Patchwork [V7,03/23] ext4: Add the basic function for inline data support.

login
register
mail settings
Submitter Tao Ma
Date Oct. 24, 2012, 2:55 a.m.
Message ID <1351047338-4963-3-git-send-email-tm@tao.ma>
Download mbox | patch
Permalink /patch/193617/
State Accepted, archived
Headers show

Comments

Tao Ma - Oct. 24, 2012, 2:55 a.m.
From: Tao Ma <boyu.mt@taobao.com>

Implement inline data with xattr.
Now we use "system.data" to store xattr, and the xattr will
be extended if the i_size is increased while we don't release
the space during truncate.

Signed-off-by: Tao Ma <boyu.mt@taobao.com>
---
 fs/ext4/Makefile |    2 +-
 fs/ext4/ext4.h   |    9 +-
 fs/ext4/inline.c |  461 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 fs/ext4/inode.c  |    5 +-
 fs/ext4/xattr.h  |   54 +++++++
 5 files changed, 528 insertions(+), 3 deletions(-)
 create mode 100644 fs/ext4/inline.c
Theodore Ts'o - Dec. 3, 2012, 1:48 a.m.
On Wed, Oct 24, 2012 at 10:55:18AM +0800, Tao Ma wrote:
> +	EXT4_STATE_MAY_INLINE_DATA,	/* may have in-inode data */

Can you write a paragraph or two about exactly what the semantics are
of this state flag --- what it means, when it should be set, and when
it should be cleared, etc.?

I'm not entirely sure I understand why you test
EXT4_STATE_MAY_INLINE_DATA versus simply calling
ext4_has_inline_data() in various places.

Thanks!!

					- Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Tao Ma - Dec. 3, 2012, 5:23 a.m.
On 12/03/2012 09:48 AM, Theodore Ts'o wrote:
> On Wed, Oct 24, 2012 at 10:55:18AM +0800, Tao Ma wrote:
>> +	EXT4_STATE_MAY_INLINE_DATA,	/* may have in-inode data */
> 
> Can you write a paragraph or two about exactly what the semantics are
> of this state flag --- what it means, when it should be set, and when
> it should be cleared, etc.?
> 
> I'm not entirely sure I understand why you test
> EXT4_STATE_MAY_INLINE_DATA versus simply calling
> ext4_has_inline_data() in various places.
OK, so in general, EXT4_STATE_MAY_INLINE_DATA is a flag that show that
the file has a potential to be inlined, but there is no guarantee it is
inlined.

So when the inode is created and the volume has inline-data support,
this flag is set. And when the write_begin comes, we have no idea
whether it is OK or not to create an inline file, so if this flag is
set, we just try to call the inline-data stuff to see whether it works.
If it succeeds, that is cool. But if not, this flag is cleared and we
will not trap to inline-data stuff for the future write_begin. The
reason why ext4_has_inline_data() doesn't work here is that a new file
has no inline data but has the potential to be inlined.

Thanks
Tao
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Tao Ma - Dec. 3, 2012, 4:17 p.m.
Hi Ted,
	As in the concall, it seems that you still have some questions about
this flag, so in general if a file has inline data,
EXT4_STAT_MAY_INLINE_DATA is definitely set(in delalloc, it is used as
another indication, will be explained later) while if a file has this
flag, it can be inlined or not. And if an inode doesn't have this flag,
it is definitely not an inline file.

I will explain one by one with the result I get from 'grep'.
grep -n EXT4_STATE_MAY_INLINE_DATA fs/ext4/*.c

fs/ext4/ialloc.c:907:		ext4_set_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
This flag is set when we init a new inode when inline data is enabled.

fs/ext4/inline.c:155:		ext4_set_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
Set when we found inline data from an inode disk read. It is OK since an
inline data file should have this flag.

fs/ext4/inline.c:287:					       EXT4_STATE_MAY_INLINE_DATA);
If we can't create the inline data for this file, clear this flag.

fs/ext4/inline.c:360:	ext4_set_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
set this flag when we update the inline data.

fs/ext4/inline.c:376:	if (!ext4_test_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA))
In ext4_prepare_inline_data, we check this flag and if the flag is
cleared, it isn't an inline one. If it is set, go on to see whether we
need to create or update it. We can't check ext4_has_inline_data here
since the inline data region may not be initialized yet.

fs/ext4/inline.c:447:	ext4_clear_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
Inline data is destroyed, so clear it.

fs/ext4/inline.c:532:		ext4_clear_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
In conversion, if this file hasn't be inlined, just clear the flag, so
that we will treat it in the normal way.

fs/ext4/inline.c:794:		ext4_clear_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
The same conversion as above, but happens in the da path.

fs/ext4/inline.c:815:	ext4_clear_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
end of da conversion. This is *tricky*. Since this is delayed
allocation, we don't want to touch the on-disk format now. Just clear
this flag so that:
1. da_write_end will know that we should handle this like normal
file(work with fs/ext4/inode.c:2713)
2. future da_write_begin will try the normal way since this flag is
cleared(work with fs/ext4/inode.c:2574:)
3. finally in da_writepages, when we see this flag is cleared while
ext4_has_inline_data is set, destroy the inline data and prepare for the
real block allocation(work with fs/ext4/inode.c:2242)

fs/ext4/inline.c:1096:	ext4_set_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
inline data is re-created, so set the flag.

fs/ext4/inline.c:1850:		ext4_clear_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA);
fallocate conversion. The same as other conversion.

fs/ext4/inode.c:906:	if (ext4_test_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA)) {
In ext4_write_begin, we have no idea of whether this can be inlined or
not, so just check this flag. And if we have the possibility(the flag
hasn't be cleared), go on the inline data path.

fs/ext4/inode.c:2242:						EXT4_STATE_MAY_INLINE_DATA));
Work with fs/ext4/inline.c:794.

fs/ext4/inode.c:2574:	if (ext4_test_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA)) {
In ext4_da_write_begin, we have no idea of whether this can be inlined
or not, so just check this flag. And if we have the possibility(the flag
hasn't be cleared), go on the inline data path.

fs/ext4/inode.c:2713:	    ext4_test_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA) &&
Work with fs/ext4/inline.c:794.

fs/ext4/namei.c:2343:	if (ext4_test_inode_state(inode,
EXT4_STATE_MAY_INLINE_DATA)) {
In dir init.

Hope I have explained all of these clearly. If you still have something
unclear, please point it out.

Thanks
Tao

On 12/03/2012 09:48 AM, Theodore Ts'o wrote:
> On Wed, Oct 24, 2012 at 10:55:18AM +0800, Tao Ma wrote:
>> +	EXT4_STATE_MAY_INLINE_DATA,	/* may have in-inode data */
> 
> Can you write a paragraph or two about exactly what the semantics are
> of this state flag --- what it means, when it should be set, and when
> it should be cleared, etc.?
> 
> I'm not entirely sure I understand why you test
> EXT4_STATE_MAY_INLINE_DATA versus simply calling
> ext4_has_inline_data() in various places.
> 
> Thanks!!
> 
> 					- Ted
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/fs/ext4/Makefile b/fs/ext4/Makefile
index 56fd8f8..e88b7a6 100644
--- a/fs/ext4/Makefile
+++ b/fs/ext4/Makefile
@@ -9,6 +9,6 @@  ext4-y	:= balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o page-io.o \
 		ext4_jbd2.o migrate.o mballoc.o block_validity.o move_extent.o \
 		mmp.o indirect.o
 
-ext4-$(CONFIG_EXT4_FS_XATTR)		+= xattr.o xattr_user.o xattr_trusted.o
+ext4-$(CONFIG_EXT4_FS_XATTR)		+= xattr.o xattr_user.o xattr_trusted.o inline.o
 ext4-$(CONFIG_EXT4_FS_POSIX_ACL)	+= acl.o
 ext4-$(CONFIG_EXT4_FS_SECURITY)		+= xattr_security.o
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index d6024d5..473c5d1 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -392,6 +392,7 @@  struct flex_groups {
 #define EXT4_EXTENTS_FL			0x00080000 /* Inode uses extents */
 #define EXT4_EA_INODE_FL	        0x00200000 /* Inode used for large EA */
 #define EXT4_EOFBLOCKS_FL		0x00400000 /* Blocks allocated beyond EOF */
+#define EXT4_INLINE_DATA_FL		0x10000000 /* Inode has inline data. */
 #define EXT4_RESERVED_FL		0x80000000 /* reserved for ext4 lib */
 
 #define EXT4_FL_USER_VISIBLE		0x004BDFFF /* User visible flags */
@@ -448,6 +449,7 @@  enum {
 	EXT4_INODE_EXTENTS	= 19,	/* Inode uses extents */
 	EXT4_INODE_EA_INODE	= 21,	/* Inode used for large EA */
 	EXT4_INODE_EOFBLOCKS	= 22,	/* Blocks allocated beyond EOF */
+	EXT4_INODE_INLINE_DATA	= 28,	/* Data in inode. */
 	EXT4_INODE_RESERVED	= 31,	/* reserved for ext4 lib */
 };
 
@@ -902,6 +904,10 @@  struct ext4_inode_info {
 	/* on-disk additional length */
 	__u16 i_extra_isize;
 
+	/* Indicate the inline data space. */
+	u16 i_inline_off;
+	u16 i_inline_size;
+
 #ifdef CONFIG_QUOTA
 	/* quota space reservation, managed internally by quota code */
 	qsize_t i_reserved_quota;
@@ -1360,6 +1366,7 @@  enum {
 	EXT4_STATE_DELALLOC_RESERVED,	/* blks already reserved for delalloc */
 	EXT4_STATE_DIOREAD_LOCK,	/* Disable support for dio read
 					   nolocking */
+	EXT4_STATE_MAY_INLINE_DATA,	/* may have in-inode data */
 };
 
 #define EXT4_INODE_BIT_FNS(name, field, offset)				\
@@ -1481,7 +1488,7 @@  static inline void ext4_clear_state_flags(struct ext4_inode_info *ei)
 #define EXT4_FEATURE_INCOMPAT_DIRDATA		0x1000 /* data in dirent */
 #define EXT4_FEATURE_INCOMPAT_BG_USE_META_CSUM	0x2000 /* use crc32c for bg */
 #define EXT4_FEATURE_INCOMPAT_LARGEDIR		0x4000 /* >2GB or 3-lvl htree */
-#define EXT4_FEATURE_INCOMPAT_INLINEDATA	0x8000 /* data in inode */
+#define EXT4_FEATURE_INCOMPAT_INLINE_DATA	0x8000 /* data in inode */
 
 #define EXT2_FEATURE_COMPAT_SUPP	EXT4_FEATURE_COMPAT_EXT_ATTR
 #define EXT2_FEATURE_INCOMPAT_SUPP	(EXT4_FEATURE_INCOMPAT_FILETYPE| \
diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c
new file mode 100644
index 0000000..aaa4461
--- /dev/null
+++ b/fs/ext4/inline.c
@@ -0,0 +1,461 @@ 
+/*
+ * Copyright (c) 2012 Taobao.
+ * Written by Tao Ma <boyu.mt@taobao.com>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of version 2.1 of the GNU Lesser General Public License
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+#include "ext4_jbd2.h"
+#include "ext4.h"
+#include "xattr.h"
+
+#define EXT4_XATTR_SYSTEM_DATA	"data"
+#define EXT4_MIN_INLINE_DATA_SIZE	((sizeof(__le32) * EXT4_N_BLOCKS))
+
+int ext4_get_inline_size(struct inode *inode)
+{
+	if (EXT4_I(inode)->i_inline_off)
+		return EXT4_I(inode)->i_inline_size;
+
+	return 0;
+}
+
+static int get_max_inline_xattr_value_size(struct inode *inode,
+					   struct ext4_iloc *iloc)
+{
+	struct ext4_xattr_ibody_header *header;
+	struct ext4_xattr_entry *entry;
+	struct ext4_inode *raw_inode;
+	int free, min_offs;
+
+	min_offs = EXT4_SB(inode->i_sb)->s_inode_size -
+			EXT4_GOOD_OLD_INODE_SIZE -
+			EXT4_I(inode)->i_extra_isize -
+			sizeof(struct ext4_xattr_ibody_header);
+
+	/*
+	 * We need to abstrace another sizeof(__u32) since an in-inode xattr
+	 * needs an empty 4 bytes to indicate the gap between the xattr entry
+	 * and the name/value pair.
+	 */
+	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
+		return EXT4_XATTR_SIZE(min_offs -
+			EXT4_XATTR_LEN(strlen(EXT4_XATTR_SYSTEM_DATA)) -
+			EXT4_XATTR_ROUND - sizeof(__u32));
+
+	raw_inode = ext4_raw_inode(iloc);
+	header = IHDR(inode, raw_inode);
+	entry = IFIRST(header);
+
+	/* Compute min_offs. */
+	for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
+		if (!entry->e_value_block && entry->e_value_size) {
+			size_t offs = le16_to_cpu(entry->e_value_offs);
+			if (offs < min_offs)
+				min_offs = offs;
+		}
+	}
+	free = min_offs -
+		((void *)entry - (void *)IFIRST(header)) - sizeof(__u32);
+
+	if (EXT4_I(inode)->i_inline_off) {
+		entry = (struct ext4_xattr_entry *)
+			((void *)raw_inode + EXT4_I(inode)->i_inline_off);
+
+		free += le32_to_cpu(entry->e_value_size);
+		goto out;
+	}
+
+	free -= EXT4_XATTR_LEN(strlen(EXT4_XATTR_SYSTEM_DATA));
+
+	if (free > EXT4_XATTR_ROUND)
+		free = EXT4_XATTR_SIZE(free - EXT4_XATTR_ROUND);
+	else
+		free = 0;
+
+out:
+	return free;
+}
+
+/*
+ * Get the maximum size we now can store in an inode.
+ * If we can't find the space for a xattr entry, don't use the space
+ * of the extents since we have no space to indicate the inline data.
+ */
+int ext4_get_max_inline_size(struct inode *inode)
+{
+	int error, max_inline_size;
+	struct ext4_iloc iloc;
+
+	if (EXT4_I(inode)->i_extra_isize == 0)
+		return 0;
+
+	error = ext4_get_inode_loc(inode, &iloc);
+	if (error) {
+		ext4_error_inode(inode, __func__, __LINE__, 0,
+				 "can't get inode location %lu",
+				 inode->i_ino);
+		return 0;
+	}
+
+	down_read(&EXT4_I(inode)->xattr_sem);
+	max_inline_size = get_max_inline_xattr_value_size(inode, &iloc);
+	up_read(&EXT4_I(inode)->xattr_sem);
+
+	brelse(iloc.bh);
+
+	if (!max_inline_size)
+		return 0;
+
+	return max_inline_size + EXT4_MIN_INLINE_DATA_SIZE;
+}
+
+int ext4_has_inline_data(struct inode *inode)
+{
+	return ext4_test_inode_flag(inode, EXT4_INODE_INLINE_DATA) &&
+	       EXT4_I(inode)->i_inline_off;
+}
+
+int ext4_find_inline_data(struct inode *inode)
+{
+	struct ext4_xattr_ibody_find is = {
+		.s = { .not_found = -ENODATA, },
+	};
+	struct ext4_xattr_info i = {
+		.name_index = EXT4_XATTR_INDEX_SYSTEM,
+		.name = EXT4_XATTR_SYSTEM_DATA,
+	};
+	int error;
+
+	if (EXT4_I(inode)->i_extra_isize == 0)
+		return 0;
+
+	error = ext4_get_inode_loc(inode, &is.iloc);
+	if (error)
+		return error;
+
+	error = ext4_xattr_ibody_find(inode, &i, &is);
+	if (error)
+		goto out;
+
+	if (!is.s.not_found) {
+		EXT4_I(inode)->i_inline_off = (u16)((void *)is.s.here -
+					(void *)ext4_raw_inode(&is.iloc));
+		EXT4_I(inode)->i_inline_size = EXT4_MIN_INLINE_DATA_SIZE +
+				le32_to_cpu(is.s.here->e_value_size);
+		ext4_set_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA);
+	}
+out:
+	brelse(is.iloc.bh);
+	return error;
+}
+
+static int ext4_read_inline_data(struct inode *inode, void *buffer,
+				 unsigned int len,
+				 struct ext4_iloc *iloc)
+{
+	struct ext4_xattr_entry *entry;
+	struct ext4_xattr_ibody_header *header;
+	int cp_len = 0;
+	struct ext4_inode *raw_inode;
+
+	if (!len)
+		return 0;
+
+	BUG_ON(len > EXT4_I(inode)->i_inline_size);
+
+	cp_len = len < EXT4_MIN_INLINE_DATA_SIZE ?
+			len : EXT4_MIN_INLINE_DATA_SIZE;
+
+	raw_inode = ext4_raw_inode(iloc);
+	memcpy(buffer, (void *)(raw_inode->i_block), cp_len);
+
+	len -= cp_len;
+	buffer += cp_len;
+
+	if (!len)
+		goto out;
+
+	header = IHDR(inode, raw_inode);
+	entry = (struct ext4_xattr_entry *)((void *)raw_inode +
+					    EXT4_I(inode)->i_inline_off);
+	len = min_t(unsigned int, len,
+		    (unsigned int)le32_to_cpu(entry->e_value_size));
+
+	memcpy(buffer,
+	       (void *)IFIRST(header) + le16_to_cpu(entry->e_value_offs), len);
+	cp_len += len;
+
+out:
+	return cp_len;
+}
+
+/*
+ * write the buffer to the inline inode.
+ * If 'create' is set, we don't need to do the extra copy in the xattr
+ * value since it is already handled by ext4_xattr_ibody_set. That saves
+ * us one memcpy.
+ */
+void ext4_write_inline_data(struct inode *inode, struct ext4_iloc *iloc,
+			    void *buffer, loff_t pos, unsigned int len)
+{
+	struct ext4_xattr_entry *entry;
+	struct ext4_xattr_ibody_header *header;
+	struct ext4_inode *raw_inode;
+	int cp_len = 0;
+
+	BUG_ON(!EXT4_I(inode)->i_inline_off);
+	BUG_ON(pos + len > EXT4_I(inode)->i_inline_size);
+
+	raw_inode = ext4_raw_inode(iloc);
+	buffer += pos;
+
+	if (pos < EXT4_MIN_INLINE_DATA_SIZE) {
+		cp_len = pos + len > EXT4_MIN_INLINE_DATA_SIZE ?
+			 EXT4_MIN_INLINE_DATA_SIZE - pos : len;
+		memcpy((void *)raw_inode->i_block + pos, buffer, cp_len);
+
+		len -= cp_len;
+		buffer += cp_len;
+		pos += cp_len;
+	}
+
+	if (!len)
+		return;
+
+	pos -= EXT4_MIN_INLINE_DATA_SIZE;
+	header = IHDR(inode, raw_inode);
+	entry = (struct ext4_xattr_entry *)((void *)raw_inode +
+					    EXT4_I(inode)->i_inline_off);
+
+	memcpy((void *)IFIRST(header) + le16_to_cpu(entry->e_value_offs) + pos,
+	       buffer, len);
+}
+
+static int ext4_create_inline_data(handle_t *handle,
+				   struct inode *inode, unsigned len)
+{
+	int error;
+	void *value = NULL;
+	struct ext4_xattr_ibody_find is = {
+		.s = { .not_found = -ENODATA, },
+	};
+	struct ext4_xattr_info i = {
+		.name_index = EXT4_XATTR_INDEX_SYSTEM,
+		.name = EXT4_XATTR_SYSTEM_DATA,
+	};
+
+	error = ext4_get_inode_loc(inode, &is.iloc);
+	if (error)
+		return error;
+
+	error = ext4_journal_get_write_access(handle, is.iloc.bh);
+	if (error)
+		goto out;
+
+	if (len > EXT4_MIN_INLINE_DATA_SIZE) {
+		value = (void *)empty_zero_page;
+		len -= EXT4_MIN_INLINE_DATA_SIZE;
+	} else {
+		value = "";
+		len = 0;
+	}
+
+	/* Insert the the xttr entry. */
+	i.value = value;
+	i.value_len = len;
+
+	error = ext4_xattr_ibody_find(inode, &i, &is);
+	if (error)
+		goto out;
+
+	BUG_ON(!is.s.not_found);
+
+	error = ext4_xattr_ibody_set(handle, inode, &i, &is);
+	if (error) {
+		if (error == -ENOSPC)
+			ext4_clear_inode_state(inode,
+					       EXT4_STATE_MAY_INLINE_DATA);
+		goto out;
+	}
+
+	memset((void *)ext4_raw_inode(&is.iloc)->i_block,
+		0, EXT4_MIN_INLINE_DATA_SIZE);
+
+	EXT4_I(inode)->i_inline_off = (u16)((void *)is.s.here -
+				      (void *)ext4_raw_inode(&is.iloc));
+	EXT4_I(inode)->i_inline_size = len + EXT4_MIN_INLINE_DATA_SIZE;
+	ext4_clear_inode_flag(inode, EXT4_INODE_EXTENTS);
+	ext4_set_inode_flag(inode, EXT4_INODE_INLINE_DATA);
+	get_bh(is.iloc.bh);
+	error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
+
+out:
+	brelse(is.iloc.bh);
+	return error;
+}
+
+static int ext4_update_inline_data(handle_t *handle, struct inode *inode,
+				   unsigned int len)
+{
+	int error;
+	void *value = NULL;
+	struct ext4_xattr_ibody_find is = {
+		.s = { .not_found = -ENODATA, },
+	};
+	struct ext4_xattr_info i = {
+		.name_index = EXT4_XATTR_INDEX_SYSTEM,
+		.name = EXT4_XATTR_SYSTEM_DATA,
+	};
+
+	/* If the old space is ok, write the data directly. */
+	if (len <= EXT4_I(inode)->i_inline_size)
+		return 0;
+
+	error = ext4_get_inode_loc(inode, &is.iloc);
+	if (error)
+		return error;
+
+	error = ext4_xattr_ibody_find(inode, &i, &is);
+	if (error)
+		goto out;
+
+	BUG_ON(is.s.not_found);
+
+	len -= EXT4_MIN_INLINE_DATA_SIZE;
+	value = kzalloc(len, GFP_NOFS);
+	if (!value)
+		goto out;
+
+	error = ext4_xattr_ibody_get(inode, i.name_index, i.name,
+				     value, len);
+	if (error == -ENODATA)
+		goto out;
+
+	error = ext4_journal_get_write_access(handle, is.iloc.bh);
+	if (error)
+		goto out;
+
+	/* Update the xttr entry. */
+	i.value = value;
+	i.value_len = len;
+
+	error = ext4_xattr_ibody_set(handle, inode, &i, &is);
+	if (error)
+		goto out;
+
+	EXT4_I(inode)->i_inline_off = (u16)((void *)is.s.here -
+				      (void *)ext4_raw_inode(&is.iloc));
+	EXT4_I(inode)->i_inline_size = EXT4_MIN_INLINE_DATA_SIZE +
+				le32_to_cpu(is.s.here->e_value_size);
+	ext4_set_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA);
+	get_bh(is.iloc.bh);
+	error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
+
+out:
+	kfree(value);
+	brelse(is.iloc.bh);
+	return error;
+}
+
+int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
+			     unsigned int len)
+{
+	int ret, size;
+	struct ext4_inode_info *ei = EXT4_I(inode);
+
+	if (!ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA))
+		return -ENOSPC;
+
+	size = ext4_get_max_inline_size(inode);
+	if (size < len)
+		return -ENOSPC;
+
+	down_write(&EXT4_I(inode)->xattr_sem);
+
+	if (ei->i_inline_off)
+		ret = ext4_update_inline_data(handle, inode, len);
+	else
+		ret = ext4_create_inline_data(handle, inode, len);
+
+	up_write(&EXT4_I(inode)->xattr_sem);
+
+	return ret;
+}
+
+static int ext4_destroy_inline_data_nolock(handle_t *handle,
+					   struct inode *inode)
+{
+	struct ext4_inode_info *ei = EXT4_I(inode);
+	struct ext4_xattr_ibody_find is = {
+		.s = { .not_found = 0, },
+	};
+	struct ext4_xattr_info i = {
+		.name_index = EXT4_XATTR_INDEX_SYSTEM,
+		.name = EXT4_XATTR_SYSTEM_DATA,
+		.value = NULL,
+		.value_len = 0,
+	};
+	int error;
+
+	if (!ei->i_inline_off)
+		return 0;
+
+	error = ext4_get_inode_loc(inode, &is.iloc);
+	if (error)
+		return error;
+
+	error = ext4_xattr_ibody_find(inode, &i, &is);
+	if (error)
+		goto out;
+
+	error = ext4_journal_get_write_access(handle, is.iloc.bh);
+	if (error)
+		goto out;
+
+	error = ext4_xattr_ibody_set(handle, inode, &i, &is);
+	if (error)
+		goto out;
+
+	memset((void *)ext4_raw_inode(&is.iloc)->i_block,
+		0, EXT4_MIN_INLINE_DATA_SIZE);
+
+	if (EXT4_HAS_INCOMPAT_FEATURE(inode->i_sb,
+				      EXT4_FEATURE_INCOMPAT_EXTENTS)) {
+		if (S_ISDIR(inode->i_mode) ||
+		    S_ISREG(inode->i_mode) || S_ISLNK(inode->i_mode)) {
+			ext4_set_inode_flag(inode, EXT4_INODE_EXTENTS);
+			ext4_ext_tree_init(handle, inode);
+		}
+	}
+	ext4_clear_inode_flag(inode, EXT4_INODE_INLINE_DATA);
+
+	get_bh(is.iloc.bh);
+	error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
+
+	EXT4_I(inode)->i_inline_off = 0;
+	EXT4_I(inode)->i_inline_size = 0;
+	ext4_clear_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA);
+out:
+	brelse(is.iloc.bh);
+	if (error == -ENODATA)
+		error = 0;
+	return error;
+}
+
+int ext4_destroy_inline_data(handle_t *handle, struct inode *inode)
+{
+	int ret;
+
+	down_write(&EXT4_I(inode)->xattr_sem);
+	ret = ext4_destroy_inline_data_nolock(handle, inode);
+	up_write(&EXT4_I(inode)->xattr_sem);
+
+	return ret;
+}
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index e8f6ce0..bcf7648 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3762,8 +3762,10 @@  static inline void ext4_iget_extra_inode(struct inode *inode,
 {
 	__le32 *magic = (void *)raw_inode +
 			EXT4_GOOD_OLD_INODE_SIZE + ei->i_extra_isize;
-	if (*magic == cpu_to_le32(EXT4_XATTR_MAGIC))
+	if (*magic == cpu_to_le32(EXT4_XATTR_MAGIC)) {
 		ext4_set_inode_state(inode, EXT4_STATE_XATTR);
+		ext4_find_inline_data(inode);
+	}
 }
 
 struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
@@ -3836,6 +3838,7 @@  struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
 	set_nlink(inode, le16_to_cpu(raw_inode->i_links_count));
 
 	ext4_clear_state_flags(ei);	/* Only relevant on 32-bit archs */
+	ei->i_inline_off = 0;
 	ei->i_dir_start_lookup = 0;
 	ei->i_dtime = le32_to_cpu(raw_inode->i_dtime);
 	/* We now have enough fields to check if the inode was active or not.
diff --git a/fs/ext4/xattr.h b/fs/ext4/xattr.h
index 8c61a41..8c51edf 100644
--- a/fs/ext4/xattr.h
+++ b/fs/ext4/xattr.h
@@ -21,6 +21,7 @@ 
 #define EXT4_XATTR_INDEX_TRUSTED		4
 #define	EXT4_XATTR_INDEX_LUSTRE			5
 #define EXT4_XATTR_INDEX_SECURITY	        6
+#define EXT4_XATTR_INDEX_SYSTEM			7
 
 struct ext4_xattr_header {
 	__le32	h_magic;	/* magic number for identification */
@@ -125,6 +126,19 @@  extern int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
 				struct ext4_xattr_info *i,
 				struct ext4_xattr_ibody_find *is);
 
+extern int ext4_has_inline_data(struct inode *inode);
+extern int ext4_get_inline_size(struct inode *inode);
+extern int ext4_get_max_inline_size(struct inode *inode);
+extern int ext4_find_inline_data(struct inode *inode);
+extern void ext4_write_inline_data(struct inode *inode,
+				   struct ext4_iloc *iloc,
+				   void *buffer, loff_t pos,
+				   unsigned int len);
+extern int ext4_prepare_inline_data(handle_t *handle, struct inode *inode,
+				    unsigned int len);
+extern int ext4_init_inline_data(handle_t *handle, struct inode *inode,
+				 unsigned int len);
+extern int ext4_destroy_inline_data(handle_t *handle, struct inode *inode);
 # else  /* CONFIG_EXT4_FS_XATTR */
 
 static inline int
@@ -200,6 +214,46 @@  extern int ext4_xattr_ibody_get(struct inode *inode, int name_index,
 	return -EOPNOTSUPP;
 }
 
+static inline int ext4_find_inline_data(struct inode *inode)
+{
+	return 0;
+}
+
+static inline int ext4_has_inline_data(struct inode *inode)
+{
+	return 0;
+}
+
+static inline int ext4_get_inline_size(struct inode *inode)
+{
+	return 0;
+}
+
+static inline int ext4_get_max_inline_size(struct inode *inode)
+{
+	return 0;
+}
+
+static inline void ext4_write_inline_data(struct inode *inode,
+					  struct ext4_iloc *iloc,
+					  void *buffer, loff_t pos,
+					  unsigned int len)
+{
+	return;
+}
+
+static inline int ext4_init_inline_data(handle_t *handle,
+					struct inode *inode,
+					unsigned int len)
+{
+	return 0;
+}
+
+static inline int ext4_destroy_inline_data(handle_t *handle,
+					   struct inode *inode)
+{
+	return 0;
+}
 # endif  /* CONFIG_EXT4_FS_XATTR */
 
 #ifdef CONFIG_EXT4_FS_SECURITY