/*
 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
 * All Rights Reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write the Free Software Foundation,
 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
 */

#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_types.h"
#include "xfs_bit.h"
#include "xfs_log.h"
#include "xfs_inum.h"
#include "xfs_trans.h"
#include "xfs_sb.h"
#include "xfs_ag.h"
#include "xfs_dir2.h"
#include "xfs_dmapi.h"
#include "xfs_mount.h"
#include "xfs_da_btree.h"
#include "xfs_bmap_btree.h"
#include "xfs_alloc_btree.h"
#include "xfs_ialloc_btree.h"
#include "xfs_dir2_sf.h"
#include "xfs_attr_sf.h"
#include "xfs_dinode.h"
#include "xfs_inode.h"
#include "xfs_inode_item.h"
#include "xfs_itable.h"
#include "xfs_btree.h"
#include "xfs_ialloc.h"
#include "xfs_alloc.h"
#include "xfs_bmap.h"
#include "xfs_attr.h"
#include "xfs_rw.h"
#include "xfs_error.h"
#include "xfs_quota.h"
#include "xfs_utils.h"
#include "xfs_rtalloc.h"
#include "xfs_refcache.h"
#include "xfs_trans_space.h"
#include "xfs_log_priv.h"
#include "xfs_filestream.h"
#include "xfs_vnodeops.h"

int
xfs_open(
        xfs_inode_t     *ip)
{
        int             mode;

        if (XFS_FORCED_SHUTDOWN(ip->i_mount))
                return XFS_ERROR(EIO);

        /*
         * If it's a directory with any blocks, read-ahead block 0
         * as we're almost certain to have the next operation be a read there.
         */
        if (S_ISDIR(ip->i_d.di_mode) && ip->i_d.di_nextents > 0) {
                mode = xfs_ilock_map_shared(ip);
                if (ip->i_d.di_nextents > 0)
                        (void)xfs_da_reada_buf(NULL, ip, 0, XFS_DATA_FORK);
                xfs_iunlock(ip, mode);
        }
        return 0;
}

/*
 * xfs_getattr
 */
int
xfs_getattr(
        xfs_inode_t     *ip,
        bhv_vattr_t     *vap,
        int             flags)
{
        bhv_vnode_t     *vp = XFS_ITOV(ip);
        xfs_mount_t     *mp = ip->i_mount;

        vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        if (!(flags & ATTR_LAZY))
                xfs_ilock(ip, XFS_ILOCK_SHARED);

        vap->va_size = XFS_ISIZE(ip);
        if (vap->va_mask == XFS_AT_SIZE)
                goto all_done;

        vap->va_nblocks =
                XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
        vap->va_nodeid = ip->i_ino;
#if XFS_BIG_INUMS
        vap->va_nodeid += mp->m_inoadd;
#endif
        vap->va_nlink = ip->i_d.di_nlink;

        /*
         * Quick exit for non-stat callers
         */
        if ((vap->va_mask &
            ~(XFS_AT_SIZE|XFS_AT_FSID|XFS_AT_NODEID|
              XFS_AT_NLINK|XFS_AT_BLKSIZE)) == 0)
                goto all_done;

        /*
         * Copy from in-core inode.
         */
        vap->va_mode = ip->i_d.di_mode;
        vap->va_uid = ip->i_d.di_uid;
        vap->va_gid = ip->i_d.di_gid;
        vap->va_projid = ip->i_d.di_projid;

        /*
         * Check vnode type block/char vs. everything else.
         */
        switch (ip->i_d.di_mode & S_IFMT) {
        case S_IFBLK:
        case S_IFCHR:
                vap->va_rdev = ip->i_df.if_u2.if_rdev;
                vap->va_blocksize = BLKDEV_IOSIZE;
                break;
        default:
                vap->va_rdev = 0;

                if (!(ip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
                        vap->va_blocksize = xfs_preferred_iosize(mp);
                } else {

                        /*
                         * If the file blocks are being allocated from a
                         * realtime partition, then return the inode's
                         * realtime extent size or the realtime volume's
                         * extent size.
                         */
                        vap->va_blocksize =
                                xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
                }
                break;
        }

        vn_atime_to_timespec(vp, &vap->va_atime);
        vap->va_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
        vap->va_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
        vap->va_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
        vap->va_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;

        /*
         * Exit for stat callers.  See if any of the rest of the fields
         * to be filled in are needed.
         */
        if ((vap->va_mask &
             (XFS_AT_XFLAGS|XFS_AT_EXTSIZE|XFS_AT_NEXTENTS|XFS_AT_ANEXTENTS|
              XFS_AT_GENCOUNT|XFS_AT_VCODE)) == 0)
                goto all_done;

        /*
         * Convert di_flags to xflags.
         */
        vap->va_xflags = xfs_ip2xflags(ip);

        /*
         * Exit for inode revalidate.  See if any of the rest of
         * the fields to be filled in are needed.
         */
        if ((vap->va_mask &
             (XFS_AT_EXTSIZE|XFS_AT_NEXTENTS|XFS_AT_ANEXTENTS|
              XFS_AT_GENCOUNT|XFS_AT_VCODE)) == 0)
                goto all_done;

        vap->va_extsize = ip->i_d.di_extsize << mp->m_sb.sb_blocklog;
        vap->va_nextents =
                (ip->i_df.if_flags & XFS_IFEXTENTS) ?
                        ip->i_df.if_bytes / sizeof(xfs_bmbt_rec_t) :
                        ip->i_d.di_nextents;
        if (ip->i_afp)
                vap->va_anextents =
                        (ip->i_afp->if_flags & XFS_IFEXTENTS) ?
                                ip->i_afp->if_bytes / sizeof(xfs_bmbt_rec_t) :
                                 ip->i_d.di_anextents;
        else
                vap->va_anextents = 0;
        vap->va_gen = ip->i_d.di_gen;

 all_done:
        if (!(flags & ATTR_LAZY))
                xfs_iunlock(ip, XFS_ILOCK_SHARED);
        return 0;
}


/*
 * xfs_setattr
 */
int
xfs_setattr(
        xfs_inode_t             *ip,
        bhv_vattr_t             *vap,
        int                     flags,
        cred_t                  *credp)
{
        bhv_vnode_t             *vp = XFS_ITOV(ip);
        xfs_mount_t             *mp = ip->i_mount;
        xfs_trans_t             *tp;
        int                     mask;
        int                     code;
        uint                    lock_flags;
        uint                    commit_flags=0;
        uid_t                   uid=0, iuid=0;
        gid_t                   gid=0, igid=0;
        int                     timeflags = 0;
        xfs_prid_t              projid=0, iprojid=0;
        int                     mandlock_before, mandlock_after;
        struct xfs_dquot        *udqp, *gdqp, *olddquot1, *olddquot2;
        int                     file_owner;
        int                     need_iolock = 1;

        vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

        if (mp->m_flags & XFS_MOUNT_RDONLY)
                return XFS_ERROR(EROFS);

        /*
         * Cannot set certain attributes.
         */
        mask = vap->va_mask;
        if (mask & XFS_AT_NOSET) {
                return XFS_ERROR(EINVAL);
        }

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        /*
         * Timestamps do not need to be logged and hence do not
         * need to be done within a transaction.
         */
        if (mask & XFS_AT_UPDTIMES) {
                ASSERT((mask & ~XFS_AT_UPDTIMES) == 0);
                timeflags = ((mask & XFS_AT_UPDATIME) ? XFS_ICHGTIME_ACC : 0) |
                            ((mask & XFS_AT_UPDCTIME) ? XFS_ICHGTIME_CHG : 0) |
                            ((mask & XFS_AT_UPDMTIME) ? XFS_ICHGTIME_MOD : 0);
                xfs_ichgtime(ip, timeflags);
                return 0;
        }

        olddquot1 = olddquot2 = NULL;
        udqp = gdqp = NULL;

        /*
         * If disk quotas is on, we make sure that the dquots do exist on disk,
         * before we start any other transactions. Trying to do this later
         * is messy. We don't care to take a readlock to look at the ids
         * in inode here, because we can't hold it across the trans_reserve.
         * If the IDs do change before we take the ilock, we're covered
         * because the i_*dquot fields will get updated anyway.
         */
        if (XFS_IS_QUOTA_ON(mp) &&
            (mask & (XFS_AT_UID|XFS_AT_GID|XFS_AT_PROJID))) {
                uint    qflags = 0;

                if ((mask & XFS_AT_UID) && XFS_IS_UQUOTA_ON(mp)) {
                        uid = vap->va_uid;
                        qflags |= XFS_QMOPT_UQUOTA;
                } else {
                        uid = ip->i_d.di_uid;
                }
                if ((mask & XFS_AT_GID) && XFS_IS_GQUOTA_ON(mp)) {
                        gid = vap->va_gid;
                        qflags |= XFS_QMOPT_GQUOTA;
                }  else {
                        gid = ip->i_d.di_gid;
                }
                if ((mask & XFS_AT_PROJID) && XFS_IS_PQUOTA_ON(mp)) {
                        projid = vap->va_projid;
                        qflags |= XFS_QMOPT_PQUOTA;
                }  else {
                        projid = ip->i_d.di_projid;
                }
                /*
                 * We take a reference when we initialize udqp and gdqp,
                 * so it is important that we never blindly double trip on
                 * the same variable. See xfs_create() for an example.
                 */
                ASSERT(udqp == NULL);
                ASSERT(gdqp == NULL);
                code = XFS_QM_DQVOPALLOC(mp, ip, uid, gid, projid, qflags,
                                         &udqp, &gdqp);
                if (code)
                        return code;
        }

        /*
         * For the other attributes, we acquire the inode lock and
         * first do an error checking pass.
         */
        tp = NULL;
        lock_flags = XFS_ILOCK_EXCL;
        if (flags & ATTR_NOLOCK)
                need_iolock = 0;
        if (!(mask & XFS_AT_SIZE)) {
                if ((mask != (XFS_AT_CTIME|XFS_AT_ATIME|XFS_AT_MTIME)) ||
                    (mp->m_flags & XFS_MOUNT_WSYNC)) {
                        tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
                        commit_flags = 0;
                        if ((code = xfs_trans_reserve(tp, 0,
                                                     XFS_ICHANGE_LOG_RES(mp), 0,
                                                     0, 0))) {
                                lock_flags = 0;
                                goto error_return;
                        }
                }
        } else {
                if (DM_EVENT_ENABLED(ip, DM_EVENT_TRUNCATE) &&
                    !(flags & ATTR_DMI)) {
                        int dmflags = AT_DELAY_FLAG(flags) | DM_SEM_FLAG_WR;
                        code = XFS_SEND_DATA(mp, DM_EVENT_TRUNCATE, vp,
                                vap->va_size, 0, dmflags, NULL);
                        if (code) {
                                lock_flags = 0;
                                goto error_return;
                        }
                }
                if (need_iolock)
                        lock_flags |= XFS_IOLOCK_EXCL;
        }

        xfs_ilock(ip, lock_flags);

        /* boolean: are we the file owner? */
        file_owner = (current_fsuid(credp) == ip->i_d.di_uid);

        /*
         * Change various properties of a file.
         * Only the owner or users with CAP_FOWNER
         * capability may do these things.
         */
        if (mask &
            (XFS_AT_MODE|XFS_AT_XFLAGS|XFS_AT_EXTSIZE|XFS_AT_UID|
             XFS_AT_GID|XFS_AT_PROJID)) {
                /*
                 * CAP_FOWNER overrides the following restrictions:
                 *
                 * The user ID of the calling process must be equal
                 * to the file owner ID, except in cases where the
                 * CAP_FSETID capability is applicable.
                 */
                if (!file_owner && !capable(CAP_FOWNER)) {
                        code = XFS_ERROR(EPERM);
                        goto error_return;
                }

                /*
                 * CAP_FSETID overrides the following restrictions:
                 *
                 * The effective user ID of the calling process shall match
                 * the file owner when setting the set-user-ID and
                 * set-group-ID bits on that file.
                 *
                 * The effective group ID or one of the supplementary group
                 * IDs of the calling process shall match the group owner of
                 * the file when setting the set-group-ID bit on that file
                 */
                if (mask & XFS_AT_MODE) {
                        mode_t m = 0;

                        if ((vap->va_mode & S_ISUID) && !file_owner)
                                m |= S_ISUID;
                        if ((vap->va_mode & S_ISGID) &&
                            !in_group_p((gid_t)ip->i_d.di_gid))
                                m |= S_ISGID;
#if 0
                        /* Linux allows this, Irix doesn't. */
                        if ((vap->va_mode & S_ISVTX) && !VN_ISDIR(vp))
                                m |= S_ISVTX;
#endif
                        if (m && !capable(CAP_FSETID))
                                vap->va_mode &= ~m;
                }
        }

        /*
         * Change file ownership.  Must be the owner or privileged.
         * If the system was configured with the "restricted_chown"
         * option, the owner is not permitted to give away the file,
         * and can change the group id only to a group of which he
         * or she is a member.
         */
        if (mask & (XFS_AT_UID|XFS_AT_GID|XFS_AT_PROJID)) {
                /*
                 * These IDs could have changed since we last looked at them.
                 * But, we're assured that if the ownership did change
                 * while we didn't have the inode locked, inode's dquot(s)
                 * would have changed also.
                 */
                iuid = ip->i_d.di_uid;
                iprojid = ip->i_d.di_projid;
                igid = ip->i_d.di_gid;
                gid = (mask & XFS_AT_GID) ? vap->va_gid : igid;
                uid = (mask & XFS_AT_UID) ? vap->va_uid : iuid;
                projid = (mask & XFS_AT_PROJID) ? (xfs_prid_t)vap->va_projid :
                         iprojid;

                /*
                 * CAP_CHOWN overrides the following restrictions:
                 *
                 * If _POSIX_CHOWN_RESTRICTED is defined, this capability
                 * shall override the restriction that a process cannot
                 * change the user ID of a file it owns and the restriction
                 * that the group ID supplied to the chown() function
                 * shall be equal to either the group ID or one of the
                 * supplementary group IDs of the calling process.
                 */
                if (restricted_chown &&
                    (iuid != uid || (igid != gid &&
                                     !in_group_p((gid_t)gid))) &&
                    !capable(CAP_CHOWN)) {
                        code = XFS_ERROR(EPERM);
                        goto error_return;
                }
                /*
                 * Do a quota reservation only if uid/projid/gid is actually
                 * going to change.
                 */
                if ((XFS_IS_UQUOTA_ON(mp) && iuid != uid) ||
                    (XFS_IS_PQUOTA_ON(mp) && iprojid != projid) ||
                    (XFS_IS_GQUOTA_ON(mp) && igid != gid)) {
                        ASSERT(tp);
                        code = XFS_QM_DQVOPCHOWNRESV(mp, tp, ip, udqp, gdqp,
                                                capable(CAP_FOWNER) ?
                                                XFS_QMOPT_FORCE_RES : 0);
                        if (code)       /* out of quota */
                                goto error_return;
                }
        }

        /*
         * Truncate file.  Must have write permission and not be a directory.
         */
        if (mask & XFS_AT_SIZE) {
                /* Short circuit the truncate case for zero length files */
                if ((vap->va_size == 0) &&
                   (ip->i_size == 0) && (ip->i_d.di_nextents == 0)) {
                        xfs_iunlock(ip, XFS_ILOCK_EXCL);
                        lock_flags &= ~XFS_ILOCK_EXCL;
                        if (mask & XFS_AT_CTIME)
                                xfs_ichgtime(ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
                        code = 0;
                        goto error_return;
                }

                if (VN_ISDIR(vp)) {
                        code = XFS_ERROR(EISDIR);
                        goto error_return;
                } else if (!VN_ISREG(vp)) {
                        code = XFS_ERROR(EINVAL);
                        goto error_return;
                }
                /*
                 * Make sure that the dquots are attached to the inode.
                 */
                if ((code = XFS_QM_DQATTACH(mp, ip, XFS_QMOPT_ILOCKED)))
                        goto error_return;
        }

        /*
         * Change file access or modified times.
         */
        if (mask & (XFS_AT_ATIME|XFS_AT_MTIME)) {
                if (!file_owner) {
                        if ((flags & ATTR_UTIME) &&
                            !capable(CAP_FOWNER)) {
                                code = XFS_ERROR(EPERM);
                                goto error_return;
                        }
                }
        }

        /*
         * Change extent size or realtime flag.
         */
        if (mask & (XFS_AT_EXTSIZE|XFS_AT_XFLAGS)) {
                /*
                 * Can't change extent size if any extents are allocated.
                 */
                if (ip->i_d.di_nextents && (mask & XFS_AT_EXTSIZE) &&
                    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) !=
                     vap->va_extsize) ) {
                        code = XFS_ERROR(EINVAL);       /* EFBIG? */
                        goto error_return;
                }

                /*
                 * Can't change realtime flag if any extents are allocated.
                 */
                if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
                    (mask & XFS_AT_XFLAGS) &&
                    (ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
                    (vap->va_xflags & XFS_XFLAG_REALTIME)) {
                        code = XFS_ERROR(EINVAL);       /* EFBIG? */
                        goto error_return;
                }
                /*
                 * Extent size must be a multiple of the appropriate block
                 * size, if set at all.
                 */
                if ((mask & XFS_AT_EXTSIZE) && vap->va_extsize != 0) {
                        xfs_extlen_t    size;

                        if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) ||
                            ((mask & XFS_AT_XFLAGS) &&
                            (vap->va_xflags & XFS_XFLAG_REALTIME))) {
                                size = mp->m_sb.sb_rextsize <<
                                       mp->m_sb.sb_blocklog;
                        } else {
                                size = mp->m_sb.sb_blocksize;
                        }
                        if (vap->va_extsize % size) {
                                code = XFS_ERROR(EINVAL);
                                goto error_return;
                        }
                }
                /*
                 * If realtime flag is set then must have realtime data.
                 */
                if ((mask & XFS_AT_XFLAGS) &&
                    (vap->va_xflags & XFS_XFLAG_REALTIME)) {
                        if ((mp->m_sb.sb_rblocks == 0) ||
                            (mp->m_sb.sb_rextsize == 0) ||
                            (ip->i_d.di_extsize % mp->m_sb.sb_rextsize)) {
                                code = XFS_ERROR(EINVAL);
                                goto error_return;
                        }
                }

                /*
                 * Can't modify an immutable/append-only file unless
                 * we have appropriate permission.
                 */
                if ((mask & XFS_AT_XFLAGS) &&
                    (ip->i_d.di_flags &
                                (XFS_DIFLAG_IMMUTABLE|XFS_DIFLAG_APPEND) ||
                     (vap->va_xflags &
                                (XFS_XFLAG_IMMUTABLE | XFS_XFLAG_APPEND))) &&
                    !capable(CAP_LINUX_IMMUTABLE)) {
                        code = XFS_ERROR(EPERM);
                        goto error_return;
                }
        }

        /*
         * Now we can make the changes.  Before we join the inode
         * to the transaction, if XFS_AT_SIZE is set then take care of
         * the part of the truncation that must be done without the
         * inode lock.  This needs to be done before joining the inode
         * to the transaction, because the inode cannot be unlocked
         * once it is a part of the transaction.
         */
        if (mask & XFS_AT_SIZE) {
                code = 0;
                if ((vap->va_size > ip->i_size) &&
                    (flags & ATTR_NOSIZETOK) == 0) {
                        code = xfs_igrow_start(ip, vap->va_size, credp);
                }
                xfs_iunlock(ip, XFS_ILOCK_EXCL);

                /*
                 * We are going to log the inode size change in this
                 * transaction so any previous writes that are beyond the on
                 * disk EOF and the new EOF that have not been written out need
                 * to be written here. If we do not write the data out, we
                 * expose ourselves to the null files problem.
                 *
                 * Only flush from the on disk size to the smaller of the in
                 * memory file size or the new size as that's the range we
                 * really care about here and prevents waiting for other data
                 * not within the range we care about here.
                 */
                if (!code &&
                    (ip->i_size != ip->i_d.di_size) &&
                    (vap->va_size > ip->i_d.di_size)) {
                        code = xfs_flush_pages(ip,
                                        ip->i_d.di_size, vap->va_size,
                                        XFS_B_ASYNC, FI_NONE);
                }

                /* wait for all I/O to complete */
                vn_iowait(ip);

                if (!code)
                        code = xfs_itruncate_data(ip, vap->va_size);
                if (code) {
                        ASSERT(tp == NULL);
                        lock_flags &= ~XFS_ILOCK_EXCL;
                        ASSERT(lock_flags == XFS_IOLOCK_EXCL);
                        goto error_return;
                }
                tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_SIZE);
                if ((code = xfs_trans_reserve(tp, 0,
                                             XFS_ITRUNCATE_LOG_RES(mp), 0,
                                             XFS_TRANS_PERM_LOG_RES,
                                             XFS_ITRUNCATE_LOG_COUNT))) {
                        xfs_trans_cancel(tp, 0);
                        if (need_iolock)
                                xfs_iunlock(ip, XFS_IOLOCK_EXCL);
                        return code;
                }
                commit_flags = XFS_TRANS_RELEASE_LOG_RES;
                xfs_ilock(ip, XFS_ILOCK_EXCL);
        }

        if (tp) {
                xfs_trans_ijoin(tp, ip, lock_flags);
                xfs_trans_ihold(tp, ip);
        }

        /* determine whether mandatory locking mode changes */
        mandlock_before = MANDLOCK(vp, ip->i_d.di_mode);

        /*
         * Truncate file.  Must have write permission and not be a directory.
         */
        if (mask & XFS_AT_SIZE) {
                if (vap->va_size > ip->i_size) {
                        xfs_igrow_finish(tp, ip, vap->va_size,
                            !(flags & ATTR_DMI));
                } else if ((vap->va_size <= ip->i_size) ||
                           ((vap->va_size == 0) && ip->i_d.di_nextents)) {
                        /*
                         * signal a sync transaction unless
                         * we're truncating an already unlinked
                         * file on a wsync filesystem
                         */
                        code = xfs_itruncate_finish(&tp, ip,
                                            (xfs_fsize_t)vap->va_size,
                                            XFS_DATA_FORK,
                                            ((ip->i_d.di_nlink != 0 ||
                                              !(mp->m_flags & XFS_MOUNT_WSYNC))
                                             ? 1 : 0));
                        if (code)
                                goto abort_return;
                        /*
                         * Truncated "down", so we're removing references
                         * to old data here - if we now delay flushing for
                         * a long time, we expose ourselves unduly to the
                         * notorious NULL files problem.  So, we mark this
                         * vnode and flush it when the file is closed, and
                         * do not wait the usual (long) time for writeout.
                         */
                        xfs_iflags_set(ip, XFS_ITRUNCATED);
                }
                /*
                 * Have to do this even if the file's size doesn't change.
                 */
                timeflags |= XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG;
        }

        /*
         * Change file access modes.
         */
        if (mask & XFS_AT_MODE) {
                ip->i_d.di_mode &= S_IFMT;
                ip->i_d.di_mode |= vap->va_mode & ~S_IFMT;

                xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
                timeflags |= XFS_ICHGTIME_CHG;
        }

        /*
         * Change file ownership.  Must be the owner or privileged.
         * If the system was configured with the "restricted_chown"
         * option, the owner is not permitted to give away the file,
         * and can change the group id only to a group of which he
         * or she is a member.
         */
        if (mask & (XFS_AT_UID|XFS_AT_GID|XFS_AT_PROJID)) {
                /*
                 * CAP_FSETID overrides the following restrictions:
                 *
                 * The set-user-ID and set-group-ID bits of a file will be
                 * cleared upon successful return from chown()
                 */
                if ((ip->i_d.di_mode & (S_ISUID|S_ISGID)) &&
                    !capable(CAP_FSETID)) {
                        ip->i_d.di_mode &= ~(S_ISUID|S_ISGID);
                }

                /*
                 * Change the ownerships and register quota modifications
                 * in the transaction.
                 */
                if (iuid != uid) {
                        if (XFS_IS_UQUOTA_ON(mp)) {
                                ASSERT(mask & XFS_AT_UID);
                                ASSERT(udqp);
                                olddquot1 = XFS_QM_DQVOPCHOWN(mp, tp, ip,
                                                        &ip->i_udquot, udqp);
                        }
                        ip->i_d.di_uid = uid;
                }
                if (igid != gid) {
                        if (XFS_IS_GQUOTA_ON(mp)) {
                                ASSERT(!XFS_IS_PQUOTA_ON(mp));
                                ASSERT(mask & XFS_AT_GID);
                                ASSERT(gdqp);
                                olddquot2 = XFS_QM_DQVOPCHOWN(mp, tp, ip,
                                                        &ip->i_gdquot, gdqp);
                        }
                        ip->i_d.di_gid = gid;
                }
                if (iprojid != projid) {
                        if (XFS_IS_PQUOTA_ON(mp)) {
                                ASSERT(!XFS_IS_GQUOTA_ON(mp));
                                ASSERT(mask & XFS_AT_PROJID);
                                ASSERT(gdqp);
                                olddquot2 = XFS_QM_DQVOPCHOWN(mp, tp, ip,
                                                        &ip->i_gdquot, gdqp);
                        }
                        ip->i_d.di_projid = projid;
                        /*
                         * We may have to rev the inode as well as
                         * the superblock version number since projids didn't
                         * exist before DINODE_VERSION_2 and SB_VERSION_NLINK.
                         */
                        if (ip->i_d.di_version == XFS_DINODE_VERSION_1)
                                xfs_bump_ino_vers2(tp, ip);
                }

                xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
                timeflags |= XFS_ICHGTIME_CHG;
        }


        /*
         * Change file access or modified times.
         */
        if (mask & (XFS_AT_ATIME|XFS_AT_MTIME)) {
                if (mask & XFS_AT_ATIME) {
                        ip->i_d.di_atime.t_sec = vap->va_atime.tv_sec;
                        ip->i_d.di_atime.t_nsec = vap->va_atime.tv_nsec;
                        ip->i_update_core = 1;
                        timeflags &= ~XFS_ICHGTIME_ACC;
                }
                if (mask & XFS_AT_MTIME) {
                        ip->i_d.di_mtime.t_sec = vap->va_mtime.tv_sec;
                        ip->i_d.di_mtime.t_nsec = vap->va_mtime.tv_nsec;
                        timeflags &= ~XFS_ICHGTIME_MOD;
                        timeflags |= XFS_ICHGTIME_CHG;
                }
                if (tp && (flags & ATTR_UTIME))
                        xfs_trans_log_inode (tp, ip, XFS_ILOG_CORE);
        }

        /*
         * Change XFS-added attributes.
         */
        if (mask & (XFS_AT_EXTSIZE|XFS_AT_XFLAGS)) {
                if (mask & XFS_AT_EXTSIZE) {
                        /*
                         * Converting bytes to fs blocks.
                         */
                        ip->i_d.di_extsize = vap->va_extsize >>
                                mp->m_sb.sb_blocklog;
                }
                if (mask & XFS_AT_XFLAGS) {
                        uint    di_flags;

                        /* can't set PREALLOC this way, just preserve it */
                        di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
                        if (vap->va_xflags & XFS_XFLAG_IMMUTABLE)
                                di_flags |= XFS_DIFLAG_IMMUTABLE;
                        if (vap->va_xflags & XFS_XFLAG_APPEND)
                                di_flags |= XFS_DIFLAG_APPEND;
                        if (vap->va_xflags & XFS_XFLAG_SYNC)
                                di_flags |= XFS_DIFLAG_SYNC;
                        if (vap->va_xflags & XFS_XFLAG_NOATIME)
                                di_flags |= XFS_DIFLAG_NOATIME;
                        if (vap->va_xflags & XFS_XFLAG_NODUMP)
                                di_flags |= XFS_DIFLAG_NODUMP;
                        if (vap->va_xflags & XFS_XFLAG_PROJINHERIT)
                                di_flags |= XFS_DIFLAG_PROJINHERIT;
                        if (vap->va_xflags & XFS_XFLAG_NODEFRAG)
                                di_flags |= XFS_DIFLAG_NODEFRAG;
                        if (vap->va_xflags & XFS_XFLAG_FILESTREAM)
                                di_flags |= XFS_DIFLAG_FILESTREAM;
                        if ((ip->i_d.di_mode & S_IFMT) == S_IFDIR) {
                                if (vap->va_xflags & XFS_XFLAG_RTINHERIT)
                                        di_flags |= XFS_DIFLAG_RTINHERIT;
                                if (vap->va_xflags & XFS_XFLAG_NOSYMLINKS)
                                        di_flags |= XFS_DIFLAG_NOSYMLINKS;
                                if (vap->va_xflags & XFS_XFLAG_EXTSZINHERIT)
                                        di_flags |= XFS_DIFLAG_EXTSZINHERIT;
                        } else if ((ip->i_d.di_mode & S_IFMT) == S_IFREG) {
                                if (vap->va_xflags & XFS_XFLAG_REALTIME) {
                                        di_flags |= XFS_DIFLAG_REALTIME;
                                        ip->i_iocore.io_flags |= XFS_IOCORE_RT;
                                } else {
                                        ip->i_iocore.io_flags &= ~XFS_IOCORE_RT;
                                }
                                if (vap->va_xflags & XFS_XFLAG_EXTSIZE)
                                        di_flags |= XFS_DIFLAG_EXTSIZE;
                        }
                        ip->i_d.di_flags = di_flags;
                }
                xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
                timeflags |= XFS_ICHGTIME_CHG;
        }

        /*
         * Change file inode change time only if XFS_AT_CTIME set
         * AND we have been called by a DMI function.
         */

        if ( (flags & ATTR_DMI) && (mask & XFS_AT_CTIME) ) {
                ip->i_d.di_ctime.t_sec = vap->va_ctime.tv_sec;
                ip->i_d.di_ctime.t_nsec = vap->va_ctime.tv_nsec;
                ip->i_update_core = 1;
                timeflags &= ~XFS_ICHGTIME_CHG;
        }

        /*
         * Send out timestamp changes that need to be set to the
         * current time.  Not done when called by a DMI function.
         */
        if (timeflags && !(flags & ATTR_DMI))
                xfs_ichgtime(ip, timeflags);

        XFS_STATS_INC(xs_ig_attrchg);

        /*
         * If this is a synchronous mount, make sure that the
         * transaction goes to disk before returning to the user.
         * This is slightly sub-optimal in that truncates require
         * two sync transactions instead of one for wsync filesystems.
         * One for the truncate and one for the timestamps since we
         * don't want to change the timestamps unless we're sure the
         * truncate worked.  Truncates are less than 1% of the laddis
         * mix so this probably isn't worth the trouble to optimize.
         */
        code = 0;
        if (tp) {
                if (mp->m_flags & XFS_MOUNT_WSYNC)
                        xfs_trans_set_sync(tp);

                code = xfs_trans_commit(tp, commit_flags);
        }

        /*
         * If the (regular) file's mandatory locking mode changed, then
         * notify the vnode.  We do this under the inode lock to prevent
         * racing calls to vop_vnode_change.
         */
        mandlock_after = MANDLOCK(vp, ip->i_d.di_mode);

        xfs_iunlock(ip, lock_flags);

        /*
         * Release any dquot(s) the inode had kept before chown.
         */
        XFS_QM_DQRELE(mp, olddquot1);
        XFS_QM_DQRELE(mp, olddquot2);
        XFS_QM_DQRELE(mp, udqp);
        XFS_QM_DQRELE(mp, gdqp);

        if (code) {
                return code;
        }

        if (DM_EVENT_ENABLED(ip, DM_EVENT_ATTRIBUTE) &&
            !(flags & ATTR_DMI)) {
                (void) XFS_SEND_NAMESP(mp, DM_EVENT_ATTRIBUTE, vp, DM_RIGHT_NULL,
                                        NULL, DM_RIGHT_NULL, NULL, NULL,
                                        0, 0, AT_DELAY_FLAG(flags));
        }
        return 0;

 abort_return:
        commit_flags |= XFS_TRANS_ABORT;
        /* FALLTHROUGH */
 error_return:
        XFS_QM_DQRELE(mp, udqp);
        XFS_QM_DQRELE(mp, gdqp);
        if (tp) {
                xfs_trans_cancel(tp, commit_flags);
        }
        if (lock_flags != 0) {
                xfs_iunlock(ip, lock_flags);
        }
        return code;
}


/*
 * xfs_access
 * Null conversion from vnode mode bits to inode mode bits, as in efs.
 */
int
xfs_access(
        xfs_inode_t     *ip,
        int             mode,
        cred_t          *credp)
{
        int             error;

        vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

        xfs_ilock(ip, XFS_ILOCK_SHARED);
        error = xfs_iaccess(ip, mode, credp);
        xfs_iunlock(ip, XFS_ILOCK_SHARED);
        return error;
}


/*
 * The maximum pathlen is 1024 bytes. Since the minimum file system
 * blocksize is 512 bytes, we can get a max of 2 extents back from
 * bmapi.
 */
#define SYMLINK_MAPS 2

STATIC int
xfs_readlink_bmap(
        xfs_inode_t     *ip,
        char            *link)
{
        xfs_mount_t     *mp = ip->i_mount;
        int             pathlen = ip->i_d.di_size;
        int             nmaps = SYMLINK_MAPS;
        xfs_bmbt_irec_t mval[SYMLINK_MAPS];
        xfs_daddr_t     d;
        int             byte_cnt;
        int             n;
        xfs_buf_t       *bp;
        int             error = 0;

        error = xfs_bmapi(NULL, ip, 0, XFS_B_TO_FSB(mp, pathlen), 0, NULL, 0,
                        mval, &nmaps, NULL, NULL);
        if (error)
                goto out;

        for (n = 0; n < nmaps; n++) {
                d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
                byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);

                bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0);
                error = XFS_BUF_GETERROR(bp);
                if (error) {
                        xfs_ioerror_alert("xfs_readlink",
                                  ip->i_mount, bp, XFS_BUF_ADDR(bp));
                        xfs_buf_relse(bp);
                        goto out;
                }
                if (pathlen < byte_cnt)
                        byte_cnt = pathlen;
                pathlen -= byte_cnt;

                memcpy(link, XFS_BUF_PTR(bp), byte_cnt);
                xfs_buf_relse(bp);
        }

        link[ip->i_d.di_size] = '\0';
        error = 0;

 out:
        return error;
}

int
xfs_readlink(
        xfs_inode_t     *ip,
        char            *link)
{
        xfs_mount_t     *mp = ip->i_mount;
        int             pathlen;
        int             error = 0;

        vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        xfs_ilock(ip, XFS_ILOCK_SHARED);

        ASSERT((ip->i_d.di_mode & S_IFMT) == S_IFLNK);
        ASSERT(ip->i_d.di_size <= MAXPATHLEN);

        pathlen = ip->i_d.di_size;

        if (!pathlen)
                goto out;

        if (ip->i_df.if_flags & XFS_IFINLINE) {
                memcpy(link, ip->i_df.if_u1.if_data, pathlen);
                link[pathlen] = '\0';
        } else {
                error = xfs_readlink_bmap(ip, link);
        }

 out:
        xfs_iunlock(ip, XFS_ILOCK_SHARED);
        return error;
}

/*
 * xfs_fsync
 *
 * This is called to sync the inode and its data out to disk.
 * We need to hold the I/O lock while flushing the data, and
 * the inode lock while flushing the inode.  The inode lock CANNOT
 * be held while flushing the data, so acquire after we're done
 * with that.
 */
int
xfs_fsync(
        xfs_inode_t     *ip,
        int             flag,
        xfs_off_t       start,
        xfs_off_t       stop)
{
        xfs_trans_t     *tp;
        int             error;
        int             log_flushed = 0, changed = 1;

        vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

        ASSERT(start >= 0 && stop >= -1);

        if (XFS_FORCED_SHUTDOWN(ip->i_mount))
                return XFS_ERROR(EIO);

        if (flag & FSYNC_DATA)
                filemap_fdatawait(vn_to_inode(XFS_ITOV(ip))->i_mapping);

        /*
         * We always need to make sure that the required inode state
         * is safe on disk.  The vnode might be clean but because
         * of committed transactions that haven't hit the disk yet.
         * Likewise, there could be unflushed non-transactional
         * changes to the inode core that have to go to disk.
         *
         * The following code depends on one assumption:  that
         * any transaction that changes an inode logs the core
         * because it has to change some field in the inode core
         * (typically nextents or nblocks).  That assumption
         * implies that any transactions against an inode will
         * catch any non-transactional updates.  If inode-altering
         * transactions exist that violate this assumption, the
         * code breaks.  Right now, it figures that if the involved
         * update_* field is clear and the inode is unpinned, the
         * inode is clean.  Either it's been flushed or it's been
         * committed and the commit has hit the disk unpinning the inode.
         * (Note that xfs_inode_item_format() called at commit clears
         * the update_* fields.)
         */
        xfs_ilock(ip, XFS_ILOCK_SHARED);

        /* If we are flushing data then we care about update_size
         * being set, otherwise we care about update_core
         */
        if ((flag & FSYNC_DATA) ?
                        (ip->i_update_size == 0) :
                        (ip->i_update_core == 0)) {
                /*
                 * Timestamps/size haven't changed since last inode
                 * flush or inode transaction commit.  That means
                 * either nothing got written or a transaction
                 * committed which caught the updates.  If the
                 * latter happened and the transaction hasn't
                 * hit the disk yet, the inode will be still
                 * be pinned.  If it is, force the log.
                 */

                xfs_iunlock(ip, XFS_ILOCK_SHARED);

                if (xfs_ipincount(ip)) {
                        _xfs_log_force(ip->i_mount, (xfs_lsn_t)0,
                                      XFS_LOG_FORCE |
                                      ((flag & FSYNC_WAIT)
                                       ? XFS_LOG_SYNC : 0),
                                      &log_flushed);
                } else {
                        /*
                         * If the inode is not pinned and nothing
                         * has changed we don't need to flush the
                         * cache.
                         */
                        changed = 0;
                }
                error = 0;
        } else  {
                /*
                 * Kick off a transaction to log the inode
                 * core to get the updates.  Make it
                 * sync if FSYNC_WAIT is passed in (which
                 * is done by everybody but specfs).  The
                 * sync transaction will also force the log.
                 */
                xfs_iunlock(ip, XFS_ILOCK_SHARED);
                tp = xfs_trans_alloc(ip->i_mount, XFS_TRANS_FSYNC_TS);
                if ((error = xfs_trans_reserve(tp, 0,
                                XFS_FSYNC_TS_LOG_RES(ip->i_mount),
                                0, 0, 0)))  {
                        xfs_trans_cancel(tp, 0);
                        return error;
                }
                xfs_ilock(ip, XFS_ILOCK_EXCL);

                /*
                 * Note - it's possible that we might have pushed
                 * ourselves out of the way during trans_reserve
                 * which would flush the inode.  But there's no
                 * guarantee that the inode buffer has actually
                 * gone out yet (it's delwri).  Plus the buffer
                 * could be pinned anyway if it's part of an
                 * inode in another recent transaction.  So we
                 * play it safe and fire off the transaction anyway.
                 */
                xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
                xfs_trans_ihold(tp, ip);
                xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
                if (flag & FSYNC_WAIT)
                        xfs_trans_set_sync(tp);
                error = _xfs_trans_commit(tp, 0, &log_flushed);

                xfs_iunlock(ip, XFS_ILOCK_EXCL);
        }

        if ((ip->i_mount->m_flags & XFS_MOUNT_BARRIER) && changed) {
                /*
                 * If the log write didn't issue an ordered tag we need
                 * to flush the disk cache for the data device now.
                 */
                if (!log_flushed)
                        xfs_blkdev_issue_flush(ip->i_mount->m_ddev_targp);

                /*
                 * If this inode is on the RT dev we need to flush that
                 * cache as well.
                 */
                if (ip->i_d.di_flags & XFS_DIFLAG_REALTIME)
                        xfs_blkdev_issue_flush(ip->i_mount->m_rtdev_targp);
        }

        return error;
}

/*
 * This is called by xfs_inactive to free any blocks beyond eof
 * when the link count isn't zero and by xfs_dm_punch_hole() when
 * punching a hole to EOF.
 */
int
xfs_free_eofblocks(
        xfs_mount_t     *mp,
        xfs_inode_t     *ip,
        int             flags)
{
        xfs_trans_t     *tp;
        int             error;
        xfs_fileoff_t   end_fsb;
        xfs_fileoff_t   last_fsb;
        xfs_filblks_t   map_len;
        int             nimaps;
        xfs_bmbt_irec_t imap;
        int             use_iolock = (flags & XFS_FREE_EOF_LOCK);

        /*
         * Figure out if there are any blocks beyond the end
         * of the file.  If not, then there is nothing to do.
         */
        end_fsb = XFS_B_TO_FSB(mp, ((xfs_ufsize_t)ip->i_size));
        last_fsb = XFS_B_TO_FSB(mp, (xfs_ufsize_t)XFS_MAXIOFFSET(mp));
        map_len = last_fsb - end_fsb;
        if (map_len <= 0)
                return 0;

        nimaps = 1;
        xfs_ilock(ip, XFS_ILOCK_SHARED);
        error = XFS_BMAPI(mp, NULL, &ip->i_iocore, end_fsb, map_len, 0,
                          NULL, 0, &imap, &nimaps, NULL, NULL);
        xfs_iunlock(ip, XFS_ILOCK_SHARED);

        if (!error && (nimaps != 0) &&
            (imap.br_startblock != HOLESTARTBLOCK ||
             ip->i_delayed_blks)) {
                /*
                 * Attach the dquots to the inode up front.
                 */
                if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
                        return error;

                /*
                 * There are blocks after the end of file.
                 * Free them up now by truncating the file to
                 * its current size.
                 */
                tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);

                /*
                 * Do the xfs_itruncate_start() call before
                 * reserving any log space because
                 * itruncate_start will call into the buffer
                 * cache and we can't
                 * do that within a transaction.
                 */
                if (use_iolock)
                        xfs_ilock(ip, XFS_IOLOCK_EXCL);
                error = xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE,
                                    ip->i_size);
                if (error) {
                        xfs_trans_cancel(tp, 0);
                        if (use_iolock)
                                xfs_iunlock(ip, XFS_IOLOCK_EXCL);
                        return error;
                }

                error = xfs_trans_reserve(tp, 0,
                                          XFS_ITRUNCATE_LOG_RES(mp),
                                          0, XFS_TRANS_PERM_LOG_RES,
                                          XFS_ITRUNCATE_LOG_COUNT);
                if (error) {
                        ASSERT(XFS_FORCED_SHUTDOWN(mp));
                        xfs_trans_cancel(tp, 0);
                        xfs_iunlock(ip, XFS_IOLOCK_EXCL);
                        return error;
                }

                xfs_ilock(ip, XFS_ILOCK_EXCL);
                xfs_trans_ijoin(tp, ip,
                                XFS_IOLOCK_EXCL |
                                XFS_ILOCK_EXCL);
                xfs_trans_ihold(tp, ip);

                error = xfs_itruncate_finish(&tp, ip,
                                             ip->i_size,
                                             XFS_DATA_FORK,
                                             0);
                /*
                 * If we get an error at this point we
                 * simply don't bother truncating the file.
                 */
                if (error) {
                        xfs_trans_cancel(tp,
                                         (XFS_TRANS_RELEASE_LOG_RES |
                                          XFS_TRANS_ABORT));
                } else {
                        error = xfs_trans_commit(tp,
                                                XFS_TRANS_RELEASE_LOG_RES);
                }
                xfs_iunlock(ip, (use_iolock ? (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)
                                            : XFS_ILOCK_EXCL));
        }
        return error;
}

/*
 * Free a symlink that has blocks associated with it.
 */
STATIC int
xfs_inactive_symlink_rmt(
        xfs_inode_t     *ip,
        xfs_trans_t     **tpp)
{
        xfs_buf_t       *bp;
        int             committed;
        int             done;
        int             error;
        xfs_fsblock_t   first_block;
        xfs_bmap_free_t free_list;
        int             i;
        xfs_mount_t     *mp;
        xfs_bmbt_irec_t mval[SYMLINK_MAPS];
        int             nmaps;
        xfs_trans_t     *ntp;
        int             size;
        xfs_trans_t     *tp;

        tp = *tpp;
        mp = ip->i_mount;
        ASSERT(ip->i_d.di_size > XFS_IFORK_DSIZE(ip));
        /*
         * We're freeing a symlink that has some
         * blocks allocated to it.  Free the
         * blocks here.  We know that we've got
         * either 1 or 2 extents and that we can
         * free them all in one bunmapi call.
         */
        ASSERT(ip->i_d.di_nextents > 0 && ip->i_d.di_nextents <= 2);
        if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
                        XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
                ASSERT(XFS_FORCED_SHUTDOWN(mp));
                xfs_trans_cancel(tp, 0);
                *tpp = NULL;
                return error;
        }
        /*
         * Lock the inode, fix the size, and join it to the transaction.
         * Hold it so in the normal path, we still have it locked for
         * the second transaction.  In the error paths we need it
         * held so the cancel won't rele it, see below.
         */
        xfs_ilock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
        size = (int)ip->i_d.di_size;
        ip->i_d.di_size = 0;
        xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
        xfs_trans_ihold(tp, ip);
        xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
        /*
         * Find the block(s) so we can inval and unmap them.
         */
        done = 0;
        XFS_BMAP_INIT(&free_list, &first_block);
        nmaps = ARRAY_SIZE(mval);
        if ((error = xfs_bmapi(tp, ip, 0, XFS_B_TO_FSB(mp, size),
                        XFS_BMAPI_METADATA, &first_block, 0, mval, &nmaps,
                        &free_list, NULL)))
                goto error0;
        /*
         * Invalidate the block(s).
         */
        for (i = 0; i < nmaps; i++) {
                bp = xfs_trans_get_buf(tp, mp->m_ddev_targp,
                        XFS_FSB_TO_DADDR(mp, mval[i].br_startblock),
                        XFS_FSB_TO_BB(mp, mval[i].br_blockcount), 0);
                xfs_trans_binval(tp, bp);
        }
        /*
         * Unmap the dead block(s) to the free_list.
         */
        if ((error = xfs_bunmapi(tp, ip, 0, size, XFS_BMAPI_METADATA, nmaps,
                        &first_block, &free_list, NULL, &done)))
                goto error1;
        ASSERT(done);
        /*
         * Commit the first transaction.  This logs the EFI and the inode.
         */
        if ((error = xfs_bmap_finish(&tp, &free_list, &committed)))
                goto error1;
        /*
         * The transaction must have been committed, since there were
         * actually extents freed by xfs_bunmapi.  See xfs_bmap_finish.
         * The new tp has the extent freeing and EFDs.
         */
        ASSERT(committed);
        /*
         * The first xact was committed, so add the inode to the new one.
         * Mark it dirty so it will be logged and moved forward in the log as
         * part of every commit.
         */
        xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
        xfs_trans_ihold(tp, ip);
        xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
        /*
         * Get a new, empty transaction to return to our caller.
         */
        ntp = xfs_trans_dup(tp);
        /*
         * Commit the transaction containing extent freeing and EFDs.
         * If we get an error on the commit here or on the reserve below,
         * we need to unlock the inode since the new transaction doesn't
         * have the inode attached.
         */
        error = xfs_trans_commit(tp, 0);
        tp = ntp;
        if (error) {
                ASSERT(XFS_FORCED_SHUTDOWN(mp));
                goto error0;
        }
        /*
         * Remove the memory for extent descriptions (just bookkeeping).
         */
        if (ip->i_df.if_bytes)
                xfs_idata_realloc(ip, -ip->i_df.if_bytes, XFS_DATA_FORK);
        ASSERT(ip->i_df.if_bytes == 0);
        /*
         * Put an itruncate log reservation in the new transaction
         * for our caller.
         */
        if ((error = xfs_trans_reserve(tp, 0, XFS_ITRUNCATE_LOG_RES(mp), 0,
                        XFS_TRANS_PERM_LOG_RES, XFS_ITRUNCATE_LOG_COUNT))) {
                ASSERT(XFS_FORCED_SHUTDOWN(mp));
                goto error0;
        }
        /*
         * Return with the inode locked but not joined to the transaction.
         */
        *tpp = tp;
        return 0;

 error1:
        xfs_bmap_cancel(&free_list);
 error0:
        /*
         * Have to come here with the inode locked and either
         * (held and in the transaction) or (not in the transaction).
         * If the inode isn't held then cancel would iput it, but
         * that's wrong since this is inactive and the vnode ref
         * count is 0 already.
         * Cancel won't do anything to the inode if held, but it still
         * needs to be locked until the cancel is done, if it was
         * joined to the transaction.
         */
        xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
        xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
        *tpp = NULL;
        return error;

}

STATIC int
xfs_inactive_symlink_local(
        xfs_inode_t     *ip,
        xfs_trans_t     **tpp)
{
        int             error;

        ASSERT(ip->i_d.di_size <= XFS_IFORK_DSIZE(ip));
        /*
         * We're freeing a symlink which fit into
         * the inode.  Just free the memory used
         * to hold the old symlink.
         */
        error = xfs_trans_reserve(*tpp, 0,
                                  XFS_ITRUNCATE_LOG_RES(ip->i_mount),
                                  0, XFS_TRANS_PERM_LOG_RES,
                                  XFS_ITRUNCATE_LOG_COUNT);

        if (error) {
                xfs_trans_cancel(*tpp, 0);
                *tpp = NULL;
                return error;
        }
        xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);

        /*
         * Zero length symlinks _can_ exist.
         */
        if (ip->i_df.if_bytes > 0) {
                xfs_idata_realloc(ip,
                                  -(ip->i_df.if_bytes),
                                  XFS_DATA_FORK);
                ASSERT(ip->i_df.if_bytes == 0);
        }
        return 0;
}

STATIC int
xfs_inactive_attrs(
        xfs_inode_t     *ip,
        xfs_trans_t     **tpp)
{
        xfs_trans_t     *tp;
        int             error;
        xfs_mount_t     *mp;

        ASSERT(ismrlocked(&ip->i_iolock, MR_UPDATE));
        tp = *tpp;
        mp = ip->i_mount;
        ASSERT(ip->i_d.di_forkoff != 0);
        xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
        xfs_iunlock(ip, XFS_ILOCK_EXCL);

        error = xfs_attr_inactive(ip);
        if (error) {
                *tpp = NULL;
                xfs_iunlock(ip, XFS_IOLOCK_EXCL);
                return error; /* goto out */
        }

        tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
        error = xfs_trans_reserve(tp, 0,
                                  XFS_IFREE_LOG_RES(mp),
                                  0, XFS_TRANS_PERM_LOG_RES,
                                  XFS_INACTIVE_LOG_COUNT);
        if (error) {
                ASSERT(XFS_FORCED_SHUTDOWN(mp));
                xfs_trans_cancel(tp, 0);
                *tpp = NULL;
                xfs_iunlock(ip, XFS_IOLOCK_EXCL);
                return error;
        }

        xfs_ilock(ip, XFS_ILOCK_EXCL);
        xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
        xfs_trans_ihold(tp, ip);
        xfs_idestroy_fork(ip, XFS_ATTR_FORK);

        ASSERT(ip->i_d.di_anextents == 0);

        *tpp = tp;
        return 0;
}

int
xfs_release(
        xfs_inode_t     *ip)
{
        bhv_vnode_t     *vp = XFS_ITOV(ip);
        xfs_mount_t     *mp = ip->i_mount;
        int             error;

        if (!VN_ISREG(vp) || (ip->i_d.di_mode == 0))
                return 0;

        /* If this is a read-only mount, don't do this (would generate I/O) */
        if (mp->m_flags & XFS_MOUNT_RDONLY)
                return 0;

        if (!XFS_FORCED_SHUTDOWN(mp)) {
                int truncated;

                /*
                 * If we are using filestreams, and we have an unlinked
                 * file that we are processing the last close on, then nothing
                 * will be able to reopen and write to this file. Purge this
                 * inode from the filestreams cache so that it doesn't delay
                 * teardown of the inode.
                 */
                if ((ip->i_d.di_nlink == 0) && xfs_inode_is_filestream(ip))
                        xfs_filestream_deassociate(ip);

                /*
                 * If we previously truncated this file and removed old data
                 * in the process, we want to initiate "early" writeout on
                 * the last close.  This is an attempt to combat the notorious
                 * NULL files problem which is particularly noticable from a
                 * truncate down, buffered (re-)write (delalloc), followed by
                 * a crash.  What we are effectively doing here is
                 * significantly reducing the time window where we'd otherwise
                 * be exposed to that problem.
                 */
                truncated = xfs_iflags_test_and_clear(ip, XFS_ITRUNCATED);
                if (truncated && VN_DIRTY(vp) && ip->i_delayed_blks > 0)
                        xfs_flush_pages(ip, 0, -1, XFS_B_ASYNC, FI_NONE);
        }

#ifdef HAVE_REFCACHE
        /* If we are in the NFS reference cache then don't do this now */
        if (ip->i_refcache)
                return 0;
#endif

        if (ip->i_d.di_nlink != 0) {
                if ((((ip->i_d.di_mode & S_IFMT) == S_IFREG) &&
                     ((ip->i_size > 0) || (VN_CACHED(vp) > 0 ||
                       ip->i_delayed_blks > 0)) &&
                     (ip->i_df.if_flags & XFS_IFEXTENTS))  &&
                    (!(ip->i_d.di_flags &
                                (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)))) {
                        error = xfs_free_eofblocks(mp, ip, XFS_FREE_EOF_LOCK);
                        if (error)
                                return error;
                        /* Update linux inode block count after free above */
                        vn_to_inode(vp)->i_blocks = XFS_FSB_TO_BB(mp,
                                ip->i_d.di_nblocks + ip->i_delayed_blks);
                }
        }

        return 0;
}

/*
 * xfs_inactive
 *
 * This is called when the vnode reference count for the vnode
 * goes to zero.  If the file has been unlinked, then it must
 * now be truncated.  Also, we clear all of the read-ahead state
 * kept for the inode here since the file is now closed.
 */
int
xfs_inactive(
        xfs_inode_t     *ip)
{
        bhv_vnode_t     *vp = XFS_ITOV(ip);
        xfs_bmap_free_t free_list;
        xfs_fsblock_t   first_block;
        int             committed;
        xfs_trans_t     *tp;
        xfs_mount_t     *mp;
        int             error;
        int             truncate;

        vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

        /*
         * If the inode is already free, then there can be nothing
         * to clean up here.
         */
        if (ip->i_d.di_mode == 0 || VN_BAD(vp)) {
                ASSERT(ip->i_df.if_real_bytes == 0);
                ASSERT(ip->i_df.if_broot_bytes == 0);
                return VN_INACTIVE_CACHE;
        }

        /*
         * Only do a truncate if it's a regular file with
         * some actual space in it.  It's OK to look at the
         * inode's fields without the lock because we're the
         * only one with a reference to the inode.
         */
        truncate = ((ip->i_d.di_nlink == 0) &&
            ((ip->i_d.di_size != 0) || (ip->i_size != 0) ||
             (ip->i_d.di_nextents > 0) || (ip->i_delayed_blks > 0)) &&
            ((ip->i_d.di_mode & S_IFMT) == S_IFREG));

        mp = ip->i_mount;

        if (ip->i_d.di_nlink == 0 && DM_EVENT_ENABLED(ip, DM_EVENT_DESTROY)) {
                (void) XFS_SEND_DESTROY(mp, vp, DM_RIGHT_NULL);
        }

        error = 0;

        /* If this is a read-only mount, don't do this (would generate I/O) */
        if (mp->m_flags & XFS_MOUNT_RDONLY)
                goto out;

        if (ip->i_d.di_nlink != 0) {
                if ((((ip->i_d.di_mode & S_IFMT) == S_IFREG) &&
                     ((ip->i_size > 0) || (VN_CACHED(vp) > 0 ||
                       ip->i_delayed_blks > 0)) &&
                      (ip->i_df.if_flags & XFS_IFEXTENTS) &&
                     (!(ip->i_d.di_flags &
                                (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)) ||
                      (ip->i_delayed_blks != 0)))) {
                        error = xfs_free_eofblocks(mp, ip, XFS_FREE_EOF_LOCK);
                        if (error)
                                return VN_INACTIVE_CACHE;
                        /* Update linux inode block count after free above */
                        vn_to_inode(vp)->i_blocks = XFS_FSB_TO_BB(mp,
                                ip->i_d.di_nblocks + ip->i_delayed_blks);
                }
                goto out;
        }

        ASSERT(ip->i_d.di_nlink == 0);

        if ((error = XFS_QM_DQATTACH(mp, ip, 0)))
                return VN_INACTIVE_CACHE;

        tp = xfs_trans_alloc(mp, XFS_TRANS_INACTIVE);
        if (truncate) {
                /*
                 * Do the xfs_itruncate_start() call before
                 * reserving any log space because itruncate_start
                 * will call into the buffer cache and we can't
                 * do that within a transaction.
                 */
                xfs_ilock(ip, XFS_IOLOCK_EXCL);

                error = xfs_itruncate_start(ip, XFS_ITRUNC_DEFINITE, 0);
                if (error) {
                        xfs_trans_cancel(tp, 0);
                        xfs_iunlock(ip, XFS_IOLOCK_EXCL);
                        return VN_INACTIVE_CACHE;
                }

                error = xfs_trans_reserve(tp, 0,
                                          XFS_ITRUNCATE_LOG_RES(mp),
                                          0, XFS_TRANS_PERM_LOG_RES,
                                          XFS_ITRUNCATE_LOG_COUNT);
                if (error) {
                        /* Don't call itruncate_cleanup */
                        ASSERT(XFS_FORCED_SHUTDOWN(mp));
                        xfs_trans_cancel(tp, 0);
                        xfs_iunlock(ip, XFS_IOLOCK_EXCL);
                        return VN_INACTIVE_CACHE;
                }

                xfs_ilock(ip, XFS_ILOCK_EXCL);
                xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
                xfs_trans_ihold(tp, ip);

                /*
                 * normally, we have to run xfs_itruncate_finish sync.
                 * But if filesystem is wsync and we're in the inactive
                 * path, then we know that nlink == 0, and that the
                 * xaction that made nlink == 0 is permanently committed
                 * since xfs_remove runs as a synchronous transaction.
                 */
                error = xfs_itruncate_finish(&tp, ip, 0, XFS_DATA_FORK,
                                (!(mp->m_flags & XFS_MOUNT_WSYNC) ? 1 : 0));

                if (error) {
                        xfs_trans_cancel(tp,
                                XFS_TRANS_RELEASE_LOG_RES | XFS_TRANS_ABORT);
                        xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
                        return VN_INACTIVE_CACHE;
                }
        } else if ((ip->i_d.di_mode & S_IFMT) == S_IFLNK) {

                /*
                 * If we get an error while cleaning up a
                 * symlink we bail out.
                 */
                error = (ip->i_d.di_size > XFS_IFORK_DSIZE(ip)) ?
                        xfs_inactive_symlink_rmt(ip, &tp) :
                        xfs_inactive_symlink_local(ip, &tp);

                if (error) {
                        ASSERT(tp == NULL);
                        return VN_INACTIVE_CACHE;
                }

                xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
                xfs_trans_ihold(tp, ip);
        } else {
                error = xfs_trans_reserve(tp, 0,
                                          XFS_IFREE_LOG_RES(mp),
                                          0, XFS_TRANS_PERM_LOG_RES,
                                          XFS_INACTIVE_LOG_COUNT);
                if (error) {
                        ASSERT(XFS_FORCED_SHUTDOWN(mp));
                        xfs_trans_cancel(tp, 0);
                        return VN_INACTIVE_CACHE;
                }

                xfs_ilock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
                xfs_trans_ijoin(tp, ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);
                xfs_trans_ihold(tp, ip);
        }

        /*
         * If there are attributes associated with the file
         * then blow them away now.  The code calls a routine
         * that recursively deconstructs the attribute fork.
         * We need to just commit the current transaction
         * because we can't use it for xfs_attr_inactive().
         */
        if (ip->i_d.di_anextents > 0) {
                error = xfs_inactive_attrs(ip, &tp);
                /*
                 * If we got an error, the transaction is already
                 * cancelled, and the inode is unlocked. Just get out.
                 */
                 if (error)
                         return VN_INACTIVE_CACHE;
        } else if (ip->i_afp) {
                xfs_idestroy_fork(ip, XFS_ATTR_FORK);
        }

        /*
         * Free the inode.
         */
        XFS_BMAP_INIT(&free_list, &first_block);
        error = xfs_ifree(tp, ip, &free_list);
        if (error) {
                /*
                 * If we fail to free the inode, shut down.  The cancel
                 * might do that, we need to make sure.  Otherwise the
                 * inode might be lost for a long time or forever.
                 */
                if (!XFS_FORCED_SHUTDOWN(mp)) {
                        cmn_err(CE_NOTE,
                "xfs_inactive:  xfs_ifree() returned an error = %d on %s",
                                error, mp->m_fsname);
                        xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR);
                }
                xfs_trans_cancel(tp, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT);
        } else {
                /*
                 * Credit the quota account(s). The inode is gone.
                 */
                XFS_TRANS_MOD_DQUOT_BYINO(mp, tp, ip, XFS_TRANS_DQ_ICOUNT, -1);

                /*
                 * Just ignore errors at this point.  There is
                 * nothing we can do except to try to keep going.
                 */
                (void) xfs_bmap_finish(&tp,  &free_list, &committed);
                (void) xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
        }
        /*
         * Release the dquots held by inode, if any.
         */
        XFS_QM_DQDETACH(mp, ip);

        xfs_iunlock(ip, XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL);

 out:
        return VN_INACTIVE_CACHE;
}


int
xfs_lookup(
        xfs_inode_t             *dp,
        bhv_vname_t             *dentry,
        bhv_vnode_t             **vpp)
{
        xfs_inode_t             *ip;
        xfs_ino_t               e_inum;
        int                     error;
        uint                    lock_mode;

        vn_trace_entry(dp, __FUNCTION__, (inst_t *)__return_address);

        if (XFS_FORCED_SHUTDOWN(dp->i_mount))
                return XFS_ERROR(EIO);

        lock_mode = xfs_ilock_map_shared(dp);
        error = xfs_dir_lookup_int(dp, lock_mode, dentry, &e_inum, &ip);
        if (!error) {
                *vpp = XFS_ITOV(ip);
                ITRACE(ip);
        }
        xfs_iunlock_map_shared(dp, lock_mode);
        return error;
}

int
xfs_create(
        xfs_inode_t             *dp,
        bhv_vname_t             *dentry,
        mode_t                  mode,
        xfs_dev_t               rdev,
        bhv_vnode_t             **vpp,
        cred_t                  *credp)
{
        char                    *name = VNAME(dentry);
        xfs_mount_t             *mp = dp->i_mount;
        bhv_vnode_t             *dir_vp = XFS_ITOV(dp);
        xfs_inode_t             *ip;
        bhv_vnode_t             *vp = NULL;
        xfs_trans_t             *tp;
        int                     error;
        xfs_bmap_free_t         free_list;
        xfs_fsblock_t           first_block;
        boolean_t               unlock_dp_on_error = B_FALSE;
        int                     dm_event_sent = 0;
        uint                    cancel_flags;
        int                     committed;
        xfs_prid_t              prid;
        struct xfs_dquot        *udqp, *gdqp;
        uint                    resblks;
        int                     namelen;

        ASSERT(!*vpp);
        vn_trace_entry(dp, __FUNCTION__, (inst_t *)__return_address);

        namelen = VNAMELEN(dentry);

        if (DM_EVENT_ENABLED(dp, DM_EVENT_CREATE)) {
                error = XFS_SEND_NAMESP(mp, DM_EVENT_CREATE,
                                dir_vp, DM_RIGHT_NULL, NULL,
                                DM_RIGHT_NULL, name, NULL,
                                mode, 0, 0);

                if (error)
                        return error;
                dm_event_sent = 1;
        }

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        /* Return through std_return after this point. */

        udqp = gdqp = NULL;
        if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
                prid = dp->i_d.di_projid;
        else
                prid = (xfs_prid_t)dfltprid;

        /*
         * Make sure that we have allocated dquot(s) on disk.
         */
        error = XFS_QM_DQVOPALLOC(mp, dp,
                        current_fsuid(credp), current_fsgid(credp), prid,
                        XFS_QMOPT_QUOTALL|XFS_QMOPT_INHERIT, &udqp, &gdqp);
        if (error)
                goto std_return;

        ip = NULL;

        tp = xfs_trans_alloc(mp, XFS_TRANS_CREATE);
        cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
        resblks = XFS_CREATE_SPACE_RES(mp, namelen);
        /*
         * Initially assume that the file does not exist and
         * reserve the resources for that case.  If that is not
         * the case we'll drop the one we have and get a more
         * appropriate transaction later.
         */
        error = xfs_trans_reserve(tp, resblks, XFS_CREATE_LOG_RES(mp), 0,
                        XFS_TRANS_PERM_LOG_RES, XFS_CREATE_LOG_COUNT);
        if (error == ENOSPC) {
                resblks = 0;
                error = xfs_trans_reserve(tp, 0, XFS_CREATE_LOG_RES(mp), 0,
                                XFS_TRANS_PERM_LOG_RES, XFS_CREATE_LOG_COUNT);
        }
        if (error) {
                cancel_flags = 0;
                goto error_return;
        }

        xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
        unlock_dp_on_error = B_TRUE;

        XFS_BMAP_INIT(&free_list, &first_block);

        ASSERT(ip == NULL);

        /*
         * Reserve disk quota and the inode.
         */
        error = XFS_TRANS_RESERVE_QUOTA(mp, tp, udqp, gdqp, resblks, 1, 0);
        if (error)
                goto error_return;

        if (resblks == 0 && (error = xfs_dir_canenter(tp, dp, name, namelen)))
                goto error_return;
        error = xfs_dir_ialloc(&tp, dp, mode, 1,
                        rdev, credp, prid, resblks > 0,
                        &ip, &committed);
        if (error) {
                if (error == ENOSPC)
                        goto error_return;
                goto abort_return;
        }
        ITRACE(ip);

        /*
         * At this point, we've gotten a newly allocated inode.
         * It is locked (and joined to the transaction).
         */

        ASSERT(ismrlocked (&ip->i_lock, MR_UPDATE));

        /*
         * Now we join the directory inode to the transaction.  We do not do it
         * earlier because xfs_dir_ialloc might commit the previous transaction
         * (and release all the locks).  An error from here on will result in
         * the transaction cancel unlocking dp so don't do it explicitly in the
         * error path.
         */
        VN_HOLD(dir_vp);
        xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
        unlock_dp_on_error = B_FALSE;

        error = xfs_dir_createname(tp, dp, name, namelen, ip->i_ino,
                                        &first_block, &free_list, resblks ?
                                        resblks - XFS_IALLOC_SPACE_RES(mp) : 0);
        if (error) {
                ASSERT(error != ENOSPC);
                goto abort_return;
        }
        xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
        xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);

        /*
         * If this is a synchronous mount, make sure that the
         * create transaction goes to disk before returning to
         * the user.
         */
        if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
                xfs_trans_set_sync(tp);
        }

        dp->i_gen++;

        /*
         * Attach the dquot(s) to the inodes and modify them incore.
         * These ids of the inode couldn't have changed since the new
         * inode has been locked ever since it was created.
         */
        XFS_QM_DQVOPCREATE(mp, tp, ip, udqp, gdqp);

        /*
         * xfs_trans_commit normally decrements the vnode ref count
         * when it unlocks the inode. Since we want to return the
         * vnode to the caller, we bump the vnode ref count now.
         */
        IHOLD(ip);
        vp = XFS_ITOV(ip);

        error = xfs_bmap_finish(&tp, &free_list, &committed);
        if (error) {
                xfs_bmap_cancel(&free_list);
                goto abort_rele;
        }

        error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
        if (error) {
                IRELE(ip);
                tp = NULL;
                goto error_return;
        }

        XFS_QM_DQRELE(mp, udqp);
        XFS_QM_DQRELE(mp, gdqp);

        *vpp = vp;

        /* Fallthrough to std_return with error = 0  */

std_return:
        if ((*vpp || (error != 0 && dm_event_sent != 0)) &&
            DM_EVENT_ENABLED(dp, DM_EVENT_POSTCREATE)) {
                (void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTCREATE,
                        dir_vp, DM_RIGHT_NULL,
                        *vpp ? vp:NULL,
                        DM_RIGHT_NULL, name, NULL,
                        mode, error, 0);
        }
        return error;

 abort_return:
        cancel_flags |= XFS_TRANS_ABORT;
        /* FALLTHROUGH */

 error_return:
        if (tp != NULL)
                xfs_trans_cancel(tp, cancel_flags);

        XFS_QM_DQRELE(mp, udqp);
        XFS_QM_DQRELE(mp, gdqp);

        if (unlock_dp_on_error)
                xfs_iunlock(dp, XFS_ILOCK_EXCL);

        goto std_return;

 abort_rele:
        /*
         * Wait until after the current transaction is aborted to
         * release the inode.  This prevents recursive transactions
         * and deadlocks from xfs_inactive.
         */
        cancel_flags |= XFS_TRANS_ABORT;
        xfs_trans_cancel(tp, cancel_flags);
        IRELE(ip);

        XFS_QM_DQRELE(mp, udqp);
        XFS_QM_DQRELE(mp, gdqp);

        goto std_return;
}

#ifdef DEBUG
/*
 * Some counters to see if (and how often) we are hitting some deadlock
 * prevention code paths.
 */

int xfs_rm_locks;
int xfs_rm_lock_delays;
int xfs_rm_attempts;
#endif

/*
 * The following routine will lock the inodes associated with the
 * directory and the named entry in the directory. The locks are
 * acquired in increasing inode number.
 *
 * If the entry is "..", then only the directory is locked. The
 * vnode ref count will still include that from the .. entry in
 * this case.
 *
 * There is a deadlock we need to worry about. If the locked directory is
 * in the AIL, it might be blocking up the log. The next inode we lock
 * could be already locked by another thread waiting for log space (e.g
 * a permanent log reservation with a long running transaction (see
 * xfs_itruncate_finish)). To solve this, we must check if the directory
 * is in the ail and use lock_nowait. If we can't lock, we need to
 * drop the inode lock on the directory and try again. xfs_iunlock will
 * potentially push the tail if we were holding up the log.
 */
STATIC int
xfs_lock_dir_and_entry(
        xfs_inode_t     *dp,
        xfs_inode_t     *ip)    /* inode of entry 'name' */
{
        int             attempts;
        xfs_ino_t       e_inum;
        xfs_inode_t     *ips[2];
        xfs_log_item_t  *lp;

#ifdef DEBUG
        xfs_rm_locks++;
#endif
        attempts = 0;

again:
        xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);

        e_inum = ip->i_ino;

        ITRACE(ip);

        /*
         * We want to lock in increasing inum. Since we've already
         * acquired the lock on the directory, we may need to release
         * if if the inum of the entry turns out to be less.
         */
        if (e_inum > dp->i_ino) {
                /*
                 * We are already in the right order, so just
                 * lock on the inode of the entry.
                 * We need to use nowait if dp is in the AIL.
                 */

                lp = (xfs_log_item_t *)dp->i_itemp;
                if (lp && (lp->li_flags & XFS_LI_IN_AIL)) {
                        if (!xfs_ilock_nowait(ip, XFS_ILOCK_EXCL)) {
                                attempts++;
#ifdef DEBUG
                                xfs_rm_attempts++;
#endif

                                /*
                                 * Unlock dp and try again.
                                 * xfs_iunlock will try to push the tail
                                 * if the inode is in the AIL.
                                 */

                                xfs_iunlock(dp, XFS_ILOCK_EXCL);

                                if ((attempts % 5) == 0) {
                                        delay(1); /* Don't just spin the CPU */
#ifdef DEBUG
                                        xfs_rm_lock_delays++;
#endif
                                }
                                goto again;
                        }
                } else {
                        xfs_ilock(ip, XFS_ILOCK_EXCL);
                }
        } else if (e_inum < dp->i_ino) {
                xfs_iunlock(dp, XFS_ILOCK_EXCL);

                ips[0] = ip;
                ips[1] = dp;
                xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
        }
        /* else  e_inum == dp->i_ino */
        /*     This can happen if we're asked to lock /x/..
         *     the entry is "..", which is also the parent directory.
         */

        return 0;
}

#ifdef DEBUG
int xfs_locked_n;
int xfs_small_retries;
int xfs_middle_retries;
int xfs_lots_retries;
int xfs_lock_delays;
#endif

/*
 * Bump the subclass so xfs_lock_inodes() acquires each lock with
 * a different value
 */
static inline int
xfs_lock_inumorder(int lock_mode, int subclass)
{
        if (lock_mode & (XFS_IOLOCK_SHARED|XFS_IOLOCK_EXCL))
                lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_IOLOCK_SHIFT;
        if (lock_mode & (XFS_ILOCK_SHARED|XFS_ILOCK_EXCL))
                lock_mode |= (subclass + XFS_LOCK_INUMORDER) << XFS_ILOCK_SHIFT;

        return lock_mode;
}

/*
 * The following routine will lock n inodes in exclusive mode.
 * We assume the caller calls us with the inodes in i_ino order.
 *
 * We need to detect deadlock where an inode that we lock
 * is in the AIL and we start waiting for another inode that is locked
 * by a thread in a long running transaction (such as truncate). This can
 * result in deadlock since the long running trans might need to wait
 * for the inode we just locked in order to push the tail and free space
 * in the log.
 */
void
xfs_lock_inodes(
        xfs_inode_t     **ips,
        int             inodes,
        int             first_locked,
        uint            lock_mode)
{
        int             attempts = 0, i, j, try_lock;
        xfs_log_item_t  *lp;

        ASSERT(ips && (inodes >= 2)); /* we need at least two */

        if (first_locked) {
                try_lock = 1;
                i = 1;
        } else {
                try_lock = 0;
                i = 0;
        }

again:
        for (; i < inodes; i++) {
                ASSERT(ips[i]);

                if (i && (ips[i] == ips[i-1]))  /* Already locked */
                        continue;

                /*
                 * If try_lock is not set yet, make sure all locked inodes
                 * are not in the AIL.
                 * If any are, set try_lock to be used later.
                 */

                if (!try_lock) {
                        for (j = (i - 1); j >= 0 && !try_lock; j--) {
                                lp = (xfs_log_item_t *)ips[j]->i_itemp;
                                if (lp && (lp->li_flags & XFS_LI_IN_AIL)) {
                                        try_lock++;
                                }
                        }
                }

                /*
                 * If any of the previous locks we have locked is in the AIL,
                 * we must TRY to get the second and subsequent locks. If
                 * we can't get any, we must release all we have
                 * and try again.
                 */

                if (try_lock) {
                        /* try_lock must be 0 if i is 0. */
                        /*
                         * try_lock means we have an inode locked
                         * that is in the AIL.
                         */
                        ASSERT(i != 0);
                        if (!xfs_ilock_nowait(ips[i], xfs_lock_inumorder(lock_mode, i))) {
                                attempts++;

                                /*
                                 * Unlock all previous guys and try again.
                                 * xfs_iunlock will try to push the tail
                                 * if the inode is in the AIL.
                                 */

                                for(j = i - 1; j >= 0; j--) {

                                        /*
                                         * Check to see if we've already
                                         * unlocked this one.
                                         * Not the first one going back,
                                         * and the inode ptr is the same.
                                         */
                                        if ((j != (i - 1)) && ips[j] ==
                                                                ips[j+1])
                                                continue;

                                        xfs_iunlock(ips[j], lock_mode);
                                }

                                if ((attempts % 5) == 0) {
                                        delay(1); /* Don't just spin the CPU */
#ifdef DEBUG
                                        xfs_lock_delays++;
#endif
                                }
                                i = 0;
                                try_lock = 0;
                                goto again;
                        }
                } else {
                        xfs_ilock(ips[i], xfs_lock_inumorder(lock_mode, i));
                }
        }

#ifdef DEBUG
        if (attempts) {
                if (attempts < 5) xfs_small_retries++;
                else if (attempts < 100) xfs_middle_retries++;
                else xfs_lots_retries++;
        } else {
                xfs_locked_n++;
        }
#endif
}

#ifdef  DEBUG
#define REMOVE_DEBUG_TRACE(x)   {remove_which_error_return = (x);}
int remove_which_error_return = 0;
#else /* ! DEBUG */
#define REMOVE_DEBUG_TRACE(x)
#endif  /* ! DEBUG */

int
xfs_remove(
        xfs_inode_t             *dp,
        bhv_vname_t             *dentry)
{
        bhv_vnode_t             *dir_vp = XFS_ITOV(dp);
        char                    *name = VNAME(dentry);
        xfs_mount_t             *mp = dp->i_mount;
        xfs_inode_t             *ip;
        xfs_trans_t             *tp = NULL;
        int                     error = 0;
        xfs_bmap_free_t         free_list;
        xfs_fsblock_t           first_block;
        int                     cancel_flags;
        int                     committed;
        int                     dm_di_mode = 0;
        int                     link_zero;
        uint                    resblks;
        int                     namelen;

        vn_trace_entry(dp, __FUNCTION__, (inst_t *)__return_address);

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        namelen = VNAMELEN(dentry);

        if (!xfs_get_dir_entry(dentry, &ip)) {
                dm_di_mode = ip->i_d.di_mode;
                IRELE(ip);
        }

        if (DM_EVENT_ENABLED(dp, DM_EVENT_REMOVE)) {
                error = XFS_SEND_NAMESP(mp, DM_EVENT_REMOVE, dir_vp,
                                        DM_RIGHT_NULL, NULL, DM_RIGHT_NULL,
                                        name, NULL, dm_di_mode, 0, 0);
                if (error)
                        return error;
        }

        /* From this point on, return through std_return */
        ip = NULL;

        /*
         * We need to get a reference to ip before we get our log
         * reservation. The reason for this is that we cannot call
         * xfs_iget for an inode for which we do not have a reference
         * once we've acquired a log reservation. This is because the
         * inode we are trying to get might be in xfs_inactive going
         * for a log reservation. Since we'll have to wait for the
         * inactive code to complete before returning from xfs_iget,
         * we need to make sure that we don't have log space reserved
         * when we call xfs_iget.  Instead we get an unlocked reference
         * to the inode before getting our log reservation.
         */
        error = xfs_get_dir_entry(dentry, &ip);
        if (error) {
                REMOVE_DEBUG_TRACE(__LINE__);
                goto std_return;
        }

        dm_di_mode = ip->i_d.di_mode;

        vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address);

        ITRACE(ip);

        error = XFS_QM_DQATTACH(mp, dp, 0);
        if (!error && dp != ip)
                error = XFS_QM_DQATTACH(mp, ip, 0);
        if (error) {
                REMOVE_DEBUG_TRACE(__LINE__);
                IRELE(ip);
                goto std_return;
        }

        tp = xfs_trans_alloc(mp, XFS_TRANS_REMOVE);
        cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
        /*
         * We try to get the real space reservation first,
         * allowing for directory btree deletion(s) implying
         * possible bmap insert(s).  If we can't get the space
         * reservation then we use 0 instead, and avoid the bmap
         * btree insert(s) in the directory code by, if the bmap
         * insert tries to happen, instead trimming the LAST
         * block from the directory.
         */
        resblks = XFS_REMOVE_SPACE_RES(mp);
        error = xfs_trans_reserve(tp, resblks, XFS_REMOVE_LOG_RES(mp), 0,
                        XFS_TRANS_PERM_LOG_RES, XFS_REMOVE_LOG_COUNT);
        if (error == ENOSPC) {
                resblks = 0;
                error = xfs_trans_reserve(tp, 0, XFS_REMOVE_LOG_RES(mp), 0,
                                XFS_TRANS_PERM_LOG_RES, XFS_REMOVE_LOG_COUNT);
        }
        if (error) {
                ASSERT(error != ENOSPC);
                REMOVE_DEBUG_TRACE(__LINE__);
                xfs_trans_cancel(tp, 0);
                IRELE(ip);
                return error;
        }

        error = xfs_lock_dir_and_entry(dp, ip);
        if (error) {
                REMOVE_DEBUG_TRACE(__LINE__);
                xfs_trans_cancel(tp, cancel_flags);
                IRELE(ip);
                goto std_return;
        }

        /*
         * At this point, we've gotten both the directory and the entry
         * inodes locked.
         */
        xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
        if (dp != ip) {
                /*
                 * Increment vnode ref count only in this case since
                 * there's an extra vnode reference in the case where
                 * dp == ip.
                 */
                IHOLD(dp);
                xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
        }

        /*
         * Entry must exist since we did a lookup in xfs_lock_dir_and_entry.
         */
        XFS_BMAP_INIT(&free_list, &first_block);
        error = xfs_dir_removename(tp, dp, name, namelen, ip->i_ino,
                                        &first_block, &free_list, 0);
        if (error) {
                ASSERT(error != ENOENT);
                REMOVE_DEBUG_TRACE(__LINE__);
                goto error1;
        }
        xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);

        dp->i_gen++;
        xfs_trans_log_inode(tp, dp, XFS_ILOG_CORE);

        error = xfs_droplink(tp, ip);
        if (error) {
                REMOVE_DEBUG_TRACE(__LINE__);
                goto error1;
        }

        /* Determine if this is the last link while
         * we are in the transaction.
         */
        link_zero = (ip)->i_d.di_nlink==0;

        /*
         * Take an extra ref on the inode so that it doesn't
         * go to xfs_inactive() from within the commit.
         */
        IHOLD(ip);

        /*
         * If this is a synchronous mount, make sure that the
         * remove transaction goes to disk before returning to
         * the user.
         */
        if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
                xfs_trans_set_sync(tp);
        }

        error = xfs_bmap_finish(&tp, &free_list, &committed);
        if (error) {
                REMOVE_DEBUG_TRACE(__LINE__);
                goto error_rele;
        }

        error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
        if (error) {
                IRELE(ip);
                goto std_return;
        }

        /*
         * Before we drop our extra reference to the inode, purge it
         * from the refcache if it is there.  By waiting until afterwards
         * to do the IRELE, we ensure that we won't go inactive in the
         * xfs_refcache_purge_ip routine (although that would be OK).
         */
        xfs_refcache_purge_ip(ip);

        /*
         * If we are using filestreams, kill the stream association.
         * If the file is still open it may get a new one but that
         * will get killed on last close in xfs_close() so we don't
         * have to worry about that.
         */
        if (link_zero && xfs_inode_is_filestream(ip))
                xfs_filestream_deassociate(ip);

        vn_trace_exit(ip, __FUNCTION__, (inst_t *)__return_address);

        IRELE(ip);

/*      Fall through to std_return with error = 0 */
 std_return:
        if (DM_EVENT_ENABLED(dp, DM_EVENT_POSTREMOVE)) {
                (void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTREMOVE,
                                dir_vp, DM_RIGHT_NULL,
                                NULL, DM_RIGHT_NULL,
                                name, NULL, dm_di_mode, error, 0);
        }
        return error;

 error1:
        xfs_bmap_cancel(&free_list);
        cancel_flags |= XFS_TRANS_ABORT;
        xfs_trans_cancel(tp, cancel_flags);
        goto std_return;

 error_rele:
        /*
         * In this case make sure to not release the inode until after
         * the current transaction is aborted.  Releasing it beforehand
         * can cause us to go to xfs_inactive and start a recursive
         * transaction which can easily deadlock with the current one.
         */
        xfs_bmap_cancel(&free_list);
        cancel_flags |= XFS_TRANS_ABORT;
        xfs_trans_cancel(tp, cancel_flags);

        /*
         * Before we drop our extra reference to the inode, purge it
         * from the refcache if it is there.  By waiting until afterwards
         * to do the IRELE, we ensure that we won't go inactive in the
         * xfs_refcache_purge_ip routine (although that would be OK).
         */
        xfs_refcache_purge_ip(ip);

        IRELE(ip);

        goto std_return;
}

int
xfs_link(
        xfs_inode_t             *tdp,
        bhv_vnode_t             *src_vp,
        bhv_vname_t             *dentry)
{
        bhv_vnode_t             *target_dir_vp = XFS_ITOV(tdp);
        xfs_mount_t             *mp = tdp->i_mount;
        xfs_inode_t             *sip = xfs_vtoi(src_vp);
        xfs_trans_t             *tp;
        xfs_inode_t             *ips[2];
        int                     error;
        xfs_bmap_free_t         free_list;
        xfs_fsblock_t           first_block;
        int                     cancel_flags;
        int                     committed;
        int                     resblks;
        char                    *target_name = VNAME(dentry);
        int                     target_namelen;

        vn_trace_entry(tdp, __FUNCTION__, (inst_t *)__return_address);
        vn_trace_entry(xfs_vtoi(src_vp), __FUNCTION__, (inst_t *)__return_address);

        target_namelen = VNAMELEN(dentry);
        ASSERT(!VN_ISDIR(src_vp));

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        if (DM_EVENT_ENABLED(tdp, DM_EVENT_LINK)) {
                error = XFS_SEND_NAMESP(mp, DM_EVENT_LINK,
                                        target_dir_vp, DM_RIGHT_NULL,
                                        src_vp, DM_RIGHT_NULL,
                                        target_name, NULL, 0, 0, 0);
                if (error)
                        return error;
        }

        /* Return through std_return after this point. */

        error = XFS_QM_DQATTACH(mp, sip, 0);
        if (!error && sip != tdp)
                error = XFS_QM_DQATTACH(mp, tdp, 0);
        if (error)
                goto std_return;

        tp = xfs_trans_alloc(mp, XFS_TRANS_LINK);
        cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
        resblks = XFS_LINK_SPACE_RES(mp, target_namelen);
        error = xfs_trans_reserve(tp, resblks, XFS_LINK_LOG_RES(mp), 0,
                        XFS_TRANS_PERM_LOG_RES, XFS_LINK_LOG_COUNT);
        if (error == ENOSPC) {
                resblks = 0;
                error = xfs_trans_reserve(tp, 0, XFS_LINK_LOG_RES(mp), 0,
                                XFS_TRANS_PERM_LOG_RES, XFS_LINK_LOG_COUNT);
        }
        if (error) {
                cancel_flags = 0;
                goto error_return;
        }

        if (sip->i_ino < tdp->i_ino) {
                ips[0] = sip;
                ips[1] = tdp;
        } else {
                ips[0] = tdp;
                ips[1] = sip;
        }

        xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);

        /*
         * Increment vnode ref counts since xfs_trans_commit &
         * xfs_trans_cancel will both unlock the inodes and
         * decrement the associated ref counts.
         */
        VN_HOLD(src_vp);
        VN_HOLD(target_dir_vp);
        xfs_trans_ijoin(tp, sip, XFS_ILOCK_EXCL);
        xfs_trans_ijoin(tp, tdp, XFS_ILOCK_EXCL);

        /*
         * If the source has too many links, we can't make any more to it.
         */
        if (sip->i_d.di_nlink >= XFS_MAXLINK) {
                error = XFS_ERROR(EMLINK);
                goto error_return;
        }

        /*
         * If we are using project inheritance, we only allow hard link
         * creation in our tree when the project IDs are the same; else
         * the tree quota mechanism could be circumvented.
         */
        if (unlikely((tdp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
                     (tdp->i_d.di_projid != sip->i_d.di_projid))) {
                error = XFS_ERROR(EXDEV);
                goto error_return;
        }

        if (resblks == 0 &&
            (error = xfs_dir_canenter(tp, tdp, target_name, target_namelen)))
                goto error_return;

        XFS_BMAP_INIT(&free_list, &first_block);

        error = xfs_dir_createname(tp, tdp, target_name, target_namelen,
                                   sip->i_ino, &first_block, &free_list,
                                   resblks);
        if (error)
                goto abort_return;
        xfs_ichgtime(tdp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
        tdp->i_gen++;
        xfs_trans_log_inode(tp, tdp, XFS_ILOG_CORE);

        error = xfs_bumplink(tp, sip);
        if (error)
                goto abort_return;

        /*
         * If this is a synchronous mount, make sure that the
         * link transaction goes to disk before returning to
         * the user.
         */
        if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
                xfs_trans_set_sync(tp);
        }

        error = xfs_bmap_finish (&tp, &free_list, &committed);
        if (error) {
                xfs_bmap_cancel(&free_list);
                goto abort_return;
        }

        error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
        if (error)
                goto std_return;

        /* Fall through to std_return with error = 0. */
std_return:
        if (DM_EVENT_ENABLED(sip, DM_EVENT_POSTLINK)) {
                (void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTLINK,
                                target_dir_vp, DM_RIGHT_NULL,
                                src_vp, DM_RIGHT_NULL,
                                target_name, NULL, 0, error, 0);
        }
        return error;

 abort_return:
        cancel_flags |= XFS_TRANS_ABORT;
        /* FALLTHROUGH */

 error_return:
        xfs_trans_cancel(tp, cancel_flags);
        goto std_return;
}


int
xfs_mkdir(
        xfs_inode_t             *dp,
        bhv_vname_t             *dentry,
        mode_t                  mode,
        bhv_vnode_t             **vpp,
        cred_t                  *credp)
{
        bhv_vnode_t             *dir_vp = XFS_ITOV(dp);
        char                    *dir_name = VNAME(dentry);
        int                     dir_namelen = VNAMELEN(dentry);
        xfs_mount_t             *mp = dp->i_mount;
        xfs_inode_t             *cdp;   /* inode of created dir */
        bhv_vnode_t             *cvp;   /* vnode of created dir */
        xfs_trans_t             *tp;
        int                     cancel_flags;
        int                     error;
        int                     committed;
        xfs_bmap_free_t         free_list;
        xfs_fsblock_t           first_block;
        boolean_t               unlock_dp_on_error = B_FALSE;
        boolean_t               created = B_FALSE;
        int                     dm_event_sent = 0;
        xfs_prid_t              prid;
        struct xfs_dquot        *udqp, *gdqp;
        uint                    resblks;

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        tp = NULL;

        if (DM_EVENT_ENABLED(dp, DM_EVENT_CREATE)) {
                error = XFS_SEND_NAMESP(mp, DM_EVENT_CREATE,
                                        dir_vp, DM_RIGHT_NULL, NULL,
                                        DM_RIGHT_NULL, dir_name, NULL,
                                        mode, 0, 0);
                if (error)
                        return error;
                dm_event_sent = 1;
        }

        /* Return through std_return after this point. */

        vn_trace_entry(dp, __FUNCTION__, (inst_t *)__return_address);

        mp = dp->i_mount;
        udqp = gdqp = NULL;
        if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
                prid = dp->i_d.di_projid;
        else
                prid = (xfs_prid_t)dfltprid;

        /*
         * Make sure that we have allocated dquot(s) on disk.
         */
        error = XFS_QM_DQVOPALLOC(mp, dp,
                        current_fsuid(credp), current_fsgid(credp), prid,
                        XFS_QMOPT_QUOTALL | XFS_QMOPT_INHERIT, &udqp, &gdqp);
        if (error)
                goto std_return;

        tp = xfs_trans_alloc(mp, XFS_TRANS_MKDIR);
        cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
        resblks = XFS_MKDIR_SPACE_RES(mp, dir_namelen);
        error = xfs_trans_reserve(tp, resblks, XFS_MKDIR_LOG_RES(mp), 0,
                                  XFS_TRANS_PERM_LOG_RES, XFS_MKDIR_LOG_COUNT);
        if (error == ENOSPC) {
                resblks = 0;
                error = xfs_trans_reserve(tp, 0, XFS_MKDIR_LOG_RES(mp), 0,
                                          XFS_TRANS_PERM_LOG_RES,
                                          XFS_MKDIR_LOG_COUNT);
        }
        if (error) {
                cancel_flags = 0;
                goto error_return;
        }

        xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
        unlock_dp_on_error = B_TRUE;

        /*
         * Check for directory link count overflow.
         */
        if (dp->i_d.di_nlink >= XFS_MAXLINK) {
                error = XFS_ERROR(EMLINK);
                goto error_return;
        }

        /*
         * Reserve disk quota and the inode.
         */
        error = XFS_TRANS_RESERVE_QUOTA(mp, tp, udqp, gdqp, resblks, 1, 0);
        if (error)
                goto error_return;

        if (resblks == 0 &&
            (error = xfs_dir_canenter(tp, dp, dir_name, dir_namelen)))
                goto error_return;
        /*
         * create the directory inode.
         */
        error = xfs_dir_ialloc(&tp, dp, mode, 2,
                        0, credp, prid, resblks > 0,
                &cdp, NULL);
        if (error) {
                if (error == ENOSPC)
                        goto error_return;
                goto abort_return;
        }
        ITRACE(cdp);

        /*
         * Now we add the directory inode to the transaction.
         * We waited until now since xfs_dir_ialloc might start
         * a new transaction.  Had we joined the transaction
         * earlier, the locks might have gotten released. An error
         * from here on will result in the transaction cancel
         * unlocking dp so don't do it explicitly in the error path.
         */
        VN_HOLD(dir_vp);
        xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
        unlock_dp_on_error = B_FALSE;

        XFS_BMAP_INIT(&free_list, &first_block);

        error = xfs_dir_createname(tp, dp, dir_name, dir_namelen, cdp->i_ino,
                                   &first_block, &free_list, resblks ?
                                   resblks - XFS_IALLOC_SPACE_RES(mp) : 0);
        if (error) {
                ASSERT(error != ENOSPC);
                goto error1;
        }
        xfs_ichgtime(dp, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);

        /*
         * Bump the in memory version number of the parent directory
         * so that other processes accessing it will recognize that
         * the directory has changed.
         */
        dp->i_gen++;

        error = xfs_dir_init(tp, cdp, dp);
        if (error)
                goto error2;

        cdp->i_gen = 1;
        error = xfs_bumplink(tp, dp);
        if (error)
                goto error2;

        cvp = XFS_ITOV(cdp);

        created = B_TRUE;

        *vpp = cvp;
        IHOLD(cdp);

        /*
         * Attach the dquots to the new inode and modify the icount incore.
         */
        XFS_QM_DQVOPCREATE(mp, tp, cdp, udqp, gdqp);

        /*
         * If this is a synchronous mount, make sure that the
         * mkdir transaction goes to disk before returning to
         * the user.
         */
        if (mp->m_flags & (XFS_MOUNT_WSYNC|XFS_MOUNT_DIRSYNC)) {
                xfs_trans_set_sync(tp);
        }

        error = xfs_bmap_finish(&tp, &free_list, &committed);
        if (error) {
                IRELE(cdp);
                goto error2;
        }

        error = xfs_trans_commit(tp, XFS_TRANS_RELEASE_LOG_RES);
        XFS_QM_DQRELE(mp, udqp);
        XFS_QM_DQRELE(mp, gdqp);
        if (error) {
                IRELE(cdp);
        }

        /* Fall through to std_return with error = 0 or errno from
         * xfs_trans_commit. */

std_return:
        if ((created || (error != 0 && dm_event_sent != 0)) &&
            DM_EVENT_ENABLED(dp, DM_EVENT_POSTCREATE)) {
                (void) XFS_SEND_NAMESP(mp, DM_EVENT_POSTCREATE,
                                        dir_vp, DM_RIGHT_NULL,
                                        created ? XFS_ITOV(cdp):NULL,
                                        DM_RIGHT_NULL,
                                        dir_name, NULL,
                                        mode, error, 0);
        }
        return error;

 error2:
 error1:
        xfs_bmap_cancel(&free_list);
 abort_return:
        cancel_flags |= XFS_TRANS_ABORT;
 error_return:
        xfs_trans_cancel(tp, cancel_flags);
        XFS_QM_DQRELE(mp, udqp);
        XFS_QM_DQRELE(mp, gdqp);

        if (unlock_dp_on_error)
                xfs_iunlock(dp, XFS_ILOCK_EXCL);

        goto std_return;
}

int
xfs_rmdir(
        xfs_inode_t             *dp,
        bhv_vname_t             *dentry)
{
        bhv_vnode_t             *dir_vp = XFS_ITOV(dp);
        char                    *name = VNAME(dentry);
        int                     namelen = VNAMELEN(dentry);
        xfs_mount_t             *mp = dp->i_mount;
        xfs_inode_t             *cdp;   /* child directory */
        xfs_trans_t             *tp;
        int                     error;
        xfs_bmap_free_t         free_list;
        xfs_fsblock_t           first_block;
        int                     cancel_flags;
        int                     committed;
        int                     dm_di_mode = S_IFDIR;
        int                     last_cdp_link;
        uint                    resblks;

        vn_trace_entry(dp, __FUNCTION__, (inst_t *)__return_address);

        if (XFS_FORCED_SHUTDOWN(mp))
                return XFS_ERROR(EIO);

        if (!xfs_get_dir_entry(dentry, &cdp)) {
                dm_di_mode = cdp->i_d.di_mode;
                IRELE(cdp);
        }

        if (DM_EVENT_ENABLED(dp, DM_EVENT_REMOVE)) {
                error = XFS_SEND_NAMESP(mp, DM_EVENT_REMOVE,
                                        dir_vp, DM_RIGHT_NULL,
                                        NULL, DM_RIGHT_NULL,
                                        name, NULL, dm_di_mode, 0, 0);
                if (error)
                        return XFS_ERROR(error);
        }

        /* Return through std_return after this point. */

        cdp = NULL;

        /*
         * We need to get a reference to cdp before we get our log
         * reservation.  The reason for this is that we cannot call
         * xfs_iget for an inode for which we do not have a reference
         * once we've acquired a log reservation.  This is because the
         * inode we are trying to get might be in xfs_inactive going
         * for a log reservation.  Since we'll have to wait for the
         * inactive code to complete before returning from xfs_iget,
         * we need to make sure that we don't have log space reserved
         * when we call xfs_iget.  Instead we get an unlocked reference
         * to the inode before getting our log reservation.
         */
        error = xfs_get_dir_entry(dentry, &cdp);
        if (error) {
                REMOVE_DEBUG_TRACE(__LINE__);
                goto std_return;
        }
        mp = dp->i_mount;
        dm_di_mode = cdp->i_d.di_mode;

        /*
         * Get the dquots for the inodes.
         */
        error = XFS_QM_DQATTACH(mp, dp, 0);
        if (!error && dp != cdp)
                error = XFS_QM_DQATTACH(mp, cdp, 0);
        if (error) {
                IRELE(cdp);
                REMOVE_DEBUG_TRACE(__LINE__);
                goto std_return;
        }

        tp = xfs_trans_alloc(mp, XFS_TRANS_RMDIR);
        cancel_flags = XFS_TRANS_RELEASE_LOG_RES;
        /*
         * We try to get the real space reservation first,
         * allowing for directory btree deletion(s) implying
         * possible bmap insert(s).  If we can't get the space
         * reservation then we use 0 instead, and avoid the bmap