]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/hfs/hfs_hotfiles.c
xnu-2782.40.9.tar.gz
[apple/xnu.git] / bsd / hfs / hfs_hotfiles.c
index e5f94cd6494628486d17244150fac5eff0b9fdab..7ba80c737bc31470fd4394d89797d58be24a9c07 100644 (file)
@@ -1,23 +1,29 @@
 /*
- * Copyright (c) 2003 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2003-2013 Apple Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
- * The contents of this file constitute Original Code as defined in and
- * are subject to the Apple Public Source License Version 1.1 (the
- * "License").  You may not use this file except in compliance with the
- * License.  Please obtain a copy of the License at
- * http://www.apple.com/publicsource and read it before using this file.
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
  * 
- * This Original Code and all software distributed under the License are
- * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ * 
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
- * License for the specific language governing rights and limitations
- * under the License.
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
  * 
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 
 #include <sys/param.h>
 #include <sys/kernel.h>
 #include <sys/malloc.h>
 #include <sys/ubc.h>
+#include <sys/ubc_internal.h>
 #include <sys/vnode.h>
+#include <sys/vnode_internal.h>
+#include <sys/kauth.h>
 
 #include <hfs/hfs.h>
 #include <hfs/hfs_endian.h>
 #define HFC_VERBOSE 0
 
 
+/*
+ * Minimum post Tiger base time.
+ * Thu Mar 31 17:00:00 2005
+ */
+#define HFC_MIN_BASE_TIME   0x424c8f00L
 
 /*
  * Hot File List (runtime).
@@ -90,6 +104,8 @@ typedef struct hotfile_data {
        hotfile_entry_t  entries[1];
 } hotfile_data_t;
 
+static int  hfs_recording_start (struct hfsmount *);
+static int  hfs_recording_stop (struct hfsmount *);
 
 
 /*
@@ -97,26 +113,32 @@ typedef struct hotfile_data {
  */
 static void              hf_insert (hotfile_data_t *, hotfile_entry_t *);
 static void              hf_delete (hotfile_data_t *, u_int32_t, u_int32_t);
-static hotfile_entry_t * hf_lookup (hotfile_data_t *, u_int32_t, u_int32_t);
 static hotfile_entry_t * hf_coldest (hotfile_data_t *);
 static hotfile_entry_t * hf_getnewentry (hotfile_data_t *);
-static int               hf_getsortedlist (hotfile_data_t *, hotfilelist_t *);
-static void              hf_printtree (hotfile_entry_t *);
+static void              hf_getsortedlist (hotfile_data_t *, hotfilelist_t *);
+
+#if HFC_DEBUG
+static hotfile_entry_t * hf_lookup (hotfile_data_t *, u_int32_t, u_int32_t);
+static void  hf_maxdepth(hotfile_entry_t *, int, int *);
+static void  hf_printtree (hotfile_entry_t *);
+#endif
 
 /*
  * Hot File misc support functions.
  */
-static int  hotfiles_collect (struct hfsmount *, struct proc *);
-static int  hotfiles_age (struct hfsmount *, struct proc *);
-static int  hotfiles_adopt (struct hfsmount *, struct proc *);
-static int  hotfiles_evict (struct hfsmount *, struct proc *);
-static int  hotfiles_refine (struct hfsmount *, struct proc *);
+static int  hotfiles_collect (struct hfsmount *);
+static int  hotfiles_age (struct hfsmount *);
+static int  hotfiles_adopt (struct hfsmount *);
+static int  hotfiles_evict (struct hfsmount *, vfs_context_t);
+static int  hotfiles_refine (struct hfsmount *);
 static int  hotextents(struct hfsmount *, HFSPlusExtentDescriptor *);
+static int  hfs_addhotfile_internal(struct vnode *);
+
 
 /*
  * Hot File Cluster B-tree (on disk) functions.
  */
-static int  hfc_btree_create (struct hfsmount *, int, int);
+static int  hfc_btree_create (struct hfsmount *, unsigned int, unsigned int);
 static int  hfc_btree_open (struct hfsmount *, struct vnode **);
 static int  hfc_btree_close (struct hfsmount *, struct vnode *);
 static int  hfc_comparekeys (HotFileKey *, HotFileKey *);
@@ -134,12 +156,13 @@ char hfc_tag[] = "CLUSTERED HOT FILES B-TREE     ";
 /*
  * Start recording the hotest files on a file system.
  *
+ * Requires that the hfc_mutex be held.
  */
-__private_extern__
-int
-hfs_recording_start(struct hfsmount *hfsmp, struct proc *p)
+static int
+hfs_recording_start(struct hfsmount *hfsmp)
 {
        hotfile_data_t *hotdata;
+       struct timeval tv;
        int maxentries;
        size_t size;
        int i;
@@ -150,7 +173,7 @@ hfs_recording_start(struct hfsmount *hfsmp, struct proc *p)
            (hfsmp->hfs_flags & HFS_METADATA_ZONE) == 0) {
                return (EPERM);
        }
-       if (HFSTOVCB(hfsmp)->freeBlocks < (2 * hfsmp->hfs_hotfile_maxblks)) {
+       if (HFSTOVCB(hfsmp)->freeBlocks < (2 * (u_int32_t)hfsmp->hfs_hotfile_maxblks)) {
                return (ENOSPC);
        }
        if (hfsmp->hfc_stage != HFC_IDLE) {
@@ -169,6 +192,8 @@ hfs_recording_start(struct hfsmount *hfsmp, struct proc *p)
                FREE(tmp, M_TEMP);
        }
 
+       microtime(&tv);  /* Times are base on GMT time. */
+
        /*
         * On first startup check for suspended recording.
         */
@@ -182,14 +207,19 @@ hfs_recording_start(struct hfsmount *hfsmp, struct proc *p)
                    (SWAP_BE32 (hotfileinfo.timeleft) > 0) &&
                    (SWAP_BE32 (hotfileinfo.timebase) > 0)) {
                        hfsmp->hfc_maxfiles = SWAP_BE32 (hotfileinfo.maxfilecnt);
-                       hfsmp->hfc_timeout = SWAP_BE32 (hotfileinfo.timeleft) + time.tv_sec ;
                        hfsmp->hfc_timebase = SWAP_BE32 (hotfileinfo.timebase);
+                       hfsmp->hfc_timeout = SWAP_BE32 (hotfileinfo.timeleft) + tv.tv_sec ;
+                       /* Fix up any bogus timebase values. */
+                       if (hfsmp->hfc_timebase < HFC_MIN_BASE_TIME) {
+                               hfsmp->hfc_timebase = hfsmp->hfc_timeout - HFC_DEFAULT_DURATION;
+                       }
 #if HFC_VERBOSE
-                       printf("HFS: resume recording hot files (%d left)\n", SWAP_BE32 (hotfileinfo.timeleft));
+                       printf("hfs: Resume recording hot files on %s (%d secs left)\n",
+                               hfsmp->vcbVN, SWAP_BE32 (hotfileinfo.timeleft));
 #endif
                } else {
                        hfsmp->hfc_maxfiles = HFC_DEFAULT_FILE_COUNT;
-                       hfsmp->hfc_timebase = time.tv_sec + 1;
+                       hfsmp->hfc_timebase = tv.tv_sec + 1;
                        hfsmp->hfc_timeout = hfsmp->hfc_timebase + HFC_DEFAULT_DURATION;
                }
                (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
@@ -210,17 +240,16 @@ hfs_recording_start(struct hfsmount *hfsmp, struct proc *p)
                        return (error);
                }
 #if HFC_VERBOSE
-               printf("HFS: begin recording hot files\n");
+               printf("hfs: begin recording hot files on %s\n", hfsmp->vcbVN);
 #endif
                hfsmp->hfc_maxfiles = HFC_DEFAULT_FILE_COUNT;
-               hfsmp->hfc_timeout = time.tv_sec + HFC_DEFAULT_DURATION;
+               hfsmp->hfc_timeout = tv.tv_sec + HFC_DEFAULT_DURATION;
 
                /* Reset time base.  */
                if (hfsmp->hfc_timebase == 0) {
-                       hfsmp->hfc_timebase = time.tv_sec + 1;
+                       hfsmp->hfc_timebase = tv.tv_sec + 1;
                } else {
-                       u_int32_t cumulativebase;
-                       u_int32_t oldbase = hfsmp->hfc_timebase;
+                       time_t cumulativebase;
 
                        cumulativebase = hfsmp->hfc_timeout - (HFC_CUMULATIVE_CYCLES * HFC_DEFAULT_DURATION);
                        hfsmp->hfc_timebase = MAX(hfsmp->hfc_timebase, cumulativebase);
@@ -235,6 +264,13 @@ hfs_recording_start(struct hfsmount *hfsmp, struct proc *p)
 
        size = sizeof(hotfile_data_t) + (maxentries * sizeof(hotfile_entry_t));
        MALLOC(hotdata, hotfile_data_t *, size, M_TEMP, M_WAITOK);
+       if (hotdata == NULL) {
+               hfsmp->hfc_recdata = NULL;
+               hfsmp->hfc_stage = HFC_IDLE;
+               wakeup((caddr_t)&hfsmp->hfc_stage);
+               return(ENOMEM);
+       }
+
        bzero(hotdata, size);
 
        for (i = 1; i < maxentries ; i++)
@@ -249,7 +285,6 @@ hfs_recording_start(struct hfsmount *hfsmp, struct proc *p)
        hotdata->hfsmp = hfsmp;
        
        hfsmp->hfc_recdata = hotdata;
-out:
        hfsmp->hfc_stage = HFC_RECORDING;
        wakeup((caddr_t)&hfsmp->hfc_stage);
        return (0);
@@ -257,28 +292,26 @@ out:
 
 /*
  * Stop recording the hotest files on a file system.
+ *
+ * Requires that the hfc_mutex be held.
  */
-__private_extern__
-int
-hfs_recording_stop(struct hfsmount *hfsmp, struct proc *p)
+static int
+hfs_recording_stop(struct hfsmount *hfsmp)
 {
        hotfile_data_t *hotdata;
        hotfilelist_t  *listp;
+       struct timeval tv;
        size_t  size;
        enum hfc_stage newstage = HFC_IDLE;
-       void * tmp;
        int  error;
 
-
        if (hfsmp->hfc_stage != HFC_RECORDING)
                return (EPERM);
 
-       hotfiles_collect(hfsmp, p);
+       hfsmp->hfc_stage = HFC_BUSY;
 
-       if (hfsmp->hfc_stage != HFC_RECORDING)
-               return (0);
+       hotfiles_collect(hfsmp);
 
-       hfsmp->hfc_stage = HFC_BUSY;
 
        /*
         * Convert hot file data into a simple file id list....
@@ -286,7 +319,7 @@ hfs_recording_stop(struct hfsmount *hfsmp, struct proc *p)
         * then dump the sample data
         */
 #if HFC_VERBOSE
-       printf("HFS: end of hot file recording\n");
+       printf("hfs: end of hot file recording on %s\n", hfsmp->vcbVN);
 #endif
        hotdata = (hotfile_data_t *)hfsmp->hfc_recdata;
        if (hotdata == NULL)
@@ -296,7 +329,7 @@ hfs_recording_stop(struct hfsmount *hfsmp, struct proc *p)
        wakeup((caddr_t)&hfsmp->hfc_stage);
 
 #if HFC_VERBOSE
-       printf("  curentries: %d\n", hotdata->activefiles);
+       printf("hfs:   curentries: %d\n", hotdata->activefiles);
 #endif
        /*
         * If no hot files recorded then we're done.
@@ -308,7 +341,7 @@ hfs_recording_stop(struct hfsmount *hfsmp, struct proc *p)
 
        /* Open the B-tree file for writing... */
        if (hfsmp->hfc_filevp)
-               panic("hfs_recording_stop: hfc_filevp exists (vp = 0x%08x)", hfsmp->hfc_filevp);
+               panic("hfs_recording_stop: hfc_filevp exists (vp = %p)", hfsmp->hfc_filevp);
 
        error = hfc_btree_open(hfsmp, &hfsmp->hfc_filevp);
        if (error) {
@@ -318,7 +351,7 @@ hfs_recording_stop(struct hfsmount *hfsmp, struct proc *p)
        /*
         * Age the previous set of clustered hot files.
         */
-       error = hotfiles_age(hfsmp, p);
+       error = hotfiles_age(hfsmp);
        if (error) {
                (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
                hfsmp->hfc_filevp = NULL;
@@ -331,16 +364,24 @@ hfs_recording_stop(struct hfsmount *hfsmp, struct proc *p)
        size = sizeof(hotfilelist_t);
        size += sizeof(hotfileinfo_t) * (hotdata->activefiles - 1);
        MALLOC(listp, hotfilelist_t *, size, M_TEMP, M_WAITOK);
+       if (listp == NULL) {
+               error = ENOMEM;
+               (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
+               hfsmp->hfc_filevp = NULL;
+               goto out;
+       }
+
        bzero(listp, size);
 
-       hf_getsortedlist(hotdata, listp);
-       listp->hfl_duration = time.tv_sec - hfsmp->hfc_timebase;
+       hf_getsortedlist(hotdata, listp);       /* NOTE: destroys hot file tree! */
+       microtime(&tv);
+       listp->hfl_duration = tv.tv_sec - hfsmp->hfc_timebase;
        hfsmp->hfc_recdata = listp;
 
        /*
         * Account for duplicates.
         */
-       error = hotfiles_refine(hfsmp, p);
+       error = hotfiles_refine(hfsmp);
        if (error) {
                (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
                hfsmp->hfc_filevp = NULL;
@@ -373,9 +414,9 @@ hfs_recording_stop(struct hfsmount *hfsmp, struct proc *p)
 out:
 #if HFC_VERBOSE
        if (newstage == HFC_EVICTION)
-               printf("HFS: evicting coldest files\n");
+               printf("hfs: evicting coldest files\n");
        else if (newstage == HFC_ADOPTION)
-               printf("HFS: adopting hotest files\n");
+               printf("hfs: adopting hotest files\n");
 #endif
        FREE(hotdata, M_TEMP);
 
@@ -387,26 +428,35 @@ out:
 /*
  * Suspend recording the hotest files on a file system.
  */
-__private_extern__
 int
-hfs_recording_suspend(struct hfsmount *hfsmp, struct proc *p)
+hfs_recording_suspend(struct hfsmount *hfsmp)
 {
        HotFilesInfo hotfileinfo;
-       hotfile_data_t *hotdata;
+       hotfile_data_t *hotdata = NULL;
+       struct timeval tv;
        int  error;
 
-       if (hfsmp->hfc_stage != HFC_RECORDING)
+       if (hfsmp->hfc_stage == HFC_DISABLED)
                return (0);
 
+       lck_mtx_lock(&hfsmp->hfc_mutex);
+
+       /*
+        * XXX NOTE
+        * A suspend can occur during eval/evict/adopt stage.
+        * In that case we would need to write out info and
+        * flush our HFBT vnode. Currently we just bail.
+        */
+
        hotdata = (hotfile_data_t *)hfsmp->hfc_recdata;
-       if (hotdata == NULL) {
-               hfsmp->hfc_stage = HFC_DISABLED;
-               return (0);
+       if (hotdata == NULL || hfsmp->hfc_stage != HFC_RECORDING) {
+               error = 0;
+               goto out;
        }
        hfsmp->hfc_stage = HFC_BUSY;
 
 #if HFC_VERBOSE
-       printf("HFS: suspend hot file recording\n");
+       printf("hfs: suspend hot file recording on %s\n", hfsmp->vcbVN);
 #endif
        error = hfc_btree_open(hfsmp, &hfsmp->hfc_filevp);
        if (error) {
@@ -414,85 +464,61 @@ hfs_recording_suspend(struct hfsmount *hfsmp, struct proc *p)
                goto out;
        }
 
-       hfs_global_shared_lock_acquire(hfsmp);
-       if (hfsmp->jnl) {
-               if (journal_start_transaction(hfsmp->jnl) != 0) {
-                       hfs_global_shared_lock_release(hfsmp);
-                       error = EINVAL;
-                       goto out;
-               }
+       if (hfs_start_transaction(hfsmp) != 0) {
+           error = EINVAL;
+           goto out;
+       }
+       if (hfs_lock(VTOC(hfsmp->hfc_filevp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT) != 0) {
+               error = EPERM;
+               goto end_transaction;
        }
-       vn_lock(hfsmp->hfc_filevp, LK_EXCLUSIVE | LK_RETRY, p);
 
+       microtime(&tv);
        hotfileinfo.magic       = SWAP_BE32 (HFC_MAGIC);
        hotfileinfo.version     = SWAP_BE32 (HFC_VERSION);
        hotfileinfo.duration    = SWAP_BE32 (HFC_DEFAULT_DURATION);
        hotfileinfo.timebase    = SWAP_BE32 (hfsmp->hfc_timebase);
-       hotfileinfo.timeleft    = SWAP_BE32 (hfsmp->hfc_timeout - time.tv_sec);
+       hotfileinfo.timeleft    = SWAP_BE32 (hfsmp->hfc_timeout - tv.tv_sec);
        hotfileinfo.threshold   = SWAP_BE32 (hotdata->threshold);
        hotfileinfo.maxfileblks = SWAP_BE32 (hotdata->maxblocks);
        hotfileinfo.maxfilecnt  = SWAP_BE32 (HFC_DEFAULT_FILE_COUNT);
-       strcpy(hotfileinfo.tag, hfc_tag);
+       strlcpy((char *)hotfileinfo.tag, hfc_tag, sizeof hotfileinfo.tag);
        (void) BTSetUserData(VTOF(hfsmp->hfc_filevp), &hotfileinfo, sizeof(hotfileinfo));
 
-       (void) VOP_UNLOCK(hfsmp->hfc_filevp, 0, p);
-       if (hfsmp->jnl) {
-               journal_end_transaction(hfsmp->jnl);
-       }
-       hfs_global_shared_lock_release(hfsmp);
-
-       (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
-       hfsmp->hfc_filevp = NULL;
-out:
-       FREE(hotdata, M_TEMP);
+       hfs_unlock(VTOC(hfsmp->hfc_filevp));
 
-       hfsmp->hfc_stage = HFC_DISABLED;
-       wakeup((caddr_t)&hfsmp->hfc_stage);
-       return (error);
-}
+end_transaction:
+       hfs_end_transaction(hfsmp);
 
-/*
- * Abort a hot file recording session.
- */
-__private_extern__
-int
-hfs_recording_abort(struct hfsmount *hfsmp, struct proc *p)
-{
-       void * tmp;
-
-       if (hfsmp->hfc_stage == HFC_DISABLED)
-               return (0);
-       
-       if (hfsmp->hfc_stage == HFC_BUSY) {
-               (void) tsleep((caddr_t)&hfsmp->hfc_stage, PINOD, "hfs_recording_abort", 0);
+out:
+       if (hfsmp->hfc_filevp) {
+               (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
+               hfsmp->hfc_filevp = NULL;
        }
-       hfsmp->hfc_stage = HFC_BUSY;
-
-       printf("HFS: terminate hot file recording\n");
-
-       if (hfsmp->hfc_recdata) {
-               tmp = hfsmp->hfc_recdata;
+       if (hotdata) {
+               FREE(hotdata, M_TEMP);
                hfsmp->hfc_recdata = NULL;
-               FREE(tmp, M_TEMP);
        }
        hfsmp->hfc_stage = HFC_DISABLED;
        wakeup((caddr_t)&hfsmp->hfc_stage);
-       return (0);
+
+       lck_mtx_unlock(&hfsmp->hfc_mutex);
+       return (error);
 }
 
+
 /*
  *
  */
-__private_extern__
 int
-hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
+hfs_recording_init(struct hfsmount *hfsmp)
 {
        CatalogKey * keyp;
        CatalogRecord * datap;
        u_int32_t  dataSize;
        HFSPlusCatalogFile *filep;
        BTScanState scanstate;
-       BTreeIterator * iterator;
+       BTreeIterator * iterator = NULL;
        FSBufferDescriptor  record;
        HotFileKey * key;
        filefork_t * filefork;
@@ -504,6 +530,24 @@ hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
        int inserted = 0;  /* debug variables */
        int filecount = 0;
 
+       /*
+        * For now, only the boot volume is supported.
+        */
+       if ((vfs_flags(HFSTOVFS(hfsmp)) & MNT_ROOTFS) == 0) {
+               hfsmp->hfc_stage = HFC_DISABLED;
+               return (EPERM);
+       }
+
+       /*
+        * Tracking of hot files requires up-to-date access times.
+        * So if access time updates are disabled, then we disable
+        * hot files, too.
+        */
+       if (vfs_flags(HFSTOVFS(hfsmp)) & MNT_NOATIME) {
+               hfsmp->hfc_stage = HFC_DISABLED;
+               return EPERM;
+       }
+       
        /*
         * If the Hot File btree exists then metadata zone is ready.
         */
@@ -513,27 +557,37 @@ hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
                        hfsmp->hfc_stage = HFC_IDLE;
                return (0);
        }
-       /*
-        * For now, only the boot volume is supported.
-        */
-       if ((HFSTOVFS(hfsmp)->mnt_flag & MNT_ROOTFS) == 0) {
-               hfsmp->hfs_flags &= ~HFS_METADATA_ZONE;
-               return (EPERM);
+
+       if (hfs_start_transaction(hfsmp) != 0) {
+               return EINVAL;
        }
+
        error = hfc_btree_create(hfsmp, HFSTOVCB(hfsmp)->blockSize, HFC_DEFAULT_FILE_COUNT);
        if (error) {
-               return (error);
+#if HFC_VERBOSE
+               printf("hfs: Error %d creating hot file b-tree on %s \n", error, hfsmp->vcbVN);
+#endif
+               goto out2;
        }
        /*
         * Open the Hot File B-tree file for writing.
         */
        if (hfsmp->hfc_filevp)
-               panic("hfs_recording_init: hfc_filevp exists (vp = 0x%08x)", hfsmp->hfc_filevp);
+               panic("hfs_recording_init: hfc_filevp exists (vp = %p)", hfsmp->hfc_filevp);
        error = hfc_btree_open(hfsmp, &hfsmp->hfc_filevp);
        if (error) {
-               return (error);
+#if HFC_VERBOSE
+               printf("hfs: Error %d opening hot file b-tree on %s \n", error, hfsmp->vcbVN);
+#endif
+               goto out2;
        }
        MALLOC(iterator, BTreeIterator *, sizeof(*iterator), M_TEMP, M_WAITOK);
+       if (iterator == NULL) {
+               error = ENOMEM;
+               (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
+               hfsmp->hfc_filevp = NULL;
+               goto out2;
+       }
        bzero(iterator, sizeof(*iterator));
        key = (HotFileKey*) &iterator->key;
        key->keyLength = HFC_KEYLENGTH;
@@ -542,7 +596,7 @@ hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
        record.itemSize = sizeof(u_int32_t);
        record.itemCount = 1;
 #if HFC_VERBOSE
-       printf("Evaluating space for \"%s\" metadata zone...\n", HFSTOVCB(hfsmp)->vcbVN);
+       printf("hfs: Evaluating space for \"%s\" metadata zone...\n", HFSTOVCB(hfsmp)->vcbVN);
 #endif
        /*
         * Get ready to scan the Catalog file.
@@ -557,15 +611,14 @@ hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
        /*
         * The writes to Hot File B-tree file are journaled.
         */
-       hfs_global_shared_lock_acquire(hfsmp);
-       if (hfsmp->jnl) {
-               if (journal_start_transaction(hfsmp->jnl) != 0) {
-                       hfs_global_shared_lock_release(hfsmp);
-                       error = EINVAL;
-                       goto out1;
-               }
+       if (hfs_start_transaction(hfsmp) != 0) {
+           error = EINVAL;
+           goto out1;
        } 
-       vn_lock(hfsmp->hfc_filevp, LK_EXCLUSIVE | LK_RETRY, p);
+       if (hfs_lock(VTOC(hfsmp->hfc_filevp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT) != 0) {
+               error = EPERM;
+               goto out0;
+       }
        filefork = VTOF(hfsmp->hfc_filevp);
 
        /*
@@ -614,7 +667,7 @@ hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
                key->fileID      = cnid;
                key->forkType    = 0;
                data = 0x3f3f3f3f;
-               error = BTInsertRecord(filefork, iterator, &record, sizeof(data));
+               error = BTInsertRecord(filefork, iterator, &record, record.itemSize);
                if (error) {
                        printf("hfs_recording_init: BTInsertRecord failed %d (fileid %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
@@ -627,7 +680,7 @@ hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
                key->fileID = cnid;
                key->forkType = 0;
                data = HFC_MINIMUM_TEMPERATURE;
-               error = BTInsertRecord(filefork, iterator, &record, sizeof(data));
+               error = BTInsertRecord(filefork, iterator, &record, record.itemSize);
                if (error) {
                        printf("hfs_recording_init: BTInsertRecord failed %d (fileid %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
@@ -636,20 +689,20 @@ hfs_recording_init(struct hfsmount *hfsmp, struct proc *p)
                inserted++;
        }
        (void) BTFlushPath(filefork);
-       (void) VOP_UNLOCK(hfsmp->hfc_filevp, 0, p);
+       hfs_unlock(VTOC(hfsmp->hfc_filevp));
 
-       if (hfsmp->jnl) {
-               journal_end_transaction(hfsmp->jnl);
-       }
-       hfs_global_shared_lock_release(hfsmp);
+out0:
+       hfs_end_transaction(hfsmp);
 #if HFC_VERBOSE
-       printf("%d files identified out of %d\n", inserted, filecount);
+       printf("hfs: %d files identified out of %d\n", inserted, filecount);
 #endif
        
 out1:
        (void) BTScanTerminate(&scanstate, &data, &data, &data);
 out2:  
-       FREE(iterator, M_TEMP);
+       hfs_end_transaction(hfsmp);
+       if (iterator)
+               FREE(iterator, M_TEMP);
        if (hfsmp->hfc_filevp) {
                (void) hfc_btree_close(hfsmp, hfsmp->hfc_filevp);
                hfsmp->hfc_filevp = NULL;
@@ -663,29 +716,37 @@ out2:
 /*
  * Use sync to perform ocassional background work.
  */
-__private_extern__
 int
-hfs_hotfilesync(struct hfsmount *hfsmp, struct proc *p)
+hfs_hotfilesync(struct hfsmount *hfsmp, vfs_context_t ctx)
 {
-       if ((HFSTOVFS(hfsmp)->mnt_kern_flag & MNTK_UNMOUNT) == 0 && hfsmp->hfc_stage) {
+       if (hfsmp->hfc_stage) {
+               struct timeval tv;
+
+               lck_mtx_lock(&hfsmp->hfc_mutex);
+
                switch (hfsmp->hfc_stage) {
                case HFC_IDLE:
-                       (void) hfs_recording_start(hfsmp, p);
+                       (void) hfs_recording_start(hfsmp);
                        break;
        
                case HFC_RECORDING:
-                       if (time.tv_sec > hfsmp->hfc_timeout)
-                               (void) hfs_recording_stop(hfsmp, p);
+                       microtime(&tv);
+                       if (tv.tv_sec > hfsmp->hfc_timeout)
+                               (void) hfs_recording_stop(hfsmp);
                        break;
        
                case HFC_EVICTION:
-                       (void) hotfiles_evict(hfsmp, p);
+                       (void) hotfiles_evict(hfsmp, ctx);
                        break;
        
                case HFC_ADOPTION:
-                       (void) hotfiles_adopt(hfsmp, p);
+                       (void) hotfiles_adopt(hfsmp);
+                       break;
+               default:
                        break;
                }
+
+               lck_mtx_unlock(&hfsmp->hfc_mutex);
        }
        return (0);
 }
@@ -699,10 +760,26 @@ hfs_hotfilesync(struct hfsmount *hfsmp, struct proc *p)
  * NOTE: Since both the data and resource fork can  be hot,
  * there can be two entries for the same file id.
  *
+ * Note: the cnode is locked on entry.
  */
-__private_extern__
 int
 hfs_addhotfile(struct vnode *vp)
+{
+       hfsmount_t *hfsmp;
+       int error;
+
+       hfsmp = VTOHFS(vp);
+       if (hfsmp->hfc_stage != HFC_RECORDING)
+               return (0);
+
+       lck_mtx_lock(&hfsmp->hfc_mutex);
+       error = hfs_addhotfile_internal(vp);
+       lck_mtx_unlock(&hfsmp->hfc_mutex);
+       return (error);
+}
+
+static int
+hfs_addhotfile_internal(struct vnode *vp)
 {
        hotfile_data_t *hotdata;
        hotfile_entry_t *entry;
@@ -714,11 +791,18 @@ hfs_addhotfile(struct vnode *vp)
        hfsmp = VTOHFS(vp);
        if (hfsmp->hfc_stage != HFC_RECORDING)
                return (0);
-       
-       if (!(vp->v_type == VREG || vp->v_type == VLNK) ||
-            (vp->v_flag & (VSYSTEM | VSWAP))) {
+
+       /* 
+        * Only regular files are eligible for hotfiles addition. 
+        * 
+        * Symlinks were previously added to the list and may exist in 
+        * extant hotfiles regions, but no new ones will be added, and no
+        * symlinks will now be relocated/evicted from the hotfiles region.
+        */
+       if (!vnode_isreg(vp) || vnode_issystem(vp)) {
                return (0);
        }
+
        /* Skip resource forks for now. */
        if (VNODE_IS_RSRC(vp)) {
                return (0);
@@ -731,9 +815,10 @@ hfs_addhotfile(struct vnode *vp)
 
        if ((ffp->ff_bytesread == 0) ||
            (ffp->ff_blocks == 0) ||
+           (ffp->ff_size == 0) ||
            (ffp->ff_blocks > hotdata->maxblocks) ||
            (cp->c_flag & (C_DELETED | C_NOEXISTS)) ||
-           (cp->c_flags & UF_NODUMP) ||
+           (cp->c_bsdflags & UF_NODUMP) ||
            (cp->c_atime < hfsmp->hfc_timebase)) {
                return (0);
        }
@@ -763,15 +848,15 @@ hfs_addhotfile(struct vnode *vp)
 }
 
 /*
- * Remove a hot file to the recording list.
+ * Remove a hot file from the recording list.
  *
  * This can happen when a hot file becomes
  * an active vnode (active hot files are
  * not kept in the recording list until the
  * end of the recording period).
  *
+ * Note: the cnode is locked on entry.
  */
-__private_extern__
 int
 hfs_removehotfile(struct vnode *vp)
 {
@@ -785,31 +870,35 @@ hfs_removehotfile(struct vnode *vp)
        if (hfsmp->hfc_stage != HFC_RECORDING)
                return (0);
 
-       if (!(vp->v_type == VREG || vp->v_type == VLNK) ||
-            (vp->v_flag & (VSYSTEM | VSWAP))) {
+       if ((!vnode_isreg(vp)) || vnode_issystem(vp)) {
                return (0);
        }
-       if ((hotdata = (hotfile_data_t *)hfsmp->hfc_recdata) == NULL)
-               return (0);
 
        ffp = VTOF(vp);
        cp = VTOC(vp);
 
        if ((ffp->ff_bytesread == 0) || (ffp->ff_blocks == 0) ||
-           (cp->c_atime < hfsmp->hfc_timebase)) {
+           (ffp->ff_size == 0) || (cp->c_atime < hfsmp->hfc_timebase)) {
                return (0);
        }
 
+       lck_mtx_lock(&hfsmp->hfc_mutex);
+       if (hfsmp->hfc_stage != HFC_RECORDING)
+               goto out;
+       if ((hotdata = (hotfile_data_t *)hfsmp->hfc_recdata) == NULL)
+               goto out;
+
        temperature = ffp->ff_bytesread / ffp->ff_size;
        if (temperature < hotdata->threshold)
-               return (0);
+               goto out;
 
        if (hotdata->coldest && (temperature >= hotdata->coldest->temperature)) {
                ++hotdata->refcount;
                hf_delete(hotdata, VTOC(vp)->c_fileid, temperature);
                --hotdata->refcount;
        }
-
+out:
+       lck_mtx_unlock(&hfsmp->hfc_mutex);
        return (0);
 }
 
@@ -820,58 +909,35 @@ hfs_removehotfile(struct vnode *vp)
  *========================================================================
  */
 
+static int
+hotfiles_collect_callback(struct vnode *vp, __unused void *cargs)
+{
+        if ((vnode_isreg(vp)) && !vnode_issystem(vp))
+               (void) hfs_addhotfile_internal(vp);
+
+       return (VNODE_RETURNED);
+}
+
 /*
  * Add all active hot files to the recording list.
  */
 static int
-hotfiles_collect(struct hfsmount *hfsmp, struct proc *p)
+hotfiles_collect(struct hfsmount *hfsmp)
 {
        struct mount *mp = HFSTOVFS(hfsmp);
-       struct vnode *nvp, *vp;
-       struct cnode *cp;
-       int error;
 
-       if (vfs_busy(mp, LK_NOWAIT, 0, p))
+       if (vfs_busy(mp, LK_NOWAIT))
                return (0);
-loop:
-       simple_lock(&mntvnode_slock);
-       for (vp = mp->mnt_vnodelist.lh_first; vp != NULL; vp = nvp) {
-               if (vp->v_mount != mp) {
-                       simple_unlock(&mntvnode_slock);
-                       goto loop;
-               }
-               simple_lock(&vp->v_interlock);
-               nvp = vp->v_mntvnodes.le_next;
 
-               if ((vp->v_flag & VSYSTEM) ||
-                   !(vp->v_type == VREG || vp->v_type == VLNK)) {
-                       simple_unlock(&vp->v_interlock);
-                       continue;
-               }
-
-               cp = VTOC(vp);
-               if (cp == NULL || vp->v_flag & (VXLOCK|VORECLAIM)) {
-                       simple_unlock(&vp->v_interlock);
-                       continue;
-               }
-
-               simple_unlock(&mntvnode_slock);
-               error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, p);
-               if (error) {
-                       if (error == ENOENT)
-                               goto loop;
-                       simple_lock(&mntvnode_slock);
-                       continue;
-               }
-               (void) hfs_addhotfile(vp);
-               vput(vp);
-
-               simple_lock(&mntvnode_slock);
-       }
-
-       simple_unlock(&mntvnode_slock);
+       /*
+        * hotfiles_collect_callback will be called for each vnode
+        * hung off of this mount point
+        * the vnode will be
+        * properly referenced and unreferenced around the callback
+        */
+       vnode_iterate(mp, 0, hotfiles_collect_callback, (void *)NULL);
 
-       vfs_unbusy(mp, p);
+       vfs_unbusy(mp);
 
        return (0);
 }
@@ -882,7 +948,7 @@ loop:
  * This is called from within BTUpdateRecord.
  */
 static int
-update_callback(const HotFileKey *key, u_int32_t *data, u_int16_t datalen, u_int32_t *state)
+update_callback(const HotFileKey *key, u_int32_t *data, u_int32_t *state)
 {
        if (key->temperature == HFC_LOOKUPTAG)
                *data = *state;
@@ -893,11 +959,10 @@ update_callback(const HotFileKey *key, u_int32_t *data, u_int16_t datalen, u_int
  * Identify files already in hot area.
  */
 static int
-hotfiles_refine(struct hfsmount *hfsmp, struct proc *p)
+hotfiles_refine(struct hfsmount *hfsmp)
 {
-       BTreeIterator * iterator;
+       BTreeIterator * iterator = NULL;
        struct mount *mp;
-       struct vnode *vp;
        filefork_t * filefork;
        hotfilelist_t  *listp;
        FSBufferDescriptor  record;
@@ -913,6 +978,10 @@ hotfiles_refine(struct hfsmount *hfsmp, struct proc *p)
        mp = HFSTOVFS(hfsmp);
 
        MALLOC(iterator, BTreeIterator *, sizeof(*iterator), M_TEMP, M_WAITOK);
+       if (iterator == NULL) {
+               error = ENOMEM;
+               goto out;
+       }
        bzero(iterator, sizeof(*iterator));
        key = (HotFileKey*) &iterator->key;
 
@@ -920,15 +989,14 @@ hotfiles_refine(struct hfsmount *hfsmp, struct proc *p)
        record.itemSize = sizeof(u_int32_t);
        record.itemCount = 1;
 
-       hfs_global_shared_lock_acquire(hfsmp);
-       if (hfsmp->jnl) {
-               if (journal_start_transaction(hfsmp->jnl) != 0) {
-                       hfs_global_shared_lock_release(hfsmp);
-                       error = EINVAL;
-                       goto out;
-               }
+       if (hfs_start_transaction(hfsmp) != 0) {
+           error = EINVAL;
+           goto out;
        } 
-       vn_lock(hfsmp->hfc_filevp, LK_EXCLUSIVE | LK_RETRY, p);
+       if (hfs_lock(VTOC(hfsmp->hfc_filevp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT) != 0) {
+               error = EPERM;
+               goto out1;
+       }
        filefork = VTOF(hfsmp->hfc_filevp);
 
        for (i = 0; i < listp->hfl_count; ++i) {
@@ -951,7 +1019,7 @@ hotfiles_refine(struct hfsmount *hfsmp, struct proc *p)
                                (IterateCallBackProcPtr)update_callback,
                                &listp->hfl_hotfile[i].hf_temperature);
                if (error) {
-                       printf("hotfiles_refine: BTUpdateRecord failed %d (file %d)\n", error, key->fileID);
+                       printf("hfs: hotfiles_refine: BTUpdateRecord failed %d (file %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
                //      break;
                }
@@ -967,7 +1035,7 @@ hotfiles_refine(struct hfsmount *hfsmp, struct proc *p)
                (void) BTSearchRecord(filefork, iterator, &record, NULL, iterator);
                error = BTDeleteRecord(filefork, iterator);
                if (error) {
-                       printf("hotfiles_refine: BTDeleteRecord failed %d (file %d)\n", error, key->fileID);
+                       printf("hfs: hotfiles_refine: BTDeleteRecord failed %d (file %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
                        break;
                }
@@ -975,9 +1043,9 @@ hotfiles_refine(struct hfsmount *hfsmp, struct proc *p)
                key->temperature = listp->hfl_hotfile[i].hf_temperature;
                key->fileID = listp->hfl_hotfile[i].hf_fileid;
                key->forkType = 0;
-               error = BTInsertRecord(filefork, iterator, &record, sizeof(data));
+               error = BTInsertRecord(filefork, iterator, &record, record.itemSize);
                if (error) {
-                       printf("hotfiles_refine: BTInsertRecord failed %d (file %d)\n", error, key->fileID);
+                       printf("hfs: hotfiles_refine: BTInsertRecord failed %d (file %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
                        break;
                }
@@ -991,25 +1059,25 @@ hotfiles_refine(struct hfsmount *hfsmp, struct proc *p)
        } /* end for */
 
        (void) BTFlushPath(filefork);
-       (void) VOP_UNLOCK(hfsmp->hfc_filevp, 0, p);
+       hfs_unlock(VTOC(hfsmp->hfc_filevp));
 
-       if (hfsmp->jnl) {
-               journal_end_transaction(hfsmp->jnl);
-       }
-       hfs_global_shared_lock_release(hfsmp);
+out1:
+       hfs_end_transaction(hfsmp);
 out:
-       FREE(iterator, M_TEMP); 
+       if (iterator)
+               FREE(iterator, M_TEMP); 
        return (error);
 }
 
 /*
  * Move new hot files into hot area.
+ *
+ * Requires that the hfc_mutex be held.
  */
 static int
-hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
+hotfiles_adopt(struct hfsmount *hfsmp)
 {
-       BTreeIterator * iterator;
-       struct mount *mp;
+       BTreeIterator * iterator = NULL;
        struct vnode *vp;
        filefork_t * filefork;
        hotfilelist_t  *listp;
@@ -1023,7 +1091,6 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
        int  last;
        int  error = 0;
        int  startedtrans = 0;
-       int  aquiredlock = 0;
 
        if ((listp = (hotfilelist_t  *)hfsmp->hfc_recdata) == NULL)
                return (0);     
@@ -1031,16 +1098,24 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
        if (hfsmp->hfc_stage != HFC_ADOPTION) {
                return (EBUSY);
        }
+       if (hfs_lock(VTOC(hfsmp->hfc_filevp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT) != 0) {
+               return (EPERM);
+       }
+
+       MALLOC(iterator, BTreeIterator *, sizeof(*iterator), M_TEMP, M_WAITOK);
+       if (iterator == NULL) {
+               hfs_unlock(VTOC(hfsmp->hfc_filevp));
+               return (ENOMEM);
+       }
+
        stage = hfsmp->hfc_stage;
        hfsmp->hfc_stage = HFC_BUSY;
 
-       mp = HFSTOVFS(hfsmp);
        blksmoved = 0;
        last = listp->hfl_next + HFC_FILESPERSYNC;
        if (last > listp->hfl_count)
                last = listp->hfl_count;
 
-       MALLOC(iterator, BTreeIterator *, sizeof(*iterator), M_TEMP, M_WAITOK);
        bzero(iterator, sizeof(*iterator));
        key = (HotFileKey*) &iterator->key;
        key->keyLength = HFC_KEYLENGTH;
@@ -1049,7 +1124,6 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
        record.itemSize = sizeof(u_int32_t);
        record.itemCount = 1;
 
-       vn_lock(hfsmp->hfc_filevp, LK_EXCLUSIVE | LK_RETRY, p);
        filefork = VTOF(hfsmp->hfc_filevp);
 
        for (i = listp->hfl_next; (i < last) && (blksmoved < HFC_BLKSPERSYNC); ++i) {
@@ -1063,7 +1137,7 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
                /*
                 * Acquire a vnode for this file.
                 */
-               error = VFS_VGET(mp, &listp->hfl_hotfile[i].hf_fileid, &vp);
+               error = hfs_vget(hfsmp, listp->hfl_hotfile[i].hf_fileid, &vp, 0, 0);
                if (error) {
                        if (error == ENOENT) {
                                error = 0;
@@ -1072,23 +1146,27 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
                        }
                        break;
                }
-               if (vp->v_type != VREG && vp->v_type != VLNK) {
-                       printf("hotfiles_adopt: huh, not a file %d (%d)\n", listp->hfl_hotfile[i].hf_fileid, VTOC(vp)->c_cnid);
-                       vput(vp);
-                       listp->hfl_hotfile[i].hf_temperature == 0;
+               if (!vnode_isreg(vp)) {
+                       /* Symlinks are ineligible for adoption into the hotfile zone.  */
+                       printf("hfs: hotfiles_adopt: huh, not a file %d (%d)\n", listp->hfl_hotfile[i].hf_fileid, VTOC(vp)->c_cnid);
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
+                       listp->hfl_hotfile[i].hf_temperature = 0;
                        listp->hfl_next++;
                        continue;  /* stale entry, go to next */
                }
                if (hotextents(hfsmp, &VTOF(vp)->ff_extents[0])) {
-                       vput(vp);
-                       listp->hfl_hotfile[i].hf_temperature == 0;
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
+                       listp->hfl_hotfile[i].hf_temperature = 0;
                        listp->hfl_next++;
                        listp->hfl_totalblocks -= listp->hfl_hotfile[i].hf_blocks;
                        continue;  /* stale entry, go to next */
                }
                fileblocks = VTOF(vp)->ff_blocks;
                if (fileblocks > hfsmp->hfs_hotfile_freeblks) {
-                       vput(vp);
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
                        listp->hfl_next++;
                        listp->hfl_totalblocks -= fileblocks;
                        continue;  /* entry too big, go to next */
@@ -1096,26 +1174,23 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
                
                if ((blksmoved > 0) &&
                    (blksmoved + fileblocks) > HFC_BLKSPERSYNC) {
-                       vput(vp);
-                       break;
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
+                       break;  /* adopt this entry the next time around */
                }
-               /* Start a new transaction. */
-               hfs_global_shared_lock_acquire(hfsmp);
-               aquiredlock = 1;
-               if (hfsmp->jnl) {
-                       if (journal_start_transaction(hfsmp->jnl) != 0) {
-                               error = EINVAL;
-                               vput(vp);
-                               break;
-                       }
-                       startedtrans = 1;
-               } 
+               if (VTOC(vp)->c_desc.cd_nameptr)
+                       data = *(const u_int32_t *)(VTOC(vp)->c_desc.cd_nameptr);
+               else
+                       data = 0x3f3f3f3f;
 
-               error = hfs_relocate(vp, hfsmp->hfs_hotfile_start, p->p_ucred, p);
-               vput(vp);
-               if (error)
-                       break;
-               
+               error = hfs_relocate(vp, hfsmp->hfs_hotfile_start, kauth_cred_get(), current_proc());
+               hfs_unlock(VTOC(vp));
+               vnode_put(vp);
+               if (error) {
+                       /* Move on to next item. */
+                       listp->hfl_next++;
+                       continue;
+               }
                /* Keep hot file free space current. */
                hfsmp->hfs_hotfile_freeblks -= fileblocks;
                listp->hfl_totalblocks -= fileblocks;
@@ -1125,14 +1200,17 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
                key->temperature = listp->hfl_hotfile[i].hf_temperature;
                key->fileID      = listp->hfl_hotfile[i].hf_fileid;
                key->forkType    = 0;
-               if (VTOC(vp)->c_desc.cd_nameptr)
-                       data = *(u_int32_t *)(VTOC(vp)->c_desc.cd_nameptr);
-               else
-                       data = 0x3f3f3f3f;
 
-               error = BTInsertRecord(filefork, iterator, &record, sizeof(data));
+               /* Start a new transaction before calling BTree code. */
+               if (hfs_start_transaction(hfsmp) != 0) {
+                   error = EINVAL;
+                   break;
+               }
+               startedtrans = 1;
+
+               error = BTInsertRecord(filefork, iterator, &record, record.itemSize);
                if (error) {
-                       printf("hotfiles_adopt: BTInsertRecord failed %d (fileid %d)\n", error, key->fileID);
+                       printf("hfs: hotfiles_adopt: BTInsertRecord failed %d (fileid %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
                        stage = HFC_IDLE;
                        break;
@@ -1144,9 +1222,9 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
                key->fileID = listp->hfl_hotfile[i].hf_fileid;
                key->forkType = 0;
                data = listp->hfl_hotfile[i].hf_temperature;
-               error = BTInsertRecord(filefork, iterator, &record, sizeof(data));
+               error = BTInsertRecord(filefork, iterator, &record, record.itemSize);
                if (error) {
-                       printf("hotfiles_adopt: BTInsertRecord failed %d (fileid %d)\n", error, key->fileID);
+                       printf("hfs: hotfiles_adopt: BTInsertRecord failed %d (fileid %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
                        stage = HFC_IDLE;
                        break;
@@ -1155,11 +1233,9 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
 
                /* Transaction complete. */
                if (startedtrans) {
-                       journal_end_transaction(hfsmp->jnl);
-                       startedtrans = 0;
+                   hfs_end_transaction(hfsmp);
+                   startedtrans = 0;
                }
-               hfs_global_shared_lock_release(hfsmp);
-               aquiredlock = 0;
 
                blksmoved += fileblocks;
                listp->hfl_next++;
@@ -1168,31 +1244,27 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
                }
                if (hfsmp->hfs_hotfile_freeblks <= 0) {
 #if HFC_VERBOSE
-                       printf("hotfiles_adopt: free space exhausted (%d)\n", hfsmp->hfs_hotfile_freeblks);
+                       printf("hfs: hotfiles_adopt: free space exhausted (%d)\n", hfsmp->hfs_hotfile_freeblks);
 #endif
                        break;
                }
        } /* end for */
 
 #if HFC_VERBOSE
-       printf("hotfiles_adopt: [%d] adopted %d blocks (%d left)\n", listp->hfl_next, blksmoved, listp->hfl_totalblocks);
+       printf("hfs: hotfiles_adopt: [%d] adopted %d blocks (%d left)\n", listp->hfl_next, blksmoved, listp->hfl_totalblocks);
 #endif
        /* Finish any outstanding transactions. */
        if (startedtrans) {
                (void) BTFlushPath(filefork);
-               journal_end_transaction(hfsmp->jnl);
+               hfs_end_transaction(hfsmp);
                startedtrans = 0;
        }
-       if (aquiredlock) {
-               hfs_global_shared_lock_release(hfsmp);
-               aquiredlock = 0;
-       }
-       (void) VOP_UNLOCK(hfsmp->hfc_filevp, 0, p);
+       hfs_unlock(VTOC(hfsmp->hfc_filevp));
 
        if ((listp->hfl_next >= listp->hfl_count) || (hfsmp->hfs_hotfile_freeblks <= 0)) {
 #if HFC_VERBOSE
-               printf("hotfiles_adopt: all done relocating %d files\n", listp->hfl_count);
-               printf("hotfiles_adopt: %d blocks free in hot file band\n", hfsmp->hfs_hotfile_freeblks);
+               printf("hfs: hotfiles_adopt: all done relocating %d files\n", listp->hfl_count);
+               printf("hfs: hotfiles_adopt: %d blocks free in hot file band\n", hfsmp->hfs_hotfile_freeblks);
 #endif
                stage = HFC_IDLE;
        }
@@ -1209,23 +1281,25 @@ hotfiles_adopt(struct hfsmount *hfsmp, struct proc *p)
 
 /*
  * Reclaim space by evicting the coldest files.
+ *
+ * Requires that the hfc_mutex be held.
  */
 static int
-hotfiles_evict(struct hfsmount *hfsmp, struct proc *p)
+hotfiles_evict(struct hfsmount *hfsmp, vfs_context_t ctx)
 {
-       BTreeIterator * iterator;
-       struct mount *mp;
+       BTreeIterator * iterator = NULL;
        struct vnode *vp;
        HotFileKey * key;
        filefork_t * filefork;
        hotfilelist_t  *listp;
        enum hfc_stage stage;
+       u_int32_t savedtemp;
        int  blksmoved;
        int  filesmoved;
        int  fileblocks;
        int  error = 0;
        int  startedtrans = 0;
-       int  aquiredlock = 0;
+       int  bt_op;
 
        if (hfsmp->hfc_stage != HFC_EVICTION) {
                return (EBUSY);
@@ -1234,17 +1308,25 @@ hotfiles_evict(struct hfsmount *hfsmp, struct proc *p)
        if ((listp = (hotfilelist_t  *)hfsmp->hfc_recdata) == NULL)
                return (0);     
 
+       if (hfs_lock(VTOC(hfsmp->hfc_filevp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT) != 0) {
+               return (EPERM);
+       }
+
+       MALLOC(iterator, BTreeIterator *, sizeof(*iterator), M_TEMP, M_WAITOK);
+       if (iterator == NULL) {
+               hfs_unlock(VTOC(hfsmp->hfc_filevp));
+               return (ENOMEM);
+       }
+
        stage = hfsmp->hfc_stage;
        hfsmp->hfc_stage = HFC_BUSY;
 
-       mp = HFSTOVFS(hfsmp);
        filesmoved = blksmoved = 0;
+       bt_op = kBTreeFirstRecord;
 
-       MALLOC(iterator, BTreeIterator *, sizeof(*iterator), M_TEMP, M_WAITOK);
        bzero(iterator, sizeof(*iterator));
        key = (HotFileKey*) &iterator->key;
 
-       vn_lock(hfsmp->hfc_filevp, LK_EXCLUSIVE | LK_RETRY, p);
        filefork = VTOF(hfsmp->hfc_filevp);
 
        while (listp->hfl_reclaimblks > 0 &&
@@ -1254,22 +1336,22 @@ hotfiles_evict(struct hfsmount *hfsmp, struct proc *p)
                /*
                 * Obtain the first record (ie the coldest one).
                 */
-               if (BTIterateRecord(filefork, kBTreeFirstRecord, iterator, NULL, NULL) != 0) {
+               if (BTIterateRecord(filefork, bt_op, iterator, NULL, NULL) != 0) {
 #if HFC_VERBOSE
-                       printf("hotfiles_evict: no more records\n");
+                       printf("hfs: hotfiles_evict: no more records\n");
 #endif
                        error = 0;
                        stage = HFC_ADOPTION;
                        break;
                }
                if (key->keyLength != HFC_KEYLENGTH) {
-                       printf("hotfiles_evict: invalid key length %d\n", key->keyLength);
+                       printf("hfs: hotfiles_evict: invalid key length %d\n", key->keyLength);
                        error = EFTYPE;
                        break;
                }               
                if (key->temperature == HFC_LOOKUPTAG) {
 #if HFC_VERBOSE
-                       printf("hotfiles_evict: ran into thread records\n");
+                       printf("hfs: hotfiles_evict: ran into thread records\n");
 #endif
                        error = 0;
                        stage = HFC_ADOPTION;
@@ -1278,44 +1360,33 @@ hotfiles_evict(struct hfsmount *hfsmp, struct proc *p)
                /*
                 * Aquire the vnode for this file.
                 */
-               error = VFS_VGET(mp, &key->fileID, &vp);
-
-               /* Start a new transaction. */
-               hfs_global_shared_lock_acquire(hfsmp);
-               aquiredlock = 1;
-               if (hfsmp->jnl) {
-                       if (journal_start_transaction(hfsmp->jnl) != 0) {
-                               if (error == 0)
-                                       vput(vp);
-                               error = EINVAL;
-                               break;
-                       }
-                       startedtrans = 1;
-               } 
+               error = hfs_vget(hfsmp, key->fileID, &vp, 0, 0);
                if (error) {
                        if (error == ENOENT) {
-                               (void) BTDeleteRecord(filefork, iterator);
-                               key->temperature = HFC_LOOKUPTAG;
-                               (void) BTDeleteRecord(filefork, iterator);
-                               goto next;  /* stale entry, go to next */
+                               goto delete;  /* stale entry, go to next */
                        } else {
-                               printf("hotfiles_evict: err %d getting file %d (%d)\n",
+                               printf("hfs: hotfiles_evict: err %d getting file %d\n",
                                       error, key->fileID);
                        }
                        break;
                }
-               if (vp->v_type != VREG && vp->v_type != VLNK) {
-                       printf("hotfiles_evict: huh, not a file %d\n", key->fileID);
-                       vput(vp);
-                       (void) BTDeleteRecord(filefork, iterator);
-                       key->temperature = HFC_LOOKUPTAG;
-                       (void) BTDeleteRecord(filefork, iterator);
-                       goto next;  /* invalid entry, go to next */
+
+               /* 
+                * Symlinks that may have been inserted into the hotfile zone during a previous OS are now stuck 
+                * here.  We do not want to move them. 
+                */
+               if (!vnode_isreg(vp)) {
+                       printf("hfs: hotfiles_evict: huh, not a file %d\n", key->fileID);
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
+                       goto delete;  /* invalid entry, go to next */
                }
+
                fileblocks = VTOF(vp)->ff_blocks;
                if ((blksmoved > 0) &&
                    (blksmoved + fileblocks) > HFC_BLKSPERSYNC) {
-                       vput(vp);
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
                        break;
                }
                /*
@@ -1323,28 +1394,34 @@ hotfiles_evict(struct hfsmount *hfsmp, struct proc *p)
                 */
                if (!hotextents(hfsmp, &VTOF(vp)->ff_extents[0])) {
 #if HFC_VERBOSE
-                       printf("hotfiles_evict: file %d isn't hot!\n", key->fileID);
+                       printf("hfs: hotfiles_evict: file %d isn't hot!\n", key->fileID);
 #endif
-                       vput(vp);
-                       (void) BTDeleteRecord(filefork, iterator);
-                       key->temperature = HFC_LOOKUPTAG;
-                       (void) BTDeleteRecord(filefork, iterator);
-                       goto next;  /* go to next */
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
+                       goto delete;  /* stale entry, go to next */
                }
                
                /*
                 * Relocate file out of hot area.
                 */
-               error = hfs_relocate(vp, HFSTOVCB(hfsmp)->nextAllocation, p->p_ucred, p);
+               error = hfs_relocate(vp, HFSTOVCB(hfsmp)->nextAllocation, vfs_context_ucred(ctx), vfs_context_proc(ctx));
                if (error) {
-                       /* XXX skip to next record here! */
-                       printf("hotfiles_evict: err % relocating file\n", error, key->fileID);
-                       vput(vp);
-                       break;
+                       printf("hfs: hotfiles_evict: err %d relocating file %d\n", error, key->fileID);
+                       hfs_unlock(VTOC(vp));
+                       vnode_put(vp);
+                       bt_op = kBTreeNextRecord;
+                       goto next;  /* go to next */
                }
-               (void) VOP_FSYNC(vp, p->p_ucred, MNT_WAIT, p);
 
-               vput(vp);
+               //
+               // We do not believe that this call to hfs_fsync() is
+               // necessary and it causes a journal transaction
+               // deadlock so we are removing it.
+               //
+               // (void) hfs_fsync(vp, MNT_WAIT, 0, p);
+
+               hfs_unlock(VTOC(vp));
+               vnode_put(vp);
 
                hfsmp->hfs_hotfile_freeblks += fileblocks;
                listp->hfl_reclaimblks -= fileblocks;
@@ -1352,47 +1429,48 @@ hotfiles_evict(struct hfsmount *hfsmp, struct proc *p)
                        listp->hfl_reclaimblks = 0;
                blksmoved += fileblocks;
                filesmoved++;
+delete:
+               /* Start a new transaction before calling BTree code. */
+               if (hfs_start_transaction(hfsmp) != 0) {
+                   error = EINVAL;
+                   break;
+               }
+               startedtrans = 1;
 
                error = BTDeleteRecord(filefork, iterator);
                if (error) {
-                       printf("hotfiles_evict: BTDeleteRecord failed %d (fileid %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
                        break;
                }
+               savedtemp = key->temperature;
                key->temperature = HFC_LOOKUPTAG;
                error = BTDeleteRecord(filefork, iterator);
                if (error) {
-                       printf("hotfiles_evict: BTDeleteRecord thread failed %d (fileid %d)\n", error, key->fileID);
                        error = MacToVFSError(error);
                        break;
                }
+               key->temperature = savedtemp;
 next:
                (void) BTFlushPath(filefork);
 
                /* Transaction complete. */
                if (startedtrans) {
-                       journal_end_transaction(hfsmp->jnl);
+                       hfs_end_transaction(hfsmp);
                        startedtrans = 0;
                }
-               hfs_global_shared_lock_release(hfsmp);
-               aquiredlock = 0;
 
        } /* end while */
 
 #if HFC_VERBOSE
-       printf("hotfiles_evict: moved %d files (%d blks, %d to go)\n", filesmoved, blksmoved, listp->hfl_reclaimblks);
+       printf("hfs: hotfiles_evict: moved %d files (%d blks, %d to go)\n", filesmoved, blksmoved, listp->hfl_reclaimblks);
 #endif
        /* Finish any outstanding transactions. */
        if (startedtrans) {
                (void) BTFlushPath(filefork);
-               journal_end_transaction(hfsmp->jnl);
+               hfs_end_transaction(hfsmp);
                startedtrans = 0;
        }
-       if (aquiredlock) {
-               hfs_global_shared_lock_release(hfsmp);
-               aquiredlock = 0;
-       }
-       (void) VOP_UNLOCK(hfsmp->hfc_filevp, 0, p);
+       hfs_unlock(VTOC(hfsmp->hfc_filevp));
 
        /*
         * Move to next stage when finished.
@@ -1400,7 +1478,7 @@ next:
        if (listp->hfl_reclaimblks <= 0) {
                stage = HFC_ADOPTION;
 #if HFC_VERBOSE
-               printf("hotfiles_evict: %d blocks free in hot file band\n", hfsmp->hfs_hotfile_freeblks);
+               printf("hfs: hotfiles_evict: %d blocks free in hot file band\n", hfsmp->hfs_hotfile_freeblks);
 #endif
        }
        FREE(iterator, M_TEMP); 
@@ -1413,10 +1491,10 @@ next:
  * Age the existing records in the hot files b-tree.
  */
 static int
-hotfiles_age(struct hfsmount *hfsmp, struct proc *p)
+hotfiles_age(struct hfsmount *hfsmp)
 {
        BTreeInfoRec  btinfo;
-       BTreeIterator * iterator;
+       BTreeIterator * iterator = NULL;
        BTreeIterator * prev_iterator;
        FSBufferDescriptor  record;
        FSBufferDescriptor  prev_record;
@@ -1434,6 +1512,10 @@ hotfiles_age(struct hfsmount *hfsmp, struct proc *p)
 
 
        MALLOC(iterator, BTreeIterator *, 2 * sizeof(*iterator), M_TEMP, M_WAITOK);
+       if (iterator == NULL) {
+               error = ENOMEM;
+               goto out2;
+       }
        bzero(iterator, 2 * sizeof(*iterator));
        key = (HotFileKey*) &iterator->key;
 
@@ -1450,15 +1532,14 @@ hotfiles_age(struct hfsmount *hfsmp, struct proc *p)
        /*
         * Capture b-tree changes inside a transaction
         */
-       hfs_global_shared_lock_acquire(hfsmp);
-       if (hfsmp->jnl) {
-               if (journal_start_transaction(hfsmp->jnl) != 0) {
-                       hfs_global_shared_lock_release(hfsmp);
-                       error = EINVAL;
-                       goto out2;
-               }
+       if (hfs_start_transaction(hfsmp) != 0) {
+           error = EINVAL;
+           goto out2;
        } 
-       vn_lock(hfsmp->hfc_filevp, LK_EXCLUSIVE | LK_RETRY, p);
+       if (hfs_lock(VTOC(hfsmp->hfc_filevp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT) != 0) {
+               error = EPERM;
+               goto out1;
+       }
        filefork = VTOF(hfsmp->hfc_filevp);
 
        error = BTGetInformation(filefork, 0, &btinfo);
@@ -1527,7 +1608,7 @@ hotfiles_age(struct hfsmount *hfsmp, struct proc *p)
                newtemp = MAX(prev_key->temperature >> 1, 4);
                prev_key->temperature = newtemp;
        
-               error = BTInsertRecord(filefork, prev_iterator, &prev_record, sizeof(data));
+               error = BTInsertRecord(filefork, prev_iterator, &prev_record, prev_record.itemSize);
                if (error) {
                        printf("hfs_agehotfiles: BTInsertRecord failed %d (file %d)\n", error, prev_key->fileID);
                        error = MacToVFSError(error);
@@ -1559,15 +1640,12 @@ hotfiles_age(struct hfsmount *hfsmp, struct proc *p)
 #endif
        (void) BTFlushPath(filefork);
 out:
-       (void) VOP_UNLOCK(hfsmp->hfc_filevp, 0, p);
-
-       if (hfsmp->jnl) {
-       //      hfs_flushvolumeheader(hfsmp, MNT_NOWAIT, 0);
-               journal_end_transaction(hfsmp->jnl);
-       }
-       hfs_global_shared_lock_release(hfsmp);
+       hfs_unlock(VTOC(hfsmp->hfc_filevp));
+out1:
+       hfs_end_transaction(hfsmp);
 out2:
-       FREE(iterator, M_TEMP); 
+       if (iterator)
+               FREE(iterator, M_TEMP); 
        return (error);
 }
 
@@ -1608,55 +1686,56 @@ hotextents(struct hfsmount *hfsmp, HFSPlusExtentDescriptor * extents)
 /*
  * Open the hot files b-tree for writing.
  *
- * On successful exit the vnode has a reference but is unlocked.
+ * On successful exit the vnode has a reference but not an iocount.
  */
 static int
 hfc_btree_open(struct hfsmount *hfsmp, struct vnode **vpp)
 {
-       struct proc *p;
+       proc_t p;
        struct vnode *vp;
-       struct cat_desc  cdesc = {0};
+       struct cat_desc  cdesc;
        struct cat_attr  cattr;
        struct cat_fork  cfork;
        static char filename[] = HFC_FILENAME;
        int  error;
        int  retry = 0;
+       int lockflags;
+       int newvnode_flags = 0;
 
        *vpp = NULL;
        p = current_proc();
 
+       bzero(&cdesc, sizeof(cdesc));
        cdesc.cd_parentcnid = kRootDirID;
-       cdesc.cd_nameptr = filename;
+       cdesc.cd_nameptr = (const u_int8_t *)filename;
        cdesc.cd_namelen = strlen(filename);
 
-       /* Lock catalog b-tree */
-       error = hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_SHARED, p);
-       if (error)
-               return (error);
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_SHARED_LOCK);
 
-       error = cat_lookup(hfsmp, &cdesc, 0, &cdesc, &cattr, &cfork);
+       error = cat_lookup(hfsmp, &cdesc, 0, 0, &cdesc, &cattr, &cfork, NULL);
 
-       /* Unlock catalog b-tree */
-       (void) hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_RELEASE, p);
+       hfs_systemfile_unlock(hfsmp, lockflags);
 
        if (error) {
-               printf("hfc_btree_open: cat_lookup error %d\n", error);
+               printf("hfs: hfc_btree_open: cat_lookup error %d\n", error);
                return (error);
        }
 again:
        cdesc.cd_flags |= CD_ISMETA;
-       error = hfs_getnewvnode(hfsmp, NULL, &cdesc, 0, &cattr, &cfork, &vp);
+       error = hfs_getnewvnode(hfsmp, NULL, NULL, &cdesc, 0, &cattr, 
+                                                       &cfork, &vp, &newvnode_flags);
        if (error) {
-               printf("hfc_btree_open: hfs_getnewvnode error %d\n", error);
+               printf("hfs: hfc_btree_open: hfs_getnewvnode error %d\n", error);
                cat_releasedesc(&cdesc);
                return (error);
        }
-       if ((vp->v_flag & VSYSTEM) == 0) {
+       if (!vnode_issystem(vp)) {
 #if HFC_VERBOSE
-               printf("hfc_btree_open: file has UBC, try again\n");
+               printf("hfs: hfc_btree_open: file has UBC, try again\n");
 #endif
-               vput(vp);
-               vgone(vp);
+               hfs_unlock(VTOC(vp));
+               vnode_recycle(vp);
+               vnode_put(vp);
                if (retry++ == 0)
                        goto again;
                else
@@ -1666,34 +1745,19 @@ again:
        /* Open the B-tree file for writing... */
        error = BTOpenPath(VTOF(vp), (KeyCompareProcPtr) hfc_comparekeys);      
        if (error) {
-               printf("hfc_btree_open: BTOpenPath error %d\n", error);
+               printf("hfs: hfc_btree_open: BTOpenPath error %d\n", error);
                error = MacToVFSError(error);
-       } else {
-#if HFC_VERBOSE
-               struct BTreeInfoRec btinfo;
-
-               if (BTGetInformation(VTOF(vp), 0, &btinfo) == 0) {
-                       printf("btinfo: nodeSize %d\n", btinfo.nodeSize);
-                       printf("btinfo: maxKeyLength %d\n", btinfo.maxKeyLength);
-                       printf("btinfo: treeDepth %d\n", btinfo.treeDepth);
-                       printf("btinfo: numRecords %d\n", btinfo.numRecords);
-                       printf("btinfo: numNodes %d\n", btinfo.numNodes);
-                       printf("btinfo: numFreeNodes %d\n", btinfo.numFreeNodes);
-               }
-#endif
        }
 
-       VOP_UNLOCK(vp, 0, p);   /* unlocked with a single reference */
-       if (error)
-               vrele(vp);
-       else
+       hfs_unlock(VTOC(vp));
+       if (error == 0) {
                *vpp = vp;
+               vnode_ref(vp);  /* keep a reference while its open */
+       }
+       vnode_put(vp);
 
-       if ((vp->v_flag & VSYSTEM) == 0)
-               panic("hfc_btree_open: not a system file (vp = 0x%08x)", vp);
-
-       if (UBCINFOEXISTS(vp))
-               panic("hfc_btree_open: has UBCInfo (vp = 0x%08x)", vp);
+       if (!vnode_issystem(vp))
+               panic("hfs: hfc_btree_open: not a system file (vp = %p)", vp);
 
        return (error);
 }
@@ -1701,31 +1765,32 @@ again:
 /*
  * Close the hot files b-tree.
  *
- * On entry the vnode is not locked but has a reference.
+ * On entry the vnode has a reference.
  */
 static int
 hfc_btree_close(struct hfsmount *hfsmp, struct vnode *vp)
 {
-       struct proc *p = current_proc();
-       int  error;
+       proc_t p = current_proc();
+       int  error = 0;
 
 
        if (hfsmp->jnl) {
-           journal_flush(hfsmp->jnl);
+           hfs_journal_flush(hfsmp, FALSE);
        }
 
-       if (vget(vp, LK_EXCLUSIVE, p) == 0) {
-               (void) VOP_FSYNC(vp, NOCRED, MNT_WAIT, p);
-               error = BTClosePath(VTOF(vp));
-               if (error)
-                       printf("hfc_btree_close: BTClosePath error %d\n", error);
-               vput(vp);
+       if (vnode_get(vp) == 0) {
+               error = hfs_lock(VTOC(vp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT);
+               if (error == 0) {
+                       (void) hfs_fsync(vp, MNT_WAIT, 0, p);
+                       error = BTClosePath(VTOF(vp));
+                       hfs_unlock(VTOC(vp));
+               }
+               vnode_rele(vp);
+               vnode_recycle(vp);
+               vnode_put(vp);
        }
-       vrele(vp);
-       vgone(vp);
-       vp = NULL;
        
-       return (0);
+       return (error);
 }
 
 /*
@@ -1733,34 +1798,67 @@ hfc_btree_close(struct hfsmount *hfsmp, struct vnode *vp)
  *
  */
 static int
-hfc_btree_create(struct hfsmount *hfsmp, int nodesize, int entries)
+hfc_btree_create(struct hfsmount *hfsmp, unsigned int nodesize, unsigned int entries)
 {
-       struct proc *p;
-       struct nameidata nd;
-       struct vnode *vp;
-       char path[128];
+       struct vnode *dvp = NULL;
+       struct vnode *vp = NULL;
+       struct cnode *cp = NULL;
+       vfs_context_t ctx = vfs_context_current();
+       struct vnode_attr va;
+       struct componentname cname;
+       static char filename[] = HFC_FILENAME;
        int  error;
 
-
        if (hfsmp->hfc_filevp)
-               panic("hfc_btree_create: hfc_filevp exists (vp = 0x%08x)", hfsmp->hfc_filevp);
+               panic("hfs: hfc_btree_create: hfc_filevp exists (vp = %p)", hfsmp->hfc_filevp);
 
-       p = current_proc();
-       snprintf(path, sizeof(path), "%s/%s",
-                hfsmp->hfs_mp->mnt_stat.f_mntonname, HFC_FILENAME);
-       NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, path, p);
-       if ((error = vn_open(&nd, O_CREAT | FWRITE, S_IRUSR | S_IWUSR)) != 0) {
+       error = VFS_ROOT(HFSTOVFS(hfsmp), &dvp, ctx);
+       if (error) {
                return (error);
        }
-       vp = nd.ni_vp;
-       
+       cname.cn_nameiop = CREATE;
+       cname.cn_flags = ISLASTCN;
+       cname.cn_context = ctx;
+       cname.cn_pnbuf = filename;
+       cname.cn_pnlen = sizeof(filename);
+       cname.cn_nameptr = filename;
+       cname.cn_namelen = strlen(filename);
+       cname.cn_hash = 0;
+       cname.cn_consume = 0;
+
+       VATTR_INIT(&va);
+       VATTR_SET(&va, va_type, VREG);
+       VATTR_SET(&va, va_mode, S_IFREG | S_IRUSR | S_IWUSR);
+       VATTR_SET(&va, va_uid, 0);
+       VATTR_SET(&va, va_gid, 0);
+
+       if (hfs_start_transaction(hfsmp) != 0) {
+           error = EINVAL;
+           goto out;
+       } 
+
+       /* call ourselves directly, ignore the higher-level VFS file creation code */
+       error = VNOP_CREATE(dvp, &vp, &cname, &va, ctx);
+       if (error) {
+               printf("hfs: error %d creating HFBT on %s\n", error, HFSTOVCB(hfsmp)->vcbVN);
+               goto out;
+       }
+       if (dvp) {
+               vnode_put(dvp);
+               dvp = NULL;
+       }
+       if ((error = hfs_lock(VTOC(vp), HFS_EXCLUSIVE_LOCK, HFS_LOCK_DEFAULT))) {
+               goto out;
+       }
+       cp = VTOC(vp);
+
        /* Don't use non-regular files or files with links. */
-       if (vp->v_type != VREG || VTOC(vp)->c_nlink != 1) {
+       if (!vnode_isreg(vp) || cp->c_linkcount != 1) {
                error = EFTYPE;
                goto out;
        }
 
-       printf("HFS: created HFBT on %s\n", HFSTOVCB(hfsmp)->vcbVN);
+       printf("hfs: created HFBT on %s\n", HFSTOVCB(hfsmp)->vcbVN);
 
        if (VTOF(vp)->ff_size < nodesize) {
                caddr_t  buffer;
@@ -1776,7 +1874,7 @@ hfc_btree_create(struct hfsmount *hfsmp, int nodesize, int entries)
                /*
                 * Mark it invisible (truncate will pull these changes).
                 */
-               ((FndrFileInfo *)&VTOC(vp)->c_finderinfo[0])->fdFlags |=
+               ((FndrFileInfo *)&cp->c_finderinfo[0])->fdFlags |=
                        SWAP_BE16 (kIsInvisible + kNameLocked);
 
                if (kmem_alloc(kernel_map, (vm_offset_t *)&buffer, nodesize)) {
@@ -1784,7 +1882,7 @@ hfc_btree_create(struct hfsmount *hfsmp, int nodesize, int entries)
                        goto out;
                }       
                bzero(buffer, nodesize);
-               index = (int16_t *)buffer;
+               index = (u_int16_t *)buffer;
        
                entirespernode = (nodesize - sizeof(BTNodeDescriptor) - 2) /
                                 (sizeof(HotFileKey) + 6);
@@ -1800,7 +1898,7 @@ hfc_btree_create(struct hfsmount *hfsmp, int nodesize, int entries)
                index[(nodesize / 2) - 1] = SWAP_BE16 (offset);
        
                /* FILL IN THE HEADER RECORD:  */
-               bthp = (BTHeaderRec *)((UInt8 *)buffer + offset);
+               bthp = (BTHeaderRec *)((u_int8_t *)buffer + offset);
                bthp->nodeSize     = SWAP_BE16 (nodesize);
                bthp->totalNodes   = SWAP_BE32 (filesize / nodesize);
                bthp->freeNodes    = SWAP_BE32 (nodecnt - 1);
@@ -1812,7 +1910,7 @@ hfc_btree_create(struct hfsmount *hfsmp, int nodesize, int entries)
                index[(nodesize / 2) - 2] = SWAP_BE16 (offset);
        
                /* FILL IN THE USER RECORD:  */
-               hotfileinfo = (HotFilesInfo *)((UInt8 *)buffer + offset);
+               hotfileinfo = (HotFilesInfo *)((u_int8_t *)buffer + offset);
                hotfileinfo->magic       = SWAP_BE32 (HFC_MAGIC);
                hotfileinfo->version     = SWAP_BE32 (HFC_VERSION);
                hotfileinfo->duration    = SWAP_BE32 (HFC_DEFAULT_DURATION);
@@ -1821,7 +1919,8 @@ hfc_btree_create(struct hfsmount *hfsmp, int nodesize, int entries)
                hotfileinfo->threshold   = SWAP_BE32 (HFC_MINIMUM_TEMPERATURE);
                hotfileinfo->maxfileblks = SWAP_BE32 (HFC_MAXIMUM_FILESIZE / HFSTOVCB(hfsmp)->blockSize);
                hotfileinfo->maxfilecnt  = SWAP_BE32 (HFC_DEFAULT_FILE_COUNT);
-               strcpy(hotfileinfo->tag, hfc_tag);
+               strlcpy((char *)hotfileinfo->tag, hfc_tag,
+                       sizeof hotfileinfo->tag);
                offset += kBTreeHeaderUserBytes;
                index[(nodesize / 2) - 3] = SWAP_BE16 (offset);
        
@@ -1831,29 +1930,50 @@ hfc_btree_create(struct hfsmount *hfsmp, int nodesize, int entries)
                                   - kBTreeHeaderUserBytes - (4 * sizeof(int16_t));
                index[(nodesize / 2) - 4] = SWAP_BE16 (offset);
 
-               vp->v_flag |= VNOFLUSH;
-               error = VOP_TRUNCATE(vp, (off_t)filesize, IO_NDELAY, NOCRED, p);
+               vnode_setnoflush(vp);
+               error = hfs_truncate(vp, (off_t)filesize, IO_NDELAY, 0, ctx);
+               if (error) {
+                       printf("hfs: error %d growing HFBT on %s\n", error, HFSTOVCB(hfsmp)->vcbVN);
+                       goto out;
+               }
+               cp->c_flag |= C_ZFWANTSYNC;
+               cp->c_zftimeout = 1;
+               
                if (error == 0) {
-                       struct iovec aiov;
-                       struct uio auio;
-
-                       auio.uio_iov = &aiov;
-                       auio.uio_iovcnt = 1;
-                       aiov.iov_base = buffer;
-                       aiov.iov_len = filesize;
-                       auio.uio_resid = nodesize;
-                       auio.uio_offset = (off_t)(0);
-                       auio.uio_segflg = UIO_SYSSPACE;
-                       auio.uio_rw = UIO_WRITE;
-                       auio.uio_procp = (struct proc *)0;
-                       error = VOP_WRITE(vp, &auio, 0, kernproc->p_ucred);
+                       struct vnop_write_args args;
+                       uio_t auio;
+
+                       auio = uio_create(1, 0, UIO_SYSSPACE, UIO_WRITE);
+                       uio_addiov(auio, (uintptr_t)buffer, nodesize);
+
+                       args.a_desc = &vnop_write_desc;
+                       args.a_vp = vp;
+                       args.a_uio = auio;
+                       args.a_ioflag = 0;
+                       args.a_context = ctx;
+
+                       hfs_unlock(cp);
+                       cp = NULL;
+
+                       error = hfs_vnop_write(&args);
+                       if (error)
+                               printf("hfs: error %d writing HFBT on %s\n", error, HFSTOVCB(hfsmp)->vcbVN);
+
+                       uio_free(auio);
                }
                kmem_free(kernel_map, (vm_offset_t)buffer, nodesize);
        }
 out:
-       (void) VOP_UNLOCK(vp, 0, p);
-       (void) vn_close(vp, FWRITE, kernproc->p_ucred, p);      
-       vgone(vp);
+       hfs_end_transaction(hfsmp);
+       if (dvp) {
+               vnode_put(dvp);
+       }
+       if (vp) {
+               if (cp)
+                       hfs_unlock(cp);
+               vnode_recycle(vp);
+               vnode_put(vp);
+       }
        return (error);
 }
 
@@ -1903,6 +2023,7 @@ hfc_comparekeys(HotFileKey *searchKey, HotFileKey *trialKey)
 /*
  * Lookup a hot file entry in the tree.
  */
+#if HFC_DEBUG
 static hotfile_entry_t *
 hf_lookup(hotfile_data_t *hotdata, u_int32_t fileid, u_int32_t temperature)
 {
@@ -1923,6 +2044,7 @@ hf_lookup(hotfile_data_t *hotdata, u_int32_t fileid, u_int32_t temperature)
        }
        return (entry);
 }
+#endif
 
 /*
  * Insert a hot file entry into the tree.
@@ -1993,6 +2115,21 @@ hf_coldest(hotfile_data_t *hotdata)
        return (entry);
 }
 
+/*
+ * Find the hottest entry in the tree.
+ */
+static hotfile_entry_t *
+hf_hottest(hotfile_data_t *hotdata)
+{
+       hotfile_entry_t *entry = hotdata->rootentry;
+
+       if (entry) {
+               while (entry->right)
+                       entry = entry->right;
+       }
+       return (entry);
+}
+
 /*
  * Delete a hot file entry from the tree.
  */
@@ -2093,38 +2230,32 @@ hf_getnewentry(hotfile_data_t *hotdata)
 
 
 /*
- * Visit the tree in desending order.
+ * Generate a sorted list of hot files (hottest to coldest).
+ *
+ * As a side effect, every node in the hot file tree will be
+ * deleted (moved to the free list).
  */
 static void
-hf_sortlist(hotfile_entry_t * root, int *index, hotfilelist_t *sortedlist)
-{
-       if (root) {
-               int i;
-
-               hf_sortlist(root->right, index, sortedlist);
-               i = *index;
-               ++(*index);
-               sortedlist->hfl_hotfile[i].hf_fileid = root->fileid;
-               sortedlist->hfl_hotfile[i].hf_temperature = root->temperature;
-               sortedlist->hfl_hotfile[i].hf_blocks = root->blocks;
-               sortedlist->hfl_totalblocks += root->blocks;
-               hf_sortlist(root->left, index, sortedlist);
-       }
-}
-
-/*
- * Generate a sorted list of hot files.
- */
-static int
 hf_getsortedlist(hotfile_data_t * hotdata, hotfilelist_t *sortedlist)
 {
-       int index = 0;
-
-       hf_sortlist(hotdata->rootentry, &index, sortedlist);
+       int i = 0;
+       hotfile_entry_t *entry;
+       
+       while ((entry = hf_hottest(hotdata)) != NULL) {
+               sortedlist->hfl_hotfile[i].hf_fileid = entry->fileid;
+               sortedlist->hfl_hotfile[i].hf_temperature = entry->temperature;
+               sortedlist->hfl_hotfile[i].hf_blocks = entry->blocks;
+               sortedlist->hfl_totalblocks += entry->blocks;
+               ++i;
 
-       sortedlist->hfl_count = hotdata->activefiles;
+               hf_delete(hotdata, entry->fileid, entry->temperature);
+       }
+       
+       sortedlist->hfl_count = i;
        
-       return (index);
+#if HFC_VERBOSE
+       printf("hfs: hf_getsortedlist returned %d entries\n", i);
+#endif
 }
 
 
@@ -2146,7 +2277,7 @@ hf_printtree(hotfile_entry_t * root)
 {
        if (root) {
                hf_printtree(root->left);
-               printf("temperature: % 8d, fileid %d\n", root->temperature, root->fileid);
+               printf("hfs: temperature: % 8d, fileid %d\n", root->temperature, root->fileid);
                hf_printtree(root->right);
        }
 }