Merge remote-tracking branch 'mfleming/elflink' into for-hpa/elflink/firmware

Conflicts:
	Makefile
	com32/elflink/ldlinux/adv.c
	com32/elflink/ldlinux/kernel.c
	com32/elflink/ldlinux/ldlinux.c
	com32/include/bitsize/stddef.h
	com32/include/bitsize/stdint.h
	com32/include/stdint.h
	com32/include/sys/module.h
	com32/include/sys/x86_64/bitops.h
	com32/include/syslinux/linux.h
	com32/lib/Makefile
	com32/lib/sys/ansicon_write.c
	com32/lib/sys/module/elfutils.h
	com32/lib/sys/vesa/efi/fill.h
	com32/lib/syslinux/load_linux.c
	com32/lib/syslinux/serial.c
	com32/lib/syslinux/shuffle.c
	core/conio.c
	core/elflink/config.c
	core/elflink/load_env32.c
	core/graphics.c
	core/include/graphics.h
	core/init.c
	core/pxelinux.asm
	mk/elf.mk
	mk/lib.mk

17 files changed, 2135 insertions, 213 deletions

diff --git a/core/fs/btrfs/btrfs.c b/core/fs/btrfs/btrfs.c
index 8dedf8ac..16386cc0 100644
--- a/core/fs/btrfs/btrfs.c
+++ b/core/fs/btrfs/btrfs.c
@@ -602,12 +602,15 @@ static void btrfs_get_fs_tree(struct fs_info *fs)
 		do {
 			do {
 				struct btrfs_root_ref *ref;
+				int pathlen;
 
 				if (btrfs_comp_keys_type(&search_key,
 							&path.item.key))
 					break;
 				ref = (struct btrfs_root_ref *)path.data;
-				if (!strcmp((char*)(ref + 1), SubvolName)) {
+				pathlen = path.item.size - sizeof(struct btrfs_root_ref);
+
+				if (!strncmp((char*)(ref + 1), SubvolName, pathlen)) {
 					subvol_ok = true;
 					break;
 				}
diff --git a/core/fs/cache.c b/core/fs/cache.c
index 0d7891be..3b21fc26 100644
--- a/core/fs/cache.c
+++ b/core/fs/cache.c
@@ -37,10 +37,10 @@ void cache_init(struct device *dev, int block_size_shift)
 
     dev->cache_head = head = (struct cache *)
 	(data + (dev->cache_entries << block_size_shift));
-    cache = dev->cache_head + 1; /* First cache descriptor */
+    cache = head + 1;		/* First cache descriptor */
 
     head->prev  = &cache[dev->cache_entries-1];
-    head->next->prev = dev->cache_head;
+    head->prev->next = head;
     head->block = -1;
     head->data  = NULL;
 
diff --git a/core/fs/chdir.c b/core/fs/chdir.c
index 9e8dfd2e..903cabce 100644
--- a/core/fs/chdir.c
+++ b/core/fs/chdir.c
@@ -1,6 +1,7 @@
 #include <stdio.h>
 #include <stdbool.h>
 #include <string.h>
+#include <dprintf.h>
 #include "fs.h"
 #include "cache.h"
 
@@ -16,57 +17,70 @@ void pm_realpath(com32sys_t *regs)
     realpath(dst, src, FILENAME_MAX);
 }
 
-#define EMIT(x)		     		\
-do {		     			\
-    if (++n < bufsize)	 		\
-    	*q++ = (x);			\
-} while (0)
-
-static size_t join_paths(char *dst, size_t bufsize,
-			 const char *s1, const char *s2)
+static size_t copy_string(char *buf, size_t ix, size_t bufsize, const char *src)
 {
-    const char *list[2];
-    int i;
     char c;
-    const char *p;
-    char *q  = dst;
-    size_t n = 0;
-    bool slash = false;
-    
-    list[0] = s1;
-    list[1] = s2;
-
-    for (i = 0; i < 2; i++) {
-	p = list[i];
-
-	while ((c = *p++)) {
-	    if (c == '/') {
-		if (!slash)
-		    EMIT(c);
-		slash = true;
-	    } else {
-		EMIT(c);
-		slash = false;
-	    }
-	}
+
+    while ((c = *src++)) {
+	if (ix+1 < bufsize)
+	    buf[ix] = c;
+	ix++;
     }
 
-    if (bufsize)
-	*q = '\0';
+    if (ix < bufsize)
+	buf[ix] = '\0';
 
-    return n;
+    return ix;
+}
+
+static size_t generic_inode_to_path(struct inode *inode, char *dst, size_t bufsize)
+{
+    size_t s = 0;
+
+    dprintf("inode %p name %s\n", inode, inode->name);
+
+    if (inode->parent) {
+	if (!inode->name)	/* Only the root should have no name */
+	    return -1;
+
+	s = generic_inode_to_path(inode->parent, dst, bufsize);
+	if (s == (size_t)-1)
+	    return s;		/* Error! */
+
+	s = copy_string(dst, s, bufsize, "/");
+	s = copy_string(dst, s, bufsize, inode->name);
+    }
+
+    return s;
 }
 
 size_t realpath(char *dst, const char *src, size_t bufsize)
 {
+    int rv;
+    struct file *file;
+    size_t s;
+
+    dprintf("realpath: input: %s\n", src);
+
     if (this_fs->fs_ops->realpath) {
-	return this_fs->fs_ops->realpath(this_fs, dst, src, bufsize);
+	s = this_fs->fs_ops->realpath(this_fs, dst, src, bufsize);
     } else {
-	/* Filesystems with "common" pathname resolution */
-	return join_paths(dst, bufsize, 
-			  src[0] == '/' ? "" : this_fs->cwd_name,
-			  src);
+	rv = searchdir(src);
+	if (rv < 0) {
+	    dprintf("realpath: searchpath failure\n");
+	    return -1;
+	}
+
+	file = handle_to_file(rv);
+	s = generic_inode_to_path(file->inode, dst, bufsize);
+	if (s == 0)
+	    s = copy_string(dst, 0, bufsize, "/");
+
+	_close_file(file);
     }
+
+    dprintf("realpath: output: %s\n", dst);
+    return s;
 }
 
 int chdir(const char *src)
@@ -74,6 +88,10 @@ int chdir(const char *src)
     int rv;
     struct file *file;
     char cwd_buf[CURRENTDIR_MAX];
+    size_t s;
+
+    dprintf("chdir: from %s (inode %p) add %s\n",
+	    this_fs->cwd_name, this_fs->cwd, src);
 
     if (this_fs->fs_ops->chdir)
 	return this_fs->fs_ops->chdir(this_fs, src);
@@ -94,10 +112,20 @@ int chdir(const char *src)
     _close_file(file);
 
     /* Save the current working directory */
-    realpath(cwd_buf, src, CURRENTDIR_MAX);
+    s = generic_inode_to_path(this_fs->cwd, cwd_buf, CURRENTDIR_MAX-1);
 
     /* Make sure the cwd_name ends in a slash, it's supposed to be a prefix */
-    join_paths(this_fs->cwd_name, CURRENTDIR_MAX, cwd_buf, "/");
+    if (s < 1 || cwd_buf[s-1] != '/')
+	cwd_buf[s++] = '/';
+
+    if (s >= CURRENTDIR_MAX)
+	s = CURRENTDIR_MAX - 1;
+
+    cwd_buf[s++] = '\0';
+    memcpy(this_fs->cwd_name, cwd_buf, s);
+
+    dprintf("chdir: final %s (inode %p)\n",
+	    this_fs->cwd_name, this_fs->cwd);
 
     return 0;
 }
diff --git a/core/fs/fat/fat.c b/core/fs/fat/fat.c
index fbc4386b..b2c20ee0 100644
--- a/core/fs/fat/fat.c
+++ b/core/fs/fat/fat.c
@@ -220,24 +220,30 @@ static sector_t next_sector(struct file *file)
     return sector;
 }
 
-/*
- * Mangle a filename pointed to by src into a buffer pointed to by dst;
- * ends on encountering any whitespace.
+/**
+ * mangle_name:
+ *
+ * Mangle a filename pointed to by src into a buffer pointed
+ * to by dst; ends on encountering any whitespace.
+ * dst is preserved.
+ *
+ * This verifies that a filename is < FILENAME_MAX characters,
+ * doesn't contain whitespace, zero-pads the output buffer,
+ * and removes redundant slashes.
+ *
+ * Unlike the generic version, this also converts backslashes to
+ * forward slashes.
  *
  */
 static void vfat_mangle_name(char *dst, const char *src)
 {
     char *p = dst;
+    int i = FILENAME_MAX-1;
     char c;
-    int i = FILENAME_MAX -1;
 
-    /*
-     * Copy the filename, converting backslash to slash and
-     * collapsing duplicate separators.
-     */
     while (not_whitespace(c = *src)) {
-        if (c == '\\')
-            c = '/';
+	if (c == '\\')
+	    c = '/';
 
         if (c == '/') {
             if (src[1] == '/' || src[1] == '\\') {
@@ -250,16 +256,13 @@ static void vfat_mangle_name(char *dst, const char *src)
         *dst++ = *src++;
     }
 
-    /* Strip terminal slashes or whitespace */
     while (1) {
         if (dst == p)
             break;
-	if (*(dst-1) == '/' && dst-1 == p) /* it's the '/' case */
-		break;
-	if (dst-2 == p && *(dst-2) == '.' && *(dst-1) == '.' )	/* the '..' case */
-		break;
-        if ((*(dst-1) != '/') && (*(dst-1) != '.'))
+        if (dst[-1] != '/')
             break;
+	if ((dst[-1] == '/') && ((dst - 1) == p))
+	    break;
 
         dst--;
         i++;
@@ -779,6 +782,34 @@ static int vfat_fs_init(struct fs_info *fs)
     return fs->block_shift;
 }
 
+static int vfat_copy_superblock(void *buf)
+{
+	struct fat_bpb fat;
+	struct disk *disk;
+	size_t sb_off;
+	void *dst;
+	int sb_len;
+
+	disk = this_fs->fs_dev->disk;
+	disk->rdwr_sectors(disk, &fat, 0, 1, 0);
+
+	/* XXX: Find better sanity checks... */
+	if (!fat.bxResSectors || !fat.bxFATs)
+		return -1;
+
+	sb_off = offsetof(struct fat_bpb, sector_size);
+	sb_len = offsetof(struct fat_bpb, fat12_16) - sb_off \
+		+ sizeof(fat.fat12_16);
+
+	/*
+	 * Only copy fields of the superblock we actually care about.
+	 */
+	dst = buf + sb_off;
+	memcpy(dst, (void *)&fat + sb_off, sb_len);
+
+	return 0;
+}
+
 const struct fs_ops vfat_fs_ops = {
     .fs_name       = "vfat",
     .fs_flags      = FS_USEMEM | FS_THISIND,
@@ -793,4 +824,5 @@ const struct fs_ops vfat_fs_ops = {
     .iget_root     = vfat_iget_root,
     .iget          = vfat_iget,
     .next_extent   = fat_next_extent,
+    .copy_super    = vfat_copy_superblock,
 };
diff --git a/core/fs/fs.c b/core/fs/fs.c
index e6f35370..d8abaa69 100644
--- a/core/fs/fs.c
+++ b/core/fs/fs.c
@@ -2,13 +2,14 @@
 #include <stdio.h>
 #include <stdbool.h>
 #include <string.h>
+#include <unistd.h>
 #include <dprintf.h>
 #include "core.h"
 #include "dev.h"
 #include "fs.h"
 #include "cache.h"
 
-char *PATH = ".:/bin/";
+char *PATH;
 
 /* The currently mounted filesystem */
 struct fs_info *this_fs = NULL;		/* Root filesystem */
@@ -42,6 +43,8 @@ void put_inode(struct inode *inode)
     while (inode && --inode->refcnt == 0) {
 	struct inode *dead = inode;
 	inode = inode->parent;
+	if (dead->name)
+	    free((char *)dead->name);
 	free(dead);
     }
 }
@@ -223,6 +226,9 @@ int searchdir(const char *name)
     char *part, *p, echar;
     int symlink_count = MAX_SYMLINK_CNT;
 
+    dprintf("searchdir: %s  root: %p  cwd: %p\n",
+	    name, this_fs->root, this_fs->cwd);
+
     if (!(file = alloc_file()))
 	goto err_no_close;
     file->fs = this_fs;
@@ -321,6 +327,9 @@ int searchdir(const char *name)
 		    goto got_link;
 		}
 
+		inode->name = strdup(part);
+		dprintf("path component: %s\n", inode->name);
+
 		inode->parent = parent;
 		parent = NULL;
 
@@ -365,6 +374,8 @@ int open_file(const char *name, struct com32_filedata *filedata)
     struct file *file;
     char mangled_name[FILENAME_MAX];
 
+    dprintf("open_file %s\n", name);
+
     mangle_name(mangled_name, name);
     rv = searchdir(mangled_name);
 
@@ -392,6 +403,8 @@ void pm_open_file(com32sys_t *regs)
     const char *name = MK_PTR(regs->es, regs->esi.w[0]);
     char mangled_name[FILENAME_MAX];
 
+    dprintf("pm_open_file %s\n", name);
+
     mangle_name(mangled_name, name);
     rv = searchdir(mangled_name);
     if (rv < 0) {
@@ -479,6 +492,7 @@ void fs_init(const struct fs_ops **ops, struct disk_private *priv)
     if (fs.fs_ops->iget_root) {
 	fs.root = fs.fs_ops->iget_root(&fs);
 	fs.cwd = get_inode(fs.root);
+	dprintf("init: root inode %p, cwd inode %p\n", fs.root, fs.cwd);
     }
 
     if (fs.fs_ops->chdir_start) {
diff --git a/core/fs/iso9660/iso9660.c b/core/fs/iso9660/iso9660.c
index 792cd99f..fe58a5b3 100644
--- a/core/fs/iso9660/iso9660.c
+++ b/core/fs/iso9660/iso9660.c
@@ -243,7 +243,7 @@ static int iso_open_config(struct com32_filedata *filedata)
 	NULL
     };
 
-    return search_config(filedata, search_directories, filenames);
+    return search_dirs(filedata, search_directories, filenames, ConfigName);
 }
 
 static int iso_fs_init(struct fs_info *fs)
diff --git a/core/fs/lib/chdir.c b/core/fs/lib/chdir.c
index 5c7b130f..715284bb 100644
--- a/core/fs/lib/chdir.c
+++ b/core/fs/lib/chdir.c
@@ -1,3 +1,4 @@
+#include <unistd.h>
 #include <core.h>
 
 int generic_chdir_start(void)
diff --git a/core/fs/lib/loadconfig.c b/core/fs/lib/loadconfig.c
index 100500c5..95e6f3f8 100644
--- a/core/fs/lib/loadconfig.c
+++ b/core/fs/lib/loadconfig.c
@@ -30,5 +30,5 @@ int generic_open_config(struct com32_filedata *filedata)
 
     dprintf("CurrentDirName: \"%s\"\n", CurrentDirName);
 
-    return search_config(filedata, search_directories, filenames);
+    return search_dirs(filedata, search_directories, filenames, ConfigName);
 }
diff --git a/core/fs/lib/searchconfig.c b/core/fs/lib/searchconfig.c
index 7fdad183..f2d740b3 100644
--- a/core/fs/lib/searchconfig.c
+++ b/core/fs/lib/searchconfig.c
@@ -5,30 +5,32 @@
 #include <fs.h>
 
 char ConfigName[FILENAME_MAX];
+char config_cwd[FILENAME_MAX];
 
 /*
- * Common implementation of load_config
- *
  * This searches for a specified set of filenames in a specified set
  * of directories.  If found, set the current working directory to
  * match.
  */
-int search_config(struct com32_filedata *filedata,
-		  const char *search_directories[], const char *filenames[])
+int search_dirs(struct com32_filedata *filedata,
+		const char *search_directories[],
+		const char *filenames[],
+		char *realname)
 {
-    char confignamebuf[FILENAME_MAX];
+    char namebuf[FILENAME_MAX];
     const char *sd, **sdp;
     const char *sf, **sfp;
 
     for (sdp = search_directories; (sd = *sdp); sdp++) {
 	for (sfp = filenames; (sf = *sfp); sfp++) {
-	    snprintf(confignamebuf, sizeof confignamebuf,
+	    snprintf(namebuf, sizeof namebuf,
 		     "%s%s%s",
 		     sd, (*sd && sd[strlen(sd)-1] == '/') ? "" : "/",
 		     sf);
-	    realpath(ConfigName, confignamebuf, FILENAME_MAX);
-	    dprintf("Config search: %s\n", ConfigName);
-	    if (open_file(ConfigName, filedata) >= 0) {
+	    if (realpath(realname, namebuf, FILENAME_MAX) == (size_t)-1)
+		continue;
+	    dprintf("Config search: %s\n", realname);
+	    if (open_file(realname, filedata) >= 0) {
 		chdir(sd);
 		return 0;	/* Got it */
 	    }
diff --git a/core/fs/loadhigh.c b/core/fs/loadhigh.c
deleted file mode 100644
index bd9d3535..00000000
--- a/core/fs/loadhigh.c
+++ /dev/null
@@ -1,112 +0,0 @@
-/*
- * -----------------------------------------------------------------------
- *
- *   Copyright 1994-2009 H. Peter Anvin - All Rights Reserved
- *   Copyright 2009-2010 Intel Corporation; author: H. Peter Anvin
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Boston MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * loadhigh.c
- *
- * An alternate interface to getfssec.
- *
- * Inputs:	SI  = file handle/cluster pointer
- *		EDI = target address in high memory
- *		EAX = maximum number of bytes to load
- *		DX  = zero-padding mask (e.g. 0003h for pad to dword)
- *		BX  = 16-bit subroutine to call at the top of each loop
- *                    (to print status and check for abort)
- *		EBP = maximum load address
- *
- * Outputs:	SI  = file handle/cluster pointer
- *		EBX = first untouched address (not including padding)
- *		EDI = first untouched address (including padding)
- *		CF  = reached EOF
- *		OF  = ran out of high memory
- */
-
-#include <com32.h>
-#include <minmax.h>
-#include "core.h"
-#include "fs.h"
-
-#define MAX_CHUNK	(1 << 20) /* 1 MB */
-
-void pm_load_high(com32sys_t *regs)
-{
-    struct fs_info *fs;
-    uint32_t bytes;
-    uint32_t zero_mask;
-    bool have_more;
-    uint32_t bytes_read;
-    char *buf, *limit;
-    struct file *file;
-    uint32_t sector_mask;
-    size_t pad;
-    uint32_t retflags = 0;
-
-    bytes     = regs->eax.l;
-    zero_mask = regs->edx.w[0];
-    buf       = (char *)regs->edi.l;
-    limit     = (char *)(regs->ebp.l & ~zero_mask);
-    file      = handle_to_file(regs->esi.w[0]);
-    fs        = file->fs;
-
-    sector_mask = SECTOR_SIZE(fs) - 1;
-
-    while (bytes) {
-	uint32_t sectors;
-	uint32_t chunk;
-
-	if (buf + SECTOR_SIZE(fs) > limit) {
-	    /* Can't fit even one more sector in... */
-	    retflags = EFLAGS_OF;
-	    break;
-	}
-
-	chunk = bytes;
-
-	if (regs->ebx.w[0]) {
-	    call16((void (*)(void))(size_t)regs->ebx.w[0], &zero_regs, NULL);
-	    chunk = min(chunk, MAX_CHUNK);
-	}
-
-	if (chunk > (((char *)limit - buf) & ~sector_mask))
-	    chunk = ((char *)limit - buf) & ~sector_mask;
-
-	sectors = (chunk + sector_mask) >> SECTOR_SHIFT(fs);
-	bytes_read = fs->fs_ops->getfssec(file, buf, sectors, &have_more);
-
-	if (bytes_read > chunk)
-	    bytes_read = chunk;
-
-	buf += bytes_read;
-	bytes -= bytes_read;
-
-	if (!have_more) {
-	    /*
-	     * If we reach EOF, the filesystem driver will have already closed
-	     * the underlying file... this really should be cleaner.
-	     */
-	    _close_file(file);
-	    regs->esi.w[0] = 0;
-	    retflags = EFLAGS_CF;
-	    break;
-	}
-    }
-
-    pad = (size_t)buf & zero_mask;
-    if (pad)
-	memset(buf, 0, pad);
-
-    regs->ebx.l = (size_t)buf;
-    regs->edi.l = (size_t)buf + pad;
-    set_flags(regs, retflags);
-}
diff --git a/core/fs/ntfs/ntfs.c b/core/fs/ntfs/ntfs.c
new file mode 100644
index 00000000..f54df7e5
--- /dev/null
+++ b/core/fs/ntfs/ntfs.c
@@ -0,0 +1,1388 @@
+/*
+ * Copyright (C) 2011-2012 Paulo Alcantara <pcacjr@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the
+ * Free Software Foundation, Inc.,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+*/
+
+/* Note: No support for compressed files */
+
+#include <dprintf.h>
+#include <stdio.h>
+#include <string.h>
+#include <sys/dirent.h>
+#include <cache.h>
+#include <core.h>
+#include <disk.h>
+#include <fs.h>
+#include <ilog2.h>
+#include <klibc/compiler.h>
+#include <ctype.h>
+
+#include "codepage.h"
+#include "ntfs.h"
+#include "runlist.h"
+
+static struct ntfs_readdir_state *readdir_state;
+
+/*** Function declarations */
+static f_mft_record_lookup ntfs_mft_record_lookup_3_0;
+static f_mft_record_lookup ntfs_mft_record_lookup_3_1;
+
+/*** Function definitions */
+
+/* Check if there are specific zero fields in an NTFS boot sector */
+static inline int ntfs_check_zero_fields(const struct ntfs_bpb *sb)
+{
+    return !sb->res_sectors && (!sb->zero_0[0] && !sb->zero_0[1] &&
+            !sb->zero_0[2]) && !sb->zero_1 && !sb->zero_2 &&
+            !sb->zero_3;
+}
+
+static inline int ntfs_check_sb_fields(const struct ntfs_bpb *sb)
+{
+    return ntfs_check_zero_fields(sb) &&
+            (!memcmp(sb->oem_name, "NTFS    ", 8) ||
+             !memcmp(sb->oem_name, "MSWIN4.0", 8) ||
+             !memcmp(sb->oem_name, "MSWIN4.1", 8));
+}
+
+static inline struct inode *new_ntfs_inode(struct fs_info *fs)
+{
+    struct inode *inode;
+
+    inode = alloc_inode(fs, 0, sizeof(struct ntfs_inode));
+    if (!inode)
+        malloc_error("inode structure");
+
+    return inode;
+}
+
+static void ntfs_fixups_writeback(struct fs_info *fs, struct ntfs_record *nrec)
+{
+    uint16_t *usa;
+    uint16_t usa_no;
+    uint16_t usa_count;
+    uint16_t *blk;
+
+    dprintf("in %s()\n", __func__);
+
+    if (nrec->magic != NTFS_MAGIC_FILE && nrec->magic != NTFS_MAGIC_INDX)
+        return;
+
+    /* get the Update Sequence Array offset */
+    usa = (uint16_t *)((uint8_t *)nrec + nrec->usa_ofs);
+    /* get the Update Sequence Array Number and skip it */
+    usa_no = *usa++;
+    /* get the Update Sequene Array count */
+    usa_count = nrec->usa_count - 1;    /* exclude the USA number */
+    /* make it to point to the last two bytes of the RECORD's first sector */
+    blk = (uint16_t *)((uint8_t *)nrec + SECTOR_SIZE(fs) - 2);
+
+    while (usa_count--) {
+        if (*blk != usa_no)
+            break;
+
+        *blk = *usa++;
+        blk = (uint16_t *)((uint8_t *)blk + SECTOR_SIZE(fs));
+    }
+}
+
+/* read content from cache */
+static int ntfs_read(struct fs_info *fs, void *buf, size_t len, uint64_t count,
+                    block_t *blk, uint64_t *blk_offset,
+                    uint64_t *blk_next_offset, uint64_t *lcn)
+{
+    uint8_t *data;
+    uint64_t offset = *blk_offset;
+    const uint32_t clust_byte_shift = NTFS_SB(fs)->clust_byte_shift;
+    const uint64_t blk_size = UINT64_C(1) << BLOCK_SHIFT(fs);
+    uint64_t bytes;
+    uint64_t lbytes;
+    uint64_t loffset;
+    uint64_t k;
+
+    dprintf("in %s()\n", __func__);
+
+    if (count > len)
+        goto out;
+
+    data = (uint8_t *)get_cache(fs->fs_dev, *blk);
+    if (!data)
+        goto out;
+
+    if (!offset)
+        offset = (*lcn << clust_byte_shift) % blk_size;
+
+    dprintf("LCN:            0x%X\n", *lcn);
+    dprintf("offset:         0x%X\n", offset);
+
+    bytes = count;              /* bytes to copy */
+    lbytes = blk_size - offset; /* bytes left to copy */
+    if (lbytes >= bytes) {
+        /* so there's room enough, then copy the whole content */
+        memcpy(buf, data + offset, bytes);
+        loffset = offset;
+        offset += count;
+    } else {
+        dprintf("bytes:             %u\n", bytes);
+        dprintf("bytes left:        %u\n", lbytes);
+        /* otherwise, let's copy it partially... */
+        k = 0;
+        while (bytes) {
+            memcpy(buf + k, data + offset, lbytes);
+            bytes -= lbytes;
+            loffset = offset;
+            offset += lbytes;
+            k += lbytes;
+            if (offset >= blk_size) {
+                /* then fetch a new FS block */
+                data = (uint8_t *)get_cache(fs->fs_dev, ++*blk);
+                if (!data)
+                    goto out;
+
+                lbytes = bytes;
+                loffset = offset;
+                offset = 0;
+            }
+        }
+    }
+
+    if (loffset >= blk_size)
+        loffset = 0;    /* it must be aligned on a block boundary */
+
+    *blk_offset = loffset;
+
+    if (blk_next_offset)
+        *blk_next_offset = offset;
+
+    *lcn += blk_size / count;   /* update LCN */
+
+    return 0;
+
+out:
+    return -1;
+}
+
+static struct ntfs_mft_record *ntfs_mft_record_lookup_3_0(struct fs_info *fs,
+                                                uint32_t file, block_t *blk)
+{
+    const uint64_t mft_record_size = NTFS_SB(fs)->mft_record_size;
+    uint8_t *buf;
+    const block_t mft_blk = NTFS_SB(fs)->mft_blk;
+    block_t cur_blk;
+    block_t right_blk;
+    uint64_t offset;
+    uint64_t next_offset;
+    const uint32_t mft_record_shift = ilog2(mft_record_size);
+    const uint32_t clust_byte_shift = NTFS_SB(fs)->clust_byte_shift;
+    uint64_t lcn;
+    int err;
+    struct ntfs_mft_record *mrec;
+
+    dprintf("in %s()\n", __func__);
+
+    buf = (uint8_t *)malloc(mft_record_size);
+    if (!buf)
+        malloc_error("uint8_t *");
+
+    /* determine MFT record's LCN and block number */
+    lcn = NTFS_SB(fs)->mft_lcn + (file << mft_record_shift >> clust_byte_shift);
+    cur_blk = (lcn << clust_byte_shift >> BLOCK_SHIFT(fs)) - mft_blk;
+    offset = (file << mft_record_shift) % BLOCK_SIZE(fs);
+    for (;;) {
+        right_blk = cur_blk + mft_blk;
+        err = ntfs_read(fs, buf, mft_record_size, mft_record_size, &right_blk,
+                        &offset, &next_offset, &lcn);
+        if (err) {
+            printf("Error while reading from cache.\n");
+            break;
+        }
+
+        ntfs_fixups_writeback(fs, (struct ntfs_record *)buf);
+
+        mrec = (struct ntfs_mft_record *)buf;
+        /* check if it has a valid magic number */
+        if (mrec->magic == NTFS_MAGIC_FILE) {
+            if (blk)
+                *blk = cur_blk;     /* update record starting block */
+
+            return mrec;            /* found MFT record */
+        }
+
+        if (next_offset >= BLOCK_SIZE(fs)) {
+            /* try the next FS block */
+            offset = 0;
+            cur_blk = right_blk - mft_blk + 1;
+        } else {
+            /* there's still content to fetch in the current block */
+            cur_blk = right_blk - mft_blk;
+            offset = next_offset;   /* update FS block offset */
+        }
+    }
+
+    free(buf);
+
+    return NULL;
+}
+
+static struct ntfs_mft_record *ntfs_mft_record_lookup_3_1(struct fs_info *fs,
+                                                uint32_t file, block_t *blk)
+{
+    const uint64_t mft_record_size = NTFS_SB(fs)->mft_record_size;
+    uint8_t *buf;
+    const block_t mft_blk = NTFS_SB(fs)->mft_blk;
+    block_t cur_blk;
+    block_t right_blk;
+    uint64_t offset;
+    uint64_t next_offset;
+    const uint32_t mft_record_shift = ilog2(mft_record_size);
+    const uint32_t clust_byte_shift = NTFS_SB(fs)->clust_byte_shift;
+    uint64_t lcn;
+    int err;
+    struct ntfs_mft_record *mrec;
+
+    dprintf("in %s()\n", __func__);
+
+    buf = (uint8_t *)malloc(mft_record_size);
+    if (!buf)
+        malloc_error("uint8_t *");
+
+    lcn = NTFS_SB(fs)->mft_lcn + (file << mft_record_shift >> clust_byte_shift);
+    cur_blk = (lcn << clust_byte_shift >> BLOCK_SHIFT(fs)) - mft_blk;
+    offset = (file << mft_record_shift) % BLOCK_SIZE(fs);
+    for (;;) {
+        right_blk = cur_blk + NTFS_SB(fs)->mft_blk;
+        err = ntfs_read(fs, buf, mft_record_size, mft_record_size, &right_blk,
+                        &offset, &next_offset, &lcn);
+        if (err) {
+            printf("Error while reading from cache.\n");
+            break;
+        }
+
+        ntfs_fixups_writeback(fs, (struct ntfs_record *)buf);
+
+        mrec = (struct ntfs_mft_record *)buf;
+        /* Check if the NTFS 3.1 MFT record number matches */
+        if (mrec->magic == NTFS_MAGIC_FILE && mrec->mft_record_no == file) {
+            if (blk)
+                *blk = cur_blk;     /* update record starting block */
+
+            return mrec;            /* found MFT record */
+        }
+
+        if (next_offset >= BLOCK_SIZE(fs)) {
+            /* try the next FS block */
+            offset = 0;
+            cur_blk = right_blk - NTFS_SB(fs)->mft_blk + 1;
+        } else {
+            /* there's still content to fetch in the current block */
+            cur_blk = right_blk - NTFS_SB(fs)->mft_blk;
+            offset = next_offset;   /* update FS block offset */
+        }
+    }
+
+    free(buf);
+
+    return NULL;
+}
+
+static bool ntfs_filename_cmp(const char *dname, struct ntfs_idx_entry *ie)
+{
+    const uint16_t *entry_fn;
+    uint8_t entry_fn_len;
+    unsigned i;
+
+    dprintf("in %s()\n", __func__);
+
+    entry_fn = ie->key.file_name.file_name;
+    entry_fn_len = ie->key.file_name.file_name_len;
+
+    if (strlen(dname) != entry_fn_len)
+        return false;
+
+    /* Do case-sensitive compares for Posix file names */
+    if (ie->key.file_name.file_name_type == FILE_NAME_POSIX) {
+        for (i = 0; i < entry_fn_len; i++)
+            if (entry_fn[i] != dname[i])
+                return false;
+    } else {
+        for (i = 0; i < entry_fn_len; i++)
+            if (tolower(entry_fn[i]) != tolower(dname[i]))
+                return false;
+    }
+
+    return true;
+}
+
+static inline uint8_t *mapping_chunk_init(struct ntfs_attr_record *attr,
+                                        struct mapping_chunk *chunk,
+                                        uint32_t *offset)
+{
+    memset(chunk, 0, sizeof *chunk);
+    *offset = 0U;
+
+    return (uint8_t *)attr + attr->data.non_resident.mapping_pairs_offset;
+}
+
+/* Parse data runs.
+ *
+ * return 0 on success or -1 on failure.
+ */
+static int parse_data_run(const void *stream, uint32_t *offset,
+                            uint8_t *attr_len, struct mapping_chunk *chunk)
+{
+    uint8_t *buf;   /* Pointer to the zero-terminated byte stream */
+    uint8_t count;  /* The count byte */
+    uint8_t v, l;   /* v is the number of changed low-order VCN bytes;
+                     * l is the number of changed low-order LCN bytes
+                     */
+    uint8_t *byte;
+    int byte_shift = 8;
+    int mask;
+    uint8_t val;
+    int64_t res;
+
+    (void)attr_len;
+
+    dprintf("in %s()\n", __func__);
+
+    chunk->flags &= ~MAP_MASK;
+
+    buf = (uint8_t *)stream + *offset;
+    if (buf > attr_len || !*buf) {
+        chunk->flags |= MAP_END;    /* we're done */
+        return 0;
+    }
+
+    if (!*offset)
+        chunk->flags |= MAP_START;  /* initial chunk */
+
+    count = *buf;
+    v = count & 0x0F;
+    l = count >> 4;
+
+    if (v > 8 || l > 8) /* more than 8 bytes ? */
+        goto out;
+
+    byte = (uint8_t *)buf + v;
+    count = v;
+
+    res = 0LL;
+    while (count--) {
+        val = *byte--;
+        mask = val >> (byte_shift - 1);
+        res = (res << byte_shift) | ((val + mask) ^ mask);
+    }
+
+    chunk->len = res;   /* get length data */
+
+    byte = (uint8_t *)buf + v + l;
+    count = l;
+
+    mask = 0xFFFFFFFF;
+    res = 0LL;
+    if (*byte & 0x80)
+        res |= (int64_t)mask;   /* sign-extend it */
+
+    while (count--)
+        res = (res << byte_shift) | *byte--;
+
+    chunk->lcn += res;
+    /* are VCNS from cur_vcn to next_vcn - 1 unallocated ? */
+    if (!chunk->lcn)
+        chunk->flags |= MAP_UNALLOCATED;
+    else
+        chunk->flags |= MAP_ALLOCATED;
+
+    *offset += v + l + 1;
+
+    return 0;
+
+out:
+    return -1;
+}
+
+static struct ntfs_mft_record *
+ntfs_attr_list_lookup(struct fs_info *fs, struct ntfs_attr_record *attr,
+                      uint32_t type, struct ntfs_mft_record *mrec)
+{
+    uint8_t *attr_len;
+    struct mapping_chunk chunk;
+    uint32_t offset;
+    uint8_t *stream;
+    int err;
+    const uint64_t blk_size = UINT64_C(1) << BLOCK_SHIFT(fs);
+    uint8_t buf[blk_size];
+    uint64_t blk_offset;
+    int64_t vcn;
+    int64_t lcn;
+    int64_t last_lcn;
+    block_t blk;
+    struct ntfs_attr_list_entry *attr_entry;
+    uint32_t len = 0;
+    struct ntfs_mft_record *retval;
+    uint64_t start_blk = 0;
+
+    dprintf("in %s()\n", __func__);
+
+    if (attr->non_resident)
+        goto handle_non_resident_attr;
+
+    attr_entry = (struct ntfs_attr_list_entry *)
+        ((uint8_t *)attr + attr->data.resident.value_offset);
+    len = attr->data.resident.value_len;
+    for (; (uint8_t *)attr_entry < (uint8_t *)attr + len;
+         attr_entry = (struct ntfs_attr_list_entry *)((uint8_t *)attr_entry +
+                                                      attr_entry->length)) {
+        dprintf("<$ATTRIBUTE_LIST> Attribute type: 0x%X\n",
+                attr_entry->type);
+        if (attr_entry->type == type)
+            goto found; /* We got the attribute! :-) */
+    }
+
+    printf("No attribute found.\n");
+    goto out;
+
+handle_non_resident_attr:
+    attr_len = (uint8_t *)attr + attr->len;
+    stream = mapping_chunk_init(attr, &chunk, &offset);
+    do {
+        err = parse_data_run(stream, &offset, attr_len, &chunk);
+        if (err) {
+            printf("parse_data_run()\n");
+            goto out;
+        }
+
+        if (chunk.flags & MAP_UNALLOCATED)
+            continue;
+        if (chunk.flags & MAP_END)
+            break;
+        if (chunk.flags & MAP_ALLOCATED) {
+            vcn = 0;
+            lcn = chunk.lcn;
+            while (vcn < chunk.len) {
+                blk = (lcn + vcn) << NTFS_SB(fs)->clust_byte_shift >>
+                    BLOCK_SHIFT(fs);
+                blk_offset = 0;
+                last_lcn = lcn;
+                lcn += vcn;
+                err = ntfs_read(fs, buf, blk_size, blk_size, &blk,
+                                &blk_offset, NULL, (uint64_t *)&lcn);
+                if (err) {
+                    printf("Error while reading from cache.\n");
+                    goto out;
+                }
+
+                attr_entry = (struct ntfs_attr_list_entry *)&buf;
+                len = attr->data.non_resident.data_size;
+                for (; (uint8_t *)attr_entry < (uint8_t *)&buf[0] + len;
+                     attr_entry = (struct ntfs_attr_list_entry *)
+                         ((uint8_t *)attr_entry + attr_entry->length)) {
+                    dprintf("<$ATTRIBUTE_LIST> Attribute type: 0x%x\n",
+                            attr_entry->type);
+                    if (attr_entry->type == type)
+                        goto found; /* We got the attribute! :-) */
+                }
+
+                lcn = last_lcn; /* restore original LCN */
+                /* go to the next VCN */
+                vcn += (blk_size / (1 << NTFS_SB(fs)->clust_byte_shift));
+            }
+        }
+    } while (!(chunk.flags & MAP_END));
+
+    printf("No attribute found.\n");
+
+out:
+    return NULL;
+
+found:
+    /* At this point we have the attribute we were looking for. Now we
+     * will look for the MFT record that stores information about this
+     * attribute.
+     */
+
+    /* Check if the attribute type we're looking for is in the same
+     * MFT record. If so, we do not need to look it up again - return it.
+     */
+    if (mrec->mft_record_no == attr_entry->mft_ref)
+        return mrec;
+
+    retval = NTFS_SB(fs)->mft_record_lookup(fs, attr_entry->mft_ref,
+                                            &start_blk);
+    if (!retval) {
+        printf("No MFT record found!\n");
+        goto out;
+    }
+
+    /* return the found MFT record */
+    return retval;
+}
+
+static struct ntfs_attr_record *
+__ntfs_attr_lookup(struct fs_info *fs, uint32_t type,
+                   struct ntfs_mft_record **mrec)
+{
+    struct ntfs_mft_record *_mrec = *mrec;
+    struct ntfs_attr_record *attr;
+    struct ntfs_attr_record *attr_list_attr;
+
+    dprintf("in %s()\n", __func__);
+
+    if (!_mrec || type == NTFS_AT_END)
+        goto out;
+
+again:
+    attr_list_attr = NULL;
+
+    attr = (struct ntfs_attr_record *)((uint8_t *)_mrec + _mrec->attrs_offset);
+    /* walk through the file attribute records */
+    for (;; attr = (struct ntfs_attr_record *)((uint8_t *)attr + attr->len)) {
+        if (attr->type == NTFS_AT_END)
+            break;
+
+        if (attr->type == NTFS_AT_ATTR_LIST) {
+            dprintf("MFT record #%lu has an $ATTRIBUTE_LIST attribute.\n",
+                    _mrec->mft_record_no);
+            attr_list_attr = attr;
+            continue;
+        }
+
+        if (attr->type == type)
+            break;
+    }
+
+    /* if the record has an $ATTRIBUTE_LIST attribute associated
+     * with it, then we need to look for the wanted attribute in
+     * it as well.
+     */
+    if (attr->type == NTFS_AT_END && attr_list_attr) {
+        struct ntfs_mft_record *retval;
+
+        retval = ntfs_attr_list_lookup(fs, attr_list_attr, type, _mrec);
+        if (!retval)
+            goto out;
+
+        _mrec = retval;
+        goto again;
+    } else if (attr->type == NTFS_AT_END && !attr_list_attr) {
+        attr = NULL;
+    }
+
+    return attr;
+
+out:
+    return NULL;
+}
+
+static inline struct ntfs_attr_record *
+ntfs_attr_lookup(struct fs_info *fs, uint32_t type,
+                 struct ntfs_mft_record **mmrec,
+                 struct ntfs_mft_record *mrec)
+{
+    struct ntfs_mft_record *_mrec = mrec;
+    struct ntfs_mft_record *other = *mmrec;
+    struct ntfs_attr_record *retval = NULL;
+
+    if (mrec == other)
+        return __ntfs_attr_lookup(fs, type, &other);
+
+    retval = __ntfs_attr_lookup(fs, type, &_mrec);
+    if (!retval) {
+        _mrec = other;
+        retval = __ntfs_attr_lookup(fs, type, &other);
+        if (!retval)
+            other = _mrec;
+    } else if (retval && (_mrec != mrec)) {
+        other = _mrec;
+    }
+
+    return retval;
+}
+
+static inline enum dirent_type get_inode_mode(struct ntfs_mft_record *mrec)
+{
+    return mrec->flags & MFT_RECORD_IS_DIRECTORY ? DT_DIR : DT_REG;
+}
+
+static int index_inode_setup(struct fs_info *fs, unsigned long mft_no,
+                            struct inode *inode)
+{
+    uint64_t start_blk = 0;
+    struct ntfs_mft_record *mrec, *lmrec;
+    struct ntfs_attr_record *attr;
+    enum dirent_type d_type;
+    uint8_t *attr_len;
+    struct mapping_chunk chunk;
+    int err;
+    uint8_t *stream;
+    uint32_t offset;
+
+    dprintf("in %s()\n", __func__);
+
+    mrec = NTFS_SB(fs)->mft_record_lookup(fs, mft_no, &start_blk);
+    if (!mrec) {
+        printf("No MFT record found.\n");
+        goto out;
+    }
+
+    lmrec = mrec;
+
+    NTFS_PVT(inode)->mft_no = mft_no;
+    NTFS_PVT(inode)->seq_no = mrec->seq_no;
+
+    NTFS_PVT(inode)->start_cluster = start_blk >> NTFS_SB(fs)->clust_shift;
+    NTFS_PVT(inode)->here = start_blk;
+
+    d_type = get_inode_mode(mrec);
+    if (d_type == DT_DIR) {    /* directory stuff */
+        dprintf("Got a directory.\n");
+        attr = ntfs_attr_lookup(fs, NTFS_AT_INDEX_ROOT, &mrec, lmrec);
+        if (!attr) {
+            printf("No attribute found.\n");
+            goto out;
+        }
+
+        /* check if we have a previous allocated state structure */
+        if (readdir_state) {
+            free(readdir_state);
+            readdir_state = NULL;
+        }
+
+        /* allocate our state structure */
+        readdir_state = malloc(sizeof *readdir_state);
+        if (!readdir_state)
+            malloc_error("ntfs_readdir_state structure");
+
+        readdir_state->mft_no = mft_no;
+        /* obviously, the ntfs_readdir() caller will start from INDEX root */
+        readdir_state->in_idx_root = true;
+    } else if (d_type == DT_REG) {        /* file stuff */
+        dprintf("Got a file.\n");
+        attr = ntfs_attr_lookup(fs, NTFS_AT_DATA, &mrec, lmrec);
+        if (!attr) {
+            printf("No attribute found.\n");
+            goto out;
+        }
+
+        NTFS_PVT(inode)->non_resident = attr->non_resident;
+        NTFS_PVT(inode)->type = attr->type;
+
+        if (!attr->non_resident) {
+            NTFS_PVT(inode)->data.resident.offset =
+                (uint32_t)((uint8_t *)attr + attr->data.resident.value_offset);
+            inode->size = attr->data.resident.value_len;
+        } else {
+            attr_len = (uint8_t *)attr + attr->len;
+
+            stream = mapping_chunk_init(attr, &chunk, &offset);
+            NTFS_PVT(inode)->data.non_resident.rlist = NULL;
+            for (;;) {
+                err = parse_data_run(stream, &offset, attr_len, &chunk);
+                if (err) {
+                    printf("parse_data_run()\n");
+                    goto out;
+                }
+
+                if (chunk.flags & MAP_UNALLOCATED)
+                    continue;
+                if (chunk.flags & MAP_END)
+                    break;
+                if (chunk.flags &  MAP_ALLOCATED) {
+                    /* append new run to the runlist */
+                    runlist_append(&NTFS_PVT(inode)->data.non_resident.rlist,
+                                    (struct runlist_element *)&chunk);
+                    /* update for next VCN */
+                    chunk.vcn += chunk.len;
+                }
+            }
+
+            if (runlist_is_empty(NTFS_PVT(inode)->data.non_resident.rlist)) {
+                printf("No mapping found\n");
+                goto out;
+            }
+
+            inode->size = attr->data.non_resident.initialized_size;
+        }
+    }
+
+    inode->mode = d_type;
+
+    free(mrec);
+
+    return 0;
+
+out:
+    free(mrec);
+
+    return -1;
+}
+
+static struct inode *ntfs_index_lookup(const char *dname, struct inode *dir)
+{
+    struct fs_info *fs = dir->fs;
+    struct ntfs_mft_record *mrec, *lmrec;
+    block_t blk;
+    uint64_t blk_offset;
+    struct ntfs_attr_record *attr;
+    struct ntfs_idx_root *ir;
+    struct ntfs_idx_entry *ie;
+    const uint64_t blk_size = UINT64_C(1) << BLOCK_SHIFT(fs);
+    uint8_t buf[blk_size];
+    struct ntfs_idx_allocation *iblk;
+    int err;
+    uint8_t *stream;
+    uint8_t *attr_len;
+    struct mapping_chunk chunk;
+    uint32_t offset;
+    int64_t vcn;
+    int64_t lcn;
+    int64_t last_lcn;
+    struct inode *inode;
+
+    dprintf("in %s()\n", __func__);
+
+    mrec = NTFS_SB(fs)->mft_record_lookup(fs, NTFS_PVT(dir)->mft_no, NULL);
+    if (!mrec) {
+        printf("No MFT record found.\n");
+        goto out;
+    }
+
+    lmrec = mrec;
+    attr = ntfs_attr_lookup(fs, NTFS_AT_INDEX_ROOT, &mrec, lmrec);
+    if (!attr) {
+        printf("No attribute found.\n");
+        goto out;
+    }
+
+    ir = (struct ntfs_idx_root *)((uint8_t *)attr +
+                            attr->data.resident.value_offset);
+    ie = (struct ntfs_idx_entry *)((uint8_t *)&ir->index +
+                                ir->index.entries_offset);
+    for (;; ie = (struct ntfs_idx_entry *)((uint8_t *)ie + ie->len)) {
+        /* bounds checks */
+        if ((uint8_t *)ie < (uint8_t *)mrec ||
+            (uint8_t *)ie + sizeof(struct ntfs_idx_entry_header) >
+            (uint8_t *)&ir->index + ir->index.index_len ||
+            (uint8_t *)ie + ie->len >
+            (uint8_t *)&ir->index + ir->index.index_len)
+            goto index_err;
+
+        /* last entry cannot contain a key. it can however contain
+         * a pointer to a child node in the B+ tree so we just break out
+         */
+        if (ie->flags & INDEX_ENTRY_END)
+            break;
+
+        if (ntfs_filename_cmp(dname, ie))
+            goto found;
+    }
+
+    /* check for the presence of a child node */
+    if (!(ie->flags & INDEX_ENTRY_NODE)) {
+        printf("No child node, aborting...\n");
+        goto out;
+    }
+
+    /* then descend into child node */
+
+    attr = ntfs_attr_lookup(fs, NTFS_AT_INDEX_ALLOCATION, &mrec, lmrec);
+    if (!attr) {
+        printf("No attribute found.\n");
+        goto out;
+    }
+
+    if (!attr->non_resident) {
+        printf("WTF ?! $INDEX_ALLOCATION isn't really resident.\n");
+        goto out;
+    }
+
+    attr_len = (uint8_t *)attr + attr->len;
+    stream = mapping_chunk_init(attr, &chunk, &offset);
+    do {
+        err = parse_data_run(stream, &offset, attr_len, &chunk);
+        if (err)
+            break;
+
+        if (chunk.flags & MAP_UNALLOCATED)
+            continue;
+
+        if (chunk.flags & MAP_ALLOCATED) {
+            dprintf("%d cluster(s) starting at 0x%08llX\n", chunk.len,
+                    chunk.lcn);
+
+            vcn = 0;
+            lcn = chunk.lcn;
+            while (vcn < chunk.len) {
+                blk = (lcn + vcn) << NTFS_SB(fs)->clust_shift <<
+                    SECTOR_SHIFT(fs) >> BLOCK_SHIFT(fs);
+
+                blk_offset = 0;
+                last_lcn = lcn;
+                lcn += vcn;
+                err = ntfs_read(fs, &buf, blk_size, blk_size, &blk,
+                                &blk_offset, NULL, (uint64_t *)&lcn);
+                if (err) {
+                    printf("Error while reading from cache.\n");
+                    goto not_found;
+                }
+
+                ntfs_fixups_writeback(fs, (struct ntfs_record *)&buf);
+
+                iblk = (struct ntfs_idx_allocation *)&buf;
+                if (iblk->magic != NTFS_MAGIC_INDX) {
+                    printf("Not a valid INDX record.\n");
+                    goto not_found;
+                }
+
+                ie = (struct ntfs_idx_entry *)((uint8_t *)&iblk->index +
+                                            iblk->index.entries_offset);
+                for (;; ie = (struct ntfs_idx_entry *)((uint8_t *)ie +
+                        ie->len)) {
+                    /* bounds checks */
+                    if ((uint8_t *)ie < (uint8_t *)iblk || (uint8_t *)ie +
+                        sizeof(struct ntfs_idx_entry_header) >
+                        (uint8_t *)&iblk->index + iblk->index.index_len ||
+                        (uint8_t *)ie + ie->len >
+                        (uint8_t *)&iblk->index + iblk->index.index_len)
+                        goto index_err;
+
+                    /* last entry cannot contain a key */
+                    if (ie->flags & INDEX_ENTRY_END)
+                        break;
+
+                    if (ntfs_filename_cmp(dname, ie))
+                        goto found;
+                }
+
+                lcn = last_lcn; /* restore the original LCN */
+                /* go to the next VCN */
+                vcn += (blk_size / (1 << NTFS_SB(fs)->clust_byte_shift));
+            }
+        }
+    } while (!(chunk.flags & MAP_END));
+
+not_found:
+    dprintf("Index not found\n");
+
+out:
+    free(mrec);
+
+    return NULL;
+
+found:
+    dprintf("Index found\n");
+    inode = new_ntfs_inode(fs);
+    err = index_inode_setup(fs, ie->data.dir.indexed_file, inode);
+    if (err) {
+        printf("Error in index_inode_setup()\n");
+        free(inode);
+        goto out;
+    }
+
+    free(mrec);
+
+    return inode;
+
+index_err:
+    printf("Corrupt index. Aborting lookup...\n");
+    goto out;
+}
+
+/* Convert an UTF-16LE LFN to OEM LFN */
+static uint8_t ntfs_cvt_filename(char *filename,
+                                const struct ntfs_idx_entry *ie)
+{
+    const uint16_t *entry_fn;
+    uint8_t entry_fn_len;
+    unsigned i;
+
+    entry_fn = ie->key.file_name.file_name;
+    entry_fn_len = ie->key.file_name.file_name_len;
+
+    for (i = 0; i < entry_fn_len; i++)
+        filename[i] = (char)entry_fn[i];
+
+    filename[i] = '\0';
+
+    return entry_fn_len;
+}
+
+static int ntfs_next_extent(struct inode *inode, uint32_t lstart)
+{
+    struct fs_info *fs = inode->fs;
+    struct ntfs_sb_info *sbi = NTFS_SB(fs);
+    sector_t pstart = 0;
+    struct runlist *rlist;
+    struct runlist *ret;
+    const uint32_t sec_size = SECTOR_SIZE(fs);
+    const uint32_t sec_shift = SECTOR_SHIFT(fs);
+
+    dprintf("in %s()\n", __func__);
+
+    if (!NTFS_PVT(inode)->non_resident) {
+        pstart = (sbi->mft_blk + NTFS_PVT(inode)->here) << BLOCK_SHIFT(fs) >>
+                sec_shift;
+        inode->next_extent.len = (inode->size + sec_size - 1) >> sec_shift;
+    } else {
+        rlist = NTFS_PVT(inode)->data.non_resident.rlist;
+
+        if (!lstart || lstart >= NTFS_PVT(inode)->here) {
+            if (runlist_is_empty(rlist))
+                goto out;   /* nothing to do ;-) */
+
+            ret = runlist_remove(&rlist);
+
+            NTFS_PVT(inode)->here =
+                ((ret->run.len << sbi->clust_byte_shift) >> sec_shift);
+
+            pstart = ret->run.lcn << sbi->clust_shift;
+            inode->next_extent.len =
+                ((ret->run.len << sbi->clust_byte_shift) + sec_size - 1) >>
+                sec_shift;
+
+            NTFS_PVT(inode)->data.non_resident.rlist = rlist;
+
+            free(ret);
+            ret = NULL;
+        }
+    }
+
+    inode->next_extent.pstart = pstart;
+
+    return 0;
+
+out:
+    return -1;
+}
+
+static uint32_t ntfs_getfssec(struct file *file, char *buf, int sectors,
+                                bool *have_more)
+{
+    uint8_t non_resident;
+    uint32_t ret;
+    struct fs_info *fs = file->fs;
+    struct inode *inode = file->inode;
+    struct ntfs_mft_record *mrec, *lmrec;
+    struct ntfs_attr_record *attr;
+    char *p;
+
+    dprintf("in %s()\n", __func__);
+
+    non_resident = NTFS_PVT(inode)->non_resident;
+
+    ret = generic_getfssec(file, buf, sectors, have_more);
+    if (!ret)
+        return ret;
+
+    if (!non_resident) {
+        mrec = NTFS_SB(fs)->mft_record_lookup(fs, NTFS_PVT(inode)->mft_no,
+                                              NULL);
+        if (!mrec) {
+            printf("No MFT record found.\n");
+            goto out;
+        }
+
+        lmrec = mrec;
+        attr = ntfs_attr_lookup(fs, NTFS_AT_DATA, &mrec, lmrec);
+        if (!attr) {
+            printf("No attribute found.\n");
+            goto out;
+        }
+
+        p = (char *)((uint8_t *)attr + attr->data.resident.value_offset);
+
+        /* p now points to the data offset, so let's copy it into buf */
+        memcpy(buf, p, inode->size);
+
+        ret = inode->size;
+
+        free(mrec);
+    }
+
+    return ret;
+
+out:
+    free(mrec);
+
+    return 0;
+}
+
+static inline bool is_filename_printable(const char *s)
+{
+    return s && (*s != '.' && *s != '$');
+}
+
+static int ntfs_readdir(struct file *file, struct dirent *dirent)
+{
+    struct fs_info *fs = file->fs;
+    struct inode *inode = file->inode;
+    struct ntfs_mft_record *mrec, *lmrec;
+    block_t blk;
+    uint64_t blk_offset;
+    const uint64_t blk_size = UINT64_C(1) << BLOCK_SHIFT(fs);
+    struct ntfs_attr_record *attr;
+    struct ntfs_idx_root *ir;
+    uint32_t count;
+    int len;
+    struct ntfs_idx_entry *ie = NULL;
+    uint8_t buf[BLOCK_SIZE(fs)];
+    struct ntfs_idx_allocation *iblk;
+    int err;
+    uint8_t *stream;
+    uint8_t *attr_len;
+    struct mapping_chunk chunk;
+    uint32_t offset;
+    int64_t vcn;
+    int64_t lcn;
+    char filename[NTFS_MAX_FILE_NAME_LEN + 1];
+
+    dprintf("in %s()\n", __func__);
+
+    mrec = NTFS_SB(fs)->mft_record_lookup(fs, NTFS_PVT(inode)->mft_no, NULL);
+    if (!mrec) {
+        printf("No MFT record found.\n");
+        goto out;
+    }
+
+    lmrec = mrec;
+    attr = ntfs_attr_lookup(fs, NTFS_AT_INDEX_ROOT, &mrec, lmrec);
+    if (!attr) {
+        printf("No attribute found.\n");
+        goto out;
+    }
+
+    ir = (struct ntfs_idx_root *)((uint8_t *)attr +
+                            attr->data.resident.value_offset);
+
+    if (!file->offset && readdir_state->in_idx_root) {
+        file->offset = (uint32_t)((uint8_t *)&ir->index +
+                                        ir->index.entries_offset);
+    }
+
+idx_root_next_entry:
+    if (readdir_state->in_idx_root) {
+        ie = (struct ntfs_idx_entry *)(uint8_t *)file->offset;
+        if (ie->flags & INDEX_ENTRY_END) {
+            file->offset = 0;
+            readdir_state->in_idx_root = false;
+            readdir_state->idx_blks_count = 1;
+            readdir_state->entries_count = 0;
+            readdir_state->last_vcn = 0;
+            goto descend_into_child_node;
+        }
+
+        file->offset = (uint32_t)((uint8_t *)ie + ie->len);
+        len = ntfs_cvt_filename(filename, ie);
+        if (!is_filename_printable(filename))
+            goto idx_root_next_entry;
+
+        goto done;
+    }
+
+descend_into_child_node:
+    if (!(ie->flags & INDEX_ENTRY_NODE))
+        goto out;
+
+    attr = ntfs_attr_lookup(fs, NTFS_AT_INDEX_ALLOCATION, &mrec, lmrec);
+    if (!attr)
+        goto out;
+
+    if (!attr->non_resident) {
+        printf("WTF ?! $INDEX_ALLOCATION isn't really resident.\n");
+        goto out;
+    }
+
+    attr_len = (uint8_t *)attr + attr->len;
+
+next_run:
+    stream = mapping_chunk_init(attr, &chunk, &offset);
+    count = readdir_state->idx_blks_count;
+    while (count--) {
+        err = parse_data_run(stream, &offset, attr_len, &chunk);
+        if (err) {
+            printf("Error while parsing data runs.\n");
+            goto out;
+        }
+
+        if (chunk.flags & MAP_UNALLOCATED)
+            break;
+        if (chunk.flags & MAP_END)
+            goto out;
+    }
+
+    if (chunk.flags & MAP_UNALLOCATED) {
+       readdir_state->idx_blks_count++;
+       goto next_run;
+    }
+
+next_vcn:
+    vcn = readdir_state->last_vcn;
+    if (vcn >= chunk.len) {
+        readdir_state->last_vcn = 0;
+        readdir_state->idx_blks_count++;
+        goto next_run;
+    }
+
+    lcn = chunk.lcn;
+    blk = (lcn + vcn) << NTFS_SB(fs)->clust_shift << SECTOR_SHIFT(fs) >>
+            BLOCK_SHIFT(fs);
+
+    blk_offset = 0;
+    err = ntfs_read(fs, &buf, blk_size, blk_size, &blk, &blk_offset, NULL,
+                    (uint64_t *)&lcn);
+    if (err) {
+        printf("Error while reading from cache.\n");
+        goto not_found;
+    }
+
+    ntfs_fixups_writeback(fs, (struct ntfs_record *)&buf);
+
+    iblk = (struct ntfs_idx_allocation *)&buf;
+    if (iblk->magic != NTFS_MAGIC_INDX) {
+        printf("Not a valid INDX record.\n");
+        goto not_found;
+    }
+
+idx_block_next_entry:
+    ie = (struct ntfs_idx_entry *)((uint8_t *)&iblk->index +
+                        iblk->index.entries_offset);
+    count = readdir_state->entries_count;
+    for ( ; count--; ie = (struct ntfs_idx_entry *)((uint8_t *)ie + ie->len)) {
+        /* bounds checks */
+        if ((uint8_t *)ie < (uint8_t *)iblk || (uint8_t *)ie +
+            sizeof(struct ntfs_idx_entry_header) >
+            (uint8_t *)&iblk->index + iblk->index.index_len ||
+            (uint8_t *)ie + ie->len >
+            (uint8_t *)&iblk->index + iblk->index.index_len)
+            goto index_err;
+
+        /* last entry cannot contain a key */
+        if (ie->flags & INDEX_ENTRY_END) {
+            /* go to the next VCN */
+            readdir_state->last_vcn += (blk_size / (1 <<
+                                NTFS_SB(fs)->clust_byte_shift));
+            readdir_state->entries_count = 0;
+            goto next_vcn;
+        }
+    }
+
+    readdir_state->entries_count++;
+
+    /* Need to check if this entry has INDEX_ENTRY_END flag set. If
+     * so, then it won't contain a indexed_file file, so continue the
+     * lookup on the next VCN/LCN (if any).
+     */
+    if (ie->flags & INDEX_ENTRY_END)
+        goto next_vcn;
+
+    len = ntfs_cvt_filename(filename, ie);
+    if (!is_filename_printable(filename))
+        goto idx_block_next_entry;
+
+    goto done;
+
+out:
+    readdir_state->in_idx_root = true;
+
+    free(mrec);
+
+    return -1;
+
+done:
+    dirent->d_ino = ie->data.dir.indexed_file;
+    dirent->d_off = file->offset;
+    dirent->d_reclen = offsetof(struct dirent, d_name) + len + 1;
+
+    free(mrec);
+
+    mrec = NTFS_SB(fs)->mft_record_lookup(fs, ie->data.dir.indexed_file, NULL);
+    if (!mrec) {
+        printf("No MFT record found.\n");
+        goto out;
+    }
+
+    dirent->d_type = get_inode_mode(mrec);
+    memcpy(dirent->d_name, filename, len + 1);
+
+    free(mrec);
+
+    return 0;
+
+not_found:
+    printf("Index not found\n");
+    goto out;
+
+index_err:
+    printf("Corrupt index. Aborting lookup...\n");
+    goto out;
+}
+
+static inline struct inode *ntfs_iget(const char *dname, struct inode *parent)
+{
+    return ntfs_index_lookup(dname, parent);
+}
+
+static struct inode *ntfs_iget_root(struct fs_info *fs)
+{
+    uint64_t start_blk;
+    struct ntfs_mft_record *mrec, *lmrec;
+    struct ntfs_attr_record *attr;
+    struct ntfs_vol_info *vol_info;
+    struct inode *inode;
+    int err;
+
+    dprintf("in %s()\n", __func__);
+
+    /* Fetch the $Volume MFT record */
+    start_blk = 0;
+    mrec = NTFS_SB(fs)->mft_record_lookup(fs, FILE_Volume, &start_blk);
+    if (!mrec) {
+        printf("Could not fetch $Volume MFT record!\n");
+        goto err_mrec;
+    }
+
+    lmrec = mrec;
+
+    /* Fetch the volume information attribute */
+    attr = ntfs_attr_lookup(fs, NTFS_AT_VOL_INFO, &mrec, lmrec);
+    if (!attr) {
+        printf("Could not find volume info attribute!\n");
+        goto err_attr;
+    }
+
+    /* Note NTFS version and choose version-dependent functions */
+    vol_info = (void *)((char *)attr + attr->data.resident.value_offset);
+    NTFS_SB(fs)->major_ver = vol_info->major_ver;
+    NTFS_SB(fs)->minor_ver = vol_info->minor_ver;
+    if (vol_info->major_ver == 3 && vol_info->minor_ver == 0)
+        NTFS_SB(fs)->mft_record_lookup = ntfs_mft_record_lookup_3_0;
+    else if (vol_info->major_ver == 3 && vol_info->minor_ver == 1 &&
+            mrec->mft_record_no == FILE_Volume)
+        NTFS_SB(fs)->mft_record_lookup = ntfs_mft_record_lookup_3_1;
+
+    /* Free MFT record */
+    free(mrec);
+    mrec = NULL;
+
+    inode = new_ntfs_inode(fs);
+    inode->fs = fs;
+
+    err = index_inode_setup(fs, FILE_root, inode);
+    if (err)
+        goto err_setup;
+
+    NTFS_PVT(inode)->start = NTFS_PVT(inode)->here;
+
+    return inode;
+
+err_setup:
+
+    free(inode);
+err_attr:
+
+    free(mrec);
+err_mrec:
+
+    return NULL;
+}
+
+/* Initialize the filesystem metadata and return blk size in bits */
+static int ntfs_fs_init(struct fs_info *fs)
+{
+    int read_count;
+    struct ntfs_bpb ntfs;
+    struct ntfs_sb_info *sbi;
+    struct disk *disk = fs->fs_dev->disk;
+    uint8_t mft_record_shift;
+
+    dprintf("in %s()\n", __func__);
+
+    read_count = disk->rdwr_sectors(disk, &ntfs, 0, 1, 0);
+    if (!read_count)
+        return -1;
+
+    if (!ntfs_check_sb_fields(&ntfs))
+        return -1;
+
+    SECTOR_SHIFT(fs) = disk->sector_shift;
+
+    /* Note: ntfs.clust_per_mft_record can be a negative number.
+     * If negative, it represents a shift count, else it represents
+     * a multiplier for the cluster size.
+     */
+    mft_record_shift = ntfs.clust_per_mft_record < 0 ?
+                    -ntfs.clust_per_mft_record :
+                    ilog2(ntfs.sec_per_clust) + SECTOR_SHIFT(fs) +
+                    ilog2(ntfs.clust_per_mft_record);
+
+    SECTOR_SIZE(fs) = 1 << SECTOR_SHIFT(fs);
+
+    sbi = malloc(sizeof *sbi);
+    if (!sbi)
+        malloc_error("ntfs_sb_info structure");
+
+    fs->fs_info = sbi;
+
+    sbi->clust_shift            = ilog2(ntfs.sec_per_clust);
+    sbi->clust_byte_shift       = sbi->clust_shift + SECTOR_SHIFT(fs);
+    sbi->clust_mask             = ntfs.sec_per_clust - 1;
+    sbi->clust_size             = ntfs.sec_per_clust << SECTOR_SHIFT(fs);
+    sbi->mft_record_size        = 1 << mft_record_shift;
+    sbi->clust_per_idx_record   = ntfs.clust_per_idx_record;
+
+    BLOCK_SHIFT(fs) = ilog2(ntfs.clust_per_idx_record) + sbi->clust_byte_shift;
+    BLOCK_SIZE(fs) = 1 << BLOCK_SHIFT(fs);
+
+    sbi->mft_lcn = ntfs.mft_lclust;
+    sbi->mft_blk = ntfs.mft_lclust << sbi->clust_shift << SECTOR_SHIFT(fs) >>
+                BLOCK_SHIFT(fs);
+    /* 16 MFT entries reserved for metadata files (approximately 16 KiB) */
+    sbi->mft_size = mft_record_shift << sbi->clust_shift << 4;
+
+    sbi->clusters = ntfs.total_sectors << SECTOR_SHIFT(fs) >> sbi->clust_shift;
+    if (sbi->clusters > 0xFFFFFFFFFFF4ULL)
+        sbi->clusters = 0xFFFFFFFFFFF4ULL;
+
+    /*
+     * Assume NTFS version 3.0 to begin with. If we find that the
+     * volume is a different version later on, we will adjust at
+     * that time.
+     */
+    sbi->major_ver = 3;
+    sbi->minor_ver = 0;
+    sbi->mft_record_lookup = ntfs_mft_record_lookup_3_0;
+
+    /* Initialize the cache */
+    cache_init(fs->fs_dev, BLOCK_SHIFT(fs));
+
+    return BLOCK_SHIFT(fs);
+}
+
+const struct fs_ops ntfs_fs_ops = {
+    .fs_name        = "ntfs",
+    .fs_flags       = FS_USEMEM | FS_THISIND,
+    .fs_init        = ntfs_fs_init,
+    .searchdir      = NULL,
+    .getfssec       = ntfs_getfssec,
+    .close_file     = generic_close_file,
+    .mangle_name    = generic_mangle_name,
+    .open_config    = generic_open_config,
+    .readdir        = ntfs_readdir,
+    .iget_root      = ntfs_iget_root,
+    .iget           = ntfs_iget,
+    .next_extent    = ntfs_next_extent,
+};
diff --git a/core/fs/ntfs/ntfs.h b/core/fs/ntfs/ntfs.h
new file mode 100644
index 00000000..721a78d7
--- /dev/null
+++ b/core/fs/ntfs/ntfs.h
@@ -0,0 +1,485 @@
+/*
+ * Copyright (C) 2011-2012 Paulo Alcantara <pcacjr@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the
+ * Free Software Foundation, Inc.,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+*/
+
+#include "runlist.h"
+
+#ifndef _NTFS_H_
+#define _NTFS_H_
+
+struct ntfs_bpb {
+    uint8_t jmp_boot[3];
+    char oem_name[8];
+    uint16_t sector_size;
+    uint8_t sec_per_clust;
+    uint16_t res_sectors;
+    uint8_t zero_0[3];
+    uint16_t zero_1;
+    uint8_t media;
+    uint16_t zero_2;
+    uint16_t unused_0;
+    uint16_t unused_1;
+    uint32_t unused_2;
+    uint32_t zero_3;
+    uint32_t unused_3;
+    uint64_t total_sectors;
+    uint64_t mft_lclust;
+    uint64_t mft_mirr_lclust;
+    int8_t clust_per_mft_record;
+    uint8_t unused_4[3];
+    uint8_t clust_per_idx_record;
+    uint8_t unused_5[3];
+    uint64_t vol_serial;
+    uint32_t unused_6;
+
+    uint8_t pad[428];       /* padding to a sector boundary (512 bytes) */
+} __attribute__((__packed__));
+
+/* Function type for an NTFS-version-dependent MFT record lookup */
+struct ntfs_mft_record;
+typedef struct ntfs_mft_record *f_mft_record_lookup(struct fs_info *,
+                                                    uint32_t, block_t *);
+
+struct ntfs_sb_info {
+    block_t mft_blk;                /* The first MFT record block */
+    uint64_t mft_lcn;               /* LCN of the first MFT record */
+    unsigned mft_size;              /* The MFT size in sectors */
+    uint64_t mft_record_size;       /* MFT record size in bytes */
+
+    uint8_t clust_per_idx_record;   /* Clusters per Index Record */
+
+    unsigned long long clusters;    /* Total number of clusters */
+
+    unsigned clust_shift;           /* Based on sectors */
+    unsigned clust_byte_shift;      /* Based on bytes */
+    unsigned clust_mask;
+    unsigned clust_size;
+
+    uint8_t major_ver;              /* Major version from $Volume */
+    uint8_t minor_ver;              /* Minor version from $Volume */
+
+    /* NTFS-version-dependent MFT record lookup function to use */
+    f_mft_record_lookup *mft_record_lookup;
+} __attribute__((__packed__));
+
+/* The NTFS in-memory inode structure */
+struct ntfs_inode {
+    int64_t initialized_size;
+    int64_t allocated_size;
+    unsigned long mft_no;       /* Number of the mft record / inode */
+    uint16_t seq_no;            /* Sequence number of the mft record */
+    uint32_t type;              /* Attribute type of this inode */
+    uint8_t non_resident;
+    union {                 /* Non-resident $DATA attribute */
+        struct {            /* Used only if non_resident flags isn't set */
+            uint32_t offset;    /* Data offset */
+        } resident;
+        struct {            /* Used only if non_resident is set */
+            struct runlist *rlist;
+        } non_resident;
+    } data;
+    uint32_t start_cluster; /* Starting cluster address */
+    sector_t start;         /* Starting sector */
+    sector_t offset;        /* Current sector offset */
+    sector_t here;          /* Sector corresponding to offset */
+};
+
+/* This is structure is used to keep a state for ntfs_readdir() callers.
+ * As NTFS stores directory entries in a complex way, this is structure
+ * ends up saving a state required to find out where we must start from
+ * for the next ntfs_readdir() call.
+ */
+struct ntfs_readdir_state {
+    unsigned long mft_no;       /* MFT record number */
+    bool in_idx_root;           /* It's true if we're still in the INDEX root */
+    uint32_t idx_blks_count;    /* Number of read INDX blocks */
+    uint32_t entries_count;     /* Number of read INDEX entries */
+    int64_t last_vcn;           /* Last VCN of the INDX block */
+};
+
+enum {
+    MAP_UNSPEC,
+    MAP_START           = 1 << 0,
+    MAP_END             = 1 << 1,
+    MAP_ALLOCATED       = 1 << 2,
+    MAP_UNALLOCATED     = 1 << 3,
+    MAP_MASK            = 0x0000000F,
+};
+
+struct mapping_chunk {
+    uint64_t vcn;
+    int64_t lcn;
+    uint64_t len;
+    uint32_t flags;
+};
+
+/* System defined attributes (32-bit)
+ * Each attribute type has a corresponding attribute name (in Unicode)
+ */
+enum {
+    NTFS_AT_UNUSED                      = 0x00,
+    NTFS_AT_STANDARD_INFORMATION        = 0x10,
+    NTFS_AT_ATTR_LIST                   = 0x20,
+    NTFS_AT_FILENAME                    = 0x30,
+    NTFS_AT_OBJ_ID                      = 0x40,
+    NTFS_AT_SECURITY_DESCP              = 0x50,
+    NTFS_AT_VOL_NAME                    = 0x60,
+    NTFS_AT_VOL_INFO                    = 0x70,
+    NTFS_AT_DATA                        = 0x80,
+    NTFS_AT_INDEX_ROOT                  = 0x90,
+    NTFS_AT_INDEX_ALLOCATION            = 0xA0,
+    NTFS_AT_BITMAP                      = 0xB0,
+    NTFS_AT_REPARSE_POINT               = 0xC0,
+    NTFS_AT_EA_INFO                     = 0xD0,
+    NTFS_AT_EA                          = 0xE0,
+    NTFS_AT_PROPERTY_SET                = 0xF0,
+    NTFS_AT_LOGGED_UTIL_STREAM          = 0x100,
+    NTFS_AT_FIRST_USER_DEFINED_ATTR     = 0x1000,
+    NTFS_AT_END                         = 0xFFFFFFFF,
+};
+
+/* NTFS File Permissions (also called attributes in DOS terminology) */
+enum {
+    NTFS_FILE_ATTR_READONLY                     = 0x00000001,
+    NTFS_FILE_ATTR_HIDDEN                       = 0x00000002,
+    NTFS_FILE_ATTR_SYSTEM                       = 0x00000004,
+    NTFS_FILE_ATTR_DIRECTORY                    = 0x00000010,
+    NTFS_FILE_ATTR_ARCHIVE                      = 0x00000020,
+    NTFS_FILE_ATTR_DEVICE                       = 0x00000040,
+    NTFS_FILE_ATTR_NORMAL                       = 0x00000080,
+    NTFS_FILE_ATTR_TEMPORARY                    = 0x00000100,
+    NTFS_FILE_ATTR_SPARSE_FILE                  = 0x00000200,
+    NTFS_FILE_ATTR_REPARSE_POINT                = 0x00000400,
+    NTFS_FILE_ATTR_COMPRESSED                   = 0x00000800,
+    NTFS_FILE_ATTR_OFFLINE                      = 0x00001000,
+    NTFS_FILE_ATTR_NOT_CONTENT_INDEXED          = 0x00002000,
+    NTFS_FILE_ATTR_ENCRYPTED                    = 0x00004000,
+    NTFS_FILE_ATTR_VALID_FLAGS                  = 0x00007FB7,
+    NTFS_FILE_ATTR_VALID_SET_FLAGS              = 0x000031A7,
+    NTFS_FILE_ATTR_DUP_FILE_NAME_INDEX_PRESENT  = 0x10000000,
+    NTFS_FILE_ATTR_DUP_VIEW_INDEX_PRESENT       = 0x20000000,
+};
+
+/*
+ * Magic identifiers present at the beginning of all ntfs record containing
+ * records (like mft records for example).
+ */
+enum {
+    /* Found in $MFT/$DATA */
+    NTFS_MAGIC_FILE     = 0x454C4946,   /* MFT entry */
+    NTFS_MAGIC_INDX     = 0x58444E49,   /* Index buffer */
+    NTFS_MAGIC_HOLE     = 0x454C4F48,
+
+    /* Found in $LogFile/$DATA */
+    NTFS_MAGIC_RSTR     = 0x52545352,
+    NTFS_MAGIC_RCRD     = 0x44524352,
+    /* Found in $LogFile/$DATA (May be found in $MFT/$DATA, also ?) */
+    NTFS_MAGIC_CHKDSK   = 0x444B4843,
+    /* Found in all ntfs record containing records. */
+    NTFS_MAGIC_BAAD     = 0x44414142,
+    NTFS_MAGIC_EMPTY    = 0xFFFFFFFF,   /* Record is empty */
+};
+
+struct ntfs_record {
+    uint32_t magic;
+    uint16_t usa_ofs;
+    uint16_t usa_count;
+} __attribute__((__packed__)) NTFS_RECORD;
+
+/* The $MFT metadata file types */
+enum ntfs_system_file {
+    FILE_MFT            = 0,
+    FILE_MFTMirr        = 1,
+    FILE_LogFile        = 2,
+    FILE_Volume         = 3,
+    FILE_AttrDef        = 4,
+    FILE_root           = 5,
+    FILE_Bitmap         = 6,
+    FILE_Boot           = 7,
+    FILE_BadClus        = 8,
+    FILE_Secure         = 9,
+    FILE_UpCase         = 10,
+    FILE_Extend         = 11,
+    FILE_reserved12     = 12,
+    FILE_reserved13     = 13,
+    FILE_reserved14     = 14,
+    FILE_reserved15     = 15,
+    FILE_reserved16     = 16,
+};
+
+enum {
+    MFT_RECORD_IN_USE       = 0x0001,
+    MFT_RECORD_IS_DIRECTORY = 0x0002,
+} __attribute__((__packed__));
+
+struct ntfs_mft_record {
+    uint32_t magic;
+    uint16_t usa_ofs;
+    uint16_t usa_count;
+    uint64_t lsn;
+    uint16_t seq_no;
+    uint16_t link_count;
+    uint16_t attrs_offset;
+    uint16_t flags;     /* MFT record flags */
+    uint32_t bytes_in_use;
+    uint32_t bytes_allocated;
+    uint64_t base_mft_record;
+    uint16_t next_attr_instance;
+    uint16_t reserved;
+    uint32_t mft_record_no;
+} __attribute__((__packed__));   /* 48 bytes */
+
+/* This is the version without the NTFS 3.1+ specific fields */
+struct ntfs_mft_record_old {
+    uint32_t magic;
+    uint16_t usa_ofs;
+    uint16_t usa_count;
+    uint64_t lsn;
+    uint16_t seq_no;
+    uint16_t link_count;
+    uint16_t attrs_offset;
+    uint16_t flags;     /* MFT record flags */
+    uint32_t bytes_in_use;
+    uint32_t bytes_allocated;
+    uint64_t base_mft_record;
+    uint16_t next_attr_instance;
+} __attribute__((__packed__));   /* 42 bytes */
+
+enum {
+    ATTR_DEF_INDEXABLE          = 0x02,
+    ATTR_DEF_MULTIPLE           = 0x04,
+    ATTR_DEF_NOT_ZERO           = 0x08,
+    ATTR_DEF_INDEXED_UNIQUE     = 0x10,
+    ATTR_DEF_NAMED_UNIQUE       = 0x20,
+    ATTR_DEF_RESIDENT           = 0x40,
+    ATTR_DEF_ALWAYS_LOG         = 0x80,
+};
+
+struct ntfs_attr_record {
+    uint32_t type;      /* Attr. type code */
+    uint32_t len;
+    uint8_t non_resident;
+    uint8_t name_len;
+    uint16_t name_offset;
+    uint16_t flags;     /* Attr. flags */
+    uint16_t instance;
+    union {
+        struct {    /* Resident attribute */
+            uint32_t value_len;
+            uint16_t value_offset;
+            uint8_t flags;  /* Flags of resident attributes */
+            int8_t reserved;
+        } __attribute__((__packed__)) resident;
+        struct {    /* Non-resident attributes */
+            uint64_t lowest_vcn;
+            uint64_t highest_vcn;
+            uint16_t mapping_pairs_offset;
+            uint8_t compression_unit;
+            uint8_t reserved[5];
+            int64_t allocated_size;
+            int64_t data_size; /* Byte size of the attribute value.
+                                * Note: it can be larger than
+                                * allocated_size if attribute value is
+                                * compressed or sparse.
+                                */
+            int64_t initialized_size;
+            int64_t compressed_size;
+        } __attribute__((__packed__)) non_resident;
+    } __attribute__((__packed__)) data;
+} __attribute__((__packed__));
+
+/* Attribute: Attribute List (0x20)
+ * Note: it can be either resident or non-resident
+ */
+struct ntfs_attr_list_entry {
+    uint32_t type;
+    uint16_t length;
+    uint8_t name_length;
+    uint8_t name_offset;
+    uint64_t lowest_vcn;
+    uint64_t mft_ref;
+    uint16_t instance;
+    uint16_t name[0];
+} __attribute__((__packed__));
+
+#define NTFS_MAX_FILE_NAME_LEN 255
+
+/* Possible namespaces for filenames in ntfs (8-bit) */
+enum {
+    FILE_NAME_POSIX             = 0x00,
+    FILE_NAME_WIN32             = 0x01,
+    FILE_NAME_DOS               = 0x02,
+    FILE_NAME_WIN32_AND_DOS     = 0x03,
+} __attribute__((__packed__));
+
+/* Attribute: Filename (0x30)
+ * Note: always resident
+ */
+struct ntfs_filename_attr {
+    uint64_t parent_directory;
+    int64_t ctime;
+    int64_t atime;
+    int64_t mtime;
+    int64_t rtime;
+    uint64_t allocated_size;
+    uint64_t data_size;
+    uint32_t file_attrs;
+    union {
+        struct {
+            uint16_t packed_ea_size;
+            uint16_t reserved;      /* reserved for alignment */
+        } __attribute__((__packed__)) ea;
+        struct {
+            uint32_t reparse_point_tag;
+        } __attribute__((__packed__)) rp;
+    } __attribute__((__packed__)) type;
+    uint8_t file_name_len;
+    uint8_t file_name_type;
+    uint16_t file_name[0];          /* File name in Unicode */
+} __attribute__((__packed__));
+
+/* Attribute: Volume Name (0x60)
+ * Note: always resident
+ * Note: Present only in FILE_volume
+ */
+struct ntfs_vol_name {
+    uint16_t name[0];       /* The name of the volume in Unicode */
+} __attribute__((__packed__));
+
+/* Attribute: Volume Information (0x70)
+ * Note: always resident
+ * Note: present only in FILE_Volume
+ */
+struct ntfs_vol_info {
+    uint64_t reserved;
+    uint8_t major_ver;
+    uint8_t minor_ver;
+    uint16_t flags;     /* Volume flags */
+} __attribute__((__packed__));
+
+/* Attribute: Data attribute (0x80)
+ * Note: can be either resident or non-resident
+ */
+struct ntfs_data_attr {
+    uint8_t data[0];
+} __attribute__((__packed__));
+
+/* Index header flags (8-bit) */
+enum {
+    SMALL_INDEX = 0,
+    LARGE_INDEX = 1,
+    LEAF_NODE   = 0,
+    INDEX_NODE  = 1,
+    NODE_MASK   = 1,
+} __attribute__((__packed__));
+
+/* Header for the indexes, describing the INDEX_ENTRY records, which
+ * follow the struct ntfs_idx_header.
+ */
+struct ntfs_idx_header {
+    uint32_t entries_offset;
+    uint32_t index_len;
+    uint32_t allocated_size;
+    uint8_t flags;              /* Index header flags */
+    uint8_t reserved[3];        /* Align to 8-byte boundary */
+} __attribute__((__packed__));
+
+/* Attribute: Index Root (0x90)
+ * Note: always resident
+ */
+struct ntfs_idx_root {
+    uint32_t type;  /* It is $FILE_NAME for directories, zero for view indexes.
+                     * No other values allowed.
+                     */
+    uint32_t collation_rule;
+    uint32_t index_block_size;
+    uint8_t clust_per_index_block;
+    uint8_t reserved[3];
+    struct ntfs_idx_header index;
+} __attribute__((__packed__));
+
+/* Attribute: Index allocation (0xA0)
+ * Note: always non-resident, of course! :-)
+ */
+struct ntfs_idx_allocation {
+    uint32_t magic;
+    uint16_t usa_ofs;           /* Update Sequence Array offsets */
+    uint16_t usa_count;         /* Update Sequence Array number in bytes */
+    int64_t lsn;
+    int64_t index_block_vcn;    /* Virtual cluster number of the index block */
+    struct ntfs_idx_header index;
+} __attribute__((__packed__));
+
+enum {
+    INDEX_ENTRY_NODE            = 1,
+    INDEX_ENTRY_END             = 2,
+    /* force enum bit width to 16-bit */
+    INDEX_ENTRY_SPACE_FILTER    = 0xFFFF,
+} __attribute__((__packed__));
+
+struct ntfs_idx_entry_header {
+    union {
+        struct { /* Only valid when INDEX_ENTRY_END is not set */
+            uint64_t indexed_file;
+        } __attribute__((__packed__)) dir;
+        struct { /* Used for views/indexes to find the entry's data */
+            uint16_t data_offset;
+            uint16_t data_len;
+            uint32_t reservedV;
+        } __attribute__((__packed__)) vi;
+    } __attribute__((__packed__)) data;
+    uint16_t len;
+    uint16_t key_len;
+    uint16_t flags;     /* Index entry flags */
+    uint16_t reserved;  /* Align to 8-byte boundary */
+} __attribute__((__packed__));
+
+struct ntfs_idx_entry {
+    union {
+        struct { /* Only valid when INDEX_ENTRY_END is not set */
+            uint64_t indexed_file;
+        } __attribute__((__packed__)) dir;
+        struct { /* Used for views/indexes to find the entry's data */
+            uint16_t data_offset;
+            uint16_t data_len;
+            uint32_t reservedV;
+        } __attribute__((__packed__)) vi;
+    } __attribute__((__packed__)) data;
+    uint16_t len;
+    uint16_t key_len;
+    uint16_t flags;     /* Index entry flags */
+    uint16_t reserved;  /* Align to 8-byte boundary */
+    union {
+        struct ntfs_filename_attr file_name;
+        //SII_INDEX_KEY sii;
+        //SDH_INDEX_KEY sdh;
+        //GUID object_id;
+        //REPARSE_INDEX_KEY reparse;
+        //SID sid;
+        uint32_t owner_id;
+    } __attribute__((__packed__)) key;
+} __attribute__((__packed__));
+
+static inline struct ntfs_sb_info *NTFS_SB(struct fs_info *fs)
+{
+    return fs->fs_info;
+}
+
+#define NTFS_PVT(i) ((struct ntfs_inode *)((i)->pvt))
+
+#endif /* _NTFS_H_ */
diff --git a/core/fs/ntfs/runlist.h b/core/fs/ntfs/runlist.h
new file mode 100644
index 00000000..4938696b
--- /dev/null
+++ b/core/fs/ntfs/runlist.h
@@ -0,0 +1,83 @@
+/*
+ * Copyright (C) 2011 Paulo Alcantara <pcacjr@gmail.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the
+ * Free Software Foundation, Inc.,
+ * 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+*/
+
+#ifndef _RUNLIST_H_
+#define _RUNLIST_H_
+
+struct runlist_element {
+    uint64_t vcn;
+    int64_t lcn;
+    uint64_t len;
+};
+
+struct runlist {
+    struct runlist_element run;
+    struct runlist *next;
+};
+
+static struct runlist *tail;
+
+static inline bool runlist_is_empty(struct runlist *rlist)
+{
+    return !rlist;
+}
+
+static inline struct runlist *runlist_alloc(void)
+{
+    struct runlist *rlist;
+
+    rlist = malloc(sizeof *rlist);
+    if (!rlist)
+        malloc_error("runlist structure");
+
+    rlist->next = NULL;
+
+    return rlist;
+}
+
+static inline void runlist_append(struct runlist **rlist,
+                                struct runlist_element *elem)
+{
+    struct runlist *n = runlist_alloc();
+
+    n->run = *elem;
+
+    if (runlist_is_empty(*rlist)) {
+        *rlist = n;
+        tail = n;
+    } else {
+        tail->next = n;
+        tail = n;
+    }
+}
+
+static inline struct runlist *runlist_remove(struct runlist **rlist)
+{
+    struct runlist *ret;
+
+    if (runlist_is_empty(*rlist))
+        return NULL;
+
+    ret = *rlist;
+    *rlist = ret->next;
+
+    return ret;
+}
+
+#endif /* _RUNLIST_H_ */
diff --git a/core/fs/pxe/dhcp_option.c b/core/fs/pxe/dhcp_option.c
index 50f2de04..f63d4a91 100644
--- a/core/fs/pxe/dhcp_option.c
+++ b/core/fs/pxe/dhcp_option.c
@@ -227,12 +227,10 @@ static void parse_dhcp_options(const void *option, int size, uint8_t opt_filter)
  * LocalDomain	- Local domain name
  * MAC_len, MAC	- Client identifier, if MAC_len == 0
  *
- * This assumes the DHCP packet is in "trackbuf".
- *
  */
-void parse_dhcp(int pkt_len)
+void parse_dhcp(const void *pkt, size_t pkt_len)
 {
-    struct bootp_t *dhcp = (struct bootp_t *)trackbuf;
+    const struct bootp_t *dhcp = (const struct bootp_t *)pkt;
     int opt_len;
 
     IPInfo.ipv4 = 4;		/* This is IPv4 only for now... */
diff --git a/core/fs/pxe/dnsresolv.c b/core/fs/pxe/dnsresolv.c
index 641ea389..fda0f815 100644
--- a/core/fs/pxe/dnsresolv.c
+++ b/core/fs/pxe/dnsresolv.c
@@ -337,11 +337,7 @@ done:
     return result;
 }
 
-
-/*
- * the one should be called from ASM file
- */
-void pxe_dns_resolv(com32sys_t *regs)
+void pm_pxe_dns_resolv(com32sys_t *regs)
 {
     const char *name = MK_PTR(regs->ds, regs->esi.w[0]);
 
diff --git a/core/fs/pxe/pxe.c b/core/fs/pxe/pxe.c
index 7b63d17b..6f490ce8 100644
--- a/core/fs/pxe/pxe.c
+++ b/core/fs/pxe/pxe.c
@@ -340,13 +340,13 @@ static void ack_packet(struct inode *inode, uint16_t ack_num)
 
 
 /**
- * Get a DHCP packet from the PXE stack into the trackbuf
+ * Get a DHCP packet from the PXE stack into a lowmem buffer
  *
  * @param:  type,  packet type
  * @return: buffer size
  *
  */
-static int pxe_get_cached_info(int type)
+static int pxe_get_cached_info(int type, void *buf, size_t bufsiz)
 {
     int err;
     static __lowmem struct s_PXENV_GET_CACHED_INFO get_cached_info;
@@ -354,8 +354,8 @@ static int pxe_get_cached_info(int type)
 
     get_cached_info.Status      = 0;
     get_cached_info.PacketType  = type;
-    get_cached_info.BufferSize  = 8192;
-    get_cached_info.Buffer      = FAR_PTR(trackbuf);
+    get_cached_info.BufferSize  = bufsiz;
+    get_cached_info.Buffer      = FAR_PTR(buf);
     err = pxe_call(PXENV_GET_CACHED_INFO, &get_cached_info);
     if (err) {
         printf("PXE API call failed, error  %04x\n", err);
@@ -1151,7 +1151,7 @@ static void make_sysuuid_string(void)
 
 /*
  * Generate an ip=<client-ip>:<boot-server-ip>:<gw-ip>:<netmask>
- * option into IPOption based on a DHCP packet in trackbuf.
+ * option into IPOption based on DHCP information in IPInfo.
  *
  */
 char __bss16 IPOption[3+4*16];
@@ -1187,9 +1187,6 @@ static void ip_init(void)
 /*
  * Print the IPAPPEND strings, in order
  */
-extern const uint16_t IPAppends[];
-extern const char numIPAppends[];
-
 static void print_ipappend(void)
 {
     size_t i;
@@ -1448,8 +1445,15 @@ static void udp_init(void)
  */
 static void network_init(void)
 {
-    struct bootp_t *bp = (struct bootp_t *)trackbuf;
     int pkt_len;
+    struct bootp_t *bp;
+    const size_t dhcp_max_packet = 4096;
+
+    bp = lmalloc(dhcp_max_packet);
+    if (!bp) {
+	printf("Out of low memory\n");
+	kaboom();
+    }
 
     *LocalDomain = 0;   /* No LocalDomain received */
 
@@ -1457,8 +1461,8 @@ static void network_init(void)
      * Get the DHCP client identifiers (query info 1)
      */
     printf("Getting cached packet ");
-    pkt_len = pxe_get_cached_info(1);
-    parse_dhcp(pkt_len);
+    pkt_len = pxe_get_cached_info(1, bp, dhcp_max_packet);
+    parse_dhcp(bp, pkt_len);
     /*
      * We don't use flags from the request packet, so
      * this is a good time to initialize DHCPMagic...
@@ -1474,8 +1478,8 @@ static void network_init(void)
      * Get the BOOTP/DHCP packet that brought us file (and an IP
      * address). This lives in the DHCPACK packet (query info 2)
      */
-    pkt_len = pxe_get_cached_info(2);
-    parse_dhcp(pkt_len);
+    pkt_len = pxe_get_cached_info(2, bp, dhcp_max_packet);
+    parse_dhcp(bp, pkt_len);
     /*
      * Save away MAC address (assume this is in query info 2. If this
      * turns out to be problematic it might be better getting it from
@@ -1489,10 +1493,12 @@ static void network_init(void)
      * Get the boot file and other info. This lives in the CACHED_REPLY
      * packet (query info 3)
      */
-    pkt_len = pxe_get_cached_info(3);
-    parse_dhcp(pkt_len);
+    pkt_len = pxe_get_cached_info(3, bp, dhcp_max_packet);
+    parse_dhcp(bp, pkt_len);
     printf("\n");
 
+    lfree(bp);
+
     make_bootif_string();
     make_sysuuid_string();
     ip_init();
@@ -1647,7 +1653,7 @@ int reset_pxe(void)
  * This function unloads the PXE and UNDI stacks and
  * unclaims the memory.
  */
-void unload_pxe(void)
+void unload_pxe(uint16_t flags)
 {
     /* PXE unload sequences */
     static const uint8_t new_api_unload[] = {
@@ -1676,7 +1682,7 @@ void unload_pxe(void)
     dprintf("FBM after reset_pxe = %d, err = %d\n", BIOS_fbm, err);
 
     /* If we want to keep PXE around, we still need to reset it */
-    if (KeepPXE || err)
+    if (flags || err)
 	return;
 
     dprintf("APIVer = %04x\n", APIVer);
diff --git a/core/fs/pxe/pxe.h b/core/fs/pxe/pxe.h
index 1e6fa76a..c754106d 100644
--- a/core/fs/pxe/pxe.h
+++ b/core/fs/pxe/pxe.h
@@ -232,14 +232,12 @@ static inline uint32_t gateway(uint32_t ip)
 
 /* pxe.c */
 bool ip_ok(uint32_t);
-int pxe_call(int, void *);
 
 /* dhcp_options.c */
-void parse_dhcp(int);
+void parse_dhcp(const void *, size_t);
 
 /* dnsresolv.c */
 int dns_mangle(char **, const char *);
-uint32_t dns_resolv(const char *);
 
 /* idle.c */
 void pxe_idle_init(void);