Merge git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable

	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.o \

	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \

	   ref-cache.o export.o tree-log.o acl.o free-space-cache.o zlib.o \

	   compression.o

	   compression.o delayed-ref.o

else

# Normal Makefile

	 */

	struct list_head delalloc_inodes;

	/*

	 * list for tracking inodes that must be sent to disk before a

	 * rename or truncate commit

	 */

	struct list_head ordered_operations;

	/* the space_info for where this inode's data allocations are done */

	struct btrfs_space_info *space_info;

	 */

	u64 logged_trans;

	/*

	 * trans that last made a change that should be fully fsync'd.  This

	 * gets reset to zero each time the inode is logged

	 */

	u64 log_dirty_trans;

	/* total number of bytes pending delalloc, used by stat to calc the

	 * real block usage of the file

	 */

	/* the start of block group preferred for allocations. */

	u64 block_group;

	/* the fsync log has some corner cases that mean we have to check

	 * directories to see if any unlinks have been done before

	 * the directory was logged.  See tree-log.c for all the

	 * details

	 */

	u64 last_unlink_trans;

	/*

	 * ordered_data_close is set by truncate when a file that used

	 * to have good data has been truncated to zero.  When it is set

	 * the btrfs file release call will add this inode to the

	 * ordered operations list so that we make sure to flush out any

	 * new data the application may have written before commit.

	 *

	 * yes, its silly to have a single bitflag, but we might grow more

	 * of these.

	 */

	unsigned ordered_data_close:1;

	struct inode vfs_inode;

};

#define BTRFS_MAX_LEVEL 8

/*

 * files bigger than this get some pre-flushing when they are added

 * to the ordered operations list.  That way we limit the total

 * work done by the commit

 */

#define BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT (8 * 1024 * 1024)

/* holds pointers to all of the tree roots */

#define BTRFS_ROOT_TREE_OBJECTID 1ULL

	int locks[BTRFS_MAX_LEVEL];

	int reada;

	/* keep some upper locks as we walk down */

	int keep_locks;

	int skip_locking;

	int lowest_level;

	/*

	 * set by btrfs_split_item, tells search_slot to keep all locks

	 * and to force calls to keep space in the nodes

	 */

	int search_for_split;

	unsigned int search_for_split:1;

	unsigned int keep_locks:1;

	unsigned int skip_locking:1;

	unsigned int leave_spinning:1;

};

/*

	struct rb_root block_group_cache_tree;

	struct extent_io_tree pinned_extents;

	struct extent_io_tree pending_del;

	struct extent_io_tree extent_ins;

	/* logical->physical extent mapping */

	struct btrfs_mapping_tree mapping_tree;

	u64 generation;

	u64 last_trans_committed;

	u64 last_trans_new_blockgroup;

	/*

	 * this is updated to the current trans every time a full commit

	 * is required instead of the faster short fsync log commits

	 */

	u64 last_trans_log_full_commit;

	u64 open_ioctl_trans;

	unsigned long mount_opt;

	u64 max_extent;

	struct mutex tree_log_mutex;

	struct mutex transaction_kthread_mutex;

	struct mutex cleaner_mutex;

	struct mutex extent_ins_mutex;

	struct mutex pinned_mutex;

	struct mutex chunk_mutex;

	struct mutex drop_mutex;

	struct mutex volume_mutex;

	struct mutex tree_reloc_mutex;

	/*

	 * this protects the ordered operations list only while we are

	 * processing all of the entries on it.  This way we make

	 * sure the commit code doesn't find the list temporarily empty

	 * because another function happens to be doing non-waiting preflush

	 * before jumping into the main commit.

	 */

	struct mutex ordered_operations_mutex;

	struct list_head trans_list;

	struct list_head hashers;

	struct list_head dead_roots;

	 * ordered extents

	 */

	spinlock_t ordered_extent_lock;

	/*

	 * all of the data=ordered extents pending writeback

	 * these can span multiple transactions and basically include

	 * every dirty data page that isn't from nodatacow

	 */

	struct list_head ordered_extents;

	/*

	 * all of the inodes that have delalloc bytes.  It is possible for

	 * this list to be empty even when there is still dirty data=ordered

	 * extents waiting to finish IO.

	 */

	struct list_head delalloc_inodes;

	/*

	 * special rename and truncate targets that must be on disk before

	 * we're allowed to commit.  This is basically the ext3 style

	 * data=ordered list.

	 */

	struct list_head ordered_operations;

	/*

	 * there is a pool of worker threads for checksumming during writes

	 * and a pool for checksumming after reads.  This is because readers

	atomic_t throttle_gen;

	u64 total_pinned;

	/* protected by the delalloc lock, used to keep from writing

	 * metadata until there is a nice batch

	 */

	u64 dirty_metadata_bytes;

	struct list_head dirty_cowonly_roots;

	struct btrfs_fs_devices *fs_devices;

}

/* extent-tree.c */

int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,

			   struct btrfs_root *root, unsigned long count);

int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len);

int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans,

			    struct btrfs_root *root, u64 bytenr,

			    u64 num_bytes, u32 *refs);

int btrfs_update_pinned_extents(struct btrfs_root *root,

				u64 bytenr, u64 num, int pin);

int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,

			struct btrfs_root *root, struct extent_buffer *leaf);

int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,

			  struct btrfs_root *root, u64 objectid, u64 bytenr);

int btrfs_extent_post_op(struct btrfs_trans_handle *trans,

			 struct btrfs_root *root);

int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy);

struct btrfs_block_group_cache *btrfs_lookup_block_group(

						 struct btrfs_fs_info *info,

			 u64 root_objectid, u64 ref_generation,

			 u64 owner_objectid);

int btrfs_update_extent_ref(struct btrfs_trans_handle *trans,

			    struct btrfs_root *root, u64 bytenr,

			    struct btrfs_root *root, u64 bytenr, u64 num_bytes,

			    u64 orig_parent, u64 parent,

			    u64 root_objectid, u64 ref_generation,

			    u64 owner_objectid);

int btrfs_cow_block(struct btrfs_trans_handle *trans,

		    struct btrfs_root *root, struct extent_buffer *buf,

		    struct extent_buffer *parent, int parent_slot,

		    struct extent_buffer **cow_ret, u64 prealloc_dest);

		    struct extent_buffer **cow_ret);

int btrfs_copy_root(struct btrfs_trans_handle *trans,

		      struct btrfs_root *root,

		      struct extent_buffer *buf,

/*

 * Copyright (C) 2009 Oracle.  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 v2 as published by the Free Software Foundation.

 *

 * This program is distributed in the hope that it will be useful,

 * but WITHOUT ANY WARRANTY; without even the implied warranty of

 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

 * General Public License for more details.

 *

 * You should have received a copy of the GNU General Public

 * License along with this program; if not, write to the

 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,

 * Boston, MA 021110-1307, USA.

 */

#include <linux/sched.h>

#include <linux/sort.h>

#include <linux/ftrace.h>

#include "ctree.h"

#include "delayed-ref.h"

#include "transaction.h"

/*

 * delayed back reference update tracking.  For subvolume trees

 * we queue up extent allocations and backref maintenance for

 * delayed processing.   This avoids deep call chains where we

 * add extents in the middle of btrfs_search_slot, and it allows

 * us to buffer up frequently modified backrefs in an rb tree instead

 * of hammering updates on the extent allocation tree.

 *

 * Right now this code is only used for reference counted trees, but

 * the long term goal is to get rid of the similar code for delayed

 * extent tree modifications.

 */

/*

 * entries in the rb tree are ordered by the byte number of the extent

 * and by the byte number of the parent block.

 */

static int comp_entry(struct btrfs_delayed_ref_node *ref,

		      u64 bytenr, u64 parent)

{

	if (bytenr < ref->bytenr)

		return -1;

	if (bytenr > ref->bytenr)

		return 1;

	if (parent < ref->parent)

		return -1;

	if (parent > ref->parent)

		return 1;

	return 0;

}

/*

 * insert a new ref into the rbtree.  This returns any existing refs

 * for the same (bytenr,parent) tuple, or NULL if the new node was properly

 * inserted.

 */

static struct btrfs_delayed_ref_node *tree_insert(struct rb_root *root,

						  u64 bytenr, u64 parent,

						  struct rb_node *node)

{

	struct rb_node **p = &root->rb_node;

	struct rb_node *parent_node = NULL;

	struct btrfs_delayed_ref_node *entry;

	int cmp;

	while (*p) {

		parent_node = *p;

		entry = rb_entry(parent_node, struct btrfs_delayed_ref_node,

				 rb_node);

		cmp = comp_entry(entry, bytenr, parent);

		if (cmp < 0)

			p = &(*p)->rb_left;

		else if (cmp > 0)

			p = &(*p)->rb_right;

		else

			return entry;

	}

	entry = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);

	rb_link_node(node, parent_node, p);

	rb_insert_color(node, root);

	return NULL;

}

/*

 * find an entry based on (bytenr,parent).  This returns the delayed

 * ref if it was able to find one, or NULL if nothing was in that spot

 */

static struct btrfs_delayed_ref_node *tree_search(struct rb_root *root,

				  u64 bytenr, u64 parent,

				  struct btrfs_delayed_ref_node **last)

{

	struct rb_node *n = root->rb_node;

	struct btrfs_delayed_ref_node *entry;

	int cmp;

	while (n) {

		entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);

		WARN_ON(!entry->in_tree);

		if (last)

			*last = entry;

		cmp = comp_entry(entry, bytenr, parent);

		if (cmp < 0)

			n = n->rb_left;

		else if (cmp > 0)

			n = n->rb_right;

		else

			return entry;

	}

	return NULL;

}

int btrfs_delayed_ref_lock(struct btrfs_trans_handle *trans,

			   struct btrfs_delayed_ref_head *head)

{

	struct btrfs_delayed_ref_root *delayed_refs;

	delayed_refs = &trans->transaction->delayed_refs;

	assert_spin_locked(&delayed_refs->lock);

	if (mutex_trylock(&head->mutex))

		return 0;

	atomic_inc(&head->node.refs);

	spin_unlock(&delayed_refs->lock);

	mutex_lock(&head->mutex);

	spin_lock(&delayed_refs->lock);

	if (!head->node.in_tree) {

		mutex_unlock(&head->mutex);

		btrfs_put_delayed_ref(&head->node);

		return -EAGAIN;

	}

	btrfs_put_delayed_ref(&head->node);

	return 0;

}

int btrfs_find_ref_cluster(struct btrfs_trans_handle *trans,

			   struct list_head *cluster, u64 start)

{

	int count = 0;

	struct btrfs_delayed_ref_root *delayed_refs;

	struct rb_node *node;

	struct btrfs_delayed_ref_node *ref;

	struct btrfs_delayed_ref_head *head;

	delayed_refs = &trans->transaction->delayed_refs;

	if (start == 0) {

		node = rb_first(&delayed_refs->root);

	} else {

		ref = NULL;

		tree_search(&delayed_refs->root, start, (u64)-1, &ref);

		if (ref) {

			struct btrfs_delayed_ref_node *tmp;

			node = rb_prev(&ref->rb_node);

			while (node) {

				tmp = rb_entry(node,

					       struct btrfs_delayed_ref_node,

					       rb_node);

				if (tmp->bytenr < start)

					break;

				ref = tmp;

				node = rb_prev(&ref->rb_node);

			}

			node = &ref->rb_node;

		} else

			node = rb_first(&delayed_refs->root);

	}

again:

	while (node && count < 32) {

		ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);

		if (btrfs_delayed_ref_is_head(ref)) {

			head = btrfs_delayed_node_to_head(ref);

			if (list_empty(&head->cluster)) {

				list_add_tail(&head->cluster, cluster);

				delayed_refs->run_delayed_start =

					head->node.bytenr;

				count++;

				WARN_ON(delayed_refs->num_heads_ready == 0);

				delayed_refs->num_heads_ready--;

			} else if (count) {

				/* the goal of the clustering is to find extents

				 * that are likely to end up in the same extent

				 * leaf on disk.  So, we don't want them spread

				 * all over the tree.  Stop now if we've hit

				 * a head that was already in use

				 */

				break;

			}

		}

		node = rb_next(node);

	}

	if (count) {

		return 0;

	} else if (start) {

		/*

		 * we've gone to the end of the rbtree without finding any

		 * clusters.  start from the beginning and try again

		 */

		start = 0;

		node = rb_first(&delayed_refs->root);

		goto again;

	}

	return 1;

}

/*

 * This checks to see if there are any delayed refs in the

 * btree for a given bytenr.  It returns one if it finds any

 * and zero otherwise.

 *

 * If it only finds a head node, it returns 0.

 *

 * The idea is to use this when deciding if you can safely delete an

 * extent from the extent allocation tree.  There may be a pending

 * ref in the rbtree that adds or removes references, so as long as this

 * returns one you need to leave the BTRFS_EXTENT_ITEM in the extent

 * allocation tree.

 */

int btrfs_delayed_ref_pending(struct btrfs_trans_handle *trans, u64 bytenr)

{

	struct btrfs_delayed_ref_node *ref;

	struct btrfs_delayed_ref_root *delayed_refs;

	struct rb_node *prev_node;

	int ret = 0;

	delayed_refs = &trans->transaction->delayed_refs;

	spin_lock(&delayed_refs->lock);

	ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL);

	if (ref) {

		prev_node = rb_prev(&ref->rb_node);

		if (!prev_node)

			goto out;

		ref = rb_entry(prev_node, struct btrfs_delayed_ref_node,

			       rb_node);

		if (ref->bytenr == bytenr)

			ret = 1;

	}

out:

	spin_unlock(&delayed_refs->lock);

	return ret;

}

/*

 * helper function to lookup reference count

 *

 * the head node for delayed ref is used to store the sum of all the

 * reference count modifications queued up in the rbtree.  This way you

 * can check to see what the reference count would be if all of the

 * delayed refs are processed.

 */

int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans,

			    struct btrfs_root *root, u64 bytenr,

			    u64 num_bytes, u32 *refs)

{

	struct btrfs_delayed_ref_node *ref;

	struct btrfs_delayed_ref_head *head;

	struct btrfs_delayed_ref_root *delayed_refs;

	struct btrfs_path *path;

	struct extent_buffer *leaf;

	struct btrfs_extent_item *ei;

	struct btrfs_key key;

	u32 num_refs;

	int ret;

	path = btrfs_alloc_path();

	if (!path)

		return -ENOMEM;

	key.objectid = bytenr;

	key.type = BTRFS_EXTENT_ITEM_KEY;

	key.offset = num_bytes;

	delayed_refs = &trans->transaction->delayed_refs;

again:

	ret = btrfs_search_slot(trans, root->fs_info->extent_root,

				&key, path, 0, 0);

	if (ret < 0)

		goto out;

	if (ret == 0) {

		leaf = path->nodes[0];

		ei = btrfs_item_ptr(leaf, path->slots[0],

				    struct btrfs_extent_item);

		num_refs = btrfs_extent_refs(leaf, ei);

	} else {

		num_refs = 0;

		ret = 0;

	}

	spin_lock(&delayed_refs->lock);

	ref = tree_search(&delayed_refs->root, bytenr, (u64)-1, NULL);

	if (ref) {

		head = btrfs_delayed_node_to_head(ref);

		if (mutex_trylock(&head->mutex)) {

			num_refs += ref->ref_mod;

			mutex_unlock(&head->mutex);

			*refs = num_refs;

			goto out;

		}

		atomic_inc(&ref->refs);

		spin_unlock(&delayed_refs->lock);

		btrfs_release_path(root->fs_info->extent_root, path);

		mutex_lock(&head->mutex);

		mutex_unlock(&head->mutex);

		btrfs_put_delayed_ref(ref);

		goto again;

	} else {

		*refs = num_refs;

	}

out:

	spin_unlock(&delayed_refs->lock);

	btrfs_free_path(path);

	return ret;

}

/*

 * helper function to update an extent delayed ref in the

 * rbtree.  existing and update must both have the same

 * bytenr and parent

 *

 * This may free existing if the update cancels out whatever

 * operation it was doing.

 */

static noinline void

update_existing_ref(struct btrfs_trans_handle *trans,

		    struct btrfs_delayed_ref_root *delayed_refs,

		    struct btrfs_delayed_ref_node *existing,

		    struct btrfs_delayed_ref_node *update)

{

	struct btrfs_delayed_ref *existing_ref;

	struct btrfs_delayed_ref *ref;

	existing_ref = btrfs_delayed_node_to_ref(existing);

	ref = btrfs_delayed_node_to_ref(update);

	if (ref->pin)

		existing_ref->pin = 1;

	if (ref->action != existing_ref->action) {

		/*

		 * this is effectively undoing either an add or a

		 * drop.  We decrement the ref_mod, and if it goes

		 * down to zero we just delete the entry without

		 * every changing the extent allocation tree.

		 */

		existing->ref_mod--;

		if (existing->ref_mod == 0) {

			rb_erase(&existing->rb_node,

				 &delayed_refs->root);

			existing->in_tree = 0;

			btrfs_put_delayed_ref(existing);

			delayed_refs->num_entries--;

			if (trans->delayed_ref_updates)

				trans->delayed_ref_updates--;

		}

	} else {

		if (existing_ref->action == BTRFS_ADD_DELAYED_REF) {

			/* if we're adding refs, make sure all the

			 * details match up.  The extent could

			 * have been totally freed and reallocated

			 * by a different owner before the delayed

			 * ref entries were removed.

			 */

			existing_ref->owner_objectid = ref->owner_objectid;

			existing_ref->generation = ref->generation;

			existing_ref->root = ref->root;

			existing->num_bytes = update->num_bytes;

		}

		/*

		 * the action on the existing ref matches

		 * the action on the ref we're trying to add.

		 * Bump the ref_mod by one so the backref that

		 * is eventually added/removed has the correct

		 * reference count

		 */

		existing->ref_mod += update->ref_mod;

	}

}

/*

 * helper function to update the accounting in the head ref

 * existing and update must have the same bytenr

 */

static noinline void

update_existing_head_ref(struct btrfs_delayed_ref_node *existing,

			 struct btrfs_delayed_ref_node *update)

{

	struct btrfs_delayed_ref_head *existing_ref;

	struct btrfs_delayed_ref_head *ref;

	existing_ref = btrfs_delayed_node_to_head(existing);

	ref = btrfs_delayed_node_to_head(update);

	if (ref->must_insert_reserved) {

		/* if the extent was freed and then

		 * reallocated before the delayed ref

		 * entries were processed, we can end up

		 * with an existing head ref without

		 * the must_insert_reserved flag set.

		 * Set it again here

		 */

		existing_ref->must_insert_reserved = ref->must_insert_reserved;

		/*

		 * update the num_bytes so we make sure the accounting

		 * is done correctly

		 */

		existing->num_bytes = update->num_bytes;

	}

	/*

	 * update the reference mod on the head to reflect this new operation

	 */

	existing->ref_mod += update->ref_mod;

}

/*

 * helper function to actually insert a delayed ref into the rbtree.

 * this does all the dirty work in terms of maintaining the correct

 * overall modification count in the head node and properly dealing

 * with updating existing nodes as new modifications are queued.

 */

static noinline int __btrfs_add_delayed_ref(struct btrfs_trans_handle *trans,

			  struct btrfs_delayed_ref_node *ref,

			  u64 bytenr, u64 num_bytes, u64 parent, u64 ref_root,

			  u64 ref_generation, u64 owner_objectid, int action,

			  int pin)

{

	struct btrfs_delayed_ref_node *existing;

	struct btrfs_delayed_ref *full_ref;

	struct btrfs_delayed_ref_head *head_ref = NULL;

	struct btrfs_delayed_ref_root *delayed_refs;

	int count_mod = 1;

	int must_insert_reserved = 0;

	/*

	 * the head node stores the sum of all the mods, so dropping a ref

	 * should drop the sum in the head node by one.

	 */

	if (parent == (u64)-1) {

		if (action == BTRFS_DROP_DELAYED_REF)

			count_mod = -1;

		else if (action == BTRFS_UPDATE_DELAYED_HEAD)

			count_mod = 0;

	}

	/*

	 * BTRFS_ADD_DELAYED_EXTENT means that we need to update

	 * the reserved accounting when the extent is finally added, or

	 * if a later modification deletes the delayed ref without ever

	 * inserting the extent into the extent allocation tree.

	 * ref->must_insert_reserved is the flag used to record

	 * that accounting mods are required.

	 *

	 * Once we record must_insert_reserved, switch the action to

	 * BTRFS_ADD_DELAYED_REF because other special casing is not required.

	 */

	if (action == BTRFS_ADD_DELAYED_EXTENT) {

		must_insert_reserved = 1;

		action = BTRFS_ADD_DELAYED_REF;

	} else {

		must_insert_reserved = 0;

	}

	delayed_refs = &trans->transaction->delayed_refs;

	/* first set the basic ref node struct up */

	atomic_set(&ref->refs, 1);

	ref->bytenr = bytenr;

	ref->parent = parent;

	ref->ref_mod = count_mod;

	ref->in_tree = 1;

	ref->num_bytes = num_bytes;

	if (btrfs_delayed_ref_is_head(ref)) {

		head_ref = btrfs_delayed_node_to_head(ref);

		head_ref->must_insert_reserved = must_insert_reserved;

		INIT_LIST_HEAD(&head_ref->cluster);

		mutex_init(&head_ref->mutex);

	} else {

		full_ref = btrfs_delayed_node_to_ref(ref);

		full_ref->root = ref_root;

		full_ref->generation = ref_generation;

		full_ref->owner_objectid = owner_objectid;

		full_ref->pin = pin;

		full_ref->action = action;

	}

	existing = tree_insert(&delayed_refs->root, bytenr,