ext4: dynamically allocate the jbd2_inode in ext4_inode_info as necessary

Replace the jbd2_inode structure (which is 48 bytes) with a pointer
and only allocate the jbd2_inode when it is needed --- that is, when
the file system has a journal present and the inode has been opened
for writing.  This allows us to further slim down the ext4_inode_info
structure.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>

5 files changed, 43 insertions, 16 deletions

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 2fb531cfd48..32b7daa41a4 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -811,7 +811,7 @@ struct ext4_inode_info {
 	 */
 	struct rw_semaphore i_data_sem;
 	struct inode vfs_inode;
-	struct jbd2_inode jinode;
+	struct jbd2_inode *jinode;
 
 	struct ext4_ext_cache i_cached_extent;
 	/*
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index b0bd792c58c..d8b992e658c 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -253,7 +253,7 @@ static inline int ext4_journal_force_commit(journal_t *journal)
 static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode)
 {
 	if (ext4_handle_valid(handle))
-		return jbd2_journal_file_inode(handle, &EXT4_I(inode)->jinode);
+		return jbd2_journal_file_inode(handle, EXT4_I(inode)->jinode);
 	return 0;
 }
 
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 5a5c55ddcee..bb003dc9fff 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -104,6 +104,7 @@ static int ext4_file_open(struct inode * inode, struct file * filp)
 {
 	struct super_block *sb = inode->i_sb;
 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
+	struct ext4_inode_info *ei = EXT4_I(inode);
 	struct vfsmount *mnt = filp->f_path.mnt;
 	struct path path;
 	char buf[64], *cp;
@@ -127,6 +128,27 @@ static int ext4_file_open(struct inode * inode, struct file * filp)
 			ext4_mark_super_dirty(sb);
 		}
 	}
+	/*
+	 * Set up the jbd2_inode if we are opening the inode for
+	 * writing and the journal is present
+	 */
+	if (sbi->s_journal && !ei->jinode && (filp->f_mode & FMODE_WRITE)) {
+		struct jbd2_inode *jinode = jbd2_alloc_inode(GFP_KERNEL);
+
+		spin_lock(&inode->i_lock);
+		if (!ei->jinode) {
+			if (!jinode) {
+				spin_unlock(&inode->i_lock);
+				return -ENOMEM;
+			}
+			ei->jinode = jinode;
+			jbd2_journal_init_jbd_inode(ei->jinode, inode);
+			jinode = NULL;
+		}
+		spin_unlock(&inode->i_lock);
+		if (unlikely(jinode != NULL))
+			jbd2_free_inode(jinode);
+	}
 	return dquot_file_open(inode, filp);
 }
 
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 0801ee6a173..2693fcda30d 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -55,10 +55,17 @@ static inline int ext4_begin_ordered_truncate(struct inode *inode,
 					      loff_t new_size)
 {
 	trace_ext4_begin_ordered_truncate(inode, new_size);
-	return jbd2_journal_begin_ordered_truncate(
-					EXT4_SB(inode->i_sb)->s_journal,
-					&EXT4_I(inode)->jinode,
-					new_size);
+	/*
+	 * If jinode is zero, then we never opened the file for
+	 * writing, so there's no need to call
+	 * jbd2_journal_begin_ordered_truncate() since there's no
+	 * outstanding writes we need to flush.
+	 */
+	if (!EXT4_I(inode)->jinode)
+		return 0;
+	return jbd2_journal_begin_ordered_truncate(EXT4_JOURNAL(inode),
+						   EXT4_I(inode)->jinode,
+						   new_size);
 }
 
 static void ext4_invalidatepage(struct page *page, unsigned long offset);
@@ -4054,7 +4061,7 @@ int ext4_block_truncate_page(handle_t *handle,
 	if (ext4_should_journal_data(inode)) {
 		err = ext4_handle_dirty_metadata(handle, inode, bh);
 	} else {
-		if (ext4_should_order_data(inode))
+		if (ext4_should_order_data(inode) && EXT4_I(inode)->jinode)
 			err = ext4_jbd2_file_inode(handle, inode);
 		mark_buffer_dirty(bh);
 	}
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index f5960d673e4..1cd4326c530 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -818,12 +818,6 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
 	memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
 	INIT_LIST_HEAD(&ei->i_prealloc_list);
 	spin_lock_init(&ei->i_prealloc_lock);
-	/*
-	 * Note:  We can be called before EXT4_SB(sb)->s_journal is set,
-	 * therefore it can be null here.  Don't check it, just initialize
-	 * jinode.
-	 */
-	jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
 	ei->i_reserved_data_blocks = 0;
 	ei->i_reserved_meta_blocks = 0;
 	ei->i_allocated_meta_blocks = 0;
@@ -832,6 +826,7 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
 #ifdef CONFIG_QUOTA
 	ei->i_reserved_quota = 0;
 #endif
+	ei->jinode = NULL;
 	INIT_LIST_HEAD(&ei->i_completed_io_list);
 	spin_lock_init(&ei->i_completed_io_lock);
 	ei->cur_aio_dio = NULL;
@@ -900,9 +895,12 @@ void ext4_clear_inode(struct inode *inode)
 	end_writeback(inode);
 	dquot_drop(inode);
 	ext4_discard_preallocations(inode);
-	if (EXT4_JOURNAL(inode))
-		jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
-				       &EXT4_I(inode)->jinode);
+	if (EXT4_I(inode)->jinode) {
+		jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
+					       EXT4_I(inode)->jinode);
+		jbd2_free_inode(EXT4_I(inode)->jinode);
+		EXT4_I(inode)->jinode = NULL;
+	}
 }
 
 static inline void ext4_show_quota_options(struct seq_file *seq,