// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
 * All Rights Reserved.
 */
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_shared.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_bit.h"
#include "xfs_mount.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_buf_item.h"
#include "xfs_btree.h"
#include "xfs_errortag.h"
#include "xfs_error.h"
#include "xfs_trace.h"
#include "xfs_alloc.h"
#include "xfs_log.h"
#include "xfs_btree_staging.h"
#include "xfs_ag.h"

/*
 * Cursor allocation zone.
 */
kmem_zone_t     *xfs_btree_cur_zone;

/*
 * Btree magic numbers.
 */
static const uint32_t xfs_magics[2][XFS_BTNUM_MAX] = {
        { XFS_ABTB_MAGIC, XFS_ABTC_MAGIC, 0, XFS_BMAP_MAGIC, XFS_IBT_MAGIC,
          XFS_FIBT_MAGIC, 0 },
        { XFS_ABTB_CRC_MAGIC, XFS_ABTC_CRC_MAGIC, XFS_RMAP_CRC_MAGIC,
          XFS_BMAP_CRC_MAGIC, XFS_IBT_CRC_MAGIC, XFS_FIBT_CRC_MAGIC,
          XFS_REFC_CRC_MAGIC }
};

uint32_t
xfs_btree_magic(
        int                     crc,
        xfs_btnum_t             btnum)
{
        uint32_t                magic = xfs_magics[crc][btnum];

        /* Ensure we asked for crc for crc-only magics. */
        ASSERT(magic != 0);
        return magic;
}

/*
 * Check a long btree block header.  Return the address of the failing check,
 * or NULL if everything is ok.
 */
xfs_failaddr_t
__xfs_btree_check_lblock(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        int                     level,
        struct xfs_buf          *bp)
{
        struct xfs_mount        *mp = cur->bc_mp;
        xfs_btnum_t             btnum = cur->bc_btnum;
        int                     crc = xfs_has_crc(mp);

        if (crc) {
                if (!uuid_equal(&block->bb_u.l.bb_uuid, &mp->m_sb.sb_meta_uuid))
                        return __this_address;
                if (block->bb_u.l.bb_blkno !=
                    cpu_to_be64(bp ? xfs_buf_daddr(bp) : XFS_BUF_DADDR_NULL))
                        return __this_address;
                if (block->bb_u.l.bb_pad != cpu_to_be32(0))
                        return __this_address;
        }

        if (be32_to_cpu(block->bb_magic) != xfs_btree_magic(crc, btnum))
                return __this_address;
        if (be16_to_cpu(block->bb_level) != level)
                return __this_address;
        if (be16_to_cpu(block->bb_numrecs) >
            cur->bc_ops->get_maxrecs(cur, level))
                return __this_address;
        if (block->bb_u.l.bb_leftsib != cpu_to_be64(NULLFSBLOCK) &&
            !xfs_btree_check_lptr(cur, be64_to_cpu(block->bb_u.l.bb_leftsib),
                        level + 1))
                return __this_address;
        if (block->bb_u.l.bb_rightsib != cpu_to_be64(NULLFSBLOCK) &&
            !xfs_btree_check_lptr(cur, be64_to_cpu(block->bb_u.l.bb_rightsib),
                        level + 1))
                return __this_address;

        return NULL;
}

/* Check a long btree block header. */
static int
xfs_btree_check_lblock(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        int                     level,
        struct xfs_buf          *bp)
{
        struct xfs_mount        *mp = cur->bc_mp;
        xfs_failaddr_t          fa;

        fa = __xfs_btree_check_lblock(cur, block, level, bp);
        if (XFS_IS_CORRUPT(mp, fa != NULL) ||
            XFS_TEST_ERROR(false, mp, XFS_ERRTAG_BTREE_CHECK_LBLOCK)) {
                if (bp)
                        trace_xfs_btree_corrupt(bp, _RET_IP_);
                return -EFSCORRUPTED;
        }
        return 0;
}

/*
 * Check a short btree block header.  Return the address of the failing check,
 * or NULL if everything is ok.
 */
xfs_failaddr_t
__xfs_btree_check_sblock(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        int                     level,
        struct xfs_buf          *bp)
{
        struct xfs_mount        *mp = cur->bc_mp;
        xfs_btnum_t             btnum = cur->bc_btnum;
        int                     crc = xfs_has_crc(mp);

        if (crc) {
                if (!uuid_equal(&block->bb_u.s.bb_uuid, &mp->m_sb.sb_meta_uuid))
                        return __this_address;
                if (block->bb_u.s.bb_blkno !=
                    cpu_to_be64(bp ? xfs_buf_daddr(bp) : XFS_BUF_DADDR_NULL))
                        return __this_address;
        }

        if (be32_to_cpu(block->bb_magic) != xfs_btree_magic(crc, btnum))
                return __this_address;
        if (be16_to_cpu(block->bb_level) != level)
                return __this_address;
        if (be16_to_cpu(block->bb_numrecs) >
            cur->bc_ops->get_maxrecs(cur, level))
                return __this_address;
        if (block->bb_u.s.bb_leftsib != cpu_to_be32(NULLAGBLOCK) &&
            !xfs_btree_check_sptr(cur, be32_to_cpu(block->bb_u.s.bb_leftsib),
                        level + 1))
                return __this_address;
        if (block->bb_u.s.bb_rightsib != cpu_to_be32(NULLAGBLOCK) &&
            !xfs_btree_check_sptr(cur, be32_to_cpu(block->bb_u.s.bb_rightsib),
                        level + 1))
                return __this_address;

        return NULL;
}

/* Check a short btree block header. */
STATIC int
xfs_btree_check_sblock(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        int                     level,
        struct xfs_buf          *bp)
{
        struct xfs_mount        *mp = cur->bc_mp;
        xfs_failaddr_t          fa;

        fa = __xfs_btree_check_sblock(cur, block, level, bp);
        if (XFS_IS_CORRUPT(mp, fa != NULL) ||
            XFS_TEST_ERROR(false, mp, XFS_ERRTAG_BTREE_CHECK_SBLOCK)) {
                if (bp)
                        trace_xfs_btree_corrupt(bp, _RET_IP_);
                return -EFSCORRUPTED;
        }
        return 0;
}

/*
 * Debug routine: check that block header is ok.
 */
int
xfs_btree_check_block(
        struct xfs_btree_cur    *cur,   /* btree cursor */
        struct xfs_btree_block  *block, /* generic btree block pointer */
        int                     level,  /* level of the btree block */
        struct xfs_buf          *bp)    /* buffer containing block, if any */
{
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
                return xfs_btree_check_lblock(cur, block, level, bp);
        else
                return xfs_btree_check_sblock(cur, block, level, bp);
}

/* Check that this long pointer is valid and points within the fs. */
bool
xfs_btree_check_lptr(
        struct xfs_btree_cur    *cur,
        xfs_fsblock_t           fsbno,
        int                     level)
{
        if (level <= 0)
                return false;
        return xfs_verify_fsbno(cur->bc_mp, fsbno);
}

/* Check that this short pointer is valid and points within the AG. */
bool
xfs_btree_check_sptr(
        struct xfs_btree_cur    *cur,
        xfs_agblock_t           agbno,
        int                     level)
{
        if (level <= 0)
                return false;
        return xfs_verify_agbno(cur->bc_mp, cur->bc_ag.pag->pag_agno, agbno);
}

/*
 * Check that a given (indexed) btree pointer at a certain level of a
 * btree is valid and doesn't point past where it should.
 */
static int
xfs_btree_check_ptr(
        struct xfs_btree_cur            *cur,
        const union xfs_btree_ptr       *ptr,
        int                             index,
        int                             level)
{
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
                if (xfs_btree_check_lptr(cur, be64_to_cpu((&ptr->l)[index]),
                                level))
                        return 0;
                xfs_err(cur->bc_mp,
"Inode %llu fork %d: Corrupt btree %d pointer at level %d index %d.",
                                cur->bc_ino.ip->i_ino,
                                cur->bc_ino.whichfork, cur->bc_btnum,
                                level, index);
        } else {
                if (xfs_btree_check_sptr(cur, be32_to_cpu((&ptr->s)[index]),
                                level))
                        return 0;
                xfs_err(cur->bc_mp,
"AG %u: Corrupt btree %d pointer at level %d index %d.",
                                cur->bc_ag.pag->pag_agno, cur->bc_btnum,
                                level, index);
        }

        return -EFSCORRUPTED;
}

#ifdef DEBUG
# define xfs_btree_debug_check_ptr      xfs_btree_check_ptr
#else
# define xfs_btree_debug_check_ptr(...) (0)
#endif

/*
 * Calculate CRC on the whole btree block and stuff it into the
 * long-form btree header.
 *
 * Prior to calculting the CRC, pull the LSN out of the buffer log item and put
 * it into the buffer so recovery knows what the last modification was that made
 * it to disk.
 */
void
xfs_btree_lblock_calc_crc(
        struct xfs_buf          *bp)
{
        struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
        struct xfs_buf_log_item *bip = bp->b_log_item;

        if (!xfs_has_crc(bp->b_mount))
                return;
        if (bip)
                block->bb_u.l.bb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
        xfs_buf_update_cksum(bp, XFS_BTREE_LBLOCK_CRC_OFF);
}

bool
xfs_btree_lblock_verify_crc(
        struct xfs_buf          *bp)
{
        struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
        struct xfs_mount        *mp = bp->b_mount;

        if (xfs_has_crc(mp)) {
                if (!xfs_log_check_lsn(mp, be64_to_cpu(block->bb_u.l.bb_lsn)))
                        return false;
                return xfs_buf_verify_cksum(bp, XFS_BTREE_LBLOCK_CRC_OFF);
        }

        return true;
}

/*
 * Calculate CRC on the whole btree block and stuff it into the
 * short-form btree header.
 *
 * Prior to calculting the CRC, pull the LSN out of the buffer log item and put
 * it into the buffer so recovery knows what the last modification was that made
 * it to disk.
 */
void
xfs_btree_sblock_calc_crc(
        struct xfs_buf          *bp)
{
        struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
        struct xfs_buf_log_item *bip = bp->b_log_item;

        if (!xfs_has_crc(bp->b_mount))
                return;
        if (bip)
                block->bb_u.s.bb_lsn = cpu_to_be64(bip->bli_item.li_lsn);
        xfs_buf_update_cksum(bp, XFS_BTREE_SBLOCK_CRC_OFF);
}

bool
xfs_btree_sblock_verify_crc(
        struct xfs_buf          *bp)
{
        struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
        struct xfs_mount        *mp = bp->b_mount;

        if (xfs_has_crc(mp)) {
                if (!xfs_log_check_lsn(mp, be64_to_cpu(block->bb_u.s.bb_lsn)))
                        return false;
                return xfs_buf_verify_cksum(bp, XFS_BTREE_SBLOCK_CRC_OFF);
        }

        return true;
}

static int
xfs_btree_free_block(
        struct xfs_btree_cur    *cur,
        struct xfs_buf          *bp)
{
        int                     error;

        error = cur->bc_ops->free_block(cur, bp);
        if (!error) {
                xfs_trans_binval(cur->bc_tp, bp);
                XFS_BTREE_STATS_INC(cur, free);
        }
        return error;
}

/*
 * Delete the btree cursor.
 */
void
xfs_btree_del_cursor(
        struct xfs_btree_cur    *cur,           /* btree cursor */
        int                     error)          /* del because of error */
{
        int                     i;              /* btree level */

        /*
         * Clear the buffer pointers and release the buffers. If we're doing
         * this because of an error, inspect all of the entries in the bc_bufs
         * array for buffers to be unlocked. This is because some of the btree
         * code works from level n down to 0, and if we get an error along the
         * way we won't have initialized all the entries down to 0.
         */
        for (i = 0; i < cur->bc_nlevels; i++) {
                if (cur->bc_bufs[i])
                        xfs_trans_brelse(cur->bc_tp, cur->bc_bufs[i]);
                else if (!error)
                        break;
        }

        ASSERT(cur->bc_btnum != XFS_BTNUM_BMAP || cur->bc_ino.allocated == 0 ||
               xfs_is_shutdown(cur->bc_mp));
        if (unlikely(cur->bc_flags & XFS_BTREE_STAGING))
                kmem_free(cur->bc_ops);
        if (!(cur->bc_flags & XFS_BTREE_LONG_PTRS) && cur->bc_ag.pag)
                xfs_perag_put(cur->bc_ag.pag);
        kmem_cache_free(xfs_btree_cur_zone, cur);
}

/*
 * Duplicate the btree cursor.
 * Allocate a new one, copy the record, re-get the buffers.
 */
int                                     /* error */
xfs_btree_dup_cursor(
        xfs_btree_cur_t *cur,           /* input cursor */
        xfs_btree_cur_t **ncur)         /* output cursor */
{
        struct xfs_buf  *bp;            /* btree block's buffer pointer */
        int             error;          /* error return value */
        int             i;              /* level number of btree block */
        xfs_mount_t     *mp;            /* mount structure for filesystem */
        xfs_btree_cur_t *new;           /* new cursor value */
        xfs_trans_t     *tp;            /* transaction pointer, can be NULL */

        tp = cur->bc_tp;
        mp = cur->bc_mp;

        /*
         * Allocate a new cursor like the old one.
         */
        new = cur->bc_ops->dup_cursor(cur);

        /*
         * Copy the record currently in the cursor.
         */
        new->bc_rec = cur->bc_rec;

        /*
         * For each level current, re-get the buffer and copy the ptr value.
         */
        for (i = 0; i < new->bc_nlevels; i++) {
                new->bc_ptrs[i] = cur->bc_ptrs[i];
                new->bc_ra[i] = cur->bc_ra[i];
                bp = cur->bc_bufs[i];
                if (bp) {
                        error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
                                                   xfs_buf_daddr(bp), mp->m_bsize,
                                                   0, &bp,
                                                   cur->bc_ops->buf_ops);
                        if (error) {
                                xfs_btree_del_cursor(new, error);
                                *ncur = NULL;
                                return error;
                        }
                }
                new->bc_bufs[i] = bp;
        }
        *ncur = new;
        return 0;
}

/*
 * XFS btree block layout and addressing:
 *
 * There are two types of blocks in the btree: leaf and non-leaf blocks.
 *
 * The leaf record start with a header then followed by records containing
 * the values.  A non-leaf block also starts with the same header, and
 * then first contains lookup keys followed by an equal number of pointers
 * to the btree blocks at the previous level.
 *
 *              +--------+-------+-------+-------+-------+-------+-------+
 * Leaf:        | header | rec 1 | rec 2 | rec 3 | rec 4 | rec 5 | rec N |
 *              +--------+-------+-------+-------+-------+-------+-------+
 *
 *              +--------+-------+-------+-------+-------+-------+-------+
 * Non-Leaf:    | header | key 1 | key 2 | key N | ptr 1 | ptr 2 | ptr N |
 *              +--------+-------+-------+-------+-------+-------+-------+
 *
 * The header is called struct xfs_btree_block for reasons better left unknown
 * and comes in different versions for short (32bit) and long (64bit) block
 * pointers.  The record and key structures are defined by the btree instances
 * and opaque to the btree core.  The block pointers are simple disk endian
 * integers, available in a short (32bit) and long (64bit) variant.
 *
 * The helpers below calculate the offset of a given record, key or pointer
 * into a btree block (xfs_btree_*_offset) or return a pointer to the given
 * record, key or pointer (xfs_btree_*_addr).  Note that all addressing
 * inside the btree block is done using indices starting at one, not zero!
 *
 * If XFS_BTREE_OVERLAPPING is set, then this btree supports keys containing
 * overlapping intervals.  In such a tree, records are still sorted lowest to
 * highest and indexed by the smallest key value that refers to the record.
 * However, nodes are different: each pointer has two associated keys -- one
 * indexing the lowest key available in the block(s) below (the same behavior
 * as the key in a regular btree) and another indexing the highest key
 * available in the block(s) below.  Because records are /not/ sorted by the
 * highest key, all leaf block updates require us to compute the highest key
 * that matches any record in the leaf and to recursively update the high keys
 * in the nodes going further up in the tree, if necessary.  Nodes look like
 * this:
 *
 *              +--------+-----+-----+-----+-----+-----+-------+-------+-----+
 * Non-Leaf:    | header | lo1 | hi1 | lo2 | hi2 | ... | ptr 1 | ptr 2 | ... |
 *              +--------+-----+-----+-----+-----+-----+-------+-------+-----+
 *
 * To perform an interval query on an overlapped tree, perform the usual
 * depth-first search and use the low and high keys to decide if we can skip
 * that particular node.  If a leaf node is reached, return the records that
 * intersect the interval.  Note that an interval query may return numerous
 * entries.  For a non-overlapped tree, simply search for the record associated
 * with the lowest key and iterate forward until a non-matching record is
 * found.  Section 14.3 ("Interval Trees") of _Introduction to Algorithms_ by
 * Cormen, Leiserson, Rivest, and Stein (2nd or 3rd ed. only) discuss this in
 * more detail.
 *
 * Why do we care about overlapping intervals?  Let's say you have a bunch of
 * reverse mapping records on a reflink filesystem:
 *
 * 1: +- file A startblock B offset C length D -----------+
 * 2:      +- file E startblock F offset G length H --------------+
 * 3:      +- file I startblock F offset J length K --+
 * 4:                                                        +- file L... --+
 *
 * Now say we want to map block (B+D) into file A at offset (C+D).  Ideally,
 * we'd simply increment the length of record 1.  But how do we find the record
 * that ends at (B+D-1) (i.e. record 1)?  A LE lookup of (B+D-1) would return
 * record 3 because the keys are ordered first by startblock.  An interval
 * query would return records 1 and 2 because they both overlap (B+D-1), and
 * from that we can pick out record 1 as the appropriate left neighbor.
 *
 * In the non-overlapped case you can do a LE lookup and decrement the cursor
 * because a record's interval must end before the next record.
 */

/*
 * Return size of the btree block header for this btree instance.
 */
static inline size_t xfs_btree_block_len(struct xfs_btree_cur *cur)
{
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
                if (cur->bc_flags & XFS_BTREE_CRC_BLOCKS)
                        return XFS_BTREE_LBLOCK_CRC_LEN;
                return XFS_BTREE_LBLOCK_LEN;
        }
        if (cur->bc_flags & XFS_BTREE_CRC_BLOCKS)
                return XFS_BTREE_SBLOCK_CRC_LEN;
        return XFS_BTREE_SBLOCK_LEN;
}

/*
 * Return size of btree block pointers for this btree instance.
 */
static inline size_t xfs_btree_ptr_len(struct xfs_btree_cur *cur)
{
        return (cur->bc_flags & XFS_BTREE_LONG_PTRS) ?
                sizeof(__be64) : sizeof(__be32);
}

/*
 * Calculate offset of the n-th record in a btree block.
 */
STATIC size_t
xfs_btree_rec_offset(
        struct xfs_btree_cur    *cur,
        int                     n)
{
        return xfs_btree_block_len(cur) +
                (n - 1) * cur->bc_ops->rec_len;
}

/*
 * Calculate offset of the n-th key in a btree block.
 */
STATIC size_t
xfs_btree_key_offset(
        struct xfs_btree_cur    *cur,
        int                     n)
{
        return xfs_btree_block_len(cur) +
                (n - 1) * cur->bc_ops->key_len;
}

/*
 * Calculate offset of the n-th high key in a btree block.
 */
STATIC size_t
xfs_btree_high_key_offset(
        struct xfs_btree_cur    *cur,
        int                     n)
{
        return xfs_btree_block_len(cur) +
                (n - 1) * cur->bc_ops->key_len + (cur->bc_ops->key_len / 2);
}

/*
 * Calculate offset of the n-th block pointer in a btree block.
 */
STATIC size_t
xfs_btree_ptr_offset(
        struct xfs_btree_cur    *cur,
        int                     n,
        int                     level)
{
        return xfs_btree_block_len(cur) +
                cur->bc_ops->get_maxrecs(cur, level) * cur->bc_ops->key_len +
                (n - 1) * xfs_btree_ptr_len(cur);
}

/*
 * Return a pointer to the n-th record in the btree block.
 */
union xfs_btree_rec *
xfs_btree_rec_addr(
        struct xfs_btree_cur    *cur,
        int                     n,
        struct xfs_btree_block  *block)
{
        return (union xfs_btree_rec *)
                ((char *)block + xfs_btree_rec_offset(cur, n));
}

/*
 * Return a pointer to the n-th key in the btree block.
 */
union xfs_btree_key *
xfs_btree_key_addr(
        struct xfs_btree_cur    *cur,
        int                     n,
        struct xfs_btree_block  *block)
{
        return (union xfs_btree_key *)
                ((char *)block + xfs_btree_key_offset(cur, n));
}

/*
 * Return a pointer to the n-th high key in the btree block.
 */
union xfs_btree_key *
xfs_btree_high_key_addr(
        struct xfs_btree_cur    *cur,
        int                     n,
        struct xfs_btree_block  *block)
{
        return (union xfs_btree_key *)
                ((char *)block + xfs_btree_high_key_offset(cur, n));
}

/*
 * Return a pointer to the n-th block pointer in the btree block.
 */
union xfs_btree_ptr *
xfs_btree_ptr_addr(
        struct xfs_btree_cur    *cur,
        int                     n,
        struct xfs_btree_block  *block)
{
        int                     level = xfs_btree_get_level(block);

        ASSERT(block->bb_level != 0);

        return (union xfs_btree_ptr *)
                ((char *)block + xfs_btree_ptr_offset(cur, n, level));
}

struct xfs_ifork *
xfs_btree_ifork_ptr(
        struct xfs_btree_cur    *cur)
{
        ASSERT(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE);

        if (cur->bc_flags & XFS_BTREE_STAGING)
                return cur->bc_ino.ifake->if_fork;
        return XFS_IFORK_PTR(cur->bc_ino.ip, cur->bc_ino.whichfork);
}

/*
 * Get the root block which is stored in the inode.
 *
 * For now this btree implementation assumes the btree root is always
 * stored in the if_broot field of an inode fork.
 */
STATIC struct xfs_btree_block *
xfs_btree_get_iroot(
        struct xfs_btree_cur    *cur)
{
        struct xfs_ifork        *ifp = xfs_btree_ifork_ptr(cur);

        return (struct xfs_btree_block *)ifp->if_broot;
}

/*
 * Retrieve the block pointer from the cursor at the given level.
 * This may be an inode btree root or from a buffer.
 */
struct xfs_btree_block *                /* generic btree block pointer */
xfs_btree_get_block(
        struct xfs_btree_cur    *cur,   /* btree cursor */
        int                     level,  /* level in btree */
        struct xfs_buf          **bpp)  /* buffer containing the block */
{
        if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
            (level == cur->bc_nlevels - 1)) {
                *bpp = NULL;
                return xfs_btree_get_iroot(cur);
        }

        *bpp = cur->bc_bufs[level];
        return XFS_BUF_TO_BLOCK(*bpp);
}

/*
 * Change the cursor to point to the first record at the given level.
 * Other levels are unaffected.
 */
STATIC int                              /* success=1, failure=0 */
xfs_btree_firstrec(
        xfs_btree_cur_t         *cur,   /* btree cursor */
        int                     level)  /* level to change */
{
        struct xfs_btree_block  *block; /* generic btree block pointer */
        struct xfs_buf          *bp;    /* buffer containing block */

        /*
         * Get the block pointer for this level.
         */
        block = xfs_btree_get_block(cur, level, &bp);
        if (xfs_btree_check_block(cur, block, level, bp))
                return 0;
        /*
         * It's empty, there is no such record.
         */
        if (!block->bb_numrecs)
                return 0;
        /*
         * Set the ptr value to 1, that's the first record/key.
         */
        cur->bc_ptrs[level] = 1;
        return 1;
}

/*
 * Change the cursor to point to the last record in the current block
 * at the given level.  Other levels are unaffected.
 */
STATIC int                              /* success=1, failure=0 */
xfs_btree_lastrec(
        xfs_btree_cur_t         *cur,   /* btree cursor */
        int                     level)  /* level to change */
{
        struct xfs_btree_block  *block; /* generic btree block pointer */
        struct xfs_buf          *bp;    /* buffer containing block */

        /*
         * Get the block pointer for this level.
         */
        block = xfs_btree_get_block(cur, level, &bp);
        if (xfs_btree_check_block(cur, block, level, bp))
                return 0;
        /*
         * It's empty, there is no such record.
         */
        if (!block->bb_numrecs)
                return 0;
        /*
         * Set the ptr value to numrecs, that's the last record/key.
         */
        cur->bc_ptrs[level] = be16_to_cpu(block->bb_numrecs);
        return 1;
}

/*
 * Compute first and last byte offsets for the fields given.
 * Interprets the offsets table, which contains struct field offsets.
 */
void
xfs_btree_offsets(
        int64_t         fields,         /* bitmask of fields */
        const short     *offsets,       /* table of field offsets */
        int             nbits,          /* number of bits to inspect */
        int             *first,         /* output: first byte offset */
        int             *last)          /* output: last byte offset */
{
        int             i;              /* current bit number */
        int64_t         imask;          /* mask for current bit number */

        ASSERT(fields != 0);
        /*
         * Find the lowest bit, so the first byte offset.
         */
        for (i = 0, imask = 1LL; ; i++, imask <<= 1) {
                if (imask & fields) {
                        *first = offsets[i];
                        break;
                }
        }
        /*
         * Find the highest bit, so the last byte offset.
         */
        for (i = nbits - 1, imask = 1LL << i; ; i--, imask >>= 1) {
                if (imask & fields) {
                        *last = offsets[i + 1] - 1;
                        break;
                }
        }
}

/*
 * Get a buffer for the block, return it read in.
 * Long-form addressing.
 */
int
xfs_btree_read_bufl(
        struct xfs_mount        *mp,            /* file system mount point */
        struct xfs_trans        *tp,            /* transaction pointer */
        xfs_fsblock_t           fsbno,          /* file system block number */
        struct xfs_buf          **bpp,          /* buffer for fsbno */
        int                     refval,         /* ref count value for buffer */
        const struct xfs_buf_ops *ops)
{
        struct xfs_buf          *bp;            /* return value */
        xfs_daddr_t             d;              /* real disk block address */
        int                     error;

        if (!xfs_verify_fsbno(mp, fsbno))
                return -EFSCORRUPTED;
        d = XFS_FSB_TO_DADDR(mp, fsbno);
        error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
                                   mp->m_bsize, 0, &bp, ops);
        if (error)
                return error;
        if (bp)
                xfs_buf_set_ref(bp, refval);
        *bpp = bp;
        return 0;
}

/*
 * Read-ahead the block, don't wait for it, don't return a buffer.
 * Long-form addressing.
 */
/* ARGSUSED */
void
xfs_btree_reada_bufl(
        struct xfs_mount        *mp,            /* file system mount point */
        xfs_fsblock_t           fsbno,          /* file system block number */
        xfs_extlen_t            count,          /* count of filesystem blocks */
        const struct xfs_buf_ops *ops)
{
        xfs_daddr_t             d;

        ASSERT(fsbno != NULLFSBLOCK);
        d = XFS_FSB_TO_DADDR(mp, fsbno);
        xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, ops);
}

/*
 * Read-ahead the block, don't wait for it, don't return a buffer.
 * Short-form addressing.
 */
/* ARGSUSED */
void
xfs_btree_reada_bufs(
        struct xfs_mount        *mp,            /* file system mount point */
        xfs_agnumber_t          agno,           /* allocation group number */
        xfs_agblock_t           agbno,          /* allocation group block number */
        xfs_extlen_t            count,          /* count of filesystem blocks */
        const struct xfs_buf_ops *ops)
{
        xfs_daddr_t             d;

        ASSERT(agno != NULLAGNUMBER);
        ASSERT(agbno != NULLAGBLOCK);
        d = XFS_AGB_TO_DADDR(mp, agno, agbno);
        xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, ops);
}

STATIC int
xfs_btree_readahead_lblock(
        struct xfs_btree_cur    *cur,
        int                     lr,
        struct xfs_btree_block  *block)
{
        int                     rval = 0;
        xfs_fsblock_t           left = be64_to_cpu(block->bb_u.l.bb_leftsib);
        xfs_fsblock_t           right = be64_to_cpu(block->bb_u.l.bb_rightsib);

        if ((lr & XFS_BTCUR_LEFTRA) && left != NULLFSBLOCK) {
                xfs_btree_reada_bufl(cur->bc_mp, left, 1,
                                     cur->bc_ops->buf_ops);
                rval++;
        }

        if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLFSBLOCK) {
                xfs_btree_reada_bufl(cur->bc_mp, right, 1,
                                     cur->bc_ops->buf_ops);
                rval++;
        }

        return rval;
}

STATIC int
xfs_btree_readahead_sblock(
        struct xfs_btree_cur    *cur,
        int                     lr,
        struct xfs_btree_block *block)
{
        int                     rval = 0;
        xfs_agblock_t           left = be32_to_cpu(block->bb_u.s.bb_leftsib);
        xfs_agblock_t           right = be32_to_cpu(block->bb_u.s.bb_rightsib);


        if ((lr & XFS_BTCUR_LEFTRA) && left != NULLAGBLOCK) {
                xfs_btree_reada_bufs(cur->bc_mp, cur->bc_ag.pag->pag_agno,
                                     left, 1, cur->bc_ops->buf_ops);
                rval++;
        }

        if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLAGBLOCK) {
                xfs_btree_reada_bufs(cur->bc_mp, cur->bc_ag.pag->pag_agno,
                                     right, 1, cur->bc_ops->buf_ops);
                rval++;
        }

        return rval;
}

/*
 * Read-ahead btree blocks, at the given level.
 * Bits in lr are set from XFS_BTCUR_{LEFT,RIGHT}RA.
 */
STATIC int
xfs_btree_readahead(
        struct xfs_btree_cur    *cur,           /* btree cursor */
        int                     lev,            /* level in btree */
        int                     lr)             /* left/right bits */
{
        struct xfs_btree_block  *block;

        /*
         * No readahead needed if we are at the root level and the
         * btree root is stored in the inode.
         */
        if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
            (lev == cur->bc_nlevels - 1))
                return 0;

        if ((cur->bc_ra[lev] | lr) == cur->bc_ra[lev])
                return 0;

        cur->bc_ra[lev] |= lr;
        block = XFS_BUF_TO_BLOCK(cur->bc_bufs[lev]);

        if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
                return xfs_btree_readahead_lblock(cur, lr, block);
        return xfs_btree_readahead_sblock(cur, lr, block);
}

STATIC int
xfs_btree_ptr_to_daddr(
        struct xfs_btree_cur            *cur,
        const union xfs_btree_ptr       *ptr,
        xfs_daddr_t                     *daddr)
{
        xfs_fsblock_t           fsbno;
        xfs_agblock_t           agbno;
        int                     error;

        error = xfs_btree_check_ptr(cur, ptr, 0, 1);
        if (error)
                return error;

        if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
                fsbno = be64_to_cpu(ptr->l);
                *daddr = XFS_FSB_TO_DADDR(cur->bc_mp, fsbno);
        } else {
                agbno = be32_to_cpu(ptr->s);
                *daddr = XFS_AGB_TO_DADDR(cur->bc_mp, cur->bc_ag.pag->pag_agno,
                                agbno);
        }

        return 0;
}

/*
 * Readahead @count btree blocks at the given @ptr location.
 *
 * We don't need to care about long or short form btrees here as we have a
 * method of converting the ptr directly to a daddr available to us.
 */
STATIC void
xfs_btree_readahead_ptr(
        struct xfs_btree_cur    *cur,
        union xfs_btree_ptr     *ptr,
        xfs_extlen_t            count)
{
        xfs_daddr_t             daddr;

        if (xfs_btree_ptr_to_daddr(cur, ptr, &daddr))
                return;
        xfs_buf_readahead(cur->bc_mp->m_ddev_targp, daddr,
                          cur->bc_mp->m_bsize * count, cur->bc_ops->buf_ops);
}

/*
 * Set the buffer for level "lev" in the cursor to bp, releasing
 * any previous buffer.
 */
STATIC void
xfs_btree_setbuf(
        xfs_btree_cur_t         *cur,   /* btree cursor */
        int                     lev,    /* level in btree */
        struct xfs_buf          *bp)    /* new buffer to set */
{
        struct xfs_btree_block  *b;     /* btree block */

        if (cur->bc_bufs[lev])
                xfs_trans_brelse(cur->bc_tp, cur->bc_bufs[lev]);
        cur->bc_bufs[lev] = bp;
        cur->bc_ra[lev] = 0;

        b = XFS_BUF_TO_BLOCK(bp);
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {

                if (b->bb_u.l.bb_leftsib == cpu_to_be64(NULLFSBLOCK))
                        cur->bc_ra[lev] |= XFS_BTCUR_LEFTRA;
                if (b->bb_u.l.bb_rightsib == cpu_to_be64(NULLFSBLOCK))
                        cur->bc_ra[lev] |= XFS_BTCUR_RIGHTRA;
        } else {
                if (b->bb_u.s.bb_leftsib == cpu_to_be32(NULLAGBLOCK))
                        cur->bc_ra[lev] |= XFS_BTCUR_LEFTRA;
                if (b->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK))
                        cur->bc_ra[lev] |= XFS_BTCUR_RIGHTRA;
        }
}

bool
xfs_btree_ptr_is_null(
        struct xfs_btree_cur            *cur,
        const union xfs_btree_ptr       *ptr)
{
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
                return ptr->l == cpu_to_be64(NULLFSBLOCK);
        else
                return ptr->s == cpu_to_be32(NULLAGBLOCK);
}

void
xfs_btree_set_ptr_null(
        struct xfs_btree_cur    *cur,
        union xfs_btree_ptr     *ptr)
{
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
                ptr->l = cpu_to_be64(NULLFSBLOCK);
        else
                ptr->s = cpu_to_be32(NULLAGBLOCK);
}

/*
 * Get/set/init sibling pointers
 */
void
xfs_btree_get_sibling(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        union xfs_btree_ptr     *ptr,
        int                     lr)
{
        ASSERT(lr == XFS_BB_LEFTSIB || lr == XFS_BB_RIGHTSIB);

        if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
                if (lr == XFS_BB_RIGHTSIB)
                        ptr->l = block->bb_u.l.bb_rightsib;
                else
                        ptr->l = block->bb_u.l.bb_leftsib;
        } else {
                if (lr == XFS_BB_RIGHTSIB)
                        ptr->s = block->bb_u.s.bb_rightsib;
                else
                        ptr->s = block->bb_u.s.bb_leftsib;
        }
}

void
xfs_btree_set_sibling(
        struct xfs_btree_cur            *cur,
        struct xfs_btree_block          *block,
        const union xfs_btree_ptr       *ptr,
        int                             lr)
{
        ASSERT(lr == XFS_BB_LEFTSIB || lr == XFS_BB_RIGHTSIB);

        if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
                if (lr == XFS_BB_RIGHTSIB)
                        block->bb_u.l.bb_rightsib = ptr->l;
                else
                        block->bb_u.l.bb_leftsib = ptr->l;
        } else {
                if (lr == XFS_BB_RIGHTSIB)
                        block->bb_u.s.bb_rightsib = ptr->s;
                else
                        block->bb_u.s.bb_leftsib = ptr->s;
        }
}

void
xfs_btree_init_block_int(
        struct xfs_mount        *mp,
        struct xfs_btree_block  *buf,
        xfs_daddr_t             blkno,
        xfs_btnum_t             btnum,
        __u16                   level,
        __u16                   numrecs,
        __u64                   owner,
        unsigned int            flags)
{
        int                     crc = xfs_has_crc(mp);
        __u32                   magic = xfs_btree_magic(crc, btnum);

        buf->bb_magic = cpu_to_be32(magic);
        buf->bb_level = cpu_to_be16(level);
        buf->bb_numrecs = cpu_to_be16(numrecs);

        if (flags & XFS_BTREE_LONG_PTRS) {
                buf->bb_u.l.bb_leftsib = cpu_to_be64(NULLFSBLOCK);
                buf->bb_u.l.bb_rightsib = cpu_to_be64(NULLFSBLOCK);
                if (crc) {
                        buf->bb_u.l.bb_blkno = cpu_to_be64(blkno);
                        buf->bb_u.l.bb_owner = cpu_to_be64(owner);
                        uuid_copy(&buf->bb_u.l.bb_uuid, &mp->m_sb.sb_meta_uuid);
                        buf->bb_u.l.bb_pad = 0;
                        buf->bb_u.l.bb_lsn = 0;
                }
        } else {
                /* owner is a 32 bit value on short blocks */
                __u32 __owner = (__u32)owner;

                buf->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
                buf->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
                if (crc) {
                        buf->bb_u.s.bb_blkno = cpu_to_be64(blkno);
                        buf->bb_u.s.bb_owner = cpu_to_be32(__owner);
                        uuid_copy(&buf->bb_u.s.bb_uuid, &mp->m_sb.sb_meta_uuid);
                        buf->bb_u.s.bb_lsn = 0;
                }
        }
}

void
xfs_btree_init_block(
        struct xfs_mount *mp,
        struct xfs_buf  *bp,
        xfs_btnum_t     btnum,
        __u16           level,
        __u16           numrecs,
        __u64           owner)
{
        xfs_btree_init_block_int(mp, XFS_BUF_TO_BLOCK(bp), xfs_buf_daddr(bp),
                                 btnum, level, numrecs, owner, 0);
}

void
xfs_btree_init_block_cur(
        struct xfs_btree_cur    *cur,
        struct xfs_buf          *bp,
        int                     level,
        int                     numrecs)
{
        __u64                   owner;

        /*
         * we can pull the owner from the cursor right now as the different
         * owners align directly with the pointer size of the btree. This may
         * change in future, but is safe for current users of the generic btree
         * code.
         */
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
                owner = cur->bc_ino.ip->i_ino;
        else
                owner = cur->bc_ag.pag->pag_agno;

        xfs_btree_init_block_int(cur->bc_mp, XFS_BUF_TO_BLOCK(bp),
                                xfs_buf_daddr(bp), cur->bc_btnum, level,
                                numrecs, owner, cur->bc_flags);
}

/*
 * Return true if ptr is the last record in the btree and
 * we need to track updates to this record.  The decision
 * will be further refined in the update_lastrec method.
 */
STATIC int
xfs_btree_is_lastrec(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        int                     level)
{
        union xfs_btree_ptr     ptr;

        if (level > 0)
                return 0;
        if (!(cur->bc_flags & XFS_BTREE_LASTREC_UPDATE))
                return 0;

        xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_RIGHTSIB);
        if (!xfs_btree_ptr_is_null(cur, &ptr))
                return 0;
        return 1;
}

STATIC void
xfs_btree_buf_to_ptr(
        struct xfs_btree_cur    *cur,
        struct xfs_buf          *bp,
        union xfs_btree_ptr     *ptr)
{
        if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
                ptr->l = cpu_to_be64(XFS_DADDR_TO_FSB(cur->bc_mp,
                                        xfs_buf_daddr(bp)));
        else {
                ptr->s = cpu_to_be32(xfs_daddr_to_agbno(cur->bc_mp,
                                        xfs_buf_daddr(bp)));
        }
}

STATIC void
xfs_btree_set_refs(
        struct xfs_btree_cur    *cur,
        struct xfs_buf          *bp)
{
        switch (cur->bc_btnum) {
        case XFS_BTNUM_BNO:
        case XFS_BTNUM_CNT:
                xfs_buf_set_ref(bp, XFS_ALLOC_BTREE_REF);
                break;
        case XFS_BTNUM_INO:
        case XFS_BTNUM_FINO:
                xfs_buf_set_ref(bp, XFS_INO_BTREE_REF);
                break;
        case XFS_BTNUM_BMAP:
                xfs_buf_set_ref(bp, XFS_BMAP_BTREE_REF);
                break;
        case XFS_BTNUM_RMAP:
                xfs_buf_set_ref(bp, XFS_RMAP_BTREE_REF);
                break;
        case XFS_BTNUM_REFC:
                xfs_buf_set_ref(bp, XFS_REFC_BTREE_REF);
                break;
        default:
                ASSERT(0);
        }
}

int
xfs_btree_get_buf_block(
        struct xfs_btree_cur            *cur,
        const union xfs_btree_ptr       *ptr,
        struct xfs_btree_block          **block,
        struct xfs_buf                  **bpp)
{
        struct xfs_mount        *mp = cur->bc_mp;
        xfs_daddr_t             d;
        int                     error;

        error = xfs_btree_ptr_to_daddr(cur, ptr, &d);
        if (error)
                return error;
        error = xfs_trans_get_buf(cur->bc_tp, mp->m_ddev_targp, d, mp->m_bsize,
                        0, bpp);
        if (error)
                return error;

        (*bpp)->b_ops = cur->bc_ops->buf_ops;
        *block = XFS_BUF_TO_BLOCK(*bpp);
        return 0;
}

/*
 * Read in the buffer at the given ptr and return the buffer and
 * the block pointer within the buffer.
 */
STATIC int
xfs_btree_read_buf_block(
        struct xfs_btree_cur            *cur,
        const union xfs_btree_ptr       *ptr,
        int                             flags,
        struct xfs_btree_block          **block,
        struct xfs_buf                  **bpp)
{
        struct xfs_mount        *mp = cur->bc_mp;
        xfs_daddr_t             d;
        int                     error;

        /* need to sort out how callers deal with failures first */
        ASSERT(!(flags & XBF_TRYLOCK));

        error = xfs_btree_ptr_to_daddr(cur, ptr, &d);
        if (error)
                return error;
        error = xfs_trans_read_buf(mp, cur->bc_tp, mp->m_ddev_targp, d,
                                   mp->m_bsize, flags, bpp,
                                   cur->bc_ops->buf_ops);
        if (error)
                return error;

        xfs_btree_set_refs(cur, *bpp);
        *block = XFS_BUF_TO_BLOCK(*bpp);
        return 0;
}

/*
 * Copy keys from one btree block to another.
 */
void
xfs_btree_copy_keys(
        struct xfs_btree_cur            *cur,
        union xfs_btree_key             *dst_key,
        const union xfs_btree_key       *src_key,
        int                             numkeys)
{
        ASSERT(numkeys >= 0);
        memcpy(dst_key, src_key, numkeys * cur->bc_ops->key_len);
}

/*
 * Copy records from one btree block to another.
 */
STATIC void
xfs_btree_copy_recs(
        struct xfs_btree_cur    *cur,
        union xfs_btree_rec     *dst_rec,
        union xfs_btree_rec     *src_rec,
        int                     numrecs)
{
        ASSERT(numrecs >= 0);
        memcpy(dst_rec, src_rec, numrecs * cur->bc_ops->rec_len);
}

/*
 * Copy block pointers from one btree block to another.
 */
void
xfs_btree_copy_ptrs(
        struct xfs_btree_cur    *cur,
        union xfs_btree_ptr     *dst_ptr,
        const union xfs_btree_ptr *src_ptr,
        int                     numptrs)
{
        ASSERT(numptrs >= 0);
        memcpy(dst_ptr, src_ptr, numptrs * xfs_btree_ptr_len(cur));
}

/*
 * Shift keys one index left/right inside a single btree block.
 */
STATIC void
xfs_btree_shift_keys(
        struct xfs_btree_cur    *cur,
        union xfs_btree_key     *key,
        int                     dir,
        int                     numkeys)
{
        char                    *dst_key;

        ASSERT(numkeys >= 0);
        ASSERT(dir == 1 || dir == -1);

        dst_key = (char *)key + (dir * cur->bc_ops->key_len);
        memmove(dst_key, key, numkeys * cur->bc_ops->key_len);
}

/*
 * Shift records one index left/right inside a single btree block.
 */
STATIC void
xfs_btree_shift_recs(
        struct xfs_btree_cur    *cur,
        union xfs_btree_rec     *rec,
        int                     dir,
        int                     numrecs)
{
        char                    *dst_rec;

        ASSERT(numrecs >= 0);
        ASSERT(dir == 1 || dir == -1);

        dst_rec = (char *)rec + (dir * cur->bc_ops->rec_len);
        memmove(dst_rec, rec, numrecs * cur->bc_ops->rec_len);
}

/*
 * Shift block pointers one index left/right inside a single btree block.
 */
STATIC void
xfs_btree_shift_ptrs(
        struct xfs_btree_cur    *cur,
        union xfs_btree_ptr     *ptr,
        int                     dir,
        int                     numptrs)
{
        char                    *dst_ptr;

        ASSERT(numptrs >= 0);
        ASSERT(dir == 1 || dir == -1);

        dst_ptr = (char *)ptr + (dir * xfs_btree_ptr_len(cur));
        memmove(dst_ptr, ptr, numptrs * xfs_btree_ptr_len(cur));
}

/*
 * Log key values from the btree block.
 */
STATIC void
xfs_btree_log_keys(
        struct xfs_btree_cur    *cur,
        struct xfs_buf          *bp,
        int                     first,
        int                     last)
{

        if (bp) {
                xfs_trans_buf_set_type(cur->bc_tp, bp, XFS_BLFT_BTREE_BUF);
                xfs_trans_log_buf(cur->bc_tp, bp,
                                  xfs_btree_key_offset(cur, first),
                                  xfs_btree_key_offset(cur, last + 1) - 1);
        } else {
                xfs_trans_log_inode(cur->bc_tp, cur->bc_ino.ip,
                                xfs_ilog_fbroot(cur->bc_ino.whichfork));
        }
}

/*
 * Log record values from the btree block.
 */
void
xfs_btree_log_recs(
        struct xfs_btree_cur    *cur,
        struct xfs_buf          *bp,
        int                     first,
        int                     last)
{

        xfs_trans_buf_set_type(cur->bc_tp, bp, XFS_BLFT_BTREE_BUF);
        xfs_trans_log_buf(cur->bc_tp, bp,
                          xfs_btree_rec_offset(cur, first),
                          xfs_btree_rec_offset(cur, last + 1) - 1);

}

/*
 * Log block pointer fields from a btree block (nonleaf).
 */
STATIC void
xfs_btree_log_ptrs(
        struct xfs_btree_cur    *cur,   /* btree cursor */
        struct xfs_buf          *bp,    /* buffer containing btree block */
        int                     first,  /* index of first pointer to log */
        int                     last)   /* index of last pointer to log */
{

        if (bp) {
                struct xfs_btree_block  *block = XFS_BUF_TO_BLOCK(bp);
                int                     level = xfs_btree_get_level(block);

                xfs_trans_buf_set_type(cur->bc_tp, bp, XFS_BLFT_BTREE_BUF);
                xfs_trans_log_buf(cur->bc_tp, bp,
                                xfs_btree_ptr_offset(cur, first, level),
                                xfs_btree_ptr_offset(cur, last + 1, level) - 1);
        } else {
                xfs_trans_log_inode(cur->bc_tp, cur->bc_ino.ip,
                        xfs_ilog_fbroot(cur->bc_ino.whichfork));
        }

}

/*
 * Log fields from a btree block header.
 */
void
xfs_btree_log_block(
        struct xfs_btree_cur    *cur,   /* btree cursor */
        struct xfs_buf          *bp,    /* buffer containing btree block */
        int                     fields) /* mask of fields: XFS_BB_... */
{
        int                     first;  /* first byte offset logged */
        int                     last;   /* last byte offset logged */
        static const short      soffsets[] = {  /* table of offsets (short) */
                offsetof(struct xfs_btree_block, bb_magic),
                offsetof(struct xfs_btree_block, bb_level),
                offsetof(struct xfs_btree_block, bb_numrecs),
                offsetof(struct xfs_btree_block, bb_u.s.bb_leftsib),
                offsetof(struct xfs_btree_block, bb_u.s.bb_rightsib),
                offsetof(struct xfs_btree_block, bb_u.s.bb_blkno),
                offsetof(struct xfs_btree_block, bb_u.s.bb_lsn),
                offsetof(struct xfs_btree_block, bb_u.s.bb_uuid),
                offsetof(struct xfs_btree_block, bb_u.s.bb_owner),
                offsetof(struct xfs_btree_block, bb_u.s.bb_crc),
                XFS_BTREE_SBLOCK_CRC_LEN
        };
        static const short      loffsets[] = {  /* table of offsets (long) */
                offsetof(struct xfs_btree_block, bb_magic),
                offsetof(struct xfs_btree_block, bb_level),
                offsetof(struct xfs_btree_block, bb_numrecs),
                offsetof(struct xfs_btree_block, bb_u.l.bb_leftsib),
                offsetof(struct xfs_btree_block, bb_u.l.bb_rightsib),
                offsetof(struct xfs_btree_block, bb_u.l.bb_blkno),
                offsetof(struct xfs_btree_block, bb_u.l.bb_lsn),
                offsetof(struct xfs_btree_block, bb_u.l.bb_uuid),
                offsetof(struct xfs_btree_block, bb_u.l.bb_owner),
                offsetof(struct xfs_btree_block, bb_u.l.bb_crc),
                offsetof(struct xfs_btree_block, bb_u.l.bb_pad),
                XFS_BTREE_LBLOCK_CRC_LEN
        };

        if (bp) {
                int nbits;

                if (cur->bc_flags & XFS_BTREE_CRC_BLOCKS) {
                        /*
                         * We don't log the CRC when updating a btree
                         * block but instead recreate it during log
                         * recovery.  As the log buffers have checksums
                         * of their own this is safe and avoids logging a crc
                         * update in a lot of places.
                         */
                        if (fields == XFS_BB_ALL_BITS)
                                fields = XFS_BB_ALL_BITS_CRC;
                        nbits = XFS_BB_NUM_BITS_CRC;
                } else {
                        nbits = XFS_BB_NUM_BITS;
                }
                xfs_btree_offsets(fields,
                                  (cur->bc_flags & XFS_BTREE_LONG_PTRS) ?
                                        loffsets : soffsets,
                                  nbits, &first, &last);
                xfs_trans_buf_set_type(cur->bc_tp, bp, XFS_BLFT_BTREE_BUF);
                xfs_trans_log_buf(cur->bc_tp, bp, first, last);
        } else {
                xfs_trans_log_inode(cur->bc_tp, cur->bc_ino.ip,
                        xfs_ilog_fbroot(cur->bc_ino.whichfork));
        }
}

/*
 * Increment cursor by one record at the level.
 * For nonzero levels the leaf-ward information is untouched.
 */
int                                             /* error */
xfs_btree_increment(
        struct xfs_btree_cur    *cur,
        int                     level,
        int                     *stat)          /* success/failure */
{
        struct xfs_btree_block  *block;
        union xfs_btree_ptr     ptr;
        struct xfs_buf          *bp;
        int                     error;          /* error return value */
        int                     lev;

        ASSERT(level < cur->bc_nlevels);

        /* Read-ahead to the right at this level. */
        xfs_btree_readahead(cur, level, XFS_BTCUR_RIGHTRA);

        /* Get a pointer to the btree block. */
        block = xfs_btree_get_block(cur, level, &bp);

#ifdef DEBUG
        error = xfs_btree_check_block(cur, block, level, bp);
        if (error)
                goto error0;
#endif

        /* We're done if we remain in the block after the increment. */
        if (++cur->bc_ptrs[level] <= xfs_btree_get_numrecs(block))
                goto out1;

        /* Fail if we just went off the right edge of the tree. */
        xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_RIGHTSIB);
        if (xfs_btree_ptr_is_null(cur, &ptr))
                goto out0;

        XFS_BTREE_STATS_INC(cur, increment);

        /*
         * March up the tree incrementing pointers.
         * Stop when we don't go off the right edge of a block.
         */
        for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
                block = xfs_btree_get_block(cur, lev, &bp);

#ifdef DEBUG
                error = xfs_btree_check_block(cur, block, lev, bp);
                if (error)
                        goto error0;
#endif

                if (++cur->bc_ptrs[lev] <= xfs_btree_get_numrecs(block))
                        break;

                /* Read-ahead the right block for the next loop. */
                xfs_btree_readahead(cur, lev, XFS_BTCUR_RIGHTRA);
        }

        /*
         * If we went off the root then we are either seriously
         * confused or have the tree root in an inode.
         */
        if (lev == cur->bc_nlevels) {
                if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE)
                        goto out0;
                ASSERT(0);
                error = -EFSCORRUPTED;
                goto error0;
        }
        ASSERT(lev < cur->bc_nlevels);

        /*
         * Now walk back down the tree, fixing up the cursor's buffer
         * pointers and key numbers.
         */
        for (block = xfs_btree_get_block(cur, lev, &bp); lev > level; ) {
                union xfs_btree_ptr     *ptrp;

                ptrp = xfs_btree_ptr_addr(cur, cur->bc_ptrs[lev], block);
                --lev;
                error = xfs_btree_read_buf_block(cur, ptrp, 0, &block, &bp);
                if (error)
                        goto error0;

                xfs_btree_setbuf(cur, lev, bp);
                cur->bc_ptrs[lev] = 1;
        }
out1:
        *stat = 1;
        return 0;

out0:
        *stat = 0;
        return 0;

error0:
        return error;
}

/*
 * Decrement cursor by one record at the level.
 * For nonzero levels the leaf-ward information is untouched.
 */
int                                             /* error */
xfs_btree_decrement(
        struct xfs_btree_cur    *cur,
        int                     level,
        int                     *stat)          /* success/failure */
{
        struct xfs_btree_block  *block;
        struct xfs_buf          *bp;
        int                     error;          /* error return value */
        int                     lev;
        union xfs_btree_ptr     ptr;

        ASSERT(level < cur->bc_nlevels);

        /* Read-ahead to the left at this level. */
        xfs_btree_readahead(cur, level, XFS_BTCUR_LEFTRA);

        /* We're done if we remain in the block after the decrement. */
        if (--cur->bc_ptrs[level] > 0)
                goto out1;

        /* Get a pointer to the btree block. */
        block = xfs_btree_get_block(cur, level, &bp);

#ifdef DEBUG
        error = xfs_btree_check_block(cur, block, level, bp);
        if (error)
                goto error0;
#endif

        /* Fail if we just went off the left edge of the tree. */
        xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_LEFTSIB);
        if (xfs_btree_ptr_is_null(cur, &ptr))
                goto out0;

        XFS_BTREE_STATS_INC(cur, decrement);

        /*
         * March up the tree decrementing pointers.
         * Stop when we don't go off the left edge of a block.
         */
        for (lev = level + 1; lev < cur->bc_nlevels; lev++) {
                if (--cur->bc_ptrs[lev] > 0)
                        break;
                /* Read-ahead the left block for the next loop. */
                xfs_btree_readahead(cur, lev, XFS_BTCUR_LEFTRA);
        }

        /*
         * If we went off the root then we are seriously confused.
         * or the root of the tree is in an inode.
         */
        if (lev == cur->bc_nlevels) {
                if (cur->bc_flags & XFS_BTREE_ROOT_IN_INODE)
                        goto out0;
                ASSERT(0);
                error = -EFSCORRUPTED;
                goto error0;
        }
        ASSERT(lev < cur->bc_nlevels);

        /*
         * Now walk back down the tree, fixing up the cursor's buffer
         * pointers and key numbers.
         */
        for (block = xfs_btree_get_block(cur, lev, &bp); lev > level; ) {
                union xfs_btree_ptr     *ptrp;

                ptrp = xfs_btree_ptr_addr(cur, cur->bc_ptrs[lev], block);
                --lev;
                error = xfs_btree_read_buf_block(cur, ptrp, 0, &block, &bp);
                if (error)
                        goto error0;
                xfs_btree_setbuf(cur, lev, bp);
                cur->bc_ptrs[lev] = xfs_btree_get_numrecs(block);
        }
out1:
        *stat = 1;
        return 0;

out0:
        *stat = 0;
        return 0;

error0:
        return error;
}

int
xfs_btree_lookup_get_block(
        struct xfs_btree_cur            *cur,   /* btree cursor */
        int                             level,  /* level in the btree */
        const union xfs_btree_ptr       *pp,    /* ptr to btree block */
        struct xfs_btree_block          **blkp) /* return btree block */
{
        struct xfs_buf          *bp;    /* buffer pointer for btree block */
        xfs_daddr_t             daddr;
        int                     error = 0;

        /* special case the root block if in an inode */
        if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
            (level == cur->bc_nlevels - 1)) {
                *blkp = xfs_btree_get_iroot(cur);
                return 0;
        }

        /*
         * If the old buffer at this level for the disk address we are
         * looking for re-use it.
         *
         * Otherwise throw it away and get a new one.
         */
        bp = cur->bc_bufs[level];
        error = xfs_btree_ptr_to_daddr(cur, pp, &daddr);
        if (error)
                return error;
        if (bp && xfs_buf_daddr(bp) == daddr) {
                *blkp = XFS_BUF_TO_BLOCK(bp);
                return 0;
        }

        error = xfs_btree_read_buf_block(cur, pp, 0, blkp, &bp);
        if (error)
                return error;

        /* Check the inode owner since the verifiers don't. */
        if (xfs_has_crc(cur->bc_mp) &&
            !(cur->bc_ino.flags & XFS_BTCUR_BMBT_INVALID_OWNER) &&
            (cur->bc_flags & XFS_BTREE_LONG_PTRS) &&
            be64_to_cpu((*blkp)->bb_u.l.bb_owner) !=
                        cur->bc_ino.ip->i_ino)
                goto out_bad;

        /* Did we get the level we were looking for? */
        if (be16_to_cpu((*blkp)->bb_level) != level)
                goto out_bad;

        /* Check that internal nodes have at least one record. */
        if (level != 0 && be16_to_cpu((*blkp)->bb_numrecs) == 0)
                goto out_bad;

        xfs_btree_setbuf(cur, level, bp);
        return 0;

out_bad:
        *blkp = NULL;
        xfs_buf_mark_corrupt(bp);
        xfs_trans_brelse(cur->bc_tp, bp);
        return -EFSCORRUPTED;
}

/*
 * Get current search key.  For level 0 we don't actually have a key
 * structure so we make one up from the record.  For all other levels
 * we just return the right key.
 */
STATIC union xfs_btree_key *
xfs_lookup_get_search_key(
        struct xfs_btree_cur    *cur,
        int                     level,
        int                     keyno,
        struct xfs_btree_block  *block,
        union xfs_btree_key     *kp)
{
        if (level == 0) {
                cur->bc_ops->init_key_from_rec(kp,
                                xfs_btree_rec_addr(cur, keyno, block));
                return kp;
        }

        return xfs_btree_key_addr(cur, keyno, block);
}

/*
 * Lookup the record.  The cursor is made to point to it, based on dir.
 * stat is set to 0 if can't find any such record, 1 for success.
 */
int                                     /* error */
xfs_btree_lookup(
        struct xfs_btree_cur    *cur,   /* btree cursor */
        xfs_lookup_t            dir,    /* <=, ==, or >= */
        int                     *stat)  /* success/failure */
{
        struct xfs_btree_block  *block; /* current btree block */
        int64_t                 diff;   /* difference for the current key */
        int                     error;  /* error return value */
        int                     keyno;  /* current key number */
        int                     level;  /* level in the btree */
        union xfs_btree_ptr     *pp;    /* ptr to btree block */
        union xfs_btree_ptr     ptr;    /* ptr to btree block */

        XFS_BTREE_STATS_INC(cur, lookup);

        /* No such thing as a zero-level tree. */
        if (XFS_IS_CORRUPT(cur->bc_mp, cur->bc_nlevels == 0))
                return -EFSCORRUPTED;

        block = NULL;
        keyno = 0;

        /* initialise start pointer from cursor */
        cur->bc_ops->init_ptr_from_cur(cur, &ptr);
        pp = &ptr;

        /*
         * Iterate over each level in the btree, starting at the root.
         * For each level above the leaves, find the key we need, based
         * on the lookup record, then follow the corresponding block
         * pointer down to the next level.
         */
        for (level = cur->bc_nlevels - 1, diff = 1; level >= 0; level--) {
                /* Get the block we need to do the lookup on. */
                error = xfs_btree_lookup_get_block(cur, level, pp, &block);
                if (error)
                        goto error0;

                if (diff == 0) {
                        /*
                         * If we already had a key match at a higher level, we
                         * know we need to use the first entry in this block.
                         */
                        keyno = 1;
                } else {
                        /* Otherwise search this block. Do a binary search. */

                        int     high;   /* high entry number */
                        int     low;    /* low entry number */

                        /* Set low and high entry numbers, 1-based. */
                        low = 1;
                        high = xfs_btree_get_numrecs(block);
                        if (!high) {
                                /* Block is empty, must be an empty leaf. */
                                if (level != 0 || cur->bc_nlevels != 1) {
                                        XFS_CORRUPTION_ERROR(__func__,
                                                        XFS_ERRLEVEL_LOW,
                                                        cur->bc_mp, block,
                                                        sizeof(*block));
                                        return -EFSCORRUPTED;
                                }

                                cur->bc_ptrs[0] = dir != XFS_LOOKUP_LE;
                                *stat = 0;
                                return 0;
                        }

                        /* Binary search the block. */
                        while (low <= high) {
                                union xfs_btree_key     key;
                                union xfs_btree_key     *kp;

                                XFS_BTREE_STATS_INC(cur, compare);

                                /* keyno is average of low and high. */
                                keyno = (low + high) >> 1;

                                /* Get current search key */
                                kp = xfs_lookup_get_search_key(cur, level,
                                                keyno, block, &key);

                                /*
                                 * Compute difference to get next direction:
                                 *  - less than, move right
                                 *  - greater than, move left
                                 *  - equal, we're done
                                 */
                                diff = cur->bc_ops->key_diff(cur, kp);
                                if (diff < 0)
                                        low = keyno + 1;
                                else if (diff > 0)
                                        high = keyno - 1;
                                else
                                        break;
                        }
                }

                /*
                 * If there are more levels, set up for the next level
                 * by getting the block number and filling in the cursor.
                 */
                if (level > 0) {
                        /*
                         * If we moved left, need the previous key number,
                         * unless there isn't one.
                         */
                        if (diff > 0 && --keyno < 1)
                                keyno = 1;
                        pp = xfs_btree_ptr_addr(cur, keyno, block);

                        error = xfs_btree_debug_check_ptr(cur, pp, 0, level);
                        if (error)
                                goto error0;

                        cur->bc_ptrs[level] = keyno;
                }
        }

        /* Done with the search. See if we need to adjust the results. */
        if (dir != XFS_LOOKUP_LE && diff < 0) {
                keyno++;
                /*
                 * If ge search and we went off the end of the block, but it's
                 * not the last block, we're in the wrong block.
                 */
                xfs_btree_get_sibling(cur, block, &ptr, XFS_BB_RIGHTSIB);
                if (dir == XFS_LOOKUP_GE &&
                    keyno > xfs_btree_get_numrecs(block) &&
                    !xfs_btree_ptr_is_null(cur, &ptr)) {
                        int     i;

                        cur->bc_ptrs[0] = keyno;
                        error = xfs_btree_increment(cur, 0, &i);
                        if (error)
                                goto error0;
                        if (XFS_IS_CORRUPT(cur->bc_mp, i != 1))
                                return -EFSCORRUPTED;
                        *stat = 1;
                        return 0;
                }
        } else if (dir == XFS_LOOKUP_LE && diff > 0)
                keyno--;
        cur->bc_ptrs[0] = keyno;

        /* Return if we succeeded or not. */
        if (keyno == 0 || keyno > xfs_btree_get_numrecs(block))
                *stat = 0;
        else if (dir != XFS_LOOKUP_EQ || diff == 0)
                *stat = 1;
        else
                *stat = 0;
        return 0;

error0:
        return error;
}

/* Find the high key storage area from a regular key. */
union xfs_btree_key *
xfs_btree_high_key_from_key(
        struct xfs_btree_cur    *cur,
        union xfs_btree_key     *key)
{
        ASSERT(cur->bc_flags & XFS_BTREE_OVERLAPPING);
        return (union xfs_btree_key *)((char *)key +
                        (cur->bc_ops->key_len / 2));
}

/* Determine the low (and high if overlapped) keys of a leaf block */
STATIC void
xfs_btree_get_leaf_keys(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        union xfs_btree_key     *key)
{
        union xfs_btree_key     max_hkey;
        union xfs_btree_key     hkey;
        union xfs_btree_rec     *rec;
        union xfs_btree_key     *high;
        int                     n;

        rec = xfs_btree_rec_addr(cur, 1, block);
        cur->bc_ops->init_key_from_rec(key, rec);

        if (cur->bc_flags & XFS_BTREE_OVERLAPPING) {

                cur->bc_ops->init_high_key_from_rec(&max_hkey, rec);
                for (n = 2; n <= xfs_btree_get_numrecs(block); n++) {
                        rec = xfs_btree_rec_addr(cur, n, block);
                        cur->bc_ops->init_high_key_from_rec(&hkey, rec);
                        if (cur->bc_ops->diff_two_keys(cur, &hkey, &max_hkey)
                                        > 0)
                                max_hkey = hkey;
                }

                high = xfs_btree_high_key_from_key(cur, key);

                memcpy(high, &max_hkey, cur->bc_ops->key_len / 2);
        }
}

/* Determine the low (and high if overlapped) keys of a node block */
STATIC void
xfs_btree_get_node_keys(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        union xfs_btree_key     *key)
{
        union xfs_btree_key     *hkey;
        union xfs_btree_key     *max_hkey;
        union xfs_btree_key     *high;
        int                     n;

        if (cur->bc_flags & XFS_BTREE_OVERLAPPING) {
                memcpy(key, xfs_btree_key_addr(cur, 1, block),
                                cur->bc_ops->key_len / 2);

                max_hkey = xfs_btree_high_key_addr(cur, 1, block);
                for (n = 2; n <= xfs_btree_get_numrecs(block); n++) {
                        hkey = xfs_btree_high_key_addr(cur, n, block);
                        if (cur->bc_ops->diff_two_keys(cur, hkey, max_hkey) > 0)
                                max_hkey = hkey;
                }

                high = xfs_btree_high_key_from_key(cur, key);
                memcpy(high, max_hkey, cur->bc_ops->key_len / 2);
        } else {
                memcpy(key, xfs_btree_key_addr(cur, 1, block),
                                cur->bc_ops->key_len);
        }
}

/* Derive the keys for any btree block. */
void
xfs_btree_get_keys(
        struct xfs_btree_cur    *cur,
        struct xfs_btree_block  *block,
        union xfs_btree_key     *key)
{
        if (be16_to_cpu(block->bb_level) == 0)
                xfs_btree_get_leaf_keys(cur, block, key);
        else
                xfs_btree_get_node_keys(cur, block, key);
}

/*
 * Decide if we need to update the parent keys of a btree block.  For
 * a standard btree this is only necessary if we're updating the first
 * record/key.  For an overlapping btree, we must always update the
 * keys because the highest key can be in any of the records or keys
 * in the block.
 */
static inline bool
xfs_btree_needs_key_update(
        struct xfs_btree_cur    *cur,
        int                     ptr)
{
        return (cur->bc_flags & XFS_BTREE_OVERLAPPING) || ptr == 1;
}

/*
 * Update the low and high parent keys of the given level, progressing
 * towards the root.  If force_all is false, stop if the keys for a given
 * level do not need updating.
 */
STATIC int
__xfs_btree_updkeys(
        struct xfs_btree_cur    *cur,
        int                     level,
        struct xfs_btree_block  *block,
        struct xfs_buf          *bp0,
        bool                    force_all)
{
        union xfs_btree_key     key;    /* keys from current level */
        union xfs_btree_key     *lkey;  /* keys from the next level up */
        union xfs_btree_key     *hkey;
        union xfs_btree_key     *nlkey; /* keys from the next level up */
        union xfs_btree_key     *nhkey;
        struct xfs_buf          *bp;
        int                     ptr;

        ASSERT(cur->bc_flags & XFS_BTREE_OVERLAPPING);

        /* Exit if there aren't any parent levels to update. */
        if (level + 1 >= cur->bc_nlevels)
                return 0;

        trace_xfs_btree_updkeys(cur, level, bp0);

        lkey = &key;
        hkey = xfs_btree_high_key_from_key(cur, lkey);
        xfs_btree_get_keys(cur, block, lkey);
        for (level++; level < cur->bc_nlevels; level++) {
#ifdef DEBUG
                int             error;
#endif
                block = xfs_btree_get_block(cur, level, &bp);
                trace_xfs_btree_updkeys(cur, level, bp);
#ifdef DEBUG
                error = xfs_btree_check_block(cur, block, level, bp);
                if (error)
                        return error;
#endif
                ptr = cur->bc_ptrs[level];
                nlkey = xfs_btree_key_addr(cur, ptr, block);
                nhkey = xfs_btree_high_key_addr(cur, ptr, block);
                if (!force_all &&
                    !(cur->bc_ops->diff_two_keys(cur, nlkey, lkey) != 0 ||
                      cur->bc_ops->diff_two_keys(cur, nhkey, hkey) != 0))
                        break;
                xfs_btree_copy_keys(cur, nlkey, lkey, 1);
                xfs_btree_log_keys(cur, bp, ptr, ptr);
                if (level + 1 >= cur->bc_nlevels)
                        break;
                xfs_btree_get_node_keys(cur, block, lkey);
        }

        return 0;
}

/* Update all the keys from some level in cursor back to the root. */
STATIC int
xfs_btree_updkeys_force(
        struct xfs_btree_cur    *cur,
        int                     level)
{
        struct xfs_buf          *bp;
        struct xfs_btree_block  *block;

        block = xfs_btree_get_block(cur, level, &bp);
        return __xfs_btree_updkeys(cur, level, block, bp, true);
}

/*
 * Update the parent keys of the given level, progressing towards the root.
 */
STATIC int
xfs_btree_update_keys(
        struct xfs_btree_cur    *cur,
        int                     level)
{
        struct xfs_btree_block  *block;
        struct xfs_buf          *bp;
        union xfs_btree_key     *kp;
        union xfs_btree_key     key;
        int                     ptr;

        ASSERT(level >= 0);

        block = xfs_btree_get_block(cur, level, &bp);
        if (cur->bc_flags & XFS_BTREE_OVERLAPPING)
                return __xfs_btree_updkeys(cur, level, block, bp, false);

        /*
         * Go up the tree from this level toward the root.
         * At each level, update the key value to the value input.
         * Stop when we reach a level where the cursor isn't pointing
         * at the first entry in the block.
         */
        xfs_btree_get_keys(cur, block, &key);
        for (level++, ptr = 1; ptr == 1 && level < cur->bc_nlevels; level++) {
#ifdef DEBUG
                int             error;
#endif
                block = xfs_btree_get_block(cur, level, &bp);
#ifdef DEBUG
                error = xfs_btree_check_block(cur, block, level, bp);
                if (error)
                        return error;
#endif
                ptr = cur->bc_ptrs[level];
                kp = xfs_btree_key_addr(cur, ptr, block);
                xfs_btree_copy_keys(cur, kp, &key, 1);
                xfs_btree_log_keys(cur, bp, ptr, ptr);
        }

        return 0;
}

/*
 * Update the record referred to by cur to the value in the
 * given record. This either works (return 0) or gets an
 * EFSCORRUPTED error.
 */
int
xfs_btree_update(
        struct xfs_btree_cur    *cur,
        union xfs_btree_rec     *rec)
{
        struct xfs_btree_block  *block;
        struct xfs_buf          *bp;
        int                     error;
        int                     ptr;
        union xfs_btree_rec     *rp;

        /* Pick up the current block. */
        block = xfs_btree_get_block(cur, 0, &bp);

#ifdef DEBUG
        error = xfs_btree_check_block(cur, block, 0, bp);
        if (error)
                goto error0;
#endif
        /* Get the address of the rec to be updated. */
        ptr = cur->bc_ptrs[0];
        rp = xfs_btree_rec_addr(cur, ptr, block);

        /* Fill in the new contents and log them. */
        xfs_btree_copy_recs(cur, rp, rec, 1);
        xfs_btree_log_recs(cur, bp, ptr, ptr);

        /*
         * If we are tracking the last record in the tree and
         * we are at the far right edge of the tree, update it.
         */
        if (xfs_btree_is_lastrec(cur, block, 0)) {
                cur->bc_ops->update_lastrec(cur, block, rec,
                                            ptr, LASTREC_UPDATE);
        }

        /* Pass new key value up to our parent. */
        if (xfs_btree_needs_key_update(cur, ptr)) {
                error = xfs_btree_update_keys(cur, 0);
                if (error)
                        goto error0;
        }

        return 0;

error0:
        return error;
}

/*
 * Move 1 record left from cur/level if possible.
 * Update cur to reflect the new path.
 */
STATIC int                                      /* error */
xfs_btree_lshift(
        struct xfs_btree_cur    *cur,
        int                     level,
        int                     *stat)          /* success/failure */
{
        struct xfs_buf          *lbp;           /* left buffer pointer */
        struct xfs_btree_block  *left;          /* left btree block */
        int                     lrecs;          /* left record count */
        struct xfs_buf          *rbp;           /* right buffer pointer */
        struct xfs_btree_block  *right;         /* right btree block */
        struct xfs_btree_cur    *tcur;          /* temporary btree cursor */
        int                     rrecs;          /* right record count */
        union xfs_btree_ptr     lptr;           /* left btree pointer */
        union xfs_btree_key     *rkp = NULL;    /* right btree key */
        union xfs_btree_ptr     *rpp = NULL;    /* right address pointer */
        union xfs_btree_rec     *rrp = NULL;    /* right record pointer */
        int                     error;          /* error return value */
        int                     i;

        if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
            level == cur->bc_nlevels - 1)
                goto out0;

        /* Set up variables for this block as "right". */
        right = xfs_btree_get_block(cur, level, &rbp);

#ifdef DEBUG
        error = xfs_btree_check_block(cur, right, level, rbp);
        if (error)
                goto error0;
#endif

        /* If we've got no left sibling then we can't shift an entry left. */
        xfs_btree_get_sibling(cur, right, &lptr, XFS_BB_LEFTSIB);
        if (xfs_btree_ptr_is_null(cur, &lptr))
                goto out0;

        /*
         * If the cursor entry is the one that would be moved, don't
         * do it... it's too complicated.
         */
        if (cur->bc_ptrs[level] <= 1)
                goto out0;

        /* Set up the left neighbor as "left". */
        error = xfs_btree_read_buf_block(cur, &lptr, 0, &left, &lbp);
        if (error)
                goto error0;

        /* If it's full, it can't take another entry. */
        lrecs = xfs_btree_get_numrecs(left);
        if (lrecs == cur->bc_ops->get_maxrecs(cur, level))
                goto out0;

        rrecs = xfs_btree_get_numrecs(right);

        /*
         * We add one entry to the left side and remove one for the right side.
         * Account for it here, the changes will be updated on disk and logged
         * later.
         */
        lrecs++;
        rrecs--;

        XFS_BTREE_STATS_INC(cur, lshift);
        XFS_BTREE_STATS_ADD(cur, moves, 1);

        /*
         * If non-leaf, copy a key and a ptr to the left block.
         * Log the changes to the left block.
         */
        if (level > 0) {
                /* It's a non-leaf.  Move keys and pointers. */
                union xfs_btree_key     *lkp;   /* left btree key */
                union xfs_btree_ptr     *lpp;   /* left address pointer */

                lkp = xfs_btree_key_addr(cur, lrecs, left);
                rkp = xfs_btree_key_addr(cur, 1, right);

                lpp = xfs_btree_ptr_addr(cur, lrecs, left);
                rpp = xfs_btree_ptr_addr(cur, 1, right);

                error = xfs_btree_debug_check_ptr(cur, rpp, 0, level);
                if (error)
                        goto error0;

                xfs_btree_copy_keys(cur, lkp, rkp, 1);
                xfs_btree_copy_ptrs(cur, lpp, rpp, 1);

                xfs_btree_log_keys(cur, lbp, lrecs, lrecs);
                xfs_btree_log_ptrs(cur, lbp, lrecs, lrecs);

                ASSERT(cur->bc_ops->keys_inorder(cur,
                        xfs_btree_key_addr(cur, lrecs - 1, left), lkp));
        } else {
                /* It's a leaf.  Move records.  */
                union xfs_btree_rec     *lrp;   /* left record pointer */

                lrp = xfs_btree_rec_addr(cur, lrecs, left);
                rrp = xfs_btree_rec_addr(cur, 1, right);

                xfs_btree_copy_recs(cur, lrp, rrp, 1);
                xfs_btree_log_recs(cur, lbp, lrecs, lrecs);

                ASSERT(cur->bc_ops->recs_inorder(cur,
                        xfs_btree_rec_addr(cur, lrecs - 1, left), lrp));
        }

        xfs_btree_set_numrecs(left, lrecs);
        xfs_btree_log_block(cur, lbp, XFS_BB_NUMRECS);

        xfs_btree_set_numrecs(right, rrecs);
        xfs_btree_log_block(cur, rbp, XFS_BB_NUMRECS);

        /*
         * Slide the contents of right down one entry.
         */
        XFS_BTREE_STATS_ADD(cur, moves, rrecs - 1);
        if (level > 0) {
                /* It's a nonleaf. operate on keys and ptrs */
                for (i = 0; i < rrecs; i++) {
                        error = xfs_btree_debug_check_ptr(cur, rpp, i + 1, level);
                        if (error)
                                goto error0;
                }

                xfs_btree_shift_keys(cur,
                                xfs_btree_key_addr(cur, 2, right),
                                -1, rrecs);
                xfs_btree_shift_ptrs(cur,
                                xfs_btree_ptr_addr(cur, 2, right),
                                -1, rrecs);

                xfs_btree_log_keys(cur, rbp, 1, rrecs);
                xfs_btree_log_ptrs(cur, rbp, 1, rrecs);
        } else {
                /* It's a leaf. operate on records */
                xfs_btree_shift_recs(cur,
                        xfs_btree_rec_addr(cur, 2, right),
                        -1, rrecs);
                xfs_btree_log_recs(cur, rbp, 1, rrecs);
        }

        /*
         * Using a temporary cursor, update the parent key values of the
         * block on the left.
         */
        if (cur->bc_flags & XFS_BTREE_OVERLAPPING) {
                error = xfs_btree_dup_cursor(cur, &tcur);
                if (error)
                        goto error0;
                i = xfs_btree_firstrec(tcur, level);
                if (XFS_IS_CORRUPT(tcur->bc_mp, i != 1)) {
                        error = -EFSCORRUPTED;
                        goto error0;
                }

                error = xfs_btree_decrement(tcur, level, &i);
                if (error)
                        goto error1;

                /* Update the parent high keys of the left block, if needed. */
                error = xfs_btree_update_keys(tcur, level);
                if (error)
                        goto error1;

                xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);
        }

        /* Update the parent keys of the right block. */
        error = xfs_btree_update_keys(cur, level);
        if (error)
                goto error0;

        /* Slide the cursor value left one. */
        cur->bc_ptrs[level]--;

        *stat = 1;
        return 0;

out0:
        *stat = 0;
        return 0;

error0:
        return error;

error1:
        xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
        return error;
}

/*
 * Move 1 record right from cur/level if possible.
 * Update cur to reflect the new path.
 */
STATIC int                                      /* error */
xfs_btree_rshift(
        struct xfs_btree_cur    *cur,
        int                     level,
        int                     *stat)          /* success/failure */
{
        struct xfs_buf          *lbp;           /* left buffer pointer */
        struct xfs_btree_block  *left;          /* left btree block */
        struct xfs_buf          *rbp;           /* right buffer pointer */
        struct xfs_btree_block  *right;         /* right btree block */
        struct xfs_btree_cur    *tcur;          /* temporary btree cursor */
        union xfs_btree_ptr     rptr;           /* right block pointer */
        union xfs_btree_key     *rkp;           /* right btree key */
        int                     rrecs;          /* right record count */
        int                     lrecs;          /* left record count */
        int                     error;          /* error return value */
        int                     i;              /* loop counter */

        if ((cur->bc_flags & XFS_BTREE_ROOT_IN_INODE) &&
            (level == cur->bc_nlevels - 1))
                goto out0;

        /* Set up variables for this block as "left". */
        left = xfs_btree_get_block(cur, level, &lbp);

#ifdef DEBUG
        error = xfs_btree_check_block(cur, left, level, lbp);
        if (error)
                goto error0;
#endif

        /* If we've got no right sibling then we can't shift an entry right. */
        xfs_btree_get_sibling(cur, left, &rptr, XFS_BB_RIGHTSIB);
        if (xfs_btree_ptr_is_null(cur, &rptr))
                goto out0;

        /*
         * If the cursor entry is the one that would be moved, don't
         * do it... it's too complicated.
         */
        lrecs = xfs_btree_get_numrecs(left);
        if (cur->bc_ptrs[level] >= lrecs)
                goto out0;

        /* Set up the right neighbor as "right". */
        error = xfs_btree_read_buf_block(cur, &rptr, 0, &right, &rbp);
        if (error)
                goto error0;

        /* If it's full, it can't take another entry. */
        rrecs = xfs_btree_get_numrecs(right);
        if (rrecs == cur->bc_ops->get_maxrecs(cur, level))
                goto out0;

        XFS_BTREE_STATS_INC(cur, rshift);
        XFS_BTREE_STATS_ADD(cur, moves, rrecs);

        /*
         * Make a hole at the start of the right neighbor block, then
         * copy the last left block entry to the hole.
         */
        if (level > 0) {
                /* It's a nonleaf. make a hole in the keys and ptrs */
                union xfs_btree_key     *lkp;
                union xfs_btree_ptr     *lpp;
                union xfs_btree_ptr     *rpp;

                lkp = xfs_btree_key_addr(cur, lrecs, left);
                lpp = xfs_btree_ptr_addr(cur, lrecs, left);
                rkp = xfs_btree_key_addr(cur, 1, right);
                rpp = xfs_btree_ptr_addr(cur, 1, right);

                for (i = rrecs - 1; i >= 0; i--) {
                        error = xfs_btree_debug_check_ptr(cur, rpp, i, level);
                        if (error)
                                goto error0;
                }

                xfs_btree_shift_keys(cur, rkp, 1, rrecs);
                xfs_btree_shift_ptrs(cur, rpp, 1, rrecs);

                error = xfs_btree_debug_check_ptr(cur, lpp, 0, level);
                if (error)
                        goto error0;

                /* Now put the new data in, and log it. */
                xfs_btree_copy_keys(cur, rkp, lkp, 1);
                xfs_btree_copy_ptrs(cur, rpp, lpp, 1);

                xfs_btree_log_keys(cur, rbp, 1, rrecs + 1);
                xfs_btree_log_ptrs(cur, rbp, 1, rrecs + 1);

                ASSERT(cur->bc_ops->keys_inorder(cur, rkp,
                        xfs_btree_key_addr(cur, 2, right)));
        } else {
                /* It's a leaf. make a hole in the records */
                union xfs_btree_rec     *lrp;
                union xfs_btree_rec     *rrp;

                lrp = xfs_btree_rec_addr(cur, lrecs, left);
                rrp = xfs_btree_rec_addr(cur, 1, right);

                xfs_btree_shift_recs(cur, rrp, 1, rrecs);

                /* Now put the new data in, and log it. */
                xfs_btree_copy_recs(cur, rrp, lrp, 1);
                xfs_btree_log_recs(cur, rbp, 1, rrecs + 1);
        }

        /*
         * Decrement and log left's numrecs, bump and log right's numrecs.
         */
        xfs_btree_set_numrecs(left, --lrecs);
        xfs_btree_log_block(cur, lbp, XFS_BB_NUMRECS);

        xfs_btree_set_numrecs(right, ++rrecs);
        xfs_btree_log_block(cur, rbp, XFS_BB_NUMRECS);

        /*
         * Using a temporary cursor, update the parent key values of the
         * block on the right.
         */
        error = xfs_btree_dup_cursor(cur, &tcur);
        if (error)
                goto error0;
        i = xfs_btree_lastrec(tcur, level);
        if (XFS_IS_CORRUPT(tcur->bc_mp, i != 1)) {
                error = -EFSCORRUPTED;
                goto error0;
        }

        error = xfs_btree_increment(tcur, level, &i);
        if (error)
                goto error1;

        /* Update the parent high keys of the left block, if needed. */
        if (cur->bc_flags & XFS_BTREE_OVERLAPPING) {
                error = xfs_btree_update_keys(cur, level);
                if (error)
                        goto error1;
        }

        /* Update the parent keys of the right block. */
        error = xfs_btree_update_keys(tcur, level);
        if (error)
                goto error1;

        xfs_btree_del_cursor(tcur, XFS_BTREE_NOERROR);

        *stat = 1;
        return 0;

out0:
        *stat = 0;
        return 0;

error0:
        return error;

error1:
        xfs_btree_del_cursor(tcur, XFS_BTREE_ERROR);
        return error;
}

/*
 * Split cur/level block in half.
 * Return new block number and the key to its first
 * record (to be inserted into parent).
 */
STATIC int                                      /* error */
__xfs_btree_split(
        struct xfs_btree_cur    *cur,
        int                     level,
        union xfs_btree_ptr     *ptrp,
        union xfs_btree_key     *key,
        struct xfs_btree_cur    **curp,
        int                     *stat)          /* success/failure */
{
        union xfs_btree_ptr     lptr;           /* left sibling block ptr */
        struct xfs_buf          *lbp;           /* left buffer pointer */
        struct xfs_btree_block  *left;          /* left btree block */
        union xfs_btree_ptr     rptr;           /* right sibling block ptr */
        struct xfs_buf          *rbp;           /* right buffer pointer */
        struct xfs_btree_block  *right;         /* right btree block */
        union xfs_btree_ptr     rrptr;          /* right-right sibling ptr */
        struct xfs_buf          *rrbp;          /* right-right buffer pointer */
        struct xfs_btree_block  *rrblock;       /* right-right btree block */
        int                     lrecs;
        int                     rrecs;
        int                     src_index;
        int                     error;          /* error return value */
        int                     i;

        XFS_BTREE_STATS_INC(cur, split);

        /* Set up left block (current one). */
        left = xfs_btree_get_block(cur, level, &lbp);

#ifdef DEBUG
        error = xfs_btree_check_block(cur, left, level, lbp);
        if (error)
                goto error0;
#endif

        xfs_btree_buf_to_ptr(cur, lbp, &lptr);

        /* Allocate the new block. If we can't do it, we're toast. Give up. */
        error = cur->bc_ops->alloc_block(cur, &lptr, &rptr, stat);
        if (error)
                goto error0;
        if (*stat == 0)
                goto out0;
        XFS_BTREE_STATS_INC(cur, alloc);

        /* Set up the new block as "right". */
        error = xfs_btree_get_buf_block(cur, &rptr, &right, &rbp);
        if (error)
                goto error0;

        /* Fill in the btree header for the new right block. */
        xfs_btree_init_block_cur(cur, rbp, xfs_btree_get_level(left), 0);

        /*
         * Split the entries between the old and the new block evenly.
         * Make sure that if there's an odd number of entries now, that
         * each new block will have the same number of entries.
         */
        lrecs = xfs_btree_get_numrecs(left);
        rrecs = lrecs / 2;
        if ((lrecs & 1) && cur->bc_ptrs[level] <= rrecs + 1)
                rrecs++;
        src_index = (lrecs - rrecs + 1);

        XFS_BTREE_STATS_ADD(cur, moves, rrecs);

        /* Adjust numrecs for the later get_*_keys() calls. */
        lrecs -= rrecs;
        xfs_btree_set_numrecs(left, lrecs);
        xfs_btree_set_numrecs(right, xfs_btree_get_numrecs(right) + rrecs);

        /*
         * Copy btree block entries from the left block over to the
         * new block, the right. Update the right block and log the
         * changes.
         */
        if (level > 0) {
                /* It's a non-leaf.  Move keys and pointers. */
                union xfs_btree_key     *lkp;   /* left btree key */
                union xfs_btree_ptr     *lpp;   /* left address pointer */
                union xfs_btree_key     *rkp;   /* right btree key */
                union xfs_btree_ptr     *rpp;   /* right address pointer */

                lkp = xfs_btree_key_addr(cur, src_index, left);
                lpp = xfs_btree_ptr_addr(cur, src_index, left);
                rkp = xfs_btree_key_addr(cur, 1, right);
                rpp = xfs_btree_ptr_addr(cur, 1, right);

                for (i = src_index; i < rrecs; i++) {
                        error = xfs_btree_debug_check_ptr(cur, lpp, i, level);
                        if (error)
                                goto error0;
                }

                /* Copy the keys & pointers to the new block. */
                xfs_btree_copy_keys(cur, rkp, lkp, rrecs);
                xfs_btree_copy_ptrs(cur, rpp, lpp, rrecs);

                xfs_btree_log_keys(cur, rbp, 1, rrecs);
                xfs_btree_log_ptrs(cur, rbp, 1, rrecs);

                /* Stash the keys of the new block for later insertion. */
                xfs_btree_get_node_keys(cur, right, key);
        } else {
                /* It's a leaf.  Move records.  */
                union xfs_btree_rec     *lrp;   /* left record pointer */
                union xfs_btree_rec     *rrp;   /* right record pointer */

                lrp = xfs_btree_rec_addr(cur, src_index, left);
                rrp = xfs_btree_rec_addr(cur, 1, right);

                /* Copy records to the new block. */
                xfs_btree_copy_recs(cur, rrp, lrp, rrecs);
                xfs_btree_log_recs(cur, rbp, 1, rrecs);

                /* Stash the keys of the new block for later insertion. */
                xfs_btree_get_leaf_keys(cur, right, key);
        }

        /*
         * Find the left block number by looking in the buffer.
         * Adjust sibling pointers.
         */
        xfs_btree_get_sibling(cur, left, &rrptr, XFS_BB_RIGHTSIB);
        xfs_btree_set_sibling(cur, right, &rrptr, XFS_BB_RIGHTSIB);
        xfs_btree_set_sibling(cur, right, &lptr, XFS_BB_LEFTSIB);
        xfs_btree_set_sibling(cur, left, &rptr, XFS_BB_RIGHTSIB);

        xfs_btree_log_block(cur, rbp, XFS_BB_ALL_BITS);
        xfs_btree_log_block(cur, lbp, XFS_BB_NUMRECS | XFS_BB_RIGHTSIB);

        /*
         * If there's a block to the new block's right, make that block
         * point back to right instead of to left.
         */
        if (!xfs_btree_ptr_is_null(cur, &rrptr)) {
                error = xfs_btree_read_buf_block(cur, &rrptr,
                                                        0, &rrblock, &rrbp);
                if (error)
                        goto error0;
                xfs_btree_set_sibling(cur, rrblock, &rptr, XFS_BB_LEFTSIB);
                xfs_btree_log_block(cur, rrbp, XFS_BB_LEFTSIB);
        }

        /* Update the parent high keys of the left block, if needed. */
        if (cur->bc_flags & XFS_BTREE_OVERLAPPING) {
                error = xfs_btree_update_keys(cur, level);
                if (error)
                        goto error0;
        }

        /*
         * If the cursor is really in the right block, move it there.
         * If it's just pointing past the last entry in left, then we'll
         * insert there, so don't change anything in that case.
         */
        if (cur->bc_ptrs[level] > lrecs + 1) {
                xfs_btree_setbuf(cur, level, rbp);
                cur->bc_ptrs[level] -= lrecs;
        }
        /*
         * If there are more levels, we'll need another cursor which refers
         * the right block, no matter where this cursor was.
         */
        if (level + 1 < cur->bc_nlevels) {
                error = xfs_btree_dup_cursor(cur, curp);
                if (error)
                        goto error0;
                (*curp)->bc_ptrs[level + 1]++;
        }
        *ptrp = rptr;
        *stat = 1;
        return 0;
out0:
        *stat = 0;
        return 0;

error0:
        return error;
}

struct xfs_btree_split_args {
        struct xfs_btree_cur    *cur;
        int                     level;
        union xfs_btree_ptr     *ptrp;
        union xfs_btree_key     *key;
        struct xfs_btree_cur    **curp;
        int                     *stat;          /* success/failure */
        int                     result;
        bool                    kswapd; /* allocation in kswapd context */
        struct completion       *done;
        struct work_struct      work;
};

/*
 * Stack switching interfaces for allocation
 */
static void
xfs_btree_split_worker(
        struct work_struct      *work)
{
        struct xfs_btree_split_args     *args = container_of(work,
                                                struct xfs_btree_split_args, work);
        unsigned long           pflags;
        unsigned long           new_pflags = 0;

        /*
         * we are in a transaction context here, but may also be doing work
         * in kswapd context, and hence we may need to inherit that state
         * temporarily to ensure that we don't block waiting for memory reclaim
         * in any way.
         */
        if (args->kswapd)
                new_pflags |= PF_MEMALLOC | PF_SWAPWRITE | PF_KSWAPD;

        current_set_flags_nested(&pflags, new_pflags);
        xfs_trans_set_context(args->cur->bc_tp);

        args->result = __xfs_btree_split(args->cur, args->level, args->ptrp,
                                         args->key, args->curp, args->stat);

        xfs_trans_clear_context(args->cur->bc_tp);
        current_restore_flags_nested(&pflags, new_pflags);

        /*
         * Do not access args after complete() has run here. We don't own args
         * and the owner may run and free args before we return here.
         */
        complete(args->done);

}

/*
 * BMBT split requests often come in with little stack to work on. Push
 * them off to a worker thread so there is lots of stack to use. For the other
 * btree types, just call directly to avoid the context switch overhead here.
 */
STATIC int                                      /* error */
xfs_btree_split(
        struct xfs_btree_cur    *cur,
        int                     level,
        union xfs_btree_ptr     *ptrp,
        union xfs_btree_key     *key,
        struct xfs_btree_cur    **curp,
        int                     *stat)          /* success/failure */
{
        struct xfs_btree_split_args     args;
        DECLARE_COMPLETION_ONSTACK(done);

        if (cur->bc_btnum != XFS_BTNUM_BMAP)
                return __xfs_btree_split(cur, level, ptrp, key, curp, stat);

        args.cur = cur;
        args.level = level;
        args.ptrp = ptrp;
        args.key = key;
        args.curp = curp;
        args.stat = stat;
        args.done = &done;
        args.kswapd = current_is_kswapd();
        INIT_WORK_ONSTACK(&args.work, xfs_btree_split_worker);
        queue_work(xfs_alloc_wq, &args.work);
        wait_for_completion(&done);
        destroy_work_on_stack(&args.work);
        return args.result;
}


/*
 * Copy the old inode root contents into a real block and make the
 * broot point to it.
 */
int                                             /* error */
xfs_btree_new_iroot(
        struct xfs_btree_cur    *cur,           /* btree cursor */
        int                     *logflags,      /* logging flags for inode */
        int                     *stat)          /* return status - 0 fail */
{
        struct xfs_buf          *cbp;           /* buffer for cblock */
        struct xfs_btree_block  *block;         /* btree block */
        struct xfs_btree_block  *cblock;        /* child btree block */
        union xfs_btree_key     *ckp;           /* child key pointer */
        union xfs_btree_ptr     *cpp;           /* child ptr pointer */
        union xfs_btree_key     *kp;            /* pointer to btree key */
        union xfs_btree_ptr     *pp;            /* pointer to block addr */
        union xfs_btree_ptr     nptr;           /* new block addr */
        int                     level;          /* btree level */
        int                     error;          /* error return code */
        int                     i;              /* loop counter */

        XFS_BTREE_STATS_INC(cur, newroot);

        ASSERT(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE);

        level = cur->bc_nlevels - 1;

        block = xfs_btree_get_iroot(cur);
        pp = xfs_btree_ptr_addr(cur, 1, block);

        /* Allocate the new block. If we can't do it, we're toast. Give up. */
        error = cur->bc_ops->alloc_block(cur, pp, &nptr, stat);
        if (error)
                goto error0;
        if (*stat == 0)
                return 0;

        XFS_BTREE_STATS_INC(cur, alloc);

        /* Copy the root into a real block. */
        error = xfs_btree_get_buf_block(cur, &nptr, &cblock, &cbp);
        if (error)
                goto error0;

        /*
         * we can't just memcpy() the root in for CRC enabled btree blocks.
         * In that case have to also ensure the blkno remains correct
         */
        memcpy(cblock, block, xfs_btree_block_len(cur));
        if (cur->bc_flags & XFS_BTREE_CRC_BLOCKS) {
                __be64 bno = cpu_to_be64(xfs_buf_daddr(cbp));
                if (cur->bc_flags & XFS_BTREE_LONG_PTRS)
                        cblock->bb_u.l.bb_blkno = bno;
                else
                        cblock->bb_u.s.bb_blkno = bno;
        }

        be16_add_cpu(&block->bb_level, 1);
        xfs_btree_set_numrecs(block, 1);
        cur->bc_nlevels++;
        cur->bc_ptrs[level + 1] = 1;

        kp = xfs_btree_key_addr(cur, 1, block);
        ckp = xfs_btree_key_addr(cur, 1, cblock);
        xfs_btree_copy_keys(cur, ckp, kp, xfs_btree_get_numrecs(cblock));

        cpp = xfs_btree_ptr_addr(cur, 1, cblock);
        for (i = 0; i < be16_to_cpu(cblock->bb_numrecs); i++) {
                error = xfs_btree_debug_check_ptr(cur, pp, i, level);
                if (error)
                        goto error0;
        }

        xfs_btree_copy_ptrs(cur, cpp, pp, xfs_btree_get_numrecs(cblock));

        error = xfs_btree_debug_check_ptr(cur, &nptr, 0, level);
        if (error)
                goto error0;

        xfs_btree_copy_ptrs(cur, pp, &nptr, 1);

        xfs_iroot_realloc(cur->bc_ino.ip,
                          1 - xfs_btree_get_numrecs(cblock),
                          cur->bc_ino.whichfork);

        xfs_btree_setbuf(cur, level, cbp);

        /*
         * Do all this logging at the end so that
         * the root is at the right level.
         */
        xfs_btree_log_block(cur, cbp, XFS_BB_ALL_BITS);
        xfs_btree_log_keys(cur, cbp, 1, be16_to_cpu(cblock->bb_numrecs));
        xfs_btree_log_ptrs(cur, cbp, 1, be16_to_cpu(cblock->bb_numrecs));

        *logflags |=
                XFS_ILOG_CORE | xfs_ilog_fbroot(cur->bc_ino.whichfork);
        *stat = 1;
        return 0;
error0:
        return error;
}

/*
 * Allocate a new root block, fill it in.
 */
STATIC int                              /* error */
xfs_btree_new_root(
        struct xfs_btree_cur    *cur,   /* btree cursor */
        int                     *stat)  /* success/failure */
{
        struct xfs_btree_block  *block; /* one half of the old root block */
        struct xfs_buf          *bp;    /* buffer containing block */
        int                     error;  /* error return value */
        struct xfs_buf          *lbp;   /* left buffer pointer */
        struct xfs_btree_block  *left;  /* left btree block */
        struct xfs_buf          *nbp;   /* new (root) buffer */
        struct xfs_btree_block  *new;   /* new (root) btree block */
        int                     nptr;   /* new value for key index, 1 or 2 */
        struct xfs_buf          *rbp;   /* right buffer pointer */
        struct xfs_btree_block  *right; /* right btree block */
        union xfs_btree_ptr     rptr;
        union xfs_btree_ptr     lptr;

        XFS_BTREE_STATS_INC(cur, newroot);