_inode(struct inode *inode, struct page *node_page)
 	ri->i_size = cpu_to_le64(i_size_read(inode));
 	ri->i_blocks = cpu_to_le64(inode->i_blocks);
 
-	read_lock(&F2FS_I(inode)->ext_lock);
-	set_raw_extent(&F2FS_I(inode)->ext, &ri->i_ext);
-	read_unlock(&F2FS_I(inode)->ext_lock);
-
+	if (F2FS_I(inode)->extent_tree)
+		set_raw_extent(&F2FS_I(inode)->extent_tree->largest,
+							&ri->i_ext);
+	else
+		memset(&ri->i_ext, 0, sizeof(ri->i_ext));
 	set_raw_inline(F2FS_I(inode), ri);
 
 	ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
@@ -331,6 +332,8 @@ void f2fs_evict_inode(struct inode *inode)
 	f2fs_bug_on(sbi, get_dirty_pages(inode));
 	remove_dirty_dir_inode(inode);
 
+	f2fs_destroy_extent_tree(inode);
+
 	if (inode->i_nlink || is_bad_inode(inode))
 		goto no_delete;
 
@@ -350,11 +353,6 @@ void f2fs_evict_inode(struct inode *inode)
 	stat_dec_inline_dir(inode);
 	stat_dec_inline_inode(inode);
 
-	/* update extent info in inode */
-	if (inode->i_nlink)
-		f2fs_preserve_extent_tree(inode);
-	f2fs_destroy_extent_tree(inode);
-
 	invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino, inode->i_ino);
 	if (xnid)
 		invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid);

fs/f2fs/namei.c

@@ -65,6 +65,8 @@ static struct inode *f2fs_new_inode(struct inode *dir, umode_t mode)
 	if (f2fs_may_inline_dentry(inode))
 		set_inode_flag(F2FS_I(inode), FI_INLINE_DENTRY);
 
+	f2fs_init_extent_tree(inode, NULL);
+
 	stat_inc_inline_inode(inode);
 	stat_inc_inline_dir(inode);
 

fs/f2fs/shrinker.c

@@ -117,6 +117,8 @@ void f2fs_join_shrinker(struct f2fs_sb_info *sbi)
 
 void f2fs_leave_shrinker(struct f2fs_sb_info *sbi)
 {
+	f2fs_shrink_extent_tree(sbi, __count_extent_cache(sbi));
+
 	spin_lock(&f2fs_list_lock);
 	list_del(&sbi->s_list);
 	spin_unlock(&f2fs_list_lock);

fs/f2fs/super.c

@@ -422,7 +422,6 @@ static struct inode *f2fs_alloc_inode(struct super_block *sb)
 	atomic_set(&fi->dirty_pages, 0);
 	fi->i_current_depth = 1;
 	fi->i_advise = 0;
-	rwlock_init(&fi->ext_lock);
 	init_rwsem(&fi->i_sem);
 	INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
 	INIT_LIST_HEAD(&fi->inmem_pages);
@@ -453,12 +452,17 @@ static int f2fs_drop_inode(struct inode *inode)
 	 */
 	if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
 		if (!inode->i_nlink && !is_bad_inode(inode)) {
+			/* to avoid evict_inode call simultaneously */
+			atomic_inc(&inode->i_count);
 			spin_unlock(&inode->i_lock);
 
 			/* some remained atomic pages should discarded */
 			if (f2fs_is_atomic_file(inode))
 				commit_inmem_pages(inode, true);
 
+			/* should remain fi->extent_tree for writepage */
+			f2fs_destroy_extent_node(inode);
+
 			sb_start_intwrite(inode->i_sb);
 			i_size_write(inode, 0);
 
@@ -473,6 +477,7 @@ static int f2fs_drop_inode(struct inode *inode)
 					F2FS_I(inode)->i_crypt_info);
 #endif
 			spin_lock(&inode->i_lock);
+			atomic_dec(&inode->i_count);
 		}
 		return 0;
 	}
@@ -721,6 +726,7 @@ static void default_options(struct f2fs_sb_info *sbi)
 
 	set_opt(sbi, BG_GC);
 	set_opt(sbi, INLINE_DATA);
+	set_opt(sbi, EXTENT_CACHE);
 
 #ifdef CONFIG_F2FS_FS_XATTR
 	set_opt(sbi, XATTR_USER);