[XFS] remove shouting-indirection macros from xfs_sb.h
[safe/jmp/linux-2.6] / fs / xfs / xfs_vfsops.c
index 00aae9c..7094caf 100644 (file)
@@ -1,88 +1,69 @@
 /*
- * XFS filesystem operations.
+ * Copyright (c) 2000-2005 Silicon Graphics, Inc.
+ * All Rights Reserved.
  *
- * Copyright (c) 2000-2005 Silicon Graphics, Inc.  All Rights Reserved.
- *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of version 2 of the GNU General Public License as
+ * 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.
- *
- * Further, this software is distributed without any warranty that it is
- * free of the rightful claim of any third person regarding infringement
- * or the like.  Any license provided herein, whether implied or
- * otherwise, applies only to this software file.  Patent licenses, if
- * any, provided herein do not apply to combinations of this program with
- * other software, or any other product whatsoever.
- *
- * You should have received a copy of the GNU General Public License along
- * with this program; if not, write the Free Software Foundation, Inc., 59
- * Temple Place - Suite 330, Boston MA 02111-1307, USA.
- *
- * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
- * Mountain View, CA  94043, or:
- *
- * http://www.sgi.com
- *
- * For further information regarding this notice, see:
+ * 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.
  *
- * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
+ * 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_macros.h"
+#include "xfs_fs.h"
 #include "xfs_types.h"
-#include "xfs_inum.h"
+#include "xfs_bit.h"
 #include "xfs_log.h"
+#include "xfs_inum.h"
 #include "xfs_trans.h"
 #include "xfs_sb.h"
-#include "xfs_dir.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_ialloc_btree.h"
 #include "xfs_alloc_btree.h"
-#include "xfs_btree.h"
-#include "xfs_alloc.h"
-#include "xfs_ialloc.h"
-#include "xfs_attr_sf.h"
-#include "xfs_dir_sf.h"
 #include "xfs_dir2_sf.h"
+#include "xfs_attr_sf.h"
 #include "xfs_dinode.h"
-#include "xfs_inode_item.h"
 #include "xfs_inode.h"
-#include "xfs_ag.h"
+#include "xfs_inode_item.h"
+#include "xfs_btree.h"
+#include "xfs_alloc.h"
+#include "xfs_ialloc.h"
+#include "xfs_quota.h"
 #include "xfs_error.h"
 #include "xfs_bmap.h"
-#include "xfs_da_btree.h"
 #include "xfs_rw.h"
 #include "xfs_refcache.h"
 #include "xfs_buf_item.h"
-#include "xfs_extfree_item.h"
-#include "xfs_quota.h"
+#include "xfs_log_priv.h"
 #include "xfs_dir2_trace.h"
+#include "xfs_extfree_item.h"
 #include "xfs_acl.h"
 #include "xfs_attr.h"
 #include "xfs_clnt.h"
-#include "xfs_log_priv.h"
+#include "xfs_mru_cache.h"
+#include "xfs_filestream.h"
+#include "xfs_fsops.h"
+#include "xfs_vnodeops.h"
+#include "xfs_vfsops.h"
 
-STATIC int xfs_sync(bhv_desc_t *, int, cred_t *);
 
-int
+int __init
 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");
+       extern spinlock_t        xfs_dabuf_global_lock;
+       spin_lock_init(&xfs_dabuf_global_lock);
 #endif
 
        /*
@@ -92,11 +73,14 @@ xfs_init(void)
                                                 "xfs_bmap_free_item");
        xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
                                            "xfs_btree_cur");
-       xfs_inode_zone = kmem_zone_init(sizeof(xfs_inode_t), "xfs_inode");
        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");
+       xfs_mru_cache_init();
+       xfs_filestream_init();
 
        /*
         * The size of the zone allocated buf log item is the maximum
@@ -108,17 +92,30 @@ xfs_init(void)
                                (((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_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_zone =
+               kmem_zone_init((sizeof(xfs_efi_log_item_t) +
+                              ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
+                                sizeof(xfs_extent_t))),
                                      "xfs_efi_item");
-       xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
-       xfs_ili_zone = kmem_zone_init(sizeof(xfs_inode_log_item_t), "xfs_ili");
-       xfs_chashlist_zone = kmem_zone_init(sizeof(xfs_chashlist_t),
-                                           "xfs_chashlist");
-       xfs_acl_zone_init(xfs_acl_zone, "xfs_acl");
+
+       /*
+        * 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_icluster_zone =
+               kmem_zone_init_flags(sizeof(xfs_icluster_t), "xfs_icluster",
+                                       KM_ZONE_SPREAD, NULL);
 
        /*
         * Allocate global trace buffers.
@@ -132,9 +129,6 @@ xfs_init(void)
 #ifdef XFS_BMBT_TRACE
        xfs_bmbt_trace_buf = ktrace_alloc(XFS_BMBT_TRACE_SIZE, KM_SLEEP);
 #endif
-#ifdef XFS_DIR_TRACE
-       xfs_dir_trace_buf = ktrace_alloc(XFS_DIR_TRACE_SIZE, KM_SLEEP);
-#endif
 #ifdef XFS_ATTR_TRACE
        xfs_attr_trace_buf = ktrace_alloc(XFS_ATTR_TRACE_SIZE, KM_SLEEP);
 #endif
@@ -153,23 +147,19 @@ xfs_init(void)
        return 0;
 }
 
-void
+void __exit
 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;
+       extern kmem_zone_t      *xfs_icluster_zone;
 
        xfs_cleanup_procfs();
        xfs_sysctl_unregister();
        xfs_refcache_destroy();
+       xfs_filestream_uninit();
+       xfs_mru_cache_uninit();
        xfs_acl_zone_destroy(xfs_acl_zone);
 
 #ifdef XFS_DIR2_TRACE
@@ -178,9 +168,6 @@ xfs_cleanup(void)
 #ifdef XFS_ATTR_TRACE
        ktrace_free(xfs_attr_trace_buf);
 #endif
-#ifdef XFS_DIR_TRACE
-       ktrace_free(xfs_dir_trace_buf);
-#endif
 #ifdef XFS_BMBT_TRACE
        ktrace_free(xfs_bmbt_trace_buf);
 #endif
@@ -191,18 +178,18 @@ xfs_cleanup(void)
        ktrace_free(xfs_alloc_trace_buf);
 #endif
 
-       kmem_cache_destroy(xfs_bmap_free_item_zone);
-       kmem_cache_destroy(xfs_btree_cur_zone);
-       kmem_cache_destroy(xfs_inode_zone);
-       kmem_cache_destroy(xfs_trans_zone);
-       kmem_cache_destroy(xfs_da_state_zone);
-       kmem_cache_destroy(xfs_dabuf_zone);
-       kmem_cache_destroy(xfs_buf_item_zone);
-       kmem_cache_destroy(xfs_efd_zone);
-       kmem_cache_destroy(xfs_efi_zone);
-       kmem_cache_destroy(xfs_ifork_zone);
-       kmem_cache_destroy(xfs_ili_zone);
-       kmem_cache_destroy(xfs_chashlist_zone);
+       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_icluster_zone);
 }
 
 /*
@@ -213,7 +200,6 @@ xfs_cleanup(void)
  */
 STATIC int
 xfs_start_flags(
-       struct vfs              *vfs,
        struct xfs_mount_args   *ap,
        struct xfs_mount        *mp)
 {
@@ -230,9 +216,7 @@ xfs_start_flags(
        }
 
        if (ap->logbufs != -1 &&
-#if defined(DEBUG) || defined(XLOG_NOLOG)
            ap->logbufs != 0 &&
-#endif
            (ap->logbufs < XLOG_MIN_ICLOGS ||
             ap->logbufs > XLOG_MAX_ICLOGS)) {
                cmn_err(CE_WARN,
@@ -242,21 +226,27 @@ xfs_start_flags(
        }
        mp->m_logbufs = ap->logbufs;
        if (ap->logbufsize != -1 &&
-           ap->logbufsize != 16 * 1024 &&
-           ap->logbufsize != 32 * 1024 &&
-           ap->logbufsize != 64 * 1024 &&
-           ap->logbufsize != 128 * 1024 &&
-           ap->logbufsize != 256 * 1024) {
+           ap->logbufsize !=  0 &&
+           (ap->logbufsize < XLOG_MIN_RECORD_BSIZE ||
+            ap->logbufsize > XLOG_MAX_RECORD_BSIZE ||
+            !is_power_of_2(ap->logbufsize))) {
                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;
@@ -266,23 +256,16 @@ xfs_start_flags(
                mp->m_inoadd = XFS_INO64_OFFSET;
        }
 #endif
-       if (ap->flags & XFSMNT_NOATIME)
-               mp->m_flags |= XFS_MOUNT_NOATIME;
-
        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 | XFS_MOUNT_32BITINOOPT);
+               mp->m_flags |= XFS_MOUNT_32BITINODES;
 
        if (ap->flags & XFSMNT_IOSIZE) {
                if (ap->iosizelog > XFS_MAX_IO_LOG ||
@@ -298,20 +281,21 @@ xfs_start_flags(
                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_IKEEP)
+               mp->m_flags |= XFS_MOUNT_IKEEP;
        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)) {
+               if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
                        cmn_err(CE_WARN,
        "XFS: tried to mount a FS read-write without recovery!");
                        return XFS_ERROR(EINVAL);
@@ -321,9 +305,16 @@ xfs_start_flags(
 
        if (ap->flags & XFSMNT_NOUUID)
                mp->m_flags |= XFS_MOUNT_NOUUID;
-       if (ap->flags & XFSMNT_NOLOGFLUSH)
-               mp->m_flags |= XFS_MOUNT_NOLOGFLUSH;
+       if (ap->flags & XFSMNT_BARRIER)
+               mp->m_flags |= XFS_MOUNT_BARRIER;
+       else
+               mp->m_flags &= ~XFS_MOUNT_BARRIER;
+
+       if (ap->flags2 & XFSMNT2_FILESTREAMS)
+               mp->m_flags |= XFS_MOUNT_FILESTREAMS;
 
+       if (ap->flags & XFSMNT_DMAPI)
+               mp->m_flags |= XFS_MOUNT_DMAPI;
        return 0;
 }
 
@@ -333,18 +324,18 @@ xfs_start_flags(
  */
 STATIC int
 xfs_finish_flags(
-       struct vfs              *vfs,
        struct xfs_mount_args   *ap,
        struct xfs_mount        *mp)
 {
-       int                     ronly = (vfs->vfs_flag & VFS_RDONLY);
+       int                     ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
 
        /* Fail a mount where the logbuf is smaller then the log stripe */
-       if (XFS_SB_VERSION_HASLOGV2(&mp->m_sb)) {
-               if ((ap->logbufsize == -1) &&
+       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 < 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);
@@ -358,6 +349,9 @@ xfs_finish_flags(
                }
        }
 
+       if (xfs_sb_version_hasattr2(&mp->m_sb))
+               mp->m_flags |= XFS_MOUNT_ATTR2;
+
        /*
         * prohibit r/w mounts of read-only filesystems
         */
@@ -368,20 +362,10 @@ xfs_finish_flags(
        }
 
        /*
-        * disallow mount attempts with (IRIX) project quota enabled
-        */
-       if (XFS_SB_VERSION_HASQUOTA(&mp->m_sb) &&
-           (mp->m_sb.sb_qflags & XFS_PQUOTA_ACCT)) {
-               cmn_err(CE_WARN,
-       "XFS: cannot mount a filesystem with IRIX project quota enabled");
-               return XFS_ERROR(ENOSYS);
-       }
-
-       /*
         * check for shared mount.
         */
        if (ap->flags & XFSMNT_SHARED) {
-               if (!XFS_SB_VERSION_HASSHARED(&mp->m_sb))
+               if (!xfs_sb_version_hasshared(&mp->m_sb))
                        return XFS_ERROR(EINVAL);
 
                /*
@@ -403,6 +387,22 @@ xfs_finish_flags(
                        return XFS_ERROR(EINVAL);
        }
 
+       if (ap->flags & XFSMNT_UQUOTA) {
+               mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
+               if (ap->flags & XFSMNT_UQUOTAENF)
+                       mp->m_qflags |= XFS_UQUOTA_ENFD;
+       }
+
+       if (ap->flags & XFSMNT_GQUOTA) {
+               mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
+               if (ap->flags & XFSMNT_GQUOTAENF)
+                       mp->m_qflags |= XFS_OQUOTA_ENFD;
+       } else if (ap->flags & XFSMNT_PQUOTA) {
+               mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
+               if (ap->flags & XFSMNT_PQUOTAENF)
+                       mp->m_qflags |= XFS_OQUOTA_ENFD;
+       }
+
        return 0;
 }
 
@@ -418,30 +418,27 @@ xfs_finish_flags(
  * 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
+int
 xfs_mount(
-       struct bhv_desc         *bhvp,
+       struct xfs_mount        *mp,
        struct xfs_mount_args   *args,
        cred_t                  *credp)
 {
-       struct 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;
+       ddev = mp->m_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;
+       error = xfs_dmops_get(mp, args);
+       if (error)
+               return error;
+       error = xfs_qmops_get(mp, args);
+       if (error)
+               return error;
+
+       if (args->flags & XFSMNT_QUIET)
+               flags |= XFS_MFSI_QUIET;
 
        /*
         * Open real time and log devices - order is important.
@@ -479,24 +476,30 @@ xfs_mount(
        }
        if (rtdev) {
                mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
-               if (!mp->m_rtdev_targp)
+               if (!mp->m_rtdev_targp) {
+                       xfs_blkdev_put(logdev);
+                       xfs_blkdev_put(rtdev);
                        goto error0;
+               }
        }
        mp->m_logdev_targp = (logdev && logdev != ddev) ?
                                xfs_alloc_buftarg(logdev, 1) : mp->m_ddev_targp;
-       if (!mp->m_logdev_targp)
+       if (!mp->m_logdev_targp) {
+               xfs_blkdev_put(logdev);
+               xfs_blkdev_put(rtdev);
                goto error0;
+       }
 
        /*
         * Setup flags based on mount(2) options and then the superblock
         */
-       error = xfs_start_flags(vfsp, args, mp);
+       error = xfs_start_flags(args, mp);
        if (error)
                goto error1;
-       error = xfs_readsb(mp);
+       error = xfs_readsb(mp, flags);
        if (error)
                goto error1;
-       error = xfs_finish_flags(vfsp, args, mp);
+       error = xfs_finish_flags(args, mp);
        if (error)
                goto error2;
 
@@ -508,7 +511,7 @@ xfs_mount(
        if (!error && logdev && logdev != ddev) {
                unsigned int    log_sector_size = BBSIZE;
 
-               if (XFS_SB_VERSION_HASSECTOR(&mp->m_sb))
+               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,
@@ -521,9 +524,20 @@ xfs_mount(
        if (error)
                goto error2;
 
-       error = XFS_IOINIT(vfsp, args, flags);
-       if (!error)
-               return 0;
+       if (mp->m_flags & XFS_MOUNT_BARRIER)
+               xfs_mountfs_check_barriers(mp);
+
+       if ((error = xfs_filestream_mount(mp)))
+               goto error2;
+
+       error = xfs_mountfs(mp, flags);
+       if (error)
+               goto error2;
+
+       XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, args->mtpt, args->fsname);
+
+       return 0;
+
 error2:
        if (mp->m_sb_bp)
                xfs_freesb(mp);
@@ -535,19 +549,19 @@ error1:
                xfs_binval(mp->m_rtdev_targp);
 error0:
        xfs_unmountfs_close(mp, credp);
+       xfs_qmops_put(mp);
+       xfs_dmops_put(mp);
        return error;
 }
 
-STATIC int
+int
 xfs_unmount(
-       bhv_desc_t      *bdp,
+       xfs_mount_t     *mp,
        int             flags,
        cred_t          *credp)
 {
-       struct vfs      *vfsp = bhvtovfs(bdp);
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
        xfs_inode_t     *rip;
-       vnode_t         *rvp;
+       bhv_vnode_t     *rvp;
        int             unmount_event_wanted = 0;
        int             unmount_event_flags = 0;
        int             xfs_unmountfs_needed = 0;
@@ -556,8 +570,9 @@ xfs_unmount(
        rip = mp->m_rootip;
        rvp = XFS_ITOV(rip);
 
-       if (vfsp->vfs_flag & VFS_DMI) {
-               error = XFS_SEND_PREUNMOUNT(mp, vfsp,
+#ifdef HAVE_DMAPI
+       if (mp->m_flags & XFS_MOUNT_DMAPI) {
+               error = XFS_SEND_PREUNMOUNT(mp,
                                rvp, DM_RIGHT_NULL, rvp, DM_RIGHT_NULL,
                                NULL, NULL, 0, 0,
                                (mp->m_dmevmask & (1<<DM_EVENT_PREUNMOUNT))?
@@ -568,13 +583,20 @@ xfs_unmount(
                unmount_event_flags = (mp->m_dmevmask & (1<<DM_EVENT_UNMOUNT))?
                                        0 : DM_FLAGS_UNWANTED;
        }
-
+#endif
        /*
         * First blow any referenced inode from this file system
         * out of the reference cache, and delete the timer.
         */
        xfs_refcache_purge_mp(mp);
 
+       /*
+        * Blow away any referenced inode in the filestreams cache.
+        * This can and will cause log traffic as inodes go inactive
+        * here.
+        */
+       xfs_filestream_unmount(mp);
+
        XFS_bflush(mp->m_ddev_targp);
        error = xfs_unmount_flush(mp, 0);
        if (error)
@@ -593,8 +615,7 @@ xfs_unmount(
         * referenced vnodes as well.
         */
        if (XFS_FORCED_SHUTDOWN(mp)) {
-               error = xfs_sync(&mp->m_bhv,
-                        (SYNC_WAIT | SYNC_CLOSE), credp);
+               error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
                ASSERT(error != EFSCORRUPTED);
        }
        xfs_unmountfs_needed = 1;
@@ -608,7 +629,7 @@ out:
                /* Note: mp structure must still exist for
                 * XFS_SEND_UNMOUNT() call.
                 */
-               XFS_SEND_UNMOUNT(mp, vfsp, error == 0 ? rvp : NULL,
+               XFS_SEND_UNMOUNT(mp, error == 0 ? rvp : NULL,
                        DM_RIGHT_NULL, 0, error, unmount_event_flags);
        }
        if (xfs_unmountfs_needed) {
@@ -617,62 +638,88 @@ out:
                 * and free the super block buffer & mount structures.
                 */
                xfs_unmountfs(mp, credp);
+               xfs_qmops_put(mp);
+               xfs_dmops_put(mp);
+               kmem_free(mp, sizeof(xfs_mount_t));
        }
 
        return XFS_ERROR(error);
 }
 
-#define REMOUNT_READONLY_FLAGS (SYNC_REMOUNT|SYNC_ATTR|SYNC_WAIT)
-
 STATIC int
-xfs_mntupdate(
-       bhv_desc_t                      *bdp,
-       int                             *flags,
-       struct xfs_mount_args           *args)
+xfs_quiesce_fs(
+       xfs_mount_t             *mp)
 {
-       struct vfs      *vfsp = bhvtovfs(bdp);
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
-       int             pincount, error;
-       int             count = 0;
+       int                     count = 0, pincount;
 
-       if (args->flags & XFSMNT_NOATIME)
-               mp->m_flags |= XFS_MOUNT_NOATIME;
-       else
-               mp->m_flags &= ~XFS_MOUNT_NOATIME;
+       xfs_refcache_purge_mp(mp);
+       xfs_flush_buftarg(mp->m_ddev_targp, 0);
+       xfs_finish_reclaim_all(mp, 0);
+
+       /* This loop must run at least twice.
+        * The first instance of the loop will flush
+        * most meta data but that will generate more
+        * meta data (typically directory updates).
+        * Which then must be flushed and logged before
+        * we can write the unmount record.
+        */
+       do {
+               xfs_syncsub(mp, SYNC_INODE_QUIESCE, NULL);
+               pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
+               if (!pincount) {
+                       delay(50);
+                       count++;
+               }
+       } while (count < 2);
 
-       if (!(vfsp->vfs_flag & VFS_RDONLY)) {
-               VFS_SYNC(vfsp, SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR, NULL, error);
-       }
+       return 0;
+}
 
-       if (*flags & MS_RDONLY) {
-               xfs_refcache_purge_mp(mp);
-               xfs_flush_buftarg(mp->m_ddev_targp, 0);
-               xfs_finish_reclaim_all(mp, 0);
-
-               /* This loop must run at least twice.
-                * The first instance of the loop will flush
-                * most meta data but that will generate more
-                * meta data (typically directory updates).
-                * Which then must be flushed and logged before
-                * we can write the unmount record.
-                */ 
-               do {
-                       VFS_SYNC(vfsp, REMOUNT_READONLY_FLAGS, NULL, error);
-                       pincount = xfs_flush_buftarg(mp->m_ddev_targp, 1);
-                       if (!pincount) {
-                               delay(50);
-                               count++;
-                       }
-               } while (count < 2);
+/*
+ * 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)
+{
+       /* wait for all modifications to complete */
+       while (atomic_read(&mp->m_active_trans) > 0)
+               delay(100);
 
-               /* Ok now write out an unmount record */
-               xfs_log_unmount_write(mp);
-               xfs_unmountfs_writesb(mp);
-               vfsp->vfs_flag |= VFS_RDONLY;
-       } else {
-               vfsp->vfs_flag &= ~VFS_RDONLY;
-       }
+       /* 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_sbcount(mp, 1);
+       xfs_log_unmount_write(mp);
+       xfs_unmountfs_writesb(mp);
+}
 
+int
+xfs_mntupdate(
+       struct xfs_mount                *mp,
+       int                             *flags,
+       struct xfs_mount_args           *args)
+{
+       if (!(*flags & MS_RDONLY)) {                    /* rw/ro -> rw */
+               if (mp->m_flags & XFS_MOUNT_RDONLY)
+                       mp->m_flags &= ~XFS_MOUNT_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 (!(mp->m_flags & XFS_MOUNT_RDONLY)) { /* rw -> ro */
+               xfs_filestream_flush(mp);
+               xfs_sync(mp, SYNC_DATA_QUIESCE);
+               xfs_attr_quiesce(mp);
+               mp->m_flags |= XFS_MOUNT_RDONLY;
+       }
        return 0;
 }
 
@@ -690,10 +737,10 @@ xfs_unmount_flush(
        xfs_inode_t     *rip = mp->m_rootip;
        xfs_inode_t     *rbmip;
        xfs_inode_t     *rsumip = NULL;
-       vnode_t         *rvp = XFS_ITOV(rip);
+       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);
 
        /*
@@ -760,81 +807,6 @@ 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,
-       vnode_t         **vpp)
-{
-       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,
-       xfs_statfs_t    *statp,
-       vnode_t         *vp)
-{
-       __uint64_t      fakeinos;
-       xfs_extlen_t    lsize;
-       xfs_mount_t     *mp;
-       xfs_sb_t        *sbp;
-       unsigned long   s;
-       u64 id;
-
-       mp = XFS_BHVTOM(bdp);
-       sbp = &(mp->m_sb);
-
-       statp->f_type = XFS_SB_MAGIC;
-
-       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;
-       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);
-
-       id = huge_encode_dev(mp->m_dev);
-       statp->f_fsid.val[0] = (u32)id;
-       statp->f_fsid.val[1] = (u32)(id >> 32);
-       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
@@ -864,49 +836,64 @@ xfs_statvfs(
  *                    determine if they should be flushed sync, async, or
  *                    delwri.
  *      SYNC_CLOSE   - This flag is passed when the system is being
- *                    unmounted.  We should sync and invalidate everthing.
+ *                    unmounted.  We should sync and invalidate everything.
  *      SYNC_FSDATA  - This indicates that the caller would like to make
  *                    sure the superblock is safe on disk.  We can ensure
- *                    this by simply makeing sure the log gets flushed
+ *                    this by simply making sure the log gets flushed
  *                    if SYNC_BDFLUSH is set, and by actually writing it
  *                    out otherwise.
+ *     SYNC_IOWAIT  - The caller wants us to wait for all data I/O to complete
+ *                    before we return (including direct I/O). Forms the drain
+ *                    side of the write barrier needed to safely quiesce the
+ *                    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;
+       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);
 
-       mp = XFS_BHVTOM(bdp);
-       return (xfs_syncsub(mp, flags, 0, NULL));
+       return xfs_syncsub(mp, flags, NULL);
 }
 
 /*
  * xfs sync routine for internal use
  *
- * This routine supports all of the flags defined for the generic VFS_SYNC
- * interface as explained above under xfs_sync.  In the interests of not
- * changing interfaces within the 6.5 family, additional internallly-
- * required functions are specified within a separate xflags parameter,
- * only available by calling this routine.
+ * This routine supports all of the flags defined for the generic vfs_sync
+ * interface as explained above under xfs_sync.
  *
  */
-STATIC int
+int
 xfs_sync_inodes(
        xfs_mount_t     *mp,
        int             flags,
-       int             xflags,
        int             *bypassed)
 {
        xfs_inode_t     *ip = NULL;
-       xfs_inode_t     *ip_next;
-       xfs_buf_t       *bp;
-       vnode_t         *vp = NULL;
-       vmap_t          vmap;
+       bhv_vnode_t     *vp = NULL;
        int             error;
        int             last_error;
        uint64_t        fflag;
@@ -915,7 +902,6 @@ 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 +954,11 @@ 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 +968,7 @@ 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)
+       if (flags & SYNC_DELWRI)
                fflag = XFS_B_DELWRI;
        if (flags & SYNC_WAIT)
                fflag = 0;              /* synchronous overrides all */
@@ -1070,24 +1058,6 @@ xfs_sync_inodes(
                }
 
                /*
-                * 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 racey, 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
@@ -1101,48 +1071,21 @@ xfs_sync_inodes(
                 * lock in xfs_ireclaim() after the inode is pulled from
                 * the mount list will sleep until we release it here.
                 * This keeps the vnode from being freed while we reference
-                * it.  It is also cheaper and simpler than actually doing
-                * a vn_get() for every inode we touch here.
+                * it.
                 */
                if (xfs_ilock_nowait(ip, lock_flags) == 0) {
-
-                       if ((flags & SYNC_BDFLUSH) || (vp == NULL)) {
+                       if (vp == NULL) {
                                ip = ip->i_mnext;
                                continue;
                        }
 
-                       /*
-                        * We need to unlock the inode list lock in order
-                        * to lock the inode. Insert a marker record into
-                        * the inode list to remember our position, dropping
-                        * the lock is now done inside the IPOINTER_INSERT
-                        * macro.
-                        *
-                        * We also use the inode list lock to protect us
-                        * in taking a snapshot of the vnode version number
-                        * for use in calling vn_get().
-                        */
-                       VMAP(vp, vmap);
-                       IPOINTER_INSERT(ip, mp);
-
-                       vp = vn_get(vp, &vmap);
+                       vp = vn_grab(vp);
                        if (vp == NULL) {
-                               /*
-                                * The vnode was reclaimed once we let go
-                                * of the inode list lock.  Skip to the
-                                * next list entry. Remove the marker.
-                                */
-
-                               XFS_MOUNT_ILOCK(mp);
-
-                               mount_locked = B_TRUE;
-                               vnode_refed  = B_FALSE;
-
-                               IPOINTER_REMOVE(ip, mp);
-
+                               ip = ip->i_mnext;
                                continue;
                        }
 
+                       IPOINTER_INSERT(ip, mp);
                        xfs_ilock(ip, lock_flags);
 
                        ASSERT(vp == XFS_ITOV(ip));
@@ -1155,208 +1098,64 @@ 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.
-                        */
+               /*
+                * 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);
 
-                       if (XFS_FORCED_SHUTDOWN(mp)) {
-                               VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
-                       } else {
-                               VOP_FLUSHINVAL_PAGES(vp, 0, -1, FI_REMAPF);
+                       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);
-
-               } 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);
-                               VOP_FLUSH_PAGES(vp, (xfs_off_t)0, -1,
-                                                       fflag, FI_NONE, error);
-                               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(ip);
 
+                       xfs_ilock(ip, XFS_ILOCK_SHARED);
                }
 
-               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);
-                                       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;
-                                       }
-
-                                       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);
-                                       }
-                               }
+               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);
 
-                       }
+                       if (flags & SYNC_WAIT) {
+                               xfs_iflock(ip);
+                               error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
 
-               } 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);
-                               } 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)++;
                        }
                }
 
@@ -1437,18 +1236,14 @@ xfs_sync_inodes(
 /*
  * xfs sync routine for internal use
  *
- * This routine supports all of the flags defined for the generic VFS_SYNC
- * interface as explained above under xfs_sync.  In the interests of not
- * changing interfaces within the 6.5 family, additional internallly-
- * required functions are specified within a separate xflags parameter,
- * only available by calling this routine.
+ * This routine supports all of the flags defined for the generic vfs_sync
+ * interface as explained above under xfs_sync.
  *
  */
 int
 xfs_syncsub(
        xfs_mount_t     *mp,
        int             flags,
-       int             xflags,
        int             *bypassed)
 {
        int             error = 0;
@@ -1470,7 +1265,7 @@ xfs_syncsub(
                if (flags & SYNC_BDFLUSH)
                        xfs_finish_reclaim_all(mp, 1);
                else
-                       error = xfs_sync_inodes(mp, flags, xflags, bypassed);
+                       error = xfs_sync_inodes(mp, flags, bypassed);
        }
 
        /*
@@ -1533,10 +1328,22 @@ xfs_syncsub(
         * eventually kicked out of the cache.
         */
        if (flags & SYNC_REFCACHE) {
-               xfs_refcache_purge_some(mp);
+               if (flags & SYNC_WAIT)
+                       xfs_refcache_purge_mp(mp);
+               else
+                       xfs_refcache_purge_some(mp);
        }
 
        /*
+        * If asked, update the disk superblock with incore counter values if we
+        * are using non-persistent counters so that they don't get too far out
+        * of sync if we crash or get a forced shutdown. We don't want to force
+        * this to disk, just get a transaction into the iclogs....
+        */
+       if (flags & SYNC_SUPER)
+               xfs_log_sbcount(mp, 0);
+
+       /*
         * Now check to see if the log needs a "dummy" transaction.
         */
 
@@ -1562,7 +1369,7 @@ xfs_syncsub(
                xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
                xfs_trans_ihold(tp, ip);
                xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
-               error = xfs_trans_commit(tp, 0, NULL);
+               error = xfs_trans_commit(tp, 0);
                xfs_iunlock(ip, XFS_ILOCK_EXCL);
                xfs_log_force(mp, (xfs_lsn_t)0, log_flags);
        }
@@ -1580,362 +1387,3 @@ 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,
-       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_IHASHSIZE    "ihashsize"    /* size of inode hash table */
-#define MNTOPT_NORECOVERY   "norecovery"   /* don't run XFS recovery */
-#define MNTOPT_NOLOGFLUSH   "nologflush"   /* don't hard flush on log writes */
-#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 */
-
-
-int
-xfs_parseargs(
-       struct bhv_desc         *bhv,
-       char                    *options,
-       struct xfs_mount_args   *args,
-       int                     update)
-{
-       struct vfs              *vfsp = bhvtovfs(bhv);
-       char                    *this_char, *value, *eov;
-       int                     dsunit, dswidth, vol_dsunit, vol_dswidth;
-       int                     iosize;
-
-#if 0  /* XXX: off by default, until some remaining issues ironed out */
-       args->flags |= XFSMNT_IDELETE; /* default to on */
-#endif
-
-       if (!options)
-               return 0;
-
-       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) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_LOGBUFS);
-                               return EINVAL;
-                       }
-                       args->logbufs = simple_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
-                       int     last, in_kilobytes = 0;
-
-                       if (!value || !*value) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_LOGBSIZE);
-                               return EINVAL;
-                       }
-                       last = strlen(value) - 1;
-                       if (value[last] == 'K' || value[last] == 'k') {
-                               in_kilobytes = 1;
-                               value[last] = '\0';
-                       }
-                       args->logbufsize = simple_strtoul(value, &eov, 10);
-                       if (in_kilobytes)
-                               args->logbufsize <<= 10;
-               } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
-                       if (!value || !*value) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_LOGDEV);
-                               return EINVAL;
-                       }
-                       strncpy(args->logname, value, MAXNAMELEN);
-               } else if (!strcmp(this_char, MNTOPT_MTPT)) {
-                       if (!value || !*value) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_MTPT);
-                               return EINVAL;
-                       }
-                       strncpy(args->mtpt, value, MAXNAMELEN);
-               } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
-                       if (!value || !*value) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_RTDEV);
-                               return EINVAL;
-                       }
-                       strncpy(args->rtname, value, MAXNAMELEN);
-               } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
-                       if (!value || !*value) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_BIOSIZE); 
-                               return EINVAL;
-                       }
-                       iosize = simple_strtoul(value, &eov, 10);
-                       args->flags |= XFSMNT_IOSIZE;
-                       args->iosizelog = (uint8_t) iosize;
-               } else if (!strcmp(this_char, MNTOPT_IHASHSIZE)) {
-                       if (!value || !*value) {
-                               printk("XFS: %s option requires an argument\n",
-                                       this_char); 
-                               return EINVAL;
-                       }
-                       args->flags |= XFSMNT_IHASHSIZE;
-                       args->ihashsize = simple_strtoul(value, &eov, 10);
-               } 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
-                       printk("XFS: %s option not allowed on this system\n",
-                               MNTOPT_INO64);
-                       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) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_SUNIT);
-                               return EINVAL;
-                       }
-                       dsunit = simple_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
-                       if (!value || !*value) {
-                               printk("XFS: %s option requires an argument\n",
-                                       MNTOPT_SWIDTH);
-                               return EINVAL;
-                       }
-                       dswidth = simple_strtoul(value, &eov, 10);
-               } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
-                       args->flags &= ~XFSMNT_32BITINODES;
-#if !XFS_BIG_INUMS
-                       printk("XFS: %s option not allowed on this system\n",
-                               MNTOPT_64BITINODE);
-                       return EINVAL;
-#endif
-               } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
-                       args->flags |= XFSMNT_NOUUID;
-               } else if (!strcmp(this_char, MNTOPT_NOLOGFLUSH)) {
-                       args->flags |= XFSMNT_NOLOGFLUSH;
-               } 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, "osyncisdsync")) {
-                       /* no-op, this is now the default */
-printk("XFS: osyncisdsync is now the default, option is deprecated.\n");
-               } else if (!strcmp(this_char, "irixsgid")) {
-printk("XFS: irixsgid is now a sysctl(2) variable, option is deprecated.\n");
-               } else {
-                       printk("XFS: unknown mount option [%s].\n", this_char);
-                       return EINVAL;
-               }
-       }
-
-       if (args->flags & XFSMNT_NORECOVERY) {
-               if ((vfsp->vfs_flag & VFS_RDONLY) == 0) {
-                       printk("XFS: no-recovery mounts must be read-only.\n");
-                       return EINVAL;
-               }
-       }
-
-       if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
-               printk(
-       "XFS: sunit and swidth options incompatible with the noalign option\n");
-               return EINVAL;
-       }
-
-       if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
-               printk("XFS: sunit and swidth must be specified together\n");
-               return EINVAL;
-       }
-
-       if (dsunit && (dswidth % dsunit != 0)) {
-               printk(
-       "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)\n",
-                       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;
-       }
-
-       return 0;
-}
-
-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 },
-               { XFS_MOUNT_NOLOGFLUSH,         "," MNTOPT_NOLOGFLUSH },
-               { XFS_MOUNT_IDELETE,            "," MNTOPT_NOIKEEP },
-               { 0, NULL }
-       };
-       struct proc_xfs_info    *xfs_infop;
-       struct xfs_mount        *mp = XFS_BHVTOM(bhv);
-
-       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", mp->m_ihsize);
-
-       if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
-               seq_printf(m, "," MNTOPT_BIOSIZE "=%d", mp->m_writeio_log);
-
-       if (mp->m_logbufs > 0)
-               seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
-
-       if (mp->m_logbsize > 0)
-               seq_printf(m, "," MNTOPT_LOGBSIZE "=%d", mp->m_logbsize);
-
-       if (mp->m_ddev_targp != mp->m_logdev_targp)
-               seq_printf(m, "," MNTOPT_LOGDEV "=%s",
-                               XFS_BUFTARG_NAME(mp->m_logdev_targp));
-
-       if (mp->m_rtdev_targp && mp->m_ddev_targp != mp->m_rtdev_targp)
-               seq_printf(m, "," MNTOPT_RTDEV "=%s",
-                               XFS_BUFTARG_NAME(mp->m_rtdev_targp));
-
-       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_32BITINOOPT))
-               seq_printf(m, "," MNTOPT_64BITINODE);
-       
-       return 0;
-}
-
-STATIC void
-xfs_freeze(
-       bhv_desc_t      *bdp)
-{
-       xfs_mount_t     *mp = XFS_BHVTOM(bdp);
-
-       while (atomic_read(&mp->m_active_trans) > 0)
-               delay(100);
-
-       /* Push the superblock and write an unmount record */
-       xfs_log_unmount_write(mp);
-       xfs_unmountfs_writesb(mp);
-}
-
-
-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,
-};