1 files changed, 221 insertions, 143 deletions

diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
index 8e58c4cc2cb9..985ed023a750 100644
--- a/fs/f2fs/data.c
+++ b/fs/f2fs/data.c
@@ -22,6 +22,7 @@
 #include "f2fs.h"
 #include "node.h"
 #include "segment.h"
+#include "trace.h"
 #include <trace/events/f2fs.h>
 
 static void f2fs_read_end_io(struct bio *bio, int err)
@@ -61,11 +62,6 @@ static void f2fs_write_end_io(struct bio *bio, int err)
 		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);
@@ -95,34 +91,16 @@ static struct bio *__bio_alloc(struct f2fs_sb_info *sbi, block_t blk_addr,
 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);
-		}
-	}
+	if (is_read_io(fio->rw))
+		trace_f2fs_submit_read_bio(io->sbi->sb, fio, io->bio);
+	else
+		trace_f2fs_submit_write_bio(io->sbi->sb, fio, io->bio);
 
+	submit_bio(fio->rw, io->bio);
 	io->bio = NULL;
 }
 
@@ -153,14 +131,15 @@ void f2fs_submit_merged_bio(struct f2fs_sb_info *sbi,
  * Return unlocked page.
  */
 int f2fs_submit_page_bio(struct f2fs_sb_info *sbi, struct page *page,
-					block_t blk_addr, int rw)
+					struct f2fs_io_info *fio)
 {
 	struct bio *bio;
 
-	trace_f2fs_submit_page_bio(page, blk_addr, rw);
+	trace_f2fs_submit_page_bio(page, fio);
+	f2fs_trace_ios(page, fio, 0);
 
 	/* Allocate a new bio */
-	bio = __bio_alloc(sbi, blk_addr, 1, is_read_io(rw));
+	bio = __bio_alloc(sbi, fio->blk_addr, 1, is_read_io(fio->rw));
 
 	if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) {
 		bio_put(bio);
@@ -168,12 +147,12 @@ int f2fs_submit_page_bio(struct f2fs_sb_info *sbi, struct page *page,
 		return -EFAULT;
 	}
 
-	submit_bio(rw, bio);
+	submit_bio(fio->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)
+					struct f2fs_io_info *fio)
 {
 	enum page_type btype = PAGE_TYPE_OF_BIO(fio->type);
 	struct f2fs_bio_info *io;
@@ -181,21 +160,21 @@ void f2fs_submit_page_mbio(struct f2fs_sb_info *sbi, struct page *page,
 
 	io = is_read ? &sbi->read_io : &sbi->write_io[btype];
 
-	verify_block_addr(sbi, blk_addr);
+	verify_block_addr(sbi, fio->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 ||
+	if (io->bio && (io->last_block_in_bio != fio->blk_addr - 1 ||
 						io->fio.rw != fio->rw))
 		__submit_merged_bio(io);
 alloc_new:
 	if (io->bio == NULL) {
 		int bio_blocks = MAX_BIO_BLOCKS(sbi);
 
-		io->bio = __bio_alloc(sbi, blk_addr, bio_blocks, is_read);
+		io->bio = __bio_alloc(sbi, fio->blk_addr, bio_blocks, is_read);
 		io->fio = *fio;
 	}
 
@@ -205,10 +184,11 @@ alloc_new:
 		goto alloc_new;
 	}
 
-	io->last_block_in_bio = blk_addr;
+	io->last_block_in_bio = fio->blk_addr;
+	f2fs_trace_ios(page, fio, 0);
 
 	up_write(&io->io_rwsem);
-	trace_f2fs_submit_page_mbio(page, fio->rw, fio->type, blk_addr);
+	trace_f2fs_submit_page_mbio(page, fio);
 }
 
 /*
@@ -217,7 +197,7 @@ alloc_new:
  *  ->node_page
  *    update block addresses in the node page
  */
-static void __set_data_blkaddr(struct dnode_of_data *dn, block_t new_addr)
+static void __set_data_blkaddr(struct dnode_of_data *dn)
 {
 	struct f2fs_node *rn;
 	__le32 *addr_array;
@@ -230,7 +210,7 @@ static void __set_data_blkaddr(struct dnode_of_data *dn, block_t new_addr)
 
 	/* Get physical address of data block */
 	addr_array = blkaddr_in_node(rn);
-	addr_array[ofs_in_node] = cpu_to_le32(new_addr);
+	addr_array[ofs_in_node] = cpu_to_le32(dn->data_blkaddr);
 	set_page_dirty(node_page);
 }
 
@@ -245,8 +225,8 @@ int reserve_new_block(struct dnode_of_data *dn)
 
 	trace_f2fs_reserve_new_block(dn->inode, dn->nid, dn->ofs_in_node);
 
-	__set_data_blkaddr(dn, NEW_ADDR);
 	dn->data_blkaddr = NEW_ADDR;
+	__set_data_blkaddr(dn);
 	mark_inode_dirty(dn->inode);
 	sync_inode_page(dn);
 	return 0;
@@ -257,9 +237,6 @@ 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(F2FS_I_SB(dn->inode), !need_put && index);
-
 	err = get_dnode_of_data(dn, index, ALLOC_NODE);
 	if (err)
 		return err;
@@ -297,7 +274,7 @@ static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
 		unsigned int blkbits = inode->i_sb->s_blocksize_bits;
 		size_t count;
 
-		clear_buffer_new(bh_result);
+		set_buffer_new(bh_result);
 		map_bh(bh_result, inode->i_sb,
 				start_blkaddr + pgofs - start_fofs);
 		count = end_fofs - pgofs + 1;
@@ -314,23 +291,24 @@ static int check_extent_cache(struct inode *inode, pgoff_t pgofs,
 	return 0;
 }
 
-void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
+void update_extent_cache(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;
 
-	f2fs_bug_on(F2FS_I_SB(dn->inode), blk_addr == NEW_ADDR);
-	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
-							dn->ofs_in_node;
+	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, blk_addr);
+	__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);
 
 	start_fofs = fi->ext.fofs;
@@ -344,16 +322,16 @@ void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
 
 	/* Initial extent */
 	if (fi->ext.len == 0) {
-		if (blk_addr != NULL_ADDR) {
+		if (dn->data_blkaddr != NULL_ADDR) {
 			fi->ext.fofs = fofs;
-			fi->ext.blk_addr = blk_addr;
+			fi->ext.blk_addr = dn->data_blkaddr;
 			fi->ext.len = 1;
 		}
 		goto end_update;
 	}
 
 	/* Front merge */
-	if (fofs == start_fofs - 1 && blk_addr == start_blkaddr - 1) {
+	if (fofs == start_fofs - 1 && dn->data_blkaddr == start_blkaddr - 1) {
 		fi->ext.fofs--;
 		fi->ext.blk_addr--;
 		fi->ext.len++;
@@ -361,7 +339,7 @@ void update_extent_cache(block_t blk_addr, struct dnode_of_data *dn)
 	}
 
 	/* Back merge */
-	if (fofs == end_fofs + 1 && blk_addr == end_blkaddr + 1) {
+	if (fofs == end_fofs + 1 && dn->data_blkaddr == end_blkaddr + 1) {
 		fi->ext.len++;
 		goto end_update;
 	}
@@ -400,6 +378,10 @@ struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync)
 	struct dnode_of_data dn;
 	struct page *page;
 	int err;
+	struct f2fs_io_info fio = {
+		.type = DATA,
+		.rw = sync ? READ_SYNC : READA,
+	};
 
 	page = find_get_page(mapping, index);
 	if (page && PageUptodate(page))
@@ -428,8 +410,8 @@ struct page *find_data_page(struct inode *inode, pgoff_t index, bool sync)
 		return page;
 	}
 
-	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, dn.data_blkaddr,
-					sync ? READ_SYNC : READA);
+	fio.blk_addr = dn.data_blkaddr;
+	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, &fio);
 	if (err)
 		return ERR_PTR(err);
 
@@ -454,7 +436,10 @@ struct page *get_lock_data_page(struct inode *inode, pgoff_t index)
 	struct dnode_of_data dn;
 	struct page *page;
 	int err;
-
+	struct f2fs_io_info fio = {
+		.type = DATA,
+		.rw = READ_SYNC,
+	};
 repeat:
 	page = grab_cache_page(mapping, index);
 	if (!page)
@@ -488,8 +473,8 @@ repeat:
 		return page;
 	}
 
-	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page,
-					dn.data_blkaddr, READ_SYNC);
+	fio.blk_addr = dn.data_blkaddr;
+	err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, &fio);
 	if (err)
 		return ERR_PTR(err);
 
@@ -539,8 +524,12 @@ repeat:
 		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
 		SetPageUptodate(page);
 	} else {
-		err = f2fs_submit_page_bio(F2FS_I_SB(inode), page,
-						dn.data_blkaddr, READ_SYNC);
+		struct f2fs_io_info fio = {
+			.type = DATA,
+			.rw = READ_SYNC,
+			.blk_addr = dn.data_blkaddr,
+		};
+		err = f2fs_submit_page_bio(F2FS_I_SB(inode), page, &fio);
 		if (err)
 			goto put_err;
 
@@ -574,30 +563,25 @@ static int __allocate_data_block(struct dnode_of_data *dn)
 	struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
 	struct f2fs_inode_info *fi = F2FS_I(dn->inode);
 	struct f2fs_summary sum;
-	block_t new_blkaddr;
 	struct node_info ni;
+	int seg = CURSEG_WARM_DATA;
 	pgoff_t fofs;
-	int type;
 
 	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;
 
-	__set_data_blkaddr(dn, NEW_ADDR);
-	dn->data_blkaddr = NEW_ADDR;
-
 	get_node_info(sbi, dn->nid, &ni);
 	set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
 
-	type = CURSEG_WARM_DATA;
+	if (dn->ofs_in_node == 0 && dn->inode_page == dn->node_page)
+		seg = CURSEG_DIRECT_IO;
 
-	allocate_data_block(sbi, NULL, NULL_ADDR, &new_blkaddr, &sum, type);
+	allocate_data_block(sbi, NULL, NULL_ADDR, &dn->data_blkaddr, &sum, seg);
 
 	/* 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);
+	__set_data_blkaddr(dn);
 
 	/* update i_size */
 	fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) +
@@ -605,10 +589,59 @@ static int __allocate_data_block(struct dnode_of_data *dn)
 	if (i_size_read(dn->inode) < ((fofs + 1) << PAGE_CACHE_SHIFT))
 		i_size_write(dn->inode, ((fofs + 1) << PAGE_CACHE_SHIFT));
 
-	dn->data_blkaddr = new_blkaddr;
 	return 0;
 }
 
+static void __allocate_data_blocks(struct inode *inode, loff_t offset,
+							size_t count)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	struct dnode_of_data dn;
+	u64 start = F2FS_BYTES_TO_BLK(offset);
+	u64 len = F2FS_BYTES_TO_BLK(count);
+	bool allocated;
+	u64 end_offset;
+
+	while (len) {
+		f2fs_balance_fs(sbi);
+		f2fs_lock_op(sbi);
+
+		/* When reading holes, we need its node page */
+		set_new_dnode(&dn, inode, NULL, NULL, 0);
+		if (get_dnode_of_data(&dn, start, ALLOC_NODE))
+			goto out;
+
+		allocated = false;
+		end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
+
+		while (dn.ofs_in_node < end_offset && len) {
+			if (dn.data_blkaddr == NULL_ADDR) {
+				if (__allocate_data_block(&dn))
+					goto sync_out;
+				allocated = true;
+			}
+			len--;
+			start++;
+			dn.ofs_in_node++;
+		}
+
+		if (allocated)
+			sync_inode_page(&dn);
+
+		f2fs_put_dnode(&dn);
+		f2fs_unlock_op(sbi);
+	}
+	return;
+
+sync_out:
+	if (allocated)
+		sync_inode_page(&dn);
+	f2fs_put_dnode(&dn);
+out:
+	f2fs_unlock_op(sbi);
+	return;
+}
+
 /*
  * get_data_block() now supported readahead/bmap/rw direct_IO with mapped bh.
  * If original data blocks are allocated, then give them to blockdev.
@@ -634,10 +667,8 @@ static int __get_data_block(struct inode *inode, sector_t iblock,
 	if (check_extent_cache(inode, pgofs, bh_result))
 		goto out;
 
-	if (create) {
-		f2fs_balance_fs(F2FS_I_SB(inode));
+	if (create)
 		f2fs_lock_op(F2FS_I_SB(inode));
-	}
 
 	/* When reading holes, we need its node page */
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
@@ -651,12 +682,14 @@ static int __get_data_block(struct inode *inode, sector_t iblock,
 		goto put_out;
 
 	if (dn.data_blkaddr != NULL_ADDR) {
+		set_buffer_new(bh_result);
 		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;
+		set_buffer_new(bh_result);
 		map_bh(bh_result, inode->i_sb, dn.data_blkaddr);
 	} else {
 		goto put_out;
@@ -740,14 +773,14 @@ int f2fs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 static int f2fs_read_data_page(struct file *file, struct page *page)
 {
 	struct inode *inode = page->mapping->host;
-	int ret;
+	int ret = -EAGAIN;
 
 	trace_f2fs_readpage(page, DATA);
 
 	/* If the file has inline data, try to read it directly */
 	if (f2fs_has_inline_data(inode))
 		ret = f2fs_read_inline_data(inode, page);
-	else
+	if (ret == -EAGAIN)
 		ret = mpage_readpage(page, get_data_block);
 
 	return ret;
@@ -769,7 +802,6 @@ static int f2fs_read_data_pages(struct file *file,
 int do_write_data_page(struct page *page, struct f2fs_io_info *fio)
 {
 	struct inode *inode = page->mapping->host;
-	block_t old_blkaddr, new_blkaddr;
 	struct dnode_of_data dn;
 	int err = 0;
 
@@ -778,10 +810,10 @@ int do_write_data_page(struct page *page, struct f2fs_io_info *fio)
 	if (err)
 		return err;
 
-	old_blkaddr = dn.data_blkaddr;
+	fio->blk_addr = dn.data_blkaddr;
 
 	/* This page is already truncated */
-	if (old_blkaddr == NULL_ADDR)
+	if (fio->blk_addr == NULL_ADDR)
 		goto out_writepage;
 
 	set_page_writeback(page);
@@ -790,14 +822,14 @@ int do_write_data_page(struct page *page, struct f2fs_io_info *fio)
 	 * If current allocation needs SSR,
 	 * it had better in-place writes for updated data.
 	 */
-	if (unlikely(old_blkaddr != NEW_ADDR &&
+	if (unlikely(fio->blk_addr != NEW_ADDR &&
 			!is_cold_data(page) &&
 			need_inplace_update(inode))) {
-		rewrite_data_page(page, old_blkaddr, fio);
+		rewrite_data_page(page, 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);
+		write_data_page(page, &dn, fio);
+		update_extent_cache(&dn);
 		set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE);
 	}
 out_writepage:
@@ -836,7 +868,12 @@ static int f2fs_write_data_page(struct page *page,
 
 	zero_user_segment(page, offset, PAGE_CACHE_SIZE);
 write:
-	if (unlikely(sbi->por_doing))
+	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
+		goto redirty_out;
+	if (f2fs_is_drop_cache(inode))
+		goto out;
+	if (f2fs_is_volatile_file(inode) && !wbc->for_reclaim &&
+			available_free_memory(sbi, BASE_CHECK))
 		goto redirty_out;
 
 	/* Dentry blocks are controlled by checkpoint */
@@ -850,7 +887,6 @@ write:
 	/* we should bypass data pages to proceed the kworkder jobs */
 	if (unlikely(f2fs_cp_error(sbi))) {
 		SetPageError(page);
-		unlock_page(page);
 		goto out;
 	}
 
@@ -859,10 +895,11 @@ write:
 	else if (has_not_enough_free_secs(sbi, 0))
 		goto redirty_out;
 
+	err = -EAGAIN;
 	f2fs_lock_op(sbi);
-	if (f2fs_has_inline_data(inode) || f2fs_may_inline(inode))
-		err = f2fs_write_inline_data(inode, page, offset);
-	else
+	if (f2fs_has_inline_data(inode))
+		err = f2fs_write_inline_data(inode, page);
+	if (err == -EAGAIN)
 		err = do_write_data_page(page, &fio);
 	f2fs_unlock_op(sbi);
 done:
@@ -951,7 +988,7 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 {
 	struct inode *inode = mapping->host;
 	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
-	struct page *page;
+	struct page *page, *ipage;
 	pgoff_t index = ((unsigned long long) pos) >> PAGE_CACHE_SHIFT;
 	struct dnode_of_data dn;
 	int err = 0;
@@ -959,45 +996,60 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping,
 	trace_f2fs_write_begin(inode, pos, len, flags);
 
 	f2fs_balance_fs(sbi);
-repeat:
-	err = f2fs_convert_inline_data(inode, pos + len, NULL);
-	if (err)
-		goto fail;
 
+	/*
+	 * We should check this at this moment to avoid deadlock on inode page
+	 * and #0 page. The locking rule for inline_data conversion should be:
+	 * lock_page(page #0) -> lock_page(inode_page)
+	 */
+	if (index != 0) {
+		err = f2fs_convert_inline_inode(inode);
+		if (err)
+			goto fail;
+	}
+repeat:
 	page = grab_cache_page_write_begin(mapping, index, flags);
 	if (!page) {
 		err = -ENOMEM;
 		goto fail;
 	}
 
-	/* to avoid latency during memory pressure */
-	unlock_page(page);
-
 	*pagep = page;
 
-	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 = f2fs_reserve_block(&dn, index);
-	f2fs_unlock_op(sbi);
-	if (err) {
-		f2fs_put_page(page, 0);
-		goto fail;
-	}
-inline_data:
-	lock_page(page);
-	if (unlikely(page->mapping != mapping)) {
-		f2fs_put_page(page, 1);
-		goto repeat;
+
+	/* check inline_data */
+	ipage = get_node_page(sbi, inode->i_ino);
+	if (IS_ERR(ipage)) {
+		err = PTR_ERR(ipage);
+		goto unlock_fail;
 	}
 
-	f2fs_wait_on_page_writeback(page, DATA);
+	set_new_dnode(&dn, inode, ipage, ipage, 0);
+
+	if (f2fs_has_inline_data(inode)) {
+		if (pos + len <= MAX_INLINE_DATA) {
+			read_inline_data(page, ipage);
+			set_inode_flag(F2FS_I(inode), FI_DATA_EXIST);
+			sync_inode_page(&dn);
+			goto put_next;
+		}
+		err = f2fs_convert_inline_page(&dn, page);
+		if (err)
+			goto put_fail;
+	}
+	err = f2fs_reserve_block(&dn, index);
+	if (err)
+		goto put_fail;
+put_next:
+	f2fs_put_dnode(&dn);
+	f2fs_unlock_op(sbi);
 
 	if ((len == PAGE_CACHE_SIZE) || PageUptodate(page))
 		return 0;
 
+	f2fs_wait_on_page_writeback(page, DATA);
+
 	if ((pos & PAGE_CACHE_MASK) >= i_size_read(inode)) {
 		unsigned start = pos & (PAGE_CACHE_SIZE - 1);
 		unsigned end = start + len;
@@ -1010,18 +1062,14 @@ inline_data:
 	if (dn.data_blkaddr == NEW_ADDR) {
 		zero_user_segment(page, 0, PAGE_CACHE_SIZE);
 	} else {
-		if (f2fs_has_inline_data(inode)) {
-			err = f2fs_read_inline_data(inode, page);
-			if (err) {
-				page_cache_release(page);
-				goto fail;
-			}
-		} else {
-			err = f2fs_submit_page_bio(sbi, page, dn.data_blkaddr,
-							READ_SYNC);
-			if (err)
-				goto fail;
-		}
+		struct f2fs_io_info fio = {
+			.type = DATA,
+			.rw = READ_SYNC,
+			.blk_addr = dn.data_blkaddr,
+		};
+		err = f2fs_submit_page_bio(sbi, page, &fio);
+		if (err)
+			goto fail;
 
 		lock_page(page);
 		if (unlikely(!PageUptodate(page))) {
@@ -1038,6 +1086,12 @@ out:
 	SetPageUptodate(page);
 	clear_cold_data(page);
 	return 0;
+
+put_fail:
+	f2fs_put_dnode(&dn);
+unlock_fail:
+	f2fs_unlock_op(sbi);
+	f2fs_put_page(page, 1);
 fail:
 	f2fs_write_failed(mapping, pos + len);
 	return err;
@@ -1052,10 +1106,7 @@ static int f2fs_write_end(struct file *file,
 
 	trace_f2fs_write_end(inode, pos, len, copied);
 
-	if (f2fs_is_atomic_file(inode) || f2fs_is_volatile_file(inode))
-		register_inmem_page(inode, page);
-	else
-		set_page_dirty(page);
+	set_page_dirty(page);
 
 	if (pos + copied > i_size_read(inode)) {
 		i_size_write(inode, pos + copied);
@@ -1093,15 +1144,21 @@ static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb,
 	size_t count = iov_iter_count(iter);
 	int err;
 
-	/* Let buffer I/O handle the inline data case. */
-	if (f2fs_has_inline_data(inode))
-		return 0;
+	/* we don't need to use inline_data strictly */
+	if (f2fs_has_inline_data(inode)) {
+		err = f2fs_convert_inline_inode(inode);
+		if (err)
+			return err;
+	}
 
 	if (check_direct_IO(inode, rw, iter, offset))
 		return 0;
 
 	trace_f2fs_direct_IO_enter(inode, offset, count, rw);
 
+	if (rw & WRITE)
+		__allocate_data_blocks(inode, offset, count);
+
 	err = blockdev_direct_IO(rw, iocb, inode, iter, offset, get_data_block);
 	if (err < 0 && (rw & WRITE))
 		f2fs_write_failed(mapping, offset + count);
@@ -1111,21 +1168,33 @@ static ssize_t f2fs_direct_IO(int rw, struct kiocb *iocb,
 	return err;
 }
 
-static void f2fs_invalidate_data_page(struct page *page, unsigned int offset,
-				      unsigned int length)
+void f2fs_invalidate_page(struct page *page, unsigned int offset,
+							unsigned int length)
 {
 	struct inode *inode = page->mapping->host;
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
 
-	if (offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE)
+	if (inode->i_ino >= F2FS_ROOT_INO(sbi) &&
+		(offset % PAGE_CACHE_SIZE || length != PAGE_CACHE_SIZE))
 		return;
 
-	if (PageDirty(page))
-		inode_dec_dirty_pages(inode);
+	if (PageDirty(page)) {
+		if (inode->i_ino == F2FS_META_INO(sbi))
+			dec_page_count(sbi, F2FS_DIRTY_META);
+		else if (inode->i_ino == F2FS_NODE_INO(sbi))
+			dec_page_count(sbi, F2FS_DIRTY_NODES);
+		else
+			inode_dec_dirty_pages(inode);
+	}
 	ClearPagePrivate(page);
 }
 
-static int f2fs_release_data_page(struct page *page, gfp_t wait)
+int f2fs_release_page(struct page *page, gfp_t wait)
 {
+	/* If this is dirty page, keep PagePrivate */
+	if (PageDirty(page))
+		return 0;
+
 	ClearPagePrivate(page);
 	return 1;
 }
@@ -1138,6 +1207,12 @@ static int f2fs_set_data_page_dirty(struct page *page)
 	trace_f2fs_set_page_dirty(page, DATA);
 
 	SetPageUptodate(page);
+
+	if (f2fs_is_atomic_file(inode)) {
+		register_inmem_page(inode, page);
+		return 1;
+	}
+
 	mark_inode_dirty(inode);
 
 	if (!PageDirty(page)) {
@@ -1152,9 +1227,12 @@ static sector_t f2fs_bmap(struct address_space *mapping, sector_t block)
 {
 	struct inode *inode = mapping->host;
 
-	if (f2fs_has_inline_data(inode))
-		return 0;
-
+	/* we don't need to use inline_data strictly */
+	if (f2fs_has_inline_data(inode)) {
+		int err = f2fs_convert_inline_inode(inode);
+		if (err)
+			return err;
+	}
 	return generic_block_bmap(mapping, block, get_data_block);
 }
 
@@ -1166,8 +1244,8 @@ const struct address_space_operations f2fs_dblock_aops = {
 	.write_begin	= f2fs_write_begin,
 	.write_end	= f2fs_write_end,
 	.set_page_dirty	= f2fs_set_data_page_dirty,
-	.invalidatepage	= f2fs_invalidate_data_page,
-	.releasepage	= f2fs_release_data_page,
+	.invalidatepage	= f2fs_invalidate_page,
+	.releasepage	= f2fs_release_page,
 	.direct_IO	= f2fs_direct_IO,
 	.bmap		= f2fs_bmap,
 };