1 files changed, 643 insertions, 237 deletions

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 7bd22a20112..f8cf619edb5 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -12,6 +12,7 @@
 #include <linux/f2fs_fs.h>
 #include <linux/buffer_head.h>
 #include <linux/mpage.h>
+#include <linux/aio.h>
 #include <linux/writeback.h>
 #include <linux/backing-dev.h>
 #include <linux/blkdev.h>
@@ -21,6 +22,191 @@
 #include "f2fs.h"
 #include "node.h"
 #include "segment.h"
+#include <trace/events/f2fs.h>
+
+static void f2fs_read_end_io(struct bio *bio, int err)
+{
+	struct bio_vec *bvec;
+	int i;
+
+	bio_for_each_segment_all(bvec, bio, i) {
+		struct page *page = bvec->bv_page;
+
+		if (!err) {
+			SetPageUptodate(page);
+		} else {
+			ClearPageUptodate(page);
+			SetPageError(page);
+		}
+		unlock_page(page);
+	}
+	bio_put(bio);
+}
+
+static void f2fs_write_end_io(struct bio *bio, int err)
+{
+	struct f2fs_sb_info *sbi = bio->bi_private;
+	struct bio_vec *bvec;
+	int i;
+
+	bio_for_each_segment_all(bvec, bio, i) {
+		struct page *page = bvec->bv_page;
+
+		if (unlikely(err)) {
+			SetPageError(page);
+			set_bit(AS_EIO, &page->mapping->flags);
+			f2fs_stop_checkpoint(sbi);
+		}
+		end_page_writeback(page);
+		dec_page_count(sbi, F2FS_WRITEBACK);
+	}
+
+	if (sbi->wait_io) {
+		complete(sbi->wait_io);
+		sbi->wait_io = NULL;
+	}
+
+	if (!get_pages(sbi, F2FS_WRITEBACK) &&
+			!list_empty(&sbi->cp_wait.task_list))
+		wake_up(&sbi->cp_wait);
+
+	bio_put(bio);
+}
+
+/*
+ * Low-level block read/write IO operations.
+ */
+static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr,
+				int npages, bool is_read)
+{
+	struct bio *bio;
+
+	/* No failure on bio allocation */
+	bio = bio_alloc(GFP_NOIO, npages);
+
+	bio->bi_bdev = sbi->sb->s_bdev;
+	bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(sbi, blk_addr);
+	bio->bi_end_io = is_read ? f2fs_read_end_io : f2fs_write_end_io;
+	bio->bi_private = sbi;
+
+	return bio;
+}
+
+static void __submit_merged_bio(struct f2fs_bio_info *io)
+{
+	struct f2fs_io_info *fio = &io->fio;
+	int rw;
+
+	if (!io->bio)
+		return;
+
+	rw = fio->rw;
+
+	if (is_read_io(rw)) {
+		trace_f2fs_submit_read_bio(io->sbi->sb, rw,
+						fio->type, io->bio);
+		submit_bio(rw, io->bio);
+	} else {
+		trace_f2fs_submit_write_bio(io->sbi->sb, rw,
+						fio->type, io->bio);
+		/*
+		 * META_FLUSH is only from the checkpoint procedure, and we
+		 * should wait this metadata bio for FS consistency.
+		 */
+		if (fio->type == META_FLUSH) {
+			DECLARE_COMPLETION_ONSTACK(wait);
+			io->sbi->wait_io = &wait;
+			submit_bio(rw, io->bio);
+			wait_for_completion(&wait);
+		} else {
+			submit_bio(rw, io->bio);
+		}
+	}
+
+	io->bio = NULL;
+}
+
+void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi,
+				enum page_type type, int rw)
+{
+	enum page_type btype = PAGE_TYPE_OF_BIO(type);
+	struct f2fs_bio_info *io;
+
+	io = is_read_io(rw) ? &sbi->read_io : &sbi->write_io[btype];
+
+	down_write(&io->io_rwsem);
+
+	/* change META to META_FLUSH in the checkpoint procedure */
+	if (type >= META_FLUSH) {
+		io->fio.type = META_FLUSH;
+		io->fio.rw = WRITE_FLUSH_FUA | REQ_META | REQ_PRIO;
+	}
+	__submit_merged_bio(io);
+	up_write(&io->io_rwsem);
+}
+
+/*
+ * Fill the locked page with data located in the block address.
+ * Return unlocked page.
+ */
+int f2fs_submit_page_bio(struct f2fs_sb_info *sbi, struct page *page,
+					block_t blk_addr, int rw)
+{
+	struct bio *bio;
+
+	trace_f2fs_submit_page_bio(page, blk_addr, rw);
+
+	/* Allocate a new bio */
+	bio = __bio_alloc(sbi, blk_addr, 1, is_read_io(rw));
+
+	if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) {
+		bio_put(bio);
+		f2fs_put_page(page, 1);
+		return -EFAULT;
+	}
+
+	submit_bio(rw, bio);
+	return 0;
+}
+
+void f2fs_submit_page_mbio(struct f2fs_sb_info *sbi, struct page *page,
+			block_t blk_addr, struct f2fs_io_info *fio)
+{
+	enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
+	struct f2fs_bio_info *io;
+	bool is_read = is_read_io(fio->rw);
+
+	io = is_read ? &sbi->read_io : &sbi->write_io[btype];
+
+	verify_block_addr(sbi, blk_addr);
+
+	down_write(&io->io_rwsem);
+
+	if (!is_read)
+		inc_page_count(sbi, F2FS_WRITEBACK);
+
+	if (io->bio && (io->last_block_in_bio != blk_addr - 1 ||
+						io->fio.rw != fio->rw))
+		__submit_merged_bio(io);
+alloc_new:
+	if (io->bio == NULL) {
+		int bio_blocks = MAX_BIO_BLOCKS(max_hw_blocks(sbi));
+
+		io->bio = __bio_alloc(sbi, blk_addr, bio_blocks, is_read);
+		io->fio = *fio;
+	}
+
+	if (bio_add_page(io->bio, page, PAGE_CACHE_SIZE, 0) <
+							PAGE_CACHE_SIZE) {
+		__submit_merged_bio(io);
+		goto alloc_new;
+	}
+
+	io->last_block_in_bio = blk_addr;
+
+	up_write(&io->io_rwsem);
+	trace_f2fs_submit_page_mbio(page, fio->rw, fio->type, blk_addr);
+}
 
 /*
  * Lock ordering for the change of data block address:
@@ -35,9 +221,9 @@ static void __set_data_blkaddr(struct dnode_of_data *dn, block_t new_addr)
 	struct page *node_page = dn->node_page;
 	unsigned int ofs_in_node = dn->ofs_in_node;
 
-	wait_on_page_writeback(node_page);
+	f2fs_wait_on_page_writeback(node_page, NODE);
 
-	rn = (struct f2fs_node *)page_address(node_page);
+	rn = F2FS_NODE(node_page);
 
 	/* Get physical address of data block */
 	addr_array = blkaddr_in_node(rn);
@@ -49,32 +235,57 @@ int reserve_new_block(struct dnode_of_data *dn)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
 
-	if (is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC))
+	if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC)))
 		return -EPERM;
-	if (!inc_valid_block_count(sbi, dn->inode, 1))
+	if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1)))
 		return -ENOSPC;
 
+	trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node);
+
 	__set_data_blkaddr(dn, NEW_ADDR);
 	dn->data_blkaddr = NEW_ADDR;
+	mark_inode_dirty(dn->inode);
 	sync_inode_page(dn);
 	return 0;
 }
 
+int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
+{
+	bool need_put = dn->inode_page ? false : true;
+	int err;
+
+	/* if inode_page exists, index should be zero */
+	f2fs_bug_on(!need_put && index);
+
+	err = get_dnode_of_data(dn, index, ALLOC_NODE);
+	if (err)
+		return err;
+
+	if (dn->data_blkaddr == NULL_ADDR)
+		err = reserve_new_block(dn);
+	if (err || need_put)
+		f2fs_put_dnode(dn);
+	return err;
+}
+
 static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
 					struct buffer_head *bh_result)
 {
 	struct f2fs_inode_info *fi = F2FS_I(inode);
-	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
 	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);
 	if (fi->ext.len == 0) {
 		read_unlock(&fi->ext.ext_lock);
 		return 0;
 	}
 
-	sbi->total_hit_ext++;
+	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;
@@ -92,7 +303,7 @@ static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
 		else
 			bh_result->b_size = UINT_MAX;
 
-		sbi->read_hit_ext++;
+		stat_inc_read_hit(inode->i_sb);
 		read_unlock(&fi->ext.ext_lock);
 		return 1;
 	}
@@ -105,13 +316,18 @@ void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
 	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
 	pgoff_t fofs, start_fofs, end_fofs;
 	block_t start_blkaddr, end_blkaddr;
+	int need_update = true;
 
-	BUG_ON(blk_addr == NEW_ADDR);
-	fofs = start_bidx_of_node(ofs_of_node(dn->node_page)) + dn->ofs_in_node;
+	f2fs_bug_on(blk_addr == NEW_ADDR);
+	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
+							dn->ofs_in_node;
 
 	/* Update the page address in the parent node */
 	__set_data_blkaddr(dn, blk_addr);
 
+	if (is_inode_flag_set(fi, FI_NO_EXTENT))
+		return;
+
 	write_lock(&fi->ext.ext_lock);
 
 	start_fofs = fi->ext.fofs;
@@ -133,7 +349,7 @@ void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
 		goto end_update;
 	}
 
-	/* Frone merge */
+	/* Front merge */
 	if (fofs == start_fofs - 1 && blk_addr == start_blkaddr - 1) {
 		fi->ext.fofs--;
 		fi->ext.blk_addr--;
@@ -158,18 +374,24 @@ void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
 					fofs - start_fofs + 1;
 			fi->ext.len -= fofs - start_fofs + 1;
 		}
-		goto end_update;
+	} else {
+		need_update = false;
 	}
-	write_unlock(&fi->ext.ext_lock);
-	return;
 
+	/* Finally, if the extent is very fragmented, let's drop the cache. */
+	if (fi->ext.len < F2FS_MIN_EXTENT_LEN) {
+		fi->ext.len = 0;
+		set_inode_flag(fi, FI_NO_EXTENT);
+		need_update = true;
+	}
 end_update:
 	write_unlock(&fi->ext.ext_lock);
-	sync_inode_page(dn);
+	if (need_update)
+		sync_inode_page(dn);
 	return;
 }
 
-struct page *find_data_page(struct inode *inode, pgoff_t index)
+struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
 	struct address_space *mapping = inode->i_mapping;
@@ -183,7 +405,7 @@ struct page *find_data_page(struct inode *inode, pgoff_t index)
 	f2fs_put_page(page, 0);
 
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
-	err = get_dnode_of_data(&dn, index, RDONLY_NODE);
+	err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
 	if (err)
 		return ERR_PTR(err);
 	f2fs_put_dnode(&dn);
@@ -192,19 +414,30 @@ struct page *find_data_page(struct inode *inode, pgoff_t index)
 		return ERR_PTR(-ENOENT);
 
 	/* By fallocate(), there is no cached page, but with NEW_ADDR */
-	if (dn.data_blkaddr == NEW_ADDR)
+	if (unlikely(dn.data_blkaddr == NEW_ADDR))
 		return ERR_PTR(-EINVAL);
 
 	page = grab_cache_page(mapping, index);
 	if (!page)
 		return ERR_PTR(-ENOMEM);
 
-	err = f2fs_readpage(sbi, page, dn.data_blkaddr, READ_SYNC);
-	if (err) {
-		f2fs_put_page(page, 1);
+	if (PageUptodate(page)) {
+		unlock_page(page);
+		return page;
+	}
+
+	err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr,
+					sync ? READ_SYNC : READA);
+	if (err)
 		return ERR_PTR(err);
+
+	if (sync) {
+		wait_on_page_locked(page);
+		if (unlikely(!PageUptodate(page))) {
+			f2fs_put_page(page, 0);
+			return ERR_PTR(-EIO);
+		}
 	}
-	unlock_page(page);
 	return page;
 }
 
@@ -221,29 +454,51 @@ struct page *get_lock_data_page(struct inode *inode, pgoff_t index)
 	struct page *page;
 	int err;
 
+repeat:
+	page = grab_cache_page(mapping, index);
+	if (!page)
+		return ERR_PTR(-ENOMEM);
+
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
-	err = get_dnode_of_data(&dn, index, RDONLY_NODE);
-	if (err)
+	err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
+	if (err) {
+		f2fs_put_page(page, 1);
 		return ERR_PTR(err);
+	}
 	f2fs_put_dnode(&dn);
 
-	if (dn.data_blkaddr == NULL_ADDR)
+	if (unlikely(dn.data_blkaddr == NULL_ADDR)) {
+		f2fs_put_page(page, 1);
 		return ERR_PTR(-ENOENT);
-
-	page = grab_cache_page(mapping, index);
-	if (!page)
-		return ERR_PTR(-ENOMEM);
+	}
 
 	if (PageUptodate(page))
 		return page;
 
-	BUG_ON(dn.data_blkaddr == NEW_ADDR);
-	BUG_ON(dn.data_blkaddr == NULL_ADDR);
+	/*
+	 * A new dentry page is allocated but not able to be written, since its
+	 * new inode page couldn't be allocated due to -ENOSPC.
+	 * In such the case, its blkaddr can be remained as NEW_ADDR.
+	 * see, f2fs_add_link -> get_new_data_page -> init_inode_metadata.
+	 */
+	if (dn.data_blkaddr == NEW_ADDR) {
+		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
+		SetPageUptodate(page);
+		return page;
+	}
 
-	err = f2fs_readpage(sbi, page, dn.data_blkaddr, READ_SYNC);
-	if (err) {
-		f2fs_put_page(page, 1);
+	err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr, READ_SYNC);
+	if (err)
 		return ERR_PTR(err);
+
+	lock_page(page);
+	if (unlikely(!PageUptodate(page))) {
+		f2fs_put_page(page, 1);
+		return ERR_PTR(-EIO);
+	}
+	if (unlikely(page->mapping != mapping)) {
+		f2fs_put_page(page, 1);
+		goto repeat;
 	}
 	return page;
 }
@@ -251,9 +506,13 @@ struct page *get_lock_data_page(struct inode *inode, pgoff_t index)
 /*
  * Caller ensures that this data page is never allocated.
  * A new zero-filled data page is allocated in the page cache.
+ *
+ * Also, caller should grab and release a rwsem by calling f2fs_lock_op() and
+ * f2fs_unlock_op().
+ * Note that, ipage is set only by make_empty_dir.
  */
-struct page *get_new_data_page(struct inode *inode, pgoff_t index,
-						bool new_i_size)
+struct page *get_new_data_page(struct inode *inode,
+		struct page *ipage, pgoff_t index, bool new_i_size)
 {
 	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
 	struct address_space *mapping = inode->i_mapping;
@@ -261,196 +520,258 @@ struct page *get_new_data_page(struct inode *inode, pgoff_t index,
 	struct dnode_of_data dn;
 	int err;
 
-	set_new_dnode(&dn, inode, NULL, NULL, 0);
-	err = get_dnode_of_data(&dn, index, 0);
+	set_new_dnode(&dn, inode, ipage, NULL, 0);
+	err = f2fs_reserve_block(&dn, index);
 	if (err)
 		return ERR_PTR(err);
-
-	if (dn.data_blkaddr == NULL_ADDR) {
-		if (reserve_new_block(&dn)) {
-			f2fs_put_dnode(&dn);
-			return ERR_PTR(-ENOSPC);
-		}
-	}
-	f2fs_put_dnode(&dn);
-
+repeat:
 	page = grab_cache_page(mapping, index);
-	if (!page)
-		return ERR_PTR(-ENOMEM);
+	if (!page) {
+		err = -ENOMEM;
+		goto put_err;
+	}
 
 	if (PageUptodate(page))
 		return page;
 
 	if (dn.data_blkaddr == NEW_ADDR) {
 		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
+		SetPageUptodate(page);
 	} else {
-		err = f2fs_readpage(sbi, page, dn.data_blkaddr, READ_SYNC);
-		if (err) {
+		err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr,
+								READ_SYNC);
+		if (err)
+			goto put_err;
+
+		lock_page(page);
+		if (unlikely(!PageUptodate(page))) {
+			f2fs_put_page(page, 1);
+			err = -EIO;
+			goto put_err;
+		}
+		if (unlikely(page->mapping != mapping)) {
 			f2fs_put_page(page, 1);
-			return ERR_PTR(err);
+			goto repeat;
 		}
 	}
-	SetPageUptodate(page);
 
 	if (new_i_size &&
 		i_size_read(inode) < ((index + 1) << PAGE_CACHE_SHIFT)) {
 		i_size_write(inode, ((index + 1) << PAGE_CACHE_SHIFT));
-		mark_inode_dirty_sync(inode);
+		/* Only the directory inode sets new_i_size */
+		set_inode_flag(F2FS_I(inode), FI_UPDATE_DIR);
 	}
 	return page;
-}
-
-static void read_end_io(struct bio *bio, int err)
-{
-	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
 
-	do {
-		struct page *page = bvec->bv_page;
-
-		if (--bvec >= bio->bi_io_vec)
-			prefetchw(&bvec->bv_page->flags);
-
-		if (uptodate) {
-			SetPageUptodate(page);
-		} else {
-			ClearPageUptodate(page);
-			SetPageError(page);
-		}
-		unlock_page(page);
-	} while (bvec >= bio->bi_io_vec);
-	kfree(bio->bi_private);
-	bio_put(bio);
+put_err:
+	f2fs_put_dnode(&dn);
+	return ERR_PTR(err);
 }
 
-/*
- * Fill the locked page with data located in the block address.
- * Read operation is synchronous, and caller must unlock the page.
- */
-int f2fs_readpage(struct f2fs_sb_info *sbi, struct page *page,
-					block_t blk_addr, int type)
+static int __allocate_data_block(struct dnode_of_data *dn)
 {
-	struct block_device *bdev = sbi->sb->s_bdev;
-	bool sync = (type == READ_SYNC);
-	struct bio *bio;
+	struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb);
+	struct f2fs_summary sum;
+	block_t new_blkaddr;
+	struct node_info ni;
+	int type;
 
-	/* This page can be already read by other threads */
-	if (PageUptodate(page)) {
-		if (!sync)
-			unlock_page(page);
-		return 0;
-	}
+	if (unlikely(is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC)))
+		return -EPERM;
+	if (unlikely(!inc_valid_block_count(sbi, dn->inode, 1)))
+		return -ENOSPC;
 
-	down_read(&sbi->bio_sem);
+	__set_data_blkaddr(dn, NEW_ADDR);
+	dn->data_blkaddr = NEW_ADDR;
 
-	/* Allocate a new bio */
-	bio = f2fs_bio_alloc(bdev, 1);
+	get_node_info(sbi, dn->nid, &ni);
+	set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
 
-	/* Initialize the bio */
-	bio->bi_sector = SECTOR_FROM_BLOCK(sbi, blk_addr);
-	bio->bi_end_io = read_end_io;
+	type = CURSEG_WARM_DATA;
 
-	if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) {
-		kfree(bio->bi_private);
-		bio_put(bio);
-		up_read(&sbi->bio_sem);
-		return -EFAULT;
-	}
+	allocate_data_block(sbi, NULL, NULL_ADDR, &new_blkaddr, &sum, type);
 
-	submit_bio(type, bio);
-	up_read(&sbi->bio_sem);
+	/* direct IO doesn't use extent cache to maximize the performance */
+	set_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT);
+	update_extent_cache(new_blkaddr, dn);
+	clear_inode_flag(F2FS_I(dn->inode), FI_NO_EXTENT);
 
-	/* wait for read completion if sync */
-	if (sync) {
-		lock_page(page);
-		if (PageError(page))
-			return -EIO;
-	}
+	dn->data_blkaddr = new_blkaddr;
 	return 0;
 }
 
 /*
- * This function should be used by the data read flow only where it
- * does not check the "create" flag that indicates block allocation.
- * The reason for this special functionality is to exploit VFS readahead
- * mechanism.
+ * get_data_block() now supported readahead/bmap/rw direct_IO with mapped bh.
+ * If original data blocks are allocated, then give them to blockdev.
+ * Otherwise,
+ *     a. preallocate requested block addresses
+ *     b. do not use extent cache for better performance
+ *     c. give the block addresses to blockdev
  */
-static int get_data_block_ro(struct inode *inode, sector_t iblock,
-			struct buffer_head *bh_result, int create)
+static int __get_data_block(struct inode *inode, sector_t iblock,
+			struct buffer_head *bh_result, int create, bool fiemap)
 {
+	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
 	unsigned int blkbits = inode->i_sb->s_blocksize_bits;
 	unsigned maxblocks = bh_result->b_size >> blkbits;
 	struct dnode_of_data dn;
-	pgoff_t pgofs;
-	int err;
+	int mode = create ? ALLOC_NODE : LOOKUP_NODE_RA;
+	pgoff_t pgofs, end_offset;
+	int err = 0, ofs = 1;
+	bool allocated = false;
 
 	/* Get the page offset from the block offset(iblock) */
 	pgofs =	(pgoff_t)(iblock >> (PAGE_CACHE_SHIFT - blkbits));
 
 	if (check_extent_cache(inode, pgofs, bh_result))
-		return 0;
+		goto out;
+
+	if (create)
+		f2fs_lock_op(sbi);
 
 	/* When reading holes, we need its node page */
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
-	err = get_dnode_of_data(&dn, pgofs, RDONLY_NODE);
-	if (err)
-		return (err == -ENOENT) ? 0 : err;
+	err = get_dnode_of_data(&dn, pgofs, mode);
+	if (err) {
+		if (err == -ENOENT)
+			err = 0;
+		goto unlock_out;
+	}
+	if (dn.data_blkaddr == NEW_ADDR && !fiemap)
+		goto put_out;
 
-	/* It does not support data allocation */
-	BUG_ON(create);
+	if (dn.data_blkaddr != NULL_ADDR) {
+		map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
+	} else if (create) {
+		err = __allocate_data_block(&dn);
+		if (err)
+			goto put_out;
+		allocated = true;
+		map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
+	} else {
+		goto put_out;
+	}
 
-	if (dn.data_blkaddr != NEW_ADDR && dn.data_blkaddr != NULL_ADDR) {
-		int i;
-		unsigned int end_offset;
+	end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
+	bh_result->b_size = (((size_t)1) << blkbits);
+	dn.ofs_in_node++;
+	pgofs++;
 
-		end_offset = IS_INODE(dn.node_page) ?
-				ADDRS_PER_INODE :
-				ADDRS_PER_BLOCK;
+get_next:
+	if (dn.ofs_in_node >= end_offset) {
+		if (allocated)
+			sync_inode_page(&dn);
+		allocated = false;
+		f2fs_put_dnode(&dn);
 
-		clear_buffer_new(bh_result);
+		set_new_dnode(&dn, inode, NULL, NULL, 0);
+		err = get_dnode_of_data(&dn, pgofs, mode);
+		if (err) {
+			if (err == -ENOENT)
+				err = 0;
+			goto unlock_out;
+		}
+		if (dn.data_blkaddr == NEW_ADDR && !fiemap)
+			goto put_out;
 
+		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
+	}
+
+	if (maxblocks > (bh_result->b_size >> blkbits)) {
+		block_t blkaddr = datablock_addr(dn.node_page, dn.ofs_in_node);
+		if (blkaddr == NULL_ADDR && create) {
+			err = __allocate_data_block(&dn);
+			if (err)
+				goto sync_out;
+			allocated = true;
+			blkaddr = dn.data_blkaddr;
+		}
 		/* Give more consecutive addresses for the read ahead */
-		for (i = 0; i < end_offset - dn.ofs_in_node; i++)
-			if (((datablock_addr(dn.node_page,
-							dn.ofs_in_node + i))
-				!= (dn.data_blkaddr + i)) || maxblocks == i)
-				break;
-		map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
-		bh_result->b_size = (i << blkbits);
+		if (blkaddr == (bh_result->b_blocknr + ofs)) {
+			ofs++;
+			dn.ofs_in_node++;
+			pgofs++;
+			bh_result->b_size += (((size_t)1) << blkbits);
+			goto get_next;
+		}
 	}
+sync_out:
+	if (allocated)
+		sync_inode_page(&dn);
+put_out:
 	f2fs_put_dnode(&dn);
-	return 0;
+unlock_out:
+	if (create)
+		f2fs_unlock_op(sbi);
+out:
+	trace_f2fs_get_data_block(inode, iblock, bh_result, err);
+	return err;
+}
+
+static int get_data_block(struct inode *inode, sector_t iblock,
+			struct buffer_head *bh_result, int create)
+{
+	return __get_data_block(inode, iblock, bh_result, create, false);
+}
+
+static int get_data_block_fiemap(struct inode *inode, sector_t iblock,
+			struct buffer_head *bh_result, int create)
+{
+	return __get_data_block(inode, iblock, bh_result, create, true);
+}
+
+int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
+		u64 start, u64 len)
+{
+	return generic_block_fiemap(inode, fieinfo,
+				start, len, get_data_block_fiemap);
 }
 
 static int f2fs_read_data_page(struct file *file, struct page *page)
 {
-	return mpage_readpage(page, get_data_block_ro);
+	struct inode *inode = page->mapping->host;
+	int ret;
+
+	trace_f2fs_readpage(page, DATA);
+
+	/* If the file has inline data, try to read it directlly */
+	if (f2fs_has_inline_data(inode))
+		ret = f2fs_read_inline_data(inode, page);
+	else
+		ret = mpage_readpage(page, get_data_block);
+
+	return ret;
 }
 
 static int f2fs_read_data_pages(struct file *file,
 			struct address_space *mapping,
 			struct list_head *pages, unsigned nr_pages)
 {
-	return mpage_readpages(mapping, pages, nr_pages, get_data_block_ro);
+	struct inode *inode = file->f_mapping->host;
+
+	/* If the file has inline data, skip readpages */
+	if (f2fs_has_inline_data(inode))
+		return 0;
+
+	return mpage_readpages(mapping, pages, nr_pages, get_data_block);
 }
 
-int do_write_data_page(struct page *page)
+int do_write_data_page(struct page *page, struct f2fs_io_info *fio)
 {
 	struct inode *inode = page->mapping->host;
-	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
-	block_t old_blk_addr, new_blk_addr;
+	block_t old_blkaddr, new_blkaddr;
 	struct dnode_of_data dn;
 	int err = 0;
 
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
-	err = get_dnode_of_data(&dn, page->index, RDONLY_NODE);
+	err = get_dnode_of_data(&dn, page->index, LOOKUP_NODE);
 	if (err)
 		return err;
 
-	old_blk_addr = dn.data_blkaddr;
+	old_blkaddr = dn.data_blkaddr;
 
 	/* This page is already truncated */
-	if (old_blk_addr == NULL_ADDR)
+	if (old_blkaddr == NULL_ADDR)
 		goto out_writepage;
 
 	set_page_writeback(page);
@@ -459,16 +780,13 @@ int do_write_data_page(struct page *page)
 	 * If current allocation needs SSR,
 	 * it had better in-place writes for updated data.
 	 */
-	if (old_blk_addr != NEW_ADDR && !is_cold_data(page) &&
-				need_inplace_update(inode)) {
-		rewrite_data_page(F2FS_SB(inode->i_sb), page,
-						old_blk_addr);
+	if (unlikely(old_blkaddr != NEW_ADDR &&
+			!is_cold_data(page) &&
+			need_inplace_update(inode))) {
+		rewrite_data_page(page, old_blkaddr, fio);
 	} else {
-		write_data_page(inode, page, &dn,
-				old_blk_addr, &new_blk_addr);
-		update_extent_cache(new_blk_addr, &dn);
-		F2FS_I(inode)->data_version =
-			le64_to_cpu(F2FS_CKPT(sbi)->checkpoint_ver);
+		write_data_page(page, &dn, &new_blkaddr, fio);
+		update_extent_cache(new_blkaddr, &dn);
 	}
 out_writepage:
 	f2fs_put_dnode(&dn);
@@ -483,70 +801,68 @@ static int f2fs_write_data_page(struct page *page,
 	loff_t i_size = i_size_read(inode);
 	const pgoff_t end_index = ((unsigned long long) i_size)
 							>> PAGE_CACHE_SHIFT;
-	unsigned offset;
+	unsigned offset = 0;
+	bool need_balance_fs = false;
 	int err = 0;
+	struct f2fs_io_info fio = {
+		.type = DATA,
+		.rw = (wbc->sync_mode == WB_SYNC_ALL) ? WRITE_SYNC : WRITE,
+	};
+
+	trace_f2fs_writepage(page, DATA);
 
 	if (page->index < end_index)
-		goto out;
+		goto write;
 
 	/*
 	 * If the offset is out-of-range of file size,
 	 * this page does not have to be written to disk.
 	 */
 	offset = i_size & (PAGE_CACHE_SIZE - 1);
-	if ((page->index >= end_index + 1) || !offset) {
-		if (S_ISDIR(inode->i_mode)) {
-			dec_page_count(sbi, F2FS_DIRTY_DENTS);
-			inode_dec_dirty_dents(inode);
-		}
-		goto unlock_out;
-	}
+	if ((page->index >= end_index + 1) || !offset)
+		goto out;
 
 	zero_user_segment(page, offset, PAGE_CACHE_SIZE);
-out:
-	if (sbi->por_doing)
-		goto redirty_out;
-
-	if (wbc->for_reclaim && !S_ISDIR(inode->i_mode) && !is_cold_data(page))
+write:
+	if (unlikely(sbi->por_doing))
 		goto redirty_out;
 
-	mutex_lock_op(sbi, DATA_WRITE);
+	/* Dentry blocks are controlled by checkpoint */
 	if (S_ISDIR(inode->i_mode)) {
-		dec_page_count(sbi, F2FS_DIRTY_DENTS);
-		inode_dec_dirty_dents(inode);
-	}
-	err = do_write_data_page(page);
-	if (err && err != -ENOENT) {
-		wbc->pages_skipped++;
-		set_page_dirty(page);
+		err = do_write_data_page(page, &fio);
+		goto done;
 	}
-	mutex_unlock_op(sbi, DATA_WRITE);
 
-	if (wbc->for_reclaim)
-		f2fs_submit_bio(sbi, DATA, true);
+	if (!wbc->for_reclaim)
+		need_balance_fs = true;
+	else if (has_not_enough_free_secs(sbi, 0))
+		goto redirty_out;
 
-	if (err == -ENOENT)
-		goto unlock_out;
+	f2fs_lock_op(sbi);
+	if (f2fs_has_inline_data(inode) || f2fs_may_inline(inode))
+		err = f2fs_write_inline_data(inode, page, offset);
+	else
+		err = do_write_data_page(page, &fio);
+	f2fs_unlock_op(sbi);
+done:
+	if (err && err != -ENOENT)
+		goto redirty_out;
 
 	clear_cold_data(page);
+out:
+	inode_dec_dirty_dents(inode);
 	unlock_page(page);
-
-	if (!wbc->for_reclaim && !S_ISDIR(inode->i_mode))
+	if (need_balance_fs)
 		f2fs_balance_fs(sbi);
+	if (wbc->for_reclaim)
+		f2fs_submit_merged_bio(sbi, DATA, WRITE);
 	return 0;
 
-unlock_out:
-	unlock_page(page);
-	return (err == -ENOENT) ? 0 : err;
-
 redirty_out:
-	wbc->pages_skipped++;
-	set_page_dirty(page);
+	redirty_page_for_writepage(wbc, page);
 	return AOP_WRITEPAGE_ACTIVATE;
 }
 
-#define MAX_DESIRED_PAGES_WP	4096
-
 static int __f2fs_writepage(struct page *page, struct writeback_control *wbc,
 			void *data)
 {
@@ -561,26 +877,41 @@ static int f2fs_write_data_pages(struct address_space *mapping,
 {
 	struct inode *inode = mapping->host;
 	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
+	bool locked = false;
 	int ret;
-	long excess_nrtw = 0, desired_nrtw;
+	long diff;
 
-	if (wbc->nr_to_write < MAX_DESIRED_PAGES_WP) {
-		desired_nrtw = MAX_DESIRED_PAGES_WP;
-		excess_nrtw = desired_nrtw - wbc->nr_to_write;
-		wbc->nr_to_write = desired_nrtw;
-	}
+	trace_f2fs_writepages(mapping->host, wbc, DATA);
+
+	/* deal with chardevs and other special file */
+	if (!mapping->a_ops->writepage)
+		return 0;
 
-	if (!S_ISDIR(inode->i_mode))
+	if (S_ISDIR(inode->i_mode) && wbc->sync_mode == WB_SYNC_NONE &&
+			get_dirty_dents(inode) < nr_pages_to_skip(sbi, DATA) &&
+			available_free_memory(sbi, DIRTY_DENTS))
+		goto skip_write;
+
+	diff = nr_pages_to_write(sbi, DATA, wbc);
+
+	if (!S_ISDIR(inode->i_mode)) {
 		mutex_lock(&sbi->writepages);
+		locked = true;
+	}
 	ret = write_cache_pages(mapping, wbc, __f2fs_writepage, mapping);
-	if (!S_ISDIR(inode->i_mode))
+	if (locked)
 		mutex_unlock(&sbi->writepages);
-	f2fs_submit_bio(sbi, DATA, (wbc->sync_mode == WB_SYNC_ALL));
+
+	f2fs_submit_merged_bio(sbi, DATA, WRITE);
 
 	remove_dirty_dir_inode(inode);
 
-	wbc->nr_to_write -= excess_nrtw;
+	wbc->nr_to_write = max((long)0, wbc->nr_to_write - diff);
 	return ret;
+
+skip_write:
+	wbc->pages_skipped += get_dirty_dents(inode);
+	return 0;
 }
 
 static int f2fs_write_begin(struct file *file, struct address_space *mapping,
@@ -594,38 +925,43 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 	struct dnode_of_data dn;
 	int err = 0;
 
-	/* for nobh_write_end */
-	*fsdata = NULL;
+	trace_f2fs_write_begin(inode, pos, len, flags);
 
 	f2fs_balance_fs(sbi);
+repeat:
+	err = f2fs_convert_inline_data(inode, pos + len);
+	if (err)
+		return err;
 
 	page = grab_cache_page_write_begin(mapping, index, flags);
 	if (!page)
 		return -ENOMEM;
+
+	/* to avoid latency during memory pressure */
+	unlock_page(page);
+
 	*pagep = page;
 
-	mutex_lock_op(sbi, DATA_NEW);
+	if (f2fs_has_inline_data(inode) && (pos + len) <= MAX_INLINE_DATA)
+		goto inline_data;
 
+	f2fs_lock_op(sbi);
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
-	err = get_dnode_of_data(&dn, index, 0);
+	err = f2fs_reserve_block(&dn, index);
+	f2fs_unlock_op(sbi);
+
 	if (err) {
-		mutex_unlock_op(sbi, DATA_NEW);
-		f2fs_put_page(page, 1);
+		f2fs_put_page(page, 0);
 		return err;
 	}
-
-	if (dn.data_blkaddr == NULL_ADDR) {
-		err = reserve_new_block(&dn);
-		if (err) {
-			f2fs_put_dnode(&dn);
-			mutex_unlock_op(sbi, DATA_NEW);
-			f2fs_put_page(page, 1);
-			return err;
-		}
+inline_data:
+	lock_page(page);
+	if (unlikely(page->mapping != mapping)) {
+		f2fs_put_page(page, 1);
+		goto repeat;
 	}
-	f2fs_put_dnode(&dn);
 
-	mutex_unlock_op(sbi, DATA_NEW);
+	f2fs_wait_on_page_writeback(page, DATA);
 
 	if ((len == PAGE_CACHE_SIZE) || PageUptodate(page))
 		return 0;
@@ -636,52 +972,113 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 
 		/* Reading beyond i_size is simple: memset to zero */
 		zero_user_segments(page, 0, start, end, PAGE_CACHE_SIZE);
-		return 0;
+		goto out;
 	}
 
 	if (dn.data_blkaddr == NEW_ADDR) {
 		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
 	} else {
-		err = f2fs_readpage(sbi, page, dn.data_blkaddr, READ_SYNC);
-		if (err) {
+		if (f2fs_has_inline_data(inode)) {
+			err = f2fs_read_inline_data(inode, page);
+			if (err) {
+				page_cache_release(page);
+				return err;
+			}
+		} else {
+			err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr,
+							READ_SYNC);
+			if (err)
+				return err;
+		}
+
+		lock_page(page);
+		if (unlikely(!PageUptodate(page))) {
 			f2fs_put_page(page, 1);
-			return err;
+			return -EIO;
+		}
+		if (unlikely(page->mapping != mapping)) {
+			f2fs_put_page(page, 1);
+			goto repeat;
 		}
 	}
+out:
 	SetPageUptodate(page);
 	clear_cold_data(page);
 	return 0;
 }
 
+static int f2fs_write_end(struct file *file,
+			struct address_space *mapping,
+			loff_t pos, unsigned len, unsigned copied,
+			struct page *page, void *fsdata)
+{
+	struct inode *inode = page->mapping->host;
+
+	trace_f2fs_write_end(inode, pos, len, copied);
+
+	SetPageUptodate(page);
+	set_page_dirty(page);
+
+	if (pos + copied > i_size_read(inode)) {
+		i_size_write(inode, pos + copied);
+		mark_inode_dirty(inode);
+		update_inode_page(inode);
+	}
+
+	f2fs_put_page(page, 1);
+	return copied;
+}
+
+static int check_direct_IO(struct inode *inode, int rw,
+		struct iov_iter *iter, loff_t offset)
+{
+	unsigned blocksize_mask = inode->i_sb->s_blocksize - 1;
+
+	if (rw == READ)
+		return 0;
+
+	if (offset & blocksize_mask)
+		return -EINVAL;
+
+	if (iov_iter_alignment(iter) & blocksize_mask)
+		return -EINVAL;
+
+	return 0;
+}
+
 static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb,
-		const struct iovec *iov, loff_t offset, unsigned long nr_segs)
+		struct iov_iter *iter, loff_t offset)
 {
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file->f_mapping->host;
 
-	if (rw == WRITE)
+	/* Let buffer I/O handle the inline data case. */
+	if (f2fs_has_inline_data(inode))
+		return 0;
+
+	if (check_direct_IO(inode, rw, iter, offset))
 		return 0;
 
-	/* Needs synchronization with the cleaner */
-	return blockdev_direct_IO(rw, iocb, inode, iov, offset, nr_segs,
-						  get_data_block_ro);
+	/* 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);
 }
 
-static void f2fs_invalidate_data_page(struct page *page, unsigned long offset)
+static void f2fs_invalidate_data_page(struct page *page, unsigned int offset,
+				      unsigned int length)
 {
 	struct inode *inode = page->mapping->host;
-	struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb);
-	if (S_ISDIR(inode->i_mode) && PageDirty(page)) {
-		dec_page_count(sbi, F2FS_DIRTY_DENTS);
+	if (PageDirty(page))
 		inode_dec_dirty_dents(inode);
-	}
 	ClearPagePrivate(page);
 }
 
 static int f2fs_release_data_page(struct page *page, gfp_t wait)
 {
 	ClearPagePrivate(page);
-	return 0;
+	return 1;
 }
 
 static int f2fs_set_data_page_dirty(struct page *page)
@@ -689,7 +1086,11 @@ static int f2fs_set_data_page_dirty(struct page *page)
 	struct address_space *mapping = page->mapping;
 	struct inode *inode = mapping->host;
 
+	trace_f2fs_set_page_dirty(page, DATA);
+
 	SetPageUptodate(page);
+	mark_inode_dirty(inode);
+
 	if (!PageDirty(page)) {
 		__set_page_dirty_nobuffers(page);
 		set_dirty_dir_page(inode, page);
@@ -700,7 +1101,12 @@ static int f2fs_set_data_page_dirty(struct page *page)
 
 static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
 {
-	return generic_block_bmap(mapping, block, get_data_block_ro);
+	struct inode *inode = mapping->host;
+
+	if (f2fs_has_inline_data(inode))
+		return 0;
+
+	return generic_block_bmap(mapping, block, get_data_block);
 }
 
 const struct address_space_operations f2fs_dblock_aops = {
@@ -709,7 +1115,7 @@ const struct address_space_operations f2fs_dblock_aops = {
 	.writepage	= f2fs_write_data_page,
 	.writepages	= f2fs_write_data_pages,
 	.write_begin	= f2fs_write_begin,
-	.write_end	= nobh_write_end,
+	.write_end	= f2fs_write_end,
 	.set_page_dirty	= f2fs_set_data_page_dirty,
 	.invalidatepage	= f2fs_invalidate_data_page,
 	.releasepage	= f2fs_release_data_page,