[PATCH V3 11/21] ext4: let add_dir_entry handle inline data properly.

[Tao Ma]

From: Tao Ma <boyu.mt@xxxxxxxxxx>

This patch let add_dir_entry handle the inline data case. So the
dir is initialized as inline dir first and then we can try to add
some files to it, when the inline space can't hold all the entries,
a dir block will be created and the dir entry will be moved to it.

Signed-off-by: Tao Ma <boyu.mt@xxxxxxxxxx>
---
 fs/ext4/ext4.h   |    2 +
 fs/ext4/inline.c |  276 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/ext4/namei.c  |   19 ++++
 fs/ext4/xattr.h  |   19 ++++
 4 files changed, 316 insertions(+), 0 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index e5847b4..e61c967 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -2262,6 +2262,8 @@ static inline void ext4_mark_super_dirty(struct super_block *sb)
 
 /* dir.c */
 extern const struct file_operations ext4_dir_operations;
+extern void ext4_show_inline_dir(struct inode *dir, struct buffer_head *bh,
+				 void *inline_start, int inline_size);
 
 /* file.c */
 extern const struct inode_operations ext4_file_inode_operations;
diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c
index eb9018b..db70d3e 100644
--- a/fs/ext4/inline.c
+++ b/fs/ext4/inline.c
@@ -927,6 +927,282 @@ int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
 	return copied;
 }
 
+#ifdef INLINE_DIR_DEBUG
+void ext4_show_inline_dir(struct inode *dir, struct buffer_head *bh,
+			  void *inline_start, int inline_size)
+{
+	int offset;
+	unsigned short de_len;
+	struct ext4_dir_entry_2 *de = inline_start;
+	void *dlimit = inline_start + inline_size;
+
+	trace_printk("inode %lu\n", dir->i_ino);
+	offset = 0;
+	while ((void *)de < dlimit) {
+		de_len = ext4_rec_len_from_disk(de->rec_len, inline_size);
+		trace_printk("de: offset %u reclen %u name %*.s "
+			     "namelen %u ino %u\n",
+			     offset, de_len, de->name_len, de->name,
+			     de->name_len, le32_to_cpu(de->inode));
+		if (ext4_check_dir_entry(dir, NULL, de, bh,
+					 inline_start, inline_size, offset))
+			BUG();
+
+		offset += de_len;
+		de = (struct ext4_dir_entry_2 *) ((char *) de + de_len);
+	}
+}
+#else
+#define ext4_show_inline_dir(dir, bh, inline_start, inline_size)
+#endif
+
+/*
+ * Add a new entry into a inline dir.
+ * It will return -ENOSPC if no space is available, and -EIO
+ * and -EEXIST if directory entry already exists.
+ */
+static int ext4_add_dirent_to_inline(handle_t *handle,
+				     struct dentry *dentry,
+				     struct inode *inode,
+				     struct ext4_iloc *iloc,
+				     void *inline_start, int inline_size)
+{
+	struct inode	*dir = dentry->d_parent->d_inode;
+	const char	*name = dentry->d_name.name;
+	int		namelen = dentry->d_name.len;
+	unsigned short	reclen;
+	int		err;
+	struct ext4_dir_entry_2 *de;
+
+	reclen = EXT4_DIR_REC_LEN(namelen);
+	err = ext4_find_dest_de(dir, inode, iloc->bh,
+				inline_start, inline_size,
+				name, namelen, &de);
+	if (err)
+		return err;
+
+	err = ext4_journal_get_write_access(handle, iloc->bh);
+	if (err)
+		return err;
+	ext4_insert_dentry(inode, de, inline_size, name, namelen);
+
+	ext4_show_inline_dir(dir, iloc->bh, inline_start, inline_size);
+
+	/*
+	 * XXX shouldn't update any times until successful
+	 * completion of syscall, but too many callers depend
+	 * on this.
+	 *
+	 * XXX similarly, too many callers depend on
+	 * ext4_new_inode() setting the times, but error
+	 * recovery deletes the inode, so the worst that can
+	 * happen is that the times are slightly out of date
+	 * and/or different from the directory change time.
+	 */
+	dir->i_mtime = dir->i_ctime = ext4_current_time(dir);
+	ext4_update_dx_flag(dir);
+	dir->i_version++;
+	ext4_mark_inode_dirty(handle, dir);
+	return 1;
+}
+
+void* ext4_get_inline_xattr_pos(struct inode *inode, struct ext4_iloc *iloc)
+{
+	struct ext4_xattr_entry *entry;
+	struct ext4_xattr_ibody_header *header;
+
+	BUG_ON(!EXT4_I(inode)->i_inline_off);
+
+	header = IHDR(inode, ext4_raw_inode(iloc));
+	entry = (struct ext4_xattr_entry *)((void *)ext4_raw_inode(iloc) +
+					    EXT4_I(inode)->i_inline_off);
+
+	return (void *)IFIRST(header) + le16_to_cpu(entry->e_value_offs);
+}
+
+/* Set the final de to cover the whole block. */
+static void ext4_update_final_de(void *de_buf, int old_size, int new_size)
+{
+	struct ext4_dir_entry_2 *de, *prev_de;
+	void *limit;
+	int de_len;
+
+	de = (struct ext4_dir_entry_2 *)de_buf;
+	if (old_size) {
+		limit = de_buf + old_size;
+		do {
+			prev_de = de;
+			de_len = ext4_rec_len_from_disk(de->rec_len, old_size);
+			de_buf += de_len;
+			de = (struct ext4_dir_entry_2 *)de_buf;
+		} while (de_buf < limit);
+
+		prev_de->rec_len = ext4_rec_len_to_disk(de_len + new_size -
+							old_size, new_size);
+	} else {
+		/* this is just created, so create an empty entry. */
+		de->inode = 0;
+		de->rec_len = ext4_rec_len_to_disk(new_size, new_size);
+	}
+}
+
+static int ext4_update_inline_dir(handle_t *handle, struct dentry *dentry,
+				  struct inode *dir,
+				  struct ext4_iloc *iloc)
+{
+	int ret, reclen = EXT4_DIR_REC_LEN(dentry->d_name.len);
+	int old_size = EXT4_I(dir)->i_inline_size - EXT4_MIN_INLINE_DATA_SIZE;
+
+	ret = ext4_update_inline_data(handle, dir,
+				EXT4_I(dir)->i_inline_size + reclen);
+	if (ret)
+		return ret;
+
+	ext4_update_final_de(ext4_get_inline_xattr_pos(dir, iloc), old_size,
+			     EXT4_I(dir)->i_inline_size -
+						EXT4_MIN_INLINE_DATA_SIZE);
+	dir->i_size = EXT4_I(dir)->i_disksize = EXT4_I(dir)->i_inline_size;
+	return 0;
+}
+
+/*
+ * Try to add the new entry to the inline data.
+ * If succeeds, return 0. If not, extended the inline dir and copied data to
+ * the new created block.
+ */
+int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
+			      struct inode *inode)
+{
+	int ret, inline_size;
+	void *inline_start, *backup_buf = NULL;
+	struct buffer_head *dir_block = NULL;
+	struct ext4_iloc iloc;
+	int blocksize = inode->i_sb->s_blocksize;
+	struct inode *dir = dentry->d_parent->d_inode;
+
+	ret = ext4_get_inode_loc(dir, &iloc);
+	if (ret)
+		return ret;
+
+	down_write(&EXT4_I(dir)->xattr_sem);
+	if (!ext4_has_inline_data(dir))
+		goto out;
+
+	inline_start = ext4_raw_inode(&iloc)->i_block;
+	inline_size = EXT4_MIN_INLINE_DATA_SIZE;
+
+	ret = ext4_add_dirent_to_inline(handle, dentry, inode, &iloc,
+					inline_start, inline_size);
+	if (ret != -ENOSPC)
+		goto out;
+
+	/* check whether it can be inserted to inline xattr space. */
+	inline_size = EXT4_I(dir)->i_inline_size -
+			EXT4_MIN_INLINE_DATA_SIZE;
+	if (inline_size > 0) {
+		inline_start = ext4_get_inline_xattr_pos(dir, &iloc);
+
+		ret = ext4_add_dirent_to_inline(handle, dentry, inode, &iloc,
+						inline_start, inline_size);
+		if (ret != -ENOSPC)
+			goto out;
+	}
+
+	/* Try to add more xattr space.*/
+	ret = ext4_update_inline_dir(handle, dentry, dir, &iloc);
+	if (ret && ret != -ENOSPC)
+		goto out;
+	else if (!ret) {
+		inline_size = EXT4_I(dir)->i_inline_size -
+				EXT4_MIN_INLINE_DATA_SIZE;
+		inline_start = ext4_get_inline_xattr_pos(dir, &iloc);
+
+		ret = ext4_add_dirent_to_inline(handle, dentry, inode, &iloc,
+						inline_start, inline_size);
+		if (ret != -ENOSPC)
+			goto out;
+	}
+
+	/*
+	 * The inline space is filled up, so create a new block for it.
+	 * As the extent tree will be created, we have to save the inline
+	 * dir first.
+	 */
+	inline_size = EXT4_I(dir)->i_inline_size;
+	backup_buf = kmalloc(inline_size, GFP_NOFS);
+	if (!backup_buf) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	memcpy(backup_buf, (void *)ext4_raw_inode(&iloc)->i_block,
+	       EXT4_MIN_INLINE_DATA_SIZE);
+	if (inline_size > EXT4_MIN_INLINE_DATA_SIZE)
+		memcpy(backup_buf + EXT4_MIN_INLINE_DATA_SIZE,
+		       ext4_get_inline_xattr_pos(dir, &iloc),
+		       inline_size - EXT4_MIN_INLINE_DATA_SIZE);
+
+	/* clear the entry and the flag in dir now. */
+	ret = ext4_destroy_inline_data_nolock(handle, dir);
+	if (ret)
+		goto out;
+
+	dir->i_size = EXT4_I(dir)->i_disksize = blocksize;
+	dir_block = ext4_bread(handle, dir, 0, 1, &ret);
+	if (!dir_block)
+		goto out;
+
+	BUFFER_TRACE(dir_block, "get_write_access");
+	ret = ext4_journal_get_write_access(handle, dir_block);
+	if (ret)
+		goto out;
+	memcpy(dir_block->b_data, backup_buf, inline_size);
+
+	/* Set the final de to cover the whole block. */
+	ext4_update_final_de(dir_block->b_data, inline_size,
+			     blocksize);
+
+	BUFFER_TRACE(dir_block, "call ext4_handle_dirty_metadata");
+	ret = ext4_handle_dirty_metadata(handle, dir, dir_block);
+
+out:
+	kfree(backup_buf);
+	brelse(dir_block);
+	if (!ret || ret == 1)
+		ext4_mark_inode_dirty(handle, dir);
+	up_write(&EXT4_I(dir)->xattr_sem);
+	brelse(iloc.bh);
+	return ret;
+}
+
+/*
+ * Try to create the inline data for the new dir.
+ * If it succeeds, return 0, otherwise return the error.
+ * In case of ENOSPC, the caller should create the normal disk layout dir.
+ */
+int ext4_try_create_inline_dir(handle_t *handle, struct inode *parent,
+			       struct inode *inode)
+{
+	int ret, inline_size = EXT4_MIN_INLINE_DATA_SIZE;
+	struct ext4_iloc iloc;
+	struct ext4_dir_entry_2 *de;
+
+	ret = ext4_get_inode_loc(inode, &iloc);
+	if (ret)
+		return ret;
+
+	ret = ext4_prepare_inline_data(handle, inode, inline_size);
+	if (ret)
+		goto out;
+
+	de = (struct ext4_dir_entry_2 *)ext4_raw_inode(&iloc)->i_block;
+	ext4_init_dot_dotdot(parent, inode, de, inline_size);
+	inode->i_size = EXT4_I(inode)->i_disksize = inline_size;
+out:
+	brelse(iloc.bh);
+	return ret;
+}
+
 int ext4_destroy_inline_data(handle_t *handle, struct inode *inode)
 {
 	int ret;
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 21acbc9..77312dc 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -1486,6 +1486,17 @@ static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
 	blocksize = sb->s_blocksize;
 	if (!dentry->d_name.len)
 		return -EINVAL;
+
+	if (ext4_has_inline_data(dir)) {
+		retval = ext4_try_add_inline_entry(handle, dentry, inode);
+		if (retval < 0)
+			return retval;
+		if (retval == 1) {
+			retval = 0;
+			return retval;
+		}
+	}
+
 	if (is_dx(dir)) {
 		retval = ext4_dx_add_entry(handle, dentry, inode);
 		if (!retval || (retval != ERR_BAD_DX_DIR))
@@ -1862,6 +1873,14 @@ static int ext4_init_new_dir(handle_t *handle, struct inode *parent,
 	int err;
 	int blocksize = inode->i_sb->s_blocksize;
 
+	if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) {
+		err = ext4_try_create_inline_dir(handle, parent, inode);
+		if (err < 0 && err != ENOSPC)
+			goto out;
+		if (!err)
+			goto out;
+	}
+
 	inode->i_size = EXT4_I(inode)->i_disksize = blocksize;
 	dir_block = ext4_bread(handle, inode, 0, 1, &err);
 	if (!dir_block)
diff --git a/fs/ext4/xattr.h b/fs/ext4/xattr.h
index a0d43c7..0229a78 100644
--- a/fs/ext4/xattr.h
+++ b/fs/ext4/xattr.h
@@ -161,6 +161,11 @@ extern int ext4_da_write_inline_data_begin(struct address_space *mapping,
 extern int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
 					 unsigned len, unsigned copied,
 					 struct page *page);
+extern int ext4_try_add_inline_entry(handle_t *handle, struct dentry *dentry,
+				     struct inode *inode);
+extern int ext4_try_create_inline_dir(handle_t *handle,
+				      struct inode *parent,
+				      struct inode *inode);
 # else  /* CONFIG_EXT4_FS_XATTR */
 
 static inline int
@@ -324,6 +329,20 @@ static inline int ext4_da_write_inline_data_end(struct inode *inode, loff_t pos,
 {
 	return 0;
 }
+
+static inline int ext4_try_add_inline_entry(handle_t *handle,
+					    struct dentry *dentry,
+					    struct inode *inode)
+{
+	return 0;
+}
+
+static inline int ext4_try_create_inline_dir(handle_t *handle,
+					     struct inode *parent,
+					     struct inode *inode)
+{
+	return 0;
+}
 # endif  /* CONFIG_EXT4_FS_XATTR */
 
 #ifdef CONFIG_EXT4_FS_SECURITY
-- 
1.7.0.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/