// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (C) 2007,2008 Oracle.  All rights reserved.
 */

#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/rbtree.h>
#include <linux/mm.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "print-tree.h"
#include "locking.h"
#include "volumes.h"

static int split_node(struct btrfs_trans_handle *trans, struct btrfs_root
                      *root, struct btrfs_path *path, int level);
static int split_leaf(struct btrfs_trans_handle *trans, struct btrfs_root *root,
                      const struct btrfs_key *ins_key, struct btrfs_path *path,
                      int data_size, int extend);
static int push_node_left(struct btrfs_trans_handle *trans,
                          struct btrfs_fs_info *fs_info,
                          struct extent_buffer *dst,
                          struct extent_buffer *src, int empty);
static int balance_node_right(struct btrfs_trans_handle *trans,
                              struct btrfs_fs_info *fs_info,
                              struct extent_buffer *dst_buf,
                              struct extent_buffer *src_buf);
static void del_ptr(struct btrfs_root *root, struct btrfs_path *path,
                    int level, int slot);

struct btrfs_path *btrfs_alloc_path(void)
{
        return kmem_cache_zalloc(btrfs_path_cachep, GFP_NOFS);
}

/*
 * set all locked nodes in the path to blocking locks.  This should
 * be done before scheduling
 */
noinline void btrfs_set_path_blocking(struct btrfs_path *p)
{
        int i;
        for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
                if (!p->nodes[i] || !p->locks[i])
                        continue;
                btrfs_set_lock_blocking_rw(p->nodes[i], p->locks[i]);
                if (p->locks[i] == BTRFS_READ_LOCK)
                        p->locks[i] = BTRFS_READ_LOCK_BLOCKING;
                else if (p->locks[i] == BTRFS_WRITE_LOCK)
                        p->locks[i] = BTRFS_WRITE_LOCK_BLOCKING;
        }
}

/* this also releases the path */
void btrfs_free_path(struct btrfs_path *p)
{
        if (!p)
                return;
        btrfs_release_path(p);
        kmem_cache_free(btrfs_path_cachep, p);
}

/*
 * path release drops references on the extent buffers in the path
 * and it drops any locks held by this path
 *
 * It is safe to call this on paths that no locks or extent buffers held.
 */
noinline void btrfs_release_path(struct btrfs_path *p)
{
        int i;

        for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
                p->slots[i] = 0;
                if (!p->nodes[i])
                        continue;
                if (p->locks[i]) {
                        btrfs_tree_unlock_rw(p->nodes[i], p->locks[i]);
                        p->locks[i] = 0;
                }
                free_extent_buffer(p->nodes[i]);
                p->nodes[i] = NULL;
        }
}

/*
 * safely gets a reference on the root node of a tree.  A lock
 * is not taken, so a concurrent writer may put a different node
 * at the root of the tree.  See btrfs_lock_root_node for the
 * looping required.
 *
 * The extent buffer returned by this has a reference taken, so
 * it won't disappear.  It may stop being the root of the tree
 * at any time because there are no locks held.
 */
struct extent_buffer *btrfs_root_node(struct btrfs_root *root)
{
        struct extent_buffer *eb;

        while (1) {
                rcu_read_lock();
                eb = rcu_dereference(root->node);

                /*
                 * RCU really hurts here, we could free up the root node because
                 * it was COWed but we may not get the new root node yet so do
                 * the inc_not_zero dance and if it doesn't work then
                 * synchronize_rcu and try again.
                 */
                if (atomic_inc_not_zero(&eb->refs)) {
                        rcu_read_unlock();
                        break;
                }
                rcu_read_unlock();
                synchronize_rcu();
        }
        return eb;
}

/* loop around taking references on and locking the root node of the
 * tree until you end up with a lock on the root.  A locked buffer
 * is returned, with a reference held.
 */
struct extent_buffer *btrfs_lock_root_node(struct btrfs_root *root)
{
        struct extent_buffer *eb;

        while (1) {
                eb = btrfs_root_node(root);
                btrfs_tree_lock(eb);
                if (eb == root->node)
                        break;
                btrfs_tree_unlock(eb);
                free_extent_buffer(eb);
        }
        return eb;
}

/* loop around taking references on and locking the root node of the
 * tree until you end up with a lock on the root.  A locked buffer
 * is returned, with a reference held.
 */
struct extent_buffer *btrfs_read_lock_root_node(struct btrfs_root *root)
{
        struct extent_buffer *eb;

        while (1) {
                eb = btrfs_root_node(root);
                btrfs_tree_read_lock(eb);
                if (eb == root->node)
                        break;
                btrfs_tree_read_unlock(eb);
                free_extent_buffer(eb);
        }
        return eb;
}

/* cowonly root (everything not a reference counted cow subvolume), just get
 * put onto a simple dirty list.  transaction.c walks this to make sure they
 * get properly updated on disk.
 */
static void add_root_to_dirty_list(struct btrfs_root *root)
{
        struct btrfs_fs_info *fs_info = root->fs_info;

        if (test_bit(BTRFS_ROOT_DIRTY, &root->state) ||
            !test_bit(BTRFS_ROOT_TRACK_DIRTY, &root->state))
                return;

        spin_lock(&fs_info->trans_lock);
        if (!test_and_set_bit(BTRFS_ROOT_DIRTY, &root->state)) {
                /* Want the extent tree to be the last on the list */
                if (root->root_key.objectid == BTRFS_EXTENT_TREE_OBJECTID)
                        list_move_tail(&root->dirty_list,
                                       &fs_info->dirty_cowonly_roots);
                else
                        list_move(&root->dirty_list,
                                  &fs_info->dirty_cowonly_roots);
        }
        spin_unlock(&fs_info->trans_lock);
}

/*
 * used by snapshot creation to make a copy of a root for a tree with
 * a given objectid.  The buffer with the new root node is returned in
 * cow_ret, and this func returns zero on success or a negative error code.
 */
int btrfs_copy_root(struct btrfs_trans_handle *trans,
                      struct btrfs_root *root,
                      struct extent_buffer *buf,
                      struct extent_buffer **cow_ret, u64 new_root_objectid)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *cow;
        int ret = 0;
        int level;
        struct btrfs_disk_key disk_key;

        WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
                trans->transid != fs_info->running_transaction->transid);
        WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
                trans->transid != root->last_trans);

        level = btrfs_header_level(buf);
        if (level == 0)
                btrfs_item_key(buf, &disk_key, 0);
        else
                btrfs_node_key(buf, &disk_key, 0);

        cow = btrfs_alloc_tree_block(trans, root, 0, new_root_objectid,
                        &disk_key, level, buf->start, 0);
        if (IS_ERR(cow))
                return PTR_ERR(cow);

        copy_extent_buffer_full(cow, buf);
        btrfs_set_header_bytenr(cow, cow->start);
        btrfs_set_header_generation(cow, trans->transid);
        btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV);
        btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN |
                                     BTRFS_HEADER_FLAG_RELOC);
        if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID)
                btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC);
        else
                btrfs_set_header_owner(cow, new_root_objectid);

        write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid);

        WARN_ON(btrfs_header_generation(buf) > trans->transid);
        if (new_root_objectid == BTRFS_TREE_RELOC_OBJECTID)
                ret = btrfs_inc_ref(trans, root, cow, 1);
        else
                ret = btrfs_inc_ref(trans, root, cow, 0);

        if (ret)
                return ret;

        btrfs_mark_buffer_dirty(cow);
        *cow_ret = cow;
        return 0;
}

enum mod_log_op {
        MOD_LOG_KEY_REPLACE,
        MOD_LOG_KEY_ADD,
        MOD_LOG_KEY_REMOVE,
        MOD_LOG_KEY_REMOVE_WHILE_FREEING,
        MOD_LOG_KEY_REMOVE_WHILE_MOVING,
        MOD_LOG_MOVE_KEYS,
        MOD_LOG_ROOT_REPLACE,
};

struct tree_mod_root {
        u64 logical;
        u8 level;
};

struct tree_mod_elem {
        struct rb_node node;
        u64 logical;
        u64 seq;
        enum mod_log_op op;

        /* this is used for MOD_LOG_KEY_* and MOD_LOG_MOVE_KEYS operations */
        int slot;

        /* this is used for MOD_LOG_KEY* and MOD_LOG_ROOT_REPLACE */
        u64 generation;

        /* those are used for op == MOD_LOG_KEY_{REPLACE,REMOVE} */
        struct btrfs_disk_key key;
        u64 blockptr;

        /* this is used for op == MOD_LOG_MOVE_KEYS */
        struct {
                int dst_slot;
                int nr_items;
        } move;

        /* this is used for op == MOD_LOG_ROOT_REPLACE */
        struct tree_mod_root old_root;
};

/*
 * Pull a new tree mod seq number for our operation.
 */
static inline u64 btrfs_inc_tree_mod_seq(struct btrfs_fs_info *fs_info)
{
        return atomic64_inc_return(&fs_info->tree_mod_seq);
}

/*
 * This adds a new blocker to the tree mod log's blocker list if the @elem
 * passed does not already have a sequence number set. So when a caller expects
 * to record tree modifications, it should ensure to set elem->seq to zero
 * before calling btrfs_get_tree_mod_seq.
 * Returns a fresh, unused tree log modification sequence number, even if no new
 * blocker was added.
 */
u64 btrfs_get_tree_mod_seq(struct btrfs_fs_info *fs_info,
                           struct seq_list *elem)
{
        write_lock(&fs_info->tree_mod_log_lock);
        spin_lock(&fs_info->tree_mod_seq_lock);
        if (!elem->seq) {
                elem->seq = btrfs_inc_tree_mod_seq(fs_info);
                list_add_tail(&elem->list, &fs_info->tree_mod_seq_list);
        }
        spin_unlock(&fs_info->tree_mod_seq_lock);
        write_unlock(&fs_info->tree_mod_log_lock);

        return elem->seq;
}

void btrfs_put_tree_mod_seq(struct btrfs_fs_info *fs_info,
                            struct seq_list *elem)
{
        struct rb_root *tm_root;
        struct rb_node *node;
        struct rb_node *next;
        struct seq_list *cur_elem;
        struct tree_mod_elem *tm;
        u64 min_seq = (u64)-1;
        u64 seq_putting = elem->seq;

        if (!seq_putting)
                return;

        spin_lock(&fs_info->tree_mod_seq_lock);
        list_del(&elem->list);
        elem->seq = 0;

        list_for_each_entry(cur_elem, &fs_info->tree_mod_seq_list, list) {
                if (cur_elem->seq < min_seq) {
                        if (seq_putting > cur_elem->seq) {
                                /*
                                 * blocker with lower sequence number exists, we
                                 * cannot remove anything from the log
                                 */
                                spin_unlock(&fs_info->tree_mod_seq_lock);
                                return;
                        }
                        min_seq = cur_elem->seq;
                }
        }
        spin_unlock(&fs_info->tree_mod_seq_lock);

        /*
         * anything that's lower than the lowest existing (read: blocked)
         * sequence number can be removed from the tree.
         */
        write_lock(&fs_info->tree_mod_log_lock);
        tm_root = &fs_info->tree_mod_log;
        for (node = rb_first(tm_root); node; node = next) {
                next = rb_next(node);
                tm = rb_entry(node, struct tree_mod_elem, node);
                if (tm->seq > min_seq)
                        continue;
                rb_erase(node, tm_root);
                kfree(tm);
        }
        write_unlock(&fs_info->tree_mod_log_lock);
}

/*
 * key order of the log:
 *       node/leaf start address -> sequence
 *
 * The 'start address' is the logical address of the *new* root node
 * for root replace operations, or the logical address of the affected
 * block for all other operations.
 *
 * Note: must be called with write lock for fs_info::tree_mod_log_lock.
 */
static noinline int
__tree_mod_log_insert(struct btrfs_fs_info *fs_info, struct tree_mod_elem *tm)
{
        struct rb_root *tm_root;
        struct rb_node **new;
        struct rb_node *parent = NULL;
        struct tree_mod_elem *cur;

        tm->seq = btrfs_inc_tree_mod_seq(fs_info);

        tm_root = &fs_info->tree_mod_log;
        new = &tm_root->rb_node;
        while (*new) {
                cur = rb_entry(*new, struct tree_mod_elem, node);
                parent = *new;
                if (cur->logical < tm->logical)
                        new = &((*new)->rb_left);
                else if (cur->logical > tm->logical)
                        new = &((*new)->rb_right);
                else if (cur->seq < tm->seq)
                        new = &((*new)->rb_left);
                else if (cur->seq > tm->seq)
                        new = &((*new)->rb_right);
                else
                        return -EEXIST;
        }

        rb_link_node(&tm->node, parent, new);
        rb_insert_color(&tm->node, tm_root);
        return 0;
}

/*
 * Determines if logging can be omitted. Returns 1 if it can. Otherwise, it
 * returns zero with the tree_mod_log_lock acquired. The caller must hold
 * this until all tree mod log insertions are recorded in the rb tree and then
 * write unlock fs_info::tree_mod_log_lock.
 */
static inline int tree_mod_dont_log(struct btrfs_fs_info *fs_info,
                                    struct extent_buffer *eb) {
        smp_mb();
        if (list_empty(&(fs_info)->tree_mod_seq_list))
                return 1;
        if (eb && btrfs_header_level(eb) == 0)
                return 1;

        write_lock(&fs_info->tree_mod_log_lock);
        if (list_empty(&(fs_info)->tree_mod_seq_list)) {
                write_unlock(&fs_info->tree_mod_log_lock);
                return 1;
        }

        return 0;
}

/* Similar to tree_mod_dont_log, but doesn't acquire any locks. */
static inline int tree_mod_need_log(const struct btrfs_fs_info *fs_info,
                                    struct extent_buffer *eb)
{
        smp_mb();
        if (list_empty(&(fs_info)->tree_mod_seq_list))
                return 0;
        if (eb && btrfs_header_level(eb) == 0)
                return 0;

        return 1;
}

static struct tree_mod_elem *
alloc_tree_mod_elem(struct extent_buffer *eb, int slot,
                    enum mod_log_op op, gfp_t flags)
{
        struct tree_mod_elem *tm;

        tm = kzalloc(sizeof(*tm), flags);
        if (!tm)
                return NULL;

        tm->logical = eb->start;
        if (op != MOD_LOG_KEY_ADD) {
                btrfs_node_key(eb, &tm->key, slot);
                tm->blockptr = btrfs_node_blockptr(eb, slot);
        }
        tm->op = op;
        tm->slot = slot;
        tm->generation = btrfs_node_ptr_generation(eb, slot);
        RB_CLEAR_NODE(&tm->node);

        return tm;
}

static noinline int tree_mod_log_insert_key(struct extent_buffer *eb, int slot,
                enum mod_log_op op, gfp_t flags)
{
        struct tree_mod_elem *tm;
        int ret;

        if (!tree_mod_need_log(eb->fs_info, eb))
                return 0;

        tm = alloc_tree_mod_elem(eb, slot, op, flags);
        if (!tm)
                return -ENOMEM;

        if (tree_mod_dont_log(eb->fs_info, eb)) {
                kfree(tm);
                return 0;
        }

        ret = __tree_mod_log_insert(eb->fs_info, tm);
        write_unlock(&eb->fs_info->tree_mod_log_lock);
        if (ret)
                kfree(tm);

        return ret;
}

static noinline int tree_mod_log_insert_move(struct extent_buffer *eb,
                int dst_slot, int src_slot, int nr_items)
{
        struct tree_mod_elem *tm = NULL;
        struct tree_mod_elem **tm_list = NULL;
        int ret = 0;
        int i;
        int locked = 0;

        if (!tree_mod_need_log(eb->fs_info, eb))
                return 0;

        tm_list = kcalloc(nr_items, sizeof(struct tree_mod_elem *), GFP_NOFS);
        if (!tm_list)
                return -ENOMEM;

        tm = kzalloc(sizeof(*tm), GFP_NOFS);
        if (!tm) {
                ret = -ENOMEM;
                goto free_tms;
        }

        tm->logical = eb->start;
        tm->slot = src_slot;
        tm->move.dst_slot = dst_slot;
        tm->move.nr_items = nr_items;
        tm->op = MOD_LOG_MOVE_KEYS;

        for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
                tm_list[i] = alloc_tree_mod_elem(eb, i + dst_slot,
                    MOD_LOG_KEY_REMOVE_WHILE_MOVING, GFP_NOFS);
                if (!tm_list[i]) {
                        ret = -ENOMEM;
                        goto free_tms;
                }
        }

        if (tree_mod_dont_log(eb->fs_info, eb))
                goto free_tms;
        locked = 1;

        /*
         * When we override something during the move, we log these removals.
         * This can only happen when we move towards the beginning of the
         * buffer, i.e. dst_slot < src_slot.
         */
        for (i = 0; i + dst_slot < src_slot && i < nr_items; i++) {
                ret = __tree_mod_log_insert(eb->fs_info, tm_list[i]);
                if (ret)
                        goto free_tms;
        }

        ret = __tree_mod_log_insert(eb->fs_info, tm);
        if (ret)
                goto free_tms;
        write_unlock(&eb->fs_info->tree_mod_log_lock);
        kfree(tm_list);

        return 0;
free_tms:
        for (i = 0; i < nr_items; i++) {
                if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
                        rb_erase(&tm_list[i]->node, &eb->fs_info->tree_mod_log);
                kfree(tm_list[i]);
        }
        if (locked)
                write_unlock(&eb->fs_info->tree_mod_log_lock);
        kfree(tm_list);
        kfree(tm);

        return ret;
}

static inline int
__tree_mod_log_free_eb(struct btrfs_fs_info *fs_info,
                       struct tree_mod_elem **tm_list,
                       int nritems)
{
        int i, j;
        int ret;

        for (i = nritems - 1; i >= 0; i--) {
                ret = __tree_mod_log_insert(fs_info, tm_list[i]);
                if (ret) {
                        for (j = nritems - 1; j > i; j--)
                                rb_erase(&tm_list[j]->node,
                                         &fs_info->tree_mod_log);
                        return ret;
                }
        }

        return 0;
}

static noinline int tree_mod_log_insert_root(struct extent_buffer *old_root,
                         struct extent_buffer *new_root, int log_removal)
{
        struct btrfs_fs_info *fs_info = old_root->fs_info;
        struct tree_mod_elem *tm = NULL;
        struct tree_mod_elem **tm_list = NULL;
        int nritems = 0;
        int ret = 0;
        int i;

        if (!tree_mod_need_log(fs_info, NULL))
                return 0;

        if (log_removal && btrfs_header_level(old_root) > 0) {
                nritems = btrfs_header_nritems(old_root);
                tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *),
                                  GFP_NOFS);
                if (!tm_list) {
                        ret = -ENOMEM;
                        goto free_tms;
                }
                for (i = 0; i < nritems; i++) {
                        tm_list[i] = alloc_tree_mod_elem(old_root, i,
                            MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS);
                        if (!tm_list[i]) {
                                ret = -ENOMEM;
                                goto free_tms;
                        }
                }
        }

        tm = kzalloc(sizeof(*tm), GFP_NOFS);
        if (!tm) {
                ret = -ENOMEM;
                goto free_tms;
        }

        tm->logical = new_root->start;
        tm->old_root.logical = old_root->start;
        tm->old_root.level = btrfs_header_level(old_root);
        tm->generation = btrfs_header_generation(old_root);
        tm->op = MOD_LOG_ROOT_REPLACE;

        if (tree_mod_dont_log(fs_info, NULL))
                goto free_tms;

        if (tm_list)
                ret = __tree_mod_log_free_eb(fs_info, tm_list, nritems);
        if (!ret)
                ret = __tree_mod_log_insert(fs_info, tm);

        write_unlock(&fs_info->tree_mod_log_lock);
        if (ret)
                goto free_tms;
        kfree(tm_list);

        return ret;

free_tms:
        if (tm_list) {
                for (i = 0; i < nritems; i++)
                        kfree(tm_list[i]);
                kfree(tm_list);
        }
        kfree(tm);

        return ret;
}

static struct tree_mod_elem *
__tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq,
                      int smallest)
{
        struct rb_root *tm_root;
        struct rb_node *node;
        struct tree_mod_elem *cur = NULL;
        struct tree_mod_elem *found = NULL;

        read_lock(&fs_info->tree_mod_log_lock);
        tm_root = &fs_info->tree_mod_log;
        node = tm_root->rb_node;
        while (node) {
                cur = rb_entry(node, struct tree_mod_elem, node);
                if (cur->logical < start) {
                        node = node->rb_left;
                } else if (cur->logical > start) {
                        node = node->rb_right;
                } else if (cur->seq < min_seq) {
                        node = node->rb_left;
                } else if (!smallest) {
                        /* we want the node with the highest seq */
                        if (found)
                                BUG_ON(found->seq > cur->seq);
                        found = cur;
                        node = node->rb_left;
                } else if (cur->seq > min_seq) {
                        /* we want the node with the smallest seq */
                        if (found)
                                BUG_ON(found->seq < cur->seq);
                        found = cur;
                        node = node->rb_right;
                } else {
                        found = cur;
                        break;
                }
        }
        read_unlock(&fs_info->tree_mod_log_lock);

        return found;
}

/*
 * this returns the element from the log with the smallest time sequence
 * value that's in the log (the oldest log item). any element with a time
 * sequence lower than min_seq will be ignored.
 */
static struct tree_mod_elem *
tree_mod_log_search_oldest(struct btrfs_fs_info *fs_info, u64 start,
                           u64 min_seq)
{
        return __tree_mod_log_search(fs_info, start, min_seq, 1);
}

/*
 * this returns the element from the log with the largest time sequence
 * value that's in the log (the most recent log item). any element with
 * a time sequence lower than min_seq will be ignored.
 */
static struct tree_mod_elem *
tree_mod_log_search(struct btrfs_fs_info *fs_info, u64 start, u64 min_seq)
{
        return __tree_mod_log_search(fs_info, start, min_seq, 0);
}

static noinline int
tree_mod_log_eb_copy(struct btrfs_fs_info *fs_info, struct extent_buffer *dst,
                     struct extent_buffer *src, unsigned long dst_offset,
                     unsigned long src_offset, int nr_items)
{
        int ret = 0;
        struct tree_mod_elem **tm_list = NULL;
        struct tree_mod_elem **tm_list_add, **tm_list_rem;
        int i;
        int locked = 0;

        if (!tree_mod_need_log(fs_info, NULL))
                return 0;

        if (btrfs_header_level(dst) == 0 && btrfs_header_level(src) == 0)
                return 0;

        tm_list = kcalloc(nr_items * 2, sizeof(struct tree_mod_elem *),
                          GFP_NOFS);
        if (!tm_list)
                return -ENOMEM;

        tm_list_add = tm_list;
        tm_list_rem = tm_list + nr_items;
        for (i = 0; i < nr_items; i++) {
                tm_list_rem[i] = alloc_tree_mod_elem(src, i + src_offset,
                    MOD_LOG_KEY_REMOVE, GFP_NOFS);
                if (!tm_list_rem[i]) {
                        ret = -ENOMEM;
                        goto free_tms;
                }

                tm_list_add[i] = alloc_tree_mod_elem(dst, i + dst_offset,
                    MOD_LOG_KEY_ADD, GFP_NOFS);
                if (!tm_list_add[i]) {
                        ret = -ENOMEM;
                        goto free_tms;
                }
        }

        if (tree_mod_dont_log(fs_info, NULL))
                goto free_tms;
        locked = 1;

        for (i = 0; i < nr_items; i++) {
                ret = __tree_mod_log_insert(fs_info, tm_list_rem[i]);
                if (ret)
                        goto free_tms;
                ret = __tree_mod_log_insert(fs_info, tm_list_add[i]);
                if (ret)
                        goto free_tms;
        }

        write_unlock(&fs_info->tree_mod_log_lock);
        kfree(tm_list);

        return 0;

free_tms:
        for (i = 0; i < nr_items * 2; i++) {
                if (tm_list[i] && !RB_EMPTY_NODE(&tm_list[i]->node))
                        rb_erase(&tm_list[i]->node, &fs_info->tree_mod_log);
                kfree(tm_list[i]);
        }
        if (locked)
                write_unlock(&fs_info->tree_mod_log_lock);
        kfree(tm_list);

        return ret;
}

static noinline int tree_mod_log_free_eb(struct extent_buffer *eb)
{
        struct tree_mod_elem **tm_list = NULL;
        int nritems = 0;
        int i;
        int ret = 0;

        if (btrfs_header_level(eb) == 0)
                return 0;

        if (!tree_mod_need_log(eb->fs_info, NULL))
                return 0;

        nritems = btrfs_header_nritems(eb);
        tm_list = kcalloc(nritems, sizeof(struct tree_mod_elem *), GFP_NOFS);
        if (!tm_list)
                return -ENOMEM;

        for (i = 0; i < nritems; i++) {
                tm_list[i] = alloc_tree_mod_elem(eb, i,
                    MOD_LOG_KEY_REMOVE_WHILE_FREEING, GFP_NOFS);
                if (!tm_list[i]) {
                        ret = -ENOMEM;
                        goto free_tms;
                }
        }

        if (tree_mod_dont_log(eb->fs_info, eb))
                goto free_tms;

        ret = __tree_mod_log_free_eb(eb->fs_info, tm_list, nritems);
        write_unlock(&eb->fs_info->tree_mod_log_lock);
        if (ret)
                goto free_tms;
        kfree(tm_list);

        return 0;

free_tms:
        for (i = 0; i < nritems; i++)
                kfree(tm_list[i]);
        kfree(tm_list);

        return ret;
}

/*
 * check if the tree block can be shared by multiple trees
 */
int btrfs_block_can_be_shared(struct btrfs_root *root,
                              struct extent_buffer *buf)
{
        /*
         * Tree blocks not in reference counted trees and tree roots
         * are never shared. If a block was allocated after the last
         * snapshot and the block was not allocated by tree relocation,
         * we know the block is not shared.
         */
        if (test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
            buf != root->node && buf != root->commit_root &&
            (btrfs_header_generation(buf) <=
             btrfs_root_last_snapshot(&root->root_item) ||
             btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)))
                return 1;

        return 0;
}

static noinline int update_ref_for_cow(struct btrfs_trans_handle *trans,
                                       struct btrfs_root *root,
                                       struct extent_buffer *buf,
                                       struct extent_buffer *cow,
                                       int *last_ref)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        u64 refs;
        u64 owner;
        u64 flags;
        u64 new_flags = 0;
        int ret;

        /*
         * Backrefs update rules:
         *
         * Always use full backrefs for extent pointers in tree block
         * allocated by tree relocation.
         *
         * If a shared tree block is no longer referenced by its owner
         * tree (btrfs_header_owner(buf) == root->root_key.objectid),
         * use full backrefs for extent pointers in tree block.
         *
         * If a tree block is been relocating
         * (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID),
         * use full backrefs for extent pointers in tree block.
         * The reason for this is some operations (such as drop tree)
         * are only allowed for blocks use full backrefs.
         */

        if (btrfs_block_can_be_shared(root, buf)) {
                ret = btrfs_lookup_extent_info(trans, fs_info, buf->start,
                                               btrfs_header_level(buf), 1,
                                               &refs, &flags);
                if (ret)
                        return ret;
                if (refs == 0) {
                        ret = -EROFS;
                        btrfs_handle_fs_error(fs_info, ret, NULL);
                        return ret;
                }
        } else {
                refs = 1;
                if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
                    btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
                        flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
                else
                        flags = 0;
        }

        owner = btrfs_header_owner(buf);
        BUG_ON(owner == BTRFS_TREE_RELOC_OBJECTID &&
               !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));

        if (refs > 1) {
                if ((owner == root->root_key.objectid ||
                     root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) &&
                    !(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)) {
                        ret = btrfs_inc_ref(trans, root, buf, 1);
                        if (ret)
                                return ret;

                        if (root->root_key.objectid ==
                            BTRFS_TREE_RELOC_OBJECTID) {
                                ret = btrfs_dec_ref(trans, root, buf, 0);
                                if (ret)
                                        return ret;
                                ret = btrfs_inc_ref(trans, root, cow, 1);
                                if (ret)
                                        return ret;
                        }
                        new_flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
                } else {

                        if (root->root_key.objectid ==
                            BTRFS_TREE_RELOC_OBJECTID)
                                ret = btrfs_inc_ref(trans, root, cow, 1);
                        else
                                ret = btrfs_inc_ref(trans, root, cow, 0);
                        if (ret)
                                return ret;
                }
                if (new_flags != 0) {
                        int level = btrfs_header_level(buf);

                        ret = btrfs_set_disk_extent_flags(trans, fs_info,
                                                          buf->start,
                                                          buf->len,
                                                          new_flags, level, 0);
                        if (ret)
                                return ret;
                }
        } else {
                if (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
                        if (root->root_key.objectid ==
                            BTRFS_TREE_RELOC_OBJECTID)
                                ret = btrfs_inc_ref(trans, root, cow, 1);
                        else
                                ret = btrfs_inc_ref(trans, root, cow, 0);
                        if (ret)
                                return ret;
                        ret = btrfs_dec_ref(trans, root, buf, 1);
                        if (ret)
                                return ret;
                }
                clean_tree_block(fs_info, buf);
                *last_ref = 1;
        }
        return 0;
}

static struct extent_buffer *alloc_tree_block_no_bg_flush(
                                          struct btrfs_trans_handle *trans,
                                          struct btrfs_root *root,
                                          u64 parent_start,
                                          const struct btrfs_disk_key *disk_key,
                                          int level,
                                          u64 hint,
                                          u64 empty_size)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *ret;

        /*
         * If we are COWing a node/leaf from the extent, chunk, device or free
         * space trees, make sure that we do not finish block group creation of
         * pending block groups. We do this to avoid a deadlock.
         * COWing can result in allocation of a new chunk, and flushing pending
         * block groups (btrfs_create_pending_block_groups()) can be triggered
         * when finishing allocation of a new chunk. Creation of a pending block
         * group modifies the extent, chunk, device and free space trees,
         * therefore we could deadlock with ourselves since we are holding a
         * lock on an extent buffer that btrfs_create_pending_block_groups() may
         * try to COW later.
         * For similar reasons, we also need to delay flushing pending block
         * groups when splitting a leaf or node, from one of those trees, since
         * we are holding a write lock on it and its parent or when inserting a
         * new root node for one of those trees.
         */
        if (root == fs_info->extent_root ||
            root == fs_info->chunk_root ||

            root == fs_info->dev_root ||
            root == fs_info->free_space_root)
                trans->can_flush_pending_bgs = false;

        ret = btrfs_alloc_tree_block(trans, root, parent_start,
                                     root->root_key.objectid, disk_key, level,
                                     hint, empty_size);
        trans->can_flush_pending_bgs = true;

        return ret;
}

/*
 * does the dirty work in cow of a single block.  The parent block (if
 * supplied) is updated to point to the new cow copy.  The new buffer is marked
 * dirty and returned locked.  If you modify the block it needs to be marked
 * dirty again.
 *
 * search_start -- an allocation hint for the new block
 *
 * empty_size -- a hint that you plan on doing more cow.  This is the size in
 * bytes the allocator should try to find free next to the block it returns.
 * This is just a hint and may be ignored by the allocator.
 */
static noinline 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 search_start, u64 empty_size)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct btrfs_disk_key disk_key;
        struct extent_buffer *cow;
        int level, ret;
        int last_ref = 0;
        int unlock_orig = 0;
        u64 parent_start = 0;

        if (*cow_ret == buf)
                unlock_orig = 1;

        btrfs_assert_tree_locked(buf);

        WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
                trans->transid != fs_info->running_transaction->transid);
        WARN_ON(test_bit(BTRFS_ROOT_REF_COWS, &root->state) &&
                trans->transid != root->last_trans);

        level = btrfs_header_level(buf);

        if (level == 0)
                btrfs_item_key(buf, &disk_key, 0);
        else
                btrfs_node_key(buf, &disk_key, 0);

        if ((root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID) && parent)
                parent_start = parent->start;

        cow = alloc_tree_block_no_bg_flush(trans, root, parent_start, &disk_key,
                                           level, search_start, empty_size);
        if (IS_ERR(cow))
                return PTR_ERR(cow);

        /* cow is set to blocking by btrfs_init_new_buffer */

        copy_extent_buffer_full(cow, buf);
        btrfs_set_header_bytenr(cow, cow->start);
        btrfs_set_header_generation(cow, trans->transid);
        btrfs_set_header_backref_rev(cow, BTRFS_MIXED_BACKREF_REV);
        btrfs_clear_header_flag(cow, BTRFS_HEADER_FLAG_WRITTEN |
                                     BTRFS_HEADER_FLAG_RELOC);
        if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID)
                btrfs_set_header_flag(cow, BTRFS_HEADER_FLAG_RELOC);
        else
                btrfs_set_header_owner(cow, root->root_key.objectid);

        write_extent_buffer_fsid(cow, fs_info->fs_devices->metadata_uuid);

        ret = update_ref_for_cow(trans, root, buf, cow, &last_ref);
        if (ret) {
                btrfs_abort_transaction(trans, ret);
                return ret;
        }

        if (test_bit(BTRFS_ROOT_REF_COWS, &root->state)) {
                ret = btrfs_reloc_cow_block(trans, root, buf, cow);
                if (ret) {
                        btrfs_abort_transaction(trans, ret);
                        return ret;
                }
        }

        if (buf == root->node) {
                WARN_ON(parent && parent != buf);
                if (root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID ||
                    btrfs_header_backref_rev(buf) < BTRFS_MIXED_BACKREF_REV)
                        parent_start = buf->start;

                extent_buffer_get(cow);
                ret = tree_mod_log_insert_root(root->node, cow, 1);
                BUG_ON(ret < 0);
                rcu_assign_pointer(root->node, cow);

                btrfs_free_tree_block(trans, root, buf, parent_start,
                                      last_ref);
                free_extent_buffer(buf);
                add_root_to_dirty_list(root);
        } else {
                WARN_ON(trans->transid != btrfs_header_generation(parent));
                tree_mod_log_insert_key(parent, parent_slot,
                                        MOD_LOG_KEY_REPLACE, GFP_NOFS);
                btrfs_set_node_blockptr(parent, parent_slot,
                                        cow->start);
                btrfs_set_node_ptr_generation(parent, parent_slot,
                                              trans->transid);
                btrfs_mark_buffer_dirty(parent);
                if (last_ref) {
                        ret = tree_mod_log_free_eb(buf);
                        if (ret) {
                                btrfs_abort_transaction(trans, ret);
                                return ret;
                        }
                }
                btrfs_free_tree_block(trans, root, buf, parent_start,
                                      last_ref);
        }
        if (unlock_orig)
                btrfs_tree_unlock(buf);
        free_extent_buffer_stale(buf);
        btrfs_mark_buffer_dirty(cow);
        *cow_ret = cow;
        return 0;
}

/*
 * returns the logical address of the oldest predecessor of the given root.
 * entries older than time_seq are ignored.
 */
static struct tree_mod_elem *__tree_mod_log_oldest_root(
                struct extent_buffer *eb_root, u64 time_seq)
{
        struct tree_mod_elem *tm;
        struct tree_mod_elem *found = NULL;
        u64 root_logical = eb_root->start;
        int looped = 0;

        if (!time_seq)
                return NULL;

        /*
         * the very last operation that's logged for a root is the
         * replacement operation (if it is replaced at all). this has
         * the logical address of the *new* root, making it the very
         * first operation that's logged for this root.
         */
        while (1) {
                tm = tree_mod_log_search_oldest(eb_root->fs_info, root_logical,
                                                time_seq);
                if (!looped && !tm)
                        return NULL;
                /*
                 * if there are no tree operation for the oldest root, we simply
                 * return it. this should only happen if that (old) root is at
                 * level 0.
                 */
                if (!tm)
                        break;

                /*
                 * if there's an operation that's not a root replacement, we
                 * found the oldest version of our root. normally, we'll find a
                 * MOD_LOG_KEY_REMOVE_WHILE_FREEING operation here.
                 */
                if (tm->op != MOD_LOG_ROOT_REPLACE)
                        break;

                found = tm;
                root_logical = tm->old_root.logical;
                looped = 1;
        }

        /* if there's no old root to return, return what we found instead */
        if (!found)
                found = tm;

        return found;
}

/*
 * tm is a pointer to the first operation to rewind within eb. then, all
 * previous operations will be rewound (until we reach something older than
 * time_seq).
 */
static void
__tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct extent_buffer *eb,
                      u64 time_seq, struct tree_mod_elem *first_tm)
{
        u32 n;
        struct rb_node *next;
        struct tree_mod_elem *tm = first_tm;
        unsigned long o_dst;
        unsigned long o_src;
        unsigned long p_size = sizeof(struct btrfs_key_ptr);

        n = btrfs_header_nritems(eb);
        read_lock(&fs_info->tree_mod_log_lock);
        while (tm && tm->seq >= time_seq) {
                /*
                 * all the operations are recorded with the operator used for
                 * the modification. as we're going backwards, we do the
                 * opposite of each operation here.
                 */
                switch (tm->op) {
                case MOD_LOG_KEY_REMOVE_WHILE_FREEING:
                        BUG_ON(tm->slot < n);
                        /* Fallthrough */
                case MOD_LOG_KEY_REMOVE_WHILE_MOVING:
                case MOD_LOG_KEY_REMOVE:
                        btrfs_set_node_key(eb, &tm->key, tm->slot);
                        btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
                        btrfs_set_node_ptr_generation(eb, tm->slot,
                                                      tm->generation);
                        n++;
                        break;
                case MOD_LOG_KEY_REPLACE:
                        BUG_ON(tm->slot >= n);
                        btrfs_set_node_key(eb, &tm->key, tm->slot);
                        btrfs_set_node_blockptr(eb, tm->slot, tm->blockptr);
                        btrfs_set_node_ptr_generation(eb, tm->slot,
                                                      tm->generation);
                        break;
                case MOD_LOG_KEY_ADD:
                        /* if a move operation is needed it's in the log */
                        n--;
                        break;
                case MOD_LOG_MOVE_KEYS:
                        o_dst = btrfs_node_key_ptr_offset(tm->slot);
                        o_src = btrfs_node_key_ptr_offset(tm->move.dst_slot);
                        memmove_extent_buffer(eb, o_dst, o_src,
                                              tm->move.nr_items * p_size);
                        break;
                case MOD_LOG_ROOT_REPLACE:
                        /*
                         * this operation is special. for roots, this must be
                         * handled explicitly before rewinding.
                         * for non-roots, this operation may exist if the node
                         * was a root: root A -> child B; then A gets empty and
                         * B is promoted to the new root. in the mod log, we'll
                         * have a root-replace operation for B, a tree block
                         * that is no root. we simply ignore that operation.
                         */
                        break;
                }
                next = rb_next(&tm->node);
                if (!next)
                        break;
                tm = rb_entry(next, struct tree_mod_elem, node);
                if (tm->logical != first_tm->logical)
                        break;
        }
        read_unlock(&fs_info->tree_mod_log_lock);
        btrfs_set_header_nritems(eb, n);
}

/*
 * Called with eb read locked. If the buffer cannot be rewound, the same buffer
 * is returned. If rewind operations happen, a fresh buffer is returned. The
 * returned buffer is always read-locked. If the returned buffer is not the
 * input buffer, the lock on the input buffer is released and the input buffer
 * is freed (its refcount is decremented).
 */
static struct extent_buffer *
tree_mod_log_rewind(struct btrfs_fs_info *fs_info, struct btrfs_path *path,
                    struct extent_buffer *eb, u64 time_seq)
{
        struct extent_buffer *eb_rewin;
        struct tree_mod_elem *tm;

        if (!time_seq)
                return eb;

        if (btrfs_header_level(eb) == 0)
                return eb;

        tm = tree_mod_log_search(fs_info, eb->start, time_seq);
        if (!tm)
                return eb;

        btrfs_set_path_blocking(path);
        btrfs_set_lock_blocking_rw(eb, BTRFS_READ_LOCK);

        if (tm->op == MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
                BUG_ON(tm->slot != 0);
                eb_rewin = alloc_dummy_extent_buffer(fs_info, eb->start);
                if (!eb_rewin) {
                        btrfs_tree_read_unlock_blocking(eb);
                        free_extent_buffer(eb);
                        return NULL;
                }
                btrfs_set_header_bytenr(eb_rewin, eb->start);
                btrfs_set_header_backref_rev(eb_rewin,
                                             btrfs_header_backref_rev(eb));
                btrfs_set_header_owner(eb_rewin, btrfs_header_owner(eb));
                btrfs_set_header_level(eb_rewin, btrfs_header_level(eb));
        } else {
                eb_rewin = btrfs_clone_extent_buffer(eb);
                if (!eb_rewin) {
                        btrfs_tree_read_unlock_blocking(eb);
                        free_extent_buffer(eb);
                        return NULL;
                }
        }

        btrfs_tree_read_unlock_blocking(eb);
        free_extent_buffer(eb);

        btrfs_tree_read_lock(eb_rewin);
        __tree_mod_log_rewind(fs_info, eb_rewin, time_seq, tm);
        WARN_ON(btrfs_header_nritems(eb_rewin) >
                BTRFS_NODEPTRS_PER_BLOCK(fs_info));

        return eb_rewin;
}

/*
 * get_old_root() rewinds the state of @root's root node to the given @time_seq
 * value. If there are no changes, the current root->root_node is returned. If
 * anything changed in between, there's a fresh buffer allocated on which the
 * rewind operations are done. In any case, the returned buffer is read locked.
 * Returns NULL on error (with no locks held).
 */
static inline struct extent_buffer *
get_old_root(struct btrfs_root *root, u64 time_seq)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct tree_mod_elem *tm;
        struct extent_buffer *eb = NULL;
        struct extent_buffer *eb_root;
        struct extent_buffer *old;
        struct tree_mod_root *old_root = NULL;
        u64 old_generation = 0;
        u64 logical;
        int level;

        eb_root = btrfs_read_lock_root_node(root);
        tm = __tree_mod_log_oldest_root(eb_root, time_seq);
        if (!tm)
                return eb_root;

        if (tm->op == MOD_LOG_ROOT_REPLACE) {
                old_root = &tm->old_root;
                old_generation = tm->generation;
                logical = old_root->logical;
                level = old_root->level;
        } else {
                logical = eb_root->start;
                level = btrfs_header_level(eb_root);
        }

        tm = tree_mod_log_search(fs_info, logical, time_seq);
        if (old_root && tm && tm->op != MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
                btrfs_tree_read_unlock(eb_root);
                free_extent_buffer(eb_root);
                old = read_tree_block(fs_info, logical, 0, level, NULL);
                if (WARN_ON(IS_ERR(old) || !extent_buffer_uptodate(old))) {
                        if (!IS_ERR(old))
                                free_extent_buffer(old);
                        btrfs_warn(fs_info,
                                   "failed to read tree block %llu from get_old_root",
                                   logical);
                } else {
                        eb = btrfs_clone_extent_buffer(old);
                        free_extent_buffer(old);
                }
        } else if (old_root) {
                btrfs_tree_read_unlock(eb_root);
                free_extent_buffer(eb_root);
                eb = alloc_dummy_extent_buffer(fs_info, logical);
        } else {
                btrfs_set_lock_blocking_rw(eb_root, BTRFS_READ_LOCK);
                eb = btrfs_clone_extent_buffer(eb_root);
                btrfs_tree_read_unlock_blocking(eb_root);
                free_extent_buffer(eb_root);
        }

        if (!eb)
                return NULL;
        btrfs_tree_read_lock(eb);
        if (old_root) {
                btrfs_set_header_bytenr(eb, eb->start);
                btrfs_set_header_backref_rev(eb, BTRFS_MIXED_BACKREF_REV);
                btrfs_set_header_owner(eb, btrfs_header_owner(eb_root));
                btrfs_set_header_level(eb, old_root->level);
                btrfs_set_header_generation(eb, old_generation);
        }
        if (tm)
                __tree_mod_log_rewind(fs_info, eb, time_seq, tm);
        else
                WARN_ON(btrfs_header_level(eb) != 0);
        WARN_ON(btrfs_header_nritems(eb) > BTRFS_NODEPTRS_PER_BLOCK(fs_info));

        return eb;
}

int btrfs_old_root_level(struct btrfs_root *root, u64 time_seq)
{
        struct tree_mod_elem *tm;
        int level;
        struct extent_buffer *eb_root = btrfs_root_node(root);

        tm = __tree_mod_log_oldest_root(eb_root, time_seq);
        if (tm && tm->op == MOD_LOG_ROOT_REPLACE) {
                level = tm->old_root.level;
        } else {
                level = btrfs_header_level(eb_root);
        }
        free_extent_buffer(eb_root);

        return level;
}

static inline int should_cow_block(struct btrfs_trans_handle *trans,
                                   struct btrfs_root *root,
                                   struct extent_buffer *buf)
{
        if (btrfs_is_testing(root->fs_info))
                return 0;

        /* Ensure we can see the FORCE_COW bit */
        smp_mb__before_atomic();

        /*
         * We do not need to cow a block if
         * 1) this block is not created or changed in this transaction;
         * 2) this block does not belong to TREE_RELOC tree;
         * 3) the root is not forced COW.
         *
         * What is forced COW:
         *    when we create snapshot during committing the transaction,
         *    after we've finished copying src root, we must COW the shared
         *    block to ensure the metadata consistency.
         */
        if (btrfs_header_generation(buf) == trans->transid &&
            !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN) &&
            !(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID &&
              btrfs_header_flag(buf, BTRFS_HEADER_FLAG_RELOC)) &&
            !test_bit(BTRFS_ROOT_FORCE_COW, &root->state))
                return 0;
        return 1;
}

/*
 * cows a single block, see __btrfs_cow_block for the real work.
 * This version of it has extra checks so that a block isn't COWed more than
 * once per transaction, as long as it hasn't been written yet
 */
noinline 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)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        u64 search_start;
        int ret;

        if (test_bit(BTRFS_ROOT_DELETING, &root->state))
                btrfs_err(fs_info,
                        "COW'ing blocks on a fs root that's being dropped");

        if (trans->transaction != fs_info->running_transaction)
                WARN(1, KERN_CRIT "trans %llu running %llu\n",
                       trans->transid,
                       fs_info->running_transaction->transid);

        if (trans->transid != fs_info->generation)
                WARN(1, KERN_CRIT "trans %llu running %llu\n",
                       trans->transid, fs_info->generation);

        if (!should_cow_block(trans, root, buf)) {
                trans->dirty = true;
                *cow_ret = buf;
                return 0;
        }

        search_start = buf->start & ~((u64)SZ_1G - 1);

        if (parent)
                btrfs_set_lock_blocking(parent);
        btrfs_set_lock_blocking(buf);

        ret = __btrfs_cow_block(trans, root, buf, parent,
                                 parent_slot, cow_ret, search_start, 0);

        trace_btrfs_cow_block(root, buf, *cow_ret);

        return ret;
}

/*
 * helper function for defrag to decide if two blocks pointed to by a
 * node are actually close by
 */
static int close_blocks(u64 blocknr, u64 other, u32 blocksize)
{
        if (blocknr < other && other - (blocknr + blocksize) < 32768)
                return 1;
        if (blocknr > other && blocknr - (other + blocksize) < 32768)
                return 1;
        return 0;
}

/*
 * compare two keys in a memcmp fashion
 */
static int comp_keys(const struct btrfs_disk_key *disk,
                     const struct btrfs_key *k2)
{
        struct btrfs_key k1;

        btrfs_disk_key_to_cpu(&k1, disk);

        return btrfs_comp_cpu_keys(&k1, k2);
}

/*
 * same as comp_keys only with two btrfs_key's
 */
int btrfs_comp_cpu_keys(const struct btrfs_key *k1, const struct btrfs_key *k2)
{
        if (k1->objectid > k2->objectid)
                return 1;
        if (k1->objectid < k2->objectid)
                return -1;
        if (k1->type > k2->type)
                return 1;
        if (k1->type < k2->type)
                return -1;
        if (k1->offset > k2->offset)
                return 1;
        if (k1->offset < k2->offset)
                return -1;
        return 0;
}

/*
 * this is used by the defrag code to go through all the
 * leaves pointed to by a node and reallocate them so that
 * disk order is close to key order
 */
int btrfs_realloc_node(struct btrfs_trans_handle *trans,
                       struct btrfs_root *root, struct extent_buffer *parent,
                       int start_slot, u64 *last_ret,
                       struct btrfs_key *progress)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *cur;
        u64 blocknr;
        u64 gen;
        u64 search_start = *last_ret;
        u64 last_block = 0;
        u64 other;
        u32 parent_nritems;
        int end_slot;
        int i;
        int err = 0;
        int parent_level;
        int uptodate;
        u32 blocksize;
        int progress_passed = 0;
        struct btrfs_disk_key disk_key;

        parent_level = btrfs_header_level(parent);

        WARN_ON(trans->transaction != fs_info->running_transaction);
        WARN_ON(trans->transid != fs_info->generation);

        parent_nritems = btrfs_header_nritems(parent);
        blocksize = fs_info->nodesize;
        end_slot = parent_nritems - 1;

        if (parent_nritems <= 1)
                return 0;

        btrfs_set_lock_blocking(parent);

        for (i = start_slot; i <= end_slot; i++) {
                struct btrfs_key first_key;
                int close = 1;

                btrfs_node_key(parent, &disk_key, i);
                if (!progress_passed && comp_keys(&disk_key, progress) < 0)
                        continue;

                progress_passed = 1;
                blocknr = btrfs_node_blockptr(parent, i);
                gen = btrfs_node_ptr_generation(parent, i);
                btrfs_node_key_to_cpu(parent, &first_key, i);
                if (last_block == 0)
                        last_block = blocknr;

                if (i > 0) {
                        other = btrfs_node_blockptr(parent, i - 1);
                        close = close_blocks(blocknr, other, blocksize);
                }
                if (!close && i < end_slot) {
                        other = btrfs_node_blockptr(parent, i + 1);
                        close = close_blocks(blocknr, other, blocksize);
                }
                if (close) {
                        last_block = blocknr;
                        continue;
                }

                cur = find_extent_buffer(fs_info, blocknr);
                if (cur)
                        uptodate = btrfs_buffer_uptodate(cur, gen, 0);
                else
                        uptodate = 0;
                if (!cur || !uptodate) {
                        if (!cur) {
                                cur = read_tree_block(fs_info, blocknr, gen,
                                                      parent_level - 1,
                                                      &first_key);
                                if (IS_ERR(cur)) {
                                        return PTR_ERR(cur);
                                } else if (!extent_buffer_uptodate(cur)) {
                                        free_extent_buffer(cur);
                                        return -EIO;
                                }
                        } else if (!uptodate) {
                                err = btrfs_read_buffer(cur, gen,
                                                parent_level - 1,&first_key);
                                if (err) {
                                        free_extent_buffer(cur);
                                        return err;
                                }
                        }
                }
                if (search_start == 0)
                        search_start = last_block;

                btrfs_tree_lock(cur);
                btrfs_set_lock_blocking(cur);
                err = __btrfs_cow_block(trans, root, cur, parent, i,
                                        &cur, search_start,
                                        min(16 * blocksize,
                                            (end_slot - i) * blocksize));
                if (err) {
                        btrfs_tree_unlock(cur);
                        free_extent_buffer(cur);
                        break;
                }
                search_start = cur->start;
                last_block = cur->start;
                *last_ret = search_start;
                btrfs_tree_unlock(cur);
                free_extent_buffer(cur);
        }
        return err;
}

/*
 * search for key in the extent_buffer.  The items start at offset p,
 * and they are item_size apart.  There are 'max' items in p.
 *
 * the slot in the array is returned via slot, and it points to
 * the place where you would insert key if it is not found in
 * the array.
 *
 * slot may point to max if the key is bigger than all of the keys
 */
static noinline int generic_bin_search(struct extent_buffer *eb,
                                       unsigned long p, int item_size,
                                       const struct btrfs_key *key,
                                       int max, int *slot)
{
        int low = 0;
        int high = max;
        int mid;
        int ret;
        struct btrfs_disk_key *tmp = NULL;
        struct btrfs_disk_key unaligned;
        unsigned long offset;
        char *kaddr = NULL;
        unsigned long map_start = 0;
        unsigned long map_len = 0;
        int err;

        if (low > high) {
                btrfs_err(eb->fs_info,
                 "%s: low (%d) > high (%d) eb %llu owner %llu level %d",
                          __func__, low, high, eb->start,
                          btrfs_header_owner(eb), btrfs_header_level(eb));
                return -EINVAL;
        }

        while (low < high) {
                mid = (low + high) / 2;
                offset = p + mid * item_size;

                if (!kaddr || offset < map_start ||
                    (offset + sizeof(struct btrfs_disk_key)) >
                    map_start + map_len) {

                        err = map_private_extent_buffer(eb, offset,
                                                sizeof(struct btrfs_disk_key),
                                                &kaddr, &map_start, &map_len);

                        if (!err) {
                                tmp = (struct btrfs_disk_key *)(kaddr + offset -
                                                        map_start);
                        } else if (err == 1) {
                                read_extent_buffer(eb, &unaligned,
                                                   offset, sizeof(unaligned));
                                tmp = &unaligned;
                        } else {
                                return err;
                        }

                } else {
                        tmp = (struct btrfs_disk_key *)(kaddr + offset -
                                                        map_start);
                }
                ret = comp_keys(tmp, key);

                if (ret < 0)
                        low = mid + 1;
                else if (ret > 0)
                        high = mid;
                else {
                        *slot = mid;
                        return 0;
                }
        }
        *slot = low;
        return 1;
}

/*
 * simple bin_search frontend that does the right thing for
 * leaves vs nodes
 */
int btrfs_bin_search(struct extent_buffer *eb, const struct btrfs_key *key,
                     int level, int *slot)
{
        if (level == 0)
                return generic_bin_search(eb,
                                          offsetof(struct btrfs_leaf, items),
                                          sizeof(struct btrfs_item),
                                          key, btrfs_header_nritems(eb),
                                          slot);
        else
                return generic_bin_search(eb,
                                          offsetof(struct btrfs_node, ptrs),
                                          sizeof(struct btrfs_key_ptr),
                                          key, btrfs_header_nritems(eb),
                                          slot);
}

static void root_add_used(struct btrfs_root *root, u32 size)
{
        spin_lock(&root->accounting_lock);
        btrfs_set_root_used(&root->root_item,
                            btrfs_root_used(&root->root_item) + size);
        spin_unlock(&root->accounting_lock);
}

static void root_sub_used(struct btrfs_root *root, u32 size)
{
        spin_lock(&root->accounting_lock);
        btrfs_set_root_used(&root->root_item,
                            btrfs_root_used(&root->root_item) - size);
        spin_unlock(&root->accounting_lock);
}

/* given a node and slot number, this reads the blocks it points to.  The
 * extent buffer is returned with a reference taken (but unlocked).
 */
static noinline struct extent_buffer *
read_node_slot(struct btrfs_fs_info *fs_info, struct extent_buffer *parent,
               int slot)
{
        int level = btrfs_header_level(parent);
        struct extent_buffer *eb;
        struct btrfs_key first_key;

        if (slot < 0 || slot >= btrfs_header_nritems(parent))
                return ERR_PTR(-ENOENT);

        BUG_ON(level == 0);

        btrfs_node_key_to_cpu(parent, &first_key, slot);
        eb = read_tree_block(fs_info, btrfs_node_blockptr(parent, slot),
                             btrfs_node_ptr_generation(parent, slot),
                             level - 1, &first_key);
        if (!IS_ERR(eb) && !extent_buffer_uptodate(eb)) {
                free_extent_buffer(eb);
                eb = ERR_PTR(-EIO);
        }

        return eb;
}

/*
 * node level balancing, used to make sure nodes are in proper order for
 * item deletion.  We balance from the top down, so we have to make sure
 * that a deletion won't leave an node completely empty later on.
 */
static noinline int balance_level(struct btrfs_trans_handle *trans,
                         struct btrfs_root *root,
                         struct btrfs_path *path, int level)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *right = NULL;
        struct extent_buffer *mid;
        struct extent_buffer *left = NULL;
        struct extent_buffer *parent = NULL;
        int ret = 0;
        int wret;
        int pslot;
        int orig_slot = path->slots[level];
        u64 orig_ptr;

        ASSERT(level > 0);

        mid = path->nodes[level];

        WARN_ON(path->locks[level] != BTRFS_WRITE_LOCK &&
                path->locks[level] != BTRFS_WRITE_LOCK_BLOCKING);
        WARN_ON(btrfs_header_generation(mid) != trans->transid);

        orig_ptr = btrfs_node_blockptr(mid, orig_slot);

        if (level < BTRFS_MAX_LEVEL - 1) {
                parent = path->nodes[level + 1];
                pslot = path->slots[level + 1];
        }

        /*
         * deal with the case where there is only one pointer in the root
         * by promoting the node below to a root
         */
        if (!parent) {
                struct extent_buffer *child;

                if (btrfs_header_nritems(mid) != 1)
                        return 0;

                /* promote the child to a root */
                child = read_node_slot(fs_info, mid, 0);
                if (IS_ERR(child)) {
                        ret = PTR_ERR(child);
                        btrfs_handle_fs_error(fs_info, ret, NULL);
                        goto enospc;
                }

                btrfs_tree_lock(child);
                btrfs_set_lock_blocking(child);
                ret = btrfs_cow_block(trans, root, child, mid, 0, &child);
                if (ret) {
                        btrfs_tree_unlock(child);
                        free_extent_buffer(child);
                        goto enospc;
                }

                ret = tree_mod_log_insert_root(root->node, child, 1);
                BUG_ON(ret < 0);
                rcu_assign_pointer(root->node, child);

                add_root_to_dirty_list(root);
                btrfs_tree_unlock(child);

                path->locks[level] = 0;
                path->nodes[level] = NULL;
                clean_tree_block(fs_info, mid);
                btrfs_tree_unlock(mid);
                /* once for the path */
                free_extent_buffer(mid);

                root_sub_used(root, mid->len);
                btrfs_free_tree_block(trans, root, mid, 0, 1);
                /* once for the root ptr */
                free_extent_buffer_stale(mid);
                return 0;
        }
        if (btrfs_header_nritems(mid) >
            BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 4)
                return 0;

        left = read_node_slot(fs_info, parent, pslot - 1);
        if (IS_ERR(left))
                left = NULL;

        if (left) {
                btrfs_tree_lock(left);
                btrfs_set_lock_blocking(left);
                wret = btrfs_cow_block(trans, root, left,
                                       parent, pslot - 1, &left);
                if (wret) {
                        ret = wret;
                        goto enospc;
                }
        }

        right = read_node_slot(fs_info, parent, pslot + 1);
        if (IS_ERR(right))
                right = NULL;

        if (right) {
                btrfs_tree_lock(right);
                btrfs_set_lock_blocking(right);
                wret = btrfs_cow_block(trans, root, right,
                                       parent, pslot + 1, &right);
                if (wret) {
                        ret = wret;
                        goto enospc;
                }
        }

        /* first, try to make some room in the middle buffer */
        if (left) {
                orig_slot += btrfs_header_nritems(left);
                wret = push_node_left(trans, fs_info, left, mid, 1);
                if (wret < 0)
                        ret = wret;
        }

        /*
         * then try to empty the right most buffer into the middle
         */
        if (right) {
                wret = push_node_left(trans, fs_info, mid, right, 1);
                if (wret < 0 && wret != -ENOSPC)
                        ret = wret;
                if (btrfs_header_nritems(right) == 0) {
                        clean_tree_block(fs_info, right);
                        btrfs_tree_unlock(right);
                        del_ptr(root, path, level + 1, pslot + 1);
                        root_sub_used(root, right->len);
                        btrfs_free_tree_block(trans, root, right, 0, 1);
                        free_extent_buffer_stale(right);
                        right = NULL;
                } else {
                        struct btrfs_disk_key right_key;
                        btrfs_node_key(right, &right_key, 0);
                        ret = tree_mod_log_insert_key(parent, pslot + 1,
                                        MOD_LOG_KEY_REPLACE, GFP_NOFS);
                        BUG_ON(ret < 0);
                        btrfs_set_node_key(parent, &right_key, pslot + 1);
                        btrfs_mark_buffer_dirty(parent);
                }
        }
        if (btrfs_header_nritems(mid) == 1) {
                /*
                 * we're not allowed to leave a node with one item in the
                 * tree during a delete.  A deletion from lower in the tree
                 * could try to delete the only pointer in this node.
                 * So, pull some keys from the left.
                 * There has to be a left pointer at this point because
                 * otherwise we would have pulled some pointers from the
                 * right
                 */
                if (!left) {
                        ret = -EROFS;
                        btrfs_handle_fs_error(fs_info, ret, NULL);
                        goto enospc;
                }
                wret = balance_node_right(trans, fs_info, mid, left);
                if (wret < 0) {
                        ret = wret;
                        goto enospc;
                }
                if (wret == 1) {
                        wret = push_node_left(trans, fs_info, left, mid, 1);
                        if (wret < 0)
                                ret = wret;
                }
                BUG_ON(wret == 1);
        }
        if (btrfs_header_nritems(mid) == 0) {
                clean_tree_block(fs_info, mid);
                btrfs_tree_unlock(mid);
                del_ptr(root, path, level + 1, pslot);
                root_sub_used(root, mid->len);
                btrfs_free_tree_block(trans, root, mid, 0, 1);
                free_extent_buffer_stale(mid);
                mid = NULL;
        } else {
                /* update the parent key to reflect our changes */
                struct btrfs_disk_key mid_key;
                btrfs_node_key(mid, &mid_key, 0);
                ret = tree_mod_log_insert_key(parent, pslot,
                                MOD_LOG_KEY_REPLACE, GFP_NOFS);
                BUG_ON(ret < 0);
                btrfs_set_node_key(parent, &mid_key, pslot);
                btrfs_mark_buffer_dirty(parent);
        }

        /* update the path */

        if (left) {
                if (btrfs_header_nritems(left) > orig_slot) {
                        extent_buffer_get(left);
                        /* left was locked after cow */
                        path->nodes[level] = left;
                        path->slots[level + 1] -= 1;
                        path->slots[level] = orig_slot;
                        if (mid) {
                                btrfs_tree_unlock(mid);
                                free_extent_buffer(mid);
                        }
                } else {
                        orig_slot -= btrfs_header_nritems(left);
                        path->slots[level] = orig_slot;
                }
        }
        /* double check we haven't messed things up */
        if (orig_ptr !=
            btrfs_node_blockptr(path->nodes[level], path->slots[level]))
                BUG();
enospc:
        if (right) {
                btrfs_tree_unlock(right);
                free_extent_buffer(right);
        }
        if (left) {
                if (path->nodes[level] != left)
                        btrfs_tree_unlock(left);
                free_extent_buffer(left);
        }
        return ret;
}

/* Node balancing for insertion.  Here we only split or push nodes around
 * when they are completely full.  This is also done top down, so we
 * have to be pessimistic.
 */
static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
                                          struct btrfs_root *root,
                                          struct btrfs_path *path, int level)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *right = NULL;
        struct extent_buffer *mid;
        struct extent_buffer *left = NULL;
        struct extent_buffer *parent = NULL;
        int ret = 0;
        int wret;
        int pslot;
        int orig_slot = path->slots[level];

        if (level == 0)
                return 1;

        mid = path->nodes[level];
        WARN_ON(btrfs_header_generation(mid) != trans->transid);

        if (level < BTRFS_MAX_LEVEL - 1) {
                parent = path->nodes[level + 1];
                pslot = path->slots[level + 1];
        }

        if (!parent)
                return 1;

        left = read_node_slot(fs_info, parent, pslot - 1);
        if (IS_ERR(left))
                left = NULL;

        /* first, try to make some room in the middle buffer */
        if (left) {
                u32 left_nr;

                btrfs_tree_lock(left);
                btrfs_set_lock_blocking(left);

                left_nr = btrfs_header_nritems(left);
                if (left_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
                        wret = 1;
                } else {
                        ret = btrfs_cow_block(trans, root, left, parent,
                                              pslot - 1, &left);
                        if (ret)
                                wret = 1;
                        else {
                                wret = push_node_left(trans, fs_info,
                                                      left, mid, 0);
                        }
                }
                if (wret < 0)
                        ret = wret;
                if (wret == 0) {
                        struct btrfs_disk_key disk_key;
                        orig_slot += left_nr;
                        btrfs_node_key(mid, &disk_key, 0);
                        ret = tree_mod_log_insert_key(parent, pslot,
                                        MOD_LOG_KEY_REPLACE, GFP_NOFS);
                        BUG_ON(ret < 0);
                        btrfs_set_node_key(parent, &disk_key, pslot);
                        btrfs_mark_buffer_dirty(parent);
                        if (btrfs_header_nritems(left) > orig_slot) {
                                path->nodes[level] = left;
                                path->slots[level + 1] -= 1;
                                path->slots[level] = orig_slot;
                                btrfs_tree_unlock(mid);
                                free_extent_buffer(mid);
                        } else {
                                orig_slot -=
                                        btrfs_header_nritems(left);
                                path->slots[level] = orig_slot;
                                btrfs_tree_unlock(left);
                                free_extent_buffer(left);
                        }
                        return 0;
                }
                btrfs_tree_unlock(left);
                free_extent_buffer(left);
        }
        right = read_node_slot(fs_info, parent, pslot + 1);
        if (IS_ERR(right))
                right = NULL;

        /*
         * then try to empty the right most buffer into the middle
         */
        if (right) {
                u32 right_nr;

                btrfs_tree_lock(right);
                btrfs_set_lock_blocking(right);

                right_nr = btrfs_header_nritems(right);
                if (right_nr >= BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 1) {
                        wret = 1;
                } else {
                        ret = btrfs_cow_block(trans, root, right,
                                              parent, pslot + 1,
                                              &right);
                        if (ret)
                                wret = 1;
                        else {
                                wret = balance_node_right(trans, fs_info,
                                                          right, mid);
                        }
                }
                if (wret < 0)
                        ret = wret;
                if (wret == 0) {
                        struct btrfs_disk_key disk_key;

                        btrfs_node_key(right, &disk_key, 0);
                        ret = tree_mod_log_insert_key(parent, pslot + 1,
                                        MOD_LOG_KEY_REPLACE, GFP_NOFS);
                        BUG_ON(ret < 0);
                        btrfs_set_node_key(parent, &disk_key, pslot + 1);
                        btrfs_mark_buffer_dirty(parent);

                        if (btrfs_header_nritems(mid) <= orig_slot) {
                                path->nodes[level] = right;
                                path->slots[level + 1] += 1;
                                path->slots[level] = orig_slot -
                                        btrfs_header_nritems(mid);
                                btrfs_tree_unlock(mid);
                                free_extent_buffer(mid);
                        } else {
                                btrfs_tree_unlock(right);
                                free_extent_buffer(right);
                        }
                        return 0;
                }
                btrfs_tree_unlock(right);
                free_extent_buffer(right);
        }
        return 1;
}

/*
 * readahead one full node of leaves, finding things that are close
 * to the block in 'slot', and triggering ra on them.
 */
static void reada_for_search(struct btrfs_fs_info *fs_info,
                             struct btrfs_path *path,
                             int level, int slot, u64 objectid)
{
        struct extent_buffer *node;
        struct btrfs_disk_key disk_key;
        u32 nritems;
        u64 search;
        u64 target;
        u64 nread = 0;
        struct extent_buffer *eb;
        u32 nr;
        u32 blocksize;
        u32 nscan = 0;

        if (level != 1)
                return;

        if (!path->nodes[level])
                return;

        node = path->nodes[level];

        search = btrfs_node_blockptr(node, slot);
        blocksize = fs_info->nodesize;
        eb = find_extent_buffer(fs_info, search);
        if (eb) {
                free_extent_buffer(eb);
                return;
        }

        target = search;

        nritems = btrfs_header_nritems(node);
        nr = slot;

        while (1) {
                if (path->reada == READA_BACK) {
                        if (nr == 0)
                                break;
                        nr--;
                } else if (path->reada == READA_FORWARD) {
                        nr++;
                        if (nr >= nritems)
                                break;
                }
                if (path->reada == READA_BACK && objectid) {
                        btrfs_node_key(node, &disk_key, nr);
                        if (btrfs_disk_key_objectid(&disk_key) != objectid)
                                break;
                }
                search = btrfs_node_blockptr(node, nr);
                if ((search <= target && target - search <= 65536) ||
                    (search > target && search - target <= 65536)) {
                        readahead_tree_block(fs_info, search);
                        nread += blocksize;
                }
                nscan++;
                if ((nread > 65536 || nscan > 32))
                        break;
        }
}

static noinline void reada_for_balance(struct btrfs_fs_info *fs_info,
                                       struct btrfs_path *path, int level)
{
        int slot;
        int nritems;
        struct extent_buffer *parent;
        struct extent_buffer *eb;
        u64 gen;
        u64 block1 = 0;
        u64 block2 = 0;

        parent = path->nodes[level + 1];
        if (!parent)
                return;

        nritems = btrfs_header_nritems(parent);
        slot = path->slots[level + 1];

        if (slot > 0) {
                block1 = btrfs_node_blockptr(parent, slot - 1);
                gen = btrfs_node_ptr_generation(parent, slot - 1);
                eb = find_extent_buffer(fs_info, block1);
                /*
                 * if we get -eagain from btrfs_buffer_uptodate, we
                 * don't want to return eagain here.  That will loop
                 * forever
                 */
                if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
                        block1 = 0;
                free_extent_buffer(eb);
        }
        if (slot + 1 < nritems) {
                block2 = btrfs_node_blockptr(parent, slot + 1);
                gen = btrfs_node_ptr_generation(parent, slot + 1);
                eb = find_extent_buffer(fs_info, block2);
                if (eb && btrfs_buffer_uptodate(eb, gen, 1) != 0)
                        block2 = 0;
                free_extent_buffer(eb);
        }

        if (block1)
                readahead_tree_block(fs_info, block1);
        if (block2)
                readahead_tree_block(fs_info, block2);
}


/*
 * when we walk down the tree, it is usually safe to unlock the higher layers
 * in the tree.  The exceptions are when our path goes through slot 0, because
 * operations on the tree might require changing key pointers higher up in the
 * tree.
 *
 * callers might also have set path->keep_locks, which tells this code to keep
 * the lock if the path points to the last slot in the block.  This is part of
 * walking through the tree, and selecting the next slot in the higher block.
 *
 * lowest_unlock sets the lowest level in the tree we're allowed to unlock.  so
 * if lowest_unlock is 1, level 0 won't be unlocked
 */
static noinline void unlock_up(struct btrfs_path *path, int level,
                               int lowest_unlock, int min_write_lock_level,
                               int *write_lock_level)
{
        int i;
        int skip_level = level;
        int no_skips = 0;
        struct extent_buffer *t;

        for (i = level; i < BTRFS_MAX_LEVEL; i++) {
                if (!path->nodes[i])
                        break;
                if (!path->locks[i])
                        break;
                if (!no_skips && path->slots[i] == 0) {
                        skip_level = i + 1;
                        continue;
                }
                if (!no_skips && path->keep_locks) {
                        u32 nritems;
                        t = path->nodes[i];
                        nritems = btrfs_header_nritems(t);
                        if (nritems < 1 || path->slots[i] >= nritems - 1) {
                                skip_level = i + 1;
                                continue;
                        }
                }
                if (skip_level < i && i >= lowest_unlock)
                        no_skips = 1;

                t = path->nodes[i];
                if (i >= lowest_unlock && i > skip_level) {
                        btrfs_tree_unlock_rw(t, path->locks[i]);
                        path->locks[i] = 0;
                        if (write_lock_level &&
                            i > min_write_lock_level &&
                            i <= *write_lock_level) {
                                *write_lock_level = i - 1;
                        }
                }
        }
}

/*
 * This releases any locks held in the path starting at level and
 * going all the way up to the root.
 *
 * btrfs_search_slot will keep the lock held on higher nodes in a few
 * corner cases, such as COW of the block at slot zero in the node.  This
 * ignores those rules, and it should only be called when there are no
 * more updates to be done higher up in the tree.
 */
noinline void btrfs_unlock_up_safe(struct btrfs_path *path, int level)
{
        int i;

        if (path->keep_locks)
                return;

        for (i = level; i < BTRFS_MAX_LEVEL; i++) {
                if (!path->nodes[i])
                        continue;
                if (!path->locks[i])
                        continue;
                btrfs_tree_unlock_rw(path->nodes[i], path->locks[i]);
                path->locks[i] = 0;
        }
}

/*
 * helper function for btrfs_search_slot.  The goal is to find a block
 * in cache without setting the path to blocking.  If we find the block
 * we return zero and the path is unchanged.
 *
 * If we can't find the block, we set the path blocking and do some
 * reada.  -EAGAIN is returned and the search must be repeated.
 */
static int
read_block_for_search(struct btrfs_root *root, struct btrfs_path *p,
                      struct extent_buffer **eb_ret, int level, int slot,
                      const struct btrfs_key *key)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        u64 blocknr;
        u64 gen;
        struct extent_buffer *b = *eb_ret;
        struct extent_buffer *tmp;
        struct btrfs_key first_key;
        int ret;
        int parent_level;

        blocknr = btrfs_node_blockptr(b, slot);
        gen = btrfs_node_ptr_generation(b, slot);
        parent_level = btrfs_header_level(b);
        btrfs_node_key_to_cpu(b, &first_key, slot);

        tmp = find_extent_buffer(fs_info, blocknr);
        if (tmp) {
                /* first we do an atomic uptodate check */
                if (btrfs_buffer_uptodate(tmp, gen, 1) > 0) {
                        *eb_ret = tmp;
                        return 0;
                }

                /* the pages were up to date, but we failed
                 * the generation number check.  Do a full
                 * read for the generation number that is correct.
                 * We must do this without dropping locks so
                 * we can trust our generation number
                 */
                btrfs_set_path_blocking(p);

                /* now we're allowed to do a blocking uptodate check */
                ret = btrfs_read_buffer(tmp, gen, parent_level - 1, &first_key);
                if (!ret) {
                        *eb_ret = tmp;
                        return 0;
                }
                free_extent_buffer(tmp);
                btrfs_release_path(p);
                return -EIO;
        }

        /*
         * reduce lock contention at high levels
         * of the btree by dropping locks before
         * we read.  Don't release the lock on the current
         * level because we need to walk this node to figure
         * out which blocks to read.
         */
        btrfs_unlock_up_safe(p, level + 1);
        btrfs_set_path_blocking(p);

        if (p->reada != READA_NONE)
                reada_for_search(fs_info, p, level, slot, key->objectid);

        ret = -EAGAIN;
        tmp = read_tree_block(fs_info, blocknr, gen, parent_level - 1,
                              &first_key);
        if (!IS_ERR(tmp)) {
                /*
                 * If the read above didn't mark this buffer up to date,
                 * it will never end up being up to date.  Set ret to EIO now
                 * and give up so that our caller doesn't loop forever
                 * on our EAGAINs.
                 */
                if (!extent_buffer_uptodate(tmp))
                        ret = -EIO;
                free_extent_buffer(tmp);
        } else {
                ret = PTR_ERR(tmp);
        }

        btrfs_release_path(p);
        return ret;
}

/*
 * helper function for btrfs_search_slot.  This does all of the checks
 * for node-level blocks and does any balancing required based on
 * the ins_len.
 *
 * If no extra work was required, zero is returned.  If we had to
 * drop the path, -EAGAIN is returned and btrfs_search_slot must
 * start over
 */
static int
setup_nodes_for_search(struct btrfs_trans_handle *trans,
                       struct btrfs_root *root, struct btrfs_path *p,
                       struct extent_buffer *b, int level, int ins_len,
                       int *write_lock_level)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        int ret;

        if ((p->search_for_split || ins_len > 0) && btrfs_header_nritems(b) >=
            BTRFS_NODEPTRS_PER_BLOCK(fs_info) - 3) {
                int sret;

                if (*write_lock_level < level + 1) {
                        *write_lock_level = level + 1;
                        btrfs_release_path(p);
                        goto again;
                }

                btrfs_set_path_blocking(p);
                reada_for_balance(fs_info, p, level);
                sret = split_node(trans, root, p, level);

                BUG_ON(sret > 0);
                if (sret) {
                        ret = sret;
                        goto done;
                }
                b = p->nodes[level];
        } else if (ins_len < 0 && btrfs_header_nritems(b) <
                   BTRFS_NODEPTRS_PER_BLOCK(fs_info) / 2) {
                int sret;

                if (*write_lock_level < level + 1) {
                        *write_lock_level = level + 1;
                        btrfs_release_path(p);
                        goto again;
                }

                btrfs_set_path_blocking(p);
                reada_for_balance(fs_info, p, level);
                sret = balance_level(trans, root, p, level);

                if (sret) {
                        ret = sret;
                        goto done;
                }
                b = p->nodes[level];
                if (!b) {
                        btrfs_release_path(p);
                        goto again;
                }
                BUG_ON(btrfs_header_nritems(b) == 1);
        }
        return 0;

again:
        ret = -EAGAIN;
done:
        return ret;
}

static void key_search_validate(struct extent_buffer *b,
                                const struct btrfs_key *key,
                                int level)
{
#ifdef CONFIG_BTRFS_ASSERT
        struct btrfs_disk_key disk_key;

        btrfs_cpu_key_to_disk(&disk_key, key);

        if (level == 0)
                ASSERT(!memcmp_extent_buffer(b, &disk_key,
                    offsetof(struct btrfs_leaf, items[0].key),
                    sizeof(disk_key)));
        else
                ASSERT(!memcmp_extent_buffer(b, &disk_key,
                    offsetof(struct btrfs_node, ptrs[0].key),
                    sizeof(disk_key)));
#endif
}

static int key_search(struct extent_buffer *b, const struct btrfs_key *key,
                      int level, int *prev_cmp, int *slot)
{
        if (*prev_cmp != 0) {
                *prev_cmp = btrfs_bin_search(b, key, level, slot);
                return *prev_cmp;
        }

        key_search_validate(b, key, level);
        *slot = 0;

        return 0;
}

int btrfs_find_item(struct btrfs_root *fs_root, struct btrfs_path *path,
                u64 iobjectid, u64 ioff, u8 key_type,
                struct btrfs_key *found_key)
{
        int ret;
        struct btrfs_key key;
        struct extent_buffer *eb;

        ASSERT(path);
        ASSERT(found_key);

        key.type = key_type;
        key.objectid = iobjectid;
        key.offset = ioff;

        ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
        if (ret < 0)
                return ret;

        eb = path->nodes[0];
        if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
                ret = btrfs_next_leaf(fs_root, path);
                if (ret)
                        return ret;
                eb = path->nodes[0];
        }

        btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
        if (found_key->type != key.type ||
                        found_key->objectid != key.objectid)
                return 1;

        return 0;
}

static struct extent_buffer *btrfs_search_slot_get_root(struct btrfs_root *root,
                                                        struct btrfs_path *p,
                                                        int write_lock_level)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *b;
        int root_lock;
        int level = 0;

        /* We try very hard to do read locks on the root */
        root_lock = BTRFS_READ_LOCK;

        if (p->search_commit_root) {
                /*
                 * The commit roots are read only so we always do read locks,
                 * and we always must hold the commit_root_sem when doing
                 * searches on them, the only exception is send where we don't
                 * want to block transaction commits for a long time, so
                 * we need to clone the commit root in order to avoid races
                 * with transaction commits that create a snapshot of one of
                 * the roots used by a send operation.
                 */
                if (p->need_commit_sem) {
                        down_read(&fs_info->commit_root_sem);
                        b = btrfs_clone_extent_buffer(root->commit_root);
                        up_read(&fs_info->commit_root_sem);
                        if (!b)
                                return ERR_PTR(-ENOMEM);

                } else {
                        b = root->commit_root;
                        extent_buffer_get(b);
                }
                level = btrfs_header_level(b);
                /*
                 * Ensure that all callers have set skip_locking when
                 * p->search_commit_root = 1.
                 */
                ASSERT(p->skip_locking == 1);

                goto out;
        }

        if (p->skip_locking) {
                b = btrfs_root_node(root);
                level = btrfs_header_level(b);
                goto out;
        }

        /*
         * If the level is set to maximum, we can skip trying to get the read
         * lock.
         */
        if (write_lock_level < BTRFS_MAX_LEVEL) {
                /*
                 * We don't know the level of the root node until we actually
                 * have it read locked
                 */
                b = btrfs_read_lock_root_node(root);
                level = btrfs_header_level(b);
                if (level > write_lock_level)
                        goto out;

                /* Whoops, must trade for write lock */
                btrfs_tree_read_unlock(b);
                free_extent_buffer(b);
        }

        b = btrfs_lock_root_node(root);
        root_lock = BTRFS_WRITE_LOCK;

        /* The level might have changed, check again */
        level = btrfs_header_level(b);

out:
        p->nodes[level] = b;
        if (!p->skip_locking)
                p->locks[level] = root_lock;
        /*
         * Callers are responsible for dropping b's references.
         */
        return b;
}


/*
 * btrfs_search_slot - look for a key in a tree and perform necessary
 * modifications to preserve tree invariants.
 *
 * @trans:      Handle of transaction, used when modifying the tree
 * @p:          Holds all btree nodes along the search path
 * @root:       The root node of the tree
 * @key:        The key we are looking for
 * @ins_len:    Indicates purpose of search, for inserts it is 1, for
 *              deletions it's -1. 0 for plain searches
 * @cow:        boolean should CoW operations be performed. Must always be 1
 *              when modifying the tree.
 *
 * If @ins_len > 0, nodes and leaves will be split as we walk down the tree.
 * If @ins_len < 0, nodes will be merged as we walk down the tree (if possible)
 *
 * If @key is found, 0 is returned and you can find the item in the leaf level
 * of the path (level 0)
 *
 * If @key isn't found, 1 is returned and the leaf level of the path (level 0)
 * points to the slot where it should be inserted
 *
 * If an error is encountered while searching the tree a negative error number
 * is returned
 */
int btrfs_search_slot(struct btrfs_trans_handle *trans, struct btrfs_root *root,
                      const struct btrfs_key *key, struct btrfs_path *p,
                      int ins_len, int cow)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *b;
        int slot;
        int ret;
        int err;
        int level;
        int lowest_unlock = 1;
        /* everything at write_lock_level or lower must be write locked */
        int write_lock_level = 0;
        u8 lowest_level = 0;
        int min_write_lock_level;
        int prev_cmp;

        lowest_level = p->lowest_level;
        WARN_ON(lowest_level && ins_len > 0);
        WARN_ON(p->nodes[0] != NULL);
        BUG_ON(!cow && ins_len);

        if (ins_len < 0) {
                lowest_unlock = 2;

                /* when we are removing items, we might have to go up to level
                 * two as we update tree pointers  Make sure we keep write
                 * for those levels as well
                 */
                write_lock_level = 2;
        } else if (ins_len > 0) {
                /*
                 * for inserting items, make sure we have a write lock on
                 * level 1 so we can update keys
                 */
                write_lock_level = 1;
        }

        if (!cow)
                write_lock_level = -1;

        if (cow && (p->keep_locks || p->lowest_level))
                write_lock_level = BTRFS_MAX_LEVEL;

        min_write_lock_level = write_lock_level;

again:
        prev_cmp = -1;
        b = btrfs_search_slot_get_root(root, p, write_lock_level);
        if (IS_ERR(b)) {
                ret = PTR_ERR(b);
                goto done;
        }

        while (b) {
                level = btrfs_header_level(b);

                /*
                 * setup the path here so we can release it under lock
                 * contention with the cow code
                 */
                if (cow) {
                        bool last_level = (level == (BTRFS_MAX_LEVEL - 1));

                        /*
                         * if we don't really need to cow this block
                         * then we don't want to set the path blocking,
                         * so we test it here
                         */
                        if (!should_cow_block(trans, root, b)) {
                                trans->dirty = true;
                                goto cow_done;
                        }

                        /*
                         * must have write locks on this node and the
                         * parent
                         */
                        if (level > write_lock_level ||
                            (level + 1 > write_lock_level &&
                            level + 1 < BTRFS_MAX_LEVEL &&
                            p->nodes[level + 1])) {
                                write_lock_level = level + 1;
                                btrfs_release_path(p);
                                goto again;
                        }

                        btrfs_set_path_blocking(p);
                        if (last_level)
                                err = btrfs_cow_block(trans, root, b, NULL, 0,
                                                      &b);
                        else
                                err = btrfs_cow_block(trans, root, b,
                                                      p->nodes[level + 1],
                                                      p->slots[level + 1], &b);
                        if (err) {
                                ret = err;
                                goto done;
                        }
                }
cow_done:
                p->nodes[level] = b;
                /*
                 * Leave path with blocking locks to avoid massive
                 * lock context switch, this is made on purpose.
                 */

                /*
                 * we have a lock on b and as long as we aren't changing
                 * the tree, there is no way to for the items in b to change.
                 * It is safe to drop the lock on our parent before we
                 * go through the expensive btree search on b.
                 *
                 * If we're inserting or deleting (ins_len != 0), then we might
                 * be changing slot zero, which may require changing the parent.
                 * So, we can't drop the lock until after we know which slot
                 * we're operating on.
                 */
                if (!ins_len && !p->keep_locks) {
                        int u = level + 1;

                        if (u < BTRFS_MAX_LEVEL && p->locks[u]) {
                                btrfs_tree_unlock_rw(p->nodes[u], p->locks[u]);
                                p->locks[u] = 0;
                        }
                }

                ret = key_search(b, key, level, &prev_cmp, &slot);
                if (ret < 0)
                        goto done;

                if (level != 0) {
                        int dec = 0;
                        if (ret && slot > 0) {
                                dec = 1;
                                slot -= 1;
                        }
                        p->slots[level] = slot;
                        err = setup_nodes_for_search(trans, root, p, b, level,
                                             ins_len, &write_lock_level);
                        if (err == -EAGAIN)
                                goto again;
                        if (err) {
                                ret = err;
                                goto done;
                        }
                        b = p->nodes[level];
                        slot = p->slots[level];

                        /*
                         * slot 0 is special, if we change the key
                         * we have to update the parent pointer
                         * which means we must have a write lock
                         * on the parent
                         */
                        if (slot == 0 && ins_len &&
                            write_lock_level < level + 1) {
                                write_lock_level = level + 1;
                                btrfs_release_path(p);
                                goto again;
                        }

                        unlock_up(p, level, lowest_unlock,
                                  min_write_lock_level, &write_lock_level);

                        if (level == lowest_level) {
                                if (dec)
                                        p->slots[level]++;
                                goto done;
                        }

                        err = read_block_for_search(root, p, &b, level,
                                                    slot, key);
                        if (err == -EAGAIN)
                                goto again;
                        if (err) {
                                ret = err;
                                goto done;
                        }

                        if (!p->skip_locking) {
                                level = btrfs_header_level(b);
                                if (level <= write_lock_level) {
                                        err = btrfs_try_tree_write_lock(b);
                                        if (!err) {
                                                btrfs_set_path_blocking(p);
                                                btrfs_tree_lock(b);
                                        }
                                        p->locks[level] = BTRFS_WRITE_LOCK;
                                } else {
                                        err = btrfs_tree_read_lock_atomic(b);
                                        if (!err) {
                                                btrfs_set_path_blocking(p);
                                                btrfs_tree_read_lock(b);
                                        }
                                        p->locks[level] = BTRFS_READ_LOCK;
                                }
                                p->nodes[level] = b;
                        }
                } else {
                        p->slots[level] = slot;
                        if (ins_len > 0 &&
                            btrfs_leaf_free_space(fs_info, b) < ins_len) {
                                if (write_lock_level < 1) {
                                        write_lock_level = 1;
                                        btrfs_release_path(p);
                                        goto again;
                                }

                                btrfs_set_path_blocking(p);
                                err = split_leaf(trans, root, key,
                                                 p, ins_len, ret == 0);

                                BUG_ON(err > 0);
                                if (err) {
                                        ret = err;
                                        goto done;
                                }
                        }
                        if (!p->search_for_split)
                                unlock_up(p, level, lowest_unlock,
                                          min_write_lock_level, NULL);
                        goto done;
                }
        }
        ret = 1;
done:
        /*
         * we don't really know what they plan on doing with the path
         * from here on, so for now just mark it as blocking
         */
        if (!p->leave_spinning)
                btrfs_set_path_blocking(p);
        if (ret < 0 && !p->skip_release_on_error)
                btrfs_release_path(p);
        return ret;
}

/*
 * Like btrfs_search_slot, this looks for a key in the given tree. It uses the
 * current state of the tree together with the operations recorded in the tree
 * modification log to search for the key in a previous version of this tree, as
 * denoted by the time_seq parameter.
 *
 * Naturally, there is no support for insert, delete or cow operations.
 *
 * The resulting path and return value will be set up as if we called
 * btrfs_search_slot at that point in time with ins_len and cow both set to 0.
 */
int btrfs_search_old_slot(struct btrfs_root *root, const struct btrfs_key *key,
                          struct btrfs_path *p, u64 time_seq)
{
        struct btrfs_fs_info *fs_info = root->fs_info;
        struct extent_buffer *b;
        int slot;
        int ret;
        int err;
        int level;
        int lowest_unlock = 1;
        u8 lowest_level = 0;
        int prev_cmp = -1;

        lowest_level = p->lowest_level;
        WARN_ON(p->nodes[0] != NULL);

        if (p->search_commit_root) {
                BUG_ON(time_seq);
                return btrfs_search_slot(NULL, root, key, p, 0, 0);
        }

again:
        b = get_old_root(root, time_seq);
        if (!b) {
                ret = -EIO;
                goto done;
        }
        level = btrfs_header_level(b);
        p->locks[level] = BTRFS_READ_LOCK;

        while (b) {
                level = btrfs_header_level(b);
                p->nodes[level] = b;

                /*
                 * we have a lock on b and as long as we aren't changing
                 * the tree, there is no way to for the items in b to change.
                 * It is safe to drop the lock on our parent before we
                 * go through the expensive btree search on b.
                 */
                btrfs_unlock_up_safe(p, level + 1);