[PATCH 0/4 v3] ext4: Speedup orphan file handling

[PATCH 0/4 v3] ext4: Speedup orphan file handling

[Jan Kara]

  Hello,

After six years, prompted by recent 0-day reports of performance issues with
orphan list handling [1], I'm sending a third revision of my series to speed up
orphan inode handling in ext4.

Orphan inode handling in ext4 is a bottleneck for workloads which heavily
excercise truncate / unlink of small files as they contend on global
s_orphan_mutex (when you have fast enough storage). This patch set implements
new way of handling orphan inodes - instead of using a linked list, we store
inode numbers of orphaned inodes in a file which is possible to implement in a
more scalable manner than linked list manipulations. See description of patch
3/4 for more details.

The patch set achieves significant gains both for a micro benchmark stressing
orphan inode handling (truncating file byte-by-byte, several threads in
parallel) and for reaim creat_clo workload. I'm happy for any review, thoughts,
ideas about the patches. I have also implemented full support in e2fsprogs
which I'll send separately.

								Honza

[1] https://lore.kernel.org/lkml/20210227120804.GB22871@xsang-OptiPlex-9020/

Changes since v2:
* Updated some comments
* Rebased onto 5.13-rc5
* Change orphan file inode from a fixed inode number to inode number stored
  in the superblock

Changes since v1:
* orphan blocks have now magic numbers
* split out orphan handling to a separate source file
* some smaller updates according to review

[PATCH 1/4] ext4: Support for checksumming from journal triggers

[Jan Kara]

JBD2 layer support triggers which are called when journaling layer moves
buffer to a certain state. We can use the frozen trigger, which gets
called when buffer data is frozen and about to be written out to the
journal, to compute block checksums for some buffer types (similarly as
does ocfs2). This avoids unnecessary repeated recomputation of the
checksum (at the cost of larger window where memory corruption won't be
caught by checksumming) and is even necessary when there are
unsynchronized updaters of the checksummed data.

So add argument to ext4_journal_get_write_access() and
ext4_journal_get_create_access() which describes buffer type so that
triggers can be set accordingly. This patch is mostly only a change of
prototype of the above mentioned functions and a few small helpers. Real
checksumming will come later.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/ext4.h        | 26 ++++++++++++--
 fs/ext4/ext4_jbd2.c   | 43 +++++++++++++++-------
 fs/ext4/ext4_jbd2.h   | 18 ++++++----
 fs/ext4/extents.c     | 12 ++++---
 fs/ext4/file.c        |  3 +-
 fs/ext4/ialloc.c      | 19 ++++++----
 fs/ext4/indirect.c    | 15 +++++---
 fs/ext4/inline.c      | 26 +++++++++-----
 fs/ext4/inode.c       | 84 +++++++++++++++++++++++++------------------
 fs/ext4/ioctl.c       |  4 ++-
 fs/ext4/mballoc.c     | 15 ++++----
 fs/ext4/namei.c       | 40 +++++++++++++--------
 fs/ext4/resize.c      | 38 ++++++++++++--------
 fs/ext4/super.c       | 16 ++++++++-
 fs/ext4/xattr.c       | 26 +++++++++-----
 fs/jbd2/transaction.c |  2 +-
 16 files changed, 259 insertions(+), 128 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 37002663d521..b81256a7e7f2 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1438,6 +1438,24 @@ struct ext4_super_block {
 
 #define EXT4_ENC_UTF8_12_1	1
 
+/* Types of ext4 journal triggers */
+enum ext4_journal_trigger_type {
+	EXT4_JTR_NONE	/* This must be the last entry for indexing to work! */
+};
+
+#define EXT4_JOURNAL_TRIGGER_COUNT EXT4_JTR_NONE
+
+struct ext4_journal_trigger {
+	struct jbd2_buffer_trigger_type tr_triggers;
+	struct super_block *sb;
+};
+
+static inline struct ext4_journal_trigger *EXT4_TRIGGER(
+				struct jbd2_buffer_trigger_type *trigger)
+{
+	return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
+}
+
 /*
  * fourth extended-fs super-block data in memory
  */
@@ -1615,6 +1633,9 @@ struct ext4_sb_info {
 	struct mb_cache *s_ea_inode_cache;
 	spinlock_t s_es_lock ____cacheline_aligned_in_smp;
 
+	/* Journal triggers for checksum computation */
+	struct ext4_journal_trigger s_journal_triggers[EXT4_JOURNAL_TRIGGER_COUNT];
+
 	/* Ratelimit ext4 messages. */
 	struct ratelimit_state s_err_ratelimit_state;
 	struct ratelimit_state s_warning_ratelimit_state;
@@ -2910,13 +2931,14 @@ int ext4_get_block(struct inode *inode, sector_t iblock,
 int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
 			   struct buffer_head *bh, int create);
 int ext4_walk_page_buffers(handle_t *handle,
+			   struct inode *inode,
 			   struct buffer_head *head,
 			   unsigned from,
 			   unsigned to,
 			   int *partial,
-			   int (*fn)(handle_t *handle,
+			   int (*fn)(handle_t *handle, struct inode *inode,
 				     struct buffer_head *bh));
-int do_journal_get_write_access(handle_t *handle,
+int do_journal_get_write_access(handle_t *handle, struct inode *inode,
 				struct buffer_head *bh);
 #define FALL_BACK_TO_NONDELALLOC 1
 #define CONVERT_INLINE_DATA	 2
diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c
index be799040a415..f601e24b6015 100644
--- a/fs/ext4/ext4_jbd2.c
+++ b/fs/ext4/ext4_jbd2.c
@@ -218,9 +218,11 @@ static void ext4_check_bdev_write_error(struct super_block *sb)
 }
 
 int __ext4_journal_get_write_access(const char *where, unsigned int line,
-				    handle_t *handle, struct buffer_head *bh)
+				    handle_t *handle, struct super_block *sb,
+				    struct buffer_head *bh,
+				    enum ext4_journal_trigger_type trigger_type)
 {
-	int err = 0;
+	int err;
 
 	might_sleep();
 
@@ -229,11 +231,18 @@ int __ext4_journal_get_write_access(const char *where, unsigned int line,
 
 	if (ext4_handle_valid(handle)) {
 		err = jbd2_journal_get_write_access(handle, bh);
-		if (err)
+		if (err) {
 			ext4_journal_abort_handle(where, line, __func__, bh,
 						  handle, err);
+			return err;
+		}
 	}
-	return err;
+	if (trigger_type == EXT4_JTR_NONE || !ext4_has_metadata_csum(sb))
+		return 0;
+	WARN_ON_ONCE(trigger_type >= EXT4_JOURNAL_TRIGGER_COUNT);
+	jbd2_journal_set_triggers(bh,
+		&EXT4_SB(sb)->s_journal_triggers[trigger_type].tr_triggers);
+	return 0;
 }
 
 /*
@@ -304,17 +313,27 @@ int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
 }
 
 int __ext4_journal_get_create_access(const char *where, unsigned int line,
-				handle_t *handle, struct buffer_head *bh)
+				handle_t *handle, struct super_block *sb,
+				struct buffer_head *bh,
+				enum ext4_journal_trigger_type trigger_type)
 {
-	int err = 0;
+	int err;
 
-	if (ext4_handle_valid(handle)) {
-		err = jbd2_journal_get_create_access(handle, bh);
-		if (err)
-			ext4_journal_abort_handle(where, line, __func__,
-						  bh, handle, err);
+	if (!ext4_handle_valid(handle))
+		return 0;
+
+	err = jbd2_journal_get_create_access(handle, bh);
+	if (err) {
+		ext4_journal_abort_handle(where, line, __func__, bh, handle,
+					  err);
+		return err;
 	}
-	return err;
+	if (trigger_type == EXT4_JTR_NONE || !ext4_has_metadata_csum(sb))
+		return 0;
+	WARN_ON_ONCE(trigger_type >= EXT4_JOURNAL_TRIGGER_COUNT);
+	jbd2_journal_set_triggers(bh,
+		&EXT4_SB(sb)->s_journal_triggers[trigger_type].tr_triggers);
+	return 0;
 }
 
 int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index 0d2fa423b7ad..0e4fa644df01 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -231,26 +231,32 @@ int ext4_expand_extra_isize(struct inode *inode,
  * Wrapper functions with which ext4 calls into JBD.
  */
 int __ext4_journal_get_write_access(const char *where, unsigned int line,
-				    handle_t *handle, struct buffer_head *bh);
+				    handle_t *handle, struct super_block *sb,
+				    struct buffer_head *bh,
+				    enum ext4_journal_trigger_type trigger_type);
 
 int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
 		  int is_metadata, struct inode *inode,
 		  struct buffer_head *bh, ext4_fsblk_t blocknr);
 
 int __ext4_journal_get_create_access(const char *where, unsigned int line,
-				handle_t *handle, struct buffer_head *bh);
+				handle_t *handle, struct super_block *sb,
+				struct buffer_head *bh,
+				enum ext4_journal_trigger_type trigger_type);
 
 int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
 				 handle_t *handle, struct inode *inode,
 				 struct buffer_head *bh);
 
-#define ext4_journal_get_write_access(handle, bh) \
-	__ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
+#define ext4_journal_get_write_access(handle, sb, bh, trigger_type) \
+	__ext4_journal_get_write_access(__func__, __LINE__, (handle), (sb), \
+					(bh), (trigger_type))
 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
 	__ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
 		      (bh), (block_nr))
-#define ext4_journal_get_create_access(handle, bh) \
-	__ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
+#define ext4_journal_get_create_access(handle, sb, bh, trigger_type) \
+	__ext4_journal_get_create_access(__func__, __LINE__, (handle), (sb), \
+					 (bh), (trigger_type))
 #define ext4_handle_dirty_metadata(handle, inode, bh) \
 	__ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
 				     (bh))
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index cbf37b2cf871..b1e479a111b4 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -139,7 +139,8 @@ static int ext4_ext_get_access(handle_t *handle, struct inode *inode,
 	if (path->p_bh) {
 		/* path points to block */
 		BUFFER_TRACE(path->p_bh, "get_write_access");
-		return ext4_journal_get_write_access(handle, path->p_bh);
+		return ext4_journal_get_write_access(handle, inode->i_sb,
+						     path->p_bh, EXT4_JTR_NONE);
 	}
 	/* path points to leaf/index in inode body */
 	/* we use in-core data, no need to protect them */
@@ -1081,7 +1082,8 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
 	}
 	lock_buffer(bh);
 
-	err = ext4_journal_get_create_access(handle, bh);
+	err = ext4_journal_get_create_access(handle, inode->i_sb, bh,
+					     EXT4_JTR_NONE);
 	if (err)
 		goto cleanup;
 
@@ -1158,7 +1160,8 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
 		}
 		lock_buffer(bh);
 
-		err = ext4_journal_get_create_access(handle, bh);
+		err = ext4_journal_get_create_access(handle, inode->i_sb, bh,
+						     EXT4_JTR_NONE);
 		if (err)
 			goto cleanup;
 
@@ -1283,7 +1286,8 @@ static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
 		return -ENOMEM;
 	lock_buffer(bh);
 
-	err = ext4_journal_get_create_access(handle, bh);
+	err = ext4_journal_get_create_access(handle, inode->i_sb, bh,
+					     EXT4_JTR_NONE);
 	if (err) {
 		unlock_buffer(bh);
 		goto out;
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 816dedcbd541..eda12bc50592 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -822,7 +822,8 @@ static int ext4_sample_last_mounted(struct super_block *sb,
 	if (IS_ERR(handle))
 		goto out;
 	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, sbi->s_sbh);
+	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto out_journal;
 	lock_buffer(sbi->s_sbh);
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index 9bab7fd4ccd5..fb1e3470f2f6 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -300,7 +300,8 @@ void ext4_free_inode(handle_t *handle, struct inode *inode)
 	}
 
 	BUFFER_TRACE(bitmap_bh, "get_write_access");
-	fatal = ext4_journal_get_write_access(handle, bitmap_bh);
+	fatal = ext4_journal_get_write_access(handle, sb, bitmap_bh,
+					      EXT4_JTR_NONE);
 	if (fatal)
 		goto error_return;
 
@@ -308,7 +309,8 @@ void ext4_free_inode(handle_t *handle, struct inode *inode)
 	gdp = ext4_get_group_desc(sb, block_group, &bh2);
 	if (gdp) {
 		BUFFER_TRACE(bh2, "get_write_access");
-		fatal = ext4_journal_get_write_access(handle, bh2);
+		fatal = ext4_journal_get_write_access(handle, sb, bh2,
+						      EXT4_JTR_NONE);
 	}
 	ext4_lock_group(sb, block_group);
 	cleared = ext4_test_and_clear_bit(bit, bitmap_bh->b_data);
@@ -1086,7 +1088,8 @@ struct inode *__ext4_new_inode(struct user_namespace *mnt_userns,
 			}
 		}
 		BUFFER_TRACE(inode_bitmap_bh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, inode_bitmap_bh);
+		err = ext4_journal_get_write_access(handle, sb, inode_bitmap_bh,
+						    EXT4_JTR_NONE);
 		if (err) {
 			ext4_std_error(sb, err);
 			goto out;
@@ -1128,7 +1131,8 @@ struct inode *__ext4_new_inode(struct user_namespace *mnt_userns,
 	}
 
 	BUFFER_TRACE(group_desc_bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, group_desc_bh);
+	err = ext4_journal_get_write_access(handle, sb, group_desc_bh,
+					    EXT4_JTR_NONE);
 	if (err) {
 		ext4_std_error(sb, err);
 		goto out;
@@ -1145,7 +1149,8 @@ struct inode *__ext4_new_inode(struct user_namespace *mnt_userns,
 			goto out;
 		}
 		BUFFER_TRACE(block_bitmap_bh, "get block bitmap access");
-		err = ext4_journal_get_write_access(handle, block_bitmap_bh);
+		err = ext4_journal_get_write_access(handle, sb, block_bitmap_bh,
+						    EXT4_JTR_NONE);
 		if (err) {
 			brelse(block_bitmap_bh);
 			ext4_std_error(sb, err);
@@ -1584,8 +1589,8 @@ int ext4_init_inode_table(struct super_block *sb, ext4_group_t group,
 	num = sbi->s_itb_per_group - used_blks;
 
 	BUFFER_TRACE(group_desc_bh, "get_write_access");
-	ret = ext4_journal_get_write_access(handle,
-					    group_desc_bh);
+	ret = ext4_journal_get_write_access(handle, sb, group_desc_bh,
+					    EXT4_JTR_NONE);
 	if (ret)
 		goto err_out;
 
diff --git a/fs/ext4/indirect.c b/fs/ext4/indirect.c
index a7bc6ad656a9..89efa78ed4b2 100644
--- a/fs/ext4/indirect.c
+++ b/fs/ext4/indirect.c
@@ -354,7 +354,8 @@ static int ext4_alloc_branch(handle_t *handle,
 		}
 		lock_buffer(bh);
 		BUFFER_TRACE(bh, "call get_create_access");
-		err = ext4_journal_get_create_access(handle, bh);
+		err = ext4_journal_get_create_access(handle, ar->inode->i_sb,
+						     bh, EXT4_JTR_NONE);
 		if (err) {
 			unlock_buffer(bh);
 			goto failed;
@@ -429,7 +430,8 @@ static int ext4_splice_branch(handle_t *handle,
 	 */
 	if (where->bh) {
 		BUFFER_TRACE(where->bh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, where->bh);
+		err = ext4_journal_get_write_access(handle, ar->inode->i_sb,
+						    where->bh, EXT4_JTR_NONE);
 		if (err)
 			goto err_out;
 	}
@@ -728,7 +730,8 @@ static int ext4_ind_truncate_ensure_credits(handle_t *handle,
 		return ret;
 	if (bh) {
 		BUFFER_TRACE(bh, "retaking write access");
-		ret = ext4_journal_get_write_access(handle, bh);
+		ret = ext4_journal_get_write_access(handle, inode->i_sb, bh,
+						    EXT4_JTR_NONE);
 		if (unlikely(ret))
 			return ret;
 	}
@@ -916,7 +919,8 @@ static void ext4_free_data(handle_t *handle, struct inode *inode,
 
 	if (this_bh) {				/* For indirect block */
 		BUFFER_TRACE(this_bh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, this_bh);
+		err = ext4_journal_get_write_access(handle, inode->i_sb,
+						    this_bh, EXT4_JTR_NONE);
 		/* Important: if we can't update the indirect pointers
 		 * to the blocks, we can't free them. */
 		if (err)
@@ -1079,7 +1083,8 @@ static void ext4_free_branches(handle_t *handle, struct inode *inode,
 				 */
 				BUFFER_TRACE(parent_bh, "get_write_access");
 				if (!ext4_journal_get_write_access(handle,
-								   parent_bh)){
+						inode->i_sb, parent_bh,
+						EXT4_JTR_NONE)) {
 					*p = 0;
 					BUFFER_TRACE(parent_bh,
 					"call ext4_handle_dirty_metadata");
diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c
index 3cf01629010d..fa85f942b884 100644
--- a/fs/ext4/inline.c
+++ b/fs/ext4/inline.c
@@ -264,7 +264,8 @@ static int ext4_create_inline_data(handle_t *handle,
 		return error;
 
 	BUFFER_TRACE(is.iloc.bh, "get_write_access");
-	error = ext4_journal_get_write_access(handle, is.iloc.bh);
+	error = ext4_journal_get_write_access(handle, inode->i_sb, is.iloc.bh,
+					      EXT4_JTR_NONE);
 	if (error)
 		goto out;
 
@@ -350,7 +351,8 @@ static int ext4_update_inline_data(handle_t *handle, struct inode *inode,
 		goto out;
 
 	BUFFER_TRACE(is.iloc.bh, "get_write_access");
-	error = ext4_journal_get_write_access(handle, is.iloc.bh);
+	error = ext4_journal_get_write_access(handle, inode->i_sb, is.iloc.bh,
+					      EXT4_JTR_NONE);
 	if (error)
 		goto out;
 
@@ -427,7 +429,8 @@ static int ext4_destroy_inline_data_nolock(handle_t *handle,
 		goto out;
 
 	BUFFER_TRACE(is.iloc.bh, "get_write_access");
-	error = ext4_journal_get_write_access(handle, is.iloc.bh);
+	error = ext4_journal_get_write_access(handle, inode->i_sb, is.iloc.bh,
+					      EXT4_JTR_NONE);
 	if (error)
 		goto out;
 
@@ -593,7 +596,7 @@ static int ext4_convert_inline_data_to_extent(struct address_space *mapping,
 		ret = __block_write_begin(page, from, to, ext4_get_block);
 
 	if (!ret && ext4_should_journal_data(inode)) {
-		ret = ext4_walk_page_buffers(handle, page_buffers(page),
+		ret = ext4_walk_page_buffers(handle, inode, page_buffers(page),
 					     from, to, NULL,
 					     do_journal_get_write_access);
 	}
@@ -682,7 +685,8 @@ int ext4_try_to_write_inline_data(struct address_space *mapping,
 		goto convert;
 	}
 
-	ret = ext4_journal_get_write_access(handle, iloc.bh);
+	ret = ext4_journal_get_write_access(handle, inode->i_sb, iloc.bh,
+					    EXT4_JTR_NONE);
 	if (ret)
 		goto out;
 
@@ -923,7 +927,8 @@ int ext4_da_write_inline_data_begin(struct address_space *mapping,
 		if (ret < 0)
 			goto out_release_page;
 	}
-	ret = ext4_journal_get_write_access(handle, iloc.bh);
+	ret = ext4_journal_get_write_access(handle, inode->i_sb, iloc.bh,
+					    EXT4_JTR_NONE);
 	if (ret)
 		goto out_release_page;
 
@@ -1028,7 +1033,8 @@ static int ext4_add_dirent_to_inline(handle_t *handle,
 		return err;
 
 	BUFFER_TRACE(iloc->bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, iloc->bh);
+	err = ext4_journal_get_write_access(handle, dir->i_sb, iloc->bh,
+					    EXT4_JTR_NONE);
 	if (err)
 		return err;
 	ext4_insert_dentry(dir, inode, de, inline_size, fname);
@@ -1223,7 +1229,8 @@ static int ext4_convert_inline_data_nolock(handle_t *handle,
 	}
 
 	lock_buffer(data_bh);
-	error = ext4_journal_get_create_access(handle, data_bh);
+	error = ext4_journal_get_create_access(handle, inode->i_sb, data_bh,
+					       EXT4_JTR_NONE);
 	if (error) {
 		unlock_buffer(data_bh);
 		error = -EIO;
@@ -1707,7 +1714,8 @@ int ext4_delete_inline_entry(handle_t *handle,
 	}
 
 	BUFFER_TRACE(bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, bh);
+	err = ext4_journal_get_write_access(handle, dir->i_sb, bh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto out;
 
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index fe6045a46599..79126bf5c09d 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -139,7 +139,6 @@ static inline int ext4_begin_ordered_truncate(struct inode *inode,
 static void ext4_invalidatepage(struct page *page, unsigned int offset,
 				unsigned int length);
 static int __ext4_journalled_writepage(struct page *page, unsigned int len);
-static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh);
 static int ext4_meta_trans_blocks(struct inode *inode, int lblocks,
 				  int pextents);
 
@@ -869,7 +868,8 @@ struct buffer_head *ext4_getblk(handle_t *handle, struct inode *inode,
 		 */
 		lock_buffer(bh);
 		BUFFER_TRACE(bh, "call get_create_access");
-		err = ext4_journal_get_create_access(handle, bh);
+		err = ext4_journal_get_create_access(handle, inode->i_sb, bh,
+						     EXT4_JTR_NONE);
 		if (unlikely(err)) {
 			unlock_buffer(bh);
 			goto errout;
@@ -954,12 +954,12 @@ int ext4_bread_batch(struct inode *inode, ext4_lblk_t block, int bh_count,
 	return err;
 }
 
-int ext4_walk_page_buffers(handle_t *handle,
+int ext4_walk_page_buffers(handle_t *handle, struct inode *inode,
 			   struct buffer_head *head,
 			   unsigned from,
 			   unsigned to,
 			   int *partial,
-			   int (*fn)(handle_t *handle,
+			   int (*fn)(handle_t *handle, struct inode *inode,
 				     struct buffer_head *bh))
 {
 	struct buffer_head *bh;
@@ -978,7 +978,7 @@ int ext4_walk_page_buffers(handle_t *handle,
 				*partial = 1;
 			continue;
 		}
-		err = (*fn)(handle, bh);
+		err = (*fn)(handle, inode, bh);
 		if (!ret)
 			ret = err;
 	}
@@ -1009,7 +1009,7 @@ int ext4_walk_page_buffers(handle_t *handle,
  * is elevated.  We'll still have enough credits for the tiny quotafile
  * write.
  */
-int do_journal_get_write_access(handle_t *handle,
+int do_journal_get_write_access(handle_t *handle, struct inode *inode,
 				struct buffer_head *bh)
 {
 	int dirty = buffer_dirty(bh);
@@ -1028,7 +1028,8 @@ int do_journal_get_write_access(handle_t *handle,
 	if (dirty)
 		clear_buffer_dirty(bh);
 	BUFFER_TRACE(bh, "get write access");
-	ret = ext4_journal_get_write_access(handle, bh);
+	ret = ext4_journal_get_write_access(handle, inode->i_sb, bh,
+					    EXT4_JTR_NONE);
 	if (!ret && dirty)
 		ret = ext4_handle_dirty_metadata(handle, NULL, bh);
 	return ret;
@@ -1208,8 +1209,8 @@ static int ext4_write_begin(struct file *file, struct address_space *mapping,
 		ret = __block_write_begin(page, pos, len, ext4_get_block);
 #endif
 	if (!ret && ext4_should_journal_data(inode)) {
-		ret = ext4_walk_page_buffers(handle, page_buffers(page),
-					     from, to, NULL,
+		ret = ext4_walk_page_buffers(handle, inode,
+					     page_buffers(page), from, to, NULL,
 					     do_journal_get_write_access);
 	}
 
@@ -1253,7 +1254,8 @@ static int ext4_write_begin(struct file *file, struct address_space *mapping,
 }
 
 /* For write_end() in data=journal mode */
-static int write_end_fn(handle_t *handle, struct buffer_head *bh)
+static int write_end_fn(handle_t *handle, struct inode *inode,
+			struct buffer_head *bh)
 {
 	int ret;
 	if (!buffer_mapped(bh) || buffer_freed(bh))
@@ -1352,6 +1354,7 @@ static int ext4_write_end(struct file *file,
  * to call ext4_handle_dirty_metadata() instead.
  */
 static void ext4_journalled_zero_new_buffers(handle_t *handle,
+					    struct inode *inode,
 					    struct page *page,
 					    unsigned from, unsigned to)
 {
@@ -1370,7 +1373,7 @@ static void ext4_journalled_zero_new_buffers(handle_t *handle,
 					size = min(to, block_end) - start;
 
 					zero_user(page, start, size);
-					write_end_fn(handle, bh);
+					write_end_fn(handle, inode, bh);
 				}
 				clear_buffer_new(bh);
 			}
@@ -1412,13 +1415,13 @@ static int ext4_journalled_write_end(struct file *file,
 		copied = ret;
 	} else if (unlikely(copied < len) && !PageUptodate(page)) {
 		copied = 0;
-		ext4_journalled_zero_new_buffers(handle, page, from, to);
+		ext4_journalled_zero_new_buffers(handle, inode, page, from, to);
 	} else {
 		if (unlikely(copied < len))
-			ext4_journalled_zero_new_buffers(handle, page,
+			ext4_journalled_zero_new_buffers(handle, inode, page,
 							 from + copied, to);
-		ret = ext4_walk_page_buffers(handle, page_buffers(page), from,
-					     from + copied, &partial,
+		ret = ext4_walk_page_buffers(handle, inode, page_buffers(page),
+					     from, from + copied, &partial,
 					     write_end_fn);
 		if (!partial)
 			SetPageUptodate(page);
@@ -1619,7 +1622,8 @@ static void ext4_print_free_blocks(struct inode *inode)
 	return;
 }
 
-static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh)
+static int ext4_bh_delay_or_unwritten(handle_t *handle, struct inode *inode,
+				      struct buffer_head *bh)
 {
 	return (buffer_delay(bh) || buffer_unwritten(bh)) && buffer_dirty(bh);
 }
@@ -1851,13 +1855,15 @@ int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
 	return 0;
 }
 
-static int bget_one(handle_t *handle, struct buffer_head *bh)
+static int bget_one(handle_t *handle, struct inode *inode,
+		    struct buffer_head *bh)
 {
 	get_bh(bh);
 	return 0;
 }
 
-static int bput_one(handle_t *handle, struct buffer_head *bh)
+static int bput_one(handle_t *handle, struct inode *inode,
+		    struct buffer_head *bh)
 {
 	put_bh(bh);
 	return 0;
@@ -1888,7 +1894,7 @@ static int __ext4_journalled_writepage(struct page *page,
 			BUG();
 			goto out;
 		}
-		ext4_walk_page_buffers(handle, page_bufs, 0, len,
+		ext4_walk_page_buffers(handle, inode, page_bufs, 0, len,
 				       NULL, bget_one);
 	}
 	/*
@@ -1920,11 +1926,11 @@ static int __ext4_journalled_writepage(struct page *page,
 	if (inline_data) {
 		ret = ext4_mark_inode_dirty(handle, inode);
 	} else {
-		ret = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
-					     do_journal_get_write_access);
+		ret = ext4_walk_page_buffers(handle, inode, page_bufs, 0, len,
+					     NULL, do_journal_get_write_access);
 
-		err = ext4_walk_page_buffers(handle, page_bufs, 0, len, NULL,
-					     write_end_fn);
+		err = ext4_walk_page_buffers(handle, inode, page_bufs, 0, len,
+					     NULL, write_end_fn);
 	}
 	if (ret == 0)
 		ret = err;
@@ -1941,7 +1947,7 @@ static int __ext4_journalled_writepage(struct page *page,
 	unlock_page(page);
 out_no_pagelock:
 	if (!inline_data && page_bufs)
-		ext4_walk_page_buffers(NULL, page_bufs, 0, len,
+		ext4_walk_page_buffers(NULL, inode, page_bufs, 0, len,
 				       NULL, bput_one);
 	brelse(inode_bh);
 	return ret;
@@ -2031,7 +2037,7 @@ static int ext4_writepage(struct page *page,
 	 * for the extremely common case, this is an optimization that
 	 * skips a useless round trip through ext4_bio_write_page().
 	 */
-	if (ext4_walk_page_buffers(NULL, page_bufs, 0, len, NULL,
+	if (ext4_walk_page_buffers(NULL, inode, page_bufs, 0, len, NULL,
 				   ext4_bh_delay_or_unwritten)) {
 		redirty_page_for_writepage(wbc, page);
 		if ((current->flags & PF_MEMALLOC) ||
@@ -3794,7 +3800,8 @@ static int __ext4_block_zero_page_range(handle_t *handle,
 	}
 	if (ext4_should_journal_data(inode)) {
 		BUFFER_TRACE(bh, "get write access");
-		err = ext4_journal_get_write_access(handle, bh);
+		err = ext4_journal_get_write_access(handle, inode->i_sb, bh,
+						    EXT4_JTR_NONE);
 		if (err)
 			goto unlock;
 	}
@@ -5142,7 +5149,9 @@ static int ext4_do_update_inode(handle_t *handle,
 	ext4_clear_inode_state(inode, EXT4_STATE_NEW);
 	if (set_large_file) {
 		BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get write access");
-		err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh);
+		err = ext4_journal_get_write_access(handle, sb,
+						    EXT4_SB(sb)->s_sbh,
+						    EXT4_JTR_NONE);
 		if (err)
 			goto out_brelse;
 		lock_buffer(EXT4_SB(sb)->s_sbh);
@@ -5743,7 +5752,8 @@ ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
 	err = ext4_get_inode_loc(inode, iloc);
 	if (!err) {
 		BUFFER_TRACE(iloc->bh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, iloc->bh);
+		err = ext4_journal_get_write_access(handle, inode->i_sb,
+						    iloc->bh, EXT4_JTR_NONE);
 		if (err) {
 			brelse(iloc->bh);
 			iloc->bh = NULL;
@@ -5866,7 +5876,8 @@ int ext4_expand_extra_isize(struct inode *inode,
 	ext4_write_lock_xattr(inode, &no_expand);
 
 	BUFFER_TRACE(iloc->bh, "get_write_access");
-	error = ext4_journal_get_write_access(handle, iloc->bh);
+	error = ext4_journal_get_write_access(handle, inode->i_sb, iloc->bh,
+					      EXT4_JTR_NONE);
 	if (error) {
 		brelse(iloc->bh);
 		goto out_unlock;
@@ -6037,7 +6048,8 @@ int ext4_change_inode_journal_flag(struct inode *inode, int val)
 	return err;
 }
 
-static int ext4_bh_unmapped(handle_t *handle, struct buffer_head *bh)
+static int ext4_bh_unmapped(handle_t *handle, struct inode *inode,
+			    struct buffer_head *bh)
 {
 	return !buffer_mapped(bh);
 }
@@ -6110,7 +6122,7 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
 	 * inode to the transaction's list to writeprotect pages on commit.
 	 */
 	if (page_has_buffers(page)) {
-		if (!ext4_walk_page_buffers(NULL, page_buffers(page),
+		if (!ext4_walk_page_buffers(NULL, inode, page_buffers(page),
 					    0, len, NULL,
 					    ext4_bh_unmapped)) {
 			/* Wait so that we don't change page under IO */
@@ -6156,11 +6168,13 @@ vm_fault_t ext4_page_mkwrite(struct vm_fault *vmf)
 		err = __block_write_begin(page, 0, len, ext4_get_block);
 		if (!err) {
 			ret = VM_FAULT_SIGBUS;
-			if (ext4_walk_page_buffers(handle, page_buffers(page),
-					0, len, NULL, do_journal_get_write_access))
+			if (ext4_walk_page_buffers(handle, inode,
+					page_buffers(page), 0, len, NULL,
+					do_journal_get_write_access))
 				goto out_error;
-			if (ext4_walk_page_buffers(handle, page_buffers(page),
-					0, len, NULL, write_end_fn))
+			if (ext4_walk_page_buffers(handle, inode,
+					page_buffers(page), 0, len, NULL,
+					write_end_fn))
 				goto out_error;
 			if (ext4_jbd2_inode_add_write(handle, inode,
 						      page_offset(page), len))
diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
index 31627f7dc5cd..752907cddbce 100644
--- a/fs/ext4/ioctl.c
+++ b/fs/ext4/ioctl.c
@@ -1104,7 +1104,9 @@ static long __ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 				err = PTR_ERR(handle);
 				goto pwsalt_err_exit;
 			}
-			err = ext4_journal_get_write_access(handle, sbi->s_sbh);
+			err = ext4_journal_get_write_access(handle, sb,
+							    sbi->s_sbh,
+							    EXT4_JTR_NONE);
 			if (err)
 				goto pwsalt_err_journal;
 			lock_buffer(sbi->s_sbh);
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index c2c22c2baac0..2e591b4c0d65 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -3725,7 +3725,8 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
 	}
 
 	BUFFER_TRACE(bitmap_bh, "getting write access");
-	err = ext4_journal_get_write_access(handle, bitmap_bh);
+	err = ext4_journal_get_write_access(handle, sb, bitmap_bh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto out_err;
 
@@ -3738,7 +3739,7 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
 			ext4_free_group_clusters(sb, gdp));
 
 	BUFFER_TRACE(gdp_bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, gdp_bh);
+	err = ext4_journal_get_write_access(handle, sb, gdp_bh, EXT4_JTR_NONE);
 	if (err)
 		goto out_err;
 
@@ -5915,7 +5916,8 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
 	}
 
 	BUFFER_TRACE(bitmap_bh, "getting write access");
-	err = ext4_journal_get_write_access(handle, bitmap_bh);
+	err = ext4_journal_get_write_access(handle, sb, bitmap_bh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto error_return;
 
@@ -5925,7 +5927,7 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
 	 * using it
 	 */
 	BUFFER_TRACE(gd_bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, gd_bh);
+	err = ext4_journal_get_write_access(handle, sb, gd_bh, EXT4_JTR_NONE);
 	if (err)
 		goto error_return;
 #ifdef AGGRESSIVE_CHECK
@@ -6106,7 +6108,8 @@ int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
 	}
 
 	BUFFER_TRACE(bitmap_bh, "getting write access");
-	err = ext4_journal_get_write_access(handle, bitmap_bh);
+	err = ext4_journal_get_write_access(handle, sb, bitmap_bh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto error_return;
 
@@ -6116,7 +6119,7 @@ int ext4_group_add_blocks(handle_t *handle, struct super_block *sb,
 	 * using it
 	 */
 	BUFFER_TRACE(gd_bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, gd_bh);
+	err = ext4_journal_get_write_access(handle, sb, gd_bh, EXT4_JTR_NONE);
 	if (err)
 		goto error_return;
 
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index a4af26d4459a..d555ffd3138c 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -70,7 +70,8 @@ static struct buffer_head *ext4_append(handle_t *handle,
 	inode->i_size += inode->i_sb->s_blocksize;
 	EXT4_I(inode)->i_disksize = inode->i_size;
 	BUFFER_TRACE(bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, bh);
+	err = ext4_journal_get_write_access(handle, inode->i_sb, bh,
+					    EXT4_JTR_NONE);
 	if (err) {
 		brelse(bh);
 		ext4_std_error(inode->i_sb, err);
@@ -1927,12 +1928,14 @@ static struct ext4_dir_entry_2 *do_split(handle_t *handle, struct inode *dir,
 	}
 
 	BUFFER_TRACE(*bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, *bh);
+	err = ext4_journal_get_write_access(handle, dir->i_sb, *bh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto journal_error;
 
 	BUFFER_TRACE(frame->bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, frame->bh);
+	err = ext4_journal_get_write_access(handle, dir->i_sb, frame->bh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto journal_error;
 
@@ -2109,7 +2112,8 @@ static int add_dirent_to_buf(handle_t *handle, struct ext4_filename *fname,
 			return err;
 	}
 	BUFFER_TRACE(bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, bh);
+	err = ext4_journal_get_write_access(handle, dir->i_sb, bh,
+					    EXT4_JTR_NONE);
 	if (err) {
 		ext4_std_error(dir->i_sb, err);
 		return err;
@@ -2167,7 +2171,8 @@ static int make_indexed_dir(handle_t *handle, struct ext4_filename *fname,
 	blocksize =  dir->i_sb->s_blocksize;
 	dxtrace(printk(KERN_DEBUG "Creating index: inode %lu\n", dir->i_ino));
 	BUFFER_TRACE(bh, "get_write_access");
-	retval = ext4_journal_get_write_access(handle, bh);
+	retval = ext4_journal_get_write_access(handle, dir->i_sb, bh,
+					       EXT4_JTR_NONE);
 	if (retval) {
 		ext4_std_error(dir->i_sb, retval);
 		brelse(bh);
@@ -2419,7 +2424,7 @@ static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname,
 	}
 
 	BUFFER_TRACE(bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, bh);
+	err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
 	if (err)
 		goto journal_error;
 
@@ -2476,7 +2481,8 @@ static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname,
 		node2->fake.rec_len = ext4_rec_len_to_disk(sb->s_blocksize,
 							   sb->s_blocksize);
 		BUFFER_TRACE(frame->bh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, frame->bh);
+		err = ext4_journal_get_write_access(handle, sb, frame->bh,
+						    EXT4_JTR_NONE);
 		if (err)
 			goto journal_error;
 		if (!add_level) {
@@ -2486,8 +2492,9 @@ static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname,
 				       icount1, icount2));
 
 			BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */
-			err = ext4_journal_get_write_access(handle,
-							     (frame - 1)->bh);
+			err = ext4_journal_get_write_access(handle, sb,
+							    (frame - 1)->bh,
+							    EXT4_JTR_NONE);
 			if (err)
 				goto journal_error;
 
@@ -2636,7 +2643,8 @@ static int ext4_delete_entry(handle_t *handle,
 		csum_size = sizeof(struct ext4_dir_entry_tail);
 
 	BUFFER_TRACE(bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, bh);
+	err = ext4_journal_get_write_access(handle, dir->i_sb, bh,
+					    EXT4_JTR_NONE);
 	if (unlikely(err))
 		goto out;
 
@@ -3088,7 +3096,8 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
 		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
 
 	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, sbi->s_sbh);
+	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto out;
 
@@ -3186,7 +3195,8 @@ int ext4_orphan_del(handle_t *handle, struct inode *inode)
 	if (prev == &sbi->s_orphan) {
 		jbd_debug(4, "superblock will point to %u\n", ino_next);
 		BUFFER_TRACE(sbi->s_sbh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, sbi->s_sbh);
+		err = ext4_journal_get_write_access(handle, inode->i_sb,
+						    sbi->s_sbh, EXT4_JTR_NONE);
 		if (err) {
 			mutex_unlock(&sbi->s_orphan_lock);
 			goto out_brelse;
@@ -3675,7 +3685,8 @@ static int ext4_rename_dir_prepare(handle_t *handle, struct ext4_renament *ent)
 	if (le32_to_cpu(ent->parent_de->inode) != ent->dir->i_ino)
 		return -EFSCORRUPTED;
 	BUFFER_TRACE(ent->dir_bh, "get_write_access");
-	return ext4_journal_get_write_access(handle, ent->dir_bh);
+	return ext4_journal_get_write_access(handle, ent->dir->i_sb,
+					     ent->dir_bh, EXT4_JTR_NONE);
 }
 
 static int ext4_rename_dir_finish(handle_t *handle, struct ext4_renament *ent,
@@ -3710,7 +3721,8 @@ static int ext4_setent(handle_t *handle, struct ext4_renament *ent,
 	int retval, retval2;
 
 	BUFFER_TRACE(ent->bh, "get write access");
-	retval = ext4_journal_get_write_access(handle, ent->bh);
+	retval = ext4_journal_get_write_access(handle, ent->dir->i_sb, ent->bh,
+					       EXT4_JTR_NONE);
 	if (retval)
 		return retval;
 	ent->de->inode = cpu_to_le32(ino);
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index bd0d185654f3..8649c0e2ed9a 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -404,7 +404,8 @@ static struct buffer_head *bclean(handle_t *handle, struct super_block *sb,
 	if (unlikely(!bh))
 		return ERR_PTR(-ENOMEM);
 	BUFFER_TRACE(bh, "get_write_access");
-	if ((err = ext4_journal_get_write_access(handle, bh))) {
+	err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
+	if (err) {
 		brelse(bh);
 		bh = ERR_PTR(err);
 	} else {
@@ -469,7 +470,8 @@ static int set_flexbg_block_bitmap(struct super_block *sb, handle_t *handle,
 			return -ENOMEM;
 
 		BUFFER_TRACE(bh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, bh);
+		err = ext4_journal_get_write_access(handle, sb, bh,
+						    EXT4_JTR_NONE);
 		if (err) {
 			brelse(bh);
 			return err;
@@ -564,7 +566,8 @@ static int setup_new_flex_group_blocks(struct super_block *sb,
 			}
 
 			BUFFER_TRACE(gdb, "get_write_access");
-			err = ext4_journal_get_write_access(handle, gdb);
+			err = ext4_journal_get_write_access(handle, sb, gdb,
+							    EXT4_JTR_NONE);
 			if (err) {
 				brelse(gdb);
 				goto out;
@@ -832,17 +835,18 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
 	}
 
 	BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh);
+	err = ext4_journal_get_write_access(handle, sb, EXT4_SB(sb)->s_sbh,
+					    EXT4_JTR_NONE);
 	if (unlikely(err))
 		goto errout;
 
 	BUFFER_TRACE(gdb_bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, gdb_bh);
+	err = ext4_journal_get_write_access(handle, sb, gdb_bh, EXT4_JTR_NONE);
 	if (unlikely(err))
 		goto errout;
 
 	BUFFER_TRACE(dind, "get_write_access");
-	err = ext4_journal_get_write_access(handle, dind);
+	err = ext4_journal_get_write_access(handle, sb, dind, EXT4_JTR_NONE);
 	if (unlikely(err)) {
 		ext4_std_error(sb, err);
 		goto errout;
@@ -951,7 +955,7 @@ static int add_new_gdb_meta_bg(struct super_block *sb,
 	n_group_desc[gdb_num] = gdb_bh;
 
 	BUFFER_TRACE(gdb_bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, gdb_bh);
+	err = ext4_journal_get_write_access(handle, sb, gdb_bh, EXT4_JTR_NONE);
 	if (err) {
 		kvfree(n_group_desc);
 		brelse(gdb_bh);
@@ -1037,7 +1041,8 @@ static int reserve_backup_gdb(handle_t *handle, struct inode *inode,
 
 	for (i = 0; i < reserved_gdb; i++) {
 		BUFFER_TRACE(primary[i], "get_write_access");
-		if ((err = ext4_journal_get_write_access(handle, primary[i])))
+		if ((err = ext4_journal_get_write_access(handle, sb, primary[i],
+							 EXT4_JTR_NONE)))
 			goto exit_bh;
 	}
 
@@ -1144,10 +1149,9 @@ static void update_backups(struct super_block *sb, sector_t blk_off, char *data,
 			   backup_block, backup_block -
 			   ext4_group_first_block_no(sb, group));
 		BUFFER_TRACE(bh, "get_write_access");
-		if ((err = ext4_journal_get_write_access(handle, bh))) {
-			brelse(bh);
+		if ((err = ext4_journal_get_write_access(handle, sb, bh,
+							 EXT4_JTR_NONE)))
 			break;
-		}
 		lock_buffer(bh);
 		memcpy(bh->b_data, data, size);
 		if (rest)
@@ -1227,7 +1231,8 @@ static int ext4_add_new_descs(handle_t *handle, struct super_block *sb,
 			gdb_bh = sbi_array_rcu_deref(sbi, s_group_desc,
 						     gdb_num);
 			BUFFER_TRACE(gdb_bh, "get_write_access");
-			err = ext4_journal_get_write_access(handle, gdb_bh);
+			err = ext4_journal_get_write_access(handle, sb, gdb_bh,
+							    EXT4_JTR_NONE);
 
 			if (!err && reserved_gdb && ext4_bg_num_gdb(sb, group))
 				err = reserve_backup_gdb(handle, resize_inode, group);
@@ -1504,7 +1509,8 @@ static int ext4_flex_group_add(struct super_block *sb,
 	}
 
 	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, sbi->s_sbh);
+	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto exit_journal;
 
@@ -1717,7 +1723,8 @@ static int ext4_group_extend_no_check(struct super_block *sb,
 	}
 
 	BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh);
+	err = ext4_journal_get_write_access(handle, sb, EXT4_SB(sb)->s_sbh,
+					    EXT4_JTR_NONE);
 	if (err) {
 		ext4_warning(sb, "error %d on journal write access", err);
 		goto errout;
@@ -1879,7 +1886,8 @@ static int ext4_convert_meta_bg(struct super_block *sb, struct inode *inode)
 		return PTR_ERR(handle);
 
 	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, sbi->s_sbh);
+	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
+					    EXT4_JTR_NONE);
 	if (err)
 		goto errout;
 
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index d29f6aa7d96e..d12982ca923b 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -4018,6 +4018,20 @@ static const char *ext4_quota_mode(struct super_block *sb)
 #endif
 }
 
+static void ext4_setup_csum_trigger(struct super_block *sb,
+				    enum ext4_journal_trigger_type type,
+				    void (*trigger)(
+					struct jbd2_buffer_trigger_type *type,
+					struct buffer_head *bh,
+					void *mapped_data,
+					size_t size))
+{
+	struct ext4_sb_info *sbi = EXT4_SB(sb);
+
+	sbi->s_journal_triggers[type].sb = sb;
+	sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
+}
+
 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 {
 	struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
@@ -6610,7 +6624,7 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type,
 	if (!bh)
 		goto out;
 	BUFFER_TRACE(bh, "get write access");
-	err = ext4_journal_get_write_access(handle, bh);
+	err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
 	if (err) {
 		brelse(bh);
 		return err;
diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index 10ba4b24a0aa..af0af30b8123 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -791,7 +791,8 @@ static void ext4_xattr_update_super_block(handle_t *handle,
 		return;
 
 	BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
-	if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
+	if (ext4_journal_get_write_access(handle, sb, EXT4_SB(sb)->s_sbh,
+					  EXT4_JTR_NONE) == 0) {
 		lock_buffer(EXT4_SB(sb)->s_sbh);
 		ext4_set_feature_xattr(sb);
 		ext4_superblock_csum_set(sb);
@@ -1169,7 +1170,8 @@ ext4_xattr_inode_dec_ref_all(handle_t *handle, struct inode *parent,
 			continue;
 		}
 		if (err > 0) {
-			err = ext4_journal_get_write_access(handle, bh);
+			err = ext4_journal_get_write_access(handle,
+					parent->i_sb, bh, EXT4_JTR_NONE);
 			if (err) {
 				ext4_warning_inode(ea_inode,
 						"Re-get write access err=%d",
@@ -1230,7 +1232,8 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode,
 	int error = 0;
 
 	BUFFER_TRACE(bh, "get_write_access");
-	error = ext4_journal_get_write_access(handle, bh);
+	error = ext4_journal_get_write_access(handle, inode->i_sb, bh,
+					      EXT4_JTR_NONE);
 	if (error)
 		goto out;
 
@@ -1371,7 +1374,8 @@ static int ext4_xattr_inode_write(handle_t *handle, struct inode *ea_inode,
 					 "ext4_getblk() return bh = NULL");
 			return -EFSCORRUPTED;
 		}
-		ret = ext4_journal_get_write_access(handle, bh);
+		ret = ext4_journal_get_write_access(handle, ea_inode->i_sb, bh,
+						   EXT4_JTR_NONE);
 		if (ret)
 			goto out;
 
@@ -1855,7 +1859,8 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 
 	if (s->base) {
 		BUFFER_TRACE(bs->bh, "get_write_access");
-		error = ext4_journal_get_write_access(handle, bs->bh);
+		error = ext4_journal_get_write_access(handle, sb, bs->bh,
+						      EXT4_JTR_NONE);
 		if (error)
 			goto cleanup;
 		lock_buffer(bs->bh);
@@ -1987,8 +1992,9 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 				if (error)
 					goto cleanup;
 				BUFFER_TRACE(new_bh, "get_write_access");
-				error = ext4_journal_get_write_access(handle,
-								      new_bh);
+				error = ext4_journal_get_write_access(
+						handle, sb, new_bh,
+						EXT4_JTR_NONE);
 				if (error)
 					goto cleanup_dquot;
 				lock_buffer(new_bh);
@@ -2092,7 +2098,8 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 			}
 
 			lock_buffer(new_bh);
-			error = ext4_journal_get_create_access(handle, new_bh);
+			error = ext4_journal_get_create_access(handle, sb,
+							new_bh, EXT4_JTR_NONE);
 			if (error) {
 				unlock_buffer(new_bh);
 				error = -EIO;
@@ -2872,7 +2879,8 @@ int ext4_xattr_delete_inode(handle_t *handle, struct inode *inode,
 			goto cleanup;
 		}
 
-		error = ext4_journal_get_write_access(handle, iloc.bh);
+		error = ext4_journal_get_write_access(handle, inode->i_sb,
+						iloc.bh, EXT4_JTR_NONE);
 		if (error) {
 			EXT4_ERROR_INODE(inode, "write access (error %d)",
 					 error);
diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c
index e8fc45fd751f..6fbbe064dd2a 100644
--- a/fs/jbd2/transaction.c
+++ b/fs/jbd2/transaction.c
@@ -1404,7 +1404,7 @@ void jbd2_journal_set_triggers(struct buffer_head *bh,
 {
 	struct journal_head *jh = jbd2_journal_grab_journal_head(bh);
 
-	if (WARN_ON(!jh))
+	if (WARN_ON_ONCE(!jh))
 		return;
 	jh->b_triggers = type;
 	jbd2_journal_put_journal_head(jh);
-- 
2.26.2

Re: [PATCH 1/4] ext4: Support for checksumming from journal triggers

[Andreas Dilger]

[-- Attachment #1: Type: text/plain, Size: 3266 bytes --]

On Jun 16, 2021, at 4:56 AM, Jan Kara <jack@suse.cz> wrote:
> 
> JBD2 layer support triggers which are called when journaling layer moves
> buffer to a certain state. We can use the frozen trigger, which gets
> called when buffer data is frozen and about to be written out to the
> journal, to compute block checksums for some buffer types (similarly as
> does ocfs2). This avoids unnecessary repeated recomputation of the
> checksum (at the cost of larger window where memory corruption won't be
> caught by checksumming) and is even necessary when there are
> unsynchronized updaters of the checksummed data.
> 
> So add argument to ext4_journal_get_write_access() and
> ext4_journal_get_create_access() which describes buffer type so that
> triggers can be set accordingly. This patch is mostly only a change of
> prototype of the above mentioned functions and a few small helpers. Real
> checksumming will come later.
> 
> Signed-off-by: Jan Kara <jack@suse.cz>
> ---

Comment inline.

> 
> diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c
> index be799040a415..f601e24b6015 100644
> --- a/fs/ext4/ext4_jbd2.c
> +++ b/fs/ext4/ext4_jbd2.c
> @@ -229,11 +231,18 @@ int __ext4_journal_get_write_access(const char *where, unsigned int line,
> 
> 	if (ext4_handle_valid(handle)) {
> 		err = jbd2_journal_get_write_access(handle, bh);
> -		if (err)
> +		if (err) {
> 			ext4_journal_abort_handle(where, line, __func__, bh,
> 						  handle, err);
> +			return err;
> +		}
> 	}
> -	return err;
> +	if (trigger_type == EXT4_JTR_NONE || !ext4_has_metadata_csum(sb))
> +		return 0;
> +	WARN_ON_ONCE(trigger_type >= EXT4_JOURNAL_TRIGGER_COUNT);

I'm not sure WARN_ON_ONCE() is enough here.  This would essentially result
in executing a random (or maybe NULL) function pointer later on.  Either
trigger_type should be checked early and return an error, or this should
be a BUG_ON() so that the crash happens here instead of in jbd context.

> +	jbd2_journal_set_triggers(bh,
> +		&EXT4_SB(sb)->s_journal_triggers[trigger_type].tr_triggers);
> +	return 0;
> }
> 
> /*
> @@ -304,17 +313,27 @@ int __ext4_forget(const char *where, unsigned int line,
> int __ext4_journal_get_create_access(const char *where, unsigned int line,
> -				handle_t *handle, struct buffer_head *bh)
> +				handle_t *handle, struct super_block *sb,
> +				struct buffer_head *bh,
> +				enum ext4_journal_trigger_type trigger_type)
> {
> -	int err = 0;
> +	int err;
> 
> -	if (ext4_handle_valid(handle)) {
> -		err = jbd2_journal_get_create_access(handle, bh);
> -		if (err)
> -			ext4_journal_abort_handle(where, line, __func__,
> -						  bh, handle, err);
> +	if (!ext4_handle_valid(handle))
> +		return 0;
> +
> +	err = jbd2_journal_get_create_access(handle, bh);
> +	if (err) {
> +		ext4_journal_abort_handle(where, line, __func__, bh, handle,
> +					  err);
> +		return err;
> 	}
> -	return err;
> +	if (trigger_type == EXT4_JTR_NONE || !ext4_has_metadata_csum(sb))
> +		return 0;
> +	WARN_ON_ONCE(trigger_type >= EXT4_JOURNAL_TRIGGER_COUNT);

Same.

> +	jbd2_journal_set_triggers(bh,
> +		&EXT4_SB(sb)->s_journal_triggers[trigger_type].tr_triggers);
> +	return 0;
> }


Cheers, Andreas






[-- Attachment #2: Message signed with OpenPGP --]
[-- Type: application/pgp-signature, Size: 873 bytes --]

Re: [PATCH 1/4] ext4: Support for checksumming from journal triggers

[Jan Kara]

On Wed 16-06-21 13:56:30, Andreas Dilger wrote:
> On Jun 16, 2021, at 4:56 AM, Jan Kara <jack@suse.cz> wrote:
> > 
> > JBD2 layer support triggers which are called when journaling layer moves
> > buffer to a certain state. We can use the frozen trigger, which gets
> > called when buffer data is frozen and about to be written out to the
> > journal, to compute block checksums for some buffer types (similarly as
> > does ocfs2). This avoids unnecessary repeated recomputation of the
> > checksum (at the cost of larger window where memory corruption won't be
> > caught by checksumming) and is even necessary when there are
> > unsynchronized updaters of the checksummed data.
> > 
> > So add argument to ext4_journal_get_write_access() and
> > ext4_journal_get_create_access() which describes buffer type so that
> > triggers can be set accordingly. This patch is mostly only a change of
> > prototype of the above mentioned functions and a few small helpers. Real
> > checksumming will come later.
> > 
> > Signed-off-by: Jan Kara <jack@suse.cz>
> > ---
> 
> Comment inline.
> 
> > 
> > diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c
> > index be799040a415..f601e24b6015 100644
> > --- a/fs/ext4/ext4_jbd2.c
> > +++ b/fs/ext4/ext4_jbd2.c
> > @@ -229,11 +231,18 @@ int __ext4_journal_get_write_access(const char *where, unsigned int line,
> > 
> > 	if (ext4_handle_valid(handle)) {
> > 		err = jbd2_journal_get_write_access(handle, bh);
> > -		if (err)
> > +		if (err) {
> > 			ext4_journal_abort_handle(where, line, __func__, bh,
> > 						  handle, err);
> > +			return err;
> > +		}
> > 	}
> > -	return err;
> > +	if (trigger_type == EXT4_JTR_NONE || !ext4_has_metadata_csum(sb))
> > +		return 0;
> > +	WARN_ON_ONCE(trigger_type >= EXT4_JOURNAL_TRIGGER_COUNT);
> 
> I'm not sure WARN_ON_ONCE() is enough here.  This would essentially result
> in executing a random (or maybe NULL) function pointer later on.  Either
> trigger_type should be checked early and return an error, or this should
> be a BUG_ON() so that the crash happens here instead of in jbd context.

Good point, I'll fix that.

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

[PATCH 2/4] ext4: Move orphan inode handling into a separate file

[Jan Kara]

Move functions for handling orphan inodes into a new file
fs/ext4/orphan.c to have them in one place and somewhat reduce size of
other files. No code changes.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/Makefile |   2 +-
 fs/ext4/ext4.h   |  11 +-
 fs/ext4/namei.c  | 182 ------------------------
 fs/ext4/orphan.c | 352 +++++++++++++++++++++++++++++++++++++++++++++++
 fs/ext4/super.c  | 173 +----------------------
 5 files changed, 364 insertions(+), 356 deletions(-)
 create mode 100644 fs/ext4/orphan.c

diff --git a/fs/ext4/Makefile b/fs/ext4/Makefile
index 49e7af6cc93f..7d89142e1421 100644
--- a/fs/ext4/Makefile
+++ b/fs/ext4/Makefile
@@ -10,7 +10,7 @@ ext4-y	:= balloc.o bitmap.o block_validity.o dir.o ext4_jbd2.o extents.o \
 		indirect.o inline.o inode.o ioctl.o mballoc.o migrate.o \
 		mmp.o move_extent.o namei.o page-io.o readpage.o resize.o \
 		super.o symlink.o sysfs.o xattr.o xattr_hurd.o xattr_trusted.o \
-		xattr_user.o fast_commit.o
+		xattr_user.o fast_commit.o orphan.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 b81256a7e7f2..33508487516f 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -2158,6 +2158,8 @@ static inline bool ext4_has_incompat_features(struct super_block *sb)
 	return (EXT4_SB(sb)->s_es->s_feature_incompat != 0);
 }
 
+extern int ext4_feature_set_ok(struct super_block *sb, int readonly);
+
 /*
  * Superblock flags
  */
@@ -3018,8 +3020,6 @@ extern int ext4_init_new_dir(handle_t *handle, struct inode *dir,
 			     struct inode *inode);
 extern int ext4_dirblock_csum_verify(struct inode *inode,
 				     struct buffer_head *bh);
-extern int ext4_orphan_add(handle_t *, struct inode *);
-extern int ext4_orphan_del(handle_t *, struct inode *);
 extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
 				__u32 start_minor_hash, __u32 *next_hash);
 extern int ext4_search_dir(struct buffer_head *bh,
@@ -3488,6 +3488,7 @@ static inline bool ext4_is_quota_journalled(struct super_block *sb)
 	return (ext4_has_feature_quota(sb) ||
 		sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]);
 }
+int ext4_enable_quotas(struct super_block *sb);
 #endif
 
 /*
@@ -3745,6 +3746,12 @@ extern int ext4_multi_mount_protect(struct super_block *, ext4_fsblk_t);
 /* verity.c */
 extern const struct fsverity_operations ext4_verityops;
 
+/* orphan.c */
+extern int ext4_orphan_add(handle_t *, struct inode *);
+extern int ext4_orphan_del(handle_t *, struct inode *);
+extern void ext4_orphan_cleanup(struct super_block *sb,
+				struct ext4_super_block *es);
+
 /*
  * Add new method to test whether block and inode bitmaps are properly
  * initialized. With uninit_bg reading the block from disk is not enough
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index d555ffd3138c..62b34b9f56f5 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -3054,188 +3054,6 @@ bool ext4_empty_dir(struct inode *inode)
 	return true;
 }
 
-/*
- * ext4_orphan_add() links an unlinked or truncated inode into a list of
- * such inodes, starting at the superblock, in case we crash before the
- * file is closed/deleted, or in case the inode truncate spans multiple
- * transactions and the last transaction is not recovered after a crash.
- *
- * At filesystem recovery time, we walk this list deleting unlinked
- * inodes and truncating linked inodes in ext4_orphan_cleanup().
- *
- * Orphan list manipulation functions must be called under i_mutex unless
- * we are just creating the inode or deleting it.
- */
-int ext4_orphan_add(handle_t *handle, struct inode *inode)
-{
-	struct super_block *sb = inode->i_sb;
-	struct ext4_sb_info *sbi = EXT4_SB(sb);
-	struct ext4_iloc iloc;
-	int err = 0, rc;
-	bool dirty = false;
-
-	if (!sbi->s_journal || is_bad_inode(inode))
-		return 0;
-
-	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
-		     !inode_is_locked(inode));
-	/*
-	 * Exit early if inode already is on orphan list. This is a big speedup
-	 * since we don't have to contend on the global s_orphan_lock.
-	 */
-	if (!list_empty(&EXT4_I(inode)->i_orphan))
-		return 0;
-
-	/*
-	 * Orphan handling is only valid for files with data blocks
-	 * being truncated, or files being unlinked. Note that we either
-	 * hold i_mutex, or the inode can not be referenced from outside,
-	 * so i_nlink should not be bumped due to race
-	 */
-	ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
-		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
-
-	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
-					    EXT4_JTR_NONE);
-	if (err)
-		goto out;
-
-	err = ext4_reserve_inode_write(handle, inode, &iloc);
-	if (err)
-		goto out;
-
-	mutex_lock(&sbi->s_orphan_lock);
-	/*
-	 * Due to previous errors inode may be already a part of on-disk
-	 * orphan list. If so skip on-disk list modification.
-	 */
-	if (!NEXT_ORPHAN(inode) || NEXT_ORPHAN(inode) >
-	    (le32_to_cpu(sbi->s_es->s_inodes_count))) {
-		/* Insert this inode at the head of the on-disk orphan list */
-		NEXT_ORPHAN(inode) = le32_to_cpu(sbi->s_es->s_last_orphan);
-		lock_buffer(sbi->s_sbh);
-		sbi->s_es->s_last_orphan = cpu_to_le32(inode->i_ino);
-		ext4_superblock_csum_set(sb);
-		unlock_buffer(sbi->s_sbh);
-		dirty = true;
-	}
-	list_add(&EXT4_I(inode)->i_orphan, &sbi->s_orphan);
-	mutex_unlock(&sbi->s_orphan_lock);
-
-	if (dirty) {
-		err = ext4_handle_dirty_metadata(handle, NULL, sbi->s_sbh);
-		rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
-		if (!err)
-			err = rc;
-		if (err) {
-			/*
-			 * We have to remove inode from in-memory list if
-			 * addition to on disk orphan list failed. Stray orphan
-			 * list entries can cause panics at unmount time.
-			 */
-			mutex_lock(&sbi->s_orphan_lock);
-			list_del_init(&EXT4_I(inode)->i_orphan);
-			mutex_unlock(&sbi->s_orphan_lock);
-		}
-	} else
-		brelse(iloc.bh);
-
-	jbd_debug(4, "superblock will point to %lu\n", inode->i_ino);
-	jbd_debug(4, "orphan inode %lu will point to %d\n",
-			inode->i_ino, NEXT_ORPHAN(inode));
-out:
-	ext4_std_error(sb, err);
-	return err;
-}
-
-/*
- * ext4_orphan_del() removes an unlinked or truncated inode from the list
- * of such inodes stored on disk, because it is finally being cleaned up.
- */
-int ext4_orphan_del(handle_t *handle, struct inode *inode)
-{
-	struct list_head *prev;
-	struct ext4_inode_info *ei = EXT4_I(inode);
-	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
-	__u32 ino_next;
-	struct ext4_iloc iloc;
-	int err = 0;
-
-	if (!sbi->s_journal && !(sbi->s_mount_state & EXT4_ORPHAN_FS))
-		return 0;
-
-	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
-		     !inode_is_locked(inode));
-	/* Do this quick check before taking global s_orphan_lock. */
-	if (list_empty(&ei->i_orphan))
-		return 0;
-
-	if (handle) {
-		/* Grab inode buffer early before taking global s_orphan_lock */
-		err = ext4_reserve_inode_write(handle, inode, &iloc);
-	}
-
-	mutex_lock(&sbi->s_orphan_lock);
-	jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino);
-
-	prev = ei->i_orphan.prev;
-	list_del_init(&ei->i_orphan);
-
-	/* If we're on an error path, we may not have a valid
-	 * transaction handle with which to update the orphan list on
-	 * disk, but we still need to remove the inode from the linked
-	 * list in memory. */
-	if (!handle || err) {
-		mutex_unlock(&sbi->s_orphan_lock);
-		goto out_err;
-	}
-
-	ino_next = NEXT_ORPHAN(inode);
-	if (prev == &sbi->s_orphan) {
-		jbd_debug(4, "superblock will point to %u\n", ino_next);
-		BUFFER_TRACE(sbi->s_sbh, "get_write_access");
-		err = ext4_journal_get_write_access(handle, inode->i_sb,
-						    sbi->s_sbh, EXT4_JTR_NONE);
-		if (err) {
-			mutex_unlock(&sbi->s_orphan_lock);
-			goto out_brelse;
-		}
-		lock_buffer(sbi->s_sbh);
-		sbi->s_es->s_last_orphan = cpu_to_le32(ino_next);
-		ext4_superblock_csum_set(inode->i_sb);
-		unlock_buffer(sbi->s_sbh);
-		mutex_unlock(&sbi->s_orphan_lock);
-		err = ext4_handle_dirty_metadata(handle, NULL, sbi->s_sbh);
-	} else {
-		struct ext4_iloc iloc2;
-		struct inode *i_prev =
-			&list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode;
-
-		jbd_debug(4, "orphan inode %lu will point to %u\n",
-			  i_prev->i_ino, ino_next);
-		err = ext4_reserve_inode_write(handle, i_prev, &iloc2);
-		if (err) {
-			mutex_unlock(&sbi->s_orphan_lock);
-			goto out_brelse;
-		}
-		NEXT_ORPHAN(i_prev) = ino_next;
-		err = ext4_mark_iloc_dirty(handle, i_prev, &iloc2);
-		mutex_unlock(&sbi->s_orphan_lock);
-	}
-	if (err)
-		goto out_brelse;
-	NEXT_ORPHAN(inode) = 0;
-	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
-out_err:
-	ext4_std_error(inode->i_sb, err);
-	return err;
-
-out_brelse:
-	brelse(iloc.bh);
-	goto out_err;
-}
-
 static int ext4_rmdir(struct inode *dir, struct dentry *dentry)
 {
 	int retval;
diff --git a/fs/ext4/orphan.c b/fs/ext4/orphan.c
new file mode 100644
index 000000000000..732b16ef655b
--- /dev/null
+++ b/fs/ext4/orphan.c
@@ -0,0 +1,352 @@
+/*
+ * Ext4 orphan inode handling
+ */
+#include <linux/fs.h>
+#include <linux/quotaops.h>
+#include <linux/buffer_head.h>
+
+#include "ext4.h"
+#include "ext4_jbd2.h"
+
+/*
+ * ext4_orphan_add() links an unlinked or truncated inode into a list of
+ * such inodes, starting at the superblock, in case we crash before the
+ * file is closed/deleted, or in case the inode truncate spans multiple
+ * transactions and the last transaction is not recovered after a crash.
+ *
+ * At filesystem recovery time, we walk this list deleting unlinked
+ * inodes and truncating linked inodes in ext4_orphan_cleanup().
+ *
+ * Orphan list manipulation functions must be called under i_mutex unless
+ * we are just creating the inode or deleting it.
+ */
+int ext4_orphan_add(handle_t *handle, struct inode *inode)
+{
+	struct super_block *sb = inode->i_sb;
+	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	struct ext4_iloc iloc;
+	int err = 0, rc;
+	bool dirty = false;
+
+	if (!sbi->s_journal || is_bad_inode(inode))
+		return 0;
+
+	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
+		     !inode_is_locked(inode));
+	/*
+	 * Exit early if inode already is on orphan list. This is a big speedup
+	 * since we don't have to contend on the global s_orphan_lock.
+	 */
+	if (!list_empty(&EXT4_I(inode)->i_orphan))
+		return 0;
+
+	/*
+	 * Orphan handling is only valid for files with data blocks
+	 * being truncated, or files being unlinked. Note that we either
+	 * hold i_mutex, or the inode can not be referenced from outside,
+	 * so i_nlink should not be bumped due to race
+	 */
+	ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
+		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
+
+	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
+	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
+					    EXT4_JTR_NONE);
+	if (err)
+		goto out;
+
+	err = ext4_reserve_inode_write(handle, inode, &iloc);
+	if (err)
+		goto out;
+
+	mutex_lock(&sbi->s_orphan_lock);
+	/*
+	 * Due to previous errors inode may be already a part of on-disk
+	 * orphan list. If so skip on-disk list modification.
+	 */
+	if (!NEXT_ORPHAN(inode) || NEXT_ORPHAN(inode) >
+	    (le32_to_cpu(sbi->s_es->s_inodes_count))) {
+		/* Insert this inode at the head of the on-disk orphan list */
+		NEXT_ORPHAN(inode) = le32_to_cpu(sbi->s_es->s_last_orphan);
+		lock_buffer(sbi->s_sbh);
+		sbi->s_es->s_last_orphan = cpu_to_le32(inode->i_ino);
+		ext4_superblock_csum_set(sb);
+		unlock_buffer(sbi->s_sbh);
+		dirty = true;
+	}
+	list_add(&EXT4_I(inode)->i_orphan, &sbi->s_orphan);
+	mutex_unlock(&sbi->s_orphan_lock);
+
+	if (dirty) {
+		err = ext4_handle_dirty_metadata(handle, NULL, sbi->s_sbh);
+		rc = ext4_mark_iloc_dirty(handle, inode, &iloc);
+		if (!err)
+			err = rc;
+		if (err) {
+			/*
+			 * We have to remove inode from in-memory list if
+			 * addition to on disk orphan list failed. Stray orphan
+			 * list entries can cause panics at unmount time.
+			 */
+			mutex_lock(&sbi->s_orphan_lock);
+			list_del_init(&EXT4_I(inode)->i_orphan);
+			mutex_unlock(&sbi->s_orphan_lock);
+		}
+	} else
+		brelse(iloc.bh);
+
+	jbd_debug(4, "superblock will point to %lu\n", inode->i_ino);
+	jbd_debug(4, "orphan inode %lu will point to %d\n",
+			inode->i_ino, NEXT_ORPHAN(inode));
+out:
+	ext4_std_error(sb, err);
+	return err;
+}
+
+/*
+ * ext4_orphan_del() removes an unlinked or truncated inode from the list
+ * of such inodes stored on disk, because it is finally being cleaned up.
+ */
+int ext4_orphan_del(handle_t *handle, struct inode *inode)
+{
+	struct list_head *prev;
+	struct ext4_inode_info *ei = EXT4_I(inode);
+	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
+	__u32 ino_next;
+	struct ext4_iloc iloc;
+	int err = 0;
+
+	if (!sbi->s_journal && !(sbi->s_mount_state & EXT4_ORPHAN_FS))
+		return 0;
+
+	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
+		     !inode_is_locked(inode));
+	/* Do this quick check before taking global s_orphan_lock. */
+	if (list_empty(&ei->i_orphan))
+		return 0;
+
+	if (handle) {
+		/* Grab inode buffer early before taking global s_orphan_lock */
+		err = ext4_reserve_inode_write(handle, inode, &iloc);
+	}
+
+	mutex_lock(&sbi->s_orphan_lock);
+	jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino);
+
+	prev = ei->i_orphan.prev;
+	list_del_init(&ei->i_orphan);
+
+	/* If we're on an error path, we may not have a valid
+	 * transaction handle with which to update the orphan list on
+	 * disk, but we still need to remove the inode from the linked
+	 * list in memory. */
+	if (!handle || err) {
+		mutex_unlock(&sbi->s_orphan_lock);
+		goto out_err;
+	}
+
+	ino_next = NEXT_ORPHAN(inode);
+	if (prev == &sbi->s_orphan) {
+		jbd_debug(4, "superblock will point to %u\n", ino_next);
+		BUFFER_TRACE(sbi->s_sbh, "get_write_access");
+		err = ext4_journal_get_write_access(handle, inode->i_sb,
+						    sbi->s_sbh, EXT4_JTR_NONE);
+		if (err) {
+			mutex_unlock(&sbi->s_orphan_lock);
+			goto out_brelse;
+		}
+		lock_buffer(sbi->s_sbh);
+		sbi->s_es->s_last_orphan = cpu_to_le32(ino_next);
+		ext4_superblock_csum_set(inode->i_sb);
+		unlock_buffer(sbi->s_sbh);
+		mutex_unlock(&sbi->s_orphan_lock);
+		err = ext4_handle_dirty_metadata(handle, NULL, sbi->s_sbh);
+	} else {
+		struct ext4_iloc iloc2;
+		struct inode *i_prev =
+			&list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode;
+
+		jbd_debug(4, "orphan inode %lu will point to %u\n",
+			  i_prev->i_ino, ino_next);
+		err = ext4_reserve_inode_write(handle, i_prev, &iloc2);
+		if (err) {
+			mutex_unlock(&sbi->s_orphan_lock);
+			goto out_brelse;
+		}
+		NEXT_ORPHAN(i_prev) = ino_next;
+		err = ext4_mark_iloc_dirty(handle, i_prev, &iloc2);
+		mutex_unlock(&sbi->s_orphan_lock);
+	}
+	if (err)
+		goto out_brelse;
+	NEXT_ORPHAN(inode) = 0;
+	err = ext4_mark_iloc_dirty(handle, inode, &iloc);
+out_err:
+	ext4_std_error(inode->i_sb, err);
+	return err;
+
+out_brelse:
+	brelse(iloc.bh);
+	goto out_err;
+}
+
+static int ext4_quota_on_mount(struct super_block *sb, int type)
+{
+	return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
+					EXT4_SB(sb)->s_jquota_fmt, type);
+}
+
+/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
+ * the superblock) which were deleted from all directories, but held open by
+ * a process at the time of a crash.  We walk the list and try to delete these
+ * inodes at recovery time (only with a read-write filesystem).
+ *
+ * In order to keep the orphan inode chain consistent during traversal (in
+ * case of crash during recovery), we link each inode into the superblock
+ * orphan list_head and handle it the same way as an inode deletion during
+ * normal operation (which journals the operations for us).
+ *
+ * We only do an iget() and an iput() on each inode, which is very safe if we
+ * accidentally point at an in-use or already deleted inode.  The worst that
+ * can happen in this case is that we get a "bit already cleared" message from
+ * ext4_free_inode().  The only reason we would point at a wrong inode is if
+ * e2fsck was run on this filesystem, and it must have already done the orphan
+ * inode cleanup for us, so we can safely abort without any further action.
+ */
+void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
+{
+	unsigned int s_flags = sb->s_flags;
+	int ret, nr_orphans = 0, nr_truncates = 0;
+#ifdef CONFIG_QUOTA
+	int quota_update = 0;
+	int i;
+#endif
+	if (!es->s_last_orphan) {
+		jbd_debug(4, "no orphan inodes to clean up\n");
+		return;
+	}
+
+	if (bdev_read_only(sb->s_bdev)) {
+		ext4_msg(sb, KERN_ERR, "write access "
+			"unavailable, skipping orphan cleanup");
+		return;
+	}
+
+	/* Check if feature set would not allow a r/w mount */
+	if (!ext4_feature_set_ok(sb, 0)) {
+		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
+			 "unknown ROCOMPAT features");
+		return;
+	}
+
+	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
+		/* don't clear list on RO mount w/ errors */
+		if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
+			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
+				  "clearing orphan list.\n");
+			es->s_last_orphan = 0;
+		}
+		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
+		return;
+	}
+
+	if (s_flags & SB_RDONLY) {
+		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
+		sb->s_flags &= ~SB_RDONLY;
+	}
+#ifdef CONFIG_QUOTA
+	/*
+	 * Turn on quotas which were not enabled for read-only mounts if
+	 * filesystem has quota feature, so that they are updated correctly.
+	 */
+	if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
+		int ret = ext4_enable_quotas(sb);
+
+		if (!ret)
+			quota_update = 1;
+		else
+			ext4_msg(sb, KERN_ERR,
+				"Cannot turn on quotas: error %d", ret);
+	}
+
+	/* Turn on journaled quotas used for old sytle */
+	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
+		if (EXT4_SB(sb)->s_qf_names[i]) {
+			int ret = ext4_quota_on_mount(sb, i);
+
+			if (!ret)
+				quota_update = 1;
+			else
+				ext4_msg(sb, KERN_ERR,
+					"Cannot turn on journaled "
+					"quota: type %d: error %d", i, ret);
+		}
+	}
+#endif
+
+	while (es->s_last_orphan) {
+		struct inode *inode;
+
+		/*
+		 * We may have encountered an error during cleanup; if
+		 * so, skip the rest.
+		 */
+		if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
+			jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
+			es->s_last_orphan = 0;
+			break;
+		}
+
+		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
+		if (IS_ERR(inode)) {
+			es->s_last_orphan = 0;
+			break;
+		}
+
+		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
+		dquot_initialize(inode);
+		if (inode->i_nlink) {
+			if (test_opt(sb, DEBUG))
+				ext4_msg(sb, KERN_DEBUG,
+					"%s: truncating inode %lu to %lld bytes",
+					__func__, inode->i_ino, inode->i_size);
+			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
+				  inode->i_ino, inode->i_size);
+			inode_lock(inode);
+			truncate_inode_pages(inode->i_mapping, inode->i_size);
+			ret = ext4_truncate(inode);
+			if (ret)
+				ext4_std_error(inode->i_sb, ret);
+			inode_unlock(inode);
+			nr_truncates++;
+		} else {
+			if (test_opt(sb, DEBUG))
+				ext4_msg(sb, KERN_DEBUG,
+					"%s: deleting unreferenced inode %lu",
+					__func__, inode->i_ino);
+			jbd_debug(2, "deleting unreferenced inode %lu\n",
+				  inode->i_ino);
+			nr_orphans++;
+		}
+		iput(inode);  /* The delete magic happens here! */
+	}
+
+#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
+
+	if (nr_orphans)
+		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
+		       PLURAL(nr_orphans));
+	if (nr_truncates)
+		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
+		       PLURAL(nr_truncates));
+#ifdef CONFIG_QUOTA
+	/* Turn off quotas if they were enabled for orphan cleanup */
+	if (quota_update) {
+		for (i = 0; i < EXT4_MAXQUOTAS; i++) {
+			if (sb_dqopt(sb)->files[i])
+				dquot_quota_off(sb, i);
+		}
+	}
+#endif
+	sb->s_flags = s_flags; /* Restore SB_RDONLY status */
+}
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index d12982ca923b..6e43c8546dc5 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -80,7 +80,6 @@ static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
 		       const char *dev_name, void *data);
 static inline int ext2_feature_set_ok(struct super_block *sb);
 static inline int ext3_feature_set_ok(struct super_block *sb);
-static int ext4_feature_set_ok(struct super_block *sb, int readonly);
 static void ext4_destroy_lazyinit_thread(void);
 static void ext4_unregister_li_request(struct super_block *sb);
 static void ext4_clear_request_list(void);
@@ -1595,14 +1594,12 @@ static int ext4_mark_dquot_dirty(struct dquot *dquot);
 static int ext4_write_info(struct super_block *sb, int type);
 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
 			 const struct path *path);
-static int ext4_quota_on_mount(struct super_block *sb, int type);
 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
 			       size_t len, loff_t off);
 static ssize_t ext4_quota_write(struct super_block *sb, int type,
 				const char *data, size_t len, loff_t off);
 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
 			     unsigned int flags);
-static int ext4_enable_quotas(struct super_block *sb);
 
 static struct dquot **ext4_get_dquots(struct inode *inode)
 {
@@ -2981,162 +2978,6 @@ static int ext4_check_descriptors(struct super_block *sb,
 	return 1;
 }
 
-/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
- * the superblock) which were deleted from all directories, but held open by
- * a process at the time of a crash.  We walk the list and try to delete these
- * inodes at recovery time (only with a read-write filesystem).
- *
- * In order to keep the orphan inode chain consistent during traversal (in
- * case of crash during recovery), we link each inode into the superblock
- * orphan list_head and handle it the same way as an inode deletion during
- * normal operation (which journals the operations for us).
- *
- * We only do an iget() and an iput() on each inode, which is very safe if we
- * accidentally point at an in-use or already deleted inode.  The worst that
- * can happen in this case is that we get a "bit already cleared" message from
- * ext4_free_inode().  The only reason we would point at a wrong inode is if
- * e2fsck was run on this filesystem, and it must have already done the orphan
- * inode cleanup for us, so we can safely abort without any further action.
- */
-static void ext4_orphan_cleanup(struct super_block *sb,
-				struct ext4_super_block *es)
-{
-	unsigned int s_flags = sb->s_flags;
-	int ret, nr_orphans = 0, nr_truncates = 0;
-#ifdef CONFIG_QUOTA
-	int quota_update = 0;
-	int i;
-#endif
-	if (!es->s_last_orphan) {
-		jbd_debug(4, "no orphan inodes to clean up\n");
-		return;
-	}
-
-	if (bdev_read_only(sb->s_bdev)) {
-		ext4_msg(sb, KERN_ERR, "write access "
-			"unavailable, skipping orphan cleanup");
-		return;
-	}
-
-	/* Check if feature set would not allow a r/w mount */
-	if (!ext4_feature_set_ok(sb, 0)) {
-		ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
-			 "unknown ROCOMPAT features");
-		return;
-	}
-
-	if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
-		/* don't clear list on RO mount w/ errors */
-		if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
-			ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
-				  "clearing orphan list.\n");
-			es->s_last_orphan = 0;
-		}
-		jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
-		return;
-	}
-
-	if (s_flags & SB_RDONLY) {
-		ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
-		sb->s_flags &= ~SB_RDONLY;
-	}
-#ifdef CONFIG_QUOTA
-	/*
-	 * Turn on quotas which were not enabled for read-only mounts if
-	 * filesystem has quota feature, so that they are updated correctly.
-	 */
-	if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
-		int ret = ext4_enable_quotas(sb);
-
-		if (!ret)
-			quota_update = 1;
-		else
-			ext4_msg(sb, KERN_ERR,
-				"Cannot turn on quotas: error %d", ret);
-	}
-
-	/* Turn on journaled quotas used for old sytle */
-	for (i = 0; i < EXT4_MAXQUOTAS; i++) {
-		if (EXT4_SB(sb)->s_qf_names[i]) {
-			int ret = ext4_quota_on_mount(sb, i);
-
-			if (!ret)
-				quota_update = 1;
-			else
-				ext4_msg(sb, KERN_ERR,
-					"Cannot turn on journaled "
-					"quota: type %d: error %d", i, ret);
-		}
-	}
-#endif
-
-	while (es->s_last_orphan) {
-		struct inode *inode;
-
-		/*
-		 * We may have encountered an error during cleanup; if
-		 * so, skip the rest.
-		 */
-		if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
-			jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
-			es->s_last_orphan = 0;
-			break;
-		}
-
-		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
-		if (IS_ERR(inode)) {
-			es->s_last_orphan = 0;
-			break;
-		}
-
-		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
-		dquot_initialize(inode);
-		if (inode->i_nlink) {
-			if (test_opt(sb, DEBUG))
-				ext4_msg(sb, KERN_DEBUG,
-					"%s: truncating inode %lu to %lld bytes",
-					__func__, inode->i_ino, inode->i_size);
-			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
-				  inode->i_ino, inode->i_size);
-			inode_lock(inode);
-			truncate_inode_pages(inode->i_mapping, inode->i_size);
-			ret = ext4_truncate(inode);
-			if (ret)
-				ext4_std_error(inode->i_sb, ret);
-			inode_unlock(inode);
-			nr_truncates++;
-		} else {
-			if (test_opt(sb, DEBUG))
-				ext4_msg(sb, KERN_DEBUG,
-					"%s: deleting unreferenced inode %lu",
-					__func__, inode->i_ino);
-			jbd_debug(2, "deleting unreferenced inode %lu\n",
-				  inode->i_ino);
-			nr_orphans++;
-		}
-		iput(inode);  /* The delete magic happens here! */
-	}
-
-#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
-
-	if (nr_orphans)
-		ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
-		       PLURAL(nr_orphans));
-	if (nr_truncates)
-		ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
-		       PLURAL(nr_truncates));
-#ifdef CONFIG_QUOTA
-	/* Turn off quotas if they were enabled for orphan cleanup */
-	if (quota_update) {
-		for (i = 0; i < EXT4_MAXQUOTAS; i++) {
-			if (sb_dqopt(sb)->files[i])
-				dquot_quota_off(sb, i);
-		}
-	}
-#endif
-	sb->s_flags = s_flags; /* Restore SB_RDONLY status */
-}
-
 /*
  * Maximal extent format file size.
  * Resulting logical blkno at s_maxbytes must fit in our on-disk
@@ -3316,7 +3157,7 @@ static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
  * Returns 1 if this filesystem can be mounted as requested,
  * 0 if it cannot be.
  */
-static int ext4_feature_set_ok(struct super_block *sb, int readonly)
+int ext4_feature_set_ok(struct super_block *sb, int readonly)
 {
 	if (ext4_has_unknown_ext4_incompat_features(sb)) {
 		ext4_msg(sb, KERN_ERR,
@@ -6327,16 +6168,6 @@ static int ext4_write_info(struct super_block *sb, int type)
 	return ret;
 }
 
-/*
- * Turn on quotas during mount time - we need to find
- * the quota file and such...
- */
-static int ext4_quota_on_mount(struct super_block *sb, int type)
-{
-	return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
-					EXT4_SB(sb)->s_jquota_fmt, type);
-}
-
 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
 {
 	struct ext4_inode_info *ei = EXT4_I(inode);
@@ -6466,7 +6297,7 @@ static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
 }
 
 /* Enable usage tracking for all quota types. */
-static int ext4_enable_quotas(struct super_block *sb)
+int ext4_enable_quotas(struct super_block *sb)
 {
 	int type, err = 0;
 	unsigned long qf_inums[EXT4_MAXQUOTAS] = {
-- 
2.26.2

Re: [PATCH 2/4] ext4: Move orphan inode handling into a separate file

[Andreas Dilger]

[-- Attachment #1: Type: text/plain, Size: 352 bytes --]


> On Jun 16, 2021, at 4:56 AM, Jan Kara <jack@suse.cz> wrote:
> 
> Move functions for handling orphan inodes into a new file
> fs/ext4/orphan.c to have them in one place and somewhat reduce size of
> other files. No code changes.
> 
> Signed-off-by: Jan Kara <jack@suse.cz>
> ---

Reviewed-by: Andreas Dilger <adilger@dilger.ca>

Cheers, Andreas






[-- Attachment #2: Message signed with OpenPGP --]
[-- Type: application/pgp-signature, Size: 873 bytes --]

Re: [PATCH 2/4] ext4: Move orphan inode handling into a separate file

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 2019 bytes --]

Hi Jan,

I love your patch! Perhaps something to improve:

[auto build test WARNING on ext4/dev]
[also build test WARNING on ext3/for_next linus/master v5.13-rc6 next-20210616]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
config: i386-randconfig-s002-20210617 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
reproduce:
        # apt-get install sparse
        # sparse version: v0.6.3-341-g8af24329-dirty
        # https://github.com/0day-ci/linux/commit/6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
        git checkout 6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        # save the attached .config to linux build tree
        make W=1 C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' W=1 ARCH=i386 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>


sparse warnings: (new ones prefixed by >>)
>> fs/ext4/orphan.c:195:64: sparse: sparse: incorrect type in argument 2 (different address spaces) @@     expected char *qf_name @@     got char [noderef] __rcu * @@
   fs/ext4/orphan.c:195:64: sparse:     expected char *qf_name
   fs/ext4/orphan.c:195:64: sparse:     got char [noderef] __rcu *

vim +195 fs/ext4/orphan.c

   192	
   193	static int ext4_quota_on_mount(struct super_block *sb, int type)
   194	{
 > 195		return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
   196						EXT4_SB(sb)->s_jquota_fmt, type);
   197	}
   198	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 39334 bytes --]

Re: [PATCH 2/4] ext4: Move orphan inode handling into a separate file

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 2067 bytes --]

Hi Jan,

I love your patch! Perhaps something to improve:

[auto build test WARNING on ext4/dev]
[also build test WARNING on ext3/for_next linus/master v5.13-rc6 next-20210616]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
config: i386-randconfig-s002-20210617 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-22) 9.3.0
reproduce:
        # apt-get install sparse
        # sparse version: v0.6.3-341-g8af24329-dirty
        # https://github.com/0day-ci/linux/commit/6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
        git checkout 6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        # save the attached .config to linux build tree
        make W=1 C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' W=1 ARCH=i386 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>


sparse warnings: (new ones prefixed by >>)
>> fs/ext4/orphan.c:195:64: sparse: sparse: incorrect type in argument 2 (different address spaces) @@     expected char *qf_name @@     got char [noderef] __rcu * @@
   fs/ext4/orphan.c:195:64: sparse:     expected char *qf_name
   fs/ext4/orphan.c:195:64: sparse:     got char [noderef] __rcu *

vim +195 fs/ext4/orphan.c

   192	
   193	static int ext4_quota_on_mount(struct super_block *sb, int type)
   194	{
 > 195		return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
   196						EXT4_SB(sb)->s_jquota_fmt, type);
   197	}
   198	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

[-- Attachment #2: config.gz --]
[-- Type: application/gzip, Size: 39334 bytes --]

Re: [PATCH 2/4] ext4: Move orphan inode handling into a separate file

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 2915 bytes --]

Hi Jan,

I love your patch! Yet something to improve:

[auto build test ERROR on ext4/dev]
[also build test ERROR on ext3/for_next linus/master v5.13-rc6 next-20210616]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
config: x86_64-randconfig-a011-20210617 (attached as .config)
compiler: clang version 13.0.0 (https://github.com/llvm/llvm-project 64720f57bea6a6bf033feef4a5751ab9c0c3b401)
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # install x86_64 cross compiling tool for clang build
        # apt-get install binutils-x86-64-linux-gnu
        # https://github.com/0day-ci/linux/commit/6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
        git checkout 6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=x86_64 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

>> fs/ext4/orphan.c:195:9: error: implicit declaration of function 'dquot_quota_on_mount' [-Werror,-Wimplicit-function-declaration]
           return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
                  ^
   fs/ext4/orphan.c:195:9: note: did you mean 'ext4_quota_on_mount'?
   fs/ext4/orphan.c:193:12: note: 'ext4_quota_on_mount' declared here
   static int ext4_quota_on_mount(struct super_block *sb, int type)
              ^
>> fs/ext4/orphan.c:195:47: error: no member named 's_qf_names' in 'struct ext4_sb_info'
           return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
                                           ~~~~~~~~~~~  ^
>> fs/ext4/orphan.c:196:19: error: no member named 's_jquota_fmt' in 'struct ext4_sb_info'
                                           EXT4_SB(sb)->s_jquota_fmt, type);
                                           ~~~~~~~~~~~  ^
   3 errors generated.


vim +/dquot_quota_on_mount +195 fs/ext4/orphan.c

   192	
   193	static int ext4_quota_on_mount(struct super_block *sb, int type)
   194	{
 > 195		return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
 > 196						EXT4_SB(sb)->s_jquota_fmt, type);
   197	}
   198	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 34134 bytes --]

Re: [PATCH 2/4] ext4: Move orphan inode handling into a separate file

[kernel test robot]

[-- Attachment #1: Type: text/plain, Size: 2977 bytes --]

Hi Jan,

I love your patch! Yet something to improve:

[auto build test ERROR on ext4/dev]
[also build test ERROR on ext3/for_next linus/master v5.13-rc6 next-20210616]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]

url:    https://github.com/0day-ci/linux/commits/Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
base:   https://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4.git dev
config: x86_64-randconfig-a011-20210617 (attached as .config)
compiler: clang version 13.0.0 (https://github.com/llvm/llvm-project 64720f57bea6a6bf033feef4a5751ab9c0c3b401)
reproduce (this is a W=1 build):
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # install x86_64 cross compiling tool for clang build
        # apt-get install binutils-x86-64-linux-gnu
        # https://github.com/0day-ci/linux/commit/6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        git remote add linux-review https://github.com/0day-ci/linux
        git fetch --no-tags linux-review Jan-Kara/ext4-Speedup-orphan-file-handling/20210617-034806
        git checkout 6b60fad4c555893cdd03e91dbfe31aa6fa9c25e7
        # save the attached .config to linux build tree
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=x86_64 

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

>> fs/ext4/orphan.c:195:9: error: implicit declaration of function 'dquot_quota_on_mount' [-Werror,-Wimplicit-function-declaration]
           return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
                  ^
   fs/ext4/orphan.c:195:9: note: did you mean 'ext4_quota_on_mount'?
   fs/ext4/orphan.c:193:12: note: 'ext4_quota_on_mount' declared here
   static int ext4_quota_on_mount(struct super_block *sb, int type)
              ^
>> fs/ext4/orphan.c:195:47: error: no member named 's_qf_names' in 'struct ext4_sb_info'
           return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
                                           ~~~~~~~~~~~  ^
>> fs/ext4/orphan.c:196:19: error: no member named 's_jquota_fmt' in 'struct ext4_sb_info'
                                           EXT4_SB(sb)->s_jquota_fmt, type);
                                           ~~~~~~~~~~~  ^
   3 errors generated.


vim +/dquot_quota_on_mount +195 fs/ext4/orphan.c

   192	
   193	static int ext4_quota_on_mount(struct super_block *sb, int type)
   194	{
 > 195		return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
 > 196						EXT4_SB(sb)->s_jquota_fmt, type);
   197	}
   198	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

[-- Attachment #2: config.gz --]
[-- Type: application/gzip, Size: 34134 bytes --]

[PATCH 3/4] ext4: Speedup ext4 orphan inode handling

[Jan Kara]

Ext4 orphan inode handling is a bottleneck for workloads which heavily
truncate / unlink small files since it contends on the global
s_orphan_mutex lock (and generally it's difficult to improve scalability
of the ondisk linked list of orphaned inodes).

This patch implements new way of handling orphan inodes. Instead of
linking orphaned inode into a linked list, we store it's inode number in
a new special file which we call "orphan file". Currently we still
protect the orphan file with a spinlock for simplicity but even in this
setting we can substantially reduce the length of the critical section
and thus speedup some workloads.

Note that the change is backwards compatible when the filesystem is
clean - the existence of the orphan file is a compat feature, we set
another ro-compat feature indicating orphan file needs scanning for
orphaned inodes when mounting filesystem read-write. This ro-compat
feature gets cleared on unmount / remount read-only.

Some performance data from 80 CPU Xeon Server with 512 GB of RAM,
filesystem located on SSD, average of 5 runs:

stress-orphan (microbenchmark truncating files byte-by-byte from N
processes in parallel)

Threads Time            Time
        Vanilla         Patched
  1       1.057200        0.945600
  2       1.680400        1.331800
  4       2.547000        1.995000
  8       7.049400        6.424200
 16      14.827800       14.937600
 32      40.948200       33.038200
 64      87.787400       60.823600
128     206.504000      122.941400

So we can see significant wins all over the board.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/ext4.h   |  70 +++++++++--
 fs/ext4/orphan.c | 319 ++++++++++++++++++++++++++++++++++++++++++-----
 fs/ext4/super.c  |  34 ++++-
 3 files changed, 379 insertions(+), 44 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 33508487516f..83298c0b6dae 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1025,7 +1025,14 @@ struct ext4_inode_info {
 	 */
 	struct rw_semaphore xattr_sem;
 
-	struct list_head i_orphan;	/* unlinked but open inodes */
+	/*
+	 * Inodes with EXT4_STATE_ORPHAN_FILE use i_orphan_idx. Otherwise
+	 * i_orphan is used.
+	 */
+	union {
+		struct list_head i_orphan;	/* unlinked but open inodes */
+		unsigned int i_orphan_idx;	/* Index in orphan file */
+	};
 
 	/* Fast commit related info */
 
@@ -1419,7 +1426,8 @@ struct ext4_super_block {
 	__u8    s_last_error_errcode;
 	__le16  s_encoding;		/* Filename charset encoding */
 	__le16  s_encoding_flags;	/* Filename charset encoding flags */
-	__le32	s_reserved[95];		/* Padding to the end of the block */
+	__le32  s_orphan_file_inum;	/* Inode for tracking orphan inodes */
+	__le32	s_reserved[94];		/* Padding to the end of the block */
 	__le32	s_checksum;		/* crc32c(superblock) */
 };
 
@@ -1440,6 +1448,7 @@ struct ext4_super_block {
 
 /* Types of ext4 journal triggers */
 enum ext4_journal_trigger_type {
+	EXT4_JTR_ORPHAN_FILE,
 	EXT4_JTR_NONE	/* This must be the last entry for indexing to work! */
 };
 
@@ -1456,6 +1465,36 @@ static inline struct ext4_journal_trigger *EXT4_TRIGGER(
 	return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
 }
 
+#define EXT4_ORPHAN_BLOCK_MAGIC 0x0b10ca04
+
+/* Structure at the tail of orphan block */
+struct ext4_orphan_block_tail {
+	__le32 ob_magic;
+	__le32 ob_checksum;
+};
+
+static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
+{
+	return (sb->s_blocksize - sizeof(struct ext4_orphan_block_tail)) /
+			sizeof(u32);
+}
+
+struct ext4_orphan_block {
+	int ob_free_entries;	/* Number of free orphan entries in block */
+	struct buffer_head *ob_bh;	/* Buffer for orphan block */
+};
+
+/*
+ * Info about orphan file.
+ */
+struct ext4_orphan_info {
+	spinlock_t of_lock;
+	int of_blocks;			/* Number of orphan blocks in a file */
+	__u32 of_csum_seed;		/* Checksum seed for orphan file */
+	struct ext4_orphan_block *of_binfo;	/* Array with info about orphan
+						 * file blocks */
+};
+
 /*
  * fourth extended-fs super-block data in memory
  */
@@ -1509,9 +1548,11 @@ struct ext4_sb_info {
 
 	/* Journaling */
 	struct journal_s *s_journal;
-	struct list_head s_orphan;
-	struct mutex s_orphan_lock;
 	unsigned long s_ext4_flags;		/* Ext4 superblock flags */
+	struct mutex s_orphan_lock;	/* Protects on disk list changes */
+	struct list_head s_orphan;	/* List of orphaned inodes in on disk
+					   list */
+	struct ext4_orphan_info s_orphan_info;
 	unsigned long s_commit_interval;
 	u32 s_max_batch_time;
 	u32 s_min_batch_time;
@@ -1846,6 +1887,7 @@ enum {
 	EXT4_STATE_LUSTRE_EA_INODE,	/* Lustre-style ea_inode */
 	EXT4_STATE_VERITY_IN_PROGRESS,	/* building fs-verity Merkle tree */
 	EXT4_STATE_FC_COMMITTING,	/* Fast commit ongoing */
+	EXT4_STATE_ORPHAN_FILE,		/* Inode orphaned in orphan file */
 };
 
 #define EXT4_INODE_BIT_FNS(name, field, offset)				\
@@ -1947,6 +1989,7 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
  */
 #define EXT4_FEATURE_COMPAT_FAST_COMMIT		0x0400
 #define EXT4_FEATURE_COMPAT_STABLE_INODES	0x0800
+#define EXT4_FEATURE_COMPAT_ORPHAN_FILE		0x1000	/* Orphan file exists */
 
 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER	0x0001
 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE	0x0002
@@ -1955,6 +1998,8 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
 #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM		0x0010
 #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK	0x0020
 #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE	0x0040
+#define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT	0x0080 /* Orphan file may be
+							  non-empty */
 #define EXT4_FEATURE_RO_COMPAT_QUOTA		0x0100
 #define EXT4_FEATURE_RO_COMPAT_BIGALLOC		0x0200
 /*
@@ -1964,6 +2009,7 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
  * GDT_CSUM bits are mutually exclusive.
  */
 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM	0x0400
+/* 0x0800 Reserved for EXT4_FEATURE_RO_COMPAT_REPLICA */
 #define EXT4_FEATURE_RO_COMPAT_READONLY		0x1000
 #define EXT4_FEATURE_RO_COMPAT_PROJECT		0x2000
 #define EXT4_FEATURE_RO_COMPAT_VERITY		0x8000
@@ -2050,6 +2096,7 @@ EXT4_FEATURE_COMPAT_FUNCS(dir_index,		DIR_INDEX)
 EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,	SPARSE_SUPER2)
 EXT4_FEATURE_COMPAT_FUNCS(fast_commit,		FAST_COMMIT)
 EXT4_FEATURE_COMPAT_FUNCS(stable_inodes,	STABLE_INODES)
+EXT4_FEATURE_COMPAT_FUNCS(orphan_file,		ORPHAN_FILE)
 
 EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,	SPARSE_SUPER)
 EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,	LARGE_FILE)
@@ -2064,6 +2111,7 @@ EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,	METADATA_CSUM)
 EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,		READONLY)
 EXT4_FEATURE_RO_COMPAT_FUNCS(project,		PROJECT)
 EXT4_FEATURE_RO_COMPAT_FUNCS(verity,		VERITY)
+EXT4_FEATURE_RO_COMPAT_FUNCS(orphan_present,	ORPHAN_PRESENT)
 
 EXT4_FEATURE_INCOMPAT_FUNCS(compression,	COMPRESSION)
 EXT4_FEATURE_INCOMPAT_FUNCS(filetype,		FILETYPE)
@@ -2097,7 +2145,8 @@ EXT4_FEATURE_INCOMPAT_FUNCS(casefold,		CASEFOLD)
 					 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
 					 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
 
-#define EXT4_FEATURE_COMPAT_SUPP	EXT4_FEATURE_COMPAT_EXT_ATTR
+#define EXT4_FEATURE_COMPAT_SUPP	(EXT4_FEATURE_COMPAT_EXT_ATTR| \
+					 EXT4_FEATURE_COMPAT_ORPHAN_FILE)
 #define EXT4_FEATURE_INCOMPAT_SUPP	(EXT4_FEATURE_INCOMPAT_FILETYPE| \
 					 EXT4_FEATURE_INCOMPAT_RECOVER| \
 					 EXT4_FEATURE_INCOMPAT_META_BG| \
@@ -2122,7 +2171,8 @@ EXT4_FEATURE_INCOMPAT_FUNCS(casefold,		CASEFOLD)
 					 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
 					 EXT4_FEATURE_RO_COMPAT_QUOTA |\
 					 EXT4_FEATURE_RO_COMPAT_PROJECT |\
-					 EXT4_FEATURE_RO_COMPAT_VERITY)
+					 EXT4_FEATURE_RO_COMPAT_VERITY |\
+					 EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
 
 #define EXTN_FEATURE_FUNCS(ver) \
 static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
@@ -2172,7 +2222,6 @@ static inline int ext4_forced_shutdown(struct ext4_sb_info *sbi)
 	return test_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
 }
 
-
 /*
  * Default values for user and/or group using reserved blocks
  */
@@ -3751,6 +3800,13 @@ extern int ext4_orphan_add(handle_t *, struct inode *);
 extern int ext4_orphan_del(handle_t *, struct inode *);
 extern void ext4_orphan_cleanup(struct super_block *sb,
 				struct ext4_super_block *es);
+extern void ext4_release_orphan_info(struct super_block *sb);
+extern int ext4_init_orphan_info(struct super_block *sb);
+extern int ext4_orphan_file_empty(struct super_block *sb);
+extern void ext4_orphan_file_block_trigger(
+				struct jbd2_buffer_trigger_type *triggers,
+				struct buffer_head *bh,
+				void *data, size_t size);
 
 /*
  * Add new method to test whether block and inode bitmaps are properly
diff --git a/fs/ext4/orphan.c b/fs/ext4/orphan.c
index 732b16ef655b..ac22667b7fd5 100644
--- a/fs/ext4/orphan.c
+++ b/fs/ext4/orphan.c
@@ -8,6 +8,52 @@
 #include "ext4.h"
 #include "ext4_jbd2.h"
 
+static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
+{
+	int i, j;
+	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
+	int ret = 0;
+	__le32 *bdata;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
+
+	spin_lock(&oi->of_lock);
+	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
+	if (i == oi->of_blocks) {
+		spin_unlock(&oi->of_lock);
+		/*
+		 * For now we don't grow or shrink orphan file. We just use
+		 * whatever was allocated at mke2fs time. The additional
+		 * credits we would have to reserve for each orphan inode
+		 * operation just don't seem worth it.
+		 */
+		return -ENOSPC;
+	}
+	oi->of_binfo[i].ob_free_entries--;
+	spin_unlock(&oi->of_lock);
+
+	/*
+	 * Get access to orphan block. We have dropped of_lock but since we
+	 * have decremented number of free entries we are guaranteed free entry
+	 * in our block.
+	 */
+	ret = ext4_journal_get_write_access(handle, inode->i_sb,
+				oi->of_binfo[i].ob_bh, EXT4_JTR_ORPHAN_FILE);
+	if (ret)
+		return ret;
+
+	bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
+	spin_lock(&oi->of_lock);
+	/* Find empty slot in a block */
+	for (j = 0; j < inodes_per_ob && bdata[j]; j++);
+	BUG_ON(j == inodes_per_ob);
+	bdata[j] = cpu_to_le32(inode->i_ino);
+	EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
+	ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
+	spin_unlock(&oi->of_lock);
+
+	return ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[i].ob_bh);
+}
+
 /*
  * ext4_orphan_add() links an unlinked or truncated inode into a list of
  * such inodes, starting at the superblock, in case we crash before the
@@ -34,10 +80,10 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
 	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
 		     !inode_is_locked(inode));
 	/*
-	 * Exit early if inode already is on orphan list. This is a big speedup
-	 * since we don't have to contend on the global s_orphan_lock.
+	 * Inode orphaned in orphan file or in orphan list?
 	 */
-	if (!list_empty(&EXT4_I(inode)->i_orphan))
+	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE) ||
+	    !list_empty(&EXT4_I(inode)->i_orphan))
 		return 0;
 
 	/*
@@ -49,6 +95,16 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
 	ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
 
+	if (sbi->s_orphan_info.of_blocks) {
+		err = ext4_orphan_file_add(handle, inode);
+		/*
+		 * Fallback to normal orphan list of orphan file is
+		 * out of space
+		 */
+		if (err != -ENOSPC)
+			return err;
+	}
+
 	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
 	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
 					    EXT4_JTR_NONE);
@@ -103,6 +159,37 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
 	return err;
 }
 
+static int ext4_orphan_file_del(handle_t *handle, struct inode *inode)
+{
+	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
+	__le32 *bdata;
+	int blk, off;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
+	int ret = 0;
+
+	if (!handle)
+		goto out;
+	blk = EXT4_I(inode)->i_orphan_idx / inodes_per_ob;
+	off = EXT4_I(inode)->i_orphan_idx % inodes_per_ob;
+
+	ret = ext4_journal_get_write_access(handle, inode->i_sb,
+				oi->of_binfo[blk].ob_bh, EXT4_JTR_ORPHAN_FILE);
+	if (ret)
+		goto out;
+
+	bdata = (__le32 *)(oi->of_binfo[blk].ob_bh->b_data);
+	spin_lock(&oi->of_lock);
+	bdata[off] = 0;
+	oi->of_binfo[blk].ob_free_entries++;
+	spin_unlock(&oi->of_lock);
+	ret = ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[blk].ob_bh);
+out:
+	ext4_clear_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
+	INIT_LIST_HEAD(&EXT4_I(inode)->i_orphan);
+
+	return ret;
+}
+
 /*
  * ext4_orphan_del() removes an unlinked or truncated inode from the list
  * of such inodes stored on disk, because it is finally being cleaned up.
@@ -121,6 +208,9 @@ int ext4_orphan_del(handle_t *handle, struct inode *inode)
 
 	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
 		     !inode_is_locked(inode));
+	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE))
+		return ext4_orphan_file_del(handle, inode);
+
 	/* Do this quick check before taking global s_orphan_lock. */
 	if (list_empty(&ei->i_orphan))
 		return 0;
@@ -196,6 +286,39 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
 					EXT4_SB(sb)->s_jquota_fmt, type);
 }
 
+static void ext4_process_orphan(struct inode *inode,
+				int *nr_truncates, int *nr_orphans)
+{
+	struct super_block *sb = inode->i_sb;
+	int ret;
+
+	dquot_initialize(inode);
+	if (inode->i_nlink) {
+		if (test_opt(sb, DEBUG))
+			ext4_msg(sb, KERN_DEBUG,
+				"%s: truncating inode %lu to %lld bytes",
+				__func__, inode->i_ino, inode->i_size);
+		jbd_debug(2, "truncating inode %lu to %lld bytes\n",
+			  inode->i_ino, inode->i_size);
+		inode_lock(inode);
+		truncate_inode_pages(inode->i_mapping, inode->i_size);
+		ret = ext4_truncate(inode);
+		if (ret)
+			ext4_std_error(inode->i_sb, ret);
+		inode_unlock(inode);
+		(*nr_truncates)++;
+	} else {
+		if (test_opt(sb, DEBUG))
+			ext4_msg(sb, KERN_DEBUG,
+				"%s: deleting unreferenced inode %lu",
+				__func__, inode->i_ino);
+		jbd_debug(2, "deleting unreferenced inode %lu\n",
+			  inode->i_ino);
+		(*nr_orphans)++;
+	}
+	iput(inode);  /* The delete magic happens here! */
+}
+
 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
  * the superblock) which were deleted from all directories, but held open by
  * a process at the time of a crash.  We walk the list and try to delete these
@@ -216,12 +339,17 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
 void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
 {
 	unsigned int s_flags = sb->s_flags;
-	int ret, nr_orphans = 0, nr_truncates = 0;
+	int nr_orphans = 0, nr_truncates = 0;
+	struct inode *inode;
+	int i, j;
 #ifdef CONFIG_QUOTA
 	int quota_update = 0;
-	int i;
 #endif
-	if (!es->s_last_orphan) {
+	__le32 *bdata;
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+
+	if (!es->s_last_orphan && !oi->of_blocks) {
 		jbd_debug(4, "no orphan inodes to clean up\n");
 		return;
 	}
@@ -285,8 +413,6 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
 #endif
 
 	while (es->s_last_orphan) {
-		struct inode *inode;
-
 		/*
 		 * We may have encountered an error during cleanup; if
 		 * so, skip the rest.
@@ -304,31 +430,21 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
 		}
 
 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
-		dquot_initialize(inode);
-		if (inode->i_nlink) {
-			if (test_opt(sb, DEBUG))
-				ext4_msg(sb, KERN_DEBUG,
-					"%s: truncating inode %lu to %lld bytes",
-					__func__, inode->i_ino, inode->i_size);
-			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
-				  inode->i_ino, inode->i_size);
-			inode_lock(inode);
-			truncate_inode_pages(inode->i_mapping, inode->i_size);
-			ret = ext4_truncate(inode);
-			if (ret)
-				ext4_std_error(inode->i_sb, ret);
-			inode_unlock(inode);
-			nr_truncates++;
-		} else {
-			if (test_opt(sb, DEBUG))
-				ext4_msg(sb, KERN_DEBUG,
-					"%s: deleting unreferenced inode %lu",
-					__func__, inode->i_ino);
-			jbd_debug(2, "deleting unreferenced inode %lu\n",
-				  inode->i_ino);
-			nr_orphans++;
+		ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
+	}
+
+	for (i = 0; i < oi->of_blocks; i++) {
+		bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
+		for (j = 0; j < inodes_per_ob; j++) {
+			if (!bdata[j])
+				continue;
+			inode = ext4_orphan_get(sb, le32_to_cpu(bdata[j]));
+			if (IS_ERR(inode))
+				continue;
+			ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
+			EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
+			ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
 		}
-		iput(inode);  /* The delete magic happens here! */
 	}
 
 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
@@ -350,3 +466,142 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
 #endif
 	sb->s_flags = s_flags; /* Restore SB_RDONLY status */
 }
+
+void ext4_release_orphan_info(struct super_block *sb)
+{
+	int i;
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+
+	if (!oi->of_blocks)
+		return;
+	for (i = 0; i < oi->of_blocks; i++)
+		brelse(oi->of_binfo[i].ob_bh);
+	kfree(oi->of_binfo);
+}
+
+static struct ext4_orphan_block_tail *ext4_orphan_block_tail(
+						struct super_block *sb,
+						struct buffer_head *bh)
+{
+	return (struct ext4_orphan_block_tail *)(bh->b_data + sb->s_blocksize -
+				sizeof(struct ext4_orphan_block_tail));
+}
+
+static int ext4_orphan_file_block_csum_verify(struct super_block *sb,
+					      struct buffer_head *bh)
+{
+	__u32 calculated;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	struct ext4_orphan_block_tail *ot;
+
+	if (!ext4_has_metadata_csum(sb))
+		return 1;
+
+	ot = ext4_orphan_block_tail(sb, bh);
+	calculated = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed,
+				 (__u8 *)bh->b_data,
+				 inodes_per_ob * sizeof(__u32));
+	return le32_to_cpu(ot->ob_checksum) == calculated;
+}
+
+/* This gets called only when checksumming is enabled */
+void ext4_orphan_file_block_trigger(struct jbd2_buffer_trigger_type *triggers,
+				    struct buffer_head *bh,
+				    void *data, size_t size)
+{
+	struct super_block *sb = EXT4_TRIGGER(triggers)->sb;
+	__u32 csum;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	struct ext4_orphan_block_tail *ot;
+
+	csum = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed, (__u8 *)data,
+			   inodes_per_ob * sizeof(__u32));
+	ot = ext4_orphan_block_tail(sb, bh);
+	ot->ob_checksum = cpu_to_le32(csum);
+}
+
+int ext4_init_orphan_info(struct super_block *sb)
+{
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	struct inode *inode;
+	int i, j;
+	int ret;
+	int free;
+	__le32 *bdata;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+	struct ext4_orphan_block_tail *ot;
+	ino_t orphan_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_orphan_file_inum);
+
+	spin_lock_init(&oi->of_lock);
+
+	if (!ext4_has_feature_orphan_file(sb))
+		return 0;
+
+	inode = ext4_iget(sb, orphan_ino, EXT4_IGET_NORMAL);
+	if (IS_ERR(inode)) {
+		ext4_msg(sb, KERN_ERR, "get orphan inode failed");
+		return PTR_ERR(inode);
+	}
+	oi->of_blocks = inode->i_size >> sb->s_blocksize_bits;
+	oi->of_csum_seed = EXT4_I(inode)->i_csum_seed;
+	oi->of_binfo = kmalloc(oi->of_blocks*sizeof(struct ext4_orphan_block),
+			       GFP_KERNEL);
+	if (!oi->of_binfo) {
+		ret = -ENOMEM;
+		goto out_put;
+	}
+	for (i = 0; i < oi->of_blocks; i++) {
+		oi->of_binfo[i].ob_bh = ext4_bread(NULL, inode, i, 0);
+		if (IS_ERR(oi->of_binfo[i].ob_bh)) {
+			ret = PTR_ERR(oi->of_binfo[i].ob_bh);
+			goto out_free;
+		}
+		if (!oi->of_binfo[i].ob_bh) {
+			ret = -EIO;
+			goto out_free;
+		}
+		ot = ext4_orphan_block_tail(sb, oi->of_binfo[i].ob_bh);
+		if (le32_to_cpu(ot->ob_magic) != EXT4_ORPHAN_BLOCK_MAGIC) {
+			ext4_error(sb, "orphan file block %d: bad magic", i);
+			ret = -EIO;
+			goto out_free;
+		}
+		if (!ext4_orphan_file_block_csum_verify(sb,
+						oi->of_binfo[i].ob_bh)) {
+			ext4_error(sb, "orphan file block %d: bad checksum", i);
+			ret = -EIO;
+			goto out_free;
+		}
+		bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
+		free = 0;
+		for (j = 0; j < inodes_per_ob; j++)
+			if (bdata[j] == 0)
+				free++;
+		oi->of_binfo[i].ob_free_entries = free;
+	}
+	iput(inode);
+	return 0;
+out_free:
+	for (i--; i >= 0; i--)
+		brelse(oi->of_binfo[i].ob_bh);
+	kfree(oi->of_binfo);
+out_put:
+	iput(inode);
+	return ret;
+}
+
+int ext4_orphan_file_empty(struct super_block *sb)
+{
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	int i;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+
+	if (!ext4_has_feature_orphan_file(sb))
+		return 1;
+	for (i = 0; i < oi->of_blocks; i++)
+		if (oi->of_binfo[i].ob_free_entries != inodes_per_ob)
+			return 0;
+	return 1;
+}
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 6e43c8546dc5..06f63b0cd988 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -1164,6 +1164,7 @@ static void ext4_put_super(struct super_block *sb)
 
 	flush_work(&sbi->s_error_work);
 	destroy_workqueue(sbi->rsv_conversion_wq);
+	ext4_release_orphan_info(sb);
 
 	/*
 	 * Unregister sysfs before destroying jbd2 journal.
@@ -1189,6 +1190,7 @@ static void ext4_put_super(struct super_block *sb)
 
 	if (!sb_rdonly(sb) && !aborted) {
 		ext4_clear_feature_journal_needs_recovery(sb);
+		ext4_clear_feature_orphan_present(sb);
 		es->s_state = cpu_to_le16(sbi->s_mount_state);
 	}
 	if (!sb_rdonly(sb))
@@ -2695,8 +2697,11 @@ static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
 		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
 	le16_add_cpu(&es->s_mnt_count, 1);
 	ext4_update_tstamp(es, s_mtime);
-	if (sbi->s_journal)
+	if (sbi->s_journal) {
 		ext4_set_feature_journal_needs_recovery(sb);
+		if (ext4_has_feature_orphan_file(sb))
+			ext4_set_feature_orphan_present(sb);
+	}
 
 	err = ext4_commit_super(sb);
 done:
@@ -3971,6 +3976,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 		silent = 1;
 		goto cantfind_ext4;
 	}
+	ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
+				ext4_orphan_file_block_trigger);
 
 	/* Load the checksum driver */
 	sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
@@ -4635,6 +4642,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	sb->s_root = NULL;
 
 	needs_recovery = (es->s_last_orphan != 0 ||
+			  ext4_has_feature_orphan_present(sb) ||
 			  ext4_has_feature_journal_needs_recovery(sb));
 
 	if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
@@ -4924,12 +4932,15 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	if (err)
 		goto failed_mount7;
 
+	err = ext4_init_orphan_info(sb);
+	if (err)
+		goto failed_mount8;
 #ifdef CONFIG_QUOTA
 	/* Enable quota usage during mount. */
 	if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
 		err = ext4_enable_quotas(sb);
 		if (err)
-			goto failed_mount8;
+			goto failed_mount9;
 	}
 #endif  /* CONFIG_QUOTA */
 
@@ -4948,7 +4959,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 		ext4_msg(sb, KERN_INFO, "recovery complete");
 		err = ext4_mark_recovery_complete(sb, es);
 		if (err)
-			goto failed_mount8;
+			goto failed_mount9;
 	}
 	if (EXT4_SB(sb)->s_journal) {
 		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
@@ -4994,6 +5005,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
 	goto failed_mount;
 
+failed_mount9:
+	ext4_release_orphan_info(sb);
 failed_mount8:
 	ext4_unregister_sysfs(sb);
 	kobject_put(&sbi->s_kobj);
@@ -5505,8 +5518,15 @@ static int ext4_mark_recovery_complete(struct super_block *sb,
 	if (err < 0)
 		goto out;
 
-	if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
+	if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
+	    ext4_has_feature_orphan_present(sb))) {
+		if (!ext4_orphan_file_empty(sb)) {
+			ext4_error(sb, "Orphan file not empty on read-only fs.");
+			err = -EFSCORRUPTED;
+			goto out;
+		}
 		ext4_clear_feature_journal_needs_recovery(sb);
+		ext4_clear_feature_orphan_present(sb);
 		ext4_commit_super(sb);
 	}
 out:
@@ -5649,6 +5669,8 @@ static int ext4_freeze(struct super_block *sb)
 
 		/* Journal blocked and flushed, clear needs_recovery flag. */
 		ext4_clear_feature_journal_needs_recovery(sb);
+		if (ext4_orphan_file_empty(sb))
+			ext4_clear_feature_orphan_present(sb);
 	}
 
 	error = ext4_commit_super(sb);
@@ -5671,6 +5693,8 @@ static int ext4_unfreeze(struct super_block *sb)
 	if (EXT4_SB(sb)->s_journal) {
 		/* Reset the needs_recovery flag before the fs is unlocked. */
 		ext4_set_feature_journal_needs_recovery(sb);
+		if (ext4_has_feature_orphan_file(sb))
+			ext4_set_feature_orphan_present(sb);
 	}
 
 	ext4_commit_super(sb);
@@ -5876,7 +5900,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
 			 * around from a previously readonly bdev mount,
 			 * require a full umount/remount for now.
 			 */
-			if (es->s_last_orphan) {
+			if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
 				ext4_msg(sb, KERN_WARNING, "Couldn't "
 				       "remount RDWR because of unprocessed "
 				       "orphan inode list.  Please "
-- 
2.26.2

Re: [PATCH 3/4] ext4: Speedup ext4 orphan inode handling

[Andreas Dilger]

[-- Attachment #1: Type: text/plain, Size: 3807 bytes --]

On Jun 16, 2021, at 4:56 AM, Jan Kara <jack@suse.cz> wrote:
> 
> Ext4 orphan inode handling is a bottleneck for workloads which heavily
> truncate / unlink small files since it contends on the global
> s_orphan_mutex lock (and generally it's difficult to improve scalability
> of the ondisk linked list of orphaned inodes).
> 
> This patch implements new way of handling orphan inodes. Instead of
> linking orphaned inode into a linked list, we store it's inode number in
> a new special file which we call "orphan file". Currently we still
> protect the orphan file with a spinlock for simplicity but even in this
> setting we can substantially reduce the length of the critical section
> and thus speedup some workloads.

Is it a single spinlock for the whole file?  Did you consider using
a per-page lock or grouplock?  With a page in the orphan file for each
CPU core, it would basically be lockless.

> Note that the change is backwards compatible when the filesystem is
> clean - the existence of the orphan file is a compat feature, we set
> another ro-compat feature indicating orphan file needs scanning for
> orphaned inodes when mounting filesystem read-write. This ro-compat
> feature gets cleared on unmount / remount read-only.
> 
> Some performance data from 80 CPU Xeon Server with 512 GB of RAM,
> filesystem located on SSD, average of 5 runs:
> 
> stress-orphan (microbenchmark truncating files byte-by-byte from N
> processes in parallel)
> 
> Threads Time            Time
>        Vanilla         Patched
>  1       1.057200        0.945600
>  2       1.680400        1.331800
>  4       2.547000        1.995000
>  8       7.049400        6.424200
> 16      14.827800       14.937600
> 32      40.948200       33.038200
> 64      87.787400       60.823600
> 128     206.504000      122.941400
> 
> So we can see significant wins all over the board.
> 
> Signed-off-by: Jan Kara <jack@suse.cz>
> 
> +static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
> +{
> 	spin_lock(&oi->of_lock);
> +	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
> +	if (i == oi->of_blocks) {
> +		spin_unlock(&oi->of_lock);
> +		/*
> +		 * For now we don't grow or shrink orphan file. We just use
> +		 * whatever was allocated at mke2fs time. The additional
> +		 * credits we would have to reserve for each orphan inode
> +		 * operation just don't seem worth it.
> +		 */
> +		return -ENOSPC;
> +	}
> +	oi->of_binfo[i].ob_free_entries--;
> +	spin_unlock(&oi->of_lock);

How do we know how large to make the orphan file at mkfs time?  What if it
becomes full during use?  It seems like reserving a fixed number of blocks
will invariably be incorrect for the actual workload on the filesystem.

> @@ -49,6 +95,16 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
> 	ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
> 		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
> 
> +	if (sbi->s_orphan_info.of_blocks) {
> +		err = ext4_orphan_file_add(handle, inode);
> +		/*
> +		 * Fallback to normal orphan list of orphan file is
> +		 * out of space
> +		 */
> +		if (err != -ENOSPC)
> +			return err;
> +	}

This could schedule a task on a workqueue to allocate a few more blocks?
That could easily reserve more credits for this action, without making
the common case more expensive.  Even if it isn't used with the current
mount, it would be available for the next mount (which presumably would
also need additional blocks).

Whether it is worth the complexity to make this fully dynamic, at least
it would auto-tune for the workload placed on this filesystem, and would
not initially be worse than the old single-linked list.

Cheers, Andreas






[-- Attachment #2: Message signed with OpenPGP --]
[-- Type: application/pgp-signature, Size: 873 bytes --]

Re: [PATCH 3/4] ext4: Speedup ext4 orphan inode handling

[Jan Kara]

On Thu 17-06-21 01:44:13, Andreas Dilger wrote:
> On Jun 16, 2021, at 4:56 AM, Jan Kara <jack@suse.cz> wrote:
> > 
> > Ext4 orphan inode handling is a bottleneck for workloads which heavily
> > truncate / unlink small files since it contends on the global
> > s_orphan_mutex lock (and generally it's difficult to improve scalability
> > of the ondisk linked list of orphaned inodes).
> > 
> > This patch implements new way of handling orphan inodes. Instead of
> > linking orphaned inode into a linked list, we store it's inode number in
> > a new special file which we call "orphan file". Currently we still
> > protect the orphan file with a spinlock for simplicity but even in this
> > setting we can substantially reduce the length of the critical section
> > and thus speedup some workloads.
> 
> Is it a single spinlock for the whole file?  Did you consider using
> a per-page lock or grouplock?  With a page in the orphan file for each
> CPU core, it would basically be lockless.

See the next patch :) I've made this one simple in terms of locking:

a) to be able to evaluate how global spinlock performs
b) to make code simpler for review

> > +static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
> > +{
> > 	spin_lock(&oi->of_lock);
> > +	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
> > +	if (i == oi->of_blocks) {
> > +		spin_unlock(&oi->of_lock);
> > +		/*
> > +		 * For now we don't grow or shrink orphan file. We just use
> > +		 * whatever was allocated at mke2fs time. The additional
> > +		 * credits we would have to reserve for each orphan inode
> > +		 * operation just don't seem worth it.
> > +		 */
> > +		return -ENOSPC;
> > +	}
> > +	oi->of_binfo[i].ob_free_entries--;
> > +	spin_unlock(&oi->of_lock);
> 
> How do we know how large to make the orphan file at mkfs time?  What if it
> becomes full during use?  It seems like reserving a fixed number of blocks
> will invariably be incorrect for the actual workload on the filesystem.

If orphan file gets full (too many orphaned inodes at this moment), we will
just fallback to using the good old orphan list. So only performance will
suffer.

In terms of number of blocks, for reasonably large filesystems we reserve
512 4k blocks for orphan file so that allows for 523776 orphaned inodes.
Sure it's possible to exhaust it but frankly I don't find it likely so I'm
not sure dynamic sizing is worth the hassle.

> > @@ -49,6 +95,16 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
> > 	ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
> > 		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
> > 
> > +	if (sbi->s_orphan_info.of_blocks) {
> > +		err = ext4_orphan_file_add(handle, inode);
> > +		/*
> > +		 * Fallback to normal orphan list of orphan file is
> > +		 * out of space
> > +		 */
> > +		if (err != -ENOSPC)
> > +			return err;
> > +	}
> 
> This could schedule a task on a workqueue to allocate a few more blocks?
> That could easily reserve more credits for this action, without making
> the common case more expensive.  Even if it isn't used with the current
> mount, it would be available for the next mount (which presumably would
> also need additional blocks).
> 
> Whether it is worth the complexity to make this fully dynamic, at least
> it would auto-tune for the workload placed on this filesystem, and would
> not initially be worse than the old single-linked list.

Adding more blocks would not be that hard as you say but if we are growing
a file there may be need to make it shorter as well (as e.g. shortlived
peak in number of orphaned inodes could have accumulated bazilion blocks
for orphan file) and that will be a bit more tricky. It can be done but I
don't think it's worth the complexity...

Thanks for the review!
								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH 3/4] ext4: Speedup ext4 orphan inode handling

[Lukas Czerner]

On Wed, Jun 16, 2021 at 12:56:54PM +0200, Jan Kara wrote:
> Ext4 orphan inode handling is a bottleneck for workloads which heavily
> truncate / unlink small files since it contends on the global
> s_orphan_mutex lock (and generally it's difficult to improve scalability
> of the ondisk linked list of orphaned inodes).
> 
> This patch implements new way of handling orphan inodes. Instead of
> linking orphaned inode into a linked list, we store it's inode number in
> a new special file which we call "orphan file". Currently we still
> protect the orphan file with a spinlock for simplicity but even in this
> setting we can substantially reduce the length of the critical section
> and thus speedup some workloads.
> 
> Note that the change is backwards compatible when the filesystem is
> clean - the existence of the orphan file is a compat feature, we set
> another ro-compat feature indicating orphan file needs scanning for
> orphaned inodes when mounting filesystem read-write. This ro-compat
> feature gets cleared on unmount / remount read-only.
> 
> Some performance data from 80 CPU Xeon Server with 512 GB of RAM,
> filesystem located on SSD, average of 5 runs:
> 
> stress-orphan (microbenchmark truncating files byte-by-byte from N
> processes in parallel)
> 
> Threads Time            Time
>         Vanilla         Patched
>   1       1.057200        0.945600
>   2       1.680400        1.331800
>   4       2.547000        1.995000
>   8       7.049400        6.424200
>  16      14.827800       14.937600
>  32      40.948200       33.038200
>  64      87.787400       60.823600
> 128     206.504000      122.941400
> 
> So we can see significant wins all over the board.

Hi Jan,

nice results! Comments below

> 
> Signed-off-by: Jan Kara <jack@suse.cz>
> ---
>  fs/ext4/ext4.h   |  70 +++++++++--
>  fs/ext4/orphan.c | 319 ++++++++++++++++++++++++++++++++++++++++++-----
>  fs/ext4/super.c  |  34 ++++-
>  3 files changed, 379 insertions(+), 44 deletions(-)
> 
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 33508487516f..83298c0b6dae 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -1025,7 +1025,14 @@ struct ext4_inode_info {
>  	 */
>  	struct rw_semaphore xattr_sem;
>  
> -	struct list_head i_orphan;	/* unlinked but open inodes */
> +	/*
> +	 * Inodes with EXT4_STATE_ORPHAN_FILE use i_orphan_idx. Otherwise
> +	 * i_orphan is used.
> +	 */
> +	union {
> +		struct list_head i_orphan;	/* unlinked but open inodes */
> +		unsigned int i_orphan_idx;	/* Index in orphan file */
> +	};
>  
>  	/* Fast commit related info */
>  
> @@ -1419,7 +1426,8 @@ struct ext4_super_block {
>  	__u8    s_last_error_errcode;
>  	__le16  s_encoding;		/* Filename charset encoding */
>  	__le16  s_encoding_flags;	/* Filename charset encoding flags */
> -	__le32	s_reserved[95];		/* Padding to the end of the block */
> +	__le32  s_orphan_file_inum;	/* Inode for tracking orphan inodes */
> +	__le32	s_reserved[94];		/* Padding to the end of the block */
>  	__le32	s_checksum;		/* crc32c(superblock) */
>  };
>  
> @@ -1440,6 +1448,7 @@ struct ext4_super_block {
>  
>  /* Types of ext4 journal triggers */
>  enum ext4_journal_trigger_type {
> +	EXT4_JTR_ORPHAN_FILE,
>  	EXT4_JTR_NONE	/* This must be the last entry for indexing to work! */
>  };
>  
> @@ -1456,6 +1465,36 @@ static inline struct ext4_journal_trigger *EXT4_TRIGGER(
>  	return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
>  }
>  
> +#define EXT4_ORPHAN_BLOCK_MAGIC 0x0b10ca04
> +
> +/* Structure at the tail of orphan block */
> +struct ext4_orphan_block_tail {
> +	__le32 ob_magic;
> +	__le32 ob_checksum;
> +};
> +
> +static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
> +{
> +	return (sb->s_blocksize - sizeof(struct ext4_orphan_block_tail)) /
> +			sizeof(u32);
> +}
> +
> +struct ext4_orphan_block {
> +	int ob_free_entries;	/* Number of free orphan entries in block */
> +	struct buffer_head *ob_bh;	/* Buffer for orphan block */
> +};
> +
> +/*
> + * Info about orphan file.
> + */
> +struct ext4_orphan_info {
> +	spinlock_t of_lock;
> +	int of_blocks;			/* Number of orphan blocks in a file */
> +	__u32 of_csum_seed;		/* Checksum seed for orphan file */
> +	struct ext4_orphan_block *of_binfo;	/* Array with info about orphan
> +						 * file blocks */
> +};
> +
>  /*
>   * fourth extended-fs super-block data in memory
>   */
> @@ -1509,9 +1548,11 @@ struct ext4_sb_info {
>  
>  	/* Journaling */
>  	struct journal_s *s_journal;
> -	struct list_head s_orphan;
> -	struct mutex s_orphan_lock;
>  	unsigned long s_ext4_flags;		/* Ext4 superblock flags */
> +	struct mutex s_orphan_lock;	/* Protects on disk list changes */
> +	struct list_head s_orphan;	/* List of orphaned inodes in on disk
> +					   list */
> +	struct ext4_orphan_info s_orphan_info;
>  	unsigned long s_commit_interval;
>  	u32 s_max_batch_time;
>  	u32 s_min_batch_time;
> @@ -1846,6 +1887,7 @@ enum {
>  	EXT4_STATE_LUSTRE_EA_INODE,	/* Lustre-style ea_inode */
>  	EXT4_STATE_VERITY_IN_PROGRESS,	/* building fs-verity Merkle tree */
>  	EXT4_STATE_FC_COMMITTING,	/* Fast commit ongoing */
> +	EXT4_STATE_ORPHAN_FILE,		/* Inode orphaned in orphan file */
>  };
>  
>  #define EXT4_INODE_BIT_FNS(name, field, offset)				\
> @@ -1947,6 +1989,7 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
>   */
>  #define EXT4_FEATURE_COMPAT_FAST_COMMIT		0x0400
>  #define EXT4_FEATURE_COMPAT_STABLE_INODES	0x0800
> +#define EXT4_FEATURE_COMPAT_ORPHAN_FILE		0x1000	/* Orphan file exists */
>  
>  #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER	0x0001
>  #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE	0x0002
> @@ -1955,6 +1998,8 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
>  #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM		0x0010
>  #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK	0x0020
>  #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE	0x0040
> +#define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT	0x0080 /* Orphan file may be
> +							  non-empty */
>  #define EXT4_FEATURE_RO_COMPAT_QUOTA		0x0100
>  #define EXT4_FEATURE_RO_COMPAT_BIGALLOC		0x0200
>  /*
> @@ -1964,6 +2009,7 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
>   * GDT_CSUM bits are mutually exclusive.
>   */
>  #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM	0x0400
> +/* 0x0800 Reserved for EXT4_FEATURE_RO_COMPAT_REPLICA */
>  #define EXT4_FEATURE_RO_COMPAT_READONLY		0x1000
>  #define EXT4_FEATURE_RO_COMPAT_PROJECT		0x2000
>  #define EXT4_FEATURE_RO_COMPAT_VERITY		0x8000
> @@ -2050,6 +2096,7 @@ EXT4_FEATURE_COMPAT_FUNCS(dir_index,		DIR_INDEX)
>  EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,	SPARSE_SUPER2)
>  EXT4_FEATURE_COMPAT_FUNCS(fast_commit,		FAST_COMMIT)
>  EXT4_FEATURE_COMPAT_FUNCS(stable_inodes,	STABLE_INODES)
> +EXT4_FEATURE_COMPAT_FUNCS(orphan_file,		ORPHAN_FILE)
>  
>  EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,	SPARSE_SUPER)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,	LARGE_FILE)
> @@ -2064,6 +2111,7 @@ EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,	METADATA_CSUM)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,		READONLY)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(project,		PROJECT)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(verity,		VERITY)
> +EXT4_FEATURE_RO_COMPAT_FUNCS(orphan_present,	ORPHAN_PRESENT)
>  
>  EXT4_FEATURE_INCOMPAT_FUNCS(compression,	COMPRESSION)
>  EXT4_FEATURE_INCOMPAT_FUNCS(filetype,		FILETYPE)
> @@ -2097,7 +2145,8 @@ EXT4_FEATURE_INCOMPAT_FUNCS(casefold,		CASEFOLD)
>  					 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
>  					 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
>  
> -#define EXT4_FEATURE_COMPAT_SUPP	EXT4_FEATURE_COMPAT_EXT_ATTR
> +#define EXT4_FEATURE_COMPAT_SUPP	(EXT4_FEATURE_COMPAT_EXT_ATTR| \
> +					 EXT4_FEATURE_COMPAT_ORPHAN_FILE)
>  #define EXT4_FEATURE_INCOMPAT_SUPP	(EXT4_FEATURE_INCOMPAT_FILETYPE| \
>  					 EXT4_FEATURE_INCOMPAT_RECOVER| \
>  					 EXT4_FEATURE_INCOMPAT_META_BG| \
> @@ -2122,7 +2171,8 @@ EXT4_FEATURE_INCOMPAT_FUNCS(casefold,		CASEFOLD)
>  					 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
>  					 EXT4_FEATURE_RO_COMPAT_QUOTA |\
>  					 EXT4_FEATURE_RO_COMPAT_PROJECT |\
> -					 EXT4_FEATURE_RO_COMPAT_VERITY)
> +					 EXT4_FEATURE_RO_COMPAT_VERITY |\
> +					 EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
>  
>  #define EXTN_FEATURE_FUNCS(ver) \
>  static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
> @@ -2172,7 +2222,6 @@ static inline int ext4_forced_shutdown(struct ext4_sb_info *sbi)
>  	return test_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
>  }
>  
> -
>  /*
>   * Default values for user and/or group using reserved blocks
>   */
> @@ -3751,6 +3800,13 @@ extern int ext4_orphan_add(handle_t *, struct inode *);
>  extern int ext4_orphan_del(handle_t *, struct inode *);
>  extern void ext4_orphan_cleanup(struct super_block *sb,
>  				struct ext4_super_block *es);
> +extern void ext4_release_orphan_info(struct super_block *sb);
> +extern int ext4_init_orphan_info(struct super_block *sb);
> +extern int ext4_orphan_file_empty(struct super_block *sb);
> +extern void ext4_orphan_file_block_trigger(
> +				struct jbd2_buffer_trigger_type *triggers,
> +				struct buffer_head *bh,
> +				void *data, size_t size);
>  
>  /*
>   * Add new method to test whether block and inode bitmaps are properly
> diff --git a/fs/ext4/orphan.c b/fs/ext4/orphan.c
> index 732b16ef655b..ac22667b7fd5 100644
> --- a/fs/ext4/orphan.c
> +++ b/fs/ext4/orphan.c
> @@ -8,6 +8,52 @@
>  #include "ext4.h"
>  #include "ext4_jbd2.h"
>  
> +static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
> +{
> +	int i, j;
> +	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
> +	int ret = 0;
> +	__le32 *bdata;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
> +
> +	spin_lock(&oi->of_lock);
> +	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
> +	if (i == oi->of_blocks) {
> +		spin_unlock(&oi->of_lock);
> +		/*
> +		 * For now we don't grow or shrink orphan file. We just use
> +		 * whatever was allocated at mke2fs time. The additional
> +		 * credits we would have to reserve for each orphan inode
> +		 * operation just don't seem worth it.
> +		 */
> +		return -ENOSPC;
> +	}
> +	oi->of_binfo[i].ob_free_entries--;
> +	spin_unlock(&oi->of_lock);
> +
> +	/*
> +	 * Get access to orphan block. We have dropped of_lock but since we
> +	 * have decremented number of free entries we are guaranteed free entry
> +	 * in our block.
> +	 */
> +	ret = ext4_journal_get_write_access(handle, inode->i_sb,
> +				oi->of_binfo[i].ob_bh, EXT4_JTR_ORPHAN_FILE);
> +	if (ret)
> +		return ret;

We've already decremented the number of free entries at this point. Shouldn't
we revert that ?

> +
> +	bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
> +	spin_lock(&oi->of_lock);
> +	/* Find empty slot in a block */
> +	for (j = 0; j < inodes_per_ob && bdata[j]; j++);
> +	BUG_ON(j == inodes_per_ob);

While BUG_ON() is probably fine here, can we do better ? AFAICT we have
not done any permanent changes yet and it should be able to recover and
let it fall back to the orphan list method. With an appropriate error
of course.

> +	bdata[j] = cpu_to_le32(inode->i_ino);
> +	EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
> +	ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> +	spin_unlock(&oi->of_lock);
> +
> +	return ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[i].ob_bh);
> +}
> +
>  /*
>   * ext4_orphan_add() links an unlinked or truncated inode into a list of
>   * such inodes, starting at the superblock, in case we crash before the
> @@ -34,10 +80,10 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
>  	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
>  		     !inode_is_locked(inode));
>  	/*
> -	 * Exit early if inode already is on orphan list. This is a big speedup
> -	 * since we don't have to contend on the global s_orphan_lock.
> +	 * Inode orphaned in orphan file or in orphan list?
>  	 */
> -	if (!list_empty(&EXT4_I(inode)->i_orphan))
> +	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE) ||
> +	    !list_empty(&EXT4_I(inode)->i_orphan))
>  		return 0;
>  
>  	/*
> @@ -49,6 +95,16 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
>  	ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
>  		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
>  
> +	if (sbi->s_orphan_info.of_blocks) {
> +		err = ext4_orphan_file_add(handle, inode);
> +		/*
> +		 * Fallback to normal orphan list of orphan file is
> +		 * out of space
> +		 */
> +		if (err != -ENOSPC)
> +			return err;
> +	}
> +
>  	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
>  	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
>  					    EXT4_JTR_NONE);
> @@ -103,6 +159,37 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
>  	return err;
>  }
>  
> +static int ext4_orphan_file_del(handle_t *handle, struct inode *inode)
> +{
> +	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
> +	__le32 *bdata;
> +	int blk, off;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
> +	int ret = 0;
> +
> +	if (!handle)
> +		goto out;
> +	blk = EXT4_I(inode)->i_orphan_idx / inodes_per_ob;
> +	off = EXT4_I(inode)->i_orphan_idx % inodes_per_ob;

Maybe we can be a bit defensive here and at least check that blk is sane
?

> +
> +	ret = ext4_journal_get_write_access(handle, inode->i_sb,
> +				oi->of_binfo[blk].ob_bh, EXT4_JTR_ORPHAN_FILE);
> +	if (ret)
> +		goto out;
> +
> +	bdata = (__le32 *)(oi->of_binfo[blk].ob_bh->b_data);
> +	spin_lock(&oi->of_lock);
> +	bdata[off] = 0;
> +	oi->of_binfo[blk].ob_free_entries++;
> +	spin_unlock(&oi->of_lock);
> +	ret = ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[blk].ob_bh);
> +out:
> +	ext4_clear_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> +	INIT_LIST_HEAD(&EXT4_I(inode)->i_orphan);
> +
> +	return ret;
> +}
> +
>  /*
>   * ext4_orphan_del() removes an unlinked or truncated inode from the list
>   * of such inodes stored on disk, because it is finally being cleaned up.
> @@ -121,6 +208,9 @@ int ext4_orphan_del(handle_t *handle, struct inode *inode)
>  
>  	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
>  		     !inode_is_locked(inode));
> +	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE))
> +		return ext4_orphan_file_del(handle, inode);
> +
>  	/* Do this quick check before taking global s_orphan_lock. */
>  	if (list_empty(&ei->i_orphan))
>  		return 0;
> @@ -196,6 +286,39 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
>  					EXT4_SB(sb)->s_jquota_fmt, type);
>  }
>  
> +static void ext4_process_orphan(struct inode *inode,
> +				int *nr_truncates, int *nr_orphans)
> +{
> +	struct super_block *sb = inode->i_sb;
> +	int ret;
> +
> +	dquot_initialize(inode);
> +	if (inode->i_nlink) {
> +		if (test_opt(sb, DEBUG))
> +			ext4_msg(sb, KERN_DEBUG,
> +				"%s: truncating inode %lu to %lld bytes",
> +				__func__, inode->i_ino, inode->i_size);
> +		jbd_debug(2, "truncating inode %lu to %lld bytes\n",
> +			  inode->i_ino, inode->i_size);
> +		inode_lock(inode);
> +		truncate_inode_pages(inode->i_mapping, inode->i_size);
> +		ret = ext4_truncate(inode);
> +		if (ret)
> +			ext4_std_error(inode->i_sb, ret);
> +		inode_unlock(inode);
> +		(*nr_truncates)++;
> +	} else {
> +		if (test_opt(sb, DEBUG))
> +			ext4_msg(sb, KERN_DEBUG,
> +				"%s: deleting unreferenced inode %lu",
> +				__func__, inode->i_ino);
> +		jbd_debug(2, "deleting unreferenced inode %lu\n",
> +			  inode->i_ino);
> +		(*nr_orphans)++;
> +	}
> +	iput(inode);  /* The delete magic happens here! */
> +}
> +
>  /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
>   * the superblock) which were deleted from all directories, but held open by
>   * a process at the time of a crash.  We walk the list and try to delete these
> @@ -216,12 +339,17 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
>  void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  {
>  	unsigned int s_flags = sb->s_flags;
> -	int ret, nr_orphans = 0, nr_truncates = 0;
> +	int nr_orphans = 0, nr_truncates = 0;
> +	struct inode *inode;
> +	int i, j;
>  #ifdef CONFIG_QUOTA
>  	int quota_update = 0;
> -	int i;
>  #endif
> -	if (!es->s_last_orphan) {
> +	__le32 *bdata;
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +
> +	if (!es->s_last_orphan && !oi->of_blocks) {
>  		jbd_debug(4, "no orphan inodes to clean up\n");
>  		return;
>  	}
> @@ -285,8 +413,6 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  #endif
>  
>  	while (es->s_last_orphan) {
> -		struct inode *inode;
> -
>  		/*
>  		 * We may have encountered an error during cleanup; if
>  		 * so, skip the rest.
> @@ -304,31 +430,21 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  		}
>  
>  		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
> -		dquot_initialize(inode);
> -		if (inode->i_nlink) {
> -			if (test_opt(sb, DEBUG))
> -				ext4_msg(sb, KERN_DEBUG,
> -					"%s: truncating inode %lu to %lld bytes",
> -					__func__, inode->i_ino, inode->i_size);
> -			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
> -				  inode->i_ino, inode->i_size);
> -			inode_lock(inode);
> -			truncate_inode_pages(inode->i_mapping, inode->i_size);
> -			ret = ext4_truncate(inode);
> -			if (ret)
> -				ext4_std_error(inode->i_sb, ret);
> -			inode_unlock(inode);
> -			nr_truncates++;
> -		} else {
> -			if (test_opt(sb, DEBUG))
> -				ext4_msg(sb, KERN_DEBUG,
> -					"%s: deleting unreferenced inode %lu",
> -					__func__, inode->i_ino);
> -			jbd_debug(2, "deleting unreferenced inode %lu\n",
> -				  inode->i_ino);
> -			nr_orphans++;
> +		ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
> +	}
> +
> +	for (i = 0; i < oi->of_blocks; i++) {
> +		bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
> +		for (j = 0; j < inodes_per_ob; j++) {
> +			if (!bdata[j])
> +				continue;
> +			inode = ext4_orphan_get(sb, le32_to_cpu(bdata[j]));
> +			if (IS_ERR(inode))
> +				continue;
> +			ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> +			EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
> +			ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
>  		}
> -		iput(inode);  /* The delete magic happens here! */
>  	}
>  
>  #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
> @@ -350,3 +466,142 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  #endif
>  	sb->s_flags = s_flags; /* Restore SB_RDONLY status */
>  }
> +
> +void ext4_release_orphan_info(struct super_block *sb)
> +{
> +	int i;
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +
> +	if (!oi->of_blocks)
> +		return;
> +	for (i = 0; i < oi->of_blocks; i++)
> +		brelse(oi->of_binfo[i].ob_bh);
> +	kfree(oi->of_binfo);
> +}
> +
> +static struct ext4_orphan_block_tail *ext4_orphan_block_tail(
> +						struct super_block *sb,
> +						struct buffer_head *bh)
> +{
> +	return (struct ext4_orphan_block_tail *)(bh->b_data + sb->s_blocksize -
> +				sizeof(struct ext4_orphan_block_tail));
> +}
> +
> +static int ext4_orphan_file_block_csum_verify(struct super_block *sb,
> +					      struct buffer_head *bh)
> +{
> +	__u32 calculated;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	struct ext4_orphan_block_tail *ot;
> +
> +	if (!ext4_has_metadata_csum(sb))
> +		return 1;
> +
> +	ot = ext4_orphan_block_tail(sb, bh);
> +	calculated = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed,
> +				 (__u8 *)bh->b_data,
> +				 inodes_per_ob * sizeof(__u32));
> +	return le32_to_cpu(ot->ob_checksum) == calculated;
> +}
> +
> +/* This gets called only when checksumming is enabled */
> +void ext4_orphan_file_block_trigger(struct jbd2_buffer_trigger_type *triggers,
> +				    struct buffer_head *bh,
> +				    void *data, size_t size)
> +{
> +	struct super_block *sb = EXT4_TRIGGER(triggers)->sb;
> +	__u32 csum;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	struct ext4_orphan_block_tail *ot;
> +
> +	csum = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed, (__u8 *)data,
> +			   inodes_per_ob * sizeof(__u32));
> +	ot = ext4_orphan_block_tail(sb, bh);
> +	ot->ob_checksum = cpu_to_le32(csum);
> +}
> +
> +int ext4_init_orphan_info(struct super_block *sb)
> +{
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	struct inode *inode;
> +	int i, j;
> +	int ret;
> +	int free;
> +	__le32 *bdata;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +	struct ext4_orphan_block_tail *ot;
> +	ino_t orphan_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_orphan_file_inum);
> +
> +	spin_lock_init(&oi->of_lock);

Do we need to init the lock even though the feature is not enabled ? Are
we using it somewhere I am missing ?

Thanks!
-Lukas

Re: [PATCH 3/4] ext4: Speedup ext4 orphan inode handling

[Darrick J. Wong]

On Wed, Jun 16, 2021 at 12:56:54PM +0200, Jan Kara wrote:
> Ext4 orphan inode handling is a bottleneck for workloads which heavily
> truncate / unlink small files since it contends on the global
> s_orphan_mutex lock (and generally it's difficult to improve scalability
> of the ondisk linked list of orphaned inodes).
> 
> This patch implements new way of handling orphan inodes. Instead of
> linking orphaned inode into a linked list, we store it's inode number in
> a new special file which we call "orphan file". Currently we still
> protect the orphan file with a spinlock for simplicity but even in this
> setting we can substantially reduce the length of the critical section
> and thus speedup some workloads.
> 
> Note that the change is backwards compatible when the filesystem is
> clean - the existence of the orphan file is a compat feature, we set
> another ro-compat feature indicating orphan file needs scanning for
> orphaned inodes when mounting filesystem read-write. This ro-compat
> feature gets cleared on unmount / remount read-only.
> 
> Some performance data from 80 CPU Xeon Server with 512 GB of RAM,
> filesystem located on SSD, average of 5 runs:
> 
> stress-orphan (microbenchmark truncating files byte-by-byte from N
> processes in parallel)
> 
> Threads Time            Time
>         Vanilla         Patched
>   1       1.057200        0.945600
>   2       1.680400        1.331800
>   4       2.547000        1.995000
>   8       7.049400        6.424200
>  16      14.827800       14.937600
>  32      40.948200       33.038200
>  64      87.787400       60.823600
> 128     206.504000      122.941400
> 
> So we can see significant wins all over the board.
> 
> Signed-off-by: Jan Kara <jack@suse.cz>
> ---
>  fs/ext4/ext4.h   |  70 +++++++++--
>  fs/ext4/orphan.c | 319 ++++++++++++++++++++++++++++++++++++++++++-----
>  fs/ext4/super.c  |  34 ++++-
>  3 files changed, 379 insertions(+), 44 deletions(-)
> 
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 33508487516f..83298c0b6dae 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -1025,7 +1025,14 @@ struct ext4_inode_info {
>  	 */
>  	struct rw_semaphore xattr_sem;
>  
> -	struct list_head i_orphan;	/* unlinked but open inodes */
> +	/*
> +	 * Inodes with EXT4_STATE_ORPHAN_FILE use i_orphan_idx. Otherwise
> +	 * i_orphan is used.
> +	 */
> +	union {
> +		struct list_head i_orphan;	/* unlinked but open inodes */
> +		unsigned int i_orphan_idx;	/* Index in orphan file */
> +	};
>  
>  	/* Fast commit related info */
>  
> @@ -1419,7 +1426,8 @@ struct ext4_super_block {
>  	__u8    s_last_error_errcode;
>  	__le16  s_encoding;		/* Filename charset encoding */
>  	__le16  s_encoding_flags;	/* Filename charset encoding flags */
> -	__le32	s_reserved[95];		/* Padding to the end of the block */
> +	__le32  s_orphan_file_inum;	/* Inode for tracking orphan inodes */
> +	__le32	s_reserved[94];		/* Padding to the end of the block */
>  	__le32	s_checksum;		/* crc32c(superblock) */
>  };
>  
> @@ -1440,6 +1448,7 @@ struct ext4_super_block {
>  
>  /* Types of ext4 journal triggers */
>  enum ext4_journal_trigger_type {
> +	EXT4_JTR_ORPHAN_FILE,
>  	EXT4_JTR_NONE	/* This must be the last entry for indexing to work! */
>  };
>  
> @@ -1456,6 +1465,36 @@ static inline struct ext4_journal_trigger *EXT4_TRIGGER(
>  	return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
>  }
>  
> +#define EXT4_ORPHAN_BLOCK_MAGIC 0x0b10ca04
> +
> +/* Structure at the tail of orphan block */
> +struct ext4_orphan_block_tail {
> +	__le32 ob_magic;
> +	__le32 ob_checksum;
> +};

Can you add the ondisk format changes to the appropriate place in
Documentation/filesystems/ext4/ please?

--D

> +
> +static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
> +{
> +	return (sb->s_blocksize - sizeof(struct ext4_orphan_block_tail)) /
> +			sizeof(u32);
> +}
> +
> +struct ext4_orphan_block {
> +	int ob_free_entries;	/* Number of free orphan entries in block */
> +	struct buffer_head *ob_bh;	/* Buffer for orphan block */
> +};
> +
> +/*
> + * Info about orphan file.
> + */
> +struct ext4_orphan_info {
> +	spinlock_t of_lock;
> +	int of_blocks;			/* Number of orphan blocks in a file */
> +	__u32 of_csum_seed;		/* Checksum seed for orphan file */
> +	struct ext4_orphan_block *of_binfo;	/* Array with info about orphan
> +						 * file blocks */
> +};
> +
>  /*
>   * fourth extended-fs super-block data in memory
>   */
> @@ -1509,9 +1548,11 @@ struct ext4_sb_info {
>  
>  	/* Journaling */
>  	struct journal_s *s_journal;
> -	struct list_head s_orphan;
> -	struct mutex s_orphan_lock;
>  	unsigned long s_ext4_flags;		/* Ext4 superblock flags */
> +	struct mutex s_orphan_lock;	/* Protects on disk list changes */
> +	struct list_head s_orphan;	/* List of orphaned inodes in on disk
> +					   list */
> +	struct ext4_orphan_info s_orphan_info;
>  	unsigned long s_commit_interval;
>  	u32 s_max_batch_time;
>  	u32 s_min_batch_time;
> @@ -1846,6 +1887,7 @@ enum {
>  	EXT4_STATE_LUSTRE_EA_INODE,	/* Lustre-style ea_inode */
>  	EXT4_STATE_VERITY_IN_PROGRESS,	/* building fs-verity Merkle tree */
>  	EXT4_STATE_FC_COMMITTING,	/* Fast commit ongoing */
> +	EXT4_STATE_ORPHAN_FILE,		/* Inode orphaned in orphan file */
>  };
>  
>  #define EXT4_INODE_BIT_FNS(name, field, offset)				\
> @@ -1947,6 +1989,7 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
>   */
>  #define EXT4_FEATURE_COMPAT_FAST_COMMIT		0x0400
>  #define EXT4_FEATURE_COMPAT_STABLE_INODES	0x0800
> +#define EXT4_FEATURE_COMPAT_ORPHAN_FILE		0x1000	/* Orphan file exists */
>  
>  #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER	0x0001
>  #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE	0x0002
> @@ -1955,6 +1998,8 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
>  #define EXT4_FEATURE_RO_COMPAT_GDT_CSUM		0x0010
>  #define EXT4_FEATURE_RO_COMPAT_DIR_NLINK	0x0020
>  #define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE	0x0040
> +#define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT	0x0080 /* Orphan file may be
> +							  non-empty */
>  #define EXT4_FEATURE_RO_COMPAT_QUOTA		0x0100
>  #define EXT4_FEATURE_RO_COMPAT_BIGALLOC		0x0200
>  /*
> @@ -1964,6 +2009,7 @@ static inline bool ext4_verity_in_progress(struct inode *inode)
>   * GDT_CSUM bits are mutually exclusive.
>   */
>  #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM	0x0400
> +/* 0x0800 Reserved for EXT4_FEATURE_RO_COMPAT_REPLICA */
>  #define EXT4_FEATURE_RO_COMPAT_READONLY		0x1000
>  #define EXT4_FEATURE_RO_COMPAT_PROJECT		0x2000
>  #define EXT4_FEATURE_RO_COMPAT_VERITY		0x8000
> @@ -2050,6 +2096,7 @@ EXT4_FEATURE_COMPAT_FUNCS(dir_index,		DIR_INDEX)
>  EXT4_FEATURE_COMPAT_FUNCS(sparse_super2,	SPARSE_SUPER2)
>  EXT4_FEATURE_COMPAT_FUNCS(fast_commit,		FAST_COMMIT)
>  EXT4_FEATURE_COMPAT_FUNCS(stable_inodes,	STABLE_INODES)
> +EXT4_FEATURE_COMPAT_FUNCS(orphan_file,		ORPHAN_FILE)
>  
>  EXT4_FEATURE_RO_COMPAT_FUNCS(sparse_super,	SPARSE_SUPER)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(large_file,	LARGE_FILE)
> @@ -2064,6 +2111,7 @@ EXT4_FEATURE_RO_COMPAT_FUNCS(metadata_csum,	METADATA_CSUM)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(readonly,		READONLY)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(project,		PROJECT)
>  EXT4_FEATURE_RO_COMPAT_FUNCS(verity,		VERITY)
> +EXT4_FEATURE_RO_COMPAT_FUNCS(orphan_present,	ORPHAN_PRESENT)
>  
>  EXT4_FEATURE_INCOMPAT_FUNCS(compression,	COMPRESSION)
>  EXT4_FEATURE_INCOMPAT_FUNCS(filetype,		FILETYPE)
> @@ -2097,7 +2145,8 @@ EXT4_FEATURE_INCOMPAT_FUNCS(casefold,		CASEFOLD)
>  					 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
>  					 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
>  
> -#define EXT4_FEATURE_COMPAT_SUPP	EXT4_FEATURE_COMPAT_EXT_ATTR
> +#define EXT4_FEATURE_COMPAT_SUPP	(EXT4_FEATURE_COMPAT_EXT_ATTR| \
> +					 EXT4_FEATURE_COMPAT_ORPHAN_FILE)
>  #define EXT4_FEATURE_INCOMPAT_SUPP	(EXT4_FEATURE_INCOMPAT_FILETYPE| \
>  					 EXT4_FEATURE_INCOMPAT_RECOVER| \
>  					 EXT4_FEATURE_INCOMPAT_META_BG| \
> @@ -2122,7 +2171,8 @@ EXT4_FEATURE_INCOMPAT_FUNCS(casefold,		CASEFOLD)
>  					 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
>  					 EXT4_FEATURE_RO_COMPAT_QUOTA |\
>  					 EXT4_FEATURE_RO_COMPAT_PROJECT |\
> -					 EXT4_FEATURE_RO_COMPAT_VERITY)
> +					 EXT4_FEATURE_RO_COMPAT_VERITY |\
> +					 EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
>  
>  #define EXTN_FEATURE_FUNCS(ver) \
>  static inline bool ext4_has_unknown_ext##ver##_compat_features(struct super_block *sb) \
> @@ -2172,7 +2222,6 @@ static inline int ext4_forced_shutdown(struct ext4_sb_info *sbi)
>  	return test_bit(EXT4_FLAGS_SHUTDOWN, &sbi->s_ext4_flags);
>  }
>  
> -
>  /*
>   * Default values for user and/or group using reserved blocks
>   */
> @@ -3751,6 +3800,13 @@ extern int ext4_orphan_add(handle_t *, struct inode *);
>  extern int ext4_orphan_del(handle_t *, struct inode *);
>  extern void ext4_orphan_cleanup(struct super_block *sb,
>  				struct ext4_super_block *es);
> +extern void ext4_release_orphan_info(struct super_block *sb);
> +extern int ext4_init_orphan_info(struct super_block *sb);
> +extern int ext4_orphan_file_empty(struct super_block *sb);
> +extern void ext4_orphan_file_block_trigger(
> +				struct jbd2_buffer_trigger_type *triggers,
> +				struct buffer_head *bh,
> +				void *data, size_t size);
>  
>  /*
>   * Add new method to test whether block and inode bitmaps are properly
> diff --git a/fs/ext4/orphan.c b/fs/ext4/orphan.c
> index 732b16ef655b..ac22667b7fd5 100644
> --- a/fs/ext4/orphan.c
> +++ b/fs/ext4/orphan.c
> @@ -8,6 +8,52 @@
>  #include "ext4.h"
>  #include "ext4_jbd2.h"
>  
> +static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
> +{
> +	int i, j;
> +	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
> +	int ret = 0;
> +	__le32 *bdata;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
> +
> +	spin_lock(&oi->of_lock);
> +	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
> +	if (i == oi->of_blocks) {
> +		spin_unlock(&oi->of_lock);
> +		/*
> +		 * For now we don't grow or shrink orphan file. We just use
> +		 * whatever was allocated at mke2fs time. The additional
> +		 * credits we would have to reserve for each orphan inode
> +		 * operation just don't seem worth it.
> +		 */
> +		return -ENOSPC;
> +	}
> +	oi->of_binfo[i].ob_free_entries--;
> +	spin_unlock(&oi->of_lock);
> +
> +	/*
> +	 * Get access to orphan block. We have dropped of_lock but since we
> +	 * have decremented number of free entries we are guaranteed free entry
> +	 * in our block.
> +	 */
> +	ret = ext4_journal_get_write_access(handle, inode->i_sb,
> +				oi->of_binfo[i].ob_bh, EXT4_JTR_ORPHAN_FILE);
> +	if (ret)
> +		return ret;
> +
> +	bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
> +	spin_lock(&oi->of_lock);
> +	/* Find empty slot in a block */
> +	for (j = 0; j < inodes_per_ob && bdata[j]; j++);
> +	BUG_ON(j == inodes_per_ob);
> +	bdata[j] = cpu_to_le32(inode->i_ino);
> +	EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
> +	ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> +	spin_unlock(&oi->of_lock);
> +
> +	return ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[i].ob_bh);
> +}
> +
>  /*
>   * ext4_orphan_add() links an unlinked or truncated inode into a list of
>   * such inodes, starting at the superblock, in case we crash before the
> @@ -34,10 +80,10 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
>  	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
>  		     !inode_is_locked(inode));
>  	/*
> -	 * Exit early if inode already is on orphan list. This is a big speedup
> -	 * since we don't have to contend on the global s_orphan_lock.
> +	 * Inode orphaned in orphan file or in orphan list?
>  	 */
> -	if (!list_empty(&EXT4_I(inode)->i_orphan))
> +	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE) ||
> +	    !list_empty(&EXT4_I(inode)->i_orphan))
>  		return 0;
>  
>  	/*
> @@ -49,6 +95,16 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
>  	ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
>  		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
>  
> +	if (sbi->s_orphan_info.of_blocks) {
> +		err = ext4_orphan_file_add(handle, inode);
> +		/*
> +		 * Fallback to normal orphan list of orphan file is
> +		 * out of space
> +		 */
> +		if (err != -ENOSPC)
> +			return err;
> +	}
> +
>  	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
>  	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
>  					    EXT4_JTR_NONE);
> @@ -103,6 +159,37 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
>  	return err;
>  }
>  
> +static int ext4_orphan_file_del(handle_t *handle, struct inode *inode)
> +{
> +	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
> +	__le32 *bdata;
> +	int blk, off;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
> +	int ret = 0;
> +
> +	if (!handle)
> +		goto out;
> +	blk = EXT4_I(inode)->i_orphan_idx / inodes_per_ob;
> +	off = EXT4_I(inode)->i_orphan_idx % inodes_per_ob;
> +
> +	ret = ext4_journal_get_write_access(handle, inode->i_sb,
> +				oi->of_binfo[blk].ob_bh, EXT4_JTR_ORPHAN_FILE);
> +	if (ret)
> +		goto out;
> +
> +	bdata = (__le32 *)(oi->of_binfo[blk].ob_bh->b_data);
> +	spin_lock(&oi->of_lock);
> +	bdata[off] = 0;
> +	oi->of_binfo[blk].ob_free_entries++;
> +	spin_unlock(&oi->of_lock);
> +	ret = ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[blk].ob_bh);
> +out:
> +	ext4_clear_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> +	INIT_LIST_HEAD(&EXT4_I(inode)->i_orphan);
> +
> +	return ret;
> +}
> +
>  /*
>   * ext4_orphan_del() removes an unlinked or truncated inode from the list
>   * of such inodes stored on disk, because it is finally being cleaned up.
> @@ -121,6 +208,9 @@ int ext4_orphan_del(handle_t *handle, struct inode *inode)
>  
>  	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
>  		     !inode_is_locked(inode));
> +	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE))
> +		return ext4_orphan_file_del(handle, inode);
> +
>  	/* Do this quick check before taking global s_orphan_lock. */
>  	if (list_empty(&ei->i_orphan))
>  		return 0;
> @@ -196,6 +286,39 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
>  					EXT4_SB(sb)->s_jquota_fmt, type);
>  }
>  
> +static void ext4_process_orphan(struct inode *inode,
> +				int *nr_truncates, int *nr_orphans)
> +{
> +	struct super_block *sb = inode->i_sb;
> +	int ret;
> +
> +	dquot_initialize(inode);
> +	if (inode->i_nlink) {
> +		if (test_opt(sb, DEBUG))
> +			ext4_msg(sb, KERN_DEBUG,
> +				"%s: truncating inode %lu to %lld bytes",
> +				__func__, inode->i_ino, inode->i_size);
> +		jbd_debug(2, "truncating inode %lu to %lld bytes\n",
> +			  inode->i_ino, inode->i_size);
> +		inode_lock(inode);
> +		truncate_inode_pages(inode->i_mapping, inode->i_size);
> +		ret = ext4_truncate(inode);
> +		if (ret)
> +			ext4_std_error(inode->i_sb, ret);
> +		inode_unlock(inode);
> +		(*nr_truncates)++;
> +	} else {
> +		if (test_opt(sb, DEBUG))
> +			ext4_msg(sb, KERN_DEBUG,
> +				"%s: deleting unreferenced inode %lu",
> +				__func__, inode->i_ino);
> +		jbd_debug(2, "deleting unreferenced inode %lu\n",
> +			  inode->i_ino);
> +		(*nr_orphans)++;
> +	}
> +	iput(inode);  /* The delete magic happens here! */
> +}
> +
>  /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
>   * the superblock) which were deleted from all directories, but held open by
>   * a process at the time of a crash.  We walk the list and try to delete these
> @@ -216,12 +339,17 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
>  void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  {
>  	unsigned int s_flags = sb->s_flags;
> -	int ret, nr_orphans = 0, nr_truncates = 0;
> +	int nr_orphans = 0, nr_truncates = 0;
> +	struct inode *inode;
> +	int i, j;
>  #ifdef CONFIG_QUOTA
>  	int quota_update = 0;
> -	int i;
>  #endif
> -	if (!es->s_last_orphan) {
> +	__le32 *bdata;
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +
> +	if (!es->s_last_orphan && !oi->of_blocks) {
>  		jbd_debug(4, "no orphan inodes to clean up\n");
>  		return;
>  	}
> @@ -285,8 +413,6 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  #endif
>  
>  	while (es->s_last_orphan) {
> -		struct inode *inode;
> -
>  		/*
>  		 * We may have encountered an error during cleanup; if
>  		 * so, skip the rest.
> @@ -304,31 +430,21 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  		}
>  
>  		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
> -		dquot_initialize(inode);
> -		if (inode->i_nlink) {
> -			if (test_opt(sb, DEBUG))
> -				ext4_msg(sb, KERN_DEBUG,
> -					"%s: truncating inode %lu to %lld bytes",
> -					__func__, inode->i_ino, inode->i_size);
> -			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
> -				  inode->i_ino, inode->i_size);
> -			inode_lock(inode);
> -			truncate_inode_pages(inode->i_mapping, inode->i_size);
> -			ret = ext4_truncate(inode);
> -			if (ret)
> -				ext4_std_error(inode->i_sb, ret);
> -			inode_unlock(inode);
> -			nr_truncates++;
> -		} else {
> -			if (test_opt(sb, DEBUG))
> -				ext4_msg(sb, KERN_DEBUG,
> -					"%s: deleting unreferenced inode %lu",
> -					__func__, inode->i_ino);
> -			jbd_debug(2, "deleting unreferenced inode %lu\n",
> -				  inode->i_ino);
> -			nr_orphans++;
> +		ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
> +	}
> +
> +	for (i = 0; i < oi->of_blocks; i++) {
> +		bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
> +		for (j = 0; j < inodes_per_ob; j++) {
> +			if (!bdata[j])
> +				continue;
> +			inode = ext4_orphan_get(sb, le32_to_cpu(bdata[j]));
> +			if (IS_ERR(inode))
> +				continue;
> +			ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> +			EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
> +			ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
>  		}
> -		iput(inode);  /* The delete magic happens here! */
>  	}
>  
>  #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
> @@ -350,3 +466,142 @@ void ext4_orphan_cleanup(struct super_block *sb, struct ext4_super_block *es)
>  #endif
>  	sb->s_flags = s_flags; /* Restore SB_RDONLY status */
>  }
> +
> +void ext4_release_orphan_info(struct super_block *sb)
> +{
> +	int i;
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +
> +	if (!oi->of_blocks)
> +		return;
> +	for (i = 0; i < oi->of_blocks; i++)
> +		brelse(oi->of_binfo[i].ob_bh);
> +	kfree(oi->of_binfo);
> +}
> +
> +static struct ext4_orphan_block_tail *ext4_orphan_block_tail(
> +						struct super_block *sb,
> +						struct buffer_head *bh)
> +{
> +	return (struct ext4_orphan_block_tail *)(bh->b_data + sb->s_blocksize -
> +				sizeof(struct ext4_orphan_block_tail));
> +}
> +
> +static int ext4_orphan_file_block_csum_verify(struct super_block *sb,
> +					      struct buffer_head *bh)
> +{
> +	__u32 calculated;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	struct ext4_orphan_block_tail *ot;
> +
> +	if (!ext4_has_metadata_csum(sb))
> +		return 1;
> +
> +	ot = ext4_orphan_block_tail(sb, bh);
> +	calculated = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed,
> +				 (__u8 *)bh->b_data,
> +				 inodes_per_ob * sizeof(__u32));
> +	return le32_to_cpu(ot->ob_checksum) == calculated;
> +}
> +
> +/* This gets called only when checksumming is enabled */
> +void ext4_orphan_file_block_trigger(struct jbd2_buffer_trigger_type *triggers,
> +				    struct buffer_head *bh,
> +				    void *data, size_t size)
> +{
> +	struct super_block *sb = EXT4_TRIGGER(triggers)->sb;
> +	__u32 csum;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	struct ext4_orphan_block_tail *ot;
> +
> +	csum = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed, (__u8 *)data,
> +			   inodes_per_ob * sizeof(__u32));
> +	ot = ext4_orphan_block_tail(sb, bh);
> +	ot->ob_checksum = cpu_to_le32(csum);
> +}
> +
> +int ext4_init_orphan_info(struct super_block *sb)
> +{
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	struct inode *inode;
> +	int i, j;
> +	int ret;
> +	int free;
> +	__le32 *bdata;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +	struct ext4_orphan_block_tail *ot;
> +	ino_t orphan_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_orphan_file_inum);
> +
> +	spin_lock_init(&oi->of_lock);
> +
> +	if (!ext4_has_feature_orphan_file(sb))
> +		return 0;
> +
> +	inode = ext4_iget(sb, orphan_ino, EXT4_IGET_NORMAL);
> +	if (IS_ERR(inode)) {
> +		ext4_msg(sb, KERN_ERR, "get orphan inode failed");
> +		return PTR_ERR(inode);
> +	}
> +	oi->of_blocks = inode->i_size >> sb->s_blocksize_bits;
> +	oi->of_csum_seed = EXT4_I(inode)->i_csum_seed;
> +	oi->of_binfo = kmalloc(oi->of_blocks*sizeof(struct ext4_orphan_block),
> +			       GFP_KERNEL);
> +	if (!oi->of_binfo) {
> +		ret = -ENOMEM;
> +		goto out_put;
> +	}
> +	for (i = 0; i < oi->of_blocks; i++) {
> +		oi->of_binfo[i].ob_bh = ext4_bread(NULL, inode, i, 0);
> +		if (IS_ERR(oi->of_binfo[i].ob_bh)) {
> +			ret = PTR_ERR(oi->of_binfo[i].ob_bh);
> +			goto out_free;
> +		}
> +		if (!oi->of_binfo[i].ob_bh) {
> +			ret = -EIO;
> +			goto out_free;
> +		}
> +		ot = ext4_orphan_block_tail(sb, oi->of_binfo[i].ob_bh);
> +		if (le32_to_cpu(ot->ob_magic) != EXT4_ORPHAN_BLOCK_MAGIC) {
> +			ext4_error(sb, "orphan file block %d: bad magic", i);
> +			ret = -EIO;
> +			goto out_free;
> +		}
> +		if (!ext4_orphan_file_block_csum_verify(sb,
> +						oi->of_binfo[i].ob_bh)) {
> +			ext4_error(sb, "orphan file block %d: bad checksum", i);
> +			ret = -EIO;
> +			goto out_free;
> +		}
> +		bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
> +		free = 0;
> +		for (j = 0; j < inodes_per_ob; j++)
> +			if (bdata[j] == 0)
> +				free++;
> +		oi->of_binfo[i].ob_free_entries = free;
> +	}
> +	iput(inode);
> +	return 0;
> +out_free:
> +	for (i--; i >= 0; i--)
> +		brelse(oi->of_binfo[i].ob_bh);
> +	kfree(oi->of_binfo);
> +out_put:
> +	iput(inode);
> +	return ret;
> +}
> +
> +int ext4_orphan_file_empty(struct super_block *sb)
> +{
> +	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
> +	int i;
> +	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
> +
> +	if (!ext4_has_feature_orphan_file(sb))
> +		return 1;
> +	for (i = 0; i < oi->of_blocks; i++)
> +		if (oi->of_binfo[i].ob_free_entries != inodes_per_ob)
> +			return 0;
> +	return 1;
> +}
> diff --git a/fs/ext4/super.c b/fs/ext4/super.c
> index 6e43c8546dc5..06f63b0cd988 100644
> --- a/fs/ext4/super.c
> +++ b/fs/ext4/super.c
> @@ -1164,6 +1164,7 @@ static void ext4_put_super(struct super_block *sb)
>  
>  	flush_work(&sbi->s_error_work);
>  	destroy_workqueue(sbi->rsv_conversion_wq);
> +	ext4_release_orphan_info(sb);
>  
>  	/*
>  	 * Unregister sysfs before destroying jbd2 journal.
> @@ -1189,6 +1190,7 @@ static void ext4_put_super(struct super_block *sb)
>  
>  	if (!sb_rdonly(sb) && !aborted) {
>  		ext4_clear_feature_journal_needs_recovery(sb);
> +		ext4_clear_feature_orphan_present(sb);
>  		es->s_state = cpu_to_le16(sbi->s_mount_state);
>  	}
>  	if (!sb_rdonly(sb))
> @@ -2695,8 +2697,11 @@ static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
>  		es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
>  	le16_add_cpu(&es->s_mnt_count, 1);
>  	ext4_update_tstamp(es, s_mtime);
> -	if (sbi->s_journal)
> +	if (sbi->s_journal) {
>  		ext4_set_feature_journal_needs_recovery(sb);
> +		if (ext4_has_feature_orphan_file(sb))
> +			ext4_set_feature_orphan_present(sb);
> +	}
>  
>  	err = ext4_commit_super(sb);
>  done:
> @@ -3971,6 +3976,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
>  		silent = 1;
>  		goto cantfind_ext4;
>  	}
> +	ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
> +				ext4_orphan_file_block_trigger);
>  
>  	/* Load the checksum driver */
>  	sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
> @@ -4635,6 +4642,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
>  	sb->s_root = NULL;
>  
>  	needs_recovery = (es->s_last_orphan != 0 ||
> +			  ext4_has_feature_orphan_present(sb) ||
>  			  ext4_has_feature_journal_needs_recovery(sb));
>  
>  	if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
> @@ -4924,12 +4932,15 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
>  	if (err)
>  		goto failed_mount7;
>  
> +	err = ext4_init_orphan_info(sb);
> +	if (err)
> +		goto failed_mount8;
>  #ifdef CONFIG_QUOTA
>  	/* Enable quota usage during mount. */
>  	if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
>  		err = ext4_enable_quotas(sb);
>  		if (err)
> -			goto failed_mount8;
> +			goto failed_mount9;
>  	}
>  #endif  /* CONFIG_QUOTA */
>  
> @@ -4948,7 +4959,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
>  		ext4_msg(sb, KERN_INFO, "recovery complete");
>  		err = ext4_mark_recovery_complete(sb, es);
>  		if (err)
> -			goto failed_mount8;
> +			goto failed_mount9;
>  	}
>  	if (EXT4_SB(sb)->s_journal) {
>  		if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
> @@ -4994,6 +5005,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
>  		ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
>  	goto failed_mount;
>  
> +failed_mount9:
> +	ext4_release_orphan_info(sb);
>  failed_mount8:
>  	ext4_unregister_sysfs(sb);
>  	kobject_put(&sbi->s_kobj);
> @@ -5505,8 +5518,15 @@ static int ext4_mark_recovery_complete(struct super_block *sb,
>  	if (err < 0)
>  		goto out;
>  
> -	if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
> +	if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
> +	    ext4_has_feature_orphan_present(sb))) {
> +		if (!ext4_orphan_file_empty(sb)) {
> +			ext4_error(sb, "Orphan file not empty on read-only fs.");
> +			err = -EFSCORRUPTED;
> +			goto out;
> +		}
>  		ext4_clear_feature_journal_needs_recovery(sb);
> +		ext4_clear_feature_orphan_present(sb);
>  		ext4_commit_super(sb);
>  	}
>  out:
> @@ -5649,6 +5669,8 @@ static int ext4_freeze(struct super_block *sb)
>  
>  		/* Journal blocked and flushed, clear needs_recovery flag. */
>  		ext4_clear_feature_journal_needs_recovery(sb);
> +		if (ext4_orphan_file_empty(sb))
> +			ext4_clear_feature_orphan_present(sb);
>  	}
>  
>  	error = ext4_commit_super(sb);
> @@ -5671,6 +5693,8 @@ static int ext4_unfreeze(struct super_block *sb)
>  	if (EXT4_SB(sb)->s_journal) {
>  		/* Reset the needs_recovery flag before the fs is unlocked. */
>  		ext4_set_feature_journal_needs_recovery(sb);
> +		if (ext4_has_feature_orphan_file(sb))
> +			ext4_set_feature_orphan_present(sb);
>  	}
>  
>  	ext4_commit_super(sb);
> @@ -5876,7 +5900,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
>  			 * around from a previously readonly bdev mount,
>  			 * require a full umount/remount for now.
>  			 */
> -			if (es->s_last_orphan) {
> +			if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
>  				ext4_msg(sb, KERN_WARNING, "Couldn't "
>  				       "remount RDWR because of unprocessed "
>  				       "orphan inode list.  Please "
> -- 
> 2.26.2
> 

[PATCH 4/4] ext4: Improve scalability of ext4 orphan file handling

[Jan Kara]

Even though the length of the critical section when adding / removing
orphaned inodes was significantly reduced by using orphan file, the
contention of lock protecting orphan file still appears high in profiles
for truncate / unlink intensive workloads with high number of threads.

This patch makes handling of orphan file completely lockless. Also to
reduce conflicts between CPUs different CPUs start searching for empty
slot in orphan file in different blocks.

Performance comparison of locked orphan file handling, lockless orphan
file handling, and completely disabled orphan inode handling
from 80 CPU Xeon Server with 526 GB of RAM, filesystem located on
SAS SSD disk, average of 5 runs:

stress-orphan (microbenchmark truncating files byte-by-byte from N
processes in parallel)

Threads Time            Time            Time
        Orphan locked   Orphan lockless No orphan
  1       0.945600       0.939400        0.891200
  2       1.331800       1.246600        1.174400
  4       1.995000       1.780600        1.713200
  8       6.424200       4.900000        4.106000
 16      14.937600       8.516400        8.138000
 32      33.038200      24.565600       24.002200
 64      60.823600      39.844600       38.440200
128     122.941400      70.950400       69.315000

So we can see that with lockless orphan file handling, addition /
deletion of orphaned inodes got almost completely out of picture even
for a microbenchmark stressing it.

For reaim creat_clo workload on ramdisk there are also noticeable gains
(average of 5 runs):

Clients         Vanilla (ops/s)        Patched (ops/s)
creat_clo-1     14705.88 (   0.00%)    14354.07 *  -2.39%*
creat_clo-3     27108.43 (   0.00%)    28301.89 (   4.40%)
creat_clo-5     37406.48 (   0.00%)    45180.73 *  20.78%*
creat_clo-7     41338.58 (   0.00%)    54687.50 *  32.29%*
creat_clo-9     45226.13 (   0.00%)    62937.07 *  39.16%*
creat_clo-11    44000.00 (   0.00%)    65088.76 *  47.93%*
creat_clo-13    36516.85 (   0.00%)    68661.97 *  88.03%*
creat_clo-15    30864.20 (   0.00%)    69551.78 * 125.35%*
creat_clo-17    27478.45 (   0.00%)    67729.08 * 146.48%*
creat_clo-19    25000.00 (   0.00%)    61621.62 * 146.49%*
creat_clo-21    18772.35 (   0.00%)    63829.79 * 240.02%*
creat_clo-23    16698.94 (   0.00%)    61938.96 * 270.92%*
creat_clo-25    14973.05 (   0.00%)    56947.61 * 280.33%*
creat_clo-27    16436.69 (   0.00%)    65008.03 * 295.51%*
creat_clo-29    13949.01 (   0.00%)    69047.62 * 395.00%*
creat_clo-31    14283.52 (   0.00%)    67982.45 * 375.95%*

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/ext4.h   |  3 +--
 fs/ext4/orphan.c | 55 +++++++++++++++++++++++++++---------------------
 2 files changed, 32 insertions(+), 26 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 83298c0b6dae..d08927e19b76 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -1480,7 +1480,7 @@ static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
 }
 
 struct ext4_orphan_block {
-	int ob_free_entries;	/* Number of free orphan entries in block */
+	atomic_t ob_free_entries;	/* Number of free orphan entries in block */
 	struct buffer_head *ob_bh;	/* Buffer for orphan block */
 };
 
@@ -1488,7 +1488,6 @@ struct ext4_orphan_block {
  * Info about orphan file.
  */
 struct ext4_orphan_info {
-	spinlock_t of_lock;
 	int of_blocks;			/* Number of orphan blocks in a file */
 	__u32 of_csum_seed;		/* Checksum seed for orphan file */
 	struct ext4_orphan_block *of_binfo;	/* Array with info about orphan
diff --git a/fs/ext4/orphan.c b/fs/ext4/orphan.c
index ac22667b7fd5..010222cde4f7 100644
--- a/fs/ext4/orphan.c
+++ b/fs/ext4/orphan.c
@@ -10,16 +10,30 @@
 
 static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
 {
-	int i, j;
+	int i, j, start;
 	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
 	int ret = 0;
+	bool found = false;
 	__le32 *bdata;
 	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
 
-	spin_lock(&oi->of_lock);
-	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
-	if (i == oi->of_blocks) {
-		spin_unlock(&oi->of_lock);
+	/*
+	 * Find block with free orphan entry. Use CPU number for a naive hash
+	 * for a search start in the orphan file
+	 */
+	start = raw_smp_processor_id()*13 % oi->of_blocks;
+	i = start;
+	do {
+		if (atomic_dec_if_positive(&oi->of_binfo[i].ob_free_entries)
+		    >= 0) {
+			found = true;
+			break;
+		}
+		if (++i >= oi->of_blocks)
+			i = 0;
+	} while (i != start);
+
+	if (!found) {
 		/*
 		 * For now we don't grow or shrink orphan file. We just use
 		 * whatever was allocated at mke2fs time. The additional
@@ -28,28 +42,24 @@ static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
 		 */
 		return -ENOSPC;
 	}
-	oi->of_binfo[i].ob_free_entries--;
-	spin_unlock(&oi->of_lock);
 
-	/*
-	 * Get access to orphan block. We have dropped of_lock but since we
-	 * have decremented number of free entries we are guaranteed free entry
-	 * in our block.
-	 */
 	ret = ext4_journal_get_write_access(handle, inode->i_sb,
 				oi->of_binfo[i].ob_bh, EXT4_JTR_ORPHAN_FILE);
 	if (ret)
 		return ret;
 
 	bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
-	spin_lock(&oi->of_lock);
 	/* Find empty slot in a block */
-	for (j = 0; j < inodes_per_ob && bdata[j]; j++);
-	BUG_ON(j == inodes_per_ob);
-	bdata[j] = cpu_to_le32(inode->i_ino);
+	j = 0;
+	do {
+		while (bdata[j]) {
+			if (++j >= inodes_per_ob)
+				j = 0;
+		}
+	} while (cmpxchg(&bdata[j], 0, cpu_to_le32(inode->i_ino)) != 0);
+
 	EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
 	ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
-	spin_unlock(&oi->of_lock);
 
 	return ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[i].ob_bh);
 }
@@ -178,10 +188,8 @@ static int ext4_orphan_file_del(handle_t *handle, struct inode *inode)
 		goto out;
 
 	bdata = (__le32 *)(oi->of_binfo[blk].ob_bh->b_data);
-	spin_lock(&oi->of_lock);
 	bdata[off] = 0;
-	oi->of_binfo[blk].ob_free_entries++;
-	spin_unlock(&oi->of_lock);
+	atomic_inc(&oi->of_binfo[blk].ob_free_entries);
 	ret = ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[blk].ob_bh);
 out:
 	ext4_clear_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
@@ -534,8 +542,6 @@ int ext4_init_orphan_info(struct super_block *sb)
 	struct ext4_orphan_block_tail *ot;
 	ino_t orphan_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_orphan_file_inum);
 
-	spin_lock_init(&oi->of_lock);
-
 	if (!ext4_has_feature_orphan_file(sb))
 		return 0;
 
@@ -579,7 +585,7 @@ int ext4_init_orphan_info(struct super_block *sb)
 		for (j = 0; j < inodes_per_ob; j++)
 			if (bdata[j] == 0)
 				free++;
-		oi->of_binfo[i].ob_free_entries = free;
+		atomic_set(&oi->of_binfo[i].ob_free_entries, free);
 	}
 	iput(inode);
 	return 0;
@@ -601,7 +607,8 @@ int ext4_orphan_file_empty(struct super_block *sb)
 	if (!ext4_has_feature_orphan_file(sb))
 		return 1;
 	for (i = 0; i < oi->of_blocks; i++)
-		if (oi->of_binfo[i].ob_free_entries != inodes_per_ob)
+		if (atomic_read(&oi->of_binfo[i].ob_free_entries) !=
+		    inodes_per_ob)
 			return 0;
 	return 1;
 }
-- 
2.26.2

Re: [PATCH 4/4] ext4: Improve scalability of ext4 orphan file handling

[Lukas Czerner]

On Wed, Jun 16, 2021 at 12:56:55PM +0200, Jan Kara wrote:
> Even though the length of the critical section when adding / removing
> orphaned inodes was significantly reduced by using orphan file, the
> contention of lock protecting orphan file still appears high in profiles
> for truncate / unlink intensive workloads with high number of threads.
> 
> This patch makes handling of orphan file completely lockless. Also to
> reduce conflicts between CPUs different CPUs start searching for empty
> slot in orphan file in different blocks.
> 
> Performance comparison of locked orphan file handling, lockless orphan
> file handling, and completely disabled orphan inode handling
> from 80 CPU Xeon Server with 526 GB of RAM, filesystem located on
> SAS SSD disk, average of 5 runs:
> 
> stress-orphan (microbenchmark truncating files byte-by-byte from N
> processes in parallel)
> 
> Threads Time            Time            Time
>         Orphan locked   Orphan lockless No orphan
>   1       0.945600       0.939400        0.891200
>   2       1.331800       1.246600        1.174400
>   4       1.995000       1.780600        1.713200
>   8       6.424200       4.900000        4.106000
>  16      14.937600       8.516400        8.138000
>  32      33.038200      24.565600       24.002200
>  64      60.823600      39.844600       38.440200
> 128     122.941400      70.950400       69.315000
> 
> So we can see that with lockless orphan file handling, addition /
> deletion of orphaned inodes got almost completely out of picture even
> for a microbenchmark stressing it.
> 
> For reaim creat_clo workload on ramdisk there are also noticeable gains
> (average of 5 runs):
> 
> Clients         Vanilla (ops/s)        Patched (ops/s)
> creat_clo-1     14705.88 (   0.00%)    14354.07 *  -2.39%*
> creat_clo-3     27108.43 (   0.00%)    28301.89 (   4.40%)
> creat_clo-5     37406.48 (   0.00%)    45180.73 *  20.78%*
> creat_clo-7     41338.58 (   0.00%)    54687.50 *  32.29%*
> creat_clo-9     45226.13 (   0.00%)    62937.07 *  39.16%*
> creat_clo-11    44000.00 (   0.00%)    65088.76 *  47.93%*
> creat_clo-13    36516.85 (   0.00%)    68661.97 *  88.03%*
> creat_clo-15    30864.20 (   0.00%)    69551.78 * 125.35%*
> creat_clo-17    27478.45 (   0.00%)    67729.08 * 146.48%*
> creat_clo-19    25000.00 (   0.00%)    61621.62 * 146.49%*
> creat_clo-21    18772.35 (   0.00%)    63829.79 * 240.02%*
> creat_clo-23    16698.94 (   0.00%)    61938.96 * 270.92%*
> creat_clo-25    14973.05 (   0.00%)    56947.61 * 280.33%*
> creat_clo-27    16436.69 (   0.00%)    65008.03 * 295.51%*
> creat_clo-29    13949.01 (   0.00%)    69047.62 * 395.00%*
> creat_clo-31    14283.52 (   0.00%)    67982.45 * 375.95%*
> 
> Signed-off-by: Jan Kara <jack@suse.cz>
> ---
>  fs/ext4/ext4.h   |  3 +--
>  fs/ext4/orphan.c | 55 +++++++++++++++++++++++++++---------------------
>  2 files changed, 32 insertions(+), 26 deletions(-)
> 
> diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
> index 83298c0b6dae..d08927e19b76 100644
> --- a/fs/ext4/ext4.h
> +++ b/fs/ext4/ext4.h
> @@ -1480,7 +1480,7 @@ static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
>  }
>  
>  struct ext4_orphan_block {
> -	int ob_free_entries;	/* Number of free orphan entries in block */
> +	atomic_t ob_free_entries;	/* Number of free orphan entries in block */
>  	struct buffer_head *ob_bh;	/* Buffer for orphan block */
>  };
>  
> @@ -1488,7 +1488,6 @@ struct ext4_orphan_block {
>   * Info about orphan file.
>   */
>  struct ext4_orphan_info {
> -	spinlock_t of_lock;
>  	int of_blocks;			/* Number of orphan blocks in a file */
>  	__u32 of_csum_seed;		/* Checksum seed for orphan file */
>  	struct ext4_orphan_block *of_binfo;	/* Array with info about orphan
> diff --git a/fs/ext4/orphan.c b/fs/ext4/orphan.c
> index ac22667b7fd5..010222cde4f7 100644
> --- a/fs/ext4/orphan.c
> +++ b/fs/ext4/orphan.c
> @@ -10,16 +10,30 @@
>  
>  static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
>  {
> -	int i, j;
> +	int i, j, start;
>  	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
>  	int ret = 0;
> +	bool found = false;
>  	__le32 *bdata;
>  	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
>  
> -	spin_lock(&oi->of_lock);
> -	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
> -	if (i == oi->of_blocks) {
> -		spin_unlock(&oi->of_lock);
> +	/*
> +	 * Find block with free orphan entry. Use CPU number for a naive hash
> +	 * for a search start in the orphan file
> +	 */
> +	start = raw_smp_processor_id()*13 % oi->of_blocks;
> +	i = start;
> +	do {
> +		if (atomic_dec_if_positive(&oi->of_binfo[i].ob_free_entries)
> +		    >= 0) {
> +			found = true;
> +			break;
> +		}
> +		if (++i >= oi->of_blocks)
> +			i = 0;
> +	} while (i != start);
> +
> +	if (!found) {
>  		/*
>  		 * For now we don't grow or shrink orphan file. We just use
>  		 * whatever was allocated at mke2fs time. The additional
> @@ -28,28 +42,24 @@ static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
>  		 */
>  		return -ENOSPC;
>  	}
> -	oi->of_binfo[i].ob_free_entries--;
> -	spin_unlock(&oi->of_lock);
>  
> -	/*
> -	 * Get access to orphan block. We have dropped of_lock but since we
> -	 * have decremented number of free entries we are guaranteed free entry
> -	 * in our block.
> -	 */
>  	ret = ext4_journal_get_write_access(handle, inode->i_sb,
>  				oi->of_binfo[i].ob_bh, EXT4_JTR_ORPHAN_FILE);
>  	if (ret)
>  		return ret;
>  
>  	bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
> -	spin_lock(&oi->of_lock);
>  	/* Find empty slot in a block */
> -	for (j = 0; j < inodes_per_ob && bdata[j]; j++);
> -	BUG_ON(j == inodes_per_ob);
> -	bdata[j] = cpu_to_le32(inode->i_ino);
> +	j = 0;
> +	do {
> +		while (bdata[j]) {
> +			if (++j >= inodes_per_ob)
> +				j = 0;
> +		}
> +	} while (cmpxchg(&bdata[j], 0, cpu_to_le32(inode->i_ino)) != 0);

In case there is any sort of corruption on disk or in memory we can
potentially get stuck here forever right ? Not sure if that matters
all that much.

Other than that it looks good and negates some of my comments on the
previous patch, sorry about that ;)

You can add

Reviewed-by: Lukas Czerner <lczerner@redhat.com>


> +
>  	EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
>  	ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> -	spin_unlock(&oi->of_lock);
>  
>  	return ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[i].ob_bh);
>  }
> @@ -178,10 +188,8 @@ static int ext4_orphan_file_del(handle_t *handle, struct inode *inode)
>  		goto out;
>  
>  	bdata = (__le32 *)(oi->of_binfo[blk].ob_bh->b_data);
> -	spin_lock(&oi->of_lock);
>  	bdata[off] = 0;
> -	oi->of_binfo[blk].ob_free_entries++;
> -	spin_unlock(&oi->of_lock);
> +	atomic_inc(&oi->of_binfo[blk].ob_free_entries);
>  	ret = ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[blk].ob_bh);
>  out:
>  	ext4_clear_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
> @@ -534,8 +542,6 @@ int ext4_init_orphan_info(struct super_block *sb)
>  	struct ext4_orphan_block_tail *ot;
>  	ino_t orphan_ino = le32_to_cpu(EXT4_SB(sb)->s_es->s_orphan_file_inum);
>  
> -	spin_lock_init(&oi->of_lock);
> -
>  	if (!ext4_has_feature_orphan_file(sb))
>  		return 0;
>  
> @@ -579,7 +585,7 @@ int ext4_init_orphan_info(struct super_block *sb)
>  		for (j = 0; j < inodes_per_ob; j++)
>  			if (bdata[j] == 0)
>  				free++;
> -		oi->of_binfo[i].ob_free_entries = free;
> +		atomic_set(&oi->of_binfo[i].ob_free_entries, free);
>  	}
>  	iput(inode);
>  	return 0;
> @@ -601,7 +607,8 @@ int ext4_orphan_file_empty(struct super_block *sb)
>  	if (!ext4_has_feature_orphan_file(sb))
>  		return 1;
>  	for (i = 0; i < oi->of_blocks; i++)
> -		if (oi->of_binfo[i].ob_free_entries != inodes_per_ob)
> +		if (atomic_read(&oi->of_binfo[i].ob_free_entries) !=
> +		    inodes_per_ob)
>  			return 0;
>  	return 1;
>  }
> -- 
> 2.26.2
> 

Re: [PATCH 4/4] ext4: Improve scalability of ext4 orphan file handling

[Jan Kara]

On Wed 30-06-21 15:46:35, Lukas Czerner wrote:
> On Wed, Jun 16, 2021 at 12:56:55PM +0200, Jan Kara wrote:
> > @@ -28,28 +42,24 @@ static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
> >  		 */
> >  		return -ENOSPC;
> >  	}
> > -	oi->of_binfo[i].ob_free_entries--;
> > -	spin_unlock(&oi->of_lock);
> >  
> > -	/*
> > -	 * Get access to orphan block. We have dropped of_lock but since we
> > -	 * have decremented number of free entries we are guaranteed free entry
> > -	 * in our block.
> > -	 */
> >  	ret = ext4_journal_get_write_access(handle, inode->i_sb,
> >  				oi->of_binfo[i].ob_bh, EXT4_JTR_ORPHAN_FILE);
> >  	if (ret)
> >  		return ret;
> >  
> >  	bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
> > -	spin_lock(&oi->of_lock);
> >  	/* Find empty slot in a block */
> > -	for (j = 0; j < inodes_per_ob && bdata[j]; j++);
> > -	BUG_ON(j == inodes_per_ob);
> > -	bdata[j] = cpu_to_le32(inode->i_ino);
> > +	j = 0;
> > +	do {
> > +		while (bdata[j]) {
> > +			if (++j >= inodes_per_ob)
> > +				j = 0;
> > +		}
> > +	} while (cmpxchg(&bdata[j], 0, cpu_to_le32(inode->i_ino)) != 0);
> 
> In case there is any sort of corruption on disk or in memory we can
> potentially get stuck here forever right ? Not sure if that matters
> all that much.
> 
> Other than that it looks good and negates some of my comments on the
> previous patch, sorry about that ;)
> 
> You can add
> 
> Reviewed-by: Lukas Czerner <lczerner@redhat.com>

Good point. I've added some limitations (and cond_resched()) to the loop so
that we cannot loop indefinitely. Thanks for review!

								Honza
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

[PATCH 0/3 v2] ext4: Speedup orphan file handling

[Jan Kara]

  Hello,

This is the second version of my patches to speed up orphan inode handling
in ext4.

Orphan inode handling in ext4 is a bottleneck for workloads which heavily
excercise truncate / unlink of small files as they contend on global
s_orphan_mutex (when you have fast enough storage). This patch set implements
new way of handling orphan inodes - instead of using a linked list, we store
inode numbers of orphaned inodes in a file which is possible to implement in a
more scalable manner than linked list manipulations. See description of patch
2/3 for more details.

The patch set achieves significant gains both for a micro benchmark stressing
orphan inode handling (truncating file byte-by-byte, several threads in
parallel) and for reaim new_fserver workload. As a highlight, microbenchmark
runtime for 128 threads is reduced from original 160 s down to 71 s, which
is also the time it takes the benchmark to run when orphan inode handling
is completely disabled. For full numbers you can check commit logs of
patches 2/3 and 3/3. You can also check my presentation from Vault at
http://events.linuxfoundation.org/sites/events/files/slides/ext4-scaling.pdf
for graphs from tests.

I'm happy for any review, thoughts, ideas about the patches.

The kernel part of the feature is complete, I have also implemented full
support in e2fsprogs. That still needs some debugging (especially the e2fsck
part) but support in mke2fs or tune2fs is fine. I'll post these as a separate
patch so that people can try this out.

For now I'm using inode 9 for orphan file. I know that is reserved as
EXT2_EXCLUDE_INO but at least for the sake of testing that should be fine.

								Honza

Changes since v1:
* orphan blocks have now magic numbers
* split out orphan handling to a separate source file
* some smaller updates according to review

[PATCH 3/4] ext4: Speedup ext4 orphan inode handling

[Jan Kara]

Ext4 orphan inode handling is a bottleneck for workloads which heavily
truncate / unlink small files since it contends on the global
s_orphan_mutex lock (and generally it's difficult to improve scalability
of the ondisk linked list of orphaned inodes).

This patch implements new way of handling orphan inodes. Instead of
linking orphaned inode into a linked list, we store it's inode number in
a new special file which we call "orphan file". Currently we still
protect the orphan file with a spinlock for simplicity but even in this
setting we can substantially reduce the length of the critical section
and thus speedup some workloads.

Note that the change is backwards compatible when the filesystem is
clean - the existence of the orphan file is a compat feature, we set
another ro-compat feature indicating orphan file needs scanning for
orphaned inodes when mounting filesystem read-write. This ro-compat
feature gets cleared on unmount / remount read-only.

Some performance data from 48 CPU Xeon Server with 32 GB of RAM,
filesystem located on ramdisk, average of 5 runs:

stress-orphan (microbenchmark truncating files byte-by-byte from N
processes in parallel)

Threads Time            Time
        Vanilla         Patched
  1       1.602800        1.260000
  2       4.292200        2.455000
  4       6.202800        3.848400
  8      10.415000        6.833000
 16      18.933600       12.883200
 32      38.517200       25.342200
 64      79.805000       50.918400
128     159.629200      102.666000

reaim new_fserver workload (tweaked to avoid calling sync(1) after every
operation)

Threads Jobs/s          Jobs/s
        Vanilla         Patched
  1      24375.00        22941.18
 25     162162.16       278571.43
 49     222209.30       331626.90
 73     280147.60       419447.52
 97     315250.00       481910.83
121     331157.90       503360.00
145     343769.00       489081.08
169     355549.56       519487.68
193     356518.65       501800.00

So in both cases we see significant wins all over the board.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/ext4.h   |  67 +++++++++++-
 fs/ext4/orphan.c | 316 +++++++++++++++++++++++++++++++++++++++++++++++++------
 fs/ext4/super.c  |  43 ++++++--
 3 files changed, 383 insertions(+), 43 deletions(-)

diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 2a69a5e9017a..b48b250e784e 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -208,6 +208,7 @@ struct ext4_io_submit {
 #define EXT4_UNDEL_DIR_INO	 6	/* Undelete directory inode */
 #define EXT4_RESIZE_INO		 7	/* Reserved group descriptors inode */
 #define EXT4_JOURNAL_INO	 8	/* Journal inode */
+#define EXT4_ORPHAN_INO		 9	/* Inode with orphan entries */
 
 /* First non-reserved inode for old ext4 filesystems */
 #define EXT4_GOOD_OLD_FIRST_INO	11
@@ -843,7 +844,14 @@ struct ext4_inode_info {
 	 */
 	struct rw_semaphore xattr_sem;
 
-	struct list_head i_orphan;	/* unlinked but open inodes */
+	/*
+	 * Inodes with EXT4_STATE_ORPHAN_FILE use i_orphan_block. Otherwise
+	 * i_orphan is used.
+	 */
+	union {
+		struct list_head i_orphan;	/* unlinked but open inodes */
+		unsigned int i_orphan_idx;	/* Index in orphan file */
+	};
 
 	/*
 	 * i_disksize keeps track of what the inode size is ON DISK, not
@@ -1211,6 +1219,7 @@ struct ext4_super_block {
 
 /* Types of ext4 journal triggers */
 enum ext4_journal_trigger_type {
+	EXT4_JTR_ORPHAN_FILE,
 	EXT4_JTR_NONE	/* This must be the last entry for indexing to work! */
 };
 
@@ -1227,6 +1236,37 @@ static inline struct ext4_journal_trigger *EXT4_TRIGGER(
 	return container_of(trigger, struct ext4_journal_trigger, tr_triggers);
 }
 
+#define EXT4_ORPHAN_BLOCK_MAGIC 0x0b10ca04
+
+/* Structure at the tail of orphan block */
+struct ext4_orphan_block_tail {
+	__le32 ob_magic;
+	__le32 ob_checksum;
+};
+
+static inline int ext4_inodes_per_orphan_block(struct super_block *sb)
+{
+	return (sb->s_blocksize - sizeof(struct ext4_orphan_block_tail)) /
+			sizeof(u32);
+}
+
+struct ext4_orphan_block {
+	int ob_free_entries;	/* Number of free orphan entries in block */
+	struct buffer_head *ob_bh;	/* Buffer for orphan block */
+};
+
+/*
+ * Info about orphan file. Some info in this structure is duplicated - once
+ * for running and once for committing transaction
+ */
+struct ext4_orphan_info {
+	spinlock_t of_lock;
+	int of_blocks;			/* Number of orphan blocks in a file */
+	__u32 of_csum_seed;		/* Checksum seed for orphan file */
+	struct ext4_orphan_block *of_binfo;	/* Array with info about orphan
+						 * file blocks */
+};
+
 /*
  * fourth extended-fs super-block data in memory
  */
@@ -1281,8 +1321,10 @@ struct ext4_sb_info {
 
 	/* Journaling */
 	struct journal_s *s_journal;
-	struct list_head s_orphan;
-	struct mutex s_orphan_lock;
+	struct mutex s_orphan_lock;	/* Protects on disk list changes */
+	struct list_head s_orphan;	/* List of orphaned inodes in on disk
+					   list */
+	struct ext4_orphan_info s_orphan_info;
 	unsigned long s_resize_flags;		/* Flags indicating if there
 						   is a resizer */
 	unsigned long s_commit_interval;
@@ -1426,6 +1468,7 @@ static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
 		ino == EXT4_BOOT_LOADER_INO ||
 		ino == EXT4_JOURNAL_INO ||
 		ino == EXT4_RESIZE_INO ||
+		ino == EXT4_ORPHAN_INO ||
 		(ino >= EXT4_FIRST_INO(sb) &&
 		 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
 }
@@ -1466,6 +1509,7 @@ enum {
 	EXT4_STATE_MAY_INLINE_DATA,	/* may have in-inode data */
 	EXT4_STATE_ORDERED_MODE,	/* data=ordered mode */
 	EXT4_STATE_EXT_PRECACHED,	/* extents have been precached */
+	EXT4_STATE_ORPHAN_FILE,		/* Inode orphaned in orphan file */
 };
 
 #define EXT4_INODE_BIT_FNS(name, field, offset)				\
@@ -1580,6 +1624,7 @@ static inline int ext4_encrypted_inode(struct inode *inode)
 #define EXT4_FEATURE_COMPAT_RESIZE_INODE	0x0010
 #define EXT4_FEATURE_COMPAT_DIR_INDEX		0x0020
 #define EXT4_FEATURE_COMPAT_SPARSE_SUPER2	0x0200
+#define EXT4_FEATURE_COMPAT_ORPHAN_FILE		0x0400	/* Orphan file exists */
 
 #define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER	0x0001
 #define EXT4_FEATURE_RO_COMPAT_LARGE_FILE	0x0002
@@ -1597,7 +1642,10 @@ static inline int ext4_encrypted_inode(struct inode *inode)
  * GDT_CSUM bits are mutually exclusive.
  */
 #define EXT4_FEATURE_RO_COMPAT_METADATA_CSUM	0x0400
+/* 0x0800 Reserved for EXT4_FEATURE_RO_COMPAT_REPLICA */
 #define EXT4_FEATURE_RO_COMPAT_READONLY		0x1000
+#define EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT	0x2000 /* Orphan file may be
+							  non-empty */
 
 #define EXT4_FEATURE_INCOMPAT_COMPRESSION	0x0001
 #define EXT4_FEATURE_INCOMPAT_FILETYPE		0x0002
@@ -1630,7 +1678,8 @@ static inline int ext4_encrypted_inode(struct inode *inode)
 					 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
 					 EXT4_FEATURE_RO_COMPAT_BTREE_DIR)
 
-#define EXT4_FEATURE_COMPAT_SUPP	EXT2_FEATURE_COMPAT_EXT_ATTR
+#define EXT4_FEATURE_COMPAT_SUPP	(EXT4_FEATURE_COMPAT_EXT_ATTR| \
+					 EXT4_FEATURE_COMPAT_ORPHAN_FILE)
 #define EXT4_FEATURE_INCOMPAT_SUPP	(EXT4_FEATURE_INCOMPAT_FILETYPE| \
 					 EXT4_FEATURE_INCOMPAT_RECOVER| \
 					 EXT4_FEATURE_INCOMPAT_META_BG| \
@@ -1649,7 +1698,8 @@ static inline int ext4_encrypted_inode(struct inode *inode)
 					 EXT4_FEATURE_RO_COMPAT_HUGE_FILE |\
 					 EXT4_FEATURE_RO_COMPAT_BIGALLOC |\
 					 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM|\
-					 EXT4_FEATURE_RO_COMPAT_QUOTA)
+					 EXT4_FEATURE_RO_COMPAT_QUOTA|\
+					 EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT)
 
 /*
  * Default values for user and/or group using reserved blocks
@@ -2972,6 +3022,13 @@ extern int ext4_orphan_add(handle_t *, struct inode *);
 extern int ext4_orphan_del(handle_t *, struct inode *);
 extern void ext4_orphan_cleanup(struct super_block *sb,
 				struct ext4_super_block *es);
+extern void ext4_release_orphan_info(struct super_block *sb);
+extern int ext4_init_orphan_info(struct super_block *sb);
+extern int ext4_orphan_file_empty(struct super_block *sb);
+extern void ext4_orphan_file_block_trigger(
+				struct jbd2_buffer_trigger_type *triggers,
+				struct buffer_head *bh,
+				void *data, size_t size);
 
 /*
  * Add new method to test whether block and inode bitmaps are properly
diff --git a/fs/ext4/orphan.c b/fs/ext4/orphan.c
index a192cc2086f9..23d504e12fcd 100644
--- a/fs/ext4/orphan.c
+++ b/fs/ext4/orphan.c
@@ -8,6 +8,52 @@
 #include "ext4.h"
 #include "ext4_jbd2.h"
 
+static int ext4_orphan_file_add(handle_t *handle, struct inode *inode)
+{
+	int i, j;
+	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
+	int ret = 0;
+	__le32 *bdata;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
+
+	spin_lock(&oi->of_lock);
+	for (i = 0; i < oi->of_blocks && !oi->of_binfo[i].ob_free_entries; i++);
+	if (i == oi->of_blocks) {
+		spin_unlock(&oi->of_lock);
+		/*
+		 * For now we don't grow or shrink orphan file. We just use
+		 * whatever was allocated at mke2fs time. The additional
+		 * credits we would have to reserve for each orphan inode
+		 * operation just don't seem worth it.
+		 */
+		return -ENOSPC;
+	}
+	oi->of_binfo[i].ob_free_entries--;
+	spin_unlock(&oi->of_lock);
+
+	/*
+	 * Get access to orphan block. We have dropped of_lock but since we
+	 * have decremented number of free entries we are guaranteed free entry
+	 * in our block.
+	 */
+	ret = ext4_journal_get_write_access(handle, inode->i_sb,
+				oi->of_binfo[i].ob_bh, EXT4_JTR_ORPHAN_FILE);
+	if (ret)
+		return ret;
+
+	bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
+	spin_lock(&oi->of_lock);
+	/* Find empty slot in a block */
+	for (j = 0; j < inodes_per_ob && bdata[j]; j++);
+	BUG_ON(j == inodes_per_ob);
+	bdata[j] = cpu_to_le32(inode->i_ino);
+	EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
+	ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
+	spin_unlock(&oi->of_lock);
+
+	return ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[i].ob_bh);
+}
+
 /*
  * ext4_orphan_add() links an unlinked or truncated inode into a list of
  * such inodes, starting at the superblock, in case we crash before the
@@ -34,10 +80,10 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
 	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
 		     !mutex_is_locked(&inode->i_mutex));
 	/*
-	 * Exit early if inode already is on orphan list. This is a big speedup
-	 * since we don't have to contend on the global s_orphan_lock.
+	 * Inode orphaned in orphan file or in orphan list?
 	 */
-	if (!list_empty(&EXT4_I(inode)->i_orphan))
+	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE) ||
+	    !list_empty(&EXT4_I(inode)->i_orphan))
 		return 0;
 
 	/*
@@ -49,6 +95,16 @@ int ext4_orphan_add(handle_t *handle, struct inode *inode)
 	J_ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
 		  S_ISLNK(inode->i_mode)) || inode->i_nlink == 0);
 
+	if (sbi->s_orphan_info.of_blocks) {
+		err = ext4_orphan_file_add(handle, inode);
+		/*
+		 * Fallback to normal orphan list of orphan file is
+		 * out of space
+		 */
+		if (err != -ENOSPC)
+			return err;
+	}
+
 	BUFFER_TRACE(sbi->s_sbh, "get_write_access");
 	err = ext4_journal_get_write_access(handle, sb, sbi->s_sbh,
 					    EXT4_JTR_NONE);
@@ -98,6 +154,37 @@ out:
 	return err;
 }
 
+static int ext4_orphan_file_del(handle_t *handle, struct inode *inode)
+{
+	struct ext4_orphan_info *oi = &EXT4_SB(inode->i_sb)->s_orphan_info;
+	__le32 *bdata;
+	int blk, off;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(inode->i_sb);
+	int ret = 0;
+
+	if (!handle)
+		goto out;
+	blk = EXT4_I(inode)->i_orphan_idx / inodes_per_ob;
+	off = EXT4_I(inode)->i_orphan_idx % inodes_per_ob;
+
+	ret = ext4_journal_get_write_access(handle, inode->i_sb,
+				oi->of_binfo[blk].ob_bh, EXT4_JTR_ORPHAN_FILE);
+	if (ret)
+		goto out;
+
+	bdata = (__le32 *)(oi->of_binfo[blk].ob_bh->b_data);
+	spin_lock(&oi->of_lock);
+	bdata[off] = 0;
+	oi->of_binfo[blk].ob_free_entries++;
+	spin_unlock(&oi->of_lock);
+	ret = ext4_handle_dirty_metadata(handle, NULL, oi->of_binfo[blk].ob_bh);
+out:
+	ext4_clear_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
+	INIT_LIST_HEAD(&EXT4_I(inode)->i_orphan);
+
+	return ret;
+}
+
 /*
  * ext4_orphan_del() removes an unlinked or truncated inode from the list
  * of such inodes stored on disk, because it is finally being cleaned up.
@@ -116,7 +203,11 @@ int ext4_orphan_del(handle_t *handle, struct inode *inode)
 
 	WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) &&
 		     !mutex_is_locked(&inode->i_mutex));
-	/* Do this quick check before taking global s_orphan_lock. */
+
+	if (ext4_test_inode_state(inode, EXT4_STATE_ORPHAN_FILE))
+		return ext4_orphan_file_del(handle, inode);
+
+	/* Do this quick check before taking global lock. */
 	if (list_empty(&ei->i_orphan))
 		return 0;
 
@@ -188,6 +279,36 @@ static int ext4_quota_on_mount(struct super_block *sb, int type)
 					EXT4_SB(sb)->s_jquota_fmt, type);
 }
 
+static void ext4_process_orphan(struct inode *inode,
+				int *nr_truncates, int *nr_orphans)
+{
+	struct super_block *sb = inode->i_sb;
+
+	dquot_initialize(inode);
+	if (inode->i_nlink) {
+		if (test_opt(sb, DEBUG))
+			ext4_msg(sb, KERN_DEBUG,
+				"%s: truncating inode %lu to %lld bytes",
+				__func__, inode->i_ino, inode->i_size);
+		jbd_debug(2, "truncating inode %lu to %lld bytes\n",
+			  inode->i_ino, inode->i_size);
+		mutex_lock(&inode->i_mutex);
+		truncate_inode_pages(inode->i_mapping, inode->i_size);
+		ext4_truncate(inode);
+		mutex_unlock(&inode->i_mutex);
+		(*nr_truncates)++;
+	} else {
+		if (test_opt(sb, DEBUG))
+			ext4_msg(sb, KERN_DEBUG,
+				"%s: deleting unreferenced inode %lu",
+				__func__, inode->i_ino);
+		jbd_debug(2, "deleting unreferenced inode %lu\n",
+			  inode->i_ino);
+		(*nr_orphans)++;
+	}
+	iput(inode);  /* The delete magic happens here! */
+}
+
 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
  * the superblock) which were deleted from all directories, but held open by
  * a process at the time of a crash.  We walk the list and try to delete these
@@ -210,10 +331,13 @@ void ext4_orphan_cleanup(struct super_block *sb,
 {
 	unsigned int s_flags = sb->s_flags;
 	int nr_orphans = 0, nr_truncates = 0;
-#ifdef CONFIG_QUOTA
-	int i;
-#endif
-	if (!es->s_last_orphan) {
+	int i, j;
+	__le32 *bdata;
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	struct inode *inode;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+
+	if (!es->s_last_orphan && !oi->of_blocks) {
 		jbd_debug(4, "no orphan inodes to clean up\n");
 		return;
 	}
@@ -262,8 +386,6 @@ void ext4_orphan_cleanup(struct super_block *sb,
 #endif
 
 	while (es->s_last_orphan) {
-		struct inode *inode;
-
 		inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
 		if (IS_ERR(inode)) {
 			es->s_last_orphan = 0;
@@ -271,29 +393,21 @@ void ext4_orphan_cleanup(struct super_block *sb,
 		}
 
 		list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
-		dquot_initialize(inode);
-		if (inode->i_nlink) {
-			if (test_opt(sb, DEBUG))
-				ext4_msg(sb, KERN_DEBUG,
-					"%s: truncating inode %lu to %lld bytes",
-					__func__, inode->i_ino, inode->i_size);
-			jbd_debug(2, "truncating inode %lu to %lld bytes\n",
-				  inode->i_ino, inode->i_size);
-			mutex_lock(&inode->i_mutex);
-			truncate_inode_pages(inode->i_mapping, inode->i_size);
-			ext4_truncate(inode);
-			mutex_unlock(&inode->i_mutex);
-			nr_truncates++;
-		} else {
-			if (test_opt(sb, DEBUG))
-				ext4_msg(sb, KERN_DEBUG,
-					"%s: deleting unreferenced inode %lu",
-					__func__, inode->i_ino);
-			jbd_debug(2, "deleting unreferenced inode %lu\n",
-				  inode->i_ino);
-			nr_orphans++;
+		ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
+	}
+
+	for (i = 0; i < oi->of_blocks; i++) {
+		bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
+		for (j = 0; j < inodes_per_ob; j++) {
+			if (!bdata[j])
+				continue;
+			inode = ext4_orphan_get(sb, le32_to_cpu(bdata[j]));
+			if (IS_ERR(inode))
+				continue;
+			ext4_set_inode_state(inode, EXT4_STATE_ORPHAN_FILE);
+			EXT4_I(inode)->i_orphan_idx = i * inodes_per_ob + j;
+			ext4_process_orphan(inode, &nr_truncates, &nr_orphans);
 		}
-		iput(inode);  /* The delete magic happens here! */
 	}
 
 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
@@ -313,3 +427,141 @@ void ext4_orphan_cleanup(struct super_block *sb,
 #endif
 	sb->s_flags = s_flags; /* Restore MS_RDONLY status */
 }
+
+void ext4_release_orphan_info(struct super_block *sb)
+{
+	int i;
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+
+	if (!oi->of_blocks)
+		return;
+	for (i = 0; i < oi->of_blocks; i++)
+		brelse(oi->of_binfo[i].ob_bh);
+	kfree(oi->of_binfo);
+}
+
+static struct ext4_orphan_block_tail *ext4_orphan_block_tail(
+						struct super_block *sb,
+						struct buffer_head *bh)
+{
+	return (struct ext4_orphan_block_tail *)(bh->b_data + sb->s_blocksize -
+				sizeof(struct ext4_orphan_block_tail));
+}
+
+static int ext4_orphan_file_block_csum_verify(struct super_block *sb,
+					      struct buffer_head *bh)
+{
+	__u32 calculated;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	struct ext4_orphan_block_tail *ot;
+
+	if (!ext4_has_metadata_csum(sb))
+		return 1;
+
+	ot = ext4_orphan_block_tail(sb, bh);
+	calculated = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed,
+				 (__u8 *)bh->b_data,
+				 inodes_per_ob * sizeof(__u32));
+	return le32_to_cpu(ot->ob_checksum) == calculated;
+}
+
+/* This gets called only when checksumming is enabled */
+void ext4_orphan_file_block_trigger(struct jbd2_buffer_trigger_type *triggers,
+				    struct buffer_head *bh,
+				    void *data, size_t size)
+{
+	struct super_block *sb = EXT4_TRIGGER(triggers)->sb;
+	__u32 csum;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	struct ext4_orphan_block_tail *ot;
+
+	csum = ext4_chksum(EXT4_SB(sb), oi->of_csum_seed, (__u8 *)data,
+			   inodes_per_ob * sizeof(__u32));
+	ot = ext4_orphan_block_tail(sb, bh);
+	ot->ob_checksum = cpu_to_le32(csum);
+}
+
+int ext4_init_orphan_info(struct super_block *sb)
+{
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	struct inode *inode;
+	int i, j;
+	int ret;
+	int free;
+	__le32 *bdata;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+	struct ext4_orphan_block_tail *ot;
+
+	spin_lock_init(&oi->of_lock);
+
+	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_ORPHAN_FILE))
+		return 0;
+
+	inode = ext4_iget(sb, EXT4_ORPHAN_INO);
+	if (IS_ERR(inode)) {
+		ext4_msg(sb, KERN_ERR, "get orphan inode failed");
+		return PTR_ERR(inode);
+	}
+	oi->of_blocks = inode->i_size >> sb->s_blocksize_bits;
+	oi->of_csum_seed = EXT4_I(inode)->i_csum_seed;
+	oi->of_binfo = kmalloc(oi->of_blocks*sizeof(struct ext4_orphan_block),
+			       GFP_KERNEL);
+	if (!oi->of_binfo) {
+		ret = -ENOMEM;
+		goto out_put;
+	}
+	for (i = 0; i < oi->of_blocks; i++) {
+		oi->of_binfo[i].ob_bh = ext4_bread(NULL, inode, i, 0);
+		if (IS_ERR(oi->of_binfo[i].ob_bh)) {
+			ret = PTR_ERR(oi->of_binfo[i].ob_bh);
+			goto out_free;
+		}
+		if (!oi->of_binfo[i].ob_bh) {
+			ret = -EIO;
+			goto out_free;
+		}
+		ot = ext4_orphan_block_tail(sb, oi->of_binfo[i].ob_bh);
+		if (le32_to_cpu(ot->ob_magic) != EXT4_ORPHAN_BLOCK_MAGIC) {
+			ext4_error(sb, "orphan file block %d: bad magic", i);
+			ret = -EIO;
+			goto out_free;
+		}
+		if (!ext4_orphan_file_block_csum_verify(sb,
+						oi->of_binfo[i].ob_bh)) {
+			ext4_error(sb, "orphan file block %d: bad checksum", i);
+			ret = -EIO;
+			goto out_free;
+		}
+		bdata = (__le32 *)(oi->of_binfo[i].ob_bh->b_data);
+		free = 0;
+		for (j = 0; j < inodes_per_ob; j++)
+			if (bdata[j] == 0)
+				free++;
+		oi->of_binfo[i].ob_free_entries = free;
+	}
+	iput(inode);
+	return 0;
+out_free:
+	for (i--; i >= 0; i--)
+		brelse(oi->of_binfo[i].ob_bh);
+	kfree(oi->of_binfo);
+out_put:
+	iput(inode);
+	return ret;
+}
+
+int ext4_orphan_file_empty(struct super_block *sb)
+{
+	struct ext4_orphan_info *oi = &EXT4_SB(sb)->s_orphan_info;
+	int i;
+	int inodes_per_ob = ext4_inodes_per_orphan_block(sb);
+
+	if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_ORPHAN_FILE))
+		return 1;
+	for (i = 0; i < oi->of_blocks; i++)
+		if (oi->of_binfo[i].ob_free_entries != inodes_per_ob)
+			return 0;
+	return 1;
+}
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index a8ea6e00b23f..e3152cddd63b 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -756,6 +756,7 @@ static void ext4_put_super(struct super_block *sb)
 
 	flush_workqueue(sbi->rsv_conversion_wq);
 	destroy_workqueue(sbi->rsv_conversion_wq);
+	ext4_release_orphan_info(sb);
 
 	if (sbi->s_journal) {
 		err = jbd2_journal_destroy(sbi->s_journal);
@@ -773,6 +774,8 @@ static void ext4_put_super(struct super_block *sb)
 
 	if (!(sb->s_flags & MS_RDONLY)) {
 		EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
+		EXT4_CLEAR_RO_COMPAT_FEATURE(sb,
+			EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT);
 		es->s_state = cpu_to_le16(sbi->s_mount_state);
 	}
 	if (!(sb->s_flags & MS_RDONLY))
@@ -1902,8 +1905,14 @@ static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
 	le16_add_cpu(&es->s_mnt_count, 1);
 	es->s_mtime = cpu_to_le32(get_seconds());
 	ext4_update_dynamic_rev(sb);
-	if (sbi->s_journal)
+	if (sbi->s_journal) {
 		EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
+		if (EXT4_HAS_COMPAT_FEATURE(sb,
+					    EXT4_FEATURE_COMPAT_ORPHAN_FILE)) {
+			EXT4_SET_RO_COMPAT_FEATURE(sb,
+				EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT);
+		}
+	}
 
 	ext4_commit_super(sb, 1);
 done:
@@ -3393,6 +3402,8 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 		silent = 1;
 		goto cantfind_ext4;
 	}
+	ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
+				ext4_orphan_file_block_trigger);
 
 	/* Load the checksum driver */
 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
@@ -3873,8 +3884,10 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	sb->s_root = NULL;
 
 	needs_recovery = (es->s_last_orphan != 0 ||
+			  EXT4_HAS_RO_COMPAT_FEATURE(sb,
+				EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT) ||
 			  EXT4_HAS_INCOMPAT_FEATURE(sb,
-				    EXT4_FEATURE_INCOMPAT_RECOVER));
+				EXT4_FEATURE_INCOMPAT_RECOVER));
 
 	if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
 	    !(sb->s_flags & MS_RDONLY))
@@ -4099,13 +4112,16 @@ no_journal:
 	if (err)
 		goto failed_mount7;
 
+	err = ext4_init_orphan_info(sb);
+	if (err)
+		goto failed_mount8;
 #ifdef CONFIG_QUOTA
 	/* Enable quota usage during mount. */
 	if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
 	    !(sb->s_flags & MS_RDONLY)) {
 		err = ext4_enable_quotas(sb);
 		if (err)
-			goto failed_mount8;
+			goto failed_mount9;
 	}
 #endif  /* CONFIG_QUOTA */
 
@@ -4155,9 +4171,11 @@ cantfind_ext4:
 	goto failed_mount;
 
 #ifdef CONFIG_QUOTA
+failed_mount9:
+	ext4_release_orphan_info(sb);
+#endif
 failed_mount8:
 	kobject_del(&sbi->s_kobj);
-#endif
 failed_mount7:
 	ext4_unregister_li_request(sb);
 failed_mount6:
@@ -4696,6 +4714,10 @@ static int ext4_freeze(struct super_block *sb)
 
 	/* Journal blocked and flushed, clear needs_recovery flag. */
 	EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
+	if (ext4_orphan_file_empty(sb)) {
+		EXT4_CLEAR_RO_COMPAT_FEATURE(sb,
+			EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT);
+	}
 	error = ext4_commit_super(sb, 1);
 out:
 	if (journal)
@@ -4715,6 +4737,10 @@ static int ext4_unfreeze(struct super_block *sb)
 
 	/* Reset the needs_recovery flag before the fs is unlocked. */
 	EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
+	if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_ORPHAN_FILE)) {
+		EXT4_SET_RO_COMPAT_FEATURE(sb,
+			EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT);
+	}
 	ext4_commit_super(sb, 1);
 	return 0;
 }
@@ -4858,8 +4884,13 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
 			    (sbi->s_mount_state & EXT4_VALID_FS))
 				es->s_state = cpu_to_le16(sbi->s_mount_state);
 
-			if (sbi->s_journal)
+			if (sbi->s_journal) {
 				ext4_mark_recovery_complete(sb, es);
+				if (ext4_orphan_file_empty(sb)) {
+					EXT4_CLEAR_RO_COMPAT_FEATURE(sb,
+						EXT4_FEATURE_RO_COMPAT_ORPHAN_PRESENT);
+				}
+			}
 		} else {
 			/* Make sure we can mount this feature set readwrite */
 			if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
@@ -4891,7 +4922,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
 			 * around from a previously readonly bdev mount,
 			 * require a full umount/remount for now.
 			 */
-			if (es->s_last_orphan) {
+			if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
 				ext4_msg(sb, KERN_WARNING, "Couldn't "
 				       "remount RDWR because of unprocessed "
 				       "orphan inode list.  Please "
-- 
2.1.4