]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/hfs/hfs_vfsops.c
xnu-1699.32.7.tar.gz
[apple/xnu.git] / bsd / hfs / hfs_vfsops.c
index 8148697b213dbbd6445697cb6efa64e87c2f0ec7..4e5b76b1457d9b84ebf44a1c14dc11869f6af7ce 100644 (file)
@@ -1,5 +1,5 @@
 /*
 /*
- * Copyright (c) 1999-2009 Apple Inc. All rights reserved.
+ * Copyright (c) 1999-2012 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
@@ -87,6 +87,7 @@
 #include <sys/utfconv.h>
 #include <sys/kdebug.h>
 #include <sys/fslog.h>
 #include <sys/utfconv.h>
 #include <sys/kdebug.h>
 #include <sys/fslog.h>
+#include <sys/ubc.h>
 
 #include <kern/locks.h>
 
 
 #include <kern/locks.h>
 
 #include "hfs_endian.h"
 #include "hfs_hotfiles.h"
 #include "hfs_quota.h"
 #include "hfs_endian.h"
 #include "hfs_hotfiles.h"
 #include "hfs_quota.h"
+#include "hfs_btreeio.h"
 
 #include "hfscommon/headers/FileMgrInternal.h"
 #include "hfscommon/headers/BTreesInternal.h"
 
 
 #include "hfscommon/headers/FileMgrInternal.h"
 #include "hfscommon/headers/BTreesInternal.h"
 
+#if CONFIG_PROTECT
+#include <sys/cprotect.h>
+#endif
+
+#if CONFIG_HFS_ALLOC_RBTREE
+#include "hfscommon/headers/HybridAllocator.h"
+#endif
+
+#define HFS_MOUNT_DEBUG 1
+
 #if    HFS_DIAGNOSTIC
 int hfs_dbg_all = 0;
 int hfs_dbg_err = 0;
 #endif
 
 #if    HFS_DIAGNOSTIC
 int hfs_dbg_all = 0;
 int hfs_dbg_err = 0;
 #endif
 
+/* Enable/disable debugging code for live volume resizing */
+int hfs_resize_debug = 0;
 
 lck_grp_attr_t *  hfs_group_attr;
 lck_attr_t *  hfs_lock_attr;
 lck_grp_t *  hfs_mutex_group;
 lck_grp_t *  hfs_rwlock_group;
 
 lck_grp_attr_t *  hfs_group_attr;
 lck_attr_t *  hfs_lock_attr;
 lck_grp_t *  hfs_mutex_group;
 lck_grp_t *  hfs_rwlock_group;
+lck_grp_t *  hfs_spinlock_group;
 
 extern struct vnodeopv_desc hfs_vnodeop_opv_desc;
 extern struct vnodeopv_desc hfs_std_vnodeop_opv_desc;
 
 extern struct vnodeopv_desc hfs_vnodeop_opv_desc;
 extern struct vnodeopv_desc hfs_std_vnodeop_opv_desc;
@@ -132,30 +147,30 @@ static int hfs_flushfiles(struct mount *, int, struct proc *);
 static int hfs_flushMDB(struct hfsmount *hfsmp, int waitfor, int altflush);
 static int hfs_getmountpoint(struct vnode *vp, struct hfsmount **hfsmpp);
 static int hfs_init(struct vfsconf *vfsp);
 static int hfs_flushMDB(struct hfsmount *hfsmp, int waitfor, int altflush);
 static int hfs_getmountpoint(struct vnode *vp, struct hfsmount **hfsmpp);
 static int hfs_init(struct vfsconf *vfsp);
-static int hfs_mount(struct mount *mp, vnode_t  devvp, user_addr_t data, vfs_context_t context);
-static int hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args, int journal_replay_only, vfs_context_t context);
-static int hfs_reload(struct mount *mp);
 static int hfs_vfs_root(struct mount *mp, struct vnode **vpp, vfs_context_t context);
 static int hfs_quotactl(struct mount *, int, uid_t, caddr_t, vfs_context_t context);
 static int hfs_start(struct mount *mp, int flags, vfs_context_t context);
 static int hfs_vfs_root(struct mount *mp, struct vnode **vpp, vfs_context_t context);
 static int hfs_quotactl(struct mount *, int, uid_t, caddr_t, vfs_context_t context);
 static int hfs_start(struct mount *mp, int flags, vfs_context_t context);
-static int hfs_statfs(struct mount *mp, register struct vfsstatfs *sbp, vfs_context_t context);
-static int hfs_sync(struct mount *mp, int waitfor, vfs_context_t context);
-static int hfs_sysctl(int *name, u_int namelen, user_addr_t oldp, size_t *oldlenp, 
-                      user_addr_t newp, size_t newlen, vfs_context_t context);
-static int hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context);
 static int hfs_vptofh(struct vnode *vp, int *fhlenp, unsigned char *fhp, vfs_context_t context);
 static int hfs_vptofh(struct vnode *vp, int *fhlenp, unsigned char *fhp, vfs_context_t context);
-
-static int hfs_reclaimspace(struct hfsmount *hfsmp, u_int32_t startblk, u_int32_t reclaimblks, vfs_context_t context);
-static int hfs_overlapped_overflow_extents(struct hfsmount *hfsmp, u_int32_t startblk,
-                                           u_int32_t catblks, u_int32_t fileID, int rsrcfork);
+static int hfs_file_extent_overlaps(struct hfsmount *hfsmp, u_int32_t allocLimit, struct HFSPlusCatalogFile *filerec);
 static int hfs_journal_replay(vnode_t devvp, vfs_context_t context);
 static int hfs_journal_replay(vnode_t devvp, vfs_context_t context);
+static int hfs_reclaimspace(struct hfsmount *hfsmp, u_int32_t allocLimit, u_int32_t reclaimblks, vfs_context_t context);
 
 
+void hfs_initialize_allocator (struct hfsmount *hfsmp);
+int hfs_teardown_allocator (struct hfsmount *hfsmp);
+
+int hfs_mount(struct mount *mp, vnode_t  devvp, user_addr_t data, vfs_context_t context);
+int hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args, int journal_replay_only, vfs_context_t context);
+int hfs_reload(struct mount *mp);
+int hfs_statfs(struct mount *mp, register struct vfsstatfs *sbp, vfs_context_t context);
+int hfs_sync(struct mount *mp, int waitfor, vfs_context_t context);
+int hfs_sysctl(int *name, u_int namelen, user_addr_t oldp, size_t *oldlenp, 
+                      user_addr_t newp, size_t newlen, vfs_context_t context);
+int hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context);
 
 /*
  * Called by vfs_mountroot when mounting HFS Plus as root.
  */
 
 
 /*
  * Called by vfs_mountroot when mounting HFS Plus as root.
  */
 
-__private_extern__
 int
 hfs_mountroot(mount_t mp, vnode_t rvp, vfs_context_t context)
 {
 int
 hfs_mountroot(mount_t mp, vnode_t rvp, vfs_context_t context)
 {
@@ -164,8 +179,13 @@ hfs_mountroot(mount_t mp, vnode_t rvp, vfs_context_t context)
        struct vfsstatfs *vfsp;
        int error;
 
        struct vfsstatfs *vfsp;
        int error;
 
-       if ((error = hfs_mountfs(rvp, mp, NULL, 0, context)))
+       if ((error = hfs_mountfs(rvp, mp, NULL, 0, context))) {
+               if (HFS_MOUNT_DEBUG) {
+                       printf("hfs_mountroot: hfs_mountfs returned %d, rvp (%p) name (%s) \n", 
+                                       error, rvp, (rvp->v_name ? rvp->v_name : "unknown device"));
+               }
                return (error);
                return (error);
+       }
 
        /* Init hfsmp */
        hfsmp = VFSTOHFS(mp);
 
        /* Init hfsmp */
        hfsmp = VFSTOHFS(mp);
@@ -193,7 +213,7 @@ hfs_mountroot(mount_t mp, vnode_t rvp, vfs_context_t context)
  * mount system call
  */
 
  * mount system call
  */
 
-static int
+int
 hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t context)
 {
        struct proc *p = vfs_context_proc(context);
 hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t context)
 {
        struct proc *p = vfs_context_proc(context);
@@ -203,6 +223,9 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
        u_int32_t cmdflags;
 
        if ((retval = copyin(data, (caddr_t)&args, sizeof(args)))) {
        u_int32_t cmdflags;
 
        if ((retval = copyin(data, (caddr_t)&args, sizeof(args)))) {
+               if (HFS_MOUNT_DEBUG) {
+                       printf("hfs_mount: copyin returned %d for fs\n", retval);
+               }
                return (retval);
        }
        cmdflags = (u_int32_t)vfs_flags(mp) & MNT_CMDFLAGS;
                return (retval);
        }
        cmdflags = (u_int32_t)vfs_flags(mp) & MNT_CMDFLAGS;
@@ -211,10 +234,19 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
 
                /* Reload incore data after an fsck. */
                if (cmdflags & MNT_RELOAD) {
 
                /* Reload incore data after an fsck. */
                if (cmdflags & MNT_RELOAD) {
-                       if (vfs_isrdonly(mp))
-                               return hfs_reload(mp);
-                       else
+                       if (vfs_isrdonly(mp)) {
+                               int error = hfs_reload(mp);
+                               if (error && HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: hfs_reload returned %d on %s \n", error, hfsmp->vcbVN);
+                               }
+                               return error;
+                       }
+                       else {
+                               if (HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: MNT_RELOAD not supported on rdwr filesystem %s\n", hfsmp->vcbVN);
+                               }
                                return (EINVAL);
                                return (EINVAL);
+                       }
                }
 
                /* Change to a read-only file system. */
                }
 
                /* Change to a read-only file system. */
@@ -226,16 +258,19 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                         * is in progress and therefore block any further 
                         * modifications to the file system.
                         */
                         * is in progress and therefore block any further 
                         * modifications to the file system.
                         */
-                       hfs_global_exclusive_lock_acquire(hfsmp);
+                       hfs_lock_global (hfsmp, HFS_EXCLUSIVE_LOCK);
                        hfsmp->hfs_flags |= HFS_RDONLY_DOWNGRADE;
                        hfsmp->hfs_downgrading_proc = current_thread();
                        hfsmp->hfs_flags |= HFS_RDONLY_DOWNGRADE;
                        hfsmp->hfs_downgrading_proc = current_thread();
-                       hfs_global_exclusive_lock_release(hfsmp);
+                       hfs_unlock_global (hfsmp);
 
                        /* use VFS_SYNC to push out System (btree) files */
                        retval = VFS_SYNC(mp, MNT_WAIT, context);
                        if (retval && ((cmdflags & MNT_FORCE) == 0)) {
                                hfsmp->hfs_flags &= ~HFS_RDONLY_DOWNGRADE;
                                hfsmp->hfs_downgrading_proc = NULL;
 
                        /* use VFS_SYNC to push out System (btree) files */
                        retval = VFS_SYNC(mp, MNT_WAIT, context);
                        if (retval && ((cmdflags & MNT_FORCE) == 0)) {
                                hfsmp->hfs_flags &= ~HFS_RDONLY_DOWNGRADE;
                                hfsmp->hfs_downgrading_proc = NULL;
+                               if (HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: VFS_SYNC returned %d during b-tree sync of %s \n", retval, hfsmp->vcbVN);
+                               }
                                goto out;
                        }
                
                                goto out;
                        }
                
@@ -246,6 +281,9 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                        if ((retval = hfs_flushfiles(mp, flags, p))) {
                                hfsmp->hfs_flags &= ~HFS_RDONLY_DOWNGRADE;
                                hfsmp->hfs_downgrading_proc = NULL;
                        if ((retval = hfs_flushfiles(mp, flags, p))) {
                                hfsmp->hfs_flags &= ~HFS_RDONLY_DOWNGRADE;
                                hfsmp->hfs_downgrading_proc = NULL;
+                               if (HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: hfs_flushfiles returned %d on %s \n", retval, hfsmp->vcbVN);
+                               }
                                goto out;
                        }
 
                                goto out;
                        }
 
@@ -265,13 +303,16 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                                }
                        }
                        if (retval) {
                                }
                        }
                        if (retval) {
+                               if (HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: FSYNC on devvp returned %d for fs %s\n", retval, hfsmp->vcbVN);
+                               }
                                hfsmp->hfs_flags &= ~HFS_RDONLY_DOWNGRADE;
                                hfsmp->hfs_downgrading_proc = NULL;
                                hfsmp->hfs_flags &= ~HFS_READ_ONLY;
                                goto out;
                        }
                        if (hfsmp->jnl) {
                                hfsmp->hfs_flags &= ~HFS_RDONLY_DOWNGRADE;
                                hfsmp->hfs_downgrading_proc = NULL;
                                hfsmp->hfs_flags &= ~HFS_READ_ONLY;
                                goto out;
                        }
                        if (hfsmp->jnl) {
-                           hfs_global_exclusive_lock_acquire(hfsmp);
+                               hfs_lock_global (hfsmp, HFS_EXCLUSIVE_LOCK);
 
                            journal_close(hfsmp->jnl);
                            hfsmp->jnl = NULL;
 
                            journal_close(hfsmp->jnl);
                            hfsmp->jnl = NULL;
@@ -280,14 +321,20 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                            //       access to the jvp because we may need
                            //       it later if we go back to being read-write.
 
                            //       access to the jvp because we may need
                            //       it later if we go back to being read-write.
 
-                           hfs_global_exclusive_lock_release(hfsmp);
+                               hfs_unlock_global (hfsmp);
                        }
 
                        }
 
+#if CONFIG_HFS_ALLOC_RBTREE
+                       (void) hfs_teardown_allocator(hfsmp);
+#endif                                         
                        hfsmp->hfs_downgrading_proc = NULL;
                }
 
                /* Change to a writable file system. */
                if (vfs_iswriteupgrade(mp)) {
                        hfsmp->hfs_downgrading_proc = NULL;
                }
 
                /* Change to a writable file system. */
                if (vfs_iswriteupgrade(mp)) {
+#if CONFIG_HFS_ALLOC_RBTREE
+                               thread_t allocator_thread;
+#endif
 
                        /*
                         * On inconsistent disks, do not allow read-write mount
 
                        /*
                         * On inconsistent disks, do not allow read-write mount
@@ -295,6 +342,9 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                         */
                        if (!(vfs_flags(mp) & MNT_ROOTFS) &&
                                        (hfsmp->vcbAtrb & kHFSVolumeInconsistentMask)) {
                         */
                        if (!(vfs_flags(mp) & MNT_ROOTFS) &&
                                        (hfsmp->vcbAtrb & kHFSVolumeInconsistentMask)) {
+                               if (HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: attempting to mount inconsistent non-root volume %s\n",  (hfsmp->vcbVN));
+                               }
                                retval = EINVAL;
                                goto out;
                        }
                                retval = EINVAL;
                                goto out;
                        }
@@ -309,39 +359,52 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
 
                            if (hfsmp->hfs_flags & HFS_NEED_JNL_RESET) {
                                        jflags = JOURNAL_RESET;
 
                            if (hfsmp->hfs_flags & HFS_NEED_JNL_RESET) {
                                        jflags = JOURNAL_RESET;
-                           } else {
+                               } else {
                                        jflags = 0;
                                        jflags = 0;
-                           }
-                           
-                           hfs_global_exclusive_lock_acquire(hfsmp);
-
-                           hfsmp->jnl = journal_open(hfsmp->jvp,
-                                                     (hfsmp->jnl_start * HFSTOVCB(hfsmp)->blockSize) + (off_t)HFSTOVCB(hfsmp)->hfsPlusIOPosOffset,
-                                                     hfsmp->jnl_size,
-                                                     hfsmp->hfs_devvp,
-                                                     hfsmp->hfs_logical_block_size,
-                                                     jflags,
-                                                     0,
-                                                     hfs_sync_metadata, hfsmp->hfs_mp);
-
-                           hfs_global_exclusive_lock_release(hfsmp);
-
-                           if (hfsmp->jnl == NULL) {
-                               retval = EINVAL;
-                               goto out;
-                           } else {
-                               hfsmp->hfs_flags &= ~HFS_NEED_JNL_RESET;
-                           }
+                               }
+
+                               hfs_lock_global (hfsmp, HFS_EXCLUSIVE_LOCK);
+
+                               hfsmp->jnl = journal_open(hfsmp->jvp,
+                                               (hfsmp->jnl_start * HFSTOVCB(hfsmp)->blockSize) + (off_t)HFSTOVCB(hfsmp)->hfsPlusIOPosOffset,
+                                               hfsmp->jnl_size,
+                                               hfsmp->hfs_devvp,
+                                               hfsmp->hfs_logical_block_size,
+                                               jflags,
+                                               0,
+                                               hfs_sync_metadata, hfsmp->hfs_mp);
+                               
+                               /*
+                                * Set up the trim callback function so that we can add
+                                * recently freed extents to the free extent cache once
+                                * the transaction that freed them is written to the
+                                * journal on disk.
+                                */
+                               if (hfsmp->jnl)
+                                       journal_trim_set_callback(hfsmp->jnl, hfs_trim_callback, hfsmp);
+                               
+                               hfs_unlock_global (hfsmp);
+
+                               if (hfsmp->jnl == NULL) {
+                                       if (HFS_MOUNT_DEBUG) {
+                                               printf("hfs_mount: journal_open == NULL; couldn't be opened on %s \n", (hfsmp->vcbVN));
+                                       }
+                                       retval = EINVAL;
+                                       goto out;
+                               } else {
+                                       hfsmp->hfs_flags &= ~HFS_NEED_JNL_RESET;
+                               }
 
                        }
 
                        /* See if we need to erase unused Catalog nodes due to <rdar://problem/6947811>. */
                        retval = hfs_erase_unused_nodes(hfsmp);
 
                        }
 
                        /* See if we need to erase unused Catalog nodes due to <rdar://problem/6947811>. */
                        retval = hfs_erase_unused_nodes(hfsmp);
-                       if (retval != E_NONE)
+                       if (retval != E_NONE) {
+                               if (HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: hfs_erase_unused_nodes returned %d for fs %s\n", retval, hfsmp->vcbVN);
+                               }
                                goto out;
                                goto out;
-                       
-                       /* Only clear HFS_READ_ONLY after a successful write */
-                       hfsmp->hfs_flags &= ~HFS_READ_ONLY;
+                       }
 
                        /* If this mount point was downgraded from read-write 
                         * to read-only, clear that information as we are now 
 
                        /* If this mount point was downgraded from read-write 
                         * to read-only, clear that information as we are now 
@@ -354,8 +417,16 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                        hfsmp->vcbAtrb &= ~kHFSVolumeUnmountedMask;
 
                        retval = hfs_flushvolumeheader(hfsmp, MNT_WAIT, 0);
                        hfsmp->vcbAtrb &= ~kHFSVolumeUnmountedMask;
 
                        retval = hfs_flushvolumeheader(hfsmp, MNT_WAIT, 0);
-                       if (retval != E_NONE)
+                       if (retval != E_NONE) {
+                               if (HFS_MOUNT_DEBUG) {
+                                       printf("hfs_mount: hfs_flushvolumeheader returned %d for fs %s\n", retval, hfsmp->vcbVN);
+                               }
                                goto out;
                                goto out;
+                       }
+               
+                       /* Only clear HFS_READ_ONLY after a successful write */
+                       hfsmp->hfs_flags &= ~HFS_READ_ONLY;
+
 
                        if (!(hfsmp->hfs_flags & (HFS_READ_ONLY | HFS_STANDARD))) {
                                /* Setup private/hidden directories for hardlinks. */
 
                        if (!(hfsmp->hfs_flags & (HFS_READ_ONLY | HFS_STANDARD))) {
                                /* Setup private/hidden directories for hardlinks. */
@@ -367,7 +438,8 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                                /*
                                 * Allow hot file clustering if conditions allow.
                                 */
                                /*
                                 * Allow hot file clustering if conditions allow.
                                 */
-                               if (hfsmp->hfs_flags & HFS_METADATA_ZONE) {
+                               if ((hfsmp->hfs_flags & HFS_METADATA_ZONE) && 
+                                               ((hfsmp->hfs_flags & HFS_SSD) == 0)) {
                                        (void) hfs_recording_init(hfsmp);
                                }
                                /* Force ACLs on HFS+ file systems. */
                                        (void) hfs_recording_init(hfsmp);
                                }
                                /* Force ACLs on HFS+ file systems. */
@@ -375,10 +447,45 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                                        vfs_setextendedsecurity(HFSTOVFS(hfsmp));
                                }
                        }
                                        vfs_setextendedsecurity(HFSTOVFS(hfsmp));
                                }
                        }
+
+#if CONFIG_HFS_ALLOC_RBTREE
+                       /* 
+                        * Like the normal mount case, we need to handle creation of the allocation red-black tree
+                        * if we're upgrading from read-only to read-write.  
+                        *
+                        * We spawn a thread to create the pair of red-black trees for this volume.
+                        * However, in so doing, we must be careful to ensure that if this thread is still
+                        * running after mount has finished, it doesn't interfere with an unmount. Specifically,
+                        * we'll need to set a bit that indicates we're in progress building the trees here.  
+                        * Unmount will check for this bit, and then if it's set, mark a corresponding bit that
+                        * notifies the tree generation code that an unmount is waiting.  Also, mark the extent
+                        * tree flags that the allocator is enabled for use before we spawn the thread that will start
+                        * scanning the RB tree.                         
+                        *
+                        * Only do this if we're operating on a read-write mount (we wouldn't care for read-only),
+                        * which has not previously encountered a bad error on the red-black tree code.  Also, don't
+                        * try to re-build a tree that already exists. 
+                        */
+                       
+                       if (hfsmp->extent_tree_flags == 0) {
+                               hfsmp->extent_tree_flags |= (HFS_ALLOC_TREEBUILD_INFLIGHT | HFS_ALLOC_RB_ENABLED);
+                               /* Initialize EOF counter so that the thread can assume it started at initial values */
+                               hfsmp->offset_block_end = 0;
+                               
+                               InitTree(hfsmp);
+                               
+                               kernel_thread_start ((thread_continue_t) hfs_initialize_allocator , hfsmp, &allocator_thread);
+                               thread_deallocate(allocator_thread);
+                       }
+
+#endif
                }
 
                /* Update file system parameters. */
                retval = hfs_changefs(mp, &args);
                }
 
                /* Update file system parameters. */
                retval = hfs_changefs(mp, &args);
+               if (retval &&  HFS_MOUNT_DEBUG) {
+                       printf("hfs_mount: hfs_changefs returned %d for %s\n", retval, hfsmp->vcbVN);
+               }
 
        } else /* not an update request */ {
 
 
        } else /* not an update request */ {
 
@@ -386,6 +493,44 @@ hfs_mount(struct mount *mp, vnode_t devvp, user_addr_t data, vfs_context_t conte
                vfs_setflags(mp, (u_int64_t)((unsigned int)MNT_DOVOLFS));
 
                retval = hfs_mountfs(devvp, mp, &args, 0, context);
                vfs_setflags(mp, (u_int64_t)((unsigned int)MNT_DOVOLFS));
 
                retval = hfs_mountfs(devvp, mp, &args, 0, context);
+               if (retval && HFS_MOUNT_DEBUG) {
+                       printf("hfs_mount: hfs_mountfs returned %d\n", retval);
+               }
+#if CONFIG_PROTECT
+               /* 
+                * If above mount call was successful, and this mount is content protection 
+                * enabled, then verify the on-disk EA on the root to ensure that the filesystem 
+                * is of a suitable vintage to allow the mount to proceed.  
+                */
+               if ((retval == 0) && (cp_fs_protected (mp))) {
+                       int err = 0;
+                       struct cp_root_xattr xattr;
+                       bzero (&xattr, sizeof(struct cp_root_xattr));
+                       hfsmp = vfs_fsprivate(mp);
+
+                       /* go get the EA to get the version information */
+                       err = cp_getrootxattr (hfsmp, &xattr);
+                       /* If there was no EA there, then write one out. */
+                       if (err == ENOATTR) {
+                               bzero(&xattr, sizeof(struct cp_root_xattr));
+                               xattr.major_version = CP_CURRENT_MAJOR_VERS;
+                               xattr.minor_version = CP_CURRENT_MINOR_VERS;
+                               xattr.flags = 0;
+
+                               err = cp_setrootxattr (hfsmp, &xattr);
+                       }       
+                       /* 
+                        * For any other error, including having an out of date CP version in the
+                        * EA, or for an error out of cp_setrootxattr, deny the mount 
+                        * and do not proceed further.
+                        */
+                       if (err || xattr.major_version != CP_CURRENT_MAJOR_VERS)  {
+                               /* Deny the mount and tear down. */
+                               retval = EPERM;
+                               (void) hfs_unmount (mp, MNT_FORCE, context);
+                       }       
+               }                                 
+#endif
        }
 out:
        if (retval == 0) {
        }
 out:
        if (retval == 0) {
@@ -627,7 +772,7 @@ hfs_reload_callback(struct vnode *vp, void *cargs)
        /*
         * Re-read cnode data for all active vnodes (non-metadata files).
         */
        /*
         * Re-read cnode data for all active vnodes (non-metadata files).
         */
-       if (!vnode_issystem(vp) && !VNODE_IS_RSRC(vp)) {
+       if (!vnode_issystem(vp) && !VNODE_IS_RSRC(vp) && (cp->c_fileid >= kHFSFirstUserCatalogNodeID)) {
                struct cat_fork *datafork;
                struct cat_desc desc;
 
                struct cat_fork *datafork;
                struct cat_desc desc;
 
@@ -661,7 +806,7 @@ hfs_reload_callback(struct vnode *vp, void *cargs)
  *     re-load B-tree header data.
  *     re-read cnode data for all active vnodes.
  */
  *     re-load B-tree header data.
  *     re-read cnode data for all active vnodes.
  */
-static int
+int
 hfs_reload(struct mount *mountp)
 {
        register struct vnode *devvp;
 hfs_reload(struct mount *mountp)
 {
        register struct vnode *devvp;
@@ -855,7 +1000,7 @@ hfs_syncer(void *arg0, void *unused)
     //
     if (hfsmp->hfs_mp->mnt_pending_write_size > hfsmp->hfs_max_pending_io) {
            int counter=0;
     //
     if (hfsmp->hfs_mp->mnt_pending_write_size > hfsmp->hfs_max_pending_io) {
            int counter=0;
-           uint64_t pending_io, start, rate;
+           uint64_t pending_io, start, rate = 0;
            
            no_max = 0;
 
            
            no_max = 0;
 
@@ -875,7 +1020,7 @@ hfs_syncer(void *arg0, void *unused)
            }
 
            if (hfsmp->jnl) {
            }
 
            if (hfsmp->jnl) {
-                   journal_flush(hfsmp->jnl);
+                   journal_flush(hfsmp->jnl, FALSE);
            } else {
                    hfs_sync(hfsmp->hfs_mp, MNT_WAIT, vfs_context_kernel());
            }
            } else {
                    hfs_sync(hfsmp->hfs_mp, MNT_WAIT, vfs_context_kernel());
            }
@@ -883,7 +1028,9 @@ hfs_syncer(void *arg0, void *unused)
            clock_get_calendar_microtime(&secs, &usecs);
            now = ((uint64_t)secs * 1000000ULL) + (uint64_t)usecs;
            hfsmp->hfs_last_sync_time = now;
            clock_get_calendar_microtime(&secs, &usecs);
            now = ((uint64_t)secs * 1000000ULL) + (uint64_t)usecs;
            hfsmp->hfs_last_sync_time = now;
-           rate = ((pending_io * 1000000ULL) / (now - start));     // yields bytes per second
+           if (now != start) {
+                   rate = ((pending_io * 1000000ULL) / (now - start));     // yields bytes per second
+           }
 
            hfs_end_transaction(hfsmp);
            
 
            hfs_end_transaction(hfsmp);
            
@@ -893,7 +1040,7 @@ hfs_syncer(void *arg0, void *unused)
            // than 2 seconds, adjust hfs_max_pending_io so that we
            // will allow about 1.5 seconds of i/o to queue up.
            //
            // than 2 seconds, adjust hfs_max_pending_io so that we
            // will allow about 1.5 seconds of i/o to queue up.
            //
-           if ((now - start) >= 300000) {
+           if (((now - start) >= 300000) && (rate != 0)) {
                    uint64_t scale = (pending_io * 100) / rate;
                    
                    if (scale < 100 || scale > 200) {
                    uint64_t scale = (pending_io * 100) / rate;
                    
                    if (scale < 100 || scale > 200) {
@@ -916,11 +1063,11 @@ hfs_syncer(void *arg0, void *unused)
            // now.  Else we defer the sync and reschedule it.
            //
            if (hfsmp->jnl) {
            // now.  Else we defer the sync and reschedule it.
            //
            if (hfsmp->jnl) {
-                   lck_rw_lock_shared(&hfsmp->hfs_global_lock);
+                       hfs_lock_global (hfsmp, HFS_SHARED_LOCK);
 
 
-                   journal_flush(hfsmp->jnl);
+                   journal_flush(hfsmp->jnl, FALSE);
 
 
-                   lck_rw_unlock_shared(&hfsmp->hfs_global_lock);
+                       hfs_unlock_global (hfsmp);
            } else {
                    hfs_sync(hfsmp->hfs_mp, MNT_WAIT, vfs_context_kernel());
            }
            } else {
                    hfs_sync(hfsmp->hfs_mp, MNT_WAIT, vfs_context_kernel());
            }
@@ -955,10 +1102,119 @@ hfs_syncer(void *arg0, void *unused)
 
 extern int IOBSDIsMediaEjectable( const char *cdev_name );
 
 
 extern int IOBSDIsMediaEjectable( const char *cdev_name );
 
+/*
+ * Initialization code for Red-Black Tree Allocator
+ * 
+ * This function will build the two red-black trees necessary for allocating space
+ * from the metadata zone as well as normal allocations.  Currently, we use 
+ * an advisory read to get most of the data into the buffer cache. 
+ * This function is intended to be run in a separate thread so as not to slow down mount.
+ * 
+ */
+
+void 
+hfs_initialize_allocator (struct hfsmount *hfsmp) {
+       
+#if CONFIG_HFS_ALLOC_RBTREE
+       u_int32_t err;
+       
+       /*
+        * Take the allocation file lock.  Journal transactions will block until
+        * we're done here. 
+        */
+       int flags = hfs_systemfile_lock(hfsmp, SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
+       
+       /*
+        * GenerateTree assumes that the bitmap lock is held when you call the function.
+        * It will drop and re-acquire the lock periodically as needed to let other allocations 
+        * through.  It returns with the bitmap lock held. Since we only maintain one tree,
+        * we don't need to specify a start block (always starts at 0).
+        */
+       err = GenerateTree(hfsmp, hfsmp->totalBlocks, &flags, 1);
+       if (err) {
+               goto bailout;
+       }
+       /* Mark offset tree as built */
+       hfsmp->extent_tree_flags |= HFS_ALLOC_RB_ACTIVE;
+       
+bailout:
+       /* 
+        * GenerateTree may drop the bitmap lock during operation in order to give other
+        * threads a chance to allocate blocks, but it will always return with the lock held, so
+        * we don't need to re-grab the lock in order to update the TREEBUILD_INFLIGHT bit.
+        */
+       hfsmp->extent_tree_flags &= ~HFS_ALLOC_TREEBUILD_INFLIGHT;
+       if (err != 0) {
+               /* Wakeup any waiters on the allocation bitmap lock */
+               wakeup((caddr_t)&hfsmp->extent_tree_flags);
+       }
+       
+       hfs_systemfile_unlock(hfsmp, flags);
+#else
+#pragma unused (hfsmp)
+#endif
+}
+
+
+/* 
+ * Teardown code for the Red-Black Tree allocator. 
+ * This function consolidates the code which serializes with respect
+ * to a thread that may be potentially still building the tree when we need to begin 
+ * tearing it down.   Since the red-black tree may not be live when we enter this function
+ * we return:
+ *             1 -> Tree was live.
+ *             0 -> Tree was not active at time of call.
+ */
+
+int 
+hfs_teardown_allocator (struct hfsmount *hfsmp) {
+       int rb_used = 0;
+       
+#if CONFIG_HFS_ALLOC_RBTREE
+       
+       int flags = 0;
+       
+       /* 
+        * Check to see if the tree-generation is still on-going.
+        * If it is, then block until it's done.
+        */
+       
+       flags = hfs_systemfile_lock(hfsmp, SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
+       
+       
+       while (hfsmp->extent_tree_flags & HFS_ALLOC_TREEBUILD_INFLIGHT) {
+               hfsmp->extent_tree_flags |= HFS_ALLOC_TEARDOWN_INFLIGHT;
+               
+               lck_rw_sleep(&(VTOC(hfsmp->hfs_allocation_vp))->c_rwlock, LCK_SLEEP_EXCLUSIVE, 
+                                        &hfsmp->extent_tree_flags, THREAD_UNINT);
+       }
+       
+       if (hfs_isrbtree_active (hfsmp)) {
+               rb_used = 1;
+       
+               /* Tear down the RB Trees while we have the bitmap locked */
+               DestroyTrees(hfsmp);
+
+       }
+
+       hfs_systemfile_unlock(hfsmp, flags);
+#else
+       #pragma unused (hfsmp)
+#endif
+       return rb_used;
+       
+}
+
+
+static int hfs_root_unmounted_cleanly = 0;
+
+SYSCTL_DECL(_vfs_generic);
+SYSCTL_INT(_vfs_generic, OID_AUTO, root_unmounted_cleanly, CTLFLAG_RD, &hfs_root_unmounted_cleanly, 0, "Root filesystem was unmounted cleanly");
+
 /*
  * Common code for mount and mountroot
  */
 /*
  * Common code for mount and mountroot
  */
-static int
+int
 hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
             int journal_replay_only, vfs_context_t context)
 {
 hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
             int journal_replay_only, vfs_context_t context)
 {
@@ -983,7 +1239,11 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        daddr64_t mdb_offset;
        int isvirtual = 0;
        int isroot = 0;
        daddr64_t mdb_offset;
        int isvirtual = 0;
        int isroot = 0;
-
+       int isssd;
+#if CONFIG_HFS_ALLOC_RBTREE
+       thread_t allocator_thread;
+#endif
+       
        if (args == NULL) {
                /* only hfs_mountroot passes us NULL as the 'args' argument */
                isroot = 1;
        if (args == NULL) {
                /* only hfs_mountroot passes us NULL as the 'args' argument */
                isroot = 1;
@@ -1004,6 +1264,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
 
        /* Get the logical block size (treated as physical block size everywhere) */
        if (VNOP_IOCTL(devvp, DKIOCGETBLOCKSIZE, (caddr_t)&log_blksize, 0, context)) {
 
        /* Get the logical block size (treated as physical block size everywhere) */
        if (VNOP_IOCTL(devvp, DKIOCGETBLOCKSIZE, (caddr_t)&log_blksize, 0, context)) {
+               if (HFS_MOUNT_DEBUG) {
+                       printf("hfs_mountfs: DKIOCGETBLOCKSIZE failed\n");
+               }
                retval = ENXIO;
                goto error_exit;
        }
                retval = ENXIO;
                goto error_exit;
        }
@@ -1017,6 +1280,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        retval = VNOP_IOCTL(devvp, DKIOCGETPHYSICALBLOCKSIZE, (caddr_t)&phys_blksize, 0, context);
        if (retval) {
                if ((retval != ENOTSUP) && (retval != ENOTTY)) {
        retval = VNOP_IOCTL(devvp, DKIOCGETPHYSICALBLOCKSIZE, (caddr_t)&phys_blksize, 0, context);
        if (retval) {
                if ((retval != ENOTSUP) && (retval != ENOTTY)) {
+                       if (HFS_MOUNT_DEBUG) {
+                               printf("hfs_mountfs: DKIOCGETPHYSICALBLOCKSIZE failed\n");
+                       }
                        retval = ENXIO;
                        goto error_exit;
                }
                        retval = ENXIO;
                        goto error_exit;
                }
@@ -1036,6 +1302,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                u_int32_t size512 = 512;
 
                if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&size512, FWRITE, context)) {
                u_int32_t size512 = 512;
 
                if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&size512, FWRITE, context)) {
+                       if (HFS_MOUNT_DEBUG) {
+                               printf("hfs_mountfs: DKIOCSETBLOCKSIZE failed \n");
+                       }
                        retval = ENXIO;
                        goto error_exit;
                }
                        retval = ENXIO;
                        goto error_exit;
                }
@@ -1044,7 +1313,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&log_blkcnt, 0, context)) {
                /* resetting block size may fail if getting block count did */
                (void)VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&log_blksize, FWRITE, context);
        if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&log_blkcnt, 0, context)) {
                /* resetting block size may fail if getting block count did */
                (void)VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&log_blksize, FWRITE, context);
-
+               if (HFS_MOUNT_DEBUG) {
+                       printf("hfs_mountfs: DKIOCGETBLOCKCOUNT failed\n");
+               }
                retval = ENXIO;
                goto error_exit;
        }
                retval = ENXIO;
                goto error_exit;
        }
@@ -1080,11 +1351,17 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        /* Now switch to our preferred physical block size. */
        if (log_blksize > 512) {
                if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&log_blksize, FWRITE, context)) {
        /* Now switch to our preferred physical block size. */
        if (log_blksize > 512) {
                if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&log_blksize, FWRITE, context)) {
+                       if (HFS_MOUNT_DEBUG) { 
+                               printf("hfs_mountfs: DKIOCSETBLOCKSIZE (2) failed\n");
+                       }
                        retval = ENXIO;
                        goto error_exit;
                }
                /* Get the count of physical blocks. */
                if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&log_blkcnt, 0, context)) {
                        retval = ENXIO;
                        goto error_exit;
                }
                /* Get the count of physical blocks. */
                if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&log_blkcnt, 0, context)) {
+                       if (HFS_MOUNT_DEBUG) { 
+                               printf("hfs_mountfs: DKIOCGETBLOCKCOUNT (2) failed\n");
+                       }
                        retval = ENXIO;
                        goto error_exit;
                }
                        retval = ENXIO;
                        goto error_exit;
                }
@@ -1100,11 +1377,17 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        if ((retval = (int)buf_meta_bread(devvp, 
                                HFS_PHYSBLK_ROUNDDOWN(mdb_offset, (phys_blksize/log_blksize)), 
                                phys_blksize, cred, &bp))) {
        if ((retval = (int)buf_meta_bread(devvp, 
                                HFS_PHYSBLK_ROUNDDOWN(mdb_offset, (phys_blksize/log_blksize)), 
                                phys_blksize, cred, &bp))) {
+               if (HFS_MOUNT_DEBUG) {
+                       printf("hfs_mountfs: buf_meta_bread failed with %d\n", retval);
+               }
                goto error_exit;
        }
        MALLOC(mdbp, HFSMasterDirectoryBlock *, kMDBSize, M_TEMP, M_WAITOK);
        if (mdbp == NULL) {
                retval = ENOMEM;
                goto error_exit;
        }
        MALLOC(mdbp, HFSMasterDirectoryBlock *, kMDBSize, M_TEMP, M_WAITOK);
        if (mdbp == NULL) {
                retval = ENOMEM;
+               if (HFS_MOUNT_DEBUG) { 
+                       printf("hfs_mountfs: MALLOC failed\n");
+               }
                goto error_exit;
        }
        bcopy((char *)buf_dataptr(bp) + HFS_PRI_OFFSET(phys_blksize), mdbp, kMDBSize);
                goto error_exit;
        }
        bcopy((char *)buf_dataptr(bp) + HFS_PRI_OFFSET(phys_blksize), mdbp, kMDBSize);
@@ -1113,6 +1396,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
 
        MALLOC(hfsmp, struct hfsmount *, sizeof(struct hfsmount), M_HFSMNT, M_WAITOK);
        if (hfsmp == NULL) {
 
        MALLOC(hfsmp, struct hfsmount *, sizeof(struct hfsmount), M_HFSMNT, M_WAITOK);
        if (hfsmp == NULL) {
+               if (HFS_MOUNT_DEBUG) { 
+                       printf("hfs_mountfs: MALLOC (2) failed\n");
+               }
                retval = ENOMEM;
                goto error_exit;
        }
                retval = ENOMEM;
                goto error_exit;
        }
@@ -1120,6 +1406,17 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        
        hfs_chashinit_finish(hfsmp);
 
        
        hfs_chashinit_finish(hfsmp);
 
+       /*
+        * See if the disk is a solid state device.  We need this to decide what to do about 
+        * hotfiles.
+        */
+       if (VNOP_IOCTL(devvp, DKIOCISSOLIDSTATE, (caddr_t)&isssd, 0, context) == 0) {
+               if (isssd) {
+                       hfsmp->hfs_flags |= HFS_SSD;
+               }
+       }
+
+
        /*
         *  Init the volume information structure
         */
        /*
         *  Init the volume information structure
         */
@@ -1128,7 +1425,8 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        lck_mtx_init(&hfsmp->hfc_mutex, hfs_mutex_group, hfs_lock_attr);
        lck_rw_init(&hfsmp->hfs_global_lock, hfs_rwlock_group, hfs_lock_attr);
        lck_rw_init(&hfsmp->hfs_insync, hfs_rwlock_group, hfs_lock_attr);
        lck_mtx_init(&hfsmp->hfc_mutex, hfs_mutex_group, hfs_lock_attr);
        lck_rw_init(&hfsmp->hfs_global_lock, hfs_rwlock_group, hfs_lock_attr);
        lck_rw_init(&hfsmp->hfs_insync, hfs_rwlock_group, hfs_lock_attr);
-
+       lck_spin_init(&hfsmp->vcbFreeExtLock, hfs_spinlock_group, hfs_lock_attr);
+       
        vfs_setfsprivate(mp, hfsmp);
        hfsmp->hfs_mp = mp;                     /* Make VFSTOHFS work */
        hfsmp->hfs_raw_dev = vnode_specrdev(devvp);
        vfs_setfsprivate(mp, hfsmp);
        hfsmp->hfs_mp = mp;                     /* Make VFSTOHFS work */
        hfsmp->hfs_raw_dev = vnode_specrdev(devvp);
@@ -1201,6 +1499,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                        retval = EROFS;
                        goto error_exit;
                }
                        retval = EROFS;
                        goto error_exit;
                }
+
+               printf("hfs_mountfs: Mounting HFS Standard volumes was deprecated in Mac OS 10.7 \n");
+
                /* Treat it as if it's read-only and not writeable */
                hfsmp->hfs_flags |= HFS_READ_ONLY;
                hfsmp->hfs_flags &= ~HFS_WRITEABLE_MEDIA;
                /* Treat it as if it's read-only and not writeable */
                hfsmp->hfs_flags |= HFS_READ_ONLY;
                hfsmp->hfs_flags &= ~HFS_WRITEABLE_MEDIA;
@@ -1272,11 +1573,18 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                                log_blksize = 512;
                                if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE,
                                    (caddr_t)&log_blksize, FWRITE, context)) {
                                log_blksize = 512;
                                if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE,
                                    (caddr_t)&log_blksize, FWRITE, context)) {
+
+                                       if (HFS_MOUNT_DEBUG) { 
+                                               printf("hfs_mountfs: DKIOCSETBLOCKSIZE (3) failed\n");
+                                       }                               
                                        retval = ENXIO;
                                        goto error_exit;
                                }
                                if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT,
                                    (caddr_t)&log_blkcnt, 0, context)) {
                                        retval = ENXIO;
                                        goto error_exit;
                                }
                                if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT,
                                    (caddr_t)&log_blkcnt, 0, context)) {
+                                       if (HFS_MOUNT_DEBUG) { 
+                                               printf("hfs_mountfs: DKIOCGETBLOCKCOUNT (3) failed\n");
+                                       }
                                        retval = ENXIO;
                                        goto error_exit;
                                }
                                        retval = ENXIO;
                                        goto error_exit;
                                }
@@ -1299,8 +1607,12 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                        mdb_offset = (daddr64_t)((embeddedOffset / log_blksize) + HFS_PRI_SECTOR(log_blksize));
                        retval = (int)buf_meta_bread(devvp, HFS_PHYSBLK_ROUNDDOWN(mdb_offset, hfsmp->hfs_log_per_phys),
                                        phys_blksize, cred, &bp);
                        mdb_offset = (daddr64_t)((embeddedOffset / log_blksize) + HFS_PRI_SECTOR(log_blksize));
                        retval = (int)buf_meta_bread(devvp, HFS_PHYSBLK_ROUNDDOWN(mdb_offset, hfsmp->hfs_log_per_phys),
                                        phys_blksize, cred, &bp);
-                       if (retval)
+                       if (retval) {
+                               if (HFS_MOUNT_DEBUG) { 
+                                       printf("hfs_mountfs: buf_meta_bread (2) failed with %d\n", retval);
+                               }
                                goto error_exit;
                                goto error_exit;
+                       }
                        bcopy((char *)buf_dataptr(bp) + HFS_PRI_OFFSET(phys_blksize), mdbp, 512);
                        buf_brelse(bp);
                        bp = NULL;
                        bcopy((char *)buf_dataptr(bp) + HFS_PRI_OFFSET(phys_blksize), mdbp, 512);
                        buf_brelse(bp);
                        bp = NULL;
@@ -1311,6 +1623,10 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                        vhp = (HFSPlusVolumeHeader*) mdbp;
                }
 
                        vhp = (HFSPlusVolumeHeader*) mdbp;
                }
 
+               if (isroot) {
+                       hfs_root_unmounted_cleanly = (SWAP_BE32(vhp->attributes) & kHFSVolumeUnmountedMask) != 0;
+               }
+
                /*
                 * On inconsistent disks, do not allow read-write mount
                 * unless it is the boot volume being mounted.  We also
                /*
                 * On inconsistent disks, do not allow read-write mount
                 * unless it is the boot volume being mounted.  We also
@@ -1323,6 +1639,10 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                   && (SWAP_BE32(vhp->attributes) & kHFSVolumeInconsistentMask)
                   && !journal_replay_only
                   && !(hfsmp->hfs_flags & HFS_READ_ONLY)) {
                   && (SWAP_BE32(vhp->attributes) & kHFSVolumeInconsistentMask)
                   && !journal_replay_only
                   && !(hfsmp->hfs_flags & HFS_READ_ONLY)) {
+                       
+                       if (HFS_MOUNT_DEBUG) { 
+                               printf("hfs_mountfs: failed to mount non-root inconsistent disk\n");
+                       }
                        retval = EINVAL;
                        goto error_exit;
                }
                        retval = EINVAL;
                        goto error_exit;
                }
@@ -1360,6 +1680,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                                        // EROFS is a special error code that means the volume has an external
                                        // journal which we couldn't find.  in that case we do not want to
                                        // rewrite the volume header - we'll just refuse to mount the volume.
                                        // EROFS is a special error code that means the volume has an external
                                        // journal which we couldn't find.  in that case we do not want to
                                        // rewrite the volume header - we'll just refuse to mount the volume.
+                                       if (HFS_MOUNT_DEBUG) { 
+                                               printf("hfs_mountfs: hfs_early_journal_init indicated external jnl \n");
+                                       }
                                        retval = EINVAL;
                                        goto error_exit;
                                }
                                        retval = EINVAL;
                                        goto error_exit;
                                }
@@ -1368,7 +1691,11 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                                // to be "FSK!" which fsck_hfs will see and force the fsck instead
                                // of just bailing out because the volume is journaled.
                                if (!ronly) {
                                // to be "FSK!" which fsck_hfs will see and force the fsck instead
                                // of just bailing out because the volume is journaled.
                                if (!ronly) {
-                                   HFSPlusVolumeHeader *jvhp;
+                                       if (HFS_MOUNT_DEBUG) { 
+                                               printf("hfs_mountfs: hfs_early_journal_init failed, setting to FSK \n");
+                                       }
+
+                                       HFSPlusVolumeHeader *jvhp;
 
                                    hfsmp->hfs_flags |= HFS_NEED_JNL_RESET;
                                    
 
                                    hfsmp->hfs_flags |= HFS_NEED_JNL_RESET;
                                    
@@ -1403,6 +1730,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                                // in the hopes that fsck_hfs will be able to
                                // fix any damage that exists on the volume.
                                if ( !(vfs_flags(mp) & MNT_ROOTFS)) {
                                // in the hopes that fsck_hfs will be able to
                                // fix any damage that exists on the volume.
                                if ( !(vfs_flags(mp) & MNT_ROOTFS)) {
+                                       if (HFS_MOUNT_DEBUG) { 
+                                               printf("hfs_mountfs: hfs_early_journal_init failed, erroring out \n");
+                                       }
                                    retval = EINVAL;
                                    goto error_exit;
                                }
                                    retval = EINVAL;
                                    goto error_exit;
                                }
@@ -1431,10 +1761,16 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                                "(%d) switching to 512\n", log_blksize);
                        log_blksize = 512;
                        if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&log_blksize, FWRITE, context)) {
                                "(%d) switching to 512\n", log_blksize);
                        log_blksize = 512;
                        if (VNOP_IOCTL(devvp, DKIOCSETBLOCKSIZE, (caddr_t)&log_blksize, FWRITE, context)) {
+                               if (HFS_MOUNT_DEBUG) { 
+                                       printf("hfs_mountfs: DKIOCSETBLOCKSIZE (4) failed \n");
+                               }
                                retval = ENXIO;
                                goto error_exit;
                        }
                        if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&log_blkcnt, 0, context)) {
                                retval = ENXIO;
                                goto error_exit;
                        }
                        if (VNOP_IOCTL(devvp, DKIOCGETBLOCKCOUNT, (caddr_t)&log_blkcnt, 0, context)) {
+                               if (HFS_MOUNT_DEBUG) { 
+                                       printf("hfs_mountfs: DKIOCGETBLOCKCOUNT (4) failed \n");
+                               }
                                retval = ENXIO;
                                goto error_exit;
                        }
                                retval = ENXIO;
                                goto error_exit;
                        }
@@ -1455,6 +1791,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                                        // to be "FSK!" which fsck_hfs will see and force the fsck instead
                                        // of just bailing out because the volume is journaled.
                                        if (!ronly) {
                                        // to be "FSK!" which fsck_hfs will see and force the fsck instead
                                        // of just bailing out because the volume is journaled.
                                        if (!ronly) {
+                                               if (HFS_MOUNT_DEBUG) { 
+                                                       printf("hfs_mountfs: hfs_early_journal_init (2) resetting.. \n");
+                                               }
                                        HFSPlusVolumeHeader *jvhp;
 
                                        hfsmp->hfs_flags |= HFS_NEED_JNL_RESET;
                                        HFSPlusVolumeHeader *jvhp;
 
                                        hfsmp->hfs_flags |= HFS_NEED_JNL_RESET;
@@ -1489,6 +1828,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                                        // in the hopes that fsck_hfs will be able to
                                        // fix any damage that exists on the volume.
                                        if ( !(vfs_flags(mp) & MNT_ROOTFS)) {
                                        // in the hopes that fsck_hfs will be able to
                                        // fix any damage that exists on the volume.
                                        if ( !(vfs_flags(mp) & MNT_ROOTFS)) {
+                                               if (HFS_MOUNT_DEBUG) { 
+                                                       printf("hfs_mountfs: hfs_early_journal_init (2) failed \n");
+                                               }
                                        retval = EINVAL;
                                        goto error_exit;
                                        }
                                        retval = EINVAL;
                                        goto error_exit;
                                        }
@@ -1497,6 +1839,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
 
                        /* Try again with a smaller block size... */
                        retval = hfs_MountHFSPlusVolume(hfsmp, vhp, embeddedOffset, disksize, p, args, cred);
 
                        /* Try again with a smaller block size... */
                        retval = hfs_MountHFSPlusVolume(hfsmp, vhp, embeddedOffset, disksize, p, args, cred);
+                       if (retval && HFS_MOUNT_DEBUG) {
+                               printf("hfs_MountHFSPlusVolume (late) returned %d\n",retval); 
+                       }
                }
                if (retval)
                        (void) hfs_relconverter(0);
                }
                if (retval)
                        (void) hfs_relconverter(0);
@@ -1507,6 +1852,9 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
        hfsmp->hfs_last_mounted_mtime = hfsmp->hfs_mtime;
 
        if ( retval ) {
        hfsmp->hfs_last_mounted_mtime = hfsmp->hfs_mtime;
 
        if ( retval ) {
+               if (HFS_MOUNT_DEBUG) { 
+                       printf("hfs_mountfs: encountered failure %d \n", retval);
+               }
                goto error_exit;
        }
 
                goto error_exit;
        }
 
@@ -1523,7 +1871,7 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                mp->mnt_vtable->vfc_vfsflags |= VFC_VFSDIRLINKS;
        } else {
                /* HFS standard doesn't support extended readdir! */
                mp->mnt_vtable->vfc_vfsflags |= VFC_VFSDIRLINKS;
        } else {
                /* HFS standard doesn't support extended readdir! */
-               mp->mnt_vtable->vfc_vfsflags &= ~VFC_VFSREADDIR_EXTENDED;
+               mount_set_noreaddirext (mp);
        }
 
        if (args) {
        }
 
        if (args) {
@@ -1548,10 +1896,10 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                /*
                 * Set the free space warning levels for the root volume:
                 *
                /*
                 * Set the free space warning levels for the root volume:
                 *
-                * Set the "danger" limit to 5% of the volume size or 125MB, whichever
-                * is less.  Set the "warning" limit to 10% of the volume size or 250MB,
+                * Set the "danger" limit to 5% of the volume size or 512MB, whichever
+                * is less.  Set the "warning" limit to 10% of the volume size or 1GB,
                 * whichever is less.  And last, set the "desired" freespace level to
                 * whichever is less.  And last, set the "desired" freespace level to
-                * to 11% of the volume size or 375MB, whichever is less.
+                * to 11% of the volume size or 1.25GB, whichever is less.
                 */
                hfsmp->hfs_freespace_notify_dangerlimit =
                        MIN(HFS_ROOTVERYLOWDISKTRIGGERLEVEL / HFSTOVCB(hfsmp)->blockSize,
                 */
                hfsmp->hfs_freespace_notify_dangerlimit =
                        MIN(HFS_ROOTVERYLOWDISKTRIGGERLEVEL / HFSTOVCB(hfsmp)->blockSize,
@@ -1583,6 +1931,32 @@ hfs_mountfs(struct vnode *devvp, struct mount *mp, struct hfs_mount_args *args,
                        }
                }
        }
                        }
                }
        }
+       
+#if CONFIG_HFS_ALLOC_RBTREE
+       /* 
+        * We spawn a thread to create the pair of red-black trees for this volume.
+        * However, in so doing, we must be careful to ensure that if this thread is still
+        * running after mount has finished, it doesn't interfere with an unmount. Specifically,
+        * we'll need to set a bit that indicates we're in progress building the trees here.  
+        * Unmount will check for this bit, and then if it's set, mark a corresponding bit that
+        * notifies the tree generation code that an unmount is waiting.  Also mark the bit that
+        * indicates the tree is live and operating.
+        *
+        * Only do this if we're operating on a read-write mount (we wouldn't care for read-only).
+        */
+       
+       if ((hfsmp->hfs_flags & HFS_READ_ONLY) == 0) {
+               hfsmp->extent_tree_flags |= (HFS_ALLOC_TREEBUILD_INFLIGHT | HFS_ALLOC_RB_ENABLED);
+               
+               /* Initialize EOF counter so that the thread can assume it started at initial values */
+               hfsmp->offset_block_end = 0;
+               InitTree(hfsmp);
+               
+               kernel_thread_start ((thread_continue_t) hfs_initialize_allocator , hfsmp, &allocator_thread);
+               thread_deallocate(allocator_thread);
+       }
+       
+#endif
 
        /*
         * Start looking for free space to drop below this level and generate a
 
        /*
         * Start looking for free space to drop below this level and generate a
@@ -1636,7 +2010,7 @@ hfs_start(__unused struct mount *mp, __unused int flags, __unused vfs_context_t
 /*
  * unmount system call
  */
 /*
  * unmount system call
  */
-static int
+int
 hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
 {
        struct proc *p = vfs_context_proc(context);
 hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
 {
        struct proc *p = vfs_context_proc(context);
@@ -1645,6 +2019,7 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
        int flags;
        int force;
        int started_tr = 0;
        int flags;
        int force;
        int started_tr = 0;
+       int rb_used = 0;
 
        flags = 0;
        force = 0;
 
        flags = 0;
        force = 0;
@@ -1691,6 +2066,10 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
                        panic("hfs_unmount: pm_sync_incomplete underflow!\n");
        }
        
                        panic("hfs_unmount: pm_sync_incomplete underflow!\n");
        }
        
+#if CONFIG_HFS_ALLOC_RBTREE
+       rb_used = hfs_teardown_allocator(hfsmp);
+#endif
+       
        /*
         * Flush out the b-trees, volume bitmap and Volume Header
         */
        /*
         * Flush out the b-trees, volume bitmap and Volume Header
         */
@@ -1753,22 +2132,31 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
                        HFSTOVCB(hfsmp)->vcbAtrb |= kHFSVolumeUnmountedMask;
                }
 
                        HFSTOVCB(hfsmp)->vcbAtrb |= kHFSVolumeUnmountedMask;
                }
 
-               if (hfsmp->hfs_flags & HFS_HAS_SPARSE_DEVICE) {
-                       int i;
-                       u_int32_t min_start = hfsmp->totalBlocks;
-
-                       // set the nextAllocation pointer to the smallest free block number
-                       // we've seen so on the next mount we won't rescan unnecessarily
-                       for(i=0; i < (int)hfsmp->vcbFreeExtCnt; i++) {
-                               if (hfsmp->vcbFreeExt[i].startBlock < min_start) {
-                                       min_start = hfsmp->vcbFreeExt[i].startBlock;
+               
+               if (rb_used) {
+                       /* If the rb-tree was live, just set min_start to 0 */
+                       hfsmp->nextAllocation = 0;
+               } 
+               else {
+                       if (hfsmp->hfs_flags & HFS_HAS_SPARSE_DEVICE) {
+                               int i;
+                               u_int32_t min_start = hfsmp->totalBlocks;
+                               
+                               // set the nextAllocation pointer to the smallest free block number
+                               // we've seen so on the next mount we won't rescan unnecessarily
+                               lck_spin_lock(&hfsmp->vcbFreeExtLock);
+                               for(i=0; i < (int)hfsmp->vcbFreeExtCnt; i++) {
+                                       if (hfsmp->vcbFreeExt[i].startBlock < min_start) {
+                                               min_start = hfsmp->vcbFreeExt[i].startBlock;
+                                       }
+                               }
+                               lck_spin_unlock(&hfsmp->vcbFreeExtLock);
+                               if (min_start < hfsmp->nextAllocation) {
+                                       hfsmp->nextAllocation = min_start;
                                }
                                }
-                       }
-                       if (min_start < hfsmp->nextAllocation) {
-                               hfsmp->nextAllocation = min_start;
                        }
                }
                        }
                }
-
+               
 
                retval = hfs_flushvolumeheader(hfsmp, MNT_WAIT, 0);
                if (retval) {
 
                retval = hfs_flushvolumeheader(hfsmp, MNT_WAIT, 0);
                if (retval) {
@@ -1784,7 +2172,7 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
        }
 
        if (hfsmp->jnl) {
        }
 
        if (hfsmp->jnl) {
-               hfs_journal_flush(hfsmp);
+               hfs_journal_flush(hfsmp, FALSE);
        }
        
        /*
        }
        
        /*
@@ -1792,11 +2180,6 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
         */
        (void) hfsUnmount(hfsmp, p);
 
         */
        (void) hfsUnmount(hfsmp, p);
 
-       /*
-        * Last chance to dump unreferenced system files.
-        */
-       (void) vflush(mp, NULLVP, FORCECLOSE);
-
        if (HFSTOVCB(hfsmp)->vcbSigWord == kHFSSigWord)
                (void) hfs_relconverter(hfsmp->hfs_encoding);
 
        if (HFSTOVCB(hfsmp)->vcbSigWord == kHFSSigWord)
                (void) hfs_relconverter(hfsmp->hfs_encoding);
 
@@ -1818,7 +2201,12 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
        }
        // XXXdbg
 
        }
        // XXXdbg
 
-#ifdef HFS_SPARSE_DEV
+       /*
+        * Last chance to dump unreferenced system files.
+        */
+       (void) vflush(mp, NULLVP, FORCECLOSE);
+
+#if HFS_SPARSE_DEV
        /* Drop our reference on the backing fs (if any). */
        if ((hfsmp->hfs_flags & HFS_HAS_SPARSE_DEVICE) && hfsmp->hfs_backingfs_rootvp) {
                struct vnode * tmpvp;
        /* Drop our reference on the backing fs (if any). */
        if ((hfsmp->hfs_flags & HFS_HAS_SPARSE_DEVICE) && hfsmp->hfs_backingfs_rootvp) {
                struct vnode * tmpvp;
@@ -1830,6 +2218,7 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
        }
 #endif /* HFS_SPARSE_DEV */
        lck_mtx_destroy(&hfsmp->hfc_mutex, hfs_mutex_group);
        }
 #endif /* HFS_SPARSE_DEV */
        lck_mtx_destroy(&hfsmp->hfc_mutex, hfs_mutex_group);
+       lck_spin_destroy(&hfsmp->vcbFreeExtLock, hfs_spinlock_group);
        vnode_rele(hfsmp->hfs_devvp);
 
        hfs_delete_chash(hfsmp);
        vnode_rele(hfsmp->hfs_devvp);
 
        hfs_delete_chash(hfsmp);
@@ -1851,7 +2240,7 @@ hfs_unmount(struct mount *mp, int mntflags, vfs_context_t context)
 static int
 hfs_vfs_root(struct mount *mp, struct vnode **vpp, __unused vfs_context_t context)
 {
 static int
 hfs_vfs_root(struct mount *mp, struct vnode **vpp, __unused vfs_context_t context)
 {
-       return hfs_vget(VFSTOHFS(mp), (cnid_t)kHFSRootFolderID, vpp, 1);
+       return hfs_vget(VFSTOHFS(mp), (cnid_t)kHFSRootFolderID, vpp, 1, 0);
 }
 
 
 }
 
 
@@ -1872,7 +2261,7 @@ hfs_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t datap, vfs_context_t
        int cmd, type, error;
 
        if (uid == ~0U)
        int cmd, type, error;
 
        if (uid == ~0U)
-               uid = vfs_context_ucred(context)->cr_ruid;
+               uid = kauth_cred_getuid(vfs_context_ucred(context));
        cmd = cmds >> SUBCMDSHIFT;
 
        switch (cmd) {
        cmd = cmds >> SUBCMDSHIFT;
 
        switch (cmd) {
@@ -1880,7 +2269,7 @@ hfs_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t datap, vfs_context_t
        case Q_QUOTASTAT:
                break;
        case Q_GETQUOTA:
        case Q_QUOTASTAT:
                break;
        case Q_GETQUOTA:
-               if (uid == vfs_context_ucred(context)->cr_ruid)
+               if (uid == kauth_cred_getuid(vfs_context_ucred(context)))
                        break;
                /* fall through */
        default:
                        break;
                /* fall through */
        default:
@@ -1943,7 +2332,7 @@ hfs_quotactl(struct mount *mp, int cmds, uid_t uid, caddr_t datap, vfs_context_t
 /*
  * Get file system statistics.
  */
 /*
  * Get file system statistics.
  */
-static int
+int
 hfs_statfs(struct mount *mp, register struct vfsstatfs *sbp, __unused vfs_context_t context)
 {
        ExtendedVCB *vcb = VFSTOVCB(mp);
 hfs_statfs(struct mount *mp, register struct vfsstatfs *sbp, __unused vfs_context_t context)
 {
        ExtendedVCB *vcb = VFSTOVCB(mp);
@@ -2084,7 +2473,7 @@ hfs_sync_callback(struct vnode *vp, void *cargs)
  *
  * Note: we are always called with the filesystem marked `MPBUSY'.
  */
  *
  * Note: we are always called with the filesystem marked `MPBUSY'.
  */
-static int
+int
 hfs_sync(struct mount *mp, int waitfor, vfs_context_t context)
 {
        struct proc *p = vfs_context_proc(context);
 hfs_sync(struct mount *mp, int waitfor, vfs_context_t context)
 {
        struct proc *p = vfs_context_proc(context);
@@ -2188,7 +2577,7 @@ hfs_sync(struct mount *mp, int waitfor, vfs_context_t context)
        }
 
        if (hfsmp->jnl) {
        }
 
        if (hfsmp->jnl) {
-           hfs_journal_flush(hfsmp);
+           hfs_journal_flush(hfsmp, FALSE);
        }
 
        {
        }
 
        {
@@ -2229,7 +2618,7 @@ hfs_fhtovp(struct mount *mp, int fhlen, unsigned char *fhp, struct vnode **vpp,
        if (fhlen < (int)sizeof(struct hfsfid))
                return (EINVAL);
 
        if (fhlen < (int)sizeof(struct hfsfid))
                return (EINVAL);
 
-       result = hfs_vget(VFSTOHFS(mp), ntohl(hfsfhp->hfsfid_cnid), &nvp, 0);
+       result = hfs_vget(VFSTOHFS(mp), ntohl(hfsfhp->hfsfid_cnid), &nvp, 0, 0);
        if (result) {
                if (result == ENOENT)
                        result = ESTALE;
        if (result) {
                if (result == ENOENT)
                        result = ESTALE;
@@ -2304,6 +2693,7 @@ hfs_init(__unused struct vfsconf *vfsp)
        hfs_group_attr   = lck_grp_attr_alloc_init();
        hfs_mutex_group  = lck_grp_alloc_init("hfs-mutex", hfs_group_attr);
        hfs_rwlock_group = lck_grp_alloc_init("hfs-rwlock", hfs_group_attr);
        hfs_group_attr   = lck_grp_attr_alloc_init();
        hfs_mutex_group  = lck_grp_alloc_init("hfs-mutex", hfs_group_attr);
        hfs_rwlock_group = lck_grp_alloc_init("hfs-rwlock", hfs_group_attr);
+       hfs_spinlock_group = lck_grp_alloc_init("hfs-spinlock", hfs_group_attr);
        
 #if HFS_COMPRESSION
     decmpfs_init();
        
 #if HFS_COMPRESSION
     decmpfs_init();
@@ -2344,7 +2734,7 @@ hfs_getmountpoint(struct vnode *vp, struct hfsmount **hfsmpp)
 /*
  * HFS filesystem related variables.
  */
 /*
  * HFS filesystem related variables.
  */
-static int
+int
 hfs_sysctl(int *name, __unused u_int namelen, user_addr_t oldp, size_t *oldlenp, 
                        user_addr_t newp, size_t newlen, vfs_context_t context)
 {
 hfs_sysctl(int *name, __unused u_int namelen, user_addr_t oldp, size_t *oldlenp, 
                        user_addr_t newp, size_t newlen, vfs_context_t context)
 {
@@ -2490,6 +2880,15 @@ encodinghint_exit:
                                                         0,
                                                         hfs_sync_metadata, hfsmp->hfs_mp);
 
                                                         0,
                                                         hfs_sync_metadata, hfsmp->hfs_mp);
 
+               /*
+                * Set up the trim callback function so that we can add
+                * recently freed extents to the free extent cache once
+                * the transaction that freed them is written to the
+                * journal on disk.
+                */
+               if (jnl)
+                       journal_trim_set_callback(jnl, hfs_trim_callback, hfsmp);
+
                if (jnl == NULL) {
                        printf("hfs: FAILED to create the journal!\n");
                        if (jvp && jvp != hfsmp->hfs_devvp) {
                if (jnl == NULL) {
                        printf("hfs: FAILED to create the journal!\n");
                        if (jvp && jvp != hfsmp->hfs_devvp) {
@@ -2501,17 +2900,17 @@ encodinghint_exit:
                        return EINVAL;
                } 
 
                        return EINVAL;
                } 
 
-               hfs_global_exclusive_lock_acquire(hfsmp);
-               
+               hfs_lock_global (hfsmp, HFS_EXCLUSIVE_LOCK);
+
                /*
                 * Flush all dirty metadata buffers.
                 */
                /*
                 * Flush all dirty metadata buffers.
                 */
-               buf_flushdirtyblks(hfsmp->hfs_devvp, MNT_WAIT, 0, "hfs_sysctl");
-               buf_flushdirtyblks(hfsmp->hfs_extents_vp, MNT_WAIT, 0, "hfs_sysctl");
-               buf_flushdirtyblks(hfsmp->hfs_catalog_vp, MNT_WAIT, 0, "hfs_sysctl");
-               buf_flushdirtyblks(hfsmp->hfs_allocation_vp, MNT_WAIT, 0, "hfs_sysctl");
+               buf_flushdirtyblks(hfsmp->hfs_devvp, TRUE, 0, "hfs_sysctl");
+               buf_flushdirtyblks(hfsmp->hfs_extents_vp, TRUE, 0, "hfs_sysctl");
+               buf_flushdirtyblks(hfsmp->hfs_catalog_vp, TRUE, 0, "hfs_sysctl");
+               buf_flushdirtyblks(hfsmp->hfs_allocation_vp, TRUE, 0, "hfs_sysctl");
                if (hfsmp->hfs_attribute_vp)
                if (hfsmp->hfs_attribute_vp)
-                       buf_flushdirtyblks(hfsmp->hfs_attribute_vp, MNT_WAIT, 0, "hfs_sysctl");
+                       buf_flushdirtyblks(hfsmp->hfs_attribute_vp, TRUE, 0, "hfs_sysctl");
 
                HFSTOVCB(hfsmp)->vcbJinfoBlock = name[1];
                HFSTOVCB(hfsmp)->vcbAtrb |= kHFSVolumeJournaledMask;
 
                HFSTOVCB(hfsmp)->vcbJinfoBlock = name[1];
                HFSTOVCB(hfsmp)->vcbAtrb |= kHFSVolumeJournaledMask;
@@ -2526,7 +2925,7 @@ encodinghint_exit:
 
                vfs_setflags(hfsmp->hfs_mp, (u_int64_t)((unsigned int)MNT_JOURNALED));
 
 
                vfs_setflags(hfsmp->hfs_mp, (u_int64_t)((unsigned int)MNT_JOURNALED));
 
-               hfs_global_exclusive_lock_release(hfsmp);
+               hfs_unlock_global (hfsmp);
                hfs_flushvolumeheader(hfsmp, MNT_WAIT, 1);
 
                {
                hfs_flushvolumeheader(hfsmp, MNT_WAIT, 1);
 
                {
@@ -2561,7 +2960,7 @@ encodinghint_exit:
 
                printf("hfs: disabling journaling for mount @ %p\n", vnode_mount(vp));
 
 
                printf("hfs: disabling journaling for mount @ %p\n", vnode_mount(vp));
 
-               hfs_global_exclusive_lock_acquire(hfsmp);
+               hfs_lock_global (hfsmp, HFS_EXCLUSIVE_LOCK);
 
                // Lights out for you buddy!
                journal_close(hfsmp->jnl);
 
                // Lights out for you buddy!
                journal_close(hfsmp->jnl);
@@ -2580,7 +2979,8 @@ encodinghint_exit:
                
                HFSTOVCB(hfsmp)->vcbAtrb &= ~kHFSVolumeJournaledMask;
                
                
                HFSTOVCB(hfsmp)->vcbAtrb &= ~kHFSVolumeJournaledMask;
                
-               hfs_global_exclusive_lock_release(hfsmp);
+               hfs_unlock_global (hfsmp);
+
                hfs_flushvolumeheader(hfsmp, MNT_WAIT, 1);
 
                {
                hfs_flushvolumeheader(hfsmp, MNT_WAIT, 1);
 
                {
@@ -2661,6 +3061,10 @@ encodinghint_exit:
                file_drop(device_fd);
                vnode_put(devvp);
                return error;
                file_drop(device_fd);
                vnode_put(devvp);
                return error;
+       } else if (name[0] == HFS_ENABLE_RESIZE_DEBUG) {
+               hfs_resize_debug = 1;
+               printf ("hfs_sysctl: Enabled volume resize debugging.\n");
+               return 0;
        }
 
        return (ENOTSUP);
        }
 
        return (ENOTSUP);
@@ -2681,7 +3085,7 @@ hfs_vfs_vget(struct mount *mp, ino64_t ino, struct vnode **vpp, __unused vfs_con
 
        hfsmp = VFSTOHFS(mp);
 
 
        hfsmp = VFSTOHFS(mp);
 
-       error = hfs_vget(hfsmp, (cnid_t)ino, vpp, 1);
+       error = hfs_vget(hfsmp, (cnid_t)ino, vpp, 1, 0);
        if (error)
                return (error);
 
        if (error)
                return (error);
 
@@ -2722,9 +3126,8 @@ hfs_vfs_vget(struct mount *mp, ino64_t ino, struct vnode **vpp, __unused vfs_con
  *
  * If the object is a file then it will represent the data fork.
  */
  *
  * If the object is a file then it will represent the data fork.
  */
-__private_extern__
 int
 int
-hfs_vget(struct hfsmount *hfsmp, cnid_t cnid, struct vnode **vpp, int skiplock)
+hfs_vget(struct hfsmount *hfsmp, cnid_t cnid, struct vnode **vpp, int skiplock, int allow_deleted)
 {
        struct vnode *vp = NULLVP;
        struct cat_desc cndesc;
 {
        struct vnode *vp = NULLVP;
        struct cat_desc cndesc;
@@ -2746,7 +3149,7 @@ hfs_vget(struct hfsmount *hfsmp, cnid_t cnid, struct vnode **vpp, int skiplock)
        /*
         * Check the hash first
         */
        /*
         * Check the hash first
         */
-       vp = hfs_chash_getvnode(hfsmp, cnid, 0, skiplock);
+       vp = hfs_chash_getvnode(hfsmp, cnid, 0, skiplock, allow_deleted);
        if (vp) {
                *vpp = vp;
                return(0);
        if (vp) {
                *vpp = vp;
                return(0);
@@ -2826,7 +3229,7 @@ hfs_vget(struct hfsmount *hfsmp, cnid_t cnid, struct vnode **vpp, int skiplock)
                 * Pick up the first link in the chain and get a descriptor for it.
                 * This allows blind volfs paths to work for hardlinks.
                 */
                 * Pick up the first link in the chain and get a descriptor for it.
                 * This allows blind volfs paths to work for hardlinks.
                 */
-               if ((hfs_lookuplink(hfsmp, linkref, &prevlinkid,  &nextlinkid) == 0) &&
+               if ((hfs_lookup_siblinglinks(hfsmp, linkref, &prevlinkid,  &nextlinkid) == 0) &&
                    (nextlinkid != 0)) {
                        lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_SHARED_LOCK);
                        error = cat_findname(hfsmp, nextlinkid, &linkdesc);
                    (nextlinkid != 0)) {
                        lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_SHARED_LOCK);
                        error = cat_findname(hfsmp, nextlinkid, &linkdesc);
@@ -2839,13 +3242,17 @@ hfs_vget(struct hfsmount *hfsmp, cnid_t cnid, struct vnode **vpp, int skiplock)
        }
 
        if (linkref) {
        }
 
        if (linkref) {
-               error = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &cnfork, &vp);
+               int newvnode_flags = 0;
+               
+               error = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr,
+                                                               &cnfork, &vp, &newvnode_flags);
                if (error == 0) {
                        VTOC(vp)->c_flag |= C_HARDLINK;
                        vnode_setmultipath(vp);
                }
        } else {
                struct componentname cn;
                if (error == 0) {
                        VTOC(vp)->c_flag |= C_HARDLINK;
                        vnode_setmultipath(vp);
                }
        } else {
                struct componentname cn;
+               int newvnode_flags = 0;
 
                /* Supply hfs_getnewvnode with a component name. */
                MALLOC_ZONE(cn.cn_pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
 
                /* Supply hfs_getnewvnode with a component name. */
                MALLOC_ZONE(cn.cn_pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
@@ -2859,7 +3266,8 @@ hfs_vget(struct hfsmount *hfsmp, cnid_t cnid, struct vnode **vpp, int skiplock)
                cn.cn_consume = 0;
                bcopy(cndesc.cd_nameptr, cn.cn_nameptr, cndesc.cd_namelen + 1);
        
                cn.cn_consume = 0;
                bcopy(cndesc.cd_nameptr, cn.cn_nameptr, cndesc.cd_namelen + 1);
        
-               error = hfs_getnewvnode(hfsmp, NULLVP, &cn, &cndesc, 0, &cnattr, &cnfork, &vp);
+               error = hfs_getnewvnode(hfsmp, NULLVP, &cn, &cndesc, 0, &cnattr, 
+                                                               &cnfork, &vp, &newvnode_flags);
 
                if (error == 0 && (VTOC(vp)->c_flag & C_HARDLINK)) {
                        hfs_savelinkorigin(VTOC(vp), cndesc.cd_parentcnid);
 
                if (error == 0 && (VTOC(vp)->c_flag & C_HARDLINK)) {
                        hfs_savelinkorigin(VTOC(vp), cndesc.cd_parentcnid);
@@ -2912,7 +3320,7 @@ hfs_flushfiles(struct mount *mp, int flags, __unused struct proc *p)
                }
 
                /* Obtain the root vnode so we can skip over it. */
                }
 
                /* Obtain the root vnode so we can skip over it. */
-               skipvp = hfs_chash_getvnode(hfsmp, kHFSRootFolderID, 0, 0);
+               skipvp = hfs_chash_getvnode(hfsmp, kHFSRootFolderID, 0, 0, 0);
        }
 #endif /* QUOTA */
 
        }
 #endif /* QUOTA */
 
@@ -2989,7 +3397,6 @@ hfs_setencodingbits(struct hfsmount *hfsmp, u_int32_t encoding)
  *
  * On journal volumes this will cause a volume header flush
  */
  *
  * On journal volumes this will cause a volume header flush
  */
-__private_extern__
 int
 hfs_volupdate(struct hfsmount *hfsmp, enum volop op, int inroot)
 {
 int
 hfs_volupdate(struct hfsmount *hfsmp, enum volop op, int inroot)
 {
@@ -3064,7 +3471,7 @@ hfs_flushMDB(struct hfsmount *hfsmp, int waitfor, int altflush)
 
        mdb = (HFSMasterDirectoryBlock *)(buf_dataptr(bp) + HFS_PRI_OFFSET(sectorsize));
     
 
        mdb = (HFSMasterDirectoryBlock *)(buf_dataptr(bp) + HFS_PRI_OFFSET(sectorsize));
     
-       mdb->drCrDate   = SWAP_BE32 (UTCToLocal(to_hfs_time(vcb->vcbCrDate)));
+       mdb->drCrDate   = SWAP_BE32 (UTCToLocal(to_hfs_time(vcb->hfs_itime)));
        mdb->drLsMod    = SWAP_BE32 (UTCToLocal(to_hfs_time(vcb->vcbLsMod)));
        mdb->drAtrb     = SWAP_BE16 (vcb->vcbAtrb);
        mdb->drNmFls    = SWAP_BE16 (vcb->vcbNmFls);
        mdb->drLsMod    = SWAP_BE32 (UTCToLocal(to_hfs_time(vcb->vcbLsMod)));
        mdb->drAtrb     = SWAP_BE16 (vcb->vcbAtrb);
        mdb->drNmFls    = SWAP_BE16 (vcb->vcbNmFls);
@@ -3141,7 +3548,6 @@ hfs_flushMDB(struct hfsmount *hfsmp, int waitfor, int altflush)
  *  not flushed since the on-disk "H+" and "HX" signatures
  *  are always stored in-memory as "H+".
  */
  *  not flushed since the on-disk "H+" and "HX" signatures
  *  are always stored in-memory as "H+".
  */
-__private_extern__
 int
 hfs_flushvolumeheader(struct hfsmount *hfsmp, int waitfor, int altflush)
 {
 int
 hfs_flushvolumeheader(struct hfsmount *hfsmp, int waitfor, int altflush)
 {
@@ -3449,7 +3855,6 @@ err_exit:
 /*
  * Extend a file system.
  */
 /*
  * Extend a file system.
  */
-__private_extern__
 int
 hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
 {
 int
 hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
 {
@@ -3470,10 +3875,11 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
        u_int32_t  phys_sectorsize;
        daddr64_t  prev_alt_sector;
        daddr_t    bitmapblks;
        u_int32_t  phys_sectorsize;
        daddr64_t  prev_alt_sector;
        daddr_t    bitmapblks;
-       int  lockflags;
+       int  lockflags = 0;
        int  error;
        int64_t oldBitmapSize;
        Boolean  usedExtendFileC = false;
        int  error;
        int64_t oldBitmapSize;
        Boolean  usedExtendFileC = false;
+       int transaction_begun = 0;
        
        devvp = hfsmp->hfs_devvp;
        vcb = HFSTOVCB(hfsmp);
        
        devvp = hfsmp->hfs_devvp;
        vcb = HFSTOVCB(hfsmp);
@@ -3493,7 +3899,7 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
         * ownership and check permissions.
         */
        if (suser(cred, NULL)) {
         * ownership and check permissions.
         */
        if (suser(cred, NULL)) {
-               error = hfs_vget(hfsmp, kHFSRootFolderID, &vp, 0);
+               error = hfs_vget(hfsmp, kHFSRootFolderID, &vp, 0, 0);
 
                if (error)
                        return (error);
 
                if (error)
                        return (error);
@@ -3546,13 +3952,32 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
 
        addblks = newblkcnt - vcb->totalBlocks;
 
 
        addblks = newblkcnt - vcb->totalBlocks;
 
-       printf("hfs_extendfs: growing %s by %d blocks\n", vcb->vcbVN, addblks);
+       if (hfs_resize_debug) {
+               printf ("hfs_extendfs: old: size=%qu, blkcnt=%u\n", oldsize, hfsmp->totalBlocks);
+               printf ("hfs_extendfs: new: size=%qu, blkcnt=%u, addblks=%u\n", newsize, (u_int32_t)newblkcnt, addblks);
+       }
+       printf("hfs_extendfs: will extend \"%s\" by %d blocks\n", vcb->vcbVN, addblks);
+
+       HFS_MOUNT_LOCK(hfsmp, TRUE);
+       if (hfsmp->hfs_flags & HFS_RESIZE_IN_PROGRESS) {
+               HFS_MOUNT_UNLOCK(hfsmp, TRUE);
+               error = EALREADY;
+               goto out;
+       }
+       hfsmp->hfs_flags |= HFS_RESIZE_IN_PROGRESS;
+       HFS_MOUNT_UNLOCK(hfsmp, TRUE);
+
+       /* Start with a clean journal. */
+       hfs_journal_flush(hfsmp, TRUE);
+
        /*
         * Enclose changes inside a transaction.
         */
        if (hfs_start_transaction(hfsmp) != 0) {
        /*
         * Enclose changes inside a transaction.
         */
        if (hfs_start_transaction(hfsmp) != 0) {
-               return (EINVAL);
+               error = EINVAL;
+               goto out;
        }
        }
+       transaction_begun = 1;
 
        /*
         * Note: we take the attributes lock in case we have an attribute data vnode
 
        /*
         * Note: we take the attributes lock in case we have an attribute data vnode
@@ -3573,6 +3998,17 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
        else
                bitmapblks = 0;
 
        else
                bitmapblks = 0;
 
+       /* 
+        * The allocation bitmap can contain unused bits that are beyond end of 
+        * current volume's allocation blocks.  Usually they are supposed to be 
+        * zero'ed out but there can be cases where they might be marked as used. 
+        * After extending the file system, those bits can represent valid 
+        * allocation blocks, so we mark all the bits from the end of current 
+        * volume to end of allocation bitmap as "free".
+        */
+       BlockMarkFreeUnused(vcb, vcb->totalBlocks, 
+                       (fp->ff_blocks * vcb->blockSize * 8) - vcb->totalBlocks);
+
        if (bitmapblks > 0) {
                daddr64_t blkno;
                daddr_t blkcnt;
        if (bitmapblks > 0) {
                daddr64_t blkno;
                daddr_t blkcnt;
@@ -3592,8 +4028,8 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                 * zone.
                 */
                error = ExtendFileC(vcb, fp, bitmapblks * vcb->blockSize, 0,
                 * zone.
                 */
                error = ExtendFileC(vcb, fp, bitmapblks * vcb->blockSize, 0,
-                               kEFAllMask | kEFNoClumpMask | kEFReserveMask | kEFMetadataMask,
-                               &bytesAdded);
+                               kEFAllMask | kEFNoClumpMask | kEFReserveMask 
+                               | kEFMetadataMask | kEFContigMask, &bytesAdded);
 
                if (error == 0) {
                        usedExtendFileC = true;
 
                if (error == 0) {
                        usedExtendFileC = true;
@@ -3705,7 +4141,8 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                 * Restore to old state.
                 */
                if (usedExtendFileC) {
                 * Restore to old state.
                 */
                if (usedExtendFileC) {
-                       (void) TruncateFileC(vcb, fp, oldBitmapSize, false);
+                       (void) TruncateFileC(vcb, fp, oldBitmapSize, 0, FORK_IS_RSRC(fp), 
+                                                                FTOC(fp)->c_fileid, false);
                } else {
                        fp->ff_blocks -= bitmapblks;
                        fp->ff_size -= (u_int64_t)bitmapblks * (u_int64_t)vcb->blockSize;
                } else {
                        fp->ff_blocks -= bitmapblks;
                        fp->ff_size -= (u_int64_t)bitmapblks * (u_int64_t)vcb->blockSize;
@@ -3721,10 +4158,15 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                hfsmp->hfs_logical_block_count = prev_phys_block_count;
                hfsmp->hfs_alt_id_sector = prev_alt_sector;
                MarkVCBDirty(vcb);
                hfsmp->hfs_logical_block_count = prev_phys_block_count;
                hfsmp->hfs_alt_id_sector = prev_alt_sector;
                MarkVCBDirty(vcb);
-               if (vcb->blockSize == 512)
-                       (void) BlockMarkAllocated(vcb, vcb->totalBlocks - 2, 2);
-               else
-                       (void) BlockMarkAllocated(vcb, vcb->totalBlocks - 1, 1);
+               if (vcb->blockSize == 512) {
+                       if (BlockMarkAllocated(vcb, vcb->totalBlocks - 2, 2)) {
+                               hfs_mark_volume_inconsistent(hfsmp);
+                       }
+               } else {
+                       if (BlockMarkAllocated(vcb, vcb->totalBlocks - 1, 1)) {
+                               hfs_mark_volume_inconsistent(hfsmp);
+                       }
+               }
                goto out;
        }
        /*
                goto out;
        }
        /*
@@ -3745,9 +4187,10 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                }
        }
        
                }
        }
        
-       /*
-        * TODO: Adjust the size of the metadata zone based on new volume size?
+       /* 
+        * Update the metadata zone size based on current volume size
         */
         */
+       hfs_metadatazone_init(hfsmp, false);
         
        /*
         * Adjust the size of hfsmp->hfs_attrdata_vp
         
        /*
         * Adjust the size of hfsmp->hfs_attrdata_vp
@@ -3769,23 +4212,48 @@ hfs_extendfs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                }
        }
 
                }
        }
 
+       /*
+        * Update the R/B Tree if necessary.  Since we don't have to drop the systemfile 
+        * locks in the middle of these operations like we do in the truncate case
+        * where we have to relocate files, we can only update the red-black tree
+        * if there were actual changes made to the bitmap.  Also, we can't really scan the 
+        * new portion of the bitmap before it has been allocated. The BlockMarkAllocated
+        * routines are smart enough to avoid the r/b tree if the portion they are manipulating is
+        * not currently controlled by the tree.  
+        *
+        * We only update hfsmp->allocLimit if totalBlocks actually increased. 
+        */
+       
+       if (error == 0) {
+               UpdateAllocLimit(hfsmp, hfsmp->totalBlocks);
+       }
+       
+       /* Log successful extending */
+       printf("hfs_extendfs: extended \"%s\" to %d blocks (was %d blocks)\n",
+              hfsmp->vcbVN, hfsmp->totalBlocks, (u_int32_t)(oldsize/hfsmp->blockSize));
+       
 out:
        if (error && fp) {
                /* Restore allocation fork. */
                bcopy(&forkdata, &fp->ff_data, sizeof(forkdata));
                VTOC(vp)->c_blocks = fp->ff_blocks;
 out:
        if (error && fp) {
                /* Restore allocation fork. */
                bcopy(&forkdata, &fp->ff_data, sizeof(forkdata));
                VTOC(vp)->c_blocks = fp->ff_blocks;
-
+               
        }
        }
-       /*
-          Regardless of whether or not the totalblocks actually increased,
-          we should reset the allocLimit field. If it changed, it will
-          get updated; if not, it will remain the same.
-       */
-       hfsmp->allocLimit = vcb->totalBlocks;
-       hfs_systemfile_unlock(hfsmp, lockflags);
-       hfs_end_transaction(hfsmp);
-
-       return (error);
+       
+       HFS_MOUNT_LOCK(hfsmp, TRUE);    
+       hfsmp->hfs_flags &= ~HFS_RESIZE_IN_PROGRESS;
+       HFS_MOUNT_UNLOCK(hfsmp, TRUE);  
+       if (lockflags) {
+               hfs_systemfile_unlock(hfsmp, lockflags);
+       }
+       if (transaction_begun) {
+               hfs_end_transaction(hfsmp);
+               hfs_journal_flush(hfsmp, FALSE);
+               /* Just to be sure, sync all data to the disk */
+               (void) VNOP_IOCTL(hfsmp->hfs_devvp, DKIOCSYNCHRONIZECACHE, NULL, FWRITE, context);
+       }
+
+       return MacToVFSError(error);
 }
 
 #define HFS_MIN_SIZE  (32LL * 1024LL * 1024LL)
 }
 
 #define HFS_MIN_SIZE  (32LL * 1024LL * 1024LL)
@@ -3793,7 +4261,6 @@ out:
 /*
  * Truncate a file system (while still mounted).
  */
 /*
  * Truncate a file system (while still mounted).
  */
-__private_extern__
 int
 hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
 {
 int
 hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
 {
@@ -3803,7 +4270,9 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
        u_int32_t reclaimblks = 0;
        int lockflags = 0;
        int transaction_begun = 0;
        u_int32_t reclaimblks = 0;
        int lockflags = 0;
        int transaction_begun = 0;
-       int error;
+       Boolean updateFreeBlocks = false;
+       Boolean disable_sparse = false;
+       int error = 0;
 
        lck_mtx_lock(&hfsmp->hfs_mutex);
        if (hfsmp->hfs_flags & HFS_RESIZE_IN_PROGRESS) {
 
        lck_mtx_lock(&hfsmp->hfs_mutex);
        if (hfsmp->hfs_flags & HFS_RESIZE_IN_PROGRESS) {
@@ -3811,8 +4280,9 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                return (EALREADY);
        }
        hfsmp->hfs_flags |= HFS_RESIZE_IN_PROGRESS;
                return (EALREADY);
        }
        hfsmp->hfs_flags |= HFS_RESIZE_IN_PROGRESS;
-       hfsmp->hfs_resize_filesmoved = 0;
-       hfsmp->hfs_resize_totalfiles = 0;
+       hfsmp->hfs_resize_blocksmoved = 0;
+       hfsmp->hfs_resize_totalblocks = 0;
+       hfsmp->hfs_resize_progress = 0;
        lck_mtx_unlock(&hfsmp->hfs_mutex);
 
        /*
        lck_mtx_unlock(&hfsmp->hfs_mutex);
 
        /*
@@ -3828,31 +4298,80 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
        newblkcnt = newsize / hfsmp->blockSize;
        reclaimblks = hfsmp->totalBlocks - newblkcnt;
 
        newblkcnt = newsize / hfsmp->blockSize;
        reclaimblks = hfsmp->totalBlocks - newblkcnt;
 
+       if (hfs_resize_debug) {
+               printf ("hfs_truncatefs: old: size=%qu, blkcnt=%u, freeblks=%u\n", oldsize, hfsmp->totalBlocks, hfs_freeblks(hfsmp, 1));
+               printf ("hfs_truncatefs: new: size=%qu, blkcnt=%u, reclaimblks=%u\n", newsize, newblkcnt, reclaimblks);
+       }
+
        /* Make sure new size is valid. */
        if ((newsize < HFS_MIN_SIZE) ||
            (newsize >= oldsize) ||
            (newsize % hfsmp->hfs_logical_block_size) ||
            (newsize % hfsmp->hfs_physical_block_size)) {
        /* Make sure new size is valid. */
        if ((newsize < HFS_MIN_SIZE) ||
            (newsize >= oldsize) ||
            (newsize % hfsmp->hfs_logical_block_size) ||
            (newsize % hfsmp->hfs_physical_block_size)) {
-               printf ("hfs_truncatefs: invalid size\n");
+               printf ("hfs_truncatefs: invalid size (newsize=%qu, oldsize=%qu)\n", newsize, oldsize);
                error = EINVAL;
                goto out;
        }
                error = EINVAL;
                goto out;
        }
-       /* Make sure there's enough space to work with. */
+
+       /* 
+        * Make sure that the file system has enough free blocks reclaim.
+        *
+        * Before resize, the disk is divided into four zones - 
+        *      A. Allocated_Stationary - These are allocated blocks that exist 
+        *         before the new end of disk.  These blocks will not be 
+        *         relocated or modified during resize.
+        *      B. Free_Stationary - These are free blocks that exist before the
+        *         new end of disk.  These blocks can be used for any new 
+        *         allocations during resize, including allocation for relocating 
+        *         data from the area of disk being reclaimed. 
+        *      C. Allocated_To-Reclaim - These are allocated blocks that exist
+        *         beyond the new end of disk.  These blocks need to be reclaimed 
+        *         during resize by allocating equal number of blocks in Free 
+        *         Stationary zone and copying the data. 
+        *      D. Free_To-Reclaim - These are free blocks that exist beyond the 
+        *         new end of disk.  Nothing special needs to be done to reclaim
+        *         them. 
+        *
+        * Total number of blocks on the disk before resize:
+        * ------------------------------------------------
+        *      Total Blocks = Allocated_Stationary + Free_Stationary + 
+        *                     Allocated_To-Reclaim + Free_To-Reclaim
+        *
+        * Total number of blocks that need to be reclaimed:
+        * ------------------------------------------------
+        *      Blocks to Reclaim = Allocated_To-Reclaim + Free_To-Reclaim 
+        *
+        * Note that the check below also makes sure that we have enough space 
+        * to relocate data from Allocated_To-Reclaim to Free_Stationary.   
+        * Therefore we do not need to check total number of blocks to relocate 
+        * later in the code.
+        *
+        * The condition below gets converted to: 
+        *
+        * Allocated To-Reclaim + Free To-Reclaim >= Free Stationary + Free To-Reclaim 
+        *
+        * which is equivalent to:
+        *
+        *              Allocated To-Reclaim >= Free Stationary
+        */
        if (reclaimblks >= hfs_freeblks(hfsmp, 1)) {
        if (reclaimblks >= hfs_freeblks(hfsmp, 1)) {
-               printf("hfs_truncatefs: insufficient space (need %u blocks; have %u blocks)\n", reclaimblks, hfs_freeblks(hfsmp, 1));
+               printf("hfs_truncatefs: insufficient space (need %u blocks; have %u free blocks)\n", reclaimblks, hfs_freeblks(hfsmp, 1));
                error = ENOSPC;
                goto out;
        }
        
        /* Start with a clean journal. */
                error = ENOSPC;
                goto out;
        }
        
        /* Start with a clean journal. */
-       hfs_journal_flush(hfsmp);
+       hfs_journal_flush(hfsmp, TRUE);
        
        if (hfs_start_transaction(hfsmp) != 0) {
                error = EINVAL;
                goto out;
        }
        transaction_begun = 1;
        
        if (hfs_start_transaction(hfsmp) != 0) {
                error = EINVAL;
                goto out;
        }
        transaction_begun = 1;
-
+       
+       /* Take the bitmap lock to update the alloc limit field */
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
+       
        /*
         * Prevent new allocations from using the part we're trying to truncate.
         *
        /*
         * Prevent new allocations from using the part we're trying to truncate.
         *
@@ -3861,18 +4380,71 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
         * interfere with allocating the new alternate volume header, and no files
         * in the allocation blocks beyond (i.e. the blocks we're trying to
         * truncate away.
         * interfere with allocating the new alternate volume header, and no files
         * in the allocation blocks beyond (i.e. the blocks we're trying to
         * truncate away.
+        *
+        * Also shrink the red-black tree if needed.
+        */
+       if (hfsmp->blockSize == 512) {
+               error = UpdateAllocLimit (hfsmp, newblkcnt - 2);
+       }
+       else {
+               error = UpdateAllocLimit (hfsmp, newblkcnt - 1);
+       }
+
+       /* Sparse devices use first fit allocation which is not ideal 
+        * for volume resize which requires best fit allocation.  If a 
+        * sparse device is being truncated, disable the sparse device 
+        * property temporarily for the duration of resize.  Also reset 
+        * the free extent cache so that it is rebuilt as sorted by 
+        * totalBlocks instead of startBlock.  
+        *
+        * Note that this will affect all allocations on the volume and 
+        * ideal fix would be just to modify resize-related allocations, 
+        * but it will result in complexity like handling of two free 
+        * extent caches sorted differently, etc.  So we stick to this 
+        * solution for now. 
+        */
+       HFS_MOUNT_LOCK(hfsmp, TRUE);    
+       if (hfsmp->hfs_flags & HFS_HAS_SPARSE_DEVICE) {
+               hfsmp->hfs_flags &= ~HFS_HAS_SPARSE_DEVICE;
+               ResetVCBFreeExtCache(hfsmp);
+               disable_sparse = true;
+       }
+       
+       /* 
+        * Update the volume free block count to reflect the total number 
+        * of free blocks that will exist after a successful resize.
+        * Relocation of extents will result in no net change in the total
+        * free space on the disk.  Therefore the code that allocates 
+        * space for new extent and deallocates the old extent explicitly 
+        * prevents updating the volume free block count.  It will also 
+        * prevent false disk full error when the number of blocks in 
+        * an extent being relocated is more than the free blocks that 
+        * will exist after the volume is resized.
         */
         */
-       lck_mtx_lock(&hfsmp->hfs_mutex);
-       if (hfsmp->blockSize == 512) 
-               hfsmp->allocLimit = newblkcnt - 2;
-       else
-               hfsmp->allocLimit = newblkcnt - 1;
        hfsmp->freeBlocks -= reclaimblks;
        hfsmp->freeBlocks -= reclaimblks;
-       lck_mtx_unlock(&hfsmp->hfs_mutex);
+       updateFreeBlocks = true;
+       HFS_MOUNT_UNLOCK(hfsmp, TRUE);
+       
+       if (lockflags) {
+               hfs_systemfile_unlock(hfsmp, lockflags);
+               lockflags = 0;  
+       }
        
        /*
        
        /*
-        * Look for files that have blocks at or beyond the location of the
-        * new alternate volume header.
+        * Update the metadata zone size to match the new volume size,
+        * and if it too less, metadata zone might be disabled.
+        */
+       hfs_metadatazone_init(hfsmp, false);
+
+       /*
+        * If some files have blocks at or beyond the location of the
+        * new alternate volume header, recalculate free blocks and 
+        * reclaim blocks.  Otherwise just update free blocks count.
+        *
+        * The current allocLimit is set to the location of new alternate 
+        * volume header, and reclaimblks are the total number of blocks 
+        * that need to be reclaimed.  So the check below is really 
+        * ignoring the blocks allocated for old alternate volume header. 
         */
        if (hfs_isallocated(hfsmp, hfsmp->allocLimit, reclaimblks)) {
                /*
         */
        if (hfs_isallocated(hfsmp, hfsmp->allocLimit, reclaimblks)) {
                /*
@@ -3883,8 +4455,9 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                transaction_begun = 0;
 
                /* Attempt to reclaim some space. */ 
                transaction_begun = 0;
 
                /* Attempt to reclaim some space. */ 
-               if (hfs_reclaimspace(hfsmp, hfsmp->allocLimit, reclaimblks, context) != 0) {
-                       printf("hfs_truncatefs: couldn't reclaim space on %s\n", hfsmp->vcbVN);
+               error = hfs_reclaimspace(hfsmp, hfsmp->allocLimit, reclaimblks, context);
+               if (error != 0) {
+                       printf("hfs_truncatefs: couldn't reclaim space on %s (error=%d)\n", hfsmp->vcbVN, error);
                        error = ENOSPC;
                        goto out;
                }
                        error = ENOSPC;
                        goto out;
                }
@@ -3895,28 +4468,20 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                transaction_begun = 1;
                
                /* Check if we're clear now. */
                transaction_begun = 1;
                
                /* Check if we're clear now. */
-               if (hfs_isallocated(hfsmp, hfsmp->allocLimit, reclaimblks)) {
-                       printf("hfs_truncatefs: didn't reclaim enough space on %s\n", hfsmp->vcbVN);
+               error = hfs_isallocated(hfsmp, hfsmp->allocLimit, reclaimblks);
+               if (error != 0) {
+                       printf("hfs_truncatefs: didn't reclaim enough space on %s (error=%d)\n", hfsmp->vcbVN, error);
                        error = EAGAIN;  /* tell client to try again */
                        goto out;
                }
                        error = EAGAIN;  /* tell client to try again */
                        goto out;
                }
-       }
-       
+       } 
+               
        /*
         * Note: we take the attributes lock in case we have an attribute data vnode
         * which needs to change size.
         */
        lockflags = hfs_systemfile_lock(hfsmp, SFL_ATTRIBUTE | SFL_EXTENTS | SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
 
        /*
         * Note: we take the attributes lock in case we have an attribute data vnode
         * which needs to change size.
         */
        lockflags = hfs_systemfile_lock(hfsmp, SFL_ATTRIBUTE | SFL_EXTENTS | SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
 
-       /*
-        * Mark the old alternate volume header as free. 
-        * We don't bother shrinking allocation bitmap file.
-        */
-       if (hfsmp->blockSize == 512) 
-               (void) BlockMarkFree(hfsmp, hfsmp->totalBlocks - 2, 2);
-       else 
-               (void) BlockMarkFree(hfsmp, hfsmp->totalBlocks - 1, 1);
-
        /*
         * Allocate last 1KB for alternate volume header.
         */
        /*
         * Allocate last 1KB for alternate volume header.
         */
@@ -3926,6 +4491,15 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
                goto out;
        }
 
                goto out;
        }
 
+       /*
+        * Mark the old alternate volume header as free. 
+        * We don't bother shrinking allocation bitmap file.
+        */
+       if (hfsmp->blockSize == 512) 
+               (void) BlockMarkFree(hfsmp, hfsmp->totalBlocks - 2, 2);
+       else 
+               (void) BlockMarkFree(hfsmp, hfsmp->totalBlocks - 1, 1);
+
        /*
         * Invalidate the existing alternate volume header.
         *
        /*
         * Invalidate the existing alternate volume header.
         *
@@ -3933,14 +4507,16 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
         * since this block will be outside of the truncated file system!
         */
        if (hfsmp->hfs_alt_id_sector) {
         * since this block will be outside of the truncated file system!
         */
        if (hfsmp->hfs_alt_id_sector) {
-               if (buf_meta_bread(hfsmp->hfs_devvp, 
+               error = buf_meta_bread(hfsmp->hfs_devvp, 
                                HFS_PHYSBLK_ROUNDDOWN(hfsmp->hfs_alt_id_sector, hfsmp->hfs_log_per_phys),
                                HFS_PHYSBLK_ROUNDDOWN(hfsmp->hfs_alt_id_sector, hfsmp->hfs_log_per_phys),
-                               hfsmp->hfs_physical_block_size, NOCRED, &bp) == 0) {
-       
+                               hfsmp->hfs_physical_block_size, NOCRED, &bp);
+               if (error == 0) {
                        bzero((void*)((char *)buf_dataptr(bp) + HFS_ALT_OFFSET(hfsmp->hfs_physical_block_size)), kMDBSize);
                        (void) VNOP_BWRITE(bp);
                        bzero((void*)((char *)buf_dataptr(bp) + HFS_ALT_OFFSET(hfsmp->hfs_physical_block_size)), kMDBSize);
                        (void) VNOP_BWRITE(bp);
-               } else if (bp) {
-                       buf_brelse(bp);
+               } else {
+                       if (bp) {
+                               buf_brelse(bp);
+                       }
                }
                bp = NULL;
        }
                }
                bp = NULL;
        }
@@ -3959,11 +4535,7 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
        error = hfs_flushvolumeheader(hfsmp, MNT_WAIT, HFS_ALTFLUSH);
        if (error)
                panic("hfs_truncatefs: unexpected error flushing volume header (%d)\n", error);
        error = hfs_flushvolumeheader(hfsmp, MNT_WAIT, HFS_ALTFLUSH);
        if (error)
                panic("hfs_truncatefs: unexpected error flushing volume header (%d)\n", error);
-       
-       /*
-        * TODO: Adjust the size of the metadata zone based on new volume size?
-        */
-        
+
        /*
         * Adjust the size of hfsmp->hfs_attrdata_vp
         */
        /*
         * Adjust the size of hfsmp->hfs_attrdata_vp
         */
@@ -3985,25 +4557,49 @@ hfs_truncatefs(struct hfsmount *hfsmp, u_int64_t newsize, vfs_context_t context)
        }
        
 out:
        }
        
 out:
-       if (error)
+       /* 
+        * Update the allocLimit to acknowledge the last one or two blocks now.
+        * Add it to the tree as well if necessary.
+        */
+       UpdateAllocLimit (hfsmp, hfsmp->totalBlocks);
+       
+       HFS_MOUNT_LOCK(hfsmp, TRUE);    
+       if (disable_sparse == true) {
+               /* Now that resize is completed, set the volume to be sparse 
+                * device again so that all further allocations will be first 
+                * fit instead of best fit.  Reset free extent cache so that 
+                * it is rebuilt.
+                */
+               hfsmp->hfs_flags |= HFS_HAS_SPARSE_DEVICE;
+               ResetVCBFreeExtCache(hfsmp);
+       }
+
+       if (error && (updateFreeBlocks == true)) {
                hfsmp->freeBlocks += reclaimblks;
                hfsmp->freeBlocks += reclaimblks;
+       }
        
        
-       lck_mtx_lock(&hfsmp->hfs_mutex);
-       hfsmp->allocLimit = hfsmp->totalBlocks;
-       if (hfsmp->nextAllocation >= hfsmp->allocLimit)
+       if (hfsmp->nextAllocation >= hfsmp->allocLimit) {
                hfsmp->nextAllocation = hfsmp->hfs_metazone_end + 1;
                hfsmp->nextAllocation = hfsmp->hfs_metazone_end + 1;
+       }
        hfsmp->hfs_flags &= ~HFS_RESIZE_IN_PROGRESS;
        hfsmp->hfs_flags &= ~HFS_RESIZE_IN_PROGRESS;
-       lck_mtx_unlock(&hfsmp->hfs_mutex);
+       HFS_MOUNT_UNLOCK(hfsmp, TRUE);
+       
+       /* On error, reset the metadata zone for original volume size */
+       if (error && (updateFreeBlocks == true)) {
+               hfs_metadatazone_init(hfsmp, false);
+       }
        
        if (lockflags) {
                hfs_systemfile_unlock(hfsmp, lockflags);
        }
        if (transaction_begun) {
                hfs_end_transaction(hfsmp);
        
        if (lockflags) {
                hfs_systemfile_unlock(hfsmp, lockflags);
        }
        if (transaction_begun) {
                hfs_end_transaction(hfsmp);
-               hfs_journal_flush(hfsmp);
+               hfs_journal_flush(hfsmp, FALSE);
+               /* Just to be sure, sync all data to the disk */
+               (void) VNOP_IOCTL(hfsmp->hfs_devvp, DKIOCSYNCHRONIZECACHE, NULL, FWRITE, context);
        }
 
        }
 
-       return (error);
+       return MacToVFSError(error);
 }
 
 
 }
 
 
@@ -4065,6 +4661,9 @@ hfs_copy_extent(
        u_int32_t ioSizeSectors;        /* Device sectors in this I/O */
        daddr64_t srcSector, destSector;
        u_int32_t sectorsPerBlock = hfsmp->blockSize / hfsmp->hfs_logical_block_size;
        u_int32_t ioSizeSectors;        /* Device sectors in this I/O */
        daddr64_t srcSector, destSector;
        u_int32_t sectorsPerBlock = hfsmp->blockSize / hfsmp->hfs_logical_block_size;
+#if CONFIG_PROTECT
+       int cpenabled = 0;
+#endif
 
        /*
         * Sanity check that we have locked the vnode of the file we're copying.
 
        /*
         * Sanity check that we have locked the vnode of the file we're copying.
@@ -4077,18 +4676,25 @@ hfs_copy_extent(
        if (cp != hfsmp->hfs_allocation_cp && cp->c_lockowner != current_thread())
                panic("hfs_copy_extent: vp=%p (cp=%p) not owned?\n", vp, cp);
 
        if (cp != hfsmp->hfs_allocation_cp && cp->c_lockowner != current_thread())
                panic("hfs_copy_extent: vp=%p (cp=%p) not owned?\n", vp, cp);
 
-       /*
-        * Wait for any in-progress writes to this vnode to complete, so that we'll
-        * be copying consistent bits.  (Otherwise, it's possible that an async
-        * write will complete to the old extent after we read from it.  That
-        * could lead to corruption.)
-        */
-       err = vnode_waitforwrites(vp, 0, 0, 0, "hfs_copy_extent");
-       if (err) {
-               printf("hfs_copy_extent: Error %d from vnode_waitforwrites\n", err);
-               return err;
+#if CONFIG_PROTECT
+       /* Prepare the CP blob and get it ready for use */
+       if (!vnode_issystem (vp) && vnode_isreg(vp) &&
+                       cp_fs_protected (hfsmp->hfs_mp)) {
+               int cp_err = 0;
+               cp_err = cp_handle_relocate (cp);
+               if (cp_err) {
+                       /* 
+                        * can't copy the file because we couldn't set up keys.
+                        * bail out 
+                        */
+                       return cp_err;
+               }
+               else {
+                       cpenabled = 1;
+               }
        }
        }
-       
+#endif
+
        /*
         * Determine the I/O size to use
         *
        /*
         * Determine the I/O size to use
         *
@@ -4118,7 +4724,14 @@ hfs_copy_extent(
                buf_setcount(bp, ioSize);
                buf_setblkno(bp, srcSector);
                buf_setlblkno(bp, srcSector);
                buf_setcount(bp, ioSize);
                buf_setblkno(bp, srcSector);
                buf_setlblkno(bp, srcSector);
-               
+
+               /* Attach the CP to the buffer */
+#if CONFIG_PROTECT
+               if (cpenabled) {
+                       buf_setcpaddr (bp, cp->c_cpentry);
+               }
+#endif
+
                /* Do the read */
                err = VNOP_STRATEGY(bp);
                if (!err)
                /* Do the read */
                err = VNOP_STRATEGY(bp);
                if (!err)
@@ -4134,8 +4747,15 @@ hfs_copy_extent(
                buf_setcount(bp, ioSize);
                buf_setblkno(bp, destSector);
                buf_setlblkno(bp, destSector);
                buf_setcount(bp, ioSize);
                buf_setblkno(bp, destSector);
                buf_setlblkno(bp, destSector);
-               if (journal_uses_fua(hfsmp->jnl))
+               if (vnode_issystem(vp) && journal_uses_fua(hfsmp->jnl))
                        buf_markfua(bp);
                        buf_markfua(bp);
+
+#if CONFIG_PROTECT
+               /* Attach the CP to the buffer */
+               if (cpenabled) {
+                       buf_setcpaddr (bp, cp->c_cpentry);
+               }
+#endif
                        
                /* Do the write */
                vnode_startwrite(hfsmp->hfs_devvp);
                        
                /* Do the write */
                vnode_startwrite(hfsmp->hfs_devvp);
@@ -4157,7 +4777,7 @@ hfs_copy_extent(
                kmem_free(kernel_map, (vm_offset_t)buffer, bufferSize);
 
        /* Make sure all writes have been flushed to disk. */
                kmem_free(kernel_map, (vm_offset_t)buffer, bufferSize);
 
        /* Make sure all writes have been flushed to disk. */
-       if (!journal_uses_fua(hfsmp->jnl)) {
+       if (vnode_issystem(vp) && !journal_uses_fua(hfsmp->jnl)) {
                err = VNOP_IOCTL(hfsmp->hfs_devvp, DKIOCSYNCHRONIZECACHE, NULL, FWRITE, context);
                if (err) {
                        printf("hfs_copy_extent: DKIOCSYNCHRONIZECACHE failed (%d)\n", err);
                err = VNOP_IOCTL(hfsmp->hfs_devvp, DKIOCSYNCHRONIZECACHE, NULL, FWRITE, context);
                if (err) {
                        printf("hfs_copy_extent: DKIOCSYNCHRONIZECACHE failed (%d)\n", err);
@@ -4172,222 +4792,1089 @@ hfs_copy_extent(
 }
 
 
 }
 
 
-/*
- * Reclaim space at the end of a volume, used by a given system file.
+/* Structure to store state of reclaiming extents from a 
+ * given file.  hfs_reclaim_file()/hfs_reclaim_xattr() 
+ * initializes the values in this structure which are then 
+ * used by code that reclaims and splits the extents.
+ */
+struct hfs_reclaim_extent_info {
+       struct vnode *vp;
+       u_int32_t fileID;
+       u_int8_t forkType;
+       u_int8_t is_dirlink;                 /* Extent belongs to directory hard link */
+       u_int8_t is_sysfile;                 /* Extent belongs to system file */
+       u_int8_t is_xattr;                   /* Extent belongs to extent-based xattr */
+       u_int8_t extent_index;
+       int lockflags;                       /* Locks that reclaim and split code should grab before modifying the extent record */
+       u_int32_t blocks_relocated;          /* Total blocks relocated for this file till now */
+       u_int32_t recStartBlock;             /* File allocation block number (FABN) for current extent record */
+       u_int32_t cur_blockCount;            /* Number of allocation blocks that have been checked for reclaim */
+       struct filefork *catalog_fp;         /* If non-NULL, extent is from catalog record */
+       union record {
+               HFSPlusExtentRecord overflow;/* Extent record from overflow extents btree */
+               HFSPlusAttrRecord xattr;     /* Attribute record for large EAs */
+       } record;
+       HFSPlusExtentDescriptor *extents;    /* Pointer to current extent record being processed.
+                                             * For catalog extent record, points to the correct 
+                                             * extent information in filefork.  For overflow extent 
+                                             * record, or xattr record, points to extent record 
+                                             * in the structure above
+                                             */
+       struct cat_desc *dirlink_desc;  
+       struct cat_attr *dirlink_attr;
+       struct filefork *dirlink_fork;        /* For directory hard links, fp points actually to this */
+       struct BTreeIterator *iterator;       /* Shared read/write iterator, hfs_reclaim_file/xattr() 
+                                               * use it for reading and hfs_reclaim_extent()/hfs_split_extent() 
+                                              * use it for writing updated extent record 
+                                              */ 
+       struct FSBufferDescriptor btdata;     /* Shared btdata for reading/writing extent record, same as iterator above */
+       u_int16_t recordlen;
+       int overflow_count;                   /* For debugging, counter for overflow extent record */
+       FCB *fcb;                             /* Pointer to the current btree being traversed */
+};
+
+/* 
+ * Split the current extent into two extents, with first extent 
+ * to contain given number of allocation blocks.  Splitting of 
+ * extent creates one new extent entry which can result in 
+ * shifting of many entries through all the extent records of a 
+ * file, and/or creating a new extent record in the overflow 
+ * extent btree. 
  *
  *
- * This routine attempts to move any extent which contains allocation blocks
- * at or after "startblk."  A separate transaction is used to do the move.
- * The contents of any moved extents are read and written via the volume's
- * device vnode -- NOT via "vp."  During the move, moved blocks which are part
- * of a transaction have their physical block numbers invalidated so they will
- * eventually be written to their new locations.
+ * Example:
+ * The diagram below represents two consecutive extent records, 
+ * for simplicity, lets call them record X and X+1 respectively.
+ * Interesting extent entries have been denoted by letters.  
+ * If the letter is unchanged before and after split, it means 
+ * that the extent entry was not modified during the split.  
+ * A '.' means that the entry remains unchanged after the split 
+ * and is not relevant for our example.  A '0' means that the 
+ * extent entry is empty.  
  *
  *
- * This routine can be used to move overflow extents for the allocation file.
+ * If there isn't sufficient contiguous free space to relocate 
+ * an extent (extent "C" below), we will have to break the one 
+ * extent into multiple smaller extents, and relocate each of 
+ * the smaller extents individually.  The way we do this is by 
+ * finding the largest contiguous free space that is currently 
+ * available (N allocation blocks), and then convert extent "C" 
+ * into two extents, C1 and C2, that occupy exactly the same 
+ * allocation blocks as extent C.  Extent C1 is the first 
+ * N allocation blocks of extent C, and extent C2 is the remainder 
+ * of extent C.  Then we can relocate extent C1 since we know 
+ * we have enough contiguous free space to relocate it in its 
+ * entirety.  We then repeat the process starting with extent C2. 
  *
  *
- * Inputs:
- *    hfsmp       The volume being resized.
- *    startblk    Blocks >= this allocation block need to be moved.
- *    locks       Which locks need to be taken for the given system file.
- *    vp          The vnode for the system file.
+ * In record X, only the entries following entry C are shifted, and 
+ * the original entry C is replaced with two entries C1 and C2 which
+ * are actually two extent entries for contiguous allocation blocks.
+ *
+ * Note that the entry E from record X is shifted into record X+1 as 
+ * the new first entry.  Since the first entry of record X+1 is updated, 
+ * the FABN will also get updated with the blockCount of entry E.  
+ * This also results in shifting of all extent entries in record X+1.  
+ * Note that the number of empty entries after the split has been 
+ * changed from 3 to 2. 
  *
  *
- * Outputs:
- *    moved       Set to true if any extents were moved.
+ * Before:
+ *               record X                           record X+1
+ *  ---------------------===---------     ---------------------------------
+ *  | A | . | . | . | B | C | D | E |     | F | . | . | . | G | 0 | 0 | 0 |
+ *  ---------------------===---------     ---------------------------------    
+ *
+ * After:
+ *  ---------------------=======-----     ---------------------------------
+ *  | A | . | . | . | B | C1| C2| D |     | E | F | . | . | . | G | 0 | 0 |
+ *  ---------------------=======-----     ---------------------------------    
+ *
+ *  C1.startBlock = C.startBlock          
+ *  C1.blockCount = N
+ *
+ *  C2.startBlock = C.startBlock + N
+ *  C2.blockCount = C.blockCount - N
+ *
+ *                                        FABN = old FABN - E.blockCount
+ *
+ * Inputs: 
+ *     extent_info - This is the structure that contains state about 
+ *                   the current file, extent, and extent record that 
+ *                   is being relocated.  This structure is shared 
+ *                   among code that traverses through all the extents 
+ *                   of the file, code that relocates extents, and 
+ *                   code that splits the extent. 
+ * Output:
+ *     Zero on success, non-zero on failure.
  */
  */
-static int
-hfs_relocate_callback(__unused HFSPlusExtentKey *key, HFSPlusExtentRecord *record, HFSPlusExtentRecord *state)
-{
-       bcopy(state, record, sizeof(HFSPlusExtentRecord));
-       return 0;
-}
-static int
-hfs_reclaim_sys_file(struct hfsmount *hfsmp, struct vnode *vp, u_long startblk, int locks, Boolean *moved, vfs_context_t context)
+static int 
+hfs_split_extent(struct hfs_reclaim_extent_info *extent_info, uint32_t newBlockCount)
 {
 {
-       int error;
-       int lockflags;
+       int error = 0;
+       int index = extent_info->extent_index;
        int i;
        int i;
-       u_long datablks;
-       u_long block;
-       u_int32_t oldStartBlock;
-       u_int32_t newStartBlock;
-       u_int32_t blockCount;
-       struct filefork *fp;
+       HFSPlusExtentDescriptor shift_extent; /* Extent entry that should be shifted into next extent record */
+       HFSPlusExtentDescriptor last_extent;
+       HFSPlusExtentDescriptor *extents; /* Pointer to current extent record being manipulated */
+       HFSPlusExtentRecord *extents_rec = NULL;
+       HFSPlusExtentKey *extents_key = NULL;
+       HFSPlusAttrRecord *xattr_rec = NULL;
+       HFSPlusAttrKey *xattr_key = NULL;
+       struct BTreeIterator iterator;
+       struct FSBufferDescriptor btdata;
+       uint16_t reclen;
+       uint32_t read_recStartBlock;    /* Starting allocation block number to read old extent record */
+       uint32_t write_recStartBlock;   /* Starting allocation block number to insert newly updated extent record */
+       Boolean create_record = false;
+       Boolean is_xattr;
+       struct cnode *cp;
+       
+       is_xattr = extent_info->is_xattr;
+       extents = extent_info->extents;
+       cp = VTOC(extent_info->vp);
 
 
-       /* If there is no vnode for this file, then there's nothing to do. */   
-       if (vp == NULL)
-               return 0;
+       if (hfs_resize_debug) {
+               printf ("hfs_split_extent: Split record:%u recStartBlock=%u %u:(%u,%u) for %u blocks\n", extent_info->overflow_count, extent_info->recStartBlock, index, extents[index].startBlock, extents[index].blockCount, newBlockCount);
+       }
 
 
-       /* printf("hfs_reclaim_sys_file: %.*s\n", VTOC(vp)->c_desc.cd_namelen, VTOC(vp)->c_desc.cd_nameptr); */
-       
-       /* We always need the allocation bitmap and extents B-tree */
-       locks |= SFL_BITMAP | SFL_EXTENTS;
-       
-       error = hfs_start_transaction(hfsmp);
-       if (error) {
-               printf("hfs_reclaim_sys_file: hfs_start_transaction returned %d\n", error);
-               return error;
+       /* Extents overflow btree can not have more than 8 extents.  
+        * No split allowed if the 8th extent is already used. 
+        */
+       if ((extent_info->fileID == kHFSExtentsFileID) && (extents[kHFSPlusExtentDensity - 1].blockCount != 0)) {
+               printf ("hfs_split_extent: Maximum 8 extents allowed for extents overflow btree, cannot split further.\n");
+               error = ENOSPC;
+               goto out;
        }
        }
-       lockflags = hfs_systemfile_lock(hfsmp, locks, HFS_EXCLUSIVE_LOCK);
-       fp = VTOF(vp);
-       datablks = 0;
 
 
-       /* Relocate non-overflow extents */
-       for (i = 0; i < kHFSPlusExtentDensity; ++i) {
-               if (fp->ff_extents[i].blockCount == 0)
+       /* Determine the starting allocation block number for the following
+        * overflow extent record, if any, before the current record 
+        * gets modified. 
+        */
+       read_recStartBlock = extent_info->recStartBlock;
+       for (i = 0; i < kHFSPlusExtentDensity; i++) {
+               if (extents[i].blockCount == 0) {
                        break;
                        break;
-               oldStartBlock = fp->ff_extents[i].startBlock;
-               blockCount = fp->ff_extents[i].blockCount;
-               datablks += blockCount;
-               block = oldStartBlock + blockCount;
-               if (block > startblk) {
-                       error = BlockAllocate(hfsmp, 1, blockCount, blockCount, true, true, &newStartBlock, &blockCount);
-                       if (error) {
-                               printf("hfs_reclaim_sys_file: BlockAllocate returned %d\n", error);
-                               goto fail;
-                       }
-                       if (blockCount != fp->ff_extents[i].blockCount) {
-                               printf("hfs_reclaim_sys_file: new blockCount=%u, original blockCount=%u", blockCount, fp->ff_extents[i].blockCount);
-                               goto free_fail;
-                       }
-                       error = hfs_copy_extent(hfsmp, vp, oldStartBlock, newStartBlock, blockCount, context);
-                       if (error) {
-                               printf("hfs_reclaim_sys_file: hfs_copy_extent returned %d\n", error);
-                               goto free_fail;
-                       }
-                       fp->ff_extents[i].startBlock = newStartBlock;
-                       VTOC(vp)->c_flag |= C_MODIFIED;
-                       *moved = true;
-                       error = BlockDeallocate(hfsmp, oldStartBlock, blockCount);
-                       if (error) {
-                               /* TODO: Mark volume inconsistent? */
-                               printf("hfs_reclaim_sys_file: BlockDeallocate returned %d\n", error);
-                               goto fail;
-                       }
-                       error = hfs_flushvolumeheader(hfsmp, MNT_WAIT, HFS_ALTFLUSH);
-                       if (error) {
-                               /* TODO: Mark volume inconsistent? */
-                               printf("hfs_reclaim_sys_file: hfs_flushvolumeheader returned %d\n", error);
-                               goto fail;
+               }
+               read_recStartBlock += extents[i].blockCount;
+       }
+
+       /* Shift and split */
+       if (index == kHFSPlusExtentDensity-1) {
+               /* The new extent created after split will go into following overflow extent record */
+               shift_extent.startBlock = extents[index].startBlock + newBlockCount;
+               shift_extent.blockCount = extents[index].blockCount - newBlockCount;
+
+               /* Last extent in the record will be split, so nothing to shift */
+       } else {
+               /* Splitting of extents can result in at most of one 
+                * extent entry to be shifted into following overflow extent 
+                * record.  So, store the last extent entry for later. 
+                */
+               shift_extent = extents[kHFSPlusExtentDensity-1];
+               if ((hfs_resize_debug) && (shift_extent.blockCount != 0)) {
+                       printf ("hfs_split_extent: Save 7:(%u,%u) to shift into overflow record\n", shift_extent.startBlock, shift_extent.blockCount);
+               }
+
+               /* Start shifting extent information from the end of the extent 
+                * record to the index where we want to insert the new extent.
+                * Note that kHFSPlusExtentDensity-1 is already saved above, and 
+                * does not need to be shifted.  The extent entry that is being 
+                * split does not get shifted.
+                */
+               for (i = kHFSPlusExtentDensity-2; i > index; i--) {
+                       if (hfs_resize_debug) {
+                               if (extents[i].blockCount) {
+                                       printf ("hfs_split_extent: Shift %u:(%u,%u) to %u:(%u,%u)\n", i, extents[i].startBlock, extents[i].blockCount, i+1, extents[i].startBlock, extents[i].blockCount);
+                               }
                        }
                        }
+                       extents[i+1] = extents[i];
                }
        }
 
                }
        }
 
-       /* Relocate overflow extents (if any) */
-       if (i == kHFSPlusExtentDensity && fp->ff_blocks > datablks) {
-               struct BTreeIterator *iterator = NULL;
-               struct FSBufferDescriptor btdata;
-               HFSPlusExtentRecord record;
-               HFSPlusExtentKey *key;
-               FCB *fcb;
-               u_int32_t fileID;
-               u_int8_t forktype;
+       if (index == kHFSPlusExtentDensity-1) {
+               /* The second half of the extent being split will be the overflow 
+                * entry that will go into following overflow extent record.  The
+                * value has been stored in 'shift_extent' above, so there is 
+                * nothing to be done here.
+                */
+       } else {
+               /* Update the values in the second half of the extent being split 
+                * before updating the first half of the split.  Note that the 
+                * extent to split or first half of the split is at index 'index' 
+                * and a new extent or second half of the split will be inserted at 
+                * 'index+1' or into following overflow extent record. 
+                */ 
+               extents[index+1].startBlock = extents[index].startBlock + newBlockCount;
+               extents[index+1].blockCount = extents[index].blockCount - newBlockCount;
+       }
+       /* Update the extent being split, only the block count will change */
+       extents[index].blockCount = newBlockCount;
+
+       if (hfs_resize_debug) {
+               printf ("hfs_split_extent: Split %u:(%u,%u) and ", index, extents[index].startBlock, extents[index].blockCount);
+               if (index != kHFSPlusExtentDensity-1) {
+                       printf ("%u:(%u,%u)\n", index+1, extents[index+1].startBlock, extents[index+1].blockCount);
+               } else {
+                       printf ("overflow:(%u,%u)\n", shift_extent.startBlock, shift_extent.blockCount);
+               }
+       }
 
 
-               forktype = VNODE_IS_RSRC(vp) ? 0xFF : 0;
-               fileID = VTOC(vp)->c_cnid;
-               if (kmem_alloc(kernel_map, (vm_offset_t*) &iterator, sizeof(*iterator))) {
-                       printf("hfs_reclaim_sys_file: kmem_alloc failed!\n");
-                       error = ENOMEM;
-                       goto fail;
+       /* Write out information about the newly split extent to the disk */
+       if (extent_info->catalog_fp) {
+               /* (extent_info->catalog_fp != NULL) means the newly split 
+                * extent exists in the catalog record.  This means that 
+                * the cnode was updated.  Therefore, to write out the changes,
+                * mark the cnode as modified.   We cannot call hfs_update()
+                * in this function because the caller hfs_reclaim_extent() 
+                * is holding the catalog lock currently.
+                */
+               cp->c_flag |= C_MODIFIED;
+       } else {
+               /* The newly split extent is for large EAs or is in overflow 
+                * extent record, so update it directly in the btree using the 
+                * iterator information from the shared extent_info structure
+                */
+               error = BTReplaceRecord(extent_info->fcb, extent_info->iterator, 
+                               &(extent_info->btdata), extent_info->recordlen);
+               if (error) {
+                       printf ("hfs_split_extent: fileID=%u BTReplaceRecord returned error=%d\n", extent_info->fileID, error);
+                       goto out;
+               }
+       }
+               
+       /* No extent entry to be shifted into another extent overflow record */
+       if (shift_extent.blockCount == 0) {
+               if (hfs_resize_debug) {
+                       printf ("hfs_split_extent: No extent entry to be shifted into overflow records\n");
                }
                }
+               error = 0;
+               goto out;
+       }
 
 
-               bzero(iterator, sizeof(*iterator));
-               key = (HFSPlusExtentKey *) &iterator->key;
-               key->keyLength = kHFSPlusExtentKeyMaximumLength;
-               key->forkType = forktype;
-               key->fileID = fileID;
-               key->startBlock = datablks;
+       /* The overflow extent entry has to be shifted into an extent 
+        * overflow record.  This means that we might have to shift 
+        * extent entries from all subsequent overflow records by one. 
+        * We start iteration from the first record to the last record, 
+        * and shift the extent entry from one record to another.  
+        * We might have to create a new extent record for the last 
+        * extent entry for the file. 
+        */
        
        
-               btdata.bufferAddress = &record;
-               btdata.itemSize = sizeof(record);
+       /* Initialize iterator to search the next record */
+       bzero(&iterator, sizeof(iterator));
+       if (is_xattr) {
+               /* Copy the key from the iterator that was used to update the modified attribute record. */
+               xattr_key = (HFSPlusAttrKey *)&(iterator.key);
+               bcopy((HFSPlusAttrKey *)&(extent_info->iterator->key), xattr_key, sizeof(HFSPlusAttrKey));
+               /* Note: xattr_key->startBlock will be initialized later in the iteration loop */
+
+               MALLOC(xattr_rec, HFSPlusAttrRecord *, 
+                               sizeof(HFSPlusAttrRecord), M_TEMP, M_WAITOK);
+               if (xattr_rec == NULL) {
+                       error = ENOMEM;
+                       goto out;
+               }
+               btdata.bufferAddress = xattr_rec;
+               btdata.itemSize = sizeof(HFSPlusAttrRecord);
                btdata.itemCount = 1;
                btdata.itemCount = 1;
-       
-               fcb = VTOF(hfsmp->hfs_extents_vp);
+               extents = xattr_rec->overflowExtents.extents;
+       } else {
+               /* Initialize the extent key for the current file */
+               extents_key = (HFSPlusExtentKey *) &(iterator.key);
+               extents_key->keyLength = kHFSPlusExtentKeyMaximumLength;
+               extents_key->forkType = extent_info->forkType;
+               extents_key->fileID = extent_info->fileID;
+               /* Note: extents_key->startBlock will be initialized later in the iteration loop */
+               
+               MALLOC(extents_rec, HFSPlusExtentRecord *, 
+                               sizeof(HFSPlusExtentRecord), M_TEMP, M_WAITOK);
+               if (extents_rec == NULL) {
+                       error = ENOMEM;
+                       goto out;
+               }
+               btdata.bufferAddress = extents_rec;
+               btdata.itemSize = sizeof(HFSPlusExtentRecord);
+               btdata.itemCount = 1;
+               extents = extents_rec[0];
+       }
 
 
-               error = BTSearchRecord(fcb, iterator, &btdata, NULL, iterator);
-               while (error == 0) {
-                       /* Stop when we encounter a different file or fork. */
-                       if ((key->fileID != fileID) ||
-                               (key->forkType != forktype)) {
-                               break;
+       /* The overflow extent entry has to be shifted into an extent 
+        * overflow record.  This means that we might have to shift 
+        * extent entries from all subsequent overflow records by one. 
+        * We start iteration from the first record to the last record, 
+        * examine one extent record in each iteration and shift one 
+        * extent entry from one record to another.  We might have to 
+        * create a new extent record for the last extent entry for the 
+        * file. 
+        *
+        * If shift_extent.blockCount is non-zero, it means that there is 
+        * an extent entry that needs to be shifted into the next 
+        * overflow extent record.  We keep on going till there are no such 
+        * entries left to be shifted.  This will also change the starting 
+        * allocation block number of the extent record which is part of 
+        * the key for the extent record in each iteration.  Note that 
+        * because the extent record key is changing while we are searching, 
+        * the record can not be updated directly, instead it has to be 
+        * deleted and inserted again.
+        */
+       while (shift_extent.blockCount) {
+               if (hfs_resize_debug) {
+                       printf ("hfs_split_extent: Will shift (%u,%u) into overflow record with startBlock=%u\n", shift_extent.startBlock, shift_extent.blockCount, read_recStartBlock);
+               }
+
+               /* Search if there is any existing overflow extent record
+                * that matches the current file and the logical start block 
+                * number.
+                *
+                * For this, the logical start block number in the key is 
+                * the value calculated based on the logical start block 
+                * number of the current extent record and the total number 
+                * of blocks existing in the current extent record.  
+                */
+               if (is_xattr) {
+                       xattr_key->startBlock = read_recStartBlock;
+               } else {
+                       extents_key->startBlock = read_recStartBlock;
+               }
+               error = BTSearchRecord(extent_info->fcb, &iterator, &btdata, &reclen, &iterator);
+               if (error) {
+                       if (error != btNotFound) {
+                               printf ("hfs_split_extent: fileID=%u startBlock=%u BTSearchRecord error=%d\n", extent_info->fileID, read_recStartBlock, error);
+                               goto out;
                        }
                        }
-                       /* 
-                        * Check if the file overlaps target space.
+                       /* No matching record was found, so create a new extent record.
+                        * Note:  Since no record was found, we can't rely on the 
+                        * btree key in the iterator any longer.  This will be initialized
+                        * later before we insert the record.
                         */
                         */
-                       for (i = 0; i < kHFSPlusExtentDensity; ++i) {
-                               if (record[i].blockCount == 0) {
-                                       goto overflow_done;
-                               }
-                               oldStartBlock = record[i].startBlock;
-                               blockCount = record[i].blockCount;
-                               block = oldStartBlock + blockCount;
-                               if (block > startblk) {
-                                       error = BlockAllocate(hfsmp, 1, blockCount, blockCount, true, true, &newStartBlock, &blockCount);
-                                       if (error) {
-                                               printf("hfs_reclaim_sys_file: BlockAllocate returned %d\n", error);
-                                               goto overflow_done;
-                                       }
-                                       if (blockCount != record[i].blockCount) {
-                                               printf("hfs_reclaim_sys_file: new blockCount=%u, original blockCount=%u", blockCount, fp->ff_extents[i].blockCount);
-                                               kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
-                                               goto free_fail;
-                                       }
-                                       error = hfs_copy_extent(hfsmp, vp, oldStartBlock, newStartBlock, blockCount, context);
-                                       if (error) {
-                                               printf("hfs_reclaim_sys_file: hfs_copy_extent returned %d\n", error);
-                                               kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
-                                               goto free_fail;
-                                       }
-                                       record[i].startBlock = newStartBlock;
-                                       VTOC(vp)->c_flag |= C_MODIFIED;
-                                       *moved = true;
-                                       /*
-                                        * NOTE: To support relocating overflow extents of the
-                                        * allocation file, we must update the BTree record BEFORE
-                                        * deallocating the old extent so that BlockDeallocate will
-                                        * use the extent's new location to calculate physical block
-                                        * numbers.  (This is for the case where the old extent's
-                                        * bitmap bits actually reside in the extent being moved.)
-                                        */
-                                       error = BTUpdateRecord(fcb, iterator, (IterateCallBackProcPtr) hfs_relocate_callback, &record);
-                                       if (error) {
-                                               /* TODO: Mark volume inconsistent? */
-                                               printf("hfs_reclaim_sys_file: BTUpdateRecord returned %d\n", error);
-                                               goto overflow_done;
-                                       }
-                                       error = BlockDeallocate(hfsmp, oldStartBlock, blockCount);
-                                       if (error) {
-                                               /* TODO: Mark volume inconsistent? */
-                                               printf("hfs_reclaim_sys_file: BlockDeallocate returned %d\n", error);
-                                               goto overflow_done;
-                                       }
-                               }
+                       create_record = true;
+               }
+       
+               /* The extra extent entry from the previous record is being inserted
+                * as the first entry in the current extent record.  This will change 
+                * the file allocation block number (FABN) of the current extent 
+                * record, which is the startBlock value from the extent record key.
+                * Since one extra entry is being inserted in the record, the new 
+                * FABN for the record will less than old FABN by the number of blocks 
+                * in the new extent entry being inserted at the start.  We have to 
+                * do this before we update read_recStartBlock to point at the 
+                * startBlock of the following record.
+                */
+               write_recStartBlock = read_recStartBlock - shift_extent.blockCount;
+               if (hfs_resize_debug) {
+                       if (create_record) {
+                               printf ("hfs_split_extent: No records found for startBlock=%u, will create new with startBlock=%u\n", read_recStartBlock, write_recStartBlock);
                        }
                        }
-                       /* Look for more records. */
-                       error = BTIterateRecord(fcb, kBTreeNextRecord, iterator, &btdata, NULL);
-                       if (error == btNotFound) {
-                               error = 0;
+               }
+
+               /* Now update the read_recStartBlock to account for total number 
+                * of blocks in this extent record.  It will now point to the 
+                * starting allocation block number for the next extent record.
+                */
+               for (i = 0; i < kHFSPlusExtentDensity; i++) {
+                       if (extents[i].blockCount == 0) {
                                break;
                        }
                                break;
                        }
+                       read_recStartBlock += extents[i].blockCount;
                }
                }
-overflow_done:
-               kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
-               if (error) {
-                       goto fail;
-               }
-       }
-       
-       hfs_systemfile_unlock(hfsmp, lockflags);
-       error = hfs_end_transaction(hfsmp);
-       if (error) {
-               printf("hfs_reclaim_sys_file: hfs_end_transaction returned %d\n", error);
-       }
 
 
-       return error;
+               if (create_record == true) {
+                       /* Initialize new record content with only one extent entry */
+                       bzero(extents, sizeof(HFSPlusExtentRecord));
+                       /* The new record will contain only one extent entry */
+                       extents[0] = shift_extent;
+                       /* There are no more overflow extents to be shifted */
+                       shift_extent.startBlock = shift_extent.blockCount = 0;
+
+                       if (is_xattr) {
+                               /* BTSearchRecord above returned btNotFound,
+                                * but since the attribute btree is never empty
+                                * if we are trying to insert new overflow 
+                                * record for the xattrs, the extents_key will
+                                * contain correct data.  So we don't need to 
+                                * re-initialize it again like below. 
+                                */
+
+                               /* Initialize the new xattr record */
+                               xattr_rec->recordType = kHFSPlusAttrExtents; 
+                               xattr_rec->overflowExtents.reserved = 0;
+                               reclen = sizeof(HFSPlusAttrExtents);
+                       } else {
+                               /* BTSearchRecord above returned btNotFound, 
+                                * which means that extents_key content might 
+                                * not correspond to the record that we are 
+                                * trying to create, especially when the extents 
+                                * overflow btree is empty.  So we reinitialize 
+                                * the extents_key again always. 
+                                */
+                               extents_key->keyLength = kHFSPlusExtentKeyMaximumLength;
+                               extents_key->forkType = extent_info->forkType;
+                               extents_key->fileID = extent_info->fileID;
+
+                               /* Initialize the new extent record */
+                               reclen = sizeof(HFSPlusExtentRecord);
+                       }
+               } else {
+                       /* The overflow extent entry from previous record will be 
+                        * the first entry in this extent record.  If the last 
+                        * extent entry in this record is valid, it will be shifted 
+                        * into the following extent record as its first entry.  So 
+                        * save the last entry before shifting entries in current 
+                        * record.
+                        */
+                       last_extent = extents[kHFSPlusExtentDensity-1];
+                       
+                       /* Shift all entries by one index towards the end */
+                       for (i = kHFSPlusExtentDensity-2; i >= 0; i--) {
+                               extents[i+1] = extents[i];
+                       }
+
+                       /* Overflow extent entry saved from previous record 
+                        * is now the first entry in the current record.
+                        */
+                       extents[0] = shift_extent;
+
+                       if (hfs_resize_debug) {
+                               printf ("hfs_split_extent: Shift overflow=(%u,%u) to record with updated startBlock=%u\n", shift_extent.startBlock, shift_extent.blockCount, write_recStartBlock);
+                       }
+
+                       /* The last entry from current record will be the 
+                        * overflow entry which will be the first entry for 
+                        * the following extent record.
+                        */
+                       shift_extent = last_extent;
+
+                       /* Since the key->startBlock is being changed for this record, 
+                        * it should be deleted and inserted with the new key.
+                        */
+                       error = BTDeleteRecord(extent_info->fcb, &iterator);
+                       if (error) {
+                               printf ("hfs_split_extent: fileID=%u startBlock=%u BTDeleteRecord error=%d\n", extent_info->fileID, read_recStartBlock, error);
+                               goto out;
+                       }
+                       if (hfs_resize_debug) {
+                               printf ("hfs_split_extent: Deleted record with startBlock=%u\n", (is_xattr ? xattr_key->startBlock : extents_key->startBlock));
+                       }
+               }
+
+               /* Insert the newly created or modified extent record */
+               bzero(&iterator.hint, sizeof(iterator.hint));
+               if (is_xattr) {
+                       xattr_key->startBlock = write_recStartBlock;
+               } else {
+                       extents_key->startBlock = write_recStartBlock;
+               }
+               error = BTInsertRecord(extent_info->fcb, &iterator, &btdata, reclen);
+               if (error) {
+                       printf ("hfs_split_extent: fileID=%u, startBlock=%u BTInsertRecord error=%d\n", extent_info->fileID, write_recStartBlock, error);
+                       goto out;
+               }
+               if (hfs_resize_debug) {
+                       printf ("hfs_split_extent: Inserted extent record with startBlock=%u\n", write_recStartBlock);
+               }
+       }
+       BTFlushPath(extent_info->fcb);
+out:
+       if (extents_rec) {
+               FREE (extents_rec, M_TEMP);
+       }
+       if (xattr_rec) {
+               FREE (xattr_rec, M_TEMP);
+       }
+       return error;
+}
+
+
+/* 
+ * Relocate an extent if it lies beyond the expected end of volume.
+ *
+ * This function is called for every extent of the file being relocated.  
+ * It allocates space for relocation, copies the data, deallocates 
+ * the old extent, and update corresponding on-disk extent.  If the function 
+ * does not find contiguous space to  relocate an extent, it splits the 
+ * extent in smaller size to be able to relocate it out of the area of 
+ * disk being reclaimed.  As an optimization, if an extent lies partially 
+ * in the area of the disk being reclaimed, it is split so that we only 
+ * have to relocate the area that was overlapping with the area of disk
+ * being reclaimed. 
+ *
+ * Note that every extent is relocated in its own transaction so that 
+ * they do not overwhelm the journal.  This function handles the extent
+ * record that exists in the catalog record, extent record from overflow 
+ * extents btree, and extents for large EAs.
+ *
+ * Inputs: 
+ *     extent_info - This is the structure that contains state about 
+ *                   the current file, extent, and extent record that 
+ *                   is being relocated.  This structure is shared 
+ *                   among code that traverses through all the extents 
+ *                   of the file, code that relocates extents, and 
+ *                   code that splits the extent. 
+ */
+static int
+hfs_reclaim_extent(struct hfsmount *hfsmp, const u_long allocLimit, struct hfs_reclaim_extent_info *extent_info, vfs_context_t context)
+{
+       int error = 0;
+       int index;
+       struct cnode *cp;
+       u_int32_t oldStartBlock;
+       u_int32_t oldBlockCount;
+       u_int32_t newStartBlock;
+       u_int32_t newBlockCount;
+       u_int32_t roundedBlockCount;
+       uint16_t node_size;
+       uint32_t remainder_blocks;
+       u_int32_t alloc_flags;
+       int blocks_allocated = false;
+
+       index = extent_info->extent_index;
+       cp = VTOC(extent_info->vp);
+
+       oldStartBlock = extent_info->extents[index].startBlock;
+       oldBlockCount = extent_info->extents[index].blockCount;
+
+       if (0 && hfs_resize_debug) {
+               printf ("hfs_reclaim_extent: Examine record:%u recStartBlock=%u, %u:(%u,%u)\n", extent_info->overflow_count, extent_info->recStartBlock, index, oldStartBlock, oldBlockCount);
+       }
+
+       /* If the current extent lies completely within allocLimit, 
+        * it does not require any relocation. 
+        */
+       if ((oldStartBlock + oldBlockCount) <= allocLimit) {
+               extent_info->cur_blockCount += oldBlockCount;
+               return error;
+       } 
+
+       /* Every extent should be relocated in its own transaction
+        * to make sure that we don't overflow the journal buffer.
+        */
+       error = hfs_start_transaction(hfsmp);
+       if (error) {
+               return error;
+       }
+       extent_info->lockflags = hfs_systemfile_lock(hfsmp, extent_info->lockflags, HFS_EXCLUSIVE_LOCK);
+
+       /* Check if the extent lies partially in the area to reclaim, 
+        * i.e. it starts before allocLimit and ends beyond allocLimit.  
+        * We have already skipped extents that lie completely within 
+        * allocLimit in the check above, so we only check for the 
+        * startBlock.  If it lies partially, split it so that we 
+        * only relocate part of the extent.
+        */
+       if (oldStartBlock < allocLimit) {
+               newBlockCount = allocLimit - oldStartBlock;
+               
+               /* If the extent belongs to a btree, check and trim 
+                * it to be multiple of the node size. 
+                */
+               if (extent_info->is_sysfile) {
+                       node_size = get_btree_nodesize(extent_info->vp);
+                       /* If the btree node size is less than the block size, 
+                        * splitting this extent will not split a node across 
+                        * different extents.  So we only check and trim if 
+                        * node size is more than the allocation block size. 
+                        */ 
+                       if (node_size > hfsmp->blockSize) {
+                               remainder_blocks = newBlockCount % (node_size / hfsmp->blockSize);
+                               if (remainder_blocks) {
+                                       newBlockCount -= remainder_blocks;
+                                       if (hfs_resize_debug) {
+                                               printf ("hfs_reclaim_extent: Fixing extent block count, node_blks=%u, old=%u, new=%u\n", node_size/hfsmp->blockSize, newBlockCount + remainder_blocks, newBlockCount);
+                                       }
+                               }
+                       }
+               }
+
+               if (hfs_resize_debug) {
+                       int idx = extent_info->extent_index;
+                       printf ("hfs_reclaim_extent: Split straddling extent %u:(%u,%u) for %u blocks\n", idx, extent_info->extents[idx].startBlock, extent_info->extents[idx].blockCount, newBlockCount);
+               }
+
+               /* Split the extents into two parts --- the first extent lies
+                * completely within allocLimit and therefore does not require
+                * relocation.  The second extent will require relocation which
+                * will be handled when the caller calls this function again 
+                * for the next extent. 
+                */
+               error = hfs_split_extent(extent_info, newBlockCount);
+               if (error == 0) {
+                       /* Split success, no relocation required */
+                       goto out;
+               }
+               /* Split failed, so try to relocate entire extent */
+               if (hfs_resize_debug) {
+                       printf ("hfs_reclaim_extent: Split straddling extent failed, reclocate full extent\n");
+               }
+       }
+
+       /* At this point, the current extent requires relocation.  
+        * We will try to allocate space equal to the size of the extent 
+        * being relocated first to try to relocate it without splitting.  
+        * If the allocation fails, we will try to allocate contiguous 
+        * blocks out of metadata zone.  If that allocation also fails, 
+        * then we will take a whatever contiguous block run is returned 
+        * by the allocation, split the extent into two parts, and then 
+        * relocate the first splitted extent. 
+        */
+       alloc_flags = HFS_ALLOC_FORCECONTIG | HFS_ALLOC_SKIPFREEBLKS; 
+       if (extent_info->is_sysfile) {
+               alloc_flags |= HFS_ALLOC_METAZONE;
+       }
+
+       error = BlockAllocate(hfsmp, 1, oldBlockCount, oldBlockCount, alloc_flags, 
+                       &newStartBlock, &newBlockCount);
+       if ((extent_info->is_sysfile == false) && 
+           ((error == dskFulErr) || (error == ENOSPC))) {
+               /* For non-system files, try reallocating space in metadata zone */
+               alloc_flags |= HFS_ALLOC_METAZONE;
+               error = BlockAllocate(hfsmp, 1, oldBlockCount, oldBlockCount, 
+                               alloc_flags, &newStartBlock, &newBlockCount);
+       } 
+       if ((error == dskFulErr) || (error == ENOSPC)) {
+               /* We did not find desired contiguous space for this extent.  
+                * So try to allocate the maximum contiguous space available.
+                */
+               alloc_flags &= ~HFS_ALLOC_FORCECONTIG;
+
+               error = BlockAllocate(hfsmp, 1, oldBlockCount, oldBlockCount, 
+                               alloc_flags, &newStartBlock, &newBlockCount);
+               if (error) {
+                       printf ("hfs_reclaim_extent: fileID=%u start=%u, %u:(%u,%u) BlockAllocate error=%d\n", extent_info->fileID, extent_info->recStartBlock, index, oldStartBlock, oldBlockCount, error);
+                       goto out;
+               }
+               blocks_allocated = true;
+
+               /* The number of blocks allocated is less than the requested 
+                * number of blocks.  For btree extents, check and trim the 
+                * extent to be multiple of the node size. 
+                */
+               if (extent_info->is_sysfile) {
+                       node_size = get_btree_nodesize(extent_info->vp);
+                       if (node_size > hfsmp->blockSize) {
+                               remainder_blocks = newBlockCount % (node_size / hfsmp->blockSize);
+                               if (remainder_blocks) {
+                                       roundedBlockCount = newBlockCount - remainder_blocks;
+                                       /* Free tail-end blocks of the newly allocated extent */
+                                       BlockDeallocate(hfsmp, newStartBlock + roundedBlockCount,
+                                                              newBlockCount - roundedBlockCount,
+                                                              HFS_ALLOC_SKIPFREEBLKS);
+                                       newBlockCount = roundedBlockCount;
+                                       if (hfs_resize_debug) {
+                                               printf ("hfs_reclaim_extent: Fixing extent block count, node_blks=%u, old=%u, new=%u\n", node_size/hfsmp->blockSize, newBlockCount + remainder_blocks, newBlockCount);
+                                       }
+                                       if (newBlockCount == 0) {
+                                               printf ("hfs_reclaim_extent: Not enough contiguous blocks available to relocate fileID=%d\n", extent_info->fileID);
+                                               error = ENOSPC;
+                                               goto out;
+                                       }
+                               }
+                       }
+               }
+
+               /* The number of blocks allocated is less than the number of 
+                * blocks requested, so split this extent --- the first extent 
+                * will be relocated as part of this function call and the caller
+                * will handle relocating the second extent by calling this 
+                * function again for the second extent. 
+                */
+               error = hfs_split_extent(extent_info, newBlockCount);
+               if (error) {
+                       printf ("hfs_reclaim_extent: fileID=%u start=%u, %u:(%u,%u) split error=%d\n", extent_info->fileID, extent_info->recStartBlock, index, oldStartBlock, oldBlockCount, error);
+                       goto out;
+               }
+               oldBlockCount = newBlockCount;
+       }
+       if (error) {
+               printf ("hfs_reclaim_extent: fileID=%u start=%u, %u:(%u,%u) contig BlockAllocate error=%d\n", extent_info->fileID, extent_info->recStartBlock, index, oldStartBlock, oldBlockCount, error);
+               goto out;
+       }
+       blocks_allocated = true;
+
+       /* Copy data from old location to new location */
+       error = hfs_copy_extent(hfsmp, extent_info->vp, oldStartBlock, 
+                       newStartBlock, newBlockCount, context);
+       if (error) {
+               printf ("hfs_reclaim_extent: fileID=%u start=%u, %u:(%u,%u)=>(%u,%u) hfs_copy_extent error=%d\n", extent_info->fileID, extent_info->recStartBlock, index, oldStartBlock, oldBlockCount, newStartBlock, newBlockCount, error);
+               goto out;
+       }
+
+       /* Update the extent record with the new start block information */
+       extent_info->extents[index].startBlock = newStartBlock;
+
+       /* Sync the content back to the disk */
+       if (extent_info->catalog_fp) {
+               /* Update the extents in catalog record */
+               if (extent_info->is_dirlink) {
+                       error = cat_update_dirlink(hfsmp, extent_info->forkType, 
+                                       extent_info->dirlink_desc, extent_info->dirlink_attr, 
+                                       &(extent_info->dirlink_fork->ff_data));
+               } else {
+                       cp->c_flag |= C_MODIFIED;
+                       /* If this is a system file, sync volume headers on disk */
+                       if (extent_info->is_sysfile) {
+                               error = hfs_flushvolumeheader(hfsmp, MNT_WAIT, HFS_ALTFLUSH);
+                       }
+               }
+       } else {
+               /* Replace record for extents overflow or extents-based xattrs */
+               error = BTReplaceRecord(extent_info->fcb, extent_info->iterator, 
+                               &(extent_info->btdata), extent_info->recordlen);
+       }
+       if (error) {
+               printf ("hfs_reclaim_extent: fileID=%u, update record error=%u\n", extent_info->fileID, error);
+               goto out;
+       }
+
+       /* Deallocate the old extent */
+       error = BlockDeallocate(hfsmp, oldStartBlock, oldBlockCount, HFS_ALLOC_SKIPFREEBLKS);
+       if (error) {
+               printf ("hfs_reclaim_extent: fileID=%u start=%u, %u:(%u,%u) BlockDeallocate error=%d\n", extent_info->fileID, extent_info->recStartBlock, index, oldStartBlock, oldBlockCount, error);
+               goto out;
+       }
+       extent_info->blocks_relocated += newBlockCount;
+
+       if (hfs_resize_debug) {
+               printf ("hfs_reclaim_extent: Relocated record:%u %u:(%u,%u) to (%u,%u)\n", extent_info->overflow_count, index, oldStartBlock, oldBlockCount, newStartBlock, newBlockCount);
+       }
+
+out:
+       if (error != 0) {
+               if (blocks_allocated == true) {
+                       BlockDeallocate(hfsmp, newStartBlock, newBlockCount, HFS_ALLOC_SKIPFREEBLKS);
+               }
+       } else {
+               /* On success, increment the total allocation blocks processed */
+               extent_info->cur_blockCount += newBlockCount;
+       }
+
+       hfs_systemfile_unlock(hfsmp, extent_info->lockflags);
+
+       /* For a non-system file, if an extent entry from catalog record 
+        * was modified, sync the in-memory changes to the catalog record
+        * on disk before ending the transaction.
+        */
+        if ((extent_info->catalog_fp) && 
+            (extent_info->is_sysfile == false)) {
+               (void) hfs_update(extent_info->vp, MNT_WAIT);
+       }
+
+       hfs_end_transaction(hfsmp);
+
+       return error;
+}
+
+/* Report intermediate progress during volume resize */
+static void 
+hfs_truncatefs_progress(struct hfsmount *hfsmp)
+{
+       u_int32_t cur_progress;
+
+       hfs_resize_progress(hfsmp, &cur_progress);
+       if (cur_progress > (hfsmp->hfs_resize_progress + 9)) {
+               printf("hfs_truncatefs: %d%% done...\n", cur_progress);
+               hfsmp->hfs_resize_progress = cur_progress;
+       }
+       return;
+}
+
+/*
+ * Reclaim space at the end of a volume for given file and forktype. 
+ *
+ * This routine attempts to move any extent which contains allocation blocks
+ * at or after "allocLimit."  A separate transaction is used for every extent 
+ * that needs to be moved.  If there is not contiguous space available for 
+ * moving an extent, it can be split into smaller extents.  The contents of 
+ * any moved extents are read and written via the volume's device vnode -- 
+ * NOT via "vp."  During the move, moved blocks which are part of a transaction 
+ * have their physical block numbers invalidated so they will eventually be 
+ * written to their new locations.
+ *
+ * This function is also called for directory hard links.  Directory hard links
+ * are regular files with no data fork and resource fork that contains alias 
+ * information for backward compatibility with pre-Leopard systems.  However 
+ * non-Mac OS X implementation can add/modify data fork or resource fork 
+ * information to directory hard links, so we check, and if required, relocate 
+ * both data fork and resource fork.  
+ *
+ * Inputs:
+ *    hfsmp       The volume being resized.
+ *    vp          The vnode for the system file.
+ *    fileID     ID of the catalog record that needs to be relocated
+ *    forktype   The type of fork that needs relocated,
+ *                     kHFSResourceForkType for resource fork,
+ *                     kHFSDataForkType for data fork
+ *    allocLimit  Allocation limit for the new volume size, 
+ *               do not use this block or beyond.  All extents 
+ *               that use this block or any blocks beyond this limit 
+ *               will be relocated.
+ *
+ * Side Effects:
+ * hfsmp->hfs_resize_blocksmoved is incremented by the number of allocation 
+ * blocks that were relocated. 
+ */
+static int
+hfs_reclaim_file(struct hfsmount *hfsmp, struct vnode *vp, u_int32_t fileID, 
+               u_int8_t forktype, u_long allocLimit, vfs_context_t context)
+{
+       int error = 0;
+       struct hfs_reclaim_extent_info *extent_info;
+       int i;
+       int lockflags = 0;
+       struct cnode *cp;
+       struct filefork *fp;
+       int took_truncate_lock = false;
+       int release_desc = false;
+       HFSPlusExtentKey *key;
+               
+       /* If there is no vnode for this file, then there's nothing to do. */   
+       if (vp == NULL) {
+               return 0;
+       }
+
+       cp = VTOC(vp);
+
+       MALLOC(extent_info, struct hfs_reclaim_extent_info *, 
+              sizeof(struct hfs_reclaim_extent_info), M_TEMP, M_WAITOK);
+       if (extent_info == NULL) {
+               return ENOMEM;
+       }
+       bzero(extent_info, sizeof(struct hfs_reclaim_extent_info));
+       extent_info->vp = vp;
+       extent_info->fileID = fileID;
+       extent_info->forkType = forktype;
+       extent_info->is_sysfile = vnode_issystem(vp);
+       if (vnode_isdir(vp) && (cp->c_flag & C_HARDLINK)) {
+               extent_info->is_dirlink = true;
+       }
+       /* We always need allocation bitmap and extent btree lock */
+       lockflags = SFL_BITMAP | SFL_EXTENTS;
+       if ((fileID == kHFSCatalogFileID) || (extent_info->is_dirlink == true)) {
+               lockflags |= SFL_CATALOG;
+       } else if (fileID == kHFSAttributesFileID) {
+               lockflags |= SFL_ATTRIBUTE;
+       } else if (fileID == kHFSStartupFileID) {
+               lockflags |= SFL_STARTUP;
+       }
+       extent_info->lockflags = lockflags;
+       extent_info->fcb = VTOF(hfsmp->hfs_extents_vp);
+
+       /* Flush data associated with current file on disk. 
+        *
+        * If the current vnode is directory hard link, no flushing of 
+        * journal or vnode is required.  The current kernel does not 
+        * modify data/resource fork of directory hard links, so nothing 
+        * will be in the cache.  If a directory hard link is newly created, 
+        * the resource fork data is written directly using devvp and 
+        * the code that actually relocates data (hfs_copy_extent()) also
+        * uses devvp for its I/O --- so they will see a consistent copy. 
+        */
+       if (extent_info->is_sysfile) {
+               /* If the current vnode is system vnode, flush journal 
+                * to make sure that all data is written to the disk.
+                */
+               error = hfs_journal_flush(hfsmp, TRUE);
+               if (error) {
+                       printf ("hfs_reclaim_file: journal_flush returned %d\n", error);
+                       goto out;
+               }
+       } else if (extent_info->is_dirlink == false) {
+               /* Flush all blocks associated with this regular file vnode.  
+                * Normally there should not be buffer cache blocks for regular 
+                * files, but for objects like symlinks, we can have buffer cache 
+                * blocks associated with the vnode.  Therefore we call
+                * buf_flushdirtyblks() also.
+                */
+               buf_flushdirtyblks(vp, 0, BUF_SKIP_LOCKED, "hfs_reclaim_file");
+
+               hfs_unlock(cp);
+               hfs_lock_truncate(cp, HFS_EXCLUSIVE_LOCK);
+               took_truncate_lock = true;
+               (void) cluster_push(vp, 0);
+               error = hfs_lock(cp, HFS_FORCE_LOCK);
+               if (error) {
+                       goto out;
+               }
+
+               /* If the file no longer exists, nothing left to do */
+               if (cp->c_flag & C_NOEXISTS) {
+                       error = 0;
+                       goto out;
+               }
+
+               /* Wait for any in-progress writes to this vnode to complete, so that we'll
+                * be copying consistent bits.  (Otherwise, it's possible that an async
+                * write will complete to the old extent after we read from it.  That
+                * could lead to corruption.)
+                */
+               error = vnode_waitforwrites(vp, 0, 0, 0, "hfs_reclaim_file");
+               if (error) {
+                       goto out;
+               }
+       }
+
+       if (hfs_resize_debug) {
+               printf("hfs_reclaim_file: === Start reclaiming %sfork for %sid=%u ===\n", (forktype ? "rsrc" : "data"), (extent_info->is_dirlink ? "dirlink" : "file"), fileID);
+       }
+
+       if (extent_info->is_dirlink) {
+               MALLOC(extent_info->dirlink_desc, struct cat_desc *, 
+                               sizeof(struct cat_desc), M_TEMP, M_WAITOK);
+               MALLOC(extent_info->dirlink_attr, struct cat_attr *, 
+                               sizeof(struct cat_attr), M_TEMP, M_WAITOK);
+               MALLOC(extent_info->dirlink_fork, struct filefork *, 
+                               sizeof(struct filefork), M_TEMP, M_WAITOK);
+               if ((extent_info->dirlink_desc == NULL) || 
+                   (extent_info->dirlink_attr == NULL) || 
+                   (extent_info->dirlink_fork == NULL)) {
+                       error = ENOMEM;
+                       goto out;
+               }
+
+               /* Lookup catalog record for directory hard link and 
+                * create a fake filefork for the value looked up from 
+                * the disk. 
+                */
+               fp = extent_info->dirlink_fork;
+               bzero(extent_info->dirlink_fork, sizeof(struct filefork));
+               extent_info->dirlink_fork->ff_cp = cp;
+               lockflags = hfs_systemfile_lock(hfsmp, lockflags, HFS_EXCLUSIVE_LOCK);
+               error = cat_lookup_dirlink(hfsmp, fileID, forktype, 
+                               extent_info->dirlink_desc, extent_info->dirlink_attr, 
+                               &(extent_info->dirlink_fork->ff_data)); 
+               hfs_systemfile_unlock(hfsmp, lockflags);
+               if (error) {
+                       printf ("hfs_reclaim_file: cat_lookup_dirlink for fileID=%u returned error=%u\n", fileID, error);
+                       goto out;
+               }
+               release_desc = true;
+       } else {
+               fp = VTOF(vp);
+       }
+
+       extent_info->catalog_fp = fp;
+       extent_info->recStartBlock = 0;
+       extent_info->extents = extent_info->catalog_fp->ff_extents;
+       /* Relocate extents from the catalog record */
+       for (i = 0; i < kHFSPlusExtentDensity; ++i) {
+               if (fp->ff_extents[i].blockCount == 0) {
+                       break;
+               }
+               extent_info->extent_index = i;
+               error = hfs_reclaim_extent(hfsmp, allocLimit, extent_info, context);
+               if (error) {
+                       printf ("hfs_reclaim_file: fileID=%u #%d %u:(%u,%u) hfs_reclaim_extent error=%d\n", fileID, extent_info->overflow_count, i, fp->ff_extents[i].startBlock, fp->ff_extents[i].blockCount, error);
+                       goto out;
+               }
+       }
+               
+       /* If the number of allocation blocks processed for reclaiming 
+        * are less than total number of blocks for the file, continuing 
+        * working on overflow extents record.
+        */
+       if (fp->ff_blocks <= extent_info->cur_blockCount) {
+               if (0 && hfs_resize_debug) {
+                       printf ("hfs_reclaim_file: Nothing more to relocate, offset=%d, ff_blocks=%u, cur_blockCount=%u\n", i, fp->ff_blocks, extent_info->cur_blockCount);
+               }
+               goto out;
+       }
+
+       if (hfs_resize_debug) {
+               printf ("hfs_reclaim_file: Will check overflow records, offset=%d, ff_blocks=%u, cur_blockCount=%u\n", i, fp->ff_blocks, extent_info->cur_blockCount);
+       }
+
+       MALLOC(extent_info->iterator, struct BTreeIterator *, sizeof(struct BTreeIterator), M_TEMP, M_WAITOK);
+       if (extent_info->iterator == NULL) {
+               error = ENOMEM;
+               goto out;
+       }
+       bzero(extent_info->iterator, sizeof(struct BTreeIterator));
+       key = (HFSPlusExtentKey *) &(extent_info->iterator->key);
+       key->keyLength = kHFSPlusExtentKeyMaximumLength;
+       key->forkType = forktype;
+       key->fileID = fileID;
+       key->startBlock = extent_info->cur_blockCount;
+
+       extent_info->btdata.bufferAddress = extent_info->record.overflow;
+       extent_info->btdata.itemSize = sizeof(HFSPlusExtentRecord);
+       extent_info->btdata.itemCount = 1;
+
+       extent_info->catalog_fp = NULL;
+
+       /* Search the first overflow extent with expected startBlock as 'cur_blockCount' */
+       lockflags = hfs_systemfile_lock(hfsmp, lockflags, HFS_EXCLUSIVE_LOCK);
+       error = BTSearchRecord(extent_info->fcb, extent_info->iterator, 
+                       &(extent_info->btdata), &(extent_info->recordlen), 
+                       extent_info->iterator);
+       hfs_systemfile_unlock(hfsmp, lockflags);
+       while (error == 0) {
+               extent_info->overflow_count++;
+               extent_info->recStartBlock = key->startBlock;
+               extent_info->extents = extent_info->record.overflow;
+               for (i = 0; i < kHFSPlusExtentDensity; i++) {
+                       if (extent_info->record.overflow[i].blockCount == 0) {
+                               goto out;
+                       }
+                       extent_info->extent_index = i;
+                       error = hfs_reclaim_extent(hfsmp, allocLimit, extent_info, context);
+                       if (error) {
+                               printf ("hfs_reclaim_file: fileID=%u #%d %u:(%u,%u) hfs_reclaim_extent error=%d\n", fileID, extent_info->overflow_count, i, extent_info->record.overflow[i].startBlock, extent_info->record.overflow[i].blockCount, error);
+                               goto out;
+                       }
+               }
+
+               /* Look for more overflow records */
+               lockflags = hfs_systemfile_lock(hfsmp, lockflags, HFS_EXCLUSIVE_LOCK);
+               error = BTIterateRecord(extent_info->fcb, kBTreeNextRecord, 
+                               extent_info->iterator, &(extent_info->btdata), 
+                               &(extent_info->recordlen));
+               hfs_systemfile_unlock(hfsmp, lockflags);
+               if (error) {
+                       break;
+               }
+               /* Stop when we encounter a different file or fork. */
+               if ((key->fileID != fileID) || (key->forkType != forktype)) {
+                       break;
+               }
+       }
+       if (error == fsBTRecordNotFoundErr || error == fsBTEndOfIterationErr) {
+               error = 0;
+       }
+       
+out:
+       /* If any blocks were relocated, account them and report progress */
+       if (extent_info->blocks_relocated) {
+               hfsmp->hfs_resize_blocksmoved += extent_info->blocks_relocated;
+               hfs_truncatefs_progress(hfsmp);
+               if (fileID < kHFSFirstUserCatalogNodeID) {
+                       printf ("hfs_reclaim_file: Relocated %u blocks from fileID=%u on \"%s\"\n", 
+                                       extent_info->blocks_relocated, fileID, hfsmp->vcbVN); 
+               }
+       }
+       if (extent_info->iterator) {
+               FREE(extent_info->iterator, M_TEMP);
+       }
+       if (release_desc == true) {
+               cat_releasedesc(extent_info->dirlink_desc);
+       }
+       if (extent_info->dirlink_desc) {
+               FREE(extent_info->dirlink_desc, M_TEMP);
+       }
+       if (extent_info->dirlink_attr) {
+               FREE(extent_info->dirlink_attr, M_TEMP);
+       }
+       if (extent_info->dirlink_fork) {
+               FREE(extent_info->dirlink_fork, M_TEMP);
+       }
+       if ((extent_info->blocks_relocated != 0) && (extent_info->is_sysfile == false)) {
+               (void) hfs_update(vp, MNT_WAIT);
+       }
+       if (took_truncate_lock) {
+               hfs_unlock_truncate(cp, 0);
+       }
+       if (extent_info) {
+               FREE(extent_info, M_TEMP);
+       }
+       if (hfs_resize_debug) {
+               printf("hfs_reclaim_file: === Finished relocating %sfork for fileid=%u (error=%d) ===\n", (forktype ? "rsrc" : "data"), fileID, error);
+       }
 
 
-free_fail:
-       (void) BlockDeallocate(hfsmp, newStartBlock, blockCount);
-fail:
-       (void) hfs_systemfile_unlock(hfsmp, lockflags);
-       (void) hfs_end_transaction(hfsmp);
        return error;
 }
 
        return error;
 }
 
@@ -4424,6 +5911,9 @@ hfs_journal_relocate_callback(void *_args)
                hfsmp->blockSize, vfs_context_ucred(args->context), &bp);
        if (error) {
                printf("hfs_reclaim_journal_file: failed to read JIB (%d)\n", error);
                hfsmp->blockSize, vfs_context_ucred(args->context), &bp);
        if (error) {
                printf("hfs_reclaim_journal_file: failed to read JIB (%d)\n", error);
+               if (bp) {
+                       buf_brelse(bp);
+               }
                return error;
        }
        jibp = (JournalInfoBlock*) buf_dataptr(bp);
                return error;
        }
        jibp = (JournalInfoBlock*) buf_dataptr(bp);
@@ -4449,10 +5939,12 @@ hfs_journal_relocate_callback(void *_args)
 
 
 static int
 
 
 static int
-hfs_reclaim_journal_file(struct hfsmount *hfsmp, vfs_context_t context)
+hfs_reclaim_journal_file(struct hfsmount *hfsmp, u_int32_t allocLimit, vfs_context_t context)
 {
        int error;
 {
        int error;
+       int journal_err;
        int lockflags;
        int lockflags;
+       u_int32_t oldStartBlock;
        u_int32_t newStartBlock;
        u_int32_t oldBlockCount;
        u_int32_t newBlockCount;
        u_int32_t newStartBlock;
        u_int32_t oldBlockCount;
        u_int32_t newBlockCount;
@@ -4461,6 +5953,11 @@ hfs_reclaim_journal_file(struct hfsmount *hfsmp, vfs_context_t context)
        struct cat_fork journal_fork;
        struct hfs_journal_relocate_args callback_args;
 
        struct cat_fork journal_fork;
        struct hfs_journal_relocate_args callback_args;
 
+       if (hfsmp->jnl_start + (hfsmp->jnl_size / hfsmp->blockSize) <= allocLimit) {
+               /* The journal does not require relocation */
+               return 0;
+       }
+
        error = hfs_start_transaction(hfsmp);
        if (error) {
                printf("hfs_reclaim_journal_file: hfs_start_transaction returned %d\n", error);
        error = hfs_start_transaction(hfsmp);
        if (error) {
                printf("hfs_reclaim_journal_file: hfs_start_transaction returned %d\n", error);
@@ -4471,7 +5968,9 @@ hfs_reclaim_journal_file(struct hfsmount *hfsmp, vfs_context_t context)
        oldBlockCount = hfsmp->jnl_size / hfsmp->blockSize;
        
        /* TODO: Allow the journal to change size based on the new volume size. */
        oldBlockCount = hfsmp->jnl_size / hfsmp->blockSize;
        
        /* TODO: Allow the journal to change size based on the new volume size. */
-       error = BlockAllocate(hfsmp, 1, oldBlockCount, oldBlockCount, true, true, &newStartBlock, &newBlockCount);
+       error = BlockAllocate(hfsmp, 1, oldBlockCount, oldBlockCount, 
+                       HFS_ALLOC_METAZONE | HFS_ALLOC_FORCECONTIG | HFS_ALLOC_SKIPFREEBLKS, 
+                        &newStartBlock, &newBlockCount);
        if (error) {
                printf("hfs_reclaim_journal_file: BlockAllocate returned %d\n", error);
                goto fail;
        if (error) {
                printf("hfs_reclaim_journal_file: BlockAllocate returned %d\n", error);
                goto fail;
@@ -4481,7 +5980,7 @@ hfs_reclaim_journal_file(struct hfsmount *hfsmp, vfs_context_t context)
                goto free_fail;
        }
        
                goto free_fail;
        }
        
-       error = BlockDeallocate(hfsmp, hfsmp->jnl_start, oldBlockCount);
+       error = BlockDeallocate(hfsmp, hfsmp->jnl_start, oldBlockCount, HFS_ALLOC_SKIPFREEBLKS);
        if (error) {
                printf("hfs_reclaim_journal_file: BlockDeallocate returned %d\n", error);
                goto free_fail;
        if (error) {
                printf("hfs_reclaim_journal_file: BlockDeallocate returned %d\n", error);
                goto free_fail;
@@ -4493,6 +5992,7 @@ hfs_reclaim_journal_file(struct hfsmount *hfsmp, vfs_context_t context)
                printf("hfs_reclaim_journal_file: cat_idlookup returned %d\n", error);
                goto free_fail;
        }
                printf("hfs_reclaim_journal_file: cat_idlookup returned %d\n", error);
                goto free_fail;
        }
+       oldStartBlock = journal_fork.cf_extents[0].startBlock;
        journal_fork.cf_size = newBlockCount * hfsmp->blockSize;
        journal_fork.cf_extents[0].startBlock = newStartBlock;
        journal_fork.cf_extents[0].blockCount = newBlockCount;
        journal_fork.cf_size = newBlockCount * hfsmp->blockSize;
        journal_fork.cf_extents[0].startBlock = newStartBlock;
        journal_fork.cf_extents[0].blockCount = newBlockCount;
@@ -4524,13 +6024,30 @@ hfs_reclaim_journal_file(struct hfsmount *hfsmp, vfs_context_t context)
                printf("hfs_reclaim_journal_file: hfs_end_transaction returned %d\n", error);
        }
        
                printf("hfs_reclaim_journal_file: hfs_end_transaction returned %d\n", error);
        }
        
+       /* Account for the blocks relocated and print progress */
+       hfsmp->hfs_resize_blocksmoved += oldBlockCount;
+       hfs_truncatefs_progress(hfsmp);
+       if (!error) {
+               printf ("hfs_reclaim_journal_file: Relocated %u blocks from journal on \"%s\"\n", 
+                               oldBlockCount, hfsmp->vcbVN);
+               if (hfs_resize_debug) {
+                       printf ("hfs_reclaim_journal_file: Successfully relocated journal from (%u,%u) to (%u,%u)\n", oldStartBlock, oldBlockCount, newStartBlock, newBlockCount);
+               }
+       }
        return error;
 
 free_fail:
        return error;
 
 free_fail:
-       (void) BlockDeallocate(hfsmp, newStartBlock, newBlockCount);
+       journal_err = BlockDeallocate(hfsmp, newStartBlock, newBlockCount, HFS_ALLOC_SKIPFREEBLKS); 
+       if (journal_err) {
+               printf("hfs_reclaim_journal_file: BlockDeallocate returned %d\n", error);
+               hfs_mark_volume_inconsistent(hfsmp);
+       }
 fail:
        hfs_systemfile_unlock(hfsmp, lockflags);
        (void) hfs_end_transaction(hfsmp);
 fail:
        hfs_systemfile_unlock(hfsmp, lockflags);
        (void) hfs_end_transaction(hfsmp);
+       if (hfs_resize_debug) {
+               printf ("hfs_reclaim_journal_file: Error relocating journal file (error=%d)\n", error);
+       }
        return error;
 }
 
        return error;
 }
 
@@ -4541,16 +6058,23 @@ fail:
  * the field in the volume header and the catalog record.
  */
 static int
  * the field in the volume header and the catalog record.
  */
 static int
-hfs_reclaim_journal_info_block(struct hfsmount *hfsmp, vfs_context_t context)
+hfs_reclaim_journal_info_block(struct hfsmount *hfsmp, u_int32_t allocLimit, vfs_context_t context)
 {
        int error;
 {
        int error;
+       int journal_err;
        int lockflags;
        int lockflags;
+       u_int32_t oldBlock;
        u_int32_t newBlock;
        u_int32_t blockCount;
        struct cat_desc jib_desc;
        struct cat_attr jib_attr;
        struct cat_fork jib_fork;
        buf_t old_bp, new_bp;
        u_int32_t newBlock;
        u_int32_t blockCount;
        struct cat_desc jib_desc;
        struct cat_attr jib_attr;
        struct cat_fork jib_fork;
        buf_t old_bp, new_bp;
+
+       if (hfsmp->vcbJinfoBlock <= allocLimit) {
+               /* The journal info block does not require relocation */
+               return 0;
+       }
        
        error = hfs_start_transaction(hfsmp);
        if (error) {
        
        error = hfs_start_transaction(hfsmp);
        if (error) {
@@ -4559,7 +6083,9 @@ hfs_reclaim_journal_info_block(struct hfsmount *hfsmp, vfs_context_t context)
        }
        lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG | SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
        
        }
        lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG | SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
        
-       error = BlockAllocate(hfsmp, 1, 1, 1, true, true, &newBlock, &blockCount);
+       error = BlockAllocate(hfsmp, 1, 1, 1, 
+                       HFS_ALLOC_METAZONE | HFS_ALLOC_FORCECONTIG | HFS_ALLOC_SKIPFREEBLKS, 
+                       &newBlock, &blockCount);
        if (error) {
                printf("hfs_reclaim_journal_info_block: BlockAllocate returned %d\n", error);
                goto fail;
        if (error) {
                printf("hfs_reclaim_journal_info_block: BlockAllocate returned %d\n", error);
                goto fail;
@@ -4568,7 +6094,7 @@ hfs_reclaim_journal_info_block(struct hfsmount *hfsmp, vfs_context_t context)
                printf("hfs_reclaim_journal_info_block: blockCount != 1 (%u)\n", blockCount);
                goto free_fail;
        }
                printf("hfs_reclaim_journal_info_block: blockCount != 1 (%u)\n", blockCount);
                goto free_fail;
        }
-       error = BlockDeallocate(hfsmp, hfsmp->vcbJinfoBlock, 1);
+       error = BlockDeallocate(hfsmp, hfsmp->vcbJinfoBlock, 1, HFS_ALLOC_SKIPFREEBLKS);
        if (error) {
                printf("hfs_reclaim_journal_info_block: BlockDeallocate returned %d\n", error);
                goto free_fail;
        if (error) {
                printf("hfs_reclaim_journal_info_block: BlockDeallocate returned %d\n", error);
                goto free_fail;
@@ -4580,6 +6106,9 @@ hfs_reclaim_journal_info_block(struct hfsmount *hfsmp, vfs_context_t context)
                hfsmp->blockSize, vfs_context_ucred(context), &old_bp);
        if (error) {
                printf("hfs_reclaim_journal_info_block: failed to read JIB (%d)\n", error);
                hfsmp->blockSize, vfs_context_ucred(context), &old_bp);
        if (error) {
                printf("hfs_reclaim_journal_info_block: failed to read JIB (%d)\n", error);
+               if (old_bp) {
+                       buf_brelse(old_bp);
+               }
                goto free_fail;
        }
        new_bp = buf_getblk(hfsmp->hfs_devvp,
                goto free_fail;
        }
        new_bp = buf_getblk(hfsmp->hfs_devvp,
@@ -4608,6 +6137,7 @@ hfs_reclaim_journal_info_block(struct hfsmount *hfsmp, vfs_context_t context)
                printf("hfs_reclaim_journal_file: cat_idlookup returned %d\n", error);
                goto fail;
        }
                printf("hfs_reclaim_journal_file: cat_idlookup returned %d\n", error);
                goto fail;
        }
+       oldBlock = jib_fork.cf_extents[0].startBlock;
        jib_fork.cf_size = hfsmp->blockSize;
        jib_fork.cf_extents[0].startBlock = newBlock;
        jib_fork.cf_extents[0].blockCount = 1;
        jib_fork.cf_size = hfsmp->blockSize;
        jib_fork.cf_extents[0].startBlock = newBlock;
        jib_fork.cf_extents[0].blockCount = 1;
@@ -4631,340 +6161,654 @@ hfs_reclaim_journal_info_block(struct hfsmount *hfsmp, vfs_context_t context)
        if (error) {
                printf("hfs_reclaim_journal_info_block: hfs_end_transaction returned %d\n", error);
        }
        if (error) {
                printf("hfs_reclaim_journal_info_block: hfs_end_transaction returned %d\n", error);
        }
-       error = hfs_journal_flush(hfsmp);
-       if (error) {
-               printf("hfs_reclaim_journal_info_block: journal_flush returned %d\n", error);
+       error = hfs_journal_flush(hfsmp, FALSE);
+       if (error) {
+               printf("hfs_reclaim_journal_info_block: journal_flush returned %d\n", error);
+       }
+
+       /* Account for the block relocated and print progress */
+       hfsmp->hfs_resize_blocksmoved += 1;
+       hfs_truncatefs_progress(hfsmp);
+       if (!error) {
+               printf ("hfs_reclaim_journal_info: Relocated 1 block from journal info on \"%s\"\n", 
+                               hfsmp->vcbVN);
+               if (hfs_resize_debug) {
+                       printf ("hfs_reclaim_journal_info_block: Successfully relocated journal info block from (%u,%u) to (%u,%u)\n", oldBlock, blockCount, newBlock, blockCount);
+               }
+       }
+       return error;
+
+free_fail:
+       journal_err = BlockDeallocate(hfsmp, newBlock, blockCount, HFS_ALLOC_SKIPFREEBLKS); 
+       if (journal_err) {
+               printf("hfs_reclaim_journal_info_block: BlockDeallocate returned %d\n", error);
+               hfs_mark_volume_inconsistent(hfsmp);
+       }
+
+fail:
+       hfs_systemfile_unlock(hfsmp, lockflags);
+       (void) hfs_end_transaction(hfsmp);
+       if (hfs_resize_debug) {
+               printf ("hfs_reclaim_journal_info_block: Error relocating journal info block (error=%d)\n", error);
+       }
+       return error;
+}
+
+
+/*
+ * This function traverses through all extended attribute records for a given 
+ * fileID, and calls function that reclaims data blocks that exist in the 
+ * area of the disk being reclaimed which in turn is responsible for allocating 
+ * new space, copying extent data, deallocating new space, and if required, 
+ * splitting the extent.
+ *
+ * Note: The caller has already acquired the cnode lock on the file.  Therefore
+ * we are assured that no other thread would be creating/deleting/modifying 
+ * extended attributes for this file.  
+ *
+ * Side Effects:
+ * hfsmp->hfs_resize_blocksmoved is incremented by the number of allocation 
+ * blocks that were relocated. 
+ *
+ * Returns: 
+ *     0 on success, non-zero on failure.
+ */
+static int 
+hfs_reclaim_xattr(struct hfsmount *hfsmp, struct vnode *vp, u_int32_t fileID, u_int32_t allocLimit, vfs_context_t context) 
+{
+       int error = 0;
+       struct hfs_reclaim_extent_info *extent_info;
+       int i;
+       HFSPlusAttrKey *key;
+       int *lockflags;
+
+       if (hfs_resize_debug) {
+               printf("hfs_reclaim_xattr: === Start reclaiming xattr for id=%u ===\n", fileID);
+       }
+
+       MALLOC(extent_info, struct hfs_reclaim_extent_info *, 
+              sizeof(struct hfs_reclaim_extent_info), M_TEMP, M_WAITOK);
+       if (extent_info == NULL) {
+               return ENOMEM;
+       }
+       bzero(extent_info, sizeof(struct hfs_reclaim_extent_info));
+       extent_info->vp = vp;
+       extent_info->fileID = fileID;
+       extent_info->is_xattr = true;
+       extent_info->is_sysfile = vnode_issystem(vp);
+       extent_info->fcb = VTOF(hfsmp->hfs_attribute_vp);
+       lockflags = &(extent_info->lockflags);
+       *lockflags = SFL_ATTRIBUTE | SFL_BITMAP;
+
+       /* Initialize iterator from the extent_info structure */
+       MALLOC(extent_info->iterator, struct BTreeIterator *, 
+              sizeof(struct BTreeIterator), M_TEMP, M_WAITOK);
+       if (extent_info->iterator == NULL) {
+               error = ENOMEM;
+               goto out;
+       }
+       bzero(extent_info->iterator, sizeof(struct BTreeIterator));
+
+       /* Build attribute key */
+       key = (HFSPlusAttrKey *)&(extent_info->iterator->key);
+       error = hfs_buildattrkey(fileID, NULL, key);
+       if (error) {
+               goto out;
+       }
+
+       /* Initialize btdata from extent_info structure.  Note that the 
+        * buffer pointer actually points to the xattr record from the 
+        * extent_info structure itself.
+        */
+       extent_info->btdata.bufferAddress = &(extent_info->record.xattr);
+       extent_info->btdata.itemSize = sizeof(HFSPlusAttrRecord);
+       extent_info->btdata.itemCount = 1;
+
+       /* 
+        * Sync all extent-based attribute data to the disk.
+        *
+        * All extent-based attribute data I/O is performed via cluster 
+        * I/O using a virtual file that spans across entire file system 
+        * space.  
+        */
+       hfs_lock_truncate(VTOC(hfsmp->hfs_attrdata_vp), HFS_EXCLUSIVE_LOCK);
+       (void)cluster_push(hfsmp->hfs_attrdata_vp, 0);
+       error = vnode_waitforwrites(hfsmp->hfs_attrdata_vp, 0, 0, 0, "hfs_reclaim_xattr");
+       hfs_unlock_truncate(VTOC(hfsmp->hfs_attrdata_vp), 0);
+       if (error) {
+               goto out;
+       }
+
+       /* Search for extended attribute for current file.  This 
+        * will place the iterator before the first matching record.
+        */
+       *lockflags = hfs_systemfile_lock(hfsmp, *lockflags, HFS_EXCLUSIVE_LOCK);
+       error = BTSearchRecord(extent_info->fcb, extent_info->iterator, 
+                       &(extent_info->btdata), &(extent_info->recordlen), 
+                       extent_info->iterator);
+       hfs_systemfile_unlock(hfsmp, *lockflags);
+       if (error) {
+               if (error != btNotFound) {
+                       goto out;
+               }
+               /* btNotFound is expected here, so just mask it */
+               error = 0;
+       } 
+
+       while (1) {
+               /* Iterate to the next record */
+               *lockflags = hfs_systemfile_lock(hfsmp, *lockflags, HFS_EXCLUSIVE_LOCK);
+               error = BTIterateRecord(extent_info->fcb, kBTreeNextRecord, 
+                               extent_info->iterator, &(extent_info->btdata), 
+                               &(extent_info->recordlen));
+               hfs_systemfile_unlock(hfsmp, *lockflags);
+
+               /* Stop the iteration if we encounter end of btree or xattr with different fileID */
+               if (error || key->fileID != fileID) {
+                       if (error == fsBTRecordNotFoundErr || error == fsBTEndOfIterationErr) {
+                               error = 0;                              
+                       }
+                       break;
+               }
+
+               /* We only care about extent-based EAs */
+               if ((extent_info->record.xattr.recordType != kHFSPlusAttrForkData) && 
+                   (extent_info->record.xattr.recordType != kHFSPlusAttrExtents)) {
+                       continue;
+               }
+
+               if (extent_info->record.xattr.recordType == kHFSPlusAttrForkData) {
+                       extent_info->overflow_count = 0;
+                       extent_info->extents = extent_info->record.xattr.forkData.theFork.extents;
+               } else if (extent_info->record.xattr.recordType == kHFSPlusAttrExtents) {
+                       extent_info->overflow_count++;
+                       extent_info->extents = extent_info->record.xattr.overflowExtents.extents;
+               }
+                       
+               extent_info->recStartBlock = key->startBlock;
+               for (i = 0; i < kHFSPlusExtentDensity; i++) {
+                       if (extent_info->extents[i].blockCount == 0) {
+                               break;
+                       } 
+                       extent_info->extent_index = i;
+                       error = hfs_reclaim_extent(hfsmp, allocLimit, extent_info, context);
+                       if (error) {
+                               printf ("hfs_reclaim_xattr: fileID=%u hfs_reclaim_extent error=%d\n", fileID, error); 
+                               goto out;
+                       }
+               }
+       }
+
+out:
+       /* If any blocks were relocated, account them and report progress */
+       if (extent_info->blocks_relocated) {
+               hfsmp->hfs_resize_blocksmoved += extent_info->blocks_relocated;
+               hfs_truncatefs_progress(hfsmp);
+       }
+       if (extent_info->iterator) {
+               FREE(extent_info->iterator, M_TEMP);
+       }
+       if (extent_info) {
+               FREE(extent_info, M_TEMP);
+       }
+       if (hfs_resize_debug) {
+               printf("hfs_reclaim_xattr: === Finished relocating xattr for fileid=%u (error=%d) ===\n", fileID, error);
+       }
+       return error;
+}
+
+/* 
+ * Reclaim any extent-based extended attributes allocation blocks from 
+ * the area of the disk that is being truncated.
+ *
+ * The function traverses the attribute btree to find out the fileIDs
+ * of the extended attributes that need to be relocated.  For every 
+ * file whose large EA requires relocation, it looks up the cnode and 
+ * calls hfs_reclaim_xattr() to do all the work for allocating 
+ * new space, copying data, deallocating old space, and if required, 
+ * splitting the extents.
+ *
+ * Inputs: 
+ *     allocLimit    - starting block of the area being reclaimed
+ *
+ * Returns:
+ *     returns 0 on success, non-zero on failure.
+ */
+static int
+hfs_reclaim_xattrspace(struct hfsmount *hfsmp, u_int32_t allocLimit, vfs_context_t context)
+{
+       int error = 0;
+       FCB *fcb;
+       struct BTreeIterator *iterator = NULL;
+       struct FSBufferDescriptor btdata;
+       HFSPlusAttrKey *key;
+       HFSPlusAttrRecord rec;
+       int lockflags = 0;
+       cnid_t prev_fileid = 0;
+       struct vnode *vp;
+       int need_relocate;
+       int btree_operation;
+       u_int32_t files_moved = 0;
+       u_int32_t prev_blocksmoved;
+       int i;
+
+       fcb = VTOF(hfsmp->hfs_attribute_vp);
+       /* Store the value to print total blocks moved by this function in end */
+       prev_blocksmoved = hfsmp->hfs_resize_blocksmoved;
+
+       if (kmem_alloc(kernel_map, (vm_offset_t *)&iterator, sizeof(*iterator))) {
+               return ENOMEM;
+       }       
+       bzero(iterator, sizeof(*iterator));
+       key = (HFSPlusAttrKey *)&iterator->key;
+       btdata.bufferAddress = &rec;
+       btdata.itemSize = sizeof(rec);
+       btdata.itemCount = 1;
+
+       need_relocate = false;
+       btree_operation = kBTreeFirstRecord;
+       /* Traverse the attribute btree to find extent-based EAs to reclaim */
+       while (1) {
+               lockflags = hfs_systemfile_lock(hfsmp, SFL_ATTRIBUTE, HFS_SHARED_LOCK);
+               error = BTIterateRecord(fcb, btree_operation, iterator, &btdata, NULL);
+               hfs_systemfile_unlock(hfsmp, lockflags);
+               if (error) {
+                       if (error == fsBTRecordNotFoundErr || error == fsBTEndOfIterationErr) {
+                               error = 0;                              
+                       }
+                       break;
+               }
+               btree_operation = kBTreeNextRecord;
+
+               /* If the extents of current fileID were already relocated, skip it */
+               if (prev_fileid == key->fileID) {
+                       continue;
+               }
+
+               /* Check if any of the extents in the current record need to be relocated */
+               need_relocate = false;
+               switch(rec.recordType) {
+                       case kHFSPlusAttrForkData:
+                               for (i = 0; i < kHFSPlusExtentDensity; i++) {
+                                       if (rec.forkData.theFork.extents[i].blockCount == 0) {
+                                               break;
+                                       }
+                                       if ((rec.forkData.theFork.extents[i].startBlock + 
+                                            rec.forkData.theFork.extents[i].blockCount) > allocLimit) {
+                                               need_relocate = true;
+                                               break;
+                                       }
+                               }
+                               break;
+
+                       case kHFSPlusAttrExtents:
+                               for (i = 0; i < kHFSPlusExtentDensity; i++) {
+                                       if (rec.overflowExtents.extents[i].blockCount == 0) {
+                                               break;
+                                       }
+                                       if ((rec.overflowExtents.extents[i].startBlock + 
+                                            rec.overflowExtents.extents[i].blockCount) > allocLimit) {
+                                               need_relocate = true;
+                                               break;
+                                       }
+                               }
+                               break;
+               };
+
+               /* Continue iterating to next attribute record */
+               if (need_relocate == false) {
+                       continue;
+               }
+
+               /* Look up the vnode for corresponding file.  The cnode 
+                * will be locked which will ensure that no one modifies 
+                * the xattrs when we are relocating them.
+                *
+                * We want to allow open-unlinked files to be moved, 
+                * so provide allow_deleted == 1 for hfs_vget().
+                */
+               if (hfs_vget(hfsmp, key->fileID, &vp, 0, 1) != 0) {
+                       continue;
+               }
+
+               error = hfs_reclaim_xattr(hfsmp, vp, key->fileID, allocLimit, context);
+               hfs_unlock(VTOC(vp));
+               vnode_put(vp);
+               if (error) {
+                       printf ("hfs_reclaim_xattrspace: Error relocating xattrs for fileid=%u (error=%d)\n", key->fileID, error);
+                       break;
+               }
+               prev_fileid = key->fileID;
+               files_moved++;
+       }
+
+       if (files_moved) {
+               printf("hfs_reclaim_xattrspace: Relocated %u xattr blocks from %u files on \"%s\"\n", 
+                               (hfsmp->hfs_resize_blocksmoved - prev_blocksmoved),
+                               files_moved, hfsmp->vcbVN);
        }
        }
-       return error;
 
 
-free_fail:
-       (void) BlockDeallocate(hfsmp, newBlock, blockCount);
-fail:
-       hfs_systemfile_unlock(hfsmp, lockflags);
-       (void) hfs_end_transaction(hfsmp);
+       kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
        return error;
 }
 
        return error;
 }
 
-
-/*
- * Reclaim space at the end of a file system.
+/* 
+ * Reclaim blocks from regular files.
+ *
+ * This function iterates over all the record in catalog btree looking 
+ * for files with extents that overlap into the space we're trying to 
+ * free up.  If a file extent requires relocation, it looks up the vnode 
+ * and calls function to relocate the data.
+ *
+ * Returns:
+ *     Zero on success, non-zero on failure. 
  */
  */
-static int
-hfs_reclaimspace(struct hfsmount *hfsmp, u_int32_t startblk, u_int32_t reclaimblks, vfs_context_t context)
+static int 
+hfs_reclaim_filespace(struct hfsmount *hfsmp, u_int32_t allocLimit, vfs_context_t context) 
 {
 {
-       struct vnode *vp = NULL;
+       int error;
        FCB *fcb;
        FCB *fcb;
-       struct BTreeIterator * iterator = NULL;
+       struct BTreeIterator *iterator = NULL;
        struct FSBufferDescriptor btdata;
        struct FSBufferDescriptor btdata;
-       struct HFSPlusCatalogFile filerec;
-       u_int32_t  saved_next_allocation;
-       cnid_t * cnidbufp;
-       size_t cnidbufsize;
-       int filecnt = 0;
-       int maxfilecnt;
-       u_int32_t block;
-       u_int32_t datablks;
-       u_int32_t rsrcblks;
-       u_int32_t blkstomove = 0;
+       int btree_operation;
        int lockflags;
        int lockflags;
-       int i;
-       int error;
-       int lastprogress = 0;
-       Boolean system_file_moved = false;
+       struct HFSPlusCatalogFile filerec;
+       struct vnode *vp;
+       struct vnode *rvp;
+       struct filefork *datafork;
+       u_int32_t files_moved = 0;
+       u_int32_t prev_blocksmoved;
 
 
-       /* Relocate extents of the Allocation file if they're in the way. */
-       error = hfs_reclaim_sys_file(hfsmp, hfsmp->hfs_allocation_vp, startblk, SFL_BITMAP, &system_file_moved, context);
-       if (error) {
-               printf("hfs_reclaimspace: reclaim allocation file returned %d\n", error);
-               return error;
-       }
-       /* Relocate extents of the Extents B-tree if they're in the way. */
-       error = hfs_reclaim_sys_file(hfsmp, hfsmp->hfs_extents_vp, startblk, SFL_EXTENTS, &system_file_moved, context);
-       if (error) {
-               printf("hfs_reclaimspace: reclaim extents b-tree returned %d\n", error);
-               return error;
-       }
-       /* Relocate extents of the Catalog B-tree if they're in the way. */
-       error = hfs_reclaim_sys_file(hfsmp, hfsmp->hfs_catalog_vp, startblk, SFL_CATALOG, &system_file_moved, context);
-       if (error) {
-               printf("hfs_reclaimspace: reclaim catalog b-tree returned %d\n", error);
-               return error;
-       }
-       /* Relocate extents of the Attributes B-tree if they're in the way. */
-       error = hfs_reclaim_sys_file(hfsmp, hfsmp->hfs_attribute_vp, startblk, SFL_ATTRIBUTE, &system_file_moved, context);
-       if (error) {
-               printf("hfs_reclaimspace: reclaim attribute b-tree returned %d\n", error);
-               return error;
-       }
-       /* Relocate extents of the Startup File if there is one and they're in the way. */
-       error = hfs_reclaim_sys_file(hfsmp, hfsmp->hfs_startup_vp, startblk, SFL_STARTUP, &system_file_moved, context);
-       if (error) {
-               printf("hfs_reclaimspace: reclaim startup file returned %d\n", error);
-               return error;
-       }
-       
-       /*
-        * We need to make sure the alternate volume header gets flushed if we moved
-        * any extents in the volume header.  But we need to do that before
-        * shrinking the size of the volume, or else the journal code will panic
-        * with an invalid (too large) block number.
-        *
-        * Note that system_file_moved will be set if ANY extent was moved, even
-        * if it was just an overflow extent.  In this case, the journal_flush isn't
-        * strictly required, but shouldn't hurt.
-        */
-       if (system_file_moved)
-               hfs_journal_flush(hfsmp);
+       fcb = VTOF(hfsmp->hfs_catalog_vp);
+       /* Store the value to print total blocks moved by this function at the end */
+       prev_blocksmoved = hfsmp->hfs_resize_blocksmoved;
 
 
-       if (hfsmp->jnl_start + (hfsmp->jnl_size / hfsmp->blockSize) > startblk) {
-               error = hfs_reclaim_journal_file(hfsmp, context);
-               if (error) {
-                       printf("hfs_reclaimspace: hfs_reclaim_journal_file failed (%d)\n", error);
-                       return error;
-               }
-       }
-       
-       if (hfsmp->vcbJinfoBlock >= startblk) {
-               error = hfs_reclaim_journal_info_block(hfsmp, context);
-               if (error) {
-                       printf("hfs_reclaimspace: hfs_reclaim_journal_info_block failed (%d)\n", error);
-                       return error;
-               }
-       }
-       
-       /* For now move a maximum of 250,000 files. */
-       maxfilecnt = MIN(hfsmp->hfs_filecount, 250000);
-       maxfilecnt = MIN((u_int32_t)maxfilecnt, reclaimblks);
-       cnidbufsize = maxfilecnt * sizeof(cnid_t);
-       if (kmem_alloc(kernel_map, (vm_offset_t *)&cnidbufp, cnidbufsize)) {
-               return (ENOMEM);
-       }       
        if (kmem_alloc(kernel_map, (vm_offset_t *)&iterator, sizeof(*iterator))) {
        if (kmem_alloc(kernel_map, (vm_offset_t *)&iterator, sizeof(*iterator))) {
-               kmem_free(kernel_map, (vm_offset_t)cnidbufp, cnidbufsize);
-               return (ENOMEM);
-       }       
-
-       saved_next_allocation = hfsmp->nextAllocation;
-       HFS_UPDATE_NEXT_ALLOCATION(hfsmp, hfsmp->hfs_metazone_start);
-
-       fcb = VTOF(hfsmp->hfs_catalog_vp);
+               return ENOMEM;
+       }
        bzero(iterator, sizeof(*iterator));
 
        btdata.bufferAddress = &filerec;
        btdata.itemSize = sizeof(filerec);
        btdata.itemCount = 1;
 
        bzero(iterator, sizeof(*iterator));
 
        btdata.bufferAddress = &filerec;
        btdata.itemSize = sizeof(filerec);
        btdata.itemCount = 1;
 
-       /* Keep the Catalog and extents files locked during iteration. */
-       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG | SFL_EXTENTS, HFS_SHARED_LOCK);
-
-       error = BTIterateRecord(fcb, kBTreeFirstRecord, iterator, NULL, NULL);
-       if (error) {
-               goto end_iteration;
-       }
-       /*
-        * Iterate over all the catalog records looking for files
-        * that overlap into the space we're trying to free up.
-        */
-       for (filecnt = 0; filecnt < maxfilecnt; ) {
-               error = BTIterateRecord(fcb, kBTreeNextRecord, iterator, &btdata, NULL);
+       btree_operation = kBTreeFirstRecord;
+       while (1) {
+               lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_SHARED_LOCK);
+               error = BTIterateRecord(fcb, btree_operation, iterator, &btdata, NULL);
+               hfs_systemfile_unlock(hfsmp, lockflags);
                if (error) {
                        if (error == fsBTRecordNotFoundErr || error == fsBTEndOfIterationErr) {
                                error = 0;                              
                        }
                        break;
                }
                if (error) {
                        if (error == fsBTRecordNotFoundErr || error == fsBTEndOfIterationErr) {
                                error = 0;                              
                        }
                        break;
                }
+               btree_operation = kBTreeNextRecord;
+
                if (filerec.recordType != kHFSPlusFileRecord) {
                        continue;
                }
                if (filerec.recordType != kHFSPlusFileRecord) {
                        continue;
                }
-               datablks = rsrcblks = 0;
-               /* 
-                * Check if either fork overlaps target space.
-                */
-               for (i = 0; i < kHFSPlusExtentDensity; ++i) {
-                       if (filerec.dataFork.extents[i].blockCount != 0) {
-                               datablks += filerec.dataFork.extents[i].blockCount;
-                               block = filerec.dataFork.extents[i].startBlock +
-                                               filerec.dataFork.extents[i].blockCount;
-                               if (block >= startblk) {
-                                       if ((filerec.fileID == hfsmp->hfs_jnlfileid) ||
-                                               (filerec.fileID == hfsmp->hfs_jnlinfoblkid)) {
-                                               printf("hfs_reclaimspace: cannot move active journal\n");
-                                               error = EPERM;
-                                               goto end_iteration;
-                                       }
-                                       cnidbufp[filecnt++] = filerec.fileID;
-                                       blkstomove += filerec.dataFork.totalBlocks;
-                                       break;
-                               }
+
+               /* Check if any of the extents require relocation */
+               if (hfs_file_extent_overlaps(hfsmp, allocLimit, &filerec) == false) {
+                       continue;
+               }
+
+               /* We want to allow open-unlinked files to be moved, so allow_deleted == 1 */
+               if (hfs_vget(hfsmp, filerec.fileID, &vp, 0, 1) != 0) {
+                       continue;
+               }
+
+               /* If data fork exists or item is a directory hard link, relocate blocks */
+               datafork = VTOF(vp);
+               if ((datafork && datafork->ff_blocks > 0) || vnode_isdir(vp)) {
+                       error = hfs_reclaim_file(hfsmp, vp, filerec.fileID, 
+                                       kHFSDataForkType, allocLimit, context);
+                       if (error)  {
+                               printf ("hfs_reclaimspace: Error reclaiming datafork blocks of fileid=%u (error=%d)\n", filerec.fileID, error);
+                               hfs_unlock(VTOC(vp));
+                               vnode_put(vp);
+                               break;
                        }
                        }
-                       if (filerec.resourceFork.extents[i].blockCount != 0) {
-                               rsrcblks += filerec.resourceFork.extents[i].blockCount;
-                               block = filerec.resourceFork.extents[i].startBlock +
-                                               filerec.resourceFork.extents[i].blockCount;
-                               if (block >= startblk) {
-                                       cnidbufp[filecnt++] = filerec.fileID;
-                                       blkstomove += filerec.resourceFork.totalBlocks;
+               }
+
+               /* If resource fork exists or item is a directory hard link, relocate blocks */
+               if (((VTOC(vp)->c_blocks - (datafork ? datafork->ff_blocks : 0)) > 0) || vnode_isdir(vp)) {
+                       if (vnode_isdir(vp)) {
+                               /* Resource fork vnode lookup is invalid for directory hard link. 
+                                * So we fake data fork vnode as resource fork vnode.
+                                */
+                               rvp = vp;
+                       } else {
+                               error = hfs_vgetrsrc(hfsmp, vp, &rvp, TRUE, FALSE);
+                               if (error) {
+                                       printf ("hfs_reclaimspace: Error looking up rvp for fileid=%u (error=%d)\n", filerec.fileID, error);
+                                       hfs_unlock(VTOC(vp));
+                                       vnode_put(vp);
                                        break;
                                }
                                        break;
                                }
+                               VTOC(rvp)->c_flag |= C_NEED_RVNODE_PUT;
                        }
                        }
-               }
-               /*
-                * Check for any overflow extents that overlap.
-                */
-               if (i == kHFSPlusExtentDensity) {
-                       if (filerec.dataFork.totalBlocks > datablks) {
-                               if (hfs_overlapped_overflow_extents(hfsmp, startblk, datablks, filerec.fileID, 0)) {
-                                       cnidbufp[filecnt++] = filerec.fileID;
-                                       blkstomove += filerec.dataFork.totalBlocks;
-                               }
-                       } else if (filerec.resourceFork.totalBlocks > rsrcblks) {
-                               if (hfs_overlapped_overflow_extents(hfsmp, startblk, rsrcblks, filerec.fileID, 1)) {
-                                       cnidbufp[filecnt++] = filerec.fileID;
-                                       blkstomove += filerec.resourceFork.totalBlocks;
-                               }
+
+                       error = hfs_reclaim_file(hfsmp, rvp, filerec.fileID, 
+                                       kHFSResourceForkType, allocLimit, context);
+                       if (error) {
+                               printf ("hfs_reclaimspace: Error reclaiming rsrcfork blocks of fileid=%u (error=%d)\n", filerec.fileID, error);
+                               hfs_unlock(VTOC(vp));
+                               vnode_put(vp);
+                               break;
                        }
                }
                        }
                }
+
+               /* The file forks were relocated successfully, now drop the 
+                * cnode lock and vnode reference, and continue iterating to 
+                * next catalog record.
+                */
+               hfs_unlock(VTOC(vp));
+               vnode_put(vp);
+               files_moved++;
        }
 
        }
 
-end_iteration:
-       if (filecnt == 0 && !system_file_moved) {
-               printf("hfs_reclaimspace: no files moved\n");
-               error = ENOSPC;
+       if (files_moved) {
+               printf("hfs_reclaim_filespace: Relocated %u blocks from %u files on \"%s\"\n", 
+                               (hfsmp->hfs_resize_blocksmoved - prev_blocksmoved),
+                               files_moved, hfsmp->vcbVN);
        }
        }
-       /* All done with catalog. */
-       hfs_systemfile_unlock(hfsmp, lockflags);
-       if (error || filecnt == 0)
-               goto out;
 
 
-       /*
-        * Double check space requirements to make sure
-        * there is enough space to relocate any files
-        * that reside in the reclaim area.
-        *
-        *                                          Blocks To Move --------------
-        *                                                            |    |    |
-        *                                                            V    V    V
-        * ------------------------------------------------------------------------
-        * |                                                        | /   ///  // |
-        * |                                                        | /   ///  // |
-        * |                                                        | /   ///  // |
-        * ------------------------------------------------------------------------
-        *
-        * <------------------- New Total Blocks ------------------><-- Reclaim -->
-        *
-        * <------------------------ Original Total Blocks ----------------------->
+       kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
+       return error;
+}
+
+/*
+ * Reclaim space at the end of a file system.
+ *
+ * Inputs - 
+ *     allocLimit      - start block of the space being reclaimed
+ *     reclaimblks     - number of allocation blocks to reclaim
+ */
+static int
+hfs_reclaimspace(struct hfsmount *hfsmp, u_int32_t allocLimit, u_int32_t reclaimblks, vfs_context_t context)
+{
+       int error = 0;
+
+       /* 
+        * Preflight the bitmap to find out total number of blocks that need 
+        * relocation. 
         *
         *
+        * Note: Since allocLimit is set to the location of new alternate volume 
+        * header, the check below does not account for blocks allocated for old 
+        * alternate volume header.
         */
         */
-       if (blkstomove >= hfs_freeblks(hfsmp, 1)) {
-               printf("hfs_truncatefs: insufficient space (need %u blocks; have %u blocks)\n", blkstomove, hfs_freeblks(hfsmp, 1));
-               error = ENOSPC;
-               goto out;
+       error = hfs_count_allocated(hfsmp, allocLimit, reclaimblks, &(hfsmp->hfs_resize_totalblocks));
+       if (error) {
+               printf ("hfs_reclaimspace: Unable to determine total blocks to reclaim error=%d\n", error);
+               return error;
+       }
+       if (hfs_resize_debug) {
+               printf ("hfs_reclaimspace: Total number of blocks to reclaim = %u\n", hfsmp->hfs_resize_totalblocks);
        }
        }
-       hfsmp->hfs_resize_filesmoved = 0;
-       hfsmp->hfs_resize_totalfiles = filecnt;
-       
-       /* Now move any files that are in the way. */
-       for (i = 0; i < filecnt; ++i) {
-               struct vnode * rvp;
-        struct cnode * cp;
 
 
-               if (hfs_vget(hfsmp, cnidbufp[i], &vp, 0) != 0)
-                       continue;
+       /* Just to be safe, sync the content of the journal to the disk before we proceed */
+       hfs_journal_flush(hfsmp, TRUE);
 
 
-        /* Relocating directory hard links is not supported, so we
-         * punt (see radar 6217026). */
-        cp = VTOC(vp);
-        if ((cp->c_flag & C_HARDLINK) && vnode_isdir(vp)) {
-            printf("hfs_reclaimspace: unable to relocate directory hard link %d\n", cp->c_cnid);
-            error = EINVAL;
-            goto out;
-        }
+       /* First, relocate journal file blocks if they're in the way.  
+        * Doing this first will make sure that journal relocate code 
+        * gets access to contiguous blocks on disk first.  The journal
+        * file has to be contiguous on the disk, otherwise resize will 
+        * fail. 
+        */
+       error = hfs_reclaim_journal_file(hfsmp, allocLimit, context);
+       if (error) {
+               printf("hfs_reclaimspace: hfs_reclaim_journal_file failed (%d)\n", error);
+               return error;
+       }
+       
+       /* Relocate journal info block blocks if they're in the way. */
+       error = hfs_reclaim_journal_info_block(hfsmp, allocLimit, context);
+       if (error) {
+               printf("hfs_reclaimspace: hfs_reclaim_journal_info_block failed (%d)\n", error);
+               return error;
+       }
 
 
-               /* Relocate any data fork blocks. */
-               if (VTOF(vp) && VTOF(vp)->ff_blocks > 0) {
-                       error = hfs_relocate(vp, hfsmp->hfs_metazone_end + 1, kauth_cred_get(), current_proc());
-               }
-               if (error) 
-                       break;
+       /* Relocate extents of the Extents B-tree if they're in the way.
+        * Relocating extents btree before other btrees is important as 
+        * this will provide access to largest contiguous block range on 
+        * the disk for relocating extents btree.  Note that extents btree 
+        * can only have maximum of 8 extents.
+        */
+       error = hfs_reclaim_file(hfsmp, hfsmp->hfs_extents_vp, kHFSExtentsFileID, 
+                       kHFSDataForkType, allocLimit, context);
+       if (error) {
+               printf("hfs_reclaimspace: reclaim extents b-tree returned %d\n", error);
+               return error;
+       }
 
 
-               /* Relocate any resource fork blocks. */
-               if ((cp->c_blocks - (VTOF(vp) ? VTOF((vp))->ff_blocks : 0)) > 0) {
-                       error = hfs_vgetrsrc(hfsmp, vp, &rvp, TRUE);
-                       if (error)
-                               break;
-                       error = hfs_relocate(rvp, hfsmp->hfs_metazone_end + 1, kauth_cred_get(), current_proc());
-                       VTOC(rvp)->c_flag |= C_NEED_RVNODE_PUT;
-                       if (error)
-                               break;
-               }
-               hfs_unlock(cp);
-               vnode_put(vp);
-               vp = NULL;
+       /* Relocate extents of the Allocation file if they're in the way. */
+       error = hfs_reclaim_file(hfsmp, hfsmp->hfs_allocation_vp, kHFSAllocationFileID, 
+                       kHFSDataForkType, allocLimit, context);
+       if (error) {
+               printf("hfs_reclaimspace: reclaim allocation file returned %d\n", error);
+               return error;
+       }
 
 
-               ++hfsmp->hfs_resize_filesmoved;
+       /* Relocate extents of the Catalog B-tree if they're in the way. */
+       error = hfs_reclaim_file(hfsmp, hfsmp->hfs_catalog_vp, kHFSCatalogFileID, 
+                       kHFSDataForkType, allocLimit, context);
+       if (error) {
+               printf("hfs_reclaimspace: reclaim catalog b-tree returned %d\n", error);
+               return error;
+       }
 
 
-               /* Report intermediate progress. */
-               if (filecnt > 100) {
-                       int progress;
+       /* Relocate extents of the Attributes B-tree if they're in the way. */
+       error = hfs_reclaim_file(hfsmp, hfsmp->hfs_attribute_vp, kHFSAttributesFileID, 
+                       kHFSDataForkType, allocLimit, context);
+       if (error) {
+               printf("hfs_reclaimspace: reclaim attribute b-tree returned %d\n", error);
+               return error;
+       }
 
 
-                       progress = (i * 100) / filecnt;
-                       if (progress > (lastprogress + 9)) {
-                               printf("hfs_reclaimspace: %d%% done...\n", progress);
-                               lastprogress = progress;
-                       }
-               }
+       /* Relocate extents of the Startup File if there is one and they're in the way. */
+       error = hfs_reclaim_file(hfsmp, hfsmp->hfs_startup_vp, kHFSStartupFileID, 
+                       kHFSDataForkType, allocLimit, context);
+       if (error) {
+               printf("hfs_reclaimspace: reclaim startup file returned %d\n", error);
+               return error;
        }
        }
-       if (vp) {
-               hfs_unlock(VTOC(vp));
-               vnode_put(vp);
-               vp = NULL;
+       
+       /*
+        * We need to make sure the alternate volume header gets flushed if we moved
+        * any extents in the volume header.  But we need to do that before
+        * shrinking the size of the volume, or else the journal code will panic
+        * with an invalid (too large) block number.
+        *
+        * Note that blks_moved will be set if ANY extent was moved, even
+        * if it was just an overflow extent.  In this case, the journal_flush isn't
+        * strictly required, but shouldn't hurt.
+        */
+       if (hfsmp->hfs_resize_blocksmoved) {
+               hfs_journal_flush(hfsmp, TRUE);
        }
        }
-       if (hfsmp->hfs_resize_filesmoved != 0) {
-               printf("hfs_reclaimspace: relocated %d files on \"%s\"\n",
-                      (int)hfsmp->hfs_resize_filesmoved, hfsmp->vcbVN);
+
+       /* Reclaim extents from catalog file records */
+       error = hfs_reclaim_filespace(hfsmp, allocLimit, context);
+       if (error) {
+               printf ("hfs_reclaimspace: hfs_reclaim_filespace returned error=%d\n", error);
+               return error;
        }
        }
-out:
-       kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
-       kmem_free(kernel_map, (vm_offset_t)cnidbufp, cnidbufsize);
 
 
-       /*
-        * Restore the roving allocation pointer on errors.
-        * (but only if we didn't move any files)
-        */
-       if (error && hfsmp->hfs_resize_filesmoved == 0) {
-               HFS_UPDATE_NEXT_ALLOCATION(hfsmp, saved_next_allocation);
+       /* Reclaim extents from extent-based extended attributes, if any */
+       error = hfs_reclaim_xattrspace(hfsmp, allocLimit, context);
+       if (error) {
+               printf ("hfs_reclaimspace: hfs_reclaim_xattrspace returned error=%d\n", error);
+               return error;
        }
        }
-       return (error);
+
+       return error;
 }
 
 
 /*
 }
 
 
 /*
- * Check if there are any overflow extents that overlap.
+ * Check if there are any extents (including overflow extents) that overlap 
+ * into the disk space that is being reclaimed.  
+ *
+ * Output - 
+ *     true  - One of the extents need to be relocated
+ *     false - No overflow extents need to be relocated, or there was an error
  */
 static int
  */
 static int
-hfs_overlapped_overflow_extents(struct hfsmount *hfsmp, u_int32_t startblk, u_int32_t catblks, u_int32_t fileID, int rsrcfork)
+hfs_file_extent_overlaps(struct hfsmount *hfsmp, u_int32_t allocLimit, struct HFSPlusCatalogFile *filerec)
 {
        struct BTreeIterator * iterator = NULL;
        struct FSBufferDescriptor btdata;
        HFSPlusExtentRecord extrec;
        HFSPlusExtentKey *extkeyptr;
        FCB *fcb;
 {
        struct BTreeIterator * iterator = NULL;
        struct FSBufferDescriptor btdata;
        HFSPlusExtentRecord extrec;
        HFSPlusExtentKey *extkeyptr;
        FCB *fcb;
-       u_int32_t block;
-       u_int8_t forktype;
-       int overlapped = 0;
-       int i;
+       int overlapped = false;
+       int i, j;
        int error;
        int error;
+       int lockflags = 0;
+       u_int32_t endblock;
+
+       /* Check if data fork overlaps the target space */
+       for (i = 0; i < kHFSPlusExtentDensity; ++i) {
+               if (filerec->dataFork.extents[i].blockCount == 0) {
+                       break;
+               }
+               endblock = filerec->dataFork.extents[i].startBlock +
+                       filerec->dataFork.extents[i].blockCount;
+               if (endblock > allocLimit) {
+                       overlapped = true;
+                       goto out;
+               }
+       }
+
+       /* Check if resource fork overlaps the target space */
+       for (j = 0; j < kHFSPlusExtentDensity; ++j) {
+               if (filerec->resourceFork.extents[j].blockCount == 0) {
+                       break;
+               }
+               endblock = filerec->resourceFork.extents[j].startBlock +
+                       filerec->resourceFork.extents[j].blockCount;
+               if (endblock > allocLimit) {
+                       overlapped = true;
+                       goto out;
+               }
+       }
+
+       /* Return back if there are no overflow extents for this file */
+       if ((i < kHFSPlusExtentDensity) && (j < kHFSPlusExtentDensity)) {
+               goto out;
+       }
 
 
-       forktype = rsrcfork ? 0xFF : 0;
        if (kmem_alloc(kernel_map, (vm_offset_t *)&iterator, sizeof(*iterator))) {
        if (kmem_alloc(kernel_map, (vm_offset_t *)&iterator, sizeof(*iterator))) {
-               return (0);
+               return 0;
        }       
        bzero(iterator, sizeof(*iterator));
        extkeyptr = (HFSPlusExtentKey *)&iterator->key;
        extkeyptr->keyLength = kHFSPlusExtentKeyMaximumLength;
        }       
        bzero(iterator, sizeof(*iterator));
        extkeyptr = (HFSPlusExtentKey *)&iterator->key;
        extkeyptr->keyLength = kHFSPlusExtentKeyMaximumLength;
-       extkeyptr->forkType = forktype;
-       extkeyptr->fileID = fileID;
-       extkeyptr->startBlock = catblks;
+       extkeyptr->forkType = 0;
+       extkeyptr->fileID = filerec->fileID;
+       extkeyptr->startBlock = 0;
 
        btdata.bufferAddress = &extrec;
        btdata.itemSize = sizeof(extrec);
 
        btdata.bufferAddress = &extrec;
        btdata.itemSize = sizeof(extrec);
@@ -4972,32 +6816,49 @@ hfs_overlapped_overflow_extents(struct hfsmount *hfsmp, u_int32_t startblk, u_in
        
        fcb = VTOF(hfsmp->hfs_extents_vp);
 
        
        fcb = VTOF(hfsmp->hfs_extents_vp);
 
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_EXTENTS, HFS_SHARED_LOCK);
+
+       /* This will position the iterator just before the first overflow 
+        * extent record for given fileID.  It will always return btNotFound, 
+        * so we special case the error code.
+        */
        error = BTSearchRecord(fcb, iterator, &btdata, NULL, iterator);
        error = BTSearchRecord(fcb, iterator, &btdata, NULL, iterator);
+       if (error && (error != btNotFound)) {
+               goto out;
+       }
+
+       /* BTIterateRecord() might return error if the btree is empty, and 
+        * therefore we return that the extent does not overflow to the caller
+        */
+       error = BTIterateRecord(fcb, kBTreeNextRecord, iterator, &btdata, NULL);
        while (error == 0) {
                /* Stop when we encounter a different file. */
        while (error == 0) {
                /* Stop when we encounter a different file. */
-               if ((extkeyptr->fileID != fileID) ||
-                   (extkeyptr->forkType != forktype)) {
+               if (extkeyptr->fileID != filerec->fileID) {
                        break;
                }
                        break;
                }
-               /* 
-                * Check if the file overlaps target space.
-                */
+               /* Check if any of the forks exist in the target space. */
                for (i = 0; i < kHFSPlusExtentDensity; ++i) {
                        if (extrec[i].blockCount == 0) {
                                break;
                        }
                for (i = 0; i < kHFSPlusExtentDensity; ++i) {
                        if (extrec[i].blockCount == 0) {
                                break;
                        }
-                       block = extrec[i].startBlock + extrec[i].blockCount;
-                       if (block >= startblk) {
-                               overlapped = 1;
-                               break;
+                       endblock = extrec[i].startBlock + extrec[i].blockCount;
+                       if (endblock > allocLimit) {
+                               overlapped = true;
+                               goto out;
                        }
                }
                /* Look for more records. */
                error = BTIterateRecord(fcb, kBTreeNextRecord, iterator, &btdata, NULL);
        }
 
                        }
                }
                /* Look for more records. */
                error = BTIterateRecord(fcb, kBTreeNextRecord, iterator, &btdata, NULL);
        }
 
-       kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
-       return (overlapped);
+out:
+       if (lockflags) {
+               hfs_systemfile_unlock(hfsmp, lockflags);
+       }
+       if (iterator) {
+               kmem_free(kernel_map, (vm_offset_t)iterator, sizeof(*iterator));
+       }
+       return overlapped;
 }
 
 
 }
 
 
@@ -5012,10 +6873,11 @@ hfs_resize_progress(struct hfsmount *hfsmp, u_int32_t *progress)
                return (ENXIO);
        }
 
                return (ENXIO);
        }
 
-       if (hfsmp->hfs_resize_totalfiles > 0)
-               *progress = (hfsmp->hfs_resize_filesmoved * 100) / hfsmp->hfs_resize_totalfiles;
-       else
+       if (hfsmp->hfs_resize_totalblocks > 0) {
+               *progress = (u_int32_t)((hfsmp->hfs_resize_blocksmoved * 100ULL) / hfsmp->hfs_resize_totalblocks);
+       } else {
                *progress = 0;
                *progress = 0;
+       }
 
        return (0);
 }
 
        return (0);
 }
@@ -5051,6 +6913,7 @@ hfs_vfs_getattr(struct mount *mp, struct vfs_attr *fsap, __unused vfs_context_t
 {
 #define HFS_ATTR_CMN_VALIDMASK (ATTR_CMN_VALIDMASK & ~(ATTR_CMN_NAMEDATTRCOUNT | ATTR_CMN_NAMEDATTRLIST))
 #define HFS_ATTR_FILE_VALIDMASK (ATTR_FILE_VALIDMASK & ~(ATTR_FILE_FILETYPE | ATTR_FILE_FORKCOUNT | ATTR_FILE_FORKLIST))
 {
 #define HFS_ATTR_CMN_VALIDMASK (ATTR_CMN_VALIDMASK & ~(ATTR_CMN_NAMEDATTRCOUNT | ATTR_CMN_NAMEDATTRLIST))
 #define HFS_ATTR_FILE_VALIDMASK (ATTR_FILE_VALIDMASK & ~(ATTR_FILE_FILETYPE | ATTR_FILE_FORKCOUNT | ATTR_FILE_FORKLIST))
+#define HFS_ATTR_CMN_VOL_VALIDMASK (ATTR_CMN_VALIDMASK & ~(ATTR_CMN_NAMEDATTRCOUNT | ATTR_CMN_NAMEDATTRLIST | ATTR_CMN_ACCTIME))
 
        ExtendedVCB *vcb = VFSTOVCB(mp);
        struct hfsmount *hfsmp = VFSTOHFS(mp);
 
        ExtendedVCB *vcb = VFSTOVCB(mp);
        struct hfsmount *hfsmp = VFSTOHFS(mp);
@@ -5177,20 +7040,20 @@ hfs_vfs_getattr(struct mount *mp, struct vfs_attr *fsap, __unused vfs_context_t
        if (VFSATTR_IS_ACTIVE(fsap, f_attributes)) {
                vol_attributes_attr_t *attrp = &fsap->f_attributes;
 
        if (VFSATTR_IS_ACTIVE(fsap, f_attributes)) {
                vol_attributes_attr_t *attrp = &fsap->f_attributes;
 
-               attrp->validattr.commonattr = HFS_ATTR_CMN_VALIDMASK;
+               attrp->validattr.commonattr = HFS_ATTR_CMN_VOL_VALIDMASK;
                attrp->validattr.volattr = ATTR_VOL_VALIDMASK & ~ATTR_VOL_INFO;
                attrp->validattr.dirattr = ATTR_DIR_VALIDMASK;
                attrp->validattr.fileattr = HFS_ATTR_FILE_VALIDMASK;
                attrp->validattr.forkattr = 0;
 
                attrp->validattr.volattr = ATTR_VOL_VALIDMASK & ~ATTR_VOL_INFO;
                attrp->validattr.dirattr = ATTR_DIR_VALIDMASK;
                attrp->validattr.fileattr = HFS_ATTR_FILE_VALIDMASK;
                attrp->validattr.forkattr = 0;
 
-               attrp->nativeattr.commonattr = HFS_ATTR_CMN_VALIDMASK;
+               attrp->nativeattr.commonattr = HFS_ATTR_CMN_VOL_VALIDMASK;
                attrp->nativeattr.volattr = ATTR_VOL_VALIDMASK & ~ATTR_VOL_INFO;
                attrp->nativeattr.dirattr = ATTR_DIR_VALIDMASK;
                attrp->nativeattr.fileattr = HFS_ATTR_FILE_VALIDMASK;
                attrp->nativeattr.forkattr = 0;
                VFSATTR_SET_SUPPORTED(fsap, f_attributes);
        }       
                attrp->nativeattr.volattr = ATTR_VOL_VALIDMASK & ~ATTR_VOL_INFO;
                attrp->nativeattr.dirattr = ATTR_DIR_VALIDMASK;
                attrp->nativeattr.fileattr = HFS_ATTR_FILE_VALIDMASK;
                attrp->nativeattr.forkattr = 0;
                VFSATTR_SET_SUPPORTED(fsap, f_attributes);
        }       
-       fsap->f_create_time.tv_sec = hfsmp->vcbCrDate;
+       fsap->f_create_time.tv_sec = hfsmp->hfs_itime;
        fsap->f_create_time.tv_nsec = 0;
        VFSATTR_SET_SUPPORTED(fsap, f_create_time);
        fsap->f_modify_time.tv_sec = hfsmp->vcbLsMod;
        fsap->f_create_time.tv_nsec = 0;
        VFSATTR_SET_SUPPORTED(fsap, f_create_time);
        fsap->f_modify_time.tv_sec = hfsmp->vcbLsMod;
@@ -5251,6 +7114,10 @@ hfs_rename_volume(struct vnode *vp, const char *name, proc_t p)
        cat_cookie_t cookie;
        int lockflags;
        int error = 0;
        cat_cookie_t cookie;
        int lockflags;
        int error = 0;
+       char converted_volname[256];
+       size_t volname_length = 0;
+       size_t conv_volname_length = 0;
+       
 
        /*
         * Ignore attempts to rename a volume to a zero-length name.
 
        /*
         * Ignore attempts to rename a volume to a zero-length name.
@@ -5285,8 +7152,16 @@ hfs_rename_volume(struct vnode *vp, const char *name, proc_t p)
                                 */
                                if (!error) {
                                        strlcpy((char *)vcb->vcbVN, name, sizeof(vcb->vcbVN));
                                 */
                                if (!error) {
                                        strlcpy((char *)vcb->vcbVN, name, sizeof(vcb->vcbVN));
+                                       volname_length = strlen ((const char*)vcb->vcbVN);
+#define DKIOCCSSETLVNAME _IOW('d', 198, char[1024])
+                                       /* Send the volume name down to CoreStorage if necessary */     
+                                       error = utf8_normalizestr(vcb->vcbVN, volname_length, (u_int8_t*)converted_volname, &conv_volname_length, 256, UTF_PRECOMPOSED);
+                                       if (error == 0) {
+                                               (void) VNOP_IOCTL (hfsmp->hfs_devvp, DKIOCCSSETLVNAME, converted_volname, 0, vfs_context_current());
+                                       }
+                                       error = 0;
                                }
                                }
-
+                               
                                hfs_systemfile_unlock(hfsmp, lockflags);
                                cat_postflight(hfsmp, &cookie, p);
                        
                                hfs_systemfile_unlock(hfsmp, lockflags);
                                cat_postflight(hfsmp, &cookie, p);
                        
@@ -5385,7 +7260,7 @@ static int hfs_journal_replay(vnode_t devvp, vfs_context_t context)
        struct hfs_mount_args *args = NULL;
 
        /* Replay allowed only on raw devices */
        struct hfs_mount_args *args = NULL;
 
        /* Replay allowed only on raw devices */
-       if (!vnode_ischr(devvp)) {
+       if (!vnode_ischr(devvp) && !vnode_isblk(devvp)) {
                retval = EINVAL;
                goto out;
        }
                retval = EINVAL;
                goto out;
        }
@@ -5407,7 +7282,10 @@ static int hfs_journal_replay(vnode_t devvp, vfs_context_t context)
        bzero(args, sizeof(struct hfs_mount_args));
 
        retval = hfs_mountfs(devvp, mp, args, 1, context);
        bzero(args, sizeof(struct hfs_mount_args));
 
        retval = hfs_mountfs(devvp, mp, args, 1, context);
-       buf_flushdirtyblks(devvp, MNT_WAIT, 0, "hfs_journal_replay");
+       buf_flushdirtyblks(devvp, TRUE, 0, "hfs_journal_replay");
+       
+       /* FSYNC the devnode to be sure all data has been flushed */
+       retval = VNOP_FSYNC(devvp, MNT_WAIT, context);
 
 out:
        if (mp) {
 
 out:
        if (mp) {