]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/vfs/vfs_lookup.c
xnu-1699.22.73.tar.gz
[apple/xnu.git] / bsd / vfs / vfs_lookup.c
index 553bb41f9805fb71214ecfa8d26111d0d5fa1009..10b885d51829894bb5e63e3a2ff6f6713a29a76c 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
@@ -111,6 +111,17 @@ static     void kdebug_lookup(struct vnode *dp, struct componentname *cnp);
 static int vfs_getrealpath(const char * path, char * realpath, size_t bufsize, vfs_context_t ctx);
 #endif
 
+boolean_t      lookup_continue_ok(struct nameidata *ndp);
+int            lookup_traverse_mountpoints(struct nameidata *ndp, struct componentname *cnp, vnode_t dp, int vbusyflags, vfs_context_t ctx);
+int            lookup_handle_symlink(struct nameidata *ndp, vnode_t *new_dp, vfs_context_t ctx);
+int            lookup_authorize_search(vnode_t dp, struct componentname *cnp, int dp_authorized_in_cache, vfs_context_t ctx);
+void           lookup_consider_update_cache(vnode_t dvp, vnode_t vp, struct componentname *cnp, int nc_generation);
+int            lookup_handle_rsrc_fork(vnode_t dp, struct nameidata *ndp, struct componentname *cnp, int wantparent, vfs_context_t ctx);
+int            lookup_handle_found_vnode(struct nameidata *ndp, struct componentname *cnp, int rdonly, 
+                       int vbusyflags, int *keep_going, int nc_generation,
+                       int wantparent, int atroot, vfs_context_t ctx);
+int            lookup_handle_emptyname(struct nameidata *ndp, struct componentname *cnp, int wantparent);
+
 /*
  * Convert a pathname into a pointer to a locked inode.
  *
@@ -150,12 +161,10 @@ int
 namei(struct nameidata *ndp)
 {
        struct filedesc *fdp;   /* pointer to file descriptor state */
-       char *cp;               /* pointer into pathname argument */
        struct vnode *dp;       /* the directory we are searching */
        struct vnode *usedvp = ndp->ni_dvp;  /* store pointer to vp in case we must loop due to
                                                                                        heavy vnode pressure */
        u_long cnpflags = ndp->ni_cnd.cn_flags; /* store in case we have to restore after loop */
-       uio_t auio;
        int error;
        struct componentname *cnp = &ndp->ni_cnd;
        vfs_context_t ctx = cnp->cn_context;
@@ -164,8 +173,8 @@ namei(struct nameidata *ndp)
 /* XXX ut should be from context */
        uthread_t ut = (struct uthread *)get_bsdthread_info(current_thread());
 #endif
-       char *tmppn;
-       char uio_buf[ UIO_SIZEOF(1) ];
+
+       fdp = p->p_fd;
 
 #if DIAGNOSTIC
        if (!vfs_context_ucred(ctx) || !p)
@@ -175,7 +184,35 @@ namei(struct nameidata *ndp)
        if (cnp->cn_flags & OPMASK)
                panic ("namei: flags contaminated with nameiops");
 #endif
-       fdp = p->p_fd;
+
+       /*
+        * A compound VNOP found something that needs further processing:
+        * either a trigger vnode, a covered directory, or a symlink.
+        */
+       if (ndp->ni_flag & NAMEI_CONTLOOKUP) {
+               int rdonly, vbusyflags, keep_going, wantparent;
+
+               rdonly = cnp->cn_flags & RDONLY;
+               vbusyflags = ((cnp->cn_flags & CN_NBMOUNTLOOK) != 0) ? LK_NOWAIT : 0;
+               keep_going = 0;
+               wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
+
+               ndp->ni_flag &= ~(NAMEI_CONTLOOKUP);
+
+               error = lookup_handle_found_vnode(ndp, &ndp->ni_cnd, rdonly, vbusyflags, 
+                               &keep_going, ndp->ni_ncgeneration, wantparent, 0, ctx);
+               if (error)
+                       goto out_drop;
+               if (keep_going) {
+                       if ((cnp->cn_flags & ISSYMLINK) == 0) {
+                               panic("We need to keep going on a continued lookup, but for vp type %d (tag %d)\n", ndp->ni_vp->v_type, ndp->ni_vp->v_tag);
+                       }
+                       goto continue_symlink;
+               }
+
+               return 0;
+
+       }
 
 vnode_recycled:
 
@@ -310,9 +347,6 @@ retry_copy:
        ndp->ni_vp  = NULLVP;
 
        for (;;) {
-               int need_newpathbuf;
-               u_int linklen;
-
                ndp->ni_startdir = dp;
 
                if ( (error = lookup(ndp)) ) {
@@ -324,104 +358,13 @@ retry_copy:
                if ((cnp->cn_flags & ISSYMLINK) == 0) {
                        return (0);
                }
-#ifndef __LP64__
-               if ((cnp->cn_flags & FSNODELOCKHELD)) {
-                       cnp->cn_flags &= ~FSNODELOCKHELD;
-                       unlock_fsnode(ndp->ni_dvp, NULL);
-               }       
-#endif /* __LP64__ */
-
-               if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
-                       error = ELOOP;
-                       break;
-               }
-#if CONFIG_MACF
-               if ((error = mac_vnode_check_readlink(ctx, ndp->ni_vp)) != 0)
-                       break;
-#endif /* MAC */
-               if (ndp->ni_pathlen > 1 || !(cnp->cn_flags & HASBUF))
-                       need_newpathbuf = 1;
-               else
-                       need_newpathbuf = 0;
-
-               if (need_newpathbuf) {
-                       MALLOC_ZONE(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
-                       if (cp == NULL) {
-                               error = ENOMEM;
-                               break;
-                       }
-               } else {
-                       cp = cnp->cn_pnbuf;
-               }
-               auio = uio_createwithbuffer(1, 0, UIO_SYSSPACE, UIO_READ, &uio_buf[0], sizeof(uio_buf));
-
-               uio_addiov(auio, CAST_USER_ADDR_T(cp), MAXPATHLEN);
-
-               error = VNOP_READLINK(ndp->ni_vp, auio, ctx);
-               if (error) {
-                       if (need_newpathbuf)
-                               FREE_ZONE(cp, MAXPATHLEN, M_NAMEI);
-                       break;
-               }
-
-               /* 
-                * Safe to set unsigned with a [larger] signed type here
-                * because 0 <= uio_resid <= MAXPATHLEN and MAXPATHLEN
-                * is only 1024.
-                */
-               linklen = MAXPATHLEN - (u_int)uio_resid(auio);
-               if (linklen + ndp->ni_pathlen > MAXPATHLEN) {
-                       if (need_newpathbuf)
-                               FREE_ZONE(cp, MAXPATHLEN, M_NAMEI);
 
-                       error = ENAMETOOLONG;
+continue_symlink:
+               /* Gives us a new path to process, and a starting dir */
+               error = lookup_handle_symlink(ndp, &dp, ctx);
+               if (error != 0) {
                        break;
                }
-               if (need_newpathbuf) {
-                       long len = cnp->cn_pnlen;
-
-                       tmppn = cnp->cn_pnbuf;
-                       bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
-                       cnp->cn_pnbuf = cp;
-                       cnp->cn_pnlen = MAXPATHLEN;
-
-                       if ( (cnp->cn_flags & HASBUF) )
-                               FREE_ZONE(tmppn, len, M_NAMEI);
-                       else
-                               cnp->cn_flags |= HASBUF;
-               } else
-                       cnp->cn_pnbuf[linklen] = '\0';
-
-               ndp->ni_pathlen += linklen;
-               cnp->cn_nameptr = cnp->cn_pnbuf;
-
-               /*
-                * starting point for 'relative'
-                * symbolic link path
-                */
-               dp = ndp->ni_dvp;
-               /*
-                * get rid of references returned via 'lookup'
-                */
-               vnode_put(ndp->ni_vp);
-               vnode_put(ndp->ni_dvp);
-
-               ndp->ni_vp = NULLVP;
-               ndp->ni_dvp = NULLVP;
-
-               /*
-                * Check if symbolic link restarts us at the root
-                */
-               if (*(cnp->cn_nameptr) == '/') {
-                       while (*(cnp->cn_nameptr) == '/') {
-                               cnp->cn_nameptr++;
-                               ndp->ni_pathlen--;
-                       }
-                       if ((dp = ndp->ni_rootdir) == NULLVP) {
-                               error = ENOENT;
-                               goto error_out;
-                       }
-               }
        }
        /*
         * only come here if we fail to handle a SYMLINK...
@@ -429,6 +372,7 @@ retry_copy:
         * we need to drop the iocount that was picked
         * up in the lookup routine
         */
+out_drop:
        if (ndp->ni_dvp)
                vnode_put(ndp->ni_dvp);
        if (ndp->ni_vp)
@@ -440,6 +384,7 @@ retry_copy:
        }
        cnp->cn_pnbuf = NULL;
        ndp->ni_vp = NULLVP;
+       ndp->ni_dvp = NULLVP;
        if (error == ERECYCLE){
                /* vnode was recycled underneath us. re-drive lookup to start at 
                   the beginning again, since recycling invalidated last lookup*/
@@ -452,143 +397,534 @@ retry_copy:
        return (error);
 }
 
+int            
+namei_compound_available(vnode_t dp, struct nameidata *ndp)
+{
+       if ((ndp->ni_flag & NAMEI_COMPOUNDOPEN) != 0) {
+               return vnode_compound_open_available(dp);
+       }
 
-/*
- * Search a pathname.
- * This is a very central and rather complicated routine.
- *
- * The pathname is pointed to by ni_ptr and is of length ni_pathlen.
- * The starting directory is taken from ni_startdir. The pathname is
- * descended until done, or a symbolic link is encountered. The variable
- * ni_more is clear if the path is completed; it is set to one if a
- * symbolic link needing interpretation is encountered.
- *
- * The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
- * whether the name is to be looked up, created, renamed, or deleted.
- * When CREATE, RENAME, or DELETE is specified, information usable in
- * creating, renaming, or deleting a directory entry may be calculated.
- * If flag has LOCKPARENT or'ed into it, the parent directory is returned
- * locked. If flag has WANTPARENT or'ed into it, the parent directory is
- * returned unlocked. Otherwise the parent directory is not returned. If
- * the target of the pathname exists and LOCKLEAF is or'ed into the flag
- * the target is returned locked, otherwise it is returned unlocked.
- * When creating or renaming and LOCKPARENT is specified, the target may not
- * be ".".  When deleting and LOCKPARENT is specified, the target may be ".".
- * 
- * Overall outline of lookup:
- *
- * dirloop:
- *     identify next component of name at ndp->ni_ptr
- *     handle degenerate case where name is null string
- *     if .. and crossing mount points and on mounted filesys, find parent
- *     call VNOP_LOOKUP routine for next component name
- *         directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
- *         component vnode returned in ni_vp (if it exists), locked.
- *     if result vnode is mounted on and crossing mount points,
- *         find mounted on vnode
- *     if more components of name, do next level at dirloop
- *     return the answer in ni_vp, locked if LOCKLEAF set
- *         if LOCKPARENT set, return locked parent in ni_dvp
- *         if WANTPARENT set, return unlocked parent in ni_dvp
- *
- * Returns:    0                       Success
- *             ENOENT                  No such file or directory
- *             EBADF                   Bad file descriptor
- *             ENOTDIR                 Not a directory
- *             EROFS                   Read-only file system [CREATE]
- *             EISDIR                  Is a directory [CREATE]
- *             cache_lookup_path:ERECYCLE  (vnode was recycled from underneath us, redrive lookup again)
- *             vnode_authorize:EROFS
- *             vnode_authorize:EACCES
- *             vnode_authorize:EPERM
- *             vnode_authorize:???
- *             VNOP_LOOKUP:ENOENT      No such file or directory
- *             VNOP_LOOKUP:EJUSTRETURN Restart system call (INTERNAL)
- *             VNOP_LOOKUP:???
- *             VFS_ROOT:ENOTSUP
- *             VFS_ROOT:ENOENT
- *             VFS_ROOT:???
- */
+       return 0;
+}
 int
-lookup(struct nameidata *ndp)
+lookup_authorize_search(vnode_t dp, struct componentname *cnp, int dp_authorized_in_cache, vfs_context_t ctx)
 {
-       char    *cp;            /* pointer into pathname argument */
-       vnode_t         tdp;            /* saved dp */
-       vnode_t         dp;             /* the directory we are searching */
-       mount_t         mp;             /* mount table entry */
-       int docache = 1;                /* == 0 do not cache last component */
-       int wantparent;                 /* 1 => wantparent or lockparent flag */
-       int rdonly;                     /* lookup read-only flag bit */
-       int trailing_slash = 0;
-       int dp_authorized = 0;
-       int error = 0;
-       struct componentname *cnp = &ndp->ni_cnd;
-       vfs_context_t ctx = cnp->cn_context;
-       int mounted_on_depth = 0;
-       int dont_cache_mp = 0;
-       vnode_t mounted_on_dp = NULLVP;
-       int current_mount_generation = 0;
-       int vbusyflags = 0;
-       int nc_generation = 0;
-       vnode_t last_dp = NULLVP;
+       int error;
 
-       /*
-        * Setup: break out flag bits into variables.
-        */
-       if (cnp->cn_flags & (NOCACHE | DOWHITEOUT)) {
-               if ((cnp->cn_flags & NOCACHE) || (cnp->cn_nameiop == DELETE))
-                       docache = 0;
+       if (!dp_authorized_in_cache) {
+               error = vnode_authorize(dp, NULL, KAUTH_VNODE_SEARCH, ctx);
+               if (error)
+                       return error;
        }
-       wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
-       rdonly = cnp->cn_flags & RDONLY;
-       cnp->cn_flags &= ~ISSYMLINK;
-       cnp->cn_consume = 0;
+#if CONFIG_MACF
+       error = mac_vnode_check_lookup(ctx, dp, cnp);
+       if (error)
+               return error;
+#endif /* CONFIG_MACF */
 
-       dp = ndp->ni_startdir;
-       ndp->ni_startdir = NULLVP;
+       return 0;
+}
 
-       if ((cnp->cn_flags & CN_NBMOUNTLOOK) != 0)
-                       vbusyflags = LK_NOWAIT;
-       cp = cnp->cn_nameptr;
+void 
+lookup_consider_update_cache(vnode_t dvp, vnode_t vp, struct componentname *cnp, int nc_generation) 
+{
+       int isdot_or_dotdot;
+       isdot_or_dotdot = (cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') || (cnp->cn_flags & ISDOTDOT);
 
-       if (*cp == '\0') {
-               if ( (vnode_getwithref(dp)) ) {
-                       dp = NULLVP;
-                       error = ENOENT;
-                       goto bad;
+       if (vp->v_name == NULL || vp->v_parent == NULLVP) {
+               int  update_flags = 0;
+
+               if (isdot_or_dotdot == 0) {
+                       if (vp->v_name == NULL)
+                               update_flags |= VNODE_UPDATE_NAME;
+                       if (dvp != NULLVP && vp->v_parent == NULLVP)
+                               update_flags |= VNODE_UPDATE_PARENT;
+
+                       if (update_flags)
+                               vnode_update_identity(vp, dvp, cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_hash, update_flags);
                }
-               goto emptyname;
        }
-dirloop: 
-       ndp->ni_vp = NULLVP;
+       if ( (cnp->cn_flags & MAKEENTRY) && (vp->v_flag & VNCACHEABLE) && LIST_FIRST(&vp->v_nclinks) == NULL) {
+               /*
+                * missing from name cache, but should
+                * be in it... this can happen if volfs
+                * causes the vnode to be created or the
+                * name cache entry got recycled but the
+                * vnode didn't...
+                * check to make sure that ni_dvp is valid
+                * cache_lookup_path may return a NULL
+                * do a quick check to see if the generation of the
+                * directory matches our snapshot... this will get
+                * rechecked behind the name cache lock, but if it
+                * already fails to match, no need to go any further
+                */
+               if (dvp != NULLVP && (nc_generation == dvp->v_nc_generation) && (!isdot_or_dotdot))
+                       cache_enter_with_gen(dvp, vp, cnp, nc_generation);
+       }
 
-       if ( (error = cache_lookup_path(ndp, cnp, dp, ctx, &trailing_slash, &dp_authorized, last_dp)) ) {
-               dp = NULLVP;
-               goto bad;
+}
+
+#if NAMEDRSRCFORK
+/*
+ * Can change ni_dvp and ni_vp.  On success, returns with iocounts on stream vnode (always) and
+ * data fork if requested.  On failure, returns with iocount data fork (always) and its parent directory 
+ * (if one was provided).
+ */
+int
+lookup_handle_rsrc_fork(vnode_t dp, struct nameidata *ndp, struct componentname *cnp, int wantparent, vfs_context_t ctx)
+{
+       vnode_t svp = NULLVP;
+       enum nsoperation nsop;
+       int error;
+
+       if (dp->v_type != VREG) {
+               error = ENOENT;
+               goto out;
        }
-       if ((cnp->cn_flags & ISLASTCN)) {
-               if (docache)
-                       cnp->cn_flags |= MAKEENTRY;
-       } else
-               cnp->cn_flags |= MAKEENTRY;
+       switch (cnp->cn_nameiop) {
+               case DELETE:
+                       if (cnp->cn_flags & CN_ALLOWRSRCFORK) {
+                               nsop = NS_DELETE;
+                       } else {
+                               error = EPERM;
+                               goto out;
+                       }
+                       break;
+               case CREATE:
+                       if (cnp->cn_flags & CN_ALLOWRSRCFORK) {
+                               nsop = NS_CREATE;
+                       } else {
+                               error = EPERM;
+                               goto out;
+                       }
+                       break;
+               case LOOKUP:
+                       /* Make sure our lookup of "/..namedfork/rsrc" is allowed. */
+                       if (cnp->cn_flags & CN_ALLOWRSRCFORK) {
+                               nsop = NS_OPEN;
+                       } else {
+                               error = EPERM;
+                               goto out;
+                       }
+                       break;
+               default:
+                       error = EPERM;
+                       goto out;
+       }
+       /* Ask the file system for the resource fork. */
+       error = vnode_getnamedstream(dp, &svp, XATTR_RESOURCEFORK_NAME, nsop, 0, ctx);
 
-       dp = ndp->ni_dvp;
+       /* During a create, it OK for stream vnode to be missing. */
+       if (error == ENOATTR || error == ENOENT) {
+               error = (nsop == NS_CREATE) ? 0 : ENOENT;
+       }               
+       if (error) {
+               goto out;
+       }
+       /* The "parent" of the stream is the file. */
+       if (wantparent) {
+               if (ndp->ni_dvp) {
+#ifndef __LP64__
+                       if (ndp->ni_cnd.cn_flags & FSNODELOCKHELD) {
+                               ndp->ni_cnd.cn_flags &= ~FSNODELOCKHELD;
+                               unlock_fsnode(ndp->ni_dvp, NULL);
+                       }       
+#endif /* __LP64__ */
+                       vnode_put(ndp->ni_dvp);
+               }
+               ndp->ni_dvp = dp;
+       } else {
+               vnode_put(dp);
+       }
+       ndp->ni_vp = svp;  /* on create this may be null */
 
-       if (ndp->ni_vp != NULLVP) {
-               /*
-                * cache_lookup_path returned a non-NULL ni_vp then,
-                * we're guaranteed that the dp is a VDIR, it's 
-                * been authorized, and vp is not ".."
-                *
-                * make sure we don't try to enter the name back into
-                * the cache if this vp is purged before we get to that
-                * check since we won't have serialized behind whatever
-                * activity is occurring in the FS that caused the purge
-                */
-               if (dp != NULLVP)
-                       nc_generation = dp->v_nc_generation - 1;
+       /* Restore the truncated pathname buffer (for audits). */
+       if (ndp->ni_pathlen == 1 && ndp->ni_next[0] == '\0') {
+               ndp->ni_next[0] = '/';
+       }
+       cnp->cn_flags  &= ~MAKEENTRY;
 
-               goto returned_from_lookup_path;
+       return 0;
+out:
+       return error;
+}
+#endif /* NAMEDRSRCFORK */
+
+/*
+ * iocounts in:
+ *     --One on ni_vp.  One on ni_dvp if there is more path, or we didn't come through the
+ *     cache, or we came through the cache and the caller doesn't want the parent.
+ *
+ * iocounts out:
+ *     --Leaves us in the correct state for the next step, whatever that might be.
+ *     --If we find a symlink, returns with iocounts on both ni_vp and ni_dvp.
+ *     --If we are to look up another component, then we have an iocount on ni_vp and
+ *     nothing else.  
+ *     --If we are done, returns an iocount on ni_vp, and possibly on ni_dvp depending on nameidata flags.
+ *     --In the event of an error, may return with ni_dvp NULL'ed out (in which case, iocount
+ *     was dropped).
+ */
+int            
+lookup_handle_found_vnode(struct nameidata *ndp, struct componentname *cnp, int rdonly, 
+               int vbusyflags, int *keep_going, int nc_generation,
+               int wantparent, int atroot, vfs_context_t ctx)
+{
+       vnode_t dp;
+       int error;
+       char *cp;
+
+       dp = ndp->ni_vp;
+       *keep_going = 0;
+
+       if (ndp->ni_vp == NULLVP) {
+               panic("NULL ni_vp in %s\n", __FUNCTION__);
+       }
+
+       if (atroot) {
+               goto nextname;
+       }
+
+#if CONFIG_TRIGGERS
+       if (dp->v_resolve) {
+               error = vnode_trigger_resolve(dp, ndp, ctx);
+               if (error) {
+                       goto out;
+               }
+       }
+#endif /* CONFIG_TRIGGERS */
+
+       /*
+        * Take into account any additional components consumed by
+        * the underlying filesystem.
+        */
+       if (cnp->cn_consume > 0) {
+               cnp->cn_nameptr += cnp->cn_consume;
+               ndp->ni_next += cnp->cn_consume;
+               ndp->ni_pathlen -= cnp->cn_consume;
+               cnp->cn_consume = 0;
+       } else {
+               lookup_consider_update_cache(ndp->ni_dvp, dp, cnp, nc_generation);
+       }
+
+       /*
+        * Check to see if the vnode has been mounted on...
+        * if so find the root of the mounted file system.
+        * Updates ndp->ni_vp.
+        */
+       error = lookup_traverse_mountpoints(ndp, cnp, dp, vbusyflags, ctx);
+       dp = ndp->ni_vp;
+       if (error) {
+               goto out;
+       }
+
+#if CONFIG_MACF
+       if (vfs_flags(vnode_mount(dp)) & MNT_MULTILABEL) {
+               error = vnode_label(vnode_mount(dp), NULL, dp, NULL, 0, ctx);
+               if (error)
+                       goto out;
+       }
+#endif
+
+       /*
+        * Check for symbolic link
+        */
+       if ((dp->v_type == VLNK) &&
+           ((cnp->cn_flags & FOLLOW) || (ndp->ni_flag & NAMEI_TRAILINGSLASH) || *ndp->ni_next == '/')) {
+               cnp->cn_flags |= ISSYMLINK;
+               *keep_going = 1;
+               return (0);
+       }
+
+       /*
+        * Check for bogus trailing slashes.
+        */
+       if ((ndp->ni_flag & NAMEI_TRAILINGSLASH)) {
+               if (dp->v_type != VDIR) {
+                       error = ENOTDIR;
+                       goto out;
+               }
+               ndp->ni_flag &= ~(NAMEI_TRAILINGSLASH);
+       } 
+       
+nextname:
+       /*
+        * Not a symbolic link.  If more pathname,
+        * continue at next component, else return.
+        *
+        * Definitely have a dvp if there's another slash 
+        */
+       if (*ndp->ni_next == '/') {
+               cnp->cn_nameptr = ndp->ni_next + 1;
+               ndp->ni_pathlen--;
+               while (*cnp->cn_nameptr == '/') {
+                       cnp->cn_nameptr++;
+                       ndp->ni_pathlen--;
+               }
+
+               cp = cnp->cn_nameptr;
+               vnode_put(ndp->ni_dvp);
+               ndp->ni_dvp = NULLVP;
+
+               if (*cp == '\0') {
+                       goto emptyname;
+               }
+
+               *keep_going = 1;
+               return 0;
+       }
+                                 
+       /*
+        * Disallow directory write attempts on read-only file systems.
+        */
+       if (rdonly &&
+           (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
+               error = EROFS;
+               goto out;
+       }
+       
+       /* If SAVESTART is set, we should have a dvp */
+       if (cnp->cn_flags & SAVESTART) {
+               /*      
+                * note that we already hold a reference
+                * on both dp and ni_dvp, but for some reason
+                * can't get another one... in this case we
+                * need to do vnode_put on dp in 'bad2'
+                */
+               if ( (vnode_get(ndp->ni_dvp)) ) {
+                       error = ENOENT;
+                       goto out;
+               }
+               ndp->ni_startdir = ndp->ni_dvp;
+       }
+       if (!wantparent && ndp->ni_dvp) {
+               vnode_put(ndp->ni_dvp);
+               ndp->ni_dvp = NULLVP;
+       }
+
+       if (cnp->cn_flags & AUDITVNPATH1)
+               AUDIT_ARG(vnpath, dp, ARG_VNODE1);
+       else if (cnp->cn_flags & AUDITVNPATH2)
+               AUDIT_ARG(vnpath, dp, ARG_VNODE2);
+
+#if NAMEDRSRCFORK
+       /*
+        * Caller wants the resource fork.
+        */
+       if ((cnp->cn_flags & CN_WANTSRSRCFORK) && (dp != NULLVP)) {
+               error = lookup_handle_rsrc_fork(dp, ndp, cnp, wantparent, ctx);
+               if (error != 0)
+                       goto out;
+
+               dp = ndp->ni_vp;
+       }
+#endif
+       if (kdebug_enable)
+               kdebug_lookup(dp, cnp);
+
+       return 0;
+
+emptyname:
+       error = lookup_handle_emptyname(ndp, cnp, wantparent);
+       if (error != 0) 
+               goto out;
+
+       return 0;
+out:
+       return error;
+
+}
+
+/*
+ * Comes in iocount on ni_vp.  May overwrite ni_dvp, but doesn't interpret incoming value.
+ */
+int 
+lookup_handle_emptyname(struct nameidata *ndp, struct componentname *cnp, int wantparent)
+{
+       vnode_t dp;
+       int error = 0;
+
+       dp = ndp->ni_vp;
+       cnp->cn_namelen = 0;
+       /*
+        * A degenerate name (e.g. / or "") which is a way of
+        * talking about a directory, e.g. like "/." or ".".
+        */
+       if (dp->v_type != VDIR) {
+               error = ENOTDIR;
+               goto out;
+       }
+       if (cnp->cn_nameiop != LOOKUP) {
+               error = EISDIR;
+               goto out;
+       }
+       if (wantparent) {
+               /*      
+                * note that we already hold a reference
+                * on dp, but for some reason can't
+                * get another one... in this case we
+                * need to do vnode_put on dp in 'bad'
+                */
+               if ( (vnode_get(dp)) ) {
+                       error = ENOENT;
+                       goto out;
+               }
+               ndp->ni_dvp = dp;
+       }
+       cnp->cn_flags &= ~ISDOTDOT;
+       cnp->cn_flags |= ISLASTCN;
+       ndp->ni_next = cnp->cn_nameptr;
+       ndp->ni_vp = dp;
+
+       if (cnp->cn_flags & AUDITVNPATH1)
+               AUDIT_ARG(vnpath, dp, ARG_VNODE1);
+       else if (cnp->cn_flags & AUDITVNPATH2)
+               AUDIT_ARG(vnpath, dp, ARG_VNODE2);
+       if (cnp->cn_flags & SAVESTART)
+               panic("lookup: SAVESTART");
+
+       return 0;
+out:
+       return error;
+}
+/*
+ * Search a pathname.
+ * This is a very central and rather complicated routine.
+ *
+ * The pathname is pointed to by ni_ptr and is of length ni_pathlen.
+ * The starting directory is taken from ni_startdir. The pathname is
+ * descended until done, or a symbolic link is encountered. The variable
+ * ni_more is clear if the path is completed; it is set to one if a
+ * symbolic link needing interpretation is encountered.
+ *
+ * The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
+ * whether the name is to be looked up, created, renamed, or deleted.
+ * When CREATE, RENAME, or DELETE is specified, information usable in
+ * creating, renaming, or deleting a directory entry may be calculated.
+ * If flag has LOCKPARENT or'ed into it, the parent directory is returned
+ * locked. If flag has WANTPARENT or'ed into it, the parent directory is
+ * returned unlocked. Otherwise the parent directory is not returned. If
+ * the target of the pathname exists and LOCKLEAF is or'ed into the flag
+ * the target is returned locked, otherwise it is returned unlocked.
+ * When creating or renaming and LOCKPARENT is specified, the target may not
+ * be ".".  When deleting and LOCKPARENT is specified, the target may be ".".
+ * 
+ * Overall outline of lookup:
+ *
+ * dirloop:
+ *     identify next component of name at ndp->ni_ptr
+ *     handle degenerate case where name is null string
+ *     if .. and crossing mount points and on mounted filesys, find parent
+ *     call VNOP_LOOKUP routine for next component name
+ *         directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
+ *         component vnode returned in ni_vp (if it exists), locked.
+ *     if result vnode is mounted on and crossing mount points,
+ *         find mounted on vnode
+ *     if more components of name, do next level at dirloop
+ *     return the answer in ni_vp, locked if LOCKLEAF set
+ *         if LOCKPARENT set, return locked parent in ni_dvp
+ *         if WANTPARENT set, return unlocked parent in ni_dvp
+ *
+ * Returns:    0                       Success
+ *             ENOENT                  No such file or directory
+ *             EBADF                   Bad file descriptor
+ *             ENOTDIR                 Not a directory
+ *             EROFS                   Read-only file system [CREATE]
+ *             EISDIR                  Is a directory [CREATE]
+ *             cache_lookup_path:ERECYCLE  (vnode was recycled from underneath us, redrive lookup again)
+ *             vnode_authorize:EROFS
+ *             vnode_authorize:EACCES
+ *             vnode_authorize:EPERM
+ *             vnode_authorize:???
+ *             VNOP_LOOKUP:ENOENT      No such file or directory
+ *             VNOP_LOOKUP:EJUSTRETURN Restart system call (INTERNAL)
+ *             VNOP_LOOKUP:???
+ *             VFS_ROOT:ENOTSUP
+ *             VFS_ROOT:ENOENT
+ *             VFS_ROOT:???
+ */
+int
+lookup(struct nameidata *ndp)
+{
+       char    *cp;            /* pointer into pathname argument */
+       vnode_t         tdp;            /* saved dp */
+       vnode_t         dp;             /* the directory we are searching */
+       int docache = 1;                /* == 0 do not cache last component */
+       int wantparent;                 /* 1 => wantparent or lockparent flag */
+       int rdonly;                     /* lookup read-only flag bit */
+       int dp_authorized = 0;
+       int error = 0;
+       struct componentname *cnp = &ndp->ni_cnd;
+       vfs_context_t ctx = cnp->cn_context;
+       int vbusyflags = 0;
+       int nc_generation = 0;
+       vnode_t last_dp = NULLVP;
+       int keep_going;
+       int atroot;
+
+       /*
+        * Setup: break out flag bits into variables.
+        */
+       if (cnp->cn_flags & (NOCACHE | DOWHITEOUT)) {
+               if ((cnp->cn_flags & NOCACHE) || (cnp->cn_nameiop == DELETE))
+                       docache = 0;
+       }
+       wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
+       rdonly = cnp->cn_flags & RDONLY;
+       cnp->cn_flags &= ~ISSYMLINK;
+       cnp->cn_consume = 0;
+
+       dp = ndp->ni_startdir;
+       ndp->ni_startdir = NULLVP;
+
+       if ((cnp->cn_flags & CN_NBMOUNTLOOK) != 0)
+                       vbusyflags = LK_NOWAIT;
+       cp = cnp->cn_nameptr;
+
+       if (*cp == '\0') {
+               if ( (vnode_getwithref(dp)) ) {
+                       dp = NULLVP;
+                       error = ENOENT;
+                       goto bad;
+               }
+               ndp->ni_vp = dp;
+               error = lookup_handle_emptyname(ndp, cnp, wantparent);
+               if (error) {
+                       goto bad;
+               }
+
+               return 0;
+       }
+dirloop: 
+       atroot = 0;
+       ndp->ni_vp = NULLVP;
+
+       if ( (error = cache_lookup_path(ndp, cnp, dp, ctx, &dp_authorized, last_dp)) ) {
+               dp = NULLVP;
+               goto bad;
+       }
+       if ((cnp->cn_flags & ISLASTCN)) {
+               if (docache)
+                       cnp->cn_flags |= MAKEENTRY;
+       } else
+               cnp->cn_flags |= MAKEENTRY;
+
+       dp = ndp->ni_dvp;
+
+       if (ndp->ni_vp != NULLVP) {
+               /*
+                * cache_lookup_path returned a non-NULL ni_vp then,
+                * we're guaranteed that the dp is a VDIR, it's 
+                * been authorized, and vp is not ".."
+                *
+                * make sure we don't try to enter the name back into
+                * the cache if this vp is purged before we get to that
+                * check since we won't have serialized behind whatever
+                * activity is occurring in the FS that caused the purge
+                */
+               if (dp != NULLVP)
+                       nc_generation = dp->v_nc_generation - 1;
+
+               goto returned_from_lookup_path;
        }
 
        /*
@@ -618,7 +954,8 @@ dirloop:
                                        error = ENOENT;
                                        goto bad;
                                }
-                               goto nextname;
+                               atroot = 1;
+                               goto returned_from_lookup_path;
                        }
                        if ((dp->v_flag & VROOT) == 0 ||
                            (cnp->cn_flags & NOCROSSMOUNT))
@@ -653,21 +990,32 @@ unionlookup:
                goto lookup_error;
        }
        if ( (cnp->cn_flags & DONOTAUTH) != DONOTAUTH ) {
-               if (!dp_authorized) {
-                       error = vnode_authorize(dp, NULL, KAUTH_VNODE_SEARCH, ctx);
-                       if (error)
-                               goto lookup_error;
-               }
-#if CONFIG_MACF
-               error = mac_vnode_check_lookup(ctx, dp, cnp);
-               if (error)
+               error = lookup_authorize_search(dp, cnp, dp_authorized, ctx);
+               if (error) {
                        goto lookup_error;
-#endif /* CONFIG_MACF */
+               }
+       }
+
+       /*
+        * Now that we've authorized a lookup, can bail out if the filesystem
+        * will be doing a batched operation.  Return an iocount on dvp.
+        */
+#if NAMEDRSRCFORK
+       if ((cnp->cn_flags & ISLASTCN) && namei_compound_available(dp, ndp) && !(cnp->cn_flags & CN_WANTSRSRCFORK)) {
+#else 
+       if ((cnp->cn_flags & ISLASTCN) && namei_compound_available(dp, ndp)) {
+#endif /* NAMEDRSRCFORK */
+               ndp->ni_flag |= NAMEI_UNFINISHED;
+               ndp->ni_ncgeneration = dp->v_nc_generation;
+               return 0;
        }
 
         nc_generation = dp->v_nc_generation;
 
-       if ( (error = VNOP_LOOKUP(dp, &ndp->ni_vp, cnp, ctx)) ) {
+       error = VNOP_LOOKUP(dp, &ndp->ni_vp, cnp, ctx);
+
+
+       if ( error ) {
 lookup_error:
                if ((error == ENOENT) &&
                    (dp->v_flag & VROOT) && (dp->v_mount != NULL) &&
@@ -699,18 +1047,9 @@ lookup_error:
                if (ndp->ni_vp != NULLVP)
                        panic("leaf should be empty");
 
-               /*
-                * If creating and at end of pathname, then can consider
-                * allowing file to be created.
-                */
-               if (rdonly) {
-                       error = EROFS;
-                       goto bad;
-               }
-               if ((cnp->cn_flags & ISLASTCN) && trailing_slash && !(cnp->cn_flags & WILLBEDIR)) {
-                       error = ENOENT;
+               error = lookup_validate_creation_path(ndp);
+               if (error)
                        goto bad;
-               }
                /*
                 * We return with ni_vp NULL to indicate that the entry
                 * doesn't currently exist, leaving a pointer to the
@@ -731,337 +1070,33 @@ lookup_error:
                return (0);
        }
 returned_from_lookup_path:
-       dp = ndp->ni_vp;
-
-       /*
-        * Take into account any additional components consumed by
-        * the underlying filesystem.
-        */
-       if (cnp->cn_consume > 0) {
-               cnp->cn_nameptr += cnp->cn_consume;
-               ndp->ni_next += cnp->cn_consume;
-               ndp->ni_pathlen -= cnp->cn_consume;
-               cnp->cn_consume = 0;
-       } else {
-               int isdot_or_dotdot;
-               isdot_or_dotdot = (cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') || (cnp->cn_flags & ISDOTDOT);
-
-               if (dp->v_name == NULL || dp->v_parent == NULLVP) {
-                       int  update_flags = 0;
-
-                       if (isdot_or_dotdot == 0) {
-                               if (dp->v_name == NULL)
-                                       update_flags |= VNODE_UPDATE_NAME;
-                               if (ndp->ni_dvp != NULLVP && dp->v_parent == NULLVP)
-                                       update_flags |= VNODE_UPDATE_PARENT;
-
-                               if (update_flags)
-                                       vnode_update_identity(dp, ndp->ni_dvp, cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_hash, update_flags);
-                       }
-               }
-               if ( (cnp->cn_flags & MAKEENTRY) && (dp->v_flag & VNCACHEABLE) && LIST_FIRST(&dp->v_nclinks) == NULL) {
-                       /*
-                        * missing from name cache, but should
-                        * be in it... this can happen if volfs
-                        * causes the vnode to be created or the
-                        * name cache entry got recycled but the
-                        * vnode didn't...
-                        * check to make sure that ni_dvp is valid
-                        * cache_lookup_path may return a NULL
-                        * do a quick check to see if the generation of the
-                        * directory matches our snapshot... this will get
-                        * rechecked behind the name cache lock, but if it
-                        * already fails to match, no need to go any further
-                        */
-                       if (ndp->ni_dvp != NULLVP && (nc_generation == ndp->ni_dvp->v_nc_generation) && (!isdot_or_dotdot))
-                               cache_enter_with_gen(ndp->ni_dvp, dp, cnp, nc_generation);
-               }
-       }
-
-       mounted_on_dp = dp;
-       mounted_on_depth = 0;
-       dont_cache_mp = 0;
-       current_mount_generation = mount_generation;
-       /*
-        * Check to see if the vnode has been mounted on...
-        * if so find the root of the mounted file system.
-        */
-check_mounted_on:
-       if ((dp->v_type == VDIR) && dp->v_mountedhere &&
-            ((cnp->cn_flags & NOCROSSMOUNT) == 0)) {
-         
-               vnode_lock(dp);
-
-               if ((dp->v_type == VDIR) && (mp = dp->v_mountedhere)) {
-                       struct uthread *uth = (struct uthread *)get_bsdthread_info(current_thread());
-
-                       mp->mnt_crossref++;
-                       vnode_unlock(dp);
-
-                               
-                       if (vfs_busy(mp, vbusyflags)) {
-                               mount_dropcrossref(mp, dp, 0);
-                               if (vbusyflags == LK_NOWAIT) {
-                                       error = ENOENT;
-                                       goto bad2;      
-                               }
-                               goto check_mounted_on;
-                       }
-
-                       /*
-                        * XXX - if this is the last component of the
-                        * pathname, and it's either not a lookup operation
-                        * or the NOTRIGGER flag is set for the operation,
-                        * set a uthread flag to let VFS_ROOT() for autofs
-                        * know it shouldn't trigger a mount.
-                        */
-                       if ((cnp->cn_flags & ISLASTCN) &&
-                           (cnp->cn_nameiop != LOOKUP ||
-                            (cnp->cn_flags & NOTRIGGER))) {
-                               uth->uu_notrigger = 1;
-                               dont_cache_mp = 1;
-                       }
-                       error = VFS_ROOT(mp, &tdp, ctx);
-                       /* XXX - clear the uthread flag */
-                       uth->uu_notrigger = 0;
-                       /*
-                        * mount_dropcrossref does a vnode_put
-                        * on dp if the 3rd arg is non-zero
-                        */
-                       mount_dropcrossref(mp, dp, 1);
-                       dp = NULL;
-                       vfs_unbusy(mp);
-
-                       if (error) {
-                               goto bad2;
-                       }
-                       ndp->ni_vp = dp = tdp;
-                       mounted_on_depth++;
-                       
-                       goto check_mounted_on;
-               } 
-               vnode_unlock(dp);
+       /* We'll always have an iocount on ni_vp when this finishes. */
+       error = lookup_handle_found_vnode(ndp, cnp, rdonly, vbusyflags, &keep_going, nc_generation, wantparent, atroot, ctx);
+       if (error != 0) {
+               goto bad2; 
        }
 
-#if CONFIG_MACF
-       if (vfs_flags(vnode_mount(dp)) & MNT_MULTILABEL) {
-               error = vnode_label(vnode_mount(dp), NULL, dp, NULL, 0, ctx);
-               if (error)
-                       goto bad2;
-       }
-#endif
-
-       if (mounted_on_depth && !dont_cache_mp) {
-               mp = mounted_on_dp->v_mountedhere;
-
-               if (mp) {
-                       mount_lock_spin(mp);
-                       mp->mnt_realrootvp_vid = dp->v_id;
-                       mp->mnt_realrootvp = dp;
-                       mp->mnt_generation = current_mount_generation;
-                       mount_unlock(mp);
-               }
-       }
-
-       /*
-        * Check for symbolic link
-        */
-       if ((dp->v_type == VLNK) &&
-           ((cnp->cn_flags & FOLLOW) || trailing_slash || *ndp->ni_next == '/')) {
-               cnp->cn_flags |= ISSYMLINK;
-               return (0);
-       }
-
-       /*
-        * Check for bogus trailing slashes.
-        */
-       if (trailing_slash) {
-               if (dp->v_type != VDIR) {
-                       error = ENOTDIR;
-                       goto bad2;
-               }
-               trailing_slash = 0;
-       }
-
-nextname:
-       /*
-        * Not a symbolic link.  If more pathname,
-        * continue at next component, else return.
-        */
-       if (*ndp->ni_next == '/') {
-               cnp->cn_nameptr = ndp->ni_next + 1;
-               ndp->ni_pathlen--;
-               while (*cnp->cn_nameptr == '/') {
-                       cnp->cn_nameptr++;
-                       ndp->ni_pathlen--;
-               }
-               vnode_put(ndp->ni_dvp);
-
-               cp = cnp->cn_nameptr;
-
-               if (*cp == '\0')
-                       goto emptyname;
-
-               /*
-                * cache_lookup_path is now responsible for dropping io ref on dp
-                * when it is called again in the dirloop.  This ensures we hold
-                * a ref on dp until we complete the next round of lookup.
-                */
-               last_dp = dp;
-               goto dirloop;
-       }
-                                 
-       /*
-        * Disallow directory write attempts on read-only file systems.
-        */
-       if (rdonly &&
-           (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
-               error = EROFS;
-               goto bad2;
-       }
-       if (cnp->cn_flags & SAVESTART) {
-               /*      
-                * note that we already hold a reference
-                * on both dp and ni_dvp, but for some reason
-                * can't get another one... in this case we
-                * need to do vnode_put on dp in 'bad2'
-                */
-               if ( (vnode_get(ndp->ni_dvp)) ) {
-                       error = ENOENT;
-                       goto bad2;
-               }
-               ndp->ni_startdir = ndp->ni_dvp;
-       }
-       if (!wantparent && ndp->ni_dvp) {
-               vnode_put(ndp->ni_dvp);
-               ndp->ni_dvp = NULLVP;
-       }
-
-       if (cnp->cn_flags & AUDITVNPATH1)
-               AUDIT_ARG(vnpath, dp, ARG_VNODE1);
-       else if (cnp->cn_flags & AUDITVNPATH2)
-               AUDIT_ARG(vnpath, dp, ARG_VNODE2);
-
-#if NAMEDRSRCFORK
-       /*
-        * Caller wants the resource fork.
-        */
-       if ((cnp->cn_flags & CN_WANTSRSRCFORK) && (dp != NULLVP)) {
-               vnode_t svp = NULLVP;
-               enum nsoperation nsop;
-
-               if (dp->v_type != VREG) {
-                       error = ENOENT;
-                       goto bad2;
-               }
-               switch (cnp->cn_nameiop) {
-               case DELETE:
-                       if (cnp->cn_flags & CN_ALLOWRSRCFORK) {
-                               nsop = NS_DELETE;
-                       } else {
-                               error = EPERM;
-                               goto bad2;
-                       }
-                       break;
-               case CREATE:
-                       if (cnp->cn_flags & CN_ALLOWRSRCFORK) {
-                               nsop = NS_CREATE;
-                       } else {
-                               error = EPERM;
-                               goto bad2;
-                       }
-                       break;
-               case LOOKUP:
-                       /* Make sure our lookup of "/..namedfork/rsrc" is allowed. */
-                       if (cnp->cn_flags & CN_ALLOWRSRCFORK) {
-                               nsop = NS_OPEN;
-                       } else {
-                               error = EPERM;
-                               goto bad2;
-                       }
-                       break;
-               default:
-                       error = EPERM;
-                       goto bad2;
-               }
-               /* Ask the file system for the resource fork. */
-               error = vnode_getnamedstream(dp, &svp, XATTR_RESOURCEFORK_NAME, nsop, 0, ctx);
-
-               /* During a create, it OK for stream vnode to be missing. */
-               if (error == ENOATTR || error == ENOENT) {
-                       error = (nsop == NS_CREATE) ? 0 : ENOENT;
-               }               
-               if (error) {
-                       goto bad2;
-               }
-               /* The "parent" of the stream is the file. */
-               if (wantparent) {
-                       if (ndp->ni_dvp) {
-#ifndef __LP64__
-                               if (ndp->ni_cnd.cn_flags & FSNODELOCKHELD) {
-                                       ndp->ni_cnd.cn_flags &= ~FSNODELOCKHELD;
-                                       unlock_fsnode(ndp->ni_dvp, NULL);
-                               }       
-#endif /* __LP64__ */
-                               vnode_put(ndp->ni_dvp);
-                       }
-                       ndp->ni_dvp = dp;
-               } else {
-                       vnode_put(dp);
-               }
-               ndp->ni_vp = dp = svp;  /* on create this may be null */
+       if (keep_going) {
+               dp = ndp->ni_vp;
 
-               /* Restore the truncated pathname buffer (for audits). */
-               if (ndp->ni_pathlen == 1 && ndp->ni_next[0] == '\0') {
-                       ndp->ni_next[0] = '/';
+               /* namei() will handle symlinks */
+               if ((dp->v_type == VLNK) &&
+                               ((cnp->cn_flags & FOLLOW) || (ndp->ni_flag & NAMEI_TRAILINGSLASH) || *ndp->ni_next == '/')) {
+                       return 0; 
                }
-               cnp->cn_flags  &= ~MAKEENTRY;
-       }
-#endif
-       if (kdebug_enable)
-               kdebug_lookup(dp, cnp);
-       return (0);
 
-emptyname:
-       cnp->cn_namelen = 0;
-       /*
-        * A degenerate name (e.g. / or "") which is a way of
-        * talking about a directory, e.g. like "/." or ".".
-        */
-       if (dp->v_type != VDIR) {
-               error = ENOTDIR;
-               goto bad;
-       }
-       if (cnp->cn_nameiop != LOOKUP) {
-               error = EISDIR;
-               goto bad;
-       }
-       if (wantparent) {
-               /*      
-                * note that we already hold a reference
-                * on dp, but for some reason can't
-                * get another one... in this case we
-                * need to do vnode_put on dp in 'bad'
+               /*
+                * Otherwise, there's more path to process.  
+                * cache_lookup_path is now responsible for dropping io ref on dp
+                * when it is called again in the dirloop.  This ensures we hold
+                * a ref on dp until we complete the next round of lookup.
                 */
-               if ( (vnode_get(dp)) ) {
-                       error = ENOENT;
-                       goto bad;
-               }
-               ndp->ni_dvp = dp;
+               last_dp = dp;
+
+               goto dirloop;
        }
-       cnp->cn_flags &= ~ISDOTDOT;
-       cnp->cn_flags |= ISLASTCN;
-       ndp->ni_next = cp;
-       ndp->ni_vp = dp;
 
-       if (cnp->cn_flags & AUDITVNPATH1)
-               AUDIT_ARG(vnpath, dp, ARG_VNODE1);
-       else if (cnp->cn_flags & AUDITVNPATH2)
-               AUDIT_ARG(vnpath, dp, ARG_VNODE2);
-       if (cnp->cn_flags & SAVESTART)
-               panic("lookup: SAVESTART");
        return (0);
-
 bad2:
 #ifndef __LP64__
        if ((cnp->cn_flags & FSNODELOCKHELD)) {
@@ -1070,9 +1105,9 @@ bad2:
        }
 #endif /* __LP64__ */
        if (ndp->ni_dvp)
-               vnode_put(ndp->ni_dvp);
-       if (dp)
-               vnode_put(dp);
+               vnode_put(ndp->ni_dvp);
+
+       vnode_put(ndp->ni_vp);
        ndp->ni_vp = NULLVP;
 
        if (kdebug_enable)
@@ -1095,6 +1130,257 @@ bad:
        return (error);
 }
 
+int 
+lookup_validate_creation_path(struct nameidata *ndp)
+{
+       struct componentname *cnp = &ndp->ni_cnd;
+
+       /*
+        * If creating and at end of pathname, then can consider
+        * allowing file to be created.
+        */
+       if (cnp->cn_flags & RDONLY) {
+               return EROFS;
+       }
+       if ((cnp->cn_flags & ISLASTCN) && (ndp->ni_flag & NAMEI_TRAILINGSLASH) && !(cnp->cn_flags & WILLBEDIR)) {
+               return ENOENT;
+       }
+       
+       return 0;
+}
+
+/*
+ * Modifies only ni_vp.  Always returns with ni_vp still valid (iocount held).
+ */
+int
+lookup_traverse_mountpoints(struct nameidata *ndp, struct componentname *cnp, vnode_t dp, 
+               int vbusyflags, vfs_context_t ctx)
+{
+       mount_t mp;
+       vnode_t tdp;
+       int error = 0;
+       uthread_t uth;
+       uint32_t depth = 0;
+       int dont_cache_mp = 0;
+       vnode_t mounted_on_dp;
+       int current_mount_generation = 0;
+       
+       mounted_on_dp = dp;
+       current_mount_generation = mount_generation;
+
+       while ((dp->v_type == VDIR) && dp->v_mountedhere &&
+                       ((cnp->cn_flags & NOCROSSMOUNT) == 0)) {
+#if CONFIG_TRIGGERS
+               /*
+                * For a trigger vnode, call its resolver when crossing its mount (if requested)
+                */
+               if (dp->v_resolve) {
+                       (void) vnode_trigger_resolve(dp, ndp, ctx);
+               }
+#endif
+               vnode_lock(dp);
+
+               if ((dp->v_type == VDIR) && (mp = dp->v_mountedhere)) {
+
+                       mp->mnt_crossref++;
+                       vnode_unlock(dp);
+
+
+                       if (vfs_busy(mp, vbusyflags)) {
+                               mount_dropcrossref(mp, dp, 0);
+                               if (vbusyflags == LK_NOWAIT) {
+                                       error = ENOENT;
+                                       goto out;
+                               }
+
+                               continue;
+                       }
+
+
+                       /*
+                        * XXX - if this is the last component of the
+                        * pathname, and it's either not a lookup operation
+                        * or the NOTRIGGER flag is set for the operation,
+                        * set a uthread flag to let VFS_ROOT() for autofs
+                        * know it shouldn't trigger a mount.
+                        */
+                       uth = (struct uthread *)get_bsdthread_info(current_thread());
+                       if ((cnp->cn_flags & ISLASTCN) &&
+                                       (cnp->cn_nameiop != LOOKUP ||
+                                        (cnp->cn_flags & NOTRIGGER))) {
+                               uth->uu_notrigger = 1;
+                               dont_cache_mp = 1;
+                       }
+
+                       error = VFS_ROOT(mp, &tdp, ctx);
+                       /* XXX - clear the uthread flag */
+                       uth->uu_notrigger = 0;
+
+                       mount_dropcrossref(mp, dp, 0);
+                       vfs_unbusy(mp);
+
+                       if (error) {
+                               goto out;
+                       }
+
+                       vnode_put(dp);
+                       ndp->ni_vp = dp = tdp;
+                       depth++;
+
+#if CONFIG_TRIGGERS
+                       /*
+                        * Check if root dir is a trigger vnode
+                        */
+                       if (dp->v_resolve) {
+                               error = vnode_trigger_resolve(dp, ndp, ctx);
+                               if (error) {
+                                       goto out;
+                               }
+                       }
+#endif                 
+
+               } else { 
+                       vnode_unlock(dp);
+                       break;
+               }
+       }
+
+       if (depth && !dont_cache_mp) {
+               mp = mounted_on_dp->v_mountedhere;
+
+               if (mp) {
+                       mount_lock_spin(mp);
+                       mp->mnt_realrootvp_vid = dp->v_id;
+                       mp->mnt_realrootvp = dp;
+                       mp->mnt_generation = current_mount_generation;
+                       mount_unlock(mp);
+               }
+       }
+
+       return 0;
+
+out:
+       return error;
+}
+
+/*
+ * Takes ni_vp and ni_dvp non-NULL.  Returns with *new_dp set to the location
+ * at which to start a lookup with a resolved path, and all other iocounts dropped.
+ */
+int 
+lookup_handle_symlink(struct nameidata *ndp, vnode_t *new_dp, vfs_context_t ctx)
+{
+       int error;
+       char *cp;               /* pointer into pathname argument */
+       uio_t auio;
+       char uio_buf[ UIO_SIZEOF(1) ];
+       int need_newpathbuf;
+       u_int linklen;
+       struct componentname *cnp = &ndp->ni_cnd;
+       vnode_t dp;
+       char *tmppn;
+
+#ifndef __LP64__
+       if ((cnp->cn_flags & FSNODELOCKHELD)) {
+               cnp->cn_flags &= ~FSNODELOCKHELD;
+               unlock_fsnode(ndp->ni_dvp, NULL);
+       }       
+#endif /* __LP64__ */
+
+       if (ndp->ni_loopcnt++ >= MAXSYMLINKS) {
+               return ELOOP;
+       }
+#if CONFIG_MACF
+       if ((error = mac_vnode_check_readlink(ctx, ndp->ni_vp)) != 0)
+               return error;
+#endif /* MAC */
+       if (ndp->ni_pathlen > 1 || !(cnp->cn_flags & HASBUF))
+               need_newpathbuf = 1;
+       else
+               need_newpathbuf = 0;
+
+       if (need_newpathbuf) {
+               MALLOC_ZONE(cp, char *, MAXPATHLEN, M_NAMEI, M_WAITOK);
+               if (cp == NULL) {
+                       return ENOMEM;
+               }
+       } else {
+               cp = cnp->cn_pnbuf;
+       }
+       auio = uio_createwithbuffer(1, 0, UIO_SYSSPACE, UIO_READ, &uio_buf[0], sizeof(uio_buf));
+
+       uio_addiov(auio, CAST_USER_ADDR_T(cp), MAXPATHLEN);
+
+       error = VNOP_READLINK(ndp->ni_vp, auio, ctx);
+       if (error) {
+               if (need_newpathbuf)
+                       FREE_ZONE(cp, MAXPATHLEN, M_NAMEI);
+               return error;
+       }
+
+       /* 
+        * Safe to set unsigned with a [larger] signed type here
+        * because 0 <= uio_resid <= MAXPATHLEN and MAXPATHLEN
+        * is only 1024.
+        */
+       linklen = MAXPATHLEN - (u_int)uio_resid(auio);
+       if (linklen + ndp->ni_pathlen > MAXPATHLEN) {
+               if (need_newpathbuf)
+                       FREE_ZONE(cp, MAXPATHLEN, M_NAMEI);
+
+               return ENAMETOOLONG;
+       }
+       if (need_newpathbuf) {
+               long len = cnp->cn_pnlen;
+
+               tmppn = cnp->cn_pnbuf;
+               bcopy(ndp->ni_next, cp + linklen, ndp->ni_pathlen);
+               cnp->cn_pnbuf = cp;
+               cnp->cn_pnlen = MAXPATHLEN;
+
+               if ( (cnp->cn_flags & HASBUF) )
+                       FREE_ZONE(tmppn, len, M_NAMEI);
+               else
+                       cnp->cn_flags |= HASBUF;
+       } else
+               cnp->cn_pnbuf[linklen] = '\0';
+
+       ndp->ni_pathlen += linklen;
+       cnp->cn_nameptr = cnp->cn_pnbuf;
+
+       /*
+        * starting point for 'relative'
+        * symbolic link path
+        */
+       dp = ndp->ni_dvp;
+
+       /*
+        * get rid of references returned via 'lookup'
+        */
+       vnode_put(ndp->ni_vp);
+       vnode_put(ndp->ni_dvp); /* ALWAYS have a dvp for a symlink */
+
+       ndp->ni_vp = NULLVP;
+       ndp->ni_dvp = NULLVP;
+
+       /*
+        * Check if symbolic link restarts us at the root
+        */
+       if (*(cnp->cn_nameptr) == '/') {
+               while (*(cnp->cn_nameptr) == '/') {
+                       cnp->cn_nameptr++;
+                       ndp->ni_pathlen--;
+               }
+               if ((dp = ndp->ni_rootdir) == NULLVP) {
+                       return ENOENT;
+               }
+       }
+
+       *new_dp = dp;
+
+       return 0;
+}
+
 /*
  * relookup - lookup a path name component
  *    Used by lookup to re-aquire things.
@@ -1205,18 +1491,27 @@ bad:
        return (error);
 }
 
-/*
- * Free pathname buffer
- */
 void
-nameidone(struct nameidata *ndp)
+namei_unlock_fsnode(struct nameidata *ndp) 
 {
 #ifndef __LP64__
        if ((ndp->ni_cnd.cn_flags & FSNODELOCKHELD)) {
                ndp->ni_cnd.cn_flags &= ~FSNODELOCKHELD;
                unlock_fsnode(ndp->ni_dvp, NULL);
        }       
+#else
+       (void)ndp;
 #endif /* __LP64__ */
+}
+
+/*
+ * Free pathname buffer
+ */
+void
+nameidone(struct nameidata *ndp)
+{
+       namei_unlock_fsnode(ndp);
+
        if (ndp->ni_cnd.cn_flags & HASBUF) {
                char *tmp = ndp->ni_cnd.cn_pnbuf;
 
@@ -1258,6 +1553,7 @@ nameidone(struct nameidata *ndp)
  * fails because /foo_bar_baz is not found will only log "/foo_bar_baz", with
  * no '>' padding.  But /foo_bar/spam would log "/foo_bar>>>>".
  */
+#if !defined(NO_KDEBUG)
 static void
 kdebug_lookup(struct vnode *dp, struct componentname *cnp)
 {
@@ -1305,7 +1601,34 @@ kdebug_lookup(struct vnode *dp, struct componentname *cnp)
                KERNEL_DEBUG_CONSTANT(code, dbg_parms[i], dbg_parms[i+1], dbg_parms[i+2], dbg_parms[i+3], 0);
        }
 }
+#else /* NO_KDEBUG */
+static void
+kdebug_lookup(struct vnode *dp __unused, struct componentname *cnp __unused)
+{
+}
+#endif /* NO_KDEBUG */
+
+int
+vfs_getbyid(fsid_t *fsid, ino64_t ino, vnode_t *vpp, vfs_context_t ctx)
+{
+       mount_t mp;
+       int error;
+       
+       mp = mount_lookupby_volfsid(fsid->val[0], 1);
+       if (mp == NULL) {
+               return EINVAL;
+       }
+
+       /* Get the target vnode. */
+       if (ino == 2) {
+               error = VFS_ROOT(mp, vpp, ctx);
+       } else {
+               error = VFS_VGET(mp, ino, vpp, ctx);
+       }
 
+       vfs_unbusy(mp);
+       return error;
+}
 /*
  * Obtain the real path from a legacy volfs style path.
  *
@@ -1384,3 +1707,59 @@ out:
        return (error);
 }
 #endif
+
+void
+lookup_compound_vnop_post_hook(int error, vnode_t dvp, vnode_t vp, struct nameidata *ndp, int did_create)
+{
+       if (error == 0 && vp == NULLVP) {
+               panic("NULL vp with error == 0.\n");
+       }
+
+       /* 
+        * We don't want to do any of this if we didn't use the compound vnop
+        * to perform the lookup... i.e. if we're allowing and using the legacy pattern,
+        * where we did a full lookup.
+        */
+       if ((ndp->ni_flag & NAMEI_COMPOUND_OP_MASK) == 0) {
+               return;
+       }
+
+       /* 
+        * If we're going to continue the lookup, we'll handle
+        * all lookup-related updates at that time.
+        */
+       if (error == EKEEPLOOKING) {
+               return;
+       }
+
+       /*
+        * Only audit or update cache for *found* vnodes.  For creation
+        * neither would happen in the non-compound-vnop case.
+        */
+       if ((vp != NULLVP) && !did_create) {
+               /* 
+                * If MAKEENTRY isn't set, and we've done a successful compound VNOP, 
+                * then we certainly don't want to update cache or identity.
+                */
+               if ((error != 0) || (ndp->ni_cnd.cn_flags & MAKEENTRY)) {
+                       lookup_consider_update_cache(dvp, vp, &ndp->ni_cnd, ndp->ni_ncgeneration);
+               }
+               if (ndp->ni_cnd.cn_flags & AUDITVNPATH1)
+                       AUDIT_ARG(vnpath, vp, ARG_VNODE1);
+               else if (ndp->ni_cnd.cn_flags & AUDITVNPATH2)
+                       AUDIT_ARG(vnpath, vp, ARG_VNODE2);
+       }
+
+       /* 
+        * If you created (whether you opened or not), cut a lookup tracepoint 
+        * for the parent dir (as would happen without a compound vnop).  Note: we may need
+        * a vnode despite failure in this case!
+        *
+        * If you did not create:
+        *      Found child (succeeded or not): cut a tracepoint for the child.  
+        *      Did not find child: cut a tracepoint with the parent.
+        */
+       if (kdebug_enable) {
+               kdebug_lookup(vp ? vp : dvp, &ndp->ni_cnd); 
+       }
+}