Merge branch 'master' of git://oss.sgi.com:8090/xfs/linux-2.6

Documentation/filesystems/xfs.txt

@@ -229,10 +229,6 @@ The following sysctls are available for the XFS filesystem:
 	ISGID bit is cleared if the irix_sgid_inherit compatibility sysctl
 	is set.
 
-  fs.xfs.restrict_chown		(Min: 0  Default: 1  Max: 1)
-  	Controls whether unprivileged users can use chown to "give away"
-	a file to another user.
-
   fs.xfs.inherit_sync		(Min: 0  Default: 1  Max: 1)
 	Setting this to "1" will cause the "sync" flag set
 	by the xfs_io(8) chattr command on a directory to be

fs/inode.c

@@ -108,84 +108,100 @@ static void wake_up_inode(struct inode *inode)
 	wake_up_bit(&inode->i_state, __I_LOCK);
 }
 
-static struct inode *alloc_inode(struct super_block *sb)
+/**
+ * inode_init_always - perform inode structure intialisation
+ * @sb		- superblock inode belongs to.
+ * @inode	- inode to initialise
+ *
+ * These are initializations that need to be done on every inode
+ * allocation as the fields are not initialised by slab allocation.
+ */
+struct inode *inode_init_always(struct super_block *sb, struct inode *inode)
 {
 	static const struct address_space_operations empty_aops;
 	static struct inode_operations empty_iops;
 	static const struct file_operations empty_fops;
-	struct inode *inode;
-
-	if (sb->s_op->alloc_inode)
-		inode = sb->s_op->alloc_inode(sb);
-	else
-		inode = (struct inode *) kmem_cache_alloc(inode_cachep, GFP_KERNEL);
 
-	if (inode) {
-		struct address_space * const mapping = &inode->i_data;
-
-		inode->i_sb = sb;
-		inode->i_blkbits = sb->s_blocksize_bits;
-		inode->i_flags = 0;
-		atomic_set(&inode->i_count, 1);
-		inode->i_op = &empty_iops;
-		inode->i_fop = &empty_fops;
-		inode->i_nlink = 1;
-		atomic_set(&inode->i_writecount, 0);
-		inode->i_size = 0;
-		inode->i_blocks = 0;
-		inode->i_bytes = 0;
-		inode->i_generation = 0;
+	struct address_space * const mapping = &inode->i_data;
+
+	inode->i_sb = sb;
+	inode->i_blkbits = sb->s_blocksize_bits;
+	inode->i_flags = 0;
+	atomic_set(&inode->i_count, 1);
+	inode->i_op = &empty_iops;
+	inode->i_fop = &empty_fops;
+	inode->i_nlink = 1;
+	atomic_set(&inode->i_writecount, 0);
+	inode->i_size = 0;
+	inode->i_blocks = 0;
+	inode->i_bytes = 0;
+	inode->i_generation = 0;
 #ifdef CONFIG_QUOTA
-		memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
+	memset(&inode->i_dquot, 0, sizeof(inode->i_dquot));
 #endif
-		inode->i_pipe = NULL;
-		inode->i_bdev = NULL;
-		inode->i_cdev = NULL;
-		inode->i_rdev = 0;
-		inode->dirtied_when = 0;
-		if (security_inode_alloc(inode)) {
-			if (inode->i_sb->s_op->destroy_inode)
-				inode->i_sb->s_op->destroy_inode(inode);
-			else
-				kmem_cache_free(inode_cachep, (inode));
-			return NULL;
-		}
+	inode->i_pipe = NULL;
+	inode->i_bdev = NULL;
+	inode->i_cdev = NULL;
+	inode->i_rdev = 0;
+	inode->dirtied_when = 0;
+	if (security_inode_alloc(inode)) {
+		if (inode->i_sb->s_op->destroy_inode)
+			inode->i_sb->s_op->destroy_inode(inode);
+		else
+			kmem_cache_free(inode_cachep, (inode));
+		return NULL;
+	}
 
-		spin_lock_init(&inode->i_lock);
-		lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
+	spin_lock_init(&inode->i_lock);
+	lockdep_set_class(&inode->i_lock, &sb->s_type->i_lock_key);
 
-		mutex_init(&inode->i_mutex);
-		lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
+	mutex_init(&inode->i_mutex);
+	lockdep_set_class(&inode->i_mutex, &sb->s_type->i_mutex_key);
 
-		init_rwsem(&inode->i_alloc_sem);
-		lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key);
+	init_rwsem(&inode->i_alloc_sem);
+	lockdep_set_class(&inode->i_alloc_sem, &sb->s_type->i_alloc_sem_key);
 
-		mapping->a_ops = &empty_aops;
- 		mapping->host = inode;
-		mapping->flags = 0;
-		mapping_set_gfp_mask(mapping, GFP_HIGHUSER_PAGECACHE);
-		mapping->assoc_mapping = NULL;
-		mapping->backing_dev_info = &default_backing_dev_info;
-		mapping->writeback_index = 0;
+	mapping->a_ops = &empty_aops;
+	mapping->host = inode;
+	mapping->flags = 0;
+	mapping_set_gfp_mask(mapping, GFP_HIGHUSER_PAGECACHE);
+	mapping->assoc_mapping = NULL;
+	mapping->backing_dev_info = &default_backing_dev_info;
+	mapping->writeback_index = 0;
 
-		/*
-		 * If the block_device provides a backing_dev_info for client
-		 * inodes then use that.  Otherwise the inode share the bdev's
-		 * backing_dev_info.
-		 */
-		if (sb->s_bdev) {
-			struct backing_dev_info *bdi;
+	/*
+	 * If the block_device provides a backing_dev_info for client
+	 * inodes then use that.  Otherwise the inode share the bdev's
+	 * backing_dev_info.
+	 */
+	if (sb->s_bdev) {
+		struct backing_dev_info *bdi;
 
-			bdi = sb->s_bdev->bd_inode_backing_dev_info;
-			if (!bdi)
-				bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
-			mapping->backing_dev_info = bdi;
-		}
-		inode->i_private = NULL;
-		inode->i_mapping = mapping;
+		bdi = sb->s_bdev->bd_inode_backing_dev_info;
+		if (!bdi)
+			bdi = sb->s_bdev->bd_inode->i_mapping->backing_dev_info;
+		mapping->backing_dev_info = bdi;
 	}
+	inode->i_private = NULL;
+	inode->i_mapping = mapping;
+
 	return inode;
 }
+EXPORT_SYMBOL(inode_init_always);
+
+static struct inode *alloc_inode(struct super_block *sb)
+{
+	struct inode *inode;
+
+	if (sb->s_op->alloc_inode)
+		inode = sb->s_op->alloc_inode(sb);
+	else
+		inode = kmem_cache_alloc(inode_cachep, GFP_KERNEL);
+
+	if (inode)
+		return inode_init_always(sb, inode);
+	return NULL;
+}
 
 void destroy_inode(struct inode *inode) 
 {
@@ -196,6 +212,7 @@ void destroy_inode(struct inode *inode)
 	else
 		kmem_cache_free(inode_cachep, (inode));
 }
+EXPORT_SYMBOL(destroy_inode);
 
 
 /*
@@ -534,6 +551,49 @@ repeat:
 	return node ? inode : NULL;
 }
 
+static unsigned long hash(struct super_block *sb, unsigned long hashval)
+{
+	unsigned long tmp;
+
+	tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
+			L1_CACHE_BYTES;
+	tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS);
+	return tmp & I_HASHMASK;
+}
+
+static inline void
+__inode_add_to_lists(struct super_block *sb, struct hlist_head *head,
+			struct inode *inode)
+{
+	inodes_stat.nr_inodes++;
+	list_add(&inode->i_list, &inode_in_use);
+	list_add(&inode->i_sb_list, &sb->s_inodes);
+	if (head)
+		hlist_add_head(&inode->i_hash, head);
+}
+
+/**
+ * inode_add_to_lists - add a new inode to relevant lists
+ * @sb		- superblock inode belongs to.
+ * @inode	- inode to mark in use
+ *
+ * When an inode is allocated it needs to be accounted for, added to the in use
+ * list, the owning superblock and the inode hash. This needs to be done under
+ * the inode_lock, so export a function to do this rather than the inode lock
+ * itself. We calculate the hash list to add to here so it is all internal
+ * which requires the caller to have already set up the inode number in the
+ * inode to add.
+ */
+void inode_add_to_lists(struct super_block *sb, struct inode *inode)
+{
+	struct hlist_head *head = inode_hashtable + hash(sb, inode->i_ino);
+
+	spin_lock(&inode_lock);
+	__inode_add_to_lists(sb, head, inode);
+	spin_unlock(&inode_lock);
+}
+EXPORT_SYMBOL_GPL(inode_add_to_lists);
+
 /**
  *	new_inode 	- obtain an inode
  *	@sb: superblock
@@ -561,9 +621,7 @@ struct inode *new_inode(struct super_block *sb)
 	inode = alloc_inode(sb);
 	if (inode) {
 		spin_lock(&inode_lock);
-		inodes_stat.nr_inodes++;
-		list_add(&inode->i_list, &inode_in_use);
-		list_add(&inode->i_sb_list, &sb->s_inodes);
+		__inode_add_to_lists(sb, NULL, inode);
 		inode->i_ino = ++last_ino;
 		inode->i_state = 0;
 		spin_unlock(&inode_lock);
@@ -622,10 +680,7 @@ static struct inode * get_new_inode(struct super_block *sb, struct hlist_head *h
 			if (set(inode, data))
 				goto set_failed;
 
-			inodes_stat.nr_inodes++;
-			list_add(&inode->i_list, &inode_in_use);
-			list_add(&inode->i_sb_list, &sb->s_inodes);
-			hlist_add_head(&inode->i_hash, head);
+			__inode_add_to_lists(sb, head, inode);
 			inode->i_state = I_LOCK|I_NEW;
 			spin_unlock(&inode_lock);
 
@@ -671,10 +726,7 @@ static struct inode * get_new_inode_fast(struct super_block *sb, struct hlist_he
 		old = find_inode_fast(sb, head, ino);
 		if (!old) {
 			inode->i_ino = ino;
-			inodes_stat.nr_inodes++;
-			list_add(&inode->i_list, &inode_in_use);
-			list_add(&inode->i_sb_list, &sb->s_inodes);
-			hlist_add_head(&inode->i_hash, head);
+			__inode_add_to_lists(sb, head, inode);
 			inode->i_state = I_LOCK|I_NEW;
 			spin_unlock(&inode_lock);
 
@@ -698,16 +750,6 @@ static struct inode * get_new_inode_fast(struct super_block *sb, struct hlist_he
 	return inode;
 }
 
-static unsigned long hash(struct super_block *sb, unsigned long hashval)
-{
-	unsigned long tmp;
-
-	tmp = (hashval * (unsigned long)sb) ^ (GOLDEN_RATIO_PRIME + hashval) /
-			L1_CACHE_BYTES;
-	tmp = tmp ^ ((tmp ^ GOLDEN_RATIO_PRIME) >> I_HASHBITS);
-	return tmp & I_HASHMASK;
-}
-
 /**
  *	iunique - get a unique inode number
  *	@sb: superblock

fs/xfs/Makefile

@@ -91,7 +91,8 @@ xfs-y				+= xfs_alloc.o \
 				   xfs_dmops.o \
 				   xfs_qmops.o
 
-xfs-$(CONFIG_XFS_TRACE)		+= xfs_dir2_trace.o
+xfs-$(CONFIG_XFS_TRACE)		+= xfs_btree_trace.o \
+				   xfs_dir2_trace.o
 
 # Objects in linux/
 xfs-y				+= $(addprefix $(XFS_LINUX)/, \
@@ -106,6 +107,7 @@ xfs-y				+= $(addprefix $(XFS_LINUX)/, \
 				   xfs_iops.o \
 				   xfs_lrw.o \
 				   xfs_super.o \
+				   xfs_sync.o \
 				   xfs_vnode.o \
 				   xfs_xattr.o)
 

fs/xfs/linux-2.6/xfs_aops.c

@@ -191,7 +191,7 @@ xfs_setfilesize(
 		ip->i_d.di_size = isize;
 		ip->i_update_core = 1;
 		ip->i_update_size = 1;
-		mark_inode_dirty_sync(ioend->io_inode);
+		xfs_mark_inode_dirty_sync(ip);
 	}
 
 	xfs_iunlock(ip, XFS_ILOCK_EXCL);

fs/xfs/linux-2.6/xfs_cred.h

@@ -24,15 +24,7 @@
  * Credentials
  */
 typedef struct cred {
-	/* EMPTY */
+       /* EMPTY */
 } cred_t;
 
-extern struct cred *sys_cred;
-
-/* this is a hack.. (assumes sys_cred is the only cred_t in the system) */
-static inline int capable_cred(cred_t *cr, int cid)
-{
-	return (cr == sys_cred) ? 1 : capable(cid);
-}
-
 #endif  /* __XFS_CRED_H__ */

fs/xfs/linux-2.6/xfs_globals.c

@@ -26,7 +26,6 @@
  */
 xfs_param_t xfs_params = {
 			  /*	MIN		DFLT		MAX	*/
-	.restrict_chown	= {	0,		1,		1	},
 	.sgid_inherit	= {	0,		0,		1	},
 	.symlink_mode	= {	0,		0,		1	},
 	.panic_mask	= {	0,		0,		255	},
@@ -43,10 +42,3 @@ xfs_param_t xfs_params = {
 	.inherit_nodfrg	= {	0,		1,		1	},
 	.fstrm_timer	= {	1,		30*100,		3600*100},
 };
-
-/*
- * Global system credential structure.
- */
-static cred_t sys_cred_val;
-cred_t *sys_cred = &sys_cred_val;
-

fs/xfs/linux-2.6/xfs_globals.h

@@ -19,6 +19,5 @@
 #define __XFS_GLOBALS_H__
 
 extern uint64_t	xfs_panic_mask;		/* set to cause more panics */
-extern struct cred *sys_cred;
 
 #endif	/* __XFS_GLOBALS_H__ */

fs/xfs/linux-2.6/xfs_ioctl.c

@@ -691,8 +691,7 @@ xfs_ioc_space(
 	if (ioflags & IO_INVIS)
 		attr_flags |= XFS_ATTR_DMI;
 
-	error = xfs_change_file_space(ip, cmd, &bf, filp->f_pos,
-					      NULL, attr_flags);
+	error = xfs_change_file_space(ip, cmd, &bf, filp->f_pos, attr_flags);
 	return -error;
 }
 
@@ -1007,7 +1006,7 @@ xfs_ioctl_setattr(
 	 * to the file owner ID, except in cases where the
 	 * CAP_FSETID capability is applicable.
 	 */
-	if (current->fsuid != ip->i_d.di_uid && !capable(CAP_FOWNER)) {
+	if (current_fsuid() != ip->i_d.di_uid && !capable(CAP_FOWNER)) {
 		code = XFS_ERROR(EPERM);
 		goto error_return;
 	}
@@ -1104,10 +1103,6 @@ xfs_ioctl_setattr(
 
 	/*
 	 * Change file ownership.  Must be the owner or privileged.
-	 * If the system was configured with the "restricted_chown"
-	 * option, the owner is not permitted to give away the file,
-	 * and can change the group id only to a group of which he
-	 * or she is a member.
 	 */
 	if (mask & FSX_PROJID) {
 		/*

fs/xfs/linux-2.6/xfs_iops.c

@@ -64,14 +64,14 @@ xfs_synchronize_atime(
 {
 	struct inode	*inode = VFS_I(ip);
 
-	if (inode) {
+	if (!(inode->i_state & I_CLEAR)) {
 		ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
 		ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
 	}
 }
 
 /*
- * If the linux inode exists, mark it dirty.
+ * If the linux inode is valid, mark it dirty.
  * Used when commiting a dirty inode into a transaction so that
  * the inode will get written back by the linux code
  */
@@ -81,7 +81,7 @@ xfs_mark_inode_dirty_sync(
 {
 	struct inode	*inode = VFS_I(ip);
 
-	if (inode)
+	if (!(inode->i_state & (I_WILL_FREE|I_FREEING|I_CLEAR)))
 		mark_inode_dirty_sync(inode);
 }
 
@@ -128,7 +128,7 @@ xfs_ichgtime(
 	if (sync_it) {
 		SYNCHRONIZE();
 		ip->i_update_core = 1;
-		mark_inode_dirty_sync(inode);
+		xfs_mark_inode_dirty_sync(ip);
 	}
 }
 
@@ -601,7 +601,7 @@ xfs_vn_setattr(
 	struct dentry	*dentry,
 	struct iattr	*iattr)
 {
-	return -xfs_setattr(XFS_I(dentry->d_inode), iattr, 0, NULL);
+	return -xfs_setattr(XFS_I(dentry->d_inode), iattr, 0);
 }
 
 /*
@@ -642,7 +642,7 @@ xfs_vn_fallocate(
 
 	xfs_ilock(ip, XFS_IOLOCK_EXCL);
 	error = xfs_change_file_space(ip, XFS_IOC_RESVSP, &bf,
-				      0, NULL, XFS_ATTR_NOLOCK);
+				      0, XFS_ATTR_NOLOCK);
 	if (!error && !(mode & FALLOC_FL_KEEP_SIZE) &&
 	    offset + len > i_size_read(inode))
 		new_size = offset + len;
@@ -653,7 +653,7 @@ xfs_vn_fallocate(
 
 		iattr.ia_valid = ATTR_SIZE;
 		iattr.ia_size = new_size;
-		error = xfs_setattr(ip, &iattr, XFS_ATTR_NOLOCK, NULL);
+		error = xfs_setattr(ip, &iattr, XFS_ATTR_NOLOCK);
 	}
 
 	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
@@ -766,12 +766,21 @@ xfs_diflags_to_iflags(
  * When reading existing inodes from disk this is called directly
  * from xfs_iget, when creating a new inode it is called from
  * xfs_ialloc after setting up the inode.
+ *
+ * We are always called with an uninitialised linux inode here.
+ * We need to initialise the necessary fields and take a reference
+ * on it.
  */
 void
 xfs_setup_inode(
 	struct xfs_inode	*ip)
 {
-	struct inode		*inode = ip->i_vnode;
+	struct inode		*inode = &ip->i_vnode;
+
+	inode->i_ino = ip->i_ino;
+	inode->i_state = I_NEW|I_LOCK;
+	inode_add_to_lists(ip->i_mount->m_super, inode);
+	ASSERT(atomic_read(&inode->i_count) == 1);
 
 	inode->i_mode	= ip->i_d.di_mode;
 	inode->i_nlink	= ip->i_d.di_nlink;

fs/xfs/linux-2.6/xfs_linux.h

@@ -77,6 +77,7 @@
 #include <linux/spinlock.h>
 #include <linux/random.h>
 #include <linux/ctype.h>
+#include <linux/writeback.h>
 
 #include <asm/page.h>
 #include <asm/div64.h>
@@ -107,7 +108,6 @@
 #undef  HAVE_PERCPU_SB	/* per cpu superblock counters are a 2.6 feature */
 #endif
 
-#define restricted_chown	xfs_params.restrict_chown.val
 #define irix_sgid_inherit	xfs_params.sgid_inherit.val
 #define irix_symlink_mode	xfs_params.symlink_mode.val
 #define xfs_panic_mask		xfs_params.panic_mask.val

fs/xfs/linux-2.6/xfs_stats.c

@@ -53,6 +53,10 @@ xfs_read_xfsstats(
 		{ "icluster",		XFSSTAT_END_INODE_CLUSTER	},
 		{ "vnodes",		XFSSTAT_END_VNODE_OPS		},
 		{ "buf",		XFSSTAT_END_BUF			},
+		{ "abtb2",		XFSSTAT_END_ABTB_V2		},
+		{ "abtc2",		XFSSTAT_END_ABTC_V2		},
+		{ "bmbt2",		XFSSTAT_END_BMBT_V2		},
+		{ "ibt2",		XFSSTAT_END_IBT_V2		},
 	};
 
 	/* Loop over all stats groups */

fs/xfs/linux-2.6/xfs_stats.h

@@ -118,6 +118,71 @@ struct xfsstats {
 	__uint32_t		xb_page_retries;
 	__uint32_t		xb_page_found;
 	__uint32_t		xb_get_read;
+/* Version 2 btree counters */
+#define XFSSTAT_END_ABTB_V2		(XFSSTAT_END_BUF+15)
+	__uint32_t		xs_abtb_2_lookup;
+	__uint32_t		xs_abtb_2_compare;
+	__uint32_t		xs_abtb_2_insrec;
+	__uint32_t		xs_abtb_2_delrec;
+	__uint32_t		xs_abtb_2_newroot;
+	__uint32_t		xs_abtb_2_killroot;
+	__uint32_t		xs_abtb_2_increment;
+	__uint32_t		xs_abtb_2_decrement;
+	__uint32_t		xs_abtb_2_lshift;
+	__uint32_t		xs_abtb_2_rshift;
+	__uint32_t		xs_abtb_2_split;
+	__uint32_t		xs_abtb_2_join;
+	__uint32_t		xs_abtb_2_alloc;
+	__uint32_t		xs_abtb_2_free;
+	__uint32_t		xs_abtb_2_moves;
+#define XFSSTAT_END_ABTC_V2		(XFSSTAT_END_ABTB_V2+15)
+	__uint32_t		xs_abtc_2_lookup;
+	__uint32_t		xs_abtc_2_compare;
+	__uint32_t		xs_abtc_2_insrec;
+	__uint32_t		xs_abtc_2_delrec;
+	__uint32_t		xs_abtc_2_newroot;
+	__uint32_t		xs_abtc_2_killroot;
+	__uint32_t		xs_abtc_2_increment;
+	__uint32_t		xs_abtc_2_decrement;
+	__uint32_t		xs_abtc_2_lshift;
+	__uint32_t		xs_abtc_2_rshift;
+	__uint32_t		xs_abtc_2_split;
+	__uint32_t		xs_abtc_2_join;
+	__uint32_t		xs_abtc_2_alloc;
+	__uint32_t		xs_abtc_2_free;
+	__uint32_t		xs_abtc_2_moves;
+#define XFSSTAT_END_BMBT_V2		(XFSSTAT_END_ABTC_V2+15)
+	__uint32_t		xs_bmbt_2_lookup;
+	__uint32_t		xs_bmbt_2_compare;
+	__uint32_t		xs_bmbt_2_insrec;
+	__uint32_t		xs_bmbt_2_delrec;
+	__uint32_t		xs_bmbt_2_newroot;
+	__uint32_t		xs_bmbt_2_killroot;
+	__uint32_t		xs_bmbt_2_increment;
+	__uint32_t		xs_bmbt_2_decrement;
+	__uint32_t		xs_bmbt_2_lshift;
+	__uint32_t		xs_bmbt_2_rshift;
+	__uint32_t		xs_bmbt_2_split;
+	__uint32_t		xs_bmbt_2_join;
+	__uint32_t		xs_bmbt_2_alloc;
+	__uint32_t		xs_bmbt_2_free;
+	__uint32_t		xs_bmbt_2_moves;
+#define XFSSTAT_END_IBT_V2		(XFSSTAT_END_BMBT_V2+15)
+	__uint32_t		xs_ibt_2_lookup;
+	__uint32_t		xs_ibt_2_compare;
+	__uint32_t		xs_ibt_2_insrec;
+	__uint32_t		xs_ibt_2_delrec;
+	__uint32_t		xs_ibt_2_newroot;
+	__uint32_t		xs_ibt_2_killroot;
+	__uint32_t		xs_ibt_2_increment;
+	__uint32_t		xs_ibt_2_decrement;
+	__uint32_t		xs_ibt_2_lshift;
+	__uint32_t		xs_ibt_2_rshift;
+	__uint32_t		xs_ibt_2_split;
+	__uint32_t		xs_ibt_2_join;
+	__uint32_t		xs_ibt_2_alloc;
+	__uint32_t		xs_ibt_2_free;
+	__uint32_t		xs_ibt_2_moves;
 /* Extra precision counters */
 	__uint64_t		xs_xstrat_bytes;
 	__uint64_t		xs_write_bytes;

fs/xfs/linux-2.6/xfs_super.h

@@ -101,9 +101,6 @@ struct block_device;
 
 extern __uint64_t xfs_max_file_offset(unsigned int);
 
-extern void xfs_flush_inode(struct xfs_inode *);
-extern void xfs_flush_device(struct xfs_inode *);
-
 extern void xfs_blkdev_issue_flush(struct xfs_buftarg *);
 
 extern const struct export_operations xfs_export_operations;

fs/xfs/linux-2.6/xfs_sync.c

@@ -0,0 +1,763 @@
+/*
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.
+ * All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_types.h"
+#include "xfs_bit.h"
+#include "xfs_log.h"
+#include "xfs_inum.h"
+#include "xfs_trans.h"
+#include "xfs_sb.h"
+#include "xfs_ag.h"
+#include "xfs_dir2.h"
+#include "xfs_dmapi.h"
+#include "xfs_mount.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_btree.h"
+#include "xfs_dir2_sf.h"
+#include "xfs_attr_sf.h"
+#include "xfs_inode.h"
+#include "xfs_dinode.h"
+#include "xfs_error.h"
+#include "xfs_mru_cache.h"
+#include "xfs_filestream.h"
+#include "xfs_vnodeops.h"
+#include "xfs_utils.h"
+#include "xfs_buf_item.h"
+#include "xfs_inode_item.h"
+#include "xfs_rw.h"
+
+#include <linux/kthread.h>
+#include <linux/freezer.h>
+
+/*
+ * Sync all the inodes in the given AG according to the
+ * direction given by the flags.
+ */
+STATIC int
+xfs_sync_inodes_ag(
+	xfs_mount_t	*mp,
+	int		ag,
+	int		flags)
+{
+	xfs_perag_t	*pag = &mp->m_perag[ag];
+	int		nr_found;
+	uint32_t	first_index = 0;
+	int		error = 0;
+	int		last_error = 0;
+	int		fflag = XFS_B_ASYNC;
+
+	if (flags & SYNC_DELWRI)
+		fflag = XFS_B_DELWRI;
+	if (flags & SYNC_WAIT)
+		fflag = 0;		/* synchronous overrides all */
+
+	do {
+		struct inode	*inode;
+		xfs_inode_t	*ip = NULL;
+		int		lock_flags = XFS_ILOCK_SHARED;
+
+		/*
+		 * use a gang lookup to find the next inode in the tree
+		 * as the tree is sparse and a gang lookup walks to find
+		 * the number of objects requested.
+		 */
+		read_lock(&pag->pag_ici_lock);
+		nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
+				(void**)&ip, first_index, 1);
+
+		if (!nr_found) {
+			read_unlock(&pag->pag_ici_lock);
+			break;
+		}
+
+		/*
+		 * Update the index for the next lookup. Catch overflows
+		 * into the next AG range which can occur if we have inodes
+		 * in the last block of the AG and we are currently
+		 * pointing to the last inode.
+		 */
+		first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
+		if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino)) {
+			read_unlock(&pag->pag_ici_lock);
+			break;
+		}
+
+		/* nothing to sync during shutdown */
+		if (XFS_FORCED_SHUTDOWN(mp)) {
+			read_unlock(&pag->pag_ici_lock);
+			return 0;
+		}
+
+		/*
+		 * If we can't get a reference on the inode, it must be
+		 * in reclaim. Leave it for the reclaim code to flush.
+		 */
+		inode = VFS_I(ip);
+		if (!igrab(inode)) {
+			read_unlock(&pag->pag_ici_lock);
+			continue;
+		}
+		read_unlock(&pag->pag_ici_lock);
+
+		/* bad inodes are dealt with elsewhere */
+		if (is_bad_inode(inode)) {
+			IRELE(ip);
+			continue;
+		}
+
+		/*
+		 * If we have to flush data or wait for I/O completion
+		 * we need to hold the iolock.
+		 */
+		if ((flags & SYNC_DELWRI) && VN_DIRTY(inode)) {
+			xfs_ilock(ip, XFS_IOLOCK_SHARED);
+			lock_flags |= XFS_IOLOCK_SHARED;
+			error = xfs_flush_pages(ip, 0, -1, fflag, FI_NONE);
+			if (flags & SYNC_IOWAIT)
+				vn_iowait(ip);
+		}
+		xfs_ilock(ip, XFS_ILOCK_SHARED);
+
+		if ((flags & SYNC_ATTR) && !xfs_inode_clean(ip)) {
+			if (flags & SYNC_WAIT) {
+				xfs_iflock(ip);
+				if (!xfs_inode_clean(ip))
+					error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
+				else
+					xfs_ifunlock(ip);
+			} else if (xfs_iflock_nowait(ip)) {
+				if (!xfs_inode_clean(ip))
+					error = xfs_iflush(ip, XFS_IFLUSH_DELWRI);
+				else
+					xfs_ifunlock(ip);
+			}
+		}
+		xfs_iput(ip, lock_flags);
+
+		if (error)
+			last_error = error;
+		/*
+		 * bail out if the filesystem is corrupted.
+		 */
+		if (error == EFSCORRUPTED)
+			return XFS_ERROR(error);
+
+	} while (nr_found);
+
+	return last_error;
+}
+
+int
+xfs_sync_inodes(
+	xfs_mount_t	*mp,
+	int		flags)
+{
+	int		error;
+	int		last_error;
+	int		i;
+	int		lflags = XFS_LOG_FORCE;
+
+	if (mp->m_flags & XFS_MOUNT_RDONLY)
+		return 0;
+	error = 0;
+	last_error = 0;
+
+	if (flags & SYNC_WAIT)
+		lflags |= XFS_LOG_SYNC;
+
+	for (i = 0; i < mp->m_sb.sb_agcount; i++) {
+		if (!mp->m_perag[i].pag_ici_init)
+			continue;
+		error = xfs_sync_inodes_ag(mp, i, flags);
+		if (error)
+			last_error = error;
+		if (error == EFSCORRUPTED)
+			break;
+	}
+	if (flags & SYNC_DELWRI)
+		xfs_log_force(mp, 0, lflags);
+
+	return XFS_ERROR(last_error);
+}
+
+STATIC int
+xfs_commit_dummy_trans(
+	struct xfs_mount	*mp,
+	uint			log_flags)
+{
+	struct xfs_inode	*ip = mp->m_rootip;
+	struct xfs_trans	*tp;
+	int			error;
+
+	/*
+	 * Put a dummy transaction in the log to tell recovery
+	 * that all others are OK.
+	 */
+	tp = xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
+	error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0);
+	if (error) {
+		xfs_trans_cancel(tp, 0);
+		return error;
+	}
+
+	xfs_ilock(ip, XFS_ILOCK_EXCL);
+
+	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
+	xfs_trans_ihold(tp, ip);
+	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
+	/* XXX(hch): ignoring the error here.. */
+	error = xfs_trans_commit(tp, 0);
+
+	xfs_iunlock(ip, XFS_ILOCK_EXCL);
+
+	xfs_log_force(mp, 0, log_flags);
+	return 0;
+}
+
+int
+xfs_sync_fsdata(
+	struct xfs_mount	*mp,
+	int			flags)
+{
+	struct xfs_buf		*bp;
+	struct xfs_buf_log_item	*bip;
+	int			error = 0;
+
+	/*
+	 * If this is xfssyncd() then only sync the superblock if we can
+	 * lock it without sleeping and it is not pinned.
+	 */
+	if (flags & SYNC_BDFLUSH) {
+		ASSERT(!(flags & SYNC_WAIT));
+
+		bp = xfs_getsb(mp, XFS_BUF_TRYLOCK);
+		if (!bp)
+			goto out;
+
+		bip = XFS_BUF_FSPRIVATE(bp, struct xfs_buf_log_item *);
+		if (!bip || !xfs_buf_item_dirty(bip) || XFS_BUF_ISPINNED(bp))
+			goto out_brelse;
+	} else {
+		bp = xfs_getsb(mp, 0);
+
+		/*
+		 * If the buffer is pinned then push on the log so we won't
+		 * get stuck waiting in the write for someone, maybe
+		 * ourselves, to flush the log.
+		 *
+		 * Even though we just pushed the log above, we did not have
+		 * the superblock buffer locked at that point so it can
+		 * become pinned in between there and here.
+		 */
+		if (XFS_BUF_ISPINNED(bp))
+			xfs_log_force(mp, 0, XFS_LOG_FORCE);
+	}
+
+
+	if (flags & SYNC_WAIT)
+		XFS_BUF_UNASYNC(bp);
+	else
+		XFS_BUF_ASYNC(bp);
+
+	return xfs_bwrite(mp, bp);
+
+ out_brelse:
+	xfs_buf_relse(bp);
+ out:
+	return error;
+}
+
+/*
+ * When remounting a filesystem read-only or freezing the filesystem, we have
+ * two phases to execute. This first phase is syncing the data before we
+ * quiesce the filesystem, and the second is flushing all the inodes out after
+ * we've waited for all the transactions created by the first phase to
+ * complete. The second phase ensures that the inodes are written to their
+ * location on disk rather than just existing in transactions in the log. This
+ * means after a quiesce there is no log replay required to write the inodes to
+ * disk (this is the main difference between a sync and a quiesce).
+ */
+/*
+ * First stage of freeze - no writers will make progress now we are here,
+ * so we flush delwri and delalloc buffers here, then wait for all I/O to
+ * complete.  Data is frozen at that point. Metadata is not frozen,
+ * transactions can still occur here so don't bother flushing the buftarg
+ * because it'll just get dirty again.
+ */
+int
+xfs_quiesce_data(
+	struct xfs_mount	*mp)
+{
+	int error;
+
+	/* push non-blocking */
+	xfs_sync_inodes(mp, SYNC_DELWRI|SYNC_BDFLUSH);
+	XFS_QM_DQSYNC(mp, SYNC_BDFLUSH);
+	xfs_filestream_flush(mp);
+
+	/* push and block */
+	xfs_sync_inodes(mp, SYNC_DELWRI|SYNC_WAIT|SYNC_IOWAIT);
+	XFS_QM_DQSYNC(mp, SYNC_WAIT);
+
+	/* write superblock and hoover up shutdown errors */
+	error = xfs_sync_fsdata(mp, 0);
+
+	/* flush data-only devices */
+	if (mp->m_rtdev_targp)
+		XFS_bflush(mp->m_rtdev_targp);
+
+	return error;
+}
+
+STATIC void
+xfs_quiesce_fs(
+	struct xfs_mount	*mp)
+{
+	int	count = 0, pincount;
+
+	xfs_flush_buftarg(mp->m_ddev_targp, 0);
+	xfs_reclaim_inodes(mp, 0, XFS_IFLUSH_DELWRI_ELSE_ASYNC);
+
+	/*
+	 * This loop must run at least twice.  The first instance of the loop
+	 * will flush most meta data but that will generate more meta data
+	 * (typically directory updates).  Which then must be flushed and
+	 * logged before we can write the unmount record.
+	 */
+	do {
+		xfs_sync_inodes(mp, SYNC_ATTR|SYNC_WAIT);
+		pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
+		if (!pincount) {
+			delay(50);
+			count++;
+		}
+	} while (count < 2);
+}
+
+/*
+ * Second stage of a quiesce. The data is already synced, now we have to take
+ * care of the metadata. New transactions are already blocked, so we need to
+ * wait for any remaining transactions to drain out before proceding.
+ */
+void
+xfs_quiesce_attr(
+	struct xfs_mount	*mp)
+{
+	int	error = 0;
+
+	/* wait for all modifications to complete */
+	while (atomic_read(&mp->m_active_trans) > 0)
+		delay(100);
+
+	/* flush inodes and push all remaining buffers out to disk */
+	xfs_quiesce_fs(mp);
+
+	ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);
+
+	/* Push the superblock and write an unmount record */
+	error = xfs_log_sbcount(mp, 1);
+	if (error)
+		xfs_fs_cmn_err(CE_WARN, mp,
+				"xfs_attr_quiesce: failed to log sb changes. "
+				"Frozen image may not be consistent.");
+	xfs_log_unmount_write(mp);
+	xfs_unmountfs_writesb(mp);
+}
+
+/*
+ * Enqueue a work item to be picked up by the vfs xfssyncd thread.
+ * Doing this has two advantages:
+ * - It saves on stack space, which is tight in certain situations
+ * - It can be used (with care) as a mechanism to avoid deadlocks.
+ * Flushing while allocating in a full filesystem requires both.
+ */
+STATIC void
+xfs_syncd_queue_work(
+	struct xfs_mount *mp,
+	void		*data,
+	void		(*syncer)(struct xfs_mount *, void *))
+{
+	struct bhv_vfs_sync_work *work;
+
+	work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
+	INIT_LIST_HEAD(&work->w_list);
+	work->w_syncer = syncer;
+	work->w_data = data;
+	work->w_mount = mp;
+	spin_lock(&mp->m_sync_lock);
+	list_add_tail(&work->w_list, &mp->m_sync_list);
+	spin_unlock(&mp->m_sync_lock);
+	wake_up_process(mp->m_sync_task);
+}
+
+/*
+ * Flush delayed allocate data, attempting to free up reserved space
+ * from existing allocations.  At this point a new allocation attempt
+ * has failed with ENOSPC and we are in the process of scratching our
+ * heads, looking about for more room...
+ */
+STATIC void
+xfs_flush_inode_work(
+	struct xfs_mount *mp,
+	void		*arg)
+{
+	struct inode	*inode = arg;
+	filemap_flush(inode->i_mapping);
+	iput(inode);
+}
+
+void
+xfs_flush_inode(
+	xfs_inode_t	*ip)
+{
+	struct inode	*inode = VFS_I(ip);
+
+	igrab(inode);
+	xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
+	delay(msecs_to_jiffies(500));
+}
+
+/*
+ * This is the "bigger hammer" version of xfs_flush_inode_work...
+ * (IOW, "If at first you don't succeed, use a Bigger Hammer").
+ */
+STATIC void
+xfs_flush_device_work(
+	struct xfs_mount *mp,
+	void		*arg)
+{
+	struct inode	*inode = arg;
+	sync_blockdev(mp->m_super->s_bdev);
+	iput(inode);
+}
+
+void
+xfs_flush_device(
+	xfs_inode_t	*ip)
+{
+	struct inode	*inode = VFS_I(ip);
+
+	igrab(inode);
+	xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
+	delay(msecs_to_jiffies(500));
+	xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
+}
+
+/*
+ * Every sync period we need to unpin all items, reclaim inodes, sync
+ * quota and write out the superblock. We might need to cover the log
+ * to indicate it is idle.
+ */
+STATIC void
+xfs_sync_worker(
+	struct xfs_mount *mp,
+	void		*unused)
+{
+	int		error;
+
+	if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
+		xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
+		xfs_reclaim_inodes(mp, 0, XFS_IFLUSH_DELWRI_ELSE_ASYNC);
+		/* dgc: errors ignored here */
+		error = XFS_QM_DQSYNC(mp, SYNC_BDFLUSH);
+		error = xfs_sync_fsdata(mp, SYNC_BDFLUSH);
+		if (xfs_log_need_covered(mp))
+			error = xfs_commit_dummy_trans(mp, XFS_LOG_FORCE);
+	}
+	mp->m_sync_seq++;
+	wake_up(&mp->m_wait_single_sync_task);
+}
+
+STATIC int
+xfssyncd(
+	void			*arg)
+{
+	struct xfs_mount	*mp = arg;
+	long			timeleft;
+	bhv_vfs_sync_work_t	*work, *n;
+	LIST_HEAD		(tmp);
+
+	set_freezable();
+	timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
+	for (;;) {
+		timeleft = schedule_timeout_interruptible(timeleft);
+		/* swsusp */
+		try_to_freeze();
+		if (kthread_should_stop() && list_empty(&mp->m_sync_list))
+			break;
+
+		spin_lock(&mp->m_sync_lock);
+		/*
+		 * We can get woken by laptop mode, to do a sync -
+		 * that's the (only!) case where the list would be
+		 * empty with time remaining.
+		 */
+		if (!timeleft || list_empty(&mp->m_sync_list)) {
+			if (!timeleft)
+				timeleft = xfs_syncd_centisecs *
+							msecs_to_jiffies(10);
+			INIT_LIST_HEAD(&mp->m_sync_work.w_list);
+			list_add_tail(&mp->m_sync_work.w_list,
+					&mp->m_sync_list);
+		}
+		list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
+			list_move(&work->w_list, &tmp);
+		spin_unlock(&mp->m_sync_lock);
+
+		list_for_each_entry_safe(work, n, &tmp, w_list) {
+			(*work->w_syncer)(mp, work->w_data);
+			list_del(&work->w_list);
+			if (work == &mp->m_sync_work)
+				continue;
+			kmem_free(work);
+		}
+	}
+
+	return 0;
+}
+
+int
+xfs_syncd_init(
+	struct xfs_mount	*mp)
+{
+	mp->m_sync_work.w_syncer = xfs_sync_worker;
+	mp->m_sync_work.w_mount = mp;
+	mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
+	if (IS_ERR(mp->m_sync_task))
+		return -PTR_ERR(mp->m_sync_task);
+	return 0;
+}
+
+void
+xfs_syncd_stop(
+	struct xfs_mount	*mp)
+{
+	kthread_stop(mp->m_sync_task);
+}
+
+int
+xfs_reclaim_inode(
+	xfs_inode_t	*ip,
+	int		locked,
+	int		sync_mode)
+{
+	xfs_perag_t	*pag = xfs_get_perag(ip->i_mount, ip->i_ino);
+
+	/* The hash lock here protects a thread in xfs_iget_core from
+	 * racing with us on linking the inode back with a vnode.
+	 * Once we have the XFS_IRECLAIM flag set it will not touch
+	 * us.
+	 */
+	write_lock(&pag->pag_ici_lock);
+	spin_lock(&ip->i_flags_lock);
+	if (__xfs_iflags_test(ip, XFS_IRECLAIM) ||
+	    !__xfs_iflags_test(ip, XFS_IRECLAIMABLE)) {
+		spin_unlock(&ip->i_flags_lock);
+		write_unlock(&pag->pag_ici_lock);
+		if (locked) {
+			xfs_ifunlock(ip);
+			xfs_iunlock(ip, XFS_ILOCK_EXCL);
+		}
+		return 1;
+	}
+	__xfs_iflags_set(ip, XFS_IRECLAIM);
+	spin_unlock(&ip->i_flags_lock);
+	write_unlock(&pag->pag_ici_lock);
+	xfs_put_perag(ip->i_mount, pag);
+
+	/*
+	 * If the inode is still dirty, then flush it out.  If the inode
+	 * is not in the AIL, then it will be OK to flush it delwri as
+	 * long as xfs_iflush() does not keep any references to the inode.
+	 * We leave that decision up to xfs_iflush() since it has the
+	 * knowledge of whether it's OK to simply do a delwri flush of
+	 * the inode or whether we need to wait until the inode is
+	 * pulled from the AIL.
+	 * We get the flush lock regardless, though, just to make sure
+	 * we don't free it while it is being flushed.
+	 */
+	if (!locked) {
+		xfs_ilock(ip, XFS_ILOCK_EXCL);
+		xfs_iflock(ip);
+	}
+
+	/*
+	 * In the case of a forced shutdown we rely on xfs_iflush() to
+	 * wait for the inode to be unpinned before returning an error.
+	 */
+	if (!is_bad_inode(VFS_I(ip)) && xfs_iflush(ip, sync_mode) == 0) {
+		/* synchronize with xfs_iflush_done */
+		xfs_iflock(ip);
+		xfs_ifunlock(ip);
+	}
+
+	xfs_iunlock(ip, XFS_ILOCK_EXCL);
+	xfs_ireclaim(ip);
+	return 0;
+}
+
+/*
+ * We set the inode flag atomically with the radix tree tag.
+ * Once we get tag lookups on the radix tree, this inode flag
+ * can go away.
+ */
+void
+xfs_inode_set_reclaim_tag(
+	xfs_inode_t	*ip)
+{
+	xfs_mount_t	*mp = ip->i_mount;
+	xfs_perag_t	*pag = xfs_get_perag(mp, ip->i_ino);
+
+	read_lock(&pag->pag_ici_lock);
+	spin_lock(&ip->i_flags_lock);
+	radix_tree_tag_set(&pag->pag_ici_root,
+			XFS_INO_TO_AGINO(mp, ip->i_ino), XFS_ICI_RECLAIM_TAG);
+	__xfs_iflags_set(ip, XFS_IRECLAIMABLE);
+	spin_unlock(&ip->i_flags_lock);
+	read_unlock(&pag->pag_ici_lock);
+	xfs_put_perag(mp, pag);
+}
+
+void
+__xfs_inode_clear_reclaim_tag(
+	xfs_mount_t	*mp,
+	xfs_perag_t	*pag,
+	xfs_inode_t	*ip)
+{
+	radix_tree_tag_clear(&pag->pag_ici_root,
+			XFS_INO_TO_AGINO(mp, ip->i_ino), XFS_ICI_RECLAIM_TAG);
+}
+
+void
+xfs_inode_clear_reclaim_tag(
+	xfs_inode_t	*ip)
+{
+	xfs_mount_t	*mp = ip->i_mount;
+	xfs_perag_t	*pag = xfs_get_perag(mp, ip->i_ino);
+
+	read_lock(&pag->pag_ici_lock);
+	spin_lock(&ip->i_flags_lock);
+	__xfs_inode_clear_reclaim_tag(mp, pag, ip);
+	spin_unlock(&ip->i_flags_lock);
+	read_unlock(&pag->pag_ici_lock);
+	xfs_put_perag(mp, pag);
+}
+
+
+STATIC void
+xfs_reclaim_inodes_ag(
+	xfs_mount_t	*mp,
+	int		ag,
+	int		noblock,
+	int		mode)
+{
+	xfs_inode_t	*ip = NULL;
+	xfs_perag_t	*pag = &mp->m_perag[ag];
+	int		nr_found;
+	uint32_t	first_index;
+	int		skipped;
+
+restart:
+	first_index = 0;
+	skipped = 0;
+	do {
+		/*
+		 * use a gang lookup to find the next inode in the tree
+		 * as the tree is sparse and a gang lookup walks to find
+		 * the number of objects requested.
+		 */
+		read_lock(&pag->pag_ici_lock);
+		nr_found = radix_tree_gang_lookup_tag(&pag->pag_ici_root,
+					(void**)&ip, first_index, 1,
+					XFS_ICI_RECLAIM_TAG);
+
+		if (!nr_found) {
+			read_unlock(&pag->pag_ici_lock);
+			break;
+		}
+
+		/*
+		 * Update the index for the next lookup. Catch overflows
+		 * into the next AG range which can occur if we have inodes
+		 * in the last block of the AG and we are currently
+		 * pointing to the last inode.
+		 */
+		first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
+		if (first_index < XFS_INO_TO_AGINO(mp, ip->i_ino)) {
+			read_unlock(&pag->pag_ici_lock);
+			break;
+		}
+
+		ASSERT(xfs_iflags_test(ip, (XFS_IRECLAIMABLE|XFS_IRECLAIM)));
+
+		/* ignore if already under reclaim */
+		if (xfs_iflags_test(ip, XFS_IRECLAIM)) {
+			read_unlock(&pag->pag_ici_lock);
+			continue;
+		}
+
+		if (noblock) {
+			if (!xfs_ilock_nowait(ip, XFS_ILOCK_EXCL)) {
+				read_unlock(&pag->pag_ici_lock);
+				continue;
+			}
+			if (xfs_ipincount(ip) ||
+			    !xfs_iflock_nowait(ip)) {
+				xfs_iunlock(ip, XFS_ILOCK_EXCL);
+				read_unlock(&pag->pag_ici_lock);
+				continue;
+			}
+		}
+		read_unlock(&pag->pag_ici_lock);
+
+		/*
+		 * hmmm - this is an inode already in reclaim. Do
+		 * we even bother catching it here?
+		 */
+		if (xfs_reclaim_inode(ip, noblock, mode))
+			skipped++;
+	} while (nr_found);
+
+	if (skipped) {
+		delay(1);
+		goto restart;
+	}
+	return;
+
+}
+
+int
+xfs_reclaim_inodes(
+	xfs_mount_t	*mp,
+	int		 noblock,
+	int		mode)
+{
+	int		i;
+
+	for (i = 0; i < mp->m_sb.sb_agcount; i++) {
+		if (!mp->m_perag[i].pag_ici_init)
+			continue;
+		xfs_reclaim_inodes_ag(mp, i, noblock, mode);
+	}
+	return 0;
+}
+
+

fs/xfs/linux-2.6/xfs_sync.h

@@ -0,0 +1,55 @@
+/*
+ * Copyright (c) 2000-2006 Silicon Graphics, Inc.
+ * All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#ifndef XFS_SYNC_H
+#define XFS_SYNC_H 1
+
+struct xfs_mount;
+
+typedef struct bhv_vfs_sync_work {
+	struct list_head	w_list;
+	struct xfs_mount	*w_mount;
+	void			*w_data;	/* syncer routine argument */
+	void			(*w_syncer)(struct xfs_mount *, void *);
+} bhv_vfs_sync_work_t;
+
+#define SYNC_ATTR		0x0001	/* sync attributes */
+#define SYNC_DELWRI		0x0002	/* look at delayed writes */
+#define SYNC_WAIT		0x0004	/* wait for i/o to complete */
+#define SYNC_BDFLUSH		0x0008	/* BDFLUSH is calling -- don't block */
+#define SYNC_IOWAIT		0x0010  /* wait for all I/O to complete */
+
+int xfs_syncd_init(struct xfs_mount *mp);
+void xfs_syncd_stop(struct xfs_mount *mp);
+
+int xfs_sync_inodes(struct xfs_mount *mp, int flags);
+int xfs_sync_fsdata(struct xfs_mount *mp, int flags);
+
+int xfs_quiesce_data(struct xfs_mount *mp);
+void xfs_quiesce_attr(struct xfs_mount *mp);
+
+void xfs_flush_inode(struct xfs_inode *ip);
+void xfs_flush_device(struct xfs_inode *ip);
+
+int xfs_reclaim_inode(struct xfs_inode *ip, int locked, int sync_mode);
+int xfs_reclaim_inodes(struct xfs_mount *mp, int noblock, int mode);
+
+void xfs_inode_set_reclaim_tag(struct xfs_inode *ip);
+void xfs_inode_clear_reclaim_tag(struct xfs_inode *ip);
+void __xfs_inode_clear_reclaim_tag(struct xfs_mount *mp, struct xfs_perag *pag,
+				struct xfs_inode *ip);
+#endif

fs/xfs/linux-2.6/xfs_sysctl.c

@@ -55,17 +55,6 @@ xfs_stats_clear_proc_handler(
 #endif /* CONFIG_PROC_FS */
 
 static ctl_table xfs_table[] = {
-	{
-		.ctl_name	= XFS_RESTRICT_CHOWN,
-		.procname	= "restrict_chown",
-		.data		= &xfs_params.restrict_chown.val,
-		.maxlen		= sizeof(int),
-		.mode		= 0644,
-		.proc_handler	= &proc_dointvec_minmax,
-		.strategy	= &sysctl_intvec,
-		.extra1		= &xfs_params.restrict_chown.min,
-		.extra2		= &xfs_params.restrict_chown.max
-	},
 	{
 		.ctl_name	= XFS_SGID_INHERIT,
 		.procname	= "irix_sgid_inherit",

fs/xfs/linux-2.6/xfs_sysctl.h

@@ -31,7 +31,6 @@ typedef struct xfs_sysctl_val {
 } xfs_sysctl_val_t;
 
 typedef struct xfs_param {
-	xfs_sysctl_val_t restrict_chown;/* Root/non-root can give away files.*/
 	xfs_sysctl_val_t sgid_inherit;	/* Inherit S_ISGID if process' GID is
 					 * not a member of parent dir GID. */
 	xfs_sysctl_val_t symlink_mode;	/* Link creat mode affected by umask */
@@ -68,7 +67,7 @@ typedef struct xfs_param {
 enum {
 	/* XFS_REFCACHE_SIZE = 1 */
 	/* XFS_REFCACHE_PURGE = 2 */
-	XFS_RESTRICT_CHOWN = 3,
+	/* XFS_RESTRICT_CHOWN = 3 */
 	XFS_SGID_INHERIT = 4,
 	XFS_SYMLINK_MODE = 5,
 	XFS_PANIC_MASK = 6,

fs/xfs/linux-2.6/xfs_vfs.h

@@ -33,37 +33,6 @@ struct xfs_mount_args;
 
 typedef struct kstatfs	bhv_statvfs_t;
 
-typedef struct bhv_vfs_sync_work {
-	struct list_head	w_list;
-	struct xfs_mount	*w_mount;
-	void			*w_data;	/* syncer routine argument */
-	void			(*w_syncer)(struct xfs_mount *, void *);
-} bhv_vfs_sync_work_t;
-
-#define SYNC_ATTR		0x0001	/* sync attributes */
-#define SYNC_CLOSE		0x0002	/* close file system down */
-#define SYNC_DELWRI		0x0004	/* look at delayed writes */
-#define SYNC_WAIT		0x0008	/* wait for i/o to complete */
-#define SYNC_BDFLUSH		0x0010	/* BDFLUSH is calling -- don't block */
-#define SYNC_FSDATA		0x0020	/* flush fs data (e.g. superblocks) */
-#define SYNC_REFCACHE		0x0040  /* prune some of the nfs ref cache */
-#define SYNC_REMOUNT		0x0080  /* remount readonly, no dummy LRs */
-#define SYNC_IOWAIT		0x0100  /* wait for all I/O to complete */
-
-/*
- * When remounting a filesystem read-only or freezing the filesystem,
- * we have two phases to execute. This first phase is syncing the data
- * before we quiesce the fielsystem, and the second is flushing all the
- * inodes out after we've waited for all the transactions created by
- * the first phase to complete. The second phase uses SYNC_INODE_QUIESCE
- * to ensure that the inodes are written to their location on disk
- * rather than just existing in transactions in the log. This means
- * after a quiesce there is no log replay required to write the inodes
- * to disk (this is the main difference between a sync and a quiesce).
- */
-#define SYNC_DATA_QUIESCE	(SYNC_DELWRI|SYNC_FSDATA|SYNC_WAIT|SYNC_IOWAIT)
-#define SYNC_INODE_QUIESCE	(SYNC_REMOUNT|SYNC_ATTR|SYNC_WAIT)
-
 #define SHUTDOWN_META_IO_ERROR	0x0001	/* write attempt to metadata failed */
 #define SHUTDOWN_LOG_IO_ERROR	0x0002	/* write attempt to the log failed */
 #define SHUTDOWN_FORCE_UMOUNT	0x0004	/* shutdown from a forced unmount */

fs/xfs/linux-2.6/xfs_vnode.c

@@ -84,25 +84,12 @@ vn_ioerror(
 
 #ifdef	XFS_INODE_TRACE
 
-/*
- * Reference count of Linux inode if present, -1 if the xfs_inode
- * has no associated Linux inode.
- */
-static inline int xfs_icount(struct xfs_inode *ip)
-{
-	struct inode *vp = VFS_I(ip);
-
-	if (vp)
-		return vn_count(vp);
-	return -1;
-}
-
 #define KTRACE_ENTER(ip, vk, s, line, ra)			\
 	ktrace_enter(	(ip)->i_trace,				\
 /*  0 */		(void *)(__psint_t)(vk),		\
 /*  1 */		(void *)(s),				\
 /*  2 */		(void *)(__psint_t) line,		\
-/*  3 */		(void *)(__psint_t)xfs_icount(ip),	\
+/*  3 */		(void *)(__psint_t)atomic_read(&VFS_I(ip)->i_count), \
 /*  4 */		(void *)(ra),				\
 /*  5 */		NULL,					\
 /*  6 */		(void *)(__psint_t)current_cpu(),	\

fs/xfs/linux-2.6/xfs_vnode.h

@@ -80,11 +80,6 @@ do { \
 	iput(VFS_I(ip)); \
 } while (0)
 
-static inline struct inode *vn_grab(struct inode *vp)
-{
-	return igrab(vp);
-}
-
 /*
  * Dealing with bad inodes
  */

fs/xfs/quota/xfs_dquot.c

@@ -101,7 +101,7 @@ xfs_qm_dqinit(
 	if (brandnewdquot) {
 		dqp->dq_flnext = dqp->dq_flprev = dqp;
 		mutex_init(&dqp->q_qlock);
-		sv_init(&dqp->q_pinwait, SV_DEFAULT, "pdq");
+		init_waitqueue_head(&dqp->q_pinwait);
 
 		/*
 		 * Because we want to use a counting completion, complete
@@ -131,7 +131,7 @@ xfs_qm_dqinit(
 		 dqp->q_res_bcount = 0;
 		 dqp->q_res_icount = 0;
 		 dqp->q_res_rtbcount = 0;
-		 dqp->q_pincount = 0;
+		 atomic_set(&dqp->q_pincount, 0);
 		 dqp->q_hash = NULL;
 		 ASSERT(dqp->dq_flnext == dqp->dq_flprev);
 
@@ -1221,16 +1221,14 @@ xfs_qm_dqflush(
 	xfs_dqtrace_entry(dqp, "DQFLUSH");
 
 	/*
-	 * If not dirty, nada.
+	 * If not dirty, or it's pinned and we are not supposed to
+	 * block, nada.
 	 */
-	if (!XFS_DQ_IS_DIRTY(dqp)) {
+	if (!XFS_DQ_IS_DIRTY(dqp) ||
+	    (!(flags & XFS_QMOPT_SYNC) && atomic_read(&dqp->q_pincount) > 0)) {
 		xfs_dqfunlock(dqp);
-		return (0);
+		return 0;
 	}
-
-	/*
-	 * Cant flush a pinned dquot. Wait for it.
-	 */
 	xfs_qm_dqunpin_wait(dqp);
 
 	/*
@@ -1274,10 +1272,8 @@ xfs_qm_dqflush(
 	dqp->dq_flags &= ~(XFS_DQ_DIRTY);
 	mp = dqp->q_mount;
 
-	/* lsn is 64 bits */
-	spin_lock(&mp->m_ail_lock);
-	dqp->q_logitem.qli_flush_lsn = dqp->q_logitem.qli_item.li_lsn;
-	spin_unlock(&mp->m_ail_lock);
+	xfs_trans_ail_copy_lsn(mp->m_ail, &dqp->q_logitem.qli_flush_lsn,
+					&dqp->q_logitem.qli_item.li_lsn);
 
 	/*
 	 * Attach an iodone routine so that we can remove this dquot from the
@@ -1323,8 +1319,10 @@ xfs_qm_dqflush_done(
 	xfs_dq_logitem_t	*qip)
 {
 	xfs_dquot_t		*dqp;
+	struct xfs_ail		*ailp;
 
 	dqp = qip->qli_dquot;
+	ailp = qip->qli_item.li_ailp;
 
 	/*
 	 * We only want to pull the item from the AIL if its
@@ -1337,15 +1335,12 @@ xfs_qm_dqflush_done(
 	if ((qip->qli_item.li_flags & XFS_LI_IN_AIL) &&
 	    qip->qli_item.li_lsn == qip->qli_flush_lsn) {
 
-		spin_lock(&dqp->q_mount->m_ail_lock);
-		/*
-		 * xfs_trans_delete_ail() drops the AIL lock.
-		 */
+		/* xfs_trans_ail_delete() drops the AIL lock. */
+		spin_lock(&ailp->xa_lock);
 		if (qip->qli_item.li_lsn == qip->qli_flush_lsn)
-			xfs_trans_delete_ail(dqp->q_mount,
-					     (xfs_log_item_t*)qip);
+			xfs_trans_ail_delete(ailp, (xfs_log_item_t*)qip);
 		else
-			spin_unlock(&dqp->q_mount->m_ail_lock);
+			spin_unlock(&ailp->xa_lock);
 	}
 
 	/*
@@ -1375,7 +1370,7 @@ xfs_dqunlock(
 	mutex_unlock(&(dqp->q_qlock));
 	if (dqp->q_logitem.qli_dquot == dqp) {
 		/* Once was dqp->q_mount, but might just have been cleared */
-		xfs_trans_unlocked_item(dqp->q_logitem.qli_item.li_mountp,
+		xfs_trans_unlocked_item(dqp->q_logitem.qli_item.li_ailp,
 					(xfs_log_item_t*)&(dqp->q_logitem));
 	}
 }
@@ -1489,7 +1484,7 @@ xfs_qm_dqpurge(
 				"xfs_qm_dqpurge: dquot %p flush failed", dqp);
 		xfs_dqflock(dqp);
 	}
-	ASSERT(dqp->q_pincount == 0);
+	ASSERT(atomic_read(&dqp->q_pincount) == 0);
 	ASSERT(XFS_FORCED_SHUTDOWN(mp) ||
 	       !(dqp->q_logitem.qli_item.li_flags & XFS_LI_IN_AIL));
 

fs/xfs/quota/xfs_dquot.h

@@ -83,8 +83,8 @@ typedef struct xfs_dquot {
 	xfs_qcnt_t	 q_res_rtbcount;/* total realtime blks used+reserved */
 	mutex_t		 q_qlock;	/* quota lock */
 	struct completion q_flush;	/* flush completion queue */
-	uint		 q_pincount;	/* pin count for this dquot */
-	sv_t		 q_pinwait;	/* sync var for pinning */
+	atomic_t          q_pincount;	/* dquot pin count */
+	wait_queue_head_t q_pinwait;	/* dquot pinning wait queue */
 #ifdef XFS_DQUOT_TRACE
 	struct ktrace	*q_trace;	/* trace header structure */
 #endif

fs/xfs/quota/xfs_dquot_item.c

@@ -88,25 +88,22 @@ xfs_qm_dquot_logitem_format(
 
 /*
  * Increment the pin count of the given dquot.
- * This value is protected by pinlock spinlock in the xQM structure.
  */
 STATIC void
 xfs_qm_dquot_logitem_pin(
 	xfs_dq_logitem_t *logitem)
 {
-	xfs_dquot_t *dqp;
+	xfs_dquot_t *dqp = logitem->qli_dquot;
 
-	dqp = logitem->qli_dquot;
 	ASSERT(XFS_DQ_IS_LOCKED(dqp));
-	spin_lock(&(XFS_DQ_TO_QINF(dqp)->qi_pinlock));
-	dqp->q_pincount++;
-	spin_unlock(&(XFS_DQ_TO_QINF(dqp)->qi_pinlock));
+	atomic_inc(&dqp->q_pincount);
 }
 
 /*
  * Decrement the pin count of the given dquot, and wake up
  * anyone in xfs_dqwait_unpin() if the count goes to 0.	 The
- * dquot must have been previously pinned with a call to xfs_dqpin().
+ * dquot must have been previously pinned with a call to
+ * xfs_qm_dquot_logitem_pin().
  */
 /* ARGSUSED */
 STATIC void
@@ -114,16 +111,11 @@ xfs_qm_dquot_logitem_unpin(
 	xfs_dq_logitem_t *logitem,
 	int		  stale)
 {
-	xfs_dquot_t *dqp;
+	xfs_dquot_t *dqp = logitem->qli_dquot;
 
-	dqp = logitem->qli_dquot;
-	ASSERT(dqp->q_pincount > 0);
-	spin_lock(&(XFS_DQ_TO_QINF(dqp)->qi_pinlock));
-	dqp->q_pincount--;
-	if (dqp->q_pincount == 0) {
-		sv_broadcast(&dqp->q_pinwait);
-	}
-	spin_unlock(&(XFS_DQ_TO_QINF(dqp)->qi_pinlock));
+	ASSERT(atomic_read(&dqp->q_pincount) > 0);
+	if (atomic_dec_and_test(&dqp->q_pincount))
+		wake_up(&dqp->q_pinwait);
 }
 
 /* ARGSUSED */
@@ -193,21 +185,14 @@ xfs_qm_dqunpin_wait(
 	xfs_dquot_t	*dqp)
 {
 	ASSERT(XFS_DQ_IS_LOCKED(dqp));
-	if (dqp->q_pincount == 0) {
+	if (atomic_read(&dqp->q_pincount) == 0)
 		return;
-	}
 
 	/*
 	 * Give the log a push so we don't wait here too long.
 	 */
 	xfs_log_force(dqp->q_mount, (xfs_lsn_t)0, XFS_LOG_FORCE);
-	spin_lock(&(XFS_DQ_TO_QINF(dqp)->qi_pinlock));
-	if (dqp->q_pincount == 0) {
-		spin_unlock(&(XFS_DQ_TO_QINF(dqp)->qi_pinlock));
-		return;
-	}
-	sv_wait(&(dqp->q_pinwait), PINOD,
-		&(XFS_DQ_TO_QINF(dqp)->qi_pinlock), s);
+	wait_event(dqp->q_pinwait, (atomic_read(&dqp->q_pincount) == 0));
 }
 
 /*
@@ -310,7 +295,7 @@ xfs_qm_dquot_logitem_trylock(
 	uint			retval;
 
 	dqp = qip->qli_dquot;
-	if (dqp->q_pincount > 0)
+	if (atomic_read(&dqp->q_pincount) > 0)
 		return (XFS_ITEM_PINNED);
 
 	if (! xfs_qm_dqlock_nowait(dqp))
@@ -568,14 +553,16 @@ xfs_qm_qoffend_logitem_committed(
 	xfs_lsn_t lsn)
 {
 	xfs_qoff_logitem_t	*qfs;
+	struct xfs_ail		*ailp;
 
 	qfs = qfe->qql_start_lip;
-	spin_lock(&qfs->qql_item.li_mountp->m_ail_lock);
+	ailp = qfs->qql_item.li_ailp;
+	spin_lock(&ailp->xa_lock);
 	/*
 	 * Delete the qoff-start logitem from the AIL.
-	 * xfs_trans_delete_ail() drops the AIL lock.
+	 * xfs_trans_ail_delete() drops the AIL lock.
 	 */
-	xfs_trans_delete_ail(qfs->qql_item.li_mountp, (xfs_log_item_t *)qfs);
+	xfs_trans_ail_delete(ailp, (xfs_log_item_t *)qfs);
 	kmem_free(qfs);
 	kmem_free(qfe);
 	return (xfs_lsn_t)-1;

fs/xfs/quota/xfs_qm.c

@@ -20,7 +20,6 @@
 #include "xfs_bit.h"
 #include "xfs_log.h"
 #include "xfs_inum.h"
-#include "xfs_clnt.h"
 #include "xfs_trans.h"
 #include "xfs_sb.h"
 #include "xfs_ag.h"
@@ -987,14 +986,10 @@ xfs_qm_dqdetach(
 }
 
 /*
- * This is called by VFS_SYNC and flags arg determines the caller,
- * and its motives, as done in xfs_sync.
- *
- * vfs_sync: SYNC_FSDATA|SYNC_ATTR|SYNC_BDFLUSH 0x31
- * syscall sync: SYNC_FSDATA|SYNC_ATTR|SYNC_DELWRI 0x25
- * umountroot : SYNC_WAIT | SYNC_CLOSE | SYNC_ATTR | SYNC_FSDATA
+ * This is called to sync quotas. We can be told to use non-blocking
+ * semantics by either the SYNC_BDFLUSH flag or the absence of the
+ * SYNC_WAIT flag.
  */
-
 int
 xfs_qm_sync(
 	xfs_mount_t	*mp,
@@ -1137,7 +1132,6 @@ xfs_qm_init_quotainfo(
 		return error;
 	}
 
-	spin_lock_init(&qinf->qi_pinlock);
 	xfs_qm_list_init(&qinf->qi_dqlist, "mpdqlist", 0);
 	qinf->qi_dqreclaims = 0;
 
@@ -1234,7 +1228,6 @@ xfs_qm_destroy_quotainfo(
 	 */
 	xfs_qm_rele_quotafs_ref(mp);
 
-	spinlock_destroy(&qi->qi_pinlock);
 	xfs_qm_list_destroy(&qi->qi_dqlist);
 
 	if (qi->qi_uquotaip) {

fs/xfs/quota/xfs_qm.h

@@ -106,7 +106,6 @@ typedef struct xfs_qm {
 typedef struct xfs_quotainfo {
 	xfs_inode_t	*qi_uquotaip;	 /* user quota inode */
 	xfs_inode_t	*qi_gquotaip;	 /* group quota inode */
-	spinlock_t	 qi_pinlock;	 /* dquot pinning lock */
 	xfs_dqlist_t	 qi_dqlist;	 /* all dquots in filesys */
 	int		 qi_dqreclaims;	 /* a change here indicates
 					    a removal in the dqlist */

fs/xfs/quota/xfs_qm_bhv.c

@@ -20,7 +20,6 @@
 #include "xfs_bit.h"
 #include "xfs_log.h"
 #include "xfs_inum.h"
-#include "xfs_clnt.h"
 #include "xfs_trans.h"
 #include "xfs_sb.h"
 #include "xfs_ag.h"

fs/xfs/quota/xfs_qm_syscalls.c

@@ -127,7 +127,7 @@ xfs_qm_quotactl(
 		break;
 
 	case Q_XQUOTASYNC:
-		return (xfs_sync_inodes(mp, SYNC_DELWRI, NULL));
+		return xfs_sync_inodes(mp, SYNC_DELWRI);
 
 	default:
 		break;
@@ -1022,101 +1022,92 @@ xfs_qm_export_flags(
 
 
 /*
- * Go thru all the inodes in the file system, releasing their dquots.
- * Note that the mount structure gets modified to indicate that quotas are off
- * AFTER this, in the case of quotaoff. This also gets called from
- * xfs_rootumount.
+ * Release all the dquots on the inodes in an AG.
  */
-void
-xfs_qm_dqrele_all_inodes(
-	struct xfs_mount *mp,
-	uint		 flags)
+STATIC void
+xfs_qm_dqrele_inodes_ag(
+	xfs_mount_t	*mp,
+	int		ag,
+	uint		flags)
 {
-	xfs_inode_t	*ip, *topino;
-	uint		ireclaims;
-	struct inode	*vp;
-	boolean_t	vnode_refd;
-
-	ASSERT(mp->m_quotainfo);
+	xfs_inode_t	*ip = NULL;
+	xfs_perag_t	*pag = &mp->m_perag[ag];
+	int		first_index = 0;
+	int		nr_found;
 
-	XFS_MOUNT_ILOCK(mp);
-again:
-	ip = mp->m_inodes;
-	if (ip == NULL) {
-		XFS_MOUNT_IUNLOCK(mp);
-		return;
-	}
 	do {
-		/* Skip markers inserted by xfs_sync */
-		if (ip->i_mount == NULL) {
-			ip = ip->i_mnext;
-			continue;
-		}
-		/* Root inode, rbmip and rsumip have associated blocks */
-		if (ip == XFS_QI_UQIP(mp) || ip == XFS_QI_GQIP(mp)) {
-			ASSERT(ip->i_udquot == NULL);
-			ASSERT(ip->i_gdquot == NULL);
-			ip = ip->i_mnext;
-			continue;
+		boolean_t	inode_refed;
+		struct inode	*inode;
+
+		/*
+		 * use a gang lookup to find the next inode in the tree
+		 * as the tree is sparse and a gang lookup walks to find
+		 * the number of objects requested.
+		 */
+		read_lock(&pag->pag_ici_lock);
+		nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
+				(void**)&ip, first_index, 1);
+
+		if (!nr_found) {
+			read_unlock(&pag->pag_ici_lock);
+			break;
 		}
-		vp = VFS_I(ip);
-		if (!vp) {
+
+		/* update the index for the next lookup */
+		first_index = XFS_INO_TO_AGINO(mp, ip->i_ino + 1);
+
+		/* skip quota inodes and those in reclaim */
+		inode = VFS_I(ip);
+		if (!inode || ip == XFS_QI_UQIP(mp) || ip == XFS_QI_GQIP(mp)) {
 			ASSERT(ip->i_udquot == NULL);
 			ASSERT(ip->i_gdquot == NULL);
-			ip = ip->i_mnext;
+			read_unlock(&pag->pag_ici_lock);
 			continue;
 		}
-		vnode_refd = B_FALSE;
 		if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
-			ireclaims = mp->m_ireclaims;
-			topino = mp->m_inodes;
-			vp = vn_grab(vp);
-			if (!vp)
-				goto again;
-
-			XFS_MOUNT_IUNLOCK(mp);
-			/* XXX restart limit ? */
+			inode = igrab(inode);
+			read_unlock(&pag->pag_ici_lock);
+			if (!inode)
+				continue;
+			inode_refed = B_TRUE;
 			xfs_ilock(ip, XFS_ILOCK_EXCL);
-			vnode_refd = B_TRUE;
 		} else {
-			ireclaims = mp->m_ireclaims;
-			topino = mp->m_inodes;
-			XFS_MOUNT_IUNLOCK(mp);
+			read_unlock(&pag->pag_ici_lock);
 		}
-
-		/*
-		 * We don't keep the mountlock across the dqrele() call,
-		 * since it can take a while..
-		 */
 		if ((flags & XFS_UQUOTA_ACCT) && ip->i_udquot) {
 			xfs_qm_dqrele(ip->i_udquot);
 			ip->i_udquot = NULL;
 		}
-		if (flags & (XFS_PQUOTA_ACCT|XFS_GQUOTA_ACCT) && ip->i_gdquot) {
+		if (flags & (XFS_PQUOTA_ACCT|XFS_GQUOTA_ACCT) &&
+		    ip->i_gdquot) {
 			xfs_qm_dqrele(ip->i_gdquot);
 			ip->i_gdquot = NULL;
 		}
 		xfs_iunlock(ip, XFS_ILOCK_EXCL);
-		/*
-		 * Wait until we've dropped the ilock and mountlock to
-		 * do the vn_rele. Or be condemned to an eternity in the
-		 * inactive code in hell.
-		 */
-		if (vnode_refd)
+		if (inode_refed)
 			IRELE(ip);
-		XFS_MOUNT_ILOCK(mp);
-		/*
-		 * If an inode was inserted or removed, we gotta
-		 * start over again.
-		 */
-		if (topino != mp->m_inodes || mp->m_ireclaims != ireclaims) {
-			/* XXX use a sentinel */
-			goto again;
-		}
-		ip = ip->i_mnext;
-	} while (ip != mp->m_inodes);
+	} while (nr_found);
+}
+
+/*
+ * Go thru all the inodes in the file system, releasing their dquots.
+ * Note that the mount structure gets modified to indicate that quotas are off
+ * AFTER this, in the case of quotaoff. This also gets called from
+ * xfs_rootumount.
+ */
+void
+xfs_qm_dqrele_all_inodes(
+	struct xfs_mount *mp,
+	uint		 flags)
+{
+	int		i;
 
-	XFS_MOUNT_IUNLOCK(mp);
+	ASSERT(mp->m_quotainfo);
+	for (i = 0; i < mp->m_sb.sb_agcount; i++) {
+		if (!mp->m_perag[i].pag_ici_init)
+			continue;
+		xfs_qm_dqrele_inodes_ag(mp, i, flags);
+	}
 }
 
 /*------------------------------------------------------------------------*/

fs/xfs/support/debug.c

@@ -84,5 +84,5 @@ assfail(char *expr, char *file, int line)
 void
 xfs_hex_dump(void *p, int length)
 {
-	print_hex_dump(KERN_ALERT, "", DUMP_PREFIX_OFFSET, 16, 1, p, length, 1);
+	print_hex_dump(KERN_ALERT, "", DUMP_PREFIX_ADDRESS, 16, 1, p, length, 1);
 }

fs/xfs/xfs.h

@@ -30,7 +30,7 @@
 #define XFS_ATTR_TRACE 1
 #define XFS_BLI_TRACE 1
 #define XFS_BMAP_TRACE 1
-#define XFS_BMBT_TRACE 1
+#define XFS_BTREE_TRACE 1
 #define XFS_DIR2_TRACE 1
 #define XFS_DQUOT_TRACE 1
 #define XFS_ILOCK_TRACE 1

fs/xfs/xfs_acl.c

@@ -366,7 +366,7 @@ xfs_acl_allow_set(
 		return ENOTDIR;
 	if (vp->i_sb->s_flags & MS_RDONLY)
 		return EROFS;
-	if (XFS_I(vp)->i_d.di_uid != current->fsuid && !capable(CAP_FOWNER))
+	if (XFS_I(vp)->i_d.di_uid != current_fsuid() && !capable(CAP_FOWNER))
 		return EPERM;
 	return 0;
 }
@@ -413,13 +413,13 @@ xfs_acl_access(
 		switch (fap->acl_entry[i].ae_tag) {
 		case ACL_USER_OBJ:
 			seen_userobj = 1;
-			if (fuid != current->fsuid)
+			if (fuid != current_fsuid())
 				continue;
 			matched.ae_tag = ACL_USER_OBJ;
 			matched.ae_perm = allows;
 			break;
 		case ACL_USER:
-			if (fap->acl_entry[i].ae_id != current->fsuid)
+			if (fap->acl_entry[i].ae_id != current_fsuid())
 				continue;
 			matched.ae_tag = ACL_USER;
 			matched.ae_perm = allows;
@@ -758,7 +758,7 @@ xfs_acl_setmode(
 	if (gap && nomask)
 		iattr.ia_mode |= gap->ae_perm << 3;
 
-	return xfs_setattr(XFS_I(vp), &iattr, 0, sys_cred);
+	return xfs_setattr(XFS_I(vp), &iattr, 0);
 }
 
 /*

fs/xfs/xfs_ag.h

@@ -192,17 +192,23 @@ typedef struct xfs_perag
 	xfs_agino_t	pagi_freecount;	/* number of free inodes */
 	xfs_agino_t	pagi_count;	/* number of allocated inodes */
 	int		pagb_count;	/* pagb slots in use */
+	xfs_perag_busy_t *pagb_list;	/* unstable blocks */
 #ifdef __KERNEL__
 	spinlock_t	pagb_lock;	/* lock for pagb_list */
-#endif
-	xfs_perag_busy_t *pagb_list;	/* unstable blocks */
+
 	atomic_t        pagf_fstrms;    /* # of filestreams active in this AG */
 
 	int		pag_ici_init;	/* incore inode cache initialised */
 	rwlock_t	pag_ici_lock;	/* incore inode lock */
 	struct radix_tree_root pag_ici_root;	/* incore inode cache root */
+#endif
 } xfs_perag_t;
 
+/*
+ * tags for inode radix tree
+ */
+#define XFS_ICI_RECLAIM_TAG	0	/* inode is to be reclaimed */
+
 #define	XFS_AG_MAXLEVELS(mp)		((mp)->m_ag_maxlevels)
 #define	XFS_MIN_FREELIST_RAW(bl,cl,mp)	\
 	(MIN(bl + 1, XFS_AG_MAXLEVELS(mp)) + MIN(cl + 1, XFS_AG_MAXLEVELS(mp)))

fs/xfs/xfs_alloc.c

@@ -89,6 +89,92 @@ STATIC int xfs_alloc_ag_vextent_small(xfs_alloc_arg_t *,
  * Internal functions.
  */
 
+/*
+ * Lookup the record equal to [bno, len] in the btree given by cur.
+ */
+STATIC int				/* error */
+xfs_alloc_lookup_eq(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agblock_t		bno,	/* starting block of extent */
+	xfs_extlen_t		len,	/* length of extent */
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.a.ar_startblock = bno;
+	cur->bc_rec.a.ar_blockcount = len;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
+}
+
+/*
+ * Lookup the first record greater than or equal to [bno, len]
+ * in the btree given by cur.
+ */
+STATIC int				/* error */
+xfs_alloc_lookup_ge(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agblock_t		bno,	/* starting block of extent */
+	xfs_extlen_t		len,	/* length of extent */
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.a.ar_startblock = bno;
+	cur->bc_rec.a.ar_blockcount = len;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
+}
+
+/*
+ * Lookup the first record less than or equal to [bno, len]
+ * in the btree given by cur.
+ */
+STATIC int				/* error */
+xfs_alloc_lookup_le(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agblock_t		bno,	/* starting block of extent */
+	xfs_extlen_t		len,	/* length of extent */
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.a.ar_startblock = bno;
+	cur->bc_rec.a.ar_blockcount = len;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_LE, stat);
+}
+
+/*
+ * Update the record referred to by cur to the value given
+ * by [bno, len].
+ * This either works (return 0) or gets an EFSCORRUPTED error.
+ */
+STATIC int				/* error */
+xfs_alloc_update(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agblock_t		bno,	/* starting block of extent */
+	xfs_extlen_t		len)	/* length of extent */
+{
+	union xfs_btree_rec	rec;
+
+	rec.alloc.ar_startblock = cpu_to_be32(bno);
+	rec.alloc.ar_blockcount = cpu_to_be32(len);
+	return xfs_btree_update(cur, &rec);
+}
+
+/*
+ * Get the data from the pointed-to record.
+ */
+STATIC int				/* error */
+xfs_alloc_get_rec(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agblock_t		*bno,	/* output: starting block of extent */
+	xfs_extlen_t		*len,	/* output: length of extent */
+	int			*stat)	/* output: success/failure */
+{
+	union xfs_btree_rec	*rec;
+	int			error;
+
+	error = xfs_btree_get_rec(cur, &rec, stat);
+	if (!error && *stat == 1) {
+		*bno = be32_to_cpu(rec->alloc.ar_startblock);
+		*len = be32_to_cpu(rec->alloc.ar_blockcount);
+	}
+	return error;
+}
+
 /*
  * Compute aligned version of the found extent.
  * Takes alignment and min length into account.
@@ -294,21 +380,20 @@ xfs_alloc_fixup_trees(
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 1);
 	}
+
 #ifdef DEBUG
-	{
-		xfs_alloc_block_t	*bnoblock;
-		xfs_alloc_block_t	*cntblock;
-
-		if (bno_cur->bc_nlevels == 1 &&
-		    cnt_cur->bc_nlevels == 1) {
-			bnoblock = XFS_BUF_TO_ALLOC_BLOCK(bno_cur->bc_bufs[0]);
-			cntblock = XFS_BUF_TO_ALLOC_BLOCK(cnt_cur->bc_bufs[0]);
-			XFS_WANT_CORRUPTED_RETURN(
-				be16_to_cpu(bnoblock->bb_numrecs) ==
-				be16_to_cpu(cntblock->bb_numrecs));
-		}
+	if (bno_cur->bc_nlevels == 1 && cnt_cur->bc_nlevels == 1) {
+		struct xfs_btree_block	*bnoblock;
+		struct xfs_btree_block	*cntblock;
+
+		bnoblock = XFS_BUF_TO_BLOCK(bno_cur->bc_bufs[0]);
+		cntblock = XFS_BUF_TO_BLOCK(cnt_cur->bc_bufs[0]);
+
+		XFS_WANT_CORRUPTED_RETURN(
+			bnoblock->bb_numrecs == cntblock->bb_numrecs);
 	}
 #endif
+
 	/*
 	 * Deal with all four cases: the allocated record is contained
 	 * within the freespace record, so we can have new freespace
@@ -333,7 +418,7 @@ xfs_alloc_fixup_trees(
 	/*
 	 * Delete the entry from the by-size btree.
 	 */
-	if ((error = xfs_alloc_delete(cnt_cur, &i)))
+	if ((error = xfs_btree_delete(cnt_cur, &i)))
 		return error;
 	XFS_WANT_CORRUPTED_RETURN(i == 1);
 	/*
@@ -343,7 +428,7 @@ xfs_alloc_fixup_trees(
 		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno1, nflen1, &i)))
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 0);
-		if ((error = xfs_alloc_insert(cnt_cur, &i)))
+		if ((error = xfs_btree_insert(cnt_cur, &i)))
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 1);
 	}
@@ -351,7 +436,7 @@ xfs_alloc_fixup_trees(
 		if ((error = xfs_alloc_lookup_eq(cnt_cur, nfbno2, nflen2, &i)))
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 0);
-		if ((error = xfs_alloc_insert(cnt_cur, &i)))
+		if ((error = xfs_btree_insert(cnt_cur, &i)))
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 1);
 	}
@@ -362,7 +447,7 @@ xfs_alloc_fixup_trees(
 		/*
 		 * No remaining freespace, just delete the by-block tree entry.
 		 */
-		if ((error = xfs_alloc_delete(bno_cur, &i)))
+		if ((error = xfs_btree_delete(bno_cur, &i)))
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 1);
 	} else {
@@ -379,7 +464,7 @@ xfs_alloc_fixup_trees(
 		if ((error = xfs_alloc_lookup_eq(bno_cur, nfbno2, nflen2, &i)))
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 0);
-		if ((error = xfs_alloc_insert(bno_cur, &i)))
+		if ((error = xfs_btree_insert(bno_cur, &i)))
 			return error;
 		XFS_WANT_CORRUPTED_RETURN(i == 1);
 	}
@@ -640,8 +725,8 @@ xfs_alloc_ag_vextent_exact(
 	/*
 	 * Allocate/initialize a cursor for the by-number freespace btree.
 	 */
-	bno_cur = xfs_btree_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_BNO, NULL, 0);
+	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
+		args->agno, XFS_BTNUM_BNO);
 	/*
 	 * Lookup bno and minlen in the btree (minlen is irrelevant, really).
 	 * Look for the closest free block <= bno, it must contain bno
@@ -696,8 +781,8 @@ xfs_alloc_ag_vextent_exact(
 	 * We are allocating agbno for rlen [agbno .. end]
 	 * Allocate/initialize a cursor for the by-size btree.
 	 */
-	cnt_cur = xfs_btree_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_CNT, NULL, 0);
+	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
+		args->agno, XFS_BTNUM_CNT);
 	ASSERT(args->agbno + args->len <=
 		be32_to_cpu(XFS_BUF_TO_AGF(args->agbp)->agf_length));
 	if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen,
@@ -759,8 +844,8 @@ xfs_alloc_ag_vextent_near(
 	/*
 	 * Get a cursor for the by-size btree.
 	 */
-	cnt_cur = xfs_btree_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_CNT, NULL, 0);
+	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
+		args->agno, XFS_BTNUM_CNT);
 	ltlen = 0;
 	bno_cur_lt = bno_cur_gt = NULL;
 	/*
@@ -818,7 +903,7 @@ xfs_alloc_ag_vextent_near(
 				XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 				if (ltlen >= args->minlen)
 					break;
-				if ((error = xfs_alloc_increment(cnt_cur, 0, &i)))
+				if ((error = xfs_btree_increment(cnt_cur, 0, &i)))
 					goto error0;
 			} while (i);
 			ASSERT(ltlen >= args->minlen);
@@ -828,7 +913,7 @@ xfs_alloc_ag_vextent_near(
 		i = cnt_cur->bc_ptrs[0];
 		for (j = 1, blen = 0, bdiff = 0;
 		     !error && j && (blen < args->maxlen || bdiff > 0);
-		     error = xfs_alloc_increment(cnt_cur, 0, &j)) {
+		     error = xfs_btree_increment(cnt_cur, 0, &j)) {
 			/*
 			 * For each entry, decide if it's better than
 			 * the previous best entry.
@@ -886,8 +971,8 @@ xfs_alloc_ag_vextent_near(
 		/*
 		 * Set up a cursor for the by-bno tree.
 		 */
-		bno_cur_lt = xfs_btree_init_cursor(args->mp, args->tp,
-			args->agbp, args->agno, XFS_BTNUM_BNO, NULL, 0);
+		bno_cur_lt = xfs_allocbt_init_cursor(args->mp, args->tp,
+			args->agbp, args->agno, XFS_BTNUM_BNO);
 		/*
 		 * Fix up the btree entries.
 		 */
@@ -914,8 +999,8 @@ xfs_alloc_ag_vextent_near(
 	/*
 	 * Allocate and initialize the cursor for the leftward search.
 	 */
-	bno_cur_lt = xfs_btree_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_BNO, NULL, 0);
+	bno_cur_lt = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
+		args->agno, XFS_BTNUM_BNO);
 	/*
 	 * Lookup <= bno to find the leftward search's starting point.
 	 */
@@ -938,7 +1023,7 @@ xfs_alloc_ag_vextent_near(
 	 * Increment the cursor, so we will point at the entry just right
 	 * of the leftward entry if any, or to the leftmost entry.
 	 */
-	if ((error = xfs_alloc_increment(bno_cur_gt, 0, &i)))
+	if ((error = xfs_btree_increment(bno_cur_gt, 0, &i)))
 		goto error0;
 	if (!i) {
 		/*
@@ -961,7 +1046,7 @@ xfs_alloc_ag_vextent_near(
 					args->minlen, &ltbnoa, &ltlena);
 			if (ltlena >= args->minlen)
 				break;
-			if ((error = xfs_alloc_decrement(bno_cur_lt, 0, &i)))
+			if ((error = xfs_btree_decrement(bno_cur_lt, 0, &i)))
 				goto error0;
 			if (!i) {
 				xfs_btree_del_cursor(bno_cur_lt,
@@ -977,7 +1062,7 @@ xfs_alloc_ag_vextent_near(
 					args->minlen, &gtbnoa, &gtlena);
 			if (gtlena >= args->minlen)
 				break;
-			if ((error = xfs_alloc_increment(bno_cur_gt, 0, &i)))
+			if ((error = xfs_btree_increment(bno_cur_gt, 0, &i)))
 				goto error0;
 			if (!i) {
 				xfs_btree_del_cursor(bno_cur_gt,
@@ -1066,7 +1151,7 @@ xfs_alloc_ag_vextent_near(
 					/*
 					 * Fell off the right end.
 					 */
-					if ((error = xfs_alloc_increment(
+					if ((error = xfs_btree_increment(
 							bno_cur_gt, 0, &i)))
 						goto error0;
 					if (!i) {
@@ -1162,7 +1247,7 @@ xfs_alloc_ag_vextent_near(
 					/*
 					 * Fell off the left end.
 					 */
-					if ((error = xfs_alloc_decrement(
+					if ((error = xfs_btree_decrement(
 							bno_cur_lt, 0, &i)))
 						goto error0;
 					if (!i) {
@@ -1267,8 +1352,8 @@ xfs_alloc_ag_vextent_size(
 	/*
 	 * Allocate and initialize a cursor for the by-size btree.
 	 */
-	cnt_cur = xfs_btree_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_CNT, NULL, 0);
+	cnt_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
+		args->agno, XFS_BTNUM_CNT);
 	bno_cur = NULL;
 	/*
 	 * Look for an entry >= maxlen+alignment-1 blocks.
@@ -1321,7 +1406,7 @@ xfs_alloc_ag_vextent_size(
 		bestflen = flen;
 		bestfbno = fbno;
 		for (;;) {
-			if ((error = xfs_alloc_decrement(cnt_cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cnt_cur, 0, &i)))
 				goto error0;
 			if (i == 0)
 				break;
@@ -1372,8 +1457,8 @@ xfs_alloc_ag_vextent_size(
 	/*
 	 * Allocate and initialize a cursor for the by-block tree.
 	 */
-	bno_cur = xfs_btree_init_cursor(args->mp, args->tp, args->agbp,
-		args->agno, XFS_BTNUM_BNO, NULL, 0);
+	bno_cur = xfs_allocbt_init_cursor(args->mp, args->tp, args->agbp,
+		args->agno, XFS_BTNUM_BNO);
 	if ((error = xfs_alloc_fixup_trees(cnt_cur, bno_cur, fbno, flen,
 			rbno, rlen, XFSA_FIXUP_CNT_OK)))
 		goto error0;
@@ -1416,7 +1501,7 @@ xfs_alloc_ag_vextent_small(
 	xfs_extlen_t	flen;
 	int		i;
 
-	if ((error = xfs_alloc_decrement(ccur, 0, &i)))
+	if ((error = xfs_btree_decrement(ccur, 0, &i)))
 		goto error0;
 	if (i) {
 		if ((error = xfs_alloc_get_rec(ccur, &fbno, &flen, &i)))
@@ -1515,8 +1600,7 @@ xfs_free_ag_extent(
 	/*
 	 * Allocate and initialize a cursor for the by-block btree.
 	 */
-	bno_cur = xfs_btree_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_BNO, NULL,
-		0);
+	bno_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_BNO);
 	cnt_cur = NULL;
 	/*
 	 * Look for a neighboring block on the left (lower block numbers)
@@ -1549,7 +1633,7 @@ xfs_free_ag_extent(
 	 * Look for a neighboring block on the right (higher block numbers)
 	 * that is contiguous with this space.
 	 */
-	if ((error = xfs_alloc_increment(bno_cur, 0, &haveright)))
+	if ((error = xfs_btree_increment(bno_cur, 0, &haveright)))
 		goto error0;
 	if (haveright) {
 		/*
@@ -1575,8 +1659,7 @@ xfs_free_ag_extent(
 	/*
 	 * Now allocate and initialize a cursor for the by-size tree.
 	 */
-	cnt_cur = xfs_btree_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_CNT, NULL,
-		0);
+	cnt_cur = xfs_allocbt_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_CNT);
 	/*
 	 * Have both left and right contiguous neighbors.
 	 * Merge all three into a single free block.
@@ -1588,7 +1671,7 @@ xfs_free_ag_extent(
 		if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		if ((error = xfs_alloc_delete(cnt_cur, &i)))
+		if ((error = xfs_btree_delete(cnt_cur, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 		/*
@@ -1597,19 +1680,19 @@ xfs_free_ag_extent(
 		if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		if ((error = xfs_alloc_delete(cnt_cur, &i)))
+		if ((error = xfs_btree_delete(cnt_cur, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 		/*
 		 * Delete the old by-block entry for the right block.
 		 */
-		if ((error = xfs_alloc_delete(bno_cur, &i)))
+		if ((error = xfs_btree_delete(bno_cur, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 		/*
 		 * Move the by-block cursor back to the left neighbor.
 		 */
-		if ((error = xfs_alloc_decrement(bno_cur, 0, &i)))
+		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 #ifdef DEBUG
@@ -1648,14 +1731,14 @@ xfs_free_ag_extent(
 		if ((error = xfs_alloc_lookup_eq(cnt_cur, ltbno, ltlen, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		if ((error = xfs_alloc_delete(cnt_cur, &i)))
+		if ((error = xfs_btree_delete(cnt_cur, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 		/*
 		 * Back up the by-block cursor to the left neighbor, and
 		 * update its length.
 		 */
-		if ((error = xfs_alloc_decrement(bno_cur, 0, &i)))
+		if ((error = xfs_btree_decrement(bno_cur, 0, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 		nbno = ltbno;
@@ -1674,7 +1757,7 @@ xfs_free_ag_extent(
 		if ((error = xfs_alloc_lookup_eq(cnt_cur, gtbno, gtlen, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		if ((error = xfs_alloc_delete(cnt_cur, &i)))
+		if ((error = xfs_btree_delete(cnt_cur, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 		/*
@@ -1693,7 +1776,7 @@ xfs_free_ag_extent(
 	else {
 		nbno = bno;
 		nlen = len;
-		if ((error = xfs_alloc_insert(bno_cur, &i)))
+		if ((error = xfs_btree_insert(bno_cur, &i)))
 			goto error0;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 	}
@@ -1705,7 +1788,7 @@ xfs_free_ag_extent(
 	if ((error = xfs_alloc_lookup_eq(cnt_cur, nbno, nlen, &i)))
 		goto error0;
 	XFS_WANT_CORRUPTED_GOTO(i == 0, error0);
-	if ((error = xfs_alloc_insert(cnt_cur, &i)))
+	if ((error = xfs_btree_insert(cnt_cur, &i)))
 		goto error0;
 	XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 	xfs_btree_del_cursor(cnt_cur, XFS_BTREE_NOERROR);
@@ -2188,6 +2271,9 @@ xfs_alloc_read_agf(
 		be32_to_cpu(agf->agf_flfirst) < XFS_AGFL_SIZE(mp) &&
 		be32_to_cpu(agf->agf_fllast) < XFS_AGFL_SIZE(mp) &&
 		be32_to_cpu(agf->agf_flcount) <= XFS_AGFL_SIZE(mp);
+	if (xfs_sb_version_haslazysbcount(&mp->m_sb))
+		agf_ok = agf_ok && be32_to_cpu(agf->agf_btreeblks) <=
+						be32_to_cpu(agf->agf_length);
 	if (unlikely(XFS_TEST_ERROR(!agf_ok, mp, XFS_ERRTAG_ALLOC_READ_AGF,
 			XFS_RANDOM_ALLOC_READ_AGF))) {
 		XFS_CORRUPTION_ERROR("xfs_alloc_read_agf",
@@ -2213,6 +2299,7 @@ xfs_alloc_read_agf(
 #ifdef DEBUG
 	else if (!XFS_FORCED_SHUTDOWN(mp)) {
 		ASSERT(pag->pagf_freeblks == be32_to_cpu(agf->agf_freeblks));
+		ASSERT(pag->pagf_btreeblks == be32_to_cpu(agf->agf_btreeblks));
 		ASSERT(pag->pagf_flcount == be32_to_cpu(agf->agf_flcount));
 		ASSERT(pag->pagf_longest == be32_to_cpu(agf->agf_longest));
 		ASSERT(pag->pagf_levels[XFS_BTNUM_BNOi] ==

fs/xfs/xfs_alloc.h

@@ -121,6 +121,19 @@ extern ktrace_t *xfs_alloc_trace_buf;
 #define	XFS_ALLOC_KTRACE_BUSYSEARCH	6
 #endif
 
+void
+xfs_alloc_mark_busy(xfs_trans_t *tp,
+		xfs_agnumber_t agno,
+		xfs_agblock_t bno,
+		xfs_extlen_t len);
+
+void
+xfs_alloc_clear_busy(xfs_trans_t *tp,
+		xfs_agnumber_t ag,
+		int idx);
+
+#endif	/* __KERNEL__ */
+
 /*
  * Compute and fill in value of m_ag_maxlevels.
  */
@@ -196,18 +209,4 @@ xfs_free_extent(
 	xfs_fsblock_t	bno,	/* starting block number of extent */
 	xfs_extlen_t	len);	/* length of extent */
 
-void
-xfs_alloc_mark_busy(xfs_trans_t *tp,
-		xfs_agnumber_t agno,
-		xfs_agblock_t bno,
-		xfs_extlen_t len);
-
-void
-xfs_alloc_clear_busy(xfs_trans_t *tp,
-		xfs_agnumber_t ag,
-		int idx);
-
-
-#endif	/* __KERNEL__ */
-
 #endif	/* __XFS_ALLOC_H__ */

fs/xfs/xfs_alloc_btree.c

@@ -35,2177 +35,464 @@
 #include "xfs_dinode.h"
 #include "xfs_inode.h"
 #include "xfs_btree.h"
+#include "xfs_btree_trace.h"
 #include "xfs_ialloc.h"
 #include "xfs_alloc.h"
 #include "xfs_error.h"
 
-/*
- * Prototypes for internal functions.
- */
-
-STATIC void xfs_alloc_log_block(xfs_trans_t *, xfs_buf_t *, int);
-STATIC void xfs_alloc_log_keys(xfs_btree_cur_t *, xfs_buf_t *, int, int);
-STATIC void xfs_alloc_log_ptrs(xfs_btree_cur_t *, xfs_buf_t *, int, int);
-STATIC void xfs_alloc_log_recs(xfs_btree_cur_t *, xfs_buf_t *, int, int);
-STATIC int xfs_alloc_lshift(xfs_btree_cur_t *, int, int *);
-STATIC int xfs_alloc_newroot(xfs_btree_cur_t *, int *);
-STATIC int xfs_alloc_rshift(xfs_btree_cur_t *, int, int *);
-STATIC int xfs_alloc_split(xfs_btree_cur_t *, int, xfs_agblock_t *,
-		xfs_alloc_key_t *, xfs_btree_cur_t **, int *);
-STATIC int xfs_alloc_updkey(xfs_btree_cur_t *, xfs_alloc_key_t *, int);
-
-/*
- * Internal functions.
- */
-
-/*
- * Single level of the xfs_alloc_delete record deletion routine.
- * Delete record pointed to by cur/level.
- * Remove the record from its block then rebalance the tree.
- * Return 0 for error, 1 for done, 2 to go on to the next level.
- */
-STATIC int				/* error */
-xfs_alloc_delrec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level removing record from */
-	int			*stat)	/* fail/done/go-on */
-{
-	xfs_agf_t		*agf;	/* allocation group freelist header */
-	xfs_alloc_block_t	*block;	/* btree block record/key lives in */
-	xfs_agblock_t		bno;	/* btree block number */
-	xfs_buf_t		*bp;	/* buffer for block */
-	int			error;	/* error return value */
-	int			i;	/* loop index */
-	xfs_alloc_key_t		key;	/* kp points here if block is level 0 */
-	xfs_agblock_t		lbno;	/* left block's block number */
-	xfs_buf_t		*lbp;	/* left block's buffer pointer */
-	xfs_alloc_block_t	*left;	/* left btree block */
-	xfs_alloc_key_t		*lkp=NULL;	/* left block key pointer */
-	xfs_alloc_ptr_t		*lpp=NULL;	/* left block address pointer */
-	int			lrecs=0;	/* number of records in left block */
-	xfs_alloc_rec_t		*lrp;	/* left block record pointer */
-	xfs_mount_t		*mp;	/* mount structure */
-	int			ptr;	/* index in btree block for this rec */
-	xfs_agblock_t		rbno;	/* right block's block number */
-	xfs_buf_t		*rbp;	/* right block's buffer pointer */
-	xfs_alloc_block_t	*right;	/* right btree block */
-	xfs_alloc_key_t		*rkp;	/* right block key pointer */
-	xfs_alloc_ptr_t		*rpp;	/* right block address pointer */
-	int			rrecs=0;	/* number of records in right block */
-	int			numrecs;
-	xfs_alloc_rec_t		*rrp;	/* right block record pointer */
-	xfs_btree_cur_t		*tcur;	/* temporary btree cursor */
-
-	/*
-	 * Get the index of the entry being deleted, check for nothing there.
-	 */
-	ptr = cur->bc_ptrs[level];
-	if (ptr == 0) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Get the buffer & block containing the record or key/ptr.
-	 */
-	bp = cur->bc_bufs[level];
-	block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
-		return error;
-#endif
-	/*
-	 * Fail if we're off the end of the block.
-	 */
-	numrecs = be16_to_cpu(block->bb_numrecs);
-	if (ptr > numrecs) {
-		*stat = 0;
-		return 0;
-	}
-	XFS_STATS_INC(xs_abt_delrec);
-	/*
-	 * It's a nonleaf.  Excise the key and ptr being deleted, by
-	 * sliding the entries past them down one.
-	 * Log the changed areas of the block.
-	 */
-	if (level > 0) {
-		lkp = XFS_ALLOC_KEY_ADDR(block, 1, cur);
-		lpp = XFS_ALLOC_PTR_ADDR(block, 1, cur);
-#ifdef DEBUG
-		for (i = ptr; i < numrecs; i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(lpp[i]), level)))
-				return error;
-		}
-#endif
-		if (ptr < numrecs) {
-			memmove(&lkp[ptr - 1], &lkp[ptr],
-				(numrecs - ptr) * sizeof(*lkp));
-			memmove(&lpp[ptr - 1], &lpp[ptr],
-				(numrecs - ptr) * sizeof(*lpp));
-			xfs_alloc_log_ptrs(cur, bp, ptr, numrecs - 1);
-			xfs_alloc_log_keys(cur, bp, ptr, numrecs - 1);
-		}
-	}
-	/*
-	 * It's a leaf.  Excise the record being deleted, by sliding the
-	 * entries past it down one.  Log the changed areas of the block.
-	 */
-	else {
-		lrp = XFS_ALLOC_REC_ADDR(block, 1, cur);
-		if (ptr < numrecs) {
-			memmove(&lrp[ptr - 1], &lrp[ptr],
-				(numrecs - ptr) * sizeof(*lrp));
-			xfs_alloc_log_recs(cur, bp, ptr, numrecs - 1);
-		}
-		/*
-		 * If it's the first record in the block, we'll need a key
-		 * structure to pass up to the next level (updkey).
-		 */
-		if (ptr == 1) {
-			key.ar_startblock = lrp->ar_startblock;
-			key.ar_blockcount = lrp->ar_blockcount;
-			lkp = &key;
-		}
-	}
-	/*
-	 * Decrement and log the number of entries in the block.
-	 */
-	numrecs--;
-	block->bb_numrecs = cpu_to_be16(numrecs);
-	xfs_alloc_log_block(cur->bc_tp, bp, XFS_BB_NUMRECS);
-	/*
-	 * See if the longest free extent in the allocation group was
-	 * changed by this operation.  True if it's the by-size btree, and
-	 * this is the leaf level, and there is no right sibling block,
-	 * and this was the last record.
-	 */
-	agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
-	mp = cur->bc_mp;
-
-	if (level == 0 &&
-	    cur->bc_btnum == XFS_BTNUM_CNT &&
-	    be32_to_cpu(block->bb_rightsib) == NULLAGBLOCK &&
-	    ptr > numrecs) {
-		ASSERT(ptr == numrecs + 1);
-		/*
-		 * There are still records in the block.  Grab the size
-		 * from the last one.
-		 */
-		if (numrecs) {
-			rrp = XFS_ALLOC_REC_ADDR(block, numrecs, cur);
-			agf->agf_longest = rrp->ar_blockcount;
-		}
-		/*
-		 * No free extents left.
-		 */
-		else
-			agf->agf_longest = 0;
-		mp->m_perag[be32_to_cpu(agf->agf_seqno)].pagf_longest =
-			be32_to_cpu(agf->agf_longest);
-		xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp,
-			XFS_AGF_LONGEST);
-	}
-	/*
-	 * Is this the root level?  If so, we're almost done.
-	 */
-	if (level == cur->bc_nlevels - 1) {
-		/*
-		 * If this is the root level,
-		 * and there's only one entry left,
-		 * and it's NOT the leaf level,
-		 * then we can get rid of this level.
-		 */
-		if (numrecs == 1 && level > 0) {
-			/*
-			 * lpp is still set to the first pointer in the block.
-			 * Make it the new root of the btree.
-			 */
-			bno = be32_to_cpu(agf->agf_roots[cur->bc_btnum]);
-			agf->agf_roots[cur->bc_btnum] = *lpp;
-			be32_add_cpu(&agf->agf_levels[cur->bc_btnum], -1);
-			mp->m_perag[be32_to_cpu(agf->agf_seqno)].pagf_levels[cur->bc_btnum]--;
-			/*
-			 * Put this buffer/block on the ag's freelist.
-			 */
-			error = xfs_alloc_put_freelist(cur->bc_tp,
-					cur->bc_private.a.agbp, NULL, bno, 1);
-			if (error)
-				return error;
-			/*
-			 * Since blocks move to the free list without the
-			 * coordination used in xfs_bmap_finish, we can't allow
-			 * block to be available for reallocation and
-			 * non-transaction writing (user data) until we know
-			 * that the transaction that moved it to the free list
-			 * is permanently on disk. We track the blocks by
-			 * declaring these blocks as "busy"; the busy list is
-			 * maintained on a per-ag basis and each transaction
-			 * records which entries should be removed when the
-			 * iclog commits to disk. If a busy block is
-			 * allocated, the iclog is pushed up to the LSN
-			 * that freed the block.
-			 */
-			xfs_alloc_mark_busy(cur->bc_tp,
-				be32_to_cpu(agf->agf_seqno), bno, 1);
-
-			xfs_trans_agbtree_delta(cur->bc_tp, -1);
-			xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp,
-				XFS_AGF_ROOTS | XFS_AGF_LEVELS);
-			/*
-			 * Update the cursor so there's one fewer level.
-			 */
-			xfs_btree_setbuf(cur, level, NULL);
-			cur->bc_nlevels--;
-		} else if (level > 0 &&
-			   (error = xfs_alloc_decrement(cur, level, &i)))
-			return error;
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * If we deleted the leftmost entry in the block, update the
-	 * key values above us in the tree.
-	 */
-	if (ptr == 1 && (error = xfs_alloc_updkey(cur, lkp, level + 1)))
-		return error;
-	/*
-	 * If the number of records remaining in the block is at least
-	 * the minimum, we're done.
-	 */
-	if (numrecs >= XFS_ALLOC_BLOCK_MINRECS(level, cur)) {
-		if (level > 0 && (error = xfs_alloc_decrement(cur, level, &i)))
-			return error;
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * Otherwise, we have to move some records around to keep the
-	 * tree balanced.  Look at the left and right sibling blocks to
-	 * see if we can re-balance by moving only one record.
-	 */
-	rbno = be32_to_cpu(block->bb_rightsib);
-	lbno = be32_to_cpu(block->bb_leftsib);
-	bno = NULLAGBLOCK;
-	ASSERT(rbno != NULLAGBLOCK || lbno != NULLAGBLOCK);
-	/*
-	 * Duplicate the cursor so our btree manipulations here won't
-	 * disrupt the next level up.
-	 */
-	if ((error = xfs_btree_dup_cursor(cur, &tcur)))
-		return error;
-	/*
-	 * If there's a right sibling, see if it's ok to shift an entry
-	 * out of it.
-	 */
-	if (rbno != NULLAGBLOCK) {
-		/*
-		 * Move the temp cursor to the last entry in the next block.
-		 * Actually any entry but the first would suffice.
-		 */
-		i = xfs_btree_lastrec(tcur, level);
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		if ((error = xfs_alloc_increment(tcur, level, &i)))
-			goto error0;
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		i = xfs_btree_lastrec(tcur, level);
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		/*
-		 * Grab a pointer to the block.
-		 */
-		rbp = tcur->bc_bufs[level];
-		right = XFS_BUF_TO_ALLOC_BLOCK(rbp);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
-			goto error0;
-#endif
-		/*
-		 * Grab the current block number, for future use.
-		 */
-		bno = be32_to_cpu(right->bb_leftsib);
-		/*
-		 * If right block is full enough so that removing one entry
-		 * won't make it too empty, and left-shifting an entry out
-		 * of right to us works, we're done.
-		 */
-		if (be16_to_cpu(right->bb_numrecs) - 1 >=
-		     XFS_ALLOC_BLOCK_MINRECS(level, cur)) {
-			if ((error = xfs_alloc_lshift(tcur, level, &i)))
-				goto error0;
-			if (i) {
-				ASSERT(be16_to_cpu(block->bb_numrecs) >=
-				       XFS_ALLOC_BLOCK_MINRECS(level, cur));
-				xfs_btree_del_cursor(tcur,
-						     XFS_BTREE_NOERROR);
-				if (level > 0 &&
-				    (error = xfs_alloc_decrement(cur, level,
-					    &i)))
-					return error;
-				*stat = 1;
-				return 0;
-			}
-		}
-		/*
-		 * Otherwise, grab the number of records in right for
-		 * future reference, and fix up the temp cursor to point
-		 * to our block again (last record).
-		 */
-		rrecs = be16_to_cpu(right->bb_numrecs);
-		if (lbno != NULLAGBLOCK) {
-			i = xfs_btree_firstrec(tcur, level);
-			XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-			if ((error = xfs_alloc_decrement(tcur, level, &i)))
-				goto error0;
-			XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		}
-	}
-	/*
-	 * If there's a left sibling, see if it's ok to shift an entry
-	 * out of it.
-	 */
-	if (lbno != NULLAGBLOCK) {
-		/*
-		 * Move the temp cursor to the first entry in the
-		 * previous block.
-		 */
-		i = xfs_btree_firstrec(tcur, level);
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		if ((error = xfs_alloc_decrement(tcur, level, &i)))
-			goto error0;
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		xfs_btree_firstrec(tcur, level);
-		/*
-		 * Grab a pointer to the block.
-		 */
-		lbp = tcur->bc_bufs[level];
-		left = XFS_BUF_TO_ALLOC_BLOCK(lbp);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-			goto error0;
-#endif
-		/*
-		 * Grab the current block number, for future use.
-		 */
-		bno = be32_to_cpu(left->bb_rightsib);
-		/*
-		 * If left block is full enough so that removing one entry
-		 * won't make it too empty, and right-shifting an entry out
-		 * of left to us works, we're done.
-		 */
-		if (be16_to_cpu(left->bb_numrecs) - 1 >=
-		     XFS_ALLOC_BLOCK_MINRECS(level, cur)) {
-			if ((error = xfs_alloc_rshift(tcur, level, &i)))
-				goto error0;
-			if (i) {
-				ASSERT(be16_to_cpu(block->bb_numrecs) >=
-				       XFS_ALLOC_BLOCK_MINRECS(level, cur));
-				xfs_btree_del_cursor(tcur,
-						     XFS_BTREE_NOERROR);
-				if (level == 0)
-					cur->bc_ptrs[0]++;
-				*stat = 1;
-				return 0;
-			}
-		}
-		/*
-		 * Otherwise, grab the number of records in right for
-		 * future reference.
-		 */
-		lrecs = be16_to_cpu(left->bb_numrecs);
-	}
-	/*
-	 * Delete the temp cursor, we're done with it.
-	 */
-	xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
-	/*
-	 * If here, we need to do a join to keep the tree balanced.
-	 */
-	ASSERT(bno != NULLAGBLOCK);
-	/*
-	 * See if we can join with the left neighbor block.
-	 */
-	if (lbno != NULLAGBLOCK &&
-	    lrecs + numrecs <= XFS_ALLOC_BLOCK_MAXRECS(level, cur)) {
-		/*
-		 * Set "right" to be the starting block,
-		 * "left" to be the left neighbor.
-		 */
-		rbno = bno;
-		right = block;
-		rrecs = be16_to_cpu(right->bb_numrecs);
-		rbp = bp;
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, lbno, 0, &lbp,
-				XFS_ALLOC_BTREE_REF)))
-			return error;
-		left = XFS_BUF_TO_ALLOC_BLOCK(lbp);
-		lrecs = be16_to_cpu(left->bb_numrecs);
-		if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-			return error;
-	}
-	/*
-	 * If that won't work, see if we can join with the right neighbor block.
-	 */
-	else if (rbno != NULLAGBLOCK &&
-		 rrecs + numrecs <= XFS_ALLOC_BLOCK_MAXRECS(level, cur)) {
-		/*
-		 * Set "left" to be the starting block,
-		 * "right" to be the right neighbor.
-		 */
-		lbno = bno;
-		left = block;
-		lrecs = be16_to_cpu(left->bb_numrecs);
-		lbp = bp;
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, rbno, 0, &rbp,
-				XFS_ALLOC_BTREE_REF)))
-			return error;
-		right = XFS_BUF_TO_ALLOC_BLOCK(rbp);
-		rrecs = be16_to_cpu(right->bb_numrecs);
-		if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
-			return error;
-	}
-	/*
-	 * Otherwise, we can't fix the imbalance.
-	 * Just return.  This is probably a logic error, but it's not fatal.
-	 */
-	else {
-		if (level > 0 && (error = xfs_alloc_decrement(cur, level, &i)))
-			return error;
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * We're now going to join "left" and "right" by moving all the stuff
-	 * in "right" to "left" and deleting "right".
-	 */
-	if (level > 0) {
-		/*
-		 * It's a non-leaf.  Move keys and pointers.
-		 */
-		lkp = XFS_ALLOC_KEY_ADDR(left, lrecs + 1, cur);
-		lpp = XFS_ALLOC_PTR_ADDR(left, lrecs + 1, cur);
-		rkp = XFS_ALLOC_KEY_ADDR(right, 1, cur);
-		rpp = XFS_ALLOC_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		for (i = 0; i < rrecs; i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(rpp[i]), level)))
-				return error;
-		}
-#endif
-		memcpy(lkp, rkp, rrecs * sizeof(*lkp));
-		memcpy(lpp, rpp, rrecs * sizeof(*lpp));
-		xfs_alloc_log_keys(cur, lbp, lrecs + 1, lrecs + rrecs);
-		xfs_alloc_log_ptrs(cur, lbp, lrecs + 1, lrecs + rrecs);
-	} else {
-		/*
-		 * It's a leaf.  Move records.
-		 */
-		lrp = XFS_ALLOC_REC_ADDR(left, lrecs + 1, cur);
-		rrp = XFS_ALLOC_REC_ADDR(right, 1, cur);
-		memcpy(lrp, rrp, rrecs * sizeof(*lrp));
-		xfs_alloc_log_recs(cur, lbp, lrecs + 1, lrecs + rrecs);
-	}
-	/*
-	 * If we joined with the left neighbor, set the buffer in the
-	 * cursor to the left block, and fix up the index.
-	 */
-	if (bp != lbp) {
-		xfs_btree_setbuf(cur, level, lbp);
-		cur->bc_ptrs[level] += lrecs;
-	}
-	/*
-	 * If we joined with the right neighbor and there's a level above
-	 * us, increment the cursor at that level.
-	 */
-	else if (level + 1 < cur->bc_nlevels &&
-		 (error = xfs_alloc_increment(cur, level + 1, &i)))
-		return error;
-	/*
-	 * Fix up the number of records in the surviving block.
-	 */
-	lrecs += rrecs;
-	left->bb_numrecs = cpu_to_be16(lrecs);
-	/*
-	 * Fix up the right block pointer in the surviving block, and log it.
-	 */
-	left->bb_rightsib = right->bb_rightsib;
-	xfs_alloc_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
-	/*
-	 * If there is a right sibling now, make it point to the
-	 * remaining block.
-	 */
-	if (be32_to_cpu(left->bb_rightsib) != NULLAGBLOCK) {
-		xfs_alloc_block_t	*rrblock;
-		xfs_buf_t		*rrbp;
-
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, be32_to_cpu(left->bb_rightsib), 0,
-				&rrbp, XFS_ALLOC_BTREE_REF)))
-			return error;
-		rrblock = XFS_BUF_TO_ALLOC_BLOCK(rrbp);
-		if ((error = xfs_btree_check_sblock(cur, rrblock, level, rrbp)))
-			return error;
-		rrblock->bb_leftsib = cpu_to_be32(lbno);
-		xfs_alloc_log_block(cur->bc_tp, rrbp, XFS_BB_LEFTSIB);
-	}
-	/*
-	 * Free the deleting block by putting it on the freelist.
-	 */
-	error = xfs_alloc_put_freelist(cur->bc_tp,
-					 cur->bc_private.a.agbp, NULL, rbno, 1);
-	if (error)
-		return error;
-	/*
-	 * Since blocks move to the free list without the coordination
-	 * used in xfs_bmap_finish, we can't allow block to be available
-	 * for reallocation and non-transaction writing (user data)
-	 * until we know that the transaction that moved it to the free
-	 * list is permanently on disk. We track the blocks by declaring
-	 * these blocks as "busy"; the busy list is maintained on a
-	 * per-ag basis and each transaction records which entries
-	 * should be removed when the iclog commits to disk. If a
-	 * busy block is allocated, the iclog is pushed up to the
-	 * LSN that freed the block.
-	 */
-	xfs_alloc_mark_busy(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1);
-	xfs_trans_agbtree_delta(cur->bc_tp, -1);
-
-	/*
-	 * Adjust the current level's cursor so that we're left referring
-	 * to the right node, after we're done.
-	 * If this leaves the ptr value 0 our caller will fix it up.
-	 */
-	if (level > 0)
-		cur->bc_ptrs[level]--;
-	/*
-	 * Return value means the next level up has something to do.
-	 */
-	*stat = 2;
-	return 0;
-
-error0:
-	xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
-	return error;
-}
-
-/*
- * Insert one record/level.  Return information to the caller
- * allowing the next level up to proceed if necessary.
- */
-STATIC int				/* error */
-xfs_alloc_insrec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to insert record at */
-	xfs_agblock_t		*bnop,	/* i/o: block number inserted */
-	xfs_alloc_rec_t		*recp,	/* i/o: record data inserted */
-	xfs_btree_cur_t		**curp,	/* output: new cursor replacing cur */
-	int			*stat)	/* output: success/failure */
-{
-	xfs_agf_t		*agf;	/* allocation group freelist header */
-	xfs_alloc_block_t	*block;	/* btree block record/key lives in */
-	xfs_buf_t		*bp;	/* buffer for block */
-	int			error;	/* error return value */
-	int			i;	/* loop index */
-	xfs_alloc_key_t		key;	/* key value being inserted */
-	xfs_alloc_key_t		*kp;	/* pointer to btree keys */
-	xfs_agblock_t		nbno;	/* block number of allocated block */
-	xfs_btree_cur_t		*ncur;	/* new cursor to be used at next lvl */
-	xfs_alloc_key_t		nkey;	/* new key value, from split */
-	xfs_alloc_rec_t		nrec;	/* new record value, for caller */
-	int			numrecs;
-	int			optr;	/* old ptr value */
-	xfs_alloc_ptr_t		*pp;	/* pointer to btree addresses */
-	int			ptr;	/* index in btree block for this rec */
-	xfs_alloc_rec_t		*rp;	/* pointer to btree records */
-
-	ASSERT(be32_to_cpu(recp->ar_blockcount) > 0);
-
-	/*
-	 * GCC doesn't understand the (arguably complex) control flow in
-	 * this function and complains about uninitialized structure fields
-	 * without this.
-	 */
-	memset(&nrec, 0, sizeof(nrec));
-
-	/*
-	 * If we made it to the root level, allocate a new root block
-	 * and we're done.
-	 */
-	if (level >= cur->bc_nlevels) {
-		XFS_STATS_INC(xs_abt_insrec);
-		if ((error = xfs_alloc_newroot(cur, &i)))
-			return error;
-		*bnop = NULLAGBLOCK;
-		*stat = i;
-		return 0;
-	}
-	/*
-	 * Make a key out of the record data to be inserted, and save it.
-	 */
-	key.ar_startblock = recp->ar_startblock;
-	key.ar_blockcount = recp->ar_blockcount;
-	optr = ptr = cur->bc_ptrs[level];
-	/*
-	 * If we're off the left edge, return failure.
-	 */
-	if (ptr == 0) {
-		*stat = 0;
-		return 0;
-	}
-	XFS_STATS_INC(xs_abt_insrec);
-	/*
-	 * Get pointers to the btree buffer and block.
-	 */
-	bp = cur->bc_bufs[level];
-	block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-	numrecs = be16_to_cpu(block->bb_numrecs);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
-		return error;
-	/*
-	 * Check that the new entry is being inserted in the right place.
-	 */
-	if (ptr <= numrecs) {
-		if (level == 0) {
-			rp = XFS_ALLOC_REC_ADDR(block, ptr, cur);
-			xfs_btree_check_rec(cur->bc_btnum, recp, rp);
-		} else {
-			kp = XFS_ALLOC_KEY_ADDR(block, ptr, cur);
-			xfs_btree_check_key(cur->bc_btnum, &key, kp);
-		}
-	}
-#endif
-	nbno = NULLAGBLOCK;
-	ncur = NULL;
-	/*
-	 * If the block is full, we can't insert the new entry until we
-	 * make the block un-full.
-	 */
-	if (numrecs == XFS_ALLOC_BLOCK_MAXRECS(level, cur)) {
-		/*
-		 * First, try shifting an entry to the right neighbor.
-		 */
-		if ((error = xfs_alloc_rshift(cur, level, &i)))
-			return error;
-		if (i) {
-			/* nothing */
-		}
-		/*
-		 * Next, try shifting an entry to the left neighbor.
-		 */
-		else {
-			if ((error = xfs_alloc_lshift(cur, level, &i)))
-				return error;
-			if (i)
-				optr = ptr = cur->bc_ptrs[level];
-			else {
-				/*
-				 * Next, try splitting the current block in
-				 * half. If this works we have to re-set our
-				 * variables because we could be in a
-				 * different block now.
-				 */
-				if ((error = xfs_alloc_split(cur, level, &nbno,
-						&nkey, &ncur, &i)))
-					return error;
-				if (i) {
-					bp = cur->bc_bufs[level];
-					block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-#ifdef DEBUG
-					if ((error =
-						xfs_btree_check_sblock(cur,
-							block, level, bp)))
-						return error;
-#endif
-					ptr = cur->bc_ptrs[level];
-					nrec.ar_startblock = nkey.ar_startblock;
-					nrec.ar_blockcount = nkey.ar_blockcount;
-				}
-				/*
-				 * Otherwise the insert fails.
-				 */
-				else {
-					*stat = 0;
-					return 0;
-				}
-			}
-		}
-	}
-	/*
-	 * At this point we know there's room for our new entry in the block
-	 * we're pointing at.
-	 */
-	numrecs = be16_to_cpu(block->bb_numrecs);
-	if (level > 0) {
-		/*
-		 * It's a non-leaf entry.  Make a hole for the new data
-		 * in the key and ptr regions of the block.
-		 */
-		kp = XFS_ALLOC_KEY_ADDR(block, 1, cur);
-		pp = XFS_ALLOC_PTR_ADDR(block, 1, cur);
-#ifdef DEBUG
-		for (i = numrecs; i >= ptr; i--) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(pp[i - 1]), level)))
-				return error;
-		}
-#endif
-		memmove(&kp[ptr], &kp[ptr - 1],
-			(numrecs - ptr + 1) * sizeof(*kp));
-		memmove(&pp[ptr], &pp[ptr - 1],
-			(numrecs - ptr + 1) * sizeof(*pp));
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sptr(cur, *bnop, level)))
-			return error;
-#endif
-		/*
-		 * Now stuff the new data in, bump numrecs and log the new data.
-		 */
-		kp[ptr - 1] = key;
-		pp[ptr - 1] = cpu_to_be32(*bnop);
-		numrecs++;
-		block->bb_numrecs = cpu_to_be16(numrecs);
-		xfs_alloc_log_keys(cur, bp, ptr, numrecs);
-		xfs_alloc_log_ptrs(cur, bp, ptr, numrecs);
-#ifdef DEBUG
-		if (ptr < numrecs)
-			xfs_btree_check_key(cur->bc_btnum, kp + ptr - 1,
-				kp + ptr);
-#endif
-	} else {
-		/*
-		 * It's a leaf entry.  Make a hole for the new record.
-		 */
-		rp = XFS_ALLOC_REC_ADDR(block, 1, cur);
-		memmove(&rp[ptr], &rp[ptr - 1],
-			(numrecs - ptr + 1) * sizeof(*rp));
-		/*
-		 * Now stuff the new record in, bump numrecs
-		 * and log the new data.
-		 */
-		rp[ptr - 1] = *recp;
-		numrecs++;
-		block->bb_numrecs = cpu_to_be16(numrecs);
-		xfs_alloc_log_recs(cur, bp, ptr, numrecs);
-#ifdef DEBUG
-		if (ptr < numrecs)
-			xfs_btree_check_rec(cur->bc_btnum, rp + ptr - 1,
-				rp + ptr);
-#endif
-	}
-	/*
-	 * Log the new number of records in the btree header.
-	 */
-	xfs_alloc_log_block(cur->bc_tp, bp, XFS_BB_NUMRECS);
-	/*
-	 * If we inserted at the start of a block, update the parents' keys.
-	 */
-	if (optr == 1 && (error = xfs_alloc_updkey(cur, &key, level + 1)))
-		return error;
-	/*
-	 * Look to see if the longest extent in the allocation group
-	 * needs to be updated.
-	 */
 
-	agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
-	if (level == 0 &&
-	    cur->bc_btnum == XFS_BTNUM_CNT &&
-	    be32_to_cpu(block->bb_rightsib) == NULLAGBLOCK &&
-	    be32_to_cpu(recp->ar_blockcount) > be32_to_cpu(agf->agf_longest)) {
-		/*
-		 * If this is a leaf in the by-size btree and there
-		 * is no right sibling block and this block is bigger
-		 * than the previous longest block, update it.
-		 */
-		agf->agf_longest = recp->ar_blockcount;
-		cur->bc_mp->m_perag[be32_to_cpu(agf->agf_seqno)].pagf_longest
-			= be32_to_cpu(recp->ar_blockcount);
-		xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp,
-			XFS_AGF_LONGEST);
-	}
-	/*
-	 * Return the new block number, if any.
-	 * If there is one, give back a record value and a cursor too.
-	 */
-	*bnop = nbno;
-	if (nbno != NULLAGBLOCK) {
-		*recp = nrec;
-		*curp = ncur;
-	}
-	*stat = 1;
-	return 0;
-}
-
-/*
- * Log header fields from a btree block.
- */
-STATIC void
-xfs_alloc_log_block(
-	xfs_trans_t		*tp,	/* transaction pointer */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			fields)	/* mask of fields: XFS_BB_... */
+STATIC struct xfs_btree_cur *
+xfs_allocbt_dup_cursor(
+	struct xfs_btree_cur	*cur)
 {
-	int			first;	/* first byte offset logged */
-	int			last;	/* last byte offset logged */
-	static const short	offsets[] = {	/* table of offsets */
-		offsetof(xfs_alloc_block_t, bb_magic),
-		offsetof(xfs_alloc_block_t, bb_level),
-		offsetof(xfs_alloc_block_t, bb_numrecs),
-		offsetof(xfs_alloc_block_t, bb_leftsib),
-		offsetof(xfs_alloc_block_t, bb_rightsib),
-		sizeof(xfs_alloc_block_t)
-	};
-
-	xfs_btree_offsets(fields, offsets, XFS_BB_NUM_BITS, &first, &last);
-	xfs_trans_log_buf(tp, bp, first, last);
+	return xfs_allocbt_init_cursor(cur->bc_mp, cur->bc_tp,
+			cur->bc_private.a.agbp, cur->bc_private.a.agno,
+			cur->bc_btnum);
 }
-
-/*
- * Log keys from a btree block (nonleaf).
- */
-STATIC void
-xfs_alloc_log_keys(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			kfirst,	/* index of first key to log */
-	int			klast)	/* index of last key to log */
-{
-	xfs_alloc_block_t	*block;	/* btree block to log from */
-	int			first;	/* first byte offset logged */
-	xfs_alloc_key_t		*kp;	/* key pointer in btree block */
-	int			last;	/* last byte offset logged */
-
-	block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-	kp = XFS_ALLOC_KEY_ADDR(block, 1, cur);
-	first = (int)((xfs_caddr_t)&kp[kfirst - 1] - (xfs_caddr_t)block);
-	last = (int)(((xfs_caddr_t)&kp[klast] - 1) - (xfs_caddr_t)block);
-	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
-}
-
-/*
- * Log block pointer fields from a btree block (nonleaf).
- */
-STATIC void
-xfs_alloc_log_ptrs(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			pfirst,	/* index of first pointer to log */
-	int			plast)	/* index of last pointer to log */
-{
-	xfs_alloc_block_t	*block;	/* btree block to log from */
-	int			first;	/* first byte offset logged */
-	int			last;	/* last byte offset logged */
-	xfs_alloc_ptr_t		*pp;	/* block-pointer pointer in btree blk */
-
-	block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-	pp = XFS_ALLOC_PTR_ADDR(block, 1, cur);
-	first = (int)((xfs_caddr_t)&pp[pfirst - 1] - (xfs_caddr_t)block);
-	last = (int)(((xfs_caddr_t)&pp[plast] - 1) - (xfs_caddr_t)block);
-	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
-}
-
-/*
- * Log records from a btree block (leaf).
- */
-STATIC void
-xfs_alloc_log_recs(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			rfirst,	/* index of first record to log */
-	int			rlast)	/* index of last record to log */
-{
-	xfs_alloc_block_t	*block;	/* btree block to log from */
-	int			first;	/* first byte offset logged */
-	int			last;	/* last byte offset logged */
-	xfs_alloc_rec_t		*rp;	/* record pointer for btree block */
-
-
-	block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-	rp = XFS_ALLOC_REC_ADDR(block, 1, cur);
-#ifdef DEBUG
-	{
-		xfs_agf_t	*agf;
-		xfs_alloc_rec_t	*p;
-
-		agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
-		for (p = &rp[rfirst - 1]; p <= &rp[rlast - 1]; p++)
-			ASSERT(be32_to_cpu(p->ar_startblock) +
-			       be32_to_cpu(p->ar_blockcount) <=
-			       be32_to_cpu(agf->agf_length));
-	}
-#endif
-	first = (int)((xfs_caddr_t)&rp[rfirst - 1] - (xfs_caddr_t)block);
-	last = (int)(((xfs_caddr_t)&rp[rlast] - 1) - (xfs_caddr_t)block);
-	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
-}
-
-/*
- * Lookup the record.  The cursor is made to point to it, based on dir.
- * Return 0 if can't find any such record, 1 for success.
- */
-STATIC int				/* error */
-xfs_alloc_lookup(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_lookup_t		dir,	/* <=, ==, or >= */
-	int			*stat)	/* success/failure */
-{
-	xfs_agblock_t		agbno;	/* a.g. relative btree block number */
-	xfs_agnumber_t		agno;	/* allocation group number */
-	xfs_alloc_block_t	*block=NULL;	/* current btree block */
-	int			diff;	/* difference for the current key */
-	int			error;	/* error return value */
-	int			keyno=0;	/* current key number */
-	int			level;	/* level in the btree */
-	xfs_mount_t		*mp;	/* file system mount point */
-
-	XFS_STATS_INC(xs_abt_lookup);
-	/*
-	 * Get the allocation group header, and the root block number.
-	 */
-	mp = cur->bc_mp;
-
-	{
-		xfs_agf_t	*agf;	/* a.g. freespace header */
-
-		agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
-		agno = be32_to_cpu(agf->agf_seqno);
-		agbno = be32_to_cpu(agf->agf_roots[cur->bc_btnum]);
-	}
-	/*
-	 * Iterate over each level in the btree, starting at the root.
-	 * For each level above the leaves, find the key we need, based
-	 * on the lookup record, then follow the corresponding block
-	 * pointer down to the next level.
-	 */
-	for (level = cur->bc_nlevels - 1, diff = 1; level >= 0; level--) {
-		xfs_buf_t	*bp;	/* buffer pointer for btree block */
-		xfs_daddr_t	d;	/* disk address of btree block */
-
-		/*
-		 * Get the disk address we're looking for.
-		 */
-		d = XFS_AGB_TO_DADDR(mp, agno, agbno);
-		/*
-		 * If the old buffer at this level is for a different block,
-		 * throw it away, otherwise just use it.
-		 */
-		bp = cur->bc_bufs[level];
-		if (bp && XFS_BUF_ADDR(bp) != d)
-			bp = NULL;
-		if (!bp) {
-			/*
-			 * Need to get a new buffer.  Read it, then
-			 * set it in the cursor, releasing the old one.
-			 */
-			if ((error = xfs_btree_read_bufs(mp, cur->bc_tp, agno,
-					agbno, 0, &bp, XFS_ALLOC_BTREE_REF)))
-				return error;
-			xfs_btree_setbuf(cur, level, bp);
-			/*
-			 * Point to the btree block, now that we have the buffer
-			 */
-			block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-			if ((error = xfs_btree_check_sblock(cur, block, level,
-					bp)))
-				return error;
-		} else
-			block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-		/*
-		 * If we already had a key match at a higher level, we know
-		 * we need to use the first entry in this block.
-		 */
-		if (diff == 0)
-			keyno = 1;
-		/*
-		 * Otherwise we need to search this block.  Do a binary search.
-		 */
-		else {
-			int		high;	/* high entry number */
-			xfs_alloc_key_t	*kkbase=NULL;/* base of keys in block */
-			xfs_alloc_rec_t	*krbase=NULL;/* base of records in block */
-			int		low;	/* low entry number */
-
-			/*
-			 * Get a pointer to keys or records.
-			 */
-			if (level > 0)
-				kkbase = XFS_ALLOC_KEY_ADDR(block, 1, cur);
-			else
-				krbase = XFS_ALLOC_REC_ADDR(block, 1, cur);
-			/*
-			 * Set low and high entry numbers, 1-based.
-			 */
-			low = 1;
-			if (!(high = be16_to_cpu(block->bb_numrecs))) {
-				/*
-				 * If the block is empty, the tree must
-				 * be an empty leaf.
-				 */
-				ASSERT(level == 0 && cur->bc_nlevels == 1);
-				cur->bc_ptrs[0] = dir != XFS_LOOKUP_LE;
-				*stat = 0;
-				return 0;
-			}
-			/*
-			 * Binary search the block.
-			 */
-			while (low <= high) {
-				xfs_extlen_t	blockcount;	/* key value */
-				xfs_agblock_t	startblock;	/* key value */
-
-				XFS_STATS_INC(xs_abt_compare);
-				/*
-				 * keyno is average of low and high.
-				 */
-				keyno = (low + high) >> 1;
-				/*
-				 * Get startblock & blockcount.
-				 */
-				if (level > 0) {
-					xfs_alloc_key_t	*kkp;
-
-					kkp = kkbase + keyno - 1;
-					startblock = be32_to_cpu(kkp->ar_startblock);
-					blockcount = be32_to_cpu(kkp->ar_blockcount);
-				} else {
-					xfs_alloc_rec_t	*krp;
-
-					krp = krbase + keyno - 1;
-					startblock = be32_to_cpu(krp->ar_startblock);
-					blockcount = be32_to_cpu(krp->ar_blockcount);
-				}
-				/*
-				 * Compute difference to get next direction.
-				 */
-				if (cur->bc_btnum == XFS_BTNUM_BNO)
-					diff = (int)startblock -
-					       (int)cur->bc_rec.a.ar_startblock;
-				else if (!(diff = (int)blockcount -
-					    (int)cur->bc_rec.a.ar_blockcount))
-					diff = (int)startblock -
-					    (int)cur->bc_rec.a.ar_startblock;
-				/*
-				 * Less than, move right.
-				 */
-				if (diff < 0)
-					low = keyno + 1;
-				/*
-				 * Greater than, move left.
-				 */
-				else if (diff > 0)
-					high = keyno - 1;
-				/*
-				 * Equal, we're done.
-				 */
-				else
-					break;
-			}
-		}
-		/*
-		 * If there are more levels, set up for the next level
-		 * by getting the block number and filling in the cursor.
-		 */
-		if (level > 0) {
-			/*
-			 * If we moved left, need the previous key number,
-			 * unless there isn't one.
-			 */
-			if (diff > 0 && --keyno < 1)
-				keyno = 1;
-			agbno = be32_to_cpu(*XFS_ALLOC_PTR_ADDR(block, keyno, cur));
-#ifdef DEBUG
-			if ((error = xfs_btree_check_sptr(cur, agbno, level)))
-				return error;
-#endif
-			cur->bc_ptrs[level] = keyno;
-		}
-	}
-	/*
-	 * Done with the search.
-	 * See if we need to adjust the results.
-	 */
-	if (dir != XFS_LOOKUP_LE && diff < 0) {
-		keyno++;
-		/*
-		 * If ge search and we went off the end of the block, but it's
-		 * not the last block, we're in the wrong block.
-		 */
-		if (dir == XFS_LOOKUP_GE &&
-		    keyno > be16_to_cpu(block->bb_numrecs) &&
-		    be32_to_cpu(block->bb_rightsib) != NULLAGBLOCK) {
-			int	i;
-
-			cur->bc_ptrs[0] = keyno;
-			if ((error = xfs_alloc_increment(cur, 0, &i)))
-				return error;
-			XFS_WANT_CORRUPTED_RETURN(i == 1);
-			*stat = 1;
-			return 0;
-		}
-	}
-	else if (dir == XFS_LOOKUP_LE && diff > 0)
-		keyno--;
-	cur->bc_ptrs[0] = keyno;
-	/*
-	 * Return if we succeeded or not.
-	 */
-	if (keyno == 0 || keyno > be16_to_cpu(block->bb_numrecs))
-		*stat = 0;
-	else
-		*stat = ((dir != XFS_LOOKUP_EQ) || (diff == 0));
-	return 0;
-}
-
-/*
- * Move 1 record left from cur/level if possible.
- * Update cur to reflect the new path.
- */
-STATIC int				/* error */
-xfs_alloc_lshift(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to shift record on */
-	int			*stat)	/* success/failure */
-{
-	int			error;	/* error return value */
-#ifdef DEBUG
-	int			i;	/* loop index */
-#endif
-	xfs_alloc_key_t		key;	/* key value for leaf level upward */
-	xfs_buf_t		*lbp;	/* buffer for left neighbor block */
-	xfs_alloc_block_t	*left;	/* left neighbor btree block */
-	int			nrec;	/* new number of left block entries */
-	xfs_buf_t		*rbp;	/* buffer for right (current) block */
-	xfs_alloc_block_t	*right;	/* right (current) btree block */
-	xfs_alloc_key_t		*rkp=NULL;	/* key pointer for right block */
-	xfs_alloc_ptr_t		*rpp=NULL;	/* address pointer for right block */
-	xfs_alloc_rec_t		*rrp=NULL;	/* record pointer for right block */
-
-	/*
-	 * Set up variables for this block as "right".
-	 */
-	rbp = cur->bc_bufs[level];
-	right = XFS_BUF_TO_ALLOC_BLOCK(rbp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
-		return error;
-#endif
-	/*
-	 * If we've got no left sibling then we can't shift an entry left.
-	 */
-	if (be32_to_cpu(right->bb_leftsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * If the cursor entry is the one that would be moved, don't
-	 * do it... it's too complicated.
-	 */
-	if (cur->bc_ptrs[level] <= 1) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Set up the left neighbor as "left".
-	 */
-	if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-			cur->bc_private.a.agno, be32_to_cpu(right->bb_leftsib),
-			0, &lbp, XFS_ALLOC_BTREE_REF)))
-		return error;
-	left = XFS_BUF_TO_ALLOC_BLOCK(lbp);
-	if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-		return error;
-	/*
-	 * If it's full, it can't take another entry.
-	 */
-	if (be16_to_cpu(left->bb_numrecs) == XFS_ALLOC_BLOCK_MAXRECS(level, cur)) {
-		*stat = 0;
-		return 0;
-	}
-	nrec = be16_to_cpu(left->bb_numrecs) + 1;
-	/*
-	 * If non-leaf, copy a key and a ptr to the left block.
-	 */
-	if (level > 0) {
-		xfs_alloc_key_t	*lkp;	/* key pointer for left block */
-		xfs_alloc_ptr_t	*lpp;	/* address pointer for left block */
-
-		lkp = XFS_ALLOC_KEY_ADDR(left, nrec, cur);
-		rkp = XFS_ALLOC_KEY_ADDR(right, 1, cur);
-		*lkp = *rkp;
-		xfs_alloc_log_keys(cur, lbp, nrec, nrec);
-		lpp = XFS_ALLOC_PTR_ADDR(left, nrec, cur);
-		rpp = XFS_ALLOC_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(*rpp), level)))
-			return error;
-#endif
-		*lpp = *rpp;
-		xfs_alloc_log_ptrs(cur, lbp, nrec, nrec);
-		xfs_btree_check_key(cur->bc_btnum, lkp - 1, lkp);
-	}
-	/*
-	 * If leaf, copy a record to the left block.
-	 */
-	else {
-		xfs_alloc_rec_t	*lrp;	/* record pointer for left block */
-
-		lrp = XFS_ALLOC_REC_ADDR(left, nrec, cur);
-		rrp = XFS_ALLOC_REC_ADDR(right, 1, cur);
-		*lrp = *rrp;
-		xfs_alloc_log_recs(cur, lbp, nrec, nrec);
-		xfs_btree_check_rec(cur->bc_btnum, lrp - 1, lrp);
-	}
-	/*
-	 * Bump and log left's numrecs, decrement and log right's numrecs.
-	 */
-	be16_add_cpu(&left->bb_numrecs, 1);
-	xfs_alloc_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS);
-	be16_add_cpu(&right->bb_numrecs, -1);
-	xfs_alloc_log_block(cur->bc_tp, rbp, XFS_BB_NUMRECS);
-	/*
-	 * Slide the contents of right down one entry.
-	 */
-	if (level > 0) {
-#ifdef DEBUG
-		for (i = 0; i < be16_to_cpu(right->bb_numrecs); i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(rpp[i + 1]),
-					level)))
-				return error;
-		}
-#endif
-		memmove(rkp, rkp + 1, be16_to_cpu(right->bb_numrecs) * sizeof(*rkp));
-		memmove(rpp, rpp + 1, be16_to_cpu(right->bb_numrecs) * sizeof(*rpp));
-		xfs_alloc_log_keys(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		xfs_alloc_log_ptrs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-	} else {
-		memmove(rrp, rrp + 1, be16_to_cpu(right->bb_numrecs) * sizeof(*rrp));
-		xfs_alloc_log_recs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		key.ar_startblock = rrp->ar_startblock;
-		key.ar_blockcount = rrp->ar_blockcount;
-		rkp = &key;
-	}
-	/*
-	 * Update the parent key values of right.
-	 */
-	if ((error = xfs_alloc_updkey(cur, rkp, level + 1)))
-		return error;
-	/*
-	 * Slide the cursor value left one.
-	 */
-	cur->bc_ptrs[level]--;
-	*stat = 1;
-	return 0;
-}
-
-/*
- * Allocate a new root block, fill it in.
- */
-STATIC int				/* error */
-xfs_alloc_newroot(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			*stat)	/* success/failure */
-{
-	int			error;	/* error return value */
-	xfs_agblock_t		lbno;	/* left block number */
-	xfs_buf_t		*lbp;	/* left btree buffer */
-	xfs_alloc_block_t	*left;	/* left btree block */
-	xfs_mount_t		*mp;	/* mount structure */
-	xfs_agblock_t		nbno;	/* new block number */
-	xfs_buf_t		*nbp;	/* new (root) buffer */
-	xfs_alloc_block_t	*new;	/* new (root) btree block */
-	int			nptr;	/* new value for key index, 1 or 2 */
-	xfs_agblock_t		rbno;	/* right block number */
-	xfs_buf_t		*rbp;	/* right btree buffer */
-	xfs_alloc_block_t	*right;	/* right btree block */
-
-	mp = cur->bc_mp;
-
-	ASSERT(cur->bc_nlevels < XFS_AG_MAXLEVELS(mp));
-	/*
-	 * Get a buffer from the freelist blocks, for the new root.
-	 */
-	error = xfs_alloc_get_freelist(cur->bc_tp,
-					cur->bc_private.a.agbp, &nbno, 1);
-	if (error)
-		return error;
-	/*
-	 * None available, we fail.
-	 */
-	if (nbno == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	xfs_trans_agbtree_delta(cur->bc_tp, 1);
-	nbp = xfs_btree_get_bufs(mp, cur->bc_tp, cur->bc_private.a.agno, nbno,
-		0);
-	new = XFS_BUF_TO_ALLOC_BLOCK(nbp);
-	/*
-	 * Set the root data in the a.g. freespace structure.
-	 */
-	{
-		xfs_agf_t	*agf;	/* a.g. freespace header */
-		xfs_agnumber_t	seqno;
-
-		agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
-		agf->agf_roots[cur->bc_btnum] = cpu_to_be32(nbno);
-		be32_add_cpu(&agf->agf_levels[cur->bc_btnum], 1);
-		seqno = be32_to_cpu(agf->agf_seqno);
-		mp->m_perag[seqno].pagf_levels[cur->bc_btnum]++;
-		xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp,
-			XFS_AGF_ROOTS | XFS_AGF_LEVELS);
-	}
-	/*
-	 * At the previous root level there are now two blocks: the old
-	 * root, and the new block generated when it was split.
-	 * We don't know which one the cursor is pointing at, so we
-	 * set up variables "left" and "right" for each case.
-	 */
-	lbp = cur->bc_bufs[cur->bc_nlevels - 1];
-	left = XFS_BUF_TO_ALLOC_BLOCK(lbp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, left, cur->bc_nlevels - 1, lbp)))
-		return error;
-#endif
-	if (be32_to_cpu(left->bb_rightsib) != NULLAGBLOCK) {
-		/*
-		 * Our block is left, pick up the right block.
-		 */
-		lbno = XFS_DADDR_TO_AGBNO(mp, XFS_BUF_ADDR(lbp));
-		rbno = be32_to_cpu(left->bb_rightsib);
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, rbno, 0, &rbp,
-				XFS_ALLOC_BTREE_REF)))
-			return error;
-		right = XFS_BUF_TO_ALLOC_BLOCK(rbp);
-		if ((error = xfs_btree_check_sblock(cur, right,
-				cur->bc_nlevels - 1, rbp)))
-			return error;
-		nptr = 1;
-	} else {
-		/*
-		 * Our block is right, pick up the left block.
-		 */
-		rbp = lbp;
-		right = left;
-		rbno = XFS_DADDR_TO_AGBNO(mp, XFS_BUF_ADDR(rbp));
-		lbno = be32_to_cpu(right->bb_leftsib);
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, lbno, 0, &lbp,
-				XFS_ALLOC_BTREE_REF)))
-			return error;
-		left = XFS_BUF_TO_ALLOC_BLOCK(lbp);
-		if ((error = xfs_btree_check_sblock(cur, left,
-				cur->bc_nlevels - 1, lbp)))
-			return error;
-		nptr = 2;
-	}
-	/*
-	 * Fill in the new block's btree header and log it.
-	 */
-	new->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
-	new->bb_level = cpu_to_be16(cur->bc_nlevels);
-	new->bb_numrecs = cpu_to_be16(2);
-	new->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-	new->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
-	xfs_alloc_log_block(cur->bc_tp, nbp, XFS_BB_ALL_BITS);
-	ASSERT(lbno != NULLAGBLOCK && rbno != NULLAGBLOCK);
-	/*
-	 * Fill in the key data in the new root.
-	 */
-	{
-		xfs_alloc_key_t		*kp;	/* btree key pointer */
-
-		kp = XFS_ALLOC_KEY_ADDR(new, 1, cur);
-		if (be16_to_cpu(left->bb_level) > 0) {
-			kp[0] = *XFS_ALLOC_KEY_ADDR(left, 1, cur);
-			kp[1] = *XFS_ALLOC_KEY_ADDR(right, 1, cur);
-		} else {
-			xfs_alloc_rec_t	*rp;	/* btree record pointer */
-
-			rp = XFS_ALLOC_REC_ADDR(left, 1, cur);
-			kp[0].ar_startblock = rp->ar_startblock;
-			kp[0].ar_blockcount = rp->ar_blockcount;
-			rp = XFS_ALLOC_REC_ADDR(right, 1, cur);
-			kp[1].ar_startblock = rp->ar_startblock;
-			kp[1].ar_blockcount = rp->ar_blockcount;
-		}
-	}
-	xfs_alloc_log_keys(cur, nbp, 1, 2);
-	/*
-	 * Fill in the pointer data in the new root.
-	 */
-	{
-		xfs_alloc_ptr_t		*pp;	/* btree address pointer */
-
-		pp = XFS_ALLOC_PTR_ADDR(new, 1, cur);
-		pp[0] = cpu_to_be32(lbno);
-		pp[1] = cpu_to_be32(rbno);
-	}
-	xfs_alloc_log_ptrs(cur, nbp, 1, 2);
-	/*
-	 * Fix up the cursor.
-	 */
-	xfs_btree_setbuf(cur, cur->bc_nlevels, nbp);
-	cur->bc_ptrs[cur->bc_nlevels] = nptr;
-	cur->bc_nlevels++;
-	*stat = 1;
-	return 0;
-}
-
-/*
- * Move 1 record right from cur/level if possible.
- * Update cur to reflect the new path.
- */
-STATIC int				/* error */
-xfs_alloc_rshift(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to shift record on */
-	int			*stat)	/* success/failure */
-{
-	int			error;	/* error return value */
-	int			i;	/* loop index */
-	xfs_alloc_key_t		key;	/* key value for leaf level upward */
-	xfs_buf_t		*lbp;	/* buffer for left (current) block */
-	xfs_alloc_block_t	*left;	/* left (current) btree block */
-	xfs_buf_t		*rbp;	/* buffer for right neighbor block */
-	xfs_alloc_block_t	*right;	/* right neighbor btree block */
-	xfs_alloc_key_t		*rkp;	/* key pointer for right block */
-	xfs_btree_cur_t		*tcur;	/* temporary cursor */
-
-	/*
-	 * Set up variables for this block as "left".
-	 */
-	lbp = cur->bc_bufs[level];
-	left = XFS_BUF_TO_ALLOC_BLOCK(lbp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-		return error;
-#endif
-	/*
-	 * If we've got no right sibling then we can't shift an entry right.
-	 */
-	if (be32_to_cpu(left->bb_rightsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * If the cursor entry is the one that would be moved, don't
-	 * do it... it's too complicated.
-	 */
-	if (cur->bc_ptrs[level] >= be16_to_cpu(left->bb_numrecs)) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Set up the right neighbor as "right".
-	 */
-	if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-			cur->bc_private.a.agno, be32_to_cpu(left->bb_rightsib),
-			0, &rbp, XFS_ALLOC_BTREE_REF)))
-		return error;
-	right = XFS_BUF_TO_ALLOC_BLOCK(rbp);
-	if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
+
+STATIC void
+xfs_allocbt_set_root(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr,
+	int			inc)
+{
+	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
+	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
+	int			btnum = cur->bc_btnum;
+
+	ASSERT(ptr->s != 0);
+
+	agf->agf_roots[btnum] = ptr->s;
+	be32_add_cpu(&agf->agf_levels[btnum], inc);
+	cur->bc_mp->m_perag[seqno].pagf_levels[btnum] += inc;
+
+	xfs_alloc_log_agf(cur->bc_tp, agbp, XFS_AGF_ROOTS | XFS_AGF_LEVELS);
+}
+
+STATIC int
+xfs_allocbt_alloc_block(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*start,
+	union xfs_btree_ptr	*new,
+	int			length,
+	int			*stat)
+{
+	int			error;
+	xfs_agblock_t		bno;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+
+	/* Allocate the new block from the freelist. If we can't, give up.  */
+	error = xfs_alloc_get_freelist(cur->bc_tp, cur->bc_private.a.agbp,
+				       &bno, 1);
+	if (error) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 		return error;
-	/*
-	 * If it's full, it can't take another entry.
-	 */
-	if (be16_to_cpu(right->bb_numrecs) == XFS_ALLOC_BLOCK_MAXRECS(level, cur)) {
+	}
+
+	if (bno == NULLAGBLOCK) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 		*stat = 0;
 		return 0;
 	}
-	/*
-	 * Make a hole at the start of the right neighbor block, then
-	 * copy the last left block entry to the hole.
-	 */
-	if (level > 0) {
-		xfs_alloc_key_t	*lkp;	/* key pointer for left block */
-		xfs_alloc_ptr_t	*lpp;	/* address pointer for left block */
-		xfs_alloc_ptr_t	*rpp;	/* address pointer for right block */
 
-		lkp = XFS_ALLOC_KEY_ADDR(left, be16_to_cpu(left->bb_numrecs), cur);
-		lpp = XFS_ALLOC_PTR_ADDR(left, be16_to_cpu(left->bb_numrecs), cur);
-		rkp = XFS_ALLOC_KEY_ADDR(right, 1, cur);
-		rpp = XFS_ALLOC_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		for (i = be16_to_cpu(right->bb_numrecs) - 1; i >= 0; i--) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(rpp[i]), level)))
-				return error;
-		}
-#endif
-		memmove(rkp + 1, rkp, be16_to_cpu(right->bb_numrecs) * sizeof(*rkp));
-		memmove(rpp + 1, rpp, be16_to_cpu(right->bb_numrecs) * sizeof(*rpp));
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(*lpp), level)))
-			return error;
-#endif
-		*rkp = *lkp;
-		*rpp = *lpp;
-		xfs_alloc_log_keys(cur, rbp, 1, be16_to_cpu(right->bb_numrecs) + 1);
-		xfs_alloc_log_ptrs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs) + 1);
-		xfs_btree_check_key(cur->bc_btnum, rkp, rkp + 1);
-	} else {
-		xfs_alloc_rec_t	*lrp;	/* record pointer for left block */
-		xfs_alloc_rec_t	*rrp;	/* record pointer for right block */
+	xfs_trans_agbtree_delta(cur->bc_tp, 1);
+	new->s = cpu_to_be32(bno);
 
-		lrp = XFS_ALLOC_REC_ADDR(left, be16_to_cpu(left->bb_numrecs), cur);
-		rrp = XFS_ALLOC_REC_ADDR(right, 1, cur);
-		memmove(rrp + 1, rrp, be16_to_cpu(right->bb_numrecs) * sizeof(*rrp));
-		*rrp = *lrp;
-		xfs_alloc_log_recs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs) + 1);
-		key.ar_startblock = rrp->ar_startblock;
-		key.ar_blockcount = rrp->ar_blockcount;
-		rkp = &key;
-		xfs_btree_check_rec(cur->bc_btnum, rrp, rrp + 1);
-	}
-	/*
-	 * Decrement and log left's numrecs, bump and log right's numrecs.
-	 */
-	be16_add_cpu(&left->bb_numrecs, -1);
-	xfs_alloc_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS);
-	be16_add_cpu(&right->bb_numrecs, 1);
-	xfs_alloc_log_block(cur->bc_tp, rbp, XFS_BB_NUMRECS);
-	/*
-	 * Using a temporary cursor, update the parent key values of the
-	 * block on the right.
-	 */
-	if ((error = xfs_btree_dup_cursor(cur, &tcur)))
-		return error;
-	i = xfs_btree_lastrec(tcur, level);
-	XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-	if ((error = xfs_alloc_increment(tcur, level, &i)) ||
-	    (error = xfs_alloc_updkey(tcur, rkp, level + 1)))
-		goto error0;
-	xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 	*stat = 1;
 	return 0;
-error0:
-	xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
-	return error;
 }
 
-/*
- * Split cur/level block in half.
- * Return new block number and its first record (to be inserted into parent).
- */
-STATIC int				/* error */
-xfs_alloc_split(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to split */
-	xfs_agblock_t		*bnop,	/* output: block number allocated */
-	xfs_alloc_key_t		*keyp,	/* output: first key of new block */
-	xfs_btree_cur_t		**curp,	/* output: new cursor */
-	int			*stat)	/* success/failure */
+STATIC int
+xfs_allocbt_free_block(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp)
 {
-	int			error;	/* error return value */
-	int			i;	/* loop index/record number */
-	xfs_agblock_t		lbno;	/* left (current) block number */
-	xfs_buf_t		*lbp;	/* buffer for left block */
-	xfs_alloc_block_t	*left;	/* left (current) btree block */
-	xfs_agblock_t		rbno;	/* right (new) block number */
-	xfs_buf_t		*rbp;	/* buffer for right block */
-	xfs_alloc_block_t	*right;	/* right (new) btree block */
+	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
+	xfs_agblock_t		bno;
+	int			error;
 
-	/*
-	 * Allocate the new block from the freelist.
-	 * If we can't do it, we're toast.  Give up.
-	 */
-	error = xfs_alloc_get_freelist(cur->bc_tp,
-					 cur->bc_private.a.agbp, &rbno, 1);
+	bno = XFS_DADDR_TO_AGBNO(cur->bc_mp, XFS_BUF_ADDR(bp));
+	error = xfs_alloc_put_freelist(cur->bc_tp, agbp, NULL, bno, 1);
 	if (error)
 		return error;
-	if (rbno == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	xfs_trans_agbtree_delta(cur->bc_tp, 1);
-	rbp = xfs_btree_get_bufs(cur->bc_mp, cur->bc_tp, cur->bc_private.a.agno,
-		rbno, 0);
-	/*
-	 * Set up the new block as "right".
-	 */
-	right = XFS_BUF_TO_ALLOC_BLOCK(rbp);
-	/*
-	 * "Left" is the current (according to the cursor) block.
-	 */
-	lbp = cur->bc_bufs[level];
-	left = XFS_BUF_TO_ALLOC_BLOCK(lbp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-		return error;
-#endif
-	/*
-	 * Fill in the btree header for the new block.
-	 */
-	right->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
-	right->bb_level = left->bb_level;
-	right->bb_numrecs = cpu_to_be16(be16_to_cpu(left->bb_numrecs) / 2);
-	/*
-	 * Make sure that if there's an odd number of entries now, that
-	 * each new block will have the same number of entries.
-	 */
-	if ((be16_to_cpu(left->bb_numrecs) & 1) &&
-	    cur->bc_ptrs[level] <= be16_to_cpu(right->bb_numrecs) + 1)
-		be16_add_cpu(&right->bb_numrecs, 1);
-	i = be16_to_cpu(left->bb_numrecs) - be16_to_cpu(right->bb_numrecs) + 1;
-	/*
-	 * For non-leaf blocks, copy keys and addresses over to the new block.
-	 */
-	if (level > 0) {
-		xfs_alloc_key_t	*lkp;	/* left btree key pointer */
-		xfs_alloc_ptr_t	*lpp;	/* left btree address pointer */
-		xfs_alloc_key_t	*rkp;	/* right btree key pointer */
-		xfs_alloc_ptr_t	*rpp;	/* right btree address pointer */
-
-		lkp = XFS_ALLOC_KEY_ADDR(left, i, cur);
-		lpp = XFS_ALLOC_PTR_ADDR(left, i, cur);
-		rkp = XFS_ALLOC_KEY_ADDR(right, 1, cur);
-		rpp = XFS_ALLOC_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		for (i = 0; i < be16_to_cpu(right->bb_numrecs); i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(lpp[i]), level)))
-				return error;
-		}
-#endif
-		memcpy(rkp, lkp, be16_to_cpu(right->bb_numrecs) * sizeof(*rkp));
-		memcpy(rpp, lpp, be16_to_cpu(right->bb_numrecs) * sizeof(*rpp));
-		xfs_alloc_log_keys(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		xfs_alloc_log_ptrs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		*keyp = *rkp;
-	}
-	/*
-	 * For leaf blocks, copy records over to the new block.
-	 */
-	else {
-		xfs_alloc_rec_t	*lrp;	/* left btree record pointer */
-		xfs_alloc_rec_t	*rrp;	/* right btree record pointer */
-
-		lrp = XFS_ALLOC_REC_ADDR(left, i, cur);
-		rrp = XFS_ALLOC_REC_ADDR(right, 1, cur);
-		memcpy(rrp, lrp, be16_to_cpu(right->bb_numrecs) * sizeof(*rrp));
-		xfs_alloc_log_recs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		keyp->ar_startblock = rrp->ar_startblock;
-		keyp->ar_blockcount = rrp->ar_blockcount;
-	}
-	/*
-	 * Find the left block number by looking in the buffer.
-	 * Adjust numrecs, sibling pointers.
-	 */
-	lbno = XFS_DADDR_TO_AGBNO(cur->bc_mp, XFS_BUF_ADDR(lbp));
-	be16_add_cpu(&left->bb_numrecs, -(be16_to_cpu(right->bb_numrecs)));
-	right->bb_rightsib = left->bb_rightsib;
-	left->bb_rightsib = cpu_to_be32(rbno);
-	right->bb_leftsib = cpu_to_be32(lbno);
-	xfs_alloc_log_block(cur->bc_tp, rbp, XFS_BB_ALL_BITS);
-	xfs_alloc_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
-	/*
-	 * If there's a block to the new block's right, make that block
-	 * point back to right instead of to left.
-	 */
-	if (be32_to_cpu(right->bb_rightsib) != NULLAGBLOCK) {
-		xfs_alloc_block_t	*rrblock;	/* rr btree block */
-		xfs_buf_t		*rrbp;		/* buffer for rrblock */
 
-		if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-				cur->bc_private.a.agno, be32_to_cpu(right->bb_rightsib), 0,
-				&rrbp, XFS_ALLOC_BTREE_REF)))
-			return error;
-		rrblock = XFS_BUF_TO_ALLOC_BLOCK(rrbp);
-		if ((error = xfs_btree_check_sblock(cur, rrblock, level, rrbp)))
-			return error;
-		rrblock->bb_leftsib = cpu_to_be32(rbno);
-		xfs_alloc_log_block(cur->bc_tp, rrbp, XFS_BB_LEFTSIB);
-	}
-	/*
-	 * If the cursor is really in the right block, move it there.
-	 * If it's just pointing past the last entry in left, then we'll
-	 * insert there, so don't change anything in that case.
-	 */
-	if (cur->bc_ptrs[level] > be16_to_cpu(left->bb_numrecs) + 1) {
-		xfs_btree_setbuf(cur, level, rbp);
-		cur->bc_ptrs[level] -= be16_to_cpu(left->bb_numrecs);
-	}
 	/*
-	 * If there are more levels, we'll need another cursor which refers to
-	 * the right block, no matter where this cursor was.
+	 * Since blocks move to the free list without the coordination used in
+	 * xfs_bmap_finish, we can't allow block to be available for
+	 * reallocation and non-transaction writing (user data) until we know
+	 * that the transaction that moved it to the free list is permanently
+	 * on disk. We track the blocks by declaring these blocks as "busy";
+	 * the busy list is maintained on a per-ag basis and each transaction
+	 * records which entries should be removed when the iclog commits to
+	 * disk. If a busy block is allocated, the iclog is pushed up to the
+	 * LSN that freed the block.
 	 */
-	if (level + 1 < cur->bc_nlevels) {
-		if ((error = xfs_btree_dup_cursor(cur, curp)))
-			return error;
-		(*curp)->bc_ptrs[level + 1]++;
-	}
-	*bnop = rbno;
-	*stat = 1;
+	xfs_alloc_mark_busy(cur->bc_tp, be32_to_cpu(agf->agf_seqno), bno, 1);
+	xfs_trans_agbtree_delta(cur->bc_tp, -1);
 	return 0;
 }
 
 /*
- * Update keys at all levels from here to the root along the cursor's path.
+ * Update the longest extent in the AGF
  */
-STATIC int				/* error */
-xfs_alloc_updkey(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_alloc_key_t		*keyp,	/* new key value to update to */
-	int			level)	/* starting level for update */
+STATIC void
+xfs_allocbt_update_lastrec(
+	struct xfs_btree_cur	*cur,
+	struct xfs_btree_block	*block,
+	union xfs_btree_rec	*rec,
+	int			ptr,
+	int			reason)
 {
-	int			ptr;	/* index of key in block */
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
+	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
+	__be32			len;
+	int			numrecs;
 
-	/*
-	 * Go up the tree from this level toward the root.
-	 * At each level, update the key value to the value input.
-	 * Stop when we reach a level where the cursor isn't pointing
-	 * at the first entry in the block.
-	 */
-	for (ptr = 1; ptr == 1 && level < cur->bc_nlevels; level++) {
-		xfs_alloc_block_t	*block;	/* btree block */
-		xfs_buf_t		*bp;	/* buffer for block */
-#ifdef DEBUG
-		int			error;	/* error return value */
-#endif
-		xfs_alloc_key_t		*kp;	/* ptr to btree block keys */
+	ASSERT(cur->bc_btnum == XFS_BTNUM_CNT);
+
+	switch (reason) {
+	case LASTREC_UPDATE:
+		/*
+		 * If this is the last leaf block and it's the last record,
+		 * then update the size of the longest extent in the AG.
+		 */
+		if (ptr != xfs_btree_get_numrecs(block))
+			return;
+		len = rec->alloc.ar_blockcount;
+		break;
+	case LASTREC_INSREC:
+		if (be32_to_cpu(rec->alloc.ar_blockcount) <=
+		    be32_to_cpu(agf->agf_longest))
+			return;
+		len = rec->alloc.ar_blockcount;
+		break;
+	case LASTREC_DELREC:
+		numrecs = xfs_btree_get_numrecs(block);
+		if (ptr <= numrecs)
+			return;
+		ASSERT(ptr == numrecs + 1);
 
-		bp = cur->bc_bufs[level];
-		block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
-			return error;
-#endif
-		ptr = cur->bc_ptrs[level];
-		kp = XFS_ALLOC_KEY_ADDR(block, ptr, cur);
-		*kp = *keyp;
-		xfs_alloc_log_keys(cur, bp, ptr, ptr);
+		if (numrecs) {
+			xfs_alloc_rec_t *rrp;
+
+			rrp = XFS_ALLOC_REC_ADDR(cur->bc_mp, block, numrecs);
+			len = rrp->ar_blockcount;
+		} else {
+			len = 0;
+		}
+
+		break;
+	default:
+		ASSERT(0);
+		return;
 	}
-	return 0;
+
+	agf->agf_longest = len;
+	cur->bc_mp->m_perag[seqno].pagf_longest = be32_to_cpu(len);
+	xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp, XFS_AGF_LONGEST);
 }
 
-/*
- * Externally visible routines.
- */
+STATIC int
+xfs_allocbt_get_minrecs(
+	struct xfs_btree_cur	*cur,
+	int			level)
+{
+	return cur->bc_mp->m_alloc_mnr[level != 0];
+}
 
-/*
- * Decrement cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-int					/* error */
-xfs_alloc_decrement(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level in btree, 0 is leaf */
-	int			*stat)	/* success/failure */
+STATIC int
+xfs_allocbt_get_maxrecs(
+	struct xfs_btree_cur	*cur,
+	int			level)
 {
-	xfs_alloc_block_t	*block;	/* btree block */
-	int			error;	/* error return value */
-	int			lev;	/* btree level */
+	return cur->bc_mp->m_alloc_mxr[level != 0];
+}
 
-	ASSERT(level < cur->bc_nlevels);
-	/*
-	 * Read-ahead to the left at this level.
-	 */
-	xfs_btree_readahead(cur, level, XFS_BTCUR_LEFTRA);
-	/*
-	 * Decrement the ptr at this level.  If we're still in the block
-	 * then we're done.
-	 */
-	if (--cur->bc_ptrs[level] > 0) {
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * Get a pointer to the btree block.
-	 */
-	block = XFS_BUF_TO_ALLOC_BLOCK(cur->bc_bufs[level]);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level,
-			cur->bc_bufs[level])))
-		return error;
-#endif
-	/*
-	 * If we just went off the left edge of the tree, return failure.
-	 */
-	if (be32_to_cpu(block->bb_leftsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * March up the tree decrementing pointers.
-	 * Stop when we don't go off the left edge of a block.
-	 */
-	for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
-		if (--cur->bc_ptrs[lev] > 0)
-			break;
-		/*
-		 * Read-ahead the left block, we're going to read it
-		 * in the next loop.
-		 */
-		xfs_btree_readahead(cur, lev, XFS_BTCUR_LEFTRA);
-	}
-	/*
-	 * If we went off the root then we are seriously confused.
-	 */
-	ASSERT(lev < cur->bc_nlevels);
-	/*
-	 * Now walk back down the tree, fixing up the cursor's buffer
-	 * pointers and key numbers.
-	 */
-	for (block = XFS_BUF_TO_ALLOC_BLOCK(cur->bc_bufs[lev]); lev > level; ) {
-		xfs_agblock_t	agbno;	/* block number of btree block */
-		xfs_buf_t	*bp;	/* buffer pointer for block */
+STATIC void
+xfs_allocbt_init_key_from_rec(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
+{
+	ASSERT(rec->alloc.ar_startblock != 0);
 
-		agbno = be32_to_cpu(*XFS_ALLOC_PTR_ADDR(block, cur->bc_ptrs[lev], cur));
-		if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-				cur->bc_private.a.agno, agbno, 0, &bp,
-				XFS_ALLOC_BTREE_REF)))
-			return error;
-		lev--;
-		xfs_btree_setbuf(cur, lev, bp);
-		block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-		if ((error = xfs_btree_check_sblock(cur, block, lev, bp)))
-			return error;
-		cur->bc_ptrs[lev] = be16_to_cpu(block->bb_numrecs);
-	}
-	*stat = 1;
-	return 0;
+	key->alloc.ar_startblock = rec->alloc.ar_startblock;
+	key->alloc.ar_blockcount = rec->alloc.ar_blockcount;
 }
 
-/*
- * Delete the record pointed to by cur.
- * The cursor refers to the place where the record was (could be inserted)
- * when the operation returns.
- */
-int					/* error */
-xfs_alloc_delete(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	int		*stat)		/* success/failure */
+STATIC void
+xfs_allocbt_init_rec_from_key(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
 {
-	int		error;		/* error return value */
-	int		i;		/* result code */
-	int		level;		/* btree level */
+	ASSERT(key->alloc.ar_startblock != 0);
 
-	/*
-	 * Go up the tree, starting at leaf level.
-	 * If 2 is returned then a join was done; go to the next level.
-	 * Otherwise we are done.
-	 */
-	for (level = 0, i = 2; i == 2; level++) {
-		if ((error = xfs_alloc_delrec(cur, level, &i)))
-			return error;
-	}
-	if (i == 0) {
-		for (level = 1; level < cur->bc_nlevels; level++) {
-			if (cur->bc_ptrs[level] == 0) {
-				if ((error = xfs_alloc_decrement(cur, level, &i)))
-					return error;
-				break;
-			}
-		}
-	}
-	*stat = i;
-	return 0;
+	rec->alloc.ar_startblock = key->alloc.ar_startblock;
+	rec->alloc.ar_blockcount = key->alloc.ar_blockcount;
 }
 
-/*
- * Get the data from the pointed-to record.
- */
-int					/* error */
-xfs_alloc_get_rec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_agblock_t		*bno,	/* output: starting block of extent */
-	xfs_extlen_t		*len,	/* output: length of extent */
-	int			*stat)	/* output: success/failure */
+STATIC void
+xfs_allocbt_init_rec_from_cur(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec)
 {
-	xfs_alloc_block_t	*block;	/* btree block */
-#ifdef DEBUG
-	int			error;	/* error return value */
-#endif
-	int			ptr;	/* record number */
+	ASSERT(cur->bc_rec.a.ar_startblock != 0);
 
-	ptr = cur->bc_ptrs[0];
-	block = XFS_BUF_TO_ALLOC_BLOCK(cur->bc_bufs[0]);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, 0, cur->bc_bufs[0])))
-		return error;
-#endif
-	/*
-	 * Off the right end or left end, return failure.
-	 */
-	if (ptr > be16_to_cpu(block->bb_numrecs) || ptr <= 0) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Point to the record and extract its data.
-	 */
-	{
-		xfs_alloc_rec_t		*rec;	/* record data */
+	rec->alloc.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
+	rec->alloc.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
+}
+
+STATIC void
+xfs_allocbt_init_ptr_from_cur(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr)
+{
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
+
+	ASSERT(cur->bc_private.a.agno == be32_to_cpu(agf->agf_seqno));
+	ASSERT(agf->agf_roots[cur->bc_btnum] != 0);
+
+	ptr->s = agf->agf_roots[cur->bc_btnum];
+}
+
+STATIC __int64_t
+xfs_allocbt_key_diff(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*key)
+{
+	xfs_alloc_rec_incore_t	*rec = &cur->bc_rec.a;
+	xfs_alloc_key_t		*kp = &key->alloc;
+	__int64_t		diff;
 
-		rec = XFS_ALLOC_REC_ADDR(block, ptr, cur);
-		*bno = be32_to_cpu(rec->ar_startblock);
-		*len = be32_to_cpu(rec->ar_blockcount);
+	if (cur->bc_btnum == XFS_BTNUM_BNO) {
+		return (__int64_t)be32_to_cpu(kp->ar_startblock) -
+				rec->ar_startblock;
 	}
-	*stat = 1;
-	return 0;
+
+	diff = (__int64_t)be32_to_cpu(kp->ar_blockcount) - rec->ar_blockcount;
+	if (diff)
+		return diff;
+
+	return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
 }
 
-/*
- * Increment cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-int					/* error */
-xfs_alloc_increment(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level in btree, 0 is leaf */
-	int			*stat)	/* success/failure */
+STATIC int
+xfs_allocbt_kill_root(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp,
+	int			level,
+	union xfs_btree_ptr	*newroot)
 {
-	xfs_alloc_block_t	*block;	/* btree block */
-	xfs_buf_t		*bp;	/* tree block buffer */
-	int			error;	/* error return value */
-	int			lev;	/* btree level */
+	int			error;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_STATS_INC(cur, killroot);
 
-	ASSERT(level < cur->bc_nlevels);
-	/*
-	 * Read-ahead to the right at this level.
-	 */
-	xfs_btree_readahead(cur, level, XFS_BTCUR_RIGHTRA);
 	/*
-	 * Get a pointer to the btree block.
+	 * Update the root pointer, decreasing the level by 1 and then
+	 * free the old root.
 	 */
-	bp = cur->bc_bufs[level];
-	block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
+	xfs_allocbt_set_root(cur, newroot, -1);
+	error = xfs_allocbt_free_block(cur, bp);
+	if (error) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 		return error;
-#endif
-	/*
-	 * Increment the ptr at this level.  If we're still in the block
-	 * then we're done.
-	 */
-	if (++cur->bc_ptrs[level] <= be16_to_cpu(block->bb_numrecs)) {
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * If we just went off the right edge of the tree, return failure.
-	 */
-	if (be32_to_cpu(block->bb_rightsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
 	}
-	/*
-	 * March up the tree incrementing pointers.
-	 * Stop when we don't go off the right edge of a block.
-	 */
-	for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
-		bp = cur->bc_bufs[lev];
-		block = XFS_BUF_TO_ALLOC_BLOCK(bp);
+
+	XFS_BTREE_STATS_INC(cur, free);
+
+	xfs_btree_setbuf(cur, level, NULL);
+	cur->bc_nlevels--;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	return 0;
+}
+
 #ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, block, lev, bp)))
-			return error;
-#endif
-		if (++cur->bc_ptrs[lev] <= be16_to_cpu(block->bb_numrecs))
-			break;
-		/*
-		 * Read-ahead the right block, we're going to read it
-		 * in the next loop.
-		 */
-		xfs_btree_readahead(cur, lev, XFS_BTCUR_RIGHTRA);
+STATIC int
+xfs_allocbt_keys_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	if (cur->bc_btnum == XFS_BTNUM_BNO) {
+		return be32_to_cpu(k1->alloc.ar_startblock) <
+		       be32_to_cpu(k2->alloc.ar_startblock);
+	} else {
+		return be32_to_cpu(k1->alloc.ar_blockcount) <
+			be32_to_cpu(k2->alloc.ar_blockcount) ||
+			(k1->alloc.ar_blockcount == k2->alloc.ar_blockcount &&
+			 be32_to_cpu(k1->alloc.ar_startblock) <
+			 be32_to_cpu(k2->alloc.ar_startblock));
 	}
-	/*
-	 * If we went off the root then we are seriously confused.
-	 */
-	ASSERT(lev < cur->bc_nlevels);
-	/*
-	 * Now walk back down the tree, fixing up the cursor's buffer
-	 * pointers and key numbers.
-	 */
-	for (bp = cur->bc_bufs[lev], block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-	     lev > level; ) {
-		xfs_agblock_t	agbno;	/* block number of btree block */
+}
 
-		agbno = be32_to_cpu(*XFS_ALLOC_PTR_ADDR(block, cur->bc_ptrs[lev], cur));
-		if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-				cur->bc_private.a.agno, agbno, 0, &bp,
-				XFS_ALLOC_BTREE_REF)))
-			return error;
-		lev--;
-		xfs_btree_setbuf(cur, lev, bp);
-		block = XFS_BUF_TO_ALLOC_BLOCK(bp);
-		if ((error = xfs_btree_check_sblock(cur, block, lev, bp)))
-			return error;
-		cur->bc_ptrs[lev] = 1;
+STATIC int
+xfs_allocbt_recs_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*r1,
+	union xfs_btree_rec	*r2)
+{
+	if (cur->bc_btnum == XFS_BTNUM_BNO) {
+		return be32_to_cpu(r1->alloc.ar_startblock) +
+			be32_to_cpu(r1->alloc.ar_blockcount) <=
+			be32_to_cpu(r2->alloc.ar_startblock);
+	} else {
+		return be32_to_cpu(r1->alloc.ar_blockcount) <
+			be32_to_cpu(r2->alloc.ar_blockcount) ||
+			(r1->alloc.ar_blockcount == r2->alloc.ar_blockcount &&
+			 be32_to_cpu(r1->alloc.ar_startblock) <
+			 be32_to_cpu(r2->alloc.ar_startblock));
 	}
-	*stat = 1;
-	return 0;
 }
+#endif	/* DEBUG */
 
-/*
- * Insert the current record at the point referenced by cur.
- * The cursor may be inconsistent on return if splits have been done.
- */
-int					/* error */
-xfs_alloc_insert(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	int		*stat)		/* success/failure */
-{
-	int		error;		/* error return value */
-	int		i;		/* result value, 0 for failure */
-	int		level;		/* current level number in btree */
-	xfs_agblock_t	nbno;		/* new block number (split result) */
-	xfs_btree_cur_t	*ncur;		/* new cursor (split result) */
-	xfs_alloc_rec_t	nrec;		/* record being inserted this level */
-	xfs_btree_cur_t	*pcur;		/* previous level's cursor */
+#ifdef XFS_BTREE_TRACE
+ktrace_t	*xfs_allocbt_trace_buf;
 
-	level = 0;
-	nbno = NULLAGBLOCK;
-	nrec.ar_startblock = cpu_to_be32(cur->bc_rec.a.ar_startblock);
-	nrec.ar_blockcount = cpu_to_be32(cur->bc_rec.a.ar_blockcount);
-	ncur = NULL;
-	pcur = cur;
-	/*
-	 * Loop going up the tree, starting at the leaf level.
-	 * Stop when we don't get a split block, that must mean that
-	 * the insert is finished with this level.
-	 */
-	do {
-		/*
-		 * Insert nrec/nbno into this level of the tree.
-		 * Note if we fail, nbno will be null.
-		 */
-		if ((error = xfs_alloc_insrec(pcur, level++, &nbno, &nrec, &ncur,
-				&i))) {
-			if (pcur != cur)
-				xfs_btree_del_cursor(pcur, XFS_BTREE_ERROR);
-			return error;
-		}
-		/*
-		 * See if the cursor we just used is trash.
-		 * Can't trash the caller's cursor, but otherwise we should
-		 * if ncur is a new cursor or we're about to be done.
-		 */
-		if (pcur != cur && (ncur || nbno == NULLAGBLOCK)) {
-			cur->bc_nlevels = pcur->bc_nlevels;
-			xfs_btree_del_cursor(pcur, XFS_BTREE_NOERROR);
-		}
-		/*
-		 * If we got a new cursor, switch to it.
-		 */
-		if (ncur) {
-			pcur = ncur;
-			ncur = NULL;
-		}
-	} while (nbno != NULLAGBLOCK);
-	*stat = i;
-	return 0;
+STATIC void
+xfs_allocbt_trace_enter(
+	struct xfs_btree_cur	*cur,
+	const char		*func,
+	char			*s,
+	int			type,
+	int			line,
+	__psunsigned_t		a0,
+	__psunsigned_t		a1,
+	__psunsigned_t		a2,
+	__psunsigned_t		a3,
+	__psunsigned_t		a4,
+	__psunsigned_t		a5,
+	__psunsigned_t		a6,
+	__psunsigned_t		a7,
+	__psunsigned_t		a8,
+	__psunsigned_t		a9,
+	__psunsigned_t		a10)
+{
+	ktrace_enter(xfs_allocbt_trace_buf, (void *)(__psint_t)type,
+		(void *)func, (void *)s, NULL, (void *)cur,
+		(void *)a0, (void *)a1, (void *)a2, (void *)a3,
+		(void *)a4, (void *)a5, (void *)a6, (void *)a7,
+		(void *)a8, (void *)a9, (void *)a10);
 }
 
-/*
- * Lookup the record equal to [bno, len] in the btree given by cur.
- */
-int					/* error */
-xfs_alloc_lookup_eq(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_agblock_t	bno,		/* starting block of extent */
-	xfs_extlen_t	len,		/* length of extent */
-	int		*stat)		/* success/failure */
+STATIC void
+xfs_allocbt_trace_cursor(
+	struct xfs_btree_cur	*cur,
+	__uint32_t		*s0,
+	__uint64_t		*l0,
+	__uint64_t		*l1)
 {
-	cur->bc_rec.a.ar_startblock = bno;
-	cur->bc_rec.a.ar_blockcount = len;
-	return xfs_alloc_lookup(cur, XFS_LOOKUP_EQ, stat);
+	*s0 = cur->bc_private.a.agno;
+	*l0 = cur->bc_rec.a.ar_startblock;
+	*l1 = cur->bc_rec.a.ar_blockcount;
 }
 
-/*
- * Lookup the first record greater than or equal to [bno, len]
- * in the btree given by cur.
- */
-int					/* error */
-xfs_alloc_lookup_ge(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_agblock_t	bno,		/* starting block of extent */
-	xfs_extlen_t	len,		/* length of extent */
-	int		*stat)		/* success/failure */
+STATIC void
+xfs_allocbt_trace_key(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*key,
+	__uint64_t		*l0,
+	__uint64_t		*l1)
 {
-	cur->bc_rec.a.ar_startblock = bno;
-	cur->bc_rec.a.ar_blockcount = len;
-	return xfs_alloc_lookup(cur, XFS_LOOKUP_GE, stat);
+	*l0 = be32_to_cpu(key->alloc.ar_startblock);
+	*l1 = be32_to_cpu(key->alloc.ar_blockcount);
 }
 
-/*
- * Lookup the first record less than or equal to [bno, len]
- * in the btree given by cur.
- */
-int					/* error */
-xfs_alloc_lookup_le(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_agblock_t	bno,		/* starting block of extent */
-	xfs_extlen_t	len,		/* length of extent */
-	int		*stat)		/* success/failure */
+STATIC void
+xfs_allocbt_trace_record(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec,
+	__uint64_t		*l0,
+	__uint64_t		*l1,
+	__uint64_t		*l2)
 {
-	cur->bc_rec.a.ar_startblock = bno;
-	cur->bc_rec.a.ar_blockcount = len;
-	return xfs_alloc_lookup(cur, XFS_LOOKUP_LE, stat);
+	*l0 = be32_to_cpu(rec->alloc.ar_startblock);
+	*l1 = be32_to_cpu(rec->alloc.ar_blockcount);
+	*l2 = 0;
 }
+#endif /* XFS_BTREE_TRACE */
+
+static const struct xfs_btree_ops xfs_allocbt_ops = {
+	.rec_len		= sizeof(xfs_alloc_rec_t),
+	.key_len		= sizeof(xfs_alloc_key_t),
+
+	.dup_cursor		= xfs_allocbt_dup_cursor,
+	.set_root		= xfs_allocbt_set_root,
+	.kill_root		= xfs_allocbt_kill_root,
+	.alloc_block		= xfs_allocbt_alloc_block,
+	.free_block		= xfs_allocbt_free_block,
+	.update_lastrec		= xfs_allocbt_update_lastrec,
+	.get_minrecs		= xfs_allocbt_get_minrecs,
+	.get_maxrecs		= xfs_allocbt_get_maxrecs,
+	.init_key_from_rec	= xfs_allocbt_init_key_from_rec,
+	.init_rec_from_key	= xfs_allocbt_init_rec_from_key,
+	.init_rec_from_cur	= xfs_allocbt_init_rec_from_cur,
+	.init_ptr_from_cur	= xfs_allocbt_init_ptr_from_cur,
+	.key_diff		= xfs_allocbt_key_diff,
+
+#ifdef DEBUG
+	.keys_inorder		= xfs_allocbt_keys_inorder,
+	.recs_inorder		= xfs_allocbt_recs_inorder,
+#endif
+
+#ifdef XFS_BTREE_TRACE
+	.trace_enter		= xfs_allocbt_trace_enter,
+	.trace_cursor		= xfs_allocbt_trace_cursor,
+	.trace_key		= xfs_allocbt_trace_key,
+	.trace_record		= xfs_allocbt_trace_record,
+#endif
+};
 
 /*
- * Update the record referred to by cur, to the value given by [bno, len].
- * This either works (return 0) or gets an EFSCORRUPTED error.
+ * Allocate a new allocation btree cursor.
  */
-int					/* error */
-xfs_alloc_update(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_agblock_t		bno,	/* starting block of extent */
-	xfs_extlen_t		len)	/* length of extent */
+struct xfs_btree_cur *			/* new alloc btree cursor */
+xfs_allocbt_init_cursor(
+	struct xfs_mount	*mp,		/* file system mount point */
+	struct xfs_trans	*tp,		/* transaction pointer */
+	struct xfs_buf		*agbp,		/* buffer for agf structure */
+	xfs_agnumber_t		agno,		/* allocation group number */
+	xfs_btnum_t		btnum)		/* btree identifier */
 {
-	xfs_alloc_block_t	*block;	/* btree block to update */
-	int			error;	/* error return value */
-	int			ptr;	/* current record number (updating) */
+	struct xfs_agf		*agf = XFS_BUF_TO_AGF(agbp);
+	struct xfs_btree_cur	*cur;
 
-	ASSERT(len > 0);
-	/*
-	 * Pick up the a.g. freelist struct and the current block.
-	 */
-	block = XFS_BUF_TO_ALLOC_BLOCK(cur->bc_bufs[0]);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, 0, cur->bc_bufs[0])))
-		return error;
-#endif
-	/*
-	 * Get the address of the rec to be updated.
-	 */
-	ptr = cur->bc_ptrs[0];
-	{
-		xfs_alloc_rec_t		*rp;	/* pointer to updated record */
+	ASSERT(btnum == XFS_BTNUM_BNO || btnum == XFS_BTNUM_CNT);
 
-		rp = XFS_ALLOC_REC_ADDR(block, ptr, cur);
-		/*
-		 * Fill in the new contents and log them.
-		 */
-		rp->ar_startblock = cpu_to_be32(bno);
-		rp->ar_blockcount = cpu_to_be32(len);
-		xfs_alloc_log_recs(cur, cur->bc_bufs[0], ptr, ptr);
-	}
-	/*
-	 * If it's the by-size btree and it's the last leaf block and
-	 * it's the last record... then update the size of the longest
-	 * extent in the a.g., which we cache in the a.g. freelist header.
-	 */
-	if (cur->bc_btnum == XFS_BTNUM_CNT &&
-	    be32_to_cpu(block->bb_rightsib) == NULLAGBLOCK &&
-	    ptr == be16_to_cpu(block->bb_numrecs)) {
-		xfs_agf_t	*agf;	/* a.g. freespace header */
-		xfs_agnumber_t	seqno;
+	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
 
-		agf = XFS_BUF_TO_AGF(cur->bc_private.a.agbp);
-		seqno = be32_to_cpu(agf->agf_seqno);
-		cur->bc_mp->m_perag[seqno].pagf_longest = len;
-		agf->agf_longest = cpu_to_be32(len);
-		xfs_alloc_log_agf(cur->bc_tp, cur->bc_private.a.agbp,
-			XFS_AGF_LONGEST);
-	}
-	/*
-	 * Updating first record in leaf. Pass new key value up to our parent.
-	 */
-	if (ptr == 1) {
-		xfs_alloc_key_t	key;	/* key containing [bno, len] */
+	cur->bc_tp = tp;
+	cur->bc_mp = mp;
+	cur->bc_nlevels = be32_to_cpu(agf->agf_levels[btnum]);
+	cur->bc_btnum = btnum;
+	cur->bc_blocklog = mp->m_sb.sb_blocklog;
 
-		key.ar_startblock = cpu_to_be32(bno);
-		key.ar_blockcount = cpu_to_be32(len);
-		if ((error = xfs_alloc_updkey(cur, &key, 1)))
-			return error;
-	}
-	return 0;
+	cur->bc_ops = &xfs_allocbt_ops;
+	if (btnum == XFS_BTNUM_CNT)
+		cur->bc_flags = XFS_BTREE_LASTREC_UPDATE;
+
+	cur->bc_private.a.agbp = agbp;
+	cur->bc_private.a.agno = agno;
+
+	return cur;
+}
+
+/*
+ * Calculate number of records in an alloc btree block.
+ */
+int
+xfs_allocbt_maxrecs(
+	struct xfs_mount	*mp,
+	int			blocklen,
+	int			leaf)
+{
+	blocklen -= XFS_ALLOC_BLOCK_LEN(mp);
+
+	if (leaf)
+		return blocklen / sizeof(xfs_alloc_rec_t);
+	return blocklen / (sizeof(xfs_alloc_key_t) + sizeof(xfs_alloc_ptr_t));
 }

fs/xfs/xfs_alloc_btree.h

@@ -24,7 +24,6 @@
 
 struct xfs_buf;
 struct xfs_btree_cur;
-struct xfs_btree_sblock;
 struct xfs_mount;
 
 /*
@@ -50,16 +49,6 @@ typedef struct xfs_alloc_rec_incore {
 
 /* btree pointer type */
 typedef __be32 xfs_alloc_ptr_t;
-/* btree block header type */
-typedef	struct xfs_btree_sblock xfs_alloc_block_t;
-
-#define	XFS_BUF_TO_ALLOC_BLOCK(bp)	((xfs_alloc_block_t *)XFS_BUF_PTR(bp))
-
-/*
- * Real block structures have a size equal to the disk block size.
- */
-#define	XFS_ALLOC_BLOCK_MAXRECS(lev,cur) ((cur)->bc_mp->m_alloc_mxr[lev != 0])
-#define	XFS_ALLOC_BLOCK_MINRECS(lev,cur) ((cur)->bc_mp->m_alloc_mnr[lev != 0])
 
 /*
  * Minimum and maximum blocksize and sectorsize.
@@ -83,73 +72,39 @@ typedef	struct xfs_btree_sblock xfs_alloc_block_t;
 #define	XFS_CNT_BLOCK(mp)	((xfs_agblock_t)(XFS_BNO_BLOCK(mp) + 1))
 
 /*
- * Record, key, and pointer address macros for btree blocks.
- */
-#define	XFS_ALLOC_REC_ADDR(bb,i,cur)	\
-	XFS_BTREE_REC_ADDR(xfs_alloc, bb, i)
-
-#define	XFS_ALLOC_KEY_ADDR(bb,i,cur)	\
-	XFS_BTREE_KEY_ADDR(xfs_alloc, bb, i)
-
-#define	XFS_ALLOC_PTR_ADDR(bb,i,cur)	\
-	XFS_BTREE_PTR_ADDR(xfs_alloc, bb, i, XFS_ALLOC_BLOCK_MAXRECS(1, cur))
-
-/*
- * Decrement cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-extern int xfs_alloc_decrement(struct xfs_btree_cur *cur, int level, int *stat);
-
-/*
- * Delete the record pointed to by cur.
- * The cursor refers to the place where the record was (could be inserted)
- * when the operation returns.
- */
-extern int xfs_alloc_delete(struct xfs_btree_cur *cur, int *stat);
-
-/*
- * Get the data from the pointed-to record.
- */
-extern int xfs_alloc_get_rec(struct xfs_btree_cur *cur,	xfs_agblock_t *bno,
-				xfs_extlen_t *len, int *stat);
-
-/*
- * Increment cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-extern int xfs_alloc_increment(struct xfs_btree_cur *cur, int level, int *stat);
-
-/*
- * Insert the current record at the point referenced by cur.
- * The cursor may be inconsistent on return if splits have been done.
- */
-extern int xfs_alloc_insert(struct xfs_btree_cur *cur, int *stat);
-
-/*
- * Lookup the record equal to [bno, len] in the btree given by cur.
- */
-extern int xfs_alloc_lookup_eq(struct xfs_btree_cur *cur, xfs_agblock_t bno,
-				xfs_extlen_t len, int *stat);
-
-/*
- * Lookup the first record greater than or equal to [bno, len]
- * in the btree given by cur.
- */
-extern int xfs_alloc_lookup_ge(struct xfs_btree_cur *cur, xfs_agblock_t bno,
-				xfs_extlen_t len, int *stat);
-
-/*
- * Lookup the first record less than or equal to [bno, len]
- * in the btree given by cur.
+ * Btree block header size depends on a superblock flag.
+ *
+ * (not quite yet, but soon)
  */
-extern int xfs_alloc_lookup_le(struct xfs_btree_cur *cur, xfs_agblock_t bno,
-				xfs_extlen_t len, int *stat);
+#define XFS_ALLOC_BLOCK_LEN(mp)	XFS_BTREE_SBLOCK_LEN
 
 /*
- * Update the record referred to by cur, to the value given by [bno, len].
- * This either works (return 0) or gets an EFSCORRUPTED error.
- */
-extern int xfs_alloc_update(struct xfs_btree_cur *cur, xfs_agblock_t bno,
-				xfs_extlen_t len);
+ * Record, key, and pointer address macros for btree blocks.
+ *
+ * (note that some of these may appear unused, but they are used in userspace)
+ */
+#define XFS_ALLOC_REC_ADDR(mp, block, index) \
+	((xfs_alloc_rec_t *) \
+		((char *)(block) + \
+		 XFS_ALLOC_BLOCK_LEN(mp) + \
+		 (((index) - 1) * sizeof(xfs_alloc_rec_t))))
+
+#define XFS_ALLOC_KEY_ADDR(mp, block, index) \
+	((xfs_alloc_key_t *) \
+		((char *)(block) + \
+		 XFS_ALLOC_BLOCK_LEN(mp) + \
+		 ((index) - 1) * sizeof(xfs_alloc_key_t)))
+
+#define XFS_ALLOC_PTR_ADDR(mp, block, index, maxrecs) \
+	((xfs_alloc_ptr_t *) \
+		((char *)(block) + \
+		 XFS_ALLOC_BLOCK_LEN(mp) + \
+		 (maxrecs) * sizeof(xfs_alloc_key_t) + \
+		 ((index) - 1) * sizeof(xfs_alloc_ptr_t)))
+
+extern struct xfs_btree_cur *xfs_allocbt_init_cursor(struct xfs_mount *,
+		struct xfs_trans *, struct xfs_buf *,
+		xfs_agnumber_t, xfs_btnum_t);
+extern int xfs_allocbt_maxrecs(struct xfs_mount *, int, int);
 
 #endif	/* __XFS_ALLOC_BTREE_H__ */

fs/xfs/xfs_arch.h

@@ -41,21 +41,36 @@
 #endif
 
 #ifdef XFS_NATIVE_HOST
-#define cpu_to_be16(val)	((__be16)(val))
-#define cpu_to_be32(val)	((__be32)(val))
-#define cpu_to_be64(val)	((__be64)(val))
-#define be16_to_cpu(val)	((__uint16_t)(val))
-#define be32_to_cpu(val)	((__uint32_t)(val))
-#define be64_to_cpu(val)	((__uint64_t)(val))
+#define cpu_to_be16(val)	((__force __be16)(__u16)(val))
+#define cpu_to_be32(val)	((__force __be32)(__u32)(val))
+#define cpu_to_be64(val)	((__force __be64)(__u64)(val))
+#define be16_to_cpu(val)	((__force __u16)(__be16)(val))
+#define be32_to_cpu(val)	((__force __u32)(__be32)(val))
+#define be64_to_cpu(val)	((__force __u64)(__be64)(val))
 #else
-#define cpu_to_be16(val)	(__swab16((__uint16_t)(val)))
-#define cpu_to_be32(val)	(__swab32((__uint32_t)(val)))
-#define cpu_to_be64(val)	(__swab64((__uint64_t)(val)))
-#define be16_to_cpu(val)	(__swab16((__be16)(val)))
-#define be32_to_cpu(val)	(__swab32((__be32)(val)))
-#define be64_to_cpu(val)	(__swab64((__be64)(val)))
+#define cpu_to_be16(val)	((__force __be16)__swab16((__u16)(val)))
+#define cpu_to_be32(val)	((__force __be32)__swab32((__u32)(val)))
+#define cpu_to_be64(val)	((__force __be64)__swab64((__u64)(val)))
+#define be16_to_cpu(val)	(__swab16((__force __u16)(__be16)(val)))
+#define be32_to_cpu(val)	(__swab32((__force __u32)(__be32)(val)))
+#define be64_to_cpu(val)	(__swab64((__force __u64)(__be64)(val)))
 #endif
 
+static inline void be16_add_cpu(__be16 *a, __s16 b)
+{
+	*a = cpu_to_be16(be16_to_cpu(*a) + b);
+}
+
+static inline void be32_add_cpu(__be32 *a, __s32 b)
+{
+	*a = cpu_to_be32(be32_to_cpu(*a) + b);
+}
+
+static inline void be64_add_cpu(__be64 *a, __s64 b)
+{
+	*a = cpu_to_be64(be64_to_cpu(*a) + b);
+}
+
 #endif	/* __KERNEL__ */
 
 /* do we need conversion? */

fs/xfs/xfs_bit.h

@@ -61,8 +61,7 @@ static inline int xfs_highbit64(__uint64_t v)
 /* Get low bit set out of 32-bit argument, -1 if none set */
 static inline int xfs_lowbit32(__uint32_t v)
 {
-	unsigned long	t = v;
-	return (v) ? find_first_bit(&t, 32) : -1;
+	return ffs(v) - 1;
 }
 
 /* Get low bit set out of 64-bit argument, -1 if none set */

fs/xfs/xfs_bmap.c

@@ -393,8 +393,8 @@ xfs_bmap_count_leaves(
 
 STATIC void
 xfs_bmap_disk_count_leaves(
-	xfs_extnum_t		idx,
-	xfs_bmbt_block_t	*block,
+	struct xfs_mount	*mp,
+	struct xfs_btree_block	*block,
 	int			numrecs,
 	int			*count);
 
@@ -402,6 +402,53 @@ xfs_bmap_disk_count_leaves(
  * Bmap internal routines.
  */
 
+STATIC int				/* error */
+xfs_bmbt_lookup_eq(
+	struct xfs_btree_cur	*cur,
+	xfs_fileoff_t		off,
+	xfs_fsblock_t		bno,
+	xfs_filblks_t		len,
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.b.br_startoff = off;
+	cur->bc_rec.b.br_startblock = bno;
+	cur->bc_rec.b.br_blockcount = len;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
+}
+
+STATIC int				/* error */
+xfs_bmbt_lookup_ge(
+	struct xfs_btree_cur	*cur,
+	xfs_fileoff_t		off,
+	xfs_fsblock_t		bno,
+	xfs_filblks_t		len,
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.b.br_startoff = off;
+	cur->bc_rec.b.br_startblock = bno;
+	cur->bc_rec.b.br_blockcount = len;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
+}
+
+/*
+* Update the record referred to by cur to the value given
+ * by [off, bno, len, state].
+ * This either works (return 0) or gets an EFSCORRUPTED error.
+ */
+STATIC int
+xfs_bmbt_update(
+	struct xfs_btree_cur	*cur,
+	xfs_fileoff_t		off,
+	xfs_fsblock_t		bno,
+	xfs_filblks_t		len,
+	xfs_exntst_t		state)
+{
+	union xfs_btree_rec	rec;
+
+	xfs_bmbt_disk_set_allf(&rec.bmbt, off, bno, len, state);
+	return xfs_btree_update(cur, &rec);
+}
+
 /*
  * Called from xfs_bmap_add_attrfork to handle btree format files.
  */
@@ -422,15 +469,14 @@ xfs_bmap_add_attrfork_btree(
 	if (ip->i_df.if_broot_bytes <= XFS_IFORK_DSIZE(ip))
 		*flags |= XFS_ILOG_DBROOT;
 	else {
-		cur = xfs_btree_init_cursor(mp, tp, NULL, 0, XFS_BTNUM_BMAP, ip,
-			XFS_DATA_FORK);
+		cur = xfs_bmbt_init_cursor(mp, tp, ip, XFS_DATA_FORK);
 		cur->bc_private.b.flist = flist;
 		cur->bc_private.b.firstblock = *firstblock;
 		if ((error = xfs_bmbt_lookup_ge(cur, 0, 0, 0, &stat)))
 			goto error0;
 		/* must be at least one entry */
 		XFS_WANT_CORRUPTED_GOTO(stat == 1, error0);
-		if ((error = xfs_bmbt_newroot(cur, flags, &stat)))
+		if ((error = xfs_btree_new_iroot(cur, flags, &stat)))
 			goto error0;
 		if (stat == 0) {
 			xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
@@ -818,10 +864,10 @@ xfs_bmap_add_extent_delay_real(
 					RIGHT.br_blockcount, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_delete(cur, &i)))
+			if ((error = xfs_btree_delete(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_decrement(cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cur, 0, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 			if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
@@ -931,7 +977,7 @@ xfs_bmap_add_extent_delay_real(
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 0, done);
 			cur->bc_rec.b.br_state = XFS_EXT_NORM;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -1007,7 +1053,7 @@ xfs_bmap_add_extent_delay_real(
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 0, done);
 			cur->bc_rec.b.br_state = XFS_EXT_NORM;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -1097,7 +1143,7 @@ xfs_bmap_add_extent_delay_real(
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 0, done);
 			cur->bc_rec.b.br_state = XFS_EXT_NORM;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -1152,7 +1198,7 @@ xfs_bmap_add_extent_delay_real(
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 0, done);
 			cur->bc_rec.b.br_state = XFS_EXT_NORM;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -1379,16 +1425,16 @@ xfs_bmap_add_extent_unwritten_real(
 					RIGHT.br_blockcount, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_delete(cur, &i)))
+			if ((error = xfs_btree_delete(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_decrement(cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cur, 0, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_delete(cur, &i)))
+			if ((error = xfs_btree_delete(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_decrement(cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cur, 0, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 			if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
@@ -1428,10 +1474,10 @@ xfs_bmap_add_extent_unwritten_real(
 					&i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_delete(cur, &i)))
+			if ((error = xfs_btree_delete(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_decrement(cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cur, 0, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 			if ((error = xfs_bmbt_update(cur, LEFT.br_startoff,
@@ -1471,10 +1517,10 @@ xfs_bmap_add_extent_unwritten_real(
 					RIGHT.br_blockcount, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_delete(cur, &i)))
+			if ((error = xfs_btree_delete(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_decrement(cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cur, 0, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 			if ((error = xfs_bmbt_update(cur, new->br_startoff,
@@ -1557,7 +1603,7 @@ xfs_bmap_add_extent_unwritten_real(
 				PREV.br_blockcount - new->br_blockcount,
 				oldext)))
 				goto done;
-			if ((error = xfs_bmbt_decrement(cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cur, 0, &i)))
 				goto done;
 			if (xfs_bmbt_update(cur, LEFT.br_startoff,
 				LEFT.br_startblock,
@@ -1605,7 +1651,7 @@ xfs_bmap_add_extent_unwritten_real(
 				oldext)))
 				goto done;
 			cur->bc_rec.b = *new;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -1647,7 +1693,7 @@ xfs_bmap_add_extent_unwritten_real(
 				PREV.br_blockcount - new->br_blockcount,
 				oldext)))
 				goto done;
-			if ((error = xfs_bmbt_increment(cur, 0, &i)))
+			if ((error = xfs_btree_increment(cur, 0, &i)))
 				goto done;
 			if ((error = xfs_bmbt_update(cur, new->br_startoff,
 				new->br_startblock,
@@ -1695,7 +1741,7 @@ xfs_bmap_add_extent_unwritten_real(
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 0, done);
 			cur->bc_rec.b.br_state = XFS_EXT_NORM;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -1743,7 +1789,7 @@ xfs_bmap_add_extent_unwritten_real(
 			cur->bc_rec.b = PREV;
 			cur->bc_rec.b.br_blockcount =
 				new->br_startoff - PREV.br_startoff;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 			/*
@@ -1758,7 +1804,7 @@ xfs_bmap_add_extent_unwritten_real(
 			XFS_WANT_CORRUPTED_GOTO(i == 0, done);
 			/* new middle extent - newext */
 			cur->bc_rec.b.br_state = new->br_state;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -2106,10 +2152,10 @@ xfs_bmap_add_extent_hole_real(
 					right.br_blockcount, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_delete(cur, &i)))
+			if ((error = xfs_btree_delete(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
-			if ((error = xfs_bmbt_decrement(cur, 0, &i)))
+			if ((error = xfs_btree_decrement(cur, 0, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 			if ((error = xfs_bmbt_update(cur, left.br_startoff,
@@ -2218,7 +2264,7 @@ xfs_bmap_add_extent_hole_real(
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 0, done);
 			cur->bc_rec.b.br_state = new->br_state;
-			if ((error = xfs_bmbt_insert(cur, &i)))
+			if ((error = xfs_btree_insert(cur, &i)))
 				goto done;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		}
@@ -2996,24 +3042,24 @@ xfs_bmap_btree_to_extents(
 	int			whichfork)  /* data or attr fork */
 {
 	/* REFERENCED */
-	xfs_bmbt_block_t	*cblock;/* child btree block */
+	struct xfs_btree_block	*cblock;/* child btree block */
 	xfs_fsblock_t		cbno;	/* child block number */
 	xfs_buf_t		*cbp;	/* child block's buffer */
 	int			error;	/* error return value */
 	xfs_ifork_t		*ifp;	/* inode fork data */
 	xfs_mount_t		*mp;	/* mount point structure */
 	__be64			*pp;	/* ptr to block address */
-	xfs_bmbt_block_t	*rblock;/* root btree block */
+	struct xfs_btree_block	*rblock;/* root btree block */
 
+	mp = ip->i_mount;
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	ASSERT(ifp->if_flags & XFS_IFEXTENTS);
 	ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE);
 	rblock = ifp->if_broot;
 	ASSERT(be16_to_cpu(rblock->bb_level) == 1);
 	ASSERT(be16_to_cpu(rblock->bb_numrecs) == 1);
-	ASSERT(XFS_BMAP_BROOT_MAXRECS(ifp->if_broot_bytes) == 1);
-	mp = ip->i_mount;
-	pp = XFS_BMAP_BROOT_PTR_ADDR(rblock, 1, ifp->if_broot_bytes);
+	ASSERT(xfs_bmbt_maxrecs(mp, ifp->if_broot_bytes, 0) == 1);
+	pp = XFS_BMAP_BROOT_PTR_ADDR(mp, rblock, 1, ifp->if_broot_bytes);
 	cbno = be64_to_cpu(*pp);
 	*logflagsp = 0;
 #ifdef DEBUG
@@ -3023,8 +3069,8 @@ xfs_bmap_btree_to_extents(
 	if ((error = xfs_btree_read_bufl(mp, tp, cbno, 0, &cbp,
 			XFS_BMAP_BTREE_REF)))
 		return error;
-	cblock = XFS_BUF_TO_BMBT_BLOCK(cbp);
-	if ((error = xfs_btree_check_lblock(cur, cblock, 0, cbp)))
+	cblock = XFS_BUF_TO_BLOCK(cbp);
+	if ((error = xfs_btree_check_block(cur, cblock, 0, cbp)))
 		return error;
 	xfs_bmap_add_free(cbno, 1, cur->bc_private.b.flist, mp);
 	ip->i_d.di_nblocks--;
@@ -3170,7 +3216,7 @@ xfs_bmap_del_extent(
 			flags |= XFS_ILOG_FEXT(whichfork);
 			break;
 		}
-		if ((error = xfs_bmbt_delete(cur, &i)))
+		if ((error = xfs_btree_delete(cur, &i)))
 			goto done;
 		XFS_WANT_CORRUPTED_GOTO(i == 1, done);
 		break;
@@ -3254,10 +3300,10 @@ xfs_bmap_del_extent(
 						got.br_startblock, temp,
 						got.br_state)))
 					goto done;
-				if ((error = xfs_bmbt_increment(cur, 0, &i)))
+				if ((error = xfs_btree_increment(cur, 0, &i)))
 					goto done;
 				cur->bc_rec.b = new;
-				error = xfs_bmbt_insert(cur, &i);
+				error = xfs_btree_insert(cur, &i);
 				if (error && error != ENOSPC)
 					goto done;
 				/*
@@ -3404,11 +3450,11 @@ xfs_bmap_extents_to_btree(
 	int			*logflagsp,	/* inode logging flags */
 	int			whichfork)	/* data or attr fork */
 {
-	xfs_bmbt_block_t	*ablock;	/* allocated (child) bt block */
+	struct xfs_btree_block	*ablock;	/* allocated (child) bt block */
 	xfs_buf_t		*abp;		/* buffer for ablock */
 	xfs_alloc_arg_t		args;		/* allocation arguments */
 	xfs_bmbt_rec_t		*arp;		/* child record pointer */
-	xfs_bmbt_block_t	*block;		/* btree root block */
+	struct xfs_btree_block	*block;		/* btree root block */
 	xfs_btree_cur_t		*cur;		/* bmap btree cursor */
 	xfs_bmbt_rec_host_t	*ep;		/* extent record pointer */
 	int			error;		/* error return value */
@@ -3428,6 +3474,7 @@ xfs_bmap_extents_to_btree(
 	 */
 	xfs_iroot_realloc(ip, 1, whichfork);
 	ifp->if_flags |= XFS_IFBROOT;
+
 	/*
 	 * Fill in the root.
 	 */
@@ -3435,14 +3482,14 @@ xfs_bmap_extents_to_btree(
 	block->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
 	block->bb_level = cpu_to_be16(1);
 	block->bb_numrecs = cpu_to_be16(1);
-	block->bb_leftsib = cpu_to_be64(NULLDFSBNO);
-	block->bb_rightsib = cpu_to_be64(NULLDFSBNO);
+	block->bb_u.l.bb_leftsib = cpu_to_be64(NULLDFSBNO);
+	block->bb_u.l.bb_rightsib = cpu_to_be64(NULLDFSBNO);
+
 	/*
 	 * Need a cursor.  Can't allocate until bb_level is filled in.
 	 */
 	mp = ip->i_mount;
-	cur = xfs_btree_init_cursor(mp, tp, NULL, 0, XFS_BTNUM_BMAP, ip,
-		whichfork);
+	cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
 	cur->bc_private.b.firstblock = *firstblock;
 	cur->bc_private.b.flist = flist;
 	cur->bc_private.b.flags = wasdel ? XFS_BTCUR_BPRV_WASDEL : 0;
@@ -3489,12 +3536,12 @@ xfs_bmap_extents_to_btree(
 	/*
 	 * Fill in the child block.
 	 */
-	ablock = XFS_BUF_TO_BMBT_BLOCK(abp);
+	ablock = XFS_BUF_TO_BLOCK(abp);
 	ablock->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
 	ablock->bb_level = 0;
-	ablock->bb_leftsib = cpu_to_be64(NULLDFSBNO);
-	ablock->bb_rightsib = cpu_to_be64(NULLDFSBNO);
-	arp = XFS_BMAP_REC_IADDR(ablock, 1, cur);
+	ablock->bb_u.l.bb_leftsib = cpu_to_be64(NULLDFSBNO);
+	ablock->bb_u.l.bb_rightsib = cpu_to_be64(NULLDFSBNO);
+	arp = XFS_BMBT_REC_ADDR(mp, ablock, 1);
 	nextents = ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t);
 	for (cnt = i = 0; i < nextents; i++) {
 		ep = xfs_iext_get_ext(ifp, i);
@@ -3505,21 +3552,24 @@ xfs_bmap_extents_to_btree(
 		}
 	}
 	ASSERT(cnt == XFS_IFORK_NEXTENTS(ip, whichfork));
-	ablock->bb_numrecs = cpu_to_be16(cnt);
+	xfs_btree_set_numrecs(ablock, cnt);
+
 	/*
 	 * Fill in the root key and pointer.
 	 */
-	kp = XFS_BMAP_KEY_IADDR(block, 1, cur);
-	arp = XFS_BMAP_REC_IADDR(ablock, 1, cur);
+	kp = XFS_BMBT_KEY_ADDR(mp, block, 1);
+	arp = XFS_BMBT_REC_ADDR(mp, ablock, 1);
 	kp->br_startoff = cpu_to_be64(xfs_bmbt_disk_get_startoff(arp));
-	pp = XFS_BMAP_PTR_IADDR(block, 1, cur);
+	pp = XFS_BMBT_PTR_ADDR(mp, block, 1, xfs_bmbt_get_maxrecs(cur,
+						be16_to_cpu(block->bb_level)));
 	*pp = cpu_to_be64(args.fsbno);
+
 	/*
 	 * Do all this logging at the end so that
 	 * the root is at the right level.
 	 */
-	xfs_bmbt_log_block(cur, abp, XFS_BB_ALL_BITS);
-	xfs_bmbt_log_recs(cur, abp, 1, be16_to_cpu(ablock->bb_numrecs));
+	xfs_btree_log_block(cur, abp, XFS_BB_ALL_BITS);
+	xfs_btree_log_recs(cur, abp, 1, be16_to_cpu(ablock->bb_numrecs));
 	ASSERT(*curp == NULL);
 	*curp = cur;
 	*logflagsp = XFS_ILOG_CORE | XFS_ILOG_FBROOT(whichfork);
@@ -4176,7 +4226,7 @@ xfs_bmap_compute_maxlevels(
 		maxleafents = MAXAEXTNUM;
 		sz = XFS_BMDR_SPACE_CALC(MINABTPTRS);
 	}
-	maxrootrecs = (int)XFS_BTREE_BLOCK_MAXRECS(sz, xfs_bmdr, 0);
+	maxrootrecs = xfs_bmdr_maxrecs(mp, sz, 0);
 	minleafrecs = mp->m_bmap_dmnr[0];
 	minnoderecs = mp->m_bmap_dmnr[1];
 	maxblocks = (maxleafents + minleafrecs - 1) / minleafrecs;
@@ -4474,6 +4524,22 @@ xfs_bmap_one_block(
 	return rval;
 }
 
+STATIC int
+xfs_bmap_sanity_check(
+	struct xfs_mount	*mp,
+	struct xfs_buf		*bp,
+	int			level)
+{
+	struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
+
+	if (be32_to_cpu(block->bb_magic) != XFS_BMAP_MAGIC ||
+	    be16_to_cpu(block->bb_level) != level ||
+	    be16_to_cpu(block->bb_numrecs) == 0 ||
+	    be16_to_cpu(block->bb_numrecs) > mp->m_bmap_dmxr[level != 0])
+		return 0;
+	return 1;
+}
+
 /*
  * Read in the extents to if_extents.
  * All inode fields are set up by caller, we just traverse the btree
@@ -4486,7 +4552,7 @@ xfs_bmap_read_extents(
 	xfs_inode_t		*ip,	/* incore inode */
 	int			whichfork) /* data or attr fork */
 {
-	xfs_bmbt_block_t	*block;	/* current btree block */
+	struct xfs_btree_block	*block;	/* current btree block */
 	xfs_fsblock_t		bno;	/* block # of "block" */
 	xfs_buf_t		*bp;	/* buffer for "block" */
 	int			error;	/* error return value */
@@ -4510,7 +4576,7 @@ xfs_bmap_read_extents(
 	 */
 	level = be16_to_cpu(block->bb_level);
 	ASSERT(level > 0);
-	pp = XFS_BMAP_BROOT_PTR_ADDR(block, 1, ifp->if_broot_bytes);
+	pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
 	bno = be64_to_cpu(*pp);
 	ASSERT(bno != NULLDFSBNO);
 	ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
@@ -4523,13 +4589,13 @@ xfs_bmap_read_extents(
 		if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
 				XFS_BMAP_BTREE_REF)))
 			return error;
-		block = XFS_BUF_TO_BMBT_BLOCK(bp);
+		block = XFS_BUF_TO_BLOCK(bp);
 		XFS_WANT_CORRUPTED_GOTO(
-			XFS_BMAP_SANITY_CHECK(mp, block, level),
+			xfs_bmap_sanity_check(mp, bp, level),
 			error0);
 		if (level == 0)
 			break;
-		pp = XFS_BTREE_PTR_ADDR(xfs_bmbt, block, 1, mp->m_bmap_dmxr[1]);
+		pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]);
 		bno = be64_to_cpu(*pp);
 		XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp, bno), error0);
 		xfs_trans_brelse(tp, bp);
@@ -4549,7 +4615,7 @@ xfs_bmap_read_extents(
 		xfs_extnum_t	start;
 
 
-		num_recs = be16_to_cpu(block->bb_numrecs);
+		num_recs = xfs_btree_get_numrecs(block);
 		if (unlikely(i + num_recs > room)) {
 			ASSERT(i + num_recs <= room);
 			xfs_fs_repair_cmn_err(CE_WARN, ip->i_mount,
@@ -4561,18 +4627,18 @@ xfs_bmap_read_extents(
 			goto error0;
 		}
 		XFS_WANT_CORRUPTED_GOTO(
-			XFS_BMAP_SANITY_CHECK(mp, block, 0),
+			xfs_bmap_sanity_check(mp, bp, 0),
 			error0);
 		/*
 		 * Read-ahead the next leaf block, if any.
 		 */
-		nextbno = be64_to_cpu(block->bb_rightsib);
+		nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
 		if (nextbno != NULLFSBLOCK)
 			xfs_btree_reada_bufl(mp, nextbno, 1);
 		/*
 		 * Copy records into the extent records.
 		 */
-		frp = XFS_BTREE_REC_ADDR(xfs_bmbt, block, 1);
+		frp = XFS_BMBT_REC_ADDR(mp, block, 1);
 		start = i;
 		for (j = 0; j < num_recs; j++, i++, frp++) {
 			xfs_bmbt_rec_host_t *trp = xfs_iext_get_ext(ifp, i);
@@ -4603,7 +4669,7 @@ xfs_bmap_read_extents(
 		if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
 				XFS_BMAP_BTREE_REF)))
 			return error;
-		block = XFS_BUF_TO_BMBT_BLOCK(bp);
+		block = XFS_BUF_TO_BLOCK(bp);
 	}
 	ASSERT(i == (ifp->if_bytes / (uint)sizeof(xfs_bmbt_rec_t)));
 	ASSERT(i == XFS_IFORK_NEXTENTS(ip, whichfork));
@@ -5029,8 +5095,7 @@ xfs_bmapi(
 				if (abno == NULLFSBLOCK)
 					break;
 				if ((ifp->if_flags & XFS_IFBROOT) && !cur) {
-					cur = xfs_btree_init_cursor(mp,
-						tp, NULL, 0, XFS_BTNUM_BMAP,
+					cur = xfs_bmbt_init_cursor(mp, tp,
 						ip, whichfork);
 					cur->bc_private.b.firstblock =
 						*firstblock;
@@ -5147,9 +5212,8 @@ xfs_bmapi(
 			 */
 			ASSERT(mval->br_blockcount <= len);
 			if ((ifp->if_flags & XFS_IFBROOT) && !cur) {
-				cur = xfs_btree_init_cursor(mp,
-					tp, NULL, 0, XFS_BTNUM_BMAP,
-					ip, whichfork);
+				cur = xfs_bmbt_init_cursor(mp,
+					tp, ip, whichfork);
 				cur->bc_private.b.firstblock =
 					*firstblock;
 				cur->bc_private.b.flist = flist;
@@ -5440,8 +5504,7 @@ xfs_bunmapi(
 	logflags = 0;
 	if (ifp->if_flags & XFS_IFBROOT) {
 		ASSERT(XFS_IFORK_FORMAT(ip, whichfork) == XFS_DINODE_FMT_BTREE);
-		cur = xfs_btree_init_cursor(mp, tp, NULL, 0, XFS_BTNUM_BMAP, ip,
-			whichfork);
+		cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
 		cur->bc_private.b.firstblock = *firstblock;
 		cur->bc_private.b.flist = flist;
 		cur->bc_private.b.flags = 0;
@@ -6131,7 +6194,7 @@ xfs_bmap_get_bp(
 
 void
 xfs_check_block(
-	xfs_bmbt_block_t        *block,
+	struct xfs_btree_block	*block,
 	xfs_mount_t		*mp,
 	int			root,
 	short			sz)
@@ -6143,36 +6206,29 @@ xfs_check_block(
 	ASSERT(be16_to_cpu(block->bb_level) > 0);
 
 	prevp = NULL;
-	for( i = 1; i <= be16_to_cpu(block->bb_numrecs); i++) {
+	for( i = 1; i <= xfs_btree_get_numrecs(block); i++) {
 		dmxr = mp->m_bmap_dmxr[0];
-
-		if (root) {
-			keyp = XFS_BMAP_BROOT_KEY_ADDR(block, i, sz);
-		} else {
-			keyp = XFS_BTREE_KEY_ADDR(xfs_bmbt, block, i);
-		}
+		keyp = XFS_BMBT_KEY_ADDR(mp, block, i);
 
 		if (prevp) {
-			xfs_btree_check_key(XFS_BTNUM_BMAP, prevp, keyp);
+			ASSERT(be64_to_cpu(prevp->br_startoff) <
+			       be64_to_cpu(keyp->br_startoff));
 		}
 		prevp = keyp;
 
 		/*
 		 * Compare the block numbers to see if there are dups.
 		 */
+		if (root)
+			pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, i, sz);
+		else
+			pp = XFS_BMBT_PTR_ADDR(mp, block, i, dmxr);
 
-		if (root) {
-			pp = XFS_BMAP_BROOT_PTR_ADDR(block, i, sz);
-		} else {
-			pp = XFS_BTREE_PTR_ADDR(xfs_bmbt, block, i, dmxr);
-		}
 		for (j = i+1; j <= be16_to_cpu(block->bb_numrecs); j++) {
-			if (root) {
-				thispa = XFS_BMAP_BROOT_PTR_ADDR(block, j, sz);
-			} else {
-				thispa = XFS_BTREE_PTR_ADDR(xfs_bmbt, block, j,
-							    dmxr);
-			}
+			if (root)
+				thispa = XFS_BMAP_BROOT_PTR_ADDR(mp, block, j, sz);
+			else
+				thispa = XFS_BMBT_PTR_ADDR(mp, block, j, dmxr);
 			if (*thispa == *pp) {
 				cmn_err(CE_WARN, "%s: thispa(%d) == pp(%d) %Ld",
 					__func__, j, i,
@@ -6195,7 +6251,7 @@ xfs_bmap_check_leaf_extents(
 	xfs_inode_t		*ip,		/* incore inode pointer */
 	int			whichfork)	/* data or attr fork */
 {
-	xfs_bmbt_block_t	*block;	/* current btree block */
+	struct xfs_btree_block	*block;	/* current btree block */
 	xfs_fsblock_t		bno;	/* block # of "block" */
 	xfs_buf_t		*bp;	/* buffer for "block" */
 	int			error;	/* error return value */
@@ -6223,7 +6279,7 @@ xfs_bmap_check_leaf_extents(
 	level = be16_to_cpu(block->bb_level);
 	ASSERT(level > 0);
 	xfs_check_block(block, mp, 1, ifp->if_broot_bytes);
-	pp = XFS_BMAP_BROOT_PTR_ADDR(block, 1, ifp->if_broot_bytes);
+	pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
 	bno = be64_to_cpu(*pp);
 
 	ASSERT(bno != NULLDFSBNO);
@@ -6245,9 +6301,9 @@ xfs_bmap_check_leaf_extents(
 		if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
 				XFS_BMAP_BTREE_REF)))
 			goto error_norelse;
-		block = XFS_BUF_TO_BMBT_BLOCK(bp);
+		block = XFS_BUF_TO_BLOCK(bp);
 		XFS_WANT_CORRUPTED_GOTO(
-			XFS_BMAP_SANITY_CHECK(mp, block, level),
+			xfs_bmap_sanity_check(mp, bp, level),
 			error0);
 		if (level == 0)
 			break;
@@ -6258,7 +6314,7 @@ xfs_bmap_check_leaf_extents(
 		 */
 
 		xfs_check_block(block, mp, 0, 0);
-		pp = XFS_BTREE_PTR_ADDR(xfs_bmbt, block, 1, mp->m_bmap_dmxr[1]);
+		pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]);
 		bno = be64_to_cpu(*pp);
 		XFS_WANT_CORRUPTED_GOTO(XFS_FSB_SANITY_CHECK(mp, bno), error0);
 		if (bp_release) {
@@ -6280,13 +6336,13 @@ xfs_bmap_check_leaf_extents(
 		xfs_extnum_t	num_recs;
 
 
-		num_recs = be16_to_cpu(block->bb_numrecs);
+		num_recs = xfs_btree_get_numrecs(block);
 
 		/*
 		 * Read-ahead the next leaf block, if any.
 		 */
 
-		nextbno = be64_to_cpu(block->bb_rightsib);
+		nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
 
 		/*
 		 * Check all the extents to make sure they are OK.
@@ -6294,13 +6350,17 @@ xfs_bmap_check_leaf_extents(
 		 * conform with the first entry in this one.
 		 */
 
-		ep = XFS_BTREE_REC_ADDR(xfs_bmbt, block, 1);
+		ep = XFS_BMBT_REC_ADDR(mp, block, 1);
 		if (i) {
-			xfs_btree_check_rec(XFS_BTNUM_BMAP, &last, ep);
+			ASSERT(xfs_bmbt_disk_get_startoff(&last) +
+			       xfs_bmbt_disk_get_blockcount(&last) <=
+			       xfs_bmbt_disk_get_startoff(ep));
 		}
 		for (j = 1; j < num_recs; j++) {
-			nextp = XFS_BTREE_REC_ADDR(xfs_bmbt, block, j + 1);
-			xfs_btree_check_rec(XFS_BTNUM_BMAP, ep, nextp);
+			nextp = XFS_BMBT_REC_ADDR(mp, block, j + 1);
+			ASSERT(xfs_bmbt_disk_get_startoff(ep) +
+			       xfs_bmbt_disk_get_blockcount(ep) <=
+			       xfs_bmbt_disk_get_startoff(nextp));
 			ep = nextp;
 		}
 
@@ -6326,7 +6386,7 @@ xfs_bmap_check_leaf_extents(
 		if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
 				XFS_BMAP_BTREE_REF)))
 			goto error_norelse;
-		block = XFS_BUF_TO_BMBT_BLOCK(bp);
+		block = XFS_BUF_TO_BLOCK(bp);
 	}
 	if (bp_release) {
 		bp_release = 0;
@@ -6356,7 +6416,7 @@ xfs_bmap_count_blocks(
 	int			whichfork,	/* data or attr fork */
 	int			*count)		/* out: count of blocks */
 {
-	xfs_bmbt_block_t	*block;	/* current btree block */
+	struct xfs_btree_block	*block;	/* current btree block */
 	xfs_fsblock_t		bno;	/* block # of "block" */
 	xfs_ifork_t		*ifp;	/* fork structure */
 	int			level;	/* btree level, for checking */
@@ -6379,7 +6439,7 @@ xfs_bmap_count_blocks(
 	block = ifp->if_broot;
 	level = be16_to_cpu(block->bb_level);
 	ASSERT(level > 0);
-	pp = XFS_BMAP_BROOT_PTR_ADDR(block, 1, ifp->if_broot_bytes);
+	pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
 	bno = be64_to_cpu(*pp);
 	ASSERT(bno != NULLDFSBNO);
 	ASSERT(XFS_FSB_TO_AGNO(mp, bno) < mp->m_sb.sb_agcount);
@@ -6413,29 +6473,29 @@ xfs_bmap_count_tree(
 	__be64			*pp;
 	xfs_fsblock_t           bno = blockno;
 	xfs_fsblock_t		nextbno;
-	xfs_bmbt_block_t        *block, *nextblock;
+	struct xfs_btree_block	*block, *nextblock;
 	int			numrecs;
 
 	if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp, XFS_BMAP_BTREE_REF)))
 		return error;
 	*count += 1;
-	block = XFS_BUF_TO_BMBT_BLOCK(bp);
+	block = XFS_BUF_TO_BLOCK(bp);
 
 	if (--level) {
 		/* Not at node above leafs, count this level of nodes */
-		nextbno = be64_to_cpu(block->bb_rightsib);
+		nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
 		while (nextbno != NULLFSBLOCK) {
 			if ((error = xfs_btree_read_bufl(mp, tp, nextbno,
 				0, &nbp, XFS_BMAP_BTREE_REF)))
 				return error;
 			*count += 1;
-			nextblock = XFS_BUF_TO_BMBT_BLOCK(nbp);
-			nextbno = be64_to_cpu(nextblock->bb_rightsib);
+			nextblock = XFS_BUF_TO_BLOCK(nbp);
+			nextbno = be64_to_cpu(nextblock->bb_u.l.bb_rightsib);
 			xfs_trans_brelse(tp, nbp);
 		}
 
 		/* Dive to the next level */
-		pp = XFS_BTREE_PTR_ADDR(xfs_bmbt, block, 1, mp->m_bmap_dmxr[1]);
+		pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]);
 		bno = be64_to_cpu(*pp);
 		if (unlikely((error =
 		     xfs_bmap_count_tree(mp, tp, ifp, bno, level, count)) < 0)) {
@@ -6448,9 +6508,9 @@ xfs_bmap_count_tree(
 	} else {
 		/* count all level 1 nodes and their leaves */
 		for (;;) {
-			nextbno = be64_to_cpu(block->bb_rightsib);
+			nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
 			numrecs = be16_to_cpu(block->bb_numrecs);
-			xfs_bmap_disk_count_leaves(0, block, numrecs, count);
+			xfs_bmap_disk_count_leaves(mp, block, numrecs, count);
 			xfs_trans_brelse(tp, bp);
 			if (nextbno == NULLFSBLOCK)
 				break;
@@ -6459,7 +6519,7 @@ xfs_bmap_count_tree(
 				XFS_BMAP_BTREE_REF)))
 				return error;
 			*count += 1;
-			block = XFS_BUF_TO_BMBT_BLOCK(bp);
+			block = XFS_BUF_TO_BLOCK(bp);
 		}
 	}
 	return 0;
@@ -6489,8 +6549,8 @@ xfs_bmap_count_leaves(
  */
 STATIC void
 xfs_bmap_disk_count_leaves(
-	xfs_extnum_t		idx,
-	xfs_bmbt_block_t	*block,
+	struct xfs_mount	*mp,
+	struct xfs_btree_block	*block,
 	int			numrecs,
 	int			*count)
 {
@@ -6498,7 +6558,7 @@ xfs_bmap_disk_count_leaves(
 	xfs_bmbt_rec_t	*frp;
 
 	for (b = 1; b <= numrecs; b++) {
-		frp = XFS_BTREE_REC_ADDR(xfs_bmbt, block, idx + b);
+		frp = XFS_BMBT_REC_ADDR(mp, block, b);
 		*count += xfs_bmbt_disk_get_blockcount(frp);
 	}
 }

fs/xfs/xfs_bmap.h

@@ -137,9 +137,7 @@ typedef struct xfs_bmalloca {
 	char			conv;	/* overwriting unwritten extents */
 } xfs_bmalloca_t;
 
-#ifdef __KERNEL__
-
-#if defined(XFS_BMAP_TRACE)
+#if defined(__KERNEL__) && defined(XFS_BMAP_TRACE)
 /*
  * Trace operations for bmap extent tracing
  */
@@ -163,9 +161,12 @@ xfs_bmap_trace_exlist(
 	int			whichfork);	/* data or attr fork */
 #define	XFS_BMAP_TRACE_EXLIST(ip,c,w)	\
 	xfs_bmap_trace_exlist(__func__,ip,c,w)
-#else
+
+#else	/* __KERNEL__ && XFS_BMAP_TRACE */
+
 #define	XFS_BMAP_TRACE_EXLIST(ip,c,w)
-#endif
+
+#endif	/* __KERNEL__ && XFS_BMAP_TRACE */
 
 /*
  * Convert inode from non-attributed to attributed.
@@ -205,20 +206,6 @@ xfs_bmap_compute_maxlevels(
 	struct xfs_mount	*mp,	/* file system mount structure */
 	int			whichfork);	/* data or attr fork */
 
-/*
- * Routine to be called at transaction's end by xfs_bmapi, xfs_bunmapi
- * caller.  Frees all the extents that need freeing, which must be done
- * last due to locking considerations.
- *
- * Return 1 if the given transaction was committed and a new one allocated,
- * and 0 otherwise.
- */
-int						/* error */
-xfs_bmap_finish(
-	struct xfs_trans	**tp,		/* transaction pointer addr */
-	xfs_bmap_free_t		*flist,		/* i/o: list extents to free */
-	int			*committed);	/* xact committed or not */
-
 /*
  * Returns the file-relative block number of the first unused block in the file.
  * This is the lowest-address hole if the file has holes, else the first block
@@ -343,6 +330,32 @@ xfs_bunmapi(
 						   extents */
 	int			*done);		/* set if not done yet */
 
+/*
+ * Check an extent list, which has just been read, for
+ * any bit in the extent flag field.
+ */
+int
+xfs_check_nostate_extents(
+	struct xfs_ifork	*ifp,
+	xfs_extnum_t		idx,
+	xfs_extnum_t		num);
+
+#ifdef __KERNEL__
+
+/*
+ * Routine to be called at transaction's end by xfs_bmapi, xfs_bunmapi
+ * caller.  Frees all the extents that need freeing, which must be done
+ * last due to locking considerations.
+ *
+ * Return 1 if the given transaction was committed and a new one allocated,
+ * and 0 otherwise.
+ */
+int						/* error */
+xfs_bmap_finish(
+	struct xfs_trans	**tp,		/* transaction pointer addr */
+	xfs_bmap_free_t		*flist,		/* i/o: list extents to free */
+	int			*committed);	/* xact committed or not */
+
 /*
  * Fcntl interface to xfs_bmapi.
  */
@@ -374,16 +387,6 @@ xfs_bmap_count_blocks(
 	int			whichfork,
 	int			*count);
 
-/*
- * Check an extent list, which has just been read, for
- * any bit in the extent flag field.
- */
-int
-xfs_check_nostate_extents(
-	struct xfs_ifork	*ifp,
-	xfs_extnum_t		idx,
-	xfs_extnum_t		num);
-
 /*
  * Search the extent records for the entry containing block bno.
  * If bno lies in a hole, point to the next entry.  If bno lies

fs/xfs/xfs_bmap_btree.h

@@ -21,9 +21,10 @@
 #define XFS_BMAP_MAGIC	0x424d4150	/* 'BMAP' */
 
 struct xfs_btree_cur;
-struct xfs_btree_lblock;
+struct xfs_btree_block;
 struct xfs_mount;
 struct xfs_inode;
+struct xfs_trans;
 
 /*
  * Bmap root header, on-disk form only.
@@ -145,71 +146,60 @@ typedef struct xfs_bmbt_key {
 /* btree pointer type */
 typedef __be64 xfs_bmbt_ptr_t, xfs_bmdr_ptr_t;
 
-/* btree block header type */
-typedef struct xfs_btree_lblock xfs_bmbt_block_t;
-
-#define XFS_BUF_TO_BMBT_BLOCK(bp)	((xfs_bmbt_block_t *)XFS_BUF_PTR(bp))
-
-#define XFS_BMAP_RBLOCK_DSIZE(lev,cur)	((cur)->bc_private.b.forksize)
-#define XFS_BMAP_RBLOCK_ISIZE(lev,cur)	\
-	((int)XFS_IFORK_PTR((cur)->bc_private.b.ip, \
-		    (cur)->bc_private.b.whichfork)->if_broot_bytes)
-
-#define XFS_BMAP_BLOCK_DMAXRECS(lev,cur) \
-	(((lev) == (cur)->bc_nlevels - 1 ? \
-		XFS_BTREE_BLOCK_MAXRECS(XFS_BMAP_RBLOCK_DSIZE(lev,cur), \
-			xfs_bmdr, (lev) == 0) : \
-		((cur)->bc_mp->m_bmap_dmxr[(lev) != 0])))
-#define XFS_BMAP_BLOCK_IMAXRECS(lev,cur) \
-	(((lev) == (cur)->bc_nlevels - 1 ? \
-			XFS_BTREE_BLOCK_MAXRECS(XFS_BMAP_RBLOCK_ISIZE(lev,cur),\
-				xfs_bmbt, (lev) == 0) : \
-			((cur)->bc_mp->m_bmap_dmxr[(lev) != 0])))
-
-#define XFS_BMAP_BLOCK_DMINRECS(lev,cur) \
-	(((lev) == (cur)->bc_nlevels - 1 ? \
-			XFS_BTREE_BLOCK_MINRECS(XFS_BMAP_RBLOCK_DSIZE(lev,cur),\
-				xfs_bmdr, (lev) == 0) : \
-			((cur)->bc_mp->m_bmap_dmnr[(lev) != 0])))
-#define XFS_BMAP_BLOCK_IMINRECS(lev,cur) \
-	(((lev) == (cur)->bc_nlevels - 1 ? \
-			XFS_BTREE_BLOCK_MINRECS(XFS_BMAP_RBLOCK_ISIZE(lev,cur),\
-				xfs_bmbt, (lev) == 0) : \
-			((cur)->bc_mp->m_bmap_dmnr[(lev) != 0])))
-
-#define XFS_BMAP_REC_DADDR(bb,i,cur)	(XFS_BTREE_REC_ADDR(xfs_bmbt, bb, i))
-
-#define XFS_BMAP_REC_IADDR(bb,i,cur)	(XFS_BTREE_REC_ADDR(xfs_bmbt, bb, i))
-
-#define XFS_BMAP_KEY_DADDR(bb,i,cur)	\
-	(XFS_BTREE_KEY_ADDR(xfs_bmbt, bb, i))
-
-#define XFS_BMAP_KEY_IADDR(bb,i,cur)	\
-	(XFS_BTREE_KEY_ADDR(xfs_bmbt, bb, i))
-
-#define XFS_BMAP_PTR_DADDR(bb,i,cur)	\
-	(XFS_BTREE_PTR_ADDR(xfs_bmbt, bb, i, XFS_BMAP_BLOCK_DMAXRECS(	\
-				be16_to_cpu((bb)->bb_level), cur)))
-#define XFS_BMAP_PTR_IADDR(bb,i,cur)	\
-	(XFS_BTREE_PTR_ADDR(xfs_bmbt, bb, i, XFS_BMAP_BLOCK_IMAXRECS(	\
-				be16_to_cpu((bb)->bb_level), cur)))
+/*
+ * Btree block header size depends on a superblock flag.
+ *
+ * (not quite yet, but soon)
+ */
+#define XFS_BMBT_BLOCK_LEN(mp)	XFS_BTREE_LBLOCK_LEN
+
+#define XFS_BMBT_REC_ADDR(mp, block, index) \
+	((xfs_bmbt_rec_t *) \
+		((char *)(block) + \
+		 XFS_BMBT_BLOCK_LEN(mp) + \
+		 ((index) - 1) * sizeof(xfs_bmbt_rec_t)))
+
+#define XFS_BMBT_KEY_ADDR(mp, block, index) \
+	((xfs_bmbt_key_t *) \
+		((char *)(block) + \
+		 XFS_BMBT_BLOCK_LEN(mp) + \
+		 ((index) - 1) * sizeof(xfs_bmbt_key_t)))
+
+#define XFS_BMBT_PTR_ADDR(mp, block, index, maxrecs) \
+	((xfs_bmbt_ptr_t *) \
+		((char *)(block) + \
+		 XFS_BMBT_BLOCK_LEN(mp) + \
+		 (maxrecs) * sizeof(xfs_bmbt_key_t) + \
+		 ((index) - 1) * sizeof(xfs_bmbt_ptr_t)))
+
+#define XFS_BMDR_REC_ADDR(block, index) \
+	((xfs_bmdr_rec_t *) \
+		((char *)(block) + \
+		 sizeof(struct xfs_bmdr_block) + \
+	         ((index) - 1) * sizeof(xfs_bmdr_rec_t)))
+
+#define XFS_BMDR_KEY_ADDR(block, index) \
+	((xfs_bmdr_key_t *) \
+		((char *)(block) + \
+		 sizeof(struct xfs_bmdr_block) + \
+		 ((index) - 1) * sizeof(xfs_bmdr_key_t)))
+
+#define XFS_BMDR_PTR_ADDR(block, index, maxrecs) \
+	((xfs_bmdr_ptr_t *) \
+		((char *)(block) + \
+		 sizeof(struct xfs_bmdr_block) + \
+		 (maxrecs) * sizeof(xfs_bmdr_key_t) + \
+		 ((index) - 1) * sizeof(xfs_bmdr_ptr_t)))
 
 /*
  * These are to be used when we know the size of the block and
  * we don't have a cursor.
  */
-#define XFS_BMAP_BROOT_REC_ADDR(bb,i,sz) \
-	(XFS_BTREE_REC_ADDR(xfs_bmbt,bb,i))
-#define XFS_BMAP_BROOT_KEY_ADDR(bb,i,sz) \
-	(XFS_BTREE_KEY_ADDR(xfs_bmbt,bb,i))
-#define XFS_BMAP_BROOT_PTR_ADDR(bb,i,sz) \
-	(XFS_BTREE_PTR_ADDR(xfs_bmbt,bb,i,XFS_BMAP_BROOT_MAXRECS(sz)))
-
-#define XFS_BMAP_BROOT_NUMRECS(bb)	be16_to_cpu((bb)->bb_numrecs)
-#define XFS_BMAP_BROOT_MAXRECS(sz)	XFS_BTREE_BLOCK_MAXRECS(sz,xfs_bmbt,0)
+#define XFS_BMAP_BROOT_PTR_ADDR(mp, bb, i, sz) \
+	XFS_BMBT_PTR_ADDR(mp, bb, i, xfs_bmbt_maxrecs(mp, sz, 0))
 
 #define XFS_BMAP_BROOT_SPACE_CALC(nrecs) \
-	(int)(sizeof(xfs_bmbt_block_t) + \
+	(int)(XFS_BTREE_LBLOCK_LEN + \
 	       ((nrecs) * (sizeof(xfs_bmbt_key_t) + sizeof(xfs_bmbt_ptr_t))))
 
 #define XFS_BMAP_BROOT_SPACE(bb) \
@@ -223,42 +213,12 @@ typedef struct xfs_btree_lblock xfs_bmbt_block_t;
  */
 #define XFS_BM_MAXLEVELS(mp,w)		((mp)->m_bm_maxlevels[(w)])
 
-#define XFS_BMAP_SANITY_CHECK(mp,bb,level) \
-	(be32_to_cpu((bb)->bb_magic) == XFS_BMAP_MAGIC && \
-	 be16_to_cpu((bb)->bb_level) == level && \
-	 be16_to_cpu((bb)->bb_numrecs) > 0 && \
-	 be16_to_cpu((bb)->bb_numrecs) <= (mp)->m_bmap_dmxr[(level) != 0])
-
-
-#ifdef __KERNEL__
-
-#if defined(XFS_BMBT_TRACE)
-/*
- * Trace buffer entry types.
- */
-#define XFS_BMBT_KTRACE_ARGBI	1
-#define XFS_BMBT_KTRACE_ARGBII	2
-#define XFS_BMBT_KTRACE_ARGFFFI 3
-#define XFS_BMBT_KTRACE_ARGI	4
-#define XFS_BMBT_KTRACE_ARGIFK	5
-#define XFS_BMBT_KTRACE_ARGIFR	6
-#define XFS_BMBT_KTRACE_ARGIK	7
-#define XFS_BMBT_KTRACE_CUR	8
-
-#define XFS_BMBT_TRACE_SIZE	4096	/* size of global trace buffer */
-#define XFS_BMBT_KTRACE_SIZE	32	/* size of per-inode trace buffer */
-extern ktrace_t	*xfs_bmbt_trace_buf;
-#endif
-
 /*
  * Prototypes for xfs_bmap.c to call.
  */
-extern void xfs_bmdr_to_bmbt(xfs_bmdr_block_t *, int, xfs_bmbt_block_t *, int);
-extern int xfs_bmbt_decrement(struct xfs_btree_cur *, int, int *);
-extern int xfs_bmbt_delete(struct xfs_btree_cur *, int *);
+extern void xfs_bmdr_to_bmbt(struct xfs_mount *, xfs_bmdr_block_t *, int,
+			struct xfs_btree_block *, int);
 extern void xfs_bmbt_get_all(xfs_bmbt_rec_host_t *r, xfs_bmbt_irec_t *s);
-extern xfs_bmbt_block_t *xfs_bmbt_get_block(struct xfs_btree_cur *cur,
-						int, struct xfs_buf **bpp);
 extern xfs_filblks_t xfs_bmbt_get_blockcount(xfs_bmbt_rec_host_t *r);
 extern xfs_fsblock_t xfs_bmbt_get_startblock(xfs_bmbt_rec_host_t *r);
 extern xfs_fileoff_t xfs_bmbt_get_startoff(xfs_bmbt_rec_host_t *r);
@@ -268,22 +228,6 @@ extern void xfs_bmbt_disk_get_all(xfs_bmbt_rec_t *r, xfs_bmbt_irec_t *s);
 extern xfs_filblks_t xfs_bmbt_disk_get_blockcount(xfs_bmbt_rec_t *r);
 extern xfs_fileoff_t xfs_bmbt_disk_get_startoff(xfs_bmbt_rec_t *r);
 
-extern int xfs_bmbt_increment(struct xfs_btree_cur *, int, int *);
-extern int xfs_bmbt_insert(struct xfs_btree_cur *, int *);
-extern void xfs_bmbt_log_block(struct xfs_btree_cur *, struct xfs_buf *, int);
-extern void xfs_bmbt_log_recs(struct xfs_btree_cur *, struct xfs_buf *, int,
-				int);
-extern int xfs_bmbt_lookup_eq(struct xfs_btree_cur *, xfs_fileoff_t,
-				xfs_fsblock_t, xfs_filblks_t, int *);
-extern int xfs_bmbt_lookup_ge(struct xfs_btree_cur *, xfs_fileoff_t,
-				xfs_fsblock_t, xfs_filblks_t, int *);
-
-/*
- * Give the bmap btree a new root block.  Copy the old broot contents
- * down into a real block and make the broot point to it.
- */
-extern int xfs_bmbt_newroot(struct xfs_btree_cur *cur, int *lflags, int *stat);
-
 extern void xfs_bmbt_set_all(xfs_bmbt_rec_host_t *r, xfs_bmbt_irec_t *s);
 extern void xfs_bmbt_set_allf(xfs_bmbt_rec_host_t *r, xfs_fileoff_t o,
 			xfs_fsblock_t b, xfs_filblks_t c, xfs_exntst_t v);
@@ -296,10 +240,15 @@ extern void xfs_bmbt_disk_set_all(xfs_bmbt_rec_t *r, xfs_bmbt_irec_t *s);
 extern void xfs_bmbt_disk_set_allf(xfs_bmbt_rec_t *r, xfs_fileoff_t o,
 			xfs_fsblock_t b, xfs_filblks_t c, xfs_exntst_t v);
 
-extern void xfs_bmbt_to_bmdr(xfs_bmbt_block_t *, int, xfs_bmdr_block_t *, int);
-extern int xfs_bmbt_update(struct xfs_btree_cur *, xfs_fileoff_t,
-				xfs_fsblock_t, xfs_filblks_t, xfs_exntst_t);
+extern void xfs_bmbt_to_bmdr(struct xfs_mount *, struct xfs_btree_block *, int,
+			xfs_bmdr_block_t *, int);
+
+extern int xfs_bmbt_get_maxrecs(struct xfs_btree_cur *, int level);
+extern int xfs_bmdr_maxrecs(struct xfs_mount *, int blocklen, int leaf);
+extern int xfs_bmbt_maxrecs(struct xfs_mount *, int blocklen, int leaf);
+
+extern struct xfs_btree_cur *xfs_bmbt_init_cursor(struct xfs_mount *,
+		struct xfs_trans *, struct xfs_inode *, int);
 
-#endif	/* __KERNEL__ */
 
 #endif	/* __XFS_BMAP_BTREE_H__ */

fs/xfs/xfs_btree.c

@@ -34,7 +34,9 @@
 #include "xfs_attr_sf.h"
 #include "xfs_dinode.h"
 #include "xfs_inode.h"
+#include "xfs_inode_item.h"
 #include "xfs_btree.h"
+#include "xfs_btree_trace.h"
 #include "xfs_ialloc.h"
 #include "xfs_error.h"
 
@@ -50,135 +52,33 @@ const __uint32_t xfs_magics[XFS_BTNUM_MAX] = {
 	XFS_ABTB_MAGIC, XFS_ABTC_MAGIC, XFS_BMAP_MAGIC, XFS_IBT_MAGIC
 };
 
-/*
- * Checking routine: return maxrecs for the block.
- */
-STATIC int				/* number of records fitting in block */
-xfs_btree_maxrecs(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_btree_block_t	*block)	/* generic btree block pointer */
-{
-	switch (cur->bc_btnum) {
-	case XFS_BTNUM_BNO:
-	case XFS_BTNUM_CNT:
-		return (int)XFS_ALLOC_BLOCK_MAXRECS(
-				be16_to_cpu(block->bb_h.bb_level), cur);
-	case XFS_BTNUM_BMAP:
-		return (int)XFS_BMAP_BLOCK_IMAXRECS(
-				be16_to_cpu(block->bb_h.bb_level), cur);
-	case XFS_BTNUM_INO:
-		return (int)XFS_INOBT_BLOCK_MAXRECS(
-				be16_to_cpu(block->bb_h.bb_level), cur);
-	default:
-		ASSERT(0);
-		return 0;
-	}
-}
-
-/*
- * External routines.
- */
-
-#ifdef DEBUG
-/*
- * Debug routine: check that block header is ok.
- */
-void
-xfs_btree_check_block(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_btree_block_t	*block,	/* generic btree block pointer */
-	int			level,	/* level of the btree block */
-	xfs_buf_t		*bp)	/* buffer containing block, if any */
-{
-	if (XFS_BTREE_LONG_PTRS(cur->bc_btnum))
-		xfs_btree_check_lblock(cur, (xfs_btree_lblock_t *)block, level,
-			bp);
-	else
-		xfs_btree_check_sblock(cur, (xfs_btree_sblock_t *)block, level,
-			bp);
-}
-
-/*
- * Debug routine: check that keys are in the right order.
- */
-void
-xfs_btree_check_key(
-	xfs_btnum_t	btnum,		/* btree identifier */
-	void		*ak1,		/* pointer to left (lower) key */
-	void		*ak2)		/* pointer to right (higher) key */
-{
-	switch (btnum) {
-	case XFS_BTNUM_BNO: {
-		xfs_alloc_key_t	*k1;
-		xfs_alloc_key_t	*k2;
-
-		k1 = ak1;
-		k2 = ak2;
-		ASSERT(be32_to_cpu(k1->ar_startblock) < be32_to_cpu(k2->ar_startblock));
-		break;
-	    }
-	case XFS_BTNUM_CNT: {
-		xfs_alloc_key_t	*k1;
-		xfs_alloc_key_t	*k2;
-
-		k1 = ak1;
-		k2 = ak2;
-		ASSERT(be32_to_cpu(k1->ar_blockcount) < be32_to_cpu(k2->ar_blockcount) ||
-		       (k1->ar_blockcount == k2->ar_blockcount &&
-			be32_to_cpu(k1->ar_startblock) < be32_to_cpu(k2->ar_startblock)));
-		break;
-	    }
-	case XFS_BTNUM_BMAP: {
-		xfs_bmbt_key_t	*k1;
-		xfs_bmbt_key_t	*k2;
-
-		k1 = ak1;
-		k2 = ak2;
-		ASSERT(be64_to_cpu(k1->br_startoff) < be64_to_cpu(k2->br_startoff));
-		break;
-	    }
-	case XFS_BTNUM_INO: {
-		xfs_inobt_key_t	*k1;
-		xfs_inobt_key_t	*k2;
-
-		k1 = ak1;
-		k2 = ak2;
-		ASSERT(be32_to_cpu(k1->ir_startino) < be32_to_cpu(k2->ir_startino));
-		break;
-	    }
-	default:
-		ASSERT(0);
-	}
-}
-#endif	/* DEBUG */
 
-/*
- * Checking routine: check that long form block header is ok.
- */
-/* ARGSUSED */
-int					/* error (0 or EFSCORRUPTED) */
+STATIC int				/* error (0 or EFSCORRUPTED) */
 xfs_btree_check_lblock(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_btree_lblock_t	*block,	/* btree long form block pointer */
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	struct xfs_btree_block	*block,	/* btree long form block pointer */
 	int			level,	/* level of the btree block */
-	xfs_buf_t		*bp)	/* buffer for block, if any */
+	struct xfs_buf		*bp)	/* buffer for block, if any */
 {
 	int			lblock_ok; /* block passes checks */
-	xfs_mount_t		*mp;	/* file system mount point */
+	struct xfs_mount	*mp;	/* file system mount point */
 
 	mp = cur->bc_mp;
 	lblock_ok =
 		be32_to_cpu(block->bb_magic) == xfs_magics[cur->bc_btnum] &&
 		be16_to_cpu(block->bb_level) == level &&
 		be16_to_cpu(block->bb_numrecs) <=
-			xfs_btree_maxrecs(cur, (xfs_btree_block_t *)block) &&
-		block->bb_leftsib &&
-		(be64_to_cpu(block->bb_leftsib) == NULLDFSBNO ||
-		 XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_leftsib))) &&
-		block->bb_rightsib &&
-		(be64_to_cpu(block->bb_rightsib) == NULLDFSBNO ||
-		 XFS_FSB_SANITY_CHECK(mp, be64_to_cpu(block->bb_rightsib)));
-	if (unlikely(XFS_TEST_ERROR(!lblock_ok, mp, XFS_ERRTAG_BTREE_CHECK_LBLOCK,
+			cur->bc_ops->get_maxrecs(cur, level) &&
+		block->bb_u.l.bb_leftsib &&
+		(be64_to_cpu(block->bb_u.l.bb_leftsib) == NULLDFSBNO ||
+		 XFS_FSB_SANITY_CHECK(mp,
+		 	be64_to_cpu(block->bb_u.l.bb_leftsib))) &&
+		block->bb_u.l.bb_rightsib &&
+		(be64_to_cpu(block->bb_u.l.bb_rightsib) == NULLDFSBNO ||
+		 XFS_FSB_SANITY_CHECK(mp,
+		 	be64_to_cpu(block->bb_u.l.bb_rightsib)));
+	if (unlikely(XFS_TEST_ERROR(!lblock_ok, mp,
+			XFS_ERRTAG_BTREE_CHECK_LBLOCK,
 			XFS_RANDOM_BTREE_CHECK_LBLOCK))) {
 		if (bp)
 			xfs_buftrace("LBTREE ERROR", bp);
@@ -189,98 +89,15 @@ xfs_btree_check_lblock(
 	return 0;
 }
 
-/*
- * Checking routine: check that (long) pointer is ok.
- */
-int					/* error (0 or EFSCORRUPTED) */
-xfs_btree_check_lptr(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_dfsbno_t	ptr,		/* btree block disk address */
-	int		level)		/* btree block level */
-{
-	xfs_mount_t	*mp;		/* file system mount point */
-
-	mp = cur->bc_mp;
-	XFS_WANT_CORRUPTED_RETURN(
-		level > 0 &&
-		ptr != NULLDFSBNO &&
-		XFS_FSB_SANITY_CHECK(mp, ptr));
-	return 0;
-}
-
-#ifdef DEBUG
-/*
- * Debug routine: check that records are in the right order.
- */
-void
-xfs_btree_check_rec(
-	xfs_btnum_t	btnum,		/* btree identifier */
-	void		*ar1,		/* pointer to left (lower) record */
-	void		*ar2)		/* pointer to right (higher) record */
-{
-	switch (btnum) {
-	case XFS_BTNUM_BNO: {
-		xfs_alloc_rec_t	*r1;
-		xfs_alloc_rec_t	*r2;
-
-		r1 = ar1;
-		r2 = ar2;
-		ASSERT(be32_to_cpu(r1->ar_startblock) +
-		       be32_to_cpu(r1->ar_blockcount) <=
-		       be32_to_cpu(r2->ar_startblock));
-		break;
-	    }
-	case XFS_BTNUM_CNT: {
-		xfs_alloc_rec_t	*r1;
-		xfs_alloc_rec_t	*r2;
-
-		r1 = ar1;
-		r2 = ar2;
-		ASSERT(be32_to_cpu(r1->ar_blockcount) < be32_to_cpu(r2->ar_blockcount) ||
-		       (r1->ar_blockcount == r2->ar_blockcount &&
-			be32_to_cpu(r1->ar_startblock) < be32_to_cpu(r2->ar_startblock)));
-		break;
-	    }
-	case XFS_BTNUM_BMAP: {
-		xfs_bmbt_rec_t	*r1;
-		xfs_bmbt_rec_t	*r2;
-
-		r1 = ar1;
-		r2 = ar2;
-		ASSERT(xfs_bmbt_disk_get_startoff(r1) +
-		       xfs_bmbt_disk_get_blockcount(r1) <=
-		       xfs_bmbt_disk_get_startoff(r2));
-		break;
-	    }
-	case XFS_BTNUM_INO: {
-		xfs_inobt_rec_t	*r1;
-		xfs_inobt_rec_t	*r2;
-
-		r1 = ar1;
-		r2 = ar2;
-		ASSERT(be32_to_cpu(r1->ir_startino) + XFS_INODES_PER_CHUNK <=
-		       be32_to_cpu(r2->ir_startino));
-		break;
-	    }
-	default:
-		ASSERT(0);
-	}
-}
-#endif	/* DEBUG */
-
-/*
- * Checking routine: check that block header is ok.
- */
-/* ARGSUSED */
-int					/* error (0 or EFSCORRUPTED) */
+STATIC int				/* error (0 or EFSCORRUPTED) */
 xfs_btree_check_sblock(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_btree_sblock_t	*block,	/* btree short form block pointer */
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	struct xfs_btree_block	*block,	/* btree short form block pointer */
 	int			level,	/* level of the btree block */
-	xfs_buf_t		*bp)	/* buffer containing block */
+	struct xfs_buf		*bp)	/* buffer containing block */
 {
-	xfs_buf_t		*agbp;	/* buffer for ag. freespace struct */
-	xfs_agf_t		*agf;	/* ag. freespace structure */
+	struct xfs_buf		*agbp;	/* buffer for ag. freespace struct */
+	struct xfs_agf		*agf;	/* ag. freespace structure */
 	xfs_agblock_t		agflen;	/* native ag. freespace length */
 	int			sblock_ok; /* block passes checks */
 
@@ -291,13 +108,13 @@ xfs_btree_check_sblock(
 		be32_to_cpu(block->bb_magic) == xfs_magics[cur->bc_btnum] &&
 		be16_to_cpu(block->bb_level) == level &&
 		be16_to_cpu(block->bb_numrecs) <=
-			xfs_btree_maxrecs(cur, (xfs_btree_block_t *)block) &&
-		(be32_to_cpu(block->bb_leftsib) == NULLAGBLOCK ||
-		 be32_to_cpu(block->bb_leftsib) < agflen) &&
-		block->bb_leftsib &&
-		(be32_to_cpu(block->bb_rightsib) == NULLAGBLOCK ||
-		 be32_to_cpu(block->bb_rightsib) < agflen) &&
-		block->bb_rightsib;
+			cur->bc_ops->get_maxrecs(cur, level) &&
+		(be32_to_cpu(block->bb_u.s.bb_leftsib) == NULLAGBLOCK ||
+		 be32_to_cpu(block->bb_u.s.bb_leftsib) < agflen) &&
+		block->bb_u.s.bb_leftsib &&
+		(be32_to_cpu(block->bb_u.s.bb_rightsib) == NULLAGBLOCK ||
+		 be32_to_cpu(block->bb_u.s.bb_rightsib) < agflen) &&
+		block->bb_u.s.bb_rightsib;
 	if (unlikely(XFS_TEST_ERROR(!sblock_ok, cur->bc_mp,
 			XFS_ERRTAG_BTREE_CHECK_SBLOCK,
 			XFS_RANDOM_BTREE_CHECK_SBLOCK))) {
@@ -311,26 +128,77 @@ xfs_btree_check_sblock(
 }
 
 /*
- * Checking routine: check that (short) pointer is ok.
+ * Debug routine: check that block header is ok.
+ */
+int
+xfs_btree_check_block(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	struct xfs_btree_block	*block,	/* generic btree block pointer */
+	int			level,	/* level of the btree block */
+	struct xfs_buf		*bp)	/* buffer containing block, if any */
+{
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		return xfs_btree_check_lblock(cur, block, level, bp);
+	else
+		return xfs_btree_check_sblock(cur, block, level, bp);
+}
+
+/*
+ * Check that (long) pointer is ok.
  */
 int					/* error (0 or EFSCORRUPTED) */
+xfs_btree_check_lptr(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_dfsbno_t		bno,	/* btree block disk address */
+	int			level)	/* btree block level */
+{
+	XFS_WANT_CORRUPTED_RETURN(
+		level > 0 &&
+		bno != NULLDFSBNO &&
+		XFS_FSB_SANITY_CHECK(cur->bc_mp, bno));
+	return 0;
+}
+
+#ifdef DEBUG
+/*
+ * Check that (short) pointer is ok.
+ */
+STATIC int				/* error (0 or EFSCORRUPTED) */
 xfs_btree_check_sptr(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_agblock_t	ptr,		/* btree block disk address */
-	int		level)		/* btree block level */
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agblock_t		bno,	/* btree block disk address */
+	int			level)	/* btree block level */
 {
-	xfs_buf_t	*agbp;		/* buffer for ag. freespace struct */
-	xfs_agf_t	*agf;		/* ag. freespace structure */
+	xfs_agblock_t		agblocks = cur->bc_mp->m_sb.sb_agblocks;
 
-	agbp = cur->bc_private.a.agbp;
-	agf = XFS_BUF_TO_AGF(agbp);
 	XFS_WANT_CORRUPTED_RETURN(
 		level > 0 &&
-		ptr != NULLAGBLOCK && ptr != 0 &&
-		ptr < be32_to_cpu(agf->agf_length));
+		bno != NULLAGBLOCK &&
+		bno != 0 &&
+		bno < agblocks);
 	return 0;
 }
 
+/*
+ * Check that block ptr is ok.
+ */
+STATIC int				/* error (0 or EFSCORRUPTED) */
+xfs_btree_check_ptr(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	union xfs_btree_ptr	*ptr,	/* btree block disk address */
+	int			index,	/* offset from ptr to check */
+	int			level)	/* btree block level */
+{
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		return xfs_btree_check_lptr(cur,
+				be64_to_cpu((&ptr->l)[index]), level);
+	} else {
+		return xfs_btree_check_sptr(cur,
+				be32_to_cpu((&ptr->s)[index]), level);
+	}
+}
+#endif
+
 /*
  * Delete the btree cursor.
  */
@@ -387,16 +255,17 @@ xfs_btree_dup_cursor(
 
 	tp = cur->bc_tp;
 	mp = cur->bc_mp;
+
 	/*
 	 * Allocate a new cursor like the old one.
 	 */
-	new = xfs_btree_init_cursor(mp, tp, cur->bc_private.a.agbp,
-		cur->bc_private.a.agno, cur->bc_btnum, cur->bc_private.b.ip,
-		cur->bc_private.b.whichfork);
+	new = cur->bc_ops->dup_cursor(cur);
+
 	/*
 	 * Copy the record currently in the cursor.
 	 */
 	new->bc_rec = cur->bc_rec;
+
 	/*
 	 * For each level current, re-get the buffer and copy the ptr value.
 	 */
@@ -416,46 +285,174 @@ xfs_btree_dup_cursor(
 		} else
 			new->bc_bufs[i] = NULL;
 	}
-	/*
-	 * For bmap btrees, copy the firstblock, flist, and flags values,
-	 * since init cursor doesn't get them.
-	 */
-	if (new->bc_btnum == XFS_BTNUM_BMAP) {
-		new->bc_private.b.firstblock = cur->bc_private.b.firstblock;
-		new->bc_private.b.flist = cur->bc_private.b.flist;
-		new->bc_private.b.flags = cur->bc_private.b.flags;
-	}
 	*ncur = new;
 	return 0;
 }
 
+/*
+ * XFS btree block layout and addressing:
+ *
+ * There are two types of blocks in the btree: leaf and non-leaf blocks.
+ *
+ * The leaf record start with a header then followed by records containing
+ * the values.  A non-leaf block also starts with the same header, and
+ * then first contains lookup keys followed by an equal number of pointers
+ * to the btree blocks at the previous level.
+ *
+ *		+--------+-------+-------+-------+-------+-------+-------+
+ * Leaf:	| header | rec 1 | rec 2 | rec 3 | rec 4 | rec 5 | rec N |
+ *		+--------+-------+-------+-------+-------+-------+-------+
+ *
+ *		+--------+-------+-------+-------+-------+-------+-------+
+ * Non-Leaf:	| header | key 1 | key 2 | key N | ptr 1 | ptr 2 | ptr N |
+ *		+--------+-------+-------+-------+-------+-------+-------+
+ *
+ * The header is called struct xfs_btree_block for reasons better left unknown
+ * and comes in different versions for short (32bit) and long (64bit) block
+ * pointers.  The record and key structures are defined by the btree instances
+ * and opaque to the btree core.  The block pointers are simple disk endian
+ * integers, available in a short (32bit) and long (64bit) variant.
+ *
+ * The helpers below calculate the offset of a given record, key or pointer
+ * into a btree block (xfs_btree_*_offset) or return a pointer to the given
+ * record, key or pointer (xfs_btree_*_addr).  Note that all addressing
+ * inside the btree block is done using indices starting at one, not zero!
+ */
+
+/*
+ * Return size of the btree block header for this btree instance.
+ */
+static inline size_t xfs_btree_block_len(struct xfs_btree_cur *cur)
+{
+	return (cur->bc_flags & XFS_BTREE_LONG_PTRS) ?
+		XFS_BTREE_LBLOCK_LEN :
+		XFS_BTREE_SBLOCK_LEN;
+}
+
+/*
+ * Return size of btree block pointers for this btree instance.
+ */
+static inline size_t xfs_btree_ptr_len(struct xfs_btree_cur *cur)
+{
+	return (cur->bc_flags & XFS_BTREE_LONG_PTRS) ?
+		sizeof(__be64) : sizeof(__be32);
+}
+
+/*
+ * Calculate offset of the n-th record in a btree block.
+ */
+STATIC size_t
+xfs_btree_rec_offset(
+	struct xfs_btree_cur	*cur,
+	int			n)
+{
+	return xfs_btree_block_len(cur) +
+		(n - 1) * cur->bc_ops->rec_len;
+}
+
+/*
+ * Calculate offset of the n-th key in a btree block.
+ */
+STATIC size_t
+xfs_btree_key_offset(
+	struct xfs_btree_cur	*cur,
+	int			n)
+{
+	return xfs_btree_block_len(cur) +
+		(n - 1) * cur->bc_ops->key_len;
+}
+
+/*
+ * Calculate offset of the n-th block pointer in a btree block.
+ */
+STATIC size_t
+xfs_btree_ptr_offset(
+	struct xfs_btree_cur	*cur,
+	int			n,
+	int			level)
+{
+	return xfs_btree_block_len(cur) +
+		cur->bc_ops->get_maxrecs(cur, level) * cur->bc_ops->key_len +
+		(n - 1) * xfs_btree_ptr_len(cur);
+}
+
+/*
+ * Return a pointer to the n-th record in the btree block.
+ */
+STATIC union xfs_btree_rec *
+xfs_btree_rec_addr(
+	struct xfs_btree_cur	*cur,
+	int			n,
+	struct xfs_btree_block	*block)
+{
+	return (union xfs_btree_rec *)
+		((char *)block + xfs_btree_rec_offset(cur, n));
+}
+
+/*
+ * Return a pointer to the n-th key in the btree block.
+ */
+STATIC union xfs_btree_key *
+xfs_btree_key_addr(
+	struct xfs_btree_cur	*cur,
+	int			n,
+	struct xfs_btree_block	*block)
+{
+	return (union xfs_btree_key *)
+		((char *)block + xfs_btree_key_offset(cur, n));
+}
+
+/*
+ * Return a pointer to the n-th block pointer in the btree block.
+ */
+STATIC union xfs_btree_ptr *
+xfs_btree_ptr_addr(
+	struct xfs_btree_cur	*cur,
+	int			n,
+	struct xfs_btree_block	*block)
+{
+	int			level = xfs_btree_get_level(block);
+
+	ASSERT(block->bb_level != 0);
+
+	return (union xfs_btree_ptr *)
+		((char *)block + xfs_btree_ptr_offset(cur, n, level));
+}
+
+/*
+ * Get a the root block which is stored in the inode.
+ *
+ * For now this btree implementation assumes the btree root is always
+ * stored in the if_broot field of an inode fork.
+ */
+STATIC struct xfs_btree_block *
+xfs_btree_get_iroot(
+       struct xfs_btree_cur    *cur)
+{
+       struct xfs_ifork        *ifp;
+
+       ifp = XFS_IFORK_PTR(cur->bc_private.b.ip, cur->bc_private.b.whichfork);
+       return (struct xfs_btree_block *)ifp->if_broot;
+}
+
 /*
  * Retrieve the block pointer from the cursor at the given level.
- * This may be a bmap btree root or from a buffer.
+ * This may be an inode btree root or from a buffer.
  */
-STATIC xfs_btree_block_t *		/* generic btree block pointer */
+STATIC struct xfs_btree_block *		/* generic btree block pointer */
 xfs_btree_get_block(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
+	struct xfs_btree_cur	*cur,	/* btree cursor */
 	int			level,	/* level in btree */
-	xfs_buf_t		**bpp)	/* buffer containing the block */
-{
-	xfs_btree_block_t	*block;	/* return value */
-	xfs_buf_t		*bp;	/* return buffer */
-	xfs_ifork_t		*ifp;	/* inode fork pointer */
-	int			whichfork; /* data or attr fork */
-
-	if (cur->bc_btnum == XFS_BTNUM_BMAP && level == cur->bc_nlevels - 1) {
-		whichfork = cur->bc_private.b.whichfork;
-		ifp = XFS_IFORK_PTR(cur->bc_private.b.ip, whichfork);
-		block = (xfs_btree_block_t *)ifp->if_broot;
-		bp = NULL;
-	} else {
-		bp = cur->bc_bufs[level];
-		block = XFS_BUF_TO_BLOCK(bp);
+	struct xfs_buf		**bpp)	/* buffer containing the block */
+{
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+	    (level == cur->bc_nlevels - 1)) {
+		*bpp = NULL;
+		return xfs_btree_get_iroot(cur);
 	}
-	ASSERT(block != NULL);
-	*bpp = bp;
-	return block;
+
+	*bpp = cur->bc_bufs[level];
+	return XFS_BUF_TO_BLOCK(*bpp);
 }
 
 /*
@@ -504,97 +501,6 @@ xfs_btree_get_bufs(
 	return bp;
 }
 
-/*
- * Allocate a new btree cursor.
- * The cursor is either for allocation (A) or bmap (B) or inodes (I).
- */
-xfs_btree_cur_t *			/* new btree cursor */
-xfs_btree_init_cursor(
-	xfs_mount_t	*mp,		/* file system mount point */
-	xfs_trans_t	*tp,		/* transaction pointer */
-	xfs_buf_t	*agbp,		/* (A only) buffer for agf structure */
-					/* (I only) buffer for agi structure */
-	xfs_agnumber_t	agno,		/* (AI only) allocation group number */
-	xfs_btnum_t	btnum,		/* btree identifier */
-	xfs_inode_t	*ip,		/* (B only) inode owning the btree */
-	int		whichfork)	/* (B only) data or attr fork */
-{
-	xfs_agf_t	*agf;		/* (A) allocation group freespace */
-	xfs_agi_t	*agi;		/* (I) allocation group inodespace */
-	xfs_btree_cur_t	*cur;		/* return value */
-	xfs_ifork_t	*ifp;		/* (I) inode fork pointer */
-	int		nlevels=0;	/* number of levels in the btree */
-
-	ASSERT(xfs_btree_cur_zone != NULL);
-	/*
-	 * Allocate a new cursor.
-	 */
-	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
-	/*
-	 * Deduce the number of btree levels from the arguments.
-	 */
-	switch (btnum) {
-	case XFS_BTNUM_BNO:
-	case XFS_BTNUM_CNT:
-		agf = XFS_BUF_TO_AGF(agbp);
-		nlevels = be32_to_cpu(agf->agf_levels[btnum]);
-		break;
-	case XFS_BTNUM_BMAP:
-		ifp = XFS_IFORK_PTR(ip, whichfork);
-		nlevels = be16_to_cpu(ifp->if_broot->bb_level) + 1;
-		break;
-	case XFS_BTNUM_INO:
-		agi = XFS_BUF_TO_AGI(agbp);
-		nlevels = be32_to_cpu(agi->agi_level);
-		break;
-	default:
-		ASSERT(0);
-	}
-	/*
-	 * Fill in the common fields.
-	 */
-	cur->bc_tp = tp;
-	cur->bc_mp = mp;
-	cur->bc_nlevels = nlevels;
-	cur->bc_btnum = btnum;
-	cur->bc_blocklog = mp->m_sb.sb_blocklog;
-	/*
-	 * Fill in private fields.
-	 */
-	switch (btnum) {
-	case XFS_BTNUM_BNO:
-	case XFS_BTNUM_CNT:
-		/*
-		 * Allocation btree fields.
-		 */
-		cur->bc_private.a.agbp = agbp;
-		cur->bc_private.a.agno = agno;
-		break;
-	case XFS_BTNUM_INO:
-		/*
-		 * Inode allocation btree fields.
-		 */
-		cur->bc_private.a.agbp = agbp;
-		cur->bc_private.a.agno = agno;
-		break;
-	case XFS_BTNUM_BMAP:
-		/*
-		 * Bmap btree fields.
-		 */
-		cur->bc_private.b.forksize = XFS_IFORK_SIZE(ip, whichfork);
-		cur->bc_private.b.ip = ip;
-		cur->bc_private.b.firstblock = NULLFSBLOCK;
-		cur->bc_private.b.flist = NULL;
-		cur->bc_private.b.allocated = 0;
-		cur->bc_private.b.flags = 0;
-		cur->bc_private.b.whichfork = whichfork;
-		break;
-	default:
-		ASSERT(0);
-	}
-	return cur;
-}
-
 /*
  * Check for the cursor referring to the last block at the given level.
  */
@@ -603,12 +509,12 @@ xfs_btree_islastblock(
 	xfs_btree_cur_t		*cur,	/* btree cursor */
 	int			level)	/* level to check */
 {
-	xfs_btree_block_t	*block;	/* generic btree block pointer */
+	struct xfs_btree_block	*block;	/* generic btree block pointer */
 	xfs_buf_t		*bp;	/* buffer containing block */
 
 	block = xfs_btree_get_block(cur, level, &bp);
 	xfs_btree_check_block(cur, block, level, bp);
-	if (XFS_BTREE_LONG_PTRS(cur->bc_btnum))
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
 		return be64_to_cpu(block->bb_u.l.bb_rightsib) == NULLDFSBNO;
 	else
 		return be32_to_cpu(block->bb_u.s.bb_rightsib) == NULLAGBLOCK;
@@ -618,12 +524,12 @@ xfs_btree_islastblock(
  * Change the cursor to point to the first record at the given level.
  * Other levels are unaffected.
  */
-int					/* success=1, failure=0 */
+STATIC int				/* success=1, failure=0 */
 xfs_btree_firstrec(
 	xfs_btree_cur_t		*cur,	/* btree cursor */
 	int			level)	/* level to change */
 {
-	xfs_btree_block_t	*block;	/* generic btree block pointer */
+	struct xfs_btree_block	*block;	/* generic btree block pointer */
 	xfs_buf_t		*bp;	/* buffer containing block */
 
 	/*
@@ -634,7 +540,7 @@ xfs_btree_firstrec(
 	/*
 	 * It's empty, there is no such record.
 	 */
-	if (!block->bb_h.bb_numrecs)
+	if (!block->bb_numrecs)
 		return 0;
 	/*
 	 * Set the ptr value to 1, that's the first record/key.
@@ -647,12 +553,12 @@ xfs_btree_firstrec(
  * Change the cursor to point to the last record in the current block
  * at the given level.  Other levels are unaffected.
  */
-int					/* success=1, failure=0 */
+STATIC int				/* success=1, failure=0 */
 xfs_btree_lastrec(
 	xfs_btree_cur_t		*cur,	/* btree cursor */
 	int			level)	/* level to change */
 {
-	xfs_btree_block_t	*block;	/* generic btree block pointer */
+	struct xfs_btree_block	*block;	/* generic btree block pointer */
 	xfs_buf_t		*bp;	/* buffer containing block */
 
 	/*
@@ -663,12 +569,12 @@ xfs_btree_lastrec(
 	/*
 	 * It's empty, there is no such record.
 	 */
-	if (!block->bb_h.bb_numrecs)
+	if (!block->bb_numrecs)
 		return 0;
 	/*
 	 * Set the ptr value to numrecs, that's the last record/key.
 	 */
-	cur->bc_ptrs[level] = be16_to_cpu(block->bb_h.bb_numrecs);
+	cur->bc_ptrs[level] = be16_to_cpu(block->bb_numrecs);
 	return 1;
 }
 
@@ -817,80 +723,98 @@ xfs_btree_reada_bufs(
 	xfs_baread(mp->m_ddev_targp, d, mp->m_bsize * count);
 }
 
-/*
- * Read-ahead btree blocks, at the given level.
- * Bits in lr are set from XFS_BTCUR_{LEFT,RIGHT}RA.
- */
-int
-xfs_btree_readahead_core(
-	xfs_btree_cur_t		*cur,		/* btree cursor */
-	int			lev,		/* level in btree */
-	int			lr)		/* left/right bits */
+STATIC int
+xfs_btree_readahead_lblock(
+	struct xfs_btree_cur	*cur,
+	int			lr,
+	struct xfs_btree_block	*block)
 {
-	xfs_alloc_block_t	*a;
-	xfs_bmbt_block_t	*b;
-	xfs_inobt_block_t	*i;
 	int			rval = 0;
+	xfs_fsblock_t		left = be64_to_cpu(block->bb_u.l.bb_leftsib);
+	xfs_fsblock_t		right = be64_to_cpu(block->bb_u.l.bb_rightsib);
 
-	ASSERT(cur->bc_bufs[lev] != NULL);
-	cur->bc_ra[lev] |= lr;
-	switch (cur->bc_btnum) {
-	case XFS_BTNUM_BNO:
-	case XFS_BTNUM_CNT:
-		a = XFS_BUF_TO_ALLOC_BLOCK(cur->bc_bufs[lev]);
-		if ((lr & XFS_BTCUR_LEFTRA) && be32_to_cpu(a->bb_leftsib) != NULLAGBLOCK) {
-			xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
-				be32_to_cpu(a->bb_leftsib), 1);
-			rval++;
-		}
-		if ((lr & XFS_BTCUR_RIGHTRA) && be32_to_cpu(a->bb_rightsib) != NULLAGBLOCK) {
-			xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
-				be32_to_cpu(a->bb_rightsib), 1);
-			rval++;
-		}
-		break;
-	case XFS_BTNUM_BMAP:
-		b = XFS_BUF_TO_BMBT_BLOCK(cur->bc_bufs[lev]);
-		if ((lr & XFS_BTCUR_LEFTRA) && be64_to_cpu(b->bb_leftsib) != NULLDFSBNO) {
-			xfs_btree_reada_bufl(cur->bc_mp, be64_to_cpu(b->bb_leftsib), 1);
-			rval++;
-		}
-		if ((lr & XFS_BTCUR_RIGHTRA) && be64_to_cpu(b->bb_rightsib) != NULLDFSBNO) {
-			xfs_btree_reada_bufl(cur->bc_mp, be64_to_cpu(b->bb_rightsib), 1);
-			rval++;
-		}
-		break;
-	case XFS_BTNUM_INO:
-		i = XFS_BUF_TO_INOBT_BLOCK(cur->bc_bufs[lev]);
-		if ((lr & XFS_BTCUR_LEFTRA) && be32_to_cpu(i->bb_leftsib) != NULLAGBLOCK) {
-			xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
-				be32_to_cpu(i->bb_leftsib), 1);
-			rval++;
-		}
-		if ((lr & XFS_BTCUR_RIGHTRA) && be32_to_cpu(i->bb_rightsib) != NULLAGBLOCK) {
-			xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
-				be32_to_cpu(i->bb_rightsib), 1);
-			rval++;
-		}
-		break;
-	default:
-		ASSERT(0);
+	if ((lr & XFS_BTCUR_LEFTRA) && left != NULLDFSBNO) {
+		xfs_btree_reada_bufl(cur->bc_mp, left, 1);
+		rval++;
+	}
+
+	if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLDFSBNO) {
+		xfs_btree_reada_bufl(cur->bc_mp, right, 1);
+		rval++;
 	}
+
 	return rval;
 }
 
-/*
- * Set the buffer for level "lev" in the cursor to bp, releasing
- * any previous buffer.
- */
-void
-xfs_btree_setbuf(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			lev,	/* level in btree */
-	xfs_buf_t		*bp)	/* new buffer to set */
+STATIC int
+xfs_btree_readahead_sblock(
+	struct xfs_btree_cur	*cur,
+	int			lr,
+	struct xfs_btree_block *block)
 {
-	xfs_btree_block_t	*b;	/* btree block */
-	xfs_buf_t		*obp;	/* old buffer pointer */
+	int			rval = 0;
+	xfs_agblock_t		left = be32_to_cpu(block->bb_u.s.bb_leftsib);
+	xfs_agblock_t		right = be32_to_cpu(block->bb_u.s.bb_rightsib);
+
+
+	if ((lr & XFS_BTCUR_LEFTRA) && left != NULLAGBLOCK) {
+		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
+				     left, 1);
+		rval++;
+	}
+
+	if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLAGBLOCK) {
+		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
+				     right, 1);
+		rval++;
+	}
+
+	return rval;
+}
+
+/*
+ * Read-ahead btree blocks, at the given level.
+ * Bits in lr are set from XFS_BTCUR_{LEFT,RIGHT}RA.
+ */
+STATIC int
+xfs_btree_readahead(
+	struct xfs_btree_cur	*cur,		/* btree cursor */
+	int			lev,		/* level in btree */
+	int			lr)		/* left/right bits */
+{
+	struct xfs_btree_block	*block;
+
+	/*
+	 * No readahead needed if we are at the root level and the
+	 * btree root is stored in the inode.
+	 */
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+	    (lev == cur->bc_nlevels - 1))
+		return 0;
+
+	if ((cur->bc_ra[lev] | lr) == cur->bc_ra[lev])
+		return 0;
+
+	cur->bc_ra[lev] |= lr;
+	block = XFS_BUF_TO_BLOCK(cur->bc_bufs[lev]);
+
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		return xfs_btree_readahead_lblock(cur, lr, block);
+	return xfs_btree_readahead_sblock(cur, lr, block);
+}
+
+/*
+ * Set the buffer for level "lev" in the cursor to bp, releasing
+ * any previous buffer.
+ */
+void
+xfs_btree_setbuf(
+	xfs_btree_cur_t		*cur,	/* btree cursor */
+	int			lev,	/* level in btree */
+	xfs_buf_t		*bp)	/* new buffer to set */
+{
+	struct xfs_btree_block	*b;	/* btree block */
+	xfs_buf_t		*obp;	/* old buffer pointer */
 
 	obp = cur->bc_bufs[lev];
 	if (obp)
@@ -900,7 +824,7 @@ xfs_btree_setbuf(
 	if (!bp)
 		return;
 	b = XFS_BUF_TO_BLOCK(bp);
-	if (XFS_BTREE_LONG_PTRS(cur->bc_btnum)) {
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
 		if (be64_to_cpu(b->bb_u.l.bb_leftsib) == NULLDFSBNO)
 			cur->bc_ra[lev] |= XFS_BTCUR_LEFTRA;
 		if (be64_to_cpu(b->bb_u.l.bb_rightsib) == NULLDFSBNO)
@@ -912,3 +836,2855 @@ xfs_btree_setbuf(
 			cur->bc_ra[lev] |= XFS_BTCUR_RIGHTRA;
 	}
 }
+
+STATIC int
+xfs_btree_ptr_is_null(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr)
+{
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		return be64_to_cpu(ptr->l) == NULLFSBLOCK;
+	else
+		return be32_to_cpu(ptr->s) == NULLAGBLOCK;
+}
+
+STATIC void
+xfs_btree_set_ptr_null(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr)
+{
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		ptr->l = cpu_to_be64(NULLFSBLOCK);
+	else
+		ptr->s = cpu_to_be32(NULLAGBLOCK);
+}
+
+/*
+ * Get/set/init sibling pointers
+ */
+STATIC void
+xfs_btree_get_sibling(
+	struct xfs_btree_cur	*cur,
+	struct xfs_btree_block	*block,
+	union xfs_btree_ptr	*ptr,
+	int			lr)
+{
+	ASSERT(lr == XFS_BB_LEFTSIB || lr == XFS_BB_RIGHTSIB);
+
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		if (lr == XFS_BB_RIGHTSIB)
+			ptr->l = block->bb_u.l.bb_rightsib;
+		else
+			ptr->l = block->bb_u.l.bb_leftsib;
+	} else {
+		if (lr == XFS_BB_RIGHTSIB)
+			ptr->s = block->bb_u.s.bb_rightsib;
+		else
+			ptr->s = block->bb_u.s.bb_leftsib;
+	}
+}
+
+STATIC void
+xfs_btree_set_sibling(
+	struct xfs_btree_cur	*cur,
+	struct xfs_btree_block	*block,
+	union xfs_btree_ptr	*ptr,
+	int			lr)
+{
+	ASSERT(lr == XFS_BB_LEFTSIB || lr == XFS_BB_RIGHTSIB);
+
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		if (lr == XFS_BB_RIGHTSIB)
+			block->bb_u.l.bb_rightsib = ptr->l;
+		else
+			block->bb_u.l.bb_leftsib = ptr->l;
+	} else {
+		if (lr == XFS_BB_RIGHTSIB)
+			block->bb_u.s.bb_rightsib = ptr->s;
+		else
+			block->bb_u.s.bb_leftsib = ptr->s;
+	}
+}
+
+STATIC void
+xfs_btree_init_block(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			numrecs,
+	struct xfs_btree_block	*new)	/* new block */
+{
+	new->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
+	new->bb_level = cpu_to_be16(level);
+	new->bb_numrecs = cpu_to_be16(numrecs);
+
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		new->bb_u.l.bb_leftsib = cpu_to_be64(NULLFSBLOCK);
+		new->bb_u.l.bb_rightsib = cpu_to_be64(NULLFSBLOCK);
+	} else {
+		new->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
+		new->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
+	}
+}
+
+/*
+ * Return true if ptr is the last record in the btree and
+ * we need to track updateѕ to this record.  The decision
+ * will be further refined in the update_lastrec method.
+ */
+STATIC int
+xfs_btree_is_lastrec(
+	struct xfs_btree_cur	*cur,
+	struct xfs_btree_block	*block,
+	int			level)
+{
+	union xfs_btree_ptr	ptr;
+
+	if (level > 0)
+		return 0;
+	if (!(cur->bc_flags & XFS_BTREE_LASTREC_UPDATE))
+		return 0;
+
+	xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_RIGHTSIB);
+	if (!xfs_btree_ptr_is_null(cur, &ptr))
+		return 0;
+	return 1;
+}
+
+STATIC void
+xfs_btree_buf_to_ptr(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp,
+	union xfs_btree_ptr	*ptr)
+{
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
+		ptr->l = cpu_to_be64(XFS_DADDR_TO_FSB(cur->bc_mp,
+					XFS_BUF_ADDR(bp)));
+	else {
+		ptr->s = cpu_to_be32(XFS_DADDR_TO_AGBNO(cur->bc_mp,
+					XFS_BUF_ADDR(bp)));
+	}
+}
+
+STATIC xfs_daddr_t
+xfs_btree_ptr_to_daddr(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr)
+{
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		ASSERT(be64_to_cpu(ptr->l) != NULLFSBLOCK);
+
+		return XFS_FSB_TO_DADDR(cur->bc_mp, be64_to_cpu(ptr->l));
+	} else {
+		ASSERT(cur->bc_private.a.agno != NULLAGNUMBER);
+		ASSERT(be32_to_cpu(ptr->s) != NULLAGBLOCK);
+
+		return XFS_AGB_TO_DADDR(cur->bc_mp, cur->bc_private.a.agno,
+					be32_to_cpu(ptr->s));
+	}
+}
+
+STATIC void
+xfs_btree_set_refs(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp)
+{
+	switch (cur->bc_btnum) {
+	case XFS_BTNUM_BNO:
+	case XFS_BTNUM_CNT:
+		XFS_BUF_SET_VTYPE_REF(*bpp, B_FS_MAP, XFS_ALLOC_BTREE_REF);
+		break;
+	case XFS_BTNUM_INO:
+		XFS_BUF_SET_VTYPE_REF(*bpp, B_FS_INOMAP, XFS_INO_BTREE_REF);
+		break;
+	case XFS_BTNUM_BMAP:
+		XFS_BUF_SET_VTYPE_REF(*bpp, B_FS_MAP, XFS_BMAP_BTREE_REF);
+		break;
+	default:
+		ASSERT(0);
+	}
+}
+
+STATIC int
+xfs_btree_get_buf_block(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr,
+	int			flags,
+	struct xfs_btree_block	**block,
+	struct xfs_buf		**bpp)
+{
+	struct xfs_mount	*mp = cur->bc_mp;
+	xfs_daddr_t		d;
+
+	/* need to sort out how callers deal with failures first */
+	ASSERT(!(flags & XFS_BUF_TRYLOCK));
+
+	d = xfs_btree_ptr_to_daddr(cur, ptr);
+	*bpp = xfs_trans_get_buf(cur->bc_tp, mp->m_ddev_targp, d,
+				 mp->m_bsize, flags);
+
+	ASSERT(*bpp);
+	ASSERT(!XFS_BUF_GETERROR(*bpp));
+
+	*block = XFS_BUF_TO_BLOCK(*bpp);
+	return 0;
+}
+
+/*
+ * Read in the buffer at the given ptr and return the buffer and
+ * the block pointer within the buffer.
+ */
+STATIC int
+xfs_btree_read_buf_block(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr,
+	int			level,
+	int			flags,
+	struct xfs_btree_block	**block,
+	struct xfs_buf		**bpp)
+{
+	struct xfs_mount	*mp = cur->bc_mp;
+	xfs_daddr_t		d;
+	int			error;
+
+	/* need to sort out how callers deal with failures first */
+	ASSERT(!(flags & XFS_BUF_TRYLOCK));
+
+	d = xfs_btree_ptr_to_daddr(cur, ptr);
+	error = xfs_trans_read_buf(mp, cur->bc_tp, mp->m_ddev_targp, d,
+				   mp->m_bsize, flags, bpp);
+	if (error)
+		return error;
+
+	ASSERT(*bpp != NULL);
+	ASSERT(!XFS_BUF_GETERROR(*bpp));
+
+	xfs_btree_set_refs(cur, *bpp);
+	*block = XFS_BUF_TO_BLOCK(*bpp);
+
+	error = xfs_btree_check_block(cur, *block, level, *bpp);
+	if (error)
+		xfs_trans_brelse(cur->bc_tp, *bpp);
+	return error;
+}
+
+/*
+ * Copy keys from one btree block to another.
+ */
+STATIC void
+xfs_btree_copy_keys(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*dst_key,
+	union xfs_btree_key	*src_key,
+	int			numkeys)
+{
+	ASSERT(numkeys >= 0);
+	memcpy(dst_key, src_key, numkeys * cur->bc_ops->key_len);
+}
+
+/*
+ * Copy records from one btree block to another.
+ */
+STATIC void
+xfs_btree_copy_recs(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*dst_rec,
+	union xfs_btree_rec	*src_rec,
+	int			numrecs)
+{
+	ASSERT(numrecs >= 0);
+	memcpy(dst_rec, src_rec, numrecs * cur->bc_ops->rec_len);
+}
+
+/*
+ * Copy block pointers from one btree block to another.
+ */
+STATIC void
+xfs_btree_copy_ptrs(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*dst_ptr,
+	union xfs_btree_ptr	*src_ptr,
+	int			numptrs)
+{
+	ASSERT(numptrs >= 0);
+	memcpy(dst_ptr, src_ptr, numptrs * xfs_btree_ptr_len(cur));
+}
+
+/*
+ * Shift keys one index left/right inside a single btree block.
+ */
+STATIC void
+xfs_btree_shift_keys(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*key,
+	int			dir,
+	int			numkeys)
+{
+	char			*dst_key;
+
+	ASSERT(numkeys >= 0);
+	ASSERT(dir == 1 || dir == -1);
+
+	dst_key = (char *)key + (dir * cur->bc_ops->key_len);
+	memmove(dst_key, key, numkeys * cur->bc_ops->key_len);
+}
+
+/*
+ * Shift records one index left/right inside a single btree block.
+ */
+STATIC void
+xfs_btree_shift_recs(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec,
+	int			dir,
+	int			numrecs)
+{
+	char			*dst_rec;
+
+	ASSERT(numrecs >= 0);
+	ASSERT(dir == 1 || dir == -1);
+
+	dst_rec = (char *)rec + (dir * cur->bc_ops->rec_len);
+	memmove(dst_rec, rec, numrecs * cur->bc_ops->rec_len);
+}
+
+/*
+ * Shift block pointers one index left/right inside a single btree block.
+ */
+STATIC void
+xfs_btree_shift_ptrs(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr,
+	int			dir,
+	int			numptrs)
+{
+	char			*dst_ptr;
+
+	ASSERT(numptrs >= 0);
+	ASSERT(dir == 1 || dir == -1);
+
+	dst_ptr = (char *)ptr + (dir * xfs_btree_ptr_len(cur));
+	memmove(dst_ptr, ptr, numptrs * xfs_btree_ptr_len(cur));
+}
+
+/*
+ * Log key values from the btree block.
+ */
+STATIC void
+xfs_btree_log_keys(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp,
+	int			first,
+	int			last)
+{
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGBII(cur, bp, first, last);
+
+	if (bp) {
+		xfs_trans_log_buf(cur->bc_tp, bp,
+				  xfs_btree_key_offset(cur, first),
+				  xfs_btree_key_offset(cur, last + 1) - 1);
+	} else {
+		xfs_trans_log_inode(cur->bc_tp, cur->bc_private.b.ip,
+				xfs_ilog_fbroot(cur->bc_private.b.whichfork));
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+}
+
+/*
+ * Log record values from the btree block.
+ */
+void
+xfs_btree_log_recs(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp,
+	int			first,
+	int			last)
+{
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGBII(cur, bp, first, last);
+
+	xfs_trans_log_buf(cur->bc_tp, bp,
+			  xfs_btree_rec_offset(cur, first),
+			  xfs_btree_rec_offset(cur, last + 1) - 1);
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+}
+
+/*
+ * Log block pointer fields from a btree block (nonleaf).
+ */
+STATIC void
+xfs_btree_log_ptrs(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	struct xfs_buf		*bp,	/* buffer containing btree block */
+	int			first,	/* index of first pointer to log */
+	int			last)	/* index of last pointer to log */
+{
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGBII(cur, bp, first, last);
+
+	if (bp) {
+		struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
+		int			level = xfs_btree_get_level(block);
+
+		xfs_trans_log_buf(cur->bc_tp, bp,
+				xfs_btree_ptr_offset(cur, first, level),
+				xfs_btree_ptr_offset(cur, last + 1, level) - 1);
+	} else {
+		xfs_trans_log_inode(cur->bc_tp, cur->bc_private.b.ip,
+			xfs_ilog_fbroot(cur->bc_private.b.whichfork));
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+}
+
+/*
+ * Log fields from a btree block header.
+ */
+void
+xfs_btree_log_block(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	struct xfs_buf		*bp,	/* buffer containing btree block */
+	int			fields)	/* mask of fields: XFS_BB_... */
+{
+	int			first;	/* first byte offset logged */
+	int			last;	/* last byte offset logged */
+	static const short	soffsets[] = {	/* table of offsets (short) */
+		offsetof(struct xfs_btree_block, bb_magic),
+		offsetof(struct xfs_btree_block, bb_level),
+		offsetof(struct xfs_btree_block, bb_numrecs),
+		offsetof(struct xfs_btree_block, bb_u.s.bb_leftsib),
+		offsetof(struct xfs_btree_block, bb_u.s.bb_rightsib),
+		XFS_BTREE_SBLOCK_LEN
+	};
+	static const short	loffsets[] = {	/* table of offsets (long) */
+		offsetof(struct xfs_btree_block, bb_magic),
+		offsetof(struct xfs_btree_block, bb_level),
+		offsetof(struct xfs_btree_block, bb_numrecs),
+		offsetof(struct xfs_btree_block, bb_u.l.bb_leftsib),
+		offsetof(struct xfs_btree_block, bb_u.l.bb_rightsib),
+		XFS_BTREE_LBLOCK_LEN
+	};
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGBI(cur, bp, fields);
+
+	if (bp) {
+		xfs_btree_offsets(fields,
+				  (cur->bc_flags & XFS_BTREE_LONG_PTRS) ?
+					loffsets : soffsets,
+				  XFS_BB_NUM_BITS, &first, &last);
+		xfs_trans_log_buf(cur->bc_tp, bp, first, last);
+	} else {
+		xfs_trans_log_inode(cur->bc_tp, cur->bc_private.b.ip,
+			xfs_ilog_fbroot(cur->bc_private.b.whichfork));
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+}
+
+/*
+ * Increment cursor by one record at the level.
+ * For nonzero levels the leaf-ward information is untouched.
+ */
+int						/* error */
+xfs_btree_increment(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			*stat)		/* success/failure */
+{
+	struct xfs_btree_block	*block;
+	union xfs_btree_ptr	ptr;
+	struct xfs_buf		*bp;
+	int			error;		/* error return value */
+	int			lev;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGI(cur, level);
+
+	ASSERT(level < cur->bc_nlevels);
+
+	/* Read-ahead to the right at this level. */
+	xfs_btree_readahead(cur, level, XFS_BTCUR_RIGHTRA);
+
+	/* Get a pointer to the btree block. */
+	block = xfs_btree_get_block(cur, level, &bp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, level, bp);
+	if (error)
+		goto error0;
+#endif
+
+	/* We're done if we remain in the block after the increment. */
+	if (++cur->bc_ptrs[level] <= xfs_btree_get_numrecs(block))
+		goto out1;
+
+	/* Fail if we just went off the right edge of the tree. */
+	xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_RIGHTSIB);
+	if (xfs_btree_ptr_is_null(cur, &ptr))
+		goto out0;
+
+	XFS_BTREE_STATS_INC(cur, increment);
+
+	/*
+	 * March up the tree incrementing pointers.
+	 * Stop when we don't go off the right edge of a block.
+	 */
+	for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
+		block = xfs_btree_get_block(cur, lev, &bp);
+
+#ifdef DEBUG
+		error = xfs_btree_check_block(cur, block, lev, bp);
+		if (error)
+			goto error0;
+#endif
+
+		if (++cur->bc_ptrs[lev] <= xfs_btree_get_numrecs(block))
+			break;
+
+		/* Read-ahead the right block for the next loop. */
+		xfs_btree_readahead(cur, lev, XFS_BTCUR_RIGHTRA);
+	}
+
+	/*
+	 * If we went off the root then we are either seriously
+	 * confused or have the tree root in an inode.
+	 */
+	if (lev == cur->bc_nlevels) {
+		if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE)
+			goto out0;
+		ASSERT(0);
+		error = EFSCORRUPTED;
+		goto error0;
+	}
+	ASSERT(lev < cur->bc_nlevels);
+
+	/*
+	 * Now walk back down the tree, fixing up the cursor's buffer
+	 * pointers and key numbers.
+	 */
+	for (block = xfs_btree_get_block(cur, lev, &bp); lev > level; ) {
+		union xfs_btree_ptr	*ptrp;
+
+		ptrp = xfs_btree_ptr_addr(cur, cur->bc_ptrs[lev], block);
+		error = xfs_btree_read_buf_block(cur, ptrp, --lev,
+							0, &block, &bp);
+		if (error)
+			goto error0;
+
+		xfs_btree_setbuf(cur, lev, bp);
+		cur->bc_ptrs[lev] = 1;
+	}
+out1:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+
+out0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 0;
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Decrement cursor by one record at the level.
+ * For nonzero levels the leaf-ward information is untouched.
+ */
+int						/* error */
+xfs_btree_decrement(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			*stat)		/* success/failure */
+{
+	struct xfs_btree_block	*block;
+	xfs_buf_t		*bp;
+	int			error;		/* error return value */
+	int			lev;
+	union xfs_btree_ptr	ptr;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGI(cur, level);
+
+	ASSERT(level < cur->bc_nlevels);
+
+	/* Read-ahead to the left at this level. */
+	xfs_btree_readahead(cur, level, XFS_BTCUR_LEFTRA);
+
+	/* We're done if we remain in the block after the decrement. */
+	if (--cur->bc_ptrs[level] > 0)
+		goto out1;
+
+	/* Get a pointer to the btree block. */
+	block = xfs_btree_get_block(cur, level, &bp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, level, bp);
+	if (error)
+		goto error0;
+#endif
+
+	/* Fail if we just went off the left edge of the tree. */
+	xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_LEFTSIB);
+	if (xfs_btree_ptr_is_null(cur, &ptr))
+		goto out0;
+
+	XFS_BTREE_STATS_INC(cur, decrement);
+
+	/*
+	 * March up the tree decrementing pointers.
+	 * Stop when we don't go off the left edge of a block.
+	 */
+	for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
+		if (--cur->bc_ptrs[lev] > 0)
+			break;
+		/* Read-ahead the left block for the next loop. */
+		xfs_btree_readahead(cur, lev, XFS_BTCUR_LEFTRA);
+	}
+
+	/*
+	 * If we went off the root then we are seriously confused.
+	 * or the root of the tree is in an inode.
+	 */
+	if (lev == cur->bc_nlevels) {
+		if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE)
+			goto out0;
+		ASSERT(0);
+		error = EFSCORRUPTED;
+		goto error0;
+	}
+	ASSERT(lev < cur->bc_nlevels);
+
+	/*
+	 * Now walk back down the tree, fixing up the cursor's buffer
+	 * pointers and key numbers.
+	 */
+	for (block = xfs_btree_get_block(cur, lev, &bp); lev > level; ) {
+		union xfs_btree_ptr	*ptrp;
+
+		ptrp = xfs_btree_ptr_addr(cur, cur->bc_ptrs[lev], block);
+		error = xfs_btree_read_buf_block(cur, ptrp, --lev,
+							0, &block, &bp);
+		if (error)
+			goto error0;
+		xfs_btree_setbuf(cur, lev, bp);
+		cur->bc_ptrs[lev] = xfs_btree_get_numrecs(block);
+	}
+out1:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+
+out0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 0;
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+STATIC int
+xfs_btree_lookup_get_block(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	int			level,	/* level in the btree */
+	union xfs_btree_ptr	*pp,	/* ptr to btree block */
+	struct xfs_btree_block	**blkp) /* return btree block */
+{
+	struct xfs_buf		*bp;	/* buffer pointer for btree block */
+	int			error = 0;
+
+	/* special case the root block if in an inode */
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+	    (level == cur->bc_nlevels - 1)) {
+		*blkp = xfs_btree_get_iroot(cur);
+		return 0;
+	}
+
+	/*
+	 * If the old buffer at this level for the disk address we are
+	 * looking for re-use it.
+	 *
+	 * Otherwise throw it away and get a new one.
+	 */
+	bp = cur->bc_bufs[level];
+	if (bp && XFS_BUF_ADDR(bp) == xfs_btree_ptr_to_daddr(cur, pp)) {
+		*blkp = XFS_BUF_TO_BLOCK(bp);
+		return 0;
+	}
+
+	error = xfs_btree_read_buf_block(cur, pp, level, 0, blkp, &bp);
+	if (error)
+		return error;
+
+	xfs_btree_setbuf(cur, level, bp);
+	return 0;
+}
+
+/*
+ * Get current search key.  For level 0 we don't actually have a key
+ * structure so we make one up from the record.  For all other levels
+ * we just return the right key.
+ */
+STATIC union xfs_btree_key *
+xfs_lookup_get_search_key(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			keyno,
+	struct xfs_btree_block	*block,
+	union xfs_btree_key	*kp)
+{
+	if (level == 0) {
+		cur->bc_ops->init_key_from_rec(kp,
+				xfs_btree_rec_addr(cur, keyno, block));
+		return kp;
+	}
+
+	return xfs_btree_key_addr(cur, keyno, block);
+}
+
+/*
+ * Lookup the record.  The cursor is made to point to it, based on dir.
+ * Return 0 if can't find any such record, 1 for success.
+ */
+int					/* error */
+xfs_btree_lookup(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_lookup_t		dir,	/* <=, ==, or >= */
+	int			*stat)	/* success/failure */
+{
+	struct xfs_btree_block	*block;	/* current btree block */
+	__int64_t		diff;	/* difference for the current key */
+	int			error;	/* error return value */
+	int			keyno;	/* current key number */
+	int			level;	/* level in the btree */
+	union xfs_btree_ptr	*pp;	/* ptr to btree block */
+	union xfs_btree_ptr	ptr;	/* ptr to btree block */
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGI(cur, dir);
+
+	XFS_BTREE_STATS_INC(cur, lookup);
+
+	block = NULL;
+	keyno = 0;
+
+	/* initialise start pointer from cursor */
+	cur->bc_ops->init_ptr_from_cur(cur, &ptr);
+	pp = &ptr;
+
+	/*
+	 * Iterate over each level in the btree, starting at the root.
+	 * For each level above the leaves, find the key we need, based
+	 * on the lookup record, then follow the corresponding block
+	 * pointer down to the next level.
+	 */
+	for (level = cur->bc_nlevels - 1, diff = 1; level >= 0; level--) {
+		/* Get the block we need to do the lookup on. */
+		error = xfs_btree_lookup_get_block(cur, level, pp, &block);
+		if (error)
+			goto error0;
+
+		if (diff == 0) {
+			/*
+			 * If we already had a key match at a higher level, we
+			 * know we need to use the first entry in this block.
+			 */
+			keyno = 1;
+		} else {
+			/* Otherwise search this block. Do a binary search. */
+
+			int	high;	/* high entry number */
+			int	low;	/* low entry number */
+
+			/* Set low and high entry numbers, 1-based. */
+			low = 1;
+			high = xfs_btree_get_numrecs(block);
+			if (!high) {
+				/* Block is empty, must be an empty leaf. */
+				ASSERT(level == 0 && cur->bc_nlevels == 1);
+
+				cur->bc_ptrs[0] = dir != XFS_LOOKUP_LE;
+				XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+				*stat = 0;
+				return 0;
+			}
+
+			/* Binary search the block. */
+			while (low <= high) {
+				union xfs_btree_key	key;
+				union xfs_btree_key	*kp;
+
+				XFS_BTREE_STATS_INC(cur, compare);
+
+				/* keyno is average of low and high. */
+				keyno = (low + high) >> 1;
+
+				/* Get current search key */
+				kp = xfs_lookup_get_search_key(cur, level,
+						keyno, block, &key);
+
+				/*
+				 * Compute difference to get next direction:
+				 *  - less than, move right
+				 *  - greater than, move left
+				 *  - equal, we're done
+				 */
+				diff = cur->bc_ops->key_diff(cur, kp);
+				if (diff < 0)
+					low = keyno + 1;
+				else if (diff > 0)
+					high = keyno - 1;
+				else
+					break;
+			}
+		}
+
+		/*
+		 * If there are more levels, set up for the next level
+		 * by getting the block number and filling in the cursor.
+		 */
+		if (level > 0) {
+			/*
+			 * If we moved left, need the previous key number,
+			 * unless there isn't one.
+			 */
+			if (diff > 0 && --keyno < 1)
+				keyno = 1;
+			pp = xfs_btree_ptr_addr(cur, keyno, block);
+
+#ifdef DEBUG
+			error = xfs_btree_check_ptr(cur, pp, 0, level);
+			if (error)
+				goto error0;
+#endif
+			cur->bc_ptrs[level] = keyno;
+		}
+	}
+
+	/* Done with the search. See if we need to adjust the results. */
+	if (dir != XFS_LOOKUP_LE && diff < 0) {
+		keyno++;
+		/*
+		 * If ge search and we went off the end of the block, but it's
+		 * not the last block, we're in the wrong block.
+		 */
+		xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_RIGHTSIB);
+		if (dir == XFS_LOOKUP_GE &&
+		    keyno > xfs_btree_get_numrecs(block) &&
+		    !xfs_btree_ptr_is_null(cur, &ptr)) {
+			int	i;
+
+			cur->bc_ptrs[0] = keyno;
+			error = xfs_btree_increment(cur, 0, &i);
+			if (error)
+				goto error0;
+			XFS_WANT_CORRUPTED_RETURN(i == 1);
+			XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+			*stat = 1;
+			return 0;
+		}
+	} else if (dir == XFS_LOOKUP_LE && diff > 0)
+		keyno--;
+	cur->bc_ptrs[0] = keyno;
+
+	/* Return if we succeeded or not. */
+	if (keyno == 0 || keyno > xfs_btree_get_numrecs(block))
+		*stat = 0;
+	else if (dir != XFS_LOOKUP_EQ || diff == 0)
+		*stat = 1;
+	else
+		*stat = 0;
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Update keys at all levels from here to the root along the cursor's path.
+ */
+STATIC int
+xfs_btree_updkey(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*keyp,
+	int			level)
+{
+	struct xfs_btree_block	*block;
+	struct xfs_buf		*bp;
+	union xfs_btree_key	*kp;
+	int			ptr;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGIK(cur, level, keyp);
+
+	ASSERT(!(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) || level >= 1);
+
+	/*
+	 * Go up the tree from this level toward the root.
+	 * At each level, update the key value to the value input.
+	 * Stop when we reach a level where the cursor isn't pointing
+	 * at the first entry in the block.
+	 */
+	for (ptr = 1; ptr == 1 && level < cur->bc_nlevels; level++) {
+#ifdef DEBUG
+		int		error;
+#endif
+		block = xfs_btree_get_block(cur, level, &bp);
+#ifdef DEBUG
+		error = xfs_btree_check_block(cur, block, level, bp);
+		if (error) {
+			XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+			return error;
+		}
+#endif
+		ptr = cur->bc_ptrs[level];
+		kp = xfs_btree_key_addr(cur, ptr, block);
+		xfs_btree_copy_keys(cur, kp, keyp, 1);
+		xfs_btree_log_keys(cur, bp, ptr, ptr);
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	return 0;
+}
+
+/*
+ * Update the record referred to by cur to the value in the
+ * given record. This either works (return 0) or gets an
+ * EFSCORRUPTED error.
+ */
+int
+xfs_btree_update(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec)
+{
+	struct xfs_btree_block	*block;
+	struct xfs_buf		*bp;
+	int			error;
+	int			ptr;
+	union xfs_btree_rec	*rp;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGR(cur, rec);
+
+	/* Pick up the current block. */
+	block = xfs_btree_get_block(cur, 0, &bp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, 0, bp);
+	if (error)
+		goto error0;
+#endif
+	/* Get the address of the rec to be updated. */
+	ptr = cur->bc_ptrs[0];
+	rp = xfs_btree_rec_addr(cur, ptr, block);
+
+	/* Fill in the new contents and log them. */
+	xfs_btree_copy_recs(cur, rp, rec, 1);
+	xfs_btree_log_recs(cur, bp, ptr, ptr);
+
+	/*
+	 * If we are tracking the last record in the tree and
+	 * we are at the far right edge of the tree, update it.
+	 */
+	if (xfs_btree_is_lastrec(cur, block, 0)) {
+		cur->bc_ops->update_lastrec(cur, block, rec,
+					    ptr, LASTREC_UPDATE);
+	}
+
+	/* Updating first rec in leaf. Pass new key value up to our parent. */
+	if (ptr == 1) {
+		union xfs_btree_key	key;
+
+		cur->bc_ops->init_key_from_rec(&key, rec);
+		error = xfs_btree_updkey(cur, &key, 1);
+		if (error)
+			goto error0;
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Move 1 record left from cur/level if possible.
+ * Update cur to reflect the new path.
+ */
+STATIC int					/* error */
+xfs_btree_lshift(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			*stat)		/* success/failure */
+{
+	union xfs_btree_key	key;		/* btree key */
+	struct xfs_buf		*lbp;		/* left buffer pointer */
+	struct xfs_btree_block	*left;		/* left btree block */
+	int			lrecs;		/* left record count */
+	struct xfs_buf		*rbp;		/* right buffer pointer */
+	struct xfs_btree_block	*right;		/* right btree block */
+	int			rrecs;		/* right record count */
+	union xfs_btree_ptr	lptr;		/* left btree pointer */
+	union xfs_btree_key	*rkp = NULL;	/* right btree key */
+	union xfs_btree_ptr	*rpp = NULL;	/* right address pointer */
+	union xfs_btree_rec	*rrp = NULL;	/* right record pointer */
+	int			error;		/* error return value */
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGI(cur, level);
+
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+	    level == cur->bc_nlevels - 1)
+		goto out0;
+
+	/* Set up variables for this block as "right". */
+	right = xfs_btree_get_block(cur, level, &rbp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, right, level, rbp);
+	if (error)
+		goto error0;
+#endif
+
+	/* If we've got no left sibling then we can't shift an entry left. */
+	xfs_btree_get_sibling(cur, right, &lptr, XFS_BB_LEFTSIB);
+	if (xfs_btree_ptr_is_null(cur, &lptr))
+		goto out0;
+
+	/*
+	 * If the cursor entry is the one that would be moved, don't
+	 * do it... it's too complicated.
+	 */
+	if (cur->bc_ptrs[level] <= 1)
+		goto out0;
+
+	/* Set up the left neighbor as "left". */
+	error = xfs_btree_read_buf_block(cur, &lptr, level, 0, &left, &lbp);
+	if (error)
+		goto error0;
+
+	/* If it's full, it can't take another entry. */
+	lrecs = xfs_btree_get_numrecs(left);
+	if (lrecs == cur->bc_ops->get_maxrecs(cur, level))
+		goto out0;
+
+	rrecs = xfs_btree_get_numrecs(right);
+
+	/*
+	 * We add one entry to the left side and remove one for the right side.
+	 * Accout for it here, the changes will be updated on disk and logged
+	 * later.
+	 */
+	lrecs++;
+	rrecs--;
+
+	XFS_BTREE_STATS_INC(cur, lshift);
+	XFS_BTREE_STATS_ADD(cur, moves, 1);
+
+	/*
+	 * If non-leaf, copy a key and a ptr to the left block.
+	 * Log the changes to the left block.
+	 */
+	if (level > 0) {
+		/* It's a non-leaf.  Move keys and pointers. */
+		union xfs_btree_key	*lkp;	/* left btree key */
+		union xfs_btree_ptr	*lpp;	/* left address pointer */
+
+		lkp = xfs_btree_key_addr(cur, lrecs, left);
+		rkp = xfs_btree_key_addr(cur, 1, right);
+
+		lpp = xfs_btree_ptr_addr(cur, lrecs, left);
+		rpp = xfs_btree_ptr_addr(cur, 1, right);
+#ifdef DEBUG
+		error = xfs_btree_check_ptr(cur, rpp, 0, level);
+		if (error)
+			goto error0;
+#endif
+		xfs_btree_copy_keys(cur, lkp, rkp, 1);
+		xfs_btree_copy_ptrs(cur, lpp, rpp, 1);
+
+		xfs_btree_log_keys(cur, lbp, lrecs, lrecs);
+		xfs_btree_log_ptrs(cur, lbp, lrecs, lrecs);
+
+		ASSERT(cur->bc_ops->keys_inorder(cur,
+			xfs_btree_key_addr(cur, lrecs - 1, left), lkp));
+	} else {
+		/* It's a leaf.  Move records.  */
+		union xfs_btree_rec	*lrp;	/* left record pointer */
+
+		lrp = xfs_btree_rec_addr(cur, lrecs, left);
+		rrp = xfs_btree_rec_addr(cur, 1, right);
+
+		xfs_btree_copy_recs(cur, lrp, rrp, 1);
+		xfs_btree_log_recs(cur, lbp, lrecs, lrecs);
+
+		ASSERT(cur->bc_ops->recs_inorder(cur,
+			xfs_btree_rec_addr(cur, lrecs - 1, left), lrp));
+	}
+
+	xfs_btree_set_numrecs(left, lrecs);
+	xfs_btree_log_block(cur, lbp, XFS_BB_NUMRECS);
+
+	xfs_btree_set_numrecs(right, rrecs);
+	xfs_btree_log_block(cur, rbp, XFS_BB_NUMRECS);
+
+	/*
+	 * Slide the contents of right down one entry.
+	 */
+	XFS_BTREE_STATS_ADD(cur, moves, rrecs - 1);
+	if (level > 0) {
+		/* It's a nonleaf. operate on keys and ptrs */
+#ifdef DEBUG
+		int			i;		/* loop index */
+
+		for (i = 0; i < rrecs; i++) {
+			error = xfs_btree_check_ptr(cur, rpp, i + 1, level);
+			if (error)
+				goto error0;
+		}
+#endif
+		xfs_btree_shift_keys(cur,
+				xfs_btree_key_addr(cur, 2, right),
+				-1, rrecs);
+		xfs_btree_shift_ptrs(cur,
+				xfs_btree_ptr_addr(cur, 2, right),
+				-1, rrecs);
+
+		xfs_btree_log_keys(cur, rbp, 1, rrecs);
+		xfs_btree_log_ptrs(cur, rbp, 1, rrecs);
+	} else {
+		/* It's a leaf. operate on records */
+		xfs_btree_shift_recs(cur,
+			xfs_btree_rec_addr(cur, 2, right),
+			-1, rrecs);
+		xfs_btree_log_recs(cur, rbp, 1, rrecs);
+
+		/*
+		 * If it's the first record in the block, we'll need a key
+		 * structure to pass up to the next level (updkey).
+		 */
+		cur->bc_ops->init_key_from_rec(&key,
+			xfs_btree_rec_addr(cur, 1, right));
+		rkp = &key;
+	}
+
+	/* Update the parent key values of right. */
+	error = xfs_btree_updkey(cur, rkp, level + 1);
+	if (error)
+		goto error0;
+
+	/* Slide the cursor value left one. */
+	cur->bc_ptrs[level]--;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+
+out0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 0;
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Move 1 record right from cur/level if possible.
+ * Update cur to reflect the new path.
+ */
+STATIC int					/* error */
+xfs_btree_rshift(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			*stat)		/* success/failure */
+{
+	union xfs_btree_key	key;		/* btree key */
+	struct xfs_buf		*lbp;		/* left buffer pointer */
+	struct xfs_btree_block	*left;		/* left btree block */
+	struct xfs_buf		*rbp;		/* right buffer pointer */
+	struct xfs_btree_block	*right;		/* right btree block */
+	struct xfs_btree_cur	*tcur;		/* temporary btree cursor */
+	union xfs_btree_ptr	rptr;		/* right block pointer */
+	union xfs_btree_key	*rkp;		/* right btree key */
+	int			rrecs;		/* right record count */
+	int			lrecs;		/* left record count */
+	int			error;		/* error return value */
+	int			i;		/* loop counter */
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGI(cur, level);
+
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+	    (level == cur->bc_nlevels - 1))
+		goto out0;
+
+	/* Set up variables for this block as "left". */
+	left = xfs_btree_get_block(cur, level, &lbp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, left, level, lbp);
+	if (error)
+		goto error0;
+#endif
+
+	/* If we've got no right sibling then we can't shift an entry right. */
+	xfs_btree_get_sibling(cur, left, &rptr, XFS_BB_RIGHTSIB);
+	if (xfs_btree_ptr_is_null(cur, &rptr))
+		goto out0;
+
+	/*
+	 * If the cursor entry is the one that would be moved, don't
+	 * do it... it's too complicated.
+	 */
+	lrecs = xfs_btree_get_numrecs(left);
+	if (cur->bc_ptrs[level] >= lrecs)
+		goto out0;
+
+	/* Set up the right neighbor as "right". */
+	error = xfs_btree_read_buf_block(cur, &rptr, level, 0, &right, &rbp);
+	if (error)
+		goto error0;
+
+	/* If it's full, it can't take another entry. */
+	rrecs = xfs_btree_get_numrecs(right);
+	if (rrecs == cur->bc_ops->get_maxrecs(cur, level))
+		goto out0;
+
+	XFS_BTREE_STATS_INC(cur, rshift);
+	XFS_BTREE_STATS_ADD(cur, moves, rrecs);
+
+	/*
+	 * Make a hole at the start of the right neighbor block, then
+	 * copy the last left block entry to the hole.
+	 */
+	if (level > 0) {
+		/* It's a nonleaf. make a hole in the keys and ptrs */
+		union xfs_btree_key	*lkp;
+		union xfs_btree_ptr	*lpp;
+		union xfs_btree_ptr	*rpp;
+
+		lkp = xfs_btree_key_addr(cur, lrecs, left);
+		lpp = xfs_btree_ptr_addr(cur, lrecs, left);
+		rkp = xfs_btree_key_addr(cur, 1, right);
+		rpp = xfs_btree_ptr_addr(cur, 1, right);
+
+#ifdef DEBUG
+		for (i = rrecs - 1; i >= 0; i--) {
+			error = xfs_btree_check_ptr(cur, rpp, i, level);
+			if (error)
+				goto error0;
+		}
+#endif
+
+		xfs_btree_shift_keys(cur, rkp, 1, rrecs);
+		xfs_btree_shift_ptrs(cur, rpp, 1, rrecs);
+
+#ifdef DEBUG
+		error = xfs_btree_check_ptr(cur, lpp, 0, level);
+		if (error)
+			goto error0;
+#endif
+
+		/* Now put the new data in, and log it. */
+		xfs_btree_copy_keys(cur, rkp, lkp, 1);
+		xfs_btree_copy_ptrs(cur, rpp, lpp, 1);
+
+		xfs_btree_log_keys(cur, rbp, 1, rrecs + 1);
+		xfs_btree_log_ptrs(cur, rbp, 1, rrecs + 1);
+
+		ASSERT(cur->bc_ops->keys_inorder(cur, rkp,
+			xfs_btree_key_addr(cur, 2, right)));
+	} else {
+		/* It's a leaf. make a hole in the records */
+		union xfs_btree_rec	*lrp;
+		union xfs_btree_rec	*rrp;
+
+		lrp = xfs_btree_rec_addr(cur, lrecs, left);
+		rrp = xfs_btree_rec_addr(cur, 1, right);
+
+		xfs_btree_shift_recs(cur, rrp, 1, rrecs);
+
+		/* Now put the new data in, and log it. */
+		xfs_btree_copy_recs(cur, rrp, lrp, 1);
+		xfs_btree_log_recs(cur, rbp, 1, rrecs + 1);
+
+		cur->bc_ops->init_key_from_rec(&key, rrp);
+		rkp = &key;
+
+		ASSERT(cur->bc_ops->recs_inorder(cur, rrp,
+			xfs_btree_rec_addr(cur, 2, right)));
+	}
+
+	/*
+	 * Decrement and log left's numrecs, bump and log right's numrecs.
+	 */
+	xfs_btree_set_numrecs(left, --lrecs);
+	xfs_btree_log_block(cur, lbp, XFS_BB_NUMRECS);
+
+	xfs_btree_set_numrecs(right, ++rrecs);
+	xfs_btree_log_block(cur, rbp, XFS_BB_NUMRECS);
+
+	/*
+	 * Using a temporary cursor, update the parent key values of the
+	 * block on the right.
+	 */
+	error = xfs_btree_dup_cursor(cur, &tcur);
+	if (error)
+		goto error0;
+	i = xfs_btree_lastrec(tcur, level);
+	XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+
+	error = xfs_btree_increment(tcur, level, &i);
+	if (error)
+		goto error1;
+
+	error = xfs_btree_updkey(tcur, rkp, level + 1);
+	if (error)
+		goto error1;
+
+	xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+
+out0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 0;
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+
+error1:
+	XFS_BTREE_TRACE_CURSOR(tcur, XBT_ERROR);
+	xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
+	return error;
+}
+
+/*
+ * Split cur/level block in half.
+ * Return new block number and the key to its first
+ * record (to be inserted into parent).
+ */
+STATIC int					/* error */
+xfs_btree_split(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	union xfs_btree_ptr	*ptrp,
+	union xfs_btree_key	*key,
+	struct xfs_btree_cur	**curp,
+	int			*stat)		/* success/failure */
+{
+	union xfs_btree_ptr	lptr;		/* left sibling block ptr */
+	struct xfs_buf		*lbp;		/* left buffer pointer */
+	struct xfs_btree_block	*left;		/* left btree block */
+	union xfs_btree_ptr	rptr;		/* right sibling block ptr */
+	struct xfs_buf		*rbp;		/* right buffer pointer */
+	struct xfs_btree_block	*right;		/* right btree block */
+	union xfs_btree_ptr	rrptr;		/* right-right sibling ptr */
+	struct xfs_buf		*rrbp;		/* right-right buffer pointer */
+	struct xfs_btree_block	*rrblock;	/* right-right btree block */
+	int			lrecs;
+	int			rrecs;
+	int			src_index;
+	int			error;		/* error return value */
+#ifdef DEBUG
+	int			i;
+#endif
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGIPK(cur, level, *ptrp, key);
+
+	XFS_BTREE_STATS_INC(cur, split);
+
+	/* Set up left block (current one). */
+	left = xfs_btree_get_block(cur, level, &lbp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, left, level, lbp);
+	if (error)
+		goto error0;
+#endif
+
+	xfs_btree_buf_to_ptr(cur, lbp, &lptr);
+
+	/* Allocate the new block. If we can't do it, we're toast. Give up. */
+	error = cur->bc_ops->alloc_block(cur, &lptr, &rptr, 1, stat);
+	if (error)
+		goto error0;
+	if (*stat == 0)
+		goto out0;
+	XFS_BTREE_STATS_INC(cur, alloc);
+
+	/* Set up the new block as "right". */
+	error = xfs_btree_get_buf_block(cur, &rptr, 0, &right, &rbp);
+	if (error)
+		goto error0;
+
+	/* Fill in the btree header for the new right block. */
+	xfs_btree_init_block(cur, xfs_btree_get_level(left), 0, right);
+
+	/*
+	 * Split the entries between the old and the new block evenly.
+	 * Make sure that if there's an odd number of entries now, that
+	 * each new block will have the same number of entries.
+	 */
+	lrecs = xfs_btree_get_numrecs(left);
+	rrecs = lrecs / 2;
+	if ((lrecs & 1) && cur->bc_ptrs[level] <= rrecs + 1)
+		rrecs++;
+	src_index = (lrecs - rrecs + 1);
+
+	XFS_BTREE_STATS_ADD(cur, moves, rrecs);
+
+	/*
+	 * Copy btree block entries from the left block over to the
+	 * new block, the right. Update the right block and log the
+	 * changes.
+	 */
+	if (level > 0) {
+		/* It's a non-leaf.  Move keys and pointers. */
+		union xfs_btree_key	*lkp;	/* left btree key */
+		union xfs_btree_ptr	*lpp;	/* left address pointer */
+		union xfs_btree_key	*rkp;	/* right btree key */
+		union xfs_btree_ptr	*rpp;	/* right address pointer */
+
+		lkp = xfs_btree_key_addr(cur, src_index, left);
+		lpp = xfs_btree_ptr_addr(cur, src_index, left);
+		rkp = xfs_btree_key_addr(cur, 1, right);
+		rpp = xfs_btree_ptr_addr(cur, 1, right);
+
+#ifdef DEBUG
+		for (i = src_index; i < rrecs; i++) {
+			error = xfs_btree_check_ptr(cur, lpp, i, level);
+			if (error)
+				goto error0;
+		}
+#endif
+
+		xfs_btree_copy_keys(cur, rkp, lkp, rrecs);
+		xfs_btree_copy_ptrs(cur, rpp, lpp, rrecs);
+
+		xfs_btree_log_keys(cur, rbp, 1, rrecs);
+		xfs_btree_log_ptrs(cur, rbp, 1, rrecs);
+
+		/* Grab the keys to the entries moved to the right block */
+		xfs_btree_copy_keys(cur, key, rkp, 1);
+	} else {
+		/* It's a leaf.  Move records.  */
+		union xfs_btree_rec	*lrp;	/* left record pointer */
+		union xfs_btree_rec	*rrp;	/* right record pointer */
+
+		lrp = xfs_btree_rec_addr(cur, src_index, left);
+		rrp = xfs_btree_rec_addr(cur, 1, right);
+
+		xfs_btree_copy_recs(cur, rrp, lrp, rrecs);
+		xfs_btree_log_recs(cur, rbp, 1, rrecs);
+
+		cur->bc_ops->init_key_from_rec(key,
+			xfs_btree_rec_addr(cur, 1, right));
+	}
+
+
+	/*
+	 * Find the left block number by looking in the buffer.
+	 * Adjust numrecs, sibling pointers.
+	 */
+	xfs_btree_get_sibling(cur, left, &rrptr, XFS_BB_RIGHTSIB);
+	xfs_btree_set_sibling(cur, right, &rrptr, XFS_BB_RIGHTSIB);
+	xfs_btree_set_sibling(cur, right, &lptr, XFS_BB_LEFTSIB);
+	xfs_btree_set_sibling(cur, left, &rptr, XFS_BB_RIGHTSIB);
+
+	lrecs -= rrecs;
+	xfs_btree_set_numrecs(left, lrecs);
+	xfs_btree_set_numrecs(right, xfs_btree_get_numrecs(right) + rrecs);
+
+	xfs_btree_log_block(cur, rbp, XFS_BB_ALL_BITS);
+	xfs_btree_log_block(cur, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
+
+	/*
+	 * If there's a block to the new block's right, make that block
+	 * point back to right instead of to left.
+	 */
+	if (!xfs_btree_ptr_is_null(cur, &rrptr)) {
+		error = xfs_btree_read_buf_block(cur, &rrptr, level,
+							0, &rrblock, &rrbp);
+		if (error)
+			goto error0;
+		xfs_btree_set_sibling(cur, rrblock, &rptr, XFS_BB_LEFTSIB);
+		xfs_btree_log_block(cur, rrbp, XFS_BB_LEFTSIB);
+	}
+	/*
+	 * If the cursor is really in the right block, move it there.
+	 * If it's just pointing past the last entry in left, then we'll
+	 * insert there, so don't change anything in that case.
+	 */
+	if (cur->bc_ptrs[level] > lrecs + 1) {
+		xfs_btree_setbuf(cur, level, rbp);
+		cur->bc_ptrs[level] -= lrecs;
+	}
+	/*
+	 * If there are more levels, we'll need another cursor which refers
+	 * the right block, no matter where this cursor was.
+	 */
+	if (level + 1 < cur->bc_nlevels) {
+		error = xfs_btree_dup_cursor(cur, curp);
+		if (error)
+			goto error0;
+		(*curp)->bc_ptrs[level + 1]++;
+	}
+	*ptrp = rptr;
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+out0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 0;
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Copy the old inode root contents into a real block and make the
+ * broot point to it.
+ */
+int						/* error */
+xfs_btree_new_iroot(
+	struct xfs_btree_cur	*cur,		/* btree cursor */
+	int			*logflags,	/* logging flags for inode */
+	int			*stat)		/* return status - 0 fail */
+{
+	struct xfs_buf		*cbp;		/* buffer for cblock */
+	struct xfs_btree_block	*block;		/* btree block */
+	struct xfs_btree_block	*cblock;	/* child btree block */
+	union xfs_btree_key	*ckp;		/* child key pointer */
+	union xfs_btree_ptr	*cpp;		/* child ptr pointer */
+	union xfs_btree_key	*kp;		/* pointer to btree key */
+	union xfs_btree_ptr	*pp;		/* pointer to block addr */
+	union xfs_btree_ptr	nptr;		/* new block addr */
+	int			level;		/* btree level */
+	int			error;		/* error return code */
+#ifdef DEBUG
+	int			i;		/* loop counter */
+#endif
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_STATS_INC(cur, newroot);
+
+	ASSERT(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE);
+
+	level = cur->bc_nlevels - 1;
+
+	block = xfs_btree_get_iroot(cur);
+	pp = xfs_btree_ptr_addr(cur, 1, block);
+
+	/* Allocate the new block. If we can't do it, we're toast. Give up. */
+	error = cur->bc_ops->alloc_block(cur, pp, &nptr, 1, stat);
+	if (error)
+		goto error0;
+	if (*stat == 0) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+		return 0;
+	}
+	XFS_BTREE_STATS_INC(cur, alloc);
+
+	/* Copy the root into a real block. */
+	error = xfs_btree_get_buf_block(cur, &nptr, 0, &cblock, &cbp);
+	if (error)
+		goto error0;
+
+	memcpy(cblock, block, xfs_btree_block_len(cur));
+
+	be16_add_cpu(&block->bb_level, 1);
+	xfs_btree_set_numrecs(block, 1);
+	cur->bc_nlevels++;
+	cur->bc_ptrs[level + 1] = 1;
+
+	kp = xfs_btree_key_addr(cur, 1, block);
+	ckp = xfs_btree_key_addr(cur, 1, cblock);
+	xfs_btree_copy_keys(cur, ckp, kp, xfs_btree_get_numrecs(cblock));
+
+	cpp = xfs_btree_ptr_addr(cur, 1, cblock);
+#ifdef DEBUG
+	for (i = 0; i < be16_to_cpu(cblock->bb_numrecs); i++) {
+		error = xfs_btree_check_ptr(cur, pp, i, level);
+		if (error)
+			goto error0;
+	}
+#endif
+	xfs_btree_copy_ptrs(cur, cpp, pp, xfs_btree_get_numrecs(cblock));
+
+#ifdef DEBUG
+	error = xfs_btree_check_ptr(cur, &nptr, 0, level);
+	if (error)
+		goto error0;
+#endif
+	xfs_btree_copy_ptrs(cur, pp, &nptr, 1);
+
+	xfs_iroot_realloc(cur->bc_private.b.ip,
+			  1 - xfs_btree_get_numrecs(cblock),
+			  cur->bc_private.b.whichfork);
+
+	xfs_btree_setbuf(cur, level, cbp);
+
+	/*
+	 * Do all this logging at the end so that
+	 * the root is at the right level.
+	 */
+	xfs_btree_log_block(cur, cbp, XFS_BB_ALL_BITS);
+	xfs_btree_log_keys(cur, cbp, 1, be16_to_cpu(cblock->bb_numrecs));
+	xfs_btree_log_ptrs(cur, cbp, 1, be16_to_cpu(cblock->bb_numrecs));
+
+	*logflags |=
+		XFS_ILOG_CORE | XFS_ILOG_FBROOT(cur->bc_private.b.whichfork);
+	*stat = 1;
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	return 0;
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Allocate a new root block, fill it in.
+ */
+STATIC int				/* error */
+xfs_btree_new_root(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	int			*stat)	/* success/failure */
+{
+	struct xfs_btree_block	*block;	/* one half of the old root block */
+	struct xfs_buf		*bp;	/* buffer containing block */
+	int			error;	/* error return value */
+	struct xfs_buf		*lbp;	/* left buffer pointer */
+	struct xfs_btree_block	*left;	/* left btree block */
+	struct xfs_buf		*nbp;	/* new (root) buffer */
+	struct xfs_btree_block	*new;	/* new (root) btree block */
+	int			nptr;	/* new value for key index, 1 or 2 */
+	struct xfs_buf		*rbp;	/* right buffer pointer */
+	struct xfs_btree_block	*right;	/* right btree block */
+	union xfs_btree_ptr	rptr;
+	union xfs_btree_ptr	lptr;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_STATS_INC(cur, newroot);
+
+	/* initialise our start point from the cursor */
+	cur->bc_ops->init_ptr_from_cur(cur, &rptr);
+
+	/* Allocate the new block. If we can't do it, we're toast. Give up. */
+	error = cur->bc_ops->alloc_block(cur, &rptr, &lptr, 1, stat);
+	if (error)
+		goto error0;
+	if (*stat == 0)
+		goto out0;
+	XFS_BTREE_STATS_INC(cur, alloc);
+
+	/* Set up the new block. */
+	error = xfs_btree_get_buf_block(cur, &lptr, 0, &new, &nbp);
+	if (error)
+		goto error0;
+
+	/* Set the root in the holding structure  increasing the level by 1. */
+	cur->bc_ops->set_root(cur, &lptr, 1);
+
+	/*
+	 * At the previous root level there are now two blocks: the old root,
+	 * and the new block generated when it was split.  We don't know which
+	 * one the cursor is pointing at, so we set up variables "left" and
+	 * "right" for each case.
+	 */
+	block = xfs_btree_get_block(cur, cur->bc_nlevels - 1, &bp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, cur->bc_nlevels - 1, bp);
+	if (error)
+		goto error0;
+#endif
+
+	xfs_btree_get_sibling(cur, block, &rptr, XFS_BB_RIGHTSIB);
+	if (!xfs_btree_ptr_is_null(cur, &rptr)) {
+		/* Our block is left, pick up the right block. */
+		lbp = bp;
+		xfs_btree_buf_to_ptr(cur, lbp, &lptr);
+		left = block;
+		error = xfs_btree_read_buf_block(cur, &rptr,
+					cur->bc_nlevels - 1, 0, &right, &rbp);
+		if (error)
+			goto error0;
+		bp = rbp;
+		nptr = 1;
+	} else {
+		/* Our block is right, pick up the left block. */
+		rbp = bp;
+		xfs_btree_buf_to_ptr(cur, rbp, &rptr);
+		right = block;
+		xfs_btree_get_sibling(cur, right, &lptr, XFS_BB_LEFTSIB);
+		error = xfs_btree_read_buf_block(cur, &lptr,
+					cur->bc_nlevels - 1, 0, &left, &lbp);
+		if (error)
+			goto error0;
+		bp = lbp;
+		nptr = 2;
+	}
+	/* Fill in the new block's btree header and log it. */
+	xfs_btree_init_block(cur, cur->bc_nlevels, 2, new);
+	xfs_btree_log_block(cur, nbp, XFS_BB_ALL_BITS);
+	ASSERT(!xfs_btree_ptr_is_null(cur, &lptr) &&
+			!xfs_btree_ptr_is_null(cur, &rptr));
+
+	/* Fill in the key data in the new root. */
+	if (xfs_btree_get_level(left) > 0) {
+		xfs_btree_copy_keys(cur,
+				xfs_btree_key_addr(cur, 1, new),
+				xfs_btree_key_addr(cur, 1, left), 1);
+		xfs_btree_copy_keys(cur,
+				xfs_btree_key_addr(cur, 2, new),
+				xfs_btree_key_addr(cur, 1, right), 1);
+	} else {
+		cur->bc_ops->init_key_from_rec(
+				xfs_btree_key_addr(cur, 1, new),
+				xfs_btree_rec_addr(cur, 1, left));
+		cur->bc_ops->init_key_from_rec(
+				xfs_btree_key_addr(cur, 2, new),
+				xfs_btree_rec_addr(cur, 1, right));
+	}
+	xfs_btree_log_keys(cur, nbp, 1, 2);
+
+	/* Fill in the pointer data in the new root. */
+	xfs_btree_copy_ptrs(cur,
+		xfs_btree_ptr_addr(cur, 1, new), &lptr, 1);
+	xfs_btree_copy_ptrs(cur,
+		xfs_btree_ptr_addr(cur, 2, new), &rptr, 1);
+	xfs_btree_log_ptrs(cur, nbp, 1, 2);
+
+	/* Fix up the cursor. */
+	xfs_btree_setbuf(cur, cur->bc_nlevels, nbp);
+	cur->bc_ptrs[cur->bc_nlevels] = nptr;
+	cur->bc_nlevels++;
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+out0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 0;
+	return 0;
+}
+
+STATIC int
+xfs_btree_make_block_unfull(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	int			level,	/* btree level */
+	int			numrecs,/* # of recs in block */
+	int			*oindex,/* old tree index */
+	int			*index,	/* new tree index */
+	union xfs_btree_ptr	*nptr,	/* new btree ptr */
+	struct xfs_btree_cur	**ncur,	/* new btree cursor */
+	union xfs_btree_rec	*nrec,	/* new record */
+	int			*stat)
+{
+	union xfs_btree_key	key;	/* new btree key value */
+	int			error = 0;
+
+	if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+	    level == cur->bc_nlevels - 1) {
+	    	struct xfs_inode *ip = cur->bc_private.b.ip;
+
+		if (numrecs < cur->bc_ops->get_dmaxrecs(cur, level)) {
+			/* A root block that can be made bigger. */
+
+			xfs_iroot_realloc(ip, 1, cur->bc_private.b.whichfork);
+		} else {
+			/* A root block that needs replacing */
+			int	logflags = 0;
+
+			error = xfs_btree_new_iroot(cur, &logflags, stat);
+			if (error || *stat == 0)
+				return error;
+
+			xfs_trans_log_inode(cur->bc_tp, ip, logflags);
+		}
+
+		return 0;
+	}
+
+	/* First, try shifting an entry to the right neighbor. */
+	error = xfs_btree_rshift(cur, level, stat);
+	if (error || *stat)
+		return error;
+
+	/* Next, try shifting an entry to the left neighbor. */
+	error = xfs_btree_lshift(cur, level, stat);
+	if (error)
+		return error;
+
+	if (*stat) {
+		*oindex = *index = cur->bc_ptrs[level];
+		return 0;
+	}
+
+	/*
+	 * Next, try splitting the current block in half.
+	 *
+	 * If this works we have to re-set our variables because we
+	 * could be in a different block now.
+	 */
+	error = xfs_btree_split(cur, level, nptr, &key, ncur, stat);
+	if (error || *stat == 0)
+		return error;
+
+
+	*index = cur->bc_ptrs[level];
+	cur->bc_ops->init_rec_from_key(&key, nrec);
+	return 0;
+}
+
+/*
+ * Insert one record/level.  Return information to the caller
+ * allowing the next level up to proceed if necessary.
+ */
+STATIC int
+xfs_btree_insrec(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	int			level,	/* level to insert record at */
+	union xfs_btree_ptr	*ptrp,	/* i/o: block number inserted */
+	union xfs_btree_rec	*recp,	/* i/o: record data inserted */
+	struct xfs_btree_cur	**curp,	/* output: new cursor replacing cur */
+	int			*stat)	/* success/failure */
+{
+	struct xfs_btree_block	*block;	/* btree block */
+	struct xfs_buf		*bp;	/* buffer for block */
+	union xfs_btree_key	key;	/* btree key */
+	union xfs_btree_ptr	nptr;	/* new block ptr */
+	struct xfs_btree_cur	*ncur;	/* new btree cursor */
+	union xfs_btree_rec	nrec;	/* new record count */
+	int			optr;	/* old key/record index */
+	int			ptr;	/* key/record index */
+	int			numrecs;/* number of records */
+	int			error;	/* error return value */
+#ifdef DEBUG
+	int			i;
+#endif
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGIPR(cur, level, *ptrp, recp);
+
+	ncur = NULL;
+
+	/*
+	 * If we have an external root pointer, and we've made it to the
+	 * root level, allocate a new root block and we're done.
+	 */
+	if (!(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
+	    (level >= cur->bc_nlevels)) {
+		error = xfs_btree_new_root(cur, stat);
+		xfs_btree_set_ptr_null(cur, ptrp);
+
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+		return error;
+	}
+
+	/* If we're off the left edge, return failure. */
+	ptr = cur->bc_ptrs[level];
+	if (ptr == 0) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+		*stat = 0;
+		return 0;
+	}
+
+	/* Make a key out of the record data to be inserted, and save it. */
+	cur->bc_ops->init_key_from_rec(&key, recp);
+
+	optr = ptr;
+
+	XFS_BTREE_STATS_INC(cur, insrec);
+
+	/* Get pointers to the btree buffer and block. */
+	block = xfs_btree_get_block(cur, level, &bp);
+	numrecs = xfs_btree_get_numrecs(block);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, level, bp);
+	if (error)
+		goto error0;
+
+	/* Check that the new entry is being inserted in the right place. */
+	if (ptr <= numrecs) {
+		if (level == 0) {
+			ASSERT(cur->bc_ops->recs_inorder(cur, recp,
+				xfs_btree_rec_addr(cur, ptr, block)));
+		} else {
+			ASSERT(cur->bc_ops->keys_inorder(cur, &key,
+				xfs_btree_key_addr(cur, ptr, block)));
+		}
+	}
+#endif
+
+	/*
+	 * If the block is full, we can't insert the new entry until we
+	 * make the block un-full.
+	 */
+	xfs_btree_set_ptr_null(cur, &nptr);
+	if (numrecs == cur->bc_ops->get_maxrecs(cur, level)) {
+		error = xfs_btree_make_block_unfull(cur, level, numrecs,
+					&optr, &ptr, &nptr, &ncur, &nrec, stat);
+		if (error || *stat == 0)
+			goto error0;
+	}
+
+	/*
+	 * The current block may have changed if the block was
+	 * previously full and we have just made space in it.
+	 */
+	block = xfs_btree_get_block(cur, level, &bp);
+	numrecs = xfs_btree_get_numrecs(block);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, level, bp);
+	if (error)
+		return error;
+#endif
+
+	/*
+	 * At this point we know there's room for our new entry in the block
+	 * we're pointing at.
+	 */
+	XFS_BTREE_STATS_ADD(cur, moves, numrecs - ptr + 1);
+
+	if (level > 0) {
+		/* It's a nonleaf. make a hole in the keys and ptrs */
+		union xfs_btree_key	*kp;
+		union xfs_btree_ptr	*pp;
+
+		kp = xfs_btree_key_addr(cur, ptr, block);
+		pp = xfs_btree_ptr_addr(cur, ptr, block);
+
+#ifdef DEBUG
+		for (i = numrecs - ptr; i >= 0; i--) {
+			error = xfs_btree_check_ptr(cur, pp, i, level);
+			if (error)
+				return error;
+		}
+#endif
+
+		xfs_btree_shift_keys(cur, kp, 1, numrecs - ptr + 1);
+		xfs_btree_shift_ptrs(cur, pp, 1, numrecs - ptr + 1);
+
+#ifdef DEBUG
+		error = xfs_btree_check_ptr(cur, ptrp, 0, level);
+		if (error)
+			goto error0;
+#endif
+
+		/* Now put the new data in, bump numrecs and log it. */
+		xfs_btree_copy_keys(cur, kp, &key, 1);
+		xfs_btree_copy_ptrs(cur, pp, ptrp, 1);
+		numrecs++;
+		xfs_btree_set_numrecs(block, numrecs);
+		xfs_btree_log_ptrs(cur, bp, ptr, numrecs);
+		xfs_btree_log_keys(cur, bp, ptr, numrecs);
+#ifdef DEBUG
+		if (ptr < numrecs) {
+			ASSERT(cur->bc_ops->keys_inorder(cur, kp,
+				xfs_btree_key_addr(cur, ptr + 1, block)));
+		}
+#endif
+	} else {
+		/* It's a leaf. make a hole in the records */
+		union xfs_btree_rec             *rp;
+
+		rp = xfs_btree_rec_addr(cur, ptr, block);
+
+		xfs_btree_shift_recs(cur, rp, 1, numrecs - ptr + 1);
+
+		/* Now put the new data in, bump numrecs and log it. */
+		xfs_btree_copy_recs(cur, rp, recp, 1);
+		xfs_btree_set_numrecs(block, ++numrecs);
+		xfs_btree_log_recs(cur, bp, ptr, numrecs);
+#ifdef DEBUG
+		if (ptr < numrecs) {
+			ASSERT(cur->bc_ops->recs_inorder(cur, rp,
+				xfs_btree_rec_addr(cur, ptr + 1, block)));
+		}
+#endif
+	}
+
+	/* Log the new number of records in the btree header. */
+	xfs_btree_log_block(cur, bp, XFS_BB_NUMRECS);
+
+	/* If we inserted at the start of a block, update the parents' keys. */
+	if (optr == 1) {
+		error = xfs_btree_updkey(cur, &key, level + 1);
+		if (error)
+			goto error0;
+	}
+
+	/*
+	 * If we are tracking the last record in the tree and
+	 * we are at the far right edge of the tree, update it.
+	 */
+	if (xfs_btree_is_lastrec(cur, block, level)) {
+		cur->bc_ops->update_lastrec(cur, block, recp,
+					    ptr, LASTREC_INSREC);
+	}
+
+	/*
+	 * Return the new block number, if any.
+	 * If there is one, give back a record value and a cursor too.
+	 */
+	*ptrp = nptr;
+	if (!xfs_btree_ptr_is_null(cur, &nptr)) {
+		*recp = nrec;
+		*curp = ncur;
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Insert the record at the point referenced by cur.
+ *
+ * A multi-level split of the tree on insert will invalidate the original
+ * cursor.  All callers of this function should assume that the cursor is
+ * no longer valid and revalidate it.
+ */
+int
+xfs_btree_insert(
+	struct xfs_btree_cur	*cur,
+	int			*stat)
+{
+	int			error;	/* error return value */
+	int			i;	/* result value, 0 for failure */
+	int			level;	/* current level number in btree */
+	union xfs_btree_ptr	nptr;	/* new block number (split result) */
+	struct xfs_btree_cur	*ncur;	/* new cursor (split result) */
+	struct xfs_btree_cur	*pcur;	/* previous level's cursor */
+	union xfs_btree_rec	rec;	/* record to insert */
+
+	level = 0;
+	ncur = NULL;
+	pcur = cur;
+
+	xfs_btree_set_ptr_null(cur, &nptr);
+	cur->bc_ops->init_rec_from_cur(cur, &rec);
+
+	/*
+	 * Loop going up the tree, starting at the leaf level.
+	 * Stop when we don't get a split block, that must mean that
+	 * the insert is finished with this level.
+	 */
+	do {
+		/*
+		 * Insert nrec/nptr into this level of the tree.
+		 * Note if we fail, nptr will be null.
+		 */
+		error = xfs_btree_insrec(pcur, level, &nptr, &rec, &ncur, &i);
+		if (error) {
+			if (pcur != cur)
+				xfs_btree_del_cursor(pcur, XFS_BTREE_ERROR);
+			goto error0;
+		}
+
+		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+		level++;
+
+		/*
+		 * See if the cursor we just used is trash.
+		 * Can't trash the caller's cursor, but otherwise we should
+		 * if ncur is a new cursor or we're about to be done.
+		 */
+		if (pcur != cur &&
+		    (ncur || xfs_btree_ptr_is_null(cur, &nptr))) {
+			/* Save the state from the cursor before we trash it */
+			if (cur->bc_ops->update_cursor)
+				cur->bc_ops->update_cursor(pcur, cur);
+			cur->bc_nlevels = pcur->bc_nlevels;
+			xfs_btree_del_cursor(pcur, XFS_BTREE_NOERROR);
+		}
+		/* If we got a new cursor, switch to it. */
+		if (ncur) {
+			pcur = ncur;
+			ncur = NULL;
+		}
+	} while (!xfs_btree_ptr_is_null(cur, &nptr));
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = i;
+	return 0;
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Try to merge a non-leaf block back into the inode root.
+ *
+ * Note: the killroot names comes from the fact that we're effectively
+ * killing the old root block.  But because we can't just delete the
+ * inode we have to copy the single block it was pointing to into the
+ * inode.
+ */
+int
+xfs_btree_kill_iroot(
+	struct xfs_btree_cur	*cur)
+{
+	int			whichfork = cur->bc_private.b.whichfork;
+	struct xfs_inode	*ip = cur->bc_private.b.ip;
+	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
+	struct xfs_btree_block	*block;
+	struct xfs_btree_block	*cblock;
+	union xfs_btree_key	*kp;
+	union xfs_btree_key	*ckp;
+	union xfs_btree_ptr	*pp;
+	union xfs_btree_ptr	*cpp;
+	struct xfs_buf		*cbp;
+	int			level;
+	int			index;
+	int			numrecs;
+#ifdef DEBUG
+	union xfs_btree_ptr	ptr;
+	int			i;
+#endif
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+
+	ASSERT(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE);
+	ASSERT(cur->bc_nlevels > 1);
+
+	/*
+	 * Don't deal with the root block needs to be a leaf case.
+	 * We're just going to turn the thing back into extents anyway.
+	 */
+	level = cur->bc_nlevels - 1;
+	if (level == 1)
+		goto out0;
+
+	/*
+	 * Give up if the root has multiple children.
+	 */
+	block = xfs_btree_get_iroot(cur);
+	if (xfs_btree_get_numrecs(block) != 1)
+		goto out0;
+
+	cblock = xfs_btree_get_block(cur, level - 1, &cbp);
+	numrecs = xfs_btree_get_numrecs(cblock);
+
+	/*
+	 * Only do this if the next level will fit.
+	 * Then the data must be copied up to the inode,
+	 * instead of freeing the root you free the next level.
+	 */
+	if (numrecs > cur->bc_ops->get_dmaxrecs(cur, level))
+		goto out0;
+
+	XFS_BTREE_STATS_INC(cur, killroot);
+
+#ifdef DEBUG
+	xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_LEFTSIB);
+	ASSERT(xfs_btree_ptr_is_null(cur, &ptr));
+	xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_RIGHTSIB);
+	ASSERT(xfs_btree_ptr_is_null(cur, &ptr));
+#endif
+
+	index = numrecs - cur->bc_ops->get_maxrecs(cur, level);
+	if (index) {
+		xfs_iroot_realloc(cur->bc_private.b.ip, index,
+				  cur->bc_private.b.whichfork);
+		block = ifp->if_broot;
+	}
+
+	be16_add_cpu(&block->bb_numrecs, index);
+	ASSERT(block->bb_numrecs == cblock->bb_numrecs);
+
+	kp = xfs_btree_key_addr(cur, 1, block);
+	ckp = xfs_btree_key_addr(cur, 1, cblock);
+	xfs_btree_copy_keys(cur, kp, ckp, numrecs);
+
+	pp = xfs_btree_ptr_addr(cur, 1, block);
+	cpp = xfs_btree_ptr_addr(cur, 1, cblock);
+#ifdef DEBUG
+	for (i = 0; i < numrecs; i++) {
+		int		error;
+
+		error = xfs_btree_check_ptr(cur, cpp, i, level - 1);
+		if (error) {
+			XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+			return error;
+		}
+	}
+#endif
+	xfs_btree_copy_ptrs(cur, pp, cpp, numrecs);
+
+	cur->bc_ops->free_block(cur, cbp);
+	XFS_BTREE_STATS_INC(cur, free);
+
+	cur->bc_bufs[level - 1] = NULL;
+	be16_add_cpu(&block->bb_level, -1);
+	xfs_trans_log_inode(cur->bc_tp, ip,
+		XFS_ILOG_CORE | XFS_ILOG_FBROOT(cur->bc_private.b.whichfork));
+	cur->bc_nlevels--;
+out0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	return 0;
+}
+
+STATIC int
+xfs_btree_dec_cursor(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			*stat)
+{
+	int			error;
+	int			i;
+
+	if (level > 0) {
+		error = xfs_btree_decrement(cur, level, &i);
+		if (error)
+			return error;
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = 1;
+	return 0;
+}
+
+/*
+ * Single level of the btree record deletion routine.
+ * Delete record pointed to by cur/level.
+ * Remove the record from its block then rebalance the tree.
+ * Return 0 for error, 1 for done, 2 to go on to the next level.
+ */
+STATIC int					/* error */
+xfs_btree_delrec(
+	struct xfs_btree_cur	*cur,		/* btree cursor */
+	int			level,		/* level removing record from */
+	int			*stat)		/* fail/done/go-on */
+{
+	struct xfs_btree_block	*block;		/* btree block */
+	union xfs_btree_ptr	cptr;		/* current block ptr */
+	struct xfs_buf		*bp;		/* buffer for block */
+	int			error;		/* error return value */
+	int			i;		/* loop counter */
+	union xfs_btree_key	key;		/* storage for keyp */
+	union xfs_btree_key	*keyp = &key;	/* passed to the next level */
+	union xfs_btree_ptr	lptr;		/* left sibling block ptr */
+	struct xfs_buf		*lbp;		/* left buffer pointer */
+	struct xfs_btree_block	*left;		/* left btree block */
+	int			lrecs = 0;	/* left record count */
+	int			ptr;		/* key/record index */
+	union xfs_btree_ptr	rptr;		/* right sibling block ptr */
+	struct xfs_buf		*rbp;		/* right buffer pointer */
+	struct xfs_btree_block	*right;		/* right btree block */
+	struct xfs_btree_block	*rrblock;	/* right-right btree block */
+	struct xfs_buf		*rrbp;		/* right-right buffer pointer */
+	int			rrecs = 0;	/* right record count */
+	struct xfs_btree_cur	*tcur;		/* temporary btree cursor */
+	int			numrecs;	/* temporary numrec count */
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_TRACE_ARGI(cur, level);
+
+	tcur = NULL;
+
+	/* Get the index of the entry being deleted, check for nothing there. */
+	ptr = cur->bc_ptrs[level];
+	if (ptr == 0) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+		*stat = 0;
+		return 0;
+	}
+
+	/* Get the buffer & block containing the record or key/ptr. */
+	block = xfs_btree_get_block(cur, level, &bp);
+	numrecs = xfs_btree_get_numrecs(block);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, level, bp);
+	if (error)
+		goto error0;
+#endif
+
+	/* Fail if we're off the end of the block. */
+	if (ptr > numrecs) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+		*stat = 0;
+		return 0;
+	}
+
+	XFS_BTREE_STATS_INC(cur, delrec);
+	XFS_BTREE_STATS_ADD(cur, moves, numrecs - ptr);
+
+	/* Excise the entries being deleted. */
+	if (level > 0) {
+		/* It's a nonleaf. operate on keys and ptrs */
+		union xfs_btree_key	*lkp;
+		union xfs_btree_ptr	*lpp;
+
+		lkp = xfs_btree_key_addr(cur, ptr + 1, block);
+		lpp = xfs_btree_ptr_addr(cur, ptr + 1, block);
+
+#ifdef DEBUG
+		for (i = 0; i < numrecs - ptr; i++) {
+			error = xfs_btree_check_ptr(cur, lpp, i, level);
+			if (error)
+				goto error0;
+		}
+#endif
+
+		if (ptr < numrecs) {
+			xfs_btree_shift_keys(cur, lkp, -1, numrecs - ptr);
+			xfs_btree_shift_ptrs(cur, lpp, -1, numrecs - ptr);
+			xfs_btree_log_keys(cur, bp, ptr, numrecs - 1);
+			xfs_btree_log_ptrs(cur, bp, ptr, numrecs - 1);
+		}
+
+		/*
+		 * If it's the first record in the block, we'll need to pass a
+		 * key up to the next level (updkey).
+		 */
+		if (ptr == 1)
+			keyp = xfs_btree_key_addr(cur, 1, block);
+	} else {
+		/* It's a leaf. operate on records */
+		if (ptr < numrecs) {
+			xfs_btree_shift_recs(cur,
+				xfs_btree_rec_addr(cur, ptr + 1, block),
+				-1, numrecs - ptr);
+			xfs_btree_log_recs(cur, bp, ptr, numrecs - 1);
+		}
+
+		/*
+		 * If it's the first record in the block, we'll need a key
+		 * structure to pass up to the next level (updkey).
+		 */
+		if (ptr == 1) {
+			cur->bc_ops->init_key_from_rec(&key,
+					xfs_btree_rec_addr(cur, 1, block));
+			keyp = &key;
+		}
+	}
+
+	/*
+	 * Decrement and log the number of entries in the block.
+	 */
+	xfs_btree_set_numrecs(block, --numrecs);
+	xfs_btree_log_block(cur, bp, XFS_BB_NUMRECS);
+
+	/*
+	 * If we are tracking the last record in the tree and
+	 * we are at the far right edge of the tree, update it.
+	 */
+	if (xfs_btree_is_lastrec(cur, block, level)) {
+		cur->bc_ops->update_lastrec(cur, block, NULL,
+					    ptr, LASTREC_DELREC);
+	}
+
+	/*
+	 * We're at the root level.  First, shrink the root block in-memory.
+	 * Try to get rid of the next level down.  If we can't then there's
+	 * nothing left to do.
+	 */
+	if (level == cur->bc_nlevels - 1) {
+		if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) {
+			xfs_iroot_realloc(cur->bc_private.b.ip, -1,
+					  cur->bc_private.b.whichfork);
+
+			error = xfs_btree_kill_iroot(cur);
+			if (error)
+				goto error0;
+
+			error = xfs_btree_dec_cursor(cur, level, stat);
+			if (error)
+				goto error0;
+			*stat = 1;
+			return 0;
+		}
+
+		/*
+		 * If this is the root level, and there's only one entry left,
+		 * and it's NOT the leaf level, then we can get rid of this
+		 * level.
+		 */
+		if (numrecs == 1 && level > 0) {
+			union xfs_btree_ptr	*pp;
+			/*
+			 * pp is still set to the first pointer in the block.
+			 * Make it the new root of the btree.
+			 */
+			pp = xfs_btree_ptr_addr(cur, 1, block);
+			error = cur->bc_ops->kill_root(cur, bp, level, pp);
+			if (error)
+				goto error0;
+		} else if (level > 0) {
+			error = xfs_btree_dec_cursor(cur, level, stat);
+			if (error)
+				goto error0;
+		}
+		*stat = 1;
+		return 0;
+	}
+
+	/*
+	 * If we deleted the leftmost entry in the block, update the
+	 * key values above us in the tree.
+	 */
+	if (ptr == 1) {
+		error = xfs_btree_updkey(cur, keyp, level + 1);
+		if (error)
+			goto error0;
+	}
+
+	/*
+	 * If the number of records remaining in the block is at least
+	 * the minimum, we're done.
+	 */
+	if (numrecs >= cur->bc_ops->get_minrecs(cur, level)) {
+		error = xfs_btree_dec_cursor(cur, level, stat);
+		if (error)
+			goto error0;
+		return 0;
+	}
+
+	/*
+	 * Otherwise, we have to move some records around to keep the
+	 * tree balanced.  Look at the left and right sibling blocks to
+	 * see if we can re-balance by moving only one record.
+	 */
+	xfs_btree_get_sibling(cur, block, &rptr, XFS_BB_RIGHTSIB);
+	xfs_btree_get_sibling(cur, block, &lptr, XFS_BB_LEFTSIB);
+
+	if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) {
+		/*
+		 * One child of root, need to get a chance to copy its contents
+		 * into the root and delete it. Can't go up to next level,
+		 * there's nothing to delete there.
+		 */
+		if (xfs_btree_ptr_is_null(cur, &rptr) &&
+		    xfs_btree_ptr_is_null(cur, &lptr) &&
+		    level == cur->bc_nlevels - 2) {
+			error = xfs_btree_kill_iroot(cur);
+			if (!error)
+				error = xfs_btree_dec_cursor(cur, level, stat);
+			if (error)
+				goto error0;
+			return 0;
+		}
+	}
+
+	ASSERT(!xfs_btree_ptr_is_null(cur, &rptr) ||
+	       !xfs_btree_ptr_is_null(cur, &lptr));
+
+	/*
+	 * Duplicate the cursor so our btree manipulations here won't
+	 * disrupt the next level up.
+	 */
+	error = xfs_btree_dup_cursor(cur, &tcur);
+	if (error)
+		goto error0;
+
+	/*
+	 * If there's a right sibling, see if it's ok to shift an entry
+	 * out of it.
+	 */
+	if (!xfs_btree_ptr_is_null(cur, &rptr)) {
+		/*
+		 * Move the temp cursor to the last entry in the next block.
+		 * Actually any entry but the first would suffice.
+		 */
+		i = xfs_btree_lastrec(tcur, level);
+		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+
+		error = xfs_btree_increment(tcur, level, &i);
+		if (error)
+			goto error0;
+		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+
+		i = xfs_btree_lastrec(tcur, level);
+		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+
+		/* Grab a pointer to the block. */
+		right = xfs_btree_get_block(tcur, level, &rbp);
+#ifdef DEBUG
+		error = xfs_btree_check_block(tcur, right, level, rbp);
+		if (error)
+			goto error0;
+#endif
+		/* Grab the current block number, for future use. */
+		xfs_btree_get_sibling(tcur, right, &cptr, XFS_BB_LEFTSIB);
+
+		/*
+		 * If right block is full enough so that removing one entry
+		 * won't make it too empty, and left-shifting an entry out
+		 * of right to us works, we're done.
+		 */
+		if (xfs_btree_get_numrecs(right) - 1 >=
+		    cur->bc_ops->get_minrecs(tcur, level)) {
+			error = xfs_btree_lshift(tcur, level, &i);
+			if (error)
+				goto error0;
+			if (i) {
+				ASSERT(xfs_btree_get_numrecs(block) >=
+				       cur->bc_ops->get_minrecs(tcur, level));
+
+				xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
+				tcur = NULL;
+
+				error = xfs_btree_dec_cursor(cur, level, stat);
+				if (error)
+					goto error0;
+				return 0;
+			}
+		}
+
+		/*
+		 * Otherwise, grab the number of records in right for
+		 * future reference, and fix up the temp cursor to point
+		 * to our block again (last record).
+		 */
+		rrecs = xfs_btree_get_numrecs(right);
+		if (!xfs_btree_ptr_is_null(cur, &lptr)) {
+			i = xfs_btree_firstrec(tcur, level);
+			XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+
+			error = xfs_btree_decrement(tcur, level, &i);
+			if (error)
+				goto error0;
+			XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+		}
+	}
+
+	/*
+	 * If there's a left sibling, see if it's ok to shift an entry
+	 * out of it.
+	 */
+	if (!xfs_btree_ptr_is_null(cur, &lptr)) {
+		/*
+		 * Move the temp cursor to the first entry in the
+		 * previous block.
+		 */
+		i = xfs_btree_firstrec(tcur, level);
+		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+
+		error = xfs_btree_decrement(tcur, level, &i);
+		if (error)
+			goto error0;
+		i = xfs_btree_firstrec(tcur, level);
+		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
+
+		/* Grab a pointer to the block. */
+		left = xfs_btree_get_block(tcur, level, &lbp);
+#ifdef DEBUG
+		error = xfs_btree_check_block(cur, left, level, lbp);
+		if (error)
+			goto error0;
+#endif
+		/* Grab the current block number, for future use. */
+		xfs_btree_get_sibling(tcur, left, &cptr, XFS_BB_RIGHTSIB);
+
+		/*
+		 * If left block is full enough so that removing one entry
+		 * won't make it too empty, and right-shifting an entry out
+		 * of left to us works, we're done.
+		 */
+		if (xfs_btree_get_numrecs(left) - 1 >=
+		    cur->bc_ops->get_minrecs(tcur, level)) {
+			error = xfs_btree_rshift(tcur, level, &i);
+			if (error)
+				goto error0;
+			if (i) {
+				ASSERT(xfs_btree_get_numrecs(block) >=
+				       cur->bc_ops->get_minrecs(tcur, level));
+				xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
+				tcur = NULL;
+				if (level == 0)
+					cur->bc_ptrs[0]++;
+				XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+				*stat = 1;
+				return 0;
+			}
+		}
+
+		/*
+		 * Otherwise, grab the number of records in right for
+		 * future reference.
+		 */
+		lrecs = xfs_btree_get_numrecs(left);
+	}
+
+	/* Delete the temp cursor, we're done with it. */
+	xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
+	tcur = NULL;
+
+	/* If here, we need to do a join to keep the tree balanced. */
+	ASSERT(!xfs_btree_ptr_is_null(cur, &cptr));
+
+	if (!xfs_btree_ptr_is_null(cur, &lptr) &&
+	    lrecs + xfs_btree_get_numrecs(block) <=
+			cur->bc_ops->get_maxrecs(cur, level)) {
+		/*
+		 * Set "right" to be the starting block,
+		 * "left" to be the left neighbor.
+		 */
+		rptr = cptr;
+		right = block;
+		rbp = bp;
+		error = xfs_btree_read_buf_block(cur, &lptr, level,
+							0, &left, &lbp);
+		if (error)
+			goto error0;
+
+	/*
+	 * If that won't work, see if we can join with the right neighbor block.
+	 */
+	} else if (!xfs_btree_ptr_is_null(cur, &rptr) &&
+		   rrecs + xfs_btree_get_numrecs(block) <=
+			cur->bc_ops->get_maxrecs(cur, level)) {
+		/*
+		 * Set "left" to be the starting block,
+		 * "right" to be the right neighbor.
+		 */
+		lptr = cptr;
+		left = block;
+		lbp = bp;
+		error = xfs_btree_read_buf_block(cur, &rptr, level,
+							0, &right, &rbp);
+		if (error)
+			goto error0;
+
+	/*
+	 * Otherwise, we can't fix the imbalance.
+	 * Just return.  This is probably a logic error, but it's not fatal.
+	 */
+	} else {
+		error = xfs_btree_dec_cursor(cur, level, stat);
+		if (error)
+			goto error0;
+		return 0;
+	}
+
+	rrecs = xfs_btree_get_numrecs(right);
+	lrecs = xfs_btree_get_numrecs(left);
+
+	/*
+	 * We're now going to join "left" and "right" by moving all the stuff
+	 * in "right" to "left" and deleting "right".
+	 */
+	XFS_BTREE_STATS_ADD(cur, moves, rrecs);
+	if (level > 0) {
+		/* It's a non-leaf.  Move keys and pointers. */
+		union xfs_btree_key	*lkp;	/* left btree key */
+		union xfs_btree_ptr	*lpp;	/* left address pointer */
+		union xfs_btree_key	*rkp;	/* right btree key */
+		union xfs_btree_ptr	*rpp;	/* right address pointer */
+
+		lkp = xfs_btree_key_addr(cur, lrecs + 1, left);
+		lpp = xfs_btree_ptr_addr(cur, lrecs + 1, left);
+		rkp = xfs_btree_key_addr(cur, 1, right);
+		rpp = xfs_btree_ptr_addr(cur, 1, right);
+#ifdef DEBUG
+		for (i = 1; i < rrecs; i++) {
+			error = xfs_btree_check_ptr(cur, rpp, i, level);
+			if (error)
+				goto error0;
+		}
+#endif
+		xfs_btree_copy_keys(cur, lkp, rkp, rrecs);
+		xfs_btree_copy_ptrs(cur, lpp, rpp, rrecs);
+
+		xfs_btree_log_keys(cur, lbp, lrecs + 1, lrecs + rrecs);
+		xfs_btree_log_ptrs(cur, lbp, lrecs + 1, lrecs + rrecs);
+	} else {
+		/* It's a leaf.  Move records.  */
+		union xfs_btree_rec	*lrp;	/* left record pointer */
+		union xfs_btree_rec	*rrp;	/* right record pointer */
+
+		lrp = xfs_btree_rec_addr(cur, lrecs + 1, left);
+		rrp = xfs_btree_rec_addr(cur, 1, right);
+
+		xfs_btree_copy_recs(cur, lrp, rrp, rrecs);
+		xfs_btree_log_recs(cur, lbp, lrecs + 1, lrecs + rrecs);
+	}
+
+	XFS_BTREE_STATS_INC(cur, join);
+
+	/*
+	 * Fix up the the number of records and right block pointer in the
+	 * surviving block, and log it.
+	 */
+	xfs_btree_set_numrecs(left, lrecs + rrecs);
+	xfs_btree_get_sibling(cur, right, &cptr, XFS_BB_RIGHTSIB),
+	xfs_btree_set_sibling(cur, left, &cptr, XFS_BB_RIGHTSIB);
+	xfs_btree_log_block(cur, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
+
+	/* If there is a right sibling, point it to the remaining block. */
+	xfs_btree_get_sibling(cur, left, &cptr, XFS_BB_RIGHTSIB);
+	if (!xfs_btree_ptr_is_null(cur, &cptr)) {
+		error = xfs_btree_read_buf_block(cur, &cptr, level,
+							0, &rrblock, &rrbp);
+		if (error)
+			goto error0;
+		xfs_btree_set_sibling(cur, rrblock, &lptr, XFS_BB_LEFTSIB);
+		xfs_btree_log_block(cur, rrbp, XFS_BB_LEFTSIB);
+	}
+
+	/* Free the deleted block. */
+	error = cur->bc_ops->free_block(cur, rbp);
+	if (error)
+		goto error0;
+	XFS_BTREE_STATS_INC(cur, free);
+
+	/*
+	 * If we joined with the left neighbor, set the buffer in the
+	 * cursor to the left block, and fix up the index.
+	 */
+	if (bp != lbp) {
+		cur->bc_bufs[level] = lbp;
+		cur->bc_ptrs[level] += lrecs;
+		cur->bc_ra[level] = 0;
+	}
+	/*
+	 * If we joined with the right neighbor and there's a level above
+	 * us, increment the cursor at that level.
+	 */
+	else if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) ||
+		   (level + 1 < cur->bc_nlevels)) {
+		error = xfs_btree_increment(cur, level + 1, &i);
+		if (error)
+			goto error0;
+	}
+
+	/*
+	 * Readjust the ptr at this level if it's not a leaf, since it's
+	 * still pointing at the deletion point, which makes the cursor
+	 * inconsistent.  If this makes the ptr 0, the caller fixes it up.
+	 * We can't use decrement because it would change the next level up.
+	 */
+	if (level > 0)
+		cur->bc_ptrs[level]--;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	/* Return value means the next level up has something to do. */
+	*stat = 2;
+	return 0;
+
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	if (tcur)
+		xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
+	return error;
+}
+
+/*
+ * Delete the record pointed to by cur.
+ * The cursor refers to the place where the record was (could be inserted)
+ * when the operation returns.
+ */
+int					/* error */
+xfs_btree_delete(
+	struct xfs_btree_cur	*cur,
+	int			*stat)	/* success/failure */
+{
+	int			error;	/* error return value */
+	int			level;
+	int			i;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+
+	/*
+	 * Go up the tree, starting at leaf level.
+	 *
+	 * If 2 is returned then a join was done; go to the next level.
+	 * Otherwise we are done.
+	 */
+	for (level = 0, i = 2; i == 2; level++) {
+		error = xfs_btree_delrec(cur, level, &i);
+		if (error)
+			goto error0;
+	}
+
+	if (i == 0) {
+		for (level = 1; level < cur->bc_nlevels; level++) {
+			if (cur->bc_ptrs[level] == 0) {
+				error = xfs_btree_decrement(cur, level, &i);
+				if (error)
+					goto error0;
+				break;
+			}
+		}
+	}
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+	*stat = i;
+	return 0;
+error0:
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
+	return error;
+}
+
+/*
+ * Get the data from the pointed-to record.
+ */
+int					/* error */
+xfs_btree_get_rec(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	union xfs_btree_rec	**recp,	/* output: btree record */
+	int			*stat)	/* output: success/failure */
+{
+	struct xfs_btree_block	*block;	/* btree block */
+	struct xfs_buf		*bp;	/* buffer pointer */
+	int			ptr;	/* record number */
+#ifdef DEBUG
+	int			error;	/* error return value */
+#endif
+
+	ptr = cur->bc_ptrs[0];
+	block = xfs_btree_get_block(cur, 0, &bp);
+
+#ifdef DEBUG
+	error = xfs_btree_check_block(cur, block, 0, bp);
+	if (error)
+		return error;
+#endif
+
+	/*
+	 * Off the right end or left end, return failure.
+	 */
+	if (ptr > xfs_btree_get_numrecs(block) || ptr <= 0) {
+		*stat = 0;
+		return 0;
+	}
+
+	/*
+	 * Point to the record and extract its data.
+	 */
+	*recp = xfs_btree_rec_addr(cur, ptr, block);
+	*stat = 1;
+	return 0;
+}

fs/xfs/xfs_btree.h

@@ -39,39 +39,19 @@ extern kmem_zone_t	*xfs_btree_cur_zone;
 #define	XFS_BTNUM_INO	((xfs_btnum_t)XFS_BTNUM_INOi)
 
 /*
- * Short form header: space allocation btrees.
- */
-typedef struct xfs_btree_sblock {
-	__be32		bb_magic;	/* magic number for block type */
-	__be16		bb_level;	/* 0 is a leaf */
-	__be16		bb_numrecs;	/* current # of data records */
-	__be32		bb_leftsib;	/* left sibling block or NULLAGBLOCK */
-	__be32		bb_rightsib;	/* right sibling block or NULLAGBLOCK */
-} xfs_btree_sblock_t;
-
-/*
- * Long form header: bmap btrees.
- */
-typedef struct xfs_btree_lblock {
-	__be32		bb_magic;	/* magic number for block type */
-	__be16		bb_level;	/* 0 is a leaf */
-	__be16		bb_numrecs;	/* current # of data records */
-	__be64		bb_leftsib;	/* left sibling block or NULLDFSBNO */
-	__be64		bb_rightsib;	/* right sibling block or NULLDFSBNO */
-} xfs_btree_lblock_t;
-
-/*
- * Combined header and structure, used by common code.
+ * Generic btree header.
+ *
+ * This is a comination of the actual format used on disk for short and long
+ * format btrees.  The first three fields are shared by both format, but
+ * the pointers are different and should be used with care.
+ *
+ * To get the size of the actual short or long form headers please use
+ * the size macros below.  Never use sizeof(xfs_btree_block).
  */
-typedef struct xfs_btree_hdr
-{
+struct xfs_btree_block {
 	__be32		bb_magic;	/* magic number for block type */
 	__be16		bb_level;	/* 0 is a leaf */
 	__be16		bb_numrecs;	/* current # of data records */
-} xfs_btree_hdr_t;
-
-typedef struct xfs_btree_block {
-	xfs_btree_hdr_t	bb_h;		/* header */
 	union {
 		struct {
 			__be32		bb_leftsib;
@@ -82,7 +62,36 @@ typedef struct xfs_btree_block {
 			__be64		bb_rightsib;
 		} l;			/* long form pointers */
 	} bb_u;				/* rest */
-} xfs_btree_block_t;
+};
+
+#define XFS_BTREE_SBLOCK_LEN	16	/* size of a short form block */
+#define XFS_BTREE_LBLOCK_LEN	24	/* size of a long form block */
+
+
+/*
+ * Generic key, ptr and record wrapper structures.
+ *
+ * These are disk format structures, and are converted where necessary
+ * by the btree specific code that needs to interpret them.
+ */
+union xfs_btree_ptr {
+	__be32			s;	/* short form ptr */
+	__be64			l;	/* long form ptr */
+};
+
+union xfs_btree_key {
+	xfs_bmbt_key_t		bmbt;
+	xfs_bmdr_key_t		bmbr;	/* bmbt root block */
+	xfs_alloc_key_t		alloc;
+	xfs_inobt_key_t		inobt;
+};
+
+union xfs_btree_rec {
+	xfs_bmbt_rec_t		bmbt;
+	xfs_bmdr_rec_t		bmbr;	/* bmbt root block */
+	xfs_alloc_rec_t		alloc;
+	xfs_inobt_rec_t		inobt;
+};
 
 /*
  * For logging record fields.
@@ -95,47 +104,132 @@ typedef struct xfs_btree_block {
 #define	XFS_BB_NUM_BITS		5
 #define	XFS_BB_ALL_BITS		((1 << XFS_BB_NUM_BITS) - 1)
 
-/*
- * Boolean to select which form of xfs_btree_block_t.bb_u to use.
- */
-#define	XFS_BTREE_LONG_PTRS(btnum)	((btnum) == XFS_BTNUM_BMAP)
-
 /*
  * Magic numbers for btree blocks.
  */
 extern const __uint32_t	xfs_magics[];
 
 /*
- * Maximum and minimum records in a btree block.
- * Given block size, type prefix, and leaf flag (0 or 1).
- * The divisor below is equivalent to lf ? (e1) : (e2) but that produces
- * compiler warnings.
- */
-#define	XFS_BTREE_BLOCK_MAXRECS(bsz,t,lf)	\
-	((int)(((bsz) - (uint)sizeof(t ## _block_t)) / \
-	 (((lf) * (uint)sizeof(t ## _rec_t)) + \
-	  ((1 - (lf)) * \
-	   ((uint)sizeof(t ## _key_t) + (uint)sizeof(t ## _ptr_t))))))
-#define	XFS_BTREE_BLOCK_MINRECS(bsz,t,lf)	\
-	(XFS_BTREE_BLOCK_MAXRECS(bsz,t,lf) / 2)
-
-/*
- * Record, key, and pointer address calculation macros.
- * Given block size, type prefix, block pointer, and index of requested entry
- * (first entry numbered 1).
- */
-#define	XFS_BTREE_REC_ADDR(t,bb,i)	\
-	((t ## _rec_t *)((char *)(bb) + sizeof(t ## _block_t) + \
-	 ((i) - 1) * sizeof(t ## _rec_t)))
-#define	XFS_BTREE_KEY_ADDR(t,bb,i)	\
-	((t ## _key_t *)((char *)(bb) + sizeof(t ## _block_t) + \
-	 ((i) - 1) * sizeof(t ## _key_t)))
-#define	XFS_BTREE_PTR_ADDR(t,bb,i,mxr)	\
-	((t ## _ptr_t *)((char *)(bb) + sizeof(t ## _block_t) + \
-	 (mxr) * sizeof(t ## _key_t) + ((i) - 1) * sizeof(t ## _ptr_t)))
+ * Generic stats interface
+ */
+#define __XFS_BTREE_STATS_INC(type, stat) \
+	XFS_STATS_INC(xs_ ## type ## _2_ ## stat)
+#define XFS_BTREE_STATS_INC(cur, stat)  \
+do {    \
+	switch (cur->bc_btnum) {  \
+	case XFS_BTNUM_BNO: __XFS_BTREE_STATS_INC(abtb, stat); break;	\
+	case XFS_BTNUM_CNT: __XFS_BTREE_STATS_INC(abtc, stat); break;	\
+	case XFS_BTNUM_BMAP: __XFS_BTREE_STATS_INC(bmbt, stat); break;	\
+	case XFS_BTNUM_INO: __XFS_BTREE_STATS_INC(ibt, stat); break;	\
+	case XFS_BTNUM_MAX: ASSERT(0); /* fucking gcc */ ; break;	\
+	}       \
+} while (0)
+
+#define __XFS_BTREE_STATS_ADD(type, stat, val) \
+	XFS_STATS_ADD(xs_ ## type ## _2_ ## stat, val)
+#define XFS_BTREE_STATS_ADD(cur, stat, val)  \
+do {    \
+	switch (cur->bc_btnum) {  \
+	case XFS_BTNUM_BNO: __XFS_BTREE_STATS_ADD(abtb, stat, val); break; \
+	case XFS_BTNUM_CNT: __XFS_BTREE_STATS_ADD(abtc, stat, val); break; \
+	case XFS_BTNUM_BMAP: __XFS_BTREE_STATS_ADD(bmbt, stat, val); break; \
+	case XFS_BTNUM_INO: __XFS_BTREE_STATS_ADD(ibt, stat, val); break; \
+	case XFS_BTNUM_MAX: ASSERT(0); /* fucking gcc */ ; break;	\
+	}       \
+} while (0)
 
 #define	XFS_BTREE_MAXLEVELS	8	/* max of all btrees */
 
+struct xfs_btree_ops {
+	/* size of the key and record structures */
+	size_t	key_len;
+	size_t	rec_len;
+
+	/* cursor operations */
+	struct xfs_btree_cur *(*dup_cursor)(struct xfs_btree_cur *);
+	void	(*update_cursor)(struct xfs_btree_cur *src,
+				 struct xfs_btree_cur *dst);
+
+	/* update btree root pointer */
+	void	(*set_root)(struct xfs_btree_cur *cur,
+				union xfs_btree_ptr *nptr, int level_change);
+	int	(*kill_root)(struct xfs_btree_cur *cur, struct xfs_buf *bp,
+				int level, union xfs_btree_ptr *newroot);
+
+	/* block allocation / freeing */
+	int	(*alloc_block)(struct xfs_btree_cur *cur,
+			       union xfs_btree_ptr *start_bno,
+			       union xfs_btree_ptr *new_bno,
+			       int length, int *stat);
+	int	(*free_block)(struct xfs_btree_cur *cur, struct xfs_buf *bp);
+
+	/* update last record information */
+	void	(*update_lastrec)(struct xfs_btree_cur *cur,
+				  struct xfs_btree_block *block,
+				  union xfs_btree_rec *rec,
+				  int ptr, int reason);
+
+	/* records in block/level */
+	int	(*get_minrecs)(struct xfs_btree_cur *cur, int level);
+	int	(*get_maxrecs)(struct xfs_btree_cur *cur, int level);
+
+	/* records on disk.  Matter for the root in inode case. */
+	int	(*get_dmaxrecs)(struct xfs_btree_cur *cur, int level);
+
+	/* init values of btree structures */
+	void	(*init_key_from_rec)(union xfs_btree_key *key,
+				     union xfs_btree_rec *rec);
+	void	(*init_rec_from_key)(union xfs_btree_key *key,
+				     union xfs_btree_rec *rec);
+	void	(*init_rec_from_cur)(struct xfs_btree_cur *cur,
+				     union xfs_btree_rec *rec);
+	void	(*init_ptr_from_cur)(struct xfs_btree_cur *cur,
+				     union xfs_btree_ptr *ptr);
+
+	/* difference between key value and cursor value */
+	__int64_t (*key_diff)(struct xfs_btree_cur *cur,
+			      union xfs_btree_key *key);
+
+#ifdef DEBUG
+	/* check that k1 is lower than k2 */
+	int	(*keys_inorder)(struct xfs_btree_cur *cur,
+				union xfs_btree_key *k1,
+				union xfs_btree_key *k2);
+
+	/* check that r1 is lower than r2 */
+	int	(*recs_inorder)(struct xfs_btree_cur *cur,
+				union xfs_btree_rec *r1,
+				union xfs_btree_rec *r2);
+#endif
+
+	/* btree tracing */
+#ifdef XFS_BTREE_TRACE
+	void		(*trace_enter)(struct xfs_btree_cur *, const char *,
+				       char *, int, int, __psunsigned_t,
+				       __psunsigned_t, __psunsigned_t,
+				       __psunsigned_t, __psunsigned_t,
+				       __psunsigned_t, __psunsigned_t,
+				       __psunsigned_t, __psunsigned_t,
+				       __psunsigned_t, __psunsigned_t);
+	void		(*trace_cursor)(struct xfs_btree_cur *, __uint32_t *,
+					__uint64_t *, __uint64_t *);
+	void		(*trace_key)(struct xfs_btree_cur *,
+				     union xfs_btree_key *, __uint64_t *,
+				     __uint64_t *);
+	void		(*trace_record)(struct xfs_btree_cur *,
+					union xfs_btree_rec *, __uint64_t *,
+					__uint64_t *, __uint64_t *);
+#endif
+};
+
+/*
+ * Reasons for the update_lastrec method to be called.
+ */
+#define LASTREC_UPDATE	0
+#define LASTREC_INSREC	1
+#define LASTREC_DELREC	2
+
+
 /*
  * Btree cursor structure.
  * This collects all information needed by the btree code in one place.
@@ -144,6 +238,8 @@ typedef struct xfs_btree_cur
 {
 	struct xfs_trans	*bc_tp;	/* transaction we're in, if any */
 	struct xfs_mount	*bc_mp;	/* file system mount struct */
+	const struct xfs_btree_ops *bc_ops;
+	uint			bc_flags; /* btree features - below */
 	union {
 		xfs_alloc_rec_incore_t	a;
 		xfs_bmbt_irec_t		b;
@@ -175,94 +271,40 @@ typedef struct xfs_btree_cur
 	}		bc_private;	/* per-btree type data */
 } xfs_btree_cur_t;
 
+/* cursor flags */
+#define XFS_BTREE_LONG_PTRS		(1<<0)	/* pointers are 64bits long */
+#define XFS_BTREE_ROOT_IN_INODE		(1<<1)	/* root may be variable size */
+#define XFS_BTREE_LASTREC_UPDATE	(1<<2)	/* track last rec externally */
+
+
 #define	XFS_BTREE_NOERROR	0
 #define	XFS_BTREE_ERROR		1
 
 /*
  * Convert from buffer to btree block header.
  */
-#define	XFS_BUF_TO_BLOCK(bp)	((xfs_btree_block_t *)XFS_BUF_PTR(bp))
-#define	XFS_BUF_TO_LBLOCK(bp)	((xfs_btree_lblock_t *)XFS_BUF_PTR(bp))
-#define	XFS_BUF_TO_SBLOCK(bp)	((xfs_btree_sblock_t *)XFS_BUF_PTR(bp))
+#define	XFS_BUF_TO_BLOCK(bp)	((struct xfs_btree_block *)XFS_BUF_PTR(bp))
 
 
-#ifdef __KERNEL__
-
-#ifdef DEBUG
 /*
- * Debug routine: check that block header is ok.
+ * Check that block header is ok.
  */
-void
+int
 xfs_btree_check_block(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_btree_block_t	*block,	/* generic btree block pointer */
-	int			level,	/* level of the btree block */
-	struct xfs_buf		*bp);	/* buffer containing block, if any */
-
-/*
- * Debug routine: check that keys are in the right order.
- */
-void
-xfs_btree_check_key(
-	xfs_btnum_t		btnum,	/* btree identifier */
-	void			*ak1,	/* pointer to left (lower) key */
-	void			*ak2);	/* pointer to right (higher) key */
-
-/*
- * Debug routine: check that records are in the right order.
- */
-void
-xfs_btree_check_rec(
-	xfs_btnum_t		btnum,	/* btree identifier */
-	void			*ar1,	/* pointer to left (lower) record */
-	void			*ar2);	/* pointer to right (higher) record */
-#else
-#define	xfs_btree_check_block(a,b,c,d)
-#define	xfs_btree_check_key(a,b,c)
-#define	xfs_btree_check_rec(a,b,c)
-#endif	/* DEBUG */
-
-/*
- * Checking routine: check that long form block header is ok.
- */
-int					/* error (0 or EFSCORRUPTED) */
-xfs_btree_check_lblock(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_btree_lblock_t	*block,	/* btree long form block pointer */
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	struct xfs_btree_block	*block,	/* generic btree block pointer */
 	int			level,	/* level of the btree block */
 	struct xfs_buf		*bp);	/* buffer containing block, if any */
 
 /*
- * Checking routine: check that (long) pointer is ok.
+ * Check that (long) pointer is ok.
  */
 int					/* error (0 or EFSCORRUPTED) */
 xfs_btree_check_lptr(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
+	struct xfs_btree_cur	*cur,	/* btree cursor */
 	xfs_dfsbno_t		ptr,	/* btree block disk address */
 	int			level);	/* btree block level */
 
-#define xfs_btree_check_lptr_disk(cur, ptr, level) \
-	xfs_btree_check_lptr(cur, be64_to_cpu(ptr), level)
-
-/*
- * Checking routine: check that short form block header is ok.
- */
-int					/* error (0 or EFSCORRUPTED) */
-xfs_btree_check_sblock(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_btree_sblock_t	*block,	/* btree short form block pointer */
-	int			level,	/* level of the btree block */
-	struct xfs_buf		*bp);	/* buffer containing block */
-
-/*
- * Checking routine: check that (short) pointer is ok.
- */
-int					/* error (0 or EFSCORRUPTED) */
-xfs_btree_check_sptr(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_agblock_t		ptr,	/* btree block disk address */
-	int			level);	/* btree block level */
-
 /*
  * Delete the btree cursor.
  */
@@ -280,15 +322,6 @@ xfs_btree_dup_cursor(
 	xfs_btree_cur_t		*cur,	/* input cursor */
 	xfs_btree_cur_t		**ncur);/* output cursor */
 
-/*
- * Change the cursor to point to the first record in the current block
- * at the given level.  Other levels are unaffected.
- */
-int					/* success=1, failure=0 */
-xfs_btree_firstrec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level);	/* level to change */
-
 /*
  * Get a buffer for the block, return it with no data read.
  * Long-form addressing.
@@ -312,20 +345,6 @@ xfs_btree_get_bufs(
 	xfs_agblock_t		agbno,	/* allocation group block number */
 	uint			lock);	/* lock flags for get_buf */
 
-/*
- * Allocate a new btree cursor.
- * The cursor is either for allocation (A) or bmap (B).
- */
-xfs_btree_cur_t *			/* new btree cursor */
-xfs_btree_init_cursor(
-	struct xfs_mount	*mp,	/* file system mount point */
-	struct xfs_trans	*tp,	/* transaction pointer */
-	struct xfs_buf		*agbp,	/* (A only) buffer for agf structure */
-	xfs_agnumber_t		agno,	/* (A only) allocation group number */
-	xfs_btnum_t		btnum,	/* btree identifier */
-	struct xfs_inode	*ip,	/* (B only) inode owning the btree */
-	int			whichfork); /* (B only) data/attr fork */
-
 /*
  * Check for the cursor referring to the last block at the given level.
  */
@@ -334,15 +353,6 @@ xfs_btree_islastblock(
 	xfs_btree_cur_t		*cur,	/* btree cursor */
 	int			level);	/* level to check */
 
-/*
- * Change the cursor to point to the last record in the current block
- * at the given level.  Other levels are unaffected.
- */
-int					/* success=1, failure=0 */
-xfs_btree_lastrec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level);	/* level to change */
-
 /*
  * Compute first and last byte offsets for the fields given.
  * Interprets the offsets table, which contains struct field offsets.
@@ -404,39 +414,53 @@ xfs_btree_reada_bufs(
 	xfs_extlen_t		count);	/* count of filesystem blocks */
 
 /*
- * Read-ahead btree blocks, at the given level.
- * Bits in lr are set from XFS_BTCUR_{LEFT,RIGHT}RA.
+ * Set the buffer for level "lev" in the cursor to bp, releasing
+ * any previous buffer.
  */
-int					/* readahead block count */
-xfs_btree_readahead_core(
+void
+xfs_btree_setbuf(
 	xfs_btree_cur_t		*cur,	/* btree cursor */
 	int			lev,	/* level in btree */
-	int			lr);	/* left/right bits */
+	struct xfs_buf		*bp);	/* new buffer to set */
 
-static inline int			/* readahead block count */
-xfs_btree_readahead(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			lev,	/* level in btree */
-	int			lr)	/* left/right bits */
-{
-	if ((cur->bc_ra[lev] | lr) == cur->bc_ra[lev])
-		return 0;
 
-	return xfs_btree_readahead_core(cur, lev, lr);
-}
+/*
+ * Common btree core entry points.
+ */
+int xfs_btree_increment(struct xfs_btree_cur *, int, int *);
+int xfs_btree_decrement(struct xfs_btree_cur *, int, int *);
+int xfs_btree_lookup(struct xfs_btree_cur *, xfs_lookup_t, int *);
+int xfs_btree_update(struct xfs_btree_cur *, union xfs_btree_rec *);
+int xfs_btree_new_iroot(struct xfs_btree_cur *, int *, int *);
+int xfs_btree_kill_iroot(struct xfs_btree_cur *);
+int xfs_btree_insert(struct xfs_btree_cur *, int *);
+int xfs_btree_delete(struct xfs_btree_cur *, int *);
+int xfs_btree_get_rec(struct xfs_btree_cur *, union xfs_btree_rec **, int *);
 
+/*
+ * Internal btree helpers also used by xfs_bmap.c.
+ */
+void xfs_btree_log_block(struct xfs_btree_cur *, struct xfs_buf *, int);
+void xfs_btree_log_recs(struct xfs_btree_cur *, struct xfs_buf *, int, int);
 
 /*
- * Set the buffer for level "lev" in the cursor to bp, releasing
- * any previous buffer.
+ * Helpers.
  */
-void
-xfs_btree_setbuf(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			lev,	/* level in btree */
-	struct xfs_buf		*bp);	/* new buffer to set */
+static inline int xfs_btree_get_numrecs(struct xfs_btree_block *block)
+{
+	return be16_to_cpu(block->bb_numrecs);
+}
+
+static inline void xfs_btree_set_numrecs(struct xfs_btree_block *block,
+		__uint16_t numrecs)
+{
+	block->bb_numrecs = cpu_to_be16(numrecs);
+}
 
-#endif	/* __KERNEL__ */
+static inline int xfs_btree_get_level(struct xfs_btree_block *block)
+{
+	return be16_to_cpu(block->bb_level);
+}
 
 
 /*

fs/xfs/xfs_btree_trace.c

@@ -0,0 +1,249 @@
+/*
+ * Copyright (c) 2008 Silicon Graphics, Inc.
+ * All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#include "xfs.h"
+#include "xfs_types.h"
+#include "xfs_inum.h"
+#include "xfs_bmap_btree.h"
+#include "xfs_alloc_btree.h"
+#include "xfs_ialloc_btree.h"
+#include "xfs_inode.h"
+#include "xfs_btree.h"
+#include "xfs_btree_trace.h"
+
+STATIC void
+xfs_btree_trace_ptr(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	ptr,
+	__psunsigned_t		*high,
+	__psunsigned_t		*low)
+{
+	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+		__u64 val = be64_to_cpu(ptr.l);
+		*high = val >> 32;
+		*low = (int)val;
+	} else {
+		*high = 0;
+		*low = be32_to_cpu(ptr.s);
+	}
+}
+
+/*
+ * Add a trace buffer entry for arguments, for a buffer & 1 integer arg.
+ */
+void
+xfs_btree_trace_argbi(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*b,
+	int			i,
+	int			line)
+{
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGBI,
+				 line, (__psunsigned_t)b, i, 0, 0, 0, 0, 0,
+				 0, 0, 0, 0);
+}
+
+/*
+ * Add a trace buffer entry for arguments, for a buffer & 2 integer args.
+ */
+void
+xfs_btree_trace_argbii(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*b,
+	int			i0,
+	int			i1,
+	int			line)
+{
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGBII,
+				 line, (__psunsigned_t)b, i0, i1, 0, 0, 0, 0,
+				 0, 0, 0, 0);
+}
+
+/*
+ * Add a trace buffer entry for arguments, for 3 block-length args
+ * and an integer arg.
+ */
+void
+xfs_btree_trace_argfffi(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	xfs_dfiloff_t		o,
+	xfs_dfsbno_t		b,
+	xfs_dfilblks_t		i,
+	int			j,
+	int			line)
+{
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGFFFI,
+				 line,
+				 o >> 32, (int)o,
+				 b >> 32, (int)b,
+				 i >> 32, (int)i,
+				 (int)j, 0, 0, 0, 0);
+}
+
+/*
+ * Add a trace buffer entry for arguments, for one integer arg.
+ */
+void
+xfs_btree_trace_argi(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	int			i,
+	int			line)
+{
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGI,
+				 line, i, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+}
+
+/*
+ * Add a trace buffer entry for arguments, for int, fsblock, key.
+ */
+void
+xfs_btree_trace_argipk(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	int			i,
+	union xfs_btree_ptr	ptr,
+	union xfs_btree_key	*key,
+	int			line)
+{
+	__psunsigned_t		high, low;
+	__uint64_t		l0, l1;
+
+	xfs_btree_trace_ptr(cur, ptr, &high, &low);
+	cur->bc_ops->trace_key(cur, key, &l0, &l1);
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGIPK,
+				 line, i, high, low,
+				 l0 >> 32, (int)l0,
+				 l1 >> 32, (int)l1,
+				 0, 0, 0, 0);
+}
+
+/*
+ * Add a trace buffer entry for arguments, for int, fsblock, rec.
+ */
+void
+xfs_btree_trace_argipr(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	int			i,
+	union xfs_btree_ptr	ptr,
+	union xfs_btree_rec	*rec,
+	int			line)
+{
+	__psunsigned_t		high, low;
+	__uint64_t		l0, l1, l2;
+
+	xfs_btree_trace_ptr(cur, ptr, &high, &low);
+	cur->bc_ops->trace_record(cur, rec, &l0, &l1, &l2);
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGIPR,
+			      line, i,
+			      high, low,
+			      l0 >> 32, (int)l0,
+			      l1 >> 32, (int)l1,
+			      l2 >> 32, (int)l2,
+			      0, 0);
+}
+
+/*
+ * Add a trace buffer entry for arguments, for int, key.
+ */
+void
+xfs_btree_trace_argik(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	int			i,
+	union xfs_btree_key	*key,
+	int			line)
+{
+	__uint64_t		l0, l1;
+
+	cur->bc_ops->trace_key(cur, key, &l0, &l1);
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGIK,
+				 line, i,
+				 l0 >> 32, (int)l0,
+				 l1 >> 32, (int)l1,
+				 0, 0, 0, 0, 0, 0);
+}
+
+/*
+ * Add a trace buffer entry for arguments, for record.
+ */
+void
+xfs_btree_trace_argr(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec,
+	int			line)
+{
+	__uint64_t		l0, l1, l2;
+
+	cur->bc_ops->trace_record(cur, rec, &l0, &l1, &l2);
+	cur->bc_ops->trace_enter(cur, func, XBT_ARGS, XFS_BTREE_KTRACE_ARGR,
+			      line,
+			      l0 >> 32, (int)l0,
+			      l1 >> 32, (int)l1,
+			      l2 >> 32, (int)l2,
+			      0, 0, 0, 0, 0);
+}
+
+/*
+ * Add a trace buffer entry for the cursor/operation.
+ */
+void
+xfs_btree_trace_cursor(
+	const char		*func,
+	struct xfs_btree_cur	*cur,
+	int			type,
+	int			line)
+{
+	__uint32_t		s0;
+	__uint64_t		l0, l1;
+	char			*s;
+
+	switch (type) {
+	case XBT_ARGS:
+		s = "args";
+		break;
+	case XBT_ENTRY:
+		s = "entry";
+		break;
+	case XBT_ERROR:
+		s = "error";
+		break;
+	case XBT_EXIT:
+		s = "exit";
+		break;
+	default:
+		s = "unknown";
+		break;
+	}
+
+	cur->bc_ops->trace_cursor(cur, &s0, &l0, &l1);
+	cur->bc_ops->trace_enter(cur, func, s, XFS_BTREE_KTRACE_CUR, line,
+				 s0,
+				 l0 >> 32, (int)l0,
+				 l1 >> 32, (int)l1,
+				 (__psunsigned_t)cur->bc_bufs[0],
+				 (__psunsigned_t)cur->bc_bufs[1],
+				 (__psunsigned_t)cur->bc_bufs[2],
+				 (__psunsigned_t)cur->bc_bufs[3],
+				 (cur->bc_ptrs[0] << 16) | cur->bc_ptrs[1],
+				 (cur->bc_ptrs[2] << 16) | cur->bc_ptrs[3]);
+}

fs/xfs/xfs_btree_trace.h

@@ -0,0 +1,116 @@
+/*
+ * Copyright (c) 2008 Silicon Graphics, Inc.
+ * All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+#ifndef __XFS_BTREE_TRACE_H__
+#define	__XFS_BTREE_TRACE_H__
+
+struct xfs_btree_cur;
+struct xfs_buf;
+
+
+/*
+ * Trace hooks.
+ * i,j = integer (32 bit)
+ * b = btree block buffer (xfs_buf_t)
+ * p = btree ptr
+ * r = btree record
+ * k = btree key
+ */
+
+#ifdef XFS_BTREE_TRACE
+
+/*
+ * Trace buffer entry types.
+ */
+#define XFS_BTREE_KTRACE_ARGBI   1
+#define XFS_BTREE_KTRACE_ARGBII  2
+#define XFS_BTREE_KTRACE_ARGFFFI 3
+#define XFS_BTREE_KTRACE_ARGI    4
+#define XFS_BTREE_KTRACE_ARGIPK  5
+#define XFS_BTREE_KTRACE_ARGIPR  6
+#define XFS_BTREE_KTRACE_ARGIK   7
+#define XFS_BTREE_KTRACE_ARGR	 8
+#define XFS_BTREE_KTRACE_CUR     9
+
+/*
+ * Sub-types for cursor traces.
+ */
+#define XBT_ARGS	0
+#define XBT_ENTRY	1
+#define XBT_ERROR	2
+#define XBT_EXIT	3
+
+void xfs_btree_trace_argbi(const char *, struct xfs_btree_cur *,
+		struct xfs_buf *, int, int);
+void xfs_btree_trace_argbii(const char *, struct xfs_btree_cur *,
+		struct xfs_buf *, int, int, int);
+void xfs_btree_trace_argfffi(const char *, struct xfs_btree_cur *,
+		xfs_dfiloff_t, xfs_dfsbno_t, xfs_dfilblks_t, int, int);
+void xfs_btree_trace_argi(const char *, struct xfs_btree_cur *, int, int);
+void xfs_btree_trace_argipk(const char *, struct xfs_btree_cur *, int,
+		union xfs_btree_ptr, union xfs_btree_key *, int);
+void xfs_btree_trace_argipr(const char *, struct xfs_btree_cur *, int,
+		union xfs_btree_ptr, union xfs_btree_rec *, int);
+void xfs_btree_trace_argik(const char *, struct xfs_btree_cur *, int,
+		union xfs_btree_key *, int);
+void xfs_btree_trace_argr(const char *, struct xfs_btree_cur *,
+		union xfs_btree_rec *, int);
+void xfs_btree_trace_cursor(const char *, struct xfs_btree_cur *, int, int);
+
+
+#define XFS_ALLOCBT_TRACE_SIZE	4096	/* size of global trace buffer */
+extern ktrace_t	*xfs_allocbt_trace_buf;
+
+#define XFS_INOBT_TRACE_SIZE	4096	/* size of global trace buffer */
+extern ktrace_t	*xfs_inobt_trace_buf;
+
+#define XFS_BMBT_TRACE_SIZE	4096	/* size of global trace buffer */
+#define XFS_BMBT_KTRACE_SIZE	32	/* size of per-inode trace buffer */
+extern ktrace_t	*xfs_bmbt_trace_buf;
+
+
+#define	XFS_BTREE_TRACE_ARGBI(c, b, i)	\
+	xfs_btree_trace_argbi(__func__, c, b, i, __LINE__)
+#define	XFS_BTREE_TRACE_ARGBII(c, b, i, j)	\
+	xfs_btree_trace_argbii(__func__, c, b, i, j, __LINE__)
+#define	XFS_BTREE_TRACE_ARGFFFI(c, o, b, i, j)	\
+	xfs_btree_trace_argfffi(__func__, c, o, b, i, j, __LINE__)
+#define	XFS_BTREE_TRACE_ARGI(c, i)	\
+	xfs_btree_trace_argi(__func__, c, i, __LINE__)
+#define	XFS_BTREE_TRACE_ARGIPK(c, i, p, k)	\
+	xfs_btree_trace_argipk(__func__, c, i, p, k, __LINE__)
+#define	XFS_BTREE_TRACE_ARGIPR(c, i, p, r)	\
+	xfs_btree_trace_argipr(__func__, c, i, p, r, __LINE__)
+#define	XFS_BTREE_TRACE_ARGIK(c, i, k)	\
+	xfs_btree_trace_argik(__func__, c, i, k, __LINE__)
+#define XFS_BTREE_TRACE_ARGR(c, r)	\
+	xfs_btree_trace_argr(__func__, c, r, __LINE__)
+#define	XFS_BTREE_TRACE_CURSOR(c, t)	\
+	xfs_btree_trace_cursor(__func__, c, t, __LINE__)
+#else
+#define	XFS_BTREE_TRACE_ARGBI(c, b, i)
+#define	XFS_BTREE_TRACE_ARGBII(c, b, i, j)
+#define	XFS_BTREE_TRACE_ARGFFFI(c, o, b, i, j)
+#define	XFS_BTREE_TRACE_ARGI(c, i)
+#define	XFS_BTREE_TRACE_ARGIPK(c, i, p, s)
+#define	XFS_BTREE_TRACE_ARGIPR(c, i, p, r)
+#define	XFS_BTREE_TRACE_ARGIK(c, i, k)
+#define XFS_BTREE_TRACE_ARGR(c, r)
+#define	XFS_BTREE_TRACE_CURSOR(c, t)
+#endif	/* XFS_BTREE_TRACE */
+
+#endif /* __XFS_BTREE_TRACE_H__ */

fs/xfs/xfs_buf_item.c

@@ -375,7 +375,7 @@ xfs_buf_item_unpin(
 	xfs_buf_log_item_t	*bip,
 	int			stale)
 {
-	xfs_mount_t	*mp;
+	struct xfs_ail	*ailp;
 	xfs_buf_t	*bp;
 	int		freed;
 
@@ -387,7 +387,7 @@ xfs_buf_item_unpin(
 	xfs_buftrace("XFS_UNPIN", bp);
 
 	freed = atomic_dec_and_test(&bip->bli_refcount);
-	mp = bip->bli_item.li_mountp;
+	ailp = bip->bli_item.li_ailp;
 	xfs_bunpin(bp);
 	if (freed && stale) {
 		ASSERT(bip->bli_flags & XFS_BLI_STALE);
@@ -399,17 +399,17 @@ xfs_buf_item_unpin(
 		xfs_buftrace("XFS_UNPIN STALE", bp);
 		/*
 		 * If we get called here because of an IO error, we may
-		 * or may not have the item on the AIL. xfs_trans_delete_ail()
+		 * or may not have the item on the AIL. xfs_trans_ail_delete()
 		 * will take care of that situation.
-		 * xfs_trans_delete_ail() drops the AIL lock.
+		 * xfs_trans_ail_delete() drops the AIL lock.
 		 */
 		if (bip->bli_flags & XFS_BLI_STALE_INODE) {
 			xfs_buf_do_callbacks(bp, (xfs_log_item_t *)bip);
 			XFS_BUF_SET_FSPRIVATE(bp, NULL);
 			XFS_BUF_CLR_IODONE_FUNC(bp);
 		} else {
-			spin_lock(&mp->m_ail_lock);
-			xfs_trans_delete_ail(mp, (xfs_log_item_t *)bip);
+			spin_lock(&ailp->xa_lock);
+			xfs_trans_ail_delete(ailp, (xfs_log_item_t *)bip);
 			xfs_buf_item_relse(bp);
 			ASSERT(XFS_BUF_FSPRIVATE(bp, void *) == NULL);
 		}
@@ -731,6 +731,7 @@ xfs_buf_item_init(
 	bip->bli_item.li_type = XFS_LI_BUF;
 	bip->bli_item.li_ops = &xfs_buf_item_ops;
 	bip->bli_item.li_mountp = mp;
+	bip->bli_item.li_ailp = mp->m_ail;
 	bip->bli_buf = bp;
 	xfs_buf_hold(bp);
 	bip->bli_format.blf_type = XFS_LI_BUF;
@@ -1122,27 +1123,23 @@ xfs_buf_iodone(
 	xfs_buf_t		*bp,
 	xfs_buf_log_item_t	*bip)
 {
-	struct xfs_mount	*mp;
+	struct xfs_ail		*ailp = bip->bli_item.li_ailp;
 
 	ASSERT(bip->bli_buf == bp);
 
 	xfs_buf_rele(bp);
-	mp = bip->bli_item.li_mountp;
 
 	/*
 	 * If we are forcibly shutting down, this may well be
 	 * off the AIL already. That's because we simulate the
 	 * log-committed callbacks to unpin these buffers. Or we may never
 	 * have put this item on AIL because of the transaction was
-	 * aborted forcibly. xfs_trans_delete_ail() takes care of these.
+	 * aborted forcibly. xfs_trans_ail_delete() takes care of these.
 	 *
 	 * Either way, AIL is useless if we're forcing a shutdown.
 	 */
-	spin_lock(&mp->m_ail_lock);
-	/*
-	 * xfs_trans_delete_ail() drops the AIL lock.
-	 */
-	xfs_trans_delete_ail(mp, (xfs_log_item_t *)bip);
+	spin_lock(&ailp->xa_lock);
+	xfs_trans_ail_delete(ailp, (xfs_log_item_t *)bip);
 	xfs_buf_item_free(bip);
 }
 

fs/xfs/xfs_clnt.h

@@ -1,105 +0,0 @@
-/*
- * Copyright (c) 2000-2005 Silicon Graphics, Inc.
- * All Rights Reserved.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it would be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write the Free Software Foundation,
- * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
- */
-#ifndef __XFS_CLNT_H__
-#define __XFS_CLNT_H__
-
-/*
- * XFS arguments structure, constructed from the arguments we
- * are passed via the mount system call.
- *
- * NOTE: The mount system call is handled differently between
- * Linux and IRIX.  In IRIX we worked work with a binary data
- * structure coming in across the syscall interface from user
- * space (the mount userspace knows about each filesystem type
- * and the set of valid options for it, and converts the users
- * argument string into a binary structure _before_ making the
- * system call), and the ABI issues that this implies.
- *
- * In Linux, we are passed a comma separated set of options;
- * ie. a NULL terminated string of characters.  Userspace mount
- * code does not have any knowledge of mount options expected by
- * each filesystem type and so each filesystem parses its mount
- * options in kernel space.
- *
- * For the Linux port, we kept this structure pretty much intact
- * and use it internally (because the existing code groks it).
- */
-struct xfs_mount_args {
-	int	flags;		/* flags -> see XFSMNT_... macros below */
-	int	flags2;		/* flags -> see XFSMNT2_... macros below */
-	int	logbufs;	/* Number of log buffers, -1 to default */
-	int	logbufsize;	/* Size of log buffers, -1 to default */
-	char	fsname[MAXNAMELEN+1];	/* data device name */
-	char	rtname[MAXNAMELEN+1];	/* realtime device filename */
-	char	logname[MAXNAMELEN+1];	/* journal device filename */
-	char	mtpt[MAXNAMELEN+1];	/* filesystem mount point */
-	int	sunit;		/* stripe unit (BBs) */
-	int	swidth;		/* stripe width (BBs), multiple of sunit */
-	uchar_t iosizelog;	/* log2 of the preferred I/O size */
-	int	ihashsize;	/* inode hash table size (buckets) */
-};
-
-/*
- * XFS mount option flags -- args->flags1
- */
-#define	XFSMNT_ATTR2		0x00000001	/* allow ATTR2 EA format */
-#define	XFSMNT_WSYNC		0x00000002	/* safe mode nfs mount
-						 * compatible */
-#define	XFSMNT_INO64		0x00000004	/* move inode numbers up
-						 * past 2^32 */
-#define XFSMNT_UQUOTA		0x00000008	/* user quota accounting */
-#define XFSMNT_PQUOTA		0x00000010	/* IRIX prj quota accounting */
-#define XFSMNT_UQUOTAENF	0x00000020	/* user quota limit
-						 * enforcement */
-#define XFSMNT_PQUOTAENF	0x00000040	/* IRIX project quota limit
-						 * enforcement */
-#define XFSMNT_QUIET		0x00000080	/* don't report mount errors */
-#define XFSMNT_NOALIGN		0x00000200	/* don't allocate at
-						 * stripe boundaries*/
-#define XFSMNT_RETERR		0x00000400	/* return error to user */
-#define XFSMNT_NORECOVERY	0x00000800	/* no recovery, implies
-						 * read-only mount */
-#define XFSMNT_SHARED		0x00001000	/* shared XFS mount */
-#define XFSMNT_IOSIZE		0x00002000	/* optimize for I/O size */
-#define XFSMNT_OSYNCISOSYNC	0x00004000	/* o_sync is REALLY o_sync */
-						/* (osyncisdsync is default) */
-#define XFSMNT_NOATTR2		0x00008000	/* turn off ATTR2 EA format */
-#define XFSMNT_32BITINODES	0x00200000	/* restrict inodes to 32
-						 * bits of address space */
-#define XFSMNT_GQUOTA		0x00400000	/* group quota accounting */
-#define XFSMNT_GQUOTAENF	0x00800000	/* group quota limit
-						 * enforcement */
-#define XFSMNT_NOUUID		0x01000000	/* Ignore fs uuid */
-#define XFSMNT_DMAPI		0x02000000	/* enable dmapi/xdsm */
-#define XFSMNT_BARRIER		0x04000000	/* use write barriers */
-#define XFSMNT_IKEEP		0x08000000	/* inode cluster delete */
-#define XFSMNT_SWALLOC		0x10000000	/* turn on stripe width
-						 * allocation */
-#define XFSMNT_DIRSYNC		0x40000000	/* sync creat,link,unlink,rename
-						 * symlink,mkdir,rmdir,mknod */
-#define XFSMNT_FLAGS2		0x80000000	/* more flags set in flags2 */
-
-/*
- * XFS mount option flags -- args->flags2
- */
-#define XFSMNT2_COMPAT_IOSIZE	0x00000001	/* don't report large preferred
-						 * I/O size in stat(2) */
-#define XFSMNT2_FILESTREAMS	0x00000002	/* enable the filestreams
-						 * allocator */
-
-#endif	/* __XFS_CLNT_H__ */

fs/xfs/xfs_da_btree.c

@@ -1566,11 +1566,14 @@ xfs_da_grow_inode(xfs_da_args_t *args, xfs_dablk_t *new_blkno)
 	int nmap, error, w, count, c, got, i, mapi;
 	xfs_trans_t *tp;
 	xfs_mount_t *mp;
+	xfs_drfsbno_t	nblks;
 
 	dp = args->dp;
 	mp = dp->i_mount;
 	w = args->whichfork;
 	tp = args->trans;
+	nblks = dp->i_d.di_nblocks;
+
 	/*
 	 * For new directories adjust the file offset and block count.
 	 */
@@ -1647,6 +1650,8 @@ xfs_da_grow_inode(xfs_da_args_t *args, xfs_dablk_t *new_blkno)
 	}
 	if (mapp != &map)
 		kmem_free(mapp);
+	/* account for newly allocated blocks in reserved blocks total */
+	args->total -= dp->i_d.di_nblocks - nblks;
 	*new_blkno = (xfs_dablk_t)bno;
 	return 0;
 }

fs/xfs/xfs_da_btree.h

@@ -72,27 +72,7 @@ typedef struct xfs_da_intnode {
 typedef struct xfs_da_node_hdr xfs_da_node_hdr_t;
 typedef struct xfs_da_node_entry xfs_da_node_entry_t;
 
-#define XFS_DA_MAXHASH	((xfs_dahash_t)-1) /* largest valid hash value */
-
 #define	XFS_LBSIZE(mp)	(mp)->m_sb.sb_blocksize
-#define	XFS_LBLOG(mp)	(mp)->m_sb.sb_blocklog
-
-#define	XFS_DA_MAKE_BNOENTRY(mp,bno,entry)	\
-	(((bno) << (mp)->m_dircook_elog) | (entry))
-#define	XFS_DA_MAKE_COOKIE(mp,bno,entry,hash)	\
-	(((xfs_off_t)XFS_DA_MAKE_BNOENTRY(mp, bno, entry) << 32) | (hash))
-#define	XFS_DA_COOKIE_HASH(mp,cookie)		((xfs_dahash_t)cookie)
-#define	XFS_DA_COOKIE_BNO(mp,cookie)		\
-	((((xfs_off_t)(cookie) >> 31) == -1LL ? \
-		(xfs_dablk_t)0 : \
-		(xfs_dablk_t)((xfs_off_t)(cookie) >> \
-				((mp)->m_dircook_elog + 32))))
-#define	XFS_DA_COOKIE_ENTRY(mp,cookie)		\
-	((((xfs_off_t)(cookie) >> 31) == -1LL ?	\
-		(xfs_dablk_t)0 : \
-		(xfs_dablk_t)(((xfs_off_t)(cookie) >> 32) & \
-				((1 << (mp)->m_dircook_elog) - 1))))
-
 
 /*========================================================================
  * Btree searching and modification structure definitions.
@@ -226,9 +206,8 @@ struct xfs_nameops {
 };
 
 
-#ifdef __KERNEL__
 /*========================================================================
- * Function prototypes for the kernel.
+ * Function prototypes.
  *========================================================================*/
 
 /*
@@ -289,6 +268,5 @@ xfs_daddr_t xfs_da_blkno(xfs_dabuf_t *dabuf);
 
 extern struct kmem_zone *xfs_da_state_zone;
 extern struct kmem_zone *xfs_dabuf_zone;
-#endif	/* __KERNEL__ */
 
 #endif	/* __XFS_DA_BTREE_H__ */

fs/xfs/xfs_dinode.h

@@ -78,8 +78,7 @@ typedef struct xfs_dinode
 	xfs_dinode_core_t	di_core;
 	/*
 	 * In adding anything between the core and the union, be
-	 * sure to update the macros like XFS_LITINO below and
-	 * XFS_BMAP_RBLOCK_DSIZE in xfs_bmap_btree.h.
+	 * sure to update the macros like XFS_LITINO below.
 	 */
 	__be32			di_next_unlinked;/* agi unlinked list ptr */
 	union {
@@ -166,7 +165,7 @@ typedef enum xfs_dinode_fmt
  */
 #define	XFS_LITINO(mp)	((mp)->m_litino)
 #define	XFS_BROOT_SIZE_ADJ	\
-	(sizeof(xfs_bmbt_block_t) - sizeof(xfs_bmdr_block_t))
+	(XFS_BTREE_LBLOCK_LEN - sizeof(xfs_bmdr_block_t))
 
 /*
  * Inode data & attribute fork sizes, per inode.

fs/xfs/xfs_dir2.c

@@ -525,11 +525,13 @@ xfs_dir2_grow_inode(
 	xfs_mount_t	*mp;
 	int		nmap;		/* number of bmap entries */
 	xfs_trans_t	*tp;
+	xfs_drfsbno_t	nblks;
 
 	xfs_dir2_trace_args_s("grow_inode", args, space);
 	dp = args->dp;
 	tp = args->trans;
 	mp = dp->i_mount;
+	nblks = dp->i_d.di_nblocks;
 	/*
 	 * Set lowest possible block in the space requested.
 	 */
@@ -622,7 +624,11 @@ xfs_dir2_grow_inode(
 	 */
 	if (mapp != &map)
 		kmem_free(mapp);
+
+	/* account for newly allocated blocks in reserved blocks total */
+	args->total -= dp->i_d.di_nblocks - nblks;
 	*dbp = xfs_dir2_da_to_db(mp, (xfs_dablk_t)bno);
+
 	/*
 	 * Update file's size if this is the data space and it grew.
 	 */

fs/xfs/xfs_dmops.c

@@ -25,7 +25,6 @@
 #include "xfs_inum.h"
 #include "xfs_ag.h"
 #include "xfs_mount.h"
-#include "xfs_clnt.h"
 
 
 static struct xfs_dmops xfs_dmcore_stub = {
@@ -38,9 +37,9 @@ static struct xfs_dmops xfs_dmcore_stub = {
 };
 
 int
-xfs_dmops_get(struct xfs_mount *mp, struct xfs_mount_args *args)
+xfs_dmops_get(struct xfs_mount *mp)
 {
-	if (args->flags & XFSMNT_DMAPI) {
+	if (mp->m_flags & XFS_MOUNT_DMAPI) {
 		cmn_err(CE_WARN,
 			"XFS: dmapi support not available in this kernel.");
 		return EINVAL;

fs/xfs/xfs_extfree_item.c

@@ -108,19 +108,16 @@ xfs_efi_item_pin(xfs_efi_log_item_t *efip)
 STATIC void
 xfs_efi_item_unpin(xfs_efi_log_item_t *efip, int stale)
 {
-	xfs_mount_t	*mp;
+	struct xfs_ail		*ailp = efip->efi_item.li_ailp;
 
-	mp = efip->efi_item.li_mountp;
-	spin_lock(&mp->m_ail_lock);
+	spin_lock(&ailp->xa_lock);
 	if (efip->efi_flags & XFS_EFI_CANCELED) {
-		/*
-		 * xfs_trans_delete_ail() drops the AIL lock.
-		 */
-		xfs_trans_delete_ail(mp, (xfs_log_item_t *)efip);
+		/* xfs_trans_ail_delete() drops the AIL lock. */
+		xfs_trans_ail_delete(ailp, (xfs_log_item_t *)efip);
 		xfs_efi_item_free(efip);
 	} else {
 		efip->efi_flags |= XFS_EFI_COMMITTED;
-		spin_unlock(&mp->m_ail_lock);
+		spin_unlock(&ailp->xa_lock);
 	}
 }
 
@@ -134,26 +131,23 @@ xfs_efi_item_unpin(xfs_efi_log_item_t *efip, int stale)
 STATIC void
 xfs_efi_item_unpin_remove(xfs_efi_log_item_t *efip, xfs_trans_t *tp)
 {
-	xfs_mount_t	*mp;
+	struct xfs_ail		*ailp = efip->efi_item.li_ailp;
 	xfs_log_item_desc_t	*lidp;
 
-	mp = efip->efi_item.li_mountp;
-	spin_lock(&mp->m_ail_lock);
+	spin_lock(&ailp->xa_lock);
 	if (efip->efi_flags & XFS_EFI_CANCELED) {
 		/*
 		 * free the xaction descriptor pointing to this item
 		 */
 		lidp = xfs_trans_find_item(tp, (xfs_log_item_t *) efip);
 		xfs_trans_free_item(tp, lidp);
-		/*
-		 * pull the item off the AIL.
-		 * xfs_trans_delete_ail() drops the AIL lock.
-		 */
-		xfs_trans_delete_ail(mp, (xfs_log_item_t *)efip);
+
+		/* xfs_trans_ail_delete() drops the AIL lock. */
+		xfs_trans_ail_delete(ailp, (xfs_log_item_t *)efip);
 		xfs_efi_item_free(efip);
 	} else {
 		efip->efi_flags |= XFS_EFI_COMMITTED;
-		spin_unlock(&mp->m_ail_lock);
+		spin_unlock(&ailp->xa_lock);
 	}
 }
 
@@ -268,6 +262,7 @@ xfs_efi_init(xfs_mount_t	*mp,
 	efip->efi_item.li_type = XFS_LI_EFI;
 	efip->efi_item.li_ops = &xfs_efi_item_ops;
 	efip->efi_item.li_mountp = mp;
+	efip->efi_item.li_ailp = mp->m_ail;
 	efip->efi_format.efi_nextents = nextents;
 	efip->efi_format.efi_id = (__psint_t)(void*)efip;
 
@@ -345,25 +340,22 @@ void
 xfs_efi_release(xfs_efi_log_item_t	*efip,
 		uint			nextents)
 {
-	xfs_mount_t	*mp;
-	int		extents_left;
+	struct xfs_ail		*ailp = efip->efi_item.li_ailp;
+	int			extents_left;
 
-	mp = efip->efi_item.li_mountp;
 	ASSERT(efip->efi_next_extent > 0);
 	ASSERT(efip->efi_flags & XFS_EFI_COMMITTED);
 
-	spin_lock(&mp->m_ail_lock);
+	spin_lock(&ailp->xa_lock);
 	ASSERT(efip->efi_next_extent >= nextents);
 	efip->efi_next_extent -= nextents;
 	extents_left = efip->efi_next_extent;
 	if (extents_left == 0) {
-		/*
-		 * xfs_trans_delete_ail() drops the AIL lock.
-		 */
-		xfs_trans_delete_ail(mp, (xfs_log_item_t *)efip);
+		/* xfs_trans_ail_delete() drops the AIL lock. */
+		xfs_trans_ail_delete(ailp, (xfs_log_item_t *)efip);
 		xfs_efi_item_free(efip);
 	} else {
-		spin_unlock(&mp->m_ail_lock);
+		spin_unlock(&ailp->xa_lock);
 	}
 }
 
@@ -565,6 +557,7 @@ xfs_efd_init(xfs_mount_t	*mp,
 	efdp->efd_item.li_type = XFS_LI_EFD;
 	efdp->efd_item.li_ops = &xfs_efd_item_ops;
 	efdp->efd_item.li_mountp = mp;
+	efdp->efd_item.li_ailp = mp->m_ail;
 	efdp->efd_efip = efip;
 	efdp->efd_format.efd_nextents = nextents;
 	efdp->efd_format.efd_efi_id = efip->efi_format.efi_id;

fs/xfs/xfs_fsops.c

@@ -126,7 +126,7 @@ xfs_growfs_data_private(
 	xfs_extlen_t		agsize;
 	xfs_extlen_t		tmpsize;
 	xfs_alloc_rec_t		*arec;
-	xfs_btree_sblock_t	*block;
+	struct xfs_btree_block	*block;
 	xfs_buf_t		*bp;
 	int			bucket;
 	int			dpct;
@@ -251,14 +251,14 @@ xfs_growfs_data_private(
 		bp = xfs_buf_get(mp->m_ddev_targp,
 			XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
 			BTOBB(mp->m_sb.sb_blocksize), 0);
-		block = XFS_BUF_TO_SBLOCK(bp);
+		block = XFS_BUF_TO_BLOCK(bp);
 		memset(block, 0, mp->m_sb.sb_blocksize);
 		block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
 		block->bb_level = 0;
 		block->bb_numrecs = cpu_to_be16(1);
-		block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-		block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
-		arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
+		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
+		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
+		arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
 		arec->ar_blockcount = cpu_to_be32(
 			agsize - be32_to_cpu(arec->ar_startblock));
@@ -272,14 +272,14 @@ xfs_growfs_data_private(
 		bp = xfs_buf_get(mp->m_ddev_targp,
 			XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
 			BTOBB(mp->m_sb.sb_blocksize), 0);
-		block = XFS_BUF_TO_SBLOCK(bp);
+		block = XFS_BUF_TO_BLOCK(bp);
 		memset(block, 0, mp->m_sb.sb_blocksize);
 		block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
 		block->bb_level = 0;
 		block->bb_numrecs = cpu_to_be16(1);
-		block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-		block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
-		arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
+		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
+		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
+		arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
 		arec->ar_blockcount = cpu_to_be32(
 			agsize - be32_to_cpu(arec->ar_startblock));
@@ -294,13 +294,13 @@ xfs_growfs_data_private(
 		bp = xfs_buf_get(mp->m_ddev_targp,
 			XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
 			BTOBB(mp->m_sb.sb_blocksize), 0);
-		block = XFS_BUF_TO_SBLOCK(bp);
+		block = XFS_BUF_TO_BLOCK(bp);
 		memset(block, 0, mp->m_sb.sb_blocksize);
 		block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
 		block->bb_level = 0;
 		block->bb_numrecs = 0;
-		block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-		block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
+		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
+		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
 		error = xfs_bwrite(mp, bp);
 		if (error) {
 			goto error0;

fs/xfs/xfs_ialloc.c

@@ -118,6 +118,102 @@ xfs_ialloc_cluster_alignment(
 	return 1;
 }
 
+/*
+ * Lookup the record equal to ino in the btree given by cur.
+ */
+STATIC int				/* error */
+xfs_inobt_lookup_eq(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agino_t		ino,	/* starting inode of chunk */
+	__int32_t		fcnt,	/* free inode count */
+	xfs_inofree_t		free,	/* free inode mask */
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.i.ir_startino = ino;
+	cur->bc_rec.i.ir_freecount = fcnt;
+	cur->bc_rec.i.ir_free = free;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_EQ, stat);
+}
+
+/*
+ * Lookup the first record greater than or equal to ino
+ * in the btree given by cur.
+ */
+int					/* error */
+xfs_inobt_lookup_ge(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agino_t		ino,	/* starting inode of chunk */
+	__int32_t		fcnt,	/* free inode count */
+	xfs_inofree_t		free,	/* free inode mask */
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.i.ir_startino = ino;
+	cur->bc_rec.i.ir_freecount = fcnt;
+	cur->bc_rec.i.ir_free = free;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_GE, stat);
+}
+
+/*
+ * Lookup the first record less than or equal to ino
+ * in the btree given by cur.
+ */
+int					/* error */
+xfs_inobt_lookup_le(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agino_t		ino,	/* starting inode of chunk */
+	__int32_t		fcnt,	/* free inode count */
+	xfs_inofree_t		free,	/* free inode mask */
+	int			*stat)	/* success/failure */
+{
+	cur->bc_rec.i.ir_startino = ino;
+	cur->bc_rec.i.ir_freecount = fcnt;
+	cur->bc_rec.i.ir_free = free;
+	return xfs_btree_lookup(cur, XFS_LOOKUP_LE, stat);
+}
+
+/*
+ * Update the record referred to by cur to the value given
+ * by [ino, fcnt, free].
+ * This either works (return 0) or gets an EFSCORRUPTED error.
+ */
+STATIC int				/* error */
+xfs_inobt_update(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agino_t		ino,	/* starting inode of chunk */
+	__int32_t		fcnt,	/* free inode count */
+	xfs_inofree_t		free)	/* free inode mask */
+{
+	union xfs_btree_rec	rec;
+
+	rec.inobt.ir_startino = cpu_to_be32(ino);
+	rec.inobt.ir_freecount = cpu_to_be32(fcnt);
+	rec.inobt.ir_free = cpu_to_be64(free);
+	return xfs_btree_update(cur, &rec);
+}
+
+/*
+ * Get the data from the pointed-to record.
+ */
+int					/* error */
+xfs_inobt_get_rec(
+	struct xfs_btree_cur	*cur,	/* btree cursor */
+	xfs_agino_t		*ino,	/* output: starting inode of chunk */
+	__int32_t		*fcnt,	/* output: number of free inodes */
+	xfs_inofree_t		*free,	/* output: free inode mask */
+	int			*stat)	/* output: success/failure */
+{
+	union xfs_btree_rec	*rec;
+	int			error;
+
+	error = xfs_btree_get_rec(cur, &rec, stat);
+	if (!error && *stat == 1) {
+		*ino = be32_to_cpu(rec->inobt.ir_startino);
+		*fcnt = be32_to_cpu(rec->inobt.ir_freecount);
+		*free = be64_to_cpu(rec->inobt.ir_free);
+	}
+	return error;
+}
+
 /*
  * Allocate new inodes in the allocation group specified by agbp.
  * Return 0 for success, else error code.
@@ -335,8 +431,7 @@ xfs_ialloc_ag_alloc(
 	/*
 	 * Insert records describing the new inode chunk into the btree.
 	 */
-	cur = xfs_btree_init_cursor(args.mp, tp, agbp, agno,
-			XFS_BTNUM_INO, (xfs_inode_t *)0, 0);
+	cur = xfs_inobt_init_cursor(args.mp, tp, agbp, agno);
 	for (thisino = newino;
 	     thisino < newino + newlen;
 	     thisino += XFS_INODES_PER_CHUNK) {
@@ -346,7 +441,7 @@ xfs_ialloc_ag_alloc(
 			return error;
 		}
 		ASSERT(i == 0);
-		if ((error = xfs_inobt_insert(cur, &i))) {
+		if ((error = xfs_btree_insert(cur, &i))) {
 			xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
 			return error;
 		}
@@ -676,8 +771,7 @@ nextag:
 	 */
 	agno = tagno;
 	*IO_agbp = NULL;
-	cur = xfs_btree_init_cursor(mp, tp, agbp, be32_to_cpu(agi->agi_seqno),
-				    XFS_BTNUM_INO, (xfs_inode_t *)0, 0);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, be32_to_cpu(agi->agi_seqno));
 	/*
 	 * If pagino is 0 (this is the root inode allocation) use newino.
 	 * This must work because we've just allocated some.
@@ -697,7 +791,7 @@ nextag:
 				goto error0;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 			freecount += rec.ir_freecount;
-			if ((error = xfs_inobt_increment(cur, 0, &i)))
+			if ((error = xfs_btree_increment(cur, 0, &i)))
 				goto error0;
 		} while (i == 1);
 
@@ -741,7 +835,7 @@ nextag:
 			/*
 			 * Search left with tcur, back up 1 record.
 			 */
-			if ((error = xfs_inobt_decrement(tcur, 0, &i)))
+			if ((error = xfs_btree_decrement(tcur, 0, &i)))
 				goto error1;
 			doneleft = !i;
 			if (!doneleft) {
@@ -755,7 +849,7 @@ nextag:
 			/*
 			 * Search right with cur, go forward 1 record.
 			 */
-			if ((error = xfs_inobt_increment(cur, 0, &i)))
+			if ((error = xfs_btree_increment(cur, 0, &i)))
 				goto error1;
 			doneright = !i;
 			if (!doneright) {
@@ -817,7 +911,7 @@ nextag:
 				 * further left.
 				 */
 				if (useleft) {
-					if ((error = xfs_inobt_decrement(tcur, 0,
+					if ((error = xfs_btree_decrement(tcur, 0,
 							&i)))
 						goto error1;
 					doneleft = !i;
@@ -837,7 +931,7 @@ nextag:
 				 * further right.
 				 */
 				else {
-					if ((error = xfs_inobt_increment(cur, 0,
+					if ((error = xfs_btree_increment(cur, 0,
 							&i)))
 						goto error1;
 					doneright = !i;
@@ -892,7 +986,7 @@ nextag:
 				XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 				if (rec.ir_freecount > 0)
 					break;
-				if ((error = xfs_inobt_increment(cur, 0, &i)))
+				if ((error = xfs_btree_increment(cur, 0, &i)))
 					goto error0;
 				XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 			}
@@ -926,7 +1020,7 @@ nextag:
 				goto error0;
 			XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
 			freecount += rec.ir_freecount;
-			if ((error = xfs_inobt_increment(cur, 0, &i)))
+			if ((error = xfs_btree_increment(cur, 0, &i)))
 				goto error0;
 		} while (i == 1);
 		ASSERT(freecount == be32_to_cpu(agi->agi_freecount) ||
@@ -1022,8 +1116,7 @@ xfs_difree(
 	/*
 	 * Initialize the cursor.
 	 */
-	cur = xfs_btree_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_INO,
-		(xfs_inode_t *)0, 0);
+	cur = xfs_inobt_init_cursor(mp, tp, agbp, agno);
 #ifdef DEBUG
 	if (cur->bc_nlevels == 1) {
 		int freecount = 0;
@@ -1036,7 +1129,7 @@ xfs_difree(
 				goto error0;
 			if (i) {
 				freecount += rec.ir_freecount;
-				if ((error = xfs_inobt_increment(cur, 0, &i)))
+				if ((error = xfs_btree_increment(cur, 0, &i)))
 					goto error0;
 			}
 		} while (i == 1);
@@ -1098,8 +1191,8 @@ xfs_difree(
 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_ICOUNT, -ilen);
 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IFREE, -(ilen - 1));
 
-		if ((error = xfs_inobt_delete(cur, &i))) {
-			cmn_err(CE_WARN, "xfs_difree: xfs_inobt_delete returned an error %d on %s.\n",
+		if ((error = xfs_btree_delete(cur, &i))) {
+			cmn_err(CE_WARN, "xfs_difree: xfs_btree_delete returned an error %d on %s.\n",
 				error, mp->m_fsname);
 			goto error0;
 		}
@@ -1141,7 +1234,7 @@ xfs_difree(
 				goto error0;
 			if (i) {
 				freecount += rec.ir_freecount;
-				if ((error = xfs_inobt_increment(cur, 0, &i)))
+				if ((error = xfs_btree_increment(cur, 0, &i)))
 					goto error0;
 			}
 		} while (i == 1);
@@ -1259,8 +1352,7 @@ xfs_dilocate(
 #endif /* DEBUG */
 			return error;
 		}
-		cur = xfs_btree_init_cursor(mp, tp, agbp, agno, XFS_BTNUM_INO,
-			(xfs_inode_t *)0, 0);
+		cur = xfs_inobt_init_cursor(mp, tp, agbp, agno);
 		if ((error = xfs_inobt_lookup_le(cur, agino, 0, 0, &i))) {
 #ifdef DEBUG
 			xfs_fs_cmn_err(CE_ALERT, mp, "xfs_dilocate: "

fs/xfs/xfs_ialloc.h

@@ -56,7 +56,6 @@ static inline int xfs_ialloc_find_free(xfs_inofree_t *fp)
 }
 
 
-#ifdef __KERNEL__
 /*
  * Allocate an inode on disk.
  * Mode is used to tell whether the new inode will need space, and whether
@@ -154,6 +153,24 @@ xfs_ialloc_pagi_init(
 	struct xfs_trans *tp,		/* transaction pointer */
         xfs_agnumber_t  agno);		/* allocation group number */
 
-#endif	/* __KERNEL__ */
+/*
+ * Lookup the first record greater than or equal to ino
+ * in the btree given by cur.
+ */
+int xfs_inobt_lookup_ge(struct xfs_btree_cur *cur, xfs_agino_t ino,
+		__int32_t fcnt,	xfs_inofree_t free, int *stat);
+
+/*
+ * Lookup the first record less than or equal to ino
+ * in the btree given by cur.
+ */
+int xfs_inobt_lookup_le(struct xfs_btree_cur *cur, xfs_agino_t ino,
+		__int32_t fcnt,	xfs_inofree_t free, int *stat);
+
+/*
+ * Get the data from the pointed-to record.
+ */
+extern int xfs_inobt_get_rec(struct xfs_btree_cur *cur, xfs_agino_t *ino,
+			     __int32_t *fcnt, xfs_inofree_t *free, int *stat);
 
 #endif	/* __XFS_IALLOC_H__ */

fs/xfs/xfs_ialloc_btree.c

@@ -35,2044 +35,349 @@
 #include "xfs_dinode.h"
 #include "xfs_inode.h"
 #include "xfs_btree.h"
+#include "xfs_btree_trace.h"
 #include "xfs_ialloc.h"
 #include "xfs_alloc.h"
 #include "xfs_error.h"
 
-STATIC void xfs_inobt_log_block(xfs_trans_t *, xfs_buf_t *, int);
-STATIC void xfs_inobt_log_keys(xfs_btree_cur_t *, xfs_buf_t *, int, int);
-STATIC void xfs_inobt_log_ptrs(xfs_btree_cur_t *, xfs_buf_t *, int, int);
-STATIC void xfs_inobt_log_recs(xfs_btree_cur_t *, xfs_buf_t *, int, int);
-STATIC int xfs_inobt_lshift(xfs_btree_cur_t *, int, int *);
-STATIC int xfs_inobt_newroot(xfs_btree_cur_t *, int *);
-STATIC int xfs_inobt_rshift(xfs_btree_cur_t *, int, int *);
-STATIC int xfs_inobt_split(xfs_btree_cur_t *, int, xfs_agblock_t *,
-		xfs_inobt_key_t *, xfs_btree_cur_t **, int *);
-STATIC int xfs_inobt_updkey(xfs_btree_cur_t *, xfs_inobt_key_t *, int);
 
-/*
- * Single level of the xfs_inobt_delete record deletion routine.
- * Delete record pointed to by cur/level.
- * Remove the record from its block then rebalance the tree.
- * Return 0 for error, 1 for done, 2 to go on to the next level.
- */
-STATIC int				/* error */
-xfs_inobt_delrec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level removing record from */
-	int			*stat)	/* fail/done/go-on */
+STATIC int
+xfs_inobt_get_minrecs(
+	struct xfs_btree_cur	*cur,
+	int			level)
 {
-	xfs_buf_t		*agbp;	/* buffer for a.g. inode header */
-	xfs_mount_t		*mp;	/* mount structure */
-	xfs_agi_t		*agi;	/* allocation group inode header */
-	xfs_inobt_block_t	*block;	/* btree block record/key lives in */
-	xfs_agblock_t		bno;	/* btree block number */
-	xfs_buf_t		*bp;	/* buffer for block */
-	int			error;	/* error return value */
-	int			i;	/* loop index */
-	xfs_inobt_key_t		key;	/* kp points here if block is level 0 */
-	xfs_inobt_key_t		*kp = NULL;	/* pointer to btree keys */
-	xfs_agblock_t		lbno;	/* left block's block number */
-	xfs_buf_t		*lbp;	/* left block's buffer pointer */
-	xfs_inobt_block_t	*left;	/* left btree block */
-	xfs_inobt_key_t		*lkp;	/* left block key pointer */
-	xfs_inobt_ptr_t		*lpp;	/* left block address pointer */
-	int			lrecs = 0;	/* number of records in left block */
-	xfs_inobt_rec_t		*lrp;	/* left block record pointer */
-	xfs_inobt_ptr_t		*pp = NULL;	/* pointer to btree addresses */
-	int			ptr;	/* index in btree block for this rec */
-	xfs_agblock_t		rbno;	/* right block's block number */
-	xfs_buf_t		*rbp;	/* right block's buffer pointer */
-	xfs_inobt_block_t	*right;	/* right btree block */
-	xfs_inobt_key_t		*rkp;	/* right block key pointer */
-	xfs_inobt_rec_t		*rp;	/* pointer to btree records */
-	xfs_inobt_ptr_t		*rpp;	/* right block address pointer */
-	int			rrecs = 0;	/* number of records in right block */
-	int			numrecs;
-	xfs_inobt_rec_t		*rrp;	/* right block record pointer */
-	xfs_btree_cur_t		*tcur;	/* temporary btree cursor */
-
-	mp = cur->bc_mp;
-
-	/*
-	 * Get the index of the entry being deleted, check for nothing there.
-	 */
-	ptr = cur->bc_ptrs[level];
-	if (ptr == 0) {
-		*stat = 0;
-		return 0;
-	}
-
-	/*
-	 * Get the buffer & block containing the record or key/ptr.
-	 */
-	bp = cur->bc_bufs[level];
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
-		return error;
-#endif
-	/*
-	 * Fail if we're off the end of the block.
-	 */
+	return cur->bc_mp->m_inobt_mnr[level != 0];
+}
 
-	numrecs = be16_to_cpu(block->bb_numrecs);
-	if (ptr > numrecs) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * It's a nonleaf.  Excise the key and ptr being deleted, by
-	 * sliding the entries past them down one.
-	 * Log the changed areas of the block.
-	 */
-	if (level > 0) {
-		kp = XFS_INOBT_KEY_ADDR(block, 1, cur);
-		pp = XFS_INOBT_PTR_ADDR(block, 1, cur);
-#ifdef DEBUG
-		for (i = ptr; i < numrecs; i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(pp[i]), level)))
-				return error;
-		}
-#endif
-		if (ptr < numrecs) {
-			memmove(&kp[ptr - 1], &kp[ptr],
-				(numrecs - ptr) * sizeof(*kp));
-			memmove(&pp[ptr - 1], &pp[ptr],
-				(numrecs - ptr) * sizeof(*kp));
-			xfs_inobt_log_keys(cur, bp, ptr, numrecs - 1);
-			xfs_inobt_log_ptrs(cur, bp, ptr, numrecs - 1);
-		}
-	}
-	/*
-	 * It's a leaf.  Excise the record being deleted, by sliding the
-	 * entries past it down one.  Log the changed areas of the block.
-	 */
-	else {
-		rp = XFS_INOBT_REC_ADDR(block, 1, cur);
-		if (ptr < numrecs) {
-			memmove(&rp[ptr - 1], &rp[ptr],
-				(numrecs - ptr) * sizeof(*rp));
-			xfs_inobt_log_recs(cur, bp, ptr, numrecs - 1);
-		}
-		/*
-		 * If it's the first record in the block, we'll need a key
-		 * structure to pass up to the next level (updkey).
-		 */
-		if (ptr == 1) {
-			key.ir_startino = rp->ir_startino;
-			kp = &key;
-		}
-	}
-	/*
-	 * Decrement and log the number of entries in the block.
-	 */
-	numrecs--;
-	block->bb_numrecs = cpu_to_be16(numrecs);
-	xfs_inobt_log_block(cur->bc_tp, bp, XFS_BB_NUMRECS);
-	/*
-	 * Is this the root level?  If so, we're almost done.
-	 */
-	if (level == cur->bc_nlevels - 1) {
-		/*
-		 * If this is the root level,
-		 * and there's only one entry left,
-		 * and it's NOT the leaf level,
-		 * then we can get rid of this level.
-		 */
-		if (numrecs == 1 && level > 0) {
-			agbp = cur->bc_private.a.agbp;
-			agi = XFS_BUF_TO_AGI(agbp);
-			/*
-			 * pp is still set to the first pointer in the block.
-			 * Make it the new root of the btree.
-			 */
-			bno = be32_to_cpu(agi->agi_root);
-			agi->agi_root = *pp;
-			be32_add_cpu(&agi->agi_level, -1);
-			/*
-			 * Free the block.
-			 */
-			if ((error = xfs_free_extent(cur->bc_tp,
-				XFS_AGB_TO_FSB(mp, cur->bc_private.a.agno, bno), 1)))
-				return error;
-			xfs_trans_binval(cur->bc_tp, bp);
-			xfs_ialloc_log_agi(cur->bc_tp, agbp,
-				XFS_AGI_ROOT | XFS_AGI_LEVEL);
-			/*
-			 * Update the cursor so there's one fewer level.
-			 */
-			cur->bc_bufs[level] = NULL;
-			cur->bc_nlevels--;
-		} else if (level > 0 &&
-			   (error = xfs_inobt_decrement(cur, level, &i)))
-			return error;
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * If we deleted the leftmost entry in the block, update the
-	 * key values above us in the tree.
-	 */
-	if (ptr == 1 && (error = xfs_inobt_updkey(cur, kp, level + 1)))
-		return error;
-	/*
-	 * If the number of records remaining in the block is at least
-	 * the minimum, we're done.
-	 */
-	if (numrecs >= XFS_INOBT_BLOCK_MINRECS(level, cur)) {
-		if (level > 0 &&
-		    (error = xfs_inobt_decrement(cur, level, &i)))
-			return error;
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * Otherwise, we have to move some records around to keep the
-	 * tree balanced.  Look at the left and right sibling blocks to
-	 * see if we can re-balance by moving only one record.
-	 */
-	rbno = be32_to_cpu(block->bb_rightsib);
-	lbno = be32_to_cpu(block->bb_leftsib);
-	bno = NULLAGBLOCK;
-	ASSERT(rbno != NULLAGBLOCK || lbno != NULLAGBLOCK);
-	/*
-	 * Duplicate the cursor so our btree manipulations here won't
-	 * disrupt the next level up.
-	 */
-	if ((error = xfs_btree_dup_cursor(cur, &tcur)))
-		return error;
-	/*
-	 * If there's a right sibling, see if it's ok to shift an entry
-	 * out of it.
-	 */
-	if (rbno != NULLAGBLOCK) {
-		/*
-		 * Move the temp cursor to the last entry in the next block.
-		 * Actually any entry but the first would suffice.
-		 */
-		i = xfs_btree_lastrec(tcur, level);
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		if ((error = xfs_inobt_increment(tcur, level, &i)))
-			goto error0;
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		i = xfs_btree_lastrec(tcur, level);
-		XFS_WANT_CORRUPTED_GOTO(i == 1, error0);
-		/*
-		 * Grab a pointer to the block.
-		 */
-		rbp = tcur->bc_bufs[level];
-		right = XFS_BUF_TO_INOBT_BLOCK(rbp);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
-			goto error0;
-#endif
-		/*
-		 * Grab the current block number, for future use.
-		 */
-		bno = be32_to_cpu(right->bb_leftsib);
-		/*
-		 * If right block is full enough so that removing one entry
-		 * won't make it too empty, and left-shifting an entry out
-		 * of right to us works, we're done.
-		 */
-		if (be16_to_cpu(right->bb_numrecs) - 1 >=
-		     XFS_INOBT_BLOCK_MINRECS(level, cur)) {
-			if ((error = xfs_inobt_lshift(tcur, level, &i)))
-				goto error0;
-			if (i) {
-				ASSERT(be16_to_cpu(block->bb_numrecs) >=
-				       XFS_INOBT_BLOCK_MINRECS(level, cur));
-				xfs_btree_del_cursor(tcur,
-						     XFS_BTREE_NOERROR);
-				if (level > 0 &&
-				    (error = xfs_inobt_decrement(cur, level,
-						&i)))
-					return error;
-				*stat = 1;
-				return 0;
-			}
-		}
-		/*
-		 * Otherwise, grab the number of records in right for
-		 * future reference, and fix up the temp cursor to point
-		 * to our block again (last record).
-		 */
-		rrecs = be16_to_cpu(right->bb_numrecs);
-		if (lbno != NULLAGBLOCK) {
-			xfs_btree_firstrec(tcur, level);
-			if ((error = xfs_inobt_decrement(tcur, level, &i)))
-				goto error0;
-		}
-	}
-	/*
-	 * If there's a left sibling, see if it's ok to shift an entry
-	 * out of it.
-	 */
-	if (lbno != NULLAGBLOCK) {
-		/*
-		 * Move the temp cursor to the first entry in the
-		 * previous block.
-		 */
-		xfs_btree_firstrec(tcur, level);
-		if ((error = xfs_inobt_decrement(tcur, level, &i)))
-			goto error0;
-		xfs_btree_firstrec(tcur, level);
-		/*
-		 * Grab a pointer to the block.
-		 */
-		lbp = tcur->bc_bufs[level];
-		left = XFS_BUF_TO_INOBT_BLOCK(lbp);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-			goto error0;
-#endif
-		/*
-		 * Grab the current block number, for future use.
-		 */
-		bno = be32_to_cpu(left->bb_rightsib);
-		/*
-		 * If left block is full enough so that removing one entry
-		 * won't make it too empty, and right-shifting an entry out
-		 * of left to us works, we're done.
-		 */
-		if (be16_to_cpu(left->bb_numrecs) - 1 >=
-		     XFS_INOBT_BLOCK_MINRECS(level, cur)) {
-			if ((error = xfs_inobt_rshift(tcur, level, &i)))
-				goto error0;
-			if (i) {
-				ASSERT(be16_to_cpu(block->bb_numrecs) >=
-				       XFS_INOBT_BLOCK_MINRECS(level, cur));
-				xfs_btree_del_cursor(tcur,
-						     XFS_BTREE_NOERROR);
-				if (level == 0)
-					cur->bc_ptrs[0]++;
-				*stat = 1;
-				return 0;
-			}
-		}
-		/*
-		 * Otherwise, grab the number of records in right for
-		 * future reference.
-		 */
-		lrecs = be16_to_cpu(left->bb_numrecs);
-	}
-	/*
-	 * Delete the temp cursor, we're done with it.
-	 */
-	xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
-	/*
-	 * If here, we need to do a join to keep the tree balanced.
-	 */
-	ASSERT(bno != NULLAGBLOCK);
-	/*
-	 * See if we can join with the left neighbor block.
-	 */
-	if (lbno != NULLAGBLOCK &&
-	    lrecs + numrecs <= XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
-		/*
-		 * Set "right" to be the starting block,
-		 * "left" to be the left neighbor.
-		 */
-		rbno = bno;
-		right = block;
-		rrecs = be16_to_cpu(right->bb_numrecs);
-		rbp = bp;
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, lbno, 0, &lbp,
-				XFS_INO_BTREE_REF)))
-			return error;
-		left = XFS_BUF_TO_INOBT_BLOCK(lbp);
-		lrecs = be16_to_cpu(left->bb_numrecs);
-		if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-			return error;
-	}
-	/*
-	 * If that won't work, see if we can join with the right neighbor block.
-	 */
-	else if (rbno != NULLAGBLOCK &&
-		 rrecs + numrecs <= XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
-		/*
-		 * Set "left" to be the starting block,
-		 * "right" to be the right neighbor.
-		 */
-		lbno = bno;
-		left = block;
-		lrecs = be16_to_cpu(left->bb_numrecs);
-		lbp = bp;
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, rbno, 0, &rbp,
-				XFS_INO_BTREE_REF)))
-			return error;
-		right = XFS_BUF_TO_INOBT_BLOCK(rbp);
-		rrecs = be16_to_cpu(right->bb_numrecs);
-		if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
-			return error;
-	}
-	/*
-	 * Otherwise, we can't fix the imbalance.
-	 * Just return.  This is probably a logic error, but it's not fatal.
-	 */
-	else {
-		if (level > 0 && (error = xfs_inobt_decrement(cur, level, &i)))
-			return error;
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * We're now going to join "left" and "right" by moving all the stuff
-	 * in "right" to "left" and deleting "right".
-	 */
-	if (level > 0) {
-		/*
-		 * It's a non-leaf.  Move keys and pointers.
-		 */
-		lkp = XFS_INOBT_KEY_ADDR(left, lrecs + 1, cur);
-		lpp = XFS_INOBT_PTR_ADDR(left, lrecs + 1, cur);
-		rkp = XFS_INOBT_KEY_ADDR(right, 1, cur);
-		rpp = XFS_INOBT_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		for (i = 0; i < rrecs; i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(rpp[i]), level)))
-				return error;
-		}
-#endif
-		memcpy(lkp, rkp, rrecs * sizeof(*lkp));
-		memcpy(lpp, rpp, rrecs * sizeof(*lpp));
-		xfs_inobt_log_keys(cur, lbp, lrecs + 1, lrecs + rrecs);
-		xfs_inobt_log_ptrs(cur, lbp, lrecs + 1, lrecs + rrecs);
-	} else {
-		/*
-		 * It's a leaf.  Move records.
-		 */
-		lrp = XFS_INOBT_REC_ADDR(left, lrecs + 1, cur);
-		rrp = XFS_INOBT_REC_ADDR(right, 1, cur);
-		memcpy(lrp, rrp, rrecs * sizeof(*lrp));
-		xfs_inobt_log_recs(cur, lbp, lrecs + 1, lrecs + rrecs);
-	}
-	/*
-	 * If we joined with the left neighbor, set the buffer in the
-	 * cursor to the left block, and fix up the index.
-	 */
-	if (bp != lbp) {
-		xfs_btree_setbuf(cur, level, lbp);
-		cur->bc_ptrs[level] += lrecs;
-	}
-	/*
-	 * If we joined with the right neighbor and there's a level above
-	 * us, increment the cursor at that level.
-	 */
-	else if (level + 1 < cur->bc_nlevels &&
-		 (error = xfs_alloc_increment(cur, level + 1, &i)))
-		return error;
-	/*
-	 * Fix up the number of records in the surviving block.
-	 */
-	lrecs += rrecs;
-	left->bb_numrecs = cpu_to_be16(lrecs);
-	/*
-	 * Fix up the right block pointer in the surviving block, and log it.
-	 */
-	left->bb_rightsib = right->bb_rightsib;
-	xfs_inobt_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
-	/*
-	 * If there is a right sibling now, make it point to the
-	 * remaining block.
-	 */
-	if (be32_to_cpu(left->bb_rightsib) != NULLAGBLOCK) {
-		xfs_inobt_block_t	*rrblock;
-		xfs_buf_t		*rrbp;
+STATIC struct xfs_btree_cur *
+xfs_inobt_dup_cursor(
+	struct xfs_btree_cur	*cur)
+{
+	return xfs_inobt_init_cursor(cur->bc_mp, cur->bc_tp,
+			cur->bc_private.a.agbp, cur->bc_private.a.agno);
+}
 
-		if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-				cur->bc_private.a.agno, be32_to_cpu(left->bb_rightsib), 0,
-				&rrbp, XFS_INO_BTREE_REF)))
-			return error;
-		rrblock = XFS_BUF_TO_INOBT_BLOCK(rrbp);
-		if ((error = xfs_btree_check_sblock(cur, rrblock, level, rrbp)))
-			return error;
-		rrblock->bb_leftsib = cpu_to_be32(lbno);
-		xfs_inobt_log_block(cur->bc_tp, rrbp, XFS_BB_LEFTSIB);
-	}
-	/*
-	 * Free the deleting block.
-	 */
-	if ((error = xfs_free_extent(cur->bc_tp, XFS_AGB_TO_FSB(mp,
-				     cur->bc_private.a.agno, rbno), 1)))
-		return error;
-	xfs_trans_binval(cur->bc_tp, rbp);
-	/*
-	 * Readjust the ptr at this level if it's not a leaf, since it's
-	 * still pointing at the deletion point, which makes the cursor
-	 * inconsistent.  If this makes the ptr 0, the caller fixes it up.
-	 * We can't use decrement because it would change the next level up.
-	 */
-	if (level > 0)
-		cur->bc_ptrs[level]--;
-	/*
-	 * Return value means the next level up has something to do.
-	 */
-	*stat = 2;
-	return 0;
+STATIC void
+xfs_inobt_set_root(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*nptr,
+	int			inc)	/* level change */
+{
+	struct xfs_buf		*agbp = cur->bc_private.a.agbp;
+	struct xfs_agi		*agi = XFS_BUF_TO_AGI(agbp);
 
-error0:
-	xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
-	return error;
+	agi->agi_root = nptr->s;
+	be32_add_cpu(&agi->agi_level, inc);
+	xfs_ialloc_log_agi(cur->bc_tp, agbp, XFS_AGI_ROOT | XFS_AGI_LEVEL);
 }
 
-/*
- * Insert one record/level.  Return information to the caller
- * allowing the next level up to proceed if necessary.
- */
-STATIC int				/* error */
-xfs_inobt_insrec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to insert record at */
-	xfs_agblock_t		*bnop,	/* i/o: block number inserted */
-	xfs_inobt_rec_t		*recp,	/* i/o: record data inserted */
-	xfs_btree_cur_t		**curp,	/* output: new cursor replacing cur */
-	int			*stat)	/* success/failure */
+STATIC int
+xfs_inobt_alloc_block(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*start,
+	union xfs_btree_ptr	*new,
+	int			length,
+	int			*stat)
 {
-	xfs_inobt_block_t	*block;	/* btree block record/key lives in */
-	xfs_buf_t		*bp;	/* buffer for block */
-	int			error;	/* error return value */
-	int			i;	/* loop index */
-	xfs_inobt_key_t		key;	/* key value being inserted */
-	xfs_inobt_key_t		*kp=NULL;	/* pointer to btree keys */
-	xfs_agblock_t		nbno;	/* block number of allocated block */
-	xfs_btree_cur_t		*ncur;	/* new cursor to be used at next lvl */
-	xfs_inobt_key_t		nkey;	/* new key value, from split */
-	xfs_inobt_rec_t		nrec;	/* new record value, for caller */
-	int			numrecs;
-	int			optr;	/* old ptr value */
-	xfs_inobt_ptr_t		*pp;	/* pointer to btree addresses */
-	int			ptr;	/* index in btree block for this rec */
-	xfs_inobt_rec_t		*rp=NULL;	/* pointer to btree records */
+	xfs_alloc_arg_t		args;		/* block allocation args */
+	int			error;		/* error return value */
+	xfs_agblock_t		sbno = be32_to_cpu(start->s);
 
-	/*
-	 * GCC doesn't understand the (arguably complex) control flow in
-	 * this function and complains about uninitialized structure fields
-	 * without this.
-	 */
-	memset(&nrec, 0, sizeof(nrec));
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
 
-	/*
-	 * If we made it to the root level, allocate a new root block
-	 * and we're done.
-	 */
-	if (level >= cur->bc_nlevels) {
-		error = xfs_inobt_newroot(cur, &i);
-		*bnop = NULLAGBLOCK;
-		*stat = i;
+	memset(&args, 0, sizeof(args));
+	args.tp = cur->bc_tp;
+	args.mp = cur->bc_mp;
+	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno, sbno);
+	args.minlen = 1;
+	args.maxlen = 1;
+	args.prod = 1;
+	args.type = XFS_ALLOCTYPE_NEAR_BNO;
+
+	error = xfs_alloc_vextent(&args);
+	if (error) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 		return error;
 	}
-	/*
-	 * Make a key out of the record data to be inserted, and save it.
-	 */
-	key.ir_startino = recp->ir_startino;
-	optr = ptr = cur->bc_ptrs[level];
-	/*
-	 * If we're off the left edge, return failure.
-	 */
-	if (ptr == 0) {
+	if (args.fsbno == NULLFSBLOCK) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 		*stat = 0;
 		return 0;
 	}
-	/*
-	 * Get pointers to the btree buffer and block.
-	 */
-	bp = cur->bc_bufs[level];
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-	numrecs = be16_to_cpu(block->bb_numrecs);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
-		return error;
-	/*
-	 * Check that the new entry is being inserted in the right place.
-	 */
-	if (ptr <= numrecs) {
-		if (level == 0) {
-			rp = XFS_INOBT_REC_ADDR(block, ptr, cur);
-			xfs_btree_check_rec(cur->bc_btnum, recp, rp);
-		} else {
-			kp = XFS_INOBT_KEY_ADDR(block, ptr, cur);
-			xfs_btree_check_key(cur->bc_btnum, &key, kp);
-		}
-	}
-#endif
-	nbno = NULLAGBLOCK;
-	ncur = NULL;
-	/*
-	 * If the block is full, we can't insert the new entry until we
-	 * make the block un-full.
-	 */
-	if (numrecs == XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
-		/*
-		 * First, try shifting an entry to the right neighbor.
-		 */
-		if ((error = xfs_inobt_rshift(cur, level, &i)))
-			return error;
-		if (i) {
-			/* nothing */
-		}
-		/*
-		 * Next, try shifting an entry to the left neighbor.
-		 */
-		else {
-			if ((error = xfs_inobt_lshift(cur, level, &i)))
-				return error;
-			if (i) {
-				optr = ptr = cur->bc_ptrs[level];
-			} else {
-				/*
-				 * Next, try splitting the current block
-				 * in half. If this works we have to
-				 * re-set our variables because
-				 * we could be in a different block now.
-				 */
-				if ((error = xfs_inobt_split(cur, level, &nbno,
-						&nkey, &ncur, &i)))
-					return error;
-				if (i) {
-					bp = cur->bc_bufs[level];
-					block = XFS_BUF_TO_INOBT_BLOCK(bp);
-#ifdef DEBUG
-					if ((error = xfs_btree_check_sblock(cur,
-							block, level, bp)))
-						return error;
-#endif
-					ptr = cur->bc_ptrs[level];
-					nrec.ir_startino = nkey.ir_startino;
-				} else {
-					/*
-					 * Otherwise the insert fails.
-					 */
-					*stat = 0;
-					return 0;
-				}
-			}
-		}
-	}
-	/*
-	 * At this point we know there's room for our new entry in the block
-	 * we're pointing at.
-	 */
-	numrecs = be16_to_cpu(block->bb_numrecs);
-	if (level > 0) {
-		/*
-		 * It's a non-leaf entry.  Make a hole for the new data
-		 * in the key and ptr regions of the block.
-		 */
-		kp = XFS_INOBT_KEY_ADDR(block, 1, cur);
-		pp = XFS_INOBT_PTR_ADDR(block, 1, cur);
-#ifdef DEBUG
-		for (i = numrecs; i >= ptr; i--) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(pp[i - 1]), level)))
-				return error;
-		}
-#endif
-		memmove(&kp[ptr], &kp[ptr - 1],
-			(numrecs - ptr + 1) * sizeof(*kp));
-		memmove(&pp[ptr], &pp[ptr - 1],
-			(numrecs - ptr + 1) * sizeof(*pp));
-		/*
-		 * Now stuff the new data in, bump numrecs and log the new data.
-		 */
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sptr(cur, *bnop, level)))
-			return error;
-#endif
-		kp[ptr - 1] = key;
-		pp[ptr - 1] = cpu_to_be32(*bnop);
-		numrecs++;
-		block->bb_numrecs = cpu_to_be16(numrecs);
-		xfs_inobt_log_keys(cur, bp, ptr, numrecs);
-		xfs_inobt_log_ptrs(cur, bp, ptr, numrecs);
-	} else {
-		/*
-		 * It's a leaf entry.  Make a hole for the new record.
-		 */
-		rp = XFS_INOBT_REC_ADDR(block, 1, cur);
-		memmove(&rp[ptr], &rp[ptr - 1],
-			(numrecs - ptr + 1) * sizeof(*rp));
-		/*
-		 * Now stuff the new record in, bump numrecs
-		 * and log the new data.
-		 */
-		rp[ptr - 1] = *recp;
-		numrecs++;
-		block->bb_numrecs = cpu_to_be16(numrecs);
-		xfs_inobt_log_recs(cur, bp, ptr, numrecs);
-	}
-	/*
-	 * Log the new number of records in the btree header.
-	 */
-	xfs_inobt_log_block(cur->bc_tp, bp, XFS_BB_NUMRECS);
-#ifdef DEBUG
-	/*
-	 * Check that the key/record is in the right place, now.
-	 */
-	if (ptr < numrecs) {
-		if (level == 0)
-			xfs_btree_check_rec(cur->bc_btnum, rp + ptr - 1,
-				rp + ptr);
-		else
-			xfs_btree_check_key(cur->bc_btnum, kp + ptr - 1,
-				kp + ptr);
-	}
-#endif
-	/*
-	 * If we inserted at the start of a block, update the parents' keys.
-	 */
-	if (optr == 1 && (error = xfs_inobt_updkey(cur, &key, level + 1)))
-		return error;
-	/*
-	 * Return the new block number, if any.
-	 * If there is one, give back a record value and a cursor too.
-	 */
-	*bnop = nbno;
-	if (nbno != NULLAGBLOCK) {
-		*recp = nrec;
-		*curp = ncur;
-	}
+	ASSERT(args.len == 1);
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
+
+	new->s = cpu_to_be32(XFS_FSB_TO_AGBNO(args.mp, args.fsbno));
 	*stat = 1;
 	return 0;
 }
 
-/*
- * Log header fields from a btree block.
- */
-STATIC void
-xfs_inobt_log_block(
-	xfs_trans_t		*tp,	/* transaction pointer */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			fields)	/* mask of fields: XFS_BB_... */
+STATIC int
+xfs_inobt_free_block(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp)
 {
-	int			first;	/* first byte offset logged */
-	int			last;	/* last byte offset logged */
-	static const short	offsets[] = {	/* table of offsets */
-		offsetof(xfs_inobt_block_t, bb_magic),
-		offsetof(xfs_inobt_block_t, bb_level),
-		offsetof(xfs_inobt_block_t, bb_numrecs),
-		offsetof(xfs_inobt_block_t, bb_leftsib),
-		offsetof(xfs_inobt_block_t, bb_rightsib),
-		sizeof(xfs_inobt_block_t)
-	};
+	xfs_fsblock_t		fsbno;
+	int			error;
 
-	xfs_btree_offsets(fields, offsets, XFS_BB_NUM_BITS, &first, &last);
-	xfs_trans_log_buf(tp, bp, first, last);
+	fsbno = XFS_DADDR_TO_FSB(cur->bc_mp, XFS_BUF_ADDR(bp));
+	error = xfs_free_extent(cur->bc_tp, fsbno, 1);
+	if (error)
+		return error;
+
+	xfs_trans_binval(cur->bc_tp, bp);
+	return error;
 }
 
-/*
- * Log keys from a btree block (nonleaf).
- */
-STATIC void
-xfs_inobt_log_keys(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			kfirst,	/* index of first key to log */
-	int			klast)	/* index of last key to log */
+STATIC int
+xfs_inobt_get_maxrecs(
+	struct xfs_btree_cur	*cur,
+	int			level)
 {
-	xfs_inobt_block_t	*block;	/* btree block to log from */
-	int			first;	/* first byte offset logged */
-	xfs_inobt_key_t		*kp;	/* key pointer in btree block */
-	int			last;	/* last byte offset logged */
-
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-	kp = XFS_INOBT_KEY_ADDR(block, 1, cur);
-	first = (int)((xfs_caddr_t)&kp[kfirst - 1] - (xfs_caddr_t)block);
-	last = (int)(((xfs_caddr_t)&kp[klast] - 1) - (xfs_caddr_t)block);
-	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
+	return cur->bc_mp->m_inobt_mxr[level != 0];
 }
 
-/*
- * Log block pointer fields from a btree block (nonleaf).
- */
 STATIC void
-xfs_inobt_log_ptrs(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			pfirst,	/* index of first pointer to log */
-	int			plast)	/* index of last pointer to log */
+xfs_inobt_init_key_from_rec(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
 {
-	xfs_inobt_block_t	*block;	/* btree block to log from */
-	int			first;	/* first byte offset logged */
-	int			last;	/* last byte offset logged */
-	xfs_inobt_ptr_t		*pp;	/* block-pointer pointer in btree blk */
-
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-	pp = XFS_INOBT_PTR_ADDR(block, 1, cur);
-	first = (int)((xfs_caddr_t)&pp[pfirst - 1] - (xfs_caddr_t)block);
-	last = (int)(((xfs_caddr_t)&pp[plast] - 1) - (xfs_caddr_t)block);
-	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
+	key->inobt.ir_startino = rec->inobt.ir_startino;
 }
 
-/*
- * Log records from a btree block (leaf).
- */
 STATIC void
-xfs_inobt_log_recs(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_buf_t		*bp,	/* buffer containing btree block */
-	int			rfirst,	/* index of first record to log */
-	int			rlast)	/* index of last record to log */
+xfs_inobt_init_rec_from_key(
+	union xfs_btree_key	*key,
+	union xfs_btree_rec	*rec)
 {
-	xfs_inobt_block_t	*block;	/* btree block to log from */
-	int			first;	/* first byte offset logged */
-	int			last;	/* last byte offset logged */
-	xfs_inobt_rec_t		*rp;	/* record pointer for btree block */
+	rec->inobt.ir_startino = key->inobt.ir_startino;
+}
 
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-	rp = XFS_INOBT_REC_ADDR(block, 1, cur);
-	first = (int)((xfs_caddr_t)&rp[rfirst - 1] - (xfs_caddr_t)block);
-	last = (int)(((xfs_caddr_t)&rp[rlast] - 1) - (xfs_caddr_t)block);
-	xfs_trans_log_buf(cur->bc_tp, bp, first, last);
+STATIC void
+xfs_inobt_init_rec_from_cur(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec)
+{
+	rec->inobt.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
+	rec->inobt.ir_freecount = cpu_to_be32(cur->bc_rec.i.ir_freecount);
+	rec->inobt.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
 }
 
 /*
- * Lookup the record.  The cursor is made to point to it, based on dir.
- * Return 0 if can't find any such record, 1 for success.
+ * intial value of ptr for lookup
  */
-STATIC int				/* error */
-xfs_inobt_lookup(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_lookup_t		dir,	/* <=, ==, or >= */
-	int			*stat)	/* success/failure */
+STATIC void
+xfs_inobt_init_ptr_from_cur(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_ptr	*ptr)
 {
-	xfs_agblock_t		agbno;	/* a.g. relative btree block number */
-	xfs_agnumber_t		agno;	/* allocation group number */
-	xfs_inobt_block_t	*block=NULL;	/* current btree block */
-	__int64_t		diff;	/* difference for the current key */
-	int			error;	/* error return value */
-	int			keyno=0;	/* current key number */
-	int			level;	/* level in the btree */
-	xfs_mount_t		*mp;	/* file system mount point */
-
-	/*
-	 * Get the allocation group header, and the root block number.
-	 */
-	mp = cur->bc_mp;
-	{
-		xfs_agi_t	*agi;	/* a.g. inode header */
-
-		agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
-		agno = be32_to_cpu(agi->agi_seqno);
-		agbno = be32_to_cpu(agi->agi_root);
-	}
-	/*
-	 * Iterate over each level in the btree, starting at the root.
-	 * For each level above the leaves, find the key we need, based
-	 * on the lookup record, then follow the corresponding block
-	 * pointer down to the next level.
-	 */
-	for (level = cur->bc_nlevels - 1, diff = 1; level >= 0; level--) {
-		xfs_buf_t	*bp;	/* buffer pointer for btree block */
-		xfs_daddr_t	d;	/* disk address of btree block */
-
-		/*
-		 * Get the disk address we're looking for.
-		 */
-		d = XFS_AGB_TO_DADDR(mp, agno, agbno);
-		/*
-		 * If the old buffer at this level is for a different block,
-		 * throw it away, otherwise just use it.
-		 */
-		bp = cur->bc_bufs[level];
-		if (bp && XFS_BUF_ADDR(bp) != d)
-			bp = NULL;
-		if (!bp) {
-			/*
-			 * Need to get a new buffer.  Read it, then
-			 * set it in the cursor, releasing the old one.
-			 */
-			if ((error = xfs_btree_read_bufs(mp, cur->bc_tp,
-					agno, agbno, 0, &bp, XFS_INO_BTREE_REF)))
-				return error;
-			xfs_btree_setbuf(cur, level, bp);
-			/*
-			 * Point to the btree block, now that we have the buffer
-			 */
-			block = XFS_BUF_TO_INOBT_BLOCK(bp);
-			if ((error = xfs_btree_check_sblock(cur, block, level,
-					bp)))
-				return error;
-		} else
-			block = XFS_BUF_TO_INOBT_BLOCK(bp);
-		/*
-		 * If we already had a key match at a higher level, we know
-		 * we need to use the first entry in this block.
-		 */
-		if (diff == 0)
-			keyno = 1;
-		/*
-		 * Otherwise we need to search this block.  Do a binary search.
-		 */
-		else {
-			int		high;	/* high entry number */
-			xfs_inobt_key_t	*kkbase=NULL;/* base of keys in block */
-			xfs_inobt_rec_t	*krbase=NULL;/* base of records in block */
-			int		low;	/* low entry number */
+	struct xfs_agi		*agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
 
-			/*
-			 * Get a pointer to keys or records.
-			 */
-			if (level > 0)
-				kkbase = XFS_INOBT_KEY_ADDR(block, 1, cur);
-			else
-				krbase = XFS_INOBT_REC_ADDR(block, 1, cur);
-			/*
-			 * Set low and high entry numbers, 1-based.
-			 */
-			low = 1;
-			if (!(high = be16_to_cpu(block->bb_numrecs))) {
-				/*
-				 * If the block is empty, the tree must
-				 * be an empty leaf.
-				 */
-				ASSERT(level == 0 && cur->bc_nlevels == 1);
-				cur->bc_ptrs[0] = dir != XFS_LOOKUP_LE;
-				*stat = 0;
-				return 0;
-			}
-			/*
-			 * Binary search the block.
-			 */
-			while (low <= high) {
-				xfs_agino_t	startino;	/* key value */
-
-				/*
-				 * keyno is average of low and high.
-				 */
-				keyno = (low + high) >> 1;
-				/*
-				 * Get startino.
-				 */
-				if (level > 0) {
-					xfs_inobt_key_t	*kkp;
-
-					kkp = kkbase + keyno - 1;
-					startino = be32_to_cpu(kkp->ir_startino);
-				} else {
-					xfs_inobt_rec_t	*krp;
-
-					krp = krbase + keyno - 1;
-					startino = be32_to_cpu(krp->ir_startino);
-				}
-				/*
-				 * Compute difference to get next direction.
-				 */
-				diff = (__int64_t)
-					startino - cur->bc_rec.i.ir_startino;
-				/*
-				 * Less than, move right.
-				 */
-				if (diff < 0)
-					low = keyno + 1;
-				/*
-				 * Greater than, move left.
-				 */
-				else if (diff > 0)
-					high = keyno - 1;
-				/*
-				 * Equal, we're done.
-				 */
-				else
-					break;
-			}
-		}
-		/*
-		 * If there are more levels, set up for the next level
-		 * by getting the block number and filling in the cursor.
-		 */
-		if (level > 0) {
-			/*
-			 * If we moved left, need the previous key number,
-			 * unless there isn't one.
-			 */
-			if (diff > 0 && --keyno < 1)
-				keyno = 1;
-			agbno = be32_to_cpu(*XFS_INOBT_PTR_ADDR(block, keyno, cur));
-#ifdef DEBUG
-			if ((error = xfs_btree_check_sptr(cur, agbno, level)))
-				return error;
-#endif
-			cur->bc_ptrs[level] = keyno;
-		}
-	}
-	/*
-	 * Done with the search.
-	 * See if we need to adjust the results.
-	 */
-	if (dir != XFS_LOOKUP_LE && diff < 0) {
-		keyno++;
-		/*
-		 * If ge search and we went off the end of the block, but it's
-		 * not the last block, we're in the wrong block.
-		 */
-		if (dir == XFS_LOOKUP_GE &&
-		    keyno > be16_to_cpu(block->bb_numrecs) &&
-		    be32_to_cpu(block->bb_rightsib) != NULLAGBLOCK) {
-			int	i;
+	ASSERT(cur->bc_private.a.agno == be32_to_cpu(agi->agi_seqno));
 
-			cur->bc_ptrs[0] = keyno;
-			if ((error = xfs_inobt_increment(cur, 0, &i)))
-				return error;
-			ASSERT(i == 1);
-			*stat = 1;
-			return 0;
-		}
-	}
-	else if (dir == XFS_LOOKUP_LE && diff > 0)
-		keyno--;
-	cur->bc_ptrs[0] = keyno;
-	/*
-	 * Return if we succeeded or not.
-	 */
-	if (keyno == 0 || keyno > be16_to_cpu(block->bb_numrecs))
-		*stat = 0;
-	else
-		*stat = ((dir != XFS_LOOKUP_EQ) || (diff == 0));
-	return 0;
+	ptr->s = agi->agi_root;
 }
 
-/*
- * Move 1 record left from cur/level if possible.
- * Update cur to reflect the new path.
- */
-STATIC int				/* error */
-xfs_inobt_lshift(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to shift record on */
-	int			*stat)	/* success/failure */
+STATIC __int64_t
+xfs_inobt_key_diff(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*key)
 {
-	int			error;	/* error return value */
-#ifdef DEBUG
-	int			i;	/* loop index */
-#endif
-	xfs_inobt_key_t		key;	/* key value for leaf level upward */
-	xfs_buf_t		*lbp;	/* buffer for left neighbor block */
-	xfs_inobt_block_t	*left;	/* left neighbor btree block */
-	xfs_inobt_key_t		*lkp=NULL;	/* key pointer for left block */
-	xfs_inobt_ptr_t		*lpp;	/* address pointer for left block */
-	xfs_inobt_rec_t		*lrp=NULL;	/* record pointer for left block */
-	int			nrec;	/* new number of left block entries */
-	xfs_buf_t		*rbp;	/* buffer for right (current) block */
-	xfs_inobt_block_t	*right;	/* right (current) btree block */
-	xfs_inobt_key_t		*rkp=NULL;	/* key pointer for right block */
-	xfs_inobt_ptr_t		*rpp=NULL;	/* address pointer for right block */
-	xfs_inobt_rec_t		*rrp=NULL;	/* record pointer for right block */
-
-	/*
-	 * Set up variables for this block as "right".
-	 */
-	rbp = cur->bc_bufs[level];
-	right = XFS_BUF_TO_INOBT_BLOCK(rbp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
-		return error;
-#endif
-	/*
-	 * If we've got no left sibling then we can't shift an entry left.
-	 */
-	if (be32_to_cpu(right->bb_leftsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * If the cursor entry is the one that would be moved, don't
-	 * do it... it's too complicated.
-	 */
-	if (cur->bc_ptrs[level] <= 1) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Set up the left neighbor as "left".
-	 */
-	if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-			cur->bc_private.a.agno, be32_to_cpu(right->bb_leftsib),
-			0, &lbp, XFS_INO_BTREE_REF)))
-		return error;
-	left = XFS_BUF_TO_INOBT_BLOCK(lbp);
-	if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-		return error;
-	/*
-	 * If it's full, it can't take another entry.
-	 */
-	if (be16_to_cpu(left->bb_numrecs) == XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
-		*stat = 0;
-		return 0;
-	}
-	nrec = be16_to_cpu(left->bb_numrecs) + 1;
-	/*
-	 * If non-leaf, copy a key and a ptr to the left block.
-	 */
-	if (level > 0) {
-		lkp = XFS_INOBT_KEY_ADDR(left, nrec, cur);
-		rkp = XFS_INOBT_KEY_ADDR(right, 1, cur);
-		*lkp = *rkp;
-		xfs_inobt_log_keys(cur, lbp, nrec, nrec);
-		lpp = XFS_INOBT_PTR_ADDR(left, nrec, cur);
-		rpp = XFS_INOBT_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(*rpp), level)))
-			return error;
-#endif
-		*lpp = *rpp;
-		xfs_inobt_log_ptrs(cur, lbp, nrec, nrec);
-	}
-	/*
-	 * If leaf, copy a record to the left block.
-	 */
-	else {
-		lrp = XFS_INOBT_REC_ADDR(left, nrec, cur);
-		rrp = XFS_INOBT_REC_ADDR(right, 1, cur);
-		*lrp = *rrp;
-		xfs_inobt_log_recs(cur, lbp, nrec, nrec);
-	}
-	/*
-	 * Bump and log left's numrecs, decrement and log right's numrecs.
-	 */
-	be16_add_cpu(&left->bb_numrecs, 1);
-	xfs_inobt_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS);
-#ifdef DEBUG
-	if (level > 0)
-		xfs_btree_check_key(cur->bc_btnum, lkp - 1, lkp);
-	else
-		xfs_btree_check_rec(cur->bc_btnum, lrp - 1, lrp);
-#endif
-	be16_add_cpu(&right->bb_numrecs, -1);
-	xfs_inobt_log_block(cur->bc_tp, rbp, XFS_BB_NUMRECS);
-	/*
-	 * Slide the contents of right down one entry.
-	 */
-	if (level > 0) {
-#ifdef DEBUG
-		for (i = 0; i < be16_to_cpu(right->bb_numrecs); i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(rpp[i + 1]),
-					level)))
-				return error;
-		}
-#endif
-		memmove(rkp, rkp + 1, be16_to_cpu(right->bb_numrecs) * sizeof(*rkp));
-		memmove(rpp, rpp + 1, be16_to_cpu(right->bb_numrecs) * sizeof(*rpp));
-		xfs_inobt_log_keys(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		xfs_inobt_log_ptrs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-	} else {
-		memmove(rrp, rrp + 1, be16_to_cpu(right->bb_numrecs) * sizeof(*rrp));
-		xfs_inobt_log_recs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		key.ir_startino = rrp->ir_startino;
-		rkp = &key;
-	}
-	/*
-	 * Update the parent key values of right.
-	 */
-	if ((error = xfs_inobt_updkey(cur, rkp, level + 1)))
-		return error;
-	/*
-	 * Slide the cursor value left one.
-	 */
-	cur->bc_ptrs[level]--;
-	*stat = 1;
-	return 0;
+	return (__int64_t)be32_to_cpu(key->inobt.ir_startino) -
+			  cur->bc_rec.i.ir_startino;
 }
 
-/*
- * Allocate a new root block, fill it in.
- */
-STATIC int				/* error */
-xfs_inobt_newroot(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			*stat)	/* success/failure */
+STATIC int
+xfs_inobt_kill_root(
+	struct xfs_btree_cur	*cur,
+	struct xfs_buf		*bp,
+	int			level,
+	union xfs_btree_ptr	*newroot)
 {
-	xfs_agi_t		*agi;	/* a.g. inode header */
-	xfs_alloc_arg_t		args;	/* allocation argument structure */
-	xfs_inobt_block_t	*block;	/* one half of the old root block */
-	xfs_buf_t		*bp;	/* buffer containing block */
-	int			error;	/* error return value */
-	xfs_inobt_key_t		*kp;	/* btree key pointer */
-	xfs_agblock_t		lbno;	/* left block number */
-	xfs_buf_t		*lbp;	/* left buffer pointer */
-	xfs_inobt_block_t	*left;	/* left btree block */
-	xfs_buf_t		*nbp;	/* new (root) buffer */
-	xfs_inobt_block_t	*new;	/* new (root) btree block */
-	int			nptr;	/* new value for key index, 1 or 2 */
-	xfs_inobt_ptr_t		*pp;	/* btree address pointer */
-	xfs_agblock_t		rbno;	/* right block number */
-	xfs_buf_t		*rbp;	/* right buffer pointer */
-	xfs_inobt_block_t	*right;	/* right btree block */
-	xfs_inobt_rec_t		*rp;	/* btree record pointer */
+	int			error;
 
-	ASSERT(cur->bc_nlevels < XFS_IN_MAXLEVELS(cur->bc_mp));
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_ENTRY);
+	XFS_BTREE_STATS_INC(cur, killroot);
 
 	/*
-	 * Get a block & a buffer.
+	 * Update the root pointer, decreasing the level by 1 and then
+	 * free the old root.
 	 */
-	agi = XFS_BUF_TO_AGI(cur->bc_private.a.agbp);
-	args.tp = cur->bc_tp;
-	args.mp = cur->bc_mp;
-	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno,
-		be32_to_cpu(agi->agi_root));
-	args.mod = args.minleft = args.alignment = args.total = args.wasdel =
-		args.isfl = args.userdata = args.minalignslop = 0;
-	args.minlen = args.maxlen = args.prod = 1;
-	args.type = XFS_ALLOCTYPE_NEAR_BNO;
-	if ((error = xfs_alloc_vextent(&args)))
+	xfs_inobt_set_root(cur, newroot, -1);
+	error = xfs_inobt_free_block(cur, bp);
+	if (error) {
+		XFS_BTREE_TRACE_CURSOR(cur, XBT_ERROR);
 		return error;
-	/*
-	 * None available, we fail.
-	 */
-	if (args.fsbno == NULLFSBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	ASSERT(args.len == 1);
-	nbp = xfs_btree_get_bufs(args.mp, args.tp, args.agno, args.agbno, 0);
-	new = XFS_BUF_TO_INOBT_BLOCK(nbp);
-	/*
-	 * Set the root data in the a.g. inode structure.
-	 */
-	agi->agi_root = cpu_to_be32(args.agbno);
-	be32_add_cpu(&agi->agi_level, 1);
-	xfs_ialloc_log_agi(args.tp, cur->bc_private.a.agbp,
-		XFS_AGI_ROOT | XFS_AGI_LEVEL);
-	/*
-	 * At the previous root level there are now two blocks: the old
-	 * root, and the new block generated when it was split.
-	 * We don't know which one the cursor is pointing at, so we
-	 * set up variables "left" and "right" for each case.
-	 */
-	bp = cur->bc_bufs[cur->bc_nlevels - 1];
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, cur->bc_nlevels - 1, bp)))
-		return error;
-#endif
-	if (be32_to_cpu(block->bb_rightsib) != NULLAGBLOCK) {
-		/*
-		 * Our block is left, pick up the right block.
-		 */
-		lbp = bp;
-		lbno = XFS_DADDR_TO_AGBNO(args.mp, XFS_BUF_ADDR(lbp));
-		left = block;
-		rbno = be32_to_cpu(left->bb_rightsib);
-		if ((error = xfs_btree_read_bufs(args.mp, args.tp, args.agno,
-				rbno, 0, &rbp, XFS_INO_BTREE_REF)))
-			return error;
-		bp = rbp;
-		right = XFS_BUF_TO_INOBT_BLOCK(rbp);
-		if ((error = xfs_btree_check_sblock(cur, right,
-				cur->bc_nlevels - 1, rbp)))
-			return error;
-		nptr = 1;
-	} else {
-		/*
-		 * Our block is right, pick up the left block.
-		 */
-		rbp = bp;
-		rbno = XFS_DADDR_TO_AGBNO(args.mp, XFS_BUF_ADDR(rbp));
-		right = block;
-		lbno = be32_to_cpu(right->bb_leftsib);
-		if ((error = xfs_btree_read_bufs(args.mp, args.tp, args.agno,
-				lbno, 0, &lbp, XFS_INO_BTREE_REF)))
-			return error;
-		bp = lbp;
-		left = XFS_BUF_TO_INOBT_BLOCK(lbp);
-		if ((error = xfs_btree_check_sblock(cur, left,
-				cur->bc_nlevels - 1, lbp)))
-			return error;
-		nptr = 2;
-	}
-	/*
-	 * Fill in the new block's btree header and log it.
-	 */
-	new->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
-	new->bb_level = cpu_to_be16(cur->bc_nlevels);
-	new->bb_numrecs = cpu_to_be16(2);
-	new->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-	new->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
-	xfs_inobt_log_block(args.tp, nbp, XFS_BB_ALL_BITS);
-	ASSERT(lbno != NULLAGBLOCK && rbno != NULLAGBLOCK);
-	/*
-	 * Fill in the key data in the new root.
-	 */
-	kp = XFS_INOBT_KEY_ADDR(new, 1, cur);
-	if (be16_to_cpu(left->bb_level) > 0) {
-		kp[0] = *XFS_INOBT_KEY_ADDR(left, 1, cur);
-		kp[1] = *XFS_INOBT_KEY_ADDR(right, 1, cur);
-	} else {
-		rp = XFS_INOBT_REC_ADDR(left, 1, cur);
-		kp[0].ir_startino = rp->ir_startino;
-		rp = XFS_INOBT_REC_ADDR(right, 1, cur);
-		kp[1].ir_startino = rp->ir_startino;
 	}
-	xfs_inobt_log_keys(cur, nbp, 1, 2);
-	/*
-	 * Fill in the pointer data in the new root.
-	 */
-	pp = XFS_INOBT_PTR_ADDR(new, 1, cur);
-	pp[0] = cpu_to_be32(lbno);
-	pp[1] = cpu_to_be32(rbno);
-	xfs_inobt_log_ptrs(cur, nbp, 1, 2);
-	/*
-	 * Fix up the cursor.
-	 */
-	xfs_btree_setbuf(cur, cur->bc_nlevels, nbp);
-	cur->bc_ptrs[cur->bc_nlevels] = nptr;
-	cur->bc_nlevels++;
-	*stat = 1;
-	return 0;
-}
 
-/*
- * Move 1 record right from cur/level if possible.
- * Update cur to reflect the new path.
- */
-STATIC int				/* error */
-xfs_inobt_rshift(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to shift record on */
-	int			*stat)	/* success/failure */
-{
-	int			error;	/* error return value */
-	int			i;	/* loop index */
-	xfs_inobt_key_t		key;	/* key value for leaf level upward */
-	xfs_buf_t		*lbp;	/* buffer for left (current) block */
-	xfs_inobt_block_t	*left;	/* left (current) btree block */
-	xfs_inobt_key_t		*lkp;	/* key pointer for left block */
-	xfs_inobt_ptr_t		*lpp;	/* address pointer for left block */
-	xfs_inobt_rec_t		*lrp;	/* record pointer for left block */
-	xfs_buf_t		*rbp;	/* buffer for right neighbor block */
-	xfs_inobt_block_t	*right;	/* right neighbor btree block */
-	xfs_inobt_key_t		*rkp;	/* key pointer for right block */
-	xfs_inobt_ptr_t		*rpp;	/* address pointer for right block */
-	xfs_inobt_rec_t		*rrp=NULL;	/* record pointer for right block */
-	xfs_btree_cur_t		*tcur;	/* temporary cursor */
+	XFS_BTREE_STATS_INC(cur, free);
 
-	/*
-	 * Set up variables for this block as "left".
-	 */
-	lbp = cur->bc_bufs[level];
-	left = XFS_BUF_TO_INOBT_BLOCK(lbp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-		return error;
-#endif
-	/*
-	 * If we've got no right sibling then we can't shift an entry right.
-	 */
-	if (be32_to_cpu(left->bb_rightsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * If the cursor entry is the one that would be moved, don't
-	 * do it... it's too complicated.
-	 */
-	if (cur->bc_ptrs[level] >= be16_to_cpu(left->bb_numrecs)) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Set up the right neighbor as "right".
-	 */
-	if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-			cur->bc_private.a.agno, be32_to_cpu(left->bb_rightsib),
-			0, &rbp, XFS_INO_BTREE_REF)))
-		return error;
-	right = XFS_BUF_TO_INOBT_BLOCK(rbp);
-	if ((error = xfs_btree_check_sblock(cur, right, level, rbp)))
-		return error;
-	/*
-	 * If it's full, it can't take another entry.
-	 */
-	if (be16_to_cpu(right->bb_numrecs) == XFS_INOBT_BLOCK_MAXRECS(level, cur)) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Make a hole at the start of the right neighbor block, then
-	 * copy the last left block entry to the hole.
-	 */
-	if (level > 0) {
-		lkp = XFS_INOBT_KEY_ADDR(left, be16_to_cpu(left->bb_numrecs), cur);
-		lpp = XFS_INOBT_PTR_ADDR(left, be16_to_cpu(left->bb_numrecs), cur);
-		rkp = XFS_INOBT_KEY_ADDR(right, 1, cur);
-		rpp = XFS_INOBT_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		for (i = be16_to_cpu(right->bb_numrecs) - 1; i >= 0; i--) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(rpp[i]), level)))
-				return error;
-		}
-#endif
-		memmove(rkp + 1, rkp, be16_to_cpu(right->bb_numrecs) * sizeof(*rkp));
-		memmove(rpp + 1, rpp, be16_to_cpu(right->bb_numrecs) * sizeof(*rpp));
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(*lpp), level)))
-			return error;
-#endif
-		*rkp = *lkp;
-		*rpp = *lpp;
-		xfs_inobt_log_keys(cur, rbp, 1, be16_to_cpu(right->bb_numrecs) + 1);
-		xfs_inobt_log_ptrs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs) + 1);
-	} else {
-		lrp = XFS_INOBT_REC_ADDR(left, be16_to_cpu(left->bb_numrecs), cur);
-		rrp = XFS_INOBT_REC_ADDR(right, 1, cur);
-		memmove(rrp + 1, rrp, be16_to_cpu(right->bb_numrecs) * sizeof(*rrp));
-		*rrp = *lrp;
-		xfs_inobt_log_recs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs) + 1);
-		key.ir_startino = rrp->ir_startino;
-		rkp = &key;
-	}
-	/*
-	 * Decrement and log left's numrecs, bump and log right's numrecs.
-	 */
-	be16_add_cpu(&left->bb_numrecs, -1);
-	xfs_inobt_log_block(cur->bc_tp, lbp, XFS_BB_NUMRECS);
-	be16_add_cpu(&right->bb_numrecs, 1);
-#ifdef DEBUG
-	if (level > 0)
-		xfs_btree_check_key(cur->bc_btnum, rkp, rkp + 1);
-	else
-		xfs_btree_check_rec(cur->bc_btnum, rrp, rrp + 1);
-#endif
-	xfs_inobt_log_block(cur->bc_tp, rbp, XFS_BB_NUMRECS);
-	/*
-	 * Using a temporary cursor, update the parent key values of the
-	 * block on the right.
-	 */
-	if ((error = xfs_btree_dup_cursor(cur, &tcur)))
-		return error;
-	xfs_btree_lastrec(tcur, level);
-	if ((error = xfs_inobt_increment(tcur, level, &i)) ||
-	    (error = xfs_inobt_updkey(tcur, rkp, level + 1))) {
-		xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
-		return error;
-	}
-	xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
-	*stat = 1;
+	cur->bc_bufs[level] = NULL;
+	cur->bc_nlevels--;
+
+	XFS_BTREE_TRACE_CURSOR(cur, XBT_EXIT);
 	return 0;
 }
 
-/*
- * Split cur/level block in half.
- * Return new block number and its first record (to be inserted into parent).
- */
-STATIC int				/* error */
-xfs_inobt_split(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level to split */
-	xfs_agblock_t		*bnop,	/* output: block number allocated */
-	xfs_inobt_key_t		*keyp,	/* output: first key of new block */
-	xfs_btree_cur_t		**curp,	/* output: new cursor */
-	int			*stat)	/* success/failure */
-{
-	xfs_alloc_arg_t		args;	/* allocation argument structure */
-	int			error;	/* error return value */
-	int			i;	/* loop index/record number */
-	xfs_agblock_t		lbno;	/* left (current) block number */
-	xfs_buf_t		*lbp;	/* buffer for left block */
-	xfs_inobt_block_t	*left;	/* left (current) btree block */
-	xfs_inobt_key_t		*lkp;	/* left btree key pointer */
-	xfs_inobt_ptr_t		*lpp;	/* left btree address pointer */
-	xfs_inobt_rec_t		*lrp;	/* left btree record pointer */
-	xfs_buf_t		*rbp;	/* buffer for right block */
-	xfs_inobt_block_t	*right;	/* right (new) btree block */
-	xfs_inobt_key_t		*rkp;	/* right btree key pointer */
-	xfs_inobt_ptr_t		*rpp;	/* right btree address pointer */
-	xfs_inobt_rec_t		*rrp;	/* right btree record pointer */
-
-	/*
-	 * Set up left block (current one).
-	 */
-	lbp = cur->bc_bufs[level];
-	args.tp = cur->bc_tp;
-	args.mp = cur->bc_mp;
-	lbno = XFS_DADDR_TO_AGBNO(args.mp, XFS_BUF_ADDR(lbp));
-	/*
-	 * Allocate the new block.
-	 * If we can't do it, we're toast.  Give up.
-	 */
-	args.fsbno = XFS_AGB_TO_FSB(args.mp, cur->bc_private.a.agno, lbno);
-	args.mod = args.minleft = args.alignment = args.total = args.wasdel =
-		args.isfl = args.userdata = args.minalignslop = 0;
-	args.minlen = args.maxlen = args.prod = 1;
-	args.type = XFS_ALLOCTYPE_NEAR_BNO;
-	if ((error = xfs_alloc_vextent(&args)))
-		return error;
-	if (args.fsbno == NULLFSBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	ASSERT(args.len == 1);
-	rbp = xfs_btree_get_bufs(args.mp, args.tp, args.agno, args.agbno, 0);
-	/*
-	 * Set up the new block as "right".
-	 */
-	right = XFS_BUF_TO_INOBT_BLOCK(rbp);
-	/*
-	 * "Left" is the current (according to the cursor) block.
-	 */
-	left = XFS_BUF_TO_INOBT_BLOCK(lbp);
 #ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, left, level, lbp)))
-		return error;
-#endif
-	/*
-	 * Fill in the btree header for the new block.
-	 */
-	right->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
-	right->bb_level = left->bb_level;
-	right->bb_numrecs = cpu_to_be16(be16_to_cpu(left->bb_numrecs) / 2);
-	/*
-	 * Make sure that if there's an odd number of entries now, that
-	 * each new block will have the same number of entries.
-	 */
-	if ((be16_to_cpu(left->bb_numrecs) & 1) &&
-	    cur->bc_ptrs[level] <= be16_to_cpu(right->bb_numrecs) + 1)
-		be16_add_cpu(&right->bb_numrecs, 1);
-	i = be16_to_cpu(left->bb_numrecs) - be16_to_cpu(right->bb_numrecs) + 1;
-	/*
-	 * For non-leaf blocks, copy keys and addresses over to the new block.
-	 */
-	if (level > 0) {
-		lkp = XFS_INOBT_KEY_ADDR(left, i, cur);
-		lpp = XFS_INOBT_PTR_ADDR(left, i, cur);
-		rkp = XFS_INOBT_KEY_ADDR(right, 1, cur);
-		rpp = XFS_INOBT_PTR_ADDR(right, 1, cur);
-#ifdef DEBUG
-		for (i = 0; i < be16_to_cpu(right->bb_numrecs); i++) {
-			if ((error = xfs_btree_check_sptr(cur, be32_to_cpu(lpp[i]), level)))
-				return error;
-		}
-#endif
-		memcpy(rkp, lkp, be16_to_cpu(right->bb_numrecs) * sizeof(*rkp));
-		memcpy(rpp, lpp, be16_to_cpu(right->bb_numrecs) * sizeof(*rpp));
-		xfs_inobt_log_keys(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		xfs_inobt_log_ptrs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		*keyp = *rkp;
-	}
-	/*
-	 * For leaf blocks, copy records over to the new block.
-	 */
-	else {
-		lrp = XFS_INOBT_REC_ADDR(left, i, cur);
-		rrp = XFS_INOBT_REC_ADDR(right, 1, cur);
-		memcpy(rrp, lrp, be16_to_cpu(right->bb_numrecs) * sizeof(*rrp));
-		xfs_inobt_log_recs(cur, rbp, 1, be16_to_cpu(right->bb_numrecs));
-		keyp->ir_startino = rrp->ir_startino;
-	}
-	/*
-	 * Find the left block number by looking in the buffer.
-	 * Adjust numrecs, sibling pointers.
-	 */
-	be16_add_cpu(&left->bb_numrecs, -(be16_to_cpu(right->bb_numrecs)));
-	right->bb_rightsib = left->bb_rightsib;
-	left->bb_rightsib = cpu_to_be32(args.agbno);
-	right->bb_leftsib = cpu_to_be32(lbno);
-	xfs_inobt_log_block(args.tp, rbp, XFS_BB_ALL_BITS);
-	xfs_inobt_log_block(args.tp, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);
-	/*
-	 * If there's a block to the new block's right, make that block
-	 * point back to right instead of to left.
-	 */
-	if (be32_to_cpu(right->bb_rightsib) != NULLAGBLOCK) {
-		xfs_inobt_block_t	*rrblock;	/* rr btree block */
-		xfs_buf_t		*rrbp;		/* buffer for rrblock */
-
-		if ((error = xfs_btree_read_bufs(args.mp, args.tp, args.agno,
-				be32_to_cpu(right->bb_rightsib), 0, &rrbp,
-				XFS_INO_BTREE_REF)))
-			return error;
-		rrblock = XFS_BUF_TO_INOBT_BLOCK(rrbp);
-		if ((error = xfs_btree_check_sblock(cur, rrblock, level, rrbp)))
-			return error;
-		rrblock->bb_leftsib = cpu_to_be32(args.agbno);
-		xfs_inobt_log_block(args.tp, rrbp, XFS_BB_LEFTSIB);
-	}
-	/*
-	 * If the cursor is really in the right block, move it there.
-	 * If it's just pointing past the last entry in left, then we'll
-	 * insert there, so don't change anything in that case.
-	 */
-	if (cur->bc_ptrs[level] > be16_to_cpu(left->bb_numrecs) + 1) {
-		xfs_btree_setbuf(cur, level, rbp);
-		cur->bc_ptrs[level] -= be16_to_cpu(left->bb_numrecs);
-	}
-	/*
-	 * If there are more levels, we'll need another cursor which refers
-	 * the right block, no matter where this cursor was.
-	 */
-	if (level + 1 < cur->bc_nlevels) {
-		if ((error = xfs_btree_dup_cursor(cur, curp)))
-			return error;
-		(*curp)->bc_ptrs[level + 1]++;
-	}
-	*bnop = args.agbno;
-	*stat = 1;
-	return 0;
+STATIC int
+xfs_inobt_keys_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*k1,
+	union xfs_btree_key	*k2)
+{
+	return be32_to_cpu(k1->inobt.ir_startino) <
+		be32_to_cpu(k2->inobt.ir_startino);
 }
 
-/*
- * Update keys at all levels from here to the root along the cursor's path.
- */
-STATIC int				/* error */
-xfs_inobt_updkey(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_inobt_key_t		*keyp,	/* new key value to update to */
-	int			level)	/* starting level for update */
+STATIC int
+xfs_inobt_recs_inorder(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*r1,
+	union xfs_btree_rec	*r2)
 {
-	int			ptr;	/* index of key in block */
-
-	/*
-	 * Go up the tree from this level toward the root.
-	 * At each level, update the key value to the value input.
-	 * Stop when we reach a level where the cursor isn't pointing
-	 * at the first entry in the block.
-	 */
-	for (ptr = 1; ptr == 1 && level < cur->bc_nlevels; level++) {
-		xfs_buf_t		*bp;	/* buffer for block */
-		xfs_inobt_block_t	*block;	/* btree block */
-#ifdef DEBUG
-		int			error;	/* error return value */
-#endif
-		xfs_inobt_key_t		*kp;	/* ptr to btree block keys */
-
-		bp = cur->bc_bufs[level];
-		block = XFS_BUF_TO_INOBT_BLOCK(bp);
-#ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
-			return error;
-#endif
-		ptr = cur->bc_ptrs[level];
-		kp = XFS_INOBT_KEY_ADDR(block, ptr, cur);
-		*kp = *keyp;
-		xfs_inobt_log_keys(cur, bp, ptr, ptr);
-	}
-	return 0;
+	return be32_to_cpu(r1->inobt.ir_startino) + XFS_INODES_PER_CHUNK <=
+		be32_to_cpu(r2->inobt.ir_startino);
 }
+#endif	/* DEBUG */
 
-/*
- * Externally visible routines.
- */
+#ifdef XFS_BTREE_TRACE
+ktrace_t	*xfs_inobt_trace_buf;
 
-/*
- * Decrement cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-int					/* error */
-xfs_inobt_decrement(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level in btree, 0 is leaf */
-	int			*stat)	/* success/failure */
+STATIC void
+xfs_inobt_trace_enter(
+	struct xfs_btree_cur	*cur,
+	const char		*func,
+	char			*s,
+	int			type,
+	int			line,
+	__psunsigned_t		a0,
+	__psunsigned_t		a1,
+	__psunsigned_t		a2,
+	__psunsigned_t		a3,
+	__psunsigned_t		a4,
+	__psunsigned_t		a5,
+	__psunsigned_t		a6,
+	__psunsigned_t		a7,
+	__psunsigned_t		a8,
+	__psunsigned_t		a9,
+	__psunsigned_t		a10)
 {
-	xfs_inobt_block_t	*block;	/* btree block */
-	int			error;
-	int			lev;	/* btree level */
-
-	ASSERT(level < cur->bc_nlevels);
-	/*
-	 * Read-ahead to the left at this level.
-	 */
-	xfs_btree_readahead(cur, level, XFS_BTCUR_LEFTRA);
-	/*
-	 * Decrement the ptr at this level.  If we're still in the block
-	 * then we're done.
-	 */
-	if (--cur->bc_ptrs[level] > 0) {
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * Get a pointer to the btree block.
-	 */
-	block = XFS_BUF_TO_INOBT_BLOCK(cur->bc_bufs[level]);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level,
-			cur->bc_bufs[level])))
-		return error;
-#endif
-	/*
-	 * If we just went off the left edge of the tree, return failure.
-	 */
-	if (be32_to_cpu(block->bb_leftsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * March up the tree decrementing pointers.
-	 * Stop when we don't go off the left edge of a block.
-	 */
-	for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
-		if (--cur->bc_ptrs[lev] > 0)
-			break;
-		/*
-		 * Read-ahead the left block, we're going to read it
-		 * in the next loop.
-		 */
-		xfs_btree_readahead(cur, lev, XFS_BTCUR_LEFTRA);
-	}
-	/*
-	 * If we went off the root then we are seriously confused.
-	 */
-	ASSERT(lev < cur->bc_nlevels);
-	/*
-	 * Now walk back down the tree, fixing up the cursor's buffer
-	 * pointers and key numbers.
-	 */
-	for (block = XFS_BUF_TO_INOBT_BLOCK(cur->bc_bufs[lev]); lev > level; ) {
-		xfs_agblock_t	agbno;	/* block number of btree block */
-		xfs_buf_t	*bp;	/* buffer containing btree block */
-
-		agbno = be32_to_cpu(*XFS_INOBT_PTR_ADDR(block, cur->bc_ptrs[lev], cur));
-		if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-				cur->bc_private.a.agno, agbno, 0, &bp,
-				XFS_INO_BTREE_REF)))
-			return error;
-		lev--;
-		xfs_btree_setbuf(cur, lev, bp);
-		block = XFS_BUF_TO_INOBT_BLOCK(bp);
-		if ((error = xfs_btree_check_sblock(cur, block, lev, bp)))
-			return error;
-		cur->bc_ptrs[lev] = be16_to_cpu(block->bb_numrecs);
-	}
-	*stat = 1;
-	return 0;
+	ktrace_enter(xfs_inobt_trace_buf, (void *)(__psint_t)type,
+		(void *)func, (void *)s, NULL, (void *)cur,
+		(void *)a0, (void *)a1, (void *)a2, (void *)a3,
+		(void *)a4, (void *)a5, (void *)a6, (void *)a7,
+		(void *)a8, (void *)a9, (void *)a10);
 }
 
-/*
- * Delete the record pointed to by cur.
- * The cursor refers to the place where the record was (could be inserted)
- * when the operation returns.
- */
-int					/* error */
-xfs_inobt_delete(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	int		*stat)		/* success/failure */
+STATIC void
+xfs_inobt_trace_cursor(
+	struct xfs_btree_cur	*cur,
+	__uint32_t		*s0,
+	__uint64_t		*l0,
+	__uint64_t		*l1)
 {
-	int		error;
-	int		i;		/* result code */
-	int		level;		/* btree level */
-
-	/*
-	 * Go up the tree, starting at leaf level.
-	 * If 2 is returned then a join was done; go to the next level.
-	 * Otherwise we are done.
-	 */
-	for (level = 0, i = 2; i == 2; level++) {
-		if ((error = xfs_inobt_delrec(cur, level, &i)))
-			return error;
-	}
-	if (i == 0) {
-		for (level = 1; level < cur->bc_nlevels; level++) {
-			if (cur->bc_ptrs[level] == 0) {
-				if ((error = xfs_inobt_decrement(cur, level, &i)))
-					return error;
-				break;
-			}
-		}
-	}
-	*stat = i;
-	return 0;
+	*s0 = cur->bc_private.a.agno;
+	*l0 = cur->bc_rec.i.ir_startino;
+	*l1 = cur->bc_rec.i.ir_free;
 }
 
-
-/*
- * Get the data from the pointed-to record.
- */
-int					/* error */
-xfs_inobt_get_rec(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_agino_t		*ino,	/* output: starting inode of chunk */
-	__int32_t		*fcnt,	/* output: number of free inodes */
-	xfs_inofree_t		*free,	/* output: free inode mask */
-	int			*stat)	/* output: success/failure */
+STATIC void
+xfs_inobt_trace_key(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_key	*key,
+	__uint64_t		*l0,
+	__uint64_t		*l1)
 {
-	xfs_inobt_block_t	*block;	/* btree block */
-	xfs_buf_t		*bp;	/* buffer containing btree block */
-#ifdef DEBUG
-	int			error;	/* error return value */
-#endif
-	int			ptr;	/* record number */
-	xfs_inobt_rec_t		*rec;	/* record data */
-
-	bp = cur->bc_bufs[0];
-	ptr = cur->bc_ptrs[0];
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, 0, bp)))
-		return error;
-#endif
-	/*
-	 * Off the right end or left end, return failure.
-	 */
-	if (ptr > be16_to_cpu(block->bb_numrecs) || ptr <= 0) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * Point to the record and extract its data.
-	 */
-	rec = XFS_INOBT_REC_ADDR(block, ptr, cur);
-	*ino = be32_to_cpu(rec->ir_startino);
-	*fcnt = be32_to_cpu(rec->ir_freecount);
-	*free = be64_to_cpu(rec->ir_free);
-	*stat = 1;
-	return 0;
+	*l0 = be32_to_cpu(key->inobt.ir_startino);
+	*l1 = 0;
 }
 
-/*
- * Increment cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-int					/* error */
-xfs_inobt_increment(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	int			level,	/* level in btree, 0 is leaf */
-	int			*stat)	/* success/failure */
+STATIC void
+xfs_inobt_trace_record(
+	struct xfs_btree_cur	*cur,
+	union xfs_btree_rec	*rec,
+	__uint64_t		*l0,
+	__uint64_t		*l1,
+	__uint64_t		*l2)
 {
-	xfs_inobt_block_t	*block;	/* btree block */
-	xfs_buf_t		*bp;	/* buffer containing btree block */
-	int			error;	/* error return value */
-	int			lev;	/* btree level */
+	*l0 = be32_to_cpu(rec->inobt.ir_startino);
+	*l1 = be32_to_cpu(rec->inobt.ir_freecount);
+	*l2 = be64_to_cpu(rec->inobt.ir_free);
+}
+#endif /* XFS_BTREE_TRACE */
+
+static const struct xfs_btree_ops xfs_inobt_ops = {
+	.rec_len		= sizeof(xfs_inobt_rec_t),
+	.key_len		= sizeof(xfs_inobt_key_t),
+
+	.dup_cursor		= xfs_inobt_dup_cursor,
+	.set_root		= xfs_inobt_set_root,
+	.kill_root		= xfs_inobt_kill_root,
+	.alloc_block		= xfs_inobt_alloc_block,
+	.free_block		= xfs_inobt_free_block,
+	.get_minrecs		= xfs_inobt_get_minrecs,
+	.get_maxrecs		= xfs_inobt_get_maxrecs,
+	.init_key_from_rec	= xfs_inobt_init_key_from_rec,
+	.init_rec_from_key	= xfs_inobt_init_rec_from_key,
+	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
+	.init_ptr_from_cur	= xfs_inobt_init_ptr_from_cur,
+	.key_diff		= xfs_inobt_key_diff,
 
-	ASSERT(level < cur->bc_nlevels);
-	/*
-	 * Read-ahead to the right at this level.
-	 */
-	xfs_btree_readahead(cur, level, XFS_BTCUR_RIGHTRA);
-	/*
-	 * Get a pointer to the btree block.
-	 */
-	bp = cur->bc_bufs[level];
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, level, bp)))
-		return error;
-#endif
-	/*
-	 * Increment the ptr at this level.  If we're still in the block
-	 * then we're done.
-	 */
-	if (++cur->bc_ptrs[level] <= be16_to_cpu(block->bb_numrecs)) {
-		*stat = 1;
-		return 0;
-	}
-	/*
-	 * If we just went off the right edge of the tree, return failure.
-	 */
-	if (be32_to_cpu(block->bb_rightsib) == NULLAGBLOCK) {
-		*stat = 0;
-		return 0;
-	}
-	/*
-	 * March up the tree incrementing pointers.
-	 * Stop when we don't go off the right edge of a block.
-	 */
-	for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
-		bp = cur->bc_bufs[lev];
-		block = XFS_BUF_TO_INOBT_BLOCK(bp);
 #ifdef DEBUG
-		if ((error = xfs_btree_check_sblock(cur, block, lev, bp)))
-			return error;
+	.keys_inorder		= xfs_inobt_keys_inorder,
+	.recs_inorder		= xfs_inobt_recs_inorder,
 #endif
-		if (++cur->bc_ptrs[lev] <= be16_to_cpu(block->bb_numrecs))
-			break;
-		/*
-		 * Read-ahead the right block, we're going to read it
-		 * in the next loop.
-		 */
-		xfs_btree_readahead(cur, lev, XFS_BTCUR_RIGHTRA);
-	}
-	/*
-	 * If we went off the root then we are seriously confused.
-	 */
-	ASSERT(lev < cur->bc_nlevels);
-	/*
-	 * Now walk back down the tree, fixing up the cursor's buffer
-	 * pointers and key numbers.
-	 */
-	for (bp = cur->bc_bufs[lev], block = XFS_BUF_TO_INOBT_BLOCK(bp);
-	     lev > level; ) {
-		xfs_agblock_t	agbno;	/* block number of btree block */
 
-		agbno = be32_to_cpu(*XFS_INOBT_PTR_ADDR(block, cur->bc_ptrs[lev], cur));
-		if ((error = xfs_btree_read_bufs(cur->bc_mp, cur->bc_tp,
-				cur->bc_private.a.agno, agbno, 0, &bp,
-				XFS_INO_BTREE_REF)))
-			return error;
-		lev--;
-		xfs_btree_setbuf(cur, lev, bp);
-		block = XFS_BUF_TO_INOBT_BLOCK(bp);
-		if ((error = xfs_btree_check_sblock(cur, block, lev, bp)))
-			return error;
-		cur->bc_ptrs[lev] = 1;
-	}
-	*stat = 1;
-	return 0;
-}
+#ifdef XFS_BTREE_TRACE
+	.trace_enter		= xfs_inobt_trace_enter,
+	.trace_cursor		= xfs_inobt_trace_cursor,
+	.trace_key		= xfs_inobt_trace_key,
+	.trace_record		= xfs_inobt_trace_record,
+#endif
+};
 
 /*
- * Insert the current record at the point referenced by cur.
- * The cursor may be inconsistent on return if splits have been done.
+ * Allocate a new inode btree cursor.
  */
-int					/* error */
-xfs_inobt_insert(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	int		*stat)		/* success/failure */
+struct xfs_btree_cur *				/* new inode btree cursor */
+xfs_inobt_init_cursor(
+	struct xfs_mount	*mp,		/* file system mount point */
+	struct xfs_trans	*tp,		/* transaction pointer */
+	struct xfs_buf		*agbp,		/* buffer for agi structure */
+	xfs_agnumber_t		agno)		/* allocation group number */
 {
-	int		error;		/* error return value */
-	int		i;		/* result value, 0 for failure */
-	int		level;		/* current level number in btree */
-	xfs_agblock_t	nbno;		/* new block number (split result) */
-	xfs_btree_cur_t	*ncur;		/* new cursor (split result) */
-	xfs_inobt_rec_t	nrec;		/* record being inserted this level */
-	xfs_btree_cur_t	*pcur;		/* previous level's cursor */
+	struct xfs_agi		*agi = XFS_BUF_TO_AGI(agbp);
+	struct xfs_btree_cur	*cur;
 
-	level = 0;
-	nbno = NULLAGBLOCK;
-	nrec.ir_startino = cpu_to_be32(cur->bc_rec.i.ir_startino);
-	nrec.ir_freecount = cpu_to_be32(cur->bc_rec.i.ir_freecount);
-	nrec.ir_free = cpu_to_be64(cur->bc_rec.i.ir_free);
-	ncur = NULL;
-	pcur = cur;
-	/*
-	 * Loop going up the tree, starting at the leaf level.
-	 * Stop when we don't get a split block, that must mean that
-	 * the insert is finished with this level.
-	 */
-	do {
-		/*
-		 * Insert nrec/nbno into this level of the tree.
-		 * Note if we fail, nbno will be null.
-		 */
-		if ((error = xfs_inobt_insrec(pcur, level++, &nbno, &nrec, &ncur,
-				&i))) {
-			if (pcur != cur)
-				xfs_btree_del_cursor(pcur, XFS_BTREE_ERROR);
-			return error;
-		}
-		/*
-		 * See if the cursor we just used is trash.
-		 * Can't trash the caller's cursor, but otherwise we should
-		 * if ncur is a new cursor or we're about to be done.
-		 */
-		if (pcur != cur && (ncur || nbno == NULLAGBLOCK)) {
-			cur->bc_nlevels = pcur->bc_nlevels;
-			xfs_btree_del_cursor(pcur, XFS_BTREE_NOERROR);
-		}
-		/*
-		 * If we got a new cursor, switch to it.
-		 */
-		if (ncur) {
-			pcur = ncur;
-			ncur = NULL;
-		}
-	} while (nbno != NULLAGBLOCK);
-	*stat = i;
-	return 0;
-}
+	cur = kmem_zone_zalloc(xfs_btree_cur_zone, KM_SLEEP);
 
-/*
- * Lookup the record equal to ino in the btree given by cur.
- */
-int					/* error */
-xfs_inobt_lookup_eq(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_agino_t	ino,		/* starting inode of chunk */
-	__int32_t	fcnt,		/* free inode count */
-	xfs_inofree_t	free,		/* free inode mask */
-	int		*stat)		/* success/failure */
-{
-	cur->bc_rec.i.ir_startino = ino;
-	cur->bc_rec.i.ir_freecount = fcnt;
-	cur->bc_rec.i.ir_free = free;
-	return xfs_inobt_lookup(cur, XFS_LOOKUP_EQ, stat);
-}
+	cur->bc_tp = tp;
+	cur->bc_mp = mp;
+	cur->bc_nlevels = be32_to_cpu(agi->agi_level);
+	cur->bc_btnum = XFS_BTNUM_INO;
+	cur->bc_blocklog = mp->m_sb.sb_blocklog;
 
-/*
- * Lookup the first record greater than or equal to ino
- * in the btree given by cur.
- */
-int					/* error */
-xfs_inobt_lookup_ge(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_agino_t	ino,		/* starting inode of chunk */
-	__int32_t	fcnt,		/* free inode count */
-	xfs_inofree_t	free,		/* free inode mask */
-	int		*stat)		/* success/failure */
-{
-	cur->bc_rec.i.ir_startino = ino;
-	cur->bc_rec.i.ir_freecount = fcnt;
-	cur->bc_rec.i.ir_free = free;
-	return xfs_inobt_lookup(cur, XFS_LOOKUP_GE, stat);
-}
+	cur->bc_ops = &xfs_inobt_ops;
 
-/*
- * Lookup the first record less than or equal to ino
- * in the btree given by cur.
- */
-int					/* error */
-xfs_inobt_lookup_le(
-	xfs_btree_cur_t	*cur,		/* btree cursor */
-	xfs_agino_t	ino,		/* starting inode of chunk */
-	__int32_t	fcnt,		/* free inode count */
-	xfs_inofree_t	free,		/* free inode mask */
-	int		*stat)		/* success/failure */
-{
-	cur->bc_rec.i.ir_startino = ino;
-	cur->bc_rec.i.ir_freecount = fcnt;
-	cur->bc_rec.i.ir_free = free;
-	return xfs_inobt_lookup(cur, XFS_LOOKUP_LE, stat);
+	cur->bc_private.a.agbp = agbp;
+	cur->bc_private.a.agno = agno;
+
+	return cur;
 }
 
 /*
- * Update the record referred to by cur, to the value given
- * by [ino, fcnt, free].
- * This either works (return 0) or gets an EFSCORRUPTED error.
+ * Calculate number of records in an inobt btree block.
  */
-int					/* error */
-xfs_inobt_update(
-	xfs_btree_cur_t		*cur,	/* btree cursor */
-	xfs_agino_t		ino,	/* starting inode of chunk */
-	__int32_t		fcnt,	/* free inode count */
-	xfs_inofree_t		free)	/* free inode mask */
+int
+xfs_inobt_maxrecs(
+	struct xfs_mount	*mp,
+	int			blocklen,
+	int			leaf)
 {
-	xfs_inobt_block_t	*block;	/* btree block to update */
-	xfs_buf_t		*bp;	/* buffer containing btree block */
-	int			error;	/* error return value */
-	int			ptr;	/* current record number (updating) */
-	xfs_inobt_rec_t		*rp;	/* pointer to updated record */
+	blocklen -= XFS_INOBT_BLOCK_LEN(mp);
 
-	/*
-	 * Pick up the current block.
-	 */
-	bp = cur->bc_bufs[0];
-	block = XFS_BUF_TO_INOBT_BLOCK(bp);
-#ifdef DEBUG
-	if ((error = xfs_btree_check_sblock(cur, block, 0, bp)))
-		return error;
-#endif
-	/*
-	 * Get the address of the rec to be updated.
-	 */
-	ptr = cur->bc_ptrs[0];
-	rp = XFS_INOBT_REC_ADDR(block, ptr, cur);
-	/*
-	 * Fill in the new contents and log them.
-	 */
-	rp->ir_startino = cpu_to_be32(ino);
-	rp->ir_freecount = cpu_to_be32(fcnt);
-	rp->ir_free = cpu_to_be64(free);
-	xfs_inobt_log_recs(cur, bp, ptr, ptr);
-	/*
-	 * Updating first record in leaf. Pass new key value up to our parent.
-	 */
-	if (ptr == 1) {
-		xfs_inobt_key_t	key;	/* key containing [ino] */
-
-		key.ir_startino = cpu_to_be32(ino);
-		if ((error = xfs_inobt_updkey(cur, &key, 1)))
-			return error;
-	}
-	return 0;
+	if (leaf)
+		return blocklen / sizeof(xfs_inobt_rec_t);
+	return blocklen / (sizeof(xfs_inobt_key_t) + sizeof(xfs_inobt_ptr_t));
 }

fs/xfs/xfs_ialloc_btree.h

@@ -24,7 +24,6 @@
 
 struct xfs_buf;
 struct xfs_btree_cur;
-struct xfs_btree_sblock;
 struct xfs_mount;
 
 /*
@@ -70,11 +69,6 @@ typedef struct xfs_inobt_key {
 /* btree pointer type */
 typedef __be32 xfs_inobt_ptr_t;
 
-/* btree block header type */
-typedef	struct xfs_btree_sblock xfs_inobt_block_t;
-
-#define	XFS_BUF_TO_INOBT_BLOCK(bp)	((xfs_inobt_block_t *)XFS_BUF_PTR(bp))
-
 /*
  * Bit manipulations for ir_free.
  */
@@ -84,14 +78,6 @@ typedef	struct xfs_btree_sblock xfs_inobt_block_t;
 #define	XFS_INOBT_SET_FREE(rp,i)	((rp)->ir_free |= XFS_INOBT_MASK(i))
 #define	XFS_INOBT_CLR_FREE(rp,i)	((rp)->ir_free &= ~XFS_INOBT_MASK(i))
 
-/*
- * Real block structures have a size equal to the disk block size.
- */
-#define	XFS_INOBT_BLOCK_MAXRECS(lev,cur) ((cur)->bc_mp->m_inobt_mxr[lev != 0])
-#define	XFS_INOBT_BLOCK_MINRECS(lev,cur) ((cur)->bc_mp->m_inobt_mnr[lev != 0])
-#define	XFS_INOBT_IS_LAST_REC(cur)	\
-	((cur)->bc_ptrs[0] == be16_to_cpu(XFS_BUF_TO_INOBT_BLOCK((cur)->bc_bufs[0])->bb_numrecs))
-
 /*
  * Maximum number of inode btree levels.
  */
@@ -104,75 +90,38 @@ typedef	struct xfs_btree_sblock xfs_inobt_block_t;
 #define	XFS_PREALLOC_BLOCKS(mp)		((xfs_agblock_t)(XFS_IBT_BLOCK(mp) + 1))
 
 /*
- * Record, key, and pointer address macros for btree blocks.
- */
-#define XFS_INOBT_REC_ADDR(bb,i,cur) \
-	(XFS_BTREE_REC_ADDR(xfs_inobt, bb, i))
-
-#define	XFS_INOBT_KEY_ADDR(bb,i,cur) \
-	(XFS_BTREE_KEY_ADDR(xfs_inobt, bb, i))
-
-#define	XFS_INOBT_PTR_ADDR(bb,i,cur) \
-	(XFS_BTREE_PTR_ADDR(xfs_inobt, bb, \
-				i, XFS_INOBT_BLOCK_MAXRECS(1, cur)))
-
-/*
- * Decrement cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-extern int xfs_inobt_decrement(struct xfs_btree_cur *cur, int level, int *stat);
-
-/*
- * Delete the record pointed to by cur.
- * The cursor refers to the place where the record was (could be inserted)
- * when the operation returns.
- */
-extern int xfs_inobt_delete(struct xfs_btree_cur *cur, int *stat);
-
-/*
- * Get the data from the pointed-to record.
- */
-extern int xfs_inobt_get_rec(struct xfs_btree_cur *cur, xfs_agino_t *ino,
-			     __int32_t *fcnt, xfs_inofree_t *free, int *stat);
-
-/*
- * Increment cursor by one record at the level.
- * For nonzero levels the leaf-ward information is untouched.
- */
-extern int xfs_inobt_increment(struct xfs_btree_cur *cur, int level, int *stat);
-
-/*
- * Insert the current record at the point referenced by cur.
- * The cursor may be inconsistent on return if splits have been done.
- */
-extern int xfs_inobt_insert(struct xfs_btree_cur *cur, int *stat);
-
-/*
- * Lookup the record equal to ino in the btree given by cur.
- */
-extern int xfs_inobt_lookup_eq(struct xfs_btree_cur *cur, xfs_agino_t ino,
-				__int32_t fcnt, xfs_inofree_t free, int *stat);
-
-/*
- * Lookup the first record greater than or equal to ino
- * in the btree given by cur.
- */
-extern int xfs_inobt_lookup_ge(struct xfs_btree_cur *cur, xfs_agino_t ino,
-				__int32_t fcnt,	xfs_inofree_t free, int *stat);
-
-/*
- * Lookup the first record less than or equal to ino
- * in the btree given by cur.
+ * Btree block header size depends on a superblock flag.
+ *
+ * (not quite yet, but soon)
  */
-extern int xfs_inobt_lookup_le(struct xfs_btree_cur *cur, xfs_agino_t ino,
-				__int32_t fcnt,	xfs_inofree_t free, int *stat);
+#define XFS_INOBT_BLOCK_LEN(mp)	XFS_BTREE_SBLOCK_LEN
 
 /*
- * Update the record referred to by cur, to the value given
- * by [ino, fcnt, free].
- * This either works (return 0) or gets an EFSCORRUPTED error.
- */
-extern int xfs_inobt_update(struct xfs_btree_cur *cur, xfs_agino_t ino,
-				__int32_t fcnt, xfs_inofree_t free);
+ * Record, key, and pointer address macros for btree blocks.
+ *
+ * (note that some of these may appear unused, but they are used in userspace)
+ */
+#define XFS_INOBT_REC_ADDR(mp, block, index) \
+	((xfs_inobt_rec_t *) \
+		((char *)(block) + \
+		 XFS_INOBT_BLOCK_LEN(mp) + \
+		 (((index) - 1) * sizeof(xfs_inobt_rec_t))))
+
+#define XFS_INOBT_KEY_ADDR(mp, block, index) \
+	((xfs_inobt_key_t *) \
+		((char *)(block) + \
+		 XFS_INOBT_BLOCK_LEN(mp) + \
+		 ((index) - 1) * sizeof(xfs_inobt_key_t)))
+
+#define XFS_INOBT_PTR_ADDR(mp, block, index, maxrecs) \
+	((xfs_inobt_ptr_t *) \
+		((char *)(block) + \
+		 XFS_INOBT_BLOCK_LEN(mp) + \
+		 (maxrecs) * sizeof(xfs_inobt_key_t) + \
+		 ((index) - 1) * sizeof(xfs_inobt_ptr_t)))
+
+extern struct xfs_btree_cur *xfs_inobt_init_cursor(struct xfs_mount *,
+		struct xfs_trans *, struct xfs_buf *, xfs_agnumber_t);
+extern int xfs_inobt_maxrecs(struct xfs_mount *, int, int);
 
 #endif	/* __XFS_IALLOC_BTREE_H__ */

fs/xfs/xfs_iget.c

@@ -38,164 +38,122 @@
 #include "xfs_ialloc.h"
 #include "xfs_quota.h"
 #include "xfs_utils.h"
+#include "xfs_trans_priv.h"
+#include "xfs_inode_item.h"
 
 /*
- * Look up an inode by number in the given file system.
- * The inode is looked up in the cache held in each AG.
- * If the inode is found in the cache, attach it to the provided
- * vnode.
- *
- * If it is not in core, read it in from the file system's device,
- * add it to the cache and attach the provided vnode.
- *
- * The inode is locked according to the value of the lock_flags parameter.
- * This flag parameter indicates how and if the inode's IO lock and inode lock
- * should be taken.
- *
- * mp -- the mount point structure for the current file system.  It points
- *       to the inode hash table.
- * tp -- a pointer to the current transaction if there is one.  This is
- *       simply passed through to the xfs_iread() call.
- * ino -- the number of the inode desired.  This is the unique identifier
- *        within the file system for the inode being requested.
- * lock_flags -- flags indicating how to lock the inode.  See the comment
- *		 for xfs_ilock() for a list of valid values.
- * bno -- the block number starting the buffer containing the inode,
- *	  if known (as by bulkstat), else 0.
+ * Check the validity of the inode we just found it the cache
  */
-STATIC int
-xfs_iget_core(
-	struct inode	*inode,
-	xfs_mount_t	*mp,
-	xfs_trans_t	*tp,
-	xfs_ino_t	ino,
-	uint		flags,
-	uint		lock_flags,
-	xfs_inode_t	**ipp,
-	xfs_daddr_t	bno)
+static int
+xfs_iget_cache_hit(
+	struct xfs_perag	*pag,
+	struct xfs_inode	*ip,
+	int			flags,
+	int			lock_flags) __releases(pag->pag_ici_lock)
 {
-	struct inode	*old_inode;
-	xfs_inode_t	*ip;
-	xfs_inode_t	*iq;
-	int		error;
-	unsigned long	first_index, mask;
-	xfs_perag_t	*pag;
-	xfs_agino_t	agino;
+	struct xfs_mount	*mp = ip->i_mount;
+	int			error = EAGAIN;
 
-	/* the radix tree exists only in inode capable AGs */
-	if (XFS_INO_TO_AGNO(mp, ino) >= mp->m_maxagi)
-		return EINVAL;
-
-	/* get the perag structure and ensure that it's inode capable */
-	pag = xfs_get_perag(mp, ino);
-	if (!pag->pagi_inodeok)
-		return EINVAL;
-	ASSERT(pag->pag_ici_init);
-	agino = XFS_INO_TO_AGINO(mp, ino);
+	/*
+	 * If INEW is set this inode is being set up
+	 * If IRECLAIM is set this inode is being torn down
+	 * Pause and try again.
+	 */
+	if (xfs_iflags_test(ip, (XFS_INEW|XFS_IRECLAIM))) {
+		XFS_STATS_INC(xs_ig_frecycle);
+		goto out_error;
+	}
 
-again:
-	read_lock(&pag->pag_ici_lock);
-	ip = radix_tree_lookup(&pag->pag_ici_root, agino);
+	/* If IRECLAIMABLE is set, we've torn down the vfs inode part */
+	if (xfs_iflags_test(ip, XFS_IRECLAIMABLE)) {
 
-	if (ip != NULL) {
 		/*
-		 * If INEW is set this inode is being set up
-		 * we need to pause and try again.
+		 * If lookup is racing with unlink, then we should return an
+		 * error immediately so we don't remove it from the reclaim
+		 * list and potentially leak the inode.
 		 */
-		if (xfs_iflags_test(ip, XFS_INEW)) {
-			read_unlock(&pag->pag_ici_lock);
-			delay(1);
-			XFS_STATS_INC(xs_ig_frecycle);
-
-			goto again;
+		if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) {
+			error = ENOENT;
+			goto out_error;
 		}
 
-		old_inode = ip->i_vnode;
-		if (old_inode == NULL) {
-			/*
-			 * If IRECLAIM is set this inode is
-			 * on its way out of the system,
-			 * we need to pause and try again.
-			 */
-			if (xfs_iflags_test(ip, XFS_IRECLAIM)) {
-				read_unlock(&pag->pag_ici_lock);
-				delay(1);
-				XFS_STATS_INC(xs_ig_frecycle);
-
-				goto again;
-			}
-			ASSERT(xfs_iflags_test(ip, XFS_IRECLAIMABLE));
-
-			/*
-			 * If lookup is racing with unlink, then we
-			 * should return an error immediately so we
-			 * don't remove it from the reclaim list and
-			 * potentially leak the inode.
-			 */
-			if ((ip->i_d.di_mode == 0) &&
-			    !(flags & XFS_IGET_CREATE)) {
-				read_unlock(&pag->pag_ici_lock);
-				xfs_put_perag(mp, pag);
-				return ENOENT;
-			}
-
-			xfs_itrace_exit_tag(ip, "xfs_iget.alloc");
-
-			XFS_STATS_INC(xs_ig_found);
-			xfs_iflags_clear(ip, XFS_IRECLAIMABLE);
-			read_unlock(&pag->pag_ici_lock);
-
-			XFS_MOUNT_ILOCK(mp);
-			list_del_init(&ip->i_reclaim);
-			XFS_MOUNT_IUNLOCK(mp);
-
-			goto finish_inode;
-
-		} else if (inode != old_inode) {
-			/* The inode is being torn down, pause and
-			 * try again.
-			 */
-			if (old_inode->i_state & (I_FREEING | I_CLEAR)) {
-				read_unlock(&pag->pag_ici_lock);
-				delay(1);
-				XFS_STATS_INC(xs_ig_frecycle);
-
-				goto again;
-			}
-/* Chances are the other vnode (the one in the inode) is being torn
-* down right now, and we landed on top of it. Question is, what do
-* we do? Unhook the old inode and hook up the new one?
-*/
-			cmn_err(CE_PANIC,
-		"xfs_iget_core: ambiguous vns: vp/0x%p, invp/0x%p",
-					old_inode, inode);
-		}
+		xfs_itrace_exit_tag(ip, "xfs_iget.alloc");
 
 		/*
-		 * Inode cache hit
+		 * We need to re-initialise the VFS inode as it has been
+		 * 'freed' by the VFS. Do this here so we can deal with
+		 * errors cleanly, then tag it so it can be set up correctly
+		 * later.
 		 */
-		read_unlock(&pag->pag_ici_lock);
-		XFS_STATS_INC(xs_ig_found);
-
-finish_inode:
-		if (ip->i_d.di_mode == 0 && !(flags & XFS_IGET_CREATE)) {
-			xfs_put_perag(mp, pag);
-			return ENOENT;
+		if (!inode_init_always(mp->m_super, VFS_I(ip))) {
+			error = ENOMEM;
+			goto out_error;
 		}
 
-		if (lock_flags != 0)
-			xfs_ilock(ip, lock_flags);
+		/*
+		 * We must set the XFS_INEW flag before clearing the
+		 * XFS_IRECLAIMABLE flag so that if a racing lookup does
+		 * not find the XFS_IRECLAIMABLE above but has the igrab()
+		 * below succeed we can safely check XFS_INEW to detect
+		 * that this inode is still being initialised.
+		 */
+		xfs_iflags_set(ip, XFS_INEW);
+		xfs_iflags_clear(ip, XFS_IRECLAIMABLE);
+
+		/* clear the radix tree reclaim flag as well. */
+		__xfs_inode_clear_reclaim_tag(mp, pag, ip);
+	} else if (!igrab(VFS_I(ip))) {
+		/* If the VFS inode is being torn down, pause and try again. */
+		XFS_STATS_INC(xs_ig_frecycle);
+		goto out_error;
+	} else if (xfs_iflags_test(ip, XFS_INEW)) {
+		/*
+		 * We are racing with another cache hit that is
+		 * currently recycling this inode out of the XFS_IRECLAIMABLE
+		 * state. Wait for the initialisation to complete before
+		 * continuing.
+		 */
+		wait_on_inode(VFS_I(ip));
+	}
 
-		xfs_iflags_clear(ip, XFS_ISTALE);
-		xfs_itrace_exit_tag(ip, "xfs_iget.found");
-		goto return_ip;
+	if (ip->i_d.di_mode == 0 && !(flags & XFS_IGET_CREATE)) {
+		error = ENOENT;
+		iput(VFS_I(ip));
+		goto out_error;
 	}
 
-	/*
-	 * Inode cache miss
-	 */
+	/* We've got a live one. */
+	read_unlock(&pag->pag_ici_lock);
+
+	if (lock_flags != 0)
+		xfs_ilock(ip, lock_flags);
+
+	xfs_iflags_clear(ip, XFS_ISTALE);
+	xfs_itrace_exit_tag(ip, "xfs_iget.found");
+	XFS_STATS_INC(xs_ig_found);
+	return 0;
+
+out_error:
 	read_unlock(&pag->pag_ici_lock);
-	XFS_STATS_INC(xs_ig_missed);
+	return error;
+}
+
+
+static int
+xfs_iget_cache_miss(
+	struct xfs_mount	*mp,
+	struct xfs_perag	*pag,
+	xfs_trans_t		*tp,
+	xfs_ino_t		ino,
+	struct xfs_inode	**ipp,
+	xfs_daddr_t		bno,
+	int			flags,
+	int			lock_flags) __releases(pag->pag_ici_lock)
+{
+	struct xfs_inode	*ip;
+	int			error;
+	unsigned long		first_index, mask;
+	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, ino);
 
 	/*
 	 * Read the disk inode attributes into a new inode structure and get
@@ -203,116 +161,85 @@ finish_inode:
 	 */
 	error = xfs_iread(mp, tp, ino, &ip, bno,
 			  (flags & XFS_IGET_BULKSTAT) ? XFS_IMAP_BULKSTAT : 0);
-	if (error) {
-		xfs_put_perag(mp, pag);
+	if (error)
 		return error;
-	}
 
 	xfs_itrace_exit_tag(ip, "xfs_iget.alloc");
 
-
-	mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
-		     "xfsino", ip->i_ino);
-	mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
-	init_waitqueue_head(&ip->i_ipin_wait);
-	atomic_set(&ip->i_pincount, 0);
-
-	/*
-	 * Because we want to use a counting completion, complete
-	 * the flush completion once to allow a single access to
-	 * the flush completion without blocking.
-	 */
-	init_completion(&ip->i_flush);
-	complete(&ip->i_flush);
+	if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) {
+		error = ENOENT;
+		goto out_destroy;
+	}
 
 	if (lock_flags)
 		xfs_ilock(ip, lock_flags);
 
-	if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) {
-		xfs_idestroy(ip);
-		xfs_put_perag(mp, pag);
-		return ENOENT;
-	}
-
 	/*
 	 * Preload the radix tree so we can insert safely under the
-	 * write spinlock.
+	 * write spinlock. Note that we cannot sleep inside the preload
+	 * region.
 	 */
 	if (radix_tree_preload(GFP_KERNEL)) {
-		xfs_idestroy(ip);
-		delay(1);
-		goto again;
+		error = EAGAIN;
+		goto out_unlock;
 	}
+
 	mask = ~(((XFS_INODE_CLUSTER_SIZE(mp) >> mp->m_sb.sb_inodelog)) - 1);
 	first_index = agino & mask;
 	write_lock(&pag->pag_ici_lock);
-	/*
-	 * insert the new inode
-	 */
+
+	/* insert the new inode */
 	error = radix_tree_insert(&pag->pag_ici_root, agino, ip);
 	if (unlikely(error)) {
-		BUG_ON(error != -EEXIST);
-		write_unlock(&pag->pag_ici_lock);
-		radix_tree_preload_end();
-		xfs_idestroy(ip);
+		WARN_ON(error != -EEXIST);
 		XFS_STATS_INC(xs_ig_dup);
-		goto again;
+		error = EAGAIN;
+		goto out_preload_end;
 	}
 
-	/*
-	 * These values _must_ be set before releasing the radix tree lock!
-	 */
+	/* These values _must_ be set before releasing the radix tree lock! */
 	ip->i_udquot = ip->i_gdquot = NULL;
 	xfs_iflags_set(ip, XFS_INEW);
 
 	write_unlock(&pag->pag_ici_lock);
 	radix_tree_preload_end();
-
-	/*
-	 * Link ip to its mount and thread it on the mount's inode list.
-	 */
-	XFS_MOUNT_ILOCK(mp);
-	if ((iq = mp->m_inodes)) {
-		ASSERT(iq->i_mprev->i_mnext == iq);
-		ip->i_mprev = iq->i_mprev;
-		iq->i_mprev->i_mnext = ip;
-		iq->i_mprev = ip;
-		ip->i_mnext = iq;
-	} else {
-		ip->i_mnext = ip;
-		ip->i_mprev = ip;
-	}
-	mp->m_inodes = ip;
-
-	XFS_MOUNT_IUNLOCK(mp);
-	xfs_put_perag(mp, pag);
-
- return_ip:
-	ASSERT(ip->i_df.if_ext_max ==
-	       XFS_IFORK_DSIZE(ip) / sizeof(xfs_bmbt_rec_t));
-
-	xfs_iflags_set(ip, XFS_IMODIFIED);
 	*ipp = ip;
-
-	/*
-	 * Set up the Linux with the Linux inode.
-	 */
-	ip->i_vnode = inode;
-	inode->i_private = ip;
-
-	/*
-	 * If we have a real type for an on-disk inode, we can set ops(&unlock)
-	 * now.	 If it's a new inode being created, xfs_ialloc will handle it.
-	 */
-	if (ip->i_d.di_mode != 0)
-		xfs_setup_inode(ip);
 	return 0;
-}
 
+out_preload_end:
+	write_unlock(&pag->pag_ici_lock);
+	radix_tree_preload_end();
+out_unlock:
+	if (lock_flags)
+		xfs_iunlock(ip, lock_flags);
+out_destroy:
+	xfs_destroy_inode(ip);
+	return error;
+}
 
 /*
- * The 'normal' internal xfs_iget, if needed it will
- * 'allocate', or 'get', the vnode.
+ * Look up an inode by number in the given file system.
+ * The inode is looked up in the cache held in each AG.
+ * If the inode is found in the cache, initialise the vfs inode
+ * if necessary.
+ *
+ * If it is not in core, read it in from the file system's device,
+ * add it to the cache and initialise the vfs inode.
+ *
+ * The inode is locked according to the value of the lock_flags parameter.
+ * This flag parameter indicates how and if the inode's IO lock and inode lock
+ * should be taken.
+ *
+ * mp -- the mount point structure for the current file system.  It points
+ *       to the inode hash table.
+ * tp -- a pointer to the current transaction if there is one.  This is
+ *       simply passed through to the xfs_iread() call.
+ * ino -- the number of the inode desired.  This is the unique identifier
+ *        within the file system for the inode being requested.
+ * lock_flags -- flags indicating how to lock the inode.  See the comment
+ *		 for xfs_ilock() for a list of valid values.
+ * bno -- the block number starting the buffer containing the inode,
+ *	  if known (as by bulkstat), else 0.
  */
 int
 xfs_iget(
@@ -324,61 +251,65 @@ xfs_iget(
 	xfs_inode_t	**ipp,
 	xfs_daddr_t	bno)
 {
-	struct inode	*inode;
 	xfs_inode_t	*ip;
 	int		error;
+	xfs_perag_t	*pag;
+	xfs_agino_t	agino;
 
-	XFS_STATS_INC(xs_ig_attempts);
-
-retry:
-	inode = iget_locked(mp->m_super, ino);
-	if (!inode)
-		/* If we got no inode we are out of memory */
-		return ENOMEM;
-
-	if (inode->i_state & I_NEW) {
-		XFS_STATS_INC(vn_active);
-		XFS_STATS_INC(vn_alloc);
-
-		error = xfs_iget_core(inode, mp, tp, ino, flags,
-				lock_flags, ipp, bno);
-		if (error) {
-			make_bad_inode(inode);
-			if (inode->i_state & I_NEW)
-				unlock_new_inode(inode);
-			iput(inode);
-		}
-		return error;
+	/* the radix tree exists only in inode capable AGs */
+	if (XFS_INO_TO_AGNO(mp, ino) >= mp->m_maxagi)
+		return EINVAL;
+
+	/* get the perag structure and ensure that it's inode capable */
+	pag = xfs_get_perag(mp, ino);
+	if (!pag->pagi_inodeok)
+		return EINVAL;
+	ASSERT(pag->pag_ici_init);
+	agino = XFS_INO_TO_AGINO(mp, ino);
+
+again:
+	error = 0;
+	read_lock(&pag->pag_ici_lock);
+	ip = radix_tree_lookup(&pag->pag_ici_root, agino);
+
+	if (ip) {
+		error = xfs_iget_cache_hit(pag, ip, flags, lock_flags);
+		if (error)
+			goto out_error_or_again;
+	} else {
+		read_unlock(&pag->pag_ici_lock);
+		XFS_STATS_INC(xs_ig_missed);
+
+		error = xfs_iget_cache_miss(mp, pag, tp, ino, &ip, bno,
+							flags, lock_flags);
+		if (error)
+			goto out_error_or_again;
 	}
+	xfs_put_perag(mp, pag);
+
+	xfs_iflags_set(ip, XFS_IMODIFIED);
+	*ipp = ip;
 
+	ASSERT(ip->i_df.if_ext_max ==
+	       XFS_IFORK_DSIZE(ip) / sizeof(xfs_bmbt_rec_t));
 	/*
-	 * If the inode is not fully constructed due to
-	 * filehandle mismatches wait for the inode to go
-	 * away and try again.
-	 *
-	 * iget_locked will call __wait_on_freeing_inode
-	 * to wait for the inode to go away.
+	 * If we have a real type for an on-disk inode, we can set ops(&unlock)
+	 * now.	 If it's a new inode being created, xfs_ialloc will handle it.
 	 */
-	if (is_bad_inode(inode)) {
-		iput(inode);
-		delay(1);
-		goto retry;
-	}
+	if (xfs_iflags_test(ip, XFS_INEW) && ip->i_d.di_mode != 0)
+		xfs_setup_inode(ip);
+	return 0;
 
-	ip = XFS_I(inode);
-	if (!ip) {
-		iput(inode);
+out_error_or_again:
+	if (error == EAGAIN) {
 		delay(1);
-		goto retry;
+		goto again;
 	}
-
-	if (lock_flags != 0)
-		xfs_ilock(ip, lock_flags);
-	XFS_STATS_INC(xs_ig_found);
-	*ipp = ip;
-	return 0;
+	xfs_put_perag(mp, pag);
+	return error;
 }
 
+
 /*
  * Look for the inode corresponding to the given ino in the hash table.
  * If it is there and its i_transp pointer matches tp, return it.
@@ -462,14 +393,13 @@ xfs_ireclaim(xfs_inode_t *ip)
 	xfs_iextract(ip);
 
 	/*
-	 * Here we do a spurious inode lock in order to coordinate with
-	 * xfs_sync().  This is because xfs_sync() references the inodes
-	 * in the mount list without taking references on the corresponding
-	 * vnodes.  We make that OK here by ensuring that we wait until
-	 * the inode is unlocked in xfs_sync() before we go ahead and
-	 * free it.  We get both the regular lock and the io lock because
-	 * the xfs_sync() code may need to drop the regular one but will
-	 * still hold the io lock.
+	 * Here we do a spurious inode lock in order to coordinate with inode
+	 * cache radix tree lookups.  This is because the lookup can reference
+	 * the inodes in the cache without taking references.  We make that OK
+	 * here by ensuring that we wait until the inode is unlocked after the
+	 * lookup before we go ahead and free it.  We get both the ilock and
+	 * the iolock because the code may need to drop the ilock one but will
+	 * still hold the iolock.
 	 */
 	xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
 
@@ -479,14 +409,6 @@ xfs_ireclaim(xfs_inode_t *ip)
 	 */
 	XFS_QM_DQDETACH(ip->i_mount, ip);
 
-	/*
-	 * Pull our behavior descriptor from the vnode chain.
-	 */
-	if (ip->i_vnode) {
-		ip->i_vnode->i_private = NULL;
-		ip->i_vnode = NULL;
-	}
-
 	/*
 	 * Free all memory associated with the inode.
 	 */
@@ -505,38 +427,13 @@ xfs_iextract(
 {
 	xfs_mount_t	*mp = ip->i_mount;
 	xfs_perag_t	*pag = xfs_get_perag(mp, ip->i_ino);
-	xfs_inode_t	*iq;
 
 	write_lock(&pag->pag_ici_lock);
 	radix_tree_delete(&pag->pag_ici_root, XFS_INO_TO_AGINO(mp, ip->i_ino));
 	write_unlock(&pag->pag_ici_lock);
 	xfs_put_perag(mp, pag);
 
-	/*
-	 * Remove from mount's inode list.
-	 */
-	XFS_MOUNT_ILOCK(mp);
-	ASSERT((ip->i_mnext != NULL) && (ip->i_mprev != NULL));
-	iq = ip->i_mnext;
-	iq->i_mprev = ip->i_mprev;
-	ip->i_mprev->i_mnext = iq;
-
-	/*
-	 * Fix up the head pointer if it points to the inode being deleted.
-	 */
-	if (mp->m_inodes == ip) {
-		if (ip == iq) {
-			mp->m_inodes = NULL;
-		} else {
-			mp->m_inodes = iq;
-		}
-	}
-
-	/* Deal with the deleted inodes list */
-	list_del_init(&ip->i_reclaim);
-
 	mp->m_ireclaims++;
-	XFS_MOUNT_IUNLOCK(mp);
 }
 
 /*
@@ -737,7 +634,7 @@ xfs_iunlock(
 		 * it is in the AIL and anyone is waiting on it.  Don't do
 		 * this if the caller has asked us not to.
 		 */
-		xfs_trans_unlocked_item(ip->i_mount,
+		xfs_trans_unlocked_item(ip->i_itemp->ili_item.li_ailp,
 					(xfs_log_item_t*)(ip->i_itemp));
 	}
 	xfs_ilock_trace(ip, 3, lock_flags, (inst_t *)__return_address);

fs/xfs/xfs_imap.h

@@ -30,11 +30,9 @@ typedef struct xfs_imap {
 	ushort		im_boffset;	/* inode offset in block in bytes */
 } xfs_imap_t;
 
-#ifdef __KERNEL__
 struct xfs_mount;
 struct xfs_trans;
 int	xfs_imap(struct xfs_mount *, struct xfs_trans *, xfs_ino_t,
 		 xfs_imap_t *, uint);
-#endif
 
 #endif	/* __XFS_IMAP_H__ */

fs/xfs/xfs_inode.c

@@ -41,6 +41,7 @@
 #include "xfs_buf_item.h"
 #include "xfs_inode_item.h"
 #include "xfs_btree.h"
+#include "xfs_btree_trace.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
 #include "xfs_bmap.h"
@@ -221,25 +222,26 @@ xfs_imap_to_bp(
  * Use xfs_imap() to determine the size and location of the
  * buffer to read from disk.
  */
-STATIC int
+int
 xfs_inotobp(
 	xfs_mount_t	*mp,
 	xfs_trans_t	*tp,
 	xfs_ino_t	ino,
 	xfs_dinode_t	**dipp,
 	xfs_buf_t	**bpp,
-	int		*offset)
+	int		*offset,
+	uint		imap_flags)
 {
 	xfs_imap_t	imap;
 	xfs_buf_t	*bp;
 	int		error;
 
 	imap.im_blkno = 0;
-	error = xfs_imap(mp, tp, ino, &imap, XFS_IMAP_LOOKUP);
+	error = xfs_imap(mp, tp, ino, &imap, imap_flags | XFS_IMAP_LOOKUP);
 	if (error)
 		return error;
 
-	error = xfs_imap_to_bp(mp, tp, &imap, &bp, XFS_BUF_LOCK, 0);
+	error = xfs_imap_to_bp(mp, tp, &imap, &bp, XFS_BUF_LOCK, imap_flags);
 	if (error)
 		return error;
 
@@ -621,7 +623,7 @@ xfs_iformat_btree(
 	ifp = XFS_IFORK_PTR(ip, whichfork);
 	dfp = (xfs_bmdr_block_t *)XFS_DFORK_PTR(dip, whichfork);
 	size = XFS_BMAP_BROOT_SPACE(dfp);
-	nrecs = XFS_BMAP_BROOT_NUMRECS(dfp);
+	nrecs = be16_to_cpu(dfp->bb_numrecs);
 
 	/*
 	 * blow out if -- fork has less extents than can fit in
@@ -649,8 +651,9 @@ xfs_iformat_btree(
 	 * Copy and convert from the on-disk structure
 	 * to the in-memory structure.
 	 */
-	xfs_bmdr_to_bmbt(dfp, XFS_DFORK_SIZE(dip, ip->i_mount, whichfork),
-		ifp->if_broot, size);
+	xfs_bmdr_to_bmbt(ip->i_mount, dfp,
+			 XFS_DFORK_SIZE(dip, ip->i_mount, whichfork),
+			 ifp->if_broot, size);
 	ifp->if_flags &= ~XFS_IFEXTENTS;
 	ifp->if_flags |= XFS_IFBROOT;
 
@@ -788,51 +791,56 @@ xfs_dic2xflags(
 }
 
 /*
- * Given a mount structure and an inode number, return a pointer
- * to a newly allocated in-core inode corresponding to the given
- * inode number.
- *
- * Initialize the inode's attributes and extent pointers if it
- * already has them (it will not if the inode has no links).
+ * Allocate and initialise an xfs_inode.
  */
-int
-xfs_iread(
-	xfs_mount_t	*mp,
-	xfs_trans_t	*tp,
-	xfs_ino_t	ino,
-	xfs_inode_t	**ipp,
-	xfs_daddr_t	bno,
-	uint		imap_flags)
+STATIC struct xfs_inode *
+xfs_inode_alloc(
+	struct xfs_mount	*mp,
+	xfs_ino_t		ino)
 {
-	xfs_buf_t	*bp;
-	xfs_dinode_t	*dip;
-	xfs_inode_t	*ip;
-	int		error;
+	struct xfs_inode	*ip;
 
-	ASSERT(xfs_inode_zone != NULL);
+	/*
+	 * if this didn't occur in transactions, we could use
+	 * KM_MAYFAIL and return NULL here on ENOMEM. Set the
+	 * code up to do this anyway.
+	 */
+	ip = kmem_zone_alloc(xfs_inode_zone, KM_SLEEP);
+	if (!ip)
+		return NULL;
 
-	ip = kmem_zone_zalloc(xfs_inode_zone, KM_SLEEP);
-	ip->i_ino = ino;
-	ip->i_mount = mp;
-	atomic_set(&ip->i_iocount, 0);
-	spin_lock_init(&ip->i_flags_lock);
+	ASSERT(atomic_read(&ip->i_iocount) == 0);
+	ASSERT(atomic_read(&ip->i_pincount) == 0);
+	ASSERT(!spin_is_locked(&ip->i_flags_lock));
+	ASSERT(completion_done(&ip->i_flush));
 
 	/*
-	 * Get pointer's to the on-disk inode and the buffer containing it.
-	 * If the inode number refers to a block outside the file system
-	 * then xfs_itobp() will return NULL.  In this case we should
-	 * return NULL as well.  Set i_blkno to 0 so that xfs_itobp() will
-	 * know that this is a new incore inode.
+	 * initialise the VFS inode here to get failures
+	 * out of the way early.
 	 */
-	error = xfs_itobp(mp, tp, ip, &dip, &bp, bno, imap_flags, XFS_BUF_LOCK);
-	if (error) {
+	if (!inode_init_always(mp->m_super, VFS_I(ip))) {
 		kmem_zone_free(xfs_inode_zone, ip);
-		return error;
+		return NULL;
 	}
 
+	/* initialise the xfs inode */
+	ip->i_ino = ino;
+	ip->i_mount = mp;
+	ip->i_blkno = 0;
+	ip->i_len = 0;
+	ip->i_boffset =0;
+	ip->i_afp = NULL;
+	memset(&ip->i_df, 0, sizeof(xfs_ifork_t));
+	ip->i_flags = 0;
+	ip->i_update_core = 0;
+	ip->i_update_size = 0;
+	ip->i_delayed_blks = 0;
+	memset(&ip->i_d, 0, sizeof(xfs_icdinode_t));
+	ip->i_size = 0;
+	ip->i_new_size = 0;
+
 	/*
 	 * Initialize inode's trace buffers.
-	 * Do this before xfs_iformat in case it adds entries.
 	 */
 #ifdef	XFS_INODE_TRACE
 	ip->i_trace = ktrace_alloc(INODE_TRACE_SIZE, KM_NOFS);
@@ -840,7 +848,7 @@ xfs_iread(
 #ifdef XFS_BMAP_TRACE
 	ip->i_xtrace = ktrace_alloc(XFS_BMAP_KTRACE_SIZE, KM_NOFS);
 #endif
-#ifdef XFS_BMBT_TRACE
+#ifdef XFS_BTREE_TRACE
 	ip->i_btrace = ktrace_alloc(XFS_BMBT_KTRACE_SIZE, KM_NOFS);
 #endif
 #ifdef XFS_RW_TRACE
@@ -853,13 +861,51 @@ xfs_iread(
 	ip->i_dir_trace = ktrace_alloc(XFS_DIR2_KTRACE_SIZE, KM_NOFS);
 #endif
 
+	return ip;
+}
+
+/*
+ * Given a mount structure and an inode number, return a pointer
+ * to a newly allocated in-core inode corresponding to the given
+ * inode number.
+ *
+ * Initialize the inode's attributes and extent pointers if it
+ * already has them (it will not if the inode has no links).
+ */
+int
+xfs_iread(
+	xfs_mount_t	*mp,
+	xfs_trans_t	*tp,
+	xfs_ino_t	ino,
+	xfs_inode_t	**ipp,
+	xfs_daddr_t	bno,
+	uint		imap_flags)
+{
+	xfs_buf_t	*bp;
+	xfs_dinode_t	*dip;
+	xfs_inode_t	*ip;
+	int		error;
+
+	ip = xfs_inode_alloc(mp, ino);
+	if (!ip)
+		return ENOMEM;
+
+	/*
+	 * Get pointer's to the on-disk inode and the buffer containing it.
+	 * If the inode number refers to a block outside the file system
+	 * then xfs_itobp() will return NULL.  In this case we should
+	 * return NULL as well.  Set i_blkno to 0 so that xfs_itobp() will
+	 * know that this is a new incore inode.
+	 */
+	error = xfs_itobp(mp, tp, ip, &dip, &bp, bno, imap_flags, XFS_BUF_LOCK);
+	if (error)
+		goto out_destroy_inode;
+
 	/*
 	 * If we got something that isn't an inode it means someone
 	 * (nfs or dmi) has a stale handle.
 	 */
 	if (be16_to_cpu(dip->di_core.di_magic) != XFS_DINODE_MAGIC) {
-		kmem_zone_free(xfs_inode_zone, ip);
-		xfs_trans_brelse(tp, bp);
 #ifdef DEBUG
 		xfs_fs_cmn_err(CE_ALERT, mp, "xfs_iread: "
 				"dip->di_core.di_magic (0x%x) != "
@@ -867,7 +913,8 @@ xfs_iread(
 				be16_to_cpu(dip->di_core.di_magic),
 				XFS_DINODE_MAGIC);
 #endif /* DEBUG */
-		return XFS_ERROR(EINVAL);
+		error = XFS_ERROR(EINVAL);
+		goto out_brelse;
 	}
 
 	/*
@@ -881,14 +928,12 @@ xfs_iread(
 		xfs_dinode_from_disk(&ip->i_d, &dip->di_core);
 		error = xfs_iformat(ip, dip);
 		if (error)  {
-			kmem_zone_free(xfs_inode_zone, ip);
-			xfs_trans_brelse(tp, bp);
 #ifdef DEBUG
 			xfs_fs_cmn_err(CE_ALERT, mp, "xfs_iread: "
 					"xfs_iformat() returned error %d",
 					error);
 #endif /* DEBUG */
-			return error;
+			goto out_brelse;
 		}
 	} else {
 		ip->i_d.di_magic = be16_to_cpu(dip->di_core.di_magic);
@@ -911,8 +956,6 @@ xfs_iread(
 			XFS_IFORK_DSIZE(ip) / (uint)sizeof(xfs_bmbt_rec_t);
 	}
 
-	INIT_LIST_HEAD(&ip->i_reclaim);
-
 	/*
 	 * The inode format changed when we moved the link count and
 	 * made it 32 bits long.  If this is an old format inode,
@@ -956,6 +999,12 @@ xfs_iread(
 	xfs_trans_brelse(tp, bp);
 	*ipp = ip;
 	return 0;
+
+ out_brelse:
+	xfs_trans_brelse(tp, bp);
+ out_destroy_inode:
+	xfs_destroy_inode(ip);
+	return error;
 }
 
 /*
@@ -1049,6 +1098,7 @@ xfs_ialloc(
 	uint		flags;
 	int		error;
 	timespec_t	tv;
+	int		filestreams = 0;
 
 	/*
 	 * Call the space management code to pick
@@ -1056,9 +1106,8 @@ xfs_ialloc(
 	 */
 	error = xfs_dialloc(tp, pip ? pip->i_ino : 0, mode, okalloc,
 			    ialloc_context, call_again, &ino);
-	if (error != 0) {
+	if (error)
 		return error;
-	}
 	if (*call_again || ino == NULLFSINO) {
 		*ipp = NULL;
 		return 0;
@@ -1072,9 +1121,8 @@ xfs_ialloc(
 	 */
 	error = xfs_trans_iget(tp->t_mountp, tp, ino,
 				XFS_IGET_CREATE, XFS_ILOCK_EXCL, &ip);
-	if (error != 0) {
+	if (error)
 		return error;
-	}
 	ASSERT(ip != NULL);
 
 	ip->i_d.di_mode = (__uint16_t)mode;
@@ -1155,13 +1203,12 @@ xfs_ialloc(
 		flags |= XFS_ILOG_DEV;
 		break;
 	case S_IFREG:
-		if (pip && xfs_inode_is_filestream(pip)) {
-			error = xfs_filestream_associate(pip, ip);
-			if (error < 0)
-				return -error;
-			if (!error)
-				xfs_iflags_set(ip, XFS_IFILESTREAM);
-		}
+		/*
+		 * we can't set up filestreams until after the VFS inode
+		 * is set up properly.
+		 */
+		if (pip && xfs_inode_is_filestream(pip))
+			filestreams = 1;
 		/* fall through */
 	case S_IFDIR:
 		if (pip && (pip->i_d.di_flags & XFS_DIFLAG_ANY)) {
@@ -1227,6 +1274,15 @@ xfs_ialloc(
 	/* now that we have an i_mode we can setup inode ops and unlock */
 	xfs_setup_inode(ip);
 
+	/* now we have set up the vfs inode we can associate the filestream */
+	if (filestreams) {
+		error = xfs_filestream_associate(pip, ip);
+		if (error < 0)
+			return -error;
+		if (!error)
+			xfs_iflags_set(ip, XFS_IFILESTREAM);
+	}
+
 	*ipp = ip;
 	return 0;
 }
@@ -1414,7 +1470,7 @@ xfs_itruncate_start(
 	mp = ip->i_mount;
 
 	/* wait for the completion of any pending DIOs */
-	if (new_size < ip->i_size)
+	if (new_size == 0 || new_size < ip->i_size)
 		vn_iowait(ip);
 
 	/*
@@ -1992,7 +2048,7 @@ xfs_iunlink_remove(
 			}
 			next_ino = XFS_AGINO_TO_INO(mp, agno, next_agino);
 			error = xfs_inotobp(mp, tp, next_ino, &last_dip,
-					    &last_ibp, &last_offset);
+					    &last_ibp, &last_offset, 0);
 			if (error) {
 				cmn_err(CE_WARN,
 			"xfs_iunlink_remove: xfs_inotobp()  returned an error %d on %s.  Returning error.",
@@ -2160,9 +2216,9 @@ xfs_ifree_cluster(
 				iip = (xfs_inode_log_item_t *)lip;
 				ASSERT(iip->ili_logged == 1);
 				lip->li_cb = (void(*)(xfs_buf_t*,xfs_log_item_t*)) xfs_istale_done;
-				spin_lock(&mp->m_ail_lock);
-				iip->ili_flush_lsn = iip->ili_item.li_lsn;
-				spin_unlock(&mp->m_ail_lock);
+				xfs_trans_ail_copy_lsn(mp->m_ail,
+							&iip->ili_flush_lsn,
+							&iip->ili_item.li_lsn);
 				xfs_iflags_set(iip->ili_inode, XFS_ISTALE);
 				pre_flushed++;
 			}
@@ -2183,9 +2239,8 @@ xfs_ifree_cluster(
 			iip->ili_last_fields = iip->ili_format.ilf_fields;
 			iip->ili_format.ilf_fields = 0;
 			iip->ili_logged = 1;
-			spin_lock(&mp->m_ail_lock);
-			iip->ili_flush_lsn = iip->ili_item.li_lsn;
-			spin_unlock(&mp->m_ail_lock);
+			xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
+						&iip->ili_item.li_lsn);
 
 			xfs_buf_attach_iodone(bp,
 				(void(*)(xfs_buf_t*,xfs_log_item_t*))
@@ -2312,9 +2367,10 @@ xfs_iroot_realloc(
 	int			rec_diff,
 	int			whichfork)
 {
+	struct xfs_mount	*mp = ip->i_mount;
 	int			cur_max;
 	xfs_ifork_t		*ifp;
-	xfs_bmbt_block_t	*new_broot;
+	struct xfs_btree_block	*new_broot;
 	int			new_max;
 	size_t			new_size;
 	char			*np;
@@ -2335,8 +2391,7 @@ xfs_iroot_realloc(
 		 */
 		if (ifp->if_broot_bytes == 0) {
 			new_size = (size_t)XFS_BMAP_BROOT_SPACE_CALC(rec_diff);
-			ifp->if_broot = (xfs_bmbt_block_t*)kmem_alloc(new_size,
-								     KM_SLEEP);
+			ifp->if_broot = kmem_alloc(new_size, KM_SLEEP);
 			ifp->if_broot_bytes = (int)new_size;
 			return;
 		}
@@ -2347,18 +2402,16 @@ xfs_iroot_realloc(
 		 * location.  The records don't change location because
 		 * they are kept butted up against the btree block header.
 		 */
-		cur_max = XFS_BMAP_BROOT_MAXRECS(ifp->if_broot_bytes);
+		cur_max = xfs_bmbt_maxrecs(mp, ifp->if_broot_bytes, 0);
 		new_max = cur_max + rec_diff;
 		new_size = (size_t)XFS_BMAP_BROOT_SPACE_CALC(new_max);
-		ifp->if_broot = (xfs_bmbt_block_t *)
-		  kmem_realloc(ifp->if_broot,
-				new_size,
+		ifp->if_broot = kmem_realloc(ifp->if_broot, new_size,
 				(size_t)XFS_BMAP_BROOT_SPACE_CALC(cur_max), /* old size */
 				KM_SLEEP);
-		op = (char *)XFS_BMAP_BROOT_PTR_ADDR(ifp->if_broot, 1,
-						      ifp->if_broot_bytes);
-		np = (char *)XFS_BMAP_BROOT_PTR_ADDR(ifp->if_broot, 1,
-						      (int)new_size);
+		op = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
+						     ifp->if_broot_bytes);
+		np = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
+						     (int)new_size);
 		ifp->if_broot_bytes = (int)new_size;
 		ASSERT(ifp->if_broot_bytes <=
 			XFS_IFORK_SIZE(ip, whichfork) + XFS_BROOT_SIZE_ADJ);
@@ -2372,7 +2425,7 @@ xfs_iroot_realloc(
 	 * records, just get rid of the root and clear the status bit.
 	 */
 	ASSERT((ifp->if_broot != NULL) && (ifp->if_broot_bytes > 0));
-	cur_max = XFS_BMAP_BROOT_MAXRECS(ifp->if_broot_bytes);
+	cur_max = xfs_bmbt_maxrecs(mp, ifp->if_broot_bytes, 0);
 	new_max = cur_max + rec_diff;
 	ASSERT(new_max >= 0);
 	if (new_max > 0)
@@ -2380,11 +2433,11 @@ xfs_iroot_realloc(
 	else
 		new_size = 0;
 	if (new_size > 0) {
-		new_broot = (xfs_bmbt_block_t *)kmem_alloc(new_size, KM_SLEEP);
+		new_broot = kmem_alloc(new_size, KM_SLEEP);
 		/*
 		 * First copy over the btree block header.
 		 */
-		memcpy(new_broot, ifp->if_broot, sizeof(xfs_bmbt_block_t));
+		memcpy(new_broot, ifp->if_broot, XFS_BTREE_LBLOCK_LEN);
 	} else {
 		new_broot = NULL;
 		ifp->if_flags &= ~XFS_IFBROOT;
@@ -2397,18 +2450,16 @@ xfs_iroot_realloc(
 		/*
 		 * First copy the records.
 		 */
-		op = (char *)XFS_BMAP_BROOT_REC_ADDR(ifp->if_broot, 1,
-						     ifp->if_broot_bytes);
-		np = (char *)XFS_BMAP_BROOT_REC_ADDR(new_broot, 1,
-						     (int)new_size);
+		op = (char *)XFS_BMBT_REC_ADDR(mp, ifp->if_broot, 1);
+		np = (char *)XFS_BMBT_REC_ADDR(mp, new_broot, 1);
 		memcpy(np, op, new_max * (uint)sizeof(xfs_bmbt_rec_t));
 
 		/*
 		 * Then copy the pointers.
 		 */
-		op = (char *)XFS_BMAP_BROOT_PTR_ADDR(ifp->if_broot, 1,
+		op = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, ifp->if_broot, 1,
 						     ifp->if_broot_bytes);
-		np = (char *)XFS_BMAP_BROOT_PTR_ADDR(new_broot, 1,
+		np = (char *)XFS_BMAP_BROOT_PTR_ADDR(mp, new_broot, 1,
 						     (int)new_size);
 		memcpy(np, op, new_max * (uint)sizeof(xfs_dfsbno_t));
 	}
@@ -2617,6 +2668,10 @@ xfs_idestroy_fork(
  * It must free the inode itself and any buffers allocated for
  * if_extents/if_data and if_broot.  It must also free the lock
  * associated with the inode.
+ *
+ * Note: because we don't initialise everything on reallocation out
+ * of the zone, we must ensure we nullify everything correctly before
+ * freeing the structure.
  */
 void
 xfs_idestroy(
@@ -2631,8 +2686,6 @@ xfs_idestroy(
 	}
 	if (ip->i_afp)
 		xfs_idestroy_fork(ip, XFS_ATTR_FORK);
-	mrfree(&ip->i_lock);
-	mrfree(&ip->i_iolock);
 
 #ifdef XFS_INODE_TRACE
 	ktrace_free(ip->i_trace);
@@ -2640,7 +2693,7 @@ xfs_idestroy(
 #ifdef XFS_BMAP_TRACE
 	ktrace_free(ip->i_xtrace);
 #endif
-#ifdef XFS_BMBT_TRACE
+#ifdef XFS_BTREE_TRACE
 	ktrace_free(ip->i_btrace);
 #endif
 #ifdef XFS_RW_TRACE
@@ -2658,20 +2711,26 @@ xfs_idestroy(
 		 * inode still in the AIL. If it is there, we should remove
 		 * it to prevent a use-after-free from occurring.
 		 */
-		xfs_mount_t	*mp = ip->i_mount;
 		xfs_log_item_t	*lip = &ip->i_itemp->ili_item;
+		struct xfs_ail	*ailp = lip->li_ailp;
 
 		ASSERT(((lip->li_flags & XFS_LI_IN_AIL) == 0) ||
 				       XFS_FORCED_SHUTDOWN(ip->i_mount));
 		if (lip->li_flags & XFS_LI_IN_AIL) {
-			spin_lock(&mp->m_ail_lock);
+			spin_lock(&ailp->xa_lock);
 			if (lip->li_flags & XFS_LI_IN_AIL)
-				xfs_trans_delete_ail(mp, lip);
+				xfs_trans_ail_delete(ailp, lip);
 			else
-				spin_unlock(&mp->m_ail_lock);
+				spin_unlock(&ailp->xa_lock);
 		}
 		xfs_inode_item_destroy(ip);
+		ip->i_itemp = NULL;
 	}
+	/* asserts to verify all state is correct here */
+	ASSERT(atomic_read(&ip->i_iocount) == 0);
+	ASSERT(atomic_read(&ip->i_pincount) == 0);
+	ASSERT(!spin_is_locked(&ip->i_flags_lock));
+	ASSERT(completion_done(&ip->i_flush));
 	kmem_zone_free(xfs_inode_zone, ip);
 }
 
@@ -2880,7 +2939,7 @@ xfs_iflush_fork(
 			ASSERT(ifp->if_broot_bytes <=
 			       (XFS_IFORK_SIZE(ip, whichfork) +
 				XFS_BROOT_SIZE_ADJ));
-			xfs_bmbt_to_bmdr(ifp->if_broot, ifp->if_broot_bytes,
+			xfs_bmbt_to_bmdr(mp, ifp->if_broot, ifp->if_broot_bytes,
 				(xfs_bmdr_block_t *)cp,
 				XFS_DFORK_SIZE(dip, mp, whichfork));
 		}
@@ -3418,10 +3477,8 @@ xfs_iflush_int(
 		iip->ili_format.ilf_fields = 0;
 		iip->ili_logged = 1;
 
-		ASSERT(sizeof(xfs_lsn_t) == 8);	/* don't lock if it shrinks */
-		spin_lock(&mp->m_ail_lock);
-		iip->ili_flush_lsn = iip->ili_item.li_lsn;
-		spin_unlock(&mp->m_ail_lock);
+		xfs_trans_ail_copy_lsn(mp->m_ail, &iip->ili_flush_lsn,
+					&iip->ili_item.li_lsn);
 
 		/*
 		 * Attach the function xfs_iflush_done to the inode's
@@ -3459,41 +3516,6 @@ corrupt_out:
 }
 
 
-/*
- * Flush all inactive inodes in mp.
- */
-void
-xfs_iflush_all(
-	xfs_mount_t	*mp)
-{
-	xfs_inode_t	*ip;
-
- again:
-	XFS_MOUNT_ILOCK(mp);
-	ip = mp->m_inodes;
-	if (ip == NULL)
-		goto out;
-
-	do {
-		/* Make sure we skip markers inserted by sync */
-		if (ip->i_mount == NULL) {
-			ip = ip->i_mnext;
-			continue;
-		}
-
-		if (!VFS_I(ip)) {
-			XFS_MOUNT_IUNLOCK(mp);
-			xfs_finish_reclaim(ip, 0, XFS_IFLUSH_ASYNC);
-			goto again;
-		}
-
-		ASSERT(vn_count(VFS_I(ip)) == 0);
-
-		ip = ip->i_mnext;
-	} while (ip != mp->m_inodes);
- out:
-	XFS_MOUNT_IUNLOCK(mp);
-}
 
 #ifdef XFS_ILOCK_TRACE
 ktrace_t	*xfs_ilock_trace_buf;

fs/xfs/xfs_inode.h

@@ -20,7 +20,7 @@
 
 struct xfs_dinode;
 struct xfs_dinode_core;
-
+struct xfs_inode;
 
 /*
  * Fork identifiers.
@@ -63,7 +63,7 @@ typedef struct xfs_ext_irec {
 typedef struct xfs_ifork {
 	int			if_bytes;	/* bytes in if_u1 */
 	int			if_real_bytes;	/* bytes allocated in if_u1 */
-	xfs_bmbt_block_t	*if_broot;	/* file's incore btree root */
+	struct xfs_btree_block	*if_broot;	/* file's incore btree root */
 	short			if_broot_bytes;	/* bytes allocated for root */
 	unsigned char		if_flags;	/* per-fork flags */
 	unsigned char		if_ext_max;	/* max # of extent records */
@@ -83,54 +83,6 @@ typedef struct xfs_ifork {
 	} if_u2;
 } xfs_ifork_t;
 
-/*
- * Flags for xfs_ichgtime().
- */
-#define	XFS_ICHGTIME_MOD	0x1	/* data fork modification timestamp */
-#define	XFS_ICHGTIME_CHG	0x2	/* inode field change timestamp */
-
-/*
- * Per-fork incore inode flags.
- */
-#define	XFS_IFINLINE	0x01	/* Inline data is read in */
-#define	XFS_IFEXTENTS	0x02	/* All extent pointers are read in */
-#define	XFS_IFBROOT	0x04	/* i_broot points to the bmap b-tree root */
-#define	XFS_IFEXTIREC	0x08	/* Indirection array of extent blocks */
-
-/*
- * Flags for xfs_itobp(), xfs_imap() and xfs_dilocate().
- */
-#define XFS_IMAP_LOOKUP		0x1
-#define XFS_IMAP_BULKSTAT	0x2
-
-#ifdef __KERNEL__
-struct bhv_desc;
-struct cred;
-struct ktrace;
-struct xfs_buf;
-struct xfs_bmap_free;
-struct xfs_bmbt_irec;
-struct xfs_bmbt_block;
-struct xfs_inode;
-struct xfs_inode_log_item;
-struct xfs_mount;
-struct xfs_trans;
-struct xfs_dquot;
-
-#if defined(XFS_ILOCK_TRACE)
-#define XFS_ILOCK_KTRACE_SIZE	32
-extern ktrace_t *xfs_ilock_trace_buf;
-extern void xfs_ilock_trace(struct xfs_inode *, int, unsigned int, inst_t *);
-#else
-#define	xfs_ilock_trace(i,n,f,ra)
-#endif
-
-typedef struct dm_attrs_s {
-	__uint32_t	da_dmevmask;	/* DMIG event mask */
-	__uint16_t	da_dmstate;	/* DMIG state info */
-	__uint16_t	da_pad;		/* DMIG extra padding */
-} dm_attrs_t;
-
 /*
  * This is the xfs in-core inode structure.
  * Most of the on-disk inode is embedded in the i_d field.
@@ -191,19 +143,98 @@ typedef struct xfs_icdinode {
 	__uint32_t	di_gen;		/* generation number */
 } xfs_icdinode_t;
 
-typedef struct {
-	struct xfs_inode	*ip_mnext;	/* next inode in mount list */
-	struct xfs_inode	*ip_mprev;	/* ptr to prev inode */
-	struct xfs_mount	*ip_mount;	/* fs mount struct ptr */
-} xfs_iptr_t;
+/*
+ * Flags for xfs_ichgtime().
+ */
+#define	XFS_ICHGTIME_MOD	0x1	/* data fork modification timestamp */
+#define	XFS_ICHGTIME_CHG	0x2	/* inode field change timestamp */
+
+/*
+ * Per-fork incore inode flags.
+ */
+#define	XFS_IFINLINE	0x01	/* Inline data is read in */
+#define	XFS_IFEXTENTS	0x02	/* All extent pointers are read in */
+#define	XFS_IFBROOT	0x04	/* i_broot points to the bmap b-tree root */
+#define	XFS_IFEXTIREC	0x08	/* Indirection array of extent blocks */
+
+/*
+ * Flags for xfs_inotobp, xfs_itobp(), xfs_imap() and xfs_dilocate().
+ */
+#define XFS_IMAP_LOOKUP		0x1
+#define XFS_IMAP_BULKSTAT	0x2
+
+/*
+ * Fork handling.
+ */
+
+#define XFS_IFORK_Q(ip)			((ip)->i_d.di_forkoff != 0)
+#define XFS_IFORK_BOFF(ip)		((int)((ip)->i_d.di_forkoff << 3))
+
+#define XFS_IFORK_PTR(ip,w)		\
+	((w) == XFS_DATA_FORK ? \
+		&(ip)->i_df : \
+		(ip)->i_afp)
+#define XFS_IFORK_DSIZE(ip) \
+	(XFS_IFORK_Q(ip) ? \
+		XFS_IFORK_BOFF(ip) : \
+		XFS_LITINO((ip)->i_mount))
+#define XFS_IFORK_ASIZE(ip) \
+	(XFS_IFORK_Q(ip) ? \
+		XFS_LITINO((ip)->i_mount) - XFS_IFORK_BOFF(ip) : \
+		0)
+#define XFS_IFORK_SIZE(ip,w) \
+	((w) == XFS_DATA_FORK ? \
+		XFS_IFORK_DSIZE(ip) : \
+		XFS_IFORK_ASIZE(ip))
+#define XFS_IFORK_FORMAT(ip,w) \
+	((w) == XFS_DATA_FORK ? \
+		(ip)->i_d.di_format : \
+		(ip)->i_d.di_aformat)
+#define XFS_IFORK_FMT_SET(ip,w,n) \
+	((w) == XFS_DATA_FORK ? \
+		((ip)->i_d.di_format = (n)) : \
+		((ip)->i_d.di_aformat = (n)))
+#define XFS_IFORK_NEXTENTS(ip,w) \
+	((w) == XFS_DATA_FORK ? \
+		(ip)->i_d.di_nextents : \
+		(ip)->i_d.di_anextents)
+#define XFS_IFORK_NEXT_SET(ip,w,n) \
+	((w) == XFS_DATA_FORK ? \
+		((ip)->i_d.di_nextents = (n)) : \
+		((ip)->i_d.di_anextents = (n)))
+
+
+
+#ifdef __KERNEL__
+
+struct bhv_desc;
+struct cred;
+struct ktrace;
+struct xfs_buf;
+struct xfs_bmap_free;
+struct xfs_bmbt_irec;
+struct xfs_inode_log_item;
+struct xfs_mount;
+struct xfs_trans;
+struct xfs_dquot;
+
+#if defined(XFS_ILOCK_TRACE)
+#define XFS_ILOCK_KTRACE_SIZE	32
+extern ktrace_t *xfs_ilock_trace_buf;
+extern void xfs_ilock_trace(struct xfs_inode *, int, unsigned int, inst_t *);
+#else
+#define	xfs_ilock_trace(i,n,f,ra)
+#endif
+
+typedef struct dm_attrs_s {
+	__uint32_t	da_dmevmask;	/* DMIG event mask */
+	__uint16_t	da_dmstate;	/* DMIG state info */
+	__uint16_t	da_pad;		/* DMIG extra padding */
+} dm_attrs_t;
 
 typedef struct xfs_inode {
 	/* Inode linking and identification information. */
-	struct xfs_inode	*i_mnext;	/* next inode in mount list */
-	struct xfs_inode	*i_mprev;	/* ptr to prev inode */
 	struct xfs_mount	*i_mount;	/* fs mount struct ptr */
-	struct list_head	i_reclaim;	/* reclaim list */
-	struct inode		*i_vnode;	/* vnode backpointer */
 	struct xfs_dquot	*i_udquot;	/* user dquot */
 	struct xfs_dquot	*i_gdquot;	/* group dquot */
 
@@ -238,6 +269,10 @@ typedef struct xfs_inode {
 	xfs_fsize_t		i_size;		/* in-memory size */
 	xfs_fsize_t		i_new_size;	/* size when write completes */
 	atomic_t		i_iocount;	/* outstanding I/O count */
+
+	/* VFS inode */
+	struct inode		i_vnode;	/* embedded VFS inode */
+
 	/* Trace buffers per inode. */
 #ifdef XFS_INODE_TRACE
 	struct ktrace		*i_trace;	/* general inode trace */
@@ -245,7 +280,7 @@ typedef struct xfs_inode {
 #ifdef XFS_BMAP_TRACE
 	struct ktrace		*i_xtrace;	/* inode extent list trace */
 #endif
-#ifdef XFS_BMBT_TRACE
+#ifdef XFS_BTREE_TRACE
 	struct ktrace		*i_btrace;	/* inode bmap btree trace */
 #endif
 #ifdef XFS_RW_TRACE
@@ -265,13 +300,30 @@ typedef struct xfs_inode {
 /* Convert from vfs inode to xfs inode */
 static inline struct xfs_inode *XFS_I(struct inode *inode)
 {
-	return (struct xfs_inode *)inode->i_private;
+	return container_of(inode, struct xfs_inode, i_vnode);
 }
 
 /* convert from xfs inode to vfs inode */
 static inline struct inode *VFS_I(struct xfs_inode *ip)
 {
-	return (struct inode *)ip->i_vnode;
+	return &ip->i_vnode;
+}
+
+/*
+ * Get rid of a partially initialized inode.
+ *
+ * We have to go through destroy_inode to make sure allocations
+ * from init_inode_always like the security data are undone.
+ *
+ * We mark the inode bad so that it takes the short cut in
+ * the reclaim path instead of going through the flush path
+ * which doesn't make sense for an inode that has never seen the
+ * light of day.
+ */
+static inline void xfs_destroy_inode(struct xfs_inode *ip)
+{
+	make_bad_inode(VFS_I(ip));
+	return destroy_inode(VFS_I(ip));
 }
 
 /*
@@ -327,50 +379,26 @@ xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
 	spin_unlock(&ip->i_flags_lock);
 	return ret;
 }
-#endif	/* __KERNEL__ */
-
 
 /*
- * Fork handling.
+ * Manage the i_flush queue embedded in the inode.  This completion
+ * queue synchronizes processes attempting to flush the in-core
+ * inode back to disk.
  */
+static inline void xfs_iflock(xfs_inode_t *ip)
+{
+	wait_for_completion(&ip->i_flush);
+}
 
-#define XFS_IFORK_Q(ip)			((ip)->i_d.di_forkoff != 0)
-#define XFS_IFORK_BOFF(ip)		((int)((ip)->i_d.di_forkoff << 3))
-
-#define XFS_IFORK_PTR(ip,w)		\
-	((w) == XFS_DATA_FORK ? \
-		&(ip)->i_df : \
-		(ip)->i_afp)
-#define XFS_IFORK_DSIZE(ip) \
-	(XFS_IFORK_Q(ip) ? \
-		XFS_IFORK_BOFF(ip) : \
-		XFS_LITINO((ip)->i_mount))
-#define XFS_IFORK_ASIZE(ip) \
-	(XFS_IFORK_Q(ip) ? \
-		XFS_LITINO((ip)->i_mount) - XFS_IFORK_BOFF(ip) : \
-		0)
-#define XFS_IFORK_SIZE(ip,w) \
-	((w) == XFS_DATA_FORK ? \
-		XFS_IFORK_DSIZE(ip) : \
-		XFS_IFORK_ASIZE(ip))
-#define XFS_IFORK_FORMAT(ip,w) \
-	((w) == XFS_DATA_FORK ? \
-		(ip)->i_d.di_format : \
-		(ip)->i_d.di_aformat)
-#define XFS_IFORK_FMT_SET(ip,w,n) \
-	((w) == XFS_DATA_FORK ? \
-		((ip)->i_d.di_format = (n)) : \
-		((ip)->i_d.di_aformat = (n)))
-#define XFS_IFORK_NEXTENTS(ip,w) \
-	((w) == XFS_DATA_FORK ? \
-		(ip)->i_d.di_nextents : \
-		(ip)->i_d.di_anextents)
-#define XFS_IFORK_NEXT_SET(ip,w,n) \
-	((w) == XFS_DATA_FORK ? \
-		((ip)->i_d.di_nextents = (n)) : \
-		((ip)->i_d.di_anextents = (n)))
+static inline int xfs_iflock_nowait(xfs_inode_t *ip)
+{
+	return try_wait_for_completion(&ip->i_flush);
+}
 
-#ifdef __KERNEL__
+static inline void xfs_ifunlock(xfs_inode_t *ip)
+{
+	complete(&ip->i_flush);
+}
 
 /*
  * In-core inode flags.
@@ -484,25 +512,15 @@ int		xfs_isilocked(xfs_inode_t *, uint);
 uint		xfs_ilock_map_shared(xfs_inode_t *);
 void		xfs_iunlock_map_shared(xfs_inode_t *, uint);
 void		xfs_ireclaim(xfs_inode_t *);
-int		xfs_finish_reclaim(xfs_inode_t *, int, int);
-int		xfs_finish_reclaim_all(struct xfs_mount *, int);
 
 /*
  * xfs_inode.c prototypes.
  */
-int		xfs_itobp(struct xfs_mount *, struct xfs_trans *,
-			  xfs_inode_t *, struct xfs_dinode **, struct xfs_buf **,
-			  xfs_daddr_t, uint, uint);
 int		xfs_iread(struct xfs_mount *, struct xfs_trans *, xfs_ino_t,
 			  xfs_inode_t **, xfs_daddr_t, uint);
-int		xfs_iread_extents(struct xfs_trans *, xfs_inode_t *, int);
 int		xfs_ialloc(struct xfs_trans *, xfs_inode_t *, mode_t,
 			   xfs_nlink_t, xfs_dev_t, struct cred *, xfs_prid_t,
 			   int, struct xfs_buf **, boolean_t *, xfs_inode_t **);
-void		xfs_dinode_from_disk(struct xfs_icdinode *,
-				     struct xfs_dinode_core *);
-void		xfs_dinode_to_disk(struct xfs_dinode_core *,
-				   struct xfs_icdinode *);
 
 uint		xfs_ip2xflags(struct xfs_inode *);
 uint		xfs_dic2xflags(struct xfs_dinode *);
@@ -513,17 +531,12 @@ int		xfs_itruncate_finish(struct xfs_trans **, xfs_inode_t *,
 				     xfs_fsize_t, int, int);
 int		xfs_iunlink(struct xfs_trans *, xfs_inode_t *);
 
-void		xfs_idestroy_fork(xfs_inode_t *, int);
 void		xfs_idestroy(xfs_inode_t *);
-void		xfs_idata_realloc(xfs_inode_t *, int, int);
 void		xfs_iextract(xfs_inode_t *);
 void		xfs_iext_realloc(xfs_inode_t *, int, int);
-void		xfs_iroot_realloc(xfs_inode_t *, int, int);
 void		xfs_ipin(xfs_inode_t *);
 void		xfs_iunpin(xfs_inode_t *);
-int		xfs_iextents_copy(xfs_inode_t *, xfs_bmbt_rec_t *, int);
 int		xfs_iflush(xfs_inode_t *, uint);
-void		xfs_iflush_all(struct xfs_mount *);
 void		xfs_ichgtime(xfs_inode_t *, int);
 xfs_fsize_t	xfs_file_last_byte(xfs_inode_t *);
 void		xfs_lock_inodes(xfs_inode_t **, int, uint);
@@ -532,6 +545,24 @@ void		xfs_lock_two_inodes(xfs_inode_t *, xfs_inode_t *, uint);
 void		xfs_synchronize_atime(xfs_inode_t *);
 void		xfs_mark_inode_dirty_sync(xfs_inode_t *);
 
+#endif /* __KERNEL__ */
+
+int		xfs_inotobp(struct xfs_mount *, struct xfs_trans *,
+			    xfs_ino_t, struct xfs_dinode **,
+			    struct xfs_buf **, int *, uint);
+int		xfs_itobp(struct xfs_mount *, struct xfs_trans *,
+			  struct xfs_inode *, struct xfs_dinode **,
+			  struct xfs_buf **, xfs_daddr_t, uint, uint);
+void		xfs_dinode_from_disk(struct xfs_icdinode *,
+				     struct xfs_dinode_core *);
+void		xfs_dinode_to_disk(struct xfs_dinode_core *,
+				   struct xfs_icdinode *);
+void		xfs_idestroy_fork(struct xfs_inode *, int);
+void		xfs_idata_realloc(struct xfs_inode *, int, int);
+void		xfs_iroot_realloc(struct xfs_inode *, int, int);
+int		xfs_iread_extents(struct xfs_trans *, struct xfs_inode *, int);
+int		xfs_iextents_copy(struct xfs_inode *, xfs_bmbt_rec_t *, int);
+
 xfs_bmbt_rec_host_t *xfs_iext_get_ext(xfs_ifork_t *, xfs_extnum_t);
 void		xfs_iext_insert(xfs_ifork_t *, xfs_extnum_t, xfs_extnum_t,
 				xfs_bmbt_irec_t *);
@@ -561,7 +592,8 @@ void		xfs_iext_irec_update_extoffs(xfs_ifork_t *, int, int);
 #define xfs_ipincount(ip)	((unsigned int) atomic_read(&ip->i_pincount))
 
 #ifdef DEBUG
-void		xfs_isize_check(struct xfs_mount *, xfs_inode_t *, xfs_fsize_t);
+void		xfs_isize_check(struct xfs_mount *, struct xfs_inode *,
+				xfs_fsize_t);
 #else	/* DEBUG */
 #define xfs_isize_check(mp, ip, isize)
 #endif	/* DEBUG */
@@ -576,26 +608,4 @@ extern struct kmem_zone	*xfs_ifork_zone;
 extern struct kmem_zone	*xfs_inode_zone;
 extern struct kmem_zone	*xfs_ili_zone;
 
-/*
- * Manage the i_flush queue embedded in the inode.  This completion
- * queue synchronizes processes attempting to flush the in-core
- * inode back to disk.
- */
-static inline void xfs_iflock(xfs_inode_t *ip)
-{
-	wait_for_completion(&ip->i_flush);
-}
-
-static inline int xfs_iflock_nowait(xfs_inode_t *ip)
-{
-	return try_wait_for_completion(&ip->i_flush);
-}
-
-static inline void xfs_ifunlock(xfs_inode_t *ip)
-{
-	complete(&ip->i_flush);
-}
-
-#endif	/* __KERNEL__ */
-
 #endif	/* __XFS_INODE_H__ */

fs/xfs/xfs_inode_item.c

@@ -932,6 +932,7 @@ xfs_inode_item_init(
 	iip->ili_item.li_type = XFS_LI_INODE;
 	iip->ili_item.li_ops = &xfs_inode_item_ops;
 	iip->ili_item.li_mountp = mp;
+	iip->ili_item.li_ailp = mp->m_ail;
 	iip->ili_inode = ip;
 
 	/*
@@ -976,9 +977,8 @@ xfs_iflush_done(
 	xfs_buf_t		*bp,
 	xfs_inode_log_item_t	*iip)
 {
-	xfs_inode_t	*ip;
-
-	ip = iip->ili_inode;
+	xfs_inode_t		*ip = iip->ili_inode;
+	struct xfs_ail		*ailp = iip->ili_item.li_ailp;
 
 	/*
 	 * We only want to pull the item from the AIL if it is
@@ -991,15 +991,12 @@ xfs_iflush_done(
 	 */
 	if (iip->ili_logged &&
 	    (iip->ili_item.li_lsn == iip->ili_flush_lsn)) {
-		spin_lock(&ip->i_mount->m_ail_lock);
+		spin_lock(&ailp->xa_lock);
 		if (iip->ili_item.li_lsn == iip->ili_flush_lsn) {
-			/*
-			 * xfs_trans_delete_ail() drops the AIL lock.
-			 */
-			xfs_trans_delete_ail(ip->i_mount,
-					     (xfs_log_item_t*)iip);
+			/* xfs_trans_ail_delete() drops the AIL lock. */
+			xfs_trans_ail_delete(ailp, (xfs_log_item_t*)iip);
 		} else {
-			spin_unlock(&ip->i_mount->m_ail_lock);
+			spin_unlock(&ailp->xa_lock);
 		}
 	}
 
@@ -1031,21 +1028,20 @@ void
 xfs_iflush_abort(
 	xfs_inode_t		*ip)
 {
-	xfs_inode_log_item_t	*iip;
+	xfs_inode_log_item_t	*iip = ip->i_itemp;
 	xfs_mount_t		*mp;
 
 	iip = ip->i_itemp;
 	mp = ip->i_mount;
 	if (iip) {
+		struct xfs_ail	*ailp = iip->ili_item.li_ailp;
 		if (iip->ili_item.li_flags & XFS_LI_IN_AIL) {
-			spin_lock(&mp->m_ail_lock);
+			spin_lock(&ailp->xa_lock);
 			if (iip->ili_item.li_flags & XFS_LI_IN_AIL) {
-				/*
-				 * xfs_trans_delete_ail() drops the AIL lock.
-				 */
-				xfs_trans_delete_ail(mp, (xfs_log_item_t *)iip);
+				/* xfs_trans_ail_delete() drops the AIL lock. */
+				xfs_trans_ail_delete(ailp, (xfs_log_item_t *)iip);
 			} else
-				spin_unlock(&mp->m_ail_lock);
+				spin_unlock(&ailp->xa_lock);
 		}
 		iip->ili_logged = 0;
 		/*

fs/xfs/xfs_inode_item.h

@@ -112,6 +112,24 @@ typedef struct xfs_inode_log_format_64 {
 #define	XFS_ILI_IOLOCKED_ANY   (XFS_ILI_IOLOCKED_EXCL | XFS_ILI_IOLOCKED_SHARED)
 
 
+#define	XFS_ILOG_FBROOT(w)	xfs_ilog_fbroot(w)
+static inline int xfs_ilog_fbroot(int w)
+{
+	return (w == XFS_DATA_FORK ? XFS_ILOG_DBROOT : XFS_ILOG_ABROOT);
+}
+
+#define	XFS_ILOG_FEXT(w)	xfs_ilog_fext(w)
+static inline int xfs_ilog_fext(int w)
+{
+	return (w == XFS_DATA_FORK ? XFS_ILOG_DEXT : XFS_ILOG_AEXT);
+}
+
+#define	XFS_ILOG_FDATA(w)	xfs_ilog_fdata(w)
+static inline int xfs_ilog_fdata(int w)
+{
+	return (w == XFS_DATA_FORK ? XFS_ILOG_DDATA : XFS_ILOG_ADATA);
+}
+
 #ifdef __KERNEL__
 
 struct xfs_buf;
@@ -148,26 +166,6 @@ typedef struct xfs_inode_log_item {
 } xfs_inode_log_item_t;
 
 
-#define	XFS_ILOG_FDATA(w)	xfs_ilog_fdata(w)
-static inline int xfs_ilog_fdata(int w)
-{
-	return (w == XFS_DATA_FORK ? XFS_ILOG_DDATA : XFS_ILOG_ADATA);
-}
-
-#endif	/* __KERNEL__ */
-
-#define	XFS_ILOG_FBROOT(w)	xfs_ilog_fbroot(w)
-static inline int xfs_ilog_fbroot(int w)
-{
-	return (w == XFS_DATA_FORK ? XFS_ILOG_DBROOT : XFS_ILOG_ABROOT);
-}
-
-#define	XFS_ILOG_FEXT(w)	xfs_ilog_fext(w)
-static inline int xfs_ilog_fext(int w)
-{
-	return (w == XFS_DATA_FORK ? XFS_ILOG_DEXT : XFS_ILOG_AEXT);
-}
-
 static inline int xfs_inode_clean(xfs_inode_t *ip)
 {
 	return (!ip->i_itemp ||
@@ -175,9 +173,6 @@ static inline int xfs_inode_clean(xfs_inode_t *ip)
 	       !ip->i_update_core;
 }
 
-
-#ifdef __KERNEL__
-
 extern void xfs_inode_item_init(struct xfs_inode *, struct xfs_mount *);
 extern void xfs_inode_item_destroy(struct xfs_inode *);
 extern void xfs_iflush_done(struct xfs_buf *, xfs_inode_log_item_t *);

fs/xfs/xfs_itable.c

@@ -359,7 +359,6 @@ xfs_bulkstat(
 	int			ubused;	/* bytes used by formatter */
 	xfs_buf_t		*bp;	/* ptr to on-disk inode cluster buf */
 	xfs_dinode_t		*dip;	/* ptr into bp for specific inode */
-	xfs_inode_t		*ip;	/* ptr to in-core inode struct */
 
 	/*
 	 * Get the last inode value, see if there's nothing to do.
@@ -416,8 +415,7 @@ xfs_bulkstat(
 		/*
 		 * Allocate and initialize a btree cursor for ialloc btree.
 		 */
-		cur = xfs_btree_init_cursor(mp, NULL, agbp, agno, XFS_BTNUM_INO,
-						(xfs_inode_t *)0, 0);
+		cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
 		irbp = irbuf;
 		irbufend = irbuf + nirbuf;
 		end_of_ag = 0;
@@ -472,7 +470,7 @@ xfs_bulkstat(
 			 * In any case, increment to the next record.
 			 */
 			if (!error)
-				error = xfs_inobt_increment(cur, 0, &tmp);
+				error = xfs_btree_increment(cur, 0, &tmp);
 		} else {
 			/*
 			 * Start of ag.  Lookup the first inode chunk.
@@ -539,7 +537,7 @@ xfs_bulkstat(
 			 * Set agino to after this chunk and bump the cursor.
 			 */
 			agino = gino + XFS_INODES_PER_CHUNK;
-			error = xfs_inobt_increment(cur, 0, &tmp);
+			error = xfs_btree_increment(cur, 0, &tmp);
 			cond_resched();
 		}
 		/*
@@ -586,6 +584,8 @@ xfs_bulkstat(
 
 					if (flags & (BULKSTAT_FG_QUICK |
 						     BULKSTAT_FG_INLINE)) {
+						int offset;
+
 						ino = XFS_AGINO_TO_INO(mp, agno,
 								       agino);
 						bno = XFS_AGB_TO_DADDR(mp, agno,
@@ -594,21 +594,15 @@ xfs_bulkstat(
 						/*
 						 * Get the inode cluster buffer
 						 */
-						ASSERT(xfs_inode_zone != NULL);
-						ip = kmem_zone_zalloc(xfs_inode_zone,
-								      KM_SLEEP);
-						ip->i_ino = ino;
-						ip->i_mount = mp;
-						spin_lock_init(&ip->i_flags_lock);
 						if (bp)
 							xfs_buf_relse(bp);
-						error = xfs_itobp(mp, NULL, ip,
-								&dip, &bp, bno,
-								XFS_IMAP_BULKSTAT,
-								XFS_BUF_LOCK);
+
+						error = xfs_inotobp(mp, NULL, ino, &dip,
+								    &bp, &offset,
+								    XFS_IMAP_BULKSTAT);
+
 						if (!error)
-							clustidx = ip->i_boffset / mp->m_sb.sb_inodesize;
-						kmem_zone_free(xfs_inode_zone, ip);
+							clustidx = offset / mp->m_sb.sb_inodesize;
 						if (XFS_TEST_ERROR(error != 0,
 								   mp, XFS_ERRTAG_BULKSTAT_READ_CHUNK,
 								   XFS_RANDOM_BULKSTAT_READ_CHUNK)) {
@@ -842,8 +836,7 @@ xfs_inumbers(
 				agino = 0;
 				continue;
 			}
-			cur = xfs_btree_init_cursor(mp, NULL, agbp, agno,
-				XFS_BTNUM_INO, (xfs_inode_t *)0, 0);
+			cur = xfs_inobt_init_cursor(mp, NULL, agbp, agno);
 			error = xfs_inobt_lookup_ge(cur, agino, 0, 0, &tmp);
 			if (error) {
 				xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
@@ -887,7 +880,7 @@ xfs_inumbers(
 			bufidx = 0;
 		}
 		if (left) {
-			error = xfs_inobt_increment(cur, 0, &tmp);
+			error = xfs_btree_increment(cur, 0, &tmp);
 			if (error) {
 				xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
 				cur = NULL;

fs/xfs/xfs_log.c

@@ -567,12 +567,12 @@ xfs_log_mount(
 	/*
 	 * Initialize the AIL now we have a log.
 	 */
-	spin_lock_init(&mp->m_ail_lock);
 	error = xfs_trans_ail_init(mp);
 	if (error) {
 		cmn_err(CE_WARN, "XFS: AIL initialisation failed: error %d", error);
 		goto error;
 	}
+	mp->m_log->l_ailp = mp->m_ail;
 
 	/*
 	 * skip log recovery on a norecovery mount.  pretend it all
@@ -900,7 +900,7 @@ xfs_log_move_tail(xfs_mount_t	*mp,
 int
 xfs_log_need_covered(xfs_mount_t *mp)
 {
-	int		needed = 0, gen;
+	int		needed = 0;
 	xlog_t		*log = mp->m_log;
 
 	if (!xfs_fs_writable(mp))
@@ -909,7 +909,7 @@ xfs_log_need_covered(xfs_mount_t *mp)
 	spin_lock(&log->l_icloglock);
 	if (((log->l_covered_state == XLOG_STATE_COVER_NEED) ||
 		(log->l_covered_state == XLOG_STATE_COVER_NEED2))
-			&& !xfs_trans_first_ail(mp, &gen)
+			&& !xfs_trans_ail_tail(log->l_ailp)
 			&& xlog_iclogs_empty(log)) {
 		if (log->l_covered_state == XLOG_STATE_COVER_NEED)
 			log->l_covered_state = XLOG_STATE_COVER_DONE;
@@ -946,7 +946,7 @@ xlog_assign_tail_lsn(xfs_mount_t *mp)
 	xfs_lsn_t tail_lsn;
 	xlog_t	  *log = mp->m_log;
 
-	tail_lsn = xfs_trans_tail_ail(mp);
+	tail_lsn = xfs_trans_ail_tail(mp->m_ail);
 	spin_lock(&log->l_grant_lock);
 	if (tail_lsn != 0) {
 		log->l_tail_lsn = tail_lsn;
@@ -1413,7 +1413,7 @@ xlog_grant_push_ail(xfs_mount_t	*mp,
      */
     if (threshold_lsn &&
 	!XLOG_FORCED_SHUTDOWN(log))
-	    xfs_trans_push_ail(mp, threshold_lsn);
+	    xfs_trans_ail_push(log->l_ailp, threshold_lsn);
 }	/* xlog_grant_push_ail */
 
 

fs/xfs/xfs_log_priv.h

@@ -404,6 +404,7 @@ typedef struct xlog_in_core {
 typedef struct log {
 	/* The following fields don't need locking */
 	struct xfs_mount	*l_mp;	        /* mount point */
+	struct xfs_ail		*l_ailp;	/* AIL log is working with */
 	struct xfs_buf		*l_xbuf;        /* extra buffer for log
 						 * wrapping */
 	struct xfs_buftarg	*l_targ;        /* buftarg of log */

fs/xfs/xfs_log_recover.c

@@ -54,10 +54,8 @@ STATIC void	xlog_recover_insert_item_backq(xlog_recover_item_t **q,
 					       xlog_recover_item_t *item);
 #if defined(DEBUG)
 STATIC void	xlog_recover_check_summary(xlog_t *);
-STATIC void	xlog_recover_check_ail(xfs_mount_t *, xfs_log_item_t *, int);
 #else
 #define	xlog_recover_check_summary(log)
-#define	xlog_recover_check_ail(mp, lip, gen)
 #endif
 
 
@@ -1419,7 +1417,13 @@ xlog_recover_add_to_trans(
 		return 0;
 	item = trans->r_itemq;
 	if (item == NULL) {
-		ASSERT(*(uint *)dp == XFS_TRANS_HEADER_MAGIC);
+		/* we need to catch log corruptions here */
+		if (*(uint *)dp != XFS_TRANS_HEADER_MAGIC) {
+			xlog_warn("XFS: xlog_recover_add_to_trans: "
+				  "bad header magic number");
+			ASSERT(0);
+			return XFS_ERROR(EIO);
+		}
 		if (len == sizeof(xfs_trans_header_t))
 			xlog_recover_add_item(&trans->r_itemq);
 		memcpy(&trans->r_theader, dp, len); /* d, s, l */
@@ -2452,8 +2456,8 @@ xlog_recover_do_inode_trans(
 		break;
 
 	case XFS_ILOG_DBROOT:
-		xfs_bmbt_to_bmdr((xfs_bmbt_block_t *)src, len,
-				 &(dip->di_u.di_bmbt),
+		xfs_bmbt_to_bmdr(mp, (struct xfs_btree_block *)src, len,
+				 &dip->di_u.di_bmbt,
 				 XFS_DFORK_DSIZE(dip, mp));
 		break;
 
@@ -2490,8 +2494,8 @@ xlog_recover_do_inode_trans(
 
 		case XFS_ILOG_ABROOT:
 			dest = XFS_DFORK_APTR(dip);
-			xfs_bmbt_to_bmdr((xfs_bmbt_block_t *)src, len,
-					 (xfs_bmdr_block_t*)dest,
+			xfs_bmbt_to_bmdr(mp, (struct xfs_btree_block *)src,
+					 len, (xfs_bmdr_block_t*)dest,
 					 XFS_DFORK_ASIZE(dip, mp));
 			break;
 
@@ -2683,11 +2687,11 @@ xlog_recover_do_efi_trans(
 	efip->efi_next_extent = efi_formatp->efi_nextents;
 	efip->efi_flags |= XFS_EFI_COMMITTED;
 
-	spin_lock(&mp->m_ail_lock);
+	spin_lock(&log->l_ailp->xa_lock);
 	/*
-	 * xfs_trans_update_ail() drops the AIL lock.
+	 * xfs_trans_ail_update() drops the AIL lock.
 	 */
-	xfs_trans_update_ail(mp, (xfs_log_item_t *)efip, lsn);
+	xfs_trans_ail_update(log->l_ailp, (xfs_log_item_t *)efip, lsn);
 	return 0;
 }
 
@@ -2706,12 +2710,12 @@ xlog_recover_do_efd_trans(
 	xlog_recover_item_t	*item,
 	int			pass)
 {
-	xfs_mount_t		*mp;
 	xfs_efd_log_format_t	*efd_formatp;
 	xfs_efi_log_item_t	*efip = NULL;
 	xfs_log_item_t		*lip;
-	int			gen;
 	__uint64_t		efi_id;
+	struct xfs_ail_cursor	cur;
+	struct xfs_ail		*ailp = log->l_ailp;
 
 	if (pass == XLOG_RECOVER_PASS1) {
 		return;
@@ -2728,25 +2732,26 @@ xlog_recover_do_efd_trans(
 	 * Search for the efi with the id in the efd format structure
 	 * in the AIL.
 	 */
-	mp = log->l_mp;
-	spin_lock(&mp->m_ail_lock);
-	lip = xfs_trans_first_ail(mp, &gen);
+	spin_lock(&ailp->xa_lock);
+	lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
 	while (lip != NULL) {
 		if (lip->li_type == XFS_LI_EFI) {
 			efip = (xfs_efi_log_item_t *)lip;
 			if (efip->efi_format.efi_id == efi_id) {
 				/*
-				 * xfs_trans_delete_ail() drops the
+				 * xfs_trans_ail_delete() drops the
 				 * AIL lock.
 				 */
-				xfs_trans_delete_ail(mp, lip);
+				xfs_trans_ail_delete(ailp, lip);
 				xfs_efi_item_free(efip);
-				return;
+				spin_lock(&ailp->xa_lock);
+				break;
 			}
 		}
-		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
+		lip = xfs_trans_ail_cursor_next(ailp, &cur);
 	}
-	spin_unlock(&mp->m_ail_lock);
+	xfs_trans_ail_cursor_done(ailp, &cur);
+	spin_unlock(&ailp->xa_lock);
 }
 
 /*
@@ -3029,33 +3034,6 @@ abort_error:
 	return error;
 }
 
-/*
- * Verify that once we've encountered something other than an EFI
- * in the AIL that there are no more EFIs in the AIL.
- */
-#if defined(DEBUG)
-STATIC void
-xlog_recover_check_ail(
-	xfs_mount_t		*mp,
-	xfs_log_item_t		*lip,
-	int			gen)
-{
-	int			orig_gen = gen;
-
-	do {
-		ASSERT(lip->li_type != XFS_LI_EFI);
-		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
-		/*
-		 * The check will be bogus if we restart from the
-		 * beginning of the AIL, so ASSERT that we don't.
-		 * We never should since we're holding the AIL lock
-		 * the entire time.
-		 */
-		ASSERT(gen == orig_gen);
-	} while (lip != NULL);
-}
-#endif	/* DEBUG */
-
 /*
  * When this is called, all of the EFIs which did not have
  * corresponding EFDs should be in the AIL.  What we do now
@@ -3080,20 +3058,23 @@ xlog_recover_process_efis(
 {
 	xfs_log_item_t		*lip;
 	xfs_efi_log_item_t	*efip;
-	int			gen;
-	xfs_mount_t		*mp;
 	int			error = 0;
+	struct xfs_ail_cursor	cur;
+	struct xfs_ail		*ailp;
 
-	mp = log->l_mp;
-	spin_lock(&mp->m_ail_lock);
-
-	lip = xfs_trans_first_ail(mp, &gen);
+	ailp = log->l_ailp;
+	spin_lock(&ailp->xa_lock);
+	lip = xfs_trans_ail_cursor_first(ailp, &cur, 0);
 	while (lip != NULL) {
 		/*
 		 * We're done when we see something other than an EFI.
+		 * There should be no EFIs left in the AIL now.
 		 */
 		if (lip->li_type != XFS_LI_EFI) {
-			xlog_recover_check_ail(mp, lip, gen);
+#ifdef DEBUG
+			for (; lip; lip = xfs_trans_ail_cursor_next(ailp, &cur))
+				ASSERT(lip->li_type != XFS_LI_EFI);
+#endif
 			break;
 		}
 
@@ -3102,18 +3083,20 @@ xlog_recover_process_efis(
 		 */
 		efip = (xfs_efi_log_item_t *)lip;
 		if (efip->efi_flags & XFS_EFI_RECOVERED) {
-			lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
+			lip = xfs_trans_ail_cursor_next(ailp, &cur);
 			continue;
 		}
 
-		spin_unlock(&mp->m_ail_lock);
-		error = xlog_recover_process_efi(mp, efip);
+		spin_unlock(&ailp->xa_lock);
+		error = xlog_recover_process_efi(log->l_mp, efip);
+		spin_lock(&ailp->xa_lock);
 		if (error)
-			return error;
-		spin_lock(&mp->m_ail_lock);
-		lip = xfs_trans_next_ail(mp, lip, &gen, NULL);
+			goto out;
+		lip = xfs_trans_ail_cursor_next(ailp, &cur);
 	}
-	spin_unlock(&mp->m_ail_lock);
+out:
+	xfs_trans_ail_cursor_done(ailp, &cur);
+	spin_unlock(&ailp->xa_lock);
 	return error;
 }
 

fs/xfs/xfs_mount.c

@@ -567,8 +567,6 @@ xfs_readsb(xfs_mount_t *mp, int flags)
 STATIC void
 xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
 {
-	int	i;
-
 	mp->m_agfrotor = mp->m_agirotor = 0;
 	spin_lock_init(&mp->m_agirotor_lock);
 	mp->m_maxagi = mp->m_sb.sb_agcount;
@@ -582,7 +580,6 @@ xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
 	mp->m_blockmask = sbp->sb_blocksize - 1;
 	mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
 	mp->m_blockwmask = mp->m_blockwsize - 1;
-	INIT_LIST_HEAD(&mp->m_del_inodes);
 
 	/*
 	 * Setup for attributes, in case they get created.
@@ -605,24 +602,20 @@ xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
 	}
 	ASSERT(mp->m_attroffset < XFS_LITINO(mp));
 
-	for (i = 0; i < 2; i++) {
-		mp->m_alloc_mxr[i] = XFS_BTREE_BLOCK_MAXRECS(sbp->sb_blocksize,
-			xfs_alloc, i == 0);
-		mp->m_alloc_mnr[i] = XFS_BTREE_BLOCK_MINRECS(sbp->sb_blocksize,
-			xfs_alloc, i == 0);
-	}
-	for (i = 0; i < 2; i++) {
-		mp->m_bmap_dmxr[i] = XFS_BTREE_BLOCK_MAXRECS(sbp->sb_blocksize,
-			xfs_bmbt, i == 0);
-		mp->m_bmap_dmnr[i] = XFS_BTREE_BLOCK_MINRECS(sbp->sb_blocksize,
-			xfs_bmbt, i == 0);
-	}
-	for (i = 0; i < 2; i++) {
-		mp->m_inobt_mxr[i] = XFS_BTREE_BLOCK_MAXRECS(sbp->sb_blocksize,
-			xfs_inobt, i == 0);
-		mp->m_inobt_mnr[i] = XFS_BTREE_BLOCK_MINRECS(sbp->sb_blocksize,
-			xfs_inobt, i == 0);
-	}
+	mp->m_alloc_mxr[0] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 1);
+	mp->m_alloc_mxr[1] = xfs_allocbt_maxrecs(mp, sbp->sb_blocksize, 0);
+	mp->m_alloc_mnr[0] = mp->m_alloc_mxr[0] / 2;
+	mp->m_alloc_mnr[1] = mp->m_alloc_mxr[1] / 2;
+
+	mp->m_inobt_mxr[0] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 1);
+	mp->m_inobt_mxr[1] = xfs_inobt_maxrecs(mp, sbp->sb_blocksize, 0);
+	mp->m_inobt_mnr[0] = mp->m_inobt_mxr[0] / 2;
+	mp->m_inobt_mnr[1] = mp->m_inobt_mxr[1] / 2;
+
+	mp->m_bmap_dmxr[0] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 1);
+	mp->m_bmap_dmxr[1] = xfs_bmbt_maxrecs(mp, sbp->sb_blocksize, 0);
+	mp->m_bmap_dmnr[0] = mp->m_bmap_dmxr[0] / 2;
+	mp->m_bmap_dmnr[1] = mp->m_bmap_dmxr[1] / 2;
 
 	mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
 	mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
@@ -1241,10 +1234,13 @@ xfs_unmountfs(
 	 * need to force the log first.
 	 */
 	xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE | XFS_LOG_SYNC);
-	xfs_iflush_all(mp);
+	xfs_reclaim_inodes(mp, 0, XFS_IFLUSH_ASYNC);
 
 	XFS_QM_DQPURGEALL(mp, XFS_QMOPT_QUOTALL | XFS_QMOPT_UMOUNTING);
 
+	if (mp->m_quotainfo)
+		XFS_QM_DONE(mp);
+
 	/*
 	 * Flush out the log synchronously so that we know for sure
 	 * that nothing is pinned.  This is important because bflush()
@@ -1285,11 +1281,6 @@ xfs_unmountfs(
 	xfs_unmountfs_wait(mp); 		/* wait for async bufs */
 	xfs_log_unmount(mp);			/* Done! No more fs ops. */
 
-	/*
-	 * All inodes from this mount point should be freed.
-	 */
-	ASSERT(mp->m_inodes == NULL);
-
 	if ((mp->m_flags & XFS_MOUNT_NOUUID) == 0)
 		uuid_table_remove(&mp->m_sb.sb_uuid);
 
@@ -1297,8 +1288,6 @@ xfs_unmountfs(
 	xfs_errortag_clearall(mp, 0);
 #endif
 	xfs_free_perag(mp);
-	if (mp->m_quotainfo)
-		XFS_QM_DONE(mp);
 }
 
 STATIC void

fs/xfs/xfs_mount.h

@@ -18,6 +18,7 @@
 #ifndef __XFS_MOUNT_H__
 #define	__XFS_MOUNT_H__
 
+#include "xfs_sync.h"
 
 typedef struct xfs_trans_reservations {
 	uint	tr_write;	/* extent alloc trans */
@@ -44,14 +45,14 @@ typedef struct xfs_trans_reservations {
 } xfs_trans_reservations_t;
 
 #ifndef __KERNEL__
-/*
- * Moved here from xfs_ag.h to avoid reordering header files
- */
+
 #define XFS_DADDR_TO_AGNO(mp,d) \
 	((xfs_agnumber_t)(XFS_BB_TO_FSBT(mp, d) / (mp)->m_sb.sb_agblocks))
 #define XFS_DADDR_TO_AGBNO(mp,d) \
 	((xfs_agblock_t)(XFS_BB_TO_FSBT(mp, d) % (mp)->m_sb.sb_agblocks))
-#else
+
+#else /* __KERNEL__ */
+
 struct cred;
 struct log;
 struct xfs_mount_args;
@@ -62,6 +63,7 @@ struct xfs_extdelta;
 struct xfs_swapext;
 struct xfs_mru_cache;
 struct xfs_nameops;
+struct xfs_ail;
 
 /*
  * Prototypes and functions for the Data Migration subsystem.
@@ -223,18 +225,10 @@ extern void	xfs_icsb_sync_counters_locked(struct xfs_mount *, int);
 #define xfs_icsb_sync_counters_locked(mp, flags) do { } while (0)
 #endif
 
-typedef struct xfs_ail {
-	struct list_head	xa_ail;
-	uint			xa_gen;
-	struct task_struct	*xa_task;
-	xfs_lsn_t		xa_target;
-} xfs_ail_t;
-
 typedef struct xfs_mount {
 	struct super_block	*m_super;
 	xfs_tid_t		m_tid;		/* next unused tid for fs */
-	spinlock_t		m_ail_lock;	/* fs AIL mutex */
-	xfs_ail_t		m_ail;		/* fs active log item list */
+	struct xfs_ail		*m_ail;		/* fs active log item list */
 	xfs_sb_t		m_sb;		/* copy of fs superblock */
 	spinlock_t		m_sb_lock;	/* sb counter lock */
 	struct xfs_buf		*m_sb_bp;	/* buffer for superblock */
@@ -247,9 +241,6 @@ typedef struct xfs_mount {
 	xfs_agnumber_t		m_agirotor;	/* last ag dir inode alloced */
 	spinlock_t		m_agirotor_lock;/* .. and lock protecting it */
 	xfs_agnumber_t		m_maxagi;	/* highest inode alloc group */
-	struct xfs_inode	*m_inodes;	/* active inode list */
-	struct list_head	m_del_inodes;	/* inodes to reclaim */
-	mutex_t			m_ilock;	/* inode list mutex */
 	uint			m_ireclaims;	/* count of calls to reclaim*/
 	uint			m_readio_log;	/* min read size log bytes */
 	uint			m_readio_blocks; /* min read size blocks */
@@ -267,7 +258,6 @@ typedef struct xfs_mount {
 	xfs_buftarg_t		*m_ddev_targp;	/* saves taking the address */
 	xfs_buftarg_t		*m_logdev_targp;/* ptr to log device */
 	xfs_buftarg_t		*m_rtdev_targp;	/* ptr to rt device */
-	__uint8_t		m_dircook_elog;	/* log d-cookie entry bits */
 	__uint8_t		m_blkbit_log;	/* blocklog + NBBY */
 	__uint8_t		m_blkbb_log;	/* blocklog - BBSHIFT */
 	__uint8_t		m_agno_log;	/* log #ag's */
@@ -276,12 +266,12 @@ typedef struct xfs_mount {
 	uint			m_blockmask;	/* sb_blocksize-1 */
 	uint			m_blockwsize;	/* sb_blocksize in words */
 	uint			m_blockwmask;	/* blockwsize-1 */
-	uint			m_alloc_mxr[2];	/* XFS_ALLOC_BLOCK_MAXRECS */
-	uint			m_alloc_mnr[2];	/* XFS_ALLOC_BLOCK_MINRECS */
-	uint			m_bmap_dmxr[2];	/* XFS_BMAP_BLOCK_DMAXRECS */
-	uint			m_bmap_dmnr[2];	/* XFS_BMAP_BLOCK_DMINRECS */
-	uint			m_inobt_mxr[2];	/* XFS_INOBT_BLOCK_MAXRECS */
-	uint			m_inobt_mnr[2];	/* XFS_INOBT_BLOCK_MINRECS */
+	uint			m_alloc_mxr[2];	/* max alloc btree records */
+	uint			m_alloc_mnr[2];	/* min alloc btree records */
+	uint			m_bmap_dmxr[2];	/* max bmap btree records */
+	uint			m_bmap_dmnr[2];	/* min bmap btree records */
+	uint			m_inobt_mxr[2];	/* max inobt btree records */
+	uint			m_inobt_mnr[2];	/* min inobt btree records */
 	uint			m_ag_maxlevels;	/* XFS_AG_MAXLEVELS */
 	uint			m_bm_maxlevels[2]; /* XFS_BM_MAXLEVELS */
 	uint			m_in_maxlevels;	/* XFS_IN_MAXLEVELS */
@@ -313,8 +303,7 @@ typedef struct xfs_mount {
 	int			m_attr_magicpct;/* 37% of the blocksize */
 	int			m_dir_magicpct;	/* 37% of the dir blocksize */
 	__uint8_t		m_mk_sharedro;	/* mark shared ro on unmount */
-	__uint8_t		m_inode_quiesce;/* call quiesce on new inodes.
-						   field governed by m_ilock */
+	__uint8_t		m_inode_quiesce;/* call quiesce on new inodes. */
 	__uint8_t		m_sectbb_log;	/* sectlog - BBSHIFT */
 	const struct xfs_nameops *m_dirnameops;	/* vector of dir name ops */
 	int			m_dirblksize;	/* directory block sz--bytes */
@@ -508,7 +497,6 @@ typedef struct xfs_mod_sb {
 #define	XFS_MOUNT_ILOCK(mp)	mutex_lock(&((mp)->m_ilock))
 #define	XFS_MOUNT_IUNLOCK(mp)	mutex_unlock(&((mp)->m_ilock))
 
-extern void	xfs_mod_sb(xfs_trans_t *, __int64_t);
 extern int	xfs_log_sbcount(xfs_mount_t *, uint);
 extern int	xfs_mountfs(xfs_mount_t *mp);
 extern void	xfs_mountfs_check_barriers(xfs_mount_t *mp);
@@ -525,20 +513,20 @@ extern struct xfs_buf *xfs_getsb(xfs_mount_t *, int);
 extern int	xfs_readsb(xfs_mount_t *, int);
 extern void	xfs_freesb(xfs_mount_t *);
 extern int	xfs_fs_writable(xfs_mount_t *);
-extern int	xfs_syncsub(xfs_mount_t *, int, int *);
-extern int	xfs_sync_inodes(xfs_mount_t *, int, int *);
-extern xfs_agnumber_t	xfs_initialize_perag(xfs_mount_t *, xfs_agnumber_t);
-extern void	xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
-extern void	xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
 extern int	xfs_sb_validate_fsb_count(struct xfs_sb *, __uint64_t);
 
-extern int	xfs_dmops_get(struct xfs_mount *, struct xfs_mount_args *);
+extern int	xfs_dmops_get(struct xfs_mount *);
 extern void	xfs_dmops_put(struct xfs_mount *);
-extern int	xfs_qmops_get(struct xfs_mount *, struct xfs_mount_args *);
+extern int	xfs_qmops_get(struct xfs_mount *);
 extern void	xfs_qmops_put(struct xfs_mount *);
 
 extern struct xfs_dmops xfs_dmcore_xfs;
 
 #endif	/* __KERNEL__ */
 
+extern void	xfs_mod_sb(struct xfs_trans *, __int64_t);
+extern xfs_agnumber_t	xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t);
+extern void	xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
+extern void	xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
+
 #endif	/* __XFS_MOUNT_H__ */

fs/xfs/xfs_qmops.c

@@ -28,7 +28,6 @@
 #include "xfs_mount.h"
 #include "xfs_quota.h"
 #include "xfs_error.h"
-#include "xfs_clnt.h"
 
 
 STATIC struct xfs_dquot *
@@ -131,9 +130,9 @@ static struct xfs_qmops xfs_qmcore_stub = {
 };
 
 int
-xfs_qmops_get(struct xfs_mount *mp, struct xfs_mount_args *args)
+xfs_qmops_get(struct xfs_mount *mp)
 {
-	if (args->flags & (XFSMNT_UQUOTA | XFSMNT_PQUOTA | XFSMNT_GQUOTA)) {
+	if (XFS_IS_QUOTA_RUNNING(mp)) {
 #ifdef CONFIG_XFS_QUOTA
 		mp->m_qm_ops = &xfs_qmcore_xfs;
 #else

fs/xfs/xfs_trans.c

@@ -1383,11 +1383,12 @@ xfs_trans_chunk_committed(
 	xfs_log_item_desc_t	*lidp;
 	xfs_log_item_t		*lip;
 	xfs_lsn_t		item_lsn;
-	struct xfs_mount	*mp;
 	int			i;
 
 	lidp = licp->lic_descs;
 	for (i = 0; i < licp->lic_unused; i++, lidp++) {
+		struct xfs_ail		*ailp;
+
 		if (xfs_lic_isfree(licp, i)) {
 			continue;
 		}
@@ -1424,19 +1425,19 @@ xfs_trans_chunk_committed(
 		 * This would cause the earlier transaction to fail
 		 * the test below.
 		 */
-		mp = lip->li_mountp;
-		spin_lock(&mp->m_ail_lock);
+		ailp = lip->li_ailp;
+		spin_lock(&ailp->xa_lock);
 		if (XFS_LSN_CMP(item_lsn, lip->li_lsn) > 0) {
 			/*
 			 * This will set the item's lsn to item_lsn
 			 * and update the position of the item in
 			 * the AIL.
 			 *
-			 * xfs_trans_update_ail() drops the AIL lock.
+			 * xfs_trans_ail_update() drops the AIL lock.
 			 */
-			xfs_trans_update_ail(mp, lip, item_lsn);
+			xfs_trans_ail_update(ailp, lip, item_lsn);
 		} else {
-			spin_unlock(&mp->m_ail_lock);
+			spin_unlock(&ailp->xa_lock);
 		}
 
 		/*

fs/xfs/xfs_trans.h

@@ -18,6 +18,8 @@
 #ifndef	__XFS_TRANS_H__
 #define	__XFS_TRANS_H__
 
+struct xfs_log_item;
+
 /*
  * This is the structure written in the log at the head of
  * every transaction. It identifies the type and id of the
@@ -98,76 +100,6 @@ typedef struct xfs_trans_header {
 #define	XFS_TRANS_TYPE_MAX		41
 /* new transaction types need to be reflected in xfs_logprint(8) */
 
-
-#ifdef __KERNEL__
-struct xfs_buf;
-struct xfs_buftarg;
-struct xfs_efd_log_item;
-struct xfs_efi_log_item;
-struct xfs_inode;
-struct xfs_item_ops;
-struct xfs_log_iovec;
-struct xfs_log_item;
-struct xfs_log_item_desc;
-struct xfs_mount;
-struct xfs_trans;
-struct xfs_dquot_acct;
-
-typedef struct xfs_log_item {
-	struct list_head		li_ail;		/* AIL pointers */
-	xfs_lsn_t			li_lsn;		/* last on-disk lsn */
-	struct xfs_log_item_desc	*li_desc;	/* ptr to current desc*/
-	struct xfs_mount		*li_mountp;	/* ptr to fs mount */
-	uint				li_type;	/* item type */
-	uint				li_flags;	/* misc flags */
-	struct xfs_log_item		*li_bio_list;	/* buffer item list */
-	void				(*li_cb)(struct xfs_buf *,
-						 struct xfs_log_item *);
-							/* buffer item iodone */
-							/* callback func */
-	struct xfs_item_ops		*li_ops;	/* function list */
-} xfs_log_item_t;
-
-#define	XFS_LI_IN_AIL	0x1
-#define XFS_LI_ABORTED	0x2
-
-typedef struct xfs_item_ops {
-	uint (*iop_size)(xfs_log_item_t *);
-	void (*iop_format)(xfs_log_item_t *, struct xfs_log_iovec *);
-	void (*iop_pin)(xfs_log_item_t *);
-	void (*iop_unpin)(xfs_log_item_t *, int);
-	void (*iop_unpin_remove)(xfs_log_item_t *, struct xfs_trans *);
-	uint (*iop_trylock)(xfs_log_item_t *);
-	void (*iop_unlock)(xfs_log_item_t *);
-	xfs_lsn_t (*iop_committed)(xfs_log_item_t *, xfs_lsn_t);
-	void (*iop_push)(xfs_log_item_t *);
-	void (*iop_pushbuf)(xfs_log_item_t *);
-	void (*iop_committing)(xfs_log_item_t *, xfs_lsn_t);
-} xfs_item_ops_t;
-
-#define IOP_SIZE(ip)		(*(ip)->li_ops->iop_size)(ip)
-#define IOP_FORMAT(ip,vp)	(*(ip)->li_ops->iop_format)(ip, vp)
-#define IOP_PIN(ip)		(*(ip)->li_ops->iop_pin)(ip)
-#define IOP_UNPIN(ip, flags)	(*(ip)->li_ops->iop_unpin)(ip, flags)
-#define IOP_UNPIN_REMOVE(ip,tp) (*(ip)->li_ops->iop_unpin_remove)(ip, tp)
-#define IOP_TRYLOCK(ip)		(*(ip)->li_ops->iop_trylock)(ip)
-#define IOP_UNLOCK(ip)		(*(ip)->li_ops->iop_unlock)(ip)
-#define IOP_COMMITTED(ip, lsn)	(*(ip)->li_ops->iop_committed)(ip, lsn)
-#define IOP_PUSH(ip)		(*(ip)->li_ops->iop_push)(ip)
-#define IOP_PUSHBUF(ip)		(*(ip)->li_ops->iop_pushbuf)(ip)
-#define IOP_COMMITTING(ip, lsn) (*(ip)->li_ops->iop_committing)(ip, lsn)
-
-/*
- * Return values for the IOP_TRYLOCK() routines.
- */
-#define	XFS_ITEM_SUCCESS	0
-#define	XFS_ITEM_PINNED		1
-#define	XFS_ITEM_LOCKED		2
-#define	XFS_ITEM_FLUSHING	3
-#define XFS_ITEM_PUSHBUF	4
-
-#endif	/* __KERNEL__ */
-
 /*
  * This structure is used to track log items associated with
  * a transaction.  It points to the log item and keeps some
@@ -176,7 +108,7 @@ typedef struct xfs_item_ops {
  * once we get to commit processing (see xfs_trans_commit()).
  */
 typedef struct xfs_log_item_desc {
-	xfs_log_item_t	*lid_item;
+	struct xfs_log_item	*lid_item;
 	ushort		lid_size;
 	unsigned char	lid_flags;
 	unsigned char	lid_index;
@@ -276,94 +208,6 @@ xfs_lic_desc_to_chunk(xfs_log_item_desc_t *dp)
 		(xfs_caddr_t)(((xfs_log_item_chunk_t*)0)->lic_descs));
 }
 
-#ifdef __KERNEL__
-/*
- * This structure is used to maintain a list of block ranges that have been
- * freed in the transaction.  The ranges are listed in the perag[] busy list
- * between when they're freed and the transaction is committed to disk.
- */
-
-typedef struct xfs_log_busy_slot {
-	xfs_agnumber_t		lbc_ag;
-	ushort			lbc_idx;	/* index in perag.busy[] */
-} xfs_log_busy_slot_t;
-
-#define XFS_LBC_NUM_SLOTS	31
-typedef struct xfs_log_busy_chunk {
-	struct xfs_log_busy_chunk	*lbc_next;
-	uint				lbc_free;	/* free slots bitmask */
-	ushort				lbc_unused;	/* first unused */
-	xfs_log_busy_slot_t		lbc_busy[XFS_LBC_NUM_SLOTS];
-} xfs_log_busy_chunk_t;
-
-#define	XFS_LBC_MAX_SLOT	(XFS_LBC_NUM_SLOTS - 1)
-#define	XFS_LBC_FREEMASK	((1U << XFS_LBC_NUM_SLOTS) - 1)
-
-#define	XFS_LBC_INIT(cp)	((cp)->lbc_free = XFS_LBC_FREEMASK)
-#define	XFS_LBC_CLAIM(cp, slot)	((cp)->lbc_free &= ~(1 << (slot)))
-#define	XFS_LBC_SLOT(cp, slot)	(&((cp)->lbc_busy[(slot)]))
-#define	XFS_LBC_VACANCY(cp)	(((cp)->lbc_free) & XFS_LBC_FREEMASK)
-#define	XFS_LBC_ISFREE(cp, slot) ((cp)->lbc_free & (1 << (slot)))
-
-/*
- * This is the type of function which can be given to xfs_trans_callback()
- * to be called upon the transaction's commit to disk.
- */
-typedef void (*xfs_trans_callback_t)(struct xfs_trans *, void *);
-
-/*
- * This is the structure maintained for every active transaction.
- */
-typedef struct xfs_trans {
-	unsigned int		t_magic;	/* magic number */
-	xfs_log_callback_t	t_logcb;	/* log callback struct */
-	unsigned int		t_type;		/* transaction type */
-	unsigned int		t_log_res;	/* amt of log space resvd */
-	unsigned int		t_log_count;	/* count for perm log res */
-	unsigned int		t_blk_res;	/* # of blocks resvd */
-	unsigned int		t_blk_res_used;	/* # of resvd blocks used */
-	unsigned int		t_rtx_res;	/* # of rt extents resvd */
-	unsigned int		t_rtx_res_used;	/* # of resvd rt extents used */
-	xfs_log_ticket_t	t_ticket;	/* log mgr ticket */
-	xfs_lsn_t		t_lsn;		/* log seq num of start of
-						 * transaction. */
-	xfs_lsn_t		t_commit_lsn;	/* log seq num of end of
-						 * transaction. */
-	struct xfs_mount	*t_mountp;	/* ptr to fs mount struct */
-	struct xfs_dquot_acct   *t_dqinfo;	/* acctg info for dquots */
-	xfs_trans_callback_t	t_callback;	/* transaction callback */
-	void			*t_callarg;	/* callback arg */
-	unsigned int		t_flags;	/* misc flags */
-	int64_t			t_icount_delta;	/* superblock icount change */
-	int64_t			t_ifree_delta;	/* superblock ifree change */
-	int64_t			t_fdblocks_delta; /* superblock fdblocks chg */
-	int64_t			t_res_fdblocks_delta; /* on-disk only chg */
-	int64_t			t_frextents_delta;/* superblock freextents chg*/
-	int64_t			t_res_frextents_delta; /* on-disk only chg */
-#ifdef DEBUG
-	int64_t			t_ag_freeblks_delta; /* debugging counter */
-	int64_t			t_ag_flist_delta; /* debugging counter */
-	int64_t			t_ag_btree_delta; /* debugging counter */
-#endif
-	int64_t			t_dblocks_delta;/* superblock dblocks change */
-	int64_t			t_agcount_delta;/* superblock agcount change */
-	int64_t			t_imaxpct_delta;/* superblock imaxpct change */
-	int64_t			t_rextsize_delta;/* superblock rextsize chg */
-	int64_t			t_rbmblocks_delta;/* superblock rbmblocks chg */
-	int64_t			t_rblocks_delta;/* superblock rblocks change */
-	int64_t			t_rextents_delta;/* superblocks rextents chg */
-	int64_t			t_rextslog_delta;/* superblocks rextslog chg */
-	unsigned int		t_items_free;	/* log item descs free */
-	xfs_log_item_chunk_t	t_items;	/* first log item desc chunk */
-	xfs_trans_header_t	t_header;	/* header for in-log trans */
-	unsigned int		t_busy_free;	/* busy descs free */
-	xfs_log_busy_chunk_t	t_busy;		/* busy/async free blocks */
-	unsigned long		t_pflags;	/* saved process flags state */
-} xfs_trans_t;
-
-#endif	/* __KERNEL__ */
-
-
 #define	XFS_TRANS_MAGIC		0x5452414E	/* 'TRAN' */
 /*
  * Values for t_flags.
@@ -906,6 +750,157 @@ typedef struct xfs_trans {
 #define	XFS_DQUOT_REF		1
 
 #ifdef __KERNEL__
+
+struct xfs_buf;
+struct xfs_buftarg;
+struct xfs_efd_log_item;
+struct xfs_efi_log_item;
+struct xfs_inode;
+struct xfs_item_ops;
+struct xfs_log_iovec;
+struct xfs_log_item_desc;
+struct xfs_mount;
+struct xfs_trans;
+struct xfs_dquot_acct;
+
+typedef struct xfs_log_item {
+	struct list_head		li_ail;		/* AIL pointers */
+	xfs_lsn_t			li_lsn;		/* last on-disk lsn */
+	struct xfs_log_item_desc	*li_desc;	/* ptr to current desc*/
+	struct xfs_mount		*li_mountp;	/* ptr to fs mount */
+	struct xfs_ail			*li_ailp;	/* ptr to AIL */
+	uint				li_type;	/* item type */
+	uint				li_flags;	/* misc flags */
+	struct xfs_log_item		*li_bio_list;	/* buffer item list */
+	void				(*li_cb)(struct xfs_buf *,
+						 struct xfs_log_item *);
+							/* buffer item iodone */
+							/* callback func */
+	struct xfs_item_ops		*li_ops;	/* function list */
+} xfs_log_item_t;
+
+#define	XFS_LI_IN_AIL	0x1
+#define XFS_LI_ABORTED	0x2
+
+typedef struct xfs_item_ops {
+	uint (*iop_size)(xfs_log_item_t *);
+	void (*iop_format)(xfs_log_item_t *, struct xfs_log_iovec *);
+	void (*iop_pin)(xfs_log_item_t *);
+	void (*iop_unpin)(xfs_log_item_t *, int);
+	void (*iop_unpin_remove)(xfs_log_item_t *, struct xfs_trans *);
+	uint (*iop_trylock)(xfs_log_item_t *);
+	void (*iop_unlock)(xfs_log_item_t *);
+	xfs_lsn_t (*iop_committed)(xfs_log_item_t *, xfs_lsn_t);
+	void (*iop_push)(xfs_log_item_t *);
+	void (*iop_pushbuf)(xfs_log_item_t *);
+	void (*iop_committing)(xfs_log_item_t *, xfs_lsn_t);
+} xfs_item_ops_t;
+
+#define IOP_SIZE(ip)		(*(ip)->li_ops->iop_size)(ip)
+#define IOP_FORMAT(ip,vp)	(*(ip)->li_ops->iop_format)(ip, vp)
+#define IOP_PIN(ip)		(*(ip)->li_ops->iop_pin)(ip)
+#define IOP_UNPIN(ip, flags)	(*(ip)->li_ops->iop_unpin)(ip, flags)
+#define IOP_UNPIN_REMOVE(ip,tp) (*(ip)->li_ops->iop_unpin_remove)(ip, tp)
+#define IOP_TRYLOCK(ip)		(*(ip)->li_ops->iop_trylock)(ip)
+#define IOP_UNLOCK(ip)		(*(ip)->li_ops->iop_unlock)(ip)
+#define IOP_COMMITTED(ip, lsn)	(*(ip)->li_ops->iop_committed)(ip, lsn)
+#define IOP_PUSH(ip)		(*(ip)->li_ops->iop_push)(ip)
+#define IOP_PUSHBUF(ip)		(*(ip)->li_ops->iop_pushbuf)(ip)
+#define IOP_COMMITTING(ip, lsn) (*(ip)->li_ops->iop_committing)(ip, lsn)
+
+/*
+ * Return values for the IOP_TRYLOCK() routines.
+ */
+#define	XFS_ITEM_SUCCESS	0
+#define	XFS_ITEM_PINNED		1
+#define	XFS_ITEM_LOCKED		2
+#define	XFS_ITEM_FLUSHING	3
+#define XFS_ITEM_PUSHBUF	4
+
+/*
+ * This structure is used to maintain a list of block ranges that have been
+ * freed in the transaction.  The ranges are listed in the perag[] busy list
+ * between when they're freed and the transaction is committed to disk.
+ */
+
+typedef struct xfs_log_busy_slot {
+	xfs_agnumber_t		lbc_ag;
+	ushort			lbc_idx;	/* index in perag.busy[] */
+} xfs_log_busy_slot_t;
+
+#define XFS_LBC_NUM_SLOTS	31
+typedef struct xfs_log_busy_chunk {
+	struct xfs_log_busy_chunk	*lbc_next;
+	uint				lbc_free;	/* free slots bitmask */
+	ushort				lbc_unused;	/* first unused */
+	xfs_log_busy_slot_t		lbc_busy[XFS_LBC_NUM_SLOTS];
+} xfs_log_busy_chunk_t;
+
+#define	XFS_LBC_MAX_SLOT	(XFS_LBC_NUM_SLOTS - 1)
+#define	XFS_LBC_FREEMASK	((1U << XFS_LBC_NUM_SLOTS) - 1)
+
+#define	XFS_LBC_INIT(cp)	((cp)->lbc_free = XFS_LBC_FREEMASK)
+#define	XFS_LBC_CLAIM(cp, slot)	((cp)->lbc_free &= ~(1 << (slot)))
+#define	XFS_LBC_SLOT(cp, slot)	(&((cp)->lbc_busy[(slot)]))
+#define	XFS_LBC_VACANCY(cp)	(((cp)->lbc_free) & XFS_LBC_FREEMASK)
+#define	XFS_LBC_ISFREE(cp, slot) ((cp)->lbc_free & (1 << (slot)))
+
+/*
+ * This is the type of function which can be given to xfs_trans_callback()
+ * to be called upon the transaction's commit to disk.
+ */
+typedef void (*xfs_trans_callback_t)(struct xfs_trans *, void *);
+
+/*
+ * This is the structure maintained for every active transaction.
+ */
+typedef struct xfs_trans {
+	unsigned int		t_magic;	/* magic number */
+	xfs_log_callback_t	t_logcb;	/* log callback struct */
+	unsigned int		t_type;		/* transaction type */
+	unsigned int		t_log_res;	/* amt of log space resvd */
+	unsigned int		t_log_count;	/* count for perm log res */
+	unsigned int		t_blk_res;	/* # of blocks resvd */
+	unsigned int		t_blk_res_used;	/* # of resvd blocks used */
+	unsigned int		t_rtx_res;	/* # of rt extents resvd */
+	unsigned int		t_rtx_res_used;	/* # of resvd rt extents used */
+	xfs_log_ticket_t	t_ticket;	/* log mgr ticket */
+	xfs_lsn_t		t_lsn;		/* log seq num of start of
+						 * transaction. */
+	xfs_lsn_t		t_commit_lsn;	/* log seq num of end of
+						 * transaction. */
+	struct xfs_mount	*t_mountp;	/* ptr to fs mount struct */
+	struct xfs_dquot_acct   *t_dqinfo;	/* acctg info for dquots */
+	xfs_trans_callback_t	t_callback;	/* transaction callback */
+	void			*t_callarg;	/* callback arg */
+	unsigned int		t_flags;	/* misc flags */
+	int64_t			t_icount_delta;	/* superblock icount change */
+	int64_t			t_ifree_delta;	/* superblock ifree change */
+	int64_t			t_fdblocks_delta; /* superblock fdblocks chg */
+	int64_t			t_res_fdblocks_delta; /* on-disk only chg */
+	int64_t			t_frextents_delta;/* superblock freextents chg*/
+	int64_t			t_res_frextents_delta; /* on-disk only chg */
+#ifdef DEBUG
+	int64_t			t_ag_freeblks_delta; /* debugging counter */
+	int64_t			t_ag_flist_delta; /* debugging counter */
+	int64_t			t_ag_btree_delta; /* debugging counter */
+#endif
+	int64_t			t_dblocks_delta;/* superblock dblocks change */
+	int64_t			t_agcount_delta;/* superblock agcount change */
+	int64_t			t_imaxpct_delta;/* superblock imaxpct change */
+	int64_t			t_rextsize_delta;/* superblock rextsize chg */
+	int64_t			t_rbmblocks_delta;/* superblock rbmblocks chg */
+	int64_t			t_rblocks_delta;/* superblock rblocks change */
+	int64_t			t_rextents_delta;/* superblocks rextents chg */
+	int64_t			t_rextslog_delta;/* superblocks rextslog chg */
+	unsigned int		t_items_free;	/* log item descs free */
+	xfs_log_item_chunk_t	t_items;	/* first log item desc chunk */
+	xfs_trans_header_t	t_header;	/* header for in-log trans */
+	unsigned int		t_busy_free;	/* busy descs free */
+	xfs_log_busy_chunk_t	t_busy;		/* busy/async free blocks */
+	unsigned long		t_pflags;	/* saved process flags state */
+} xfs_trans_t;
+
 /*
  * XFS transaction mechanism exported interfaces that are
  * actually macros.
@@ -928,7 +923,6 @@ typedef struct xfs_trans {
 /*
  * XFS transaction mechanism exported interfaces.
  */
-void		xfs_trans_init(struct xfs_mount *);
 xfs_trans_t	*xfs_trans_alloc(struct xfs_mount *, uint);
 xfs_trans_t	*_xfs_trans_alloc(struct xfs_mount *, uint);
 xfs_trans_t	*xfs_trans_dup(xfs_trans_t *);
@@ -975,13 +969,8 @@ int		_xfs_trans_commit(xfs_trans_t *,
 				  int *);
 #define xfs_trans_commit(tp, flags)	_xfs_trans_commit(tp, flags, NULL)
 void		xfs_trans_cancel(xfs_trans_t *, int);
-int		xfs_trans_roll(struct xfs_trans **, struct xfs_inode *);
 int		xfs_trans_ail_init(struct xfs_mount *);
 void		xfs_trans_ail_destroy(struct xfs_mount *);
-void		xfs_trans_push_ail(struct xfs_mount *, xfs_lsn_t);
-xfs_lsn_t	xfs_trans_tail_ail(struct xfs_mount *);
-void		xfs_trans_unlocked_item(struct xfs_mount *,
-					xfs_log_item_t *);
 xfs_log_busy_slot_t *xfs_trans_add_busy(xfs_trans_t *tp,
 					xfs_agnumber_t ag,
 					xfs_extlen_t idx);
@@ -990,4 +979,7 @@ extern kmem_zone_t	*xfs_trans_zone;
 
 #endif	/* __KERNEL__ */
 
+void		xfs_trans_init(struct xfs_mount *);
+int		xfs_trans_roll(struct xfs_trans **, struct xfs_inode *);
+
 #endif	/* __XFS_TRANS_H__ */

fs/xfs/xfs_trans_ail.c

@@ -1,5 +1,6 @@
 /*
  * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
+ * Copyright (c) 2008 Dave Chinner
  * All Rights Reserved.
  *
  * This program is free software; you can redistribute it and/or
@@ -28,13 +29,13 @@
 #include "xfs_trans_priv.h"
 #include "xfs_error.h"
 
-STATIC void xfs_ail_insert(xfs_ail_t *, xfs_log_item_t *);
-STATIC xfs_log_item_t * xfs_ail_delete(xfs_ail_t *, xfs_log_item_t *);
-STATIC xfs_log_item_t * xfs_ail_min(xfs_ail_t *);
-STATIC xfs_log_item_t * xfs_ail_next(xfs_ail_t *, xfs_log_item_t *);
+STATIC void xfs_ail_insert(struct xfs_ail *, xfs_log_item_t *);
+STATIC xfs_log_item_t * xfs_ail_delete(struct xfs_ail *, xfs_log_item_t *);
+STATIC xfs_log_item_t * xfs_ail_min(struct xfs_ail *);
+STATIC xfs_log_item_t * xfs_ail_next(struct xfs_ail *, xfs_log_item_t *);
 
 #ifdef DEBUG
-STATIC void xfs_ail_check(xfs_ail_t *, xfs_log_item_t *);
+STATIC void xfs_ail_check(struct xfs_ail *, xfs_log_item_t *);
 #else
 #define	xfs_ail_check(a,l)
 #endif /* DEBUG */
@@ -50,20 +51,20 @@ STATIC void xfs_ail_check(xfs_ail_t *, xfs_log_item_t *);
  * lsn of the last item in the AIL.
  */
 xfs_lsn_t
-xfs_trans_tail_ail(
-	xfs_mount_t	*mp)
+xfs_trans_ail_tail(
+	struct xfs_ail	*ailp)
 {
 	xfs_lsn_t	lsn;
 	xfs_log_item_t	*lip;
 
-	spin_lock(&mp->m_ail_lock);
-	lip = xfs_ail_min(&mp->m_ail);
+	spin_lock(&ailp->xa_lock);
+	lip = xfs_ail_min(ailp);
 	if (lip == NULL) {
 		lsn = (xfs_lsn_t)0;
 	} else {
 		lsn = lip->li_lsn;
 	}
-	spin_unlock(&mp->m_ail_lock);
+	spin_unlock(&ailp->xa_lock);
 
 	return lsn;
 }
@@ -85,16 +86,125 @@ xfs_trans_tail_ail(
  * any of the objects, so the lock is not needed.
  */
 void
-xfs_trans_push_ail(
-	xfs_mount_t		*mp,
-	xfs_lsn_t		threshold_lsn)
+xfs_trans_ail_push(
+	struct xfs_ail	*ailp,
+	xfs_lsn_t	threshold_lsn)
 {
-	xfs_log_item_t		*lip;
+	xfs_log_item_t	*lip;
+
+	lip = xfs_ail_min(ailp);
+	if (lip && !XFS_FORCED_SHUTDOWN(ailp->xa_mount)) {
+		if (XFS_LSN_CMP(threshold_lsn, ailp->xa_target) > 0)
+			xfsaild_wakeup(ailp, threshold_lsn);
+	}
+}
+
+/*
+ * AIL traversal cursor initialisation.
+ *
+ * The cursor keeps track of where our current traversal is up
+ * to by tracking the next ƣtem in the list for us. However, for
+ * this to be safe, removing an object from the AIL needs to invalidate
+ * any cursor that points to it. hence the traversal cursor needs to
+ * be linked to the struct xfs_ail so that deletion can search all the
+ * active cursors for invalidation.
+ *
+ * We don't link the push cursor because it is embedded in the struct
+ * xfs_ail and hence easily findable.
+ */
+STATIC void
+xfs_trans_ail_cursor_init(
+	struct xfs_ail		*ailp,
+	struct xfs_ail_cursor	*cur)
+{
+	cur->item = NULL;
+	if (cur == &ailp->xa_cursors)
+		return;
+
+	cur->next = ailp->xa_cursors.next;
+	ailp->xa_cursors.next = cur;
+}
+
+/*
+ * Set the cursor to the next item, because when we look
+ * up the cursor the current item may have been freed.
+ */
+STATIC void
+xfs_trans_ail_cursor_set(
+	struct xfs_ail		*ailp,
+	struct xfs_ail_cursor	*cur,
+	struct xfs_log_item	*lip)
+{
+	if (lip)
+		cur->item = xfs_ail_next(ailp, lip);
+}
+
+/*
+ * Get the next item in the traversal and advance the cursor.
+ * If the cursor was invalidated (inidicated by a lip of 1),
+ * restart the traversal.
+ */
+struct xfs_log_item *
+xfs_trans_ail_cursor_next(
+	struct xfs_ail		*ailp,
+	struct xfs_ail_cursor	*cur)
+{
+	struct xfs_log_item	*lip = cur->item;
+
+	if ((__psint_t)lip & 1)
+		lip = xfs_ail_min(ailp);
+	xfs_trans_ail_cursor_set(ailp, cur, lip);
+	return lip;
+}
+
+/*
+ * Now that the traversal is complete, we need to remove the cursor
+ * from the list of traversing cursors. Avoid removing the embedded
+ * push cursor, but use the fact it is alway present to make the
+ * list deletion simple.
+ */
+void
+xfs_trans_ail_cursor_done(
+	struct xfs_ail		*ailp,
+	struct xfs_ail_cursor	*done)
+{
+	struct xfs_ail_cursor	*prev = NULL;
+	struct xfs_ail_cursor	*cur;
+
+	done->item = NULL;
+	if (done == &ailp->xa_cursors)
+		return;
+	prev = &ailp->xa_cursors;
+	for (cur = prev->next; cur; prev = cur, cur = prev->next) {
+		if (cur == done) {
+			prev->next = cur->next;
+			break;
+		}
+	}
+	ASSERT(cur);
+}
+
+/*
+ * Invalidate any cursor that is pointing to this item. This is
+ * called when an item is removed from the AIL. Any cursor pointing
+ * to this object is now invalid and the traversal needs to be
+ * terminated so it doesn't reference a freed object. We set the
+ * cursor item to a value of 1 so we can distinguish between an
+ * invalidation and the end of the list when getting the next item
+ * from the cursor.
+ */
+STATIC void
+xfs_trans_ail_cursor_clear(
+	struct xfs_ail		*ailp,
+	struct xfs_log_item	*lip)
+{
+	struct xfs_ail_cursor	*cur;
 
-	lip = xfs_ail_min(&mp->m_ail);
-	if (lip && !XFS_FORCED_SHUTDOWN(mp)) {
-		if (XFS_LSN_CMP(threshold_lsn, mp->m_ail.xa_target) > 0)
-			xfsaild_wakeup(mp, threshold_lsn);
+	/* need to search all cursors */
+	for (cur = &ailp->xa_cursors; cur; cur = cur->next) {
+		if (cur->item == lip)
+			cur->item = (struct xfs_log_item *)
+					((__psint_t)cur->item | 1);
 	}
 }
 
@@ -103,25 +213,27 @@ xfs_trans_push_ail(
  * Return the current tree generation number for use
  * in calls to xfs_trans_next_ail().
  */
-STATIC xfs_log_item_t *
-xfs_trans_first_push_ail(
-	xfs_mount_t	*mp,
-	int		*gen,
-	xfs_lsn_t	lsn)
+xfs_log_item_t *
+xfs_trans_ail_cursor_first(
+	struct xfs_ail		*ailp,
+	struct xfs_ail_cursor	*cur,
+	xfs_lsn_t		lsn)
 {
-	xfs_log_item_t	*lip;
+	xfs_log_item_t		*lip;
 
-	lip = xfs_ail_min(&mp->m_ail);
-	*gen = (int)mp->m_ail.xa_gen;
+	xfs_trans_ail_cursor_init(ailp, cur);
+	lip = xfs_ail_min(ailp);
 	if (lsn == 0)
-		return lip;
+		goto out;
 
-	list_for_each_entry(lip, &mp->m_ail.xa_ail, li_ail) {
+	list_for_each_entry(lip, &ailp->xa_ail, li_ail) {
 		if (XFS_LSN_CMP(lip->li_lsn, lsn) >= 0)
-			return lip;
+			goto out;
 	}
-
-	return NULL;
+	lip = NULL;
+out:
+	xfs_trans_ail_cursor_set(ailp, cur, lip);
+	return lip;
 }
 
 /*
@@ -129,29 +241,29 @@ xfs_trans_first_push_ail(
  */
 long
 xfsaild_push(
-	xfs_mount_t	*mp,
+	struct xfs_ail	*ailp,
 	xfs_lsn_t	*last_lsn)
 {
 	long		tout = 1000; /* milliseconds */
 	xfs_lsn_t	last_pushed_lsn = *last_lsn;
-	xfs_lsn_t	target =  mp->m_ail.xa_target;
+	xfs_lsn_t	target =  ailp->xa_target;
 	xfs_lsn_t	lsn;
 	xfs_log_item_t	*lip;
-	int		gen;
-	int		restarts;
 	int		flush_log, count, stuck;
+	xfs_mount_t	*mp = ailp->xa_mount;
+	struct xfs_ail_cursor	*cur = &ailp->xa_cursors;
 
-#define	XFS_TRANS_PUSH_AIL_RESTARTS	10
-
-	spin_lock(&mp->m_ail_lock);
-	lip = xfs_trans_first_push_ail(mp, &gen, *last_lsn);
+	spin_lock(&ailp->xa_lock);
+	xfs_trans_ail_cursor_init(ailp, cur);
+	lip = xfs_trans_ail_cursor_first(ailp, cur, *last_lsn);
 	if (!lip || XFS_FORCED_SHUTDOWN(mp)) {
 		/*
 		 * AIL is empty or our push has reached the end.
 		 */
-		spin_unlock(&mp->m_ail_lock);
+		xfs_trans_ail_cursor_done(ailp, cur);
+		spin_unlock(&ailp->xa_lock);
 		last_pushed_lsn = 0;
-		goto out;
+		return tout;
 	}
 
 	XFS_STATS_INC(xs_push_ail);
@@ -169,7 +281,7 @@ xfsaild_push(
 	 */
 	tout = 10;
 	lsn = lip->li_lsn;
-	flush_log = stuck = count = restarts = 0;
+	flush_log = stuck = count = 0;
 	while ((XFS_LSN_CMP(lip->li_lsn, target) < 0)) {
 		int	lock_result;
 		/*
@@ -184,7 +296,7 @@ xfsaild_push(
 		 * skip to the next item in the list.
 		 */
 		lock_result = IOP_TRYLOCK(lip);
-		spin_unlock(&mp->m_ail_lock);
+		spin_unlock(&ailp->xa_lock);
 		switch (lock_result) {
 		case XFS_ITEM_SUCCESS:
 			XFS_STATS_INC(xs_push_ail_success);
@@ -221,7 +333,7 @@ xfsaild_push(
 			break;
 		}
 
-		spin_lock(&mp->m_ail_lock);
+		spin_lock(&ailp->xa_lock);
 		/* should we bother continuing? */
 		if (XFS_FORCED_SHUTDOWN(mp))
 			break;
@@ -244,14 +356,13 @@ xfsaild_push(
 		if (stuck > 100)
 			break;
 
-		lip = xfs_trans_next_ail(mp, lip, &gen, &restarts);
+		lip = xfs_trans_ail_cursor_next(ailp, cur);
 		if (lip == NULL)
 			break;
-		if (restarts > XFS_TRANS_PUSH_AIL_RESTARTS)
-			break;
 		lsn = lip->li_lsn;
 	}
-	spin_unlock(&mp->m_ail_lock);
+	xfs_trans_ail_cursor_done(ailp, cur);
+	spin_unlock(&ailp->xa_lock);
 
 	if (flush_log) {
 		/*
@@ -274,8 +385,7 @@ xfsaild_push(
 		 */
 		tout += 20;
 		last_pushed_lsn = 0;
-	} else if ((restarts > XFS_TRANS_PUSH_AIL_RESTARTS) ||
-		   ((stuck * 100) / count > 90)) {
+	} else if ((stuck * 100) / count > 90) {
 		/*
 		 * Either there is a lot of contention on the AIL or we
 		 * are stuck due to operations in progress. "Stuck" in this
@@ -287,7 +397,6 @@ xfsaild_push(
 		 */
 		tout += 10;
 	}
-out:
 	*last_lsn = last_pushed_lsn;
 	return tout;
 }	/* xfsaild_push */
@@ -303,7 +412,7 @@ out:
  */
 void
 xfs_trans_unlocked_item(
-	xfs_mount_t	*mp,
+	struct xfs_ail	*ailp,
 	xfs_log_item_t	*lip)
 {
 	xfs_log_item_t	*min_lip;
@@ -315,7 +424,7 @@ xfs_trans_unlocked_item(
 	 * over some potentially valid data.
 	 */
 	if (!(lip->li_flags & XFS_LI_IN_AIL) ||
-	    XFS_FORCED_SHUTDOWN(mp)) {
+	    XFS_FORCED_SHUTDOWN(ailp->xa_mount)) {
 		return;
 	}
 
@@ -331,10 +440,10 @@ xfs_trans_unlocked_item(
 	 * the call to xfs_log_move_tail() doesn't do anything if there's
 	 * not enough free space to wake people up so we're safe calling it.
 	 */
-	min_lip = xfs_ail_min(&mp->m_ail);
+	min_lip = xfs_ail_min(ailp);
 
 	if (min_lip == lip)
-		xfs_log_move_tail(mp, 1);
+		xfs_log_move_tail(ailp->xa_mount, 1);
 }	/* xfs_trans_unlocked_item */
 
 
@@ -347,41 +456,37 @@ xfs_trans_unlocked_item(
  * we move in the AIL is the minimum one, update the tail lsn in the
  * log manager.
  *
- * Increment the AIL's generation count to indicate that the tree
- * has changed.
- *
  * This function must be called with the AIL lock held.  The lock
  * is dropped before returning.
  */
 void
-xfs_trans_update_ail(
-	xfs_mount_t	*mp,
+xfs_trans_ail_update(
+	struct xfs_ail	*ailp,
 	xfs_log_item_t	*lip,
-	xfs_lsn_t	lsn) __releases(mp->m_ail_lock)
+	xfs_lsn_t	lsn) __releases(ailp->xa_lock)
 {
-	xfs_log_item_t		*dlip=NULL;
+	xfs_log_item_t		*dlip = NULL;
 	xfs_log_item_t		*mlip;	/* ptr to minimum lip */
 
-	mlip = xfs_ail_min(&mp->m_ail);
+	mlip = xfs_ail_min(ailp);
 
 	if (lip->li_flags & XFS_LI_IN_AIL) {
-		dlip = xfs_ail_delete(&mp->m_ail, lip);
+		dlip = xfs_ail_delete(ailp, lip);
 		ASSERT(dlip == lip);
+		xfs_trans_ail_cursor_clear(ailp, dlip);
 	} else {
 		lip->li_flags |= XFS_LI_IN_AIL;
 	}
 
 	lip->li_lsn = lsn;
-
-	xfs_ail_insert(&mp->m_ail, lip);
-	mp->m_ail.xa_gen++;
+	xfs_ail_insert(ailp, lip);
 
 	if (mlip == dlip) {
-		mlip = xfs_ail_min(&mp->m_ail);
-		spin_unlock(&mp->m_ail_lock);
-		xfs_log_move_tail(mp, mlip->li_lsn);
+		mlip = xfs_ail_min(ailp);
+		spin_unlock(&ailp->xa_lock);
+		xfs_log_move_tail(ailp->xa_mount, mlip->li_lsn);
 	} else {
-		spin_unlock(&mp->m_ail_lock);
+		spin_unlock(&ailp->xa_lock);
 	}
 
 
@@ -403,29 +508,30 @@ xfs_trans_update_ail(
  * is dropped before returning.
  */
 void
-xfs_trans_delete_ail(
-	xfs_mount_t	*mp,
-	xfs_log_item_t	*lip) __releases(mp->m_ail_lock)
+xfs_trans_ail_delete(
+	struct xfs_ail	*ailp,
+	xfs_log_item_t	*lip) __releases(ailp->xa_lock)
 {
 	xfs_log_item_t		*dlip;
 	xfs_log_item_t		*mlip;
 
 	if (lip->li_flags & XFS_LI_IN_AIL) {
-		mlip = xfs_ail_min(&mp->m_ail);
-		dlip = xfs_ail_delete(&mp->m_ail, lip);
+		mlip = xfs_ail_min(ailp);
+		dlip = xfs_ail_delete(ailp, lip);
 		ASSERT(dlip == lip);
+		xfs_trans_ail_cursor_clear(ailp, dlip);
 
 
 		lip->li_flags &= ~XFS_LI_IN_AIL;
 		lip->li_lsn = 0;
-		mp->m_ail.xa_gen++;
 
 		if (mlip == dlip) {
-			mlip = xfs_ail_min(&mp->m_ail);
-			spin_unlock(&mp->m_ail_lock);
-			xfs_log_move_tail(mp, (mlip ? mlip->li_lsn : 0));
+			mlip = xfs_ail_min(ailp);
+			spin_unlock(&ailp->xa_lock);
+			xfs_log_move_tail(ailp->xa_mount,
+						(mlip ? mlip->li_lsn : 0));
 		} else {
-			spin_unlock(&mp->m_ail_lock);
+			spin_unlock(&ailp->xa_lock);
 		}
 	}
 	else {
@@ -433,13 +539,13 @@ xfs_trans_delete_ail(
 		 * If the file system is not being shutdown, we are in
 		 * serious trouble if we get to this stage.
 		 */
-		if (XFS_FORCED_SHUTDOWN(mp))
-			spin_unlock(&mp->m_ail_lock);
-		else {
+		struct xfs_mount	*mp = ailp->xa_mount;
+
+		spin_unlock(&ailp->xa_lock);
+		if (!XFS_FORCED_SHUTDOWN(mp)) {
 			xfs_cmn_err(XFS_PTAG_AILDELETE, CE_ALERT, mp,
 		"%s: attempting to delete a log item that is not in the AIL",
 					__func__);
-			spin_unlock(&mp->m_ail_lock);
 			xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
 		}
 	}
@@ -447,56 +553,6 @@ xfs_trans_delete_ail(
 
 
 
-/*
- * Return the item in the AIL with the smallest lsn.
- * Return the current tree generation number for use
- * in calls to xfs_trans_next_ail().
- */
-xfs_log_item_t *
-xfs_trans_first_ail(
-	xfs_mount_t	*mp,
-	int		*gen)
-{
-	xfs_log_item_t	*lip;
-
-	lip = xfs_ail_min(&mp->m_ail);
-	*gen = (int)mp->m_ail.xa_gen;
-
-	return lip;
-}
-
-/*
- * If the generation count of the tree has not changed since the
- * caller last took something from the AIL, then return the elmt
- * in the tree which follows the one given.  If the count has changed,
- * then return the minimum elmt of the AIL and bump the restarts counter
- * if one is given.
- */
-xfs_log_item_t *
-xfs_trans_next_ail(
-	xfs_mount_t	*mp,
-	xfs_log_item_t	*lip,
-	int		*gen,
-	int		*restarts)
-{
-	xfs_log_item_t	*nlip;
-
-	ASSERT(mp && lip && gen);
-	if (mp->m_ail.xa_gen == *gen) {
-		nlip = xfs_ail_next(&mp->m_ail, lip);
-	} else {
-		nlip = xfs_ail_min(&mp->m_ail);
-		*gen = (int)mp->m_ail.xa_gen;
-		if (restarts != NULL) {
-			XFS_STATS_INC(xs_push_ail_restarts);
-			(*restarts)++;
-		}
-	}
-
-	return (nlip);
-}
-
-
 /*
  * The active item list (AIL) is a doubly linked list of log
  * items sorted by ascending lsn.  The base of the list is
@@ -515,15 +571,35 @@ int
 xfs_trans_ail_init(
 	xfs_mount_t	*mp)
 {
-	INIT_LIST_HEAD(&mp->m_ail.xa_ail);
-	return xfsaild_start(mp);
+	struct xfs_ail	*ailp;
+	int		error;
+
+	ailp = kmem_zalloc(sizeof(struct xfs_ail), KM_MAYFAIL);
+	if (!ailp)
+		return ENOMEM;
+
+	ailp->xa_mount = mp;
+	INIT_LIST_HEAD(&ailp->xa_ail);
+	spin_lock_init(&ailp->xa_lock);
+	error = xfsaild_start(ailp);
+	if (error)
+		goto out_free_ailp;
+	mp->m_ail = ailp;
+	return 0;
+
+out_free_ailp:
+	kmem_free(ailp);
+	return error;
 }
 
 void
 xfs_trans_ail_destroy(
 	xfs_mount_t	*mp)
 {
-	xfsaild_stop(mp);
+	struct xfs_ail	*ailp = mp->m_ail;
+
+	xfsaild_stop(ailp);
+	kmem_free(ailp);
 }
 
 /*
@@ -534,7 +610,7 @@ xfs_trans_ail_destroy(
  */
 STATIC void
 xfs_ail_insert(
-	xfs_ail_t	*ailp,
+	struct xfs_ail	*ailp,
 	xfs_log_item_t	*lip)
 /* ARGSUSED */
 {
@@ -568,7 +644,7 @@ xfs_ail_insert(
 /*ARGSUSED*/
 STATIC xfs_log_item_t *
 xfs_ail_delete(
-	xfs_ail_t	*ailp,
+	struct xfs_ail	*ailp,
 	xfs_log_item_t	*lip)
 /* ARGSUSED */
 {
@@ -585,7 +661,7 @@ xfs_ail_delete(
  */
 STATIC xfs_log_item_t *
 xfs_ail_min(
-	xfs_ail_t	*ailp)
+	struct xfs_ail	*ailp)
 /* ARGSUSED */
 {
 	if (list_empty(&ailp->xa_ail))
@@ -601,7 +677,7 @@ xfs_ail_min(
  */
 STATIC xfs_log_item_t *
 xfs_ail_next(
-	xfs_ail_t	*ailp,
+	struct xfs_ail	*ailp,
 	xfs_log_item_t	*lip)
 /* ARGSUSED */
 {
@@ -617,7 +693,7 @@ xfs_ail_next(
  */
 STATIC void
 xfs_ail_check(
-	xfs_ail_t 	*ailp,
+	struct xfs_ail	*ailp,
 	xfs_log_item_t	*lip)
 {
 	xfs_log_item_t	*prev_lip;

fs/xfs/xfs_trans_buf.c

@@ -527,9 +527,8 @@ xfs_trans_brelse(xfs_trans_t	*tp,
 			lip = XFS_BUF_FSPRIVATE(bp, xfs_log_item_t *);
 			if (lip->li_type == XFS_LI_BUF) {
 				bip = XFS_BUF_FSPRIVATE(bp,xfs_buf_log_item_t*);
-				xfs_trans_unlocked_item(
-						bip->bli_item.li_mountp,
-						lip);
+				xfs_trans_unlocked_item(bip->bli_item.li_ailp,
+							lip);
 			}
 		}
 		xfs_buf_relse(bp);
@@ -626,7 +625,7 @@ xfs_trans_brelse(xfs_trans_t	*tp,
 	 * tell the AIL that the buffer is being unlocked.
 	 */
 	if (bip != NULL) {
-		xfs_trans_unlocked_item(bip->bli_item.li_mountp,
+		xfs_trans_unlocked_item(bip->bli_item.li_ailp,
 					(xfs_log_item_t*)bip);
 	}
 

fs/xfs/xfs_trans_item.c

@@ -22,6 +22,14 @@
 #include "xfs_inum.h"
 #include "xfs_trans.h"
 #include "xfs_trans_priv.h"
+/* XXX: from here down needed until struct xfs_trans has it's own ailp */
+#include "xfs_bit.h"
+#include "xfs_buf_item.h"
+#include "xfs_sb.h"
+#include "xfs_ag.h"
+#include "xfs_dir2.h"
+#include "xfs_dmapi.h"
+#include "xfs_mount.h"
 
 STATIC int	xfs_trans_unlock_chunk(xfs_log_item_chunk_t *,
 					int, int, xfs_lsn_t);
@@ -79,6 +87,7 @@ xfs_trans_add_item(xfs_trans_t *tp, xfs_log_item_t *lip)
 		lidp->lid_size = 0;
 		lip->li_desc = lidp;
 		lip->li_mountp = tp->t_mountp;
+		lip->li_ailp = tp->t_mountp->m_ail;
 		return lidp;
 	}
 
@@ -120,6 +129,7 @@ xfs_trans_add_item(xfs_trans_t *tp, xfs_log_item_t *lip)
 	lidp->lid_size = 0;
 	lip->li_desc = lidp;
 	lip->li_mountp = tp->t_mountp;
+	lip->li_ailp = tp->t_mountp->m_ail;
 	return lidp;
 }
 

fs/xfs/xfs_trans_priv.h

@@ -44,25 +44,93 @@ xfs_log_busy_slot_t		*xfs_trans_add_busy(xfs_trans_t *tp,
 						    xfs_extlen_t idx);
 
 /*
- * From xfs_trans_ail.c
+ * AIL traversal cursor.
+ *
+ * Rather than using a generation number for detecting changes in the ail, use
+ * a cursor that is protected by the ail lock. The aild cursor exists in the
+ * struct xfs_ail, but other traversals can declare it on the stack and link it
+ * to the ail list.
+ *
+ * When an object is deleted from or moved int the AIL, the cursor list is
+ * searched to see if the object is a designated cursor item. If it is, it is
+ * deleted from the cursor so that the next time the cursor is used traversal
+ * will return to the start.
+ *
+ * This means a traversal colliding with a removal will cause a restart of the
+ * list scan, rather than any insertion or deletion anywhere in the list. The
+ * low bit of the item pointer is set if the cursor has been invalidated so
+ * that we can tell the difference between invalidation and reaching the end
+ * of the list to trigger traversal restarts.
  */
-void			xfs_trans_update_ail(struct xfs_mount *mp,
-				     struct xfs_log_item *lip, xfs_lsn_t lsn)
-				     __releases(mp->m_ail_lock);
-void			xfs_trans_delete_ail(struct xfs_mount *mp,
-				     struct xfs_log_item *lip)
-				     __releases(mp->m_ail_lock);
-struct xfs_log_item	*xfs_trans_first_ail(struct xfs_mount *, int *);
-struct xfs_log_item	*xfs_trans_next_ail(struct xfs_mount *,
-				     struct xfs_log_item *, int *, int *);
+struct xfs_ail_cursor {
+	struct xfs_ail_cursor	*next;
+	struct xfs_log_item	*item;
+};
 
+/*
+ * Private AIL structures.
+ *
+ * Eventually we need to drive the locking in here as well.
+ */
+struct xfs_ail {
+	struct xfs_mount	*xa_mount;
+	struct list_head	xa_ail;
+	uint			xa_gen;
+	struct task_struct	*xa_task;
+	xfs_lsn_t		xa_target;
+	struct xfs_ail_cursor	xa_cursors;
+	spinlock_t		xa_lock;
+};
 
 /*
- * AIL push thread support
+ * From xfs_trans_ail.c
  */
-long	xfsaild_push(struct xfs_mount *, xfs_lsn_t *);
-void	xfsaild_wakeup(struct xfs_mount *, xfs_lsn_t);
-int	xfsaild_start(struct xfs_mount *);
-void	xfsaild_stop(struct xfs_mount *);
+void			xfs_trans_ail_update(struct xfs_ail *ailp,
+					struct xfs_log_item *lip, xfs_lsn_t lsn)
+					__releases(ailp->xa_lock);
+void			xfs_trans_ail_delete(struct xfs_ail *ailp,
+					struct xfs_log_item *lip)
+					__releases(ailp->xa_lock);
+void			xfs_trans_ail_push(struct xfs_ail *, xfs_lsn_t);
+void			xfs_trans_unlocked_item(struct xfs_ail *,
+					xfs_log_item_t *);
+
+xfs_lsn_t		xfs_trans_ail_tail(struct xfs_ail *ailp);
+
+struct xfs_log_item	*xfs_trans_ail_cursor_first(struct xfs_ail *ailp,
+					struct xfs_ail_cursor *cur,
+					xfs_lsn_t lsn);
+struct xfs_log_item	*xfs_trans_ail_cursor_next(struct xfs_ail *ailp,
+					struct xfs_ail_cursor *cur);
+void			xfs_trans_ail_cursor_done(struct xfs_ail *ailp,
+					struct xfs_ail_cursor *cur);
+
+long	xfsaild_push(struct xfs_ail *, xfs_lsn_t *);
+void	xfsaild_wakeup(struct xfs_ail *, xfs_lsn_t);
+int	xfsaild_start(struct xfs_ail *);
+void	xfsaild_stop(struct xfs_ail *);
 
+#if BITS_PER_LONG != 64
+static inline void
+xfs_trans_ail_copy_lsn(
+	struct xfs_ail	*ailp,
+	xfs_lsn_t	*dst,
+	xfs_lsn_t	*src)
+{
+	ASSERT(sizeof(xfs_lsn_t) == 8);	/* don't lock if it shrinks */
+	spin_lock(&ailp->xa_lock);
+	*dst = *src;
+	spin_unlock(&ailp->xa_lock);
+}
+#else
+static inline void
+xfs_trans_ail_copy_lsn(
+	struct xfs_ail	*ailp,
+	xfs_lsn_t	*dst,
+	xfs_lsn_t	*src)
+{
+	ASSERT(sizeof(xfs_lsn_t) == 8);
+	*dst = *src;
+}
+#endif
 #endif	/* __XFS_TRANS_PRIV_H__ */

fs/xfs/xfs_vfsops.c

@@ -49,71 +49,15 @@
 #include "xfs_extfree_item.h"
 #include "xfs_acl.h"
 #include "xfs_attr.h"
-#include "xfs_clnt.h"
 #include "xfs_mru_cache.h"
 #include "xfs_filestream.h"
 #include "xfs_fsops.h"
 #include "xfs_vnodeops.h"
 #include "xfs_vfsops.h"
 #include "xfs_utils.h"
+#include "xfs_sync.h"
 
 
-STATIC void
-xfs_quiesce_fs(
-	xfs_mount_t		*mp)
-{
-	int			count = 0, pincount;
-
-	xfs_flush_buftarg(mp->m_ddev_targp, 0);
-	xfs_finish_reclaim_all(mp, 0);
-
-	/* This loop must run at least twice.
-	 * The first instance of the loop will flush
-	 * most meta data but that will generate more
-	 * meta data (typically directory updates).
-	 * Which then must be flushed and logged before
-	 * we can write the unmount record.
-	 */
-	do {
-		xfs_syncsub(mp, SYNC_INODE_QUIESCE, NULL);
-		pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
-		if (!pincount) {
-			delay(50);
-			count++;
-		}
-	} while (count < 2);
-}
-
-/*
- * Second stage of a quiesce. The data is already synced, now we have to take
- * care of the metadata. New transactions are already blocked, so we need to
- * wait for any remaining transactions to drain out before proceding.
- */
-void
-xfs_attr_quiesce(
-	xfs_mount_t	*mp)
-{
-	int	error = 0;
-
-	/* wait for all modifications to complete */
-	while (atomic_read(&mp->m_active_trans) > 0)
-		delay(100);
-
-	/* flush inodes and push all remaining buffers out to disk */
-	xfs_quiesce_fs(mp);
-
-	ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);
-
-	/* Push the superblock and write an unmount record */
-	error = xfs_log_sbcount(mp, 1);
-	if (error)
-		xfs_fs_cmn_err(CE_WARN, mp,
-				"xfs_attr_quiesce: failed to log sb changes. "
-				"Frozen image may not be consistent.");
-	xfs_log_unmount_write(mp);
-	xfs_unmountfs_writesb(mp);
-}
-
 /*
  * xfs_unmount_flush implements a set of flush operation on special
  * inodes, which are needed as a separate set of operations so that
@@ -196,562 +140,3 @@ fscorrupt_out2:
 	return XFS_ERROR(EFSCORRUPTED);
 }
 
-/*
- * xfs_sync flushes any pending I/O to file system vfsp.
- *
- * This routine is called by vfs_sync() to make sure that things make it
- * out to disk eventually, on sync() system calls to flush out everything,
- * and when the file system is unmounted.  For the vfs_sync() case, all
- * we really need to do is sync out the log to make all of our meta-data
- * updates permanent (except for timestamps).  For calls from pflushd(),
- * dirty pages are kept moving by calling pdflush() on the inodes
- * containing them.  We also flush the inodes that we can lock without
- * sleeping and the superblock if we can lock it without sleeping from
- * vfs_sync() so that items at the tail of the log are always moving out.
- *
- * Flags:
- *      SYNC_BDFLUSH - We're being called from vfs_sync() so we don't want
- *		       to sleep if we can help it.  All we really need
- *		       to do is ensure that the log is synced at least
- *		       periodically.  We also push the inodes and
- *		       superblock if we can lock them without sleeping
- *			and they are not pinned.
- *      SYNC_ATTR    - We need to flush the inodes.  If SYNC_BDFLUSH is not
- *		       set, then we really want to lock each inode and flush
- *		       it.
- *      SYNC_WAIT    - All the flushes that take place in this call should
- *		       be synchronous.
- *      SYNC_DELWRI  - This tells us to push dirty pages associated with
- *		       inodes.  SYNC_WAIT and SYNC_BDFLUSH are used to
- *		       determine if they should be flushed sync, async, or
- *		       delwri.
- *      SYNC_CLOSE   - This flag is passed when the system is being
- *		       unmounted.  We should sync and invalidate everything.
- *      SYNC_FSDATA  - This indicates that the caller would like to make
- *		       sure the superblock is safe on disk.  We can ensure
- *		       this by simply making sure the log gets flushed
- *		       if SYNC_BDFLUSH is set, and by actually writing it
- *		       out otherwise.
- *	SYNC_IOWAIT  - The caller wants us to wait for all data I/O to complete
- *		       before we return (including direct I/O). Forms the drain
- *		       side of the write barrier needed to safely quiesce the
- *		       filesystem.
- *
- */
-int
-xfs_sync(
-	xfs_mount_t	*mp,
-	int		flags)
-{
-	int		error;
-
-	/*
-	 * Get the Quota Manager to flush the dquots.
-	 *
-	 * If XFS quota support is not enabled or this filesystem
-	 * instance does not use quotas XFS_QM_DQSYNC will always
-	 * return zero.
-	 */
-	error = XFS_QM_DQSYNC(mp, flags);
-	if (error) {
-		/*
-		 * If we got an IO error, we will be shutting down.
-		 * So, there's nothing more for us to do here.
-		 */
-		ASSERT(error != EIO || XFS_FORCED_SHUTDOWN(mp));
-		if (XFS_FORCED_SHUTDOWN(mp))
-			return XFS_ERROR(error);
-	}
-
-	if (flags & SYNC_IOWAIT)
-		xfs_filestream_flush(mp);
-
-	return xfs_syncsub(mp, flags, NULL);
-}
-
-/*
- * xfs sync routine for internal use
- *
- * This routine supports all of the flags defined for the generic vfs_sync
- * interface as explained above under xfs_sync.
- *
- */
-int
-xfs_sync_inodes(
-	xfs_mount_t	*mp,
-	int		flags,
-	int             *bypassed)
-{
-	xfs_inode_t	*ip = NULL;
-	struct inode	*vp = NULL;
-	int		error;
-	int		last_error;
-	uint64_t	fflag;
-	uint		lock_flags;
-	uint		base_lock_flags;
-	boolean_t	mount_locked;
-	boolean_t	vnode_refed;
-	int		preempt;
-	xfs_iptr_t	*ipointer;
-#ifdef DEBUG
-	boolean_t	ipointer_in = B_FALSE;
-
-#define IPOINTER_SET	ipointer_in = B_TRUE
-#define IPOINTER_CLR	ipointer_in = B_FALSE
-#else
-#define IPOINTER_SET
-#define IPOINTER_CLR
-#endif
-
-
-/* Insert a marker record into the inode list after inode ip. The list
- * must be locked when this is called. After the call the list will no
- * longer be locked.
- */
-#define IPOINTER_INSERT(ip, mp)	{ \
-		ASSERT(ipointer_in == B_FALSE); \
-		ipointer->ip_mnext = ip->i_mnext; \
-		ipointer->ip_mprev = ip; \
-		ip->i_mnext = (xfs_inode_t *)ipointer; \
-		ipointer->ip_mnext->i_mprev = (xfs_inode_t *)ipointer; \
-		preempt = 0; \
-		XFS_MOUNT_IUNLOCK(mp); \
-		mount_locked = B_FALSE; \
-		IPOINTER_SET; \
-	}
-
-/* Remove the marker from the inode list. If the marker was the only item
- * in the list then there are no remaining inodes and we should zero out
- * the whole list. If we are the current head of the list then move the head
- * past us.
- */
-#define IPOINTER_REMOVE(ip, mp)	{ \
-		ASSERT(ipointer_in == B_TRUE); \
-		if (ipointer->ip_mnext != (xfs_inode_t *)ipointer) { \
-			ip = ipointer->ip_mnext; \
-			ip->i_mprev = ipointer->ip_mprev; \
-			ipointer->ip_mprev->i_mnext = ip; \
-			if (mp->m_inodes == (xfs_inode_t *)ipointer) { \
-				mp->m_inodes = ip; \
-			} \
-		} else { \
-			ASSERT(mp->m_inodes == (xfs_inode_t *)ipointer); \
-			mp->m_inodes = NULL; \
-			ip = NULL; \
-		} \
-		IPOINTER_CLR; \
-	}
-
-#define XFS_PREEMPT_MASK	0x7f
-
-	ASSERT(!(flags & SYNC_BDFLUSH));
-
-	if (bypassed)
-		*bypassed = 0;
-	if (mp->m_flags & XFS_MOUNT_RDONLY)
-		return 0;
-	error = 0;
-	last_error = 0;
-	preempt = 0;
-
-	/* Allocate a reference marker */
-	ipointer = (xfs_iptr_t *)kmem_zalloc(sizeof(xfs_iptr_t), KM_SLEEP);
-
-	fflag = XFS_B_ASYNC;		/* default is don't wait */
-	if (flags & SYNC_DELWRI)
-		fflag = XFS_B_DELWRI;
-	if (flags & SYNC_WAIT)
-		fflag = 0;		/* synchronous overrides all */
-
-	base_lock_flags = XFS_ILOCK_SHARED;
-	if (flags & (SYNC_DELWRI | SYNC_CLOSE)) {
-		/*
-		 * We need the I/O lock if we're going to call any of
-		 * the flush/inval routines.
-		 */
-		base_lock_flags |= XFS_IOLOCK_SHARED;
-	}
-
-	XFS_MOUNT_ILOCK(mp);
-
-	ip = mp->m_inodes;
-
-	mount_locked = B_TRUE;
-	vnode_refed  = B_FALSE;
-
-	IPOINTER_CLR;
-
-	do {
-		ASSERT(ipointer_in == B_FALSE);
-		ASSERT(vnode_refed == B_FALSE);
-
-		lock_flags = base_lock_flags;
-
-		/*
-		 * There were no inodes in the list, just break out
-		 * of the loop.
-		 */
-		if (ip == NULL) {
-			break;
-		}
-
-		/*
-		 * We found another sync thread marker - skip it
-		 */
-		if (ip->i_mount == NULL) {
-			ip = ip->i_mnext;
-			continue;
-		}
-
-		vp = VFS_I(ip);
-
-		/*
-		 * If the vnode is gone then this is being torn down,
-		 * call reclaim if it is flushed, else let regular flush
-		 * code deal with it later in the loop.
-		 */
-
-		if (vp == NULL) {
-			/* Skip ones already in reclaim */
-			if (ip->i_flags & XFS_IRECLAIM) {
-				ip = ip->i_mnext;
-				continue;
-			}
-			if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0) {
-				ip = ip->i_mnext;
-			} else if ((xfs_ipincount(ip) == 0) &&
-				    xfs_iflock_nowait(ip)) {
-				IPOINTER_INSERT(ip, mp);
-
-				xfs_finish_reclaim(ip, 1,
-						XFS_IFLUSH_DELWRI_ELSE_ASYNC);
-
-				XFS_MOUNT_ILOCK(mp);
-				mount_locked = B_TRUE;
-				IPOINTER_REMOVE(ip, mp);
-			} else {
-				xfs_iunlock(ip, XFS_ILOCK_EXCL);
-				ip = ip->i_mnext;
-			}
-			continue;
-		}
-
-		if (VN_BAD(vp)) {
-			ip = ip->i_mnext;
-			continue;
-		}
-
-		if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
-			XFS_MOUNT_IUNLOCK(mp);
-			kmem_free(ipointer);
-			return 0;
-		}
-
-		/*
-		 * Try to lock without sleeping.  We're out of order with
-		 * the inode list lock here, so if we fail we need to drop
-		 * the mount lock and try again.  If we're called from
-		 * bdflush() here, then don't bother.
-		 *
-		 * The inode lock here actually coordinates with the
-		 * almost spurious inode lock in xfs_ireclaim() to prevent
-		 * the vnode we handle here without a reference from
-		 * being freed while we reference it.  If we lock the inode
-		 * while it's on the mount list here, then the spurious inode
-		 * lock in xfs_ireclaim() after the inode is pulled from
-		 * the mount list will sleep until we release it here.
-		 * This keeps the vnode from being freed while we reference
-		 * it.
-		 */
-		if (xfs_ilock_nowait(ip, lock_flags) == 0) {
-			if (vp == NULL) {
-				ip = ip->i_mnext;
-				continue;
-			}
-
-			vp = vn_grab(vp);
-			if (vp == NULL) {
-				ip = ip->i_mnext;
-				continue;
-			}
-
-			IPOINTER_INSERT(ip, mp);
-			xfs_ilock(ip, lock_flags);
-
-			ASSERT(vp == VFS_I(ip));
-			ASSERT(ip->i_mount == mp);
-
-			vnode_refed = B_TRUE;
-		}
-
-		/* From here on in the loop we may have a marker record
-		 * in the inode list.
-		 */
-
-		/*
-		 * If we have to flush data or wait for I/O completion
-		 * we need to drop the ilock that we currently hold.
-		 * If we need to drop the lock, insert a marker if we
-		 * have not already done so.
-		 */
-		if ((flags & (SYNC_CLOSE|SYNC_IOWAIT)) ||
-		    ((flags & SYNC_DELWRI) && VN_DIRTY(vp))) {
-			if (mount_locked) {
-				IPOINTER_INSERT(ip, mp);
-			}
-			xfs_iunlock(ip, XFS_ILOCK_SHARED);
-
-			if (flags & SYNC_CLOSE) {
-				/* Shutdown case. Flush and invalidate. */
-				if (XFS_FORCED_SHUTDOWN(mp))
-					xfs_tosspages(ip, 0, -1,
-							     FI_REMAPF);
-				else
-					error = xfs_flushinval_pages(ip,
-							0, -1, FI_REMAPF);
-			} else if ((flags & SYNC_DELWRI) && VN_DIRTY(vp)) {
-				error = xfs_flush_pages(ip, 0,
-							-1, fflag, FI_NONE);
-			}
-
-			/*
-			 * When freezing, we need to wait ensure all I/O (including direct
-			 * I/O) is complete to ensure no further data modification can take
-			 * place after this point
-			 */
-			if (flags & SYNC_IOWAIT)
-				vn_iowait(ip);
-
-			xfs_ilock(ip, XFS_ILOCK_SHARED);
-		}
-
-		if ((flags & SYNC_ATTR) &&
-		    (ip->i_update_core ||
-		     (ip->i_itemp && ip->i_itemp->ili_format.ilf_fields))) {
-			if (mount_locked)
-				IPOINTER_INSERT(ip, mp);
-
-			if (flags & SYNC_WAIT) {
-				xfs_iflock(ip);
-				error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
-
-			/*
-			 * If we can't acquire the flush lock, then the inode
-			 * is already being flushed so don't bother waiting.
-			 *
-			 * If we can lock it then do a delwri flush so we can
-			 * combine multiple inode flushes in each disk write.
-			 */
-			} else if (xfs_iflock_nowait(ip)) {
-				error = xfs_iflush(ip, XFS_IFLUSH_DELWRI);
-			} else if (bypassed) {
-				(*bypassed)++;
-			}
-		}
-
-		if (lock_flags != 0) {
-			xfs_iunlock(ip, lock_flags);
-		}
-
-		if (vnode_refed) {
-			/*
-			 * If we had to take a reference on the vnode
-			 * above, then wait until after we've unlocked
-			 * the inode to release the reference.  This is
-			 * because we can be already holding the inode
-			 * lock when IRELE() calls xfs_inactive().
-			 *
-			 * Make sure to drop the mount lock before calling
-			 * IRELE() so that we don't trip over ourselves if
-			 * we have to go for the mount lock again in the
-			 * inactive code.
-			 */
-			if (mount_locked) {
-				IPOINTER_INSERT(ip, mp);
-			}
-
-			IRELE(ip);
-
-			vnode_refed = B_FALSE;
-		}
-
-		if (error) {
-			last_error = error;
-		}
-
-		/*
-		 * bail out if the filesystem is corrupted.
-		 */
-		if (error == EFSCORRUPTED)  {
-			if (!mount_locked) {
-				XFS_MOUNT_ILOCK(mp);
-				IPOINTER_REMOVE(ip, mp);
-			}
-			XFS_MOUNT_IUNLOCK(mp);
-			ASSERT(ipointer_in == B_FALSE);
-			kmem_free(ipointer);
-			return XFS_ERROR(error);
-		}
-
-		/* Let other threads have a chance at the mount lock
-		 * if we have looped many times without dropping the
-		 * lock.
-		 */
-		if ((++preempt & XFS_PREEMPT_MASK) == 0) {
-			if (mount_locked) {
-				IPOINTER_INSERT(ip, mp);
-			}
-		}
-
-		if (mount_locked == B_FALSE) {
-			XFS_MOUNT_ILOCK(mp);
-			mount_locked = B_TRUE;
-			IPOINTER_REMOVE(ip, mp);
-			continue;
-		}
-
-		ASSERT(ipointer_in == B_FALSE);
-		ip = ip->i_mnext;
-
-	} while (ip != mp->m_inodes);
-
-	XFS_MOUNT_IUNLOCK(mp);
-
-	ASSERT(ipointer_in == B_FALSE);
-
-	kmem_free(ipointer);
-	return XFS_ERROR(last_error);
-}
-
-/*
- * xfs sync routine for internal use
- *
- * This routine supports all of the flags defined for the generic vfs_sync
- * interface as explained above under xfs_sync.
- *
- */
-int
-xfs_syncsub(
-	xfs_mount_t	*mp,
-	int		flags,
-	int             *bypassed)
-{
-	int		error = 0;
-	int		last_error = 0;
-	uint		log_flags = XFS_LOG_FORCE;
-	xfs_buf_t	*bp;
-	xfs_buf_log_item_t	*bip;
-
-	/*
-	 * Sync out the log.  This ensures that the log is periodically
-	 * flushed even if there is not enough activity to fill it up.
-	 */
-	if (flags & SYNC_WAIT)
-		log_flags |= XFS_LOG_SYNC;
-
-	xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
-
-	if (flags & (SYNC_ATTR|SYNC_DELWRI)) {
-		if (flags & SYNC_BDFLUSH)
-			xfs_finish_reclaim_all(mp, 1);
-		else
-			error = xfs_sync_inodes(mp, flags, bypassed);
-	}
-
-	/*
-	 * Flushing out dirty data above probably generated more
-	 * log activity, so if this isn't vfs_sync() then flush
-	 * the log again.
-	 */
-	if (flags & SYNC_DELWRI) {
-		xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
-	}
-
-	if (flags & SYNC_FSDATA) {
-		/*
-		 * If this is vfs_sync() then only sync the superblock
-		 * if we can lock it without sleeping and it is not pinned.
-		 */
-		if (flags & SYNC_BDFLUSH) {
-			bp = xfs_getsb(mp, XFS_BUF_TRYLOCK);
-			if (bp != NULL) {
-				bip = XFS_BUF_FSPRIVATE(bp,xfs_buf_log_item_t*);
-				if ((bip != NULL) &&
-				    xfs_buf_item_dirty(bip)) {
-					if (!(XFS_BUF_ISPINNED(bp))) {
-						XFS_BUF_ASYNC(bp);
-						error = xfs_bwrite(mp, bp);
-					} else {
-						xfs_buf_relse(bp);
-					}
-				} else {
-					xfs_buf_relse(bp);
-				}
-			}
-		} else {
-			bp = xfs_getsb(mp, 0);
-			/*
-			 * If the buffer is pinned then push on the log so
-			 * we won't get stuck waiting in the write for
-			 * someone, maybe ourselves, to flush the log.
-			 * Even though we just pushed the log above, we
-			 * did not have the superblock buffer locked at
-			 * that point so it can become pinned in between
-			 * there and here.
-			 */
-			if (XFS_BUF_ISPINNED(bp))
-				xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE);
-			if (flags & SYNC_WAIT)
-				XFS_BUF_UNASYNC(bp);
-			else
-				XFS_BUF_ASYNC(bp);
-			error = xfs_bwrite(mp, bp);
-		}
-		if (error) {
-			last_error = error;
-		}
-	}
-
-	/*
-	 * Now check to see if the log needs a "dummy" transaction.
-	 */
-	if (!(flags & SYNC_REMOUNT) && xfs_log_need_covered(mp)) {
-		xfs_trans_t *tp;
-		xfs_inode_t *ip;
-
-		/*
-		 * Put a dummy transaction in the log to tell
-		 * recovery that all others are OK.
-		 */
-		tp = xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
-		if ((error = xfs_trans_reserve(tp, 0,
-				XFS_ICHANGE_LOG_RES(mp),
-				0, 0, 0)))  {
-			xfs_trans_cancel(tp, 0);
-			return error;
-		}
-
-		ip = mp->m_rootip;
-		xfs_ilock(ip, XFS_ILOCK_EXCL);
-
-		xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
-		xfs_trans_ihold(tp, ip);
-		xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
-		error = xfs_trans_commit(tp, 0);
-		xfs_iunlock(ip, XFS_ILOCK_EXCL);
-		xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
-	}
-
-	/*
-	 * When shutting down, we need to insure that the AIL is pushed
-	 * to disk or the filesystem can appear corrupt from the PROM.
-	 */
-	if ((flags & (SYNC_CLOSE|SYNC_WAIT)) == (SYNC_CLOSE|SYNC_WAIT)) {
-		XFS_bflush(mp->m_ddev_targp);
-		if (mp->m_rtdev_targp) {
-			XFS_bflush(mp->m_rtdev_targp);
-		}
-	}
-
-	return XFS_ERROR(last_error);
-}

fs/xfs/xfs_vfsops.h

@@ -8,9 +8,7 @@ struct kstatfs;
 struct xfs_mount;
 struct xfs_mount_args;
 
-int xfs_sync(struct xfs_mount *mp, int flags);
 void xfs_do_force_shutdown(struct xfs_mount *mp, int flags, char *fname,
 		int lnnum);
-void xfs_attr_quiesce(struct xfs_mount *mp);
 
 #endif /* _XFS_VFSOPS_H */

fs/xfs/xfs_vnodeops.c

@@ -79,8 +79,7 @@ int
 xfs_setattr(
 	struct xfs_inode	*ip,
 	struct iattr		*iattr,
-	int			flags,
-	cred_t			*credp)
+	int			flags)
 {
 	xfs_mount_t		*mp = ip->i_mount;
 	struct inode		*inode = VFS_I(ip);
@@ -233,10 +232,6 @@ xfs_setattr(
 
 	/*
 	 * Change file ownership.  Must be the owner or privileged.
-	 * If the system was configured with the "restricted_chown"
-	 * option, the owner is not permitted to give away the file,
-	 * and can change the group id only to a group of which he
-	 * or she is a member.
 	 */
 	if (mask & (ATTR_UID|ATTR_GID)) {
 		/*
@@ -260,9 +255,8 @@ xfs_setattr(
 		 * shall be equal to either the group ID or one of the
 		 * supplementary group IDs of the calling process.
 		 */
-		if (restricted_chown &&
-		    (iuid != uid || (igid != gid &&
-				     !in_group_p((gid_t)gid))) &&
+		if ((iuid != uid ||
+		     (igid != gid && !in_group_p((gid_t)gid))) &&
 		    !capable(CAP_CHOWN)) {
 			code = XFS_ERROR(EPERM);
 			goto error_return;
@@ -456,10 +450,6 @@ xfs_setattr(
 
 	/*
 	 * Change file ownership.  Must be the owner or privileged.
-	 * If the system was configured with the "restricted_chown"
-	 * option, the owner is not permitted to give away the file,
-	 * and can change the group id only to a group of which he
-	 * or she is a member.
 	 */
 	if (mask & (ATTR_UID|ATTR_GID)) {
 		/*
@@ -2009,7 +1999,7 @@ xfs_remove(
 			goto out_bmap_cancel;
 
 		/*
-		 * Drop the link from dp to ip.
+		 * Drop the "." link from ip to self.
 		 */
 		error = xfs_droplink(tp, ip);
 		if (error)
@@ -2024,7 +2014,7 @@ xfs_remove(
 	}
 
 	/*
-	 * Drop the "." link from ip to self.
+	 * Drop the link from dp to ip.
 	 */
 	error = xfs_droplink(tp, ip);
 	if (error)
@@ -2833,122 +2823,10 @@ xfs_reclaim(
 	if (!ip->i_update_core && (ip->i_itemp == NULL)) {
 		xfs_ilock(ip, XFS_ILOCK_EXCL);
 		xfs_iflock(ip);
-		return xfs_finish_reclaim(ip, 1, XFS_IFLUSH_DELWRI_ELSE_SYNC);
-	} else {
-		xfs_mount_t	*mp = ip->i_mount;
-
-		/* Protect sync and unpin from us */
-		XFS_MOUNT_ILOCK(mp);
-		spin_lock(&ip->i_flags_lock);
-		__xfs_iflags_set(ip, XFS_IRECLAIMABLE);
-		VFS_I(ip)->i_private = NULL;
-		ip->i_vnode = NULL;
-		spin_unlock(&ip->i_flags_lock);
-		list_add_tail(&ip->i_reclaim, &mp->m_del_inodes);
-		XFS_MOUNT_IUNLOCK(mp);
+		xfs_iflags_set(ip, XFS_IRECLAIMABLE);
+		return xfs_reclaim_inode(ip, 1, XFS_IFLUSH_DELWRI_ELSE_SYNC);
 	}
-	return 0;
-}
-
-int
-xfs_finish_reclaim(
-	xfs_inode_t	*ip,
-	int		locked,
-	int		sync_mode)
-{
-	xfs_perag_t	*pag = xfs_get_perag(ip->i_mount, ip->i_ino);
-	struct inode	*vp = VFS_I(ip);
-
-	if (vp && VN_BAD(vp))
-		goto reclaim;
-
-	/* The hash lock here protects a thread in xfs_iget_core from
-	 * racing with us on linking the inode back with a vnode.
-	 * Once we have the XFS_IRECLAIM flag set it will not touch
-	 * us.
-	 */
-	write_lock(&pag->pag_ici_lock);
-	spin_lock(&ip->i_flags_lock);
-	if (__xfs_iflags_test(ip, XFS_IRECLAIM) ||
-	    (!__xfs_iflags_test(ip, XFS_IRECLAIMABLE) && vp == NULL)) {
-		spin_unlock(&ip->i_flags_lock);
-		write_unlock(&pag->pag_ici_lock);
-		if (locked) {
-			xfs_ifunlock(ip);
-			xfs_iunlock(ip, XFS_ILOCK_EXCL);
-		}
-		return 1;
-	}
-	__xfs_iflags_set(ip, XFS_IRECLAIM);
-	spin_unlock(&ip->i_flags_lock);
-	write_unlock(&pag->pag_ici_lock);
-	xfs_put_perag(ip->i_mount, pag);
-
-	/*
-	 * If the inode is still dirty, then flush it out.  If the inode
-	 * is not in the AIL, then it will be OK to flush it delwri as
-	 * long as xfs_iflush() does not keep any references to the inode.
-	 * We leave that decision up to xfs_iflush() since it has the
-	 * knowledge of whether it's OK to simply do a delwri flush of
-	 * the inode or whether we need to wait until the inode is
-	 * pulled from the AIL.
-	 * We get the flush lock regardless, though, just to make sure
-	 * we don't free it while it is being flushed.
-	 */
-	if (!locked) {
-		xfs_ilock(ip, XFS_ILOCK_EXCL);
-		xfs_iflock(ip);
-	}
-
-	/*
-	 * In the case of a forced shutdown we rely on xfs_iflush() to
-	 * wait for the inode to be unpinned before returning an error.
-	 */
-	if (xfs_iflush(ip, sync_mode) == 0) {
-		/* synchronize with xfs_iflush_done */
-		xfs_iflock(ip);
-		xfs_ifunlock(ip);
-	}
-
-	xfs_iunlock(ip, XFS_ILOCK_EXCL);
-
- reclaim:
-	xfs_ireclaim(ip);
-	return 0;
-}
-
-int
-xfs_finish_reclaim_all(xfs_mount_t *mp, int noblock)
-{
-	int		purged;
-	xfs_inode_t	*ip, *n;
-	int		done = 0;
-
-	while (!done) {
-		purged = 0;
-		XFS_MOUNT_ILOCK(mp);
-		list_for_each_entry_safe(ip, n, &mp->m_del_inodes, i_reclaim) {
-			if (noblock) {
-				if (xfs_ilock_nowait(ip, XFS_ILOCK_EXCL) == 0)
-					continue;
-				if (xfs_ipincount(ip) ||
-				    !xfs_iflock_nowait(ip)) {
-					xfs_iunlock(ip, XFS_ILOCK_EXCL);
-					continue;
-				}
-			}
-			XFS_MOUNT_IUNLOCK(mp);
-			if (xfs_finish_reclaim(ip, noblock,
-					XFS_IFLUSH_DELWRI_ELSE_ASYNC))
-				delay(1);
-			purged = 1;
-			break;
-		}
-
-		done = !purged;
-	}
-
-	XFS_MOUNT_IUNLOCK(mp);
+	xfs_inode_set_reclaim_tag(ip);
 	return 0;
 }
 
@@ -3474,7 +3352,6 @@ xfs_change_file_space(
 	int		cmd,
 	xfs_flock64_t	*bf,
 	xfs_off_t	offset,
-	cred_t		*credp,
 	int		attr_flags)
 {
 	xfs_mount_t	*mp = ip->i_mount;
@@ -3562,7 +3439,7 @@ xfs_change_file_space(
 		iattr.ia_valid = ATTR_SIZE;
 		iattr.ia_size = startoffset;
 
-		error = xfs_setattr(ip, &iattr, attr_flags, credp);
+		error = xfs_setattr(ip, &iattr, attr_flags);
 
 		if (error)
 			return error;

fs/xfs/xfs_vnodeops.h

@@ -15,8 +15,7 @@ struct xfs_iomap;
 
 
 int xfs_open(struct xfs_inode *ip);
-int xfs_setattr(struct xfs_inode *ip, struct iattr *vap, int flags,
-		struct cred *credp);
+int xfs_setattr(struct xfs_inode *ip, struct iattr *vap, int flags);
 #define	XFS_ATTR_DMI		0x01	/* invocation from a DMI function */
 #define	XFS_ATTR_NONBLOCK	0x02	/* return EAGAIN if operation would block */
 #define XFS_ATTR_NOLOCK		0x04	/* Don't grab any conflicting locks */
@@ -44,8 +43,7 @@ int xfs_inode_flush(struct xfs_inode *ip, int flags);
 int xfs_set_dmattrs(struct xfs_inode *ip, u_int evmask, u_int16_t state);
 int xfs_reclaim(struct xfs_inode *ip);
 int xfs_change_file_space(struct xfs_inode *ip, int cmd,
-		xfs_flock64_t *bf, xfs_off_t offset,
-		struct cred *credp, int	attr_flags);
+		xfs_flock64_t *bf, xfs_off_t offset, int attr_flags);
 int xfs_rename(struct xfs_inode *src_dp, struct xfs_name *src_name,
 		struct xfs_inode *src_ip, struct xfs_inode *target_dp,
 		struct xfs_name *target_name, struct xfs_inode *target_ip);

include/linux/fs.h

@@ -1874,7 +1874,9 @@ extern loff_t default_llseek(struct file *file, loff_t offset, int origin);
 
 extern loff_t vfs_llseek(struct file *file, loff_t offset, int origin);
 
+extern struct inode * inode_init_always(struct super_block *, struct inode *);
 extern void inode_init_once(struct inode *);
+extern void inode_add_to_lists(struct super_block *, struct inode *);
 extern void iput(struct inode *);
 extern struct inode * igrab(struct inode *);
 extern ino_t iunique(struct super_block *, ino_t);

kernel/sysctl_check.c

@@ -730,7 +730,6 @@ static const struct trans_ctl_table trans_fs_quota_table[] = {
 };
 
 static const struct trans_ctl_table trans_fs_xfs_table[] = {
-	{ XFS_RESTRICT_CHOWN,	"restrict_chown" },
 	{ XFS_SGID_INHERIT,	"irix_sgid_inherit" },
 	{ XFS_SYMLINK_MODE,	"irix_symlink_mode" },
 	{ XFS_PANIC_MASK,	"panic_mask" },