Merge tag 'for-f2fs-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs

Pull f2fs updates from Jaegeuk Kim:
 "New features:
   - in-memory extent_cache
   - fs_shutdown to test power-off-recovery
   - use inline_data to store symlink path
   - show f2fs as a non-misc filesystem

  Major fixes:
   - avoid CPU stalls on sync_dirty_dir_inodes
   - fix some power-off-recovery procedure
   - fix handling of broken symlink correctly
   - fix missing dot and dotdot made by sudden power cuts
   - handle wrong data index during roll-forward recovery
   - preallocate data blocks for direct_io

  ... and a bunch of minor bug fixes and cleanups"

* tag 'for-f2fs-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (71 commits)
  f2fs: pass checkpoint reason on roll-forward recovery
  f2fs: avoid abnormal behavior on broken symlink
  f2fs: flush symlink path to avoid broken symlink after POR
  f2fs: change 0 to false for bool type
  f2fs: do not recover wrong data index
  f2fs: do not increase link count during recovery
  f2fs: assign parent's i_mode for empty dir
  f2fs: add F2FS_INLINE_DOTS to recover missing dot dentries
  f2fs: fix mismatching lock and unlock pages for roll-forward recovery
  f2fs: fix sparse warnings
  f2fs: limit b_size of mapped bh in f2fs_map_bh
  f2fs: persist system.advise into on-disk inode
  f2fs: avoid NULL pointer dereference in f2fs_xattr_advise_get
  f2fs: preallocate fallocated blocks for direct IO
  f2fs: enable inline data by default
  f2fs: preserve extent info for extent cache
  f2fs: initialize extent tree with on-disk extent info of inode
  f2fs: introduce __{find,grab}_extent_tree
  f2fs: split set_data_blkaddr from f2fs_update_extent_cache
  f2fs: enable fast symlink by utilizing inline data
  ...

23 files changed, 1399 insertions, 263 deletions

diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt
index dac11d7fef27..e9e750e59efc 100644
--- a/Documentation/filesystems/f2fs.txt
+++ b/Documentation/filesystems/f2fs.txt
@@ -140,6 +140,12 @@ nobarrier              This option can be used if underlying storage guarantees
 fastboot               This option is used when a system wants to reduce mount
                        time as much as possible, even though normal performance
 		       can be sacrificed.
+extent_cache           Enable an extent cache based on rb-tree, it can cache
+                       as many as extent which map between contiguous logical
+                       address and physical address per inode, resulting in
+                       increasing the cache hit ratio.
+noinline_data          Disable the inline data feature, inline data feature is
+                       enabled by default.
 
 ================================================================================
 DEBUGFS ENTRIES
diff --git a/fs/Kconfig b/fs/Kconfig
index ec35851e5b71..011f43365d7b 100644
--- a/fs/Kconfig
+++ b/fs/Kconfig
@@ -32,6 +32,7 @@ source "fs/gfs2/Kconfig"
 source "fs/ocfs2/Kconfig"
 source "fs/btrfs/Kconfig"
 source "fs/nilfs2/Kconfig"
+source "fs/f2fs/Kconfig"
 
 config FS_DAX
 	bool "Direct Access (DAX) support"
@@ -217,7 +218,6 @@ source "fs/pstore/Kconfig"
 source "fs/sysv/Kconfig"
 source "fs/ufs/Kconfig"
 source "fs/exofs/Kconfig"
-source "fs/f2fs/Kconfig"
 
 endif # MISC_FILESYSTEMS
 
diff --git a/fs/f2fs/Kconfig b/fs/f2fs/Kconfig
index 94e2d2ffabe1..05f0f663f14c 100644
--- a/fs/f2fs/Kconfig
+++ b/fs/f2fs/Kconfig
@@ -1,5 +1,5 @@
 config F2FS_FS
-	tristate "F2FS filesystem support (EXPERIMENTAL)"
+	tristate "F2FS filesystem support"
 	depends on BLOCK
 	help
 	  F2FS is based on Log-structured File System (LFS), which supports
diff --git a/fs/f2fs/acl.c b/fs/f2fs/acl.c
index 742202779bd5..4320ffab3495 100644
--- a/fs/f2fs/acl.c
+++ b/fs/f2fs/acl.c
@@ -351,13 +351,11 @@ static int f2fs_acl_create(struct inode *dir, umode_t *mode,
 
 	*acl = f2fs_acl_clone(p, GFP_NOFS);
 	if (!*acl)
-		return -ENOMEM;
+		goto no_mem;
 
 	ret = f2fs_acl_create_masq(*acl, mode);
-	if (ret < 0) {
-		posix_acl_release(*acl);
-		return -ENOMEM;
-	}
+	if (ret < 0)
+		goto no_mem_clone;
 
 	if (ret == 0) {
 		posix_acl_release(*acl);
@@ -378,6 +376,12 @@ no_acl:
 	*default_acl = NULL;
 	*acl = NULL;
 	return 0;
+
+no_mem_clone:
+	posix_acl_release(*acl);
+no_mem:
+	posix_acl_release(p);
+	return -ENOMEM;
 }
 
 int f2fs_init_acl(struct inode *inode, struct inode *dir, struct page *ipage,
diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index 7f794b72b3b7..a5e17a2a0781 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -276,7 +276,7 @@ continue_unlock:
 			if (!clear_page_dirty_for_io(page))
 				goto continue_unlock;
 
-			if (f2fs_write_meta_page(page, &wbc)) {
+			if (mapping->a_ops->writepage(page, &wbc)) {
 				unlock_page(page);
 				break;
 			}
@@ -464,20 +464,19 @@ static void recover_orphan_inode(struct f2fs_sb_info *sbi, nid_t ino)
 
 void recover_orphan_inodes(struct f2fs_sb_info *sbi)
 {
-	block_t start_blk, orphan_blkaddr, i, j;
+	block_t start_blk, orphan_blocks, i, j;
 
 	if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG))
 		return;
 
 	set_sbi_flag(sbi, SBI_POR_DOING);
 
-	start_blk = __start_cp_addr(sbi) + 1 +
-		le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload);
-	orphan_blkaddr = __start_sum_addr(sbi) - 1;
+	start_blk = __start_cp_addr(sbi) + 1 + __cp_payload(sbi);
+	orphan_blocks = __start_sum_addr(sbi) - 1 - __cp_payload(sbi);
 
-	ra_meta_pages(sbi, start_blk, orphan_blkaddr, META_CP);
+	ra_meta_pages(sbi, start_blk, orphan_blocks, META_CP);
 
-	for (i = 0; i < orphan_blkaddr; i++) {
+	for (i = 0; i < orphan_blocks; i++) {
 		struct page *page = get_meta_page(sbi, start_blk + i);
 		struct f2fs_orphan_block *orphan_blk;
 
@@ -615,7 +614,7 @@ int get_valid_checkpoint(struct f2fs_sb_info *sbi)
 	unsigned long blk_size = sbi->blocksize;
 	unsigned long long cp1_version = 0, cp2_version = 0;
 	unsigned long long cp_start_blk_no;
-	unsigned int cp_blks = 1 + le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload);
+	unsigned int cp_blks = 1 + __cp_payload(sbi);
 	block_t cp_blk_no;
 	int i;
 
@@ -796,6 +795,7 @@ retry:
 		 * wribacking dentry pages in the freeing inode.
 		 */
 		f2fs_submit_merged_bio(sbi, DATA, WRITE);
+		cond_resched();
 	}
 	goto retry;
 }
@@ -884,7 +884,7 @@ static void do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	__u32 crc32 = 0;
 	void *kaddr;
 	int i;
-	int cp_payload_blks = le32_to_cpu(F2FS_RAW_SUPER(sbi)->cp_payload);
+	int cp_payload_blks = __cp_payload(sbi);
 
 	/*
 	 * This avoids to conduct wrong roll-forward operations and uses
@@ -1048,17 +1048,18 @@ void write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
 	unsigned long long ckpt_ver;
 
-	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops");
-
 	mutex_lock(&sbi->cp_mutex);
 
 	if (!is_sbi_flag_set(sbi, SBI_IS_DIRTY) &&
-			cpc->reason != CP_DISCARD && cpc->reason != CP_UMOUNT)
+		(cpc->reason == CP_FASTBOOT || cpc->reason == CP_SYNC))
 		goto out;
 	if (unlikely(f2fs_cp_error(sbi)))
 		goto out;
 	if (f2fs_readonly(sbi->sb))
 		goto out;
+
+	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "start block_ops");
+
 	if (block_operations(sbi))
 		goto out;
 
@@ -1085,6 +1086,10 @@ void write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 
 	unblock_operations(sbi);
 	stat_inc_cp_count(sbi->stat_info);
+
+	if (cpc->reason == CP_RECOVERY)
+		f2fs_msg(sbi->sb, KERN_NOTICE,
+			"checkpoint: version = %llx", ckpt_ver);
 out:
 	mutex_unlock(&sbi->cp_mutex);
 	trace_f2fs_write_checkpoint(sbi->sb, cpc->reason, "finish checkpoint");
@@ -1103,14 +1108,9 @@ void init_ino_entry_info(struct f2fs_sb_info *sbi)
 		im->ino_num = 0;
 	}
 
-	/*
-	 * 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->max_orphans = (sbi->blocks_per_seg - F2FS_CP_PACKS -
-			NR_CURSEG_TYPE) * F2FS_ORPHANS_PER_BLOCK;
+			NR_CURSEG_TYPE - __cp_payload(sbi)) *
+				F2FS_ORPHANS_PER_BLOCK;
 }
 
 int __init create_checkpoint_caches(void)
diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 319eda511c4f..b91b0e10678e 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -25,6 +25,9 @@
 #include "trace.h"
 #include <trace/events/f2fs.h>
 
+static struct kmem_cache *extent_tree_slab;
+static struct kmem_cache *extent_node_slab;
+
 static void f2fs_read_end_io(struct bio *bio, int err)
 {
 	struct bio_vec *bvec;
@@ -197,7 +200,7 @@ alloc_new:
  *  ->node_page
  *    update block addresses in the node page
  */
-static void __set_data_blkaddr(struct dnode_of_data *dn)
+void set_data_blkaddr(struct dnode_of_data *dn)
 {
 	struct f2fs_node *rn;
 	__le32 *addr_array;
@@ -226,7 +229,7 @@ int reserve_new_block(struct dnode_of_data *dn)
 	trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node);
 
 	dn->data_blkaddr = NEW_ADDR;
-	__set_data_blkaddr(dn);
+	set_data_blkaddr(dn);
 	mark_inode_dirty(dn->inode);
 	sync_inode_page(dn);
 	return 0;
@@ -248,73 +251,62 @@ int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
 	return err;
 }
 
-static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
-					struct buffer_head *bh_result)
+static void f2fs_map_bh(struct super_block *sb, pgoff_t pgofs,
+			struct extent_info *ei, struct buffer_head *bh_result)
+{
+	unsigned int blkbits = sb->s_blocksize_bits;
+	size_t max_size = bh_result->b_size;
+	size_t mapped_size;
+
+	clear_buffer_new(bh_result);
+	map_bh(bh_result, sb, ei->blk + pgofs - ei->fofs);
+	mapped_size = (ei->fofs + ei->len - pgofs) << blkbits;
+	bh_result->b_size = min(max_size, mapped_size);
+}
+
+static bool lookup_extent_info(struct inode *inode, pgoff_t pgofs,
+							struct extent_info *ei)
 {
 	struct f2fs_inode_info *fi = F2FS_I(inode);
 	pgoff_t start_fofs, end_fofs;
 	block_t start_blkaddr;
 
-	if (is_inode_flag_set(fi, FI_NO_EXTENT))
-		return 0;
-
-	read_lock(&fi->ext.ext_lock);
+	read_lock(&fi->ext_lock);
 	if (fi->ext.len == 0) {
-		read_unlock(&fi->ext.ext_lock);
-		return 0;
+		read_unlock(&fi->ext_lock);
+		return false;
 	}
 
 	stat_inc_total_hit(inode->i_sb);
 
 	start_fofs = fi->ext.fofs;
 	end_fofs = fi->ext.fofs + fi->ext.len - 1;
-	start_blkaddr = fi->ext.blk_addr;
+	start_blkaddr = fi->ext.blk;
 
 	if (pgofs >= start_fofs && pgofs <= end_fofs) {
-		unsigned int blkbits = inode->i_sb->s_blocksize_bits;
-		size_t count;
-
-		set_buffer_new(bh_result);
-		map_bh(bh_result, inode->i_sb,
-				start_blkaddr + pgofs - start_fofs);
-		count = end_fofs - pgofs + 1;
-		if (count < (UINT_MAX >> blkbits))
-			bh_result->b_size = (count << blkbits);
-		else
-			bh_result->b_size = UINT_MAX;
-
+		*ei = fi->ext;
 		stat_inc_read_hit(inode->i_sb);
-		read_unlock(&fi->ext.ext_lock);
-		return 1;
+		read_unlock(&fi->ext_lock);
+		return true;
 	}
-	read_unlock(&fi->ext.ext_lock);
-	return 0;
+	read_unlock(&fi->ext_lock);
+	return false;
 }
 
-void update_extent_cache(struct dnode_of_data *dn)
+static bool update_extent_info(struct inode *inode, pgoff_t fofs,
+								block_t blkaddr)
 {
-	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
-	pgoff_t fofs, start_fofs, end_fofs;
+	struct f2fs_inode_info *fi = F2FS_I(inode);
+	pgoff_t start_fofs, end_fofs;
 	block_t start_blkaddr, end_blkaddr;
 	int need_update = true;
 
-	f2fs_bug_on(F2FS_I_SB(dn->inode), dn->data_blkaddr == NEW_ADDR);
-
-	/* Update the page address in the parent node */
-	__set_data_blkaddr(dn);
-
-	if (is_inode_flag_set(fi, FI_NO_EXTENT))
-		return;
-
-	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
-							dn->ofs_in_node;
-
-	write_lock(&fi->ext.ext_lock);
+	write_lock(&fi->ext_lock);
 
 	start_fofs = fi->ext.fofs;
 	end_fofs = fi->ext.fofs + fi->ext.len - 1;
-	start_blkaddr = fi->ext.blk_addr;
-	end_blkaddr = fi->ext.blk_addr + fi->ext.len - 1;
+	start_blkaddr = fi->ext.blk;
+	end_blkaddr = fi->ext.blk + fi->ext.len - 1;
 
 	/* Drop and initialize the matched extent */
 	if (fi->ext.len == 1 && fofs == start_fofs)
@@ -322,24 +314,24 @@ void update_extent_cache(struct dnode_of_data *dn)
 
 	/* Initial extent */
 	if (fi->ext.len == 0) {
-		if (dn->data_blkaddr != NULL_ADDR) {
+		if (blkaddr != NULL_ADDR) {
 			fi->ext.fofs = fofs;
-			fi->ext.blk_addr = dn->data_blkaddr;
+			fi->ext.blk = blkaddr;
 			fi->ext.len = 1;
 		}
 		goto end_update;
 	}
 
 	/* Front merge */
-	if (fofs == start_fofs - 1 && dn->data_blkaddr == start_blkaddr - 1) {
+	if (fofs == start_fofs - 1 && blkaddr == start_blkaddr - 1) {
 		fi->ext.fofs--;
-		fi->ext.blk_addr--;
+		fi->ext.blk--;
 		fi->ext.len++;
 		goto end_update;
 	}
 
 	/* Back merge */
-	if (fofs == end_fofs + 1 && dn->data_blkaddr == end_blkaddr + 1) {
+	if (fofs == end_fofs + 1 && blkaddr == end_blkaddr + 1) {
 		fi->ext.len++;
 		goto end_update;
 	}
@@ -351,8 +343,7 @@ void update_extent_cache(struct dnode_of_data *dn)
 			fi->ext.len = fofs - start_fofs;
 		} else {
 			fi->ext.fofs = fofs + 1;
-			fi->ext.blk_addr = start_blkaddr +
-					fofs - start_fofs + 1;
+			fi->ext.blk = start_blkaddr + fofs - start_fofs + 1;
 			fi->ext.len -= fofs - start_fofs + 1;
 		}
 	} else {
@@ -366,27 +357,583 @@ void update_extent_cache(struct dnode_of_data *dn)
 		need_update = true;
 	}
 end_update:
-	write_unlock(&fi->ext.ext_lock);
-	if (need_update)
-		sync_inode_page(dn);
+	write_unlock(&fi->ext_lock);
+	return need_update;
+}
+
+static struct extent_node *__attach_extent_node(struct f2fs_sb_info *sbi,
+				struct extent_tree *et, struct extent_info *ei,
+				struct rb_node *parent, struct rb_node **p)
+{
+	struct extent_node *en;
+
+	en = kmem_cache_alloc(extent_node_slab, GFP_ATOMIC);
+	if (!en)
+		return NULL;
+
+	en->ei = *ei;
+	INIT_LIST_HEAD(&en->list);
+
+	rb_link_node(&en->rb_node, parent, p);
+	rb_insert_color(&en->rb_node, &et->root);
+	et->count++;
+	atomic_inc(&sbi->total_ext_node);
+	return en;
+}
+
+static void __detach_extent_node(struct f2fs_sb_info *sbi,
+				struct extent_tree *et, struct extent_node *en)
+{
+	rb_erase(&en->rb_node, &et->root);
+	et->count--;
+	atomic_dec(&sbi->total_ext_node);
+
+	if (et->cached_en == en)
+		et->cached_en = NULL;
+}
+
+static struct extent_tree *__find_extent_tree(struct f2fs_sb_info *sbi,
+							nid_t ino)
+{
+	struct extent_tree *et;
+
+	down_read(&sbi->extent_tree_lock);
+	et = radix_tree_lookup(&sbi->extent_tree_root, ino);
+	if (!et) {
+		up_read(&sbi->extent_tree_lock);
+		return NULL;
+	}
+	atomic_inc(&et->refcount);
+	up_read(&sbi->extent_tree_lock);
+
+	return et;
+}
+
+static struct extent_tree *__grab_extent_tree(struct inode *inode)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	struct extent_tree *et;
+	nid_t ino = inode->i_ino;
+
+	down_write(&sbi->extent_tree_lock);
+	et = radix_tree_lookup(&sbi->extent_tree_root, ino);
+	if (!et) {
+		et = f2fs_kmem_cache_alloc(extent_tree_slab, GFP_NOFS);
+		f2fs_radix_tree_insert(&sbi->extent_tree_root, ino, et);
+		memset(et, 0, sizeof(struct extent_tree));
+		et->ino = ino;
+		et->root = RB_ROOT;
+		et->cached_en = NULL;
+		rwlock_init(&et->lock);
+		atomic_set(&et->refcount, 0);
+		et->count = 0;
+		sbi->total_ext_tree++;
+	}
+	atomic_inc(&et->refcount);
+	up_write(&sbi->extent_tree_lock);
+
+	return et;
+}
+
+static struct extent_node *__lookup_extent_tree(struct extent_tree *et,
+							unsigned int fofs)
+{
+	struct rb_node *node = et->root.rb_node;
+	struct extent_node *en;
+
+	if (et->cached_en) {
+		struct extent_info *cei = &et->cached_en->ei;
+
+		if (cei->fofs <= fofs && cei->fofs + cei->len > fofs)
+			return et->cached_en;
+	}
+
+	while (node) {
+		en = rb_entry(node, struct extent_node, rb_node);
+
+		if (fofs < en->ei.fofs) {
+			node = node->rb_left;
+		} else if (fofs >= en->ei.fofs + en->ei.len) {
+			node = node->rb_right;
+		} else {
+			et->cached_en = en;
+			return en;
+		}
+	}
+	return NULL;
+}
+
+static struct extent_node *__try_back_merge(struct f2fs_sb_info *sbi,
+				struct extent_tree *et, struct extent_node *en)
+{
+	struct extent_node *prev;
+	struct rb_node *node;
+
+	node = rb_prev(&en->rb_node);
+	if (!node)
+		return NULL;
+
+	prev = rb_entry(node, struct extent_node, rb_node);
+	if (__is_back_mergeable(&en->ei, &prev->ei)) {
+		en->ei.fofs = prev->ei.fofs;
+		en->ei.blk = prev->ei.blk;
+		en->ei.len += prev->ei.len;
+		__detach_extent_node(sbi, et, prev);
+		return prev;
+	}
+	return NULL;
+}
+
+static struct extent_node *__try_front_merge(struct f2fs_sb_info *sbi,
+				struct extent_tree *et, struct extent_node *en)
+{
+	struct extent_node *next;
+	struct rb_node *node;
+
+	node = rb_next(&en->rb_node);
+	if (!node)
+		return NULL;
+
+	next = rb_entry(node, struct extent_node, rb_node);
+	if (__is_front_mergeable(&en->ei, &next->ei)) {
+		en->ei.len += next->ei.len;
+		__detach_extent_node(sbi, et, next);
+		return next;
+	}
+	return NULL;
+}
+
+static struct extent_node *__insert_extent_tree(struct f2fs_sb_info *sbi,
+				struct extent_tree *et, struct extent_info *ei,
+				struct extent_node **den)
+{
+	struct rb_node **p = &et->root.rb_node;
+	struct rb_node *parent = NULL;
+	struct extent_node *en;
+
+	while (*p) {
+		parent = *p;
+		en = rb_entry(parent, struct extent_node, rb_node);
+
+		if (ei->fofs < en->ei.fofs) {
+			if (__is_front_mergeable(ei, &en->ei)) {
+				f2fs_bug_on(sbi, !den);
+				en->ei.fofs = ei->fofs;
+				en->ei.blk = ei->blk;
+				en->ei.len += ei->len;
+				*den = __try_back_merge(sbi, et, en);
+				return en;
+			}
+			p = &(*p)->rb_left;
+		} else if (ei->fofs >= en->ei.fofs + en->ei.len) {
+			if (__is_back_mergeable(ei, &en->ei)) {
+				f2fs_bug_on(sbi, !den);
+				en->ei.len += ei->len;
+				*den = __try_front_merge(sbi, et, en);
+				return en;
+			}
+			p = &(*p)->rb_right;
+		} else {
+			f2fs_bug_on(sbi, 1);
+		}
+	}
+
+	return __attach_extent_node(sbi, et, ei, parent, p);
+}
+
+static unsigned int __free_extent_tree(struct f2fs_sb_info *sbi,
+					struct extent_tree *et, bool free_all)
+{
+	struct rb_node *node, *next;
+	struct extent_node *en;
+	unsigned int count = et->count;
+
+	node = rb_first(&et->root);
+	while (node) {
+		next = rb_next(node);
+		en = rb_entry(node, struct extent_node, rb_node);
+
+		if (free_all) {
+			spin_lock(&sbi->extent_lock);
+			if (!list_empty(&en->list))
+				list_del_init(&en->list);
+			spin_unlock(&sbi->extent_lock);
+		}
+
+		if (free_all || list_empty(&en->list)) {
+			__detach_extent_node(sbi, et, en);
+			kmem_cache_free(extent_node_slab, en);
+		}
+		node = next;
+	}
+
+	return count - et->count;
+}
+
+static void f2fs_init_extent_tree(struct inode *inode,
+						struct f2fs_extent *i_ext)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	struct extent_tree *et;
+	struct extent_node *en;
+	struct extent_info ei;
+
+	if (le32_to_cpu(i_ext->len) < F2FS_MIN_EXTENT_LEN)
+		return;
+
+	et = __grab_extent_tree(inode);
+
+	write_lock(&et->lock);
+	if (et->count)
+		goto out;
+
+	set_extent_info(&ei, le32_to_cpu(i_ext->fofs),
+		le32_to_cpu(i_ext->blk), le32_to_cpu(i_ext->len));
+
+	en = __insert_extent_tree(sbi, et, &ei, NULL);
+	if (en) {
+		et->cached_en = en;
+
+		spin_lock(&sbi->extent_lock);
+		list_add_tail(&en->list, &sbi->extent_list);
+		spin_unlock(&sbi->extent_lock);
+	}
+out:
+	write_unlock(&et->lock);
+	atomic_dec(&et->refcount);
+}
+
+static bool f2fs_lookup_extent_tree(struct inode *inode, pgoff_t pgofs,
+							struct extent_info *ei)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	struct extent_tree *et;
+	struct extent_node *en;
+
+	trace_f2fs_lookup_extent_tree_start(inode, pgofs);
+
+	et = __find_extent_tree(sbi, inode->i_ino);
+	if (!et)
+		return false;
+
+	read_lock(&et->lock);
+	en = __lookup_extent_tree(et, pgofs);
+	if (en) {
+		*ei = en->ei;
+		spin_lock(&sbi->extent_lock);
+		if (!list_empty(&en->list))
+			list_move_tail(&en->list, &sbi->extent_list);
+		spin_unlock(&sbi->extent_lock);
+		stat_inc_read_hit(sbi->sb);
+	}
+	stat_inc_total_hit(sbi->sb);
+	read_unlock(&et->lock);
+
+	trace_f2fs_lookup_extent_tree_end(inode, pgofs, en);
+
+	atomic_dec(&et->refcount);
+	return en ? true : false;
+}
+
+static void f2fs_update_extent_tree(struct inode *inode, pgoff_t fofs,
+							block_t blkaddr)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	struct extent_tree *et;
+	struct extent_node *en = NULL, *en1 = NULL, *en2 = NULL, *en3 = NULL;
+	struct extent_node *den = NULL;
+	struct extent_info ei, dei;
+	unsigned int endofs;
+
+	trace_f2fs_update_extent_tree(inode, fofs, blkaddr);
+
+	et = __grab_extent_tree(inode);
+
+	write_lock(&et->lock);
+
+	/* 1. lookup and remove existing extent info in cache */
+	en = __lookup_extent_tree(et, fofs);
+	if (!en)
+		goto update_extent;
+
+	dei = en->ei;
+	__detach_extent_node(sbi, et, en);
+
+	/* 2. if extent can be split more, split and insert the left part */
+	if (dei.len > 1) {
+		/*  insert left part of split extent into cache */
+		if (fofs - dei.fofs >= F2FS_MIN_EXTENT_LEN) {
+			set_extent_info(&ei, dei.fofs, dei.blk,
+							fofs - dei.fofs);
+			en1 = __insert_extent_tree(sbi, et, &ei, NULL);
+		}
+
+		/* insert right part of split extent into cache */
+		endofs = dei.fofs + dei.len - 1;
+		if (endofs - fofs >= F2FS_MIN_EXTENT_LEN) {
+			set_extent_info(&ei, fofs + 1,
+				fofs - dei.fofs + dei.blk, endofs - fofs);
+			en2 = __insert_extent_tree(sbi, et, &ei, NULL);
+		}
+	}
+
+update_extent:
+	/* 3. update extent in extent cache */
+	if (blkaddr) {
+		set_extent_info(&ei, fofs, blkaddr, 1);
+		en3 = __insert_extent_tree(sbi, et, &ei, &den);
+	}
+
+	/* 4. update in global extent list */
+	spin_lock(&sbi->extent_lock);
+	if (en && !list_empty(&en->list))
+		list_del(&en->list);
+	/*
+	 * en1 and en2 split from en, they will become more and more smaller
+	 * fragments after splitting several times. So if the length is smaller
+	 * than F2FS_MIN_EXTENT_LEN, we will not add them into extent tree.
+	 */
+	if (en1)
+		list_add_tail(&en1->list, &sbi->extent_list);
+	if (en2)
+		list_add_tail(&en2->list, &sbi->extent_list);
+	if (en3) {
+		if (list_empty(&en3->list))
+			list_add_tail(&en3->list, &sbi->extent_list);
+		else
+			list_move_tail(&en3->list, &sbi->extent_list);
+	}
+	if (den && !list_empty(&den->list))
+		list_del(&den->list);
+	spin_unlock(&sbi->extent_lock);
+
+	/* 5. release extent node */
+	if (en)
+		kmem_cache_free(extent_node_slab, en);
+	if (den)
+		kmem_cache_free(extent_node_slab, den);
+
+	write_unlock(&et->lock);
+	atomic_dec(&et->refcount);
+}
+
+void f2fs_preserve_extent_tree(struct inode *inode)
+{
+	struct extent_tree *et;
+	struct extent_info *ext = &F2FS_I(inode)->ext;
+	bool sync = false;
+
+	if (!test_opt(F2FS_I_SB(inode), EXTENT_CACHE))
+		return;
+
+	et = __find_extent_tree(F2FS_I_SB(inode), inode->i_ino);
+	if (!et) {
+		if (ext->len) {
+			ext->len = 0;
+			update_inode_page(inode);
+		}
+		return;
+	}
+
+	read_lock(&et->lock);
+	if (et->count) {
+		struct extent_node *en;
+
+		if (et->cached_en) {
+			en = et->cached_en;
+		} else {
+			struct rb_node *node = rb_first(&et->root);
+
+			if (!node)
+				node = rb_last(&et->root);
+			en = rb_entry(node, struct extent_node, rb_node);
+		}
+
+		if (__is_extent_same(ext, &en->ei))
+			goto out;
+
+		*ext = en->ei;
+		sync = true;
+	} else if (ext->len) {
+		ext->len = 0;
+		sync = true;
+	}
+out:
+	read_unlock(&et->lock);
+	atomic_dec(&et->refcount);
+
+	if (sync)
+		update_inode_page(inode);
+}
+
+void f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink)
+{
+	struct extent_tree *treevec[EXT_TREE_VEC_SIZE];
+	struct extent_node *en, *tmp;
+	unsigned long ino = F2FS_ROOT_INO(sbi);
+	struct radix_tree_iter iter;
+	void **slot;
+	unsigned int found;
+	unsigned int node_cnt = 0, tree_cnt = 0;
+
+	if (!test_opt(sbi, EXTENT_CACHE))
+		return;
+
+	if (available_free_memory(sbi, EXTENT_CACHE))
+		return;
+
+	spin_lock(&sbi->extent_lock);
+	list_for_each_entry_safe(en, tmp, &sbi->extent_list, list) {
+		if (!nr_shrink--)
+			break;
+		list_del_init(&en->list);
+	}
+	spin_unlock(&sbi->extent_lock);
+
+	down_read(&sbi->extent_tree_lock);
+	while ((found = radix_tree_gang_lookup(&sbi->extent_tree_root,
+				(void **)treevec, ino, EXT_TREE_VEC_SIZE))) {
+		unsigned i;
+
+		ino = treevec[found - 1]->ino + 1;
+		for (i = 0; i < found; i++) {
+			struct extent_tree *et = treevec[i];
+
+			atomic_inc(&et->refcount);
+			write_lock(&et->lock);
+			node_cnt += __free_extent_tree(sbi, et, false);
+			write_unlock(&et->lock);
+			atomic_dec(&et->refcount);
+		}
+	}
+	up_read(&sbi->extent_tree_lock);
+
+	down_write(&sbi->extent_tree_lock);
+	radix_tree_for_each_slot(slot, &sbi->extent_tree_root, &iter,
+							F2FS_ROOT_INO(sbi)) {
+		struct extent_tree *et = (struct extent_tree *)*slot;
+
+		if (!atomic_read(&et->refcount) && !et->count) {
+			radix_tree_delete(&sbi->extent_tree_root, et->ino);
+			kmem_cache_free(extent_tree_slab, et);
+			sbi->total_ext_tree--;
+			tree_cnt++;
+		}
+	}
+	up_write(&sbi->extent_tree_lock);
+
+	trace_f2fs_shrink_extent_tree(sbi, node_cnt, tree_cnt);
+}
+
+void f2fs_destroy_extent_tree(struct inode *inode)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	struct extent_tree *et;
+	unsigned int node_cnt = 0;
+
+	if (!test_opt(sbi, EXTENT_CACHE))
+		return;
+
+	et = __find_extent_tree(sbi, inode->i_ino);
+	if (!et)
+		goto out;
+
+	/* free all extent info belong to this extent tree */
+	write_lock(&et->lock);
+	node_cnt = __free_extent_tree(sbi, et, true);
+	write_unlock(&et->lock);
+
+	atomic_dec(&et->refcount);
+
+	/* try to find and delete extent tree entry in radix tree */
+	down_write(&sbi->extent_tree_lock);
+	et = radix_tree_lookup(&sbi->extent_tree_root, inode->i_ino);
+	if (!et) {
+		up_write(&sbi->extent_tree_lock);
+		goto out;
+	}
+	f2fs_bug_on(sbi, atomic_read(&et->refcount) || et->count);
+	radix_tree_delete(&sbi->extent_tree_root, inode->i_ino);
+	kmem_cache_free(extent_tree_slab, et);
+	sbi->total_ext_tree--;
+	up_write(&sbi->extent_tree_lock);
+out:
+	trace_f2fs_destroy_extent_tree(inode, node_cnt);
 	return;
 }
 
+void f2fs_init_extent_cache(struct inode *inode, struct f2fs_extent *i_ext)
+{
+	if (test_opt(F2FS_I_SB(inode), EXTENT_CACHE))
+		f2fs_init_extent_tree(inode, i_ext);
+
+	write_lock(&F2FS_I(inode)->ext_lock);
+	get_extent_info(&F2FS_I(inode)->ext, *i_ext);
+	write_unlock(&F2FS_I(inode)->ext_lock);
+}
+
+static bool f2fs_lookup_extent_cache(struct inode *inode, pgoff_t pgofs,
+							struct extent_info *ei)
+{
+	if (is_inode_flag_set(F2FS_I(inode), FI_NO_EXTENT))
+		return false;
+
+	if (test_opt(F2FS_I_SB(inode), EXTENT_CACHE))
+		return f2fs_lookup_extent_tree(inode, pgofs, ei);
+
+	return lookup_extent_info(inode, pgofs, ei);
+}
+
+void f2fs_update_extent_cache(struct dnode_of_data *dn)
+{
+	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
+	pgoff_t fofs;
+
+	f2fs_bug_on(F2FS_I_SB(dn->inode), dn->data_blkaddr == NEW_ADDR);
+
+	if (is_inode_flag_set(fi, FI_NO_EXTENT))
+		return;
+
+	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
+							dn->ofs_in_node;
+
+	if (test_opt(F2FS_I_SB(dn->inode), EXTENT_CACHE))
+		return f2fs_update_extent_tree(dn->inode, fofs,
+							dn->data_blkaddr);
+
+	if (update_extent_info(dn->inode, fofs, dn->data_blkaddr))
+		sync_inode_page(dn);
+}
+
 struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync)
 {
 	struct address_space *mapping = inode->i_mapping;
 	struct dnode_of_data dn;
 	struct page *page;
+	struct extent_info ei;
 	int err;
 	struct f2fs_io_info fio = {
 		.type = DATA,
 		.rw = sync ? READ_SYNC : READA,
 	};
 
+	/*
+	 * If sync is false, it needs to check its block allocation.
+	 * This is need and triggered by two flows:
+	 *   gc and truncate_partial_data_page.
+	 */
+	if (!sync)
+		goto searc