[XFS] implement generic xfs_btree_split
[safe/jmp/linux-2.6] / fs / xfs / xfs_vfsops.c
index 22a0927..439dd39 100644 (file)
@@ -43,7 +43,6 @@
 #include "xfs_error.h"
 #include "xfs_bmap.h"
 #include "xfs_rw.h"
-#include "xfs_refcache.h"
 #include "xfs_buf_item.h"
 #include "xfs_log_priv.h"
 #include "xfs_dir2_trace.h"
 #include "xfs_acl.h"
 #include "xfs_attr.h"
 #include "xfs_clnt.h"
+#include "xfs_mru_cache.h"
+#include "xfs_filestream.h"
 #include "xfs_fsops.h"
+#include "xfs_vnodeops.h"
+#include "xfs_vfsops.h"
+#include "xfs_utils.h"
 
-STATIC int     xfs_sync(bhv_desc_t *, int, cred_t *);
 
-int
-xfs_init(void)
-{
-       extern kmem_zone_t      *xfs_bmap_free_item_zone;
-       extern kmem_zone_t      *xfs_btree_cur_zone;
-       extern kmem_zone_t      *xfs_trans_zone;
-       extern kmem_zone_t      *xfs_buf_item_zone;
-       extern kmem_zone_t      *xfs_dabuf_zone;
-#ifdef XFS_DABUF_DEBUG
-       extern lock_t           xfs_dabuf_global_lock;
-       spinlock_init(&xfs_dabuf_global_lock, "xfsda");
-#endif
-
-       /*
-        * Initialize all of the zone allocators we use.
-        */
-       xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
-                                                "xfs_bmap_free_item");
-       xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
-                                           "xfs_btree_cur");
-       xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
-       xfs_da_state_zone =
-               kmem_zone_init(sizeof(xfs_da_state_t), "xfs_da_state");
-       xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
-       xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
-       xfs_acl_zone_init(xfs_acl_zone, "xfs_acl");
-
-       /*
-        * The size of the zone allocated buf log item is the maximum
-        * size possible under XFS.  This wastes a little bit of memory,
-        * but it is much faster.
-        */
-       xfs_buf_item_zone =
-               kmem_zone_init((sizeof(xfs_buf_log_item_t) +
-                               (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
-                                 NBWORD) * sizeof(int))),
-                              "xfs_buf_item");
-       xfs_efd_zone =
-               kmem_zone_init((sizeof(xfs_efd_log_item_t) +
-                              ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
-                                sizeof(xfs_extent_t))),
-                                     "xfs_efd_item");
-       xfs_efi_zone =
-               kmem_zone_init((sizeof(xfs_efi_log_item_t) +
-                              ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
-                                sizeof(xfs_extent_t))),
-                                     "xfs_efi_item");
-
-       /*
-        * These zones warrant special memory allocator hints
-        */
-       xfs_inode_zone =
-               kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
-                                       KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
-                                       KM_ZONE_SPREAD, NULL);
-       xfs_ili_zone =
-               kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
-                                       KM_ZONE_SPREAD, NULL);
-       xfs_chashlist_zone =
-               kmem_zone_init_flags(sizeof(xfs_chashlist_t), "xfs_chashlist",
-                                       KM_ZONE_SPREAD, NULL);
-
-       /*
-        * Allocate global trace buffers.
-        */
-#ifdef XFS_ALLOC_TRACE
-       xfs_alloc_trace_buf = ktrace_alloc(XFS_ALLOC_TRACE_SIZE, KM_SLEEP);
-#endif
-#ifdef XFS_BMAP_TRACE
-       xfs_bmap_trace_buf = ktrace_alloc(XFS_BMAP_TRACE_SIZE, KM_SLEEP);
-#endif
-#ifdef XFS_BMBT_TRACE
-       xfs_bmbt_trace_buf = ktrace_alloc(XFS_BMBT_TRACE_SIZE, KM_SLEEP);
-#endif
-#ifdef XFS_ATTR_TRACE
-       xfs_attr_trace_buf = ktrace_alloc(XFS_ATTR_TRACE_SIZE, KM_SLEEP);
-#endif
-#ifdef XFS_DIR2_TRACE
-       xfs_dir2_trace_buf = ktrace_alloc(XFS_DIR2_GTRACE_SIZE, KM_SLEEP);
-#endif
-
-       xfs_dir_startup();
-
-#if (defined(DEBUG) || defined(INDUCE_IO_ERROR))
-       xfs_error_test_init();
-#endif /* DEBUG || INDUCE_IO_ERROR */
-
-       xfs_init_procfs();
-       xfs_sysctl_register();
-       return 0;
-}
-
-void
-xfs_cleanup(void)
-{
-       extern kmem_zone_t      *xfs_bmap_free_item_zone;
-       extern kmem_zone_t      *xfs_btree_cur_zone;
-       extern kmem_zone_t      *xfs_inode_zone;
-       extern kmem_zone_t      *xfs_trans_zone;
-       extern kmem_zone_t      *xfs_da_state_zone;
-       extern kmem_zone_t      *xfs_dabuf_zone;
-       extern kmem_zone_t      *xfs_efd_zone;
-       extern kmem_zone_t      *xfs_efi_zone;
-       extern kmem_zone_t      *xfs_buf_item_zone;
-       extern kmem_zone_t      *xfs_chashlist_zone;
-
-       xfs_cleanup_procfs();
-       xfs_sysctl_unregister();
-       xfs_refcache_destroy();
-       xfs_acl_zone_destroy(xfs_acl_zone);
-
-#ifdef XFS_DIR2_TRACE
-       ktrace_free(xfs_dir2_trace_buf);
-#endif
-#ifdef XFS_ATTR_TRACE
-       ktrace_free(xfs_attr_trace_buf);
-#endif
-#ifdef XFS_BMBT_TRACE
-       ktrace_free(xfs_bmbt_trace_buf);
-#endif
-#ifdef XFS_BMAP_TRACE
-       ktrace_free(xfs_bmap_trace_buf);
-#endif
-#ifdef XFS_ALLOC_TRACE
-       ktrace_free(xfs_alloc_trace_buf);
-#endif
-
-       kmem_zone_destroy(xfs_bmap_free_item_zone);
-       kmem_zone_destroy(xfs_btree_cur_zone);
-       kmem_zone_destroy(xfs_inode_zone);
-       kmem_zone_destroy(xfs_trans_zone);
-       kmem_zone_destroy(xfs_da_state_zone);
-       kmem_zone_destroy(xfs_dabuf_zone);
-       kmem_zone_destroy(xfs_buf_item_zone);
-       kmem_zone_destroy(xfs_efd_zone);
-       kmem_zone_destroy(xfs_efi_zone);
-       kmem_zone_destroy(xfs_ifork_zone);
-       kmem_zone_destroy(xfs_ili_zone);
-       kmem_zone_destroy(xfs_chashlist_zone);
-}
-
-/*
- * xfs_start_flags
- *
- * This function fills in xfs_mount_t fields based on mount args.
- * Note: the superblock has _not_ yet been read in.
- */
-STATIC int
-xfs_start_flags(
-       struct bhv_vfs          *vfs,
-       struct xfs_mount_args   *ap,
-       struct xfs_mount        *mp)
-{
-       /* Values are in BBs */
-       if ((ap->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
-               /*
-                * At this point the superblock has not been read
-                * in, therefore we do not know the block size.
-                * Before the mount call ends we will convert
-                * these to FSBs.
-                */
-               mp->m_dalign = ap->sunit;
-               mp->m_swidth = ap->swidth;
-       }
-
-       if (ap->logbufs != -1 &&
-           ap->logbufs != 0 &&
-           (ap->logbufs < XLOG_MIN_ICLOGS ||
-            ap->logbufs > XLOG_MAX_ICLOGS)) {
-               cmn_err(CE_WARN,
-                       "XFS: invalid logbufs value: %d [not %d-%d]",
-                       ap->logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
-               return XFS_ERROR(EINVAL);
-       }
-       mp->m_logbufs = ap->logbufs;
-       if (ap->logbufsize != -1 &&
-           ap->logbufsize !=  0 &&
-           ap->logbufsize != 16 * 1024 &&
-           ap->logbufsize != 32 * 1024 &&
-           ap->logbufsize != 64 * 1024 &&
-           ap->logbufsize != 128 * 1024 &&
-           ap->logbufsize != 256 * 1024) {
-               cmn_err(CE_WARN,
-       "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
-                       ap->logbufsize);
-               return XFS_ERROR(EINVAL);
-       }
-       mp->m_ihsize = ap->ihashsize;
-       mp->m_logbsize = ap->logbufsize;
-       mp->m_fsname_len = strlen(ap->fsname) + 1;
-       mp->m_fsname = kmem_alloc(mp->m_fsname_len, KM_SLEEP);
-       strcpy(mp->m_fsname, ap->fsname);
-       if (ap->rtname[0]) {
-               mp->m_rtname = kmem_alloc(strlen(ap->rtname) + 1, KM_SLEEP);
-               strcpy(mp->m_rtname, ap->rtname);
-       }
-       if (ap->logname[0]) {
-               mp->m_logname = kmem_alloc(strlen(ap->logname) + 1, KM_SLEEP);
-               strcpy(mp->m_logname, ap->logname);
-       }
-
-       if (ap->flags & XFSMNT_WSYNC)
-               mp->m_flags |= XFS_MOUNT_WSYNC;
-#if XFS_BIG_INUMS
-       if (ap->flags & XFSMNT_INO64) {
-               mp->m_flags |= XFS_MOUNT_INO64;
-               mp->m_inoadd = XFS_INO64_OFFSET;
-       }
-#endif
-       if (ap->flags & XFSMNT_RETERR)
-               mp->m_flags |= XFS_MOUNT_RETERR;
-       if (ap->flags & XFSMNT_NOALIGN)
-               mp->m_flags |= XFS_MOUNT_NOALIGN;
-       if (ap->flags & XFSMNT_SWALLOC)
-               mp->m_flags |= XFS_MOUNT_SWALLOC;
-       if (ap->flags & XFSMNT_OSYNCISOSYNC)
-               mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
-       if (ap->flags & XFSMNT_32BITINODES)
-               mp->m_flags |= XFS_MOUNT_32BITINODES;
-
-       if (ap->flags & XFSMNT_IOSIZE) {
-               if (ap->iosizelog > XFS_MAX_IO_LOG ||
-                   ap->iosizelog < XFS_MIN_IO_LOG) {
-                       cmn_err(CE_WARN,
-               "XFS: invalid log iosize: %d [not %d-%d]",
-                               ap->iosizelog, XFS_MIN_IO_LOG,
-                               XFS_MAX_IO_LOG);
-                       return XFS_ERROR(EINVAL);
-               }
-
-               mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
-               mp->m_readio_log = mp->m_writeio_log = ap->iosizelog;
-       }
-
-       if (ap->flags & XFSMNT_IHASHSIZE)
-               mp->m_flags |= XFS_MOUNT_IHASHSIZE;
-       if (ap->flags & XFSMNT_IDELETE)
-               mp->m_flags |= XFS_MOUNT_IDELETE;
-       if (ap->flags & XFSMNT_DIRSYNC)
-               mp->m_flags |= XFS_MOUNT_DIRSYNC;
-       if (ap->flags & XFSMNT_ATTR2)
-               mp->m_flags |= XFS_MOUNT_ATTR2;
-
-       if (ap->flags2 & XFSMNT2_COMPAT_IOSIZE)
-               mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
-
-       /*
-        * no recovery flag requires a read-only mount
-        */
-       if (ap->flags & XFSMNT_NORECOVERY) {
-               if (!(vfs->vfs_flag & VFS_RDONLY)) {
-                       cmn_err(CE_WARN,
-       "XFS: tried to mount a FS read-write without recovery!");
-                       return XFS_ERROR(EINVAL);
-               }
-               mp->m_flags |= XFS_MOUNT_NORECOVERY;
-       }
-
-       if (ap->flags & XFSMNT_NOUUID)
-               mp->m_flags |= XFS_MOUNT_NOUUID;
-       if (ap->flags & XFSMNT_BARRIER)
-               mp->m_flags |= XFS_MOUNT_BARRIER;
-       else
-               mp->m_flags &= ~XFS_MOUNT_BARRIER;
-
-       return 0;
-}
-
-/*
- * This function fills in xfs_mount_t fields based on mount args.
- * Note: the superblock _has_ now been read in.
- */
-STATIC int
-xfs_finish_flags(
-       struct bhv_vfs          *vfs,
-       struct xfs_mount_args   *ap,
-       struct xfs_mount        *mp)
-{
-       int                     ronly = (vfs->vfs_flag & VFS_RDONLY);
-
-       /* Fail a mount where the logbuf is smaller then the log stripe */
-       if (XFS_SB_VERSION_HASLOGV2(&mp->m_sb)) {
-               if ((ap->logbufsize <= 0) &&
-                   (mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE)) {
-                       mp->m_logbsize = mp->m_sb.sb_logsunit;
-               } else if (ap->logbufsize > 0 &&
-                          ap->logbufsize < mp->m_sb.sb_logsunit) {
-                       cmn_err(CE_WARN,
-       "XFS: logbuf size must be greater than or equal to log stripe size");
-                       return XFS_ERROR(EINVAL);
-               }
-       } else {
-               /* Fail a mount if the logbuf is larger than 32K */
-               if (ap->logbufsize > XLOG_BIG_RECORD_BSIZE) {
-                       cmn_err(CE_WARN,
-       "XFS: logbuf size for version 1 logs must be 16K or 32K");
-                       return XFS_ERROR(EINVAL);
-               }
-       }
-
-       if (XFS_SB_VERSION_HASATTR2(&mp->m_sb)) {
-               mp->m_flags |= XFS_MOUNT_ATTR2;
-       }
-
-       /*
-        * prohibit r/w mounts of read-only filesystems
-        */
-       if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
-               cmn_err(CE_WARN,
-       "XFS: cannot mount a read-only filesystem as read-write");
-               return XFS_ERROR(EROFS);
-       }
-
-       /*
-        * check for shared mount.
-        */
-       if (ap->flags & XFSMNT_SHARED) {
-               if (!XFS_SB_VERSION_HASSHARED(&mp->m_sb))
-                       return XFS_ERROR(EINVAL);
-
-               /*
-                * For IRIX 6.5, shared mounts must have the shared
-                * version bit set, have the persistent readonly
-                * field set, must be version 0 and can only be mounted
-                * read-only.
-                */
-               if (!ronly || !(mp->m_sb.sb_flags & XFS_SBF_READONLY) ||
-                    (mp->m_sb.sb_shared_vn != 0))
-                       return XFS_ERROR(EINVAL);
-
-               mp->m_flags |= XFS_MOUNT_SHARED;
-
-               /*
-                * Shared XFS V0 can't deal with DMI.  Return EINVAL.
-                */
-               if (mp->m_sb.sb_shared_vn == 0 && (ap->flags & XFSMNT_DMAPI))
-                       return XFS_ERROR(EINVAL);
-       }
-
-       return 0;
-}
-
-/*
- * xfs_mount
- *
- * The file system configurations are:
- *     (1) device (partition) with data and internal log
- *     (2) logical volume with data and log subvolumes.
- *     (3) logical volume with data, log, and realtime subvolumes.
- *
- * We only have to handle opening the log and realtime volumes here if
- * they are present.  The data subvolume has already been opened by
- * get_sb_bdev() and is stored in vfsp->vfs_super->s_bdev.
- */
-STATIC int
-xfs_mount(
-       struct bhv_desc         *bhvp,
-       struct xfs_mount_args   *args,
-       cred_t                  *credp)
-{
-       struct bhv_vfs          *vfsp = bhvtovfs(bhvp);
-       struct bhv_desc         *p;
-       struct xfs_mount        *mp = XFS_BHVTOM(bhvp);
-       struct block_device     *ddev, *logdev, *rtdev;
-       int                     flags = 0, error;
-
-       ddev = vfsp->vfs_super->s_bdev;
-       logdev = rtdev = NULL;
-
-       /*
-        * Setup xfs_mount function vectors from available behaviors
-        */
-       p = vfs_bhv_lookup(vfsp, VFS_POSITION_DM);
-       mp->m_dm_ops = p ? *(xfs_dmops_t *) vfs_bhv_custom(p) : xfs_dmcore_stub;
-       p = vfs_bhv_lookup(vfsp, VFS_POSITION_QM);
-       mp->m_qm_ops = p ? *(xfs_qmops_t *) vfs_bhv_custom(p) : xfs_qmcore_stub;
-       p = vfs_bhv_lookup(vfsp, VFS_POSITION_IO);
-       mp->m_io_ops = p ? *(xfs_ioops_t *) vfs_bhv_custom(p) : xfs_iocore_xfs;
-
-       if (args->flags & XFSMNT_QUIET)
-               flags |= XFS_MFSI_QUIET;
-
-       /*
-        * Open real time and log devices - order is important.
-        */
-       if (args->logname[0]) {
-               error = xfs_blkdev_get(mp, args->logname, &logdev);
-               if (error)
-                       return error;
-       }
-       if (args->rtname[0]) {
-               error = xfs_blkdev_get(mp, args->rtname, &rtdev);
-               if (error) {
-                       xfs_blkdev_put(logdev);
-                       return error;
-               }
-
-               if (rtdev == ddev || rtdev == logdev) {
-                       cmn_err(CE_WARN,
-       "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
-                       xfs_blkdev_put(logdev);
-                       xfs_blkdev_put(rtdev);
-                       return EINVAL;
-               }
-       }
-
-       /*
-        * Setup xfs_mount buffer target pointers
-        */
-       error = ENOMEM;
-       mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
-       if (!mp->m_ddev_targp) {
-               xfs_blkdev_put(logdev);
-               xfs_blkdev_put(rtdev);
-               return error;
-       }
-       if (rtdev) {
-               mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
-               if (!mp->m_rtdev_targp)
-                       goto error0;
-       }
-       mp->m_logdev_targp = (logdev && logdev != ddev) ?
-                               xfs_alloc_buftarg(logdev, 1) : mp->m_ddev_targp;
-       if (!mp->m_logdev_targp)
-               goto error0;
-
-       /*
-        * Setup flags based on mount(2) options and then the superblock
-        */
-       error = xfs_start_flags(vfsp, args, mp);
-       if (error)
-               goto error1;
-       error = xfs_readsb(mp, flags);
-       if (error)
-               goto error1;
-       error = xfs_finish_flags(vfsp, args, mp);
-       if (error)
-               goto error2;
-
-       /*
-        * Setup xfs_mount buffer target pointers based on superblock
-        */
-       error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
-                                   mp->m_sb.sb_sectsize);
-       if (!error && logdev && logdev != ddev) {
-               unsigned int    log_sector_size = BBSIZE;
-
-               if (XFS_SB_VERSION_HASSECTOR(&mp->m_sb))
-                       log_sector_size = mp->m_sb.sb_logsectsize;
-               error = xfs_setsize_buftarg(mp->m_logdev_targp,
-                                           mp->m_sb.sb_blocksize,
-                                           log_sector_size);
-       }
-       if (!error && rtdev)
-               error = xfs_setsize_buftarg(mp->m_rtdev_targp,
-                                           mp->m_sb.sb_blocksize,
-                                           mp->m_sb.sb_sectsize);
-       if (error)
-               goto error2;
-
-       if (mp->m_flags & XFS_MOUNT_BARRIER)
-               xfs_mountfs_check_barriers(mp);
-
-       error = XFS_IOINIT(vfsp, args, flags);
-       if (error)
-               goto error2;
-
-       return 0;
-
-error2:
-       if (mp->m_sb_bp)
-               xfs_freesb(mp);
-error1:
-       xfs_binval(mp->m_ddev_targp);
-       if (logdev && logdev != ddev)
-               xfs_binval(mp->m_logdev_targp);
-       if (rtdev)
-               xfs_binval(mp->m_rtdev_targp);
-error0:
-       xfs_unmountfs_close(mp, credp);
-       return error;
-}
-
-STATIC int
-xfs_unmount(
-       bhv_desc_t      *bdp,
-       int             flags,
-       cred_t          *credp)
-{
-       bhv_vfs_t       *vfsp = bhvtovfs(bdp);
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
-       xfs_inode_t     *rip;
-       bhv_vnode_t     *rvp;
-       int             unmount_event_wanted = 0;
-       int             unmount_event_flags = 0;
-       int             xfs_unmountfs_needed = 0;
-       int             error;
-
-       rip = mp->m_rootip;
-       rvp = XFS_ITOV(rip);
-
-       if (vfsp->vfs_flag & VFS_DMI) {
-               error = XFS_SEND_PREUNMOUNT(mp, vfsp,
-                               rvp, DM_RIGHT_NULL, rvp, DM_RIGHT_NULL,
-                               NULL, NULL, 0, 0,
-                               (mp->m_dmevmask & (1<<DM_EVENT_PREUNMOUNT))?
-                                       0:DM_FLAGS_UNWANTED);
-                       if (error)
-                               return XFS_ERROR(error);
-               unmount_event_wanted = 1;
-               unmount_event_flags = (mp->m_dmevmask & (1<<DM_EVENT_UNMOUNT))?
-                                       0 : DM_FLAGS_UNWANTED;
-       }
-
-       /*
-        * First blow any referenced inode from this file system
-        * out of the reference cache, and delete the timer.
-        */
-       xfs_refcache_purge_mp(mp);
-
-       XFS_bflush(mp->m_ddev_targp);
-       error = xfs_unmount_flush(mp, 0);
-       if (error)
-               goto out;
-
-       ASSERT(vn_count(rvp) == 1);
-
-       /*
-        * Drop the reference count
-        */
-       VN_RELE(rvp);
-
-       /*
-        * If we're forcing a shutdown, typically because of a media error,
-        * we want to make sure we invalidate dirty pages that belong to
-        * referenced vnodes as well.
-        */
-       if (XFS_FORCED_SHUTDOWN(mp)) {
-               error = xfs_sync(&mp->m_bhv,
-                        (SYNC_WAIT | SYNC_CLOSE), credp);
-               ASSERT(error != EFSCORRUPTED);
-       }
-       xfs_unmountfs_needed = 1;
-
-out:
-       /*      Send DMAPI event, if required.
-        *      Then do xfs_unmountfs() if needed.
-        *      Then return error (or zero).
-        */
-       if (unmount_event_wanted) {
-               /* Note: mp structure must still exist for
-                * XFS_SEND_UNMOUNT() call.
-                */
-               XFS_SEND_UNMOUNT(mp, vfsp, error == 0 ? rvp : NULL,
-                       DM_RIGHT_NULL, 0, error, unmount_event_flags);
-       }
-       if (xfs_unmountfs_needed) {
-               /*
-                * Call common unmount function to flush to disk
-                * and free the super block buffer & mount structures.
-                */
-               xfs_unmountfs(mp, credp);
-       }
-
-       return XFS_ERROR(error);
-}
-
-STATIC int
+STATIC void
 xfs_quiesce_fs(
        xfs_mount_t             *mp)
 {
        int                     count = 0, pincount;
 
-       xfs_refcache_purge_mp(mp);
        xfs_flush_buftarg(mp->m_ddev_targp, 0);
        xfs_finish_reclaim_all(mp, 0);
 
@@ -640,43 +75,43 @@ xfs_quiesce_fs(
         * we can write the unmount record.
         */
        do {
-               xfs_syncsub(mp, SYNC_REMOUNT|SYNC_ATTR|SYNC_WAIT, NULL);
+               xfs_syncsub(mp, SYNC_INODE_QUIESCE, NULL);
                pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
                if (!pincount) {
                        delay(50);
                        count++;
                }
        } while (count < 2);
-
-       return 0;
 }
 
-STATIC int
-xfs_mntupdate(
-       bhv_desc_t                      *bdp,
-       int                             *flags,
-       struct xfs_mount_args           *args)
+/*
+ * Second stage of a quiesce. The data is already synced, now we have to take
+ * care of the metadata. New transactions are already blocked, so we need to
+ * wait for any remaining transactions to drain out before proceding.
+ */
+void
+xfs_attr_quiesce(
+       xfs_mount_t     *mp)
 {
-       bhv_vfs_t       *vfsp = bhvtovfs(bdp);
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
-
-       if (!(*flags & MS_RDONLY)) {                    /* rw/ro -> rw */
-               if (vfsp->vfs_flag & VFS_RDONLY)
-                       vfsp->vfs_flag &= ~VFS_RDONLY;
-               if (args->flags & XFSMNT_BARRIER) {
-                       mp->m_flags |= XFS_MOUNT_BARRIER;
-                       xfs_mountfs_check_barriers(mp);
-               } else {
-                       mp->m_flags &= ~XFS_MOUNT_BARRIER;
-               }
-       } else if (!(vfsp->vfs_flag & VFS_RDONLY)) {    /* rw -> ro */
-               bhv_vfs_sync(vfsp, SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR, NULL);
-               xfs_quiesce_fs(mp);
-               xfs_log_unmount_write(mp);
-               xfs_unmountfs_writesb(mp);
-               vfsp->vfs_flag |= VFS_RDONLY;
-       }
-       return 0;
+       int     error = 0;
+
+       /* wait for all modifications to complete */
+       while (atomic_read(&mp->m_active_trans) > 0)
+               delay(100);
+
+       /* flush inodes and push all remaining buffers out to disk */
+       xfs_quiesce_fs(mp);
+
+       ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);
+
+       /* Push the superblock and write an unmount record */
+       error = xfs_log_sbcount(mp, 1);
+       if (error)
+               xfs_fs_cmn_err(CE_WARN, mp,
+                               "xfs_attr_quiesce: failed to log sb changes. "
+                               "Frozen image may not be consistent.");
+       xfs_log_unmount_write(mp);
+       xfs_unmountfs_writesb(mp);
 }
 
 /*
@@ -693,10 +128,9 @@ xfs_unmount_flush(
        xfs_inode_t     *rip = mp->m_rootip;
        xfs_inode_t     *rbmip;
        xfs_inode_t     *rsumip = NULL;
-       bhv_vnode_t     *rvp = XFS_ITOV(rip);
        int             error;
 
-       xfs_ilock(rip, XFS_ILOCK_EXCL);
+       xfs_ilock(rip, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
        xfs_iflock(rip);
 
        /*
@@ -711,7 +145,7 @@ xfs_unmount_flush(
                if (error == EFSCORRUPTED)
                        goto fscorrupt_out;
 
-               ASSERT(vn_count(XFS_ITOV(rbmip)) == 1);
+               ASSERT(vn_count(VFS_I(rbmip)) == 1);
 
                rsumip = mp->m_rsumip;
                xfs_ilock(rsumip, XFS_ILOCK_EXCL);
@@ -722,7 +156,7 @@ xfs_unmount_flush(
                if (error == EFSCORRUPTED)
                        goto fscorrupt_out;
 
-               ASSERT(vn_count(XFS_ITOV(rsumip)) == 1);
+               ASSERT(vn_count(VFS_I(rsumip)) == 1);
        }
 
        /*
@@ -732,7 +166,7 @@ xfs_unmount_flush(
        if (error == EFSCORRUPTED)
                goto fscorrupt_out2;
 
-       if (vn_count(rvp) != 1 && !relocation) {
+       if (vn_count(VFS_I(rip)) != 1 && !relocation) {
                xfs_iunlock(rip, XFS_ILOCK_EXCL);
                return XFS_ERROR(EBUSY);
        }
@@ -746,8 +180,8 @@ xfs_unmount_flush(
                goto fscorrupt_out2;
 
        if (rbmip) {
-               VN_RELE(XFS_ITOV(rbmip));
-               VN_RELE(XFS_ITOV(rsumip));
+               IRELE(rbmip);
+               IRELE(rsumip);
        }
 
        xfs_iunlock(rip, XFS_ILOCK_EXCL);
@@ -763,80 +197,6 @@ fscorrupt_out2:
 }
 
 /*
- * xfs_root extracts the root vnode from a vfs.
- *
- * vfsp -- the vfs struct for the desired file system
- * vpp  -- address of the caller's vnode pointer which should be
- *         set to the desired fs root vnode
- */
-STATIC int
-xfs_root(
-       bhv_desc_t      *bdp,
-       bhv_vnode_t     **vpp)
-{
-       bhv_vnode_t     *vp;
-
-       vp = XFS_ITOV((XFS_BHVTOM(bdp))->m_rootip);
-       VN_HOLD(vp);
-       *vpp = vp;
-       return 0;
-}
-
-/*
- * xfs_statvfs
- *
- * Fill in the statvfs structure for the given file system.  We use
- * the superblock lock in the mount structure to ensure a consistent
- * snapshot of the counters returned.
- */
-STATIC int
-xfs_statvfs(
-       bhv_desc_t      *bdp,
-       bhv_statvfs_t   *statp,
-       bhv_vnode_t     *vp)
-{
-       __uint64_t      fakeinos;
-       xfs_extlen_t    lsize;
-       xfs_mount_t     *mp;
-       xfs_sb_t        *sbp;
-       unsigned long   s;
-
-       mp = XFS_BHVTOM(bdp);
-       sbp = &(mp->m_sb);
-
-       statp->f_type = XFS_SB_MAGIC;
-
-       xfs_icsb_sync_counters_flags(mp, XFS_ICSB_LAZY_COUNT);
-       s = XFS_SB_LOCK(mp);
-       statp->f_bsize = sbp->sb_blocksize;
-       lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
-       statp->f_blocks = sbp->sb_dblocks - lsize;
-       statp->f_bfree = statp->f_bavail =
-                               sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
-       fakeinos = statp->f_bfree << sbp->sb_inopblog;
-#if XFS_BIG_INUMS
-       fakeinos += mp->m_inoadd;
-#endif
-       statp->f_files =
-           MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
-       if (mp->m_maxicount)
-#if XFS_BIG_INUMS
-               if (!mp->m_inoadd)
-#endif
-                       statp->f_files = min_t(typeof(statp->f_files),
-                                               statp->f_files,
-                                               mp->m_maxicount);
-       statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
-       XFS_SB_UNLOCK(mp, s);
-
-       xfs_statvfs_fsid(statp, mp);
-       statp->f_namelen = MAXNAMELEN - 1;
-
-       return 0;
-}
-
-
-/*
  * xfs_sync flushes any pending I/O to file system vfsp.
  *
  * This routine is called by vfs_sync() to make sure that things make it
@@ -878,14 +238,33 @@ xfs_statvfs(
  *                    filesystem.
  *
  */
-/*ARGSUSED*/
-STATIC int
+int
 xfs_sync(
-       bhv_desc_t      *bdp,
-       int             flags,
-       cred_t          *credp)
+       xfs_mount_t     *mp,
+       int             flags)
 {
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
+       int             error;
+
+       /*
+        * Get the Quota Manager to flush the dquots.
+        *
+        * If XFS quota support is not enabled or this filesystem
+        * instance does not use quotas XFS_QM_DQSYNC will always
+        * return zero.
+        */
+       error = XFS_QM_DQSYNC(mp, flags);
+       if (error) {
+               /*
+                * If we got an IO error, we will be shutting down.
+                * So, there's nothing more for us to do here.
+                */
+               ASSERT(error != EIO || XFS_FORCED_SHUTDOWN(mp));
+               if (XFS_FORCED_SHUTDOWN(mp))
+                       return XFS_ERROR(error);
+       }
+
+       if (flags & SYNC_IOWAIT)
+               xfs_filestream_flush(mp);
 
        return xfs_syncsub(mp, flags, NULL);
 }
@@ -904,9 +283,7 @@ xfs_sync_inodes(
        int             *bypassed)
 {
        xfs_inode_t     *ip = NULL;
-       xfs_inode_t     *ip_next;
-       xfs_buf_t       *bp;
-       bhv_vnode_t     *vp = NULL;
+       struct inode    *vp = NULL;
        int             error;
        int             last_error;
        uint64_t        fflag;
@@ -915,7 +292,6 @@ xfs_sync_inodes(
        boolean_t       mount_locked;
        boolean_t       vnode_refed;
        int             preempt;
-       xfs_dinode_t    *dip;
        xfs_iptr_t      *ipointer;
 #ifdef DEBUG
        boolean_t       ipointer_in = B_FALSE;
@@ -968,9 +344,11 @@ xfs_sync_inodes(
 
 #define XFS_PREEMPT_MASK       0x7f
 
+       ASSERT(!(flags & SYNC_BDFLUSH));
+
        if (bypassed)
                *bypassed = 0;
-       if (XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY)
+       if (mp->m_flags & XFS_MOUNT_RDONLY)
                return 0;
        error = 0;
        last_error = 0;
@@ -980,7 +358,7 @@ xfs_sync_inodes(
        ipointer = (xfs_iptr_t *)kmem_zalloc(sizeof(xfs_iptr_t), KM_SLEEP);
 
        fflag = XFS_B_ASYNC;            /* default is don't wait */
-       if (flags & (SYNC_BDFLUSH | SYNC_DELWRI))
+       if (flags & SYNC_DELWRI)
                fflag = XFS_B_DELWRI;
        if (flags & SYNC_WAIT)
                fflag = 0;              /* synchronous overrides all */
@@ -1025,7 +403,7 @@ xfs_sync_inodes(
                        continue;
                }
 
-               vp = XFS_ITOV_NULL(ip);
+               vp = VFS_I(ip);
 
                /*
                 * If the vnode is gone then this is being torn down,
@@ -1065,29 +443,11 @@ xfs_sync_inodes(
 
                if (XFS_FORCED_SHUTDOWN(mp) && !(flags & SYNC_CLOSE)) {
                        XFS_MOUNT_IUNLOCK(mp);
-                       kmem_free(ipointer, sizeof(xfs_iptr_t));
+                       kmem_free(ipointer);
                        return 0;
                }
 
                /*
-                * If this is just vfs_sync() or pflushd() calling
-                * then we can skip inodes for which it looks like
-                * there is nothing to do.  Since we don't have the
-                * inode locked this is racy, but these are periodic
-                * calls so it doesn't matter.  For the others we want
-                * to know for sure, so we at least try to lock them.
-                */
-               if (flags & SYNC_BDFLUSH) {
-                       if (((ip->i_itemp == NULL) ||
-                            !(ip->i_itemp->ili_format.ilf_fields &
-                              XFS_ILOG_ALL)) &&
-                           (ip->i_update_core == 0)) {
-                               ip = ip->i_mnext;
-                               continue;
-                       }
-               }
-
-               /*
                 * Try to lock without sleeping.  We're out of order with
                 * the inode list lock here, so if we fail we need to drop
                 * the mount lock and try again.  If we're called from
@@ -1104,7 +464,7 @@ xfs_sync_inodes(
                 * it.
                 */
                if (xfs_ilock_nowait(ip, lock_flags) == 0) {
-                       if ((flags & SYNC_BDFLUSH) || (vp == NULL)) {
+                       if (vp == NULL) {
                                ip = ip->i_mnext;
                                continue;
                        }
@@ -1118,7 +478,7 @@ xfs_sync_inodes(
                        IPOINTER_INSERT(ip, mp);
                        xfs_ilock(ip, lock_flags);
 
-                       ASSERT(vp == XFS_ITOV(ip));
+                       ASSERT(vp == VFS_I(ip));
                        ASSERT(ip->i_mount == mp);
 
                        vnode_refed = B_TRUE;
@@ -1128,213 +488,64 @@ xfs_sync_inodes(
                 * in the inode list.
                 */
 
-               if ((flags & SYNC_CLOSE)  && (vp != NULL)) {
-                       /*
-                        * This is the shutdown case.  We just need to
-                        * flush and invalidate all the pages associated
-                        * with the inode.  Drop the inode lock since
-                        * we can't hold it across calls to the buffer
-                        * cache.
-                        *
-                        * We don't set the VREMAPPING bit in the vnode
-                        * here, because we don't hold the vnode lock
-                        * exclusively.  It doesn't really matter, though,
-                        * because we only come here when we're shutting
-                        * down anyway.
-                        */
-                       xfs_iunlock(ip, XFS_ILOCK_SHARED);
-
-                       if (XFS_FORCED_SHUTDOWN(mp)) {
-                               bhv_vop_toss_pages(vp, 0, -1, FI_REMAPF);
-                       } else {
-                               bhv_vop_flushinval_pages(vp, 0, -1, FI_REMAPF);
+               /*
+                * If we have to flush data or wait for I/O completion
+                * we need to drop the ilock that we currently hold.
+                * If we need to drop the lock, insert a marker if we
+                * have not already done so.
+                */
+               if ((flags & (SYNC_CLOSE|SYNC_IOWAIT)) ||
+                   ((flags & SYNC_DELWRI) && VN_DIRTY(vp))) {
+                       if (mount_locked) {
+                               IPOINTER_INSERT(ip, mp);
                        }
+                       xfs_iunlock(ip, XFS_ILOCK_SHARED);
 
-                       xfs_ilock(ip, XFS_ILOCK_SHARED);
-
-               } else if ((flags & SYNC_DELWRI) && (vp != NULL)) {
-                       if (VN_DIRTY(vp)) {
-                               /* We need to have dropped the lock here,
-                                * so insert a marker if we have not already
-                                * done so.
-                                */
-                               if (mount_locked) {
-                                       IPOINTER_INSERT(ip, mp);
-                               }
-
-                               /*
-                                * Drop the inode lock since we can't hold it
-                                * across calls to the buffer cache.
-                                */
-                               xfs_iunlock(ip, XFS_ILOCK_SHARED);
-                               error = bhv_vop_flush_pages(vp, (xfs_off_t)0,
+                       if (flags & SYNC_CLOSE) {
+                               /* Shutdown case. Flush and invalidate. */
+                               if (XFS_FORCED_SHUTDOWN(mp))
+                                       xfs_tosspages(ip, 0, -1,
+                                                            FI_REMAPF);
+                               else
+                                       error = xfs_flushinval_pages(ip,
+                                                       0, -1, FI_REMAPF);
+                       } else if ((flags & SYNC_DELWRI) && VN_DIRTY(vp)) {
+                               error = xfs_flush_pages(ip, 0,
                                                        -1, fflag, FI_NONE);
-                               xfs_ilock(ip, XFS_ILOCK_SHARED);
                        }
 
-               }
-               /*
-                * When freezing, we need to wait ensure all I/O (including direct
-                * I/O) is complete to ensure no further data modification can take
-                * place after this point
-                */
-               if (flags & SYNC_IOWAIT)
-                       vn_iowait(vp);
-
-               if (flags & SYNC_BDFLUSH) {
-                       if ((flags & SYNC_ATTR) &&
-                           ((ip->i_update_core) ||
-                            ((ip->i_itemp != NULL) &&
-                             (ip->i_itemp->ili_format.ilf_fields != 0)))) {
-
-                               /* Insert marker and drop lock if not already
-                                * done.
-                                */
-                               if (mount_locked) {
-                                       IPOINTER_INSERT(ip, mp);
-                               }
-
-                               /*
-                                * We don't want the periodic flushing of the
-                                * inodes by vfs_sync() to interfere with
-                                * I/O to the file, especially read I/O
-                                * where it is only the access time stamp
-                                * that is being flushed out.  To prevent
-                                * long periods where we have both inode
-                                * locks held shared here while reading the
-                                * inode's buffer in from disk, we drop the
-                                * inode lock while reading in the inode
-                                * buffer.  We have to release the buffer
-                                * and reacquire the inode lock so that they
-                                * are acquired in the proper order (inode
-                                * locks first).  The buffer will go at the
-                                * end of the lru chain, though, so we can
-                                * expect it to still be there when we go
-                                * for it again in xfs_iflush().
-                                */
-                               if ((xfs_ipincount(ip) == 0) &&
-                                   xfs_iflock_nowait(ip)) {
-
-                                       xfs_ifunlock(ip);
-                                       xfs_iunlock(ip, XFS_ILOCK_SHARED);
-
-                                       error = xfs_itobp(mp, NULL, ip,
-                                                         &dip, &bp, 0, 0);
-                                       if (!error) {
-                                               xfs_buf_relse(bp);
-                                       } else {
-                                               /* Bailing out, remove the
-                                                * marker and free it.
-                                                */
-                                               XFS_MOUNT_ILOCK(mp);
-                                               IPOINTER_REMOVE(ip, mp);
-                                               XFS_MOUNT_IUNLOCK(mp);
-
-                                               ASSERT(!(lock_flags &
-                                                       XFS_IOLOCK_SHARED));
-
-                                               kmem_free(ipointer,
-                                                       sizeof(xfs_iptr_t));
-                                               return (0);
-                                       }
-
-                                       /*
-                                        * Since we dropped the inode lock,
-                                        * the inode may have been reclaimed.
-                                        * Therefore, we reacquire the mount
-                                        * lock and check to see if we were the
-                                        * inode reclaimed. If this happened
-                                        * then the ipointer marker will no
-                                        * longer point back at us. In this
-                                        * case, move ip along to the inode
-                                        * after the marker, remove the marker
-                                        * and continue.
-                                        */
-                                       XFS_MOUNT_ILOCK(mp);
-                                       mount_locked = B_TRUE;
-
-                                       if (ip != ipointer->ip_mprev) {
-                                               IPOINTER_REMOVE(ip, mp);
-
-                                               ASSERT(!vnode_refed);
-                                               ASSERT(!(lock_flags &
-                                                       XFS_IOLOCK_SHARED));
-                                               continue;
-                                       }
+                       /*
+                        * When freezing, we need to wait ensure all I/O (including direct
+                        * I/O) is complete to ensure no further data modification can take
+                        * place after this point
+                        */
+                       if (flags & SYNC_IOWAIT)
+                               vn_iowait(ip);
 
-                                       ASSERT(ip->i_mount == mp);
-
-                                       if (xfs_ilock_nowait(ip,
-                                                   XFS_ILOCK_SHARED) == 0) {
-                                               ASSERT(ip->i_mount == mp);
-                                               /*
-                                                * We failed to reacquire
-                                                * the inode lock without
-                                                * sleeping, so just skip
-                                                * the inode for now.  We
-                                                * clear the ILOCK bit from
-                                                * the lock_flags so that we
-                                                * won't try to drop a lock
-                                                * we don't hold below.
-                                                */
-                                               lock_flags &= ~XFS_ILOCK_SHARED;
-                                               IPOINTER_REMOVE(ip_next, mp);
-                                       } else if ((xfs_ipincount(ip) == 0) &&
-                                                  xfs_iflock_nowait(ip)) {
-                                               ASSERT(ip->i_mount == mp);
-                                               /*
-                                                * Since this is vfs_sync()
-                                                * calling we only flush the
-                                                * inode out if we can lock
-                                                * it without sleeping and
-                                                * it is not pinned.  Drop
-                                                * the mount lock here so
-                                                * that we don't hold it for
-                                                * too long. We already have
-                                                * a marker in the list here.
-                                                */
-                                               XFS_MOUNT_IUNLOCK(mp);
-                                               mount_locked = B_FALSE;
-                                               error = xfs_iflush(ip,
-                                                          XFS_IFLUSH_DELWRI);
-                                       } else {
-                                               ASSERT(ip->i_mount == mp);
-                                               IPOINTER_REMOVE(ip_next, mp);
-                                       }
-                               }
+                       xfs_ilock(ip, XFS_ILOCK_SHARED);
+               }
 
-                       }
+               if ((flags & SYNC_ATTR) &&
+                   (ip->i_update_core ||
+                    (ip->i_itemp && ip->i_itemp->ili_format.ilf_fields))) {
+                       if (mount_locked)
+                               IPOINTER_INSERT(ip, mp);
 
-               } else {
-                       if ((flags & SYNC_ATTR) &&
-                           ((ip->i_update_core) ||
-                            ((ip->i_itemp != NULL) &&
-                             (ip->i_itemp->ili_format.ilf_fields != 0)))) {
-                               if (mount_locked) {
-                                       IPOINTER_INSERT(ip, mp);
-                               }
+                       if (flags & SYNC_WAIT) {
+                               xfs_iflock(ip);
+                               error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
 
-                               if (flags & SYNC_WAIT) {
-                                       xfs_iflock(ip);
-                                       error = xfs_iflush(ip,
-                                                          XFS_IFLUSH_SYNC);
-                               } else {
-                                       /*
-                                        * If we can't acquire the flush
-                                        * lock, then the inode is already
-                                        * being flushed so don't bother
-                                        * waiting.  If we can lock it then
-                                        * do a delwri flush so we can
-                                        * combine multiple inode flushes
-                                        * in each disk write.
-                                        */
-                                       if (xfs_iflock_nowait(ip)) {
-                                               error = xfs_iflush(ip,
-                                                          XFS_IFLUSH_DELWRI);
-                                       }
-                                       else if (bypassed)
-                                               (*bypassed)++;
-                               }
+                       /*
+                        * If we can't acquire the flush lock, then the inode
+                        * is already being flushed so don't bother waiting.
+                        *
+                        * If we can lock it then do a delwri flush so we can
+                        * combine multiple inode flushes in each disk write.
+                        */
+                       } else if (xfs_iflock_nowait(ip)) {
+                               error = xfs_iflush(ip, XFS_IFLUSH_DELWRI);
+                       } else if (bypassed) {
+                               (*bypassed)++;
                        }
                }
 
@@ -1348,10 +559,10 @@ xfs_sync_inodes(
                         * above, then wait until after we've unlocked
                         * the inode to release the reference.  This is
                         * because we can be already holding the inode
-                        * lock when VN_RELE() calls xfs_inactive().
+                        * lock when IRELE() calls xfs_inactive().
                         *
                         * Make sure to drop the mount lock before calling
-                        * VN_RELE() so that we don't trip over ourselves if
+                        * IRELE() so that we don't trip over ourselves if
                         * we have to go for the mount lock again in the
                         * inactive code.
                         */
@@ -1359,7 +570,7 @@ xfs_sync_inodes(
                                IPOINTER_INSERT(ip, mp);
                        }
 
-                       VN_RELE(vp);
+                       IRELE(ip);
 
                        vnode_refed = B_FALSE;
                }
@@ -1378,7 +589,7 @@ xfs_sync_inodes(
                        }
                        XFS_MOUNT_IUNLOCK(mp);
                        ASSERT(ipointer_in == B_FALSE);
-                       kmem_free(ipointer, sizeof(xfs_iptr_t));
+                       kmem_free(ipointer);
                        return XFS_ERROR(error);
                }
 
@@ -1408,7 +619,7 @@ xfs_sync_inodes(
 
        ASSERT(ipointer_in == B_FALSE);
 
-       kmem_free(ipointer, sizeof(xfs_iptr_t));
+       kmem_free(ipointer);
        return XFS_ERROR(last_error);
 }
 
@@ -1502,21 +713,8 @@ xfs_syncsub(
        }
 
        /*
-        * If this is the periodic sync, then kick some entries out of
-        * the reference cache.  This ensures that idle entries are
-        * eventually kicked out of the cache.
-        */
-       if (flags & SYNC_REFCACHE) {
-               if (flags & SYNC_WAIT)
-                       xfs_refcache_purge_mp(mp);
-               else
-                       xfs_refcache_purge_some(mp);
-       }
-
-       /*
         * Now check to see if the log needs a "dummy" transaction.
         */
-
        if (!(flags & SYNC_REMOUNT) && xfs_log_need_covered(mp)) {
                xfs_trans_t *tp;
                xfs_inode_t *ip;
@@ -1557,448 +755,3 @@ xfs_syncsub(
 
        return XFS_ERROR(last_error);
 }
-
-/*
- * xfs_vget - called by DMAPI and NFSD to get vnode from file handle
- */
-STATIC int
-xfs_vget(
-       bhv_desc_t      *bdp,
-       bhv_vnode_t     **vpp,
-       fid_t           *fidp)
-{
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
-       xfs_fid_t       *xfid = (struct xfs_fid *)fidp;
-       xfs_inode_t     *ip;
-       int             error;
-       xfs_ino_t       ino;
-       unsigned int    igen;
-
-       /*
-        * Invalid.  Since handles can be created in user space and passed in
-        * via gethandle(), this is not cause for a panic.
-        */
-       if (xfid->xfs_fid_len != sizeof(*xfid) - sizeof(xfid->xfs_fid_len))
-               return XFS_ERROR(EINVAL);
-
-       ino  = xfid->xfs_fid_ino;
-       igen = xfid->xfs_fid_gen;
-
-       /*
-        * NFS can sometimes send requests for ino 0.  Fail them gracefully.
-        */
-       if (ino == 0)
-               return XFS_ERROR(ESTALE);
-
-       error = xfs_iget(mp, NULL, ino, 0, XFS_ILOCK_SHARED, &ip, 0);
-       if (error) {
-               *vpp = NULL;
-               return error;
-       }
-
-       if (ip == NULL) {
-               *vpp = NULL;
-               return XFS_ERROR(EIO);
-       }
-
-       if (ip->i_d.di_mode == 0 || ip->i_d.di_gen != igen) {
-               xfs_iput_new(ip, XFS_ILOCK_SHARED);
-               *vpp = NULL;
-               return XFS_ERROR(ENOENT);
-       }
-
-       *vpp = XFS_ITOV(ip);
-       xfs_iunlock(ip, XFS_ILOCK_SHARED);
-       return 0;
-}
-
-
-#define MNTOPT_LOGBUFS "logbufs"       /* number of XFS log buffers */
-#define MNTOPT_LOGBSIZE        "logbsize"      /* size of XFS log buffers */
-#define MNTOPT_LOGDEV  "logdev"        /* log device */
-#define MNTOPT_RTDEV   "rtdev"         /* realtime I/O device */
-#define MNTOPT_BIOSIZE "biosize"       /* log2 of preferred buffered io size */
-#define MNTOPT_WSYNC   "wsync"         /* safe-mode nfs compatible mount */
-#define MNTOPT_INO64   "ino64"         /* force inodes into 64-bit range */
-#define MNTOPT_NOALIGN "noalign"       /* turn off stripe alignment */
-#define MNTOPT_SWALLOC "swalloc"       /* turn on stripe width allocation */
-#define MNTOPT_SUNIT   "sunit"         /* data volume stripe unit */
-#define MNTOPT_SWIDTH  "swidth"        /* data volume stripe width */
-#define MNTOPT_NOUUID  "nouuid"        /* ignore filesystem UUID */
-#define MNTOPT_MTPT    "mtpt"          /* filesystem mount point */
-#define MNTOPT_GRPID   "grpid"         /* group-ID from parent directory */
-#define MNTOPT_NOGRPID "nogrpid"       /* group-ID from current process */
-#define MNTOPT_BSDGROUPS    "bsdgroups"    /* group-ID from parent directory */
-#define MNTOPT_SYSVGROUPS   "sysvgroups"   /* group-ID from current process */
-#define MNTOPT_ALLOCSIZE    "allocsize"    /* preferred allocation size */
-#define MNTOPT_IHASHSIZE    "ihashsize"    /* size of inode hash table */
-#define MNTOPT_NORECOVERY   "norecovery"   /* don't run XFS recovery */
-#define MNTOPT_BARRIER "barrier"       /* use writer barriers for log write and
-                                        * unwritten extent conversion */
-#define MNTOPT_NOBARRIER "nobarrier"   /* .. disable */
-#define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
-#define MNTOPT_64BITINODE   "inode64"  /* inodes can be allocated anywhere */
-#define MNTOPT_IKEEP   "ikeep"         /* do not free empty inode clusters */
-#define MNTOPT_NOIKEEP "noikeep"       /* free empty inode clusters */
-#define MNTOPT_LARGEIO    "largeio"    /* report large I/O sizes in stat() */
-#define MNTOPT_NOLARGEIO   "nolargeio" /* do not report large I/O sizes
-                                        * in stat(). */
-#define MNTOPT_ATTR2   "attr2"         /* do use attr2 attribute format */
-#define MNTOPT_NOATTR2 "noattr2"       /* do not use attr2 attribute format */
-
-STATIC unsigned long
-suffix_strtoul(char *s, char **endp, unsigned int base)
-{
-       int     last, shift_left_factor = 0;
-       char    *value = s;
-
-       last = strlen(value) - 1;
-       if (value[last] == 'K' || value[last] == 'k') {
-               shift_left_factor = 10;
-               value[last] = '\0';
-       }
-       if (value[last] == 'M' || value[last] == 'm') {
-               shift_left_factor = 20;
-               value[last] = '\0';
-       }
-       if (value[last] == 'G' || value[last] == 'g') {
-               shift_left_factor = 30;
-               value[last] = '\0';
-       }
-
-       return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
-}
-
-STATIC int
-xfs_parseargs(
-       struct bhv_desc         *bhv,
-       char                    *options,
-       struct xfs_mount_args   *args,
-       int                     update)
-{
-       bhv_vfs_t               *vfsp = bhvtovfs(bhv);
-       char                    *this_char, *value, *eov;
-       int                     dsunit, dswidth, vol_dsunit, vol_dswidth;
-       int                     iosize;
-
-       args->flags |= XFSMNT_IDELETE;
-       args->flags |= XFSMNT_BARRIER;
-       args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
-
-       if (!options)
-               goto done;
-
-       iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
-
-       while ((this_char = strsep(&options, ",")) != NULL) {
-               if (!*this_char)
-                       continue;
-               if ((value = strchr(this_char, '=')) != NULL)
-                       *value++ = 0;
-
-               if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       args->logbufs = simple_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       args->logbufsize = suffix_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       strncpy(args->logname, value, MAXNAMELEN);
-               } else if (!strcmp(this_char, MNTOPT_MTPT)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       strncpy(args->mtpt, value, MAXNAMELEN);
-               } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       strncpy(args->rtname, value, MAXNAMELEN);
-               } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       iosize = simple_strtoul(value, &eov, 10);
-                       args->flags |= XFSMNT_IOSIZE;
-                       args->iosizelog = (uint8_t) iosize;
-               } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       iosize = suffix_strtoul(value, &eov, 10);
-                       args->flags |= XFSMNT_IOSIZE;
-                       args->iosizelog = ffs(iosize) - 1;
-               } else if (!strcmp(this_char, MNTOPT_IHASHSIZE)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       args->flags |= XFSMNT_IHASHSIZE;
-                       args->ihashsize = simple_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_GRPID) ||
-                          !strcmp(this_char, MNTOPT_BSDGROUPS)) {
-                       vfsp->vfs_flag |= VFS_GRPID;
-               } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
-                          !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
-                       vfsp->vfs_flag &= ~VFS_GRPID;
-               } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
-                       args->flags |= XFSMNT_WSYNC;
-               } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
-                       args->flags |= XFSMNT_OSYNCISOSYNC;
-               } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
-                       args->flags |= XFSMNT_NORECOVERY;
-               } else if (!strcmp(this_char, MNTOPT_INO64)) {
-                       args->flags |= XFSMNT_INO64;
-#if !XFS_BIG_INUMS
-                       cmn_err(CE_WARN,
-                               "XFS: %s option not allowed on this system",
-                               this_char);
-                       return EINVAL;
-#endif
-               } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
-                       args->flags |= XFSMNT_NOALIGN;
-               } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
-                       args->flags |= XFSMNT_SWALLOC;
-               } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       dsunit = simple_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
-                       if (!value || !*value) {
-                               cmn_err(CE_WARN,
-                                       "XFS: %s option requires an argument",
-                                       this_char);
-                               return EINVAL;
-                       }
-                       dswidth = simple_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
-                       args->flags &= ~XFSMNT_32BITINODES;
-#if !XFS_BIG_INUMS
-                       cmn_err(CE_WARN,
-                               "XFS: %s option not allowed on this system",
-                               this_char);
-                       return EINVAL;
-#endif
-               } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
-                       args->flags |= XFSMNT_NOUUID;
-               } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
-                       args->flags |= XFSMNT_BARRIER;
-               } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
-                       args->flags &= ~XFSMNT_BARRIER;
-               } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
-                       args->flags &= ~XFSMNT_IDELETE;
-               } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
-                       args->flags |= XFSMNT_IDELETE;
-               } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
-                       args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
-               } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
-                       args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
-               } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
-                       args->flags |= XFSMNT_ATTR2;
-               } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
-                       args->flags &= ~XFSMNT_ATTR2;
-               } else if (!strcmp(this_char, "osyncisdsync")) {
-                       /* no-op, this is now the default */
-                       cmn_err(CE_WARN,
-       "XFS: osyncisdsync is now the default, option is deprecated.");
-               } else if (!strcmp(this_char, "irixsgid")) {
-                       cmn_err(CE_WARN,
-       "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
-               } else {
-                       cmn_err(CE_WARN,
-                               "XFS: unknown mount option [%s].", this_char);
-                       return EINVAL;
-               }
-       }
-
-       if (args->flags & XFSMNT_NORECOVERY) {
-               if ((vfsp->vfs_flag & VFS_RDONLY) == 0) {
-                       cmn_err(CE_WARN,
-                               "XFS: no-recovery mounts must be read-only.");
-                       return EINVAL;
-               }
-       }
-
-       if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
-               cmn_err(CE_WARN,
-       "XFS: sunit and swidth options incompatible with the noalign option");
-               return EINVAL;
-       }
-
-       if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
-               cmn_err(CE_WARN,
-                       "XFS: sunit and swidth must be specified together");
-               return EINVAL;
-       }
-
-       if (dsunit && (dswidth % dsunit != 0)) {
-               cmn_err(CE_WARN,
-       "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
-                       dswidth, dsunit);
-               return EINVAL;
-       }
-
-       if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
-               if (dsunit) {
-                       args->sunit = dsunit;
-                       args->flags |= XFSMNT_RETERR;
-               } else {
-                       args->sunit = vol_dsunit;
-               }
-               dswidth ? (args->swidth = dswidth) :
-                         (args->swidth = vol_dswidth);
-       } else {
-               args->sunit = args->swidth = 0;
-       }
-
-done:
-       if (args->flags & XFSMNT_32BITINODES)
-               vfsp->vfs_flag |= VFS_32BITINODES;
-       if (args->flags2)
-               args->flags |= XFSMNT_FLAGS2;
-       return 0;
-}
-
-STATIC int
-xfs_showargs(
-       struct bhv_desc         *bhv,
-       struct seq_file         *m)
-{
-       static struct proc_xfs_info {
-               int     flag;
-               char    *str;
-       } xfs_info[] = {
-               /* the few simple ones we can get from the mount struct */
-               { XFS_MOUNT_WSYNC,              "," MNTOPT_WSYNC },
-               { XFS_MOUNT_INO64,              "," MNTOPT_INO64 },
-               { XFS_MOUNT_NOALIGN,            "," MNTOPT_NOALIGN },
-               { XFS_MOUNT_SWALLOC,            "," MNTOPT_SWALLOC },
-               { XFS_MOUNT_NOUUID,             "," MNTOPT_NOUUID },
-               { XFS_MOUNT_NORECOVERY,         "," MNTOPT_NORECOVERY },
-               { XFS_MOUNT_OSYNCISOSYNC,       "," MNTOPT_OSYNCISOSYNC },
-               { 0, NULL }
-       };
-       struct proc_xfs_info    *xfs_infop;
-       struct xfs_mount        *mp = XFS_BHVTOM(bhv);
-       struct bhv_vfs          *vfsp = XFS_MTOVFS(mp);
-
-       for (xfs_infop = xfs_info; xfs_infop->flag; xfs_infop++) {
-               if (mp->m_flags & xfs_infop->flag)
-                       seq_puts(m, xfs_infop->str);
-       }
-
-       if (mp->m_flags & XFS_MOUNT_IHASHSIZE)
-               seq_printf(m, "," MNTOPT_IHASHSIZE "=%d", (int)mp->m_ihsize);
-
-       if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
-               seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
-                               (int)(1 << mp->m_writeio_log) >> 10);
-
-       if (mp->m_logbufs > 0)
-               seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
-       if (mp->m_logbsize > 0)
-               seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
-
-       if (mp->m_logname)
-               seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
-       if (mp->m_rtname)
-               seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
-
-       if (mp->m_dalign > 0)
-               seq_printf(m, "," MNTOPT_SUNIT "=%d",
-                               (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
-       if (mp->m_swidth > 0)
-               seq_printf(m, "," MNTOPT_SWIDTH "=%d",
-                               (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
-
-       if (!(mp->m_flags & XFS_MOUNT_IDELETE))
-               seq_printf(m, "," MNTOPT_IKEEP);
-       if (!(mp->m_flags & XFS_MOUNT_COMPAT_IOSIZE))
-               seq_printf(m, "," MNTOPT_LARGEIO);
-
-       if (!(vfsp->vfs_flag & VFS_32BITINODES))
-               seq_printf(m, "," MNTOPT_64BITINODE);
-       if (vfsp->vfs_flag & VFS_GRPID)
-               seq_printf(m, "," MNTOPT_GRPID);
-
-       return 0;
-}
-
-/*
- * Second stage of a freeze. The data is already frozen, now we have to take
- * care of the metadata. New transactions are already blocked, so we need to
- * wait for any remaining transactions to drain out before proceding.
- */
-STATIC void
-xfs_freeze(
-       bhv_desc_t      *bdp)
-{
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
-
-       /* wait for all modifications to complete */
-       while (atomic_read(&mp->m_active_trans) > 0)
-               delay(100);
-
-       /* flush inodes and push all remaining buffers out to disk */
-       xfs_quiesce_fs(mp);
-
-       ASSERT_ALWAYS(atomic_read(&mp->m_active_trans) == 0);
-
-       /* Push the superblock and write an unmount record */
-       xfs_log_unmount_write(mp);
-       xfs_unmountfs_writesb(mp);
-       xfs_fs_log_dummy(mp);
-}
-
-
-bhv_vfsops_t xfs_vfsops = {
-       BHV_IDENTITY_INIT(VFS_BHV_XFS,VFS_POSITION_XFS),
-       .vfs_parseargs          = xfs_parseargs,
-       .vfs_showargs           = xfs_showargs,
-       .vfs_mount              = xfs_mount,
-       .vfs_unmount            = xfs_unmount,
-       .vfs_mntupdate          = xfs_mntupdate,
-       .vfs_root               = xfs_root,
-       .vfs_statvfs            = xfs_statvfs,
-       .vfs_sync               = xfs_sync,
-       .vfs_vget               = xfs_vget,
-       .vfs_dmapiops           = (vfs_dmapiops_t)fs_nosys,
-       .vfs_quotactl           = (vfs_quotactl_t)fs_nosys,
-       .vfs_init_vnode         = xfs_initialize_vnode,
-       .vfs_force_shutdown     = xfs_do_force_shutdown,
-       .vfs_freeze             = xfs_freeze,
-};