Performance improvements in SEEK_DATA and xattr scalability

improvements, plus a lot of clean ups and bug fixes.
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v2
 
 iQEcBAABCAAGBQJW6c9mAAoJEPL5WVaVDYGjWsEIAJkWUvKB3GgGgP82sKDBP2P8
 IbWegO1ICMrSY78BqLI7mLCqggH5JClBgYU3O4VFv8Brj1L9mS5X+vflaDE1j9jj
 Ik1KZKtZl1opOwO1L3D4l/ipZAiENUp7NehTtpsFousmz6nMZ5vo6x4t3QSwbUIm
 YXpxUIxHEhBcW5i3EDkfYG8305V5oj8HsVf6T98OlWGpBO5VGNMAHvA7CQdQe7Rd
 chv70rij5V684bJAEoosEFXVAuOUrxcBqbFA3Nlb432YOPj0ISLx76kw0GIjUYtf
 yjoSClbRgwxGzh0jm+yaoYjjm83xbsYbHSsBmh3+/QLMbKTLXeCqR/BiqJavmcM=
 =bWpz
 -----END PGP SIGNATURE-----

Merge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4

Pull ext4 updates from Ted Ts'o:
 "Performance improvements in SEEK_DATA and xattr scalability
  improvements, plus a lot of clean ups and bug fixes"

* tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (38 commits)
  ext4: clean up error handling in the MMP support
  jbd2: do not fail journal because of frozen_buffer allocation failure
  ext4: use __GFP_NOFAIL in ext4_free_blocks()
  ext4: fix compile error while opening the macro DOUBLE_CHECK
  ext4: print ext4 mount option data_err=abort correctly
  ext4: fix NULL pointer dereference in ext4_mark_inode_dirty()
  ext4: drop unneeded BUFFER_TRACE in ext4_delete_inline_entry()
  ext4: fix misspellings in comments.
  jbd2: fix FS corruption possibility in jbd2_journal_destroy() on umount path
  ext4: more efficient SEEK_DATA implementation
  ext4: cleanup handling of bh->b_state in DAX mmap
  ext4: return hole from ext4_map_blocks()
  ext4: factor out determining of hole size
  ext4: fix setting of referenced bit in ext4_es_lookup_extent()
  ext4: remove i_ioend_count
  ext4: simplify io_end handling for AIO DIO
  ext4: move trans handling and completion deferal out of _ext4_get_block
  ext4: rename and split get blocks functions
  ext4: use i_mutex to serialize unaligned AIO DIO
  ext4: pack ioend structure better
  ...

fs/ext2/ext2.h

@@ -61,6 +61,8 @@ struct ext2_block_alloc_info {
 #define rsv_start rsv_window._rsv_start
 #define rsv_end rsv_window._rsv_end
 
+struct mb_cache;
+
 /*
  * second extended-fs super-block data in memory
  */
@@ -111,6 +113,7 @@ struct ext2_sb_info {
 	 * of the mount options.
 	 */
 	spinlock_t s_lock;
+	struct mb_cache *s_mb_cache;
 };
 
 static inline spinlock_t *

fs/ext2/super.c

@@ -131,7 +131,10 @@ static void ext2_put_super (struct super_block * sb)
 
 	dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
 
-	ext2_xattr_put_super(sb);
+	if (sbi->s_mb_cache) {
+		ext2_xattr_destroy_cache(sbi->s_mb_cache);
+		sbi->s_mb_cache = NULL;
+	}
 	if (!(sb->s_flags & MS_RDONLY)) {
 		struct ext2_super_block *es = sbi->s_es;
 
@@ -1104,6 +1107,14 @@ static int ext2_fill_super(struct super_block *sb, void *data, int silent)
 		ext2_msg(sb, KERN_ERR, "error: insufficient memory");
 		goto failed_mount3;
 	}
+
+#ifdef CONFIG_EXT2_FS_XATTR
+	sbi->s_mb_cache = ext2_xattr_create_cache();
+	if (!sbi->s_mb_cache) {
+		ext2_msg(sb, KERN_ERR, "Failed to create an mb_cache");
+		goto failed_mount3;
+	}
+#endif
 	/*
 	 * set up enough so that it can read an inode
 	 */
@@ -1149,6 +1160,8 @@ cantfind_ext2:
 			sb->s_id);
 	goto failed_mount;
 failed_mount3:
+	if (sbi->s_mb_cache)
+		ext2_xattr_destroy_cache(sbi->s_mb_cache);
 	percpu_counter_destroy(&sbi->s_freeblocks_counter);
 	percpu_counter_destroy(&sbi->s_freeinodes_counter);
 	percpu_counter_destroy(&sbi->s_dirs_counter);
@@ -1555,20 +1568,17 @@ MODULE_ALIAS_FS("ext2");
 
 static int __init init_ext2_fs(void)
 {
-	int err = init_ext2_xattr();
+	int err;
-	if (err)
+
-		return err;
 	err = init_inodecache();
 	if (err)
-		goto out1;
+		return err;
         err = register_filesystem(&ext2_fs_type);
 	if (err)
 		goto out;
 	return 0;
 out:
 	destroy_inodecache();
-out1:
-	exit_ext2_xattr();
 	return err;
 }
 
@@ -1576,7 +1586,6 @@ static void __exit exit_ext2_fs(void)
 {
 	unregister_filesystem(&ext2_fs_type);
 	destroy_inodecache();
-	exit_ext2_xattr();
 }
 
 MODULE_AUTHOR("Remy Card and others");

fs/ext2/xattr.c

@@ -90,14 +90,12 @@
 static int ext2_xattr_set2(struct inode *, struct buffer_head *,
 			   struct ext2_xattr_header *);
 
-static int ext2_xattr_cache_insert(struct buffer_head *);
+static int ext2_xattr_cache_insert(struct mb_cache *, struct buffer_head *);
 static struct buffer_head *ext2_xattr_cache_find(struct inode *,
 						 struct ext2_xattr_header *);
 static void ext2_xattr_rehash(struct ext2_xattr_header *,
 			      struct ext2_xattr_entry *);
 
-static struct mb_cache *ext2_xattr_cache;
-
 static const struct xattr_handler *ext2_xattr_handler_map[] = {
 	[EXT2_XATTR_INDEX_USER]		     = &ext2_xattr_user_handler,
 #ifdef CONFIG_EXT2_FS_POSIX_ACL
@@ -152,6 +150,7 @@ ext2_xattr_get(struct inode *inode, int name_index, const char *name,
 	size_t name_len, size;
 	char *end;
 	int error;
+	struct mb_cache *ext2_mb_cache = EXT2_SB(inode->i_sb)->s_mb_cache;
 
 	ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
 		  name_index, name, buffer, (long)buffer_size);
@@ -196,7 +195,7 @@ bad_block:	ext2_error(inode->i_sb, "ext2_xattr_get",
 			goto found;
 		entry = next;
 	}
-	if (ext2_xattr_cache_insert(bh))
+	if (ext2_xattr_cache_insert(ext2_mb_cache, bh))
 		ea_idebug(inode, "cache insert failed");
 	error = -ENODATA;
 	goto cleanup;
@@ -209,7 +208,7 @@ found:
 	    le16_to_cpu(entry->e_value_offs) + size > inode->i_sb->s_blocksize)
 		goto bad_block;
 
-	if (ext2_xattr_cache_insert(bh))
+	if (ext2_xattr_cache_insert(ext2_mb_cache, bh))
 		ea_idebug(inode, "cache insert failed");
 	if (buffer) {
 		error = -ERANGE;
@@ -247,6 +246,7 @@ ext2_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
 	char *end;
 	size_t rest = buffer_size;
 	int error;
+	struct mb_cache *ext2_mb_cache = EXT2_SB(inode->i_sb)->s_mb_cache;
 
 	ea_idebug(inode, "buffer=%p, buffer_size=%ld",
 		  buffer, (long)buffer_size);
@@ -281,7 +281,7 @@ bad_block:	ext2_error(inode->i_sb, "ext2_xattr_list",
 			goto bad_block;
 		entry = next;
 	}
-	if (ext2_xattr_cache_insert(bh))
+	if (ext2_xattr_cache_insert(ext2_mb_cache, bh))
 		ea_idebug(inode, "cache insert failed");
 
 	/* list the attribute names */
@@ -483,22 +483,23 @@ bad_block:		ext2_error(sb, "ext2_xattr_set",
 	/* Here we know that we can set the new attribute. */
 
 	if (header) {
-		struct mb_cache_entry *ce;
-
 		/* assert(header == HDR(bh)); */
-		ce = mb_cache_entry_get(ext2_xattr_cache, bh->b_bdev,
-					bh->b_blocknr);
 		lock_buffer(bh);
 		if (header->h_refcount == cpu_to_le32(1)) {
+			__u32 hash = le32_to_cpu(header->h_hash);
+
 			ea_bdebug(bh, "modifying in-place");
-			if (ce)
+			/*
-				mb_cache_entry_free(ce);
+			 * This must happen under buffer lock for
+			 * ext2_xattr_set2() to reliably detect modified block
+			 */
+			mb_cache_entry_delete_block(EXT2_SB(sb)->s_mb_cache,
+						    hash, bh->b_blocknr);
+
 			/* keep the buffer locked while modifying it. */
 		} else {
 			int offset;
 
-			if (ce)
-				mb_cache_entry_release(ce);
 			unlock_buffer(bh);
 			ea_bdebug(bh, "cloning");
 			header = kmalloc(bh->b_size, GFP_KERNEL);
@@ -626,6 +627,7 @@ ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh,
 	struct super_block *sb = inode->i_sb;
 	struct buffer_head *new_bh = NULL;
 	int error;
+	struct mb_cache *ext2_mb_cache = EXT2_SB(sb)->s_mb_cache;
 
 	if (header) {
 		new_bh = ext2_xattr_cache_find(inode, header);
@@ -653,7 +655,7 @@ ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh,
 			   don't need to change the reference count. */
 			new_bh = old_bh;
 			get_bh(new_bh);
-			ext2_xattr_cache_insert(new_bh);
+			ext2_xattr_cache_insert(ext2_mb_cache, new_bh);
 		} else {
 			/* We need to allocate a new block */
 			ext2_fsblk_t goal = ext2_group_first_block_no(sb,
@@ -674,7 +676,7 @@ ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh,
 			memcpy(new_bh->b_data, header, new_bh->b_size);
 			set_buffer_uptodate(new_bh);
 			unlock_buffer(new_bh);
-			ext2_xattr_cache_insert(new_bh);
+			ext2_xattr_cache_insert(ext2_mb_cache, new_bh);
 			
 			ext2_xattr_update_super_block(sb);
 		}
@@ -707,19 +709,21 @@ ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh,
 
 	error = 0;
 	if (old_bh && old_bh != new_bh) {
-		struct mb_cache_entry *ce;
-
 		/*
 		 * If there was an old block and we are no longer using it,
 		 * release the old block.
 		 */
-		ce = mb_cache_entry_get(ext2_xattr_cache, old_bh->b_bdev,
-					old_bh->b_blocknr);
 		lock_buffer(old_bh);
 		if (HDR(old_bh)->h_refcount == cpu_to_le32(1)) {
+			__u32 hash = le32_to_cpu(HDR(old_bh)->h_hash);
+
+			/*
+			 * This must happen under buffer lock for
+			 * ext2_xattr_set2() to reliably detect freed block
+			 */
+			mb_cache_entry_delete_block(ext2_mb_cache,
+						    hash, old_bh->b_blocknr);
 			/* Free the old block. */
-			if (ce)
-				mb_cache_entry_free(ce);
 			ea_bdebug(old_bh, "freeing");
 			ext2_free_blocks(inode, old_bh->b_blocknr, 1);
 			mark_inode_dirty(inode);
@@ -730,8 +734,6 @@ ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh,
 		} else {
 			/* Decrement the refcount only. */
 			le32_add_cpu(&HDR(old_bh)->h_refcount, -1);
-			if (ce)
-				mb_cache_entry_release(ce);
 			dquot_free_block_nodirty(inode, 1);
 			mark_inode_dirty(inode);
 			mark_buffer_dirty(old_bh);
@@ -757,7 +759,6 @@ void
 ext2_xattr_delete_inode(struct inode *inode)
 {
 	struct buffer_head *bh = NULL;
-	struct mb_cache_entry *ce;
 
 	down_write(&EXT2_I(inode)->xattr_sem);
 	if (!EXT2_I(inode)->i_file_acl)
@@ -777,19 +778,22 @@ ext2_xattr_delete_inode(struct inode *inode)
 			EXT2_I(inode)->i_file_acl);
 		goto cleanup;
 	}
-	ce = mb_cache_entry_get(ext2_xattr_cache, bh->b_bdev, bh->b_blocknr);
 	lock_buffer(bh);
 	if (HDR(bh)->h_refcount == cpu_to_le32(1)) {
-		if (ce)
+		__u32 hash = le32_to_cpu(HDR(bh)->h_hash);
-			mb_cache_entry_free(ce);
+
+		/*
+		 * This must happen under buffer lock for ext2_xattr_set2() to
+		 * reliably detect freed block
+		 */
+		mb_cache_entry_delete_block(EXT2_SB(inode->i_sb)->s_mb_cache,
+					    hash, bh->b_blocknr);
 		ext2_free_blocks(inode, EXT2_I(inode)->i_file_acl, 1);
 		get_bh(bh);
 		bforget(bh);
 		unlock_buffer(bh);
 	} else {
 		le32_add_cpu(&HDR(bh)->h_refcount, -1);
-		if (ce)
-			mb_cache_entry_release(ce);
 		ea_bdebug(bh, "refcount now=%d",
 			le32_to_cpu(HDR(bh)->h_refcount));
 		unlock_buffer(bh);
@@ -805,18 +809,6 @@ cleanup:
 	up_write(&EXT2_I(inode)->xattr_sem);
 }
 
-/*
- * ext2_xattr_put_super()
- *
- * This is called when a file system is unmounted.
- */
-void
-ext2_xattr_put_super(struct super_block *sb)
-{
-	mb_cache_shrink(sb->s_bdev);
-}
-
-
 /*
  * ext2_xattr_cache_insert()
  *
@@ -826,28 +818,20 @@ ext2_xattr_put_super(struct super_block *sb)
  * Returns 0, or a negative error number on failure.
  */
 static int
-ext2_xattr_cache_insert(struct buffer_head *bh)
+ext2_xattr_cache_insert(struct mb_cache *cache, struct buffer_head *bh)
 {
 	__u32 hash = le32_to_cpu(HDR(bh)->h_hash);
-	struct mb_cache_entry *ce;
 	int error;
 
-	ce = mb_cache_entry_alloc(ext2_xattr_cache, GFP_NOFS);
+	error = mb_cache_entry_create(cache, GFP_NOFS, hash, bh->b_blocknr, 1);
-	if (!ce)
-		return -ENOMEM;
-	error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
 	if (error) {
-		mb_cache_entry_free(ce);
 		if (error == -EBUSY) {
 			ea_bdebug(bh, "already in cache (%d cache entries)",
 				atomic_read(&ext2_xattr_cache->c_entry_count));
 			error = 0;
 		}
-	} else {
+	} else
-		ea_bdebug(bh, "inserting [%x] (%d cache entries)", (int)hash,
+		ea_bdebug(bh, "inserting [%x]", (int)hash);
-			  atomic_read(&ext2_xattr_cache->c_entry_count));
-		mb_cache_entry_release(ce);
-	}
 	return error;
 }
 
@@ -904,22 +888,16 @@ ext2_xattr_cache_find(struct inode *inode, struct ext2_xattr_header *header)
 {
 	__u32 hash = le32_to_cpu(header->h_hash);
 	struct mb_cache_entry *ce;
+	struct mb_cache *ext2_mb_cache = EXT2_SB(inode->i_sb)->s_mb_cache;
 
 	if (!header->h_hash)
 		return NULL;  /* never share */
 	ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
 again:
-	ce = mb_cache_entry_find_first(ext2_xattr_cache, inode->i_sb->s_bdev,
+	ce = mb_cache_entry_find_first(ext2_mb_cache, hash);
-				       hash);
 	while (ce) {
 		struct buffer_head *bh;
 
-		if (IS_ERR(ce)) {
-			if (PTR_ERR(ce) == -EAGAIN)
-				goto again;
-			break;
-		}
-
 		bh = sb_bread(inode->i_sb, ce->e_block);
 		if (!bh) {
 			ext2_error(inode->i_sb, "ext2_xattr_cache_find",
@@ -927,7 +905,21 @@ again:
 				inode->i_ino, (unsigned long) ce->e_block);
 		} else {
 			lock_buffer(bh);
-			if (le32_to_cpu(HDR(bh)->h_refcount) >
+			/*
+			 * We have to be careful about races with freeing or
+			 * rehashing of xattr block. Once we hold buffer lock
+			 * xattr block's state is stable so we can check
+			 * whether the block got freed / rehashed or not.
+			 * Since we unhash mbcache entry under buffer lock when
+			 * freeing / rehashing xattr block, checking whether
+			 * entry is still hashed is reliable.
+			 */
+			if (hlist_bl_unhashed(&ce->e_hash_list)) {
+				mb_cache_entry_put(ext2_mb_cache, ce);
+				unlock_buffer(bh);
+				brelse(bh);
+				goto again;
+			} else if (le32_to_cpu(HDR(bh)->h_refcount) >
 				   EXT2_XATTR_REFCOUNT_MAX) {
 				ea_idebug(inode, "block %ld refcount %d>%d",
 					  (unsigned long) ce->e_block,
@@ -936,13 +928,14 @@ again:
 			} else if (!ext2_xattr_cmp(header, HDR(bh))) {
 				ea_bdebug(bh, "b_count=%d",
 					  atomic_read(&(bh->b_count)));
-				mb_cache_entry_release(ce);
+				mb_cache_entry_touch(ext2_mb_cache, ce);
+				mb_cache_entry_put(ext2_mb_cache, ce);
 				return bh;
 			}
 			unlock_buffer(bh);
 			brelse(bh);
 		}
-		ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
+		ce = mb_cache_entry_find_next(ext2_mb_cache, ce);
 	}
 	return NULL;
 }
@@ -1015,17 +1008,15 @@ static void ext2_xattr_rehash(struct ext2_xattr_header *header,
 
 #undef BLOCK_HASH_SHIFT
 
-int __init
+#define HASH_BUCKET_BITS 10
-init_ext2_xattr(void)
+
+struct mb_cache *ext2_xattr_create_cache(void)
 {
-	ext2_xattr_cache = mb_cache_create("ext2_xattr", 6);
+	return mb_cache_create(HASH_BUCKET_BITS);
-	if (!ext2_xattr_cache)
-		return -ENOMEM;
-	return 0;
 }
 
-void
+void ext2_xattr_destroy_cache(struct mb_cache *cache)
-exit_ext2_xattr(void)
 {
-	mb_cache_destroy(ext2_xattr_cache);
+	if (cache)
+		mb_cache_destroy(cache);
 }

fs/ext2/xattr.h

@@ -53,6 +53,8 @@ struct ext2_xattr_entry {
 #define EXT2_XATTR_SIZE(size) \
 	(((size) + EXT2_XATTR_ROUND) & ~EXT2_XATTR_ROUND)
 
+struct mb_cache;
+
 # ifdef CONFIG_EXT2_FS_XATTR
 
 extern const struct xattr_handler ext2_xattr_user_handler;
@@ -65,10 +67,9 @@ extern int ext2_xattr_get(struct inode *, int, const char *, void *, size_t);
 extern int ext2_xattr_set(struct inode *, int, const char *, const void *, size_t, int);
 
 extern void ext2_xattr_delete_inode(struct inode *);
-extern void ext2_xattr_put_super(struct super_block *);
 
-extern int init_ext2_xattr(void);
+extern struct mb_cache *ext2_xattr_create_cache(void);
-extern void exit_ext2_xattr(void);
+extern void ext2_xattr_destroy_cache(struct mb_cache *cache);
 
 extern const struct xattr_handler *ext2_xattr_handlers[];
 
@@ -93,19 +94,7 @@ ext2_xattr_delete_inode(struct inode *inode)
 {
 }
 
-static inline void
+static inline void ext2_xattr_destroy_cache(struct mb_cache *cache)
-ext2_xattr_put_super(struct super_block *sb)
-{
-}
-
-static inline int
-init_ext2_xattr(void)
-{
-	return 0;
-}
-
-static inline void
-exit_ext2_xattr(void)
 {
 }
 

fs/ext4/ext4.h

@@ -41,6 +41,18 @@
  * The fourth extended filesystem constants/structures
  */
 
+/*
+ * with AGGRESSIVE_CHECK allocator runs consistency checks over
+ * structures. these checks slow things down a lot
+ */
+#define AGGRESSIVE_CHECK__
+
+/*
+ * with DOUBLE_CHECK defined mballoc creates persistent in-core
+ * bitmaps, maintains and uses them to check for double allocations
+ */
+#define DOUBLE_CHECK__
+
 /*
  * Define EXT4FS_DEBUG to produce debug messages
  */
@@ -182,9 +194,9 @@ typedef struct ext4_io_end {
 	struct bio		*bio;		/* Linked list of completed
 						 * bios covering the extent */
 	unsigned int		flag;		/* unwritten or not */
+	atomic_t		count;		/* reference counter */
 	loff_t			offset;		/* offset in the file */
 	ssize_t			size;		/* size of the extent */
-	atomic_t		count;		/* reference counter */
 } ext4_io_end_t;
 
 struct ext4_io_submit {
@@ -1024,13 +1036,8 @@ struct ext4_inode_info {
 	 * transaction reserved
 	 */
 	struct list_head i_rsv_conversion_list;
-	/*
-	 * Completed IOs that need unwritten extents handling and don't have
-	 * transaction reserved
-	 */
-	atomic_t i_ioend_count;	/* Number of outstanding io_end structs */
-	atomic_t i_unwritten; /* Nr. of inflight conversions pending */
 	struct work_struct i_rsv_conversion_work;
+	atomic_t i_unwritten; /* Nr. of inflight conversions pending */
 
 	spinlock_t i_block_reservation_lock;
 
@@ -1513,16 +1520,6 @@ static inline void ext4_set_io_unwritten_flag(struct inode *inode,
 	}
 }
 
-static inline ext4_io_end_t *ext4_inode_aio(struct inode *inode)
-{
-	return inode->i_private;
-}
-
-static inline void ext4_inode_aio_set(struct inode *inode, ext4_io_end_t *io)
-{
-	inode->i_private = io;
-}
-
 /*
  * Inode dynamic state flags
  */
@@ -2506,12 +2503,14 @@ extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
 int ext4_inode_is_fast_symlink(struct inode *inode);
 struct buffer_head *ext4_getblk(handle_t *, struct inode *, ext4_lblk_t, int);
 struct buffer_head *ext4_bread(handle_t *, struct inode *, ext4_lblk_t, int);
-int ext4_get_block_write(struct inode *inode, sector_t iblock,
+int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
-			 struct buffer_head *bh_result, int create);
+			     struct buffer_head *bh_result, int create);
 int ext4_dax_mmap_get_block(struct inode *inode, sector_t iblock,
 			    struct buffer_head *bh_result, int create);
 int ext4_get_block(struct inode *inode, sector_t iblock,
-				struct buffer_head *bh_result, int create);
+		   struct buffer_head *bh_result, int create);
+int ext4_dio_get_block(struct inode *inode, sector_t iblock,
+		       struct buffer_head *bh_result, int create);
 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,
@@ -2559,6 +2558,9 @@ extern void ext4_da_update_reserve_space(struct inode *inode,
 					int used, int quota_claim);
 extern int ext4_issue_zeroout(struct inode *inode, ext4_lblk_t lblk,
 			      ext4_fsblk_t pblk, ext4_lblk_t len);
+extern int ext4_get_next_extent(struct inode *inode, ext4_lblk_t lblk,
+				unsigned int map_len,
+				struct extent_status *result);
 
 /* indirect.c */
 extern int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
@@ -3285,10 +3287,7 @@ static inline void ext4_inode_resume_unlocked_dio(struct inode *inode)
 #define EXT4_WQ_HASH_SZ		37
 #define ext4_ioend_wq(v)   (&ext4__ioend_wq[((unsigned long)(v)) %\
 					    EXT4_WQ_HASH_SZ])
-#define ext4_aio_mutex(v)  (&ext4__aio_mutex[((unsigned long)(v)) %\
-					     EXT4_WQ_HASH_SZ])
 extern wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
-extern struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
 
 #define EXT4_RESIZING	0
 extern int ext4_resize_begin(struct super_block *sb);

fs/ext4/ext4_extents.h

@@ -11,7 +11,7 @@
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  *
- * You should have received a copy of the GNU General Public Licens
+ * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-
  */

fs/ext4/extents.c

@@ -15,7 +15,7 @@
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  *
- * You should have received a copy of the GNU General Public Licens
+ * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-
  */
@@ -1736,6 +1736,12 @@ ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
 	 */
 	if (ext1_ee_len + ext2_ee_len > EXT_INIT_MAX_LEN)
 		return 0;
+	/*
+	 * The check for IO to unwritten extent is somewhat racy as we
+	 * increment i_unwritten / set EXT4_STATE_DIO_UNWRITTEN only after
+	 * dropping i_data_sem. But reserved blocks should save us in that
+	 * case.
+	 */
 	if (ext4_ext_is_unwritten(ex1) &&
 	    (ext4_test_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN) ||
 	     atomic_read(&EXT4_I(inode)->i_unwritten) ||
@@ -2292,60 +2298,70 @@ static int ext4_fill_fiemap_extents(struct inode *inode,
 	return err;
 }
 
+/*
+ * ext4_ext_determine_hole - determine hole around given block
+ * @inode:	inode we lookup in
+ * @path:	path in extent tree to @lblk
+ * @lblk:	pointer to logical block around which we want to determine hole
+ *
+ * Determine hole length (and start if easily possible) around given logical
+ * block. We don't try too hard to find the beginning of the hole but @path
+ * actually points to extent before @lblk, we provide it.
+ *
+ * The function returns the length of a hole starting at @lblk. We update @lblk
+ * to the beginning of the hole if we managed to find it.
+ */
+static ext4_lblk_t ext4_ext_determine_hole(struct inode *inode,
+					   struct ext4_ext_path *path,
+					   ext4_lblk_t *lblk)
+{
+	int depth = ext_depth(inode);
+	struct ext4_extent *ex;
+	ext4_lblk_t len;
+
+	ex = path[depth].p_ext;
+	if (ex == NULL) {
+		/* there is no extent yet, so gap is [0;-] */
+		*lblk = 0;
+		len = EXT_MAX_BLOCKS;
+	} else if (*lblk < le32_to_cpu(ex->ee_block)) {
+		len = le32_to_cpu(ex->ee_block) - *lblk;
+	} else if (*lblk >= le32_to_cpu(ex->ee_block)
+			+ ext4_ext_get_actual_len(ex)) {
+		ext4_lblk_t next;
+
+		*lblk = le32_to_cpu(ex->ee_block) + ext4_ext_get_actual_len(ex);
+		next = ext4_ext_next_allocated_block(path);
+		BUG_ON(next == *lblk);
+		len = next - *lblk;
+	} else {
+		BUG();
+	}
+	return len;
+}
+
 /*
  * ext4_ext_put_gap_in_cache:
  * calculate boundaries of the gap that the requested block fits into
  * and cache this gap
  */
 static void
-ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
+ext4_ext_put_gap_in_cache(struct inode *inode, ext4_lblk_t hole_start,
-				ext4_lblk_t block)
+			  ext4_lblk_t hole_len)
 {
-	int depth = ext_depth(inode);
-	ext4_lblk_t len;
-	ext4_lblk_t lblock;
-	struct ext4_extent *ex;
 	struct extent_status es;
 
-	ex = path[depth].p_ext;
+	ext4_es_find_delayed_extent_range(inode, hole_start,
-	if (ex == NULL) {
+					  hole_start + hole_len - 1, &es);
-		/* there is no extent yet, so gap is [0;-] */
-		lblock = 0;
-		len = EXT_MAX_BLOCKS;
-		ext_debug("cache gap(whole file):");
-	} else if (block < le32_to_cpu(ex->ee_block)) {
-		lblock = block;
-		len = le32_to_cpu(ex->ee_block) - block;
-		ext_debug("cache gap(before): %u [%u:%u]",
-				block,
-				le32_to_cpu(ex->ee_block),
-				 ext4_ext_get_actual_len(ex));
-	} else if (block >= le32_to_cpu(ex->ee_block)
-			+ ext4_ext_get_actual_len(ex)) {
-		ext4_lblk_t next;
-		lblock = le32_to_cpu(ex->ee_block)
-			+ ext4_ext_get_actual_len(ex);
-
-		next = ext4_ext_next_allocated_block(path);
-		ext_debug("cache gap(after): [%u:%u] %u",
-				le32_to_cpu(ex->ee_block),
-				ext4_ext_get_actual_len(ex),
-				block);
-		BUG_ON(next == lblock);
-		len = next - lblock;
-	} else {
-		BUG();
-	}
-
-	ext4_es_find_delayed_extent_range(inode, lblock, lblock + len - 1, &es);
 	if (es.es_len) {
 		/* There's delayed extent containing lblock? */
-		if (es.es_lblk <= lblock)
+		if (es.es_lblk <= hole_start)
 			return;
-		len = min(es.es_lblk - lblock, len);
+		hole_len = min(es.es_lblk - hole_start, hole_len);
 	}
-	ext_debug(" -> %u:%u\n", lblock, len);
+	ext_debug(" -> %u:%u\n", hole_start, hole_len);
-	ext4_es_insert_extent(inode, lblock, len, ~0, EXTENT_STATUS_HOLE);
+	ext4_es_insert_extent(inode, hole_start, hole_len, ~0,
+			      EXTENT_STATUS_HOLE);
 }
 
 /*
@@ -3927,7 +3943,7 @@ get_reserved_cluster_alloc(struct inode *inode, ext4_lblk_t lblk_start,
 static int
 convert_initialized_extent(handle_t *handle, struct inode *inode,
 			   struct ext4_map_blocks *map,
-			   struct ext4_ext_path **ppath, int flags,
+			   struct ext4_ext_path **ppath,
 			   unsigned int allocated)
 {
 	struct ext4_ext_path *path = *ppath;
@@ -4007,7 +4023,6 @@ ext4_ext_handle_unwritten_extents(handle_t *handle, struct inode *inode,
 	struct ext4_ext_path *path = *ppath;
 	int ret = 0;
 	int err = 0;
-	ext4_io_end_t *io = ext4_inode_aio(inode);
 
 	ext_debug("ext4_ext_handle_unwritten_extents: inode %lu, logical "
 		  "block %llu, max_blocks %u, flags %x, allocated %u\n",
@@ -4030,15 +4045,6 @@ ext4_ext_handle_unwritten_extents(handle_t *handle, struct inode *inode,
 					 flags | EXT4_GET_BLOCKS_CONVERT);
 		if (ret <= 0)
 			goto out;
-		/*
-		 * Flag the inode(non aio case) or end_io struct (aio case)
-		 * that this IO needs to conversion to written when IO is
-		 * completed
-		 */
-		if (io)
-			ext4_set_io_unwritten_flag(inode, io);
-		else
-			ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
 		map->m_flags |= EXT4_MAP_UNWRITTEN;
 		goto out;
 	}
@@ -4283,9 +4289,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 	unsigned int allocated = 0, offset = 0;
 	unsigned int allocated_clusters = 0;
 	struct ext4_allocation_request ar;
-	ext4_io_end_t *io = ext4_inode_aio(inode);
 	ext4_lblk_t cluster_offset;
-	int set_unwritten = 0;
 	bool map_from_cluster = false;
 
 	ext_debug("blocks %u/%u requested for inode %lu\n",
@@ -4347,7 +4351,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 			    (flags & EXT4_GET_BLOCKS_CONVERT_UNWRITTEN)) {
 				allocated = convert_initialized_extent(
 						handle, inode, map, &path,
-						flags, allocated);
+						allocated);
 				goto out2;
 			} else if (!ext4_ext_is_unwritten(ex))
 				goto out;
@@ -4368,11 +4372,22 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
 	 * we couldn't try to create block if create flag is zero
 	 */
 	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
+		ext4_lblk_t hole_start, hole_len;
+
+		hole_start = map->m_lblk;
+		hole_len = ext4_ext_determine_hole(inode, path, &hole_start);
 		/*
 		 * put just found gap into cache to speed up
 		 * subsequent requests
 		 */
-		ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
+		ext4_ext_put_gap_in_cache(inode, hole_start, hole_len);
+
+		/* Update hole_len to reflect hole size after map->m_lblk */
+		if (hole_start != map->m_lblk)
+			hole_len -= map->m_lblk - hole_start;
+		map->m_pblk = 0;
+		map->m_len = min_t(unsigned int, map->m_len, hole_len);
+
 		goto out2;
 	}
 
@@ -4482,15 +4497,6 @@ got_allocated_blocks:
 	if (flags & EXT4_GET_BLOCKS_UNWRIT_EXT){
 		ext4_ext_mark_unwritten(&newex);
 		map->m_flags |= EXT4_MAP_UNWRITTEN;
-		/*
-		 * io_end structure was created for every IO write to an
-		 * unwritten extent. To avoid unnecessary conversion,
-		 * here we flag the IO that really needs the conversion.
-		 * For non asycn direct IO case, flag the inode state
-		 * that we need to perform conversion when IO is done.
-		 */
-		if (flags & EXT4_GET_BLOCKS_PRE_IO)
-			set_unwritten = 1;
 	}
 
 	err = 0;
@@ -4501,14 +4507,6 @@ got_allocated_blocks:
 		err = ext4_ext_insert_extent(handle, inode, &path,
 					     &newex, flags);
 
-	if (!err && set_unwritten) {
-		if (io)
-			ext4_set_io_unwritten_flag(inode, io);
-		else
-			ext4_set_inode_state(inode,
-					     EXT4_STATE_DIO_UNWRITTEN);
-	}
-
 	if (err && free_on_err) {
 		int fb_flags = flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE ?
 			EXT4_FREE_BLOCKS_NO_QUOT_UPDATE : 0;

fs/ext4/extents_status.c

@@ -823,8 +823,8 @@ out:
 		es->es_lblk = es1->es_lblk;
 		es->es_len = es1->es_len;
 		es->es_pblk = es1->es_pblk;
-		if (!ext4_es_is_referenced(es))
+		if (!ext4_es_is_referenced(es1))
-			ext4_es_set_referenced(es);
+			ext4_es_set_referenced(es1);
 		stats->es_stats_cache_hits++;
 	} else {
 		stats->es_stats_cache_misses++;

fs/ext4/file.c

@@ -93,31 +93,29 @@ ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 {
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file_inode(iocb->ki_filp);
-	struct mutex *aio_mutex = NULL;
 	struct blk_plug plug;
 	int o_direct = iocb->ki_flags & IOCB_DIRECT;
+	int unaligned_aio = 0;
 	int overwrite = 0;
 	ssize_t ret;
 
-	/*
-	 * Unaligned direct AIO must be serialized; see comment above
-	 * In the case of O_APPEND, assume that we must always serialize
-	 */
-	if (o_direct &&
-	    ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) &&
-	    !is_sync_kiocb(iocb) &&
-	    (iocb->ki_flags & IOCB_APPEND ||
-	     ext4_unaligned_aio(inode, from, iocb->ki_pos))) {
-		aio_mutex = ext4_aio_mutex(inode);
-		mutex_lock(aio_mutex);
-		ext4_unwritten_wait(inode);
-	}
-
 	inode_lock(inode);
 	ret = generic_write_checks(iocb, from);
 	if (ret <= 0)
 		goto out;
 
+	/*
+	 * Unaligned direct AIO must be serialized among each other as zeroing
+	 * of partial blocks of two competing unaligned AIOs can result in data
+	 * corruption.
+	 */
+	if (o_direct && ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS) &&
+	    !is_sync_kiocb(iocb) &&
+	    ext4_unaligned_aio(inode, from, iocb->ki_pos)) {
+		unaligned_aio = 1;
+		ext4_unwritten_wait(inode);
+	}
+
 	/*
 	 * If we have encountered a bitmap-format file, the size limit
 	 * is smaller than s_maxbytes, which is for extent-mapped files.
@@ -139,7 +137,7 @@ ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 		blk_start_plug(&plug);
 
 		/* check whether we do a DIO overwrite or not */
-		if (ext4_should_dioread_nolock(inode) && !aio_mutex &&
+		if (ext4_should_dioread_nolock(inode) && !unaligned_aio &&
 		    !file->f_mapping->nrpages && pos + length <= i_size_read(inode)) {
 			struct ext4_map_blocks map;
 			unsigned int blkbits = inode->i_blkbits;
@@ -181,14 +179,10 @@ ext4_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 	if (o_direct)
 		blk_finish_plug(&plug);
 
-	if (aio_mutex)
-		mutex_unlock(aio_mutex);
 	return ret;
 
 out:
 	inode_unlock(inode);
-	if (aio_mutex)
-		mutex_unlock(aio_mutex);
 	return ret;
 }
 
@@ -417,7 +411,7 @@ static int ext4_file_open(struct inode * inode, struct file * filp)
  */
 static int ext4_find_unwritten_pgoff(struct inode *inode,
 				     int whence,
-				     struct ext4_map_blocks *map,
+				     ext4_lblk_t end_blk,
 				     loff_t *offset)
 {
 	struct pagevec pvec;
@@ -432,7 +426,7 @@ static int ext4_find_unwritten_pgoff(struct inode *inode,
 	blkbits = inode->i_sb->s_blocksize_bits;
 	startoff = *offset;
 	lastoff = startoff;
-	endoff = (loff_t)(map->m_lblk + map->m_len) << blkbits;
+	endoff = (loff_t)end_blk << blkbits;
 
 	index = startoff >> PAGE_CACHE_SHIFT;
 	end = endoff >> PAGE_CACHE_SHIFT;
@@ -550,12 +544,11 @@ out:
 static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
 {
 	struct inode *inode = file->f_mapping->host;
-	struct ext4_map_blocks map;
 	struct extent_status es;
 	ext4_lblk_t start, last, end;
 	loff_t dataoff, isize;
 	int blkbits;
-	int ret = 0;
+	int ret;
 
 	inode_lock(inode);
 
@@ -572,41 +565,32 @@ static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
 	dataoff = offset;
 
 	do {
-		map.m_lblk = last;
+		ret = ext4_get_next_extent(inode, last, end - last + 1, &es);
-		map.m_len = end - last + 1;
+		if (ret <= 0) {
-		ret = ext4_map_blocks(NULL, inode, &map, 0);
+			/* No extent found -> no data */
-		if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) {
+			if (ret == 0)
-			if (last != start)
+				ret = -ENXIO;
-				dataoff = (loff_t)last << blkbits;
+			inode_unlock(inode);
-			break;
+			return ret;
 		}
 
-		/*
+		last = es.es_lblk;
-		 * If there is a delay extent at this offset,
+		if (last != start)
-		 * it will be as a data.
+			dataoff = (loff_t)last << blkbits;
-		 */
+		if (!ext4_es_is_unwritten(&es))
-		ext4_es_find_delayed_extent_range(inode, last, last, &es);
-		if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) {
-			if (last != start)
-				dataoff = (loff_t)last << blkbits;
 			break;
-		}
 
 		/*
 		 * If there is a unwritten extent at this offset,
 		 * it will be as a data or a hole according to page
 		 * cache that has data or not.
 		 */
-		if (map.m_flags & EXT4_MAP_UNWRITTEN) {
+		if (ext4_find_unwritten_pgoff(inode, SEEK_DATA,
-			int unwritten;
+					      es.es_lblk + es.es_len, &dataoff))
-			unwritten = ext4_find_unwritten_pgoff(inode, SEEK_DATA,
+			break;
-							      &map, &dataoff);
+		last += es.es_len;
-			if (unwritten)
-				break;
-		}
-
-		last++;
 		dataoff = (loff_t)last << blkbits;
+		cond_resched();
 	} while (last <= end);
 
 	inode_unlock(inode);
@@ -623,12 +607,11 @@ static loff_t ext4_seek_data(struct file *file, loff_t offset, loff_t maxsize)
 static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize)
 {
 	struct inode *inode = file->f_mapping->host;
-	struct ext4_map_blocks map;
 	struct extent_status es;
 	ext4_lblk_t start, last, end;
 	loff_t holeoff, isize;
 	int blkbits;
-	int ret = 0;
+	int ret;
 
 	inode_lock(inode);
 
@@ -645,44 +628,30 @@ static loff_t ext4_seek_hole(struct file *file, loff_t offset, loff_t maxsize)
 	holeoff = offset;
 
 	do {
-		map.m_lblk = last;
+		ret = ext4_get_next_extent(inode, last, end - last + 1, &es);
-		map.m_len = end - last + 1;
+		if (ret < 0) {
-		ret = ext4_map_blocks(NULL, inode, &map, 0);
+			inode_unlock(inode);
-		if (ret > 0 && !(map.m_flags & EXT4_MAP_UNWRITTEN)) {
+			return ret;
-			last += ret;
-			holeoff = (loff_t)last << blkbits;
-			continue;
 		}
-
+		/* Found a hole? */
-		/*
+		if (ret == 0 || es.es_lblk > last) {
-		 * If there is a delay extent at this offset,
+			if (last != start)
-		 * we will skip this extent.
+				holeoff = (loff_t)last << blkbits;
-		 */
+			break;
-		ext4_es_find_delayed_extent_range(inode, last, last, &es);
-		if (es.es_len != 0 && in_range(last, es.es_lblk, es.es_len)) {
-			last = es.es_lblk + es.es_len;
-			holeoff = (loff_t)last << blkbits;
-			continue;
 		}
-
 		/*
 		 * If there is a unwritten extent at this offset,
 		 * it will be as a data or a hole according to page
 		 * cache that has data or not.
 		 */
-		if (map.m_flags & EXT4_MAP_UNWRITTEN) {
+		if (ext4_es_is_unwritten(&es) &&
-			int unwritten;
+		    ext4_find_unwritten_pgoff(inode, SEEK_HOLE,
-			unwritten = ext4_find_unwritten_pgoff(inode, SEEK_HOLE,
+					      last + es.es_len, &holeoff))
-							      &map, &holeoff);
+			break;
-			if (!unwritten) {
-				last += ret;
-				holeoff = (loff_t)last << blkbits;
-				continue;
-			}
-		}
 
-		/* find a hole */
+		last += es.es_len;
-		break;
+		holeoff = (loff_t)last << blkbits;
+		cond_resched();
 	} while (last <= end);
 
 	inode_unlock(inode);

fs/ext4/ialloc.c

@@ -787,7 +787,7 @@ struct inode *__ext4_new_inode(handle_t *handle, struct inode *dir,
 	sbi = EXT4_SB(sb);
 
 	/*
-	 * Initalize owners and quota early so that we don't have to account
+	 * Initialize owners and quota early so that we don't have to account
 	 * for quota initialization worst case in standard inode creating
 	 * transaction
 	 */

fs/ext4/indirect.c

@@ -555,8 +555,23 @@ int ext4_ind_map_blocks(handle_t *handle, struct inode *inode,
 		goto got_it;
 	}
 
-	/* Next simple case - plain lookup or failed read of indirect block */
+	/* Next simple case - plain lookup failed */
-	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0 || err == -EIO)
+	if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
+		unsigned epb = inode->i_sb->s_blocksize / sizeof(u32);
+		int i;
+
+		/* Count number blocks in a subtree under 'partial' */
+		count = 1;
+		for (i = 0; partial + i != chain + depth - 1; i++)
+			count *= epb;
+		/* Fill in size of a hole we found */
+		map->m_pblk = 0;
+		map->m_len = min_t(unsigned int, map->m_len, count);
+		goto cleanup;
+	}
+
+	/* Failed read of indirect block */
+	if (err == -EIO)
 		goto cleanup;
 
 	/*
@@ -693,21 +708,21 @@ retry:
 		}
 		if (IS_DAX(inode))
 			ret = dax_do_io(iocb, inode, iter, offset,
-					ext4_get_block, NULL, 0);
+					ext4_dio_get_block, NULL, 0);
 		else
 			ret = __blockdev_direct_IO(iocb, inode,
 						   inode->i_sb->s_bdev, iter,
-						   offset, ext4_get_block, NULL,
+						   offset, ext4_dio_get_block,
-						   NULL, 0);
+						   NULL, NULL, 0);
 		inode_dio_end(inode);
 	} else {
 locked:
 		if (IS_DAX(inode))
 			ret = dax_do_io(iocb, inode, iter, offset,
-					ext4_get_block, NULL, DIO_LOCKING);
+					ext4_dio_get_block, NULL, DIO_LOCKING);
 		else
 			ret = blockdev_direct_IO(iocb, inode, iter, offset,
-						 ext4_get_block);
+						 ext4_dio_get_block);
 
 		if (unlikely(iov_iter_rw(iter) == WRITE && ret < 0)) {
 			loff_t isize = i_size_read(inode);

fs/ext4/inline.c

@@ -581,9 +581,10 @@ retry:
 	if (ret)
 		goto out;
 
-	if (ext4_should_dioread_nolock(inode))
+	if (ext4_should_dioread_nolock(inode)) {
-		ret = __block_write_begin(page, from, to, ext4_get_block_write);
+		ret = __block_write_begin(page, from, to,
-	else
+					  ext4_get_block_unwritten);
+	} else
 		ret = __block_write_begin(page, from, to, ext4_get_block);
 
 	if (!ret && ext4_should_journal_data(inode)) {
@@ -1696,7 +1697,6 @@ int ext4_delete_inline_entry(handle_t *handle,
 	if (err)
 		goto out;
 
-	BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
 	err = ext4_mark_inode_dirty(handle, dir);
 	if (unlikely(err))
 		goto out;

fs/ext4/inode.c

@@ -216,7 +216,6 @@ void ext4_evict_inode(struct inode *inode)
 		}
 		truncate_inode_pages_final(&inode->i_data);
 
-		WARN_ON(atomic_read(&EXT4_I(inode)->i_ioend_count));
 		goto no_delete;
 	}
 
@@ -228,8 +227,6 @@ void ext4_evict_inode(struct inode *inode)
 		ext4_begin_ordered_truncate(inode, 0);
 	truncate_inode_pages_final(&inode->i_data);
 
-	WARN_ON(atomic_read(&EXT4_I(inode)->i_ioend_count));
-
 	/*
 	 * Protect us against freezing - iput() caller didn't have to have any
 	 * protection against it
@@ -458,13 +455,13 @@ static void ext4_map_blocks_es_recheck(handle_t *handle,
  * Otherwise, call with ext4_ind_map_blocks() to handle indirect mapping
  * based files
  *
- * On success, it returns the number of blocks being mapped or allocated.
+ * On success, it returns the number of blocks being mapped or allocated.  if
- * if create==0 and the blocks are pre-allocated and unwritten block,
+ * create==0 and the blocks are pre-allocated and unwritten, the resulting @map
- * the result buffer head is unmapped. If the create ==1, it will make sure
+ * is marked as unwritten. If the create == 1, it will mark @map as mapped.
- * the buffer head is mapped.
  *
  * It returns 0 if plain look up failed (blocks have not been allocated), in
- * that case, buffer head is unmapped
+ * that case, @map is returned as unmapped but we still do fill map->m_len to
+ * indicate the length of a hole starting at map->m_lblk.
  *
  * It returns the error in case of allocation failure.
  */
@@ -507,6 +504,11 @@ int ext4_map_blocks(handle_t *handle, struct inode *inode,
 				retval = map->m_len;
 			map->m_len = retval;
 		} else if (ext4_es_is_delayed(&es) || ext4_es_is_hole(&es)) {
+			map->m_pblk = 0;
+			retval = es.es_len - (map->m_lblk - es.es_lblk);
+			if (retval > map->m_len)
+				retval = map->m_len;
+			map->m_len = retval;
 			retval = 0;
 		} else {
 			BUG_ON(1);
@@ -714,16 +716,11 @@ static void ext4_update_bh_state(struct buffer_head *bh, unsigned long flags)
 		 cmpxchg(&bh->b_state, old_state, new_state) != old_state));
 }
 
-/* Maximum number of blocks we map for direct IO at once. */
-#define DIO_MAX_BLOCKS 4096
-
 static int _ext4_get_block(struct inode *inode, sector_t iblock,
 			   struct buffer_head *bh, int flags)
 {
-	handle_t *handle = ext4_journal_current_handle();
 	struct ext4_map_blocks map;
-	int ret = 0, started = 0;
+	int ret = 0;
-	int dio_credits;
 
 	if (ext4_has_inline_data(inode))
 		return -ERANGE;
@@ -731,33 +728,14 @@ static int _ext4_get_block(struct inode *inode, sector_t iblock,
 	map.m_lblk = iblock;
 	map.m_len = bh->b_size >> inode->i_blkbits;
 
-	if (flags && !handle) {
+	ret = ext4_map_blocks(ext4_journal_current_handle(), inode, &map,
-		/* Direct IO write... */
+			      flags);
-		if (map.m_len > DIO_MAX_BLOCKS)
-			map.m_len = DIO_MAX_BLOCKS;
-		dio_credits = ext4_chunk_trans_blocks(inode, map.m_len);
-		handle = ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS,
-					    dio_credits);
-		if (IS_ERR(handle)) {
-			ret = PTR_ERR(handle);
-			return ret;
-		}
-		started = 1;
-	}
-
-	ret = ext4_map_blocks(handle, inode, &map, flags);
 	if (ret > 0) {
-		ext4_io_end_t *io_end = ext4_inode_aio(inode);
-
 		map_bh(bh, inode->i_sb, map.m_pblk);
 		ext4_update_bh_state(bh, map.m_flags);
-		if (io_end && io_end->flag & EXT4_IO_END_UNWRITTEN)
-			set_buffer_defer_completion(bh);
 		bh->b_size = inode->i_sb->s_blocksize * map.m_len;
 		ret = 0;
 	}
-	if (started)
-		ext4_journal_stop(handle);
 	return ret;
 }
 
@@ -768,6 +746,155 @@ int ext4_get_block(struct inode *inode, sector_t iblock,
 			       create ? EXT4_GET_BLOCKS_CREATE : 0);
 }
 
+/*
+ * Get block function used when preparing for buffered write if we require
+ * creating an unwritten extent if blocks haven't been allocated.  The extent
+ * will be converted to written after the IO is complete.
+ */
+int ext4_get_block_unwritten(struct inode *inode, sector_t iblock,
+			     struct buffer_head *bh_result, int create)
+{
+	ext4_debug("ext4_get_block_unwritten: inode %lu, create flag %d\n",
+		   inode->i_ino, create);
+	return _ext4_get_block(inode, iblock, bh_result,
+			       EXT4_GET_BLOCKS_IO_CREATE_EXT);
+}
+
+/* Maximum number of blocks we map for direct IO at once. */
+#define DIO_MAX_BLOCKS 4096
+
+static handle_t *start_dio_trans(struct inode *inode,
+				 struct buffer_head *bh_result)
+{
+	int dio_credits;
+
+	/* Trim mapping request to maximum we can map at once for DIO */
+	if (bh_result->b_size >> inode->i_blkbits > DIO_MAX_BLOCKS)
+		bh_result->b_size = DIO_MAX_BLOCKS << inode->i_blkbits;
+	dio_credits = ext4_chunk_trans_blocks(inode,
+				      bh_result->b_size >> inode->i_blkbits);
+	return ext4_journal_start(inode, EXT4_HT_MAP_BLOCKS, dio_credits);
+}
+
+/* Get block function for DIO reads and writes to inodes without extents */
+int ext4_dio_get_block(struct inode *inode, sector_t iblock,
+		       struct buffer_head *bh, int create)
+{
+	handle_t *handle;
+	int ret;
+
+	/* We don't expect handle for direct IO */
+	WARN_ON_ONCE(ext4_journal_current_handle());
+
+	if (create) {
+		handle = start_dio_trans(inode, bh);
+		if (IS_ERR(handle))
+			return PTR_ERR(handle);
+	}
+	ret = _ext4_get_block(inode, iblock, bh,
+			      create ? EXT4_GET_BLOCKS_CREATE : 0);
+	if (create)
+		ext4_journal_stop(handle);
+	return ret;
+}
+
+/*
+ * Get block function for AIO DIO writes when we create unwritten extent if
+ * blocks are not allocated yet. The extent will be converted to written
+ * after IO is complete.
+ */
+static int ext4_dio_get_block_unwritten_async(struct inode *inode,
+		sector_t iblock, struct buffer_head *bh_result,	int create)
+{
+	handle_t *handle;
+	int ret;
+
+	/* We don't expect handle for direct IO */
+	WARN_ON_ONCE(ext4_journal_current_handle());
+
+	handle = start_dio_trans(inode, bh_result);
+	if (IS_ERR(handle))
+		return PTR_ERR(handle);
+	ret = _ext4_get_block(inode, iblock, bh_result,
+			      EXT4_GET_BLOCKS_IO_CREATE_EXT);
+	ext4_journal_stop(handle);
+
+	/*
+	 * When doing DIO using unwritten extents, we need io_end to convert
+	 * unwritten extents to written on IO completion. We allocate io_end
+	 * once we spot unwritten extent and store it in b_private. Generic
+	 * DIO code keeps b_private set and furthermore passes the value to
+	 * our completion callback in 'private' argument.
+	 */
+	if (!ret && buffer_unwritten(bh_result)) {
+		if (!bh_result->b_private) {
+			ext4_io_end_t *io_end;
+
+			io_end = ext4_init_io_end(inode, GFP_KERNEL);
+			if (!io_end)
+				return -ENOMEM;
+			bh_result->b_private = io_end;
+			ext4_set_io_unwritten_flag(inode, io_end);
+		}
+		set_buffer_defer_completion(bh_result);
+	}
+
+	return ret;
+}
+
+/*
+ * Get block function for non-AIO DIO writes when we create unwritten extent if
+ * blocks are not allocated yet. The extent will be converted to written
+ * after IO is complete from ext4_ext_direct_IO() function.
+ */
+static int ext4_dio_get_block_unwritten_sync(struct inode *inode,
+		sector_t iblock, struct buffer_head *bh_result,	int create)
+{
+	handle_t *handle;
+	int ret;
+
+	/* We don't expect handle for direct IO */
+	WARN_ON_ONCE(ext4_journal_current_handle());
+
+	handle = start_dio_trans(inode, bh_result);
+	if (IS_ERR(handle))
+		return PTR_ERR(handle);
+	ret = _ext4_get_block(inode, iblock, bh_result,
+			      EXT4_GET_BLOCKS_IO_CREATE_EXT);
+	ext4_journal_stop(handle);
+
+	/*
+	 * Mark inode as having pending DIO writes to unwritten extents.
+	 * ext4_ext_direct_IO() checks this flag and converts extents to
+	 * written.
+	 */
+	if (!ret && buffer_unwritten(bh_result))
+		ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
+
+	return ret;
+}
+
+static int ext4_dio_get_block_overwrite(struct inode *inode, sector_t iblock,
+		   struct buffer_head *bh_result, int create)
+{
+	int ret;
+
+	ext4_debug("ext4_dio_get_block_overwrite: inode %lu, create flag %d\n",
+		   inode->i_ino, create);
+	/* We don't expect handle for direct IO */
+	WARN_ON_ONCE(ext4_journal_current_handle());
+
+	ret = _ext4_get_block(inode, iblock, bh_result, 0);
+	/*
+	 * Blocks should have been preallocated! ext4_file_write_iter() checks
+	 * that.
+	 */
+	WARN_ON_ONCE(!buffer_mapped(bh_result) || buffer_unwritten(bh_result));
+
+	return ret;
+}
+
+
 /*
  * `handle' can be NULL if create is zero
  */
@@ -1079,13 +1206,14 @@ retry_journal:
 #ifdef CONFIG_EXT4_FS_ENCRYPTION
 	if (ext4_should_dioread_nolock(inode))
 		ret = ext4_block_write_begin(page, pos, len,
-					     ext4_get_block_write);
+					     ext4_get_block_unwritten);
 	else
 		ret = ext4_block_write_begin(page, pos, len,
 					     ext4_get_block);
 #else
 	if (ext4_should_dioread_nolock(inode))
-		ret = __block_write_begin(page, pos, len, ext4_get_block_write);
+		ret = __block_write_begin(page, pos, len,
+					  ext4_get_block_unwritten);
 	else
 		ret = __block_write_begin(page, pos, len, ext4_get_block);
 #endif
@@ -3088,37 +3216,6 @@ static int ext4_releasepage(struct page *page, gfp_t wait)
 		return try_to_free_buffers(page);
 }
 
-/*
- * ext4_get_block used when preparing for a DIO write or buffer write.
- * We allocate an uinitialized extent if blocks haven't been allocated.
- * The extent will be converted to initialized after the IO is complete.
- */
-int ext4_get_block_write(struct inode *inode, sector_t iblock,
-		   struct buffer_head *bh_result, int create)
-{
-	ext4_debug("ext4_get_block_write: inode %lu, create flag %d\n",
-		   inode->i_ino, create);
-	return _ext4_get_block(inode, iblock, bh_result,
-			       EXT4_GET_BLOCKS_IO_CREATE_EXT);
-}
-
-static int ext4_get_block_overwrite(struct inode *inode, sector_t iblock,
-		   struct buffer_head *bh_result, int create)
-{
-	int ret;
-
-	ext4_debug("ext4_get_block_overwrite: inode %lu, create flag %d\n",
-		   inode->i_ino, create);
-	ret = _ext4_get_block(inode, iblock, bh_result, 0);
-	/*
-	 * Blocks should have been preallocated! ext4_file_write_iter() checks
-	 * that.
-	 */
-	WARN_ON_ONCE(!buffer_mapped(bh_result));
-
-	return ret;
-}
-
 #ifdef CONFIG_FS_DAX
 int ext4_dax_mmap_get_block(struct inode *inode, sector_t iblock,
 			    struct buffer_head *bh_result, int create)
@@ -3179,13 +3276,12 @@ out:
 	WARN_ON_ONCE(ret == 0 && create);
 	if (ret > 0) {
 		map_bh(bh_result, inode->i_sb, map.m_pblk);
-		bh_result->b_state = (bh_result->b_state & ~EXT4_MAP_FLAGS) |
-					map.m_flags;
 		/*
 		 * At least for now we have to clear BH_New so that DAX code
 		 * doesn't attempt to zero blocks again in a racy way.
 		 */
-		bh_result->b_state &= ~(1 << BH_New);
+		map.m_flags &= ~EXT4_MAP_NEW;
+		ext4_update_bh_state(bh_result, map.m_flags);
 		bh_result->b_size = map.m_len << inode->i_blkbits;
 		ret = 0;
 	}
@@ -3196,7 +3292,7 @@ out:
 static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
 			    ssize_t size, void *private)
 {
-        ext4_io_end_t *io_end = iocb->private;
+        ext4_io_end_t *io_end = private;
 
 	/* if not async direct IO just return */
 	if (!io_end)
@@ -3204,10 +3300,8 @@ static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
 
 	ext_debug("ext4_end_io_dio(): io_end 0x%p "
 		  "for inode %lu, iocb 0x%p, offset %llu, size %zd\n",
- 		  iocb->private, io_end->inode->i_ino, iocb, offset,
+		  io_end, io_end->inode->i_ino, iocb, offset, size);
-		  size);
 
-	iocb->private = NULL;
 	io_end->offset = offset;
 	io_end->size = size;
 	ext4_put_io_end(io_end);
@@ -3243,7 +3337,6 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 	get_block_t *get_block_func = NULL;
 	int dio_flags = 0;
 	loff_t final_size = offset + count;
-	ext4_io_end_t *io_end = NULL;
 
 	/* Use the old path for reads and writes beyond i_size. */
 	if (iov_iter_rw(iter) != WRITE || final_size > inode->i_size)
@@ -3268,16 +3361,17 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 	/*
 	 * We could direct write to holes and fallocate.
 	 *
-	 * Allocated blocks to fill the hole are marked as
+	 * Allocated blocks to fill the hole are marked as unwritten to prevent
-	 * unwritten to prevent parallel buffered read to expose
+	 * parallel buffered read to expose the stale data before DIO complete
-	 * the stale data before DIO complete the data IO.
+	 * the data IO.
 	 *
-	 * As to previously fallocated extents, ext4 get_block will
+	 * As to previously fallocated extents, ext4 get_block will just simply
-	 * just simply mark the buffer mapped but still keep the
+	 * mark the buffer mapped but still keep the extents unwritten.
-	 * extents unwritten.
 	 *
-	 * For non AIO case, we will convert those unwritten extents
+	 * For non AIO case, we will convert those unwritten extents to written
-	 * to written after return back from blockdev_direct_IO.
+	 * after return back from blockdev_direct_IO. That way we save us from
+	 * allocating io_end structure and also the overhead of offloading
+	 * the extent convertion to a workqueue.
 	 *
 	 * For async DIO, the conversion needs to be deferred when the
 	 * IO is completed. The ext4 end_io callback function will be
@@ -3285,30 +3379,13 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 	 * case, we allocate an io_end structure to hook to the iocb.
 	 */
 	iocb->private = NULL;
-	if (overwrite) {
+	if (overwrite)
-		get_block_func = ext4_get_block_overwrite;
+		get_block_func = ext4_dio_get_block_overwrite;
+	else if (is_sync_kiocb(iocb)) {
+		get_block_func = ext4_dio_get_block_unwritten_sync;
+		dio_flags = DIO_LOCKING;
 	} else {
-		ext4_inode_aio_set(inode, NULL);
+		get_block_func = ext4_dio_get_block_unwritten_async;
-		if (!is_sync_kiocb(iocb)) {
-			io_end = ext4_init_io_end(inode, GFP_NOFS);
-			if (!io_end) {
-				ret = -ENOMEM;
-				goto retake_lock;
-			}
-			/*
-			 * Grab reference for DIO. Will be dropped in
-			 * ext4_end_io_dio()
-			 */
-			iocb->private = ext4_get_io_end(io_end);
-			/*
-			 * we save the io structure for current async direct
-			 * IO, so that later ext4_map_blocks() could flag the
-			 * io structure whether there is a unwritten extents
-			 * needs to be converted when IO is completed.
-			 */
-			ext4_inode_aio_set(inode, io_end);
-		}
-		get_block_func = ext4_get_block_write;
 		dio_flags = DIO_LOCKING;
 	}
 #ifdef CONFIG_EXT4_FS_ENCRYPTION
@@ -3323,27 +3400,6 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 					   get_block_func,
 					   ext4_end_io_dio, NULL, dio_flags);
 
-	/*
-	 * Put our reference to io_end. This can free the io_end structure e.g.
-	 * in sync IO case or in case of error. It can even perform extent
-	 * conversion if all bios we submitted finished before we got here.
-	 * Note that in that case iocb->private can be already set to NULL
-	 * here.
-	 */
-	if (io_end) {
-		ext4_inode_aio_set(inode, NULL);
-		ext4_put_io_end(io_end);
-		/*
-		 * When no IO was submitted ext4_end_io_dio() was not
-		 * called so we have to put iocb's reference.
-		 */
-		if (ret <= 0 && ret != -EIOCBQUEUED && iocb->private) {
-			WARN_ON(iocb->private != io_end);
-			WARN_ON(io_end->flag & EXT4_IO_END_UNWRITTEN);
-			ext4_put_io_end(io_end);
-			iocb->private = NULL;
-		}
-	}
 	if (ret > 0 && !overwrite && ext4_test_inode_state(inode,
 						EXT4_STATE_DIO_UNWRITTEN)) {
 		int err;
@@ -3358,7 +3414,6 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 		ext4_clear_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
 	}
 
-retake_lock:
 	if (iov_iter_rw(iter) == WRITE)
 		inode_dio_end(inode);
 	/* take i_mutex locking again if we do a ovewrite dio */
@@ -5261,6 +5316,8 @@ int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
 	might_sleep();
 	trace_ext4_mark_inode_dirty(inode, _RET_IP_);
 	err = ext4_reserve_inode_write(handle, inode, &iloc);
+	if (err)
+		return err;
 	if (ext4_handle_valid(handle) &&
 	    EXT4_I(inode)->i_extra_isize < sbi->s_want_extra_isize &&
 	    !ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND)) {
@@ -5291,9 +5348,7 @@ int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
 			}
 		}
 	}
-	if (!err)
+	return ext4_mark_iloc_dirty(handle, inode, &iloc);
-		err = ext4_mark_iloc_dirty(handle, inode, &iloc);
-	return err;
 }
 
 /*
@@ -5502,7 +5557,7 @@ int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	unlock_page(page);
 	/* OK, we need to fill the hole... */
 	if (ext4_should_dioread_nolock(inode))
-		get_block = ext4_get_block_write;
+		get_block = ext4_get_block_unwritten;
 	else
 		get_block = ext4_get_block;
 retry_alloc:
@@ -5545,3 +5600,70 @@ int ext4_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
 
 	return err;
 }
+
+/*
+ * Find the first extent at or after @lblk in an inode that is not a hole.
+ * Search for @map_len blocks at most. The extent is returned in @result.
+ *
+ * The function returns 1 if we found an extent. The function returns 0 in
+ * case there is no extent at or after @lblk and in that case also sets
+ * @result->es_len to 0. In case of error, the error code is returned.
+ */
+int ext4_get_next_extent(struct inode *inode, ext4_lblk_t lblk,
+			 unsigned int map_len, struct extent_status *result)
+{
+	struct ext4_map_blocks map;
+	struct extent_status es = {};
+	int ret;
+
+	map.m_lblk = lblk;
+	map.m_len = map_len;
+
+	/*
+	 * For non-extent based files this loop may iterate several times since
+	 * we do not determine full hole size.
+	 */
+	while (map.m_len > 0) {
+		ret = ext4_map_blocks(NULL, inode, &map, 0);
+		if (ret < 0)
+			return ret;
+		/* There's extent covering m_lblk? Just return it. */
+		if (ret > 0) {
+			int status;
+
+			ext4_es_store_pblock(result, map.m_pblk);
+			result->es_lblk = map.m_lblk;
+			result->es_len = map.m_len;
+			if (map.m_flags & EXT4_MAP_UNWRITTEN)
+				status = EXTENT_STATUS_UNWRITTEN;
+			else
+				status = EXTENT_STATUS_WRITTEN;
+			ext4_es_store_status(result, status);
+			return 1;
+		}
+		ext4_es_find_delayed_extent_range(inode, map.m_lblk,
+						  map.m_lblk + map.m_len - 1,
+						  &es);
+		/* Is delalloc data before next block in extent tree? */
+		if (es.es_len && es.es_lblk < map.m_lblk + map.m_len) {
+			ext4_lblk_t offset = 0;
+
+			if (es.es_lblk < lblk)
+				offset = lblk - es.es_lblk;
+			result->es_lblk = es.es_lblk + offset;
+			ext4_es_store_pblock(result,
+					     ext4_es_pblock(&es) + offset);
+			result->es_len = es.es_len - offset;
+			ext4_es_store_status(result, ext4_es_status(&es));
+
+			return 1;
+		}
+		/* There's a hole at m_lblk, advance us after it */
+		map.m_lblk += map.m_len;
+		map_len -= map.m_len;
+		map.m_len = map_len;
+		cond_resched();
+	}
+	result->es_len = 0;
+	return 0;
+}

fs/ext4/mballoc.c

@@ -11,7 +11,7 @@
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  *
- * You should have received a copy of the GNU General Public Licens
+ * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-
  */
@@ -815,7 +815,7 @@ static void mb_regenerate_buddy(struct ext4_buddy *e4b)
  * for this page; do not hold this lock when calling this routine!
  */
 
-static int ext4_mb_init_cache(struct page *page, char *incore)
+static int ext4_mb_init_cache(struct page *page, char *incore, gfp_t gfp)
 {
 	ext4_group_t ngroups;
 	int blocksize;
@@ -848,7 +848,7 @@ static int ext4_mb_init_cache(struct page *page, char *incore)
 	/* allocate buffer_heads to read bitmaps */
 	if (groups_per_page > 1) {
 		i = sizeof(struct buffer_head *) * groups_per_page;
-		bh = kzalloc(i, GFP_NOFS);
+		bh = kzalloc(i, gfp);
 		if (bh == NULL) {
 			err = -ENOMEM;
 			goto out;
@@ -983,7 +983,7 @@ out:
  * are on the same page e4b->bd_buddy_page is NULL and return value is 0.
  */
 static int ext4_mb_get_buddy_page_lock(struct super_block *sb,
-		ext4_group_t group, struct ext4_buddy *e4b)
+		ext4_group_t group, struct ext4_buddy *e4b, gfp_t gfp)
 {
 	struct inode *inode = EXT4_SB(sb)->s_buddy_cache;
 	int block, pnum, poff;
@@ -1002,7 +1002,7 @@ static int ext4_mb_get_buddy_page_lock(struct super_block *sb,
 	block = group * 2;
 	pnum = block / blocks_per_page;
 	poff = block % blocks_per_page;
-	page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+	page = find_or_create_page(inode->i_mapping, pnum, gfp);
 	if (!page)
 		return -ENOMEM;
 	BUG_ON(page->mapping != inode->i_mapping);
@@ -1016,7 +1016,7 @@ static int ext4_mb_get_buddy_page_lock(struct super_block *sb,
 
 	block++;
 	pnum = block / blocks_per_page;
-	page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+	page = find_or_create_page(inode->i_mapping, pnum, gfp);
 	if (!page)
 		return -ENOMEM;
 	BUG_ON(page->mapping != inode->i_mapping);
@@ -1042,7 +1042,7 @@ static void ext4_mb_put_buddy_page_lock(struct ext4_buddy *e4b)
  * calling this routine!
  */
 static noinline_for_stack
-int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
+int ext4_mb_init_group(struct super_block *sb, ext4_group_t group, gfp_t gfp)
 {
 
 	struct ext4_group_info *this_grp;
@@ -1062,7 +1062,7 @@ int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
 	 * The call to ext4_mb_get_buddy_page_lock will mark the
 	 * page accessed.
 	 */
-	ret = ext4_mb_get_buddy_page_lock(sb, group, &e4b);
+	ret = ext4_mb_get_buddy_page_lock(sb, group, &e4b, gfp);
 	if (ret || !EXT4_MB_GRP_NEED_INIT(this_grp)) {
 		/*
 		 * somebody initialized the group
@@ -1072,7 +1072,7 @@ int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
 	}
 
 	page = e4b.bd_bitmap_page;
-	ret = ext4_mb_init_cache(page, NULL);
+	ret = ext4_mb_init_cache(page, NULL, gfp);
 	if (ret)
 		goto err;
 	if (!PageUptodate(page)) {
@@ -1091,7 +1091,7 @@ int ext4_mb_init_group(struct super_block *sb, ext4_group_t group)
 	}
 	/* init buddy cache */
 	page = e4b.bd_buddy_page;
-	ret = ext4_mb_init_cache(page, e4b.bd_bitmap);
+	ret = ext4_mb_init_cache(page, e4b.bd_bitmap, gfp);
 	if (ret)
 		goto err;
 	if (!PageUptodate(page)) {
@@ -1109,8 +1109,8 @@ err:
  * calling this routine!
  */
 static noinline_for_stack int
-ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
+ext4_mb_load_buddy_gfp(struct super_block *sb, ext4_group_t group,
-					struct ext4_buddy *e4b)
+		       struct ext4_buddy *e4b, gfp_t gfp)
 {
 	int blocks_per_page;
 	int block;
@@ -1140,7 +1140,7 @@ ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
 		 * we need full data about the group
 		 * to make a good selection
 		 */
-		ret = ext4_mb_init_group(sb, group);
+		ret = ext4_mb_init_group(sb, group, gfp);
 		if (ret)
 			return ret;
 	}
@@ -1168,11 +1168,11 @@ ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
 			 * wait for it to initialize.
 			 */
 			page_cache_release(page);
-		page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+		page = find_or_create_page(inode->i_mapping, pnum, gfp);
 		if (page) {
 			BUG_ON(page->mapping != inode->i_mapping);
 			if (!PageUptodate(page)) {
-				ret = ext4_mb_init_cache(page, NULL);
+				ret = ext4_mb_init_cache(page, NULL, gfp);
 				if (ret) {
 					unlock_page(page);
 					goto err;
@@ -1204,11 +1204,12 @@ ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
 	if (page == NULL || !PageUptodate(page)) {
 		if (page)
 			page_cache_release(page);
-		page = find_or_create_page(inode->i_mapping, pnum, GFP_NOFS);
+		page = find_or_create_page(inode->i_mapping, pnum, gfp);
 		if (page) {
 			BUG_ON(page->mapping != inode->i_mapping);
 			if (!PageUptodate(page)) {
-				ret = ext4_mb_init_cache(page, e4b->bd_bitmap);
+				ret = ext4_mb_init_cache(page, e4b->bd_bitmap,
+							 gfp);
 				if (ret) {
 					unlock_page(page);
 					goto err;
@@ -1247,6 +1248,12 @@ err:
 	return ret;
 }
 
+static int ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
+			      struct ext4_buddy *e4b)
+{
+	return ext4_mb_load_buddy_gfp(sb, group, e4b, GFP_NOFS);
+}
+
 static void ext4_mb_unload_buddy(struct ext4_buddy *e4b)
 {
 	if (e4b->bd_bitmap_page)
@@ -2045,7 +2052,7 @@ static int ext4_mb_good_group(struct ext4_allocation_context *ac,
 
 	/* We only do this if the grp has never been initialized */
 	if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) {
-		int ret = ext4_mb_init_group(ac->ac_sb, group);
+		int ret = ext4_mb_init_group(ac->ac_sb, group, GFP_NOFS);
 		if (ret)
 			return ret;
 	}
@@ -4694,16 +4701,6 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
 			    inode, bh, block);
 	}
 
-	/*
-	 * We need to make sure we don't reuse the freed block until
-	 * after the transaction is committed, which we can do by
-	 * treating the block as metadata, below.  We make an
-	 * exception if the inode is to be written in writeback mode
-	 * since writeback mode has weak data consistency guarantees.
-	 */
-	if (!ext4_should_writeback_data(inode))
-		flags |= EXT4_FREE_BLOCKS_METADATA;
-
 	/*
 	 * If the extent to be freed does not begin on a cluster
 	 * boundary, we need to deal with partial clusters at the
@@ -4738,14 +4735,13 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
 
 	if (!bh && (flags & EXT4_FREE_BLOCKS_FORGET)) {
 		int i;
+		int is_metadata = flags & EXT4_FREE_BLOCKS_METADATA;
 
 		for (i = 0; i < count; i++) {
 			cond_resched();
-			bh = sb_find_get_block(inode->i_sb, block + i);
+			if (is_metadata)
-			if (!bh)
+				bh = sb_find_get_block(inode->i_sb, block + i);
-				continue;
+			ext4_forget(handle, is_metadata, inode, bh, block + i);
-			ext4_forget(handle, flags & EXT4_FREE_BLOCKS_METADATA,
-				    inode, bh, block + i);
 		}
 	}
 
@@ -4815,16 +4811,23 @@ do_more:
 #endif
 	trace_ext4_mballoc_free(sb, inode, block_group, bit, count_clusters);
 
-	err = ext4_mb_load_buddy(sb, block_group, &e4b);
+	/* __GFP_NOFAIL: retry infinitely, ignore TIF_MEMDIE and memcg limit. */
+	err = ext4_mb_load_buddy_gfp(sb, block_group, &e4b,
+				     GFP_NOFS|__GFP_NOFAIL);
 	if (err)
 		goto error_return;
 
-	if ((flags & EXT4_FREE_BLOCKS_METADATA) && ext4_handle_valid(handle)) {
+	/*
+	 * We need to make sure we don't reuse the freed block until after the
+	 * transaction is committed. We make an exception if the inode is to be
+	 * written in writeback mode since writeback mode has weak data
+	 * consistency guarantees.
+	 */
+	if (ext4_handle_valid(handle) &&
+	    ((flags & EXT4_FREE_BLOCKS_METADATA) ||
+	     !ext4_should_writeback_data(inode))) {
 		struct ext4_free_data *new_entry;
 		/*
-		 * blocks being freed are metadata. these blocks shouldn't
-		 * be used until this transaction is committed
-		 *
 		 * We use __GFP_NOFAIL because ext4_free_blocks() is not allowed
 		 * to fail.
 		 */
@@ -5217,7 +5220,7 @@ int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
 		grp = ext4_get_group_info(sb, group);
 		/* We only do this if the grp has never been initialized */
 		if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) {
-			ret = ext4_mb_init_group(sb, group);
+			ret = ext4_mb_init_group(sb, group, GFP_NOFS);
 			if (ret)
 				break;
 		}

fs/ext4/mballoc.h

@@ -22,18 +22,6 @@
 #include "ext4_jbd2.h"
 #include "ext4.h"
 
-/*
- * with AGGRESSIVE_CHECK allocator runs consistency checks over
- * structures. these checks slow things down a lot
- */
-#define AGGRESSIVE_CHECK__
-
-/*
- * with DOUBLE_CHECK defined mballoc creates persistent in-core
- * bitmaps, maintains and uses them to check for double allocations
- */
-#define DOUBLE_CHECK__
-
 /*
  */
 #ifdef CONFIG_EXT4_DEBUG

fs/ext4/migrate.c

@@ -361,7 +361,7 @@ static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
 	 * blocks.
 	 *
 	 * While converting to extents we need not
-	 * update the orignal inode i_blocks for extent blocks
+	 * update the original inode i_blocks for extent blocks
 	 * via quota APIs. The quota update happened via tmp_inode already.
 	 */
 	spin_lock(&inode->i_lock);

fs/ext4/mmp.c

@@ -91,21 +91,22 @@ static int read_mmp_block(struct super_block *sb, struct buffer_head **bh,
 	submit_bh(READ_SYNC | REQ_META | REQ_PRIO, *bh);
 	wait_on_buffer(*bh);
 	if (!buffer_uptodate(*bh)) {
-		brelse(*bh);
-		*bh = NULL;
 		ret = -EIO;
 		goto warn_exit;
 	}
-
 	mmp = (struct mmp_struct *)((*bh)->b_data);
-	if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC)
+	if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC) {
 		ret = -EFSCORRUPTED;
-	else if (!ext4_mmp_csum_verify(sb, mmp))
+		goto warn_exit;
+	}
+	if (!ext4_mmp_csum_verify(sb, mmp)) {
 		ret = -EFSBADCRC;
-	else
+		goto warn_exit;
-		return 0;
+	}
-
+	return 0;
 warn_exit:
+	brelse(*bh);
+	*bh = NULL;
 	ext4_warning(sb, "Error %d while reading MMP block %llu",
 		     ret, mmp_block);
 	return ret;
@@ -181,15 +182,13 @@ static int kmmpd(void *data)
 		    EXT4_FEATURE_INCOMPAT_MMP)) {
 			ext4_warning(sb, "kmmpd being stopped since MMP feature"
 				     " has been disabled.");
-			EXT4_SB(sb)->s_mmp_tsk = NULL;
+			goto exit_thread;
-			goto failed;
 		}
 
 		if (sb->s_flags & MS_RDONLY) {
 			ext4_warning(sb, "kmmpd being stopped since filesystem "
 				     "has been remounted as readonly.");
-			EXT4_SB(sb)->s_mmp_tsk = NULL;
+			goto exit_thread;
-			goto failed;
 		}
 
 		diff = jiffies - last_update_time;
@@ -211,9 +210,7 @@ static int kmmpd(void *data)
 			if (retval) {
 				ext4_error(sb, "error reading MMP data: %d",
 					   retval);
-
+				goto exit_thread;
-				EXT4_SB(sb)->s_mmp_tsk = NULL;
-				goto failed;
 			}
 
 			mmp_check = (struct mmp_struct *)(bh_check->b_data);
@@ -225,7 +222,9 @@ static int kmmpd(void *data)
 					     "The filesystem seems to have been"
 					     " multiply mounted.");
 				ext4_error(sb, "abort");
-				goto failed;
+				put_bh(bh_check);
+				retval = -EBUSY;
+				goto exit_thread;
 			}
 			put_bh(bh_check);
 		}
@@ -248,7 +247,8 @@ static int kmmpd(void *data)
 
 	retval = write_mmp_block(sb, bh);
 
-failed:
+exit_thread:
+	EXT4_SB(sb)->s_mmp_tsk = NULL;
 	kfree(data);
 	brelse(bh);
 	return retval;

fs/ext4/page-io.c

@@ -128,9 +128,6 @@ static void ext4_release_io_end(ext4_io_end_t *io_end)
 	BUG_ON(io_end->flag & EXT4_IO_END_UNWRITTEN);
 	WARN_ON(io_end->handle);
 
-	if (atomic_dec_and_test(&EXT4_I(io_end->inode)->i_ioend_count))
-		wake_up_all(ext4_ioend_wq(io_end->inode));
-
 	for (bio = io_end->bio; bio; bio = next_bio) {
 		next_bio = bio->bi_private;
 		ext4_finish_bio(bio);
@@ -265,7 +262,6 @@ ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags)
 {
 	ext4_io_end_t *io = kmem_cache_zalloc(io_end_cachep, flags);
 	if (io) {
-		atomic_inc(&EXT4_I(inode)->i_ioend_count);
 		io->inode = inode;
 		INIT_LIST_HEAD(&io->list);
 		atomic_set(&io->count, 1);

fs/ext4/super.c

@@ -55,7 +55,6 @@
 
 static struct ext4_lazy_init *ext4_li_info;
 static struct mutex ext4_li_mtx;
-static int ext4_mballoc_ready;
 static struct ratelimit_state ext4_mount_msg_ratelimit;
 
 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
@@ -844,7 +843,6 @@ static void ext4_put_super(struct super_block *sb)
 	ext4_release_system_zone(sb);
 	ext4_mb_release(sb);
 	ext4_ext_release(sb);
-	ext4_xattr_put_super(sb);
 
 	if (!(sb->s_flags & MS_RDONLY)) {
 		ext4_clear_feature_journal_needs_recovery(sb);
@@ -944,7 +942,6 @@ static struct inode *ext4_alloc_inode(struct super_block *sb)
 	spin_lock_init(&ei->i_completed_io_lock);
 	ei->i_sync_tid = 0;
 	ei->i_datasync_tid = 0;
-	atomic_set(&ei->i_ioend_count, 0);
 	atomic_set(&ei->i_unwritten, 0);
 	INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
 #ifdef CONFIG_EXT4_FS_ENCRYPTION
@@ -1425,9 +1422,9 @@ static const struct mount_opts {
 	{Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
 	{Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
 	{Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
-	 MOPT_NO_EXT2 | MOPT_SET},
+	 MOPT_NO_EXT2},
 	{Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
-	 MOPT_NO_EXT2 | MOPT_CLEAR},
+	 MOPT_NO_EXT2},
 	{Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
 	{Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
 	{Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
@@ -1705,6 +1702,10 @@ static int handle_mount_opt(struct super_block *sb, char *opt, int token,
 		ext4_msg(sb, KERN_INFO, "dax option not supported");
 		return -1;
 #endif
+	} else if (token == Opt_data_err_abort) {
+		sbi->s_mount_opt |= m->mount_opt;
+	} else if (token == Opt_data_err_ignore) {
+		sbi->s_mount_opt &= ~m->mount_opt;
 	} else {
 		if (!args->from)
 			arg = 1;
@@ -1914,6 +1915,8 @@ static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
 		SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
 	if (nodefs || sbi->s_max_dir_size_kb)
 		SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
+	if (test_opt(sb, DATA_ERR_ABORT))
+		SEQ_OPTS_PUTS("data_err=abort");
 
 	ext4_show_quota_options(seq, sb);
 	return 0;
@@ -3796,12 +3799,10 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
 	sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
 
 no_journal:
-	if (ext4_mballoc_ready) {
+	sbi->s_mb_cache = ext4_xattr_create_cache();
-		sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
+	if (!sbi->s_mb_cache) {
-		if (!sbi->s_mb_cache) {
+		ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
-			ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
+		goto failed_mount_wq;
-			goto failed_mount_wq;
-		}
 	}
 
 	if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
@@ -4027,6 +4028,10 @@ failed_mount4:
 	if (EXT4_SB(sb)->rsv_conversion_wq)
 		destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
 failed_mount_wq:
+	if (sbi->s_mb_cache) {
+		ext4_xattr_destroy_cache(sbi->s_mb_cache);
+		sbi->s_mb_cache = NULL;
+	}
 	if (sbi->s_journal) {
 		jbd2_journal_destroy(sbi->s_journal);
 		sbi->s_journal = NULL;
@@ -5321,7 +5326,6 @@ MODULE_ALIAS_FS("ext4");
 
 /* Shared across all ext4 file systems */
 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
-struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
 
 static int __init ext4_init_fs(void)
 {
@@ -5334,10 +5338,8 @@ static int __init ext4_init_fs(void)
 	/* Build-time check for flags consistency */
 	ext4_check_flag_values();
 
-	for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
+	for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
-		mutex_init(&ext4__aio_mutex[i]);
 		init_waitqueue_head(&ext4__ioend_wq[i]);
-	}
 
 	err = ext4_init_es();
 	if (err)
@@ -5358,8 +5360,6 @@ static int __init ext4_init_fs(void)
 	err = ext4_init_mballoc();
 	if (err)
 		goto out2;
-	else
-		ext4_mballoc_ready = 1;
 	err = init_inodecache();
 	if (err)
 		goto out1;
@@ -5375,7 +5375,6 @@ out:
 	unregister_as_ext3();
 	destroy_inodecache();
 out1:
-	ext4_mballoc_ready = 0;
 	ext4_exit_mballoc();
 out2:
 	ext4_exit_sysfs();

fs/ext4/xattr.c

@@ -545,30 +545,44 @@ static void
 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
 			 struct buffer_head *bh)
 {
-	struct mb_cache_entry *ce = NULL;
-	int error = 0;
 	struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
+	u32 hash, ref;
+	int error = 0;
 
-	ce = mb_cache_entry_get(ext4_mb_cache, bh->b_bdev, bh->b_blocknr);
 	BUFFER_TRACE(bh, "get_write_access");
 	error = ext4_journal_get_write_access(handle, bh);
 	if (error)
 		goto out;
 
 	lock_buffer(bh);
-	if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
+	hash = le32_to_cpu(BHDR(bh)->h_hash);
+	ref = le32_to_cpu(BHDR(bh)->h_refcount);
+	if (ref == 1) {
 		ea_bdebug(bh, "refcount now=0; freeing");
-		if (ce)
+		/*
-			mb_cache_entry_free(ce);
+		 * This must happen under buffer lock for
+		 * ext4_xattr_block_set() to reliably detect freed block
+		 */
+		mb_cache_entry_delete_block(ext4_mb_cache, hash, bh->b_blocknr);
 		get_bh(bh);
 		unlock_buffer(bh);
 		ext4_free_blocks(handle, inode, bh, 0, 1,
 				 EXT4_FREE_BLOCKS_METADATA |
 				 EXT4_FREE_BLOCKS_FORGET);
 	} else {
-		le32_add_cpu(&BHDR(bh)->h_refcount, -1);
+		ref--;
-		if (ce)
+		BHDR(bh)->h_refcount = cpu_to_le32(ref);
-			mb_cache_entry_release(ce);
+		if (ref == EXT4_XATTR_REFCOUNT_MAX - 1) {
+			struct mb_cache_entry *ce;
+
+			ce = mb_cache_entry_get(ext4_mb_cache, hash,
+						bh->b_blocknr);
+			if (ce) {
+				ce->e_reusable = 1;
+				mb_cache_entry_put(ext4_mb_cache, ce);
+			}
+		}
+
 		/*
 		 * Beware of this ugliness: Releasing of xattr block references
 		 * from different inodes can race and so we have to protect
@@ -790,8 +804,6 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 	if (i->value && i->value_len > sb->s_blocksize)
 		return -ENOSPC;
 	if (s->base) {
-		ce = mb_cache_entry_get(ext4_mb_cache, bs->bh->b_bdev,
-					bs->bh->b_blocknr);
 		BUFFER_TRACE(bs->bh, "get_write_access");
 		error = ext4_journal_get_write_access(handle, bs->bh);
 		if (error)
@@ -799,10 +811,15 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 		lock_buffer(bs->bh);
 
 		if (header(s->base)->h_refcount == cpu_to_le32(1)) {
-			if (ce) {
+			__u32 hash = le32_to_cpu(BHDR(bs->bh)->h_hash);
-				mb_cache_entry_free(ce);
+
-				ce = NULL;
+			/*
-			}
+			 * This must happen under buffer lock for
+			 * ext4_xattr_block_set() to reliably detect modified
+			 * block
+			 */
+			mb_cache_entry_delete_block(ext4_mb_cache, hash,
+						    bs->bh->b_blocknr);
 			ea_bdebug(bs->bh, "modifying in-place");
 			error = ext4_xattr_set_entry(i, s);
 			if (!error) {
@@ -826,10 +843,6 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 			int offset = (char *)s->here - bs->bh->b_data;
 
 			unlock_buffer(bs->bh);
-			if (ce) {
-				mb_cache_entry_release(ce);
-				ce = NULL;
-			}
 			ea_bdebug(bs->bh, "cloning");
 			s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
 			error = -ENOMEM;
@@ -872,6 +885,8 @@ inserted:
 			if (new_bh == bs->bh)
 				ea_bdebug(new_bh, "keeping");
 			else {
+				u32 ref;
+
 				/* The old block is released after updating
 				   the inode. */
 				error = dquot_alloc_block(inode,
@@ -884,9 +899,40 @@ inserted:
 				if (error)
 					goto cleanup_dquot;
 				lock_buffer(new_bh);
-				le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
+				/*
+				 * We have to be careful about races with
+				 * freeing, rehashing or adding references to
+				 * xattr block. Once we hold buffer lock xattr
+				 * block's state is stable so we can check
+				 * whether the block got freed / rehashed or
+				 * not.  Since we unhash mbcache entry under
+				 * buffer lock when freeing / rehashing xattr
+				 * block, checking whether entry is still
+				 * hashed is reliable. Same rules hold for
+				 * e_reusable handling.
+				 */
+				if (hlist_bl_unhashed(&ce->e_hash_list) ||
+				    !ce->e_reusable) {
+					/*
+					 * Undo everything and check mbcache
+					 * again.
+					 */
+					unlock_buffer(new_bh);
+					dquot_free_block(inode,
+							 EXT4_C2B(EXT4_SB(sb),
+								  1));
+					brelse(new_bh);
+					mb_cache_entry_put(ext4_mb_cache, ce);
+					ce = NULL;
+					new_bh = NULL;
+					goto inserted;
+				}
+				ref = le32_to_cpu(BHDR(new_bh)->h_refcount) + 1;
+				BHDR(new_bh)->h_refcount = cpu_to_le32(ref);
+				if (ref >= EXT4_XATTR_REFCOUNT_MAX)
+					ce->e_reusable = 0;
 				ea_bdebug(new_bh, "reusing; refcount now=%d",
-					le32_to_cpu(BHDR(new_bh)->h_refcount));
+					  ref);
 				unlock_buffer(new_bh);
 				error = ext4_handle_dirty_xattr_block(handle,
 								      inode,
@@ -894,7 +940,8 @@ inserted:
 				if (error)
 					goto cleanup_dquot;
 			}
-			mb_cache_entry_release(ce);
+			mb_cache_entry_touch(ext4_mb_cache, ce);
+			mb_cache_entry_put(ext4_mb_cache, ce);
 			ce = NULL;
 		} else if (bs->bh && s->base == bs->bh->b_data) {
 			/* We were modifying this block in-place. */
@@ -959,7 +1006,7 @@ getblk_failed:
 
 cleanup:
 	if (ce)
-		mb_cache_entry_release(ce);
+		mb_cache_entry_put(ext4_mb_cache, ce);
 	brelse(new_bh);
 	if (!(bs->bh && s->base == bs->bh->b_data))
 		kfree(s->base);
@@ -1070,6 +1117,17 @@ static int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
 	return 0;
 }
 
+static int ext4_xattr_value_same(struct ext4_xattr_search *s,
+				 struct ext4_xattr_info *i)
+{
+	void *value;
+
+	if (le32_to_cpu(s->here->e_value_size) != i->value_len)
+		return 0;
+	value = ((void *)s->base) + le16_to_cpu(s->here->e_value_offs);
+	return !memcmp(value, i->value, i->value_len);
+}
+
 /*
  * ext4_xattr_set_handle()
  *
@@ -1146,6 +1204,13 @@ ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
 		else if (!bs.s.not_found)
 			error = ext4_xattr_block_set(handle, inode, &i, &bs);
 	} else {
+		error = 0;
+		/* Xattr value did not change? Save us some work and bail out */
+		if (!is.s.not_found && ext4_xattr_value_same(&is.s, &i))
+			goto cleanup;
+		if (!bs.s.not_found && ext4_xattr_value_same(&bs.s, &i))
+			goto cleanup;
+
 		error = ext4_xattr_ibody_set(handle, inode, &i, &is);
 		if (!error && !bs.s.not_found) {
 			i.value = NULL;
@@ -1511,17 +1576,6 @@ cleanup:
 	brelse(bh);
 }
 
-/*
- * ext4_xattr_put_super()
- *
- * This is called when a file system is unmounted.
- */
-void
-ext4_xattr_put_super(struct super_block *sb)
-{
-	mb_cache_shrink(sb->s_bdev);
-}
-
 /*
  * ext4_xattr_cache_insert()
  *
@@ -1533,26 +1587,19 @@ ext4_xattr_put_super(struct super_block *sb)
 static void
 ext4_xattr_cache_insert(struct mb_cache *ext4_mb_cache, struct buffer_head *bh)
 {
-	__u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
+	struct ext4_xattr_header *header = BHDR(bh);
-	struct mb_cache_entry *ce;
+	__u32 hash = le32_to_cpu(header->h_hash);
+	int reusable = le32_to_cpu(header->h_refcount) <
+		       EXT4_XATTR_REFCOUNT_MAX;
 	int error;
 
-	ce = mb_cache_entry_alloc(ext4_mb_cache, GFP_NOFS);
+	error = mb_cache_entry_create(ext4_mb_cache, GFP_NOFS, hash,
-	if (!ce) {
+				      bh->b_blocknr, reusable);
-		ea_bdebug(bh, "out of memory");
-		return;
-	}
-	error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
 	if (error) {
-		mb_cache_entry_free(ce);
+		if (error == -EBUSY)
-		if (error == -EBUSY) {
 			ea_bdebug(bh, "already in cache");
-			error = 0;
+	} else
-		}
-	} else {
 		ea_bdebug(bh, "inserting [%x]", (int)hash);
-		mb_cache_entry_release(ce);
-	}
 }
 
 /*
@@ -1614,33 +1661,20 @@ ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
 	if (!header->h_hash)
 		return NULL;  /* never share */
 	ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
-again:
+	ce = mb_cache_entry_find_first(ext4_mb_cache, hash);
-	ce = mb_cache_entry_find_first(ext4_mb_cache, inode->i_sb->s_bdev,
-				       hash);
 	while (ce) {
 		struct buffer_head *bh;
 
-		if (IS_ERR(ce)) {
-			if (PTR_ERR(ce) == -EAGAIN)
-				goto again;
-			break;
-		}
 		bh = sb_bread(inode->i_sb, ce->e_block);
 		if (!bh) {
 			EXT4_ERROR_INODE(inode, "block %lu read error",
 					 (unsigned long) ce->e_block);
-		} else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
-				EXT4_XATTR_REFCOUNT_MAX) {
-			ea_idebug(inode, "block %lu refcount %d>=%d",
-				  (unsigned long) ce->e_block,
-				  le32_to_cpu(BHDR(bh)->h_refcount),
-					  EXT4_XATTR_REFCOUNT_MAX);
 		} else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
 			*pce = ce;
 			return bh;
 		}
 		brelse(bh);
-		ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
+		ce = mb_cache_entry_find_next(ext4_mb_cache, ce);
 	}
 	return NULL;
 }
@@ -1716,9 +1750,9 @@ static void ext4_xattr_rehash(struct ext4_xattr_header *header,
 #define	HASH_BUCKET_BITS	10
 
 struct mb_cache *
-ext4_xattr_create_cache(char *name)
+ext4_xattr_create_cache(void)
 {
-	return mb_cache_create(name, HASH_BUCKET_BITS);
+	return mb_cache_create(HASH_BUCKET_BITS);
 }
 
 void ext4_xattr_destroy_cache(struct mb_cache *cache)

fs/ext4/xattr.h

@@ -108,7 +108,6 @@ extern int ext4_xattr_set(struct inode *, int, const char *, const void *, size_
 extern int ext4_xattr_set_handle(handle_t *, struct inode *, int, const char *, const void *, size_t, int);
 
 extern void ext4_xattr_delete_inode(handle_t *, struct inode *);
-extern void ext4_xattr_put_super(struct super_block *);
 
 extern int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
 			    struct ext4_inode *raw_inode, handle_t *handle);
@@ -124,7 +123,7 @@ extern int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
 				       struct ext4_xattr_info *i,
 				       struct ext4_xattr_ibody_find *is);
 
-extern struct mb_cache *ext4_xattr_create_cache(char *name);
+extern struct mb_cache *ext4_xattr_create_cache(void);
 extern void ext4_xattr_destroy_cache(struct mb_cache *);
 
 #ifdef CONFIG_EXT4_FS_SECURITY

fs/jbd2/commit.c

@@ -131,14 +131,12 @@ static int journal_submit_commit_record(journal_t *journal,
 	if (is_journal_aborted(journal))
 		return 0;
 
-	bh = jbd2_journal_get_descriptor_buffer(journal);
+	bh = jbd2_journal_get_descriptor_buffer(commit_transaction,
+						JBD2_COMMIT_BLOCK);
 	if (!bh)
 		return 1;
 
 	tmp = (struct commit_header *)bh->b_data;
-	tmp->h_magic = cpu_to_be32(JBD2_MAGIC_NUMBER);
-	tmp->h_blocktype = cpu_to_be32(JBD2_COMMIT_BLOCK);
-	tmp->h_sequence = cpu_to_be32(commit_transaction->t_tid);
 	tmp->h_commit_sec = cpu_to_be64(now.tv_sec);
 	tmp->h_commit_nsec = cpu_to_be32(now.tv_nsec);
 
@@ -222,7 +220,7 @@ static int journal_submit_data_buffers(journal_t *journal,
 	spin_lock(&journal->j_list_lock);
 	list_for_each_entry(jinode, &commit_transaction->t_inode_list, i_list) {
 		mapping = jinode->i_vfs_inode->i_mapping;
-		set_bit(__JI_COMMIT_RUNNING, &jinode->i_flags);
+		jinode->i_flags |= JI_COMMIT_RUNNING;
 		spin_unlock(&journal->j_list_lock);
 		/*
 		 * submit the inode data buffers. We use writepage
@@ -236,8 +234,8 @@ static int journal_submit_data_buffers(journal_t *journal,
 			ret = err;
 		spin_lock(&journal->j_list_lock);
 		J_ASSERT(jinode->i_transaction == commit_transaction);
-		clear_bit(__JI_COMMIT_RUNNING, &jinode->i_flags);
+		jinode->i_flags &= ~JI_COMMIT_RUNNING;
-		smp_mb__after_atomic();
+		smp_mb();
 		wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
 	}
 	spin_unlock(&journal->j_list_lock);
@@ -258,7 +256,7 @@ static int journal_finish_inode_data_buffers(journal_t *journal,
 	/* For locking, see the comment in journal_submit_data_buffers() */
 	spin_lock(&journal->j_list_lock);
 	list_for_each_entry(jinode, &commit_transaction->t_inode_list, i_list) {
-		set_bit(__JI_COMMIT_RUNNING, &jinode->i_flags);
+		jinode->i_flags |= JI_COMMIT_RUNNING;
 		spin_unlock(&journal->j_list_lock);
 		err = filemap_fdatawait(jinode->i_vfs_inode->i_mapping);
 		if (err) {
@@ -274,8 +272,8 @@ static int journal_finish_inode_data_buffers(journal_t *journal,
 				ret = err;
 		}
 		spin_lock(&journal->j_list_lock);
-		clear_bit(__JI_COMMIT_RUNNING, &jinode->i_flags);
+		jinode->i_flags &= ~JI_COMMIT_RUNNING;
-		smp_mb__after_atomic();
+		smp_mb();
 		wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
 	}
 
@@ -319,22 +317,6 @@ static void write_tag_block(journal_t *j, journal_block_tag_t *tag,
 		tag->t_blocknr_high = cpu_to_be32((block >> 31) >> 1);
 }
 
-static void jbd2_descr_block_csum_set(journal_t *j,
-				      struct buffer_head *bh)
-{
-	struct jbd2_journal_block_tail *tail;
-	__u32 csum;
-
-	if (!jbd2_journal_has_csum_v2or3(j))
-		return;
-
-	tail = (struct jbd2_journal_block_tail *)(bh->b_data + j->j_blocksize -
-			sizeof(struct jbd2_journal_block_tail));
-	tail->t_checksum = 0;
-	csum = jbd2_chksum(j, j->j_csum_seed, bh->b_data, j->j_blocksize);
-	tail->t_checksum = cpu_to_be32(csum);
-}
-
 static void jbd2_block_tag_csum_set(journal_t *j, journal_block_tag_t *tag,
 				    struct buffer_head *bh, __u32 sequence)
 {
@@ -379,7 +361,6 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 	ktime_t start_time;
 	u64 commit_time;
 	char *tagp = NULL;
-	journal_header_t *header;
 	journal_block_tag_t *tag = NULL;
 	int space_left = 0;
 	int first_tag = 0;
@@ -554,8 +535,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 		jbd2_journal_abort(journal, err);
 
 	blk_start_plug(&plug);
-	jbd2_journal_write_revoke_records(journal, commit_transaction,
+	jbd2_journal_write_revoke_records(commit_transaction, &log_bufs);
-					  &log_bufs, WRITE_SYNC);
 
 	jbd_debug(3, "JBD2: commit phase 2b\n");
 
@@ -616,7 +596,9 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 
 			jbd_debug(4, "JBD2: get descriptor\n");
 
-			descriptor = jbd2_journal_get_descriptor_buffer(journal);
+			descriptor = jbd2_journal_get_descriptor_buffer(
+							commit_transaction,
+							JBD2_DESCRIPTOR_BLOCK);
 			if (!descriptor) {
 				jbd2_journal_abort(journal, -EIO);
 				continue;
@@ -625,11 +607,6 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 			jbd_debug(4, "JBD2: got buffer %llu (%p)\n",
 				(unsigned long long)descriptor->b_blocknr,
 				descriptor->b_data);
-			header = (journal_header_t *)descriptor->b_data;
-			header->h_magic     = cpu_to_be32(JBD2_MAGIC_NUMBER);
-			header->h_blocktype = cpu_to_be32(JBD2_DESCRIPTOR_BLOCK);
-			header->h_sequence  = cpu_to_be32(commit_transaction->t_tid);
-
 			tagp = &descriptor->b_data[sizeof(journal_header_t)];
 			space_left = descriptor->b_size -
 						sizeof(journal_header_t);
@@ -721,7 +698,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 
 			tag->t_flags |= cpu_to_be16(JBD2_FLAG_LAST_TAG);
 
-			jbd2_descr_block_csum_set(journal, descriptor);
+			jbd2_descriptor_block_csum_set(journal, descriptor);
 start_journal_io:
 			for (i = 0; i < bufs; i++) {
 				struct buffer_head *bh = wbuf[i];

fs/jbd2/journal.c

@@ -805,10 +805,13 @@ int jbd2_journal_bmap(journal_t *journal, unsigned long blocknr,
  * But we don't bother doing that, so there will be coherency problems with
  * mmaps of blockdevs which hold live JBD-controlled filesystems.
  */
-struct buffer_head *jbd2_journal_get_descriptor_buffer(journal_t *journal)
+struct buffer_head *
+jbd2_journal_get_descriptor_buffer(transaction_t *transaction, int type)
 {
+	journal_t *journal = transaction->t_journal;
 	struct buffer_head *bh;
 	unsigned long long blocknr;
+	journal_header_t *header;
 	int err;
 
 	err = jbd2_journal_next_log_block(journal, &blocknr);
@@ -821,12 +824,31 @@ struct buffer_head *jbd2_journal_get_descriptor_buffer(journal_t *journal)
 		return NULL;
 	lock_buffer(bh);
 	memset(bh->b_data, 0, journal->j_blocksize);
+	header = (journal_header_t *)bh->b_data;
+	header->h_magic = cpu_to_be32(JBD2_MAGIC_NUMBER);
+	header->h_blocktype = cpu_to_be32(type);
+	header->h_sequence = cpu_to_be32(transaction->t_tid);
 	set_buffer_uptodate(bh);
 	unlock_buffer(bh);
 	BUFFER_TRACE(bh, "return this buffer");
 	return bh;
 }
 
+void jbd2_descriptor_block_csum_set(journal_t *j, struct buffer_head *bh)
+{
+	struct jbd2_journal_block_tail *tail;
+	__u32 csum;
+
+	if (!jbd2_journal_has_csum_v2or3(j))
+		return;
+
+	tail = (struct jbd2_journal_block_tail *)(bh->b_data + j->j_blocksize -
+			sizeof(struct jbd2_journal_block_tail));
+	tail->t_checksum = 0;
+	csum = jbd2_chksum(j, j->j_csum_seed, bh->b_data, j->j_blocksize);
+	tail->t_checksum = cpu_to_be32(csum);
+}
+
 /*
  * Return tid of the oldest transaction in the journal and block in the journal
  * where the transaction starts.
@@ -1408,11 +1430,12 @@ out:
 /**
  * jbd2_mark_journal_empty() - Mark on disk journal as empty.
  * @journal: The journal to update.
+ * @write_op: With which operation should we write the journal sb
  *
  * Update a journal's dynamic superblock fields to show that journal is empty.
  * Write updated superblock to disk waiting for IO to complete.
  */
-static void jbd2_mark_journal_empty(journal_t *journal)
+static void jbd2_mark_journal_empty(journal_t *journal, int write_op)
 {
 	journal_superblock_t *sb = journal->j_superblock;
 
@@ -1430,7 +1453,7 @@ static void jbd2_mark_journal_empty(journal_t *journal)
 	sb->s_start    = cpu_to_be32(0);
 	read_unlock(&journal->j_state_lock);
 
-	jbd2_write_superblock(journal, WRITE_FUA);
+	jbd2_write_superblock(journal, write_op);
 
 	/* Log is no longer empty */
 	write_lock(&journal->j_state_lock);
@@ -1716,7 +1739,13 @@ int jbd2_journal_destroy(journal_t *journal)
 	if (journal->j_sb_buffer) {
 		if (!is_journal_aborted(journal)) {
 			mutex_lock(&journal->j_checkpoint_mutex);
-			jbd2_mark_journal_empty(journal);
+
+			write_lock(&journal->j_state_lock);
+			journal->j_tail_sequence =
+				++journal->j_transaction_sequence;
+			write_unlock(&journal->j_state_lock);
+
+			jbd2_mark_journal_empty(journal, WRITE_FLUSH_FUA);
 			mutex_unlock(&journal->j_checkpoint_mutex);
 		} else
 			err = -EIO;
@@ -1975,7 +2004,7 @@ int jbd2_journal_flush(journal_t *journal)
 	 * the magic code for a fully-recovered superblock.  Any future
 	 * commits of data to the journal will restore the current
 	 * s_start value. */
-	jbd2_mark_journal_empty(journal);
+	jbd2_mark_journal_empty(journal, WRITE_FUA);
 	mutex_unlock(&journal->j_checkpoint_mutex);
 	write_lock(&journal->j_state_lock);
 	J_ASSERT(!journal->j_running_transaction);
@@ -2021,7 +2050,7 @@ int jbd2_journal_wipe(journal_t *journal, int write)
 	if (write) {
 		/* Lock to make assertions happy... */
 		mutex_lock(&journal->j_checkpoint_mutex);
-		jbd2_mark_journal_empty(journal);
+		jbd2_mark_journal_empty(journal, WRITE_FUA);
 		mutex_unlock(&journal->j_checkpoint_mutex);
 	}
 
@@ -2565,7 +2594,7 @@ void jbd2_journal_release_jbd_inode(journal_t *journal,
 restart:
 	spin_lock(&journal->j_list_lock);
 	/* Is commit writing out inode - we have to wait */
-	if (test_bit(__JI_COMMIT_RUNNING, &jinode->i_flags)) {
+	if (jinode->i_flags & JI_COMMIT_RUNNING) {
 		wait_queue_head_t *wq;
 		DEFINE_WAIT_BIT(wait, &jinode->i_flags, __JI_COMMIT_RUNNING);
 		wq = bit_waitqueue(&jinode->i_flags, __JI_COMMIT_RUNNING);

fs/jbd2/recovery.c

@@ -174,8 +174,7 @@ static int jread(struct buffer_head **bhp, journal_t *journal,
 	return 0;
 }
 
-static int jbd2_descr_block_csum_verify(journal_t *j,
+static int jbd2_descriptor_block_csum_verify(journal_t *j, void *buf)
-					void *buf)
 {
 	struct jbd2_journal_block_tail *tail;
 	__be32 provided;
@@ -522,8 +521,8 @@ static int do_one_pass(journal_t *journal,
 				descr_csum_size =
 					sizeof(struct jbd2_journal_block_tail);
 			if (descr_csum_size > 0 &&
-			    !jbd2_descr_block_csum_verify(journal,
+			    !jbd2_descriptor_block_csum_verify(journal,
-							  bh->b_data)) {
+							       bh->b_data)) {
 				printk(KERN_ERR "JBD2: Invalid checksum "
 				       "recovering block %lu in log\n",
 				       next_log_block);
@@ -811,26 +810,6 @@ static int do_one_pass(journal_t *journal,
 	return err;
 }
 
-static int jbd2_revoke_block_csum_verify(journal_t *j,
-					 void *buf)
-{
-	struct jbd2_journal_revoke_tail *tail;
-	__be32 provided;
-	__u32 calculated;
-
-	if (!jbd2_journal_has_csum_v2or3(j))
-		return 1;
-
-	tail = (struct jbd2_journal_revoke_tail *)(buf + j->j_blocksize -
-			sizeof(struct jbd2_journal_revoke_tail));
-	provided = tail->r_checksum;
-	tail->r_checksum = 0;
-	calculated = jbd2_chksum(j, j->j_csum_seed, buf, j->j_blocksize);
-	tail->r_checksum = provided;
-
-	return provided == cpu_to_be32(calculated);
-}
-
 /* Scan a revoke record, marking all blocks mentioned as revoked. */
 
 static int scan_revoke_records(journal_t *journal, struct buffer_head *bh,
@@ -846,11 +825,11 @@ static int scan_revoke_records(journal_t *journal, struct buffer_head *bh,
 	offset = sizeof(jbd2_journal_revoke_header_t);
 	rcount = be32_to_cpu(header->r_count);
 
-	if (!jbd2_revoke_block_csum_verify(journal, header))
+	if (!jbd2_descriptor_block_csum_verify(journal, header))
 		return -EFSBADCRC;
 
 	if (jbd2_journal_has_csum_v2or3(journal))
-		csum_size = sizeof(struct jbd2_journal_revoke_tail);
+		csum_size = sizeof(struct jbd2_journal_block_tail);
 	if (rcount > journal->j_blocksize - csum_size)
 		return -EINVAL;
 	max = rcount;

fs/jbd2/revoke.c

@@ -122,11 +122,11 @@ struct jbd2_revoke_table_s
 
 
 #ifdef __KERNEL__
-static void write_one_revoke_record(journal_t *, transaction_t *,
+static void write_one_revoke_record(transaction_t *,
 				    struct list_head *,
 				    struct buffer_head **, int *,
-				    struct jbd2_revoke_record_s *, int);
+				    struct jbd2_revoke_record_s *);
-static void flush_descriptor(journal_t *, struct buffer_head *, int, int);
+static void flush_descriptor(journal_t *, struct buffer_head *, int);
 #endif
 
 /* Utility functions to maintain the revoke table */
@@ -519,11 +519,10 @@ void jbd2_journal_switch_revoke_table(journal_t *journal)
  * Write revoke records to the journal for all entries in the current
  * revoke hash, deleting the entries as we go.
  */
-void jbd2_journal_write_revoke_records(journal_t *journal,
+void jbd2_journal_write_revoke_records(transaction_t *transaction,
-				       transaction_t *transaction,
+				       struct list_head *log_bufs)
-				       struct list_head *log_bufs,
-				       int write_op)
 {
+	journal_t *journal = transaction->t_journal;
 	struct buffer_head *descriptor;
 	struct jbd2_revoke_record_s *record;
 	struct jbd2_revoke_table_s *revoke;
@@ -544,16 +543,15 @@ void jbd2_journal_write_revoke_records(journal_t *journal,
 		while (!list_empty(hash_list)) {
 			record = (struct jbd2_revoke_record_s *)
 				hash_list->next;
-			write_one_revoke_record(journal, transaction, log_bufs,
+			write_one_revoke_record(transaction, log_bufs,
-						&descriptor, &offset,
+						&descriptor, &offset, record);
-						record, write_op);
 			count++;
 			list_del(&record->hash);
 			kmem_cache_free(jbd2_revoke_record_cache, record);
 		}
 	}
 	if (descriptor)
-		flush_descriptor(journal, descriptor, offset, write_op);
+		flush_descriptor(journal, descriptor, offset);
 	jbd_debug(1, "Wrote %d revoke records\n", count);
 }
 
@@ -562,18 +560,16 @@ void jbd2_journal_write_revoke_records(journal_t *journal,
  * block if the old one is full or if we have not already created one.
  */
 
-static void write_one_revoke_record(journal_t *journal,
+static void write_one_revoke_record(transaction_t *transaction,
-				    transaction_t *transaction,
 				    struct list_head *log_bufs,
 				    struct buffer_head **descriptorp,
 				    int *offsetp,
-				    struct jbd2_revoke_record_s *record,
+				    struct jbd2_revoke_record_s *record)
-				    int write_op)
 {
+	journal_t *journal = transaction->t_journal;
 	int csum_size = 0;
 	struct buffer_head *descriptor;
 	int sz, offset;
-	journal_header_t *header;
 
 	/* If we are already aborting, this all becomes a noop.  We
            still need to go round the loop in
@@ -587,7 +583,7 @@ static void write_one_revoke_record(journal_t *journal,
 
 	/* Do we need to leave space at the end for a checksum? */
 	if (jbd2_journal_has_csum_v2or3(journal))
-		csum_size = sizeof(struct jbd2_journal_revoke_tail);
+		csum_size = sizeof(struct jbd2_journal_block_tail);
 
 	if (jbd2_has_feature_64bit(journal))
 		sz = 8;
@@ -597,19 +593,16 @@ static void write_one_revoke_record(journal_t *journal,
 	/* Make sure we have a descriptor with space left for the record */
 	if (descriptor) {
 		if (offset + sz > journal->j_blocksize - csum_size) {
-			flush_descriptor(journal, descriptor, offset, write_op);
+			flush_descriptor(journal, descriptor, offset);
 			descriptor = NULL;
 		}
 	}
 
 	if (!descriptor) {
-		descriptor = jbd2_journal_get_descriptor_buffer(journal);
+		descriptor = jbd2_journal_get_descriptor_buffer(transaction,
+							JBD2_REVOKE_BLOCK);
 		if (!descriptor)
 			return;
-		header = (journal_header_t *)descriptor->b_data;
-		header->h_magic     = cpu_to_be32(JBD2_MAGIC_NUMBER);
-		header->h_blocktype = cpu_to_be32(JBD2_REVOKE_BLOCK);
-		header->h_sequence  = cpu_to_be32(transaction->t_tid);
 
 		/* Record it so that we can wait for IO completion later */
 		BUFFER_TRACE(descriptor, "file in log_bufs");
@@ -630,21 +623,6 @@ static void write_one_revoke_record(journal_t *journal,
 	*offsetp = offset;
 }
 
-static void jbd2_revoke_csum_set(journal_t *j, struct buffer_head *bh)
-{
-	struct jbd2_journal_revoke_tail *tail;
-	__u32 csum;
-
-	if (!jbd2_journal_has_csum_v2or3(j))
-		return;
-
-	tail = (struct jbd2_journal_revoke_tail *)(bh->b_data + j->j_blocksize -
-			sizeof(struct jbd2_journal_revoke_tail));
-	tail->r_checksum = 0;
-	csum = jbd2_chksum(j, j->j_csum_seed, bh->b_data, j->j_blocksize);
-	tail->r_checksum = cpu_to_be32(csum);
-}
-
 /*
  * Flush a revoke descriptor out to the journal.  If we are aborting,
  * this is a noop; otherwise we are generating a buffer which needs to
@@ -654,7 +632,7 @@ static void jbd2_revoke_csum_set(journal_t *j, struct buffer_head *bh)
 
 static void flush_descriptor(journal_t *journal,
 			     struct buffer_head *descriptor,
-			     int offset, int write_op)
+			     int offset)
 {
 	jbd2_journal_revoke_header_t *header;
 
@@ -665,12 +643,12 @@ static void flush_descriptor(journal_t *journal,
 
 	header = (jbd2_journal_revoke_header_t *)descriptor->b_data;
 	header->r_count = cpu_to_be32(offset);
-	jbd2_revoke_csum_set(journal, descriptor);
+	jbd2_descriptor_block_csum_set(journal, descriptor);
 
 	set_buffer_jwrite(descriptor);
 	BUFFER_TRACE(descriptor, "write");
 	set_buffer_dirty(descriptor);
-	write_dirty_buffer(descriptor, write_op);
+	write_dirty_buffer(descriptor, WRITE_SYNC);
 }
 #endif
 

fs/jbd2/transaction.c

@@ -966,14 +966,8 @@ repeat:
 		if (!frozen_buffer) {
 			JBUFFER_TRACE(jh, "allocate memory for buffer");
 			jbd_unlock_bh_state(bh);
-			frozen_buffer = jbd2_alloc(jh2bh(jh)->b_size, GFP_NOFS);
+			frozen_buffer = jbd2_alloc(jh2bh(jh)->b_size,
-			if (!frozen_buffer) {
+						   GFP_NOFS | __GFP_NOFAIL);
-				printk(KERN_ERR "%s: OOM for frozen_buffer\n",
-				       __func__);
-				JBUFFER_TRACE(jh, "oom!");
-				error = -ENOMEM;
-				goto out;
-			}
 			goto repeat;
 		}
 		jh->b_frozen_data = frozen_buffer;
@@ -1226,15 +1220,9 @@ int jbd2_journal_get_undo_access(handle_t *handle, struct buffer_head *bh)
 		goto out;
 
 repeat:
-	if (!jh->b_committed_data) {
+	if (!jh->b_committed_data)
-		committed_data = jbd2_alloc(jh2bh(jh)->b_size, GFP_NOFS);
+		committed_data = jbd2_alloc(jh2bh(jh)->b_size,
-		if (!committed_data) {
+					    GFP_NOFS|__GFP_NOFAIL);
-			printk(KERN_ERR "%s: No memory for committed data\n",
-				__func__);
-			err = -ENOMEM;
-			goto out;
-		}
-	}
 
 	jbd_lock_bh_state(bh);
 	if (!jh->b_committed_data) {

include/linux/jbd2.h

@@ -200,7 +200,7 @@ typedef struct journal_block_tag_s
 	__be32		t_blocknr_high; /* most-significant high 32bits. */
 } journal_block_tag_t;
 
-/* Tail of descriptor block, for checksumming */
+/* Tail of descriptor or revoke block, for checksumming */
 struct jbd2_journal_block_tail {
 	__be32		t_checksum;	/* crc32c(uuid+descr_block) */
 };
@@ -215,11 +215,6 @@ typedef struct jbd2_journal_revoke_header_s
 	__be32		 r_count;	/* Count of bytes used in the block */
 } jbd2_journal_revoke_header_t;
 
-/* Tail of revoke block, for checksumming */
-struct jbd2_journal_revoke_tail {
-	__be32		r_checksum;	/* crc32c(uuid+revoke_block) */
-};
-
 /* Definitions for the journal tag flags word: */
 #define JBD2_FLAG_ESCAPE		1	/* on-disk block is escaped */
 #define JBD2_FLAG_SAME_UUID	2	/* block has same uuid as previous */
@@ -1137,7 +1132,8 @@ static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
 }
 
 /* Log buffer allocation */
-struct buffer_head *jbd2_journal_get_descriptor_buffer(journal_t *journal);
+struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
+void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
 			      unsigned long *block);
@@ -1327,10 +1323,8 @@ extern int	   jbd2_journal_init_revoke_caches(void);
 extern void	   jbd2_journal_destroy_revoke(journal_t *);
 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
-extern void	   jbd2_journal_write_revoke_records(journal_t *journal,
+extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
-						     transaction_t *transaction,
+						     struct list_head *log_bufs);
-						     struct list_head *log_bufs,
-						     int write_op);
 
 /* Recovery revoke support */
 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);

include/linux/mbcache.h

@@ -1,55 +1,52 @@
-/*
+#ifndef _LINUX_MBCACHE_H
-  File: linux/mbcache.h
+#define _LINUX_MBCACHE_H
+
+#include <linux/hash.h>
+#include <linux/list_bl.h>
+#include <linux/list.h>
+#include <linux/atomic.h>
+#include <linux/fs.h>
+
+struct mb_cache;
 
-  (C) 2001 by Andreas Gruenbacher, <a.gruenbacher@computer.org>
-*/
 struct mb_cache_entry {
-	struct list_head		e_lru_list;
+	/* List of entries in cache - protected by cache->c_list_lock */
-	struct mb_cache			*e_cache;
+	struct list_head	e_list;
-	unsigned short			e_used;
+	/* Hash table list - protected by hash chain bitlock */
-	unsigned short			e_queued;
+	struct hlist_bl_node	e_hash_list;
-	atomic_t			e_refcnt;
+	atomic_t		e_refcnt;
-	struct block_device		*e_bdev;
+	/* Key in hash - stable during lifetime of the entry */
-	sector_t			e_block;
+	u32			e_key;
-	struct hlist_bl_node		e_block_list;
+	u32			e_referenced:1;
-	struct {
+	u32			e_reusable:1;
-		struct hlist_bl_node	o_list;
+	/* Block number of hashed block - stable during lifetime of the entry */
-		unsigned int		o_key;
+	sector_t		e_block;
-	} e_index;
-	struct hlist_bl_head		*e_block_hash_p;
-	struct hlist_bl_head		*e_index_hash_p;
 };
 
-struct mb_cache {
+struct mb_cache *mb_cache_create(int bucket_bits);
-	struct list_head		c_cache_list;
+void mb_cache_destroy(struct mb_cache *cache);
-	const char			*c_name;
-	atomic_t			c_entry_count;
-	int				c_max_entries;
-	int				c_bucket_bits;
-	struct kmem_cache		*c_entry_cache;
-	struct hlist_bl_head		*c_block_hash;
-	struct hlist_bl_head		*c_index_hash;
-};
 
-/* Functions on caches */
+int mb_cache_entry_create(struct mb_cache *cache, gfp_t mask, u32 key,
+			  sector_t block, bool reusable);
+void __mb_cache_entry_free(struct mb_cache_entry *entry);
+static inline int mb_cache_entry_put(struct mb_cache *cache,
+				     struct mb_cache_entry *entry)
+{
+	if (!atomic_dec_and_test(&entry->e_refcnt))
+		return 0;
+	__mb_cache_entry_free(entry);
+	return 1;
+}
 
-struct mb_cache *mb_cache_create(const char *, int);
+void mb_cache_entry_delete_block(struct mb_cache *cache, u32 key,
-void mb_cache_shrink(struct block_device *);
+				  sector_t block);
-void mb_cache_destroy(struct mb_cache *);
+struct mb_cache_entry *mb_cache_entry_get(struct mb_cache *cache, u32 key,
-
+					  sector_t block);
-/* Functions on cache entries */
-
-struct mb_cache_entry *mb_cache_entry_alloc(struct mb_cache *, gfp_t);
-int mb_cache_entry_insert(struct mb_cache_entry *, struct block_device *,
-			  sector_t, unsigned int);
-void mb_cache_entry_release(struct mb_cache_entry *);
-void mb_cache_entry_free(struct mb_cache_entry *);
-struct mb_cache_entry *mb_cache_entry_get(struct mb_cache *,
-					  struct block_device *,
-					  sector_t);
 struct mb_cache_entry *mb_cache_entry_find_first(struct mb_cache *cache,
-						 struct block_device *, 
+						 u32 key);
-						 unsigned int);
+struct mb_cache_entry *mb_cache_entry_find_next(struct mb_cache *cache,
-struct mb_cache_entry *mb_cache_entry_find_next(struct mb_cache_entry *,
+						struct mb_cache_entry *entry);
-						struct block_device *, 
+void mb_cache_entry_touch(struct mb_cache *cache,
-						unsigned int);
+			  struct mb_cache_entry *entry);
+
+#endif	/* _LINUX_MBCACHE_H */