]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/hfs/hfs_vfsutils.c
xnu-792.6.22.tar.gz
[apple/xnu.git] / bsd / hfs / hfs_vfsutils.c
index 63aa2a90ea671d19a392e081ca03df519bf74425..4eed699c9223f227dbd622a98f4b71a0cdc0c1c8 100644 (file)
@@ -1,24 +1,21 @@
 /*
- * Copyright (c) 2000-2002 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved.
  *
  * @APPLE_LICENSE_HEADER_START@
  * 
- * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
+ * 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. 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
+ * This 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, QUIET ENJOYMENT OR NON-INFRINGEMENT.
- * Please see the License for the specific language governing rights and
- * limitations under the License.
+ * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
+ * License for the specific language governing rights and limitations
+ * under the License.
  * 
  * @APPLE_LICENSE_HEADER_END@
  */
 #include <sys/malloc.h>
 #include <sys/stat.h>
 #include <sys/mount.h>
-#include <sys/namei.h>
-#include <sys/lock.h>
 #include <sys/buf.h>
 #include <sys/ubc.h>
 #include <sys/unistd.h>
+#include <sys/utfconv.h>
+#include <sys/kauth.h>
 
 #include "hfs.h"
 #include "hfs_catalog.h"
 #include "hfscommon/headers/HFSUnicodeWrappers.h"
 
 
-extern int count_lock_queue __P((void));
-extern uid_t console_user;
+extern int count_lock_queue(void);
 
 
 static void ReleaseMetaFileVNode(struct vnode *vp);
 static int  hfs_late_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp, void *_args);
 
+static void hfs_metadatazone_init(struct hfsmount *);
+static u_int32_t hfs_hotfile_freeblocks(struct hfsmount *);
+
+
+
 u_int32_t GetLogicalBlockSize(struct vnode *vp);
 
-/* BTree accessor routines */
-extern OSStatus GetBTreeBlock(FileReference vp, UInt32 blockNum, GetBlockOptions options, BlockDescriptor *block);
-extern OSStatus SetBTreeBlockSize(FileReference vp, ByteCount blockSize, ItemCount minBlockCount);
-extern OSStatus ExtendBTreeFile(FileReference vp, FSSize minEOF, FSSize maxEOF);
-extern OSStatus ReleaseBTreeBlock(FileReference vp, BlockDescPtr blockPtr, ReleaseBlockOptions options);
+extern int hfs_attrkeycompare(HFSPlusAttrKey *searchKey, HFSPlusAttrKey *trialKey);
+
 
 //*******************************************************************************
 // Note: Finder information in the HFS/HFS+ metadata are considered opaque and
@@ -82,10 +80,12 @@ extern OSStatus ReleaseBTreeBlock(FileReference vp, BlockDescPtr blockPtr, Relea
 char hfs_catname[] = "Catalog B-tree";
 char hfs_extname[] = "Extents B-tree";
 char hfs_vbmname[] = "Volume Bitmap";
+char hfs_attrname[] = "Attribute B-tree";
 
 char hfs_privdirname[] =
        "\xE2\x90\x80\xE2\x90\x80\xE2\x90\x80\xE2\x90\x80HFS+ Private Data";
 
+__private_extern__
 OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
                struct proc *p)
 {
@@ -102,9 +102,11 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
                return (EINVAL);
 
        /* don't mount a writeable volume if its dirty, it must be cleaned by fsck_hfs */
-       if ((hfsmp->hfs_fs_ronly == 0) && ((SWAP_BE16(mdb->drAtrb) & kHFSVolumeUnmountedMask) == 0))
+       if (((hfsmp->hfs_flags & HFS_READ_ONLY) == 0) &&
+           ((SWAP_BE16(mdb->drAtrb) & kHFSVolumeUnmountedMask) == 0)) {
                return (EINVAL);
-               
+       }
+       hfsmp->hfs_flags |= HFS_STANDARD;
        /*
         * The MDB seems OK: transfer info from it into VCB
         * Note - the VCB starts out clear (all zeros)
@@ -130,7 +132,7 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
        vcb->vcbFilCnt          = SWAP_BE32 (mdb->drFilCnt);
        vcb->vcbDirCnt          = SWAP_BE32 (mdb->drDirCnt);
        bcopy(mdb->drFndrInfo, vcb->vcbFndrInfo, sizeof(vcb->vcbFndrInfo));
-       if (!hfsmp->hfs_fs_ronly)
+       if ((hfsmp->hfs_flags & HFS_READ_ONLY) == 0)
                vcb->vcbWrCnt++;        /* Compensate for write of MDB on last flush */
 
        /* convert hfs encoded name into UTF-8 string */
@@ -145,10 +147,12 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
        hfsmp->hfs_logBlockSize = BestBlockSizeFit(vcb->blockSize, MAXBSIZE, hfsmp->hfs_phys_block_size);
        vcb->vcbVBMIOSize = kHFSBlockSize;
 
-       VCB_LOCK_INIT(vcb);
+       hfsmp->hfs_alt_id_sector = HFS_ALT_SECTOR(hfsmp->hfs_phys_block_size,
+                                                 hfsmp->hfs_phys_block_count);
 
        bzero(&cndesc, sizeof(cndesc));
-       cndesc.cd_parentcnid = kRootParID;
+       cndesc.cd_parentcnid = kHFSRootParentID;
+       cndesc.cd_flags |= CD_ISMETA;
        bzero(&cnattr, sizeof(cnattr));
        cnattr.ca_nlink = 1;
        cnattr.ca_mode = S_IFREG;
@@ -163,6 +167,7 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
        fork.cf_size = SWAP_BE32(mdb->drXTFlSize);
        fork.cf_blocks = fork.cf_size / vcb->blockSize;
        fork.cf_clump = SWAP_BE32(mdb->drXTClpSiz);
+       fork.cf_vblocks = 0;
        fork.cf_extents[0].startBlock = SWAP_BE16(mdb->drXTExtRec[0].startBlock);
        fork.cf_extents[0].blockCount = SWAP_BE16(mdb->drXTExtRec[0].blockCount);
        fork.cf_extents[1].startBlock = SWAP_BE16(mdb->drXTExtRec[1].startBlock);
@@ -171,15 +176,13 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
        fork.cf_extents[2].blockCount = SWAP_BE16(mdb->drXTExtRec[2].blockCount);
        cnattr.ca_blocks = fork.cf_blocks;
 
-       error = hfs_getnewvnode(hfsmp, NULL, &cndesc, 0, &cnattr, &fork,
-                               &vcb->extentsRefNum);
+       error = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &fork,
+                               &hfsmp->hfs_extents_vp);
        if (error) goto MtVolErr;
-       error = MacToVFSError(BTOpenPath(VTOF(vcb->extentsRefNum),
-                                        (KeyCompareProcPtr)CompareExtentKeys,
-                                        GetBTreeBlock, ReleaseBTreeBlock, 
-                                        ExtendBTreeFile, SetBTreeBlockSize));
+       error = MacToVFSError(BTOpenPath(VTOF(hfsmp->hfs_extents_vp),
+                                        (KeyCompareProcPtr)CompareExtentKeys));
        if (error) {
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto MtVolErr;
        }
 
@@ -192,6 +195,7 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
        fork.cf_size = SWAP_BE32(mdb->drCTFlSize);
        fork.cf_blocks = fork.cf_size / vcb->blockSize;
        fork.cf_clump = SWAP_BE32(mdb->drCTClpSiz);
+       fork.cf_vblocks = 0;
        fork.cf_extents[0].startBlock = SWAP_BE16(mdb->drCTExtRec[0].startBlock);
        fork.cf_extents[0].blockCount = SWAP_BE16(mdb->drCTExtRec[0].blockCount);
        fork.cf_extents[1].startBlock = SWAP_BE16(mdb->drCTExtRec[1].startBlock);
@@ -200,19 +204,34 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
        fork.cf_extents[2].blockCount = SWAP_BE16(mdb->drCTExtRec[2].blockCount);
        cnattr.ca_blocks = fork.cf_blocks;
 
-       error = hfs_getnewvnode(hfsmp, NULL, &cndesc, 0, &cnattr, &fork,
-                               &vcb->catalogRefNum);
+       error = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &fork,
+                               &hfsmp->hfs_catalog_vp);
+       if (error) {
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
+               goto MtVolErr;
+       }
+       error = MacToVFSError(BTOpenPath(VTOF(hfsmp->hfs_catalog_vp),
+                                        (KeyCompareProcPtr)CompareCatalogKeys));
        if (error) {
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto MtVolErr;
        }
-       error = MacToVFSError(BTOpenPath(VTOF(vcb->catalogRefNum),
-                                        (KeyCompareProcPtr)CompareCatalogKeys,
-                                        GetBTreeBlock, ReleaseBTreeBlock,
-                                        ExtendBTreeFile, SetBTreeBlockSize));
+
+       /*
+        * Set up dummy Allocation file vnode (used only for locking bitmap)
+        */  
+       cndesc.cd_nameptr = hfs_vbmname;
+       cndesc.cd_namelen = strlen(hfs_vbmname);
+       cndesc.cd_cnid = cnattr.ca_fileid = kHFSAllocationFileID;
+       bzero(&fork, sizeof(fork));
+       cnattr.ca_blocks = 0;
+
+       error = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &fork,
+                                &hfsmp->hfs_allocation_vp);
        if (error) {
-               VOP_UNLOCK(vcb->catalogRefNum, 0, p);
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto MtVolErr;
        }
 
@@ -220,10 +239,11 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
        vcb->vcbAtrb &= ~kHFSVolumeUnmountedMask;
 
        /*
-        * all done with b-trees so we can unlock now...
+        * all done with system files so we can unlock now...
         */
-       VOP_UNLOCK(vcb->catalogRefNum, 0, p);
-       VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+       hfs_unlock(VTOC(hfsmp->hfs_allocation_vp));
+       hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+       hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
 
     if ( error == noErr )
       {
@@ -236,8 +256,8 @@ OSErr hfs_MountHFSVolume(struct hfsmount *hfsmp, HFSMasterDirectoryBlock *mdb,
 
     //--       Release any resources allocated so far before exiting with an error:
 MtVolErr:
-       ReleaseMetaFileVNode(vcb->catalogRefNum);
-       ReleaseMetaFileVNode(vcb->extentsRefNum);
+       ReleaseMetaFileVNode(hfsmp->hfs_catalog_vp);
+       ReleaseMetaFileVNode(hfsmp->hfs_extents_vp);
 
 CmdDone:
     return (error);
@@ -249,38 +269,59 @@ CmdDone:
 //
 //*******************************************************************************
 
+__private_extern__
 OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
-       off_t embeddedOffset, u_int64_t disksize, struct proc *p, void *args)
+       off_t embeddedOffset, u_int64_t disksize, struct proc *p, void *args, kauth_cred_t cred)
 {
        register ExtendedVCB *vcb;
        struct cat_desc cndesc;
        struct cat_attr cnattr;
+       struct cat_fork cfork;
        UInt32 blockSize;
+       daddr64_t spare_sectors;
+       struct BTreeInfoRec btinfo;
+       u_int16_t  signature;
+       u_int16_t  version;
+       int  i;
        OSErr retval;
 
-       // XXXdbg - added the kHFSJSigWord case
-       if ((SWAP_BE16(vhp->signature) != kHFSPlusSigWord &&
-                SWAP_BE16(vhp->signature) != kHFSJSigWord) ||
-           SWAP_BE16(vhp->version) != kHFSPlusVersion) {
-               // XXXdbg
-               printf("hfs: mount: sig 0x%x and version 0x%x are not HFS or HFS+.\n",
-                          vhp->signature, vhp->version);
+       signature = SWAP_BE16(vhp->signature);
+       version = SWAP_BE16(vhp->version);
+
+       if (signature == kHFSPlusSigWord) {
+               if (version != kHFSPlusVersion) {
+                       printf("hfs_mount: invalid HFS+ version: %d\n", version);
+                       return (EINVAL);
+               }
+       } else if (signature == kHFSXSigWord) {
+               if (version != kHFSXVersion) {
+                       printf("hfs_mount: invalid HFSX version: %d\n", version);
+                       return (EINVAL);
+               }
+               /* The in-memory signature is always 'H+'. */
+               signature = kHFSPlusSigWord;
+               hfsmp->hfs_flags |= HFS_X;
+       } else {
+               /* Removed printf for invalid HFS+ signature because it gives
+                * false error for UFS root volume 
+                */
                return (EINVAL);
        }
 
        /* Block size must be at least 512 and a power of 2 */
        blockSize = SWAP_BE32(vhp->blockSize);
-       if (blockSize < 512 || (blockSize & (blockSize-1)) != 0)
+       if (blockSize < 512 || !powerof2(blockSize))
                return (EINVAL);
    
        /* don't mount a writable volume if its dirty, it must be cleaned by fsck_hfs */
-       if (hfsmp->hfs_fs_ronly == 0 && hfsmp->jnl == NULL && (SWAP_BE32(vhp->attributes) & kHFSVolumeUnmountedMask) == 0)
+       if ((hfsmp->hfs_flags & HFS_READ_ONLY) == 0 && hfsmp->jnl == NULL &&
+           (SWAP_BE32(vhp->attributes) & kHFSVolumeUnmountedMask) == 0)
                return (EINVAL);
 
        /* Make sure we can live with the physical block size. */
        if ((disksize & (hfsmp->hfs_phys_block_size - 1)) ||
            (embeddedOffset & (hfsmp->hfs_phys_block_size - 1)) ||
-           (SWAP_BE32(vhp->blockSize) < hfsmp->hfs_phys_block_size)) {
+           (blockSize < hfsmp->hfs_phys_block_size)) {
                return (ENXIO);
        }
        /*
@@ -289,16 +330,10 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
         */
        vcb = HFSTOVCB(hfsmp);
 
-       vcb->vcbSigWord = SWAP_BE16(vhp->signature);
-
-       // XXXdbg - remap this in case we've mounted a dirty journaled volume
-       if (vcb->vcbSigWord == kHFSJSigWord) {
-               vcb->vcbSigWord = kHFSPlusSigWord;
-       }
-
+       vcb->vcbSigWord = signature;
        vcb->vcbJinfoBlock = SWAP_BE32(vhp->journalInfoBlock);
        vcb->vcbLsMod   = to_bsd_time(SWAP_BE32(vhp->modifyDate));
-       vcb->vcbAtrb    = (UInt16)SWAP_BE32(vhp->attributes);
+       vcb->vcbAtrb    = SWAP_BE32(vhp->attributes);
        vcb->vcbClpSiz  = SWAP_BE32(vhp->rsrcClumpSize);
        vcb->vcbNxtCNID = SWAP_BE32(vhp->nextCatalogID);
        vcb->vcbVolBkUp = to_bsd_time(SWAP_BE32(vhp->backupDate));
@@ -310,16 +345,14 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
        bcopy(vhp->finderInfo, vcb->vcbFndrInfo, sizeof(vhp->finderInfo));    
 
        vcb->vcbAlBlSt = 0;             /* hfs+ allocation blocks start at first block of volume */
-       if (!hfsmp->hfs_fs_ronly)
+       if ((hfsmp->hfs_flags & HFS_READ_ONLY) == 0)
                vcb->vcbWrCnt++;        /* compensate for write of Volume Header on last flush */
 
-       VCB_LOCK_INIT(vcb);
-
        /* Now fill in the Extended VCB info */
        vcb->nextAllocation     = SWAP_BE32(vhp->nextAllocation);
        vcb->totalBlocks        = SWAP_BE32(vhp->totalBlocks);
        vcb->freeBlocks         = SWAP_BE32(vhp->freeBlocks);
-       vcb->blockSize          = SWAP_BE32(vhp->blockSize);
+       vcb->blockSize          = blockSize;
        vcb->encodingsBitmap    = SWAP_BE64(vhp->encodingsBitmap);
        vcb->localCreateDate    = SWAP_BE32(vhp->createDate);
        
@@ -336,8 +369,24 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
        hfsmp->hfs_logBlockSize = BestBlockSizeFit(vcb->blockSize, MAXBSIZE, hfsmp->hfs_phys_block_size);
        vcb->vcbVBMIOSize = min(vcb->blockSize, MAXPHYSIO);
 
+       /*
+        * Validate and initialize the location of the alternate volume header.
+        */
+       spare_sectors = hfsmp->hfs_phys_block_count -
+                       (((daddr64_t)vcb->totalBlocks * blockSize) /
+                          hfsmp->hfs_phys_block_size);
+
+       if (spare_sectors > (blockSize / hfsmp->hfs_phys_block_size)) {
+               hfsmp->hfs_alt_id_sector = 0;  /* partition has grown! */
+       } else {
+               hfsmp->hfs_alt_id_sector = (hfsmp->hfsPlusIOPosOffset / hfsmp->hfs_phys_block_size) +
+                                          HFS_ALT_SECTOR(hfsmp->hfs_phys_block_size,
+                                                         hfsmp->hfs_phys_block_count);
+       }
+
        bzero(&cndesc, sizeof(cndesc));
-       cndesc.cd_parentcnid = kRootParID;
+       cndesc.cd_parentcnid = kHFSRootParentID;
+       cndesc.cd_flags |= CD_ISMETA;
        bzero(&cnattr, sizeof(cnattr));
        cnattr.ca_nlink = 1;
        cnattr.ca_mode = S_IFREG;
@@ -349,21 +398,25 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
        cndesc.cd_namelen = strlen(hfs_extname);
        cndesc.cd_cnid = cnattr.ca_fileid = kHFSExtentsFileID;
 
-       SWAP_HFS_PLUS_FORK_DATA (&vhp->extentsFile);
-       cnattr.ca_blocks = vhp->extentsFile.totalBlocks;
-
-       retval = hfs_getnewvnode(hfsmp, NULL, &cndesc, 0, &cnattr,
-                                (struct cat_fork *)&vhp->extentsFile,
-                                &vcb->extentsRefNum);
-       SWAP_HFS_PLUS_FORK_DATA (&vhp->extentsFile);
+       cfork.cf_size    = SWAP_BE64 (vhp->extentsFile.logicalSize);
+       cfork.cf_clump   = SWAP_BE32 (vhp->extentsFile.clumpSize);
+       cfork.cf_blocks  = SWAP_BE32 (vhp->extentsFile.totalBlocks);
+       cfork.cf_vblocks = 0;
+       cnattr.ca_blocks = cfork.cf_blocks;
+       for (i = 0; i < kHFSPlusExtentDensity; i++) {
+               cfork.cf_extents[i].startBlock =
+                               SWAP_BE32 (vhp->extentsFile.extents[i].startBlock);
+               cfork.cf_extents[i].blockCount =
+                               SWAP_BE32 (vhp->extentsFile.extents[i].blockCount);
+       }
+       retval = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &cfork,
+                                &hfsmp->hfs_extents_vp);
 
        if (retval) goto ErrorExit;
-       retval = MacToVFSError(BTOpenPath(VTOF(vcb->extentsRefNum),
-                                         (KeyCompareProcPtr) CompareExtentKeysPlus,
-                                         GetBTreeBlock, ReleaseBTreeBlock,
-                                         ExtendBTreeFile, SetBTreeBlockSize));
+       retval = MacToVFSError(BTOpenPath(VTOF(hfsmp->hfs_extents_vp),
+                                         (KeyCompareProcPtr) CompareExtentKeysPlus));
        if (retval) {
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto ErrorExit;
        }
 
@@ -374,26 +427,39 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
        cndesc.cd_namelen = strlen(hfs_catname);
        cndesc.cd_cnid = cnattr.ca_fileid = kHFSCatalogFileID;
 
-       SWAP_HFS_PLUS_FORK_DATA(&vhp->catalogFile);
-       cnattr.ca_blocks = vhp->catalogFile.totalBlocks;
-
-       retval = hfs_getnewvnode(hfsmp, NULL, &cndesc, 0, &cnattr,
-                                (struct cat_fork *)&vhp->catalogFile,
-                                &vcb->catalogRefNum);
-       SWAP_HFS_PLUS_FORK_DATA(&vhp->catalogFile);
+       cfork.cf_size    = SWAP_BE64 (vhp->catalogFile.logicalSize);
+       cfork.cf_clump   = SWAP_BE32 (vhp->catalogFile.clumpSize);
+       cfork.cf_blocks  = SWAP_BE32 (vhp->catalogFile.totalBlocks);
+       cfork.cf_vblocks = 0;
+       cnattr.ca_blocks = cfork.cf_blocks;
+       for (i = 0; i < kHFSPlusExtentDensity; i++) {
+               cfork.cf_extents[i].startBlock =
+                               SWAP_BE32 (vhp->catalogFile.extents[i].startBlock);
+               cfork.cf_extents[i].blockCount =
+                               SWAP_BE32 (vhp->catalogFile.extents[i].blockCount);
+       }
+       retval = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &cfork,
+                                &hfsmp->hfs_catalog_vp);
        if (retval) {
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto ErrorExit;
        }
-       retval = MacToVFSError(BTOpenPath(VTOF(vcb->catalogRefNum),
-                                         (KeyCompareProcPtr) CompareExtendedCatalogKeys,
-                                         GetBTreeBlock, ReleaseBTreeBlock,
-                                         ExtendBTreeFile, SetBTreeBlockSize));
+       retval = MacToVFSError(BTOpenPath(VTOF(hfsmp->hfs_catalog_vp),
+                                         (KeyCompareProcPtr) CompareExtendedCatalogKeys));
        if (retval) {
-               VOP_UNLOCK(vcb->catalogRefNum, 0, p);
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto ErrorExit;
        }
+       if ((hfsmp->hfs_flags & HFS_X) &&
+           BTGetInformation(VTOF(hfsmp->hfs_catalog_vp), 0, &btinfo) == 0) {
+               if (btinfo.keyCompareType == kHFSBinaryCompare) {
+                       hfsmp->hfs_flags |= HFS_CASE_SENSITIVE;
+                       /* Install a case-sensitive key compare */
+                       (void) BTOpenPath(VTOF(hfsmp->hfs_catalog_vp),
+                                         (KeyCompareProcPtr)cat_binarykeycompare);
+               }
+       }
 
        /*
         * Set up Allocation file vnode
@@ -402,25 +468,70 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
        cndesc.cd_namelen = strlen(hfs_vbmname);
        cndesc.cd_cnid = cnattr.ca_fileid = kHFSAllocationFileID;
 
-       SWAP_HFS_PLUS_FORK_DATA(&vhp->allocationFile);
-       cnattr.ca_blocks = vhp->allocationFile.totalBlocks;
-
-       retval = hfs_getnewvnode(hfsmp, NULL, &cndesc, 0, &cnattr,
-                                (struct cat_fork *)&vhp->allocationFile,
-                                &vcb->allocationsRefNum);
-       SWAP_HFS_PLUS_FORK_DATA(&vhp->allocationFile);
+       cfork.cf_size    = SWAP_BE64 (vhp->allocationFile.logicalSize);
+       cfork.cf_clump   = SWAP_BE32 (vhp->allocationFile.clumpSize);
+       cfork.cf_blocks  = SWAP_BE32 (vhp->allocationFile.totalBlocks);
+       cfork.cf_vblocks = 0;
+       cnattr.ca_blocks = cfork.cf_blocks;
+       for (i = 0; i < kHFSPlusExtentDensity; i++) {
+               cfork.cf_extents[i].startBlock =
+                               SWAP_BE32 (vhp->allocationFile.extents[i].startBlock);
+               cfork.cf_extents[i].blockCount =
+                               SWAP_BE32 (vhp->allocationFile.extents[i].blockCount);
+       }
+       retval = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &cfork,
+                                &hfsmp->hfs_allocation_vp);
        if (retval) {
-               VOP_UNLOCK(vcb->catalogRefNum, 0, p);
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto ErrorExit;
        }
 
+       /*
+        * Set up Attribute B-tree vnode
+        */
+       if (vhp->attributesFile.totalBlocks != 0) {
+               cndesc.cd_nameptr = hfs_attrname;
+               cndesc.cd_namelen = strlen(hfs_attrname);
+               cndesc.cd_cnid = cnattr.ca_fileid = kHFSAttributesFileID;
+       
+               cfork.cf_size    = SWAP_BE64 (vhp->attributesFile.logicalSize);
+               cfork.cf_clump   = SWAP_BE32 (vhp->attributesFile.clumpSize);
+               cfork.cf_blocks  = SWAP_BE32 (vhp->attributesFile.totalBlocks);
+               cfork.cf_vblocks = 0;
+               cnattr.ca_blocks = cfork.cf_blocks;
+               for (i = 0; i < kHFSPlusExtentDensity; i++) {
+                       cfork.cf_extents[i].startBlock =
+                                       SWAP_BE32 (vhp->attributesFile.extents[i].startBlock);
+                       cfork.cf_extents[i].blockCount =
+                                       SWAP_BE32 (vhp->attributesFile.extents[i].blockCount);
+               }
+               retval = hfs_getnewvnode(hfsmp, NULL, NULL, &cndesc, 0, &cnattr, &cfork,
+                                        &hfsmp->hfs_attribute_vp);
+               if (retval) {
+                       hfs_unlock(VTOC(hfsmp->hfs_allocation_vp));
+                       hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+                       hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
+                       goto ErrorExit;
+               }
+               retval = MacToVFSError(BTOpenPath(VTOF(hfsmp->hfs_attribute_vp),
+                                                 (KeyCompareProcPtr) hfs_attrkeycompare));
+               if (retval) {
+                       hfs_unlock(VTOC(hfsmp->hfs_attribute_vp));
+                       hfs_unlock(VTOC(hfsmp->hfs_allocation_vp));
+                       hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+                       hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
+                       goto ErrorExit;
+               }
+       }
+
+
        /* Pick up volume name and create date */
        retval = cat_idlookup(hfsmp, kHFSRootFolderID, &cndesc, &cnattr, NULL);
        if (retval) {
-               VOP_UNLOCK(vcb->allocationsRefNum, 0, p);
-               VOP_UNLOCK(vcb->catalogRefNum, 0, p);
-               VOP_UNLOCK(vcb->extentsRefNum, 0, p);
+               hfs_unlock(VTOC(hfsmp->hfs_allocation_vp));
+               hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
                goto ErrorExit;
        }
        vcb->vcbCrDate = cnattr.ca_itime;
@@ -430,27 +541,18 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
 
        /* mark the volume dirty (clear clean unmount bit) */
        vcb->vcbAtrb &= ~kHFSVolumeUnmountedMask;
-       if (hfsmp->jnl && hfsmp->hfs_fs_ronly == 0) {
-               hfs_flushvolumeheader(hfsmp, TRUE, TRUE);
+       if (hfsmp->jnl && (hfsmp->hfs_flags & HFS_READ_ONLY) == 0) {
+               hfs_flushvolumeheader(hfsmp, TRUE, 0);
        }
 
        /*
         * all done with metadata files so we can unlock now...
         */
-       VOP_UNLOCK(vcb->allocationsRefNum, 0, p);
-       VOP_UNLOCK(vcb->catalogRefNum, 0, p);
-       VOP_UNLOCK(vcb->extentsRefNum, 0, p);
-
-       /* setup private/hidden directory for unlinked files */
-       hfsmp->hfs_private_metadata_dir = FindMetaDataDirectory(vcb);
-       if (hfsmp->jnl && (hfsmp->hfs_fs_ronly == 0))
-               hfs_remove_orphans(hfsmp);
-
-       if ( !(vcb->vcbAtrb & kHFSVolumeHardwareLockMask) )     // if the disk is not write protected
-       {
-               MarkVCBDirty( vcb );    // mark VCB dirty so it will be written
-       }
-
+       if (hfsmp->hfs_attribute_vp)
+               hfs_unlock(VTOC(hfsmp->hfs_attribute_vp));
+       hfs_unlock(VTOC(hfsmp->hfs_allocation_vp));
+       hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+       hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
 
        //
        // Check if we need to do late journal initialization.  This only
@@ -465,9 +567,42 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
                retval = hfs_late_journal_init(hfsmp, vhp, args);
                if (retval != 0) {
                        hfsmp->jnl = NULL;
+                       
+                       // if the journal failed to open, then set the lastMountedVersion
+                       // to be "FSK!" which fsck_hfs will see and force the fsck instead
+                       // of just bailing out because the volume is journaled.
+                       if (!(hfsmp->hfs_flags & HFS_READ_ONLY)) {
+                               HFSPlusVolumeHeader *jvhp;
+                               daddr64_t mdb_offset;
+                               struct buf *bp = NULL;
+                               
+                               hfsmp->hfs_flags |= HFS_NEED_JNL_RESET;
+                                   
+                               mdb_offset = (daddr64_t)((embeddedOffset / blockSize) + HFS_PRI_SECTOR(blockSize));
+
+                               retval = (int)buf_meta_bread(hfsmp->hfs_devvp, mdb_offset, blockSize, cred, &bp);
+                               if (retval == 0) {
+                                       jvhp = (HFSPlusVolumeHeader *)(buf_dataptr(bp) + HFS_PRI_OFFSET(blockSize));
+                                           
+                                       if (SWAP_BE16(jvhp->signature) == kHFSPlusSigWord || SWAP_BE16(jvhp->signature) == kHFSXSigWord) {
+                                               printf ("hfs(3): Journal replay fail.  Writing lastMountVersion as FSK!\n");
+                                               jvhp->lastMountedVersion = SWAP_BE32(kFSKMountVersion);
+                                               buf_bwrite(bp);
+                                       } else {
+                                               buf_brelse(bp);
+                                       }
+                                       bp = NULL;
+                               } else if (bp) {
+                                       buf_brelse(bp);
+                                       // clear this so the error exit path won't try to use it
+                                       bp = NULL;
+                           }
+                       }
+
+                       retval = EINVAL;
                        goto ErrorExit;
                } else if (hfsmp->jnl) {
-                       hfsmp->hfs_mp->mnt_flag |= MNT_JOURNALED;
+                       vfs_setflags(hfsmp->hfs_mp, (uint64_t)((unsigned int)MNT_JOURNALED));
                }
        } else if (hfsmp->jnl) {
                struct cat_attr jinfo_attr, jnl_attr;
@@ -482,19 +617,54 @@ OSErr hfs_MountHFSPlusVolume(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
                }
        }
 
+       /*
+        * Establish a metadata allocation zone.
+        */
+       hfs_metadatazone_init(hfsmp);
+
+       /*
+        * Make any metadata zone adjustments.
+        */
+       if (hfsmp->hfs_flags & HFS_METADATA_ZONE) {
+               /* Keep the roving allocator out of the metadata zone. */
+               if (vcb->nextAllocation >= hfsmp->hfs_metazone_start &&
+                   vcb->nextAllocation <= hfsmp->hfs_metazone_end) {       
+                       vcb->nextAllocation = hfsmp->hfs_metazone_end + 1;
+               }
+       }
+
+       /* setup private/hidden directory for unlinked files */
+       FindMetaDataDirectory(vcb);
+       if ((hfsmp->hfs_flags & HFS_READ_ONLY) == 0) 
+               hfs_remove_orphans(hfsmp);
+
+       if ( !(vcb->vcbAtrb & kHFSVolumeHardwareLockMask) )     // if the disk is not write protected
+       {
+               MarkVCBDirty( vcb );    // mark VCB dirty so it will be written
+       }
+
+       /*
+        * Allow hot file clustering if conditions allow.
+        */
+       if ((hfsmp->hfs_flags & HFS_METADATA_ZONE)  &&
+           ((hfsmp->hfs_flags & HFS_READ_ONLY) == 0)) {
+               (void) hfs_recording_init(hfsmp);
+       }
+
+       hfs_checkextendedsecurity(hfsmp);
 
        return (0);
 
 ErrorExit:
        /*
-        * A fatal error occured and the volume cannot be mounted
+        * A fatal error occurred and the volume cannot be mounted
         * release any resources that we aquired...
         */
-
-       InvalidateCatalogCache(vcb);   
-       ReleaseMetaFileVNode(vcb->allocationsRefNum);
-       ReleaseMetaFileVNode(vcb->catalogRefNum);
-       ReleaseMetaFileVNode(vcb->extentsRefNum);
+       if (hfsmp->hfs_attribute_vp)
+               ReleaseMetaFileVNode(hfsmp->hfs_attribute_vp);
+       ReleaseMetaFileVNode(hfsmp->hfs_allocation_vp);
+       ReleaseMetaFileVNode(hfsmp->hfs_catalog_vp);
+       ReleaseMetaFileVNode(hfsmp->hfs_extents_vp);
 
        return (retval);
 }
@@ -510,12 +680,15 @@ static void ReleaseMetaFileVNode(struct vnode *vp)
        struct filefork *fp;
 
        if (vp && (fp = VTOF(vp))) {
-               if (fp->fcbBTCBPtr != NULL)
+               if (fp->fcbBTCBPtr != NULL) {
+                       (void)hfs_lock(VTOC(vp), HFS_EXCLUSIVE_LOCK);
                        (void) BTClosePath(fp);
+                       hfs_unlock(VTOC(vp));
+               }
 
                /* release the node even if BTClosePath fails */
-               vrele(vp);
-               vgone(vp);
+               vnode_recycle(vp);
+               vnode_put(vp);
        }
 }
 
@@ -527,34 +700,29 @@ static void ReleaseMetaFileVNode(struct vnode *vp)
 *
 *************************************************************/
 
-short hfsUnmount( register struct hfsmount *hfsmp, struct proc *p)
+__private_extern__
+int
+hfsUnmount( register struct hfsmount *hfsmp, struct proc *p)
 {
-       ExtendedVCB *vcb = HFSTOVCB(hfsmp);
-       int retval = E_NONE;
-
-       InvalidateCatalogCache( vcb );
+       if (hfsmp->hfs_allocation_vp)
+               ReleaseMetaFileVNode(hfsmp->hfs_allocation_vp);
 
-       if (vcb->vcbSigWord == kHFSPlusSigWord)
-               ReleaseMetaFileVNode(vcb->allocationsRefNum);
+       if (hfsmp->hfs_attribute_vp)
+               ReleaseMetaFileVNode(hfsmp->hfs_attribute_vp);
 
-       ReleaseMetaFileVNode(vcb->catalogRefNum);
-       ReleaseMetaFileVNode(vcb->extentsRefNum);
+       ReleaseMetaFileVNode(hfsmp->hfs_catalog_vp);
+       ReleaseMetaFileVNode(hfsmp->hfs_extents_vp);
 
-       return (retval);
+       return (0);
 }
 
 
 /*
- * Some 3rd party kexts link against hfs_getcatalog so keep a stub for now.
+ * Test if fork has overflow extents.
  */
-short
-hfs_getcatalog(void *p1, u_long p2, void *p3, short p4, void *p5)
-{
-       return ENOENT;
-}
-
-
-int overflow_extents(struct filefork *fp)
+__private_extern__
+int
+overflow_extents(struct filefork *fp)
 {
        u_long blocks;
 
@@ -583,53 +751,129 @@ int overflow_extents(struct filefork *fp)
 }
 
 
-/* __private_extern__ */
+/*
+ * Lock HFS system file(s).
+ */
+__private_extern__
 int
-hfs_metafilelocking(struct hfsmount *hfsmp, u_long fileID, u_int flags, struct proc *p)
+hfs_systemfile_lock(struct hfsmount *hfsmp, int flags, enum hfslocktype locktype)
 {
-       ExtendedVCB             *vcb;
-       struct vnode    *vp = NULL;
-       int                             numOfLockedBuffs;
-       int     retval = 0;
-
-       vcb = HFSTOVCB(hfsmp);
-
-       switch (fileID) {
-       case kHFSExtentsFileID:
-               vp = vcb->extentsRefNum;
-               break;
+       if (flags & ~SFL_VALIDMASK)
+               panic("hfs_systemfile_lock: invalid lock request (0x%x)", (unsigned long) flags);
+       /*
+        * Locking order is Catalog file, Attributes file, Bitmap file, Extents file
+        */
+       if (flags & SFL_CATALOG) {
+               (void) hfs_lock(VTOC(hfsmp->hfs_catalog_vp), locktype);
+               /*
+                * When the catalog file has overflow extents then
+                * also acquire the extents b-tree lock if its not
+                * already requested.
+                */
+               if ((flags & SFL_EXTENTS) == 0 &&
+                   overflow_extents(VTOF(hfsmp->hfs_catalog_vp))) {
+                       flags |= SFL_EXTENTS;
+               }
+       }
+       if (flags & SFL_ATTRIBUTE) {
+               if (hfsmp->hfs_attribute_vp) {
+                       (void) hfs_lock(VTOC(hfsmp->hfs_attribute_vp), locktype);
+                       /*
+                        * When the attribute file has overflow extents then
+                        * also acquire the extents b-tree lock if its not
+                        * already requested.
+                        */
+                       if ((flags & SFL_EXTENTS) == 0 &&
+                           overflow_extents(VTOF(hfsmp->hfs_attribute_vp))) {
+                               flags |= SFL_EXTENTS;
+                       }
+               } else {
+                       flags &= ~SFL_ATTRIBUTE;
+               }
+       }
+       if (flags & SFL_BITMAP) {
+               /*
+                * Since the only bitmap operations are clearing and
+                * setting bits we always need exclusive access. And
+                * when we have a journal, we can "hide" behind that
+                * lock since we can only change the bitmap from
+                * within a transaction.
+                */
+               if (hfsmp->jnl) {
+                       flags &= ~SFL_BITMAP;
+               } else {
+                       (void) hfs_lock(VTOC(hfsmp->hfs_allocation_vp), HFS_EXCLUSIVE_LOCK);
+               }
+       }
+       if (flags & SFL_EXTENTS) {
+               /*
+                * Since the extents btree lock is recursive we always
+                * need exclusive access.
+                */
+               (void) hfs_lock(VTOC(hfsmp->hfs_extents_vp), HFS_EXCLUSIVE_LOCK);
+       }
+       return (flags);
+}
 
-       case kHFSCatalogFileID:
-               vp = vcb->catalogRefNum;
-               break;
+/*
+ * unlock HFS system file(s).
+ */
+__private_extern__
+void
+hfs_systemfile_unlock(struct hfsmount *hfsmp, int flags)
+{
+       struct timeval tv;
+       u_int32_t lastfsync;
+       int numOfLockedBuffs;
 
-       case kHFSAllocationFileID:
-               /* bitmap is covered by Extents B-tree locking */
-               /* FALL THROUGH */
-       default:
-               panic("hfs_lockmetafile: invalid fileID");
+       microuptime(&tv);
+       lastfsync = tv.tv_sec;
+       
+       if (flags & ~SFL_VALIDMASK)
+               panic("hfs_systemfile_unlock: invalid lock request (0x%x)", (unsigned long) flags);
+
+       if (flags & SFL_ATTRIBUTE && hfsmp->hfs_attribute_vp) {
+               if (hfsmp->jnl == NULL) {
+                       BTGetLastSync((FCB*)VTOF(hfsmp->hfs_attribute_vp), &lastfsync);
+                       numOfLockedBuffs = count_lock_queue();
+                       if ((numOfLockedBuffs > kMaxLockedMetaBuffers) ||
+                           ((numOfLockedBuffs > 1) && ((tv.tv_sec - lastfsync) >
+                             kMaxSecsForFsync))) {
+                               hfs_btsync(hfsmp->hfs_attribute_vp, HFS_SYNCTRANS);
+                       }
+               }
+               hfs_unlock(VTOC(hfsmp->hfs_attribute_vp));
        }
-
-       /* Release, if necesary any locked buffer caches */
-       if ((flags & LK_TYPE_MASK) == LK_RELEASE) {
-               struct timeval tv = time;
-               u_int32_t               lastfsync = tv.tv_sec; 
-               
-               (void) BTGetLastSync((FCB*)VTOF(vp), &lastfsync);
-               
-               numOfLockedBuffs = count_lock_queue();
-               if ((numOfLockedBuffs > kMaxLockedMetaBuffers) || ((numOfLockedBuffs>1) && ((tv.tv_sec - lastfsync) > kMaxSecsForFsync))) {
-                       hfs_btsync(vp, HFS_SYNCTRANS);
+       if (flags & SFL_CATALOG) {
+               if (hfsmp->jnl == NULL) {
+                       BTGetLastSync((FCB*)VTOF(hfsmp->hfs_catalog_vp), &lastfsync);
+                       numOfLockedBuffs = count_lock_queue();
+                       if ((numOfLockedBuffs > kMaxLockedMetaBuffers) ||
+                           ((numOfLockedBuffs > 1) && ((tv.tv_sec - lastfsync) >
+                             kMaxSecsForFsync))) {
+                               hfs_btsync(hfsmp->hfs_catalog_vp, HFS_SYNCTRANS);
+                       }
                }
-       } else {
-               flags |= LK_RETRY;
+               hfs_unlock(VTOC(hfsmp->hfs_catalog_vp));
+       }
+       if (flags & SFL_BITMAP) {
+               hfs_unlock(VTOC(hfsmp->hfs_allocation_vp));
+       }
+       if (flags & SFL_EXTENTS) {
+               if (hfsmp->jnl == NULL) {
+                       BTGetLastSync((FCB*)VTOF(hfsmp->hfs_extents_vp), &lastfsync);
+                       numOfLockedBuffs = count_lock_queue();
+                       if ((numOfLockedBuffs > kMaxLockedMetaBuffers) ||
+                           ((numOfLockedBuffs > 1) && ((tv.tv_sec - lastfsync) >
+                             kMaxSecsForFsync))) {
+                               hfs_btsync(hfsmp->hfs_extents_vp, HFS_SYNCTRANS);
+                       }
+               }
+               hfs_unlock(VTOC(hfsmp->hfs_extents_vp));
        }
-       
-       retval = lockmgr(&VTOC(vp)->c_lock, flags, &vp->v_interlock, p);
-
-       return (retval);
 }
 
+
 /*
  * RequireFileLock
  *
@@ -639,37 +883,32 @@ hfs_metafilelocking(struct hfsmount *hfsmp, u_long fileID, u_int flags, struct p
 #if HFS_DIAGNOSTIC
 void RequireFileLock(FileReference vp, int shareable)
 {
-       struct lock__bsd__ *lkp;
-       int locked = false;
-       pid_t pid;
-       void * self;
+       int locked;
 
-       pid = current_proc()->p_pid;
-       self = (void *) current_thread();
-       lkp = &VTOC(vp)->c_lock;
-
-       simple_lock(&lkp->lk_interlock);
+       /* The extents btree and allocation bitmap are always exclusive. */
+       if (VTOC(vp)->c_fileid == kHFSExtentsFileID ||
+           VTOC(vp)->c_fileid == kHFSAllocationFileID) {
+               shareable = 0;
+       }
        
-       if (shareable && (lkp->lk_sharecount > 0) && (lkp->lk_lockholder == LK_NOPROC))
-               locked = true;
-       else if ((lkp->lk_exclusivecount > 0) && (lkp->lk_lockholder == pid) && (lkp->lk_lockthread == self))
-               locked = true;
-
-       simple_unlock(&lkp->lk_interlock);
+       locked = VTOC(vp)->c_lockowner == (void *)current_thread();
        
-       if (!locked) {
+       if (!locked && !shareable) {
                switch (VTOC(vp)->c_fileid) {
-                       case 3:
-                               DEBUG_BREAK_MSG((" #\n # RequireFileLock: extent btree vnode not locked! v: 0x%08X\n #\n", (u_int)vp));
-                               break;
-
-                       case 4:
-                               DEBUG_BREAK_MSG((" #\n # RequireFileLock: catalog btree vnode not locked! v: 0x%08X\n #\n", (u_int)vp));
-                               break;
-
-                       default:
-                               DEBUG_BREAK_MSG((" #\n # RequireFileLock: file (%d) not locked! v: 0x%08X\n #\n", VTOC(vp)->c_fileid, (u_int)vp));
-                               break;
+               case kHFSExtentsFileID:
+                       panic("extents btree not locked! v: 0x%08X\n #\n", (u_int)vp);
+                       break;
+               case kHFSCatalogFileID:
+                       panic("catalog btree not locked! v: 0x%08X\n #\n", (u_int)vp);
+                       break;
+               case kHFSAllocationFileID:
+                       /* The allocation file can hide behind the jornal lock. */
+                       if (VTOHFS(vp)->jnl == NULL)
+                               panic("allocation file not locked! v: 0x%08X\n #\n", (u_int)vp);
+                       break;
+               case kHFSAttributesFileID:
+                       panic("attributes btree not locked! v: 0x%08X\n #\n", (u_int)vp);
+                       break;
                }
        }
 }
@@ -680,27 +919,24 @@ void RequireFileLock(FileReference vp, int shareable)
  * There are three ways to qualify for ownership rights on an object:
  *
  * 1. (a) Your UID matches the cnode's UID.
- *    (b) The object in question is owned by "unknown" and
- *        your UID matches the console user's UID.
+ *    (b) The object in question is owned by "unknown"
  * 2. (a) Permissions on the filesystem are being ignored and
  *        your UID matches the replacement UID.
  *    (b) Permissions on the filesystem are being ignored and
- *        the replacement UID is "unknown" and
- *        your UID matches the console user UID.
+ *        the replacement UID is "unknown".
  * 3. You are root.
  *
  */
 int
-hfs_owner_rights(struct hfsmount *hfsmp, uid_t cnode_uid, struct ucred *cred,
+hfs_owner_rights(struct hfsmount *hfsmp, uid_t cnode_uid, kauth_cred_t cred,
                struct proc *p, int invokesuperuserstatus)
 {
-       if ((cred->cr_uid == cnode_uid) ||                                    /* [1a] */
-           ((cnode_uid == UNKNOWNUID) && (cred->cr_uid == console_user)) ||  /* [1b] */
-           ((HFSTOVFS(hfsmp)->mnt_flag & MNT_UNKNOWNPERMISSIONS) &&          /* [2] */
-             ((cred->cr_uid == hfsmp->hfs_uid) ||                            /* [2a] */
-               ((hfsmp->hfs_uid == UNKNOWNUID) &&                            /* [2b] */
-                 (cred->cr_uid == console_user)))) ||
-           (invokesuperuserstatus && (suser(cred, &p->p_acflag) == 0))) {    /* [3] */
+       if ((kauth_cred_getuid(cred) == cnode_uid) ||                                    /* [1a] */
+           (cnode_uid == UNKNOWNUID) ||                                                                          /* [1b] */
+           ((((unsigned int)vfs_flags(HFSTOVFS(hfsmp))) & MNT_UNKNOWNPERMISSIONS) &&          /* [2] */
+             ((kauth_cred_getuid(cred) == hfsmp->hfs_uid) ||                            /* [2a] */
+               (hfsmp->hfs_uid == UNKNOWNUID))) ||                           /* [2b] */
+           (invokesuperuserstatus && (suser(cred, 0) == 0))) {    /* [3] */
                return (0);
        } else {        
                return (EPERM);
@@ -755,8 +991,9 @@ unsigned long BestBlockSizeFit(unsigned long allocationBlockSize,
  * To make the HFS Plus filesystem follow UFS unlink semantics, a remove
  * of an active vnode is translated to a move/rename so the file appears
  * deleted. The destination folder for these move/renames is setup here
- * and a reference to it is place in hfsmp->hfs_private_metadata_dir.
+ * and a reference to it is place in hfsmp->hfs_privdir_desc.
  */
+__private_extern__
 u_long
 FindMetaDataDirectory(ExtendedVCB *vcb)
 {
@@ -765,7 +1002,10 @@ FindMetaDataDirectory(ExtendedVCB *vcb)
        struct cnode * dcp = NULL;
        struct FndrDirInfo * fndrinfo;
        struct cat_desc out_desc = {0};
+       struct proc *p = current_proc();
        struct timeval tv;
+       cat_cookie_t cookie;
+       int lockflags;
        int error;
        
        if (vcb->vcbSigWord != kHFSPlusSigWord)
@@ -780,32 +1020,48 @@ FindMetaDataDirectory(ExtendedVCB *vcb)
                hfsmp->hfs_privdir_desc.cd_flags = CD_ISDIR;
        }
 
-       /* Lock catalog b-tree */
-       error = hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_EXCLUSIVE, current_proc());    
-       if (error)
-               return (0);
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_SHARED_LOCK);
 
        error = cat_lookup(hfsmp, &hfsmp->hfs_privdir_desc, 0, NULL,
-                       &hfsmp->hfs_privdir_attr, NULL);
+                       &hfsmp->hfs_privdir_attr, NULL, NULL);
+
+       hfs_systemfile_unlock(hfsmp, lockflags);
 
        if (error == 0) {
-               /* Unlock catalog b-tree */
-               (void) hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_RELEASE, current_proc());
                hfsmp->hfs_metadata_createdate = hfsmp->hfs_privdir_attr.ca_itime;
+               hfsmp->hfs_privdir_desc.cd_cnid = hfsmp->hfs_privdir_attr.ca_fileid;
+               /*
+                * Clear the system immutable flag if set...
+                */
+               if ((hfsmp->hfs_privdir_attr.ca_flags & SF_IMMUTABLE) &&
+                   (hfsmp->hfs_flags & HFS_READ_ONLY) == 0) {
+                       hfsmp->hfs_privdir_attr.ca_flags &= ~SF_IMMUTABLE;
+
+                       if ((error = hfs_start_transaction(hfsmp)) != 0) {
+                           return (hfsmp->hfs_privdir_attr.ca_fileid);
+                       }
+
+                       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_SHARED_LOCK);
+                       (void) cat_update(hfsmp, &hfsmp->hfs_privdir_desc,
+                                        &hfsmp->hfs_privdir_attr, NULL, NULL);
+                       hfs_systemfile_unlock(hfsmp, lockflags);
+
+                       hfs_end_transaction(hfsmp);
+               }
                return (hfsmp->hfs_privdir_attr.ca_fileid);
-       } else if (hfsmp->hfs_fs_ronly) {
-               /* Unlock catalog b-tree */
-               (void) hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_RELEASE, current_proc());
+
+       } else if (hfsmp->hfs_flags & HFS_READ_ONLY) {
+
                return (0);
        }
     
        /* Setup the default attributes */
        bzero(&hfsmp->hfs_privdir_attr, sizeof(struct cat_attr));
        hfsmp->hfs_privdir_attr.ca_mode = S_IFDIR;
-       hfsmp->hfs_privdir_attr.ca_flags = SF_IMMUTABLE;
        hfsmp->hfs_privdir_attr.ca_nlink = 2;
        hfsmp->hfs_privdir_attr.ca_itime = vcb->vcbCrDate;
-       hfsmp->hfs_privdir_attr.ca_mtime = time.tv_sec;
+       microtime(&tv);
+       hfsmp->hfs_privdir_attr.ca_mtime = tv.tv_sec;
 
        /* hidden and off the desktop view */
        fndrinfo = (struct FndrDirInfo *)&hfsmp->hfs_privdir_attr.ca_finderinfo;
@@ -813,49 +1069,51 @@ FindMetaDataDirectory(ExtendedVCB *vcb)
        fndrinfo->frLocation.h = SWAP_BE16 (22460);
        fndrinfo->frFlags |= SWAP_BE16 (kIsInvisible + kNameLocked);            
 
-       // XXXdbg
-       hfs_global_shared_lock_acquire(hfsmp);
-       if (hfsmp->jnl) {
-           if ((error = journal_start_transaction(hfsmp->jnl)) != 0) {
-                       hfs_global_shared_lock_release(hfsmp);
-                       return (0);
-           }
+       if ((error = hfs_start_transaction(hfsmp)) != 0) {
+           return (0);
        }
+       /* Reserve some space in the Catalog file. */
+       if (cat_preflight(hfsmp, CAT_CREATE, &cookie, p) != 0) {
+           hfs_end_transaction(hfsmp);
 
-       error = cat_create(hfsmp, &hfsmp->hfs_privdir_desc,
-                       &hfsmp->hfs_privdir_attr, &out_desc);
+           return (0);
+       }
+
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_EXCLUSIVE_LOCK);
 
-       /* Unlock catalog b-tree */
-       (void) hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_RELEASE, current_proc());
+               error = cat_create(hfsmp, &hfsmp->hfs_privdir_desc,
+                               &hfsmp->hfs_privdir_attr, &out_desc);
+
+       hfs_systemfile_unlock(hfsmp, lockflags);
+
+       cat_postflight(hfsmp, &cookie, p);
+       
        if (error) {
-           if (hfsmp->jnl) {
-                       journal_end_transaction(hfsmp->jnl);
-           }
-               hfs_global_shared_lock_release(hfsmp);
+               hfs_volupdate(hfsmp, VOL_UPDATE, 0);
 
-           return (0);
+               hfs_end_transaction(hfsmp);
+
+               return (0);
        }
 
        hfsmp->hfs_privdir_desc.cd_hint = out_desc.cd_hint;
        hfsmp->hfs_privdir_desc.cd_cnid = out_desc.cd_cnid;
        hfsmp->hfs_privdir_attr.ca_fileid = out_desc.cd_cnid;
        hfsmp->hfs_metadata_createdate = vcb->vcbCrDate;
-       
-       if (VFS_ROOT(HFSTOVFS(hfsmp), &dvp) == 0) {
+
+       if (hfs_vget(hfsmp, kRootDirID, &dvp, 0) == 0) {
                dcp = VTOC(dvp);
                dcp->c_childhint = out_desc.cd_hint;
                dcp->c_nlink++;
                dcp->c_entries++;
-               dcp->c_flag |= C_CHANGE | C_UPDATE;
-               tv = time;
-               (void) VOP_UPDATE(dvp, &tv, &tv, 0);
-               vput(dvp);
+               dcp->c_touch_chgtime = TRUE;
+               dcp->c_touch_modtime = TRUE;
+               (void) hfs_update(dvp, 0);
+               hfs_unlock(dcp);
+               vnode_put(dvp);
        }
        hfs_volupdate(hfsmp, VOL_MKDIR, 1);
-       if (hfsmp->jnl) {
-           journal_end_transaction(hfsmp->jnl);
-       } 
-       hfs_global_shared_lock_release(hfsmp);
+       hfs_end_transaction(hfsmp);
 
        cat_releasedesc(&out_desc);
 
@@ -864,7 +1122,7 @@ FindMetaDataDirectory(ExtendedVCB *vcb)
 
 __private_extern__
 u_long
-GetFileInfo(ExtendedVCB *vcb, u_int32_t dirid, char *name,
+GetFileInfo(ExtendedVCB *vcb, u_int32_t dirid, const char *name,
                        struct cat_attr *fattr, struct cat_fork *forkinfo)
 {
        struct hfsmount * hfsmp;
@@ -872,7 +1130,7 @@ GetFileInfo(ExtendedVCB *vcb, u_int32_t dirid, char *name,
        struct cnode * dcp = NULL;
        struct FndrDirInfo * fndrinfo;
        struct cat_desc jdesc;
-       struct timeval tv;
+       int lockflags;
        int error;
        
        if (vcb->vcbSigWord != kHFSPlusSigWord)
@@ -885,25 +1143,22 @@ GetFileInfo(ExtendedVCB *vcb, u_int32_t dirid, char *name,
        jdesc.cd_nameptr = name;
        jdesc.cd_namelen = strlen(name);
 
-       /* Lock catalog b-tree */
-       error = hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_EXCLUSIVE, current_proc());    
-       if (error)
-               return (0);
-
-       error = cat_lookup(hfsmp, &jdesc, 0, NULL, fattr, forkinfo);
-
-       (void) hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_RELEASE, current_proc());
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_SHARED_LOCK);
+       error = cat_lookup(hfsmp, &jdesc, 0, NULL, fattr, forkinfo, NULL);
+       hfs_systemfile_unlock(hfsmp, lockflags);
 
        if (error == 0) {
                return (fattr->ca_fileid);
-       } else if (hfsmp->hfs_fs_ronly) {
+       } else if (hfsmp->hfs_flags & HFS_READ_ONLY) {
                return (0);
        }
+
+       return (0);     /* XXX what callers expect on an error */
 }
 
 
 /*
- * On Journaled HFS, there can be orphaned files.  These
+ * On HFS Plus Volume, there can be orphaned files.  These
  * are files that were unlinked while busy. If the volume
  * was not cleanly unmounted then some of these files may
  * have persisted and need to be removed.
@@ -916,19 +1171,27 @@ hfs_remove_orphans(struct hfsmount * hfsmp)
        struct FSBufferDescriptor btdata;
        struct HFSPlusCatalogFile filerec;
        struct HFSPlusCatalogKey * keyp;
+       struct proc *p = current_proc();
        FCB *fcb;
        ExtendedVCB *vcb;
        char filename[32];
        char tempname[32];
        size_t namelen;
+       cat_cookie_t cookie;
        int catlock = 0;
-       int result, started_tr = 0;
-       
-       if (hfsmp->hfs_orphans_cleaned)
+       int catreserve = 0;
+       int started_tr = 0;
+       int lockflags;
+       int result;
+       int orphanedlinks = 0;
+
+       bzero(&cookie, sizeof(cookie));
+
+       if (hfsmp->hfs_flags & HFS_CLEANED_ORPHANS)
                return;
 
        vcb = HFSTOVCB(hfsmp);
-       fcb = VTOF(vcb->catalogRefNum);
+       fcb = VTOF(hfsmp->hfs_catalog_vp);
 
        btdata.bufferAddress = &filerec;
        btdata.itemSize = sizeof(filerec);
@@ -936,39 +1199,32 @@ hfs_remove_orphans(struct hfsmount * hfsmp)
 
        MALLOC(iterator, struct BTreeIterator *, sizeof(*iterator), M_TEMP, M_WAITOK);
        bzero(iterator, sizeof(*iterator));
+       
+       /* Build a key to "temp" */
        keyp = (HFSPlusCatalogKey*)&iterator->key;
-       keyp->parentID = hfsmp->hfs_private_metadata_dir;
-
-       // XXXdbg
-       hfs_global_shared_lock_acquire(hfsmp);
-       if (hfsmp->jnl) {
-           if (journal_start_transaction(hfsmp->jnl) != 0) {
-                       hfs_global_shared_lock_release(hfsmp);
-                       return;
-           }
-               started_tr = 1;
-       }
-
-       /* Lock catalog b-tree */
-       result = hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_EXCLUSIVE, current_proc());   
-       if (result)
-               goto exit;
-       catlock = 1;
+       keyp->parentID = hfsmp->hfs_privdir_desc.cd_cnid;
+       keyp->nodeName.length = 4;  /* "temp" */
+       keyp->keyLength = kHFSPlusCatalogKeyMinimumLength + keyp->nodeName.length * 2;
+       keyp->nodeName.unicode[0] = 't';
+       keyp->nodeName.unicode[1] = 'e';
+       keyp->nodeName.unicode[2] = 'm';
+       keyp->nodeName.unicode[3] = 'p';
 
        /*
-        * Position the iterator at the folder thread record.
-        * (i.e. one record before first child)
+        * Position the iterator just before the first real temp file.
         */
-       result = BTSearchRecord(fcb, iterator, NULL, NULL, iterator);
-       if (result)
-               goto exit;
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_EXCLUSIVE_LOCK);
+       (void) BTSearchRecord(fcb, iterator, NULL, NULL, iterator);
+       hfs_systemfile_unlock(hfsmp, lockflags);
 
-       /* Visit all the children in the HFS+ private directory. */
+       /* Visit all the temp files in the HFS+ private directory. */
        for (;;) {
+               lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG, HFS_EXCLUSIVE_LOCK);
                result = BTIterateRecord(fcb, kBTreeNextRecord, iterator, &btdata, NULL);
+               hfs_systemfile_unlock(hfsmp, lockflags);
                if (result)
                        break;
-               if (keyp->parentID != hfsmp->hfs_private_metadata_dir)
+               if (keyp->parentID != hfsmp->hfs_privdir_desc.cd_cnid)
                        break;
                if (filerec.recordType != kHFSPlusFileRecord)
                        continue;
@@ -982,46 +1238,89 @@ hfs_remove_orphans(struct hfsmount * hfsmp)
                 * Delete all files named "tempxxx", where
                 * xxx is the file's cnid in decimal.
                 *
-                * Delete all files named "iNodexxx", that
-                * have a link count of zero.
                 */
                if (bcmp(tempname, filename, namelen) == 0) {
-                       struct filefork fork = {0};
-                       struct cnode cnode = {0};
+                       struct filefork dfork;
+                       struct filefork rfork;
+                       struct cnode cnode;
 
-                       // XXXdebug
-                       //printf("hfs_remove_orphans: removing %s\n", filename);
+                       bzero(&dfork, sizeof(dfork));
+                       bzero(&rfork, sizeof(rfork));
+                       bzero(&cnode, sizeof(cnode));
+                       
+                       if (hfs_start_transaction(hfsmp) != 0) {
+                           printf("hfs_remove_orphans: failed to start transaction\n");
+                           goto exit;
+                       }
+                       started_tr = 1;
+               
+                       /*
+                        * Reserve some space in the Catalog file.
+                        */
+                       if (cat_preflight(hfsmp, CAT_DELETE, &cookie, p) != 0) {
+                           printf("hfs_remove_orphans: cat_preflight failed\n");
+                               goto exit;
+                       }
+                       catreserve = 1;
+
+                       lockflags = hfs_systemfile_lock(hfsmp, SFL_CATALOG | SFL_ATTRIBUTE | SFL_EXTENTS | SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
+                       catlock = 1;
 
                        /* Build a fake cnode */
-                       cnode.c_desc.cd_parentcnid = hfsmp->hfs_private_metadata_dir;
+                       cat_convertattr(hfsmp, (CatalogRecord *)&filerec, &cnode.c_attr,
+                                       &dfork.ff_data, &rfork.ff_data);
+                       cnode.c_desc.cd_parentcnid = hfsmp->hfs_privdir_desc.cd_cnid;
                        cnode.c_desc.cd_nameptr = filename;
                        cnode.c_desc.cd_namelen = namelen;
-                       cnode.c_desc.cd_cnid = filerec.fileID;
-                       cnode.c_attr.ca_fileid = filerec.fileID;
-                       cnode.c_blocks = filerec.dataFork.totalBlocks +
-                                        filerec.resourceFork.totalBlocks;
+                       cnode.c_desc.cd_cnid = cnode.c_attr.ca_fileid;
+                       cnode.c_blocks = dfork.ff_blocks + rfork.ff_blocks;
 
                        /* Position iterator at previous entry */
                        if (BTIterateRecord(fcb, kBTreePrevRecord, iterator,
-                           NULL, NULL) != 0)
+                           NULL, NULL) != 0) {
                                break;
-                       
+                       }
+
                        /* Truncate the file to zero (both forks) */
-                       if (filerec.dataFork.totalBlocks > 0) {
-                               fork.ff_cp = &cnode;
-                               cnode.c_datafork = &fork;
-                               bcopy(&filerec.dataFork, &fork.ff_data, sizeof(struct cat_fork));
-                               if (TruncateFileC(vcb, (FCB*)&fork, 0, false) != 0) {
-                                       printf("error truncting data fork!\n");
-                                       break;
+                       if (dfork.ff_blocks > 0) {
+                               u_int64_t fsize;
+                               
+                               dfork.ff_cp = &cnode;
+                               cnode.c_datafork = &dfork;
+                               cnode.c_rsrcfork = NULL;
+                               fsize = (u_int64_t)dfork.ff_blocks * (u_int64_t)HFSTOVCB(hfsmp)->blockSize;
+                               while (fsize > 0) {
+                                       if (fsize > HFS_BIGFILE_SIZE) {
+                                               fsize -= HFS_BIGFILE_SIZE;
+                                       } else {
+                                               fsize = 0;
+                                       }
+
+                                       if (TruncateFileC(vcb, (FCB*)&dfork, fsize, false) != 0) {
+                                               printf("error truncting data fork!\n");
+                                               break;
+                                       }
+
+                                       //
+                                       // if we're iteratively truncating this file down,
+                                       // then end the transaction and start a new one so
+                                       // that no one transaction gets too big.
+                                       //
+                                       if (fsize > 0 && started_tr) {
+                                               hfs_end_transaction(hfsmp);
+                                               if (hfs_start_transaction(hfsmp) != 0) {
+                                                       started_tr = 0;
+                                                       break;
+                                               }
+                                       }
                                }
                        }
-                       if (filerec.resourceFork.totalBlocks > 0) {
-                               fork.ff_cp = &cnode;
+
+                       if (rfork.ff_blocks > 0) {
+                               rfork.ff_cp = &cnode;
                                cnode.c_datafork = NULL;
-                               cnode.c_rsrcfork = &fork;
-                               bcopy(&filerec.resourceFork, &fork.ff_data, sizeof(struct cat_fork));
-                               if (TruncateFileC(vcb, (FCB*)&fork, 0, false) != 0) {
+                               cnode.c_rsrcfork = &rfork;
+                               if (TruncateFileC(vcb, (FCB*)&rfork, 0, false) != 0) {
                                        printf("error truncting rsrc fork!\n");
                                        break;
                                }
@@ -1029,30 +1328,47 @@ hfs_remove_orphans(struct hfsmount * hfsmp)
 
                        /* Remove the file record from the Catalog */   
                        if (cat_delete(hfsmp, &cnode.c_desc, &cnode.c_attr) != 0) {
-                               printf("error deleting cat rec!\n");
+                               printf("hfs_remove_oprhans: error deleting cat rec for id %d!\n", cnode.c_desc.cd_cnid);
+                               hfs_volupdate(hfsmp, VOL_UPDATE, 0);
                                break;
                        }
+                       ++orphanedlinks;
+
+                       /* Delete any attributes, ignore errors */
+                       (void) hfs_removeallattr(hfsmp, cnode.c_fileid);
                        
                        /* Update parent and volume counts */   
                        hfsmp->hfs_privdir_attr.ca_entries--;
                        (void)cat_update(hfsmp, &hfsmp->hfs_privdir_desc,
                                         &hfsmp->hfs_privdir_attr, NULL, NULL);
                        hfs_volupdate(hfsmp, VOL_RMFILE, 0);
-               }
-       }
-       
-exit:
-       /* Unlock catalog b-tree */
-       if (catlock)
-               (void) hfs_metafilelocking(hfsmp, kHFSCatalogFileID, LK_RELEASE, current_proc());
 
+                       /* Drop locks and end the transaction */
+                       hfs_systemfile_unlock(hfsmp, lockflags);
+                       cat_postflight(hfsmp, &cookie, p);
+                       catlock = catreserve = 0;
+                       if (started_tr) {
+                               hfs_end_transaction(hfsmp);
+                               started_tr = 0;
+                       }
+
+               } /* end if */
+       } /* end for */
+       if (orphanedlinks > 0)
+               printf("HFS: Removed %d orphaned unlinked files\n", orphanedlinks);
+exit:
+       if (catlock) {
+               hfs_systemfile_unlock(hfsmp, lockflags);
+       }
+       if (catreserve) {
+               cat_postflight(hfsmp, &cookie, p);
+       }
        if (started_tr) {
-               journal_end_transaction(hfsmp->jnl);
+               hfs_end_transaction(hfsmp);
        }
-       hfs_global_shared_lock_release(hfsmp);
 
        FREE(iterator, M_TEMP);
-       hfsmp->hfs_orphans_cleaned = 1;
+       hfsmp->hfs_flags |= HFS_CLEANED_ORPHANS;
 }
 
 
@@ -1071,7 +1387,7 @@ u_int32_t logBlockSize;
        /* start with default */
        logBlockSize = VTOHFS(vp)->hfs_logBlockSize;
 
-       if (vp->v_flag & VSYSTEM) {
+       if (vnode_issystem(vp)) {
                if (VTOF(vp)->fcbBTCBPtr != NULL) {
                        BTreeInfoRec                    bTreeInfo;
        
@@ -1101,9 +1417,10 @@ __private_extern__
 u_int32_t
 hfs_freeblks(struct hfsmount * hfsmp, int wantreserve)
 {
-       struct vcb_t *vcb = HFSTOVCB(hfsmp);
+       ExtendedVCB *vcb = HFSTOVCB(hfsmp);
        u_int32_t freeblks;
 
+       HFS_MOUNT_LOCK(hfsmp, TRUE);
        freeblks = vcb->freeBlocks;
        if (wantreserve) {
                if (freeblks > vcb->reserveBlocks)
@@ -1111,8 +1428,50 @@ hfs_freeblks(struct hfsmount * hfsmp, int wantreserve)
                else
                        freeblks = 0;
        }
+       if (freeblks > vcb->loanedBlocks)
+               freeblks -= vcb->loanedBlocks;
+       else
+               freeblks = 0;
+       HFS_MOUNT_UNLOCK(hfsmp, TRUE);
+
+#ifdef HFS_SPARSE_DEV
+       /* 
+        * When the underlying device is sparse, check the
+        * available space on the backing store volume.
+        */
+       if ((hfsmp->hfs_flags & HFS_HAS_SPARSE_DEVICE) && hfsmp->hfs_backingfs_rootvp) {
+               struct vfsstatfs *vfsp;  /* 272 bytes */
+               u_int32_t vfreeblks;
+               u_int32_t loanedblks;
+               struct mount * backingfs_mp;
+
+               backingfs_mp = vnode_mount(hfsmp->hfs_backingfs_rootvp);
+
+               if (vfsp = vfs_statfs(backingfs_mp)) {
+                       HFS_MOUNT_LOCK(hfsmp, TRUE);
+                       vfreeblks = (u_int32_t)vfsp->f_bavail;
+                       /* Normalize block count if needed. */
+                       if (vfsp->f_bsize != vcb->blockSize) {
+                               vfreeblks = ((u_int64_t)vfreeblks * (u_int64_t)(vfsp->f_bsize)) / vcb->blockSize;
+                       }
+                       if (vfreeblks > hfsmp->hfs_sparsebandblks)
+                               vfreeblks -= hfsmp->hfs_sparsebandblks;
+                       else
+                               vfreeblks = 0;
+                       
+                       /* Take into account any delayed allocations. */
+                       loanedblks = 2 * vcb->loanedBlocks;
+                       if (vfreeblks > loanedblks)
+                               vfreeblks -= loanedblks;
+                       else
+                               vfreeblks = 0;
+
+                       freeblks = MIN(vfreeblks, freeblks);
+                       HFS_MOUNT_UNLOCK(hfsmp, TRUE);
+               }
+       }
+#endif /* HFS_SPARSE_DEV */
 
-       freeblks -= vcb->loanedBlocks;
        return (freeblks);
 }
 
@@ -1127,14 +1486,13 @@ short MacToVFSError(OSErr err)
 
        switch (err) {
        case dskFulErr:                 /*    -34 */
+       case btNoSpaceAvail:            /* -32733 */
                return ENOSPC;
-       case btNoSpaceAvail:    /* -32733 */
-               return EFBIG;
        case fxOvFlErr:                 /* -32750 */
                return EOVERFLOW;
        
        case btBadNode:                 /* -32731 */
-               return EBADF;
+               return EIO;
        
        case memFullErr:                /*  -108 */
                return ENOMEM;          /*   +12 */
@@ -1173,93 +1531,120 @@ short MacToVFSError(OSErr err)
 
 
 /*
- * Get the directory entry name hint for a given index.
- * The directory cnode (dcp) must be locked.
+ * Find the current thread's directory hint for a given index.
+ *
+ * Requires an exclusive lock on directory cnode.
  */
 __private_extern__
-char *
-hfs_getnamehint(struct cnode *dcp, int index)
+directoryhint_t *
+hfs_getdirhint(struct cnode *dcp, int index)
 {
-       struct hfs_index *entry;
-       void *self;
-
-       if (index > 0) {
-               self = current_thread();
-               SLIST_FOREACH(entry, &dcp->c_indexlist, hi_link) {
-                       if ((entry->hi_index == index)
-                       &&  (entry->hi_thread == self))
-                               return (entry->hi_name);
+       struct timeval tv;
+       directoryhint_t *hint;
+       boolean_t need_remove, need_init;
+       char * name;
+
+       microuptime(&tv);
+
+       /*
+        *  Look for an existing hint first.  If not found, create a new one (when
+        *  the list is not full) or recycle the oldest hint.  Since new hints are
+        *  always added to the head of the list, the last hint is always the
+        *  oldest.
+        */
+       TAILQ_FOREACH(hint, &dcp->c_hintlist, dh_link) {
+               if (hint->dh_index == index)
+                       break;
+       }
+       if (hint != NULL) { /* found an existing hint */
+               need_init = false;
+               need_remove = true;
+       } else { /* cannot find an existing hint */
+               need_init = true;
+               if (dcp->c_dirhintcnt < HFS_MAXDIRHINTS) { /* we don't need recycling */
+                       /* Create a default directory hint */
+                       MALLOC_ZONE(hint, directoryhint_t *, sizeof(directoryhint_t), M_HFSDIRHINT, M_WAITOK);
+                       ++dcp->c_dirhintcnt;
+                       need_remove = false;
+               } else {                                /* recycle the last (i.e., the oldest) hint */
+                       hint = TAILQ_LAST(&dcp->c_hintlist, hfs_hinthead);
+                       if ((name = hint->dh_desc.cd_nameptr)) {
+                               hint->dh_desc.cd_nameptr = NULL;
+                               vfs_removename(name);
+                       }
+                       need_remove = true;
                }
        }
 
-       return (NULL);
-}
+       if (need_remove)
+               TAILQ_REMOVE(&dcp->c_hintlist, hint, dh_link);
 
-/*
- * Save a directory entry name hint for a given index.
- * The directory cnode (dcp) must be locked.
- */
-__private_extern__
-void
-hfs_savenamehint(struct cnode *dcp, int index, const char * namehint)
-{
-       struct hfs_index *entry;
-       int len;
-
-       if (index > 0) {
-               len = strlen(namehint);
-               MALLOC(entry, struct hfs_index *, len + sizeof(struct hfs_index),
-                       M_TEMP, M_WAITOK);
-               entry->hi_index = index;
-               entry->hi_thread = current_thread();
-               bcopy(namehint, entry->hi_name, len + 1);
-               SLIST_INSERT_HEAD(&dcp->c_indexlist, entry, hi_link);
+       TAILQ_INSERT_HEAD(&dcp->c_hintlist, hint, dh_link);
+
+       if (need_init) {
+               hint->dh_index = index;
+               hint->dh_desc.cd_flags = 0;
+               hint->dh_desc.cd_encoding = 0;
+               hint->dh_desc.cd_namelen = 0;
+               hint->dh_desc.cd_nameptr = NULL;
+               hint->dh_desc.cd_parentcnid = dcp->c_cnid;
+               hint->dh_desc.cd_hint = dcp->c_childhint;
+               hint->dh_desc.cd_cnid = 0;
        }
+       hint->dh_time = tv.tv_sec;
+       return (hint);
 }
 
 /*
- * Release the directory entry name hint for a given index.
- * The directory cnode (dcp) must be locked.
+ * Release a single directory hint.
+ *
+ * Requires an exclusive lock on directory cnode.
  */
 __private_extern__
 void
-hfs_relnamehint(struct cnode *dcp, int index)
+hfs_reldirhint(struct cnode *dcp, directoryhint_t * relhint)
 {
-       struct hfs_index *entry;
-       void *self;
-
-       if (index > 0) {
-               self = current_thread();
-               SLIST_FOREACH(entry, &dcp->c_indexlist, hi_link) {
-                       if ((entry->hi_index == index)
-                       &&  (entry->hi_thread == self)) {
-                               SLIST_REMOVE(&dcp->c_indexlist, entry, hfs_index,
-                                       hi_link);
-                               FREE(entry, M_TEMP);
-                               break;
-                       }
-               }
+       char * name;
+
+       TAILQ_REMOVE(&dcp->c_hintlist, relhint, dh_link);
+       name = relhint->dh_desc.cd_nameptr;
+       if (name != NULL) {
+               relhint->dh_desc.cd_nameptr = NULL;
+               vfs_removename(name);
        }
+       FREE_ZONE(relhint, sizeof(directoryhint_t), M_HFSDIRHINT);
+       --dcp->c_dirhintcnt;
 }
 
 /*
- * Release all directory entry name hints.
+ * Release directory hints for given directory
+ *
+ * Requires an exclusive lock on directory cnode.
  */
 __private_extern__
 void
-hfs_relnamehints(struct cnode *dcp)
+hfs_reldirhints(struct cnode *dcp, int stale_hints_only)
 {
-       struct hfs_index *entry;
-       struct hfs_index *next;
-
-       if (!SLIST_EMPTY(&dcp->c_indexlist)) {
-               for(entry = SLIST_FIRST(&dcp->c_indexlist);
-                   entry != NULL;
-                   entry = next) {
-                       next = SLIST_NEXT(entry, hi_link);
-                       SLIST_REMOVE(&dcp->c_indexlist, entry, hfs_index, hi_link);
-                       FREE(entry, M_TEMP);
+       struct timeval tv;
+       directoryhint_t *hint, *prev;
+       char * name;
+
+       if (stale_hints_only)
+               microuptime(&tv);
+
+       /* searching from the oldest to the newest, so we can stop early when releasing stale hints only */
+       for (hint = TAILQ_LAST(&dcp->c_hintlist, hfs_hinthead); hint != NULL; hint = prev) {
+               if (stale_hints_only && (tv.tv_sec - hint->dh_time) < HFS_DIRHINT_TTL)
+                       break;  /* stop here if this entry is too new */
+               name = hint->dh_desc.cd_nameptr;
+               if (name != NULL) {
+                       hint->dh_desc.cd_nameptr = NULL;
+                       vfs_removename(name);
                }
+               prev = TAILQ_PREV(hint, hfs_hinthead, dh_link); /* must save this pointer before calling FREE_ZONE on this node */
+               TAILQ_REMOVE(&dcp->c_hintlist, hint, dh_link);
+               FREE_ZONE(hint, sizeof(directoryhint_t), M_HFSDIRHINT);
+               --dcp->c_dirhintcnt;
        }
 }
 
@@ -1300,8 +1685,8 @@ out:
 __private_extern__
 int
 hfs_early_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
-                                          void *_args, int embeddedOffset, int mdb_offset,
-                                          HFSMasterDirectoryBlock *mdbp, struct ucred *cred)
+                                          void *_args, off_t embeddedOffset, daddr64_t mdb_offset,
+                                          HFSMasterDirectoryBlock *mdbp, kauth_cred_t cred)
 {
        JournalInfoBlock *jibp;
        struct buf       *jinfo_bp, *bp;
@@ -1319,14 +1704,14 @@ hfs_early_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
 
        sectors_per_fsblock = SWAP_BE32(vhp->blockSize) / blksize;
                                
-       retval = meta_bread(devvp,
-                                               embeddedOffset/blksize + 
-                                               (SWAP_BE32(vhp->journalInfoBlock)*sectors_per_fsblock),
+       retval = (int)buf_meta_bread(devvp,
+                                               (daddr64_t)((embeddedOffset/blksize) + 
+                                               (SWAP_BE32(vhp->journalInfoBlock)*sectors_per_fsblock)),
                                                SWAP_BE32(vhp->blockSize), cred, &jinfo_bp);
        if (retval)
                return retval;
 
-       jibp = (JournalInfoBlock *)jinfo_bp->b_data;
+       jibp = (JournalInfoBlock *)buf_dataptr(jinfo_bp);
        jibp->flags  = SWAP_BE32(jibp->flags);
        jibp->offset = SWAP_BE64(jibp->offset);
        jibp->size   = SWAP_BE64(jibp->size);
@@ -1335,18 +1720,19 @@ hfs_early_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
                hfsmp->jvp = hfsmp->hfs_devvp;
        } else {
                printf("hfs: journal not stored in fs! don't know what to do.\n");
-               brelse(jinfo_bp);
+               buf_brelse(jinfo_bp);
                return EINVAL;
        }
 
        // save this off for the hack-y check in hfs_remove()
        hfsmp->jnl_start = jibp->offset / SWAP_BE32(vhp->blockSize);
+       hfsmp->jnl_size  = jibp->size;
 
        if (jibp->flags & kJIJournalNeedInitMask) {
                printf("hfs: Initializing the journal (joffset 0x%llx sz 0x%llx)...\n",
-                          jibp->offset + (off_t)embeddedOffset, jibp->size);
+                          jibp->offset + embeddedOffset, jibp->size);
                hfsmp->jnl = journal_create(hfsmp->jvp,
-                                                                       jibp->offset + (off_t)embeddedOffset,
+                                                                       jibp->offset + embeddedOffset,
                                                                        jibp->size,
                                                                        devvp,
                                                                        blksize,
@@ -1358,16 +1744,18 @@ hfs_early_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
                // we'd just re-init it on the next mount.
                jibp->flags &= ~kJIJournalNeedInitMask;
                jibp->flags  = SWAP_BE32(jibp->flags);
-               bwrite(jinfo_bp);
+               jibp->offset = SWAP_BE64(jibp->offset);
+               jibp->size   = SWAP_BE64(jibp->size);
+               buf_bwrite(jinfo_bp);
                jinfo_bp = NULL;
                jibp     = NULL;
        } else { 
                //printf("hfs: Opening the journal (joffset 0x%llx sz 0x%llx vhp_blksize %d)...\n",
-               //         jibp->offset + (off_t)embeddedOffset,
+               //         jibp->offset + embeddedOffset,
                //         jibp->size, SWAP_BE32(vhp->blockSize));
                                
                hfsmp->jnl = journal_open(hfsmp->jvp,
-                                                                 jibp->offset + (off_t)embeddedOffset,
+                                                                 jibp->offset + embeddedOffset,
                                                                  jibp->size,
                                                                  devvp,
                                                                  blksize,
@@ -1375,22 +1763,25 @@ hfs_early_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
                                                                  arg_tbufsz,
                                                                  hfs_sync_metadata, hfsmp->hfs_mp);
 
-               brelse(jinfo_bp);
+               buf_brelse(jinfo_bp);
                jinfo_bp = NULL;
                jibp     = NULL;
 
                if (hfsmp->jnl && mdbp) {
                        // reload the mdb because it could have changed
                        // if the journal had to be replayed.
-                       retval = meta_bread(devvp, mdb_offset, blksize, cred, &bp);
+                       if (mdb_offset == 0) {
+                               mdb_offset = (daddr64_t)((embeddedOffset / blksize) + HFS_PRI_SECTOR(blksize));
+                       }
+                       retval = (int)buf_meta_bread(devvp, mdb_offset, blksize, cred, &bp);
                        if (retval) {
-                               brelse(bp);
+                               buf_brelse(bp);
                                printf("hfs: failed to reload the mdb after opening the journal (retval %d)!\n",
                                           retval);
                                return retval;
                        }
-                       bcopy(bp->b_data + HFS_PRI_OFFSET(blksize), mdbp, 512);
-                       brelse(bp);
+                       bcopy((char *)buf_dataptr(bp) + HFS_PRI_OFFSET(blksize), mdbp, 512);
+                       buf_brelse(bp);
                        bp = NULL;
                }
        }
@@ -1401,9 +1792,7 @@ hfs_early_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp,
        // if we expected the journal to be there and we couldn't
        // create it or open it then we have to bail out.
        if (hfsmp->jnl == NULL) {
-               hfsmp->jnl_start = 0;
-               
-               printf("hfs: failed to open/create the journal (retval %d).\n", retval);
+               printf("hfs: early jnl init: failed to open/create the journal (retval %d).\n", retval);
                return EINVAL;
        }
 
@@ -1471,9 +1860,9 @@ hfs_late_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp, void *_a
 
 
        sectors_per_fsblock = SWAP_BE32(vhp->blockSize) / hfsmp->hfs_phys_block_size;
-       retval = meta_bread(devvp,
-                                               vcb->hfsPlusIOPosOffset / hfsmp->hfs_phys_block_size + 
-                                               (SWAP_BE32(vhp->journalInfoBlock)*sectors_per_fsblock),
+       retval = (int)buf_meta_bread(devvp,
+                                               (daddr64_t)(vcb->hfsPlusIOPosOffset / hfsmp->hfs_phys_block_size + 
+                                               (SWAP_BE32(vhp->journalInfoBlock)*sectors_per_fsblock)),
                                                SWAP_BE32(vhp->blockSize), NOCRED, &jinfo_bp);
        if (retval) {
                printf("hfs: can't read journal info block. disabling journaling.\n");
@@ -1481,7 +1870,7 @@ hfs_late_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp, void *_a
                return 0;
        }
 
-       jibp = (JournalInfoBlock *)jinfo_bp->b_data;
+       jibp = (JournalInfoBlock *)buf_dataptr(jinfo_bp);
        jibp->flags  = SWAP_BE32(jibp->flags);
        jibp->offset = SWAP_BE64(jibp->offset);
        jibp->size   = SWAP_BE64(jibp->size);
@@ -1490,7 +1879,7 @@ hfs_late_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp, void *_a
        if (fid == 0 || jfork.cf_extents[0].startBlock == 0 || jfork.cf_size == 0) {
                printf("hfs: can't find the journal file! disabling journaling (start: %d)\n",
                           jfork.cf_extents[0].startBlock);
-               brelse(jinfo_bp);
+               buf_brelse(jinfo_bp);
                vcb->vcbAtrb &= ~kHFSVolumeJournaledMask;
                return 0;
        }
@@ -1518,12 +1907,13 @@ hfs_late_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp, void *_a
                hfsmp->jvp = hfsmp->hfs_devvp;
        } else {
                printf("hfs: journal not stored in fs! don't know what to do.\n");
-               brelse(jinfo_bp);
+               buf_brelse(jinfo_bp);
                return EINVAL;
        }
 
        // save this off for the hack-y check in hfs_remove()
        hfsmp->jnl_start = jibp->offset / SWAP_BE32(vhp->blockSize);
+       hfsmp->jnl_size  = jibp->size;
 
        if (jibp->flags & kJIJournalNeedInitMask) {
                printf("hfs: Initializing the journal (joffset 0x%llx sz 0x%llx)...\n",
@@ -1573,9 +1963,9 @@ hfs_late_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp, void *_a
                jibp->offset = SWAP_BE64(jibp->offset);
                jibp->size   = SWAP_BE64(jibp->size);
 
-               bwrite(jinfo_bp);
+               buf_bwrite(jinfo_bp);
        } else {
-               brelse(jinfo_bp);
+               buf_brelse(jinfo_bp);
        } 
        jinfo_bp = NULL;
        jibp     = NULL;
@@ -1585,11 +1975,282 @@ hfs_late_journal_init(struct hfsmount *hfsmp, HFSPlusVolumeHeader *vhp, void *_a
        // if we expected the journal to be there and we couldn't
        // create it or open it then we have to bail out.
        if (hfsmp->jnl == NULL) {
-               hfsmp->jnl_start = 0;
-               
-               printf("hfs: failed to open/create the journal (retval %d).\n", retval);
+               printf("hfs: late jnl init: failed to open/create the journal (retval %d).\n", retval);
                return EINVAL;
        }
 
        return 0;
 }
+
+/*
+ * Calculate the allocation zone for metadata.
+ *
+ * This zone includes the following:
+ *     Allocation Bitmap file
+ *     Overflow Extents file
+ *     Journal file
+ *     Quota files
+ *     Clustered Hot files
+ *     Catalog file
+ *
+ *                          METADATA ALLOCATION ZONE
+ * ____________________________________________________________________________
+ * |    |    |     |               |                              |           |
+ * | BM | JF | OEF |    CATALOG    |--->                          | HOT FILES |
+ * |____|____|_____|_______________|______________________________|___________|
+ *
+ * <------------------------------- N * 128 MB ------------------------------->
+ *
+ */
+#define GIGABYTE  (u_int64_t)(1024*1024*1024)
+
+#define OVERFLOW_DEFAULT_SIZE (4*1024*1024)
+#define OVERFLOW_MAXIMUM_SIZE (128*1024*1024)
+#define JOURNAL_DEFAULT_SIZE  (8*1024*1024)
+#define JOURNAL_MAXIMUM_SIZE  (512*1024*1024)
+#define HOTBAND_MINIMUM_SIZE  (10*1024*1024)
+#define HOTBAND_MAXIMUM_SIZE  (512*1024*1024)
+
+static void
+hfs_metadatazone_init(struct hfsmount *hfsmp)
+{
+       ExtendedVCB  *vcb;
+       u_int64_t  fs_size;
+       u_int64_t  zonesize;
+       u_int64_t  temp;
+       u_int64_t  filesize;
+       u_int32_t  blk;
+       int  items;
+
+       vcb = HFSTOVCB(hfsmp);
+       fs_size = (u_int64_t)vcb->blockSize * (u_int64_t)vcb->totalBlocks;
+
+       /*
+        * For volumes less than 10 GB, don't bother.
+        */
+       if (fs_size < ((u_int64_t)10 * GIGABYTE))
+               return;
+       /*
+        * Skip non-journaled volumes as well.
+        */
+       if (hfsmp->jnl == NULL)
+               return;
+
+       /*
+        * Start with allocation bitmap (a fixed size).
+        */
+       zonesize = roundup(vcb->totalBlocks / 8, vcb->vcbVBMIOSize);
+
+       /*
+        * Overflow Extents file gets 4 MB per 100 GB.
+        */
+       items = fs_size / ((u_int64_t)100 * GIGABYTE);
+       filesize = (u_int64_t)(items + 1) * OVERFLOW_DEFAULT_SIZE;
+       if (filesize > OVERFLOW_MAXIMUM_SIZE)
+               filesize = OVERFLOW_MAXIMUM_SIZE;
+       zonesize += filesize;
+       hfsmp->hfs_overflow_maxblks = filesize / vcb->blockSize;
+
+       /*
+        * Plan for at least 8 MB of journal for each
+        * 100 GB of disk space (up to a 512 MB).
+        */
+       items = fs_size / ((u_int64_t)100 * GIGABYTE);
+       filesize = (u_int64_t)(items + 1) * JOURNAL_DEFAULT_SIZE;
+       if (filesize > JOURNAL_MAXIMUM_SIZE)
+               filesize = JOURNAL_MAXIMUM_SIZE;
+       zonesize += filesize;
+
+       /*
+        * Catalog file gets 10 MB per 1 GB.
+        *
+        * How about considering the current catalog size (used nodes * node size)
+        * and the current file data size to help estimate the required
+        * catalog size.
+        */
+       filesize = MIN((fs_size / 1024) * 10, GIGABYTE);
+       hfsmp->hfs_catalog_maxblks = filesize / vcb->blockSize;
+       zonesize += filesize;
+
+       /*
+        * Add space for hot file region.
+        *
+        * ...for now, use 5 MB per 1 GB (0.5 %)
+        */
+       filesize = (fs_size / 1024) * 5;
+       if (filesize > HOTBAND_MAXIMUM_SIZE)
+               filesize = HOTBAND_MAXIMUM_SIZE;
+       else if (filesize < HOTBAND_MINIMUM_SIZE)
+               filesize = HOTBAND_MINIMUM_SIZE;
+       /*
+        * Calculate user quota file requirements.
+        */
+       items = QF_USERS_PER_GB * (fs_size / GIGABYTE);
+       if (items < QF_MIN_USERS)
+               items = QF_MIN_USERS;
+       else if (items > QF_MAX_USERS)
+               items = QF_MAX_USERS;
+       if (!powerof2(items)) {
+               int x = items;
+               items = 4;
+               while (x>>1 != 1) {
+                       x = x >> 1;
+                       items = items << 1;
+               }
+       }
+       filesize += (items + 1) * sizeof(struct dqblk);
+       /*
+        * Calculate group quota file requirements.
+        *
+        */
+       items = QF_GROUPS_PER_GB * (fs_size / GIGABYTE);
+       if (items < QF_MIN_GROUPS)
+               items = QF_MIN_GROUPS;
+       else if (items > QF_MAX_GROUPS)
+               items = QF_MAX_GROUPS;
+       if (!powerof2(items)) {
+               int x = items;
+               items = 4;
+               while (x>>1 != 1) {
+                       x = x >> 1;
+                       items = items << 1;
+               }
+       }
+       filesize += (items + 1) * sizeof(struct dqblk);
+       zonesize += filesize;
+
+       /*
+        * Round up entire zone to a bitmap block's worth.
+        * The extra space goes to the catalog file and hot file area.
+        */
+       temp = zonesize;
+       zonesize = roundup(zonesize, vcb->vcbVBMIOSize * 8 * vcb->blockSize);
+       temp = zonesize - temp;  /* temp has extra space */
+       filesize += temp / 3;
+       hfsmp->hfs_catalog_maxblks += (temp - (temp / 3)) / vcb->blockSize;
+
+       hfsmp->hfs_hotfile_maxblks = filesize / vcb->blockSize;
+
+       /* Convert to allocation blocks. */
+       blk = zonesize / vcb->blockSize;
+
+       /* The default metadata zone location is at the start of volume. */
+       hfsmp->hfs_metazone_start = 1;
+       hfsmp->hfs_metazone_end = blk - 1;
+       
+       /* The default hotfile area is at the end of the zone. */
+       hfsmp->hfs_hotfile_start = blk - (filesize / vcb->blockSize);
+       hfsmp->hfs_hotfile_end = hfsmp->hfs_metazone_end;
+       hfsmp->hfs_hotfile_freeblks = hfs_hotfile_freeblocks(hfsmp);
+#if 0
+       printf("HFS: metadata zone is %d to %d\n", hfsmp->hfs_metazone_start, hfsmp->hfs_metazone_end);
+       printf("HFS: hot file band is %d to %d\n", hfsmp->hfs_hotfile_start, hfsmp->hfs_hotfile_end);
+       printf("HFS: hot file band free blocks = %d\n", hfsmp->hfs_hotfile_freeblks);
+#endif
+       hfsmp->hfs_flags |= HFS_METADATA_ZONE;
+}
+
+
+static u_int32_t
+hfs_hotfile_freeblocks(struct hfsmount *hfsmp)
+{
+       ExtendedVCB  *vcb = HFSTOVCB(hfsmp);
+       int  lockflags;
+       int  freeblocks;
+
+       lockflags = hfs_systemfile_lock(hfsmp, SFL_BITMAP, HFS_EXCLUSIVE_LOCK);
+       freeblocks = MetaZoneFreeBlocks(vcb);
+       hfs_systemfile_unlock(hfsmp, lockflags);
+
+       /* Minus Extents overflow file reserve. */
+       freeblocks -=
+               hfsmp->hfs_overflow_maxblks - VTOF(hfsmp->hfs_extents_vp)->ff_blocks;
+       /* Minus catalog file reserve. */
+       freeblocks -=
+               hfsmp->hfs_catalog_maxblks - VTOF(hfsmp->hfs_catalog_vp)->ff_blocks;
+       if (freeblocks < 0)
+               freeblocks = 0;
+
+       return MIN(freeblocks, hfsmp->hfs_hotfile_maxblks);
+}
+
+/*
+ * Determine if a file is a "virtual" metadata file.
+ * This includes journal and quota files.
+ */
+__private_extern__
+int
+hfs_virtualmetafile(struct cnode *cp)
+{
+       char * filename;
+
+
+       if (cp->c_parentcnid != kHFSRootFolderID)
+               return (0);
+
+       filename = cp->c_desc.cd_nameptr;
+       if (filename == NULL)
+               return (0);
+
+       if ((strcmp(filename, ".journal") == 0) ||
+           (strcmp(filename, ".journal_info_block") == 0) ||
+           (strcmp(filename, ".quota.user") == 0) ||
+           (strcmp(filename, ".quota.group") == 0) ||
+           (strcmp(filename, ".hotfiles.btree") == 0))
+               return (1);
+
+       return (0);
+}
+
+
+__private_extern__
+int
+hfs_start_transaction(struct hfsmount *hfsmp)
+{
+    int ret;
+
+    if (hfsmp->jnl == NULL || journal_owner(hfsmp->jnl) != current_thread()) {
+       lck_rw_lock_shared(&hfsmp->hfs_global_lock);
+    }
+
+    if (hfsmp->jnl) {
+       ret = journal_start_transaction(hfsmp->jnl);
+       if (ret == 0) {
+           OSAddAtomic(1, &hfsmp->hfs_global_lock_nesting);
+       }
+    } else {
+       ret = 0;
+    }
+
+    if (ret != 0) {
+       lck_rw_done(&hfsmp->hfs_global_lock);
+    }
+
+    return ret;
+}
+
+__private_extern__
+int
+hfs_end_transaction(struct hfsmount *hfsmp)
+{
+    int need_unlock=0, ret;
+
+    if (    hfsmp->jnl == NULL
+       || (   journal_owner(hfsmp->jnl) == current_thread()
+           && (OSAddAtomic(-1, &hfsmp->hfs_global_lock_nesting) == 1)) ) {
+
+           need_unlock = 1;
+    } 
+
+    if (hfsmp->jnl) {
+       ret = journal_end_transaction(hfsmp->jnl);
+    } else {
+       ret = 0;
+    }
+
+    if (need_unlock) {
+       lck_rw_done(&hfsmp->hfs_global_lock);
+    }
+
+    return ret;
+}